diff --git a/CMakeLists.txt b/CMakeLists.txt
index b215cce5d76251005d880a2d9a5e25531ac689c8..61a85bf4637f99157c7762b69e9d884edf90b874 100755
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -171,6 +171,7 @@ if( WITH_CUDA STREQUAL "yes" )
endif( CUDA_FOUND )
endif( WITH_CUDA STREQUAL "yes" )
+
####
# Check for OpenMP
#
diff --git a/Copyright b/Copyright
index 1acdabf71aab5f9fa39db4324b5df03c2d4628f8..5582a49998d29ad3d7552e697c4eeb26a09ea461 100644
--- a/Copyright
+++ b/Copyright
@@ -1,6 +1,6 @@
MIT License
-Copyright (c) 2004-2016 Tomáš Oberhuber et al.
+Copyright (c) 2004-2017 Tomáš Oberhuber et al.
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
diff --git a/TODO b/TODO
index e83b80a67e00b0c6fa1911f2d37919b410aa03ba..089a312dd155e7817f8fc5ab5b3a0c823f832be8 100644
--- a/TODO
+++ b/TODO
@@ -16,9 +16,6 @@ TODO:
TODO:
- - zrejme bude potreba udrzovat ke kazdemu objektu jeho obraz na GPU/MIC
- - to by zarizovala metoda syncToDevice() napr. kazdy objekt by mel promennou modified, ktera by rikala, jestli se zmenil a zda je nutne ho
- prekopirovavat
TODO:
- zavest namespaces
@@ -31,6 +28,11 @@ TODO: CUDA unified memory
se s nimi pracovat postaru
- bylo by dobre to obalit unique poinetry, aby se nemusela delat dealokace rucne
+TODO: shared pointery
+ - mohli bysme pomoci nich odstranit Shared objekty
+ - asi by bylo lepsi datcounter z shared pointeru primo do array a tento counter by se alokoval az po porvnim sdileni dat
+ - diky tomu by se array mohlo vytvaret i na gpu bez nutnosti dynamicke alokace, jen by nebylo mozne delat bind (nebo nejaky zjednoduseny)
+
TODO: Mesh
* vsechny traits zkusit presunout do jednotneho MeshTraits, tj. temer MeshConfigTraits ale pojmenovat jako MeshTraits
* omezit tnlDimesnionsTag - asi to ale nepujde
diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt
index 0f00654e037cfb1ab253fa4276a57fc1f0f25126..dca68ee58a2075836a177be031be115907577cc8 100755
--- a/examples/CMakeLists.txt
+++ b/examples/CMakeLists.txt
@@ -1,5 +1,15 @@
add_subdirectory( heat-equation )
+add_subdirectory( transport-equation )
add_subdirectory( navier-stokes )
-add_subdirectory( advection )
+
+#add_subdirectory( mean-curvature-flow )
+#add_subdirectory( hamilton-jacobi )
+#add_subdirectory( hamilton-jacobi-parallel )
+#add_subdirectory( fast-sweeping )
+#add_subdirectory( hamilton-jacobi-parallel-map )
+#add_subdirectory( fast-sweeping-map )
+#add_subdirectory( narrow-band )
+
add_subdirectory( inviscid-flow )
#add_subdirectory( mean-curvature-flow )
+
diff --git a/examples/advection/CMakeLists.txt b/examples/advection/CMakeLists.txt
deleted file mode 100644
index a09abadcdf99acf50e41f0909dfe02f4dfc31ea9..0000000000000000000000000000000000000000
--- a/examples/advection/CMakeLists.txt
+++ /dev/null
@@ -1,20 +0,0 @@
-set( tnl_heat_equation_SOURCES
- advection.cpp
- advection.cu )
-
-IF( BUILD_CUDA )
- CUDA_ADD_EXECUTABLE( tnl-advection${debugExt} advection.cu)
- target_link_libraries( tnl-advection${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE( tnl-advection${debugExt} advection.cpp)
- target_link_libraries( tnl-advection${debugExt} tnl${debugExt}-${tnlVersion} )
-ENDIF( BUILD_CUDA )
-
-
-INSTALL( TARGETS tnl-advection${debugExt}
- RUNTIME DESTINATION bin
- PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
-
-INSTALL( FILES tnl-run-advection
- ${tnl_heat_equation_SOURCES}
- DESTINATION share/tnl-${tnlVersion}/examples/advection )
diff --git a/examples/advection/LaxFridrichs.h b/examples/advection/LaxFridrichs.h
deleted file mode 100644
index 46e63b77a2fa7661b8f7dd83cae964eab0c61272..0000000000000000000000000000000000000000
--- a/examples/advection/LaxFridrichs.h
+++ /dev/null
@@ -1,222 +0,0 @@
-#ifndef LaxFridrichs_H
-#define LaxFridrichs_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichs
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichs< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
- Real tau;
- Real artificalViscosity;
- Real advectionSpeedX;
- Real advectionSpeedY;
-
- void setAdvectionSpeedY(const Real& advectionSpeed)
- {
- this->advectionSpeedY = advectionSpeed;
- }
-
-
- void setAdvectionSpeedX(const Real& advectionSpeed)
- {
- this->advectionSpeedX = advectionSpeed;
- }
-
- void setViscosity(const Real& artificalViscosity)
- {
- this->artificalViscosity = artificalViscosity;
- }
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- static String getType();
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichs< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
- Real tau;
- Real artificalViscosity;
- Real advectionSpeedX;
- Real advectionSpeedY;
-
- void setAdvectionSpeedY(const Real& advectionSpeed)
- {
- this->advectionSpeedY = advectionSpeed;
- }
-
-
- void setAdvectionSpeedX(const Real& advectionSpeed)
- {
- this->advectionSpeedX = advectionSpeed;
- }
-
- void setViscosity(const Real& artificalViscosity)
- {
- this->artificalViscosity = artificalViscosity;
- }
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- static String getType();
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichs< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
- Real tau;
- Real artificalViscosity;
- Real advectionSpeedX;
- Real advectionSpeedY;
-
- void setAdvectionSpeedY(const Real& advectionSpeed)
- {
- this->advectionSpeedY = advectionSpeed;
- }
-
-
- void setAdvectionSpeedX(const Real& advectionSpeed)
- {
- this->advectionSpeedX = advectionSpeed;
- }
-
- void setViscosity(const Real& artificalViscosity)
- {
- this->artificalViscosity = artificalViscosity;
- }
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- static String getType();
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-
-#include "LaxFridrichs_impl.h"
-
-#endif /* LaxFridrichs_H */
diff --git a/examples/advection/LaxFridrichs_impl.h b/examples/advection/LaxFridrichs_impl.h
deleted file mode 100644
index dc65664b3b5b8ef599d61988924b3a705b6f94ca..0000000000000000000000000000000000000000
--- a/examples/advection/LaxFridrichs_impl.h
+++ /dev/null
@@ -1,361 +0,0 @@
-#ifndef LaxFridrichs_IMPL_H
-#define LaxFridrichs_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichs< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichs< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichs< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- double a;
- a = this->advectionSpeedX;
- return (0.5 / this->tau ) * this->artificalViscosity *
- ( u[ west ]- 2.0 * u[ center ] + u[ east ] )
- - (a = this->advectionSpeedX * ( u[ east ] - u[west] ) ) * hxInverse * 0.5;
-
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichs< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichs< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichs< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichs< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichs< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
- const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- double a;
- double b;
- a = this->advectionSpeedX;
- b = this->advectionSpeedY;
- return ( 0.25 / this->tau ) * this->artificalViscosity *
- ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4 * u[ center ] ) -
- (a * ( u[ east ] - u[west] ) ) * hxInverse * 0.5 -
- (b * ( u[ north ] - u[ south ] ) ) * hyInverse * 0.5;
-
-
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichs< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichs< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichs< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichs< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichs< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichs< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- //return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichs< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- /* const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );*/
-}
-
-} // namespace TNL
-
-#endif /* LaxFridrichsIMPL_H */
-
diff --git a/examples/advection/advection.cpp b/examples/advection/advection.cpp
deleted file mode 100644
index 374d4714086168f7ebf1ed2a62664c4aaae0c4b7..0000000000000000000000000000000000000000
--- a/examples/advection/advection.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "advection.h"
diff --git a/examples/advection/advection.cu b/examples/advection/advection.cu
deleted file mode 100644
index 374d4714086168f7ebf1ed2a62664c4aaae0c4b7..0000000000000000000000000000000000000000
--- a/examples/advection/advection.cu
+++ /dev/null
@@ -1 +0,0 @@
-#include "advection.h"
diff --git a/examples/advection/advection.h b/examples/advection/advection.h
deleted file mode 100644
index dc8d52ae085b277225634637969e72e2d3ff6a40..0000000000000000000000000000000000000000
--- a/examples/advection/advection.h
+++ /dev/null
@@ -1,122 +0,0 @@
-#include <TNL/tnlConfig.h>
-#include <TNL/Solvers/Solver.h>
-#include <TNL/Solvers/BuildConfigTags.h>
-#include <TNL/Operators/DirichletBoundaryConditions.h>
-#include <TNL/Operators/NeumannBoundaryConditions.h>
-#include <TNL/Functions/Analytic/Constant.h>
-#include "advectionProblem.h"
-#include "LaxFridrichs.h"
-#include "advectionRhs.h"
-#include "advectionBuildConfigTag.h"
-
-using namespace TNL;
-
-typedef advectionBuildConfigTag BuildConfig;
-
-/****
- * Uncoment the following (and comment the previous line) for the complete build.
- * This will include support for all floating point precisions, all indexing types
- * and more solvers. You may then choose between them from the command line.
- * The compile time may, however, take tens of minutes or even several hours,
- * esppecially if CUDA is enabled. Use this, if you want, only for the final build,
- * not in the development phase.
- */
-//typedef tnlDefaultConfigTag BuildConfig;
-
-template< typename ConfigTag >class advectionConfig
-{
- public:
- static void configSetup( Config::ConfigDescription & config )
- {
- config.addDelimiter( "advection settings:" );
- config.addEntry< String >( "boundary-conditions-type", "Choose the boundary conditions type.", "dirichlet");
- config.addEntryEnum< String >( "dirichlet" );
- config.addEntryEnum< String >( "neumann" );
- config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions." );
- config.addEntry< double >( "artifical-viscosity", "This sets value of artifical viscosity (default 1)", 1.0);
- config.addEntry< String >( "begin", "choose begin type", "sin");
- config.addEntryEnum< String >( "sin");
- config.addEntryEnum< String >( "sin_square");
- config.addEntry< double >( "advection-speedX", "This sets value of advection speed in X direction (default 1)" , 1.0);
- config.addEntry< double >( "advection-speedY", "This sets value of advection speed in Y direction (default 1)" , 1.0);
- config.addEntry< String >( "move", "choose movement type", "advection");
- config.addEntryEnum< String >( "advection");
- config.addEntryEnum< String >( "rotation");
- config.addEntry< int >( "dimension", "choose movement typeproblem dimension", 1);
- config.addEntryEnum< int >( 1 );
- config.addEntryEnum< int >( 2 );
- config.addEntry< double >( "realSize", "Real size of scheme", 1.0);
-
- /****
- * Add definition of your solver command line arguments.
- */
-
- }
-};
-
-template< typename Real,
- typename Device,
- typename Index,
- typename MeshType,
- typename ConfigTag,
- typename SolverStarter >
-class advectionSetter
-{
- public:
-
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
-
- static bool run( const Config::ParameterContainer & parameters )
- {
- enum { Dimensions = MeshType::getMeshDimensions() };
- typedef LaxFridrichs< MeshType, Real, Index > ApproximateOperator;
- typedef advectionRhs< MeshType, Real > RightHandSide;
- typedef Containers::StaticVector < MeshType::getMeshDimensions(), Real > Vertex;
-
- /****
- * Resolve the template arguments of your solver here.
- * The following code is for the Dirichlet and the Neumann boundary conditions.
- * Both can be constant or defined as descrete values of Vector.
- */
- String boundaryConditionsType = parameters.getParameter< String >( "boundary-conditions-type" );
- if( parameters.checkParameter( "boundary-conditions-constant" ) )
- {
- typedef Functions::Analytic::Constant< Dimensions, Real > Constant;
- if( boundaryConditionsType == "dirichlet" )
- {
- typedef Operators::DirichletBoundaryConditions< MeshType, Constant, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
- typedef advectionProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Operators::NeumannBoundaryConditions< MeshType, Constant, Real, Index > BoundaryConditions;
- typedef advectionProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Functions::MeshFunction< MeshType > MeshFunction;
- if( boundaryConditionsType == "dirichlet" )
- {
- typedef Operators::DirichletBoundaryConditions< MeshType, MeshFunction, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
- typedef advectionProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Operators::NeumannBoundaryConditions< MeshType, MeshFunction, Real, Index > BoundaryConditions;
- typedef advectionProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
-
-};
-
-int main( int argc, char* argv[] )
-{
- Solvers::Solver< advectionSetter, advectionConfig, BuildConfig > solver;
- if( ! solver. run( argc, argv ) )
- return EXIT_FAILURE;
- return EXIT_SUCCESS;
-}
-
diff --git a/examples/advection/advectionBuildConfigTag.h b/examples/advection/advectionBuildConfigTag.h
deleted file mode 100644
index 98371b43a3fce8743b7945bbec10ef04ce0d219c..0000000000000000000000000000000000000000
--- a/examples/advection/advectionBuildConfigTag.h
+++ /dev/null
@@ -1,50 +0,0 @@
-#ifndef advectionBUILDCONFIGTAG_H_
-#define advectionBUILDCONFIGTAG_H_
-
-#include <TNL/Solvers/BuildConfigTags.h>
-
-namespace TNL {
-
-class advectionBuildConfigTag{};
-
-namespace Solvers {
-
-/****
- * Turn off support for float and long double.
- */
-template<> struct ConfigTagReal< advectionBuildConfigTag, float > { enum { enabled = false }; };
-template<> struct ConfigTagReal< advectionBuildConfigTag, long double > { enum { enabled = false }; };
-
-/****
- * Turn off support for short int and long int indexing.
- */
-template<> struct ConfigTagIndex< advectionBuildConfigTag, short int >{ enum { enabled = false }; };
-template<> struct ConfigTagIndex< advectionBuildConfigTag, long int >{ enum { enabled = false }; };
-
-/****
- * Use of Grid is enabled for allowed dimensions and Real, Device and Index types.
- */
-
-template< int Dimensions, typename Real, typename Device, typename Index >
- struct ConfigTagMesh< advectionBuildConfigTag, Meshes::Grid< Dimensions, Real, Device, Index > >
- { enum { enabled = ConfigTagDimensions< advectionBuildConfigTag, Dimensions >::enabled &&
- ConfigTagReal< advectionBuildConfigTag, Real >::enabled &&
- ConfigTagDevice< advectionBuildConfigTag, Device >::enabled &&
- ConfigTagIndex< advectionBuildConfigTag, Index >::enabled }; };
-
-/****
- * Please, chose your preferred time discretisation here.
- */
-template<> struct ConfigTagTimeDiscretisation< advectionBuildConfigTag, ExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
-template<> struct ConfigTagTimeDiscretisation< advectionBuildConfigTag, SemiImplicitTimeDiscretisationTag >{ enum { enabled = true }; };
-template<> struct ConfigTagTimeDiscretisation< advectionBuildConfigTag, ImplicitTimeDiscretisationTag >{ enum { enabled = true }; };
-
-/****
- * Only the Runge-Kutta-Merson solver is enabled by default.
- */
-template<> struct ConfigTagExplicitSolver< advectionBuildConfigTag, Solvers::ExplicitEulerSolverTag >{ enum { enabled = true }; };
-
-} // namespace Solvers
-} // namespace TNL
-
-#endif /* advectionBUILDCONFIGTAG_H_ */
diff --git a/examples/advection/advectionProblem_impl.h b/examples/advection/advectionProblem_impl.h
deleted file mode 100644
index cd37f52eff7454b863d14eb245b2b9d7236a5a37..0000000000000000000000000000000000000000
--- a/examples/advection/advectionProblem_impl.h
+++ /dev/null
@@ -1,342 +0,0 @@
-#ifndef advectionPROBLEM_IMPL_H_
-#define advectionPROBLEM_IMPL_H_
-
-#include <TNL/FileName.h>
-#include <TNL/Matrices/MatrixSetter.h>
-#include <TNL/Solvers/PDE/ExplicitUpdater.h>
-#include <TNL/Solvers/PDE/LinearSystemAssembler.h>
-#include <TNL/Solvers/PDE/BackwardTimeDiscretisation.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-String
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getTypeStatic()
-{
- return String( "advectionProblem< " ) + Mesh :: getTypeStatic() + " >";
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-String
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getPrologHeader() const
-{
- return String( "advection" );
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-writeProlog( Logger& logger, const Config::ParameterContainer& parameters ) const
-{
- /****
- * Add data you want to have in the computation report (log) as follows:
- * logger.writeParameter< double >( "Parameter description", parameter );
- */
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setup( const MeshPointer& meshPointer,
- const Config::ParameterContainer& parameters,
- const String& prefix )
-{
- if( ! this->boundaryConditionPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) ||
- ! this->rightHandSidePointer->setup( parameters, prefix + "right-hand-side-" ) )
- return false;
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-typename advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::IndexType
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getDofs( const MeshPointer& mesh ) const
-{
- /****
- * Return number of DOFs (degrees of freedom) i.e. number
- * of unknowns to be resolved by the main solver.
- */
- return mesh->template getEntitiesCount< typename MeshType::Cell >();
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-bindDofs( const MeshPointer& mesh,
- DofVectorPointer& dofVector )
-{
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setInitialCondition( const Config::ParameterContainer& parameters,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData )
-{
- std::cout << "vaules adding";
- typedef typename MeshType::Cell Cell;
- int dimensions = parameters.getParameter< int >( "dimension" );
- int count = mesh->template getEntitiesCount< Cell >();
- const RealType& size = parameters.getParameter< double >( "realSize" ) / ::pow(count, 1.0/dimensions);
- const String& beginChoice = parameters.getParameter< String >( "begin" );
- std::cout << beginChoice << " " << dimensions << " " << size << " " << count << " "<< 1/dimensions << std::endl;
- getchar();
- if (beginChoice == "sin_square")
- {
- double constantFunction;
- if (dimensions == 1)
- {
- std::cout << "adding DOFS" << std::endl;
- ( *dofs )[0] = 0;
- double expValue;
- for (IndexType i = 1; i < count-2; i++)
- {
- expValue = std::exp(-std::pow(size*i-2,2));
- if( (i>0.4*count) && (i<0.5*count))
- constantFunction=1;
- else
- constantFunction=0;
- if(expValue>constantFunction)
- ( *dofs )[i] = expValue;
- else ( *dofs )[i] = constantFunction;
- };
- ( *dofs )[count-1] = 0;
- }
- else if (dimensions == 2)
- {
- count = std::sqrt(count);
- double expValue;
- for (IndexType i = 0; i < count-1; i++)
- for (IndexType j = 0; j < count-1; j++)
- {
- expValue = std::exp(-std::pow(size*i-2,2)-std::pow(size*j-2,2));
- if( (i>0.4*count) && (i<0.5*count) && (j>0.4*count) && (j<0.5*count) )
- constantFunction=1;
- else constantFunction=0;
- if( expValue>constantFunction)
- ( *dofs )[i * count + j] = expValue;
- else ( *dofs )[i * count + j] = constantFunction;
- };
- };
- }
- else if (beginChoice == "sin")
- {
- if (dimensions == 1)
- {
- ( *dofs )[0] = 0;
- for (IndexType i = 1; i < count-2; i++)
- {
- ( *dofs )[i] = std::exp(-std::pow(size*i-2,2));
- };
- ( *dofs )[count-1] = 0;
- }
- else if (dimensions == 2)
- {
- count = ::sqrt(count);
- for (IndexType i = 1; i < count-1; i++)
- for (IndexType j = 1; j < count-1; j++)
- {
- ( *dofs )[i * count + j] = std::exp(-std::pow(size*i-2,2)-std::pow(size*j-2,2));
- };
- };
- };
- //setting velocity field
- std::cout << *dofs << std::endl;
- getchar();
- /*const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );
- if( ! dofs.load( initialConditionFile ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << initialConditionFile << "." << std::endl;
- return false;
- }
- return true;*/
- dofs->save( "dofs.tnl" );
- this->velocityType = parameters.getParameter< String >( "move" );
- const double artificalViscosity = parameters.getParameter< double >( "artifical-viscosity" );
- differentialOperatorPointer->setViscosity(artificalViscosity);
- const double advectionSpeedX = parameters.getParameter< double >( "advection-speedX" );
- differentialOperatorPointer->setAdvectionSpeedX(advectionSpeedX);
- const double advectionSpeedY = parameters.getParameter< double >( "advection-speedY" );
- differentialOperatorPointer->setAdvectionSpeedY(advectionSpeedY);
- std::cout << "vaules added";
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
- template< typename Matrix >
-bool
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setupLinearSystem( const MeshPointer& mesh,
- Matrix& matrix )
-{
- const IndexType dofs = this->getDofs( mesh );
- typedef typename Matrix::ObjectType::CompressedRowsLengthsVector CompressedRowsLengthsVectorType;
- SharedPointer< CompressedRowsLengthsVectorType > rowLengths;
- if( ! rowLengths->setSize( dofs ) )
- return false;
- Matrices::MatrixSetter< MeshType, DifferentialOperator, BoundaryCondition, CompressedRowsLengthsVectorType > matrixSetter;
- matrixSetter.template getCompressedRowsLengths< typename Mesh::Cell >( mesh,
- differentialOperatorPointer,
- boundaryConditionPointer,
- rowLengths );
- matrix->setDimensions( dofs, dofs );
- if( ! matrix->setCompressedRowsLengths( *rowLengths ) )
- return false;
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-makeSnapshot( const RealType& time,
- const IndexType& step,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData )
-{
- std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
- this->bindDofs( mesh, dofs );
- FileName fileName;
- fileName.setFileNameBase( "u-" );
- fileName.setExtension( "tnl" );
- fileName.setIndex( step );
- if( ! dofs->save( fileName.getFileName() ) )
- return false;
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getExplicitRHS( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& _u,
- DofVectorPointer& _fu,
- MeshDependentDataPointer& meshDependentData )
-{
- /****
- * If you use an explicit solver like Euler or Merson, you
- * need to implement this method. Compute the right-hand side of
- *
- * d/dt u(x) = fu( x, u )
- *
- * You may use supporting mesh dependent data if you need.
- */
- typedef typename MeshType::Cell Cell;
- int count = ::sqrt(mesh->template getEntitiesCount< Cell >());
-// const RealType& size = parameters.getParameter< double >( "realSize" ) / ::pow(count, 0.5);
-/* if (this->velocityType == "rotation")
- {
- double radius;
- for (int i =1; i < count; i++)
- for (int j =1; j < count; j++)
- {
- radius = ::sqrt(pow(i-1-(count/2.0),2) + ::pow(j-1-(count/2.0),2));
- if (radius != 0.0)
- _fu[(i-1)*count+j-1] =(0.25*tau)*differentialOperator.artificalViscosity* //smoothening part
- (_u[(i-1)*count-2+j]+_u[(i-1)*count+j]+
- _u[i*count+j-1]+_u[(i-2)*count+j-1]-
- 4.0*_u[(i-1)*count+j-1])
- -((1.0/(2.0*count))*differentialOperator.advectionSpeedX //X addition
- *radius*(-1)*((j-1-(count/2.0))/radius)
- *(_u[(i-1)*count+j]-_u[(i-1)*count+j-2]))
- -((1.0/(2.0*count))*differentialOperator.advectionSpeedY //Y addition
- *radius*((i-1-(count/2.0))/radius)
- *(_u[i*count+j-1]-_u[(i-2)*count+j-1]))
- ;}
- }
- else if (this->velocityType == "advection")
-*/ {
- this->bindDofs( mesh, _u );
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, DifferentialOperator, BoundaryCondition, RightHandSide > explicitUpdater;
- SharedPointer< MeshFunctionType > u( mesh, _u );
- SharedPointer< MeshFunctionType > fu( mesh, _fu );
- differentialOperatorPointer->setTau(tau);
- explicitUpdater.template update< typename Mesh::Cell >( time,
- mesh,
- this->differentialOperatorPointer,
- this->boundaryConditionPointer,
- this->rightHandSidePointer,
- u,
- fu );
-/* BoundaryConditionsSetter< MeshFunctionType, BoundaryCondition > boundaryConditionsSetter;
- boundaryConditionsSetter.template apply< typename Mesh::Cell >(
- this->boundaryCondition,
- time + tau,
- u ); */
- }
-}
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
- template< typename Matrix >
-void
-advectionProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-assemblyLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& _u,
- Matrix& matrix,
- DofVectorPointer& b,
- MeshDependentDataPointer& meshDependentData )
-{
- /*LinearSystemAssembler< Mesh,
- MeshFunctionType,
- DifferentialOperator,
- BoundaryCondition,
- RightHandSide,
- BackwardTimeDiscretisation,
- Matrix,
- DofVectorType > systemAssembler;
-
- MeshFunction< Mesh > u( mesh, _u );
- systemAssembler.template assembly< typename Mesh::Cell >( time,
- tau,
- mesh,
- this->differentialOperator,
- this->boundaryCondition,
- this->rightHandSide,
- u,
- matrix,
- b );*/
-}
-
-} // namespace TNL
-
-#endif /* advectionPROBLEM_IMPL_H_ */
diff --git a/examples/advection/advectionRhs.h b/examples/advection/advectionRhs.h
deleted file mode 100644
index 1ddb255a09906deece99fbf31df54659f3740a92..0000000000000000000000000000000000000000
--- a/examples/advection/advectionRhs.h
+++ /dev/null
@@ -1,33 +0,0 @@
-#pragma once
-
-#include <TNL/Functions/Domain.h>
-
-namespace TNL {
-
-template< typename Mesh, typename Real >class advectionRhs
- : public Functions::Domain< Mesh::meshDimensions, Functions::MeshDomain >
-{
- public:
-
- typedef Mesh MeshType;
- typedef Real RealType;
-
- bool setup( const Config::ParameterContainer& parameters,
- const String& prefix = "" )
- {
- return true;
- }
-
- template< typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshEntity& entity,
- const Real& time = 0.0 ) const
- {
- typedef typename MeshEntity::MeshType::VertexType VertexType;
- VertexType v = entity.getCenter();
- return 0.0;
- }
-};
-
-} // namespace TNL
-
diff --git a/examples/advection/tnl-run-advection b/examples/advection/tnl-run-advection
deleted file mode 100644
index 4ce2c6be3f3aea3c49e3619155cc489335bb2e25..0000000000000000000000000000000000000000
--- a/examples/advection/tnl-run-advection
+++ /dev/null
@@ -1,20 +0,0 @@
-#!/usr/bin/env bash
-
-tnl-grid-setup --dimensions 1 \
- --origin-x 0.0 \
- --proportions-x 10.0 \
- --size-x 500 \
-
-#tnl-init --test-function sin-wave \
-# --output-file init.tnl
-tnl-advection-dbg --time-discretisation explicit \
- --time-step 1.0e-3 \
- --boundary-conditions-constant 0 \
- --discrete-solver merson \
- --snapshot-period 0.1 \
- --final-time 1.0 \
- --artifical-viscosity 1.0 \
- --begin sin \
- --advection-speedX 2.0 \
-
-tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/advection/tnl-run-advection1 b/examples/advection/tnl-run-advection1
deleted file mode 100644
index fb5e81faf2e19ab8e6b19c02c0c441dd96a9c807..0000000000000000000000000000000000000000
--- a/examples/advection/tnl-run-advection1
+++ /dev/null
@@ -1,24 +0,0 @@
-#!/usr/bin/env bash
-
-tnl-grid-setup --dimensions 2 \
- --origin-x 0.0 \
- --origin-y 0.0 \
- --proportions-x 10.0 \
- --proportions-y 10.0 \
- --size-x 500 \
- --size-y 500 \
-
-#tnl-init --test-function sin-wave \
-# --output-file init.tnl
-./advection --time-discretisation explicit \
- --time-step 1.0e-3 \
- --boundary-conditions-constant 0 \
- --discrete-solver euler \
- --snapshot-period 0.1 \
- --final-time 1.0 \
- --artifical-viscosity 0.2 \
- --begin sin_square \
- --advection-speedX 2.0 \
- --advection-speedY 2.0 \
-
-tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/advection/tnl-run-advectionrot1 b/examples/advection/tnl-run-advectionrot1
deleted file mode 100644
index 5d3e3990a98ef64233a6b629b4002cc50b26a923..0000000000000000000000000000000000000000
--- a/examples/advection/tnl-run-advectionrot1
+++ /dev/null
@@ -1,25 +0,0 @@
-#!/usr/bin/env bash
-
-tnl-grid-setup --dimensions 2 \
- --origin-x 0.0 \
- --origin-y 0.0 \
- --proportions-x 10.0 \
- --proportions-y 10.0 \
- --size-x 100 \
- --size-y 100 \
-
-#tnl-init --test-function sin-wave \
-# --output-file init.tnl
-./advection --time-discretisation explicit \
- --time-step 1.0e-2 \
- --boundary-conditions-constant 0 \
- --discrete-solver euler \
- --snapshot-period 0.1 \
- --final-time 1.0 \
- --artifical-viscosity 5.0 \
- --begin sin_square \
- --advection-speedX 2.0 \
- --advection-speedY 2.0 \
- --move rotation \
-
-tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/heat-equation/tnl-heat-equation.h b/examples/heat-equation/tnl-heat-equation.h
index bb9de072e09465d93ba8bacdda947cfc474c1438..382ecdb3ba247c4aa4746b78b4752555f91c6100 100644
--- a/examples/heat-equation/tnl-heat-equation.h
+++ b/examples/heat-equation/tnl-heat-equation.h
@@ -46,7 +46,7 @@ class heatEquationConfig
config.addEntry< String >( "boundary-conditions-file", "File with the values of the boundary conditions.", "boundary.tnl" );
config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions." );
config.addEntry< double >( "right-hand-side-constant", "This sets a constant value for the right-hand side.", 0.0 );
- //config.addEntry< String >( "initial-condition", "File with the initial condition.", "initial.tnl");
+ config.addEntry< String >( "initial-condition", "File with the initial condition.", "initial.tnl");
};
};
diff --git a/examples/inviscid-flow/1d/CMakeLists.txt b/examples/inviscid-flow/1d/CMakeLists.txt
deleted file mode 100644
index 6fe0766eec90a76d7888c56f615122804abf4bcc..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/CMakeLists.txt
+++ /dev/null
@@ -1,20 +0,0 @@
-set( tnl_heat_equation_SOURCES
- euler.cpp
- euler.cu )
-
-IF( BUILD_CUDA )
- CUDA_ADD_EXECUTABLE(tnl-euler-1d${debugExt} euler.cu)
- target_link_libraries (tnl-euler-1d${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(tnl-euler-1d${debugExt} euler.cpp)
- target_link_libraries (tnl-euler-1d${debugExt} tnl${debugExt}-${tnlVersion} )
-ENDIF( BUILD_CUDA )
-
-
-INSTALL( TARGETS tnl-euler-1d${debugExt}
- RUNTIME DESTINATION bin
- PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
-
-INSTALL( FILES tnl-run-euler-1d
- ${tnl_heat_equation_SOURCES}
- DESTINATION share/tnl-${tnlVersion}/examples/inviscid-flow-1d )
diff --git a/examples/inviscid-flow/1d/EulerPressureGetter.h b/examples/inviscid-flow/1d/EulerPressureGetter.h
deleted file mode 100644
index fc98847a276c165b9352a42767e70e43ec5bcd98..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/EulerPressureGetter.h
+++ /dev/null
@@ -1,63 +0,0 @@
-#ifndef EulerPressureGetter_H
-#define EulerPressureGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-#include <TNL/Functions/Domain.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerPressureGetter
-: public Functions::Domain< Mesh::getMeshDimensions(), Functions::MeshDomain >
-{
- public:
-
- typedef Mesh MeshType;
- typedef typename MeshType::DeviceType DeviceType;
- typedef Real RealType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
-
- EulerPressureGetter( const MeshFunctionType& velocity,
- const MeshFunctionType& rhoVel,
- const MeshFunctionType& energy,
- const RealType& gamma )
- : rho( rho ), rhoVel( rhoVel ), energy( energy ), gamma( gamma )
- {}
-
- template< typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshEntity& entity,
- const RealType& time = 0.0 ) const
- {
- return this->operator[]( entity.getIndex() );
- }
-
- __cuda_callable__
- Real operator[]( const IndexType& idx ) const
- {
- return ( this->gamma - 1.0 ) * ( this->energy[ idx ] - 0.5 * this->rhoVel[ idx ] * this->rhoVel[ idx ] / this->rho[ idx ]);
- }
-
-
- protected:
-
- Real gamma;
-
- const MeshFunctionType& rho;
-
- const MeshFunctionType& rhoVel;
-
- const MeshFunctionType& energy;
-
-};
-
-} //namespace TNL
-
-#endif /* EulerPressureGetter_H */
diff --git a/examples/inviscid-flow/1d/EulerVelGetter.h b/examples/inviscid-flow/1d/EulerVelGetter.h
deleted file mode 100644
index 7e58ae77d6058d27c6e21169603ed09efe68f089..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/EulerVelGetter.h
+++ /dev/null
@@ -1,56 +0,0 @@
-#ifndef EulerVelGetter_H
-#define EulerVelGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerVelGetter
-: public Functions::Domain< Mesh::getMeshDimensions(), Functions::MeshDomain >
-{
- public:
-
- typedef Mesh MeshType;
- typedef typename MeshType::DeviceType DeviceType;
- typedef Real RealType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
-
- EulerVelGetter( const MeshFunctionType& rho,
- const MeshFunctionType& rhoVel)
- : rho( rho ), rhoVel( rhoVel )
- {}
-
- template< typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshEntity& entity,
- const RealType& time = 0.0 ) const
- {
- return this->operator[]( entity.getIndex() );
- }
-
- __cuda_callable__
- Real operator[]( const IndexType& idx ) const
- {
- return this->rho[ idx ] / this->rhoVel[ idx ];
- }
-
-
- protected:
-
- const MeshFunctionType& rho;
-
- const MeshFunctionType& rhoVel;
-
-};
-
-} // namespace TNL
-
-#endif /* EulerVelGetter_H */
diff --git a/examples/inviscid-flow/1d/LaxFridrichs.h b/examples/inviscid-flow/1d/LaxFridrichs.h
deleted file mode 100644
index a5cb7eea5c6183e07facebb0750d02b8fc04d694..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichs.h
+++ /dev/null
@@ -1,36 +0,0 @@
-#ifndef LaxFridrichs_H
-#define LaxFridrichs_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-#include "LaxFridrichsContinuity.h"
-#include "LaxFridrichsMomentum.h"
-#include "LaxFridrichsEnergy.h"
-#include "EulerVelGetter.h"
-#include "EulerPressureGetter.h"
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichs
-{
- public:
- typedef Real RealType;
- typedef typename Mesh::DeviceType DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< Mesh > MeshFunctionType;
-
- typedef LaxFridrichsContinuity< Mesh, Real, Index > Continuity;
- typedef LaxFridrichsMomentum< Mesh, Real, Index > Momentum;
- typedef LaxFridrichsEnergy< Mesh, Real, Index > Energy;
- typedef EulerVelGetter< Mesh, Real, Index > Velocity;
- typedef EulerPressureGetter< Mesh, Real, Index > Pressure;
-
-};
-
-} // namespace TNL
-
-#endif /* LaxFridrichs_H */
diff --git a/examples/inviscid-flow/1d/LaxFridrichsContinuity.h b/examples/inviscid-flow/1d/LaxFridrichsContinuity.h
deleted file mode 100644
index 8053895eda0787a52d12335ad4d9a1acc87e114a..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsContinuity.h
+++ /dev/null
@@ -1,184 +0,0 @@
-#ifndef LaxFridrichsContinuity_H
-#define LaxFridrichsContinuity_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsContinuity
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType velocity)
- {
- //this->velocity = velocity;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-#include "LaxFridrichsContinuity_impl.h"
-
-#endif /* LaxFridrichsContinuity_H */
diff --git a/examples/inviscid-flow/1d/LaxFridrichsContinuity_impl.h b/examples/inviscid-flow/1d/LaxFridrichsContinuity_impl.h
deleted file mode 100644
index 3be1a5b6e4cbcdafcab3cf81c55ac51304123a8e..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsContinuity_impl.h
+++ /dev/null
@@ -1,345 +0,0 @@
-#ifndef LaxFridrichsContinuity_IMPL_H
-#define LaxFridrichsContinuity_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
-
- //rho
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return (0.5 * this->tau) * ( u[ west ] - 2.0 * u[ center ] + u[ east ] )
- - 0.5 * hxInverse * ( u[ west ] * this->velocity[ west ] - u[ east ] * this->velocity[ east ] );
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} // namespace TNL
-
-#endif /* LaxFridrichsContinuityIMPL_H */
-
diff --git a/examples/inviscid-flow/1d/LaxFridrichsEnergy.h b/examples/inviscid-flow/1d/LaxFridrichsEnergy.h
deleted file mode 100644
index fa0012eb8d006aa452b702c4d0d38eed6b04ccdd..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsEnergy.h
+++ /dev/null
@@ -1,200 +0,0 @@
-#ifndef LaxFridrichsEnergy_H
-#define LaxFridrichsEnergy_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsEnergy
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-#include "LaxFridrichsEnergy_impl.h"
-
-#endif /* LaxFridrichsEnergy_H */
diff --git a/examples/inviscid-flow/1d/LaxFridrichsEnergy_impl.h b/examples/inviscid-flow/1d/LaxFridrichsEnergy_impl.h
deleted file mode 100644
index ba8092e807211da1470873252ad269c98d69a990..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsEnergy_impl.h
+++ /dev/null
@@ -1,352 +0,0 @@
-#ifndef LaxFridrichsEnergy_IMPL_H
-#define LaxFridrichsEnergy_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
-
- //energy
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return (0.5 * this->tau) * ( u[ west ] - 2.0 * u[ center ] + u[ east ] )
- - 0.5 * hxInverse *
- (( u[ west ] + this->pressure[ west ] ) * velocity[ west ]
- - (u[ east ] + this->pressure[ east ] ) * velocity[ east ] );
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* LaxFridrichsEnergyIMPL_H */
-
diff --git a/examples/inviscid-flow/1d/LaxFridrichsMomentum.h b/examples/inviscid-flow/1d/LaxFridrichsMomentum.h
deleted file mode 100644
index 5bb8818833b32f14415c4d2c107dfe0025d44a7e..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsMomentum.h
+++ /dev/null
@@ -1,201 +0,0 @@
-#ifndef LaxFridrichsMomentum_H
-#define LaxFridrichsMomentum_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsMomentum
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentum< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentum< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentum< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocity;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-
-#include "LaxFridrichsMomentum_impl.h"
-
-#endif /* LaxFridrichsMomentum_H */
diff --git a/examples/inviscid-flow/1d/LaxFridrichsMomentum_impl.h b/examples/inviscid-flow/1d/LaxFridrichsMomentum_impl.h
deleted file mode 100644
index 75c7a7fea080bc13201c19dd843f017b8052e4bc..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/LaxFridrichsMomentum_impl.h
+++ /dev/null
@@ -1,347 +0,0 @@
-#ifndef LaxFridrichsMomentum_IMPL_H
-#define LaxFridrichsMomentum_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentum< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentum< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentum< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
-
- //rhoVelocity
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return (0.5 * this->tau) * ( u[ west ] - 2.0 * u[ center ] + u[ east ] )
- - 0.5 * hxInverse *
- (( u[ west ] * this -> velocity[ west ] + this -> pressure [ west ] )
- - (u[ east ] * this -> velocity[ east ] + this -> pressure [ east ] ));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentum< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentum< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentum< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentum< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentum< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentum< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentum< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentum< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentum< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentum< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentum< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentum< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} // namespace TNL
-
-#endif /* LaxFridrichsMomentumIMPL_H */
-
diff --git a/examples/inviscid-flow/1d/eulerProblem_impl.h b/examples/inviscid-flow/1d/eulerProblem_impl.h
deleted file mode 100644
index 0ed9d4f1a5da37c32faa7fc0e41ef4bc92e84521..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/eulerProblem_impl.h
+++ /dev/null
@@ -1,383 +0,0 @@
-#ifndef eulerPROBLEM_IMPL_H_
-#define eulerPROBLEM_IMPL_H_
-
-#include <TNL/FileName.h>
-#include <TNL/Matrices/MatrixSetter.h>
-#include <TNL/Solvers/PDE/ExplicitUpdater.h>
-#include <TNL/Solvers/PDE/LinearSystemAssembler.h>
-#include <TNL/Solvers/PDE/BackwardTimeDiscretisation.h>
-#include "LaxFridrichsContinuity.h"
-#include "LaxFridrichsMomentum.h"
-#include "LaxFridrichsEnergy.h"
-#include "EulerVelGetter.h"
-#include "EulerPressureGetter.h"
-
-namespace TNL {
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-String
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getTypeStatic()
-{
- return String( "eulerProblem< " ) + Mesh :: getTypeStatic() + " >";
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-String
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getPrologHeader() const
-{
- return String( "euler" );
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-writeProlog( Logger& logger, const Config::ParameterContainer& parameters ) const
-{
- /****
- * Add data you want to have in the computation report (log) as follows:
- * logger.writeParameter< double >( "Parameter description", parameter );
- */
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setup( const MeshPointer& meshPointer,
- const Config::ParameterContainer& parameters,
- const String& prefix )
-{
- if( ! this->boundaryConditionsPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) ||
- ! this->rightHandSidePointer->setup( parameters, prefix + "right-hand-side-" ) )
- return false;
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-typename eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::IndexType
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getDofs( const MeshPointer& mesh ) const
-{
- /****
- * Return number of DOFs (degrees of freedom) i.e. number
- * of unknowns to be resolved by the main solver.
- */
- return 3*mesh->template getEntitiesCount< typename MeshType::Cell >();
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-bindDofs( const MeshPointer& mesh,
- DofVectorPointer& dofVector )
-{
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setInitialCondition( const Config::ParameterContainer& parameters,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData )
-{
- std::cout << std::endl << "get conditions from CML";
- typedef typename MeshType::Cell Cell;
- this->gamma = parameters.getParameter< RealType >( "gamma" );
- RealType rhoL = parameters.getParameter< RealType >( "left-density" );
- RealType velL = parameters.getParameter< RealType >( "left-velocity" );
- RealType preL = parameters.getParameter< RealType >( "left-pressure" );
- RealType eL = ( preL / (gamma - 1) ) + 0.5 * rhoL * velL * velL;
- RealType rhoR = parameters.getParameter< RealType >( "right-density" );
- RealType velR = parameters.getParameter< RealType >( "right-velocity" );
- RealType preR = parameters.getParameter< RealType >( "right-pressure" );
- RealType eR = ( preR / (gamma - 1) ) + 0.5 * rhoR * velR * velR;
- RealType x0 = parameters.getParameter< RealType >( "riemann-border" );
- std::cout << std::endl << gamma << " " << rhoL << " " << velL << " " << preL << " " << eL << " " << rhoR << " " << velR << " " << preR << " " << eR << " " << x0 << " " << gamma << std::endl;
- int count = mesh->template getEntitiesCount< Cell >();
- std::cout << count << std::endl;
- uRho->bind( mesh, *dofs, 0);
- uRhoVelocity->bind( mesh, *dofs, count);
- uEnergy->bind( mesh, *dofs, 2 * count);
- Containers::Vector < RealType, DeviceType, IndexType > data;
- data.setSize(2*count);
- velocity->bind( mesh, data, 0);
- pressure->bind( mesh, data, count );
- std::cout << std::endl << "set conditions from CML"<< std::endl;
- for(IndexType i = 0; i < count; i++)
- if (i < x0 * count )
- {
- ( *uRho )[i] = rhoL;
- ( *uRhoVelocity )[i] = rhoL * velL;
- ( *uEnergy )[i] = eL;
- ( *velocity )[i] = velL;
- ( *pressure )[i] = preL;
- }
- else
- {
- ( *uRho )[i] = rhoR;
- ( *uRhoVelocity )[i] = rhoR * velR;
- ( *uEnergy )[i] = eR;
- ( *velocity )[i] = velR;
- ( *pressure )[i] = preR;
- };
- std::cout << "dofs = " << *dofs << std::endl;
- getchar();
-
-
- /*
- const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );
- if( ! dofs.load( initialConditionFile ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << initialConditionFile << "." << std::endl;
- return false;
- }
- */
- return true;
-}
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
- template< typename Matrix >
-bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-setupLinearSystem( const MeshPointer& mesh,
- Matrix& matrix )
-{
-/* const IndexType dofs = this->getDofs( mesh );
- typedef typename Matrix::CompressedRowsLengthsVector CompressedRowsLengthsVectorType;
- CompressedRowsLengthsVectorType rowLengths;
- if( ! rowLengths.setSize( dofs ) )
- return false;
- MatrixSetter< MeshType, DifferentialOperator, BoundaryCondition, CompressedRowsLengthsVectorType > matrixSetter;
- matrixSetter.template getCompressedRowsLengths< typename Mesh::Cell >( mesh,
- differentialOperator,
- boundaryCondition,
- rowLengths );
- matrix.setDimensions( dofs, dofs );
- if( ! matrix.setCompressedRowsLengths( rowLengths ) )
- return false;*/
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-makeSnapshot( const RealType& time,
- const IndexType& step,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData )
-{
- std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
- this->bindDofs( mesh, dofs );
- typedef typename MeshType::Cell Cell;
- int count = mesh->template getEntitiesCount< Cell >();
- std::ofstream vysledek;
-/* std::cout << "pressure:" << std::endl;
- for (IndexType i = 0; i<count; i++)std::cout << this->pressure[i] << " " << i ;
- vysledek.open("pressure" + to_string(step) + ".txt");
- for (IndexType i = 0; i<count; i++)
- vysledek << 0.01*i << " " << pressure[i] << std::endl;
- vysledek.close();
- std::cout << " " << std::endl;
- std::cout << "velocity:" << std::endl;
- for (IndexType i = 0; i<count; i++)std::cout << this->velocity[i] << " " ;
- vysledek.open("velocity" + to_string(step) + ".txt");
- for (IndexType i = 0; i<count; i++)
- vysledek << 0.01*i << " " << pressure[i] << std::endl;
- vysledek.close();
- std::cout << "energy:" << std::endl;
- for (IndexType i = 0; i<count; i++)std::cout << this->uEnergy[i] << " " ;
- vysledek.open("energy" + to_string(step) + ".txt");
- for (IndexType i = 0; i<count; i++)
- vysledek << 0.01*i << " " << uEnergy[i] << std::endl;
- vysledek.close();
- std::cout << " " << std::endl;
- std::cout << "density:" << std::endl;
- for (IndexType i = 0; i<count; i++)std::cout << this->uRho[i] << " " ;
- vysledek.open("density" + to_string(step) + ".txt");
- for (IndexType i = 0; i<count; i++)
- vysledek << 0.01*i << " " << uRho[i] << std::endl;
- vysledek.close();
-*/ getchar();
-
- FileName fileName;
- fileName.setExtension( "tnl" );
- fileName.setIndex( step );
- fileName.setFileNameBase( "rho-" );
-
- if( ! uRho->save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "rhoVel-" );
- if( ! uRhoVelocity->save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "energy-" );
- if( ! uEnergy->save( fileName.getFileName() ) )
- return false;
- return true;
-}
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-getExplicitRHS( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& _u,
- DofVectorPointer& _fu,
- MeshDependentDataPointer& meshDependentData )
-{
- std::cout << "explicitRHS" << std::endl;
- typedef typename MeshType::Cell Cell;
- int count = mesh->template getEntitiesCount< Cell >();
- //bind _u
- this->uRho->bind(mesh, _u, 0);
- this->uRhoVelocity->bind(mesh, _u ,count);
- this->uEnergy->bind(mesh, _u, 2 * count);
-
- //bind _fu
- this->fuRho->bind(mesh, _u, 0);
- this->fuRhoVelocity->bind(mesh, _u, count);
- this->fuEnergy->bind(mesh, _u, 2 * count);
-
- //generating Differential operator object
- SharedPointer< Continuity > lF1DContinuity;
- SharedPointer< Momentum > lF1DMomentum;
- SharedPointer< Energy > lF1DEnergy;
-
-
-
- std::cout << "explicitRHSrho" << std::endl;
- //rho
- this->bindDofs( mesh, _u );
- lF1DContinuity->setTau(tau);
- lF1DContinuity->setVelocity( *velocity);
- /*ExplicitUpdater< Mesh, MeshFunctionType, Continuity, BoundaryCondition, RightHandSide > explicitUpdaterContinuity;
- explicitUpdaterContinuity.template update< typename Mesh::Cell >( time,
- mesh,
- lF1DContinuity,
- this->boundaryCondition,
- this->rightHandSide,
- uRho,
- fuRho );*/
-
- std::cout << "explicitRHSrhovel" << std::endl;
- //rhoVelocity
- lF1DMomentum->setTau(tau);
- lF1DMomentum->setVelocity( *velocity);
- lF1DMomentum->setPressure( *pressure);
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, Momentum, BoundaryCondition, RightHandSide > explicitUpdaterMomentum;
- explicitUpdaterMomentum.template update< typename Mesh::Cell >( time,
- mesh,
- lF1DMomentum,
- this->boundaryConditionsPointer,
- this->rightHandSidePointer,
- uRhoVelocity,
- fuRhoVelocity );
-
- std::cout << "explicitRHSenergy" << std::endl;
- //energy
- lF1DEnergy->setTau(tau);
- lF1DEnergy->setPressure( *pressure);
- lF1DEnergy->setVelocity( *velocity);
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, Energy, BoundaryCondition, RightHandSide > explicitUpdaterEnergy;
- explicitUpdaterEnergy.template update< typename Mesh::Cell >( time,
- mesh,
- lF1DEnergy,
- this->boundaryConditionsPointer,
- this->rightHandSidePointer,
- uEnergy,
- fuEnergy );
-
- }
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
- template< typename Matrix >
-void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-assemblyLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& _u,
- Matrix& matrix,
- DofVectorPointer& b,
- MeshDependentDataPointer& meshDependentData )
-{
-/* LinearSystemAssembler< Mesh,
- MeshFunctionType,
- DifferentialOperator,
- BoundaryCondition,
- RightHandSide,
- BackwardTimeDiscretisation,
- Matrix,
- DofVectorType > systemAssembler;
-
- MeshFunction< Mesh > u( mesh, _u );
- systemAssembler.template assembly< typename Mesh::Cell >( time,
- tau,
- mesh,
- this->differentialOperator,
- this->boundaryCondition,
- this->rightHandSide,
- u,
- matrix,
- b );*/
-}
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
-postIterate( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData )
-{
- //velocity
- this->velocity->setMesh( mesh );
- Velocity velocityGetter( *uRho, *uRhoVelocity );
- *this->velocity = velocityGetter;
- //pressure
- this->pressure->setMesh( mesh );
- Pressure pressureGetter( *uRho, *uRhoVelocity, *uEnergy, gamma );
- *this->pressure = pressureGetter;
-}
-
-} // namespace TNL
-
-#endif /* eulerPROBLEM_IMPL_H_ */
diff --git a/examples/inviscid-flow/1d/tnl-run-euler-1d b/examples/inviscid-flow/1d/tnl-run-euler-1d
deleted file mode 100644
index a45f664042d2f4dab646319e506dd130b6238cfc..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/1d/tnl-run-euler-1d
+++ /dev/null
@@ -1,25 +0,0 @@
-#!/usr/bin/env bash
-
-tnl-grid-setup --dimensions 1 \
- --origin-x 0.0 \
- --proportions-x 1.0 \
- --size-x 100
-
-#tnl-init --test-function sin-wave \
-# --output-file init.tnl
-tnl-euler-1d-dbg --time-discretisation explicit \
- --time-step 1.0e-3 \
- --boundary-conditions-constant 0 \
- --discrete-solver euler \
- --snapshot-period 0.1 \
- --final-time 1.0 \
- --left-density 1.0 \
- --left-velocity 0.75 \
- --left-pressure 1.0 \
- --right-density 0.125 \
- --right-velocity 0 \
- --right-pressure 0.1 \
- --gamma 1.4 \
- --riemann-border 0.3 \
-
-tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/inviscid-flow/2d/CMakeLists.txt b/examples/inviscid-flow/2d/CMakeLists.txt
deleted file mode 100644
index d753d50afcc11a2fccbe04e30410c77e8a5c1f81..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/CMakeLists.txt
+++ /dev/null
@@ -1,21 +0,0 @@
-set( tnl_heat_equation_SOURCES
- euler.cpp
- euler-cuda.cu )
-
-IF( BUILD_CUDA )
- CUDA_ADD_EXECUTABLE(tnl-euler-2d${debugExt} euler-cuda.cu)
- target_link_libraries (tnl-euler-2d${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(tnl-euler-2d${debugExt} euler.cpp)
- target_link_libraries (tnl-euler-2d${debugExt} tnl${debugExt}-${tnlVersion} )
-ENDIF( BUILD_CUDA )
-
-
-INSTALL( TARGETS tnl-euler-2d${debugExt}
- RUNTIME DESTINATION bin
- PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
-
-INSTALL( FILES run-euler
- ${tnl_heat_equation_SOURCES}
- DESTINATION share/tnl-${tnlVersion}/examples/inviscid-flow-2d )
-
diff --git a/examples/inviscid-flow/2d/EulerPressureGetter.h b/examples/inviscid-flow/2d/EulerPressureGetter.h
deleted file mode 100644
index 49e50d72576c270dc1f94f661aae70444ca00677..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerPressureGetter.h
+++ /dev/null
@@ -1,222 +0,0 @@
-#ifndef EulerPressureGetter_H
-#define EulerPressureGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerPressureGetter
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerPressureGetter< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real gamma;
- MeshFunctionType velocity;
- MeshFunctionType rho;
- MeshFunctionType energy;
-
- void setGamma(const Real& gamma)
- {
- this->gamma = gamma;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- void setEnergy(const MeshFunctionType& energy)
- {
- this->energy = energy;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerPressureGetter< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real gamma;
- MeshFunctionType velocity;
- MeshFunctionType rho;
- MeshFunctionType energy;
-
- void setGamma(const Real& gamma)
- {
- this->gamma = gamma;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- void setEnergy(const MeshFunctionType& energy)
- {
- this->energy = energy;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerPressureGetter< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real gamma;
- MeshFunctionType velocity;
- MeshFunctionType rho;
- MeshFunctionType energy;
-
- void setGamma(const Real& gamma)
- {
- this->gamma = gamma;
- };
-
- void setVelocity(const MeshFunctionType& velocity)
- {
- this->velocity = velocity;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- void setEnergy(const MeshFunctionType& energy)
- {
- this->energy = energy;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-
-} //namespace TNL
-
-#include "EulerPressureGetter_impl.h"
-
-#endif /* EulerPressureGetter_H */
diff --git a/examples/inviscid-flow/2d/EulerPressureGetter_impl.h b/examples/inviscid-flow/2d/EulerPressureGetter_impl.h
deleted file mode 100644
index 7d6b9310df3ddd19fbb261d7eb8dca00b72c82b9..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerPressureGetter_impl.h
+++ /dev/null
@@ -1,338 +0,0 @@
-#ifndef EulerPressureGetter_IMPL_H
-#define EulerPressureGetter_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerPressureGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerPressureGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerPressureGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerPressureGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerPressureGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerPressureGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerPressureGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerPressureGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- //pressure
- const IndexType& center = entity.getIndex();
-
- return ( this->gamma - 1 ) * ( energy[ center ] - 0.5 * rho[ center ] * ::pow( velocity[ center ],2 ));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerPressureGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerPressureGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerPressureGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerPressureGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerPressureGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerPressureGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerPressureGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} // namespace TNL
-
-#endif /* EulerPressureGetterIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/EulerVelGetter.h b/examples/inviscid-flow/2d/EulerVelGetter.h
deleted file mode 100644
index 2da0052b87492d6dd27a2b6def5fa3ead42318fa..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelGetter.h
+++ /dev/null
@@ -1,186 +0,0 @@
-#ifndef EulerVelGetter_H
-#define EulerVelGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-#include <TNL/Functions/Domain.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerVelGetter
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
- : public TNL::Functions::Domain< 1, TNL::Functions::MeshDomain >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType velX;
- MeshFunctionType velY;
-
- void setVelX(const MeshFunctionType& velX)
- {
- this->velX = velX;
- };
-
- void setVelY(const MeshFunctionType& velY)
- {
- this->velY = velY;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
- : public TNL::Functions::Domain< 2, TNL::Functions::MeshDomain >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType velX;
- MeshFunctionType velY;
-
- void setVelX(const MeshFunctionType& velX)
- {
- this->velX = velX;
- };
-
- void setVelY(const MeshFunctionType& velY)
- {
- this->velY = velY;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
- : public TNL::Functions::Domain< 3, TNL::Functions::MeshDomain >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType velX;
- MeshFunctionType velY;
-
- void setVelX(const MeshFunctionType& velX)
- {
- this->velX = velX;
- };
-
- void setVelY(const MeshFunctionType& velY)
- {
- this->velY = velY;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} //namespace TNL
-
-#include "EulerVelGetter_impl.h"
-
-#endif /* EulerVelGetter_H */
diff --git a/examples/inviscid-flow/2d/EulerVelGetter_impl.h b/examples/inviscid-flow/2d/EulerVelGetter_impl.h
deleted file mode 100644
index eeac07e57011bac7bd9769de63687f2f9bbc0852..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelGetter_impl.h
+++ /dev/null
@@ -1,340 +0,0 @@
-#ifndef EulerVelGetter_IMPL_H
-#define EulerVelGetter_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- //vel
- const IndexType& center = entity.getIndex();
- return ::sqrt(pow( velX[ center ],2)+pow( velY[ center ],2));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* EulerVelGetterIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/EulerVelXGetter.h b/examples/inviscid-flow/2d/EulerVelXGetter.h
deleted file mode 100644
index 93e3363bfab2f2d7eca6b375d9013c661bd47427..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelXGetter.h
+++ /dev/null
@@ -1,183 +0,0 @@
-#ifndef EulerVelXGetter_H
-#define EulerVelXGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerVelXGetter
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelXGetter< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelX;
- MeshFunctionType rho;
-
- void setRhoVelX(const MeshFunctionType& rhoVelX)
- {
- this->rhoVelX = rhoVelX;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelXGetter< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelX;
- MeshFunctionType rho;
-
- void setRhoVelX(const MeshFunctionType& rhoVelX)
- {
- this->rhoVelX = rhoVelX;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelXGetter< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelX;
- MeshFunctionType rho;
-
- void setRhoVelX(const MeshFunctionType& rhoVelX)
- {
- this->rhoVelX = rhoVelX;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-#include "EulerVelXGetter_impl.h"
-
-#endif /* EulerVelXGetter_H */
diff --git a/examples/inviscid-flow/2d/EulerVelXGetter_impl.h b/examples/inviscid-flow/2d/EulerVelXGetter_impl.h
deleted file mode 100644
index 6008509056afbadfbd99dddc3c0a57bad4185b28..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelXGetter_impl.h
+++ /dev/null
@@ -1,339 +0,0 @@
-#pragma once
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelXGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelXGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelXGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelXGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelXGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelXGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelXGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelXGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- //velX
- const IndexType& center = entity.getIndex();
- return ( rhoVelX[ center ] / rho[ center ]);
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelXGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelXGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelXGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelXGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelXGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelXGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelXGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-}// namespace TNL
-
-
-
diff --git a/examples/inviscid-flow/2d/EulerVelYGetter.h b/examples/inviscid-flow/2d/EulerVelYGetter.h
deleted file mode 100644
index 49474808d38d60baa3a0625ca690b953f714c5c7..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelYGetter.h
+++ /dev/null
@@ -1,183 +0,0 @@
-#ifndef EulerVelYGetter_H
-#define EulerVelYGetter_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class EulerVelYGetter
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelYGetter< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelY;
- MeshFunctionType rho;
-
- void setRhoVelY(const MeshFunctionType& rhoVelY)
- {
- this->rhoVelY = rhoVelY;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelYGetter< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelY;
- MeshFunctionType rho;
-
- void setRhoVelY(const MeshFunctionType& rhoVelY)
- {
- this->rhoVelY = rhoVelY;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class EulerVelYGetter< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- MeshFunctionType rhoVelY;
- MeshFunctionType rho;
-
- void setRhoVelY(const MeshFunctionType& rhoVelY)
- {
- this->rhoVelY = rhoVelY;
- };
-
- void setRho(const MeshFunctionType& rho)
- {
- this->rho = rho;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} //namespace TNL
-
-
-#include "EulerVelYGetter_impl.h"
-
-#endif /* EulerVelYGetter_H */
diff --git a/examples/inviscid-flow/2d/EulerVelYGetter_impl.h b/examples/inviscid-flow/2d/EulerVelYGetter_impl.h
deleted file mode 100644
index 337f4208c3ee6c0feef0e77e4d136b7704113671..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/EulerVelYGetter_impl.h
+++ /dev/null
@@ -1,332 +0,0 @@
-#ifndef EulerVelYGetter_IMPL_H
-#define EulerVelYGetter_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelYGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelYGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelYGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelYGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelYGetter< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelYGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelYGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelYGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- //velY
- const IndexType& center = entity.getIndex();
- return ( rhoVelY[ center ] / rho[ center ]);
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelYGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelYGetter< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-EulerVelYGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "EulerVelYGetter< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-EulerVelYGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-EulerVelYGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-EulerVelYGetter< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} // namespace TNL
-
-#endif /* EulerVelYGetterIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/LaxFridrichs.h b/examples/inviscid-flow/2d/LaxFridrichs.h
deleted file mode 100644
index b689e002122e0c885a53585ebe982f281cfbf43e..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichs.h
+++ /dev/null
@@ -1,42 +0,0 @@
-#ifndef LaxFridrichs_H
-#define LaxFridrichs_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-#include "LaxFridrichsContinuity.h"
-#include "LaxFridrichsEnergy.h"
-#include "LaxFridrichsMomentumX.h"
-#include "LaxFridrichsMomentumY.h"
-#include "EulerPressureGetter.h"
-#include "EulerVelXGetter.h"
-#include "EulerVelYGetter.h"
-#include "EulerVelGetter.h"
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichs
-{
- public:
- typedef Real RealType;
- typedef typename Mesh::DeviceType DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< Mesh > MeshFunctionType;
-
- typedef LaxFridrichsContinuity< Mesh, Real, Index > Continuity;
- typedef LaxFridrichsMomentumX< Mesh, Real, Index > MomentumX;
- typedef LaxFridrichsMomentumY< Mesh, Real, Index > MomentumY;
- typedef LaxFridrichsEnergy< Mesh, Real, Index > Energy;
- typedef EulerVelXGetter< Mesh, Real, Index > VelocityX;
- typedef EulerVelYGetter< Mesh, Real, Index > VelocityY;
- typedef EulerVelGetter< Mesh, Real, Index > Velocity;
- typedef EulerPressureGetter< Mesh, Real, Index > Pressure;
-
-};
-
-} //namespace TNL
-
-#endif /* LaxFridrichs_H */
diff --git a/examples/inviscid-flow/2d/LaxFridrichsContinuity.h b/examples/inviscid-flow/2d/LaxFridrichsContinuity.h
deleted file mode 100644
index 433d6a3512cbd8493ac16d0181d2255bb7692974..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsContinuity.h
+++ /dev/null
@@ -1,204 +0,0 @@
-#ifndef LaxFridrichsContinuity_H
-#define LaxFridrichsContinuity_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsContinuity
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsContinuity< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-
-} //namespace TNL
-
-#include "LaxFridrichsContinuity_impl .h"
-
-#endif /* LaxFridrichsContinuity_H */
diff --git a/examples/inviscid-flow/2d/LaxFridrichsContinuity_impl .h b/examples/inviscid-flow/2d/LaxFridrichsContinuity_impl .h
deleted file mode 100644
index 237b1a5845614a16ec708cf495ad057b297830b2..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsContinuity_impl .h
+++ /dev/null
@@ -1,349 +0,0 @@
-#ifndef LaxFridrichsContinuity_IMPL_H
-#define LaxFridrichsContinuity_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- //rho
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
- const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return (0.5 * this->tau) * ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4.0 * u[ center ] )
- - 0.5 * hxInverse * ( u[ west ] * this->velocityX[ west ] - u[ east ] * this->velocityX[ east ] )
- - 0.5 * hyInverse * ( u[ north ] * this->velocityY[ north ] - u[ south ] * this->velocityY[ east ] );
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsContinuity< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* LaxFridrichsContinuityIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/LaxFridrichsEnergy.h b/examples/inviscid-flow/2d/LaxFridrichsEnergy.h
deleted file mode 100644
index 8685697295d485d03c57e3748d2b6320e84f3ac5..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsEnergy.h
+++ /dev/null
@@ -1,219 +0,0 @@
-#ifndef LaxFridrichsEnergy_H
-#define LaxFridrichsEnergy_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsEnergy
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsEnergy< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} //namespace TNL
-
-
-#include "LaxFridrichsEnergy_impl.h"
-
-#endif /* LaxFridrichsEnergy_H */
diff --git a/examples/inviscid-flow/2d/LaxFridrichsEnergy_impl.h b/examples/inviscid-flow/2d/LaxFridrichsEnergy_impl.h
deleted file mode 100644
index e4f9b2e176d9acdb50157c6cf5f77a6c86801dd8..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsEnergy_impl.h
+++ /dev/null
@@ -1,350 +0,0 @@
-#ifndef LaxFridrichsEnergy_IMPL_H
-#define LaxFridrichsEnergy_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- //energy
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
- const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return (0.5 * this->tau) * ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4.0 * u[ center ] )
- - 0.5 * hxInverse * ((( u[ west ] + this->pressure[ west ] ) * this->velocityX[ west ] )
- -(( u[ east ] + this->pressure[ east ] ) * this->velocityX[ east ] ))
- - 0.5 * hyInverse * ((( u[ north ] + this->pressure[ north ] ) * this->velocityY[ north ] )
- -(( u[ south ] + this->pressure[ south ] ) * this->velocityY[ south ] ));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsEnergy< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* LaxFridrichsEnergyIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/LaxFridrichsMomentumX.h b/examples/inviscid-flow/2d/LaxFridrichsMomentumX.h
deleted file mode 100644
index b82758fc7c24cebfcc08580d73c0fd142a82dedc..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsMomentumX.h
+++ /dev/null
@@ -1,220 +0,0 @@
-#ifndef LaxFridrichsMomentumX_H
-#define LaxFridrichsMomentumX_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsMomentumX
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumX< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumX< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumX< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-
-} // namespace TNL
-
-#include "LaxFridrichsMomentumX_impl.h"
-
-#endif /* LaxFridrichsMomentumX_H */
diff --git a/examples/inviscid-flow/2d/LaxFridrichsMomentumX_impl.h b/examples/inviscid-flow/2d/LaxFridrichsMomentumX_impl.h
deleted file mode 100644
index bffb1fe02b4bdc505cb2214954a159f4b8f141bc..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsMomentumX_impl.h
+++ /dev/null
@@ -1,351 +0,0 @@
-#ifndef LaxFridrichsMomentumX_IMPL_H
-#define LaxFridrichsMomentumX_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumX< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumX< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumX< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumX< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumX< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumX< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumX< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumX< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- //rhoVelX
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
- const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return (0.5 * this->tau) * ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4.0 * u[ center ] )
- - 0.5 * hxInverse * (( u[ west ] * this->velocityX[ west ] + this->pressure[ west ] )
- -( u[ east ] * this->velocityX[ east ] + this->pressure[ east ] ))
- - 0.5 * hyInverse * (( u[ north ] * this->velocityY[ north ] )
- -( u[ south ] * this->velocityY[ south ] ));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumX< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumX< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumX< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumX< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumX< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumX< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumX< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* LaxFridrichsMomentumXIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/LaxFridrichsMomentumY.h b/examples/inviscid-flow/2d/LaxFridrichsMomentumY.h
deleted file mode 100644
index 309051829f7b973ed3009ae12eef2c8ec425f647..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsMomentumY.h
+++ /dev/null
@@ -1,218 +0,0 @@
-#ifndef LaxFridrichsMomentumY_H
-#define LaxFridrichsMomentumY_H
-
-#include <TNL/Containers/Vector.h>
-#include <TNL/Meshes/Grid.h>
-
-namespace TNL {
-
-template< typename Mesh,
- typename Real = typename Mesh::RealType,
- typename Index = typename Mesh::IndexType >
-class LaxFridrichsMomentumY
-{
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumY< Meshes::Grid< 1,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumY< Meshes::Grid< 2,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-class LaxFridrichsMomentumY< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
-{
- public:
- typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
- typedef typename MeshType::CoordinatesType CoordinatesType;
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
- typedef Functions::MeshFunction< MeshType > MeshFunctionType;
- enum { Dimensions = MeshType::getMeshDimensions() };
-
- static String getType();
- Real tau;
- MeshFunctionType velocityX;
- MeshFunctionType velocityY;
- MeshFunctionType pressure;
-
- void setTau(const Real& tau)
- {
- this->tau = tau;
- };
-
- void setVelocityX(const MeshFunctionType& velocityX)
- {
- this->velocityX = velocityX;
- };
-
- void setVelocityY(const MeshFunctionType& velocityY)
- {
- this->velocityY = velocityY;
- };
-
- void setPressure(const MeshFunctionType& pressure)
- {
- this->pressure = pressure;
- };
-
- template< typename MeshFunction, typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const RealType& time = 0.0 ) const;
-
- template< typename MeshEntity >
- __cuda_callable__
- Index getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const;
-
- template< typename MeshEntity, typename Vector, typename MatrixRow >
- __cuda_callable__
- void updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const;
-};
-
-} // namespace TNL
-
-#include "LaxFridrichsMomentumY_impl.h"
-
-#endif /* LaxFridrichsMomentumY_H */
diff --git a/examples/inviscid-flow/2d/LaxFridrichsMomentumY_impl.h b/examples/inviscid-flow/2d/LaxFridrichsMomentumY_impl.h
deleted file mode 100644
index 0624e19aa88750ceb7c78f940b287e515e34ee45..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/LaxFridrichsMomentumY_impl.h
+++ /dev/null
@@ -1,351 +0,0 @@
-#ifndef LaxFridrichsMomentumY_IMPL_H
-#define LaxFridrichsMomentumY_IMPL_H
-
-namespace TNL {
-
-/****
- * 1D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumY< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumY< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumY< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumY< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumY< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
- matrixRow.setElement( 0, west, - lambdaX );
- matrixRow.setElement( 1, center, 2.0 * lambdaX );
- matrixRow.setElement( 2, east, - lambdaX );
-}
-
-/****
- * 2D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumY< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumY< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumY< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
-
- //rhoVelY
- const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
- const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- return (0.5 * this->tau) * ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4.0 * u[ center ] )
- - 0.5 * hyInverse * (( u[ west ] * this->velocityX[ west ] )
- -( u[ east ] * this->velocityX[ east ] ))
- - 0.5 * hxInverse * (( u[ north ] * this->velocityY[ north ] + this->pressure[ north ] )
- -( u[ south ] * this->velocityY[ south ] + this->pressure[ south ]));
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumY< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumY< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
- matrixRow.setElement( 0, south, -lambdaY );
- matrixRow.setElement( 1, west, -lambdaX );
- matrixRow.setElement( 2, center, 2.0 * ( lambdaX + lambdaY ) );
- matrixRow.setElement( 3, east, -lambdaX );
- matrixRow.setElement( 4, north, -lambdaY );
-}
-
-/****
- * 3D problem
- */
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-String
-LaxFridrichsMomentumY< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getType()
-{
- return String( "LaxFridrichsMomentumY< " ) +
- MeshType::getType() + ", " +
- TNL::getType< Real >() + ", " +
- TNL::getType< Index >() + " >";
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshFunction, typename MeshEntity >
-__cuda_callable__
-Real
-LaxFridrichsMomentumY< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-operator()( const MeshFunction& u,
- const MeshEntity& entity,
- const Real& time ) const
-{
- /****
- * Implement your explicit form of the differential operator here.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
- static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
- static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
-
- const RealType& hxSquareInverse = entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& hySquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& hzSquareInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- return ( u[ west ] - 2.0 * u[ center ] + u[ east ] ) * hxSquareInverse +
- ( u[ south ] - 2.0 * u[ center ] + u[ north ] ) * hySquareInverse +
- ( u[ up ] - 2.0 * u[ center ] + u[ down ] ) * hzSquareInverse;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-template< typename MeshEntity >
-__cuda_callable__
-Index
-LaxFridrichsMomentumY< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-getLinearSystemRowLength( const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity ) const
-{
- /****
- * Return a number of non-zero elements in a line (associated with given grid element) of
- * the linear system.
- * The following example is the Laplace operator approximated
- * by the Finite difference method.
- */
-
- return 2*Dimensions + 1;
-}
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
- template< typename MeshEntity, typename Vector, typename MatrixRow >
-__cuda_callable__
-void
-LaxFridrichsMomentumY< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
-updateLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshType& mesh,
- const IndexType& index,
- const MeshEntity& entity,
- const MeshFunctionType& u,
- Vector& b,
- MatrixRow& matrixRow ) const
-{
- /****
- * Setup the non-zero elements of the linear system here.
- * The following example is the Laplace operator appriximated
- * by the Finite difference method.
- */
-
- const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
- const RealType& lambdaX = tau * entity.getMesh().template getSpaceStepsProducts< -2, 0, 0 >();
- const RealType& lambdaY = tau * entity.getMesh().template getSpaceStepsProducts< 0, -2, 0 >();
- const RealType& lambdaZ = tau * entity.getMesh().template getSpaceStepsProducts< 0, 0, -2 >();
- const IndexType& center = entity.getIndex();
- const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
- const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
- const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
- const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
- const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
- const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
- matrixRow.setElement( 0, down, -lambdaZ );
- matrixRow.setElement( 1, south, -lambdaY );
- matrixRow.setElement( 2, west, -lambdaX );
- matrixRow.setElement( 3, center, 2.0 * ( lambdaX + lambdaY + lambdaZ ) );
- matrixRow.setElement( 4, east, -lambdaX );
- matrixRow.setElement( 5, north, -lambdaY );
- matrixRow.setElement( 6, up, -lambdaZ );
-}
-
-} //namespace TNL
-
-#endif /* LaxFridrichsMomentumYIMPL_H */
-
diff --git a/examples/inviscid-flow/2d/euler-cuda.cu b/examples/inviscid-flow/2d/euler-cuda.cu
deleted file mode 100644
index 4d76005cb1f70724be978ff0fa6fec63c4a8a76f..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/euler-cuda.cu
+++ /dev/null
@@ -1 +0,0 @@
-#include "euler.h"
diff --git a/examples/inviscid-flow/2d/euler.cpp b/examples/inviscid-flow/2d/euler.cpp
deleted file mode 100644
index 4d76005cb1f70724be978ff0fa6fec63c4a8a76f..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/euler.cpp
+++ /dev/null
@@ -1 +0,0 @@
-#include "euler.h"
diff --git a/examples/inviscid-flow/2d/euler.h b/examples/inviscid-flow/2d/euler.h
deleted file mode 100644
index a888f928e86227dd3126c32408ba69fe46c89b7d..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/euler.h
+++ /dev/null
@@ -1,118 +0,0 @@
-#include <TNL/tnlConfig.h>
-#include <TNL/Solvers/Solver.h>
-#include <TNL/Solvers/BuildConfigTags.h>
-#include <TNL/Operators/DirichletBoundaryConditions.h>
-#include <TNL/Operators/NeumannBoundaryConditions.h>
-#include <TNL/Functions/Analytic/Constant.h>
-#include "eulerProblem.h"
-#include "LaxFridrichs.h"
-#include "eulerRhs.h"
-#include "eulerBuildConfigTag.h"
-
-using namespace TNL;
-
-typedef eulerBuildConfigTag BuildConfig;
-
-/****
- * Uncoment the following (and comment the previous line) for the complete build.
- * This will include support for all floating point precisions, all indexing types
- * and more solvers. You may then choose between them from the command line.
- * The compile time may, however, take tens of minutes or even several hours,
- * esppecially if CUDA is enabled. Use this, if you want, only for the final build,
- * not in the development phase.
- */
-//typedef tnlDefaultConfigTag BuildConfig;
-
-template< typename ConfigTag >class eulerConfig
-{
- public:
- static void configSetup( Config::ConfigDescription & config )
- {
- config.addDelimiter( "euler2D settings:" );
- config.addEntry< String >( "boundary-conditions-type", "Choose the boundary conditions type.", "dirichlet");
- config.addEntryEnum< String >( "dirichlet" );
- config.addEntryEnum< String >( "neumann" );
- config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions." );
- config.addEntry< double >( "left-density", "This sets a value of left density." );
- config.addEntry< double >( "left-velocityX", "This sets a value of left_x velocity." );
- config.addEntry< double >( "left-velocityY", "This sets a value of left_y velocity." );
- config.addEntry< double >( "left-pressure", "This sets a value of left pressure." );
- config.addEntry< double >( "riemann-border", "This sets a position of discontinuity." );
- config.addEntry< double >( "right-density", "This sets a value of right density." );
- config.addEntry< double >( "right-velocityX", "This sets a value of right_x velocity." );
- config.addEntry< double >( "right-velocityY", "This sets a value of right_y velocity." );
- config.addEntry< double >( "right-pressure", "This sets a value of right pressure." );
- config.addEntry< double >( "gamma", "This sets a value of gamma constant." );
-
- /****
- * Add definition of your solver command line arguments.
- */
-
- }
-};
-
-template< typename Real,
- typename Device,
- typename Index,
- typename MeshType,
- typename ConfigTag,
- typename SolverStarter >
-class eulerSetter
-{
- public:
-
- typedef Real RealType;
- typedef Device DeviceType;
- typedef Index IndexType;
-
- static bool run( const Config::ParameterContainer & parameters )
- {
- enum { Dimensions = MeshType::getMeshDimensions() };
- typedef LaxFridrichs< MeshType, Real, Index > ApproximateOperator;
- typedef eulerRhs< MeshType, Real > RightHandSide;
- typedef Containers::StaticVector < MeshType::getMeshDimensions(), Real > Vertex;
-
- /****
- * Resolve the template arguments of your solver here.
- * The following code is for the Dirichlet and the Neumann boundary conditions.
- * Both can be constant or defined as descrete values of Vector.
- */
- String boundaryConditionsType = parameters.getParameter< String >( "boundary-conditions-type" );
- if( parameters.checkParameter( "boundary-conditions-constant" ) )
- {
- typedef Functions::Analytic::Constant< Dimensions, Real > Constant;
- if( boundaryConditionsType == "dirichlet" )
- {
- typedef Operators::DirichletBoundaryConditions< MeshType, Constant, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
- typedef eulerProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Operators::NeumannBoundaryConditions< MeshType, Constant, Real, Index > BoundaryConditions;
- typedef eulerProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Functions::MeshFunction< MeshType > MeshFunction;
- if( boundaryConditionsType == "dirichlet" )
- {
- typedef Operators::DirichletBoundaryConditions< MeshType, MeshFunction, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
- typedef eulerProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
- typedef Operators::NeumannBoundaryConditions< MeshType, MeshFunction, Real, Index > BoundaryConditions;
- typedef eulerProblem< MeshType, BoundaryConditions, RightHandSide, ApproximateOperator > Problem;
- SolverStarter solverStarter;
- return solverStarter.template run< Problem >( parameters );
- }
-
-};
-
-int main( int argc, char* argv[] )
-{
- Solvers::Solver< eulerSetter, eulerConfig, BuildConfig > solver;
- if( ! solver. run( argc, argv ) )
- return EXIT_FAILURE;
- return EXIT_SUCCESS;
-}
diff --git a/examples/inviscid-flow/2d/eulerBuildConfigTag.h b/examples/inviscid-flow/2d/eulerBuildConfigTag.h
deleted file mode 100644
index fef4dfffce5400b1ef97786bc92b8a57ac246c1d..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/eulerBuildConfigTag.h
+++ /dev/null
@@ -1,51 +0,0 @@
-#ifndef eulerBUILDCONFIGTAG_H_
-#define eulerBUILDCONFIGTAG_H_
-
-#include <TNL/Solvers/BuildConfigTags.h>
-
-namespace TNL {
-
-class eulerBuildConfigTag{};
-
-namespace Solvers {
-
-/****
- * Turn off support for float and long double.
- */
-template<> struct ConfigTagReal< eulerBuildConfigTag, float > { enum { enabled = false }; };
-template<> struct ConfigTagReal< eulerBuildConfigTag, long double > { enum { enabled = false }; };
-
-/****
- * Turn off support for short int and long int indexing.
- */
-template<> struct ConfigTagIndex< eulerBuildConfigTag, short int >{ enum { enabled = false }; };
-template<> struct ConfigTagIndex< eulerBuildConfigTag, long int >{ enum { enabled = false }; };
-
-template< int Dimensions > struct ConfigTagDimensions< eulerBuildConfigTag, Dimensions >{ enum { enabled = ( Dimensions == 2 ) }; };
-
-/****
- * Use of Grid is enabled for allowed dimensions and Real, Device and Index types.
- */
-template< int Dimensions, typename Real, typename Device, typename Index >
- struct ConfigTagMesh< eulerBuildConfigTag, Meshes::Grid< Dimensions, Real, Device, Index > >
- { enum { enabled = ConfigTagDimensions< eulerBuildConfigTag, Dimensions >::enabled &&
- ConfigTagReal< eulerBuildConfigTag, Real >::enabled &&
- ConfigTagDevice< eulerBuildConfigTag, Device >::enabled &&
- ConfigTagIndex< eulerBuildConfigTag, Index >::enabled }; };
-
-/****
- * Please, chose your preferred time discretisation here.
- */
-template<> struct ConfigTagTimeDiscretisation< eulerBuildConfigTag, ExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
-template<> struct ConfigTagTimeDiscretisation< eulerBuildConfigTag, SemiImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
-template<> struct ConfigTagTimeDiscretisation< eulerBuildConfigTag, ImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
-
-/****
- * Only the Runge-Kutta-Merson solver is enabled by default.
- */
-template<> struct ConfigTagExplicitSolver< eulerBuildConfigTag, ExplicitEulerSolverTag >{ enum { enabled = false }; };
-
-} // namespace Solvers
-} // namespace TNL
-
-#endif /* eulerBUILDCONFIGTAG_H_ */
diff --git a/examples/inviscid-flow/2d/eulerProblem.h b/examples/inviscid-flow/2d/eulerProblem.h
deleted file mode 100644
index 71a9a7a7c1b71f96afd3c289258f27225337f854..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/eulerProblem.h
+++ /dev/null
@@ -1,135 +0,0 @@
-#pragma once
-
-#include <TNL/Problems/PDEProblem.h>
-#include <TNL/Functions/MeshFunction.h>
-
-using namespace TNL::Problems;
-
-namespace TNL {
-
-template< typename Mesh,
- typename BoundaryCondition,
- typename RightHandSide,
- typename DifferentialOperator >
-class eulerProblem:
- public PDEProblem< Mesh,
- typename DifferentialOperator::RealType,
- typename Mesh::DeviceType,
- typename DifferentialOperator::IndexType >
-{
- public:
-
- typedef typename DifferentialOperator::RealType RealType;
- typedef typename Mesh::DeviceType DeviceType;
- typedef typename DifferentialOperator::IndexType IndexType;
- typedef Functions::MeshFunction< Mesh > MeshFunctionType;
- typedef SharedPointer< MeshFunctionType, DeviceType > MeshFunctionPointer;
- typedef SharedPointer< DifferentialOperator > DifferentialOperatorPointer;
- typedef SharedPointer< BoundaryCondition > BoundaryConditionPointer;
- typedef SharedPointer< RightHandSide, DeviceType > RightHandSidePointer;
- typedef PDEProblem< Mesh, RealType, DeviceType, IndexType > BaseType;
-
- using typename BaseType::MeshType;
- using typename BaseType::MeshPointer;
- using typename BaseType::DofVectorType;
- using typename BaseType::DofVectorPointer;
- using typename BaseType::MeshDependentDataType;
- using typename BaseType::MeshDependentDataPointer;
-
- typedef typename DifferentialOperator::Continuity Continuity;
- typedef typename DifferentialOperator::MomentumX MomentumX;
- typedef typename DifferentialOperator::MomentumY MomentumY;
- typedef typename DifferentialOperator::Energy Energy;
- typedef typename DifferentialOperator::Velocity Velocity;
- typedef typename DifferentialOperator::VelocityX VelocityX;
- typedef typename DifferentialOperator::VelocityY VelocityY;
- typedef typename DifferentialOperator::Pressure Pressure;
-
- static String getTypeStatic();
-
- String getPrologHeader() const;
-
- void writeProlog( Logger& logger,
- const Config::ParameterContainer& parameters ) const;
-
- bool setup( const MeshPointer& meshPointer,
- const Config::ParameterContainer& parameters,
- const String& prefix = "" );
-
- bool setInitialCondition( const Config::ParameterContainer& parameters,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData );
-
- template< typename Matrix >
- bool setupLinearSystem( const MeshPointer& mesh,
- Matrix& matrix );
-
- bool makeSnapshot( const RealType& time,
- const IndexType& step,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData );
-
- IndexType getDofs( const MeshPointer& mesh ) const;
-
- void bindDofs( const MeshPointer& mesh,
- DofVectorPointer& dofs );
-
- void getExplicitRHS( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& _u,
- DofVectorPointer& _fu,
- MeshDependentDataPointer& meshDependentData );
-
- template< typename Matrix >
- void assemblyLinearSystem( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- Matrix& matrix,
- DofVectorPointer& rightHandSide,
- MeshDependentDataPointer& meshDependentData );
-
- bool postIterate( const RealType& time,
- const RealType& tau,
- const MeshPointer& mesh,
- DofVectorPointer& dofs,
- MeshDependentDataPointer& meshDependentData );
-
- protected:
-
- DifferentialOperatorPointer differentialOperatorPointer;
- BoundaryConditionPointer boundaryConditionPointer;
- RightHandSidePointer rightHandSidePointer;
-
- //definition
- Containers::Vector< RealType, DeviceType, IndexType > _uRho;
- Containers::Vector< RealType, DeviceType, IndexType > _uRhoVelocityX;
- Containers::Vector< RealType, DeviceType, IndexType > _uRhoVelocityY;
- Containers::Vector< RealType, DeviceType, IndexType > _uEnergy;
-
- Containers::Vector< RealType, DeviceType, IndexType > _fuRho;
- Containers::Vector< RealType, DeviceType, IndexType > _fuRhoVelocityX;
- Containers::Vector< RealType, DeviceType, IndexType > _fuRhoVelocityY;
- Containers::Vector< RealType, DeviceType, IndexType > _fuEnergy;
-
- Containers::Vector< RealType, DeviceType, IndexType > rho;
- Containers::Vector< RealType, DeviceType, IndexType > rhoVelX;
- Containers::Vector< RealType, DeviceType, IndexType > rhoVelY;
- Containers::Vector< RealType, DeviceType, IndexType > energy;
- Containers::Vector< RealType, DeviceType, IndexType > data;
- Containers::Vector< RealType, DeviceType, IndexType > pressure;
- Containers::Vector< RealType, DeviceType, IndexType > velocity;
- Containers::Vector< RealType, DeviceType, IndexType > velocityX;
- Containers::Vector< RealType, DeviceType, IndexType > velocityY;
- double gamma;
-
-
-};
-
-} // namespace TNL
-
-#include "eulerProblem_impl.h"
-
diff --git a/examples/inviscid-flow/2d/eulerRhs.h b/examples/inviscid-flow/2d/eulerRhs.h
deleted file mode 100644
index 1b46dc831fe9daa6133ff32ee4bea5faf4eb8d1c..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/eulerRhs.h
+++ /dev/null
@@ -1,35 +0,0 @@
-#ifndef eulerRHS_H_
-#define eulerRHS_H_
-
-#include <TNL/Functions/Domain.h>
-
-namespace TNL {
-
-template< typename Mesh, typename Real >class eulerRhs
- : public Functions::Domain< Mesh::meshDimensions, Functions::MeshDomain >
- {
- public:
-
- typedef Mesh MeshType;
- typedef Real RealType;
-
- bool setup( const Config::ParameterContainer& parameters,
- const String& prefix = "" )
- {
- return true;
- }
-
- template< typename MeshEntity >
- __cuda_callable__
- Real operator()( const MeshEntity& entity,
- const Real& time = 0.0 ) const
- {
- typedef typename MeshEntity::MeshType::VertexType VertexType;
- VertexType v = entity.getCenter();
- return 0.0;
- }
-};
-
-} //namespace TNL
-
-#endif /* eulerRHS_H_ */
diff --git a/examples/inviscid-flow/2d/run-euler b/examples/inviscid-flow/2d/run-euler
deleted file mode 100644
index f68b98a8406dea2f2142526283ec60248551c56d..0000000000000000000000000000000000000000
--- a/examples/inviscid-flow/2d/run-euler
+++ /dev/null
@@ -1,19 +0,0 @@
-#!/usr/bin/env bash
-
-tnl-grid-setup --dimensions 2 \
- --origin-x 0.0 \
- --origin-y 0.0 \
- --proportions-x 1.0 \
- --proportions-y 1.0 \
- --size-x 100 \
- --size-y 100
-
-tnl-init --test-function sin-wave \
- --output-file init.tnl
-./euler --time-discretisation explicit \
- --boundary-conditions-constant 0 \
- --discrete-solver merson \
- --snapshot-period 0.01 \
- --final-time 1.0
-
-tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/inviscid-flow/CMakeLists.txt b/examples/inviscid-flow/CMakeLists.txt
index 10af3c9c773bc36bca44191359b4a4e244d1faa4..634cce16141410c692254c2b7385b2cac5acd5a7 100644
--- a/examples/inviscid-flow/CMakeLists.txt
+++ b/examples/inviscid-flow/CMakeLists.txt
@@ -1,2 +1,24 @@
-add_subdirectory( 1d )
-add_subdirectory( 2d )
+set( tnl_inviscid_flow_HEADERS
+ CompressibleConservativeVariables.h )
+
+set( tnl_inviscid_flow_SOURCES
+ euler.cpp
+ euler.cu )
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(tnl-euler-2d${debugExt} euler.cu)
+ target_link_libraries (tnl-euler-2d${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(tnl-euler-2d${debugExt} euler.cpp)
+ target_link_libraries (tnl-euler-2d${debugExt} tnl${debugExt}-${tnlVersion} )
+ENDIF( BUILD_CUDA )
+
+
+INSTALL( TARGETS tnl-euler-2d${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+INSTALL( FILES run-euler
+ ${tnl_inviscid_flow_SOURCES}
+ DESTINATION share/tnl-${tnlVersion}/examples/inviscid-flow-2d )
+
diff --git a/examples/inviscid-flow/CompressibleConservativeVariables.h b/examples/inviscid-flow/CompressibleConservativeVariables.h
new file mode 100644
index 0000000000000000000000000000000000000000..04f1214999de5e6154131461098a061c2cfca311
--- /dev/null
+++ b/examples/inviscid-flow/CompressibleConservativeVariables.h
@@ -0,0 +1,147 @@
+/***************************************************************************
+ CompressibleConservativeVariables.h - description
+ -------------------
+ begin : Feb 12, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Functions/MeshFunction.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/SharedPointer.h>
+
+namespace TNL {
+
+template< typename Mesh >
+class CompressibleConservativeVariables
+{
+ public:
+ typedef Mesh MeshType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef typename MeshType::RealType RealType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::IndexType IndexType;
+ typedef Functions::MeshFunction< Mesh > MeshFunctionType;
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< MeshType > MeshPointer;
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef SharedPointer< VelocityFieldType > MomentumFieldPointer;
+
+ CompressibleConservativeVariables(){};
+
+ CompressibleConservativeVariables( const MeshPointer& meshPointer )
+ : density( meshPointer ),
+ momentum( meshPointer ),
+ //pressure( meshPointer ),
+ energy( meshPointer ){};
+
+ void setMesh( const MeshPointer& meshPointer )
+ {
+ this->density->setMesh( meshPointer );
+ this->momentum->setMesh( meshPointer );
+ //this->pressure.setMesh( meshPointer );
+ this->energy->setMesh( meshPointer );
+ }
+
+ template< typename Vector >
+ void bind( const MeshPointer& meshPointer,
+ const Vector& data,
+ IndexType offset = 0 )
+ {
+ IndexType currentOffset( offset );
+ this->density->bind( meshPointer, data, currentOffset );
+ currentOffset += this->density->getDofs( meshPointer );
+ for( IndexType i = 0; i < Dimensions; i++ )
+ {
+ ( *this->momentum )[ i ]->bind( meshPointer, data, currentOffset );
+ currentOffset += ( *this->momentum )[ i ]->getDofs( meshPointer );
+ }
+ this->energy->bind( meshPointer, data, currentOffset );
+ }
+
+ IndexType getDofs( const MeshPointer& meshPointer ) const
+ {
+ return this->density->getDofs( meshPointer ) +
+ this->momentum->getDofs( meshPointer ) +
+ this->energy->getDofs( meshPointer );
+ }
+
+ MeshFunctionPointer& getDensity()
+ {
+ return this->density;
+ }
+
+ const MeshFunctionPointer& getDensity() const
+ {
+ return this->density;
+ }
+
+ void setDensity( MeshFunctionPointer& density )
+ {
+ this->density = density;
+ }
+
+ MomentumFieldPointer& getMomentum()
+ {
+ return this->momentum;
+ }
+
+ const MomentumFieldPointer& getMomentum() const
+ {
+ return this->momentum;
+ }
+
+ void setMomentum( MomentumFieldPointer& momentum )
+ {
+ this->momentum = momentum;
+ }
+
+ /*MeshFunctionPointer& getPressure()
+ {
+ return this->pressure;
+ }
+
+ const MeshFunctionPointer& getPressure() const
+ {
+ return this->pressure;
+ }
+
+ void setPressure( MeshFunctionPointer& pressure )
+ {
+ this->pressure = pressure;
+ }*/
+
+ MeshFunctionPointer& getEnergy()
+ {
+ return this->energy;
+ }
+
+ const MeshFunctionPointer& getEnergy() const
+ {
+ return this->energy;
+ }
+
+ void setEnergy( MeshFunctionPointer& energy )
+ {
+ this->energy = energy;
+ }
+
+ void getVelocityField( VelocityFieldType& velocityField )
+ {
+
+ }
+
+ protected:
+
+ MeshFunctionPointer density;
+ MomentumFieldPointer momentum;
+ MeshFunctionPointer energy;
+
+};
+
+} // namespace TN
\ No newline at end of file
diff --git a/examples/inviscid-flow/LaxFridrichs.h b/examples/inviscid-flow/LaxFridrichs.h
new file mode 100644
index 0000000000000000000000000000000000000000..42ea19c145fd164d020af76b974c58020e8184e5
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichs.h
@@ -0,0 +1,141 @@
+/***************************************************************************
+ LaxFridrichs.h - description
+ -------------------
+ begin : Feb 18, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include <TNL/Functions/VectorField.h>
+
+#include "LaxFridrichsContinuity.h"
+#include "LaxFridrichsEnergy.h"
+#include "LaxFridrichsMomentumX.h"
+#include "LaxFridrichsMomentumY.h"
+#include "LaxFridrichsMomentumZ.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichs
+{
+ public:
+ typedef Mesh MeshType;
+ typedef Real RealType;
+ typedef typename Mesh::DeviceType DeviceType;
+ typedef Index IndexType;
+ typedef Functions::MeshFunction< Mesh > MeshFunctionType;
+ static const int Dimensions = Mesh::getMeshDimensions();
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VectorFieldType;
+
+ typedef LaxFridrichsContinuity< Mesh, Real, Index > ContinuityOperatorType;
+ typedef LaxFridrichsMomentumX< Mesh, Real, Index > MomentumXOperatorType;
+ typedef LaxFridrichsMomentumY< Mesh, Real, Index > MomentumYOperatorType;
+ typedef LaxFridrichsMomentumZ< Mesh, Real, Index > MomentumZOperatorType;
+ typedef LaxFridrichsEnergy< Mesh, Real, Index > EnergyOperatorType;
+
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef SharedPointer< VectorFieldType > VectorFieldPointer;
+ typedef SharedPointer< MeshType > MeshPointer;
+
+ typedef SharedPointer< ContinuityOperatorType > ContinuityOperatorPointer;
+ typedef SharedPointer< MomentumXOperatorType > MomentumXOperatorPointer;
+ typedef SharedPointer< MomentumYOperatorType > MomentumYOperatorPointer;
+ typedef SharedPointer< MomentumZOperatorType > MomentumZOperatorPointer;
+ typedef SharedPointer< EnergyOperatorType > EnergyOperatorPointer;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "numerical-viscosity", "Value of artificial (numerical) viscosity in the Lax-Fridrichs scheme", 1.0 );
+ }
+
+ LaxFridrichs()
+ : artificialViscosity( 1.0 ) {}
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->artificialViscosity = parameters.getParameter< double >( prefix + "numerical-viscosity" );
+ this->continuityOperatorPointer->setArtificialViscosity( artificialViscosity );
+ this->momentumXOperatorPointer->setArtificialViscosity( artificialViscosity );
+ this->momentumYOperatorPointer->setArtificialViscosity( artificialViscosity );
+ this->momentumZOperatorPointer->setArtificialViscosity( artificialViscosity );
+ this->energyOperatorPointer->setArtificialViscosity( artificialViscosity );
+
+ return true;
+ }
+
+ void setTau( const RealType& tau )
+ {
+ this->continuityOperatorPointer->setTau( tau );
+ this->momentumXOperatorPointer->setTau( tau );
+ this->momentumYOperatorPointer->setTau( tau );
+ this->momentumZOperatorPointer->setTau( tau );
+ this->energyOperatorPointer->setTau( tau );
+ }
+
+ void setPressure( const MeshFunctionPointer& pressure )
+ {
+ this->momentumXOperatorPointer->setPressure( pressure );
+ this->momentumYOperatorPointer->setPressure( pressure );
+ this->momentumZOperatorPointer->setPressure( pressure );
+ this->energyOperatorPointer->setPressure( pressure );
+ }
+
+ void setVelocity( const VectorFieldPointer& velocity )
+ {
+ this->continuityOperatorPointer->setVelocity( velocity );
+ this->momentumXOperatorPointer->setVelocity( velocity );
+ this->momentumYOperatorPointer->setVelocity( velocity );
+ this->momentumZOperatorPointer->setVelocity( velocity );
+ this->energyOperatorPointer->setVelocity( velocity );
+ }
+
+ const ContinuityOperatorPointer& getContinuityOperator() const
+ {
+ return this->continuityOperatorPointer;
+ }
+
+ const MomentumXOperatorPointer& getMomentumXOperator() const
+ {
+ return this->momentumXOperatorPointer;
+ }
+
+ const MomentumYOperatorPointer& getMomentumYOperator() const
+ {
+ return this->momentumYOperatorPointer;
+ }
+
+ const MomentumZOperatorPointer& getMomentumZOperator() const
+ {
+ return this->momentumZOperatorPointer;
+ }
+
+ const EnergyOperatorPointer& getEnergyOperator() const
+ {
+ return this->energyOperatorPointer;
+ }
+
+ protected:
+
+ ContinuityOperatorPointer continuityOperatorPointer;
+ MomentumXOperatorPointer momentumXOperatorPointer;
+ MomentumYOperatorPointer momentumYOperatorPointer;
+ MomentumZOperatorPointer momentumZOperatorPointer;
+ EnergyOperatorPointer energyOperatorPointer;
+
+ RealType artificialViscosity;
+};
+
+} //namespace TNL
diff --git a/examples/inviscid-flow/LaxFridrichsContinuity.h b/examples/inviscid-flow/LaxFridrichsContinuity.h
new file mode 100644
index 0000000000000000000000000000000000000000..9cc5ddc7f0ec7ec3aca8452eaecea7187d632963
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsContinuity.h
@@ -0,0 +1,288 @@
+/***************************************************************************
+ LaxFridrichsContinuity.h - description
+ -------------------
+ begin : Feb 17, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/SharedPointer.h>
+
+namespace TNL {
+
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsContinuityBase
+{
+ public:
+
+ typedef Real RealType;
+ typedef Index IndexType;
+ typedef Mesh MeshType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType > VelocityFieldPointer;
+
+ LaxFridrichsContinuityBase()
+ : artificialViscosity( 1.0 ){};
+
+ static String getType()
+ {
+ return String( "LaxFridrichsContinuity< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocity( const VelocityFieldPointer& velocity )
+ {
+ this->velocity = velocity;
+ };
+
+ void setArtificialViscosity( const RealType& artificialViscosity )
+ {
+ this->artificialViscosity = artificialViscosity;
+ }
+
+
+ protected:
+
+ RealType tau;
+
+ VelocityFieldPointer velocity;
+
+ RealType artificialViscosity;
+};
+
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsContinuity
+{
+};
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsContinuity< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsContinuityBase< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsContinuityBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ return 1.0 / 2.0 * this->tau * this->artificialViscosity * ( u[ west ] - 2.0 * u[ center ] + u[ east ] )
+ - 0.5 * ( u[ west ] * velocity_x_west - u[ east ] * velocity_x_east ) * hxInverse;
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsContinuity< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsContinuityBase< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsContinuityBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+ //rho
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+
+ return 1.0 / 4.0 * this->tau * this->artificialViscosity * ( u[ west ] + u[ east ] + u[ south ] + u[ north ] - 4.0 * u[ center ] )
+ - 0.5 * ( ( u[ west ] * velocity_x_west - u[ east ] * velocity_x_east ) * hxInverse
+ + ( u[ north ] * velocity_y_north - u[ south ] * velocity_y_south ) * hyInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsContinuity< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsContinuityBase< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsContinuityBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ //rho
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+ const RealType& velocity_z_up = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ up ];
+ const RealType& velocity_z_down = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ down ];
+
+ return 1.0 / 6.0 * this->tau * this->artificialViscosity *
+ ( u[ west ] + u[ east ] + u[ south ] + u[ north ] + u[ up ] + u[ down ]- 6.0 * u[ center ] )
+ - 0.5 * ( ( u[ west ] * velocity_x_west - u[ east ] * velocity_x_east ) * hxInverse
+ + ( u[ north ] * velocity_y_north - u[ south ] * velocity_y_south ) * hyInverse
+ + ( u[ up ] * velocity_z_up - u[ down ] * velocity_z_down ) * hzInverse );
+
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+
+} //namespace TNL
diff --git a/examples/inviscid-flow/LaxFridrichsEnergy.h b/examples/inviscid-flow/LaxFridrichsEnergy.h
new file mode 100644
index 0000000000000000000000000000000000000000..df1570f19ca13f32bcb408c83a658a5e8f0cad81
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsEnergy.h
@@ -0,0 +1,309 @@
+/***************************************************************************
+ LaxFridrichsEnergy.h - description
+ -------------------
+ begin : Feb 17, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsEnergyBase
+{
+ public:
+
+ typedef Real RealType;
+ typedef Index IndexType;
+ typedef Mesh MeshType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef SharedPointer< VelocityFieldType > VelocityFieldPointer;
+
+ LaxFridrichsEnergyBase()
+ : artificialViscosity( 1.0 ){};
+
+ static String getType()
+ {
+ return String( "LaxFridrichsEnergy< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocity( const VelocityFieldPointer& velocity )
+ {
+ this->velocity = velocity;
+ };
+
+ void setPressure( const MeshFunctionPointer& pressure )
+ {
+ this->pressure = pressure;
+ };
+
+ void setArtificialViscosity( const RealType& artificialViscosity )
+ {
+ this->artificialViscosity = artificialViscosity;
+ }
+
+ protected:
+
+ RealType tau;
+
+ VelocityFieldPointer velocity;
+
+ MeshFunctionPointer pressure;
+
+ RealType artificialViscosity;
+};
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsEnergy
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsEnergy< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsEnergyBase< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsEnergyBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& e,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ return 1.0 / 2.0 * this->tau * this->artificialViscosity * ( e[ west ] - 2.0 * e[ center ] + e[ east ] )
+ - 0.5 * ( ( e[ west ] + pressure_west ) * velocity_x_west
+ - ( e[ east ] + pressure_east ) * velocity_x_east ) * hxInverse;
+
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsEnergy< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsEnergyBase< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsEnergyBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& e,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ const RealType& pressure_north = this->pressure.template getData< DeviceType >()[ north ];
+ const RealType& pressure_south = this->pressure.template getData< DeviceType >()[ south ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+
+ return 1.0 / 4.0 * this->tau * this->artificialViscosity * ( e[ west ] + e[ east ] + e[ south ] + e[ north ] - 4.0 * e[ center ] )
+ - 0.5 * ( ( ( ( e[ west ] + pressure_west ) * velocity_x_west )
+ -( ( e[ east ] + pressure_east ) * velocity_x_east ) ) * hxInverse
+ + ( ( ( e[ north ] + pressure_north ) * velocity_y_north )
+ -( ( e[ south ] + pressure_south ) * velocity_y_south ) ) * hyInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsEnergy< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsEnergyBase< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsEnergyBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& e,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ const RealType& pressure_north = this->pressure.template getData< DeviceType >()[ north ];
+ const RealType& pressure_south = this->pressure.template getData< DeviceType >()[ south ];
+ const RealType& pressure_up = this->pressure.template getData< DeviceType >()[ up ];
+ const RealType& pressure_down = this->pressure.template getData< DeviceType >()[ down ];
+
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+ const RealType& velocity_z_up = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ up ];
+ const RealType& velocity_z_down = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ down ];
+
+ return 1.0 / 6.0 * this->tau * this->artificialViscosity *
+ ( e[ west ] + e[ east ] + e[ south ] + e[ north ] + e[ up ] + e[ down ] - 6.0 * e[ center ] )
+ - 0.5 * ( ( ( ( e[ west ] + pressure_west ) * velocity_x_west )
+ -( ( e[ east ] + pressure_east ) * velocity_x_east ) ) * hxInverse
+ + ( ( ( e[ north ] + pressure_north ) * velocity_y_north )
+ -( ( e[ south ] + pressure_south ) * velocity_y_south ) ) * hyInverse
+ + ( ( ( e[ up ] + pressure_up ) * velocity_z_up )
+ -( ( e[ down ] + pressure_down ) * velocity_z_down ) ) * hzInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+} //namespace TNL
diff --git a/examples/inviscid-flow/LaxFridrichsMomentumBase.h b/examples/inviscid-flow/LaxFridrichsMomentumBase.h
new file mode 100644
index 0000000000000000000000000000000000000000..b3eb777b890cca2eb22b7a988edfa4d0dde5c956
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsMomentumBase.h
@@ -0,0 +1,68 @@
+/***************************************************************************
+ LaxFridrichsMomentumBase.h - description
+ -------------------
+ begin : Feb 17, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsMomentumBase
+{
+ public:
+
+ typedef Real RealType;
+ typedef Index IndexType;
+ typedef Mesh MeshType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef SharedPointer< VelocityFieldType > VelocityFieldPointer;
+
+ LaxFridrichsMomentumBase()
+ : artificialViscosity( 1.0 ){};
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocity( const VelocityFieldPointer& velocity )
+ {
+ this->velocity = velocity;
+ };
+
+ void setPressure( const MeshFunctionPointer& pressure )
+ {
+ this->pressure = pressure;
+ };
+
+ void setArtificialViscosity( const RealType& artificialViscosity )
+ {
+ this->artificialViscosity = artificialViscosity;
+ }
+
+ protected:
+
+ RealType tau;
+
+ VelocityFieldPointer velocity;
+
+ MeshFunctionPointer pressure;
+
+ RealType artificialViscosity;
+};
+
+} //namespace TNL
diff --git a/examples/inviscid-flow/LaxFridrichsMomentumX.h b/examples/inviscid-flow/LaxFridrichsMomentumX.h
new file mode 100644
index 0000000000000000000000000000000000000000..c4f004fa4eb447f06bc5e0b95e65be50f6347f8b
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsMomentumX.h
@@ -0,0 +1,276 @@
+/***************************************************************************
+ LaxFridrichsMomentumX.h - description
+ -------------------
+ begin : Feb 18, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include "LaxFridrichsMomentumBase.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsMomentumX
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumX< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumX< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+
+ return 1.0 / 2.0 * this->tau * this->artificialViscosity * ( rho_u[ west ] + rho_u[ east ] - 2.0 * rho_u[ center ] )
+ - 0.5 * ( ( rho_u[ west ] * velocity_x_west + pressure_west )
+ -( rho_u[ east ] * velocity_x_east + pressure_east ) ) * hxInverse;
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumX< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumX< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
+
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+
+ return 1.0 / 4.0 * this->tau * this->artificialViscosity * ( rho_u[ west ] + rho_u[ east ] + rho_u[ south ] + rho_u[ north ] - 4.0 * rho_u[ center ] )
+ - 0.5 * ( ( ( rho_u[ west ] * velocity_x_west + pressure_west )
+ - ( rho_u[ east ] * velocity_x_east + pressure_east ) ) * hxInverse
+ + ( ( rho_u[ north ] * velocity_y_north )
+ - ( rho_u[ south ] * velocity_y_south ) ) * hyInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumX< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumX< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ const RealType& pressure_west = this->pressure.template getData< DeviceType >()[ west ];
+ const RealType& pressure_east = this->pressure.template getData< DeviceType >()[ east ];
+ //const RealType& pressure_north = this->pressure.template getData< DeviceType >()[ north ];
+ //const RealType& pressure_south = this->pressure.template getData< DeviceType >()[ south ];
+ //const RealType& pressure_up = this->pressure.template getData< DeviceType >()[ up ];
+ //const RealType& pressure_down = this->pressure.template getData< DeviceType >()[ down ];
+
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+ const RealType& velocity_z_up = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ up ];
+ const RealType& velocity_z_down = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ down ];
+ return 1.0 / 6.0 * this->tau * this->artificialViscosity *
+ ( rho_u[ west ] + rho_u[ east ] + rho_u[ south ] + rho_u[ north ] + rho_u[ up ] + rho_u[ down ] - 6.0 * rho_u[ center ] )
+ - 0.5 * ( ( ( rho_u[ west ] * velocity_x_west + pressure_west )
+ - ( rho_u[ east ] * velocity_x_east + pressure_east ) )* hxInverse
+ + ( ( rho_u[ north ] * velocity_y_north )
+ - ( rho_u[ south ] * velocity_y_south ) )* hyInverse
+ + ( ( rho_u[ up ] * velocity_z_up )
+ - ( rho_u[ down ] * velocity_z_down ) )* hzInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+
+} // namespace TNL
+
diff --git a/examples/inviscid-flow/LaxFridrichsMomentumY.h b/examples/inviscid-flow/LaxFridrichsMomentumY.h
new file mode 100644
index 0000000000000000000000000000000000000000..2b931be883e81dbb066c404d22d292482e85ff14
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsMomentumY.h
@@ -0,0 +1,260 @@
+/***************************************************************************
+ LaxFridrichsMomentumY.h - description
+ -------------------
+ begin : Feb 18, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include "LaxFridrichsMomentumBase.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsMomentumY
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumY< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumY< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_v,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ //const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ return 0.0;
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumY< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumY< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_v,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
+
+ const RealType& pressure_north = this->pressure.template getData< DeviceType >()[ north ];
+ const RealType& pressure_south = this->pressure.template getData< DeviceType >()[ south ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+
+ return 1.0 / 4.0 * this->tau * this->artificialViscosity * ( rho_v[ west ] + rho_v[ east ] + rho_v[ south ] + rho_v[ north ] - 4.0 * rho_v[ center ] )
+ - 0.5 * ( ( ( rho_v[ west ] * velocity_x_west )
+ - ( rho_v[ east ] * velocity_x_east ) )* hxInverse
+ + ( ( rho_v[ north ] * velocity_y_north + pressure_north )
+ - ( rho_v[ south ] * velocity_y_south + pressure_south ) )* hyInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumY< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumY< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_v,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ const RealType& pressure_north = this->pressure.template getData< DeviceType >()[ north ];
+ const RealType& pressure_south = this->pressure.template getData< DeviceType >()[ south ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+ const RealType& velocity_z_up = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ up ];
+ const RealType& velocity_z_down = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ down ];
+ return 1.0 / 6.0 * this->tau * this->artificialViscosity *
+ ( rho_v[ west ] + rho_v[ east ] + rho_v[ south ] + rho_v[ north ] + rho_v[ up ] + rho_v[ down ] - 6.0 * rho_v[ center ] )
+ - 0.5 * ( ( ( rho_v[ west ] * velocity_x_west )
+ - ( rho_v[ east ] * velocity_x_east ) ) * hxInverse
+ + ( ( rho_v[ north ] * velocity_y_north + pressure_north )
+ - ( rho_v[ south ] * velocity_y_south + pressure_south ) ) * hyInverse
+ + ( ( rho_v[ up ] * velocity_z_up )
+ - ( rho_v[ down ] * velocity_z_down ) ) * hzInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+
+} // namespace TNL
+
diff --git a/examples/inviscid-flow/LaxFridrichsMomentumZ.h b/examples/inviscid-flow/LaxFridrichsMomentumZ.h
new file mode 100644
index 0000000000000000000000000000000000000000..0ab7da1234570de3e5ab0edabafdca17ce8e719d
--- /dev/null
+++ b/examples/inviscid-flow/LaxFridrichsMomentumZ.h
@@ -0,0 +1,240 @@
+/***************************************************************************
+ LaxFridrichsMomentumZ.h - description
+ -------------------
+ begin : Feb 18, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include "LaxFridrichsMomentumBase.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class LaxFridrichsMomentumZ
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumZ< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumZ< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_w,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ //const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ return 0.0;
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumZ< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumZ< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_w,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ //const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+ return 0.0;
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class LaxFridrichsMomentumZ< Meshes::Grid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+ : public LaxFridrichsMomentumBase< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef LaxFridrichsMomentumBase< MeshType, Real, Index > BaseType;
+
+ using typename BaseType::RealType;
+ using typename BaseType::IndexType;
+ using typename BaseType::DeviceType;
+ using typename BaseType::CoordinatesType;
+ using typename BaseType::MeshFunctionType;
+ using typename BaseType::MeshFunctionPointer;
+ using typename BaseType::VelocityFieldType;
+ using typename BaseType::VelocityFieldPointer;
+ using BaseType::Dimensions;
+
+ static String getType()
+ {
+ return String( "LaxFridrichsMomentumZ< " ) +
+ MeshType::getType() + ", " +
+ TNL::getType< Real >() + ", " +
+ TNL::getType< Index >() + " >";
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& rho_w,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ const RealType& pressure_up = this->pressure.template getData< DeviceType >()[ up ];
+ const RealType& pressure_down = this->pressure.template getData< DeviceType >()[ down ];
+ const RealType& velocity_x_east = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ east ];
+ const RealType& velocity_x_west = this->velocity.template getData< DeviceType >()[ 0 ].template getData< DeviceType >()[ west ];
+ const RealType& velocity_y_north = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ north ];
+ const RealType& velocity_y_south = this->velocity.template getData< DeviceType >()[ 1 ].template getData< DeviceType >()[ south ];
+ const RealType& velocity_z_up = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ up ];
+ const RealType& velocity_z_down = this->velocity.template getData< DeviceType >()[ 2 ].template getData< DeviceType >()[ down ];
+ return 1.0 / 6.0 * this->tau * this->artificialViscosity *
+ ( rho_w[ west ] + rho_w[ east ] + rho_w[ south ] + rho_w[ north ] + rho_w[ up ] + rho_w[ down ] - 6.0 * rho_w[ center ] )
+ -0.5 * ( ( ( rho_w[ west ] * velocity_x_west )
+ - ( rho_w[ east ] * velocity_x_east ) )* hxInverse
+ + ( ( rho_w[ north ] * velocity_y_north )
+ - ( rho_w[ south ] * velocity_y_south ) )* hyInverse
+ + ( ( rho_w[ up ] * velocity_z_up + pressure_up )
+ - ( rho_w[ down ] * velocity_z_down + pressure_down ) )* hzInverse );
+ }
+
+ /*template< typename MeshEntity >
+ __cuda_callable__
+ Index getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity ) const;
+
+ template< typename MeshEntity, typename Vector, typename MatrixRow >
+ __cuda_callable__
+ void updateLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ const IndexType& index,
+ const MeshEntity& entity,
+ const MeshFunctionType& u,
+ Vector& b,
+ MatrixRow& matrixRow ) const;*/
+};
+
+
+} // namespace TNL
+
diff --git a/examples/inviscid-flow/PhysicalVariablesGetter.h b/examples/inviscid-flow/PhysicalVariablesGetter.h
new file mode 100644
index 0000000000000000000000000000000000000000..25108da0791e266f48f32e0ae0c7296d4e2ef18b
--- /dev/null
+++ b/examples/inviscid-flow/PhysicalVariablesGetter.h
@@ -0,0 +1,116 @@
+/***************************************************************************
+ CompressibleConservativeVariables.h - description
+ -------------------
+ begin : Feb 12, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/SharedPointer.h>
+#include <TNL/Functions/MeshFunction.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/Functions/MeshFunctionEvaluator.h>
+#include "CompressibleConservativeVariables.h"
+
+namespace TNL {
+
+template< typename Mesh >
+class PhysicalVariablesGetter
+{
+ public:
+
+ typedef Mesh MeshType;
+ typedef typename MeshType::RealType RealType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::IndexType IndexType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef CompressibleConservativeVariables< MeshType > ConservativeVariablesType;
+ typedef SharedPointer< ConservativeVariablesType > ConservativeVariablesPointer;
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType > VelocityFieldPointer;
+
+ class VelocityGetter : public Functions::Domain< Dimensions, Functions::MeshDomain >
+ {
+ public:
+ typedef typename MeshType::RealType RealType;
+
+ VelocityGetter( MeshFunctionPointer density,
+ MeshFunctionPointer momentum )
+ : density( density ), momentum( momentum ) {}
+
+ template< typename EntityType >
+ __cuda_callable__
+ RealType operator()( const EntityType& meshEntity,
+ const RealType& time = 0.0 ) const
+ {
+ return momentum.template getData< DeviceType >()( meshEntity ) /
+ density.template getData< DeviceType >()( meshEntity );
+ }
+
+ protected:
+ const MeshFunctionPointer density, momentum;
+ };
+
+ class PressureGetter : public Functions::Domain< Dimensions, Functions::MeshDomain >
+ {
+ public:
+ typedef typename MeshType::RealType RealType;
+
+ PressureGetter( MeshFunctionPointer density,
+ MeshFunctionPointer energy,
+ VelocityFieldPointer momentum,
+ const RealType& gamma )
+ : density( density ), energy( energy ), momentum( momentum ), gamma( gamma ) {}
+
+ template< typename EntityType >
+ __cuda_callable__
+ RealType operator()( const EntityType& meshEntity,
+ const RealType& time = 0.0 ) const
+ {
+ const RealType e = energy.template getData< DeviceType >()( meshEntity );
+ const RealType rho = density.template getData< DeviceType >()( meshEntity );
+ const RealType momentumNorm = momentum.template getData< DeviceType >().getVector( meshEntity ).lpNorm( 2.0 );
+ return ( gamma - 1.0 ) * ( e - 0.5 * momentumNorm * momentumNorm / rho );
+ }
+
+ protected:
+ const MeshFunctionPointer density, energy;
+ const VelocityFieldPointer momentum;
+ const RealType gamma;
+ };
+
+
+ void getVelocity( const ConservativeVariablesPointer& conservativeVariables,
+ VelocityFieldPointer& velocity )
+ {
+ Functions::MeshFunctionEvaluator< MeshFunctionType, VelocityGetter > evaluator;
+ for( int i = 0; i < Dimensions; i++ )
+ {
+ SharedPointer< VelocityGetter, DeviceType > velocityGetter( conservativeVariables->getDensity(),
+ ( *conservativeVariables->getMomentum() )[ i ] );
+ evaluator.evaluate( ( *velocity )[ i ], velocityGetter );
+ }
+ }
+
+ void getPressure( const ConservativeVariablesPointer& conservativeVariables,
+ const RealType& gamma,
+ MeshFunctionPointer& pressure )
+ {
+ Functions::MeshFunctionEvaluator< MeshFunctionType, PressureGetter > evaluator;
+ SharedPointer< PressureGetter, DeviceType > pressureGetter( conservativeVariables->getDensity(),
+ conservativeVariables->getEnergy(),
+ conservativeVariables->getMomentum(),
+ gamma );
+ evaluator.evaluate( pressure, pressureGetter );
+ }
+
+};
+
+} //namespace TNL
diff --git a/examples/inviscid-flow/RiemannProblemInitialCondition.h b/examples/inviscid-flow/RiemannProblemInitialCondition.h
new file mode 100644
index 0000000000000000000000000000000000000000..0de73687f12ef63d54026b9366687f794dc4d373
--- /dev/null
+++ b/examples/inviscid-flow/RiemannProblemInitialCondition.h
@@ -0,0 +1,209 @@
+/***************************************************************************
+ RiemannProblemInitialCondition.h - description
+ -------------------
+ begin : Feb 13, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Containers/StaticVector.h>
+#include <TNL/Operators/Analytic/Sign.h>
+#include <TNL/Functions/MeshFunctionEvaluator.h>
+#include <TNL/Operators/Analytic/Sign.h>
+#include "CompressibleConservativeVariables.h"
+
+namespace TNL {
+
+template< typename Mesh >
+class RiemannProblemInitialCondition
+{
+ public:
+
+ typedef Mesh MeshType;
+ typedef typename MeshType::RealType RealType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::IndexType IndexType;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef Containers::StaticVector< Dimensions, RealType > VertexType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ typedef SharedPointer< MeshFunctionType > MeshFunctionPointer;
+ typedef Functions::VectorField< Dimensions, MeshType > VectorFieldType;
+
+ RiemannProblemInitialCondition()
+ : discontinuityPlacement( 0.5 ),
+ leftDensity( 1.0 ), rightDensity( 1.0 ),
+ leftVelocity( -2.0 ), rightVelocity( 2.0 ),
+ leftPressure( 0.4 ), rightPressure( 0.4 ),
+ gamma( 1.67 ){}
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "discontinuity-placement-0", "x-coordinate of the discontinuity placement.", 0.5 );
+ config.addEntry< double >( prefix + "discontinuity-placement-1", "y-coordinate of the discontinuity placement.", 0.5 );
+ config.addEntry< double >( prefix + "discontinuity-placement-2", "z-coordinate of the discontinuity placement.", 0.5 );
+ config.addEntry< double >( prefix + "left-density", "Density on the left side of the discontinuity.", 1.0 );
+ config.addEntry< double >( prefix + "right-density", "Density on the right side of the discontinuity.", 0.0 );
+ config.addEntry< double >( prefix + "left-velocity-0", "x-coordinate of the velocity on the left side of the discontinuity.", 1.0 );
+ config.addEntry< double >( prefix + "left-velocity-1", "y-coordinate of the velocity on the left side of the discontinuity.", 1.0 );
+ config.addEntry< double >( prefix + "left-velocity-2", "z-coordinate of the velocity on the left side of the discontinuity.", 1.0 );
+ config.addEntry< double >( prefix + "right-velocity-0", "x-coordinate of the velocity on the right side of the discontinuity.", 0.0 );
+ config.addEntry< double >( prefix + "right-velocity-1", "y-coordinate of the velocity on the right side of the discontinuity.", 0.0 );
+ config.addEntry< double >( prefix + "right-velocity-2", "z-coordinate of the velocity on the right side of the discontinuity.", 0.0 );
+ config.addEntry< double >( prefix + "left-pressure", "Pressure on the left side of the discontinuity.", 1.0 );
+ config.addEntry< double >( prefix + "right-pressure", "Pressure on the right side of the discontinuity.", 0.0 );
+ config.addEntry< double >( prefix + "gamma", "Gamma in the ideal gas state equation.", 1.67 );
+ }
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->discontinuityPlacement.setup( parameters, prefix + "discontinuity-placement-" );
+ this->leftVelocity.setup( parameters, prefix + "left-velocity-" );
+ this->rightVelocity.setup( parameters, prefix + "right-velocity-" );
+ this->leftDensity = parameters.getParameter< double >( prefix + "left-density" );
+ this->rightDensity = parameters.getParameter< double >( prefix + "right-density" );
+ this->leftPressure = parameters.getParameter< double >( prefix + "left-pressure" );
+ this->rightPressure = parameters.getParameter< double >( prefix + "right-pressure" );
+ this->gamma = parameters.getParameter< double >( prefix + "gamma" );
+ return true;
+ };
+
+ void setDiscontinuityPlacement( const VertexType& v )
+ {
+ this->discontinuityPlacement = v;
+ }
+
+ const VertexType& getDiscontinuityPlasement() const
+ {
+ return this->discontinuityPlacement;
+ }
+
+ void setLeftDensity( const RealType& leftDensity )
+ {
+ this->leftDensity = leftDensity;
+ }
+
+ const RealType& getLeftDensity() const
+ {
+ return this->leftDensity;
+ }
+
+ void setRightDensity( const RealType& rightDensity )
+ {
+ this->rightDensity = rightDensity;
+ }
+
+ const RealType& getRightDensity() const
+ {
+ return this->rightDensity;
+ }
+
+ void setLeftVelocity( const VertexType& leftVelocity )
+ {
+ this->leftVelocity = leftVelocity;
+ }
+
+ const VertexType& getLeftVelocity() const
+ {
+ return this->leftVelocity;
+ }
+
+ void setRightVelocity( const RealType& rightVelocity )
+ {
+ this->rightVelocity = rightVelocity;
+ }
+
+ const VertexType& getRightVelocity() const
+ {
+ return this->rightVelocity;
+ }
+
+ void setLeftPressure( const RealType& leftPressure )
+ {
+ this->leftPressure = leftPressure;
+ }
+
+ const RealType& getLeftPressure() const
+ {
+ return this->leftPressure;
+ }
+
+ void setRightPressure( const RealType& rightPressure )
+ {
+ this->rightPressure = rightPressure;
+ }
+
+ const RealType& getRightPressure() const
+ {
+ return this->rightPressure;
+ }
+
+ void setInitialCondition( CompressibleConservativeVariables< MeshType >& conservativeVariables,
+ const VertexType& center = VertexType( 0.0 ) )
+ {
+ typedef Functions::Analytic::VectorNorm< Dimensions, RealType > VectorNormType;
+ typedef Operators::Analytic::Sign< Dimensions, RealType > SignType;
+ typedef Functions::OperatorFunction< SignType, VectorNormType > InitialConditionType;
+ typedef SharedPointer< InitialConditionType, DeviceType > InitialConditionPointer;
+
+ InitialConditionPointer initialCondition;
+ initialCondition->getFunction().setCenter( center );
+ initialCondition->getFunction().setMaxNorm( true );
+ initialCondition->getFunction().setRadius( discontinuityPlacement[ 0 ] );
+ discontinuityPlacement *= 1.0 / discontinuityPlacement[ 0 ];
+ for( int i = 1; i < Dimensions; i++ )
+ discontinuityPlacement[ i ] = 1.0 / discontinuityPlacement[ i ];
+ initialCondition->getFunction().setAnisotropy( discontinuityPlacement );
+ initialCondition->getFunction().setMultiplicator( -1.0 );
+
+ Functions::MeshFunctionEvaluator< MeshFunctionType, InitialConditionType > evaluator;
+
+ /****
+ * Density
+ */
+ initialCondition->getOperator().setPositiveValue( leftDensity );
+ initialCondition->getOperator().setNegativeValue( rightDensity );
+ evaluator.evaluate( conservativeVariables.getDensity(), initialCondition );
+ conservativeVariables.getDensity()->write( "density.gplt", "gnuplot" );
+
+ /****
+ * Momentum
+ */
+ for( int i = 0; i < Dimensions; i++ )
+ {
+ initialCondition->getOperator().setPositiveValue( leftDensity * leftVelocity[ i ] );
+ initialCondition->getOperator().setNegativeValue( rightDensity * rightVelocity[ i ] );
+ evaluator.evaluate( ( *conservativeVariables.getMomentum() )[ i ], initialCondition );
+ }
+
+ /****
+ * Energy
+ */
+ const RealType leftKineticEnergy = leftVelocity.lpNorm( 2.0 );
+ const RealType rightKineticEnergy = rightVelocity.lpNorm( 2.0 );
+ const RealType leftEnergy = leftPressure / ( gamma - 1.0 ) + 0.5 * leftDensity * leftKineticEnergy * leftKineticEnergy;
+ const RealType rightEnergy = rightPressure / ( gamma - 1.0 ) + 0.5 * rightDensity * rightKineticEnergy * rightKineticEnergy;
+ initialCondition->getOperator().setPositiveValue( leftEnergy );
+ initialCondition->getOperator().setNegativeValue( rightEnergy );
+ evaluator.evaluate( conservativeVariables.getEnergy(), initialCondition );
+ conservativeVariables.getEnergy()->write( "energy-init", "gnuplot" );
+ }
+
+
+ protected:
+
+ VertexType discontinuityPlacement;
+
+ RealType leftDensity, rightDensity;
+ VertexType leftVelocity, rightVelocity;
+ RealType leftPressure, rightPressure;
+
+ RealType gamma; // gamma in the ideal gas state equation
+};
+
+} //namespace TNL
\ No newline at end of file
diff --git a/examples/inviscid-flow/1d/euler.cpp b/examples/inviscid-flow/euler.cpp
similarity index 100%
rename from examples/inviscid-flow/1d/euler.cpp
rename to examples/inviscid-flow/euler.cpp
diff --git a/examples/inviscid-flow/1d/euler.cu b/examples/inviscid-flow/euler.cu
similarity index 100%
rename from examples/inviscid-flow/1d/euler.cu
rename to examples/inviscid-flow/euler.cu
diff --git a/examples/inviscid-flow/1d/euler.h b/examples/inviscid-flow/euler.h
similarity index 82%
rename from examples/inviscid-flow/1d/euler.h
rename to examples/inviscid-flow/euler.h
index 65bf55820659f8ce67d3a1cfd5622e1f558dab72..baa8941346311cd651c3f7443c11c56f2c690651 100644
--- a/examples/inviscid-flow/1d/euler.h
+++ b/examples/inviscid-flow/euler.h
@@ -6,20 +6,21 @@
#include <TNL/Functions/Analytic/Constant.h>
#include "eulerProblem.h"
#include "LaxFridrichs.h"
-
#include "eulerRhs.h"
#include "eulerBuildConfigTag.h"
+#include "RiemannProblemInitialCondition.h"
+
using namespace TNL;
typedef eulerBuildConfigTag BuildConfig;
/****
- * Uncoment the following (and comment the previous line) for the complete build.
+ * Uncomment the following (and comment the previous line) for the complete build.
* This will include support for all floating point precisions, all indexing types
* and more solvers. You may then choose between them from the command line.
* The compile time may, however, take tens of minutes or even several hours,
- * esppecially if CUDA is enabled. Use this, if you want, only for the final build,
+ * especially if CUDA is enabled. Use this, if you want, only for the final build,
* not in the development phase.
*/
//typedef tnlDefaultConfigTag BuildConfig;
@@ -29,19 +30,14 @@ template< typename ConfigTag >class eulerConfig
public:
static void configSetup( Config::ConfigDescription & config )
{
- config.addDelimiter( "euler settings:" );
+ config.addDelimiter( "Inviscid flow settings:" );
config.addEntry< String >( "boundary-conditions-type", "Choose the boundary conditions type.", "dirichlet");
config.addEntryEnum< String >( "dirichlet" );
config.addEntryEnum< String >( "neumann" );
config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions." );
- config.addEntry< double >( "left-density", "This sets a value of left density." );
- config.addEntry< double >( "left-velocity", "This sets a value of left velocity." );
- config.addEntry< double >( "left-pressure", "This sets a value of left pressure." );
- config.addEntry< double >( "riemann-border", "This sets a position of discontinuity." );
- config.addEntry< double >( "right-density", "This sets a value of right density." );
- config.addEntry< double >( "right-velocity", "This sets a value of right velocity." );
- config.addEntry< double >( "right-pressure", "This sets a value of right pressure." );
- config.addEntry< double >( "gamma", "This sets a value of gamma constant." );
+ typedef Meshes::Grid< 3 > Mesh;
+ LaxFridrichs< Mesh >::configSetup( config, "inviscid-operators-" );
+ RiemannProblemInitialCondition< Mesh >::configSetup( config );
/****
* Add definition of your solver command line arguments.
@@ -115,5 +111,3 @@ int main( int argc, char* argv[] )
return EXIT_FAILURE;
return EXIT_SUCCESS;
}
-
-
diff --git a/examples/inviscid-flow/1d/eulerBuildConfigTag.h b/examples/inviscid-flow/eulerBuildConfigTag.h
similarity index 96%
rename from examples/inviscid-flow/1d/eulerBuildConfigTag.h
rename to examples/inviscid-flow/eulerBuildConfigTag.h
index abede16343a81cde5e75386f8b2e15117fd6672d..2bb81e8cb3cd81772c4c5722ca9c5a518ffea25a 100644
--- a/examples/inviscid-flow/1d/eulerBuildConfigTag.h
+++ b/examples/inviscid-flow/eulerBuildConfigTag.h
@@ -21,7 +21,7 @@ template<> struct ConfigTagReal< eulerBuildConfigTag, long double > { enum { ena
template<> struct ConfigTagIndex< eulerBuildConfigTag, short int >{ enum { enabled = false }; };
template<> struct ConfigTagIndex< eulerBuildConfigTag, long int >{ enum { enabled = false }; };
-template< int Dimensions > struct ConfigTagDimensions< eulerBuildConfigTag, Dimensions >{ enum { enabled = ( Dimensions == 1 ) }; };
+template< int Dimensions > struct ConfigTagDimensions< eulerBuildConfigTag, Dimensions >{ enum { enabled = true }; };
/****
* Use of Grid is enabled for allowed dimensions and Real, Device and Index types.
diff --git a/examples/inviscid-flow/1d/eulerProblem.h b/examples/inviscid-flow/eulerProblem.h
similarity index 64%
rename from examples/inviscid-flow/1d/eulerProblem.h
rename to examples/inviscid-flow/eulerProblem.h
index f687ee5460accc7a593bb8f6e034872f8d8e6eda..79cff960a4e6801900cf43990f3cae0d23178ee2 100644
--- a/examples/inviscid-flow/1d/eulerProblem.h
+++ b/examples/inviscid-flow/eulerProblem.h
@@ -1,33 +1,40 @@
-#ifndef eulerPROBLEM_H_
-#define eulerPROBLEM_H_
+/***************************************************************************
+ eulerProblem.h - description
+ -------------------
+ begin : Feb 13, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
#include <TNL/Problems/PDEProblem.h>
#include <TNL/Functions/MeshFunction.h>
+#include "CompressibleConservativeVariables.h"
+
using namespace TNL::Problems;
+
namespace TNL {
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
class eulerProblem:
public PDEProblem< Mesh,
- typename DifferentialOperator::RealType,
- typename Mesh::DeviceType,
- typename DifferentialOperator::IndexType >
+ typename InviscidOperators::RealType,
+ typename Mesh::DeviceType,
+ typename InviscidOperators::IndexType >
{
public:
-
- typedef typename DifferentialOperator::RealType RealType;
+
+ typedef typename InviscidOperators::RealType RealType;
typedef typename Mesh::DeviceType DeviceType;
- typedef typename DifferentialOperator::IndexType IndexType;
- typedef Functions::MeshFunction< Mesh > MeshFunctionType;
- typedef SharedPointer< MeshFunctionType, DeviceType > MeshFunctionPointer;
- typedef SharedPointer< DifferentialOperator > DifferentialOperatorPointer;
- typedef SharedPointer< BoundaryCondition > BoundaryConditionPointer;
- typedef SharedPointer< RightHandSide, DeviceType > RightHandSidePointer;
- typedef PDEProblem< Mesh, RealType, DeviceType, IndexType > BaseType;
+ typedef typename InviscidOperators::IndexType IndexType;
+ typedef PDEProblem< Mesh, RealType, DeviceType, IndexType > BaseType;
using typename BaseType::MeshType;
using typename BaseType::MeshPointer;
@@ -36,13 +43,17 @@ class eulerProblem:
using typename BaseType::MeshDependentDataType;
using typename BaseType::MeshDependentDataPointer;
- typedef typename DifferentialOperator::Continuity Continuity;
- typedef typename DifferentialOperator::Momentum Momentum;
- typedef typename DifferentialOperator::Energy Energy;
- typedef typename DifferentialOperator::Velocity Velocity;
- typedef typename DifferentialOperator::Pressure Pressure;
-
+ static const int Dimensions = Mesh::getMeshDimensions();
+ typedef Functions::MeshFunction< Mesh > MeshFunctionType;
+ typedef CompressibleConservativeVariables< MeshType > ConservativeVariablesType;
+ typedef Functions::VectorField< Dimensions, MeshFunctionType > VelocityFieldType;
+ typedef SharedPointer< MeshFunctionType, DeviceType > MeshFunctionPointer;
+ typedef SharedPointer< ConservativeVariablesType > ConservativeVariablesPointer;
+ typedef SharedPointer< VelocityFieldType > VelocityFieldPointer;
+ typedef SharedPointer< InviscidOperators > InviscidOperatorsPointer;
+ typedef SharedPointer< BoundaryCondition > BoundaryConditionPointer;
+ typedef SharedPointer< RightHandSide, DeviceType > RightHandSidePointer;
static String getTypeStatic();
@@ -51,9 +62,9 @@ class eulerProblem:
void writeProlog( Logger& logger,
const Config::ParameterContainer& parameters ) const;
- bool setup( const MeshPointer& meshPointer,
+ bool setup( const MeshPointer& meshPointer,
const Config::ParameterContainer& parameters,
- const String& prefix );
+ const String& prefix = "" );
bool setInitialCondition( const Config::ParameterContainer& parameters,
const MeshPointer& mesh,
@@ -90,7 +101,7 @@ class eulerProblem:
Matrix& matrix,
DofVectorPointer& rightHandSide,
MeshDependentDataPointer& meshDependentData );
-
+
bool postIterate( const RealType& time,
const RealType& tau,
const MeshPointer& mesh,
@@ -99,22 +110,21 @@ class eulerProblem:
protected:
- DifferentialOperatorPointer differentialOperatorPointer;
- BoundaryConditionPointer boundaryConditionsPointer;
+ InviscidOperatorsPointer inviscidOperatorsPointer;
+
+ BoundaryConditionPointer boundaryConditionPointer;
RightHandSidePointer rightHandSidePointer;
- MeshFunctionPointer uRho, uRhoVelocity, uEnergy;
- MeshFunctionPointer fuRho, fuRhoVelocity, fuEnergy;
+ ConservativeVariablesPointer conservativeVariables,
+ conservativeVariablesRHS;
- MeshFunctionPointer pressure, velocity, rho, rhoVel, energy;
+ VelocityFieldPointer velocity;
+ MeshFunctionPointer pressure;
- RealType gamma;
-
-
+ RealType gamma;
};
-} // namepsace TNL
+} // namespace TNL
#include "eulerProblem_impl.h"
-#endif /* eulerPROBLEM_H_ */
diff --git a/examples/inviscid-flow/2d/eulerProblem_impl.h b/examples/inviscid-flow/eulerProblem_impl.h
similarity index 51%
rename from examples/inviscid-flow/2d/eulerProblem_impl.h
rename to examples/inviscid-flow/eulerProblem_impl.h
index 27d4aa4ae67b7cacf7fdfb13214349a8a996a874..08b12e344e7551f2c0f3b8dead00f8dbe8c2c891 100644
--- a/examples/inviscid-flow/2d/eulerProblem_impl.h
+++ b/examples/inviscid-flow/eulerProblem_impl.h
@@ -1,3 +1,13 @@
+/***************************************************************************
+ eulerProblem_impl.h - description
+ -------------------
+ begin : Feb 13, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
#pragma once
#include <TNL/FileName.h>
@@ -5,24 +15,27 @@
#include <TNL/Solvers/PDE/ExplicitUpdater.h>
#include <TNL/Solvers/PDE/LinearSystemAssembler.h>
#include <TNL/Solvers/PDE/BackwardTimeDiscretisation.h>
+#include <TNL/Functions/Analytic/VectorNorm.h>
+
+#include "RiemannProblemInitialCondition.h"
+#include "CompressibleConservativeVariables.h"
+#include "PhysicalVariablesGetter.h"
+#include "eulerProblem.h"
#include "LaxFridrichsContinuity.h"
#include "LaxFridrichsEnergy.h"
#include "LaxFridrichsMomentumX.h"
#include "LaxFridrichsMomentumY.h"
-#include "EulerPressureGetter.h"
-#include "EulerVelXGetter.h"
-#include "EulerVelYGetter.h"
-#include "EulerVelGetter.h"
+#include "LaxFridrichsMomentumZ.h"
namespace TNL {
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
String
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
getTypeStatic()
{
return String( "eulerProblem< " ) + Mesh :: getTypeStatic() + " >";
@@ -31,20 +44,20 @@ getTypeStatic()
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
String
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
getPrologHeader() const
{
- return String( "euler2D" );
+ return String( "Inviscid flow solver" );
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
writeProlog( Logger& logger, const Config::ParameterContainer& parameters ) const
{
/****
@@ -56,115 +69,83 @@ writeProlog( Logger& logger, const Config::ParameterContainer& parameters ) cons
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
setup( const MeshPointer& meshPointer,
const Config::ParameterContainer& parameters,
const String& prefix )
{
- if( ! this->boundaryConditionPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) ||
+ if( ! this->inviscidOperatorsPointer->setup( meshPointer, parameters, prefix + "inviscid-operators-" ) ||
+ ! this->boundaryConditionPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) ||
! this->rightHandSidePointer->setup( parameters, prefix + "right-hand-side-" ) )
return false;
+ this->gamma = parameters.getParameter< double >( "gamma" );
+ velocity->setMesh( meshPointer );
+ pressure->setMesh( meshPointer );
return true;
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
-typename eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::IndexType
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+ typename InviscidOperators >
+typename eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::IndexType
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
getDofs( const MeshPointer& mesh ) const
{
/****
* Return number of DOFs (degrees of freedom) i.e. number
* of unknowns to be resolved by the main solver.
*/
- return 4*mesh->template getEntitiesCount< typename MeshType::Cell >();
+ return this->conservativeVariables->getDofs( mesh );
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
bindDofs( const MeshPointer& mesh,
DofVectorPointer& dofVector )
{
+ this->conservativeVariables->bind( mesh, dofVector );
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
setInitialCondition( const Config::ParameterContainer& parameters,
const MeshPointer& mesh,
DofVectorPointer& dofs,
MeshDependentDataPointer& meshDependentData )
{
- typedef typename MeshType::Cell Cell;
- double gamma = parameters.getParameter< double >( "gamma" );
- double rhoL = parameters.getParameter< double >( "left-density" );
- double velLX = parameters.getParameter< double >( "left-velocityX" );
- double velLY = parameters.getParameter< double >( "left-velocityY" );
- double preL = parameters.getParameter< double >( "left-pressure" );
- double eL = ( preL / (gamma - 1) ) + 0.5 * rhoL * ::pow(velLX,2) + ::pow(velLY,2);
- double rhoR = parameters.getParameter< double >( "right-density" );
- double velRX = parameters.getParameter< double >( "right-velocityX" );
- double velRY = parameters.getParameter< double >( "right-velocityY" );
- double preR = parameters.getParameter< double >( "right-pressure" );
- double eR = ( preR / (gamma - 1) ) + 0.5 * rhoR * ::pow(velRX,2) + ::pow(velRY,2);
- double x0 = parameters.getParameter< double >( "riemann-border" );
- int size = mesh->template getEntitiesCount< Cell >();
- int size2 = size * size;
- this->rho.bind( *dofs, 0, size2 );
- this->rhoVelX.bind( *dofs, size2,size2);
- this->rhoVelY.bind( *dofs, 2*size2,size2);
- this->energy.bind( *dofs,3*size2,size2);
- this->data.setSize( 4*size2);
- this->pressure.bind(this->data,0,size2);
- this->velocity.bind(this->data,size2,size2);
- this->velocityX.bind(this->data,2*size2,size2);
- this->velocityY.bind(this->data,3*size2,size2);
- for(long int j = 0; j < size; j++)
- for(long int i = 0; i < size; i++)
- if ((i < x0 * size)&&(j < x0 * size) )
- {
- this->rho[j*size+i] = rhoL;
- this->rhoVelX[j*size+i] = rhoL * velLX;
- this->rhoVelY[j*size+i] = rhoL * velLY;
- this->energy[j*size+i] = eL;
- this->velocity[j*size+i] = ::sqrt( ::pow(velLX,2) + ::pow(velLY,2) );
- this->velocityX[j*size+i] = velLX;
- this->velocityY[j*size+i] = velLY;
- this->pressure[j*size+i] = preL;
- }
- else
- {
- this->rho[j*size+i] = rhoR;
- this->rhoVelX[j*size+i] = rhoR * velRX;
- this->rhoVelY[j*size+i] = rhoR * velRY;
- this->energy[j*size+i] = eR;
- this->velocity[j*size+i] = ::sqrt( ::pow(velRX,2) + :: pow(velRY,2) );
- this->velocityX[j*size+i] = velRX;
- this->velocityY[j*size+i] = velRY;
- this->pressure[j*size+i] = preR;
- };
- this->gamma = gamma;
- return true;
+ CompressibleConservativeVariables< MeshType > conservativeVariables;
+ conservativeVariables.bind( mesh, dofs );
+ const String& initialConditionType = parameters.getParameter< String >( "initial-condition" );
+ if( initialConditionType == "riemann-problem" )
+ {
+ RiemannProblemInitialCondition< MeshType > initialCondition;
+ if( ! initialCondition.setup( parameters ) )
+ return false;
+ initialCondition.setInitialCondition( conservativeVariables );
+ return true;
+ }
+ std::cerr << "Unknown initial condition " << initialConditionType << std::endl;
+ return false;
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
template< typename Matrix >
bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
setupLinearSystem( const MeshPointer& mesh,
Matrix& matrix )
{
@@ -173,7 +154,7 @@ setupLinearSystem( const MeshPointer& mesh,
CompressedRowsLengthsVectorType rowLengths;
if( ! rowLengths.setSize( dofs ) )
return false;
- MatrixSetter< MeshType, DifferentialOperator, BoundaryCondition, CompressedRowsLengthsVectorType > matrixSetter;
+ MatrixSetter< MeshType, InviscidOperators, BoundaryCondition, CompressedRowsLengthsVectorType > matrixSetter;
matrixSetter.template getCompressedRowsLengths< typename Mesh::Cell >( mesh,
differentialOperator,
boundaryCondition,
@@ -187,9 +168,9 @@ setupLinearSystem( const MeshPointer& mesh,
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
makeSnapshot( const RealType& time,
const IndexType& step,
const MeshPointer& mesh,
@@ -197,44 +178,40 @@ makeSnapshot( const RealType& time,
MeshDependentDataPointer& meshDependentData )
{
std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
- this->bindDofs( mesh, dofs );
+
+ this->bindDofs( mesh, dofs );
+ PhysicalVariablesGetter< MeshType > physicalVariablesGetter;
+ physicalVariablesGetter.getVelocity( this->conservativeVariables, this->velocity );
+ physicalVariablesGetter.getPressure( this->conservativeVariables, this->gamma, this->pressure );
+
FileName fileName;
fileName.setExtension( "tnl" );
fileName.setIndex( step );
- fileName.setFileNameBase( "rho-" );
- if( ! this->rho.save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "rhoVelX-" );
- if( ! this->rhoVelX.save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "rhoVelY-" );
- if( ! this->rhoVelY.save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "energy-" );
- if( ! this->energy.save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "velocityX-" );
- if( ! this->velocityX.save( fileName.getFileName() ) )
- return false;
- fileName.setFileNameBase( "velocityY-" );
- if( ! this->velocityY.save( fileName.getFileName() ) )
+ fileName.setFileNameBase( "density-" );
+ if( ! this->conservativeVariables->getDensity()->save( fileName.getFileName() ) )
return false;
+
fileName.setFileNameBase( "velocity-" );
- if( ! this->velocity.save( fileName.getFileName() ) )
+ if( ! this->velocity->save( fileName.getFileName() ) )
return false;
- fileName.setFileNameBase( "pressue-" );
- if( ! this->pressure.save( fileName.getFileName() ) )
+
+ fileName.setFileNameBase( "pressure-" );
+ if( ! this->pressure->save( fileName.getFileName() ) )
return false;
+ fileName.setFileNameBase( "energy-" );
+ if( ! this->conservativeVariables->getEnergy()->save( fileName.getFileName() ) )
+ return false;
+
return true;
}
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
getExplicitRHS( const RealType& time,
const RealType& tau,
const MeshPointer& mesh,
@@ -243,92 +220,100 @@ getExplicitRHS( const RealType& time,
MeshDependentDataPointer& meshDependentData )
{
typedef typename MeshType::Cell Cell;
- int count = mesh->template getEntitiesCount< Cell >()/4;
- //bind _u
- this->_uRho.bind( *_u,0,count);
- this->_uRhoVelocityX.bind( *_u,count,count);
- this->_uRhoVelocityY.bind( *_u,2 * count,count);
- this->_uEnergy.bind( *_u,3 * count,count);
-
- //bind _fu
- this->_fuRho.bind( *_u,0,count);
- this->_fuRhoVelocityX.bind( *_u,count,count);
- this->_fuRhoVelocityY.bind( *_u,2 * count,count);
- this->_fuEnergy.bind( *_u,3 * count,count);
- //bind MeshFunctionType
- MeshFunctionPointer velocity( mesh, this->velocity );
- MeshFunctionPointer velocityX( mesh, this->velocityX );
- MeshFunctionPointer velocityY( mesh, this->velocityY );
- MeshFunctionPointer pressure( mesh, this->pressure );
- MeshFunctionPointer uRho( mesh, _uRho );
- MeshFunctionPointer fuRho( mesh, _fuRho );
- MeshFunctionPointer uRhoVelocityX( mesh, _uRhoVelocityX );
- MeshFunctionPointer fuRhoVelocityX( mesh, _fuRhoVelocityX );
- MeshFunctionPointer uRhoVelocityY( mesh, _uRhoVelocityY );
- MeshFunctionPointer fuRhoVelocityY( mesh, _fuRhoVelocityY );
- MeshFunctionPointer uEnergy( mesh, _uEnergy );
- MeshFunctionPointer fuEnergy( mesh, _fuEnergy );
- //generate Operators
- SharedPointer< Continuity > lF2DContinuity;
- SharedPointer< MomentumX > lF2DMomentumX;
- SharedPointer< MomentumY > lF2DMomentumY;
- SharedPointer< Energy > lF2DEnergy;
-
- this->bindDofs( mesh, _u );
- //rho
- lF2DContinuity->setTau(tau);
- lF2DContinuity->setVelocityX( *velocityX );
- lF2DContinuity->setVelocityY( *velocityY );
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, Continuity, BoundaryCondition, RightHandSide > explicitUpdaterContinuity;
+
+ /****
+ * Bind DOFs
+ */
+ this->conservativeVariables->bind( mesh, _u );
+ this->conservativeVariablesRHS->bind( mesh, _fu );
+ this->velocity->setMesh( mesh );
+ this->pressure->setMesh( mesh );
+
+ /****
+ * Resolve the physical variables
+ */
+ PhysicalVariablesGetter< typename MeshPointer::ObjectType > physicalVariables;
+ physicalVariables.getVelocity( this->conservativeVariables, this->velocity );
+ physicalVariables.getPressure( this->conservativeVariables, this->gamma, this->pressure );
+
+ /****
+ * Set-up operators
+ */
+ typedef typename InviscidOperators::ContinuityOperatorType ContinuityOperatorType;
+ typedef typename InviscidOperators::MomentumXOperatorType MomentumXOperatorType;
+ typedef typename InviscidOperators::MomentumYOperatorType MomentumYOperatorType;
+ typedef typename InviscidOperators::MomentumZOperatorType MomentumZOperatorType;
+ typedef typename InviscidOperators::EnergyOperatorType EnergyOperatorType;
+
+ this->inviscidOperatorsPointer->setTau( tau );
+ this->inviscidOperatorsPointer->setVelocity( this->velocity );
+ this->inviscidOperatorsPointer->setPressure( this->pressure );
+
+ /****
+ * Continuity equation
+ */
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, ContinuityOperatorType, BoundaryCondition, RightHandSide > explicitUpdaterContinuity;
explicitUpdaterContinuity.template update< typename Mesh::Cell >( time,
mesh,
- lF2DContinuity,
+ this->inviscidOperatorsPointer->getContinuityOperator(),
this->boundaryConditionPointer,
this->rightHandSidePointer,
- uRho,
- fuRho );
-
- //rhoVelocityX
- lF2DMomentumX->setTau(tau);
- lF2DMomentumX->setVelocityX( *velocityX );
- lF2DMomentumX->setVelocityY( *velocityY );
- lF2DMomentumX->setPressure( *pressure );
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, MomentumX, BoundaryCondition, RightHandSide > explicitUpdaterMomentumX;
+ this->conservativeVariables->getDensity(),
+ this->conservativeVariablesRHS->getDensity() );
+
+ /****
+ * Momentum equations
+ */
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, MomentumXOperatorType, BoundaryCondition, RightHandSide > explicitUpdaterMomentumX;
explicitUpdaterMomentumX.template update< typename Mesh::Cell >( time,
mesh,
- lF2DMomentumX,
- this->boundaryConditionPointer,
- this->rightHandSidePointer,
- uRhoVelocityX,
- fuRhoVelocityX );
-
- //rhoVelocityY
- lF2DMomentumY->setTau(tau);
- lF2DMomentumY->setVelocityX( *velocityX );
- lF2DMomentumY->setVelocityY( *velocityY );
- lF2DMomentumY->setPressure( *pressure );
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, MomentumY, BoundaryCondition, RightHandSide > explicitUpdaterMomentumY;
- explicitUpdaterMomentumY.template update< typename Mesh::Cell >( time,
- mesh,
- lF2DMomentumY,
+ this->inviscidOperatorsPointer->getMomentumXOperator(),
this->boundaryConditionPointer,
this->rightHandSidePointer,
- uRhoVelocityY,
- fuRhoVelocityY );
+ ( *this->conservativeVariables->getMomentum() )[ 0 ], // uRhoVelocityX,
+ ( *this->conservativeVariablesRHS->getMomentum() )[ 0 ] ); //, fuRhoVelocityX );
+
+ if( Dimensions > 1 )
+ {
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, MomentumYOperatorType, BoundaryCondition, RightHandSide > explicitUpdaterMomentumY;
+ explicitUpdaterMomentumY.template update< typename Mesh::Cell >( time,
+ mesh,
+ this->inviscidOperatorsPointer->getMomentumYOperator(),
+ this->boundaryConditionPointer,
+ this->rightHandSidePointer,
+ ( *this->conservativeVariables->getMomentum() )[ 1 ], // uRhoVelocityX,
+ ( *this->conservativeVariablesRHS->getMomentum() )[ 1 ] ); //, fuRhoVelocityX );
+ }
+
+ if( Dimensions > 2 )
+ {
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, MomentumZOperatorType, BoundaryCondition, RightHandSide > explicitUpdaterMomentumZ;
+ explicitUpdaterMomentumZ.template update< typename Mesh::Cell >( time,
+ mesh,
+ this->inviscidOperatorsPointer->getMomentumZOperator(),
+ this->boundaryConditionPointer,
+ this->rightHandSidePointer,
+ ( *this->conservativeVariables->getMomentum() )[ 2 ], // uRhoVelocityX,
+ ( *this->conservativeVariablesRHS->getMomentum() )[ 2 ] ); //, fuRhoVelocityX );
+ }
+
- //energy
- lF2DEnergy->setTau(tau);
- lF2DEnergy->setVelocityX( *velocityX );
- lF2DEnergy->setVelocityY( *velocityY );
- lF2DEnergy->setPressure( *pressure );
- Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, Energy, BoundaryCondition, RightHandSide > explicitUpdaterEnergy;
+ /****
+ * Energy equation
+ */
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, EnergyOperatorType, BoundaryCondition, RightHandSide > explicitUpdaterEnergy;
explicitUpdaterEnergy.template update< typename Mesh::Cell >( time,
mesh,
- lF2DEnergy,
+ this->inviscidOperatorsPointer->getEnergyOperator(),
this->boundaryConditionPointer,
this->rightHandSidePointer,
- uEnergy,
- fuEnergy );
+ this->conservativeVariables->getEnergy(), // uRhoVelocityX,
+ this->conservativeVariablesRHS->getEnergy() ); //, fuRhoVelocityX );
+
+ /*this->conservativeVariablesRHS->getDensity()->write( "density", "gnuplot" );
+ this->conservativeVariablesRHS->getEnergy()->write( "energy", "gnuplot" );
+ this->conservativeVariablesRHS->getMomentum()->write( "momentum", "gnuplot", 0.05 );
+ getchar();*/
/*
BoundaryConditionsSetter< MeshFunctionType, BoundaryCondition > boundaryConditionsSetter;
@@ -341,10 +326,10 @@ getExplicitRHS( const RealType& time,
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
template< typename Matrix >
void
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
assemblyLinearSystem( const RealType& time,
const RealType& tau,
const MeshPointer& mesh,
@@ -355,7 +340,7 @@ assemblyLinearSystem( const RealType& time,
{
/* LinearSystemAssembler< Mesh,
MeshFunctionType,
- DifferentialOperator,
+ InviscidOperators,
BoundaryCondition,
RightHandSide,
BackwardTimeDiscretisation,
@@ -377,15 +362,16 @@ assemblyLinearSystem( const RealType& time,
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
+ typename InviscidOperators >
bool
-eulerProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+eulerProblem< Mesh, BoundaryCondition, RightHandSide, InviscidOperators >::
postIterate( const RealType& time,
const RealType& tau,
const MeshPointer& mesh,
DofVectorPointer& dofs,
MeshDependentDataPointer& meshDependentData )
{
+ /*
typedef typename MeshType::Cell Cell;
int count = mesh->template getEntitiesCount< Cell >()/4;
//bind _u
@@ -433,7 +419,8 @@ postIterate( const RealType& time,
euler2DPressure.setRho(uRho);
// OperatorFunction< euler2DPressure, MeshFunction, void, time > OFPressure;
// pressure = OFPressure;
-
+ */
+ return true;
}
} // namespace TNL
diff --git a/examples/inviscid-flow/1d/eulerRhs.h b/examples/inviscid-flow/eulerRhs.h
similarity index 100%
rename from examples/inviscid-flow/1d/eulerRhs.h
rename to examples/inviscid-flow/eulerRhs.h
diff --git a/examples/inviscid-flow/run-euler b/examples/inviscid-flow/run-euler
new file mode 100644
index 0000000000000000000000000000000000000000..7af3493805c96e19c65a3f475e2ed63aaf59db92
--- /dev/null
+++ b/examples/inviscid-flow/run-euler
@@ -0,0 +1,35 @@
+#!/usr/bin/env bash
+
+tnl-grid-setup --dimensions 2 \
+ --origin-x 0.0 \
+ --origin-y 0.0 \
+ --proportions-x 1.0 \
+ --proportions-y 1.0 \
+ --size-x 100 \
+ --size-y 100
+
+tnl-init --test-function sin-wave \
+ --output-file init.tnl
+tnl-euler-2d --initial-condition riemann-problem \
+ --discontinuity-placement-0 0.3 \
+ --discontinuity-placement-1 0.3 \
+ --discontinuity-placement-2 0.3 \
+ --left-density 1.0 \
+ --right-density 1.0 \
+ --left-velocity-0 -0.2 \
+ --left-velocity-1 -0.2 \
+ --left-velocity-2 -0.2 \
+ --right-velocity-0 0.2 \
+ --right-velocity-1 0.2 \
+ --right-velocity-2 0.2 \
+ --left-pressure 0.4 \
+ --right-pressure 0.4 \
+ --time-discretisation explicit \
+ --boundary-conditions-type neumann \
+ --boundary-conditions-constant 0 \
+ --discrete-solver euler \
+ --time-step 0.0001 \
+ --snapshot-period 0.01 \
+ --final-time 1.0
+
+tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/narrow-band/CMakeLists.txt b/examples/narrow-band/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..748343b58e4c5bcca2fe5de2160d3e4a2eeb2e76
--- /dev/null
+++ b/examples/narrow-band/CMakeLists.txt
@@ -0,0 +1,22 @@
+set( tnl_narrow_band_SOURCES
+# MainBuildConfig.h
+# tnlNarrowBand2D_impl.h
+# tnlNarrowBand.h
+# narrowBandConfig.h
+ main.cpp)
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(narrow-band${debugExt} main.cu)
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(narrow-band${debugExt} main.cpp)
+ENDIF( BUILD_CUDA )
+target_link_libraries (narrow-band${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS narrow-band${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+#INSTALL( FILES ${tnl_narrow_band_SOURCES}
+# DESTINATION share/tnl-${tnlVersion}/examples/narrow-band )
diff --git a/examples/narrow-band/MainBuildConfig.h b/examples/narrow-band/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/examples/narrow-band/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/examples/narrow-band/main.cpp b/examples/narrow-band/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/examples/narrow-band/main.cpp
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/examples/narrow-band/main.cu b/examples/narrow-band/main.cu
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/examples/narrow-band/main.cu
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/examples/narrow-band/main.h b/examples/narrow-band/main.h
new file mode 100644
index 0000000000000000000000000000000000000000..47d8648bbf7646cd9580134f55be6002359cf21e
--- /dev/null
+++ b/examples/narrow-band/main.h
@@ -0,0 +1,88 @@
+/***************************************************************************
+ main.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+
+#include "MainBuildConfig.h"
+ //for HOST versions:
+//#include "tnlNarrowBand.h"
+ //for DEVICE versions:
+#include "tnlNarrowBand_CUDA.h"
+#include "narrowBandConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+
+#include <mesh/tnlGrid.h>
+#include <core/tnlDevice.h>
+#include <time.h>
+#include <ctime>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ time_t start;
+ time_t stop;
+ time(&start);
+ std::clock_t start2= std::clock();
+ Config::ParameterContainer parameters;
+ tnlConfigDescription configDescription;
+ narrowBandConfig< BuildConfig >::configSetup( configDescription );
+
+ if( ! parseCommandLine( argc, argv, configDescription, parameters ) )
+ return false;
+
+ const int& dim = parameters.getParameter< int >( "dim" );
+
+ if(dim == 2)
+ {
+ tnlNarrowBand<tnlGrid<2,double,tnlHost, int>, double, int> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ checkCudaDevice;
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver..." << endl;
+ solver.run();
+ }
+// else if(dim == 3)
+// {
+// tnlNarrowBand<tnlGrid<3,double,tnlHost, int>, double, int> solver;
+// if(!solver.init(parameters))
+// {
+// cerr << "Solver failed to initialize." << endl;
+// return EXIT_FAILURE;
+// }
+// checkCudaDevice;
+// cout << "-------------------------------------------------------------" << endl;
+// cout << "Starting solver..." << endl;
+// solver.run();
+// }
+ else
+ {
+ cerr << "Unsupported number of dimensions: " << dim << "!" << endl;
+ return EXIT_FAILURE;
+ }
+
+
+ time(&stop);
+ cout << "Solver stopped..." << endl;
+ cout << endl;
+ cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) << endl;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/examples/narrow-band/narrowBandConfig.h b/examples/narrow-band/narrowBandConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..bab58ceac46bf9c766b697ed79c2c676111323a2
--- /dev/null
+++ b/examples/narrow-band/narrowBandConfig.h
@@ -0,0 +1,40 @@
+/***************************************************************************
+ narrowBandConfig.h - description
+ -------------------
+ begin : Oct 15, 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef NARROWBANDCONFIG_H_
+#define NARROWBANDCONFIG_H_
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class narrowBandConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Narrow Band Solver solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "fast-sweeping" );
+ config.addRequiredEntry < String > ( "initial-condition", "Initial condition for solver");
+ config.addRequiredEntry < int > ( "dim", "Dimension of problem.");
+ config.addRequiredEntry < double > ( "tau", "Time step.");
+ config.addRequiredEntry < double > ( "final-time", "Final time.");
+ config.addEntry < String > ( "mesh", "Name of mesh.", "mesh.tnl" );
+ config.addEntry < String > ( "exact-input", "Are the function values near the curve equal to the SDF? (yes/no)", "no" );
+ }
+};
+
+#endif /* NARROWBANDCONFIG_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand.h b/examples/narrow-band/tnlNarrowBand.h
new file mode 100644
index 0000000000000000000000000000000000000000..e2817424a1feb0f24bcbf8a759d37e8c9697977d
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand.h
@@ -0,0 +1,186 @@
+/***************************************************************************
+ tnlNarrowBand.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND_H_
+#define TNLNARROWBAND_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlMeshFunction.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+#ifdef HAVE_OPENMP
+#include <omp.h>
+#endif
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlNarrowBand
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ tnlNarrowBand();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+
+ void setupSquare1000(Index i, Index j);
+ void setupSquare1100(Index i, Index j);
+ void setupSquare1010(Index i, Index j);
+ void setupSquare1001(Index i, Index j);
+ void setupSquare1110(Index i, Index j);
+ void setupSquare1101(Index i, Index j);
+ void setupSquare1011(Index i, Index j);
+ void setupSquare1111(Index i, Index j);
+ void setupSquare0000(Index i, Index j);
+ void setupSquare0100(Index i, Index j);
+ void setupSquare0010(Index i, Index j);
+ void setupSquare0001(Index i, Index j);
+ void setupSquare0110(Index i, Index j);
+ void setupSquare0101(Index i, Index j);
+ void setupSquare0011(Index i, Index j);
+ void setupSquare0111(Index i, Index j);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage > Entity;
+
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlNarrowBand();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j, const Index k);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage > Entity;
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+ //for single core version use this implementation:
+#include "tnlNarrowBand2D_impl.h"
+ //for parallel version use this one instead:
+// #include "tnlNarrowBand2D_openMP_impl.h"
+
+#include "tnlNarrowBand3D_impl.h"
+
+#endif /* TNLNARROWBAND_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h b/examples/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..e532bca7774c9c42648bf43f648a07ebbf20eb78
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h
@@ -0,0 +1,1317 @@
+/***************************************************************************
+ tnlNarrowBand2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND2D_IMPL_H_
+#define TNLNARROWBAND2D_IMPL_H_
+
+#define NARROWBAND_SUBGRID_SIZE 32
+
+#include "tnlNarrowBand.h"
+
+#ifdef HAVE_CUDA
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+#endif
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return -arg;
+ return 0.0;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlNarrowBand< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlNarrowBand()
+:dofVector(Mesh)
+{
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ //Entity.refresh();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+ tau = parameters.getParameter< double >( "tau" );
+
+ finalTime = parameters.getParameter< double >( "final-time" );
+
+ statusGridSize = ((Mesh.getDimensions().x() + NARROWBAND_SUBGRID_SIZE-1 ) / NARROWBAND_SUBGRID_SIZE);
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaStatusVector), statusGridSize*statusGridSize*sizeof(int));
+// cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), statusGridSize*statusGridSize* sizeof(int)), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&reinitialize, sizeof(int));
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+#endif
+
+ int n = Mesh.getDimensions().x();
+
+ dim3 threadsPerBlock2(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
+ dim3 numBlocks2(statusGridSize ,statusGridSize);
+ initSetupGridCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initSetupGrid2CUDA<<<numBlocks2,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+
+ /*dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);*/
+ initCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+
+ cout << "Solver initialized." << endl;
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlockFS(1, 512);
+ dim3 numBlocksFS(4,1);
+ dim3 threadsPerBlockNB(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
+ dim3 numBlocksNB(n/NARROWBAND_SUBGRID_SIZE + 1,n/NARROWBAND_SUBGRID_SIZE + 1);
+
+ double time = 0.0;
+ int reinit = 0;
+
+ cout << "Hi!" << endl;
+ runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ cout << "Hi2!" << endl;
+ while(time < finalTime)
+ {
+ if(tau+time > finalTime)
+ tau=finalTime-time;
+
+ runNarrowBandCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver,tau);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ time += tau;
+
+
+ cudaMemcpy(&reinit, this->reinitialize, sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ if(reinit != 0 /*&& time != finalTime */)
+ {
+ cout << time << endl;
+
+ initSetupGridCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initSetupGrid2CUDA<<<numBlocksNB,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ }
+ }
+
+ //data.setLike(dofVector.getData());
+ //cudaMemcpy(data.getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaMemcpy(dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ //data.save("u-00001.tnl");
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ // 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+ // 8 - to the east of curve, 16 - to the west of curve.
+ int subgridID = i/NARROWBAND_SUBGRID_SIZE + (j/NARROWBAND_SUBGRID_SIZE) * statusGridSize;
+ if(cudaStatusVector[subgridID] != 0 && i<Mesh.getDimensions().x() && j < Mesh.getDimensions().y())
+ {
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real a,b, tmp;
+
+ if( i == 0 /*|| (i/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 9))*/ )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 /*|| (i/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 17))*/ )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 /*|| (j/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 3))*/ )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 /* || (j/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 5)) */)
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ // cudaDofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+ atomicFabsMin(&(cudaDofVector2[Entity.getIndex()]), tmp);
+ }
+
+}
+
+
+__global__ void initCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+
+
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int i = threadIdx.x + blockDim.x*blockIdx.x;
+ int j = blockDim.y*blockIdx.y + threadIdx.y;
+
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ int gid = Entity.getIndex();
+
+ if(abs(cudaDofVector2[gid]) > 1.5*h)
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector2[gid]);
+
+// if (i >0 && j > 0 && i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y())
+// {
+// if(cudaDofVector2[gid]*cudaDofVector2[gid+1] <= 0 )
+// {
+// cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+// cudaDofVector2[gid+1] = sign(cudaDofVector2[gid+1])*0.5*h;
+// }
+// if( cudaDofVector2[gid]*cudaDofVector2[gid+Mesh.getDimensions().x()] <= 0 )
+// {
+// cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+// cudaDofVector2[gid+Mesh.getDimensions().x()] = sign(cudaDofVector2[gid+Mesh.getDimensions().x()])*0.5*h;
+// }
+//
+// if(cudaDofVector2[gid]*cudaDofVector2[gid-1] <= 0 )
+// {
+// cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+// cudaDofVector2[gid-1] = sign(cudaDofVector2[gid-1])*0.5*h;
+// }
+// if( cudaDofVector2[gid]*cudaDofVector2[gid-Mesh.getDimensions().x()] <= 0 )
+// {
+// cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+// cudaDofVector2[gid-Mesh.getDimensions().x()] = sign(cudaDofVector2[gid-Mesh.getDimensions().x()])*0.5*h;
+// }
+// }
+
+
+//
+
+
+
+
+
+
+// if(i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y() )
+// {
+// if(cudaDofVector[Entity.getIndex()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1111(i,j);
+// else
+// setupSquare1110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1101(i,j);
+// else
+// setupSquare1100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1011(i,j);
+// else
+// setupSquare1010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1001(i,j);
+// else
+// setupSquare1000(i,j);
+// }
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0111(i,j);
+// else
+// setupSquare0110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0101(i,j);
+// else
+// setupSquare0100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0011(i,j);
+// else
+// setupSquare0010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0001(i,j);
+// else
+// setupSquare0000(i,j);
+// }
+// }
+// }
+//
+// }
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ //Real fy = abs(y);
+
+ //Real tmpMin = Min(fx,abs(y));
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+
+ int gx = 0;
+ int gy = threadIdx.y;
+ //if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ // return;
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ //int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ //int id1 = gx+gy;
+ //int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+ if(blockIdx.x==0)
+ {
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+
+}
+
+
+
+
+__global__ void initSetupGridCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ __shared__ double u0;
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+
+// printf("Hello from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+// printf("Hello from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+
+ if(blockIdx.x+blockIdx.y == 0)
+ *(solver->reinitialize) = 0;
+
+ solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y] = 0;
+
+ u0 = solver->cudaDofVector2[(blockDim.y*blockIdx.y + 0)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + 0];
+ }
+ __syncthreads();
+
+ double u = solver->cudaDofVector2[(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x];
+
+ if(u*u0 <=0.0)
+ atomicMax(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y]),1);
+ }
+// if(threadIdx.x+threadIdx.y == 0)
+
+// printf("Bye from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+
+
+}
+
+
+
+// run this with one thread per block
+__global__ void initSetupGrid2CUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+// printf("Hello\n");
+ if(solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y] == 1)
+ {
+// 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+// 8 - to the east of curve, 16 - to the west of curve.
+ if(blockIdx.x > 0)
+ {
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x - 1 + gridDim.x*blockIdx.y]), 16);
+ }
+
+ if(blockIdx.x < gridDim.x - 1)
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + 1 + gridDim.x*blockIdx.y]), 8);
+
+ if(blockIdx.y > 0 )
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*(blockIdx.y - 1)]), 4);
+
+ if(blockIdx.y < gridDim.y - 1)
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*(blockIdx.y + 1)]), 2);
+ }
+
+
+}
+
+
+
+
+
+__global__ void runNarrowBandCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, double tau)
+{
+ int gid = (blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x()+ threadIdx.x;
+ int i = threadIdx.x + blockIdx.x*blockDim.x;
+ int j = threadIdx.y + blockIdx.y*blockDim.y;
+
+// if(i+j == 0)
+// printf("Hello\n");
+
+ int blockID = blockIdx.x + blockIdx.y*gridDim.x; /*i/NARROWBAND_SUBGRID_SIZE + (j/NARROWBAND_SUBGRID_SIZE) * ((Mesh.getDimensions().x() + NARROWBAND_SUBGRID_SIZE-1 ) / NARROWBAND_SUBGRID_SIZE);*/
+
+ int status = solver->cudaStatusVector[blockID];
+
+ if(solver->Mesh.getDimensions().x() > i && solver->Mesh.getDimensions().y() > j)
+ {
+
+ if(status != 0)
+ {
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(solver->Mesh);
+ Entity.setCoordinates(Containers::StaticVector<2,double>(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ double value = solver->cudaDofVector2[Entity.getIndex()];
+ double xf,xb,yf,yb, grad, fu, a,b;
+ a = b = 0.0;
+
+ if( i == 0 || (threadIdx.x == 0 && !(status & 9)) )
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ xf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] - value;
+ }
+ else if( i == solver->Mesh.getDimensions().x() - 1 || (threadIdx.x == blockDim.x - 1 && !(status & 17)) )
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ xf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()] - value;
+ }
+ else
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ xf = solver-> cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] - value;
+ }
+
+ if( j == 0 || (threadIdx.y == 0 && !(status & 3)) )
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] ;
+ yf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] - value;
+ }
+ else if( j == solver->Mesh.getDimensions().y() - 1 || (threadIdx.y == blockDim.y - 1 && !(status & 5)) )
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ yf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()] - value;
+ }
+ else
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ yf = solver-> cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] - value;
+ }
+ __syncthreads();
+
+
+
+
+
+ if(sign(value) >= 0.0)
+ {
+ xf = solver->negativePart(xf);
+
+ xb = solver->positivePart(xb);
+
+ yf = solver->negativePart(yf);
+
+ yb = solver->positivePart(yb);
+
+ }
+ else
+ {
+
+ xb = solver->negativePart(xb);
+
+ xf = solver->positivePart(xf);
+
+ yb = solver->negativePart(yb);
+
+ yf = solver->positivePart(yf);
+ }
+
+
+ if(xb > xf)
+ a = xb*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+ else
+ a = xf*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+
+ if(yb > yf)
+ b = yb*solver->Mesh.template getSpaceStepsProducts< 0, -1 >();
+ else
+ b = yf*solver->Mesh.template getSpaceStepsProducts< 0, -1 >();
+
+
+
+// grad = sqrt(0.5 * (xf*xf + xb*xb + yf*yf + yb*yb ) )*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+
+ grad = sqrt(/*0.5 **/ (a*a + b*b ) );
+
+ fu = -1.0 * grad;
+
+ if((tau*fu+value)*value <=0 )
+ {
+ // 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+ // 8 - to the east of curve, 16 - to the west of curve.
+
+ if((threadIdx.x == 6 && !(status & 9)) && (blockIdx.x > 0) )
+ atomicMax(solver->reinitialize,1);
+ else if((threadIdx.x == blockDim.x - 7 && !(status & 17)) && (blockIdx.x < gridDim.x - 1) )
+ atomicMax(solver->reinitialize,1);
+ else if((threadIdx.y == 6 && !(status & 3)) && (blockIdx.y > 0) )
+ atomicMax(solver->reinitialize,1);
+ else if((threadIdx.y == blockDim.y - 7 && !(status & 5)) && (blockIdx.y < gridDim.y - 1) )
+ atomicMax(solver->reinitialize,1);
+ }
+
+ solver->cudaDofVector2[Entity.getIndex()] += tau*fu;
+ }
+ }
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+#endif
+
+
+
+
+#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h b/examples/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..2d1296058391a0685d9385047469acb41a40b765
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h
@@ -0,0 +1,1313 @@
+/***************************************************************************
+ tnlNarrowBand2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND2D_IMPL_H_
+#define TNLNARROWBAND2D_IMPL_H_
+
+#define NARROWBAND_SUBGRID_SIZE 32
+
+#include "tnlNarrowBand.h"
+
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return -arg;
+ return 0.0;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlNarrowBand< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlNarrowBand()
+:dofVector(Mesh)
+{
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ //Entity.refresh();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+ tau = parameters.getParameter< double >( "tau" );
+
+ finalTime = parameters.getParameter< double >( "final-time" );
+
+ statusGridSize = ((Mesh.getDimensions().x() + NARROWBAND_SUBGRID_SIZE-1 ) / NARROWBAND_SUBGRID_SIZE);
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaStatusVector), statusGridSize*statusGridSize*sizeof(int));
+// cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), statusGridSize*statusGridSize* sizeof(int)), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&reinitialize, sizeof(int));
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+#endif
+
+ int n = Mesh.getDimensions().x();
+
+ dim3 threadsPerBlock2(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
+ dim3 numBlocks2(statusGridSize ,statusGridSize);
+ initSetupGridCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initSetupGrid2CUDA<<<numBlocks2,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+
+ /*dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);*/
+ initCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+
+ cout << "Solver initialized." << endl;
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlockFS(1, 512);
+ dim3 numBlocksFS(4,1);
+ dim3 threadsPerBlockNB(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
+ dim3 numBlocksNB(n/NARROWBAND_SUBGRID_SIZE + 1,n/NARROWBAND_SUBGRID_SIZE + 1);
+
+ double time = 0.0;
+ int reinit = 0;
+
+ cout << "Hi!" << endl;
+ runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ cout << "Hi2!" << endl;
+ while(time < finalTime)
+ {
+ if(tau+time > finalTime)
+ tau=finalTime-time;
+
+ runNarrowBandCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver,tau);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ time += tau;
+
+
+ cudaMemcpy(&reinit, this->reinitialize, sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ if(reinit != 0 /*&& time != finalTime */)
+ {
+ cout << time << endl;
+
+ initSetupGridCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initSetupGrid2CUDA<<<numBlocksNB,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ }
+ }
+
+ //data.setLike(dofVector.getData());
+ //cudaMemcpy(data.getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaMemcpy(dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ //data.save("u-00001.tnl");
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ // 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+ // 8 - to the east of curve, 16 - to the west of curve.
+ int subgridID = i/NARROWBAND_SUBGRID_SIZE + (j/NARROWBAND_SUBGRID_SIZE) * ((Mesh.getDimensions().x() + NARROWBAND_SUBGRID_SIZE-1 ) / NARROWBAND_SUBGRID_SIZE);
+ if(/*cudaStatusVector[subgridID] != 0 &&*/ i<Mesh.getDimensions().x() && Mesh.getDimensions().y())
+ {
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real a,b, tmp;
+
+ if( i == 0 /*|| (i/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 9)) */)
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 /*|| (i/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 17)) */)
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0/* || (j/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 3)) */)
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 /* || (j/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 5))*/ )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ // cudaDofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+ atomicFabsMin(&(cudaDofVector2[Entity.getIndex()]), tmp);
+ }
+
+}
+
+
+__global__ void initCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+
+
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int i = threadIdx.x + blockDim.x*blockIdx.x;
+ int j = blockDim.y*blockIdx.y + threadIdx.y;
+
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ int gid = Entity.getIndex();
+
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector2[gid]);
+
+ if (i >0 && j > 0 && i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y())
+ {
+ if(cudaDofVector2[gid]*cudaDofVector2[gid+1] <= 0 )
+ {
+ cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+ cudaDofVector2[gid+1] = sign(cudaDofVector2[gid+1])*0.5*h;
+ }
+ if( cudaDofVector2[gid]*cudaDofVector2[gid+Mesh.getDimensions().x()] <= 0 )
+ {
+ cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+ cudaDofVector2[gid+Mesh.getDimensions().x()] = sign(cudaDofVector2[gid+Mesh.getDimensions().x()])*0.5*h;
+ }
+
+ if(cudaDofVector2[gid]*cudaDofVector2[gid-1] <= 0 )
+ {
+ cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+ cudaDofVector2[gid-1] = sign(cudaDofVector2[gid-1])*0.5*h;
+ }
+ if( cudaDofVector2[gid]*cudaDofVector2[gid-Mesh.getDimensions().x()] <= 0 )
+ {
+ cudaDofVector2[gid] = sign(cudaDofVector2[gid])*0.5*h;
+ cudaDofVector2[gid-Mesh.getDimensions().x()] = sign(cudaDofVector2[gid-Mesh.getDimensions().x()])*0.5*h;
+ }
+ }
+
+
+//
+
+
+
+
+
+
+// if(i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y() )
+// {
+// if(cudaDofVector[Entity.getIndex()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1111(i,j);
+// else
+// setupSquare1110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1101(i,j);
+// else
+// setupSquare1100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1011(i,j);
+// else
+// setupSquare1010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1001(i,j);
+// else
+// setupSquare1000(i,j);
+// }
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0111(i,j);
+// else
+// setupSquare0110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0101(i,j);
+// else
+// setupSquare0100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0011(i,j);
+// else
+// setupSquare0010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0001(i,j);
+// else
+// setupSquare0000(i,j);
+// }
+// }
+// }
+//
+// }
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ //Real fy = abs(y);
+
+ //Real tmpMin = Min(fx,abs(y));
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+
+ int gx = 0;
+ int gy = threadIdx.y;
+ //if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ // return;
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ //int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ //int id1 = gx+gy;
+ //int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+ if(blockIdx.x==0)
+ {
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+
+}
+
+
+
+
+__global__ void initSetupGridCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ __shared__ double u0;
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+
+// printf("Hello from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+// printf("Hello from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+
+ if(blockIdx.x+blockIdx.y == 0)
+ *(solver->reinitialize) = 0;
+
+ solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y] = 0;
+
+ u0 = solver->cudaDofVector2[(blockDim.y*blockIdx.y + 0)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + 0];
+ }
+ __syncthreads();
+
+ double u = solver->cudaDofVector2[(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x];
+
+ if(u*u0 <=0.0)
+ atomicMax(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y]),1);
+ }
+// if(threadIdx.x+threadIdx.y == 0)
+
+// printf("Bye from block = %d, thread = %d, x = %d, y = %d\n", blockIdx.x + gridDim.x*blockIdx.y,(blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x() + blockDim.x*blockIdx.x + threadIdx.x, threadIdx.x, threadIdx.y);
+
+
+}
+
+
+
+// run this with one thread per block
+__global__ void initSetupGrid2CUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+// printf("Hello\n");
+ if(solver->cudaStatusVector[blockIdx.x + gridDim.x*blockIdx.y] == 1)
+ {
+// 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+// 8 - to the east of curve, 16 - to the west of curve.
+ if(blockIdx.x > 0)
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x - 1 + gridDim.x*blockIdx.y]), 16);
+
+ if(blockIdx.x < gridDim.x - 1)
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + 1 + gridDim.x*blockIdx.y]), 8);
+
+ if(blockIdx.y > 0 )
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*(blockIdx.y - 1)]), 4);
+
+ if(blockIdx.y < gridDim.y - 1)
+ atomicAdd(&(solver->cudaStatusVector[blockIdx.x + gridDim.x*(blockIdx.y + 1)]), 2);
+ }
+
+
+}
+
+
+
+
+
+__global__ void runNarrowBandCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, double tau)
+{
+ int gid = (blockDim.y*blockIdx.y + threadIdx.y)*solver->Mesh.getDimensions().x()+ threadIdx.x;
+ int i = threadIdx.x + blockIdx.x*blockDim.x;
+ int j = threadIdx.y + blockIdx.y*blockDim.y;
+
+// if(i+j == 0)
+// printf("Hello\n");
+
+ int blockID = blockIdx.x + blockIdx.y*gridDim.x; /*i/NARROWBAND_SUBGRID_SIZE + (j/NARROWBAND_SUBGRID_SIZE) * ((Mesh.getDimensions().x() + NARROWBAND_SUBGRID_SIZE-1 ) / NARROWBAND_SUBGRID_SIZE);*/
+
+ int status = solver->cudaStatusVector[blockID];
+
+ if(solver->Mesh.getDimensions().x() > i && solver->Mesh.getDimensions().y() > j)
+ {
+
+// if(status != 0)
+ {
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(solver->Mesh);
+ Entity.setCoordinates(Containers::StaticVector<2,double>(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ double value = solver->cudaDofVector2[Entity.getIndex()];
+ double xf,xb,yf,yb, grad, fu, a,b;
+ a = b = 0.0;
+
+ if( i == 0 /*|| (threadIdx.x == 0 && !(status & 9)) */)
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ xf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] - value;
+ }
+ else if( i == solver->Mesh.getDimensions().x() - 1 /*|| (threadIdx.x == blockDim.x - 1 && !(status & 17)) */)
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ xf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()] - value;
+ }
+ else
+ {
+ xb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ xf = solver-> cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] - value;
+ }
+
+ if( j == 0/* || (threadIdx.y == 0 && !(status & 3))*/ )
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] ;
+ yf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] - value;
+ }
+ else if( j == solver->Mesh.getDimensions().y() - 1 /*|| (threadIdx.y == blockDim.y - 1 && !(status & 5)) */)
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ yf = solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()] - value;
+ }
+ else
+ {
+ yb = value - solver->cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ yf = solver-> cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] - value;
+ }
+ __syncthreads();
+
+
+
+
+
+ if(sign(value) > 0.0)
+ {
+ xf = solver->negativePart(xf);
+
+ xb = solver->positivePart(xb);
+
+ yf = solver->negativePart(yf);
+
+ yb = solver->positivePart(yb);
+
+ }
+ else
+ {
+
+ xb = solver->negativePart(xb);
+
+ xf = solver->positivePart(xf);
+
+ yb = solver->negativePart(yb);
+
+ yf = solver->positivePart(yf);
+ }
+
+
+ if(xb > xf)
+ a = xb*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+ else
+ a = xf*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+
+ if(yb > yf)
+ b = yb*solver->Mesh.template getSpaceStepsProducts< 0, -1 >();
+ else
+ b = yf*solver->Mesh.template getSpaceStepsProducts< 0, -1 >();
+
+
+
+// grad = sqrt(0.5 * (xf*xf + xb*xb + yf*yf + yb*yb ) )*solver->Mesh.template getSpaceStepsProducts< -1, 0 >();
+
+ grad = sqrt(/*0.5 **/ (a*a + b*b ) );
+
+ fu = -1.0 * grad;
+
+// if((tau*fu+value)*value <=0 )
+// {
+// // 1 - with curve, 2 - to the north of curve, 4 - to the south of curve,
+// // 8 - to the east of curve, 16 - to the west of curve.
+//
+// if((threadIdx.x == 1 && !(status & 9)) && (blockIdx.x > 0) )
+// atomicMax(solver->reinitialize,1);
+// else if((threadIdx.x == blockDim.x - 2 && !(status & 17)) && (blockIdx.x < gridDim.x - 1) )
+// atomicMax(solver->reinitialize,1);
+// else if((threadIdx.y == 1 && !(status & 3)) && (blockIdx.y > 0) )
+// atomicMax(solver->reinitialize,1);
+// else if((threadIdx.y == blockDim.y - 2 && !(status & 5)) && (blockIdx.y < gridDim.y - 1) )
+// atomicMax(solver->reinitialize,1);
+// }
+
+ solver->cudaDofVector2[Entity.getIndex()] += tau*fu;
+ }
+ }
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+#endif
+
+
+
+
+#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand2D_impl.h b/examples/narrow-band/tnlNarrowBand2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..8248baa0949862de6ebcee797e6916c9634de72e
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand2D_impl.h
@@ -0,0 +1,927 @@
+/***************************************************************************
+ tnlNarrowBand2D_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND2D_IMPL_H_
+#define TNLNARROWBAND2D_IMPL_H_
+
+#include "tnlNarrowBand.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlNarrowBand< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlNarrowBand()
+:Entity(Mesh),
+ dofVector(Mesh),
+ dofVector2(Mesh)
+{
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+ dofVector2.load(initialCondition);
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ Entity.refresh();
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+ cout << "a" << endl;
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ for(int i=0; i< Mesh.getDimensions().x()*Mesh.getDimensions().x();i++)
+ {
+ dofVector2[i]=INT_MAX*sign(dofVector[i]);
+ }
+
+ for(int i = 0 ; i < Mesh.getDimensions().x()-1; i++)
+ {
+ for(int j = 0 ; j < Mesh.getDimensions().x()-1; j++)
+ {
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ neighbourEntities.refresh(Mesh,Entity.getIndex());
+
+ if(dofVector[this->Entity.getIndex()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1111(i,j);
+ else
+ setupSquare1110(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1101(i,j);
+ else
+ setupSquare1100(i,j);
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1011(i,j);
+ else
+ setupSquare1010(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1001(i,j);
+ else
+ setupSquare1000(i,j);
+ }
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0111(i,j);
+ else
+ setupSquare0110(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0101(i,j);
+ else
+ setupSquare0100(i,j);
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0011(i,j);
+ else
+ setupSquare0010(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0001(i,j);
+ else
+ setupSquare0000(i,j);
+ }
+ }
+ }
+
+ }
+ }
+ cout << "a" << endl;
+
+// Real tmp = 0.0;
+// Real ax=0.5/sqrt(2.0);
+//
+// if(!exactInput)
+// {
+// for(Index i = 0; i < Mesh.getDimensions().x()*Mesh.getDimensions().y(); i++)
+// dofVector[i]=0.5*h*sign(dofVector[i]);
+// }
+//
+//
+// for(Index i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// for(Index j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+// }
+//
+//
+//
+// for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// Index j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// Index i = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+//
+// Index i = Mesh.getDimensions().x() - 1;
+// Index j = Mesh.getDimensions().y() - 1;
+//
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// i = 0;
+// j = Mesh.getDimensions().y() -1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+ //data.setLike(dofVector2.getData());
+ //data=dofVector2.getData();
+ //cout << data.getType() << endl;
+ dofVector2.save("u-00000.tnl");
+ //dofVector2.getData().save("u-00000.tnl");
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+// data.setLike(dofVector2.getData());
+// data = dofVector2.getData();
+// cout << data.getType() << endl;
+ dofVector2.save("u-00001.tnl");
+ //dofVector2.getData().save("u-00001.tnl");
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ Real value = dofVector2[Entity.getIndex()];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(fabs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+
+ dofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+
+// if(dofVector2[Entity.getIndex()] > 1.0)
+// cout << value << " " << tmp << " " << dofVector2[Entity.getIndex()] << endl;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+
+
+
+
+#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand3D_CUDA_impl.h b/examples/narrow-band/tnlNarrowBand3D_CUDA_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..ac8a023d7d93060f63858892c0541f5eda0ee706
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand3D_CUDA_impl.h
@@ -0,0 +1,961 @@
+/***************************************************************************
+ tnlNarrowBand2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND3D_IMPL_H_
+#define TNLNARROWBAND3D_IMPL_H_
+
+#include "tnlNarrowBand.h"
+
+//__device__
+//double fabsMin( double x, double y)
+//{
+// double fx = abs(x);
+//
+// if(Min(fx,abs(y)) == fx)
+// return x;
+// else
+// return y;
+//}
+//
+//__device__
+//double atomicFabsMin(double* address, double val)
+//{
+// unsigned long long int* address_as_ull =
+// (unsigned long long int*)address;
+// unsigned long long int old = *address_as_ull, assumed;
+// do {
+// assumed = old;
+// old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(assumed,val) ));
+// } while (assumed != old);
+// return __longlong_as_double(old);
+//}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlNarrowBand< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ this->h = Mesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(8, 8,8);
+ dim3 numBlocks(n/8 + 1, n/8 +1, n/8 +1);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(1, 512);
+ dim3 numBlocks(8,1);
+
+
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,0,0);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, Index k)
+{
+ tnlGridEntity< tnlGrid< 3,double, tnlHost, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j,k));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real a,b,c, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0 >()] );
+ }
+
+ if( k == 0 )
+ c = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+ else if( k == Mesh.getDimensions().z() - 1 )
+ c = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+ else
+ {
+ c = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1 >()] );
+ }
+
+ Real hD = 3.0*h*h - 2.0*(a*a + b*b + c*c - a*b - a*c - b*c);
+
+ if(hD < 0.0)
+ tmp = fabsMin(a,fabsMin(b,c)) + sign(value)*h;
+ else
+ tmp = (1.0/3.0) * ( a + b + c + sign(value)*sqrt(hD) );
+
+ atomicFabsMin(&cudaDofVector2[Entity.getIndex()],tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid(int i, int j, int k)
+{
+ tnlGridEntity< tnlGrid< 3,double, tnlHost, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j,k));
+ Entity.refresh();
+ int gid = Entity.getIndex();
+
+ if(abs(cudaDofVector[gid]) < 1.8*h)
+ cudaDofVector2[gid] = cudaDofVector[gid];
+ else
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector[gid]);
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlNarrowBand< tnlGrid< 3,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+ int gx = 0;
+ int gy = threadIdx.y;
+
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+
+ if(blockIdx.x==0)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx = 0;
+ gy = threadIdx.y;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+
+
+
+
+ else if(blockIdx.x==4)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx = 0;
+ gy = threadIdx.y;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==5)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==6)
+ {
+
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==7)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlNarrowBand< tnlGrid< 3,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gz = blockDim.z*blockIdx.z + threadIdx.z;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && solver->Mesh.getDimensions().z() > gz)
+ {
+ solver->initGrid(gx,gy,gz);
+ }
+
+
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(INT_MAX,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(-INT_MAX,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-al;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(cudaDofVector[index],cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//
+//
+//
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-al;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(cudaDofVector[index],cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//}
+#endif
+
+
+
+
+#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand3D_impl.h b/examples/narrow-band/tnlNarrowBand3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..eb446d9d2831967e9016b3eaaf326baf6751ab7c
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand3D_impl.h
@@ -0,0 +1,307 @@
+/***************************************************************************
+ tnlNarrowBand2D_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND3D_IMPL_H_
+#define TNLNARROWBAND3D_IMPL_H_
+
+#include "tnlNarrowBand.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlNarrowBand< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: tnlNarrowBand()
+:Entity(Mesh),
+ dofVector(Mesh),
+ dofVector2(Mesh)
+{
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+ dofVector2.load(initialCondition);
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ Entity.refresh();
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+// cout << "bla "<<endl;
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ for(int i=0; i< Mesh.getDimensions().x()*Mesh.getDimensions().y()*Mesh.getDimensions().z();i++)
+ {
+
+ if (abs(dofVector[i]) < 1.8*h)
+ dofVector2[i]=dofVector[i];
+ else
+ dofVector2[i]=INT_MAX*sign(dofVector[i]);
+ }
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+
+
+
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+ dofVector2.save("u-00001.tnl");
+
+ cout << "bla 3"<<endl;
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, Index k)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j,k));
+ this->Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ Real value = dofVector2[Entity.getIndex()];
+ Real a,b,c, tmp;
+
+ if( i == 0 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0>()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ {
+ a = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0>()] );
+ }
+
+ if( j == 0 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0>()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0>()];
+ else
+ {
+ b = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0>()] );
+ }
+
+ if( k == 0 )
+ c = dofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1>()];
+ else if( k == Mesh.getDimensions().z() - 1 )
+ c = dofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1>()];
+ else
+ {
+ c = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1>()] );
+ }
+
+ Real hD = 3.0*h*h - 2.0*(a*a+b*b+c*c-a*b-a*c-b*c);
+
+ if(hD < 0.0)
+ tmp = fabsMin(a,fabsMin(b,c)) + sign(value)*h;
+ else
+ tmp = (1.0/3.0) * ( a + b + c + sign(value)*sqrt(hD) );
+
+
+ dofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+}
+
+
+
+#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/examples/narrow-band/tnlNarrowBand_CUDA.h b/examples/narrow-band/tnlNarrowBand_CUDA.h
new file mode 100644
index 0000000000000000000000000000000000000000..8da92f5fc570275c85ebba69113f0e59a4be7988
--- /dev/null
+++ b/examples/narrow-band/tnlNarrowBand_CUDA.h
@@ -0,0 +1,203 @@
+/***************************************************************************
+ tnlNarrowBand_CUDA.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLNARROWBAND_H_
+#define TNLNARROWBAND_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+
+#include <functions/tnlMeshFunction.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlNarrowBand
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlNarrowBand();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+ bool run();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid();
+ __device__ void updateValue(const Index i, const Index j);
+ __device__ void updateValue(const Index i, const Index j, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ int* cudaStatusVector;
+ int counter;
+ int* reinitialize;
+ __device__ void setupSquare1000(Index i, Index j);
+ __device__ void setupSquare1100(Index i, Index j);
+ __device__ void setupSquare1010(Index i, Index j);
+ __device__ void setupSquare1001(Index i, Index j);
+ __device__ void setupSquare1110(Index i, Index j);
+ __device__ void setupSquare1101(Index i, Index j);
+ __device__ void setupSquare1011(Index i, Index j);
+ __device__ void setupSquare1111(Index i, Index j);
+ __device__ void setupSquare0000(Index i, Index j);
+ __device__ void setupSquare0100(Index i, Index j);
+ __device__ void setupSquare0010(Index i, Index j);
+ __device__ void setupSquare0001(Index i, Index j);
+ __device__ void setupSquare0110(Index i, Index j);
+ __device__ void setupSquare0101(Index i, Index j);
+ __device__ void setupSquare0011(Index i, Index j);
+ __device__ void setupSquare0111(Index i, Index j);
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+ int statusGridSize;
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data;
+
+
+ RealType h, tau, finalTime;
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+ bool run();
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid(int i, int j, int k);
+ __device__ void updateValue(const Index i, const Index j, const Index k);
+ __device__ void updateValue(const Index i, const Index j, const Index k, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlNarrowBand< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ int counter;
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data;
+
+ RealType h;
+
+
+};
+
+
+
+
+
+
+
+#ifdef HAVE_CUDA
+//template<int sweep_t>
+__global__ void runCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i);
+//__global__ void runCUDA(tnlNarrowBand< tnlGrid< 3,double, tnlHost, int >, double, int >* solver, int sweep, int i);
+
+__global__ void initCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+
+__global__ void initSetupGridCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+__global__ void initSetupGrid2CUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+__global__ void initSetupGrid1_2CUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+__global__ void runNarrowBandCUDA(tnlNarrowBand< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, double tau);
+//__global__ void initCUDA(tnlNarrowBand< tnlGrid< 3,double, tnlHost, int >, double, int >* solver);
+#endif
+
+
+
+#include "tnlNarrowBand2D_CUDA_v4_impl.h"
+// #include "tnlNarrowBand3D_CUDA_impl.h"
+
+#endif /* TNLNARROWBAND_H_ */
diff --git a/examples/navier-stokes/navierStokesBoundaryConditions_impl.h b/examples/navier-stokes/navierStokesBoundaryConditions_impl.h
index c6674391bc24cd313f78b6c703c6d786d91a181f..3eb66e8555419614e134d0dc6370461336dc3301 100644
--- a/examples/navier-stokes/navierStokesBoundaryConditions_impl.h
+++ b/examples/navier-stokes/navierStokesBoundaryConditions_impl.h
@@ -31,11 +31,19 @@ navierStokesBoundaryConditions< Mesh >::navierStokesBoundaryConditions()
template< typename Mesh >
bool navierStokesBoundaryConditions< Mesh >::setup( const Config::ParameterContainer& parameters )
{
+<<<<<<< HEAD
+ this -> maxInflowVelocity = parameters. getParameter< double >( "max-inflow-velocity" );
+ //this -> maxOutflowVelocity = parameters. getParameter< double >( "max-outflow-velocity" );
+ this -> startUp = parameters. getParameter< double >( "start-up" );
+ this -> T = parameters. getParameter< double >( "T" );
+ this -> R = parameters. getParameter< double >( "R" );
+=======
this->maxInflowVelocity = parameters. getParameter< double >( "max-inflow-velocity" );
//this->maxOutflowVelocity = parameters. getParameter< double >( "max-outflow-velocity" );
this->startUp = parameters. getParameter< double >( "start-up" );
this->T = parameters. getParameter< double >( "T" );
this->R = parameters. getParameter< double >( "R" );
+>>>>>>> develop
this->p0 = parameters. getParameter< double >( "p0" );
this->gamma = parameters. getParameter< double >( "gamma" );
return true;
diff --git a/examples/navier-stokes/navierStokesSolverMonitor_impl.h b/examples/navier-stokes/navierStokesSolverMonitor_impl.h
index 47e82e3bba940c430b5d16e377a6c1751a96b808..462d31257aabda14745385f13bb70a412b90c6ce 100644
--- a/examples/navier-stokes/navierStokesSolverMonitor_impl.h
+++ b/examples/navier-stokes/navierStokesSolverMonitor_impl.h
@@ -30,7 +30,11 @@ navierStokesSolverMonitor< Real, Index > :: navierStokesSolverMonitor()
template< typename Real, typename Index >
void navierStokesSolverMonitor< Real, Index > :: refresh()
{
+<<<<<<< HEAD
+ if( this -> verbose > 0 && this -> refresRate % this -> refreshRate == 0 )
+=======
if( this->verbose > 0 && this->refresRate % this->refreshRate == 0 )
+>>>>>>> develop
{
std::cout << "V=( " << uMax
<< " , " << uAvg
diff --git a/examples/transport-equation/CMakeLists.txt b/examples/transport-equation/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..9c4b587d180a06142149459094b3ca2e8c3e9150
--- /dev/null
+++ b/examples/transport-equation/CMakeLists.txt
@@ -0,0 +1,34 @@
+set( tnl_transport_equation_SOURCES
+ tnl-transport-equation.cpp
+ tnl-transport-equation.cu )
+
+set( tnl_transport_equation_HEADERS
+ tnl-transport-equation.h
+ transportEquationBuildConfigTag.h
+ transportEquationProblem.h
+ transportEquationProblem_impl.h )
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE( tnl-transport-equation${debugExt} tnl-transport-equation.cu)
+ target_link_libraries( tnl-transport-equation${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
+ CUDA_ADD_EXECUTABLE( tnl-transport-equation-eoc${debugExt} tnl-transport-equation-eoc.cu)
+ target_link_libraries( tnl-transport-equation-eoc${debugExt} tnl${debugExt}-${tnlVersion} ${CUSPARSE_LIBRARY} )
+
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE( tnl-transport-equation${debugExt} tnl-transport-equation.cpp)
+ target_link_libraries( tnl-transport-equation${debugExt} tnl${debugExt}-${tnlVersion} )
+ ADD_EXECUTABLE( tnl-transport-equation-eoc${debugExt} tnl-transport-equation-eoc.cpp)
+ target_link_libraries( tnl-transport-equation-eoc${debugExt} tnl${debugExt}-${tnlVersion} )
+ENDIF( BUILD_CUDA )
+
+
+INSTALL( TARGETS tnl-transport-equation${debugExt}
+ tnl-transport-equation-eoc${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+INSTALL( FILES tnl-run-transport-equation
+ ${tnl_transport_equation_SOURCES}
+ ${tnl_transport_equation_HEADERS}
+ DESTINATION share/tnl-${tnlVersion}/examples/transport-equation )
diff --git a/examples/transport-equation/tnl-run-transport-equation b/examples/transport-equation/tnl-run-transport-equation
new file mode 100644
index 0000000000000000000000000000000000000000..aa33723a4336b7c947c448e6486a2e324ef4edc7
--- /dev/null
+++ b/examples/transport-equation/tnl-run-transport-equation
@@ -0,0 +1,34 @@
+#!/usr/bin/env bash
+
+tnl-grid-setup --dimensions 2 \
+ --origin-x 0.0 \
+ --origin-x 0.0 \
+ --proportions-x 1.0 \
+ --proportions-y 1.0 \
+ --size-x 100 \
+ --size-y 100
+
+tnl-init --test-function heaviside-of-vector-norm \
+ --output-file init.tnl \
+ --coefficient -1.0 \
+ --vector-norm-center-0 0.5 \
+ --vector-norm-center-1 0.5 \
+ --vector-norm-center-2 0.5 \
+ --vector-norm-radius 0.2 \
+ --vector-norm-power 0.6 \
+ --vector-norm-max-norm false
+
+tnl-transport-equation --device host \
+ --initial-condition init.tnl \
+ --time-discretisation explicit \
+ --time-step 1.0e-5 \
+ --boundary-conditions-constant 0.0 \
+ --discrete-solver euler \
+ --snapshot-period 0.005 \
+ --final-time 0.1 \
+ --numerical-viscosity 1.0 \
+ --velocity-field-0-constant 1.0 \
+ --velocity-field-1-constant 1.0 \
+ --velocity-field-2-constant 1.0
+
+tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/transport-equation/tnl-run-transport-equation-eoc b/examples/transport-equation/tnl-run-transport-equation-eoc
new file mode 100755
index 0000000000000000000000000000000000000000..ec08d44172fdb38be9d9349013871fce7e6a6abf
--- /dev/null
+++ b/examples/transport-equation/tnl-run-transport-equation-eoc
@@ -0,0 +1,32 @@
+#!/usr/bin/env bash
+
+tnl-grid-setup --dimensions 2 \
+ --origin-x 0.0 \
+ --origin-x 0.0 \
+ --proportions-x 1.0 \
+ --proportions-y 1.0 \
+ --size-x 100 \
+ --size-y 100
+
+tnl-transport-equation-eoc --device host \
+ --initial-condition heaviside-vector-norm \
+ --heaviside-multiplicator 1.0 \
+ --vector-norm-multiplicator -1.0 \
+ --vector-norm-center-0 0.5 \
+ --vector-norm-center-1 0.5 \
+ --vector-norm-center-2 0.5 \
+ --vector-norm-radius 0.2 \
+ --vector-norm-power 0.6 \
+ --vector-norm-max-norm false \
+ --time-discretisation explicit \
+ --time-step 1.0e-5 \
+ --boundary-conditions-constant 0.0 \
+ --discrete-solver euler \
+ --snapshot-period 0.005 \
+ --final-time 0.1 \
+ --numerical-viscosity 1.0 \
+ --velocity-field-0-constant 1.0 \
+ --velocity-field-1-constant 1.0 \
+ --velocity-field-2-constant 1.0
+
+tnl-view --mesh mesh.tnl --input-files *tnl
diff --git a/examples/transport-equation/tnl-transport-equation-eoc.cpp b/examples/transport-equation/tnl-transport-equation-eoc.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..00e2e867a1c34e3a71073144e0209f15cba9c150
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation-eoc.cpp
@@ -0,0 +1,11 @@
+/***************************************************************************
+ tnl-transport-equation-eoc.cpp - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#include "tnl-transport-equation-eoc.h"
diff --git a/examples/transport-equation/tnl-transport-equation-eoc.cu b/examples/transport-equation/tnl-transport-equation-eoc.cu
new file mode 100644
index 0000000000000000000000000000000000000000..9a17dc4e1f6297eb2c1ec04dca8dc67535c8a762
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation-eoc.cu
@@ -0,0 +1,12 @@
+/***************************************************************************
+ tnl-transport-equation-eoc.cu - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#include "tnl-transport-equation-eoc.h"
diff --git a/examples/transport-equation/tnl-transport-equation-eoc.h b/examples/transport-equation/tnl-transport-equation-eoc.h
new file mode 100644
index 0000000000000000000000000000000000000000..08f0a6fbeb66a74177d52584344af718a911a074
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation-eoc.h
@@ -0,0 +1,145 @@
+/***************************************************************************
+ tnl-transport-equation-eoc.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#include <TNL/tnlConfig.h>
+#include <TNL/Solvers/Solver.h>
+#include <TNL/Solvers/BuildConfigTags.h>
+#include <TNL/Operators/DirichletBoundaryConditions.h>
+#include <TNL/Operators/NeumannBoundaryConditions.h>
+#include <TNL/Operators/Advection/LaxFridrichs.h>
+#include <TNL/Functions/Analytic/Constant.h>
+#include <TNL/Functions/Analytic/VectorNorm.h>
+#include <TNL/Operators/Analytic/Heaviside.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/Meshes/Grid.h>
+#include "transportEquationProblemEoc.h"
+#include "transportEquationBuildConfigTag.h"
+
+using namespace TNL;
+
+typedef transportEquationBuildConfigTag BuildConfig;
+
+/****
+ * Uncomment the following (and comment the previous line) for the complete build.
+ * This will include support for all floating point precisions, all indexing types
+ * and more solvers. You may then choose between them from the command line.
+ * The compile time may, however, take tens of minutes or even several hours,
+ * especially if CUDA is enabled. Use this, if you want, only for the final build,
+ * not in the development phase.
+ */
+//typedef tnlDefaultConfigTag BuildConfig;
+
+template< typename ConfigTag >class advectionConfig
+{
+ public:
+ static void configSetup( Config::ConfigDescription& config )
+ {
+ config.addDelimiter( "Transport equation settings:" );
+ config.addDelimiter( "Initial condition" );
+ config.addEntry< String >( "initial-condition", "Set type of initial condition.", "heaviside-vector-norm" );
+ config.addEntryEnum< String >( "heaviside-vector-norm" );
+ Functions::Analytic::VectorNorm< 3, double >::configSetup( config, "vector-norm-" );
+ Operators::Analytic::Heaviside< 3, double >::configSetup( config, "heaviside-" );
+ Operators::Analytic::Shift< 3, double >::configSetup( config, "heaviside-" );
+
+ config.addDelimiter( "Velocity field" );
+ config.addEntry< String >( "velocity-field", "Type of velocity field.", "constant" );
+ config.addEntryEnum< String >( "constant" );
+ Functions::VectorField< 3, Functions::Analytic::Constant< 3 > >::configSetup( config, "velocity-field-" );
+
+ config.addDelimiter( "Numerical scheme" );
+ typedef Meshes::Grid< 3 > MeshType;
+ Operators::Advection::LaxFridrichs< MeshType >::configSetup( config );
+
+ config.addDelimiter( "Boundary conditions" );
+ config.addEntry< String >( "boundary-conditions-type", "Choose the boundary conditions type.", "dirichlet");
+ config.addEntryEnum< String >( "dirichlet" );
+ config.addEntryEnum< String >( "neumann" );
+ config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions.", 0.0 );
+ }
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename MeshType,
+ typename ConfigTag,
+ typename SolverStarter >
+class advectionSetter
+{
+ public:
+
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ static const int Dimensions = MeshType::getMeshDimensions();
+
+ template< typename Problem >
+ static bool callSolverStarter( const Config::ParameterContainer& parameters )
+ {
+ SolverStarter solverStarter;
+ return solverStarter.template run< Problem >( parameters );
+ }
+
+ template< typename DifferentialOperatorType >
+ static bool setBoundaryConditionsType( const Config::ParameterContainer& parameters )
+ {
+ typedef Functions::Analytic::Constant< Dimensions, Real > ConstantFunctionType;
+ String boundaryConditionsType = parameters.getParameter< String >( "boundary-conditions-type" );
+ if( boundaryConditionsType == "dirichlet" )
+ {
+ typedef Operators::DirichletBoundaryConditions< MeshType, ConstantFunctionType, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
+ typedef transportEquationProblemEoc< MeshType, BoundaryConditions, ConstantFunctionType, DifferentialOperatorType > Problem;
+ return callSolverStarter< Problem >( parameters );
+ }
+ if( boundaryConditionsType == "neumann" )
+ {
+ typedef Operators::DirichletBoundaryConditions< MeshType, ConstantFunctionType, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
+ typedef transportEquationProblemEoc< MeshType, BoundaryConditions, ConstantFunctionType, DifferentialOperatorType > Problem;
+ return callSolverStarter< Problem >( parameters );
+ }
+ std::cerr << "Unknown boundary conditions type: " << boundaryConditionsType << "." << std::endl;
+ return false;
+ }
+
+ template< typename VelocityFieldType >
+ static bool setDifferentialOperatorType( const Config::ParameterContainer& parameters )
+ {
+ typedef Operators::Advection::LaxFridrichs< MeshType, Real, Index, VelocityFieldType > DifferentialOperatorType;
+ return setBoundaryConditionsType< DifferentialOperatorType >( parameters );
+ }
+
+ static bool setVelocityFieldType( const Config::ParameterContainer& parameters )
+ {
+ String velocityFieldType = parameters.getParameter< String >( "velocity-field" );
+ if( velocityFieldType == "constant" )
+ {
+ typedef Functions::Analytic::Constant< Dimensions, RealType > VelocityFieldType;
+ return setDifferentialOperatorType< VelocityFieldType >( parameters );
+ }
+ return false;
+ }
+
+ static bool run( const Config::ParameterContainer& parameters )
+ {
+ return setVelocityFieldType( parameters );
+ }
+};
+
+int main( int argc, char* argv[] )
+{
+ Solvers::Solver< advectionSetter, advectionConfig, BuildConfig > solver;
+ if( ! solver. run( argc, argv ) )
+ return EXIT_FAILURE;
+ return EXIT_SUCCESS;
+}
+
diff --git a/examples/transport-equation/tnl-transport-equation.cpp b/examples/transport-equation/tnl-transport-equation.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..f8cc396d31d0592dc8626c370015886f1d3b2dc4
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation.cpp
@@ -0,0 +1,11 @@
+/***************************************************************************
+ tnl-transport-equation.cpp - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#include "tnl-transport-equation.h"
diff --git a/examples/transport-equation/tnl-transport-equation.cu b/examples/transport-equation/tnl-transport-equation.cu
new file mode 100644
index 0000000000000000000000000000000000000000..068984a891fdc9df2c58ad24ae7a5c9b87ef015d
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation.cu
@@ -0,0 +1,11 @@
+/***************************************************************************
+ tnl-transport-equation.cu - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#include "tnl-transport-equation.h"
diff --git a/examples/transport-equation/tnl-transport-equation.h b/examples/transport-equation/tnl-transport-equation.h
new file mode 100644
index 0000000000000000000000000000000000000000..ad9cc970a3000384cd4a49e5d369f1326c59453d
--- /dev/null
+++ b/examples/transport-equation/tnl-transport-equation.h
@@ -0,0 +1,132 @@
+/***************************************************************************
+ tnl-transport-equation.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#include <TNL/tnlConfig.h>
+#include <TNL/Solvers/Solver.h>
+#include <TNL/Solvers/BuildConfigTags.h>
+#include <TNL/Operators/DirichletBoundaryConditions.h>
+#include <TNL/Operators/NeumannBoundaryConditions.h>
+#include <TNL/Operators/Advection/LaxFridrichs.h>
+#include <TNL/Functions/Analytic/Constant.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/Meshes/Grid.h>
+#include "transportEquationProblem.h"
+#include "transportEquationBuildConfigTag.h"
+
+using namespace TNL;
+
+typedef transportEquationBuildConfigTag BuildConfig;
+
+/****
+ * Uncomment the following (and comment the previous line) for the complete build.
+ * This will include support for all floating point precisions, all indexing types
+ * and more solvers. You may then choose between them from the command line.
+ * The compile time may, however, take tens of minutes or even several hours,
+ * especially if CUDA is enabled. Use this, if you want, only for the final build,
+ * not in the development phase.
+ */
+//typedef tnlDefaultConfigTag BuildConfig;
+
+template< typename ConfigTag >class advectionConfig
+{
+ public:
+ static void configSetup( Config::ConfigDescription& config )
+ {
+ config.addDelimiter( "Transport equation settings:" );
+ config.addEntry< String >( "velocity-field", "Type of velocity field.", "constant" );
+ config.addEntryEnum< String >( "constant" );
+ Functions::VectorField< 3, Functions::Analytic::Constant< 3 > >::configSetup( config, "velocity-field-" );
+
+ typedef Meshes::Grid< 3 > MeshType;
+ Operators::Advection::LaxFridrichs< MeshType >::configSetup( config );
+
+ config.addEntry< String >( "boundary-conditions-type", "Choose the boundary conditions type.", "dirichlet");
+ config.addEntryEnum< String >( "dirichlet" );
+ config.addEntryEnum< String >( "neumann" );
+ config.addEntry< double >( "boundary-conditions-constant", "This sets a value in case of the constant boundary conditions." );
+ }
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename MeshType,
+ typename ConfigTag,
+ typename SolverStarter >
+class advectionSetter
+{
+ public:
+
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ static const int Dimensions = MeshType::getMeshDimensions();
+
+ template< typename Problem >
+ static bool callSolverStarter( const Config::ParameterContainer& parameters )
+ {
+ SolverStarter solverStarter;
+ return solverStarter.template run< Problem >( parameters );
+ }
+
+ template< typename DifferentialOperatorType >
+ static bool setBoundaryConditionsType( const Config::ParameterContainer& parameters )
+ {
+ typedef Functions::Analytic::Constant< Dimensions, Real > ConstantFunctionType;
+ String boundaryConditionsType = parameters.getParameter< String >( "boundary-conditions-type" );
+ if( boundaryConditionsType == "dirichlet" )
+ {
+ typedef Operators::DirichletBoundaryConditions< MeshType, ConstantFunctionType, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
+ typedef transportEquationProblem< MeshType, BoundaryConditions, ConstantFunctionType, DifferentialOperatorType > Problem;
+ return callSolverStarter< Problem >( parameters );
+ }
+ if( boundaryConditionsType == "neumann" )
+ {
+ typedef Operators::DirichletBoundaryConditions< MeshType, ConstantFunctionType, MeshType::getMeshDimensions(), Real, Index > BoundaryConditions;
+ typedef transportEquationProblem< MeshType, BoundaryConditions, ConstantFunctionType, DifferentialOperatorType > Problem;
+ return callSolverStarter< Problem >( parameters );
+ }
+ std::cerr << "Unknown boundary conditions type: " << boundaryConditionsType << "." << std::endl;
+ return false;
+ }
+
+ template< typename VelocityFieldType >
+ static bool setDifferentialOperatorType( const Config::ParameterContainer& parameters )
+ {
+ typedef Operators::Advection::LaxFridrichs< MeshType, Real, Index, VelocityFieldType > DifferentialOperatorType;
+ return setBoundaryConditionsType< DifferentialOperatorType >( parameters );
+ }
+
+ static bool setVelocityFieldType( const Config::ParameterContainer& parameters )
+ {
+ String velocityFieldType = parameters.getParameter< String >( "velocity-field" );
+ if( velocityFieldType == "constant" )
+ {
+ typedef Functions::Analytic::Constant< Dimensions, RealType > VelocityFieldType;
+ return setDifferentialOperatorType< VelocityFieldType >( parameters );
+ }
+ return false;
+ }
+
+ static bool run( const Config::ParameterContainer& parameters )
+ {
+ return setVelocityFieldType( parameters );
+ }
+};
+
+int main( int argc, char* argv[] )
+{
+ Solvers::Solver< advectionSetter, advectionConfig, BuildConfig > solver;
+ if( ! solver. run( argc, argv ) )
+ return EXIT_FAILURE;
+ return EXIT_SUCCESS;
+}
+
diff --git a/examples/transport-equation/transportEquationBuildConfigTag.h b/examples/transport-equation/transportEquationBuildConfigTag.h
new file mode 100644
index 0000000000000000000000000000000000000000..df87a06b2b818ec466449ccab911cbcccd8bfb4a
--- /dev/null
+++ b/examples/transport-equation/transportEquationBuildConfigTag.h
@@ -0,0 +1,58 @@
+/***************************************************************************
+ transportEquationBuildConfigTag.cpp - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+
+#pragma once
+
+#include <TNL/Solvers/BuildConfigTags.h>
+
+namespace TNL {
+
+class transportEquationBuildConfigTag{};
+
+namespace Solvers {
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct ConfigTagReal< transportEquationBuildConfigTag, float > { enum { enabled = false }; };
+template<> struct ConfigTagReal< transportEquationBuildConfigTag, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct ConfigTagIndex< transportEquationBuildConfigTag, short int >{ enum { enabled = false }; };
+template<> struct ConfigTagIndex< transportEquationBuildConfigTag, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of Grid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct ConfigTagMesh< transportEquationBuildConfigTag, Meshes::Grid< Dimensions, Real, Device, Index > >
+ { enum { enabled = ConfigTagDimensions< transportEquationBuildConfigTag, Dimensions >::enabled &&
+ ConfigTagReal< transportEquationBuildConfigTag, Real >::enabled &&
+ ConfigTagDevice< transportEquationBuildConfigTag, Device >::enabled &&
+ ConfigTagIndex< transportEquationBuildConfigTag, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct ConfigTagTimeDiscretisation< transportEquationBuildConfigTag, ExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct ConfigTagTimeDiscretisation< transportEquationBuildConfigTag, SemiImplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct ConfigTagTimeDiscretisation< transportEquationBuildConfigTag, ImplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct ConfigTagExplicitSolver< transportEquationBuildConfigTag, Solvers::ExplicitEulerSolverTag >{ enum { enabled = true }; };
+
+} // namespace Solvers
+} // namespace TNL
diff --git a/examples/advection/advectionProblem.h b/examples/transport-equation/transportEquationProblem.h
similarity index 76%
rename from examples/advection/advectionProblem.h
rename to examples/transport-equation/transportEquationProblem.h
index a44048c08c7899a47a8ee27a6f91a5d758eff1f4..8e7e164ab2688e600ad8a1578985a1f90d834138 100644
--- a/examples/advection/advectionProblem.h
+++ b/examples/transport-equation/transportEquationProblem.h
@@ -1,5 +1,14 @@
-#ifndef advectionPROBLEM_H_
-#define advectionPROBLEM_H_
+/***************************************************************************
+ transportEquationProblem.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
#include <TNL/Problems/PDEProblem.h>
#include <TNL/Functions/MeshFunction.h>
@@ -12,12 +21,12 @@ namespace TNL {
template< typename Mesh,
typename BoundaryCondition,
typename RightHandSide,
- typename DifferentialOperator >
-class advectionProblem:
- public PDEProblem< Mesh,
- typename DifferentialOperator::RealType,
- typename Mesh::DeviceType,
- typename DifferentialOperator::IndexType >
+ typename DifferentialOperator >
+class transportEquationProblem:
+public PDEProblem< Mesh,
+ typename DifferentialOperator::RealType,
+ typename Mesh::DeviceType,
+ typename DifferentialOperator::IndexType >
{
public:
@@ -30,7 +39,9 @@ class advectionProblem:
typedef SharedPointer< DifferentialOperator > DifferentialOperatorPointer;
typedef SharedPointer< BoundaryCondition > BoundaryConditionPointer;
typedef SharedPointer< RightHandSide, DeviceType > RightHandSidePointer;
-
+ typedef typename DifferentialOperator::VelocityFieldType VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType, DeviceType > VelocityFieldPointer;
+
using typename BaseType::MeshType;
using typename BaseType::MeshPointer;
using typename BaseType::DofVectorType;
@@ -93,13 +104,11 @@ class advectionProblem:
BoundaryConditionPointer boundaryConditionPointer;
- RightHandSidePointer rightHandSidePointer;
+ RightHandSidePointer rightHandSidePointer;
- String velocityType;
+ VelocityFieldPointer velocityField;
};
} // namespace TNL
-#include "advectionProblem_impl.h"
-
-#endif /* advectionPROBLEM_H_ */
+#include "transportEquationProblem_impl.h"
diff --git a/examples/transport-equation/transportEquationProblemEoc.h b/examples/transport-equation/transportEquationProblemEoc.h
new file mode 100644
index 0000000000000000000000000000000000000000..0f2c8ea3ae6749b2f5c3b9896c2f4a8b5ad35500
--- /dev/null
+++ b/examples/transport-equation/transportEquationProblemEoc.h
@@ -0,0 +1,95 @@
+/***************************************************************************
+ transportEquationProblemEoc.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Problems/PDEProblem.h>
+#include <TNL/Functions/MeshFunction.h>
+#include <TNL/SharedPointer.h>
+#include "transportEquationProblem.h"
+
+using namespace TNL::Problems;
+
+namespace TNL {
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+class transportEquationProblemEoc:
+public transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >
+{
+ public:
+
+ typedef typename DifferentialOperator::RealType RealType;
+ typedef typename Mesh::DeviceType DeviceType;
+ typedef typename DifferentialOperator::IndexType IndexType;
+ typedef Functions::MeshFunction< Mesh > MeshFunctionType;
+ typedef transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator > BaseType;
+ typedef SharedPointer< MeshFunctionType, DeviceType > MeshFunctionPointer;
+ typedef SharedPointer< DifferentialOperator > DifferentialOperatorPointer;
+ typedef SharedPointer< BoundaryCondition > BoundaryConditionPointer;
+ typedef SharedPointer< RightHandSide, DeviceType > RightHandSidePointer;
+ typedef typename DifferentialOperator::VelocityFieldType VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType, DeviceType > VelocityFieldPointer;
+
+
+ using typename BaseType::MeshType;
+ using typename BaseType::MeshPointer;
+ using typename BaseType::DofVectorType;
+ using typename BaseType::DofVectorPointer;
+ using typename BaseType::MeshDependentDataType;
+ using typename BaseType::MeshDependentDataPointer;
+
+ //using BaseType::getExplicitRHS;
+
+ static String getTypeStatic();
+
+ String getPrologHeader() const;
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ bool setInitialCondition( const Config::ParameterContainer& parameters,
+ const MeshPointer& mesh,
+ DofVectorPointer& dofs,
+ MeshDependentDataPointer& meshDependentData );
+
+ /*template< typename Matrix >
+ bool setupLinearSystem( const MeshPointer& mesh,
+ Matrix& matrix );
+
+ bool makeSnapshot( const RealType& time,
+ const IndexType& step,
+ const MeshPointer& mesh,
+ DofVectorPointer& dofs,
+ MeshDependentDataPointer& meshDependentData );
+
+ IndexType getDofs( const MeshPointer& mesh ) const;
+
+ void bindDofs( const MeshPointer& mesh,
+ DofVectorPointer& dofs );
+
+ void getExplicitRHS( const RealType& time,
+ const RealType& tau,
+ const MeshPointer& mesh,
+ DofVectorPointer& _u,
+ DofVectorPointer& _fu,
+ MeshDependentDataPointer& meshDependentData );*/
+
+
+ protected:
+};
+
+} // namespace TNL
+
+#include "transportEquationProblemEoc_impl.h"
+
diff --git a/examples/transport-equation/transportEquationProblemEoc_impl.h b/examples/transport-equation/transportEquationProblemEoc_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..c238bc311b5ea0e25745eef64e58ad9d1e3caa57
--- /dev/null
+++ b/examples/transport-equation/transportEquationProblemEoc_impl.h
@@ -0,0 +1,151 @@
+/***************************************************************************
+ transportEquationProblemEoc_impl.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/FileName.h>
+#include <TNL/Matrices/MatrixSetter.h>
+#include <TNL/Solvers/PDE/ExplicitUpdater.h>
+#include <TNL/Solvers/PDE/LinearSystemAssembler.h>
+#include <TNL/Solvers/PDE/BackwardTimeDiscretisation.h>
+#include <TNL/Functions/Analytic/Paraboloid.h>
+#include <TNL/Operators/Analytic/Heaviside.h>
+#include <TNL/Operators/Analytic/Shift.h>
+
+#include "transportEquationProblem.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+String
+transportEquationProblemEoc< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getTypeStatic()
+{
+ return String( "transportEquationProblemEoc< " ) + Mesh :: getTypeStatic() + " >";
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+String
+transportEquationProblemEoc< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getPrologHeader() const
+{
+ return String( "Transport Equation EOC" );
+}
+
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+bool
+transportEquationProblemEoc< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ if( ! this->velocityField->setup( meshPointer, parameters, prefix + "velocity-field-" ) ||
+ ! this->differentialOperatorPointer->setup( meshPointer, parameters, prefix ) ||
+ ! this->boundaryConditionPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) )
+ return false;
+
+ /****
+ * Render the exact solution
+ */
+ const String& initialCondition = parameters.getParameter< String >( "initial-condition" );
+ const double& finalTime = parameters.getParameter< double >( "final-time" );
+ const double& snapshotPeriod = parameters.getParameter< double >( "snapshot-period" );
+ static const int Dimensions = Mesh::getMeshDimensions();
+ typedef typename MeshPointer::ObjectType MeshType;
+ typedef Functions::MeshFunction< MeshType > MeshFunction;
+ SharedPointer< MeshFunction > u( meshPointer );
+ if( initialCondition == "heaviside-vector-norm" )
+ {
+ typedef Functions::Analytic::VectorNorm< Dimensions, RealType > VectorNormType;
+ typedef Operators::Analytic::Heaviside< Dimensions, RealType > HeavisideType;
+ typedef Functions::OperatorFunction< HeavisideType, VectorNormType > InitialConditionType;
+ String velocityFieldType = parameters.getParameter< String >( "velocity-field" );
+ if( velocityFieldType == "constant" )
+ {
+ typedef Operators::Analytic::Shift< Dimensions, RealType > ShiftOperatorType;
+ typedef Functions::OperatorFunction< ShiftOperatorType, InitialConditionType > ExactSolutionType;
+ SharedPointer< ExactSolutionType, Devices::Host > exactSolution;
+ if( ! exactSolution->getFunction().setup( parameters, prefix + "vector-norm-" ) ||
+ ! exactSolution->getOperator().setup( parameters, prefix + "heaviside-" ) )
+ return false;
+ Containers::StaticVector< Dimensions, RealType > velocity;
+ for( int i = 0; i < Dimensions; i++ )
+ velocity[ i ] = parameters.getParameter< double >( "velocity-field-" + String( i ) + "-constant" );
+
+ Functions::MeshFunctionEvaluator< MeshFunction, ExactSolutionType > evaluator;
+ RealType time( 0.0 );
+ int step( 0 );
+ exactSolution->getOperator().setShift( 0.0 * velocity );
+ evaluator.evaluate( u, exactSolution, time );
+ FileName fileName;
+ fileName.setFileNameBase( "exact-u-" );
+ fileName.setExtension( "tnl" );
+ fileName.setIndex( step );
+ if( ! u->save( fileName.getFileName() ) )
+ return false;
+ while( time < finalTime )
+ {
+ time += snapshotPeriod;
+ if( time > finalTime )
+ time = finalTime;
+ exactSolution->getOperator().setShift( time * velocity );
+ std::cerr << time * velocity << std::endl;
+ std::cerr << exactSolution->getOperator().getShift() << std::endl;
+ evaluator.evaluate( u, exactSolution, time );
+ fileName.setIndex( ++step );
+ if( ! u->save( fileName.getFileName() ) )
+ return false;
+ }
+ }
+ if( velocityFieldType == "rotation" )
+ {
+ // TODO: implement this using RotationXY operator
+ }
+ }
+
+ return true;
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+bool
+transportEquationProblemEoc< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+setInitialCondition( const Config::ParameterContainer& parameters,
+ const MeshPointer& meshPointer,
+ DofVectorPointer& dofs,
+ MeshDependentDataPointer& meshDependentData )
+{
+ this->bindDofs( meshPointer, dofs );
+ //const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );
+ FileName fileName;
+ fileName.setFileNameBase( "exact-u-" );
+ fileName.setExtension( "tnl" );
+ fileName.setIndex( 0 );
+ if( ! this->uPointer->boundLoad( fileName.getFileName() ) )
+ {
+ std::cerr << "I am not able to load the initial condition from the file " << fileName.getFileName() << "." << std::endl;
+ return false;
+ }
+ return true;
+}
+
+} // namespace TNL
diff --git a/examples/transport-equation/transportEquationProblem_impl.h b/examples/transport-equation/transportEquationProblem_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..7e2023ad2fb39a2381b71ecca0705fa0d09ff56f
--- /dev/null
+++ b/examples/transport-equation/transportEquationProblem_impl.h
@@ -0,0 +1,233 @@
+/***************************************************************************
+ transportEquationProblem_impl.h - description
+ -------------------
+ begin : Feb 10, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/FileName.h>
+#include <TNL/Matrices/MatrixSetter.h>
+#include <TNL/Solvers/PDE/ExplicitUpdater.h>
+#include <TNL/Solvers/PDE/LinearSystemAssembler.h>
+#include <TNL/Solvers/PDE/BackwardTimeDiscretisation.h>
+
+#include "transportEquationProblem.h"
+
+namespace TNL {
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+String
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getTypeStatic()
+{
+ return String( "transportEquationProblem< " ) + Mesh :: getTypeStatic() + " >";
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+String
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getPrologHeader() const
+{
+ return String( "Transport Equation" );
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+void
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+writeProlog( Logger& logger, const Config::ParameterContainer& parameters ) const
+{
+ /****
+ * Add data you want to have in the computation report (log) as follows:
+ * logger.writeParameter< double >( "Parameter description", parameter );
+ */
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+bool
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ if( ! this->velocityField->setup( meshPointer, parameters, prefix + "velocity-field-" ) ||
+ ! this->differentialOperatorPointer->setup( meshPointer, parameters, prefix ) ||
+ ! this->boundaryConditionPointer->setup( meshPointer, parameters, prefix + "boundary-conditions-" ) )
+ return false;
+ return true;
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+typename transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::IndexType
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getDofs( const MeshPointer& mesh ) const
+{
+ /****
+ * Return number of DOFs (degrees of freedom) i.e. number
+ * of unknowns to be resolved by the main solver.
+ */
+ return mesh->template getEntitiesCount< typename MeshType::Cell >();
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+void
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+bindDofs( const MeshPointer& meshPointer,
+ DofVectorPointer& dofVector )
+{
+ const IndexType dofs = meshPointer->template getEntitiesCount< typename MeshType::Cell >();
+ this->uPointer->bind( meshPointer, dofVector );
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+bool
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+setInitialCondition( const Config::ParameterContainer& parameters,
+ const MeshPointer& meshPointer,
+ DofVectorPointer& dofs,
+ MeshDependentDataPointer& meshDependentData )
+{
+ this->bindDofs( meshPointer, dofs );
+ const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );
+ if( ! this->uPointer->boundLoad( initialConditionFile ) )
+ {
+ std::cerr << "I am not able to load the initial condition from the file " << initialConditionFile << "." << std::endl;
+ return false;
+ }
+ return true;
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+ template< typename Matrix >
+bool
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+setupLinearSystem( const MeshPointer& mesh,
+ Matrix& matrix )
+{
+ /*const IndexType dofs = this->getDofs( mesh );
+ typedef typename Matrix::ObjectType::CompressedRowsLengthsVector CompressedRowsLengthsVectorType;
+ SharedPointer< CompressedRowsLengthsVectorType > rowLengths;
+ if( ! rowLengths->setSize( dofs ) )
+ return false;
+ Matrices::MatrixSetter< MeshType, DifferentialOperator, BoundaryCondition, CompressedRowsLengthsVectorType > matrixSetter;
+ matrixSetter.template getCompressedRowsLengths< typename Mesh::Cell >( mesh,
+ differentialOperatorPointer,
+ boundaryConditionPointer,
+ rowLengths );
+ matrix->setDimensions( dofs, dofs );
+ if( ! matrix->setCompressedRowsLengths( *rowLengths ) )
+ return false;*/
+ return true;
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+bool
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+makeSnapshot( const RealType& time,
+ const IndexType& step,
+ const MeshPointer& mesh,
+ DofVectorPointer& dofs,
+ MeshDependentDataPointer& meshDependentData )
+{
+ std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
+ this->bindDofs( mesh, dofs );
+ FileName fileName;
+ fileName.setFileNameBase( "u-" );
+ fileName.setExtension( "tnl" );
+ fileName.setIndex( step );
+ if( ! dofs->save( fileName.getFileName() ) )
+ return false;
+ return true;
+}
+
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+void
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+getExplicitRHS( const RealType& time,
+ const RealType& tau,
+ const MeshPointer& mesh,
+ DofVectorPointer& _u,
+ DofVectorPointer& _fu,
+ MeshDependentDataPointer& meshDependentData )
+{
+ /****
+ * If you use an explicit solver like Euler or Merson, you
+ * need to implement this method. Compute the right-hand side of
+ *
+ * d/dt u(x) = fu( x, u )
+ *
+ * You may use supporting mesh dependent data if you need.
+ */
+ typedef typename MeshType::Cell Cell;
+ int count = ::sqrt(mesh->template getEntitiesCount< Cell >());
+ this->bindDofs( mesh, _u );
+ Solvers::PDE::ExplicitUpdater< Mesh, MeshFunctionType, DifferentialOperator, BoundaryCondition, RightHandSide > explicitUpdater;
+ SharedPointer< MeshFunctionType > u( mesh, _u );
+ SharedPointer< MeshFunctionType > fu( mesh, _fu );
+ differentialOperatorPointer->setTau(tau);
+ differentialOperatorPointer->setVelocityField( this->velocityField );
+ explicitUpdater.template update< typename Mesh::Cell >( time,
+ mesh,
+ this->differentialOperatorPointer,
+ this->boundaryConditionPointer,
+ this->rightHandSidePointer,
+ u,
+ fu );
+ /*BoundaryConditionsSetter< MeshFunctionType, BoundaryCondition > boundaryConditionsSetter;
+ boundaryConditionsSetter.template apply< typename Mesh::Cell >(
+ this->boundaryCondition,
+ time + tau,
+ u ); */
+}
+template< typename Mesh,
+ typename BoundaryCondition,
+ typename RightHandSide,
+ typename DifferentialOperator >
+ template< typename Matrix >
+void
+transportEquationProblem< Mesh, BoundaryCondition, RightHandSide, DifferentialOperator >::
+assemblyLinearSystem( const RealType& time,
+ const RealType& tau,
+ const MeshPointer& mesh,
+ DofVectorPointer& _u,
+ Matrix& matrix,
+ DofVectorPointer& b,
+ MeshDependentDataPointer& meshDependentData )
+{
+}
+
+} // namespace TNL
diff --git a/src/TNL/Config/ConfigEntryBase.h b/src/TNL/Config/ConfigEntryBase.h
index 44b6d4e8f5b6a9abb848be88fd69b8ba7e81e4ff..f69729ef5aaaa0d8120307eecfc8608d12742dfc 100644
--- a/src/TNL/Config/ConfigEntryBase.h
+++ b/src/TNL/Config/ConfigEntryBase.h
@@ -43,8 +43,7 @@ struct ConfigEntryBase
virtual void printEnumValues() const {};
- // TODO: Fix this -- uncommenting leads to SIGSEGV (for example in tnl-init)
- //virtual ~ConfigEntryBase() {};
+ virtual ~ConfigEntryBase() {};
};
} // namespace Config
diff --git a/src/TNL/Config/ParameterContainer.h b/src/TNL/Config/ParameterContainer.h
index 5f6df343000fe3a28eb1afe5edc1f54156ba451d..5ac93da80dfda6eac8333d6f08540ea6683f4d71 100644
--- a/src/TNL/Config/ParameterContainer.h
+++ b/src/TNL/Config/ParameterContainer.h
@@ -14,6 +14,7 @@
#include <TNL/Config/ConfigDescription.h>
#include <TNL/mpi-supp.h>
#include <TNL/param-types.h>
+//#include <TNL/Debugging/StackBacktrace.h>
namespace TNL {
namespace Config {
@@ -62,7 +63,7 @@ class ParameterContainer
template< class T > bool getParameter( const String& name,
T& value,
- bool verbose = false ) const
+ bool verbose = true ) const
{
int i;
const int size = parameters. getSize();
@@ -73,7 +74,10 @@ class ParameterContainer
return true;
}
if( verbose )
+ {
std::cerr << "Missing parameter '" << name << "'." << std::endl;
+ throw(0); //PrintStackBacktrace;
+ }
return false;
}
diff --git a/src/TNL/Containers/Algorithms/CudaReductionKernel.h b/src/TNL/Containers/Algorithms/CudaReductionKernel.h
index e2ab7e2cebc348e0797cf11a88775a3e7b5c388d..98a8841a36917cbb420dd54ef25c619a699853fb 100644
--- a/src/TNL/Containers/Algorithms/CudaReductionKernel.h
+++ b/src/TNL/Containers/Algorithms/CudaReductionKernel.h
@@ -123,7 +123,7 @@ CudaReductionKernel( Operation operation,
/***
* This runs in one warp so it is synchronized implicitly.
- */
+ */
if( tid < 32 )
{
volatile ResultType* vsdata = sdata;
@@ -132,6 +132,8 @@ CudaReductionKernel( Operation operation,
operation.commonReductionOnDevice( vsdata[ tid ], vsdata[ tid + 32 ] );
//printf( "4: tid %d data %f \n", tid, sdata[ tid ] );
}
+ // TODO: If blocksize == 32, the following does not work
+ // We do not check if tid < 16. Fix it!!!
if( blockSize >= 32 )
{
operation.commonReductionOnDevice( vsdata[ tid ], vsdata[ tid + 16 ] );
diff --git a/src/TNL/Containers/Array_impl.h b/src/TNL/Containers/Array_impl.h
index 65c2bc58f487460137b3bc4996264552345ab062..e9d34f860fa6c0dcff23e9847c311a3fe4d85e45 100644
--- a/src/TNL/Containers/Array_impl.h
+++ b/src/TNL/Containers/Array_impl.h
@@ -22,6 +22,8 @@
namespace TNL {
namespace Containers {
+using namespace std;
+
template< typename Element,
typename Device,
typename Index >
@@ -177,9 +179,9 @@ setSize( const Index size )
this->size = size;
if( size > 0 && ! this->allocationPointer )
{
- std::cerr << "I am not able to allocate new array with size "
- << ( double ) this->size * sizeof( ElementType ) / 1.0e9 << " GB." << std::endl;
- this->size = 0;
+ cerr << "I am not able to allocate new array with size "
+ << ( double ) this->size * sizeof( ElementType ) / 1.0e9 << " GB." << endl;
+ this -> size = 0;
return false;
}
return true;
@@ -299,7 +301,7 @@ Index
Array< Element, Device, Index >::
getSize() const
{
- return this->size;
+ return this -> size;
}
template< typename Element,
@@ -411,7 +413,7 @@ bool
Array< Element, Device, Index >::
operator == ( const ArrayT& array ) const
{
- if( array. getSize() != this->getSize() )
+ if( array. getSize() != this -> getSize() )
return false;
if( this->getSize() == 0 )
return true;
@@ -449,7 +451,7 @@ template< typename Element,
__cuda_callable__
const Element* Array< Element, Device, Index > :: getData() const
{
- return this->data;
+ return this -> data;
}
template< typename Element,
@@ -458,7 +460,7 @@ template< typename Element,
__cuda_callable__
Element* Array< Element, Device, Index > :: getData()
{
- return this->data;
+ return this -> data;
}
template< typename Element,
@@ -495,8 +497,8 @@ bool Array< Element, Device, Index > :: save( File& file ) const
#endif
if( this->size != 0 && ! ArrayIO< Element, Device, Index >::save( file, this->data, this->size ) )
{
- std::cerr << "I was not able to save " << this->getType()
- << " with size " << this->getSize() << std::endl;
+ cerr << "I was not able to save " << this->getType()
+ << " with size " << this -> getSize() << endl;
return false;
}
return true;
@@ -521,7 +523,7 @@ load( File& file )
#endif
if( _size < 0 )
{
- std::cerr << "Error: The size " << _size << " of the file is not a positive number or zero." << std::endl;
+ cerr << "Error: The size " << _size << " of the file is not a positive number or zero." << endl;
return false;
}
setSize( _size );
@@ -529,8 +531,8 @@ load( File& file )
{
if( ! ArrayIO< Element, Device, Index >::load( file, this->data, this->size ) )
{
- std::cerr << "I was not able to load " << this->getType()
- << " with size " << this->getSize() << std::endl;
+ cerr << "I was not able to load " << this->getType()
+ << " with size " << this -> getSize() << endl;
return false;
}
}
@@ -556,7 +558,7 @@ boundLoad( File& file )
#endif
if( _size < 0 )
{
- std::cerr << "Error: The size " << _size << " of the file is not a positive number or zero." << std::endl;
+ cerr << "Error: The size " << _size << " of the file is not a positive number or zero." << endl;
return false;
}
if( this->getSize() != 0 )
@@ -573,8 +575,8 @@ boundLoad( File& file )
{
if( ! ArrayIO< Element, Device, Index >::load( file, this->data, this->size ) )
{
- std::cerr << "I was not able to load " << this->getType()
- << " with size " << this->getSize() << std::endl;
+ cerr << "I was not able to load " << this->getType()
+ << " with size " << this -> getSize() << endl;
return false;
}
}
@@ -591,14 +593,14 @@ boundLoad( const String& fileName )
File file;
if( ! file. open( fileName, tnlReadMode ) )
{
- std::cerr << "I am not bale to open the file " << fileName << " for reading." << std::endl;
+ cerr << "I am not bale to open the file " << fileName << " for reading." << endl;
return false;
}
if( ! this->boundLoad( file ) )
return false;
if( ! file. close() )
{
- std::cerr << "An error occurred when I was closing the file " << fileName << "." << std::endl;
+ cerr << "An error occurred when I was closing the file " << fileName << "." << endl;
return false;
}
return true;
diff --git a/src/TNL/Containers/StaticVector.h b/src/TNL/Containers/StaticVector.h
index 2b0e7c7fdeb013e47d578a4ff33053a05aecde2e..a75f34d3b88f0390caec5ea0e10bd60c810bb8f0 100644
--- a/src/TNL/Containers/StaticVector.h
+++ b/src/TNL/Containers/StaticVector.h
@@ -23,6 +23,8 @@ class StaticVector : public Containers::StaticArray< Size, Real >
typedef StaticVector< Size, Real > ThisType;
enum { size = Size };
+ using Containers::StaticArray< Size, Real >::operator=;
+
__cuda_callable__
StaticVector();
@@ -36,6 +38,9 @@ class StaticVector : public Containers::StaticArray< Size, Real >
//! Copy constructor
__cuda_callable__
StaticVector( const StaticVector< Size, Real >& v );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
static String getType();
@@ -85,6 +90,9 @@ class StaticVector : public Containers::StaticArray< Size, Real >
__cuda_callable__
ThisType abs() const;
+
+ __cuda_callable__
+ Real lpNorm( const Real& p ) const;
};
template< typename Real >
@@ -94,7 +102,7 @@ class StaticVector< 1, Real > : public Containers::StaticArray< 1, Real >
typedef Real RealType;
typedef StaticVector< 1, Real > ThisType;
enum { size = 1 };
-
+
__cuda_callable__
StaticVector();
@@ -105,6 +113,9 @@ class StaticVector< 1, Real > : public Containers::StaticArray< 1, Real >
//! Copy constructor
__cuda_callable__
StaticVector( const StaticVector< 1, Real >& v );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
static String getType();
@@ -154,6 +165,9 @@ class StaticVector< 1, Real > : public Containers::StaticArray< 1, Real >
__cuda_callable__
ThisType abs() const;
+
+ __cuda_callable__
+ Real lpNorm( const Real& p ) const;
};
template< typename Real >
@@ -163,7 +177,7 @@ class StaticVector< 2, Real > : public Containers::StaticArray< 2, Real >
typedef Real RealType;
typedef StaticVector< 2, Real > ThisType;
enum { size = 2 };
-
+
__cuda_callable__
StaticVector();
@@ -180,6 +194,9 @@ class StaticVector< 2, Real > : public Containers::StaticArray< 2, Real >
//! Copy constructor
__cuda_callable__
StaticVector( const StaticVector< 2, Real >& v );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
static String getType();
@@ -229,6 +246,9 @@ class StaticVector< 2, Real > : public Containers::StaticArray< 2, Real >
__cuda_callable__
ThisType abs() const;
+
+ __cuda_callable__
+ Real lpNorm( const Real& p ) const;
};
template< typename Real >
@@ -238,7 +258,7 @@ class StaticVector< 3, Real > : public Containers::StaticArray< 3, Real >
typedef Real RealType;
typedef StaticVector< 3, Real > ThisType;
enum { size = 3 };
-
+
__cuda_callable__
StaticVector();
@@ -255,6 +275,9 @@ class StaticVector< 3, Real > : public Containers::StaticArray< 3, Real >
//! Copy constructor
__cuda_callable__
StaticVector( const StaticVector< 3, Real >& v );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
static String getType();
@@ -304,6 +327,9 @@ class StaticVector< 3, Real > : public Containers::StaticArray< 3, Real >
__cuda_callable__
ThisType abs() const;
+
+ __cuda_callable__
+ Real lpNorm( const Real& p ) const;
};
template< int Size, typename Real, typename Scalar >
diff --git a/src/TNL/Containers/StaticVector1D_impl.h b/src/TNL/Containers/StaticVector1D_impl.h
index a8e4151a352a71b455f7404977d3e347874a46d2..ab2391eddacb43a5667d7c82d9c120a940bb1e6e 100644
--- a/src/TNL/Containers/StaticVector1D_impl.h
+++ b/src/TNL/Containers/StaticVector1D_impl.h
@@ -33,6 +33,15 @@ StaticVector< 1, Real >::StaticVector( const StaticVector< 1, Real >& v )
{
}
+template< typename Real >
+bool
+StaticVector< 1, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->data[ 0 ] = parameters.getParameter< double >( prefix + "0" );
+ return true;
+}
+
template< typename Real >
String StaticVector< 1, Real >::getType()
{
@@ -148,6 +157,14 @@ StaticVector< 1, Real >::abs() const
return StaticVector< 1, Real >( TNL::abs( this->data[ 0 ] ) );
}
+template< typename Real >
+__cuda_callable__
+Real
+StaticVector< 1, Real >::lpNorm( const Real& p ) const
+{
+ return TNL::abs( this->data[ 0 ] );
+}
+
#ifdef UNDEF //TEMPLATE_EXPLICIT_INSTANTIATION
#ifndef HAVE_CUDA
diff --git a/src/TNL/Containers/StaticVector2D_impl.h b/src/TNL/Containers/StaticVector2D_impl.h
index 60b0ffa452d7c3eb8028b4f504d11b39645de5ef..d5f63f283129ae11e50fcdd4352cff91eda26ea3 100644
--- a/src/TNL/Containers/StaticVector2D_impl.h
+++ b/src/TNL/Containers/StaticVector2D_impl.h
@@ -49,6 +49,16 @@ StaticVector< 2, Real >::StaticVector( const StaticVector< 2, Real >& v )
{
}
+template< typename Real >
+bool
+StaticVector< 2, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->data[ 0 ] = parameters.getParameter< double >( prefix + "0" );
+ this->data[ 1 ] = parameters.getParameter< double >( prefix + "1" );
+ return true;
+}
+
template< typename Real >
String StaticVector< 2, Real >::getType()
{
@@ -177,6 +187,19 @@ StaticVector< 2, Real >::abs() const
::abs( this->data[ 1 ] ) );
}
+template< typename Real >
+__cuda_callable__
+Real
+StaticVector< 2, Real >::lpNorm( const Real& p ) const
+{
+ if( p == 1.0 )
+ return TNL::abs( this->data[ 0 ] ) + TNL::abs( this->data[ 1 ] );
+ if( p == 2.0 )
+ return std::sqrt( this->data[ 0 ] * this->data[ 0 ] +
+ this->data[ 1 ] * this->data[ 1 ] );
+ return std::pow( std::pow( TNL::abs( this->data[ 0 ] ), p ) +
+ std::pow( TNL::abs( this->data[ 1 ] ), p ), 1.0 / p );
+}
#ifdef UNDEF //TEMPLATE_EXPLICIT_INSTANTIATION
@@ -195,3 +218,4 @@ extern template class StaticVector< 2, long double >;
} // namespace Containers
} // namespace TNL
+
diff --git a/src/TNL/Containers/StaticVector3D_impl.h b/src/TNL/Containers/StaticVector3D_impl.h
index 91de4c92266453c7b5660b64644c6e5a89f3e049..68f5802e2772884305e3a6e6e50db913e1b8319d 100644
--- a/src/TNL/Containers/StaticVector3D_impl.h
+++ b/src/TNL/Containers/StaticVector3D_impl.h
@@ -47,6 +47,17 @@ StaticVector< 3, Real >::StaticVector( const StaticVector< 3, Real >& v )
{
}
+template< typename Real >
+bool
+StaticVector< 3, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->data[ 0 ] = parameters.getParameter< double >( prefix + "0" );
+ this->data[ 1 ] = parameters.getParameter< double >( prefix + "1" );
+ this->data[ 2 ] = parameters.getParameter< double >( prefix + "2" );
+ return true;
+}
+
template< typename Real >
String StaticVector< 3, Real >::getType()
{
@@ -187,6 +198,25 @@ StaticVector< 3, Real >::abs() const
::abs( this->data[ 2 ] ) );
}
+template< typename Real >
+__cuda_callable__
+Real
+StaticVector< 3, Real >::lpNorm( const Real& p ) const
+{
+ if( p == 1.0 )
+ return TNL::abs( this->data[ 0 ] ) +
+ TNL::abs( this->data[ 1 ] ) +
+ TNL::abs( this->data[ 2 ] );
+ if( p == 2.0 )
+ return std::sqrt( this->data[ 0 ] * this->data[ 0 ] +
+ this->data[ 1 ] * this->data[ 1 ] +
+ this->data[ 2 ] * this->data[ 2 ] );
+ return std::pow( std::pow( TNL::abs( this->data[ 0 ] ), p ) +
+ std::pow( TNL::abs( this->data[ 1 ] ), p ) +
+ std::pow( TNL::abs( this->data[ 2 ] ), p ), 1.0 / p );
+}
+
+
#ifdef UNDEF //TEMPLATE_EXPLICIT_INSTANTIATION
diff --git a/src/TNL/Containers/StaticVector_impl.h b/src/TNL/Containers/StaticVector_impl.h
index 32b8f4b95ece7bdc9c6e0a435c072e17e58e8979..8c8d37b9c0c0e5b3f58a5f182f75f2837eddf156 100644
--- a/src/TNL/Containers/StaticVector_impl.h
+++ b/src/TNL/Containers/StaticVector_impl.h
@@ -10,6 +10,8 @@
#pragma once
+#include <TNL/Config/ParameterContainer.h>
+
namespace TNL {
namespace Containers {
@@ -40,6 +42,17 @@ StaticVector< Size, Real >::StaticVector( const StaticVector< Size, Real >& v )
{
}
+template< int Size, typename Real >
+bool
+StaticVector< Size, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ for( int i = 0; i < Size; i++ )
+ if( ! parameters.template getParameter< double >( prefix + String( i ), this->data[ i ] ) )
+ return false;
+ return true;
+}
+
template< int Size, typename Real >
String StaticVector< Size, Real >::getType()
{
@@ -180,6 +193,30 @@ StaticVector< Size, Real >::abs() const
return v;
}
+template< int Size, typename Real >
+__cuda_callable__
+Real
+StaticVector< Size, Real >::lpNorm( const Real& p ) const
+{
+ if( p == 1.0 )
+ {
+ Real aux = TNL::abs( this->data[ 0 ] );
+ for( int i = 1; i < Size; i++ )
+ aux += TNL::abs( this->data[ i ] );
+ return aux;
+ }
+ if( p == 2.0 )
+ {
+ Real aux = this->data[ 0 ] * this->data[ 0 ];
+ for( int i = 1; i < Size; i++ )
+ aux += this->data[ i ] * this->data[ i ];
+ return std::sqrt( aux );
+ }
+ Real aux = std::pow( TNL::abs( this->data[ 0 ] ), p );
+ for( int i = 1; i < Size; i++ )
+ aux += std::pow( TNL::abs( this->data[ i ] ), p );
+ return std::pow( aux, 1.0 / p );
+}
template< int Size, typename Real, typename Scalar >
__cuda_callable__
diff --git a/src/TNL/Debugging/StackBacktrace.h b/src/TNL/Debugging/StackBacktrace.h
index 17e576fd7219e6da8b4b7e3bfc5beab4510d42cc..ee419922637899c7187e25d187abc6a427ebff68 100644
--- a/src/TNL/Debugging/StackBacktrace.h
+++ b/src/TNL/Debugging/StackBacktrace.h
@@ -18,6 +18,7 @@
namespace TNL {
namespace Debugging {
+#ifndef NDEBUG
/*
* Print a demangled stack backtrace of the caller function to FILE* out.
*
@@ -99,6 +100,13 @@ printStackBacktrace( FILE *out = stderr, unsigned int max_frames = 63 )
free(funcname);
free(symbollist);
}
+#endif
} // namespace Debugging
} // namespace TNL
+
+#ifdef NDEBUG
+#define PrintStackBacktrace
+#else
+#define PrintStackBacktrace TNL::Debugging::printStackBacktrace();
+#endif
diff --git a/src/TNL/Experimental/CMakeLists.txt b/src/TNL/Experimental/CMakeLists.txt
index acb1a43c37120ab468bea9a3e1126c319d2dca92..47fc9efa44f92699b64e247af1778ad83e6cebfa 100755
--- a/src/TNL/Experimental/CMakeLists.txt
+++ b/src/TNL/Experimental/CMakeLists.txt
@@ -1,2 +1,3 @@
ADD_SUBDIRECTORY( Arithmetics )
ADD_SUBDIRECTORY( Multimaps )
+ADD_SUBDIRECTORY( Hamilton-Jacobi )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..538e5d99ccaacd6892be2a378da2f19349ec3fa4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/CMakeLists.txt
@@ -0,0 +1,3 @@
+ADD_SUBDIRECTORY( Operators )
+#ADD_SUBDIRECTORY( Solvers )
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Operators/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..8ee63a0ac16a88e13ecb964013435160464e38b1
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/CMakeLists.txt
@@ -0,0 +1 @@
+ADD_SUBDIRECTORY( Hamilton-Jacobi )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..bce61025f2bae120464881216c62091e40604836
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/CMakeLists.txt
@@ -0,0 +1,2 @@
+ADD_SUBDIRECTORY( Godunov )
+ADD_SUBDIRECTORY( Godunov-Eikonal )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/CMakeLists.txt
new file mode 100644
index 0000000000000000000000000000000000000000..202550179b61cbf4a47bca43f96f751d0e2510eb
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/CMakeLists.txt
@@ -0,0 +1,18 @@
+set( headers godunov.h
+ godunov1D_impl.h
+ godunov2D_impl.h
+ godunov3D_impl.h)
+
+SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/implementation/operators/godunov )
+
+if( BUILD_CUDA)
+ set( tnl_implementation_operators_godunov_CUDA__SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+endif()
+
+set( tnl_implementation_operators_godunov_SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+
+INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/implementation/operators/godunov )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/godunovEikonal.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/godunovEikonal.h
new file mode 100644
index 0000000000000000000000000000000000000000..a90c7258438a52c0afebf2d0e9838e1491f16a13
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Eikonal/godunovEikonal.h
@@ -0,0 +1,264 @@
+/***************************************************************************
+ godunov.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <mesh/tnlGrid.h>
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1 >() ];
+
+ if( f > 0.0 )
+ {
+ RealType xf = negativePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = positivePart( ( u_c - u_w ) * hx_inv );
+
+ if( xb + xf > 0.0 )
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ return sqrt( xf * xf + xb * xb );
+ }
+ else if( f < 0.0 )
+ {
+ RealType xf = positivePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = negativePart( ( u_c - u_w ) * hx_inv );
+
+ if( xb + xf > 0.0 )
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ return sqrt( xf * xf + xb * xb );
+
+ }
+ else
+ {
+ return 0.0;
+ }
+ }
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hy_inv = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1, 0 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1, 0 >() ];
+ const RealType& u_n = u[ neighbourEntities.template getEntityIndex< 0, 1 >() ];
+ const RealType& u_s = u[ neighbourEntities.template getEntityIndex< 0, -1 >() ];
+
+ if( f > 0.0 )
+ {
+ RealType xf = negativePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = positivePart( ( u_c - u_w ) * hx_inv );
+ RealType yf = negativePart( ( u_n - u_c ) * hy_inv );
+ RealType yb = positivePart( ( u_c - u_s ) * hy_inv );
+
+ if( xb + xf > 0.0 )
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if( yb + yf > 0.0 )
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb );
+ }
+ else if( f < 0.0 )
+ {
+ RealType xf = positivePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = negativePart( ( u_c - u_w ) * hx_inv );
+ RealType yf = positivePart( ( u_n - u_c ) * hy_inv );
+ RealType yb = negativePart( ( u_c - u_s ) * hy_inv );
+
+ if( xb + xf > 0.0 )
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if( yb + yf > 0.0 )
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb );
+
+ }
+ else
+ {
+ return 0.0;
+ }
+ }
+};
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hy_inv = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hz_inv = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1, 0, 0 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1, 0, 0 >() ];
+ const RealType& u_n = u[ neighbourEntities.template getEntityIndex< 0, 1, 0 >() ];
+ const RealType& u_s = u[ neighbourEntities.template getEntityIndex< 0, -1, 0 >() ];
+ const RealType& u_t = u[ neighbourEntities.template getEntityIndex< 0, 0, 1 >() ];
+ const RealType& u_b = u[ neighbourEntities.template getEntityIndex< 0, 0, -1 >() ];
+
+ if( f > 0.0 )
+ {
+ RealType xf = negativePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = positivePart( ( u_c - u_w ) * hx_inv );
+ RealType yf = negativePart( ( u_n - u_c ) * hy_inv );
+ RealType yb = positivePart( ( u_c - u_s ) * hy_inv );
+ RealType zf = negativePart( ( u_t - u_c ) * hz_inv );
+ RealType zb = positivePart( ( u_c - u_b ) * hz_inv );
+
+ if( xb + xf > 0.0 )
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if( yb + yf > 0.0 )
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ if( zb + zf > 0.0 )
+ zf = 0.0;
+ else
+ zb = 0.0;
+
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb + zf * zf + zb * zb );
+ }
+ else if( f < 0.0 )
+ {
+ RealType xf = positivePart( ( u_e - u_c ) * hx_inv );
+ RealType xb = negativePart( ( u_c - u_w ) * hx_inv );
+ RealType yf = positivePart( ( u_n - u_c ) * hy_inv );
+ RealType yb = negativePart( ( u_c - u_s ) * hy_inv );
+ RealType zf = positivePart( ( u_t - u_c ) * hz_inv );
+ RealType zb = negativePart( ( u_c - u_b ) * hz_inv );
+
+ if( xb + xf > 0.0 )
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if( yb + yf > 0.0 )
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ if( zb + zf > 0.0 )
+ zb = 0.0;
+ else
+ zf = 0.0;
+
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb + zf * zf + zb * zb );
+ }
+ else
+ {
+ return 0.0;
+ }
+ }
+};
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..8b74f755ed0a74e7001f1c155c60f61c650679d2
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/CMakeLists.txt
@@ -0,0 +1,24 @@
+set( headers godunovEikonal.h
+ parallelGodunovEikonal.h
+ parallelGodunovMap.h
+ godunovEikonal1D_impl.h
+ godunovEikonal2D_impl.h
+ godunovEikonal3D_impl.h
+ parallelGodunovEikonal1D_impl.h
+ parallelGodunovEikonal2D_impl.h
+ parallelGodunovEikonal3D_impl.h
+ parallelGodunovMap2D_impl.h)
+
+SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/implementation/operators/godunov-eikonal )
+
+if( BUILD_CUDA)
+ set( tnl_implementation_operators_godunov-eikonal_CUDA__SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+endif()
+
+set( tnl_implementation_operators_godunov-eikonal_SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+
+INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/implementation/operators/godunov-eikonal )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal.h
new file mode 100644
index 0000000000000000000000000000000000000000..d4941ba38a992fd76047a7ead2f538b79af2f079
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal.h
@@ -0,0 +1,186 @@
+/***************************************************************************
+ godunov.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <matrices/tnlCSRMatrix.h>
+#include <solvers/preconditioners/tnlDummyPreconditioner.h>
+#include <solvers/tnlSolverMonitor.h>
+#include <core/tnlLogger.h>
+#include <core/vectors/tnlVector.h>
+#include <core/vectors/tnlSharedVector.h>
+#include <core/mfilename.h>
+#include <mesh/tnlGrid.h>
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme
+{
+};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+
+ static String getType();
+
+ RealType positivePart(const RealType arg) const;
+
+ RealType negativePart(const RealType arg) const;
+
+ RealType sign(const RealType x, const RealType eps) const;
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const;
+
+ bool init( const Config::ParameterContainer& parameters );
+
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType h;
+
+ RealType epsilon;
+
+
+};
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ static String getType();
+
+ RealType positivePart(const RealType arg) const;
+
+ RealType negativePart(const RealType arg) const;
+
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const;
+
+ bool init( const Config::ParameterContainer& parameters );
+
+ protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx;
+ RealType hy;
+
+ RealType epsilon;
+};
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ static String getType();
+
+ RealType positivePart(const RealType arg) const;
+
+ RealType negativePart(const RealType arg) const;
+
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const;
+
+
+ bool init( const Config::ParameterContainer& parameters );
+
+
+ protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx;
+ RealType hy;
+ RealType hz;
+
+ RealType epsilon;
+
+};
+
+#include <operators/hamilton-jacobi/godunov-eikonal/godunovEikonal1D_impl.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/godunovEikonal2D_impl.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/godunovEikonal3D_impl.h>
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal1D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal1D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..308cfe93e7292e3660d305e9bd91e7e7b72716c5
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal1D_impl.h
@@ -0,0 +1,165 @@
+/***************************************************************************
+ godunov1D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if(eps == 0.0)
+ return 0.0;
+
+ return sin( ( M_PI * x ) / (2 * eps) );
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool godunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ h = originalMesh.template getSpaceSteps().x();
+ cout << "h = " << h << endl;
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=h;
+
+ /*dofVector. setSize( this->mesh.getDofs() );*/
+
+ return true;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String godunovEikonalScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "godunovEikonalScheme< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovEikonalScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+/*
+ RealType nabla, xb, xf, signui;
+
+ signui = sign(u[cellIndex],epsilon);
+
+ if(signui > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return signui*(1.0 - nabla);
+ }
+ else if (signui < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return signui*(1.0 - nabla);
+ }
+ else
+*/ {
+ return 0.0;
+ }
+
+}
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..2b4dfdcef01ac7adec0be39f0b8f088bfbed758a
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal2D_impl.h
@@ -0,0 +1,186 @@
+/***************************************************************************
+ godunov2D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef GODUNOVEIKONAL2D_IMPL_H_
+#define GODUNOVEIKONAL2D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if ( eps == 0.0)
+ return 0.0;
+
+ return sin((M_PI*x)/(2.0*eps));
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool godunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+
+
+ hx = originalMesh.template getSpaceStepsProducts< 1, 0 >();
+ hy = originalMesh.template getSpaceStepsProducts< 0, 1 >();
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=sqrt( hx*hx + hy*hy );
+
+// dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String godunovEikonalScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlLinearDiffusion< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovEikonalScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+/*
+ RealType nabla, xb, xf, yb, yf, signui;
+
+ signui = sign(u[cellIndex],epsilon);
+
+ if(signui > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = negativePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = positivePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if(yb + yf > 0.0)
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+
+ return signui*(1.0 - nabla);
+ }
+ else if (signui < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = positivePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = negativePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if(yb + yf > 0.0)
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+
+ return signui*(1.0 - nabla);
+ }
+ else
+*/ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* GODUNOVEIKONAL2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..2d5bd0442a0ecd9c920df520cb41ccd8f83b7932
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/godunovEikonal3D_impl.h
@@ -0,0 +1,196 @@
+/***************************************************************************
+ godunov3D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef GODUNOVEIKONAL3D_IMPL_H_
+#define GODUNOVEIKONAL3D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if ( eps == 0.0)
+ return 0.0;
+
+ return sin((M_PI*x)/(2.0*eps));
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool godunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ hx = originalMesh.getSpaceSteps().x();
+ hy = originalMesh.getSpaceSteps().y();
+ hz = originalMesh.getSpaceSteps().z();
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=sqrt( hx*hx + hy*hy +hz*hz );
+
+ // dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String godunovEikonalScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlLinearDiffusion< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovEikonalScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+/*
+ RealType nabla, xb, xf, yb, yf, zb, zf, signui;
+
+ signui = sign(u[cellIndex],epsilon);
+
+ if(signui > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = negativePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = positivePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+ zf = negativePart((u[mesh.getCellZSuccessor( cellIndex )] - u[cellIndex])/hz);
+ zb = positivePart((u[cellIndex] - u[mesh.getCellZPredecessor( cellIndex )])/hz);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if(yb + yf > 0.0)
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ if(zb + zf > 0.0)
+ zf = 0.0;
+ else
+ zb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb );
+
+ return signui*(1.0 - nabla);
+ }
+ else if (signui < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = positivePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = negativePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+ zf = positivePart((u[mesh.getCellZSuccessor( cellIndex )] - u[cellIndex])/hz);
+ zb = negativePart((u[cellIndex] - u[mesh.getCellZPredecessor( cellIndex )])/hz);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if(yb + yf > 0.0)
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ if(zb + zf > 0.0)
+ zb = 0.0;
+ else
+ zf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb );
+
+ return signui*(1.0 - nabla);
+ }
+ else
+*/ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* GODUNOVEIKONAL3D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal.h
new file mode 100644
index 0000000000000000000000000000000000000000..e2f1859231a8096ae38e0dcaff664f490efa5cfc
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal.h
@@ -0,0 +1,269 @@
+/***************************************************************************
+ parallelGodunovEikonal.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVEIKONAL_H_
+#define PARALLELGODUNOVEIKONAL_H_
+
+#include <matrices/tnlCSRMatrix.h>
+#include <solvers/preconditioners/tnlDummyPreconditioner.h>
+#include <solvers/tnlSolverMonitor.h>
+#include <core/tnlLogger.h>
+#include <core/vectors/tnlVector.h>
+#include <core/vectors/tnlSharedVector.h>
+#include <core/mfilename.h>
+#include <mesh/tnlGrid.h>
+#include <core/tnlCuda.h>
+#include <mesh/grids/tnlGridEntity.h>
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class parallelGodunovEikonalScheme
+{
+};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const RealType eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType h;
+
+
+
+
+};
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities) const;
+
+ #ifdef HAVE_CUDA
+ __device__
+ #endif
+ Real getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const RealType* u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx, ihx;
+ RealType hy, ihy;
+
+
+
+
+};
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities ) const;
+
+#ifdef HAVE_CUDA
+ __device__
+#endif
+ Real getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const RealType* u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities ) const;
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx, ihx;
+ RealType hy, ihy;
+ RealType hz, ihz;
+
+};
+
+
+
+//#include <operators/godunov-eikonal/parallelGodunovEikonal1D_impl.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/parallelGodunovEikonal2D_impl.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/parallelGodunovEikonal3D_impl.h>
+
+
+#endif /* PARALLELGODUNOVEIKONAL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal1D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal1D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..b2e5c549fdf97c68ef3f4c4adf82e261dd85f576
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal1D_impl.h
@@ -0,0 +1,170 @@
+/***************************************************************************
+ parallelGodunovEikonal1D_impl.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVEIKONAL1D_IMPL_H_
+#define PARALLELGODUNOVEIKONAL1D_IMPL_H_
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if(eps == 0.0)
+ return 0.0;
+
+ return sin( ( M_PI * x ) / (2 * eps) );
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool parallelGodunovEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ h = originalMesh.template getSpaceStepsProducts< 1, 0 >();
+ cout << "h = " << h << endl;
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=h;
+
+ /*dofVector. setSize( this->mesh.getDofs() );*/
+
+ return true;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String parallelGodunovEikonalScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "parallelGodunovEikonalScheme< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time,
+ const IndexType boundaryCondition ) const
+{
+/*
+ RealType nabla, xb, xf, signui;
+
+ signui = sign(u[cellIndex],epsilon);
+
+ if(signui > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return signui*(1.0 - nabla);
+ }
+ else if (signui < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return signui*(1.0 - nabla);
+ }
+ else
+*/ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* PARALLELGODUNOVEIKONAL1D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..29738ab330e3b951d6b8a9bd6be89b5606c0f5b0
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal2D_impl.h
@@ -0,0 +1,484 @@
+/***************************************************************************
+ parallelGodunovEikonal2D_impl.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVEIKONAL2D_IMPL_H_
+#define PARALLELGODUNOVEIKONAL2D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return -arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ if (x < -eps)
+ return (-1.0);
+
+ if ( x == 0.0)
+ return 0.0;
+
+ return sin(/*(M_PI*x)/(2.0*eps) */(M_PI/2.0)*(x/eps));
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+bool parallelGodunovEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ //MeshType originalMesh;
+ if( ! originalMesh.load( meshFile ) )
+ {
+ //cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+
+
+ hx = originalMesh.template getSpaceStepsProducts< 1, 0 >();
+ ihx = 1.0/hx;
+ hy = originalMesh.template getSpaceStepsProducts< 0, 1 >();
+ ihy = 1.0/hy;
+
+ this->epsilon = parameters. getParameter< double >( "epsilon" );
+
+ this->epsilon *=sqrt( hx*hx + hy*hy );
+
+// dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+String parallelGodunovEikonalScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "parallelGodunovEikonalScheme< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities ) const
+{
+
+
+ if ( ((coordinates.x() == 0 && (boundaryCondition & 4)) or
+ (coordinates.x() == mesh.getDimensions().x() - 1 && (boundaryCondition & 2)) or
+ (coordinates.y() == 0 && (boundaryCondition & 8)) or
+ (coordinates.y() == mesh.getDimensions().y() - 1 && (boundaryCondition & 1)))
+ /*and
+ !( (coordinates.y() == 0 or coordinates.y() == mesh.getDimensions().y() - 1)
+ and
+ ( coordinates.x() == 0 or coordinates.x() == mesh.getDimensions().x() - 1)
+ )*/
+ )
+ {
+ return 0.0;
+ }
+
+
+ //-----------------------------------
+
+ RealType signui;
+ signui = sign(u[cellIndex],this->epsilon);
+
+
+#ifdef HAVE_CUDA
+ //printf("%d : %d ;;;; %d : %d , %f \n",threadIdx.x, mesh.getDimensions().x() , threadIdx.y,mesh.getDimensions().y(), epsilon );
+#endif
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType a,b,c;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+
+
+ //xb *= ihx;
+ //xf *= ihx;
+ // yb *= ihy;
+ //yf *= ihy;
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+ }
+
+
+ if(xb - xf > 0.0)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb - yf > 0.0)
+ b = yb;
+ else
+ b = yf;
+
+ c =(1.0 - sqrt(a*a+b*b)*ihx );
+
+ if(sign(c) > 0.0 )
+ return sign(u[cellIndex])*c;
+ else
+ return signui*c;
+
+ //--------------------------------------------------
+
+//
+//
+//
+// RealType acc = hx*hy*hx*hy;
+//
+// RealType nabla, xb, xf, yb, yf, signui;
+//
+// signui = sign(u[cellIndex],this->epsilon);
+//
+//
+// //if(fabs(u[cellIndex]) < acc) return 0.0;
+//
+// if(signui > 0.0)
+// {
+// /**/ /* if(boundaryCondition & 2)
+// xf = (u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx;
+// else *//*if(boundaryCondition & 4)
+// xf = 0.0;
+// else /**/if(coordinates.x() == mesh.getDimensions().x() - 1)
+// xf = negativePart((u[neighbourEntities.template getEntityIndex< -1, 0 >()] - u[cellIndex])*ihx);
+// else
+// xf = negativePart((u[neighbourEntities.template getEntityIndex< 1, 0 >()] - u[cellIndex])*ihx);
+//
+// /**/ /* if(boundaryCondition & 4)
+// xb = (u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])*ihx;
+// else *//*if(boundaryCondition & 2)
+// xb = 0.0;
+// else /**/if(coordinates.x() == 0)
+// xb = positivePart((u[cellIndex] - u[mesh.template getCellNextToCell<+1,0>( cellIndex )])*ihx);
+// else
+// xb = positivePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< -1, 0 >()])*ihx);
+//
+// /**/ /* if(boundaryCondition & 1)
+// yf = (u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])*ihy;
+// else *//*if(boundaryCondition & 8)
+// yf = 0.0;
+// else /**/if(coordinates.y() == mesh.getDimensions().y() - 1)
+// yf = negativePart((u[neighbourEntities.template getEntityIndex< 0, -1 >()] - u[cellIndex])*ihy);
+// else
+// yf = negativePart((u[neighbourEntities.template getEntityIndex< 0, 1 >()] - u[cellIndex])*ihy);
+//
+// /**/ /* if(boundaryCondition & 8)
+// yb = (u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])*ihy;
+// else *//*if(boundaryCondition & 1)
+// yb = 0.0;
+// else /**/if(coordinates.y() == 0)
+// yb = positivePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< 0, 1 >()])*ihy);
+// else
+// yb = positivePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< 0, -1 >()])*ihy);
+//
+// if(xb - xf > 0.0)
+// xf = 0.0;
+// else
+// xb = 0.0;
+//
+// if(yb - yf > 0.0)
+// yf = 0.0;
+// else
+// yb = 0.0;
+//
+// nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+// if(fabs(1.0-nabla) < acc)
+// return 0.0;
+// return signui*(1.0 - nabla);
+// }
+// else if (signui < 0.0)
+// {
+//
+// /**/ /* if(boundaryCondition & 2)
+// xf = (u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])*ihx;
+// else*//* if(boundaryCondition & 4)
+// xf = 0.0;
+// else /**/if(coordinates.x() == mesh.getDimensions().x() - 1)
+// xf = positivePart((u[neighbourEntities.template getEntityIndex< -1, 0 >()] - u[cellIndex])*ihx);
+// else
+// xf = positivePart((u[neighbourEntities.template getEntityIndex< 1, 0 >()] - u[cellIndex])*ihx);
+//
+// /**/ /* if(boundaryCondition & 4)
+// xb = (u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])*ihx;
+// else*//* if(boundaryCondition & 2)
+// xb = 0.0;
+// else /**/if(coordinates.x() == 0)
+// xb = negativePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< 1, 0 >()])*ihx);
+// else
+// xb = negativePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< -1, 0 >()])*ihx);
+//
+// /**/ /* if(boundaryCondition & 1)
+// yf = (u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])*ihy;
+// else *//*if(boundaryCondition & 8)
+// yf = 0.0;
+// else /**/if(coordinates.y() == mesh.getDimensions().y() - 1)
+// yf = positivePart((u[neighbourEntities.template getEntityIndex< 0, -1 >()] - u[cellIndex])*ihy);
+// else
+// yf = positivePart((u[neighbourEntities.template getEntityIndex< 0, 1 >()] - u[cellIndex])*ihy);
+//
+// /**/ /* if(boundaryCondition & 8)
+// yb = (u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])*ihy;
+// else*//* if(boundaryCondition & 1)
+// yb = 0.0;
+// else /**/if(coordinates.y() == 0)
+// yb = negativePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< 0, 1 >()])*ihy);
+// else
+// yb = negativePart((u[cellIndex] - u[neighbourEntities.template getEntityIndex< 0, -1 >()])*ihy);
+//
+//
+// if(xb - xf > 0.0)
+// xf = 0.0;
+// else
+// xb = 0.0;
+//
+// if(yb - yf > 0.0)
+// yf = 0.0;
+// else
+// yb = 0.0;
+//
+// nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+//
+// if(fabs(1.0-nabla) < acc)
+// return 0.0;
+// return signui*(1.0 - nabla);
+// }
+// else
+// {
+// return 0.0;
+// }
+
+}
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+
+#ifdef HAVE_CUDA
+__device__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >:: getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Real* u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities) const
+{
+
+ RealType signui;
+ if(boundaryCondition == 0)
+ signui = sign(u[cellIndex],/*(boundaryCondition != 0) * */this->epsilon);
+ else
+ signui = sign(u[cellIndex]);
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType a,b/*,c*/;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+ }
+
+
+ if(xb > xf)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb > yf)
+ b = yb;
+ else
+ b = yf;
+
+// c =(1.0 - sqrt(a*a+b*b)*ihx );
+
+// if(c > 0.0 )
+// return sign(u[cellIndex])*c;
+// else
+ return signui*(1.0 - sqrt(a*a+b*b)*ihx );
+}
+
+
+#endif /* PARALLELGODUNOVEIKONAL2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..3aa8aeed5357e7a4f0b0f176e9002c89a4295a84
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovEikonal3D_impl.h
@@ -0,0 +1,370 @@
+/***************************************************************************
+ parallelGodunovEikonal3D_impl.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVEIKONAL3D_IMPL_H_
+#define PARALLELGODUNOVEIKONAL3D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return -arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real parallelGodunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ if (x < -eps)
+ return (-1.0);
+
+ if ( x == 0.0)
+ return 0.0;
+
+ return sin((M_PI/2.0)*(x/eps));
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool parallelGodunovEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ hx = originalMesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ hy = originalMesh.template getSpaceStepsProducts< 0, 1, 0 >();
+ hz = originalMesh.template getSpaceStepsProducts< 0, 0, 1 >();
+ ihx = 1.0/hx;
+ ihy = 1.0/hy;
+ ihz = 1.0/hz;
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=sqrt( hx*hx + hy*hy + hz*hz );
+
+ // dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String parallelGodunovEikonalScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlLinearDiffusion< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities ) const
+{
+ if ( ((coordinates.x() == 0 && (boundaryCondition & 4)) or
+ (coordinates.x() == mesh.getDimensions().x() - 1 && (boundaryCondition & 2)) or
+ (coordinates.y() == 0 && (boundaryCondition & 8)) or
+ (coordinates.y() == mesh.getDimensions().y() - 1 && (boundaryCondition & 1)) or
+ (coordinates.z() == 0 && (boundaryCondition & 32)) or
+ (coordinates.z() == mesh.getDimensions().y() - 1 && (boundaryCondition & 16)))
+
+ )
+ {
+ return 0.0;
+ }
+
+
+ //-----------------------------------
+
+ RealType signui;
+ signui = sign(u[cellIndex], this->epsilon);
+
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType zb = u[cellIndex];
+ RealType zf = -u[cellIndex];
+ RealType a,b,c,d;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+
+
+ if(coordinates.z() == mesh.getDimensions().z() - 1)
+ zf += u[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+ else
+ zf += u[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+
+ if(coordinates.z() == 0)
+ zb -= u[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+ else
+ zb -= u[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ zf = negativePart(zf);
+
+ zb = positivePart(zb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+
+ zb = negativePart(zb);
+
+ zf = positivePart(zf);
+ }
+
+
+ if(xb - xf > 0.0)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb - yf > 0.0)
+ b = yb;
+ else
+ b = yf;
+
+ if(zb - zf > 0.0)
+ c = zb;
+ else
+ c = zf;
+
+ d = ( 1.0 - sqrt(a*a + b*b + c*c)*ihx );
+
+// d = 1.0 - sqrt(xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb)*ihx; /*upwind*/
+
+ if(sign(d) > 0.0 )
+ return sign(u[cellIndex])*d;
+ else
+ return signui*d;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+
+#ifdef HAVE_CUDA
+__device__
+#endif
+Real parallelGodunovEikonalScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >:: getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Real* u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities
+ ) const
+{
+ RealType signui;
+ signui = sign(u[cellIndex], this->epsilon);
+
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType zb = u[cellIndex];
+ RealType zf = -u[cellIndex];
+ RealType a,b,c,d;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+
+
+ if(coordinates.z() == mesh.getDimensions().z() - 1)
+ zf += u[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+ else
+ zf += u[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+
+ if(coordinates.z() == 0)
+ zb -= u[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+ else
+ zb -= u[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ zf = negativePart(zf);
+
+ zb = positivePart(zb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+
+ zb = negativePart(zb);
+
+ zf = positivePart(zf);
+ }
+
+
+ if(xb - xf > 0.0)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb - yf > 0.0)
+ b = yb;
+ else
+ b = yf;
+
+ if(zb - zf > 0.0)
+ c = zb;
+ else
+ c = zf;
+
+ d = ( 1.0 - sqrt(a*a + b*b + c*c)*ihx );
+
+ // d = 1.0 - sqrt(xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb)*ihx; /*upwind*/
+
+ if(sign(d) > 0.0 )
+ return sign(u[cellIndex])*d;
+ else
+ return signui*d;
+}
+
+
+#endif /* PARALLELGODUNOVEIKONAL3D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap.h
new file mode 100644
index 0000000000000000000000000000000000000000..db7a087224b933150fcda1ed47119d2e23448000
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap.h
@@ -0,0 +1,271 @@
+/***************************************************************************
+ parallelGodunovMap.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVMAP_H_
+#define PARALLELGODUNOVMAP_H_
+
+#include <matrices/tnlCSRMatrix.h>
+#include <solvers/preconditioners/tnlDummyPreconditioner.h>
+#include <solvers/tnlSolverMonitor.h>
+#include <core/tnlLogger.h>
+#include <core/vectors/tnlVector.h>
+#include <core/vectors/tnlSharedVector.h>
+#include <core/mfilename.h>
+#include <mesh/tnlGrid.h>
+#include <core/tnlCuda.h>
+#include <mesh/grids/tnlGridEntity.h>
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class parallelGodunovMapScheme
+{
+};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovMapScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const RealType eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType h;
+
+
+
+
+};
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovMapScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities,
+ const Vector& map) const;
+
+ #ifdef HAVE_CUDA
+ __device__
+ #endif
+ Real getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const RealType* u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities,
+ const RealType* map) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx, ihx;
+ RealType hy, ihy;
+
+
+
+
+};
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class parallelGodunovMapScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ static String getType();
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType positivePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType negativePart(const RealType arg) const;
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ RealType sign(const RealType x, const Real eps) const;
+
+ template< typename Vector >
+ #ifdef HAVE_CUDA
+ __device__ __host__
+ #endif
+ Real getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities ) const;
+
+#ifdef HAVE_CUDA
+ __device__
+#endif
+ Real getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const RealType* u,
+ const RealType& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities) const;
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+ bool init( const Config::ParameterContainer& parameters );
+
+ RealType epsilon;
+
+protected:
+
+ MeshType originalMesh;
+
+ DofVectorType dofVector;
+
+ RealType hx, ihx;
+ RealType hy, ihy;
+ RealType hz, ihz;
+
+};
+
+
+
+//#include <operators/godunov-eikonal/parallelGodunovMap1D_impl.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/parallelGodunovMap2D_impl.h>
+//#include <operators/godunov-eikonal/parallelGodunovMap3D_impl.h>
+
+
+#endif /* PARALLELGODUNOVMAP_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..c5f7d613497e7f61b449a2e16190c18988a12c3f
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov-Eikonal/parallelGodunovMap2D_impl.h
@@ -0,0 +1,391 @@
+/***************************************************************************
+ parallelGodunovMap2D_impl.h - description
+ -------------------
+ begin : Dec 1 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef PARALLELGODUNOVMAP2D_IMPL_H_
+#define PARALLELGODUNOVMAP2D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovMapScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovMapScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return -arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+Real parallelGodunovMapScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ if (x < -eps)
+ return (-1.0);
+
+ if ( x == 0.0)
+ return 0.0;
+
+ return sin(/*(M_PI*x)/(2.0*eps) */(M_PI/2.0)*(x/eps));
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+bool parallelGodunovMapScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ //MeshType originalMesh;
+ if( ! originalMesh.load( meshFile ) )
+ {
+ //cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+
+
+ hx = originalMesh.template getSpaceStepsProducts< 1, 0 >();
+ ihx = 1.0/hx;
+ hy = originalMesh.template getSpaceStepsProducts< 0, 1 >();
+ ihy = 1.0/hy;
+
+ this->epsilon = parameters. getParameter< double >( "epsilon" );
+
+ this->epsilon *=sqrt( hx*hx + hy*hy );
+
+// dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+String parallelGodunovMapScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "parallelGodunovMapScheme< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real parallelGodunovMapScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities,
+ const Vector& map) const
+{
+
+
+ if ( ((coordinates.x() == 0 && (boundaryCondition & 4)) or
+ (coordinates.x() == mesh.getDimensions().x() - 1 && (boundaryCondition & 2)) or
+ (coordinates.y() == 0 && (boundaryCondition & 8)) or
+ (coordinates.y() == mesh.getDimensions().y() - 1 && (boundaryCondition & 1)))
+ /*and
+ !( (coordinates.y() == 0 or coordinates.y() == mesh.getDimensions().y() - 1)
+ and
+ ( coordinates.x() == 0 or coordinates.x() == mesh.getDimensions().x() - 1)
+ )*/
+ )
+ {
+ return 0.0;
+ }
+
+
+ RealType signui;
+// if(boundaryCondition == 0)
+ signui = sign(u[cellIndex],/*(boundaryCondition != 0) * */this->epsilon);
+// else
+// signui = sign(u[cellIndex]);
+
+ RealType value;
+// if(map[cellIndex] == 0.0)
+// {
+//// value = INT_MAX;
+// u[cellIndex] = sign(u[cellIndex])*INT_MAX;
+// return 0.0;
+// }
+// else
+ value = 1.0/ map[cellIndex];
+
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType a,b,c;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+ }
+
+ /*if(boundaryCondition &1)
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ if(boundaryCondition & 2)
+ xb = 0.0;
+ else
+ xf = 0.0;*/
+
+ if(xb - xf > 0.0)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb - yf > 0.0)
+ b = yb;
+ else
+ b = yf;
+
+ c =(value - sqrt(a*a+b*b)*ihx );
+
+ if(c > 0.0 )
+ return sign(u[cellIndex])*c;
+ else
+
+ return signui*c;
+
+}
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+
+#ifdef HAVE_CUDA
+__device__
+#endif
+Real parallelGodunovMapScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >:: getValueDev( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Real* u,
+ const Real& time,
+ const IndexType boundaryCondition,
+ const tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities,
+ const Real* map) const
+{
+// int gid = (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x;
+
+ RealType signui;
+ if(boundaryCondition == 0)
+ signui = sign(u[cellIndex],/*(boundaryCondition != 0) * */this->epsilon);
+ else
+ signui = sign(u[cellIndex]);
+
+// RealType value;
+// if(map[cellIndex] == 0.0)
+// {
+//// value = INT_MAX;
+// u[cellIndex] = sign(u[cellIndex])*INT_MAX;
+// return 0.0;
+// }
+// else
+ RealType value = /*1.0/*/ map[cellIndex];
+
+
+ RealType xb = u[cellIndex];
+ RealType xf = -u[cellIndex];
+ RealType yb = u[cellIndex];
+ RealType yf = -u[cellIndex];
+ RealType a,b,c;
+
+
+ if(coordinates.x() == mesh.getDimensions().x() - 1)
+ xf += u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ xf += u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+
+ if(coordinates.x() == 0)
+ xb -= u[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else
+ xb -= u[neighbourEntities.template getEntityIndex< -1, 0 >()];
+
+ if(coordinates.y() == mesh.getDimensions().y() - 1)
+ yf += u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ yf += u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+
+ if(coordinates.y() == 0)
+ yb -= u[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else
+ yb -= u[neighbourEntities.template getEntityIndex< 0, -1 >()];
+
+
+ if(signui > 0.0)
+ {
+ xf = negativePart(xf);
+
+ xb = positivePart(xb);
+
+ yf = negativePart(yf);
+
+ yb = positivePart(yb);
+
+ }
+ else if(signui < 0.0)
+ {
+
+ xb = negativePart(xb);
+
+ xf = positivePart(xf);
+
+ yb = negativePart(yb);
+
+ yf = positivePart(yf);
+ }
+
+// if(boundaryCondition &1)
+// yb = 0.0;
+// else
+// yf = 0.0;
+//
+// if(boundaryCondition & 2)
+// xb = 0.0;
+// else
+// xf = 0.0;
+
+ if(xb > xf)
+ a = xb;
+ else
+ a = xf;
+
+ if(yb > yf)
+ b = yb;
+ else
+ b = yf;
+
+ c = (value - sqrt(a*a+b*b)*ihx );
+
+// if(c > 0.0 )
+// return sign(u[cellIndex])*c;
+// else
+ return signui*c;
+}
+
+
+#endif /* PARALLELGODUNOVMAP2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..74f7510d72329f60305bb366604e72b7201a3ea0
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/CMakeLists.txt
@@ -0,0 +1,17 @@
+set( headers godunov1D_impl.h
+ godunov2D_impl.h
+ godunov3D_impl.h)
+
+SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/implementation/operators/godunov )
+
+if( BUILD_CUDA)
+ set( tnl_implementation_operators_godunov_CUDA__SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+endif()
+
+set( tnl_implementation_operators_godunov_SOURCES
+ ${common_SOURCES}
+ PARENT_SCOPE )
+
+INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/implementation/operators/godunov )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov1D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov1D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..2f902ad7d8dc81273a7e0067f29c459a4bbd1dfc
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov1D_impl.h
@@ -0,0 +1,174 @@
+/***************************************************************************
+ godunov1D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef GODUNOV1D_IMPL_H_
+#define GODUNOV1D_IMPL_H_
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index, Function > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index, Function > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index, Function > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if(eps == 0.0)
+ return 0.0;
+
+ return sin( ( M_PI * x ) / (2 * eps) );
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+bool godunovScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index, Function > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ h = originalMesh.getSpaceSteps().x();
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=h;
+
+ f.setup( parameters );
+
+ /*dofVector. setSize( this->mesh.getDofs() );*/
+
+ return true;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+String godunovScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index, Function > :: getType()
+{
+ return String( "godunovScheme< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovScheme< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index, Function >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+
+ RealType nabla, xb, xf, fi;
+
+ fi = f.getValue(mesh.getCellCenter( coordinates ),time);
+
+ if(fi > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return -fi*( nabla);
+ }
+ else if (fi < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/h);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/h);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb );
+
+ return -fi*( nabla);
+ }
+ else
+ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* GODUNOV1D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..fdd7e671630ec4c2eb40230f80b779c1f5058f89
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov2D_impl.h
@@ -0,0 +1,192 @@
+/***************************************************************************
+ godunov2D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef GODUNOV2D_IMPL_H_
+#define GODUNOV2D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index, Function >:: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index, Function > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index, Function > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if ( eps == 0.0)
+ return 0.0;
+
+ return sin((M_PI*x)/(2.0*eps));
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+bool godunovScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index, Function > :: init( const Config::ParameterContainer& parameters )
+{
+
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ hx = originalMesh.getSpaceSteps().x();
+ hy = originalMesh.getSpaceSteps().y();
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=sqrt( hx*hx + hy*hy );
+
+ f.setup( parameters );
+
+// dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+String godunovScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index, Function > :: getType()
+{
+ return String( "tnlLinearDiffusion< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovScheme< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index, Function >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+
+
+ RealType nabla, xb, xf, yb, yf, fi;
+
+ fi = f.getValue(mesh.getCellCenter( coordinates ),time);
+
+ if(fi > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = negativePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = positivePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if(yb + yf > 0.0)
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+
+ return -1.0*fi*( nabla);
+ }
+ else if (fi < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = positivePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = negativePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if(yb + yf > 0.0)
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb );
+
+ return -1.0*fi*( nabla);
+ }
+ else
+ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* GODUNOV2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..d25f354bc0a4cc448166ad44d3400f16132682f9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/Godunov/godunov3D_impl.h
@@ -0,0 +1,204 @@
+/***************************************************************************
+ godunov3D_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef GODUNOV3D_IMPL_H_
+#define GODUNOV3D_IMPL_H_
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index, Function > :: positivePart(const Real arg) const
+{
+ if(arg > 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index, Function > :: negativePart(const Real arg) const
+{
+ if(arg < 0.0)
+ return arg;
+ return 0.0;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+Real godunovScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index, Function > :: sign(const Real x, const Real eps) const
+{
+ if(x > eps)
+ return 1.0;
+ else if (x < -eps)
+ return (-1.0);
+
+ if ( eps == 0.0)
+ return 0.0;
+
+ return sin((M_PI*x)/(2.0*eps));
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+bool godunovScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index, Function > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ if( ! this->originalMesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ hx = originalMesh.getSpaceSteps().x();
+ hy = originalMesh.getSpaceSteps().y();
+ hz = originalMesh.getSpaceSteps().z();
+
+ epsilon = parameters. getParameter< double >( "epsilon" );
+
+ if(epsilon != 0.0)
+ epsilon *=sqrt( hx*hx + hy*hy +hz*hz );
+
+ f.setup( parameters );
+
+ // dofVector. setSize( this->mesh.getDofs() );
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+String godunovScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index, Function > :: getType()
+{
+ return String( "tnlLinearDiffusion< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename Function >
+template< typename Vector >
+#ifdef HAVE_CUDA
+__device__ __host__
+#endif
+Real godunovScheme< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index, Function >:: getValue( const MeshType& mesh,
+ const IndexType cellIndex,
+ const CoordinatesType& coordinates,
+ const Vector& u,
+ const Real& time ) const
+{
+
+ RealType nabla, xb, xf, yb, yf, zb, zf, fi;
+
+ fi = f.getValue(mesh.getCellCenter( coordinates ),time);
+
+ if(fi > 0.0)
+ {
+ xf = negativePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = positivePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = negativePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = positivePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+ zf = negativePart((u[mesh.getCellZSuccessor( cellIndex )] - u[cellIndex])/hz);
+ zb = positivePart((u[cellIndex] - u[mesh.getCellZPredecessor( cellIndex )])/hz);
+
+ if(xb + xf > 0.0)
+ xf = 0.0;
+ else
+ xb = 0.0;
+
+ if(yb + yf > 0.0)
+ yf = 0.0;
+ else
+ yb = 0.0;
+
+ if(zb + zf > 0.0)
+ zf = 0.0;
+ else
+ zb = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb );
+
+ return -fi*( nabla);
+ }
+ else if (fi < 0.0)
+ {
+ xf = positivePart((u[mesh.getCellXSuccessor( cellIndex )] - u[cellIndex])/hx);
+ xb = negativePart((u[cellIndex] - u[mesh.getCellXPredecessor( cellIndex )])/hx);
+ yf = positivePart((u[mesh.getCellYSuccessor( cellIndex )] - u[cellIndex])/hy);
+ yb = negativePart((u[cellIndex] - u[mesh.getCellYPredecessor( cellIndex )])/hy);
+ zf = positivePart((u[mesh.getCellZSuccessor( cellIndex )] - u[cellIndex])/hz);
+ zb = negativePart((u[cellIndex] - u[mesh.getCellZPredecessor( cellIndex )])/hz);
+
+ if(xb + xf > 0.0)
+ xb = 0.0;
+ else
+ xf = 0.0;
+
+ if(yb + yf > 0.0)
+ yb = 0.0;
+ else
+ yf = 0.0;
+
+ if(zb + zf > 0.0)
+ zb = 0.0;
+ else
+ zf = 0.0;
+
+ nabla = sqrt (xf*xf + xb*xb + yf*yf + yb*yb + zf*zf + zb*zb );
+
+ return -fi*( nabla);
+ }
+ else
+ {
+ return 0.0;
+ }
+
+}
+
+
+#endif /* GODUNOV3D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/tnlEikonalOperator.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/tnlEikonalOperator.h
new file mode 100644
index 0000000000000000000000000000000000000000..3beb7e997e0f7aa6759df7c571b3a398445630b9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/tnlEikonalOperator.h
@@ -0,0 +1,66 @@
+/*
+ * File: tnlEikonalOperator.h
+ * Author: oberhuber
+ *
+ * Created on July 11, 2016, 6:33 PM
+ */
+
+#pragma once
+
+#include <functions/tnlConstantFunction.h>
+#include <functions/tnlFunctionAdapter.h>
+
+template< typename GradientNormOperator,
+ typename Anisotropy =
+ tnlConstantFunction< GradientNormOperator::MeshType::getMeshDimensions(),
+ typename GradientNormOperator::MeshType::RealType > >
+class tnlEikonalOperator
+ : public tnlOperator< typename GradientNormOperator::MeshType, MeshDomain >
+{
+ public:
+
+ typedef typename GradientNormOperator::MeshType MeshType;
+ typedef typename MeshType::RealType RealType;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::IndexType IndexType;
+ typedef GradientNormOperator GradientNormOperatorType;
+ typedef Anisotropy AnisotropyType;
+ //typedef tnlExactLinearDiffusion< 1 > ExactOperatorType;
+
+ static const int Dimensions = MeshType::meshDimensions;
+
+ static constexpr int getMeshDimensions() { return Dimensions; }
+
+ static String getType();
+
+ AnisotropyType& getAnisotropy() { return this->anisotropy; };
+
+ const AnisotropyType& getAnisotropy() const { return this->anisotropy; };
+
+ const RealType getSmoothing() const { return this->smoothing; };
+
+ void setSmoothing( const RealType& smoothing ) { this->smoothing = smoothing; };
+
+ template< typename PreimageFunction,
+ typename MeshEntity >
+ __cuda_callable__
+ RealType operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ const RealType signU = sign( u( entity ), smoothing * entity.getMesh().getSmallestSpaceStep() );
+ const RealType f = tnlFunctionAdapter< MeshType, AnisotropyType >::getValue( anisotropy, entity, 0.0 );
+ return signU * ( f - gradientNorm( u, entity, signU ) );
+ };
+
+ protected:
+
+ GradientNormOperatorType gradientNorm;
+
+ AnisotropyType anisotropy;
+
+ RealType smoothing;
+
+};
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/upwindEikonal.h b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/upwindEikonal.h
new file mode 100644
index 0000000000000000000000000000000000000000..e8f793260999eba3f7e2082a5237ee3daba6e86a
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Operators/Hamilton-Jacobi/upwindEikonal.h
@@ -0,0 +1,202 @@
+/***************************************************************************
+ upwindEikonal.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <mesh/tnlGrid.h>
+#include <functions/tnlFunctions.h>
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class upwindEikonalScheme
+{
+};
+
+/****
+ * 1D scheme
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class upwindEikonalScheme< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1 >() ];
+
+ if( f > 0.0 )
+ {
+ const RealType& xf = negativePart( ( u_e - u_c ) * hx_inv );
+ const RealType& xb = positivePart( ( u_c - u_w ) * hx_inv );
+ return sqrt( xf * xf + xb * xb );
+ }
+ else if( f < 0.0 )
+ {
+ const RealType& xf = negativePart( ( u_c - u_w ) * hx_inv );
+ const RealType& xb = positivePart( ( u_e - u_c ) * hx_inv );
+ return sqrt( xf * xf + xb * xb );
+ }
+ else
+ {
+ return 0.0;
+ }
+ };
+};
+
+/****
+ * 2D scheme
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class upwindEikonalScheme< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hy_inv = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1, 0 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1, 0 >() ];
+ const RealType& u_n = u[ neighbourEntities.template getEntityIndex< 0, 1 >() ];
+ const RealType& u_s = u[ neighbourEntities.template getEntityIndex< 0, -1 >() ];
+ if( f > 0.0 )
+ {
+ const RealType xf = negativePart( ( u_e - u_c ) * hx_inv );
+ const RealType xb = positivePart( ( u_c - u_w ) * hx_inv );
+ const RealType yf = negativePart( ( u_n - u_c ) * hy_inv );
+ const RealType yb = positivePart( ( u_c - u_s ) * hy_inv );
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb );
+ }
+ else if( f < 0.0 )
+ {
+ const RealType xf = positivePart( ( u_e - u_c ) * hx_inv );
+ const RealType xb = negativePart( ( u_c - u_w ) * hx_inv );
+ const RealType yf = positivePart( ( u_n - u_c ) * hy_inv );
+ const RealType yb = negativePart( ( u_c - u_s ) * hy_inv );
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb );
+ }
+ else
+ return 0.0;
+ }
+};
+
+/****
+ * 3D scheme
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class upwindEikonalScheme< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+ public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ static String getType();
+
+ template< typename PreimageFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const PreimageFunction& u,
+ const MeshEntity& entity,
+ const RealType& f ) const
+ {
+ const RealType& hx_inv = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hy_inv = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hz_inv = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+ const RealType& u_c = u[ entity.getIndex() ];
+ const RealType& u_e = u[ neighbourEntities.template getEntityIndex< 1, 0, 0 >() ];
+ const RealType& u_w = u[ neighbourEntities.template getEntityIndex< -1, 0, 0 >() ];
+ const RealType& u_n = u[ neighbourEntities.template getEntityIndex< 0, 1, 0 >() ];
+ const RealType& u_s = u[ neighbourEntities.template getEntityIndex< 0, -1, 0 >() ];
+ const RealType& u_t = u[ neighbourEntities.template getEntityIndex< 0, 0, 1 >() ];
+ const RealType& u_b = u[ neighbourEntities.template getEntityIndex< 0, 0, -1 >() ];
+
+ if( f > 0.0 )
+ {
+ const RealType xf = negativePart( ( u_e - u_c ) * hx_inv );
+ const RealType xb = positivePart( ( u_c - u_w ) * hx_inv );
+ const RealType yf = negativePart( ( u_n - u_c ) * hy_inv );
+ const RealType yb = positivePart( ( u_c - u_s ) * hy_inv );
+ const RealType zf = negativePart( ( u_t - u_c ) * hz_inv );
+ const RealType zb = positivePart( ( u_c - u_b ) * hz_inv );
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb + zf * zf + zb * zb );
+ }
+ else if( f < 0.0 )
+ {
+ const RealType xf = positivePart( ( u_e - u_c ) * hx_inv );
+ const RealType xb = negativePart( ( u_c - u_w ) * hx_inv );
+ const RealType yf = positivePart( ( u_n - u_c ) * hy_inv );
+ const RealType yb = negativePart( ( u_c - u_s ) * hy_inv );
+ const RealType zf = positivePart( ( u_t - u_c ) * hz_inv );
+ const RealType zb = negativePart( ( u_c - u_b ) * hz_inv );
+ return sqrt( xf * xf + xb * xb + yf * yf + yb * yb + zf * zf + zb * zb );
+ }
+ else
+ return 0.0;
+ };
+};
\ No newline at end of file
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..b4ec5170b50f7b9bfdf6b81f261f396c0c59d624
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/CMakeLists.txt
@@ -0,0 +1,3 @@
+ADD_SUBDIRECTORY( hamilton-jacobi )
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..73cc05e2646fd6f5ad77227a73722d981580f7c6
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/CMakeLists.txt
@@ -0,0 +1,22 @@
+set( tnl_fast_sweeping_map_SOURCES
+# MainBuildConfig.h
+# tnlFastSweepingMap2D_impl.h
+# tnlFastSweepingMap.h
+# fastSweepingMapConfig.h
+ main.cpp)
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(fast-sweeping-map${debugExt} main.cu)
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(fast-sweeping-map${debugExt} main.cpp)
+ENDIF( BUILD_CUDA )
+target_link_libraries (fast-sweeping-map${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS fast-sweeping-map${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+#INSTALL( FILES ${tnl_fast_sweeping_map_SOURCES}
+# DESTINATION share/tnl-${tnlVersion}/examples/fast-sweeping-map )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/MainBuildConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/fastSweepingMapConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/fastSweepingMapConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..9251deca876e821dace59682ca1a151555095c69
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/fastSweepingMapConfig.h
@@ -0,0 +1,39 @@
+/***************************************************************************
+ fastSweepingConfig.h - description
+ -------------------
+ begin : Oct 15, 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef FASTSWEEPINGCONFIG_H_
+#define FASTSWEEPINGCONFIG_H_
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class fastSweepingMapConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Parallel Eikonal solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "fast-sweeping" );
+ config.addRequiredEntry < String > ( "initial-condition", "Initial condition for solver");
+ config.addRequiredEntry < int > ( "dim", "Dimension of problem.");
+ config.addEntry < String > ( "mesh", "Name of mesh.", "mesh.tnl" );
+ config.addEntry < String > ( "exact-input", "Are the function values near the curve equal to the SDF? (yes/no)", "no" );
+ config.addRequiredEntry < String > ( "map", "Gradient map for solver");
+ }
+};
+
+#endif /* FASTSWEEPINGCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cpp
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cu b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cu
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.cu
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.h
new file mode 100644
index 0000000000000000000000000000000000000000..d17b47ac5125bebf9ac9619a9496b5dc4b46b394
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/main.h
@@ -0,0 +1,88 @@
+/***************************************************************************
+ main.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+
+#include "MainBuildConfig.h"
+ //for HOST versions:
+#include "tnlFastSweepingMap.h"
+ //for DEVICE versions:
+//#include "tnlFastSweepingMap_CUDA.h"
+#include "fastSweepingMapConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+
+#include <mesh/tnlGrid.h>
+#include <core/tnlDevice.h>
+#include <time.h>
+#include <ctime>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ time_t start;
+ time_t stop;
+ time(&start);
+ std::clock_t start2= std::clock();
+ Config::ParameterContainer parameters;
+ tnlConfigDescription configDescription;
+ fastSweepingMapConfig< BuildConfig >::configSetup( configDescription );
+
+ if( ! parseCommandLine( argc, argv, configDescription, parameters ) )
+ return false;
+
+ const int& dim = parameters.getParameter< int >( "dim" );
+
+ if(dim == 2)
+ {
+ tnlFastSweepingMap<tnlGrid<2,double,tnlHost, int>, double, int> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ checkCudaDevice;
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver..." << endl;
+ solver.run();
+ }
+// else if(dim == 3)
+// {
+// tnlFastSweepingMap<tnlGrid<3,double,tnlHost, int>, double, int> solver;
+// if(!solver.init(parameters))
+// {
+// cerr << "Solver failed to initialize." << endl;
+// return EXIT_FAILURE;
+// }
+// checkCudaDevice;
+// cout << "-------------------------------------------------------------" << endl;
+// cout << "Starting solver..." << endl;
+// solver.run();
+// }
+ else
+ {
+ cerr << "Unsupported number of dimensions: " << dim << "!" << endl;
+ return EXIT_FAILURE;
+ }
+
+
+ time(&stop);
+ cout << "Solver stopped..." << endl;
+ cout << endl;
+ cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) << endl;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap.h
new file mode 100644
index 0000000000000000000000000000000000000000..f6f2150a4df3ae6f0fe85efeeb6ea9abac0007d3
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap.h
@@ -0,0 +1,188 @@
+/***************************************************************************
+ tnlFastSweepingMap.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING_H_
+#define TNLFASTSWEEPING_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlMeshFunction.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+#ifdef HAVE_OPENMP
+#include <omp.h>
+#endif
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ tnlFastSweepingMap();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+
+ void setupSquare1000(Index i, Index j);
+ void setupSquare1100(Index i, Index j);
+ void setupSquare1010(Index i, Index j);
+ void setupSquare1001(Index i, Index j);
+ void setupSquare1110(Index i, Index j);
+ void setupSquare1101(Index i, Index j);
+ void setupSquare1011(Index i, Index j);
+ void setupSquare1111(Index i, Index j);
+ void setupSquare0000(Index i, Index j);
+ void setupSquare0100(Index i, Index j);
+ void setupSquare0010(Index i, Index j);
+ void setupSquare0001(Index i, Index j);
+ void setupSquare0110(Index i, Index j);
+ void setupSquare0101(Index i, Index j);
+ void setupSquare0011(Index i, Index j);
+ void setupSquare0111(Index i, Index j);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+ int something_changed;
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data,map;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage > Entity;
+
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlFastSweepingMap();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j, const Index k);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage > Entity;
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+ //for single core version use this implementation:
+#include "tnlFastSweepingMap2D_impl.h"
+ //for parallel version use this one instead:
+// #include "tnlFastSweepingMap2D_openMP_impl.h"
+
+// #include "tnlFastSweepingMap3D_impl.h"
+
+#endif /* TNLFASTSWEEPING_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..5e9d11b5264e301d8eb9a30b625861bf6a79ad2b
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h
@@ -0,0 +1,1051 @@
+/***************************************************************************
+ tnlFastSweepingMap2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweepingMap.h"
+
+#define MAP_SOLVER_MAX_VALUE 3
+
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweepingMap< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlFastSweepingMap()
+:dofVector(Mesh)
+{
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ const String& mapFile = parameters.getParameter <String>("map");
+ if(! this->map.load( mapFile ))
+ cout << "Failed to load map file : " << mapFile << endl;
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ //Entity.refresh();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(map_cuda), this->map.getSize()*sizeof(double));
+ cudaMemcpy(map_cuda, this->map.getData(), this->map.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(changed), sizeof(int));
+ //counter == 0 --> setting changed to 0
+ cudaMemcpy(changed, &counter, sizeof(int), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(1, 1024);
+ dim3 numBlocks(4,1);
+
+ int run = 1;
+ int zero = 0;
+ int cntr = 0;
+
+ while(run != 0)
+ {
+ cudaMemcpy(this->changed, &zero, sizeof(int), cudaMemcpyHostToDevice);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,0,0, this->changed);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(&run, this->changed,sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ cntr++;
+ cout << "Finished set of sweeps #" << cntr << " " << run << endl;
+ }
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ //data.setLike(dofVector.getData());
+ //cudaMemcpy(data.getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaMemcpy(dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ //data.save("u-00001.tnl");
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, Index* something_changed)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+
+ if(map_cuda[Entity.getIndex()] != 0.0)
+ {
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real im = abs(1.0/map_cuda[Entity.getIndex()]);
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(abs(a-b) >= im*h)
+ tmp = fabsMin(a,b) + sign(value)*im*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * im * h * im * h - (a - b) * (a - b) ) );
+
+ // cudaDofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+ atomicFabsMin(&(cudaDofVector2[Entity.getIndex()]), tmp);
+
+ if(abs(value)-abs(tmp) > 0.0)
+ atomicMax(something_changed,1);
+ }
+ else
+ {
+ atomicFabsMin(&(cudaDofVector2[Entity.getIndex()]), MAP_SOLVER_MAX_VALUE);
+ }
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int i = threadIdx.x + blockDim.x*blockIdx.x;
+ int j = blockDim.y*blockIdx.y + threadIdx.y;
+
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ int gid = Entity.getIndex();
+
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector[gid]);
+
+ if(abs(cudaDofVector[gid]) < 1.01*h)
+ {
+ cudaDofVector2[gid] = cudaDofVector[gid];
+ if(map_cuda[gid] != 0.0)
+ cudaDofVector2[gid] /=map_cuda[gid];
+ }
+
+
+
+
+
+// if(i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y() )
+// {
+// if(cudaDofVector[Entity.getIndex()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1111(i,j);
+// else
+// setupSquare1110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1101(i,j);
+// else
+// setupSquare1100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1011(i,j);
+// else
+// setupSquare1010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1001(i,j);
+// else
+// setupSquare1000(i,j);
+// }
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0111(i,j);
+// else
+// setupSquare0110(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0101(i,j);
+// else
+// setupSquare0100(i,j);
+// }
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0011(i,j);
+// else
+// setupSquare0010(i,j);
+// }
+// else
+// {
+// if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0001(i,j);
+// else
+// setupSquare0000(i,j);
+// }
+// }
+// }
+//
+// }
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ //Real fy = abs(y);
+
+ //Real tmpMin = Min(fx,abs(y));
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweepingMap< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i, int* changed)
+{
+
+ __shared__ int something_changed;
+ if(threadIdx.x+threadIdx.y == 0)
+ something_changed = 0;
+
+ int gx = 0;
+ int gy = threadIdx.y;
+ //if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ // return;
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ //int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ //int id1 = gx+gy;
+ //int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+ __syncthreads();
+ if(blockIdx.x==0)
+ {
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,&something_changed);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,&something_changed);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,&something_changed);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,&something_changed);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+
+
+ if(threadIdx.x+threadIdx.y == 0)
+ atomicMax(changed, something_changed);
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweepingMap< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+
+
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..adff615fd43cd012ba5e45095140d7c2e0e5fbf0
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap2D_impl.h
@@ -0,0 +1,823 @@
+/***************************************************************************
+ tnlFastSweepingMap2D_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+
+#define MAP_SOLVER_MAX_VALUE 3
+
+
+#include "tnlFastSweepingMap.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweepingMap< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlFastSweepingMap()
+:Entity(Mesh),
+ dofVector(Mesh),
+ dofVector2(Mesh)
+{
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+ dofVector2.load(initialCondition);
+
+ const String& mapFile = parameters.getParameter <String>("map");
+ if(! this->map.load( mapFile ))
+ cout << "Failed to load map file : " << mapFile << endl;
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ Entity.refresh();
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+ cout << "a" << endl;
+
+ something_changed = 1;
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ for(int i=0; i< Mesh.getDimensions().x()*Mesh.getDimensions().x();i++)
+ {
+ dofVector2[i]=INT_MAX*sign(dofVector[i]);
+
+ if(abs(dofVector[i]) < 1.01*h)
+ {
+ dofVector2[i] = dofVector[i];
+ if(map[i] != 0.0)
+ dofVector2[i] /= map[i];
+ }
+ }
+
+// for(int i = 0 ; i < Mesh.getDimensions().x()-1; i++)
+// {
+// for(int j = 0 ; j < Mesh.getDimensions().x()-1; j++)
+// {
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// neighbourEntities.refresh(Mesh,Entity.getIndex());
+//
+// if(dofVector[this->Entity.getIndex()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1111(i,j);
+// else
+// setupSquare1110(i,j);
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1101(i,j);
+// else
+// setupSquare1100(i,j);
+// }
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1011(i,j);
+// else
+// setupSquare1010(i,j);
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare1001(i,j);
+// else
+// setupSquare1000(i,j);
+// }
+// }
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0111(i,j);
+// else
+// setupSquare0110(i,j);
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0101(i,j);
+// else
+// setupSquare0100(i,j);
+// }
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0011(i,j);
+// else
+// setupSquare0010(i,j);
+// }
+// else
+// {
+// if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+// setupSquare0001(i,j);
+// else
+// setupSquare0000(i,j);
+// }
+// }
+// }
+//
+// }
+// }
+ cout << "a" << endl;
+
+ //data.setLike(dofVector2.getData());
+ //data=dofVector2.getData();
+ //cout << data.getType() << endl;
+ dofVector2.save("u-00000.tnl");
+ //dofVector2.getData().save("u-00000.tnl");
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+ int cntr = 0;
+ while(something_changed != 0)
+ {
+ something_changed = 0;
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ cntr++;
+ cout << "Finished set of sweeps #" << cntr << " " << something_changed << endl;
+ }
+
+
+
+ dofVector2.save("u-00001.tnl");
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ if(map[Entity.getIndex()] != 0.0)
+ {
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ Real value = dofVector2[Entity.getIndex()];
+ Real im = abs(1.0/map[Entity.getIndex()]);
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(fabs(a-b) >= im*h)
+ tmp = fabsMin(a,b) + sign(value)*im*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * im * h * im * h - (a - b) * (a - b) ) );
+
+ if(abs(value)-abs(tmp) > 0.0)
+ something_changed = 1;
+
+ dofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+
+ }
+ else
+ {
+ dofVector2[Entity.getIndex()] = MAP_SOLVER_MAX_VALUE;
+ }
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap_CUDA.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap_CUDA.h
new file mode 100644
index 0000000000000000000000000000000000000000..aa606ea47999ce739fb23d7ad6c38f937cf23f26
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping-map/tnlFastSweepingMap_CUDA.h
@@ -0,0 +1,196 @@
+/***************************************************************************
+ tnlFastSweepingMap_CUDA.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING_H_
+#define TNLFASTSWEEPING_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+
+#include <functions/tnlMeshFunction.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlFastSweepingMap();
+
+ __host__ static String getType();
+ __host__ bool init( const Config::ParameterContainer& parameters );
+ __host__ bool run();
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid();
+ __device__ void updateValue(const Index i, const Index j, Index* something_changed);
+ __device__ void updateValue(const Index i, const Index j, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlFastSweepingMap< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ double* map_cuda;
+ int counter;
+ int* changed;
+ __device__ void setupSquare1000(Index i, Index j);
+ __device__ void setupSquare1100(Index i, Index j);
+ __device__ void setupSquare1010(Index i, Index j);
+ __device__ void setupSquare1001(Index i, Index j);
+ __device__ void setupSquare1110(Index i, Index j);
+ __device__ void setupSquare1101(Index i, Index j);
+ __device__ void setupSquare1011(Index i, Index j);
+ __device__ void setupSquare1111(Index i, Index j);
+ __device__ void setupSquare0000(Index i, Index j);
+ __device__ void setupSquare0100(Index i, Index j);
+ __device__ void setupSquare0010(Index i, Index j);
+ __device__ void setupSquare0001(Index i, Index j);
+ __device__ void setupSquare0110(Index i, Index j);
+ __device__ void setupSquare0101(Index i, Index j);
+ __device__ void setupSquare0011(Index i, Index j);
+ __device__ void setupSquare0111(Index i, Index j);
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data, map;
+
+
+ RealType h;
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweepingMap< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+
+ __host__ static String getType();
+ __host__ bool init( const Config::ParameterContainer& parameters );
+ __host__ bool run();
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid(int i, int j, int k);
+ __device__ void updateValue(const Index i, const Index j, const Index k);
+ __device__ void updateValue(const Index i, const Index j, const Index k, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlFastSweepingMap< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ int counter;
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data;
+
+ RealType h;
+
+
+};
+
+
+
+
+
+
+
+#ifdef HAVE_CUDA
+//template<int sweep_t>
+__global__ void runCUDA(tnlFastSweepingMap< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i, int* changed);
+//__global__ void runCUDA(tnlFastSweepingMap< tnlGrid< 3,double, tnlHost, int >, double, int >* solver, int sweep, int i);
+
+__global__ void initCUDA(tnlFastSweepingMap< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+//__global__ void initCUDA(tnlFastSweepingMap< tnlGrid< 3,double, tnlHost, int >, double, int >* solver);
+#endif
+
+/*various implementtions.... choose one*/
+//#include "tnlFastSweepingMap2D_CUDA_impl.h"
+//#include "tnlFastSweepingMap2D_CUDA_v2_impl.h"
+//#include "tnlFastSweepingMap2D_CUDA_v3_impl.h"
+#include "tnlFastSweepingMap2D_CUDA_v4_impl.h"
+//#include "tnlFastSweepingMap2D_CUDA_v5_impl.h"
+
+
+// #include "tnlFastSweepingMap3D_CUDA_impl.h"
+
+#endif /* TNLFASTSWEEPING_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..7a2963cc680e3a5c83f9116c19c715b3d44d7a42
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/CMakeLists.txt
@@ -0,0 +1,22 @@
+set( tnl_fast_sweeping_SOURCES
+# MainBuildConfig.h
+# tnlFastSweeping2D_impl.h
+# tnlFastSweeping.h
+# fastSweepingConfig.h
+ main.cpp)
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(fast-sweeping${debugExt} main.cu)
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(fast-sweeping${debugExt} main.cpp)
+ENDIF( BUILD_CUDA )
+target_link_libraries (fast-sweeping${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS fast-sweeping${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+#INSTALL( FILES ${tnl_fast_sweeping_SOURCES}
+# DESTINATION share/tnl-${tnlVersion}/examples/fast-sweeping )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/MainBuildConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/fastSweepingConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/fastSweepingConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..3df2c1e889050448fc07baf7dcd0e32feab3f778
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/fastSweepingConfig.h
@@ -0,0 +1,38 @@
+/***************************************************************************
+ fastSweepingConfig.h - description
+ -------------------
+ begin : Oct 15, 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef FASTSWEEPINGCONFIG_H_
+#define FASTSWEEPINGCONFIG_H_
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class fastSweepingConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Parallel Eikonal solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "fast-sweeping" );
+ config.addRequiredEntry < String > ( "initial-condition", "Initial condition for solver");
+ config.addRequiredEntry < int > ( "dim", "Dimension of problem.");
+ config.addEntry < String > ( "mesh", "Name of mesh.", "mesh.tnl" );
+ config.addEntry < String > ( "exact-input", "Are the function values near the curve equal to the SDF? (yes/no)", "no" );
+ }
+};
+
+#endif /* FASTSWEEPINGCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cpp
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cu b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cu
new file mode 100644
index 0000000000000000000000000000000000000000..8849008ff630db0400a6d7d98e789099e5fbb5d9
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.cu
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.h
new file mode 100644
index 0000000000000000000000000000000000000000..217377f3b77645f2ad2ede8db68bdada5f0885fb
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/main.h
@@ -0,0 +1,88 @@
+/***************************************************************************
+ main.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+
+#include "MainBuildConfig.h"
+ //for HOST versions:
+#include "tnlFastSweeping.h"
+ //for DEVICE versions:
+//#include "tnlFastSweeping_CUDA.h"
+#include "fastSweepingConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+
+#include <mesh/tnlGrid.h>
+#include <core/tnlDevice.h>
+#include <time.h>
+#include <ctime>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ time_t start;
+ time_t stop;
+ time(&start);
+ std::clock_t start2= std::clock();
+ Config::ParameterContainer parameters;
+ tnlConfigDescription configDescription;
+ fastSweepingConfig< BuildConfig >::configSetup( configDescription );
+
+ if( ! parseCommandLine( argc, argv, configDescription, parameters ) )
+ return false;
+
+ const int& dim = parameters.getParameter< int >( "dim" );
+
+ if(dim == 2)
+ {
+ tnlFastSweeping<tnlGrid<2,double,tnlHost, int>, double, int> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ checkCudaDevice;
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver..." << endl;
+ solver.run();
+ }
+ else if(dim == 3)
+ {
+ tnlFastSweeping<tnlGrid<3,double,tnlHost, int>, double, int> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ checkCudaDevice;
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver..." << endl;
+ solver.run();
+ }
+ else
+ {
+ cerr << "Unsupported number of dimensions: " << dim << "!" << endl;
+ return EXIT_FAILURE;
+ }
+
+
+ time(&stop);
+ cout << "Solver stopped..." << endl;
+ cout << endl;
+ cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) << endl;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping.h
new file mode 100644
index 0000000000000000000000000000000000000000..55f145af95deba9752fb5c4f564bbbfa331bf153
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping.h
@@ -0,0 +1,186 @@
+/***************************************************************************
+ tnlFastSweeping.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING_H_
+#define TNLFASTSWEEPING_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlMeshFunction.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+#ifdef HAVE_OPENMP
+#include <omp.h>
+#endif
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlFastSweeping
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ tnlFastSweeping();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+
+ void setupSquare1000(Index i, Index j);
+ void setupSquare1100(Index i, Index j);
+ void setupSquare1010(Index i, Index j);
+ void setupSquare1001(Index i, Index j);
+ void setupSquare1110(Index i, Index j);
+ void setupSquare1101(Index i, Index j);
+ void setupSquare1011(Index i, Index j);
+ void setupSquare1111(Index i, Index j);
+ void setupSquare0000(Index i, Index j);
+ void setupSquare0100(Index i, Index j);
+ void setupSquare0010(Index i, Index j);
+ void setupSquare0001(Index i, Index j);
+ void setupSquare0110(Index i, Index j);
+ void setupSquare0101(Index i, Index j);
+ void setupSquare0011(Index i, Index j);
+ void setupSquare0111(Index i, Index j);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage > Entity;
+
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlFastSweeping();
+
+ static String getType();
+ bool init( const Config::ParameterContainer& parameters );
+
+ bool initGrid();
+ bool run();
+
+ //for single core version use this implementation:
+ void updateValue(const Index i, const Index j, const Index k);
+ //for parallel version use this one instead:
+// void updateValue(const Index i, const Index j, DofVectorType* grid);
+
+ Real fabsMin(const Real x, const Real y);
+
+
+protected:
+
+ MeshType Mesh;
+
+ bool exactInput;
+
+
+ tnlMeshFunction<MeshType> dofVector, dofVector2;
+ DofVectorType data;
+
+ RealType h;
+
+ tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage > Entity;
+
+#ifdef HAVE_OPENMP
+// omp_lock_t* gridLock;
+#endif
+
+
+};
+
+
+ //for single core version use this implementation:
+#include "tnlFastSweeping2D_impl.h"
+ //for parallel version use this one instead:
+// #include "tnlFastSweeping2D_openMP_impl.h"
+
+#include "tnlFastSweeping3D_impl.h"
+
+#endif /* TNLFASTSWEEPING_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..48cbba5cb2e247296d41d3b7c2f935c8f75ce960
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h
@@ -0,0 +1,522 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.getSpaceSteps().x();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+//
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+//
+// dofVector.save("u-00001.tnl");
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(32, 32);
+ dim3 numBlocks(n/32 + 1 ,n/32 +1);
+
+ for(int i = 2*n - 1; i > -1; i--)
+ {
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,4,i);
+ cudaDeviceSynchronize();
+ }
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ for(int i = 0; i < 2*n ; i++)
+ {
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,i);
+ cudaDeviceSynchronize();
+ }
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ for(int i = 0; i < 2*n ; i++)
+ {
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,i);
+ cudaDeviceSynchronize();
+ }
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ for(int i = 2*n - 1; i > -1; i--)
+ {
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,3,i);
+ cudaDeviceSynchronize();
+ }
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData(), cudaDofVector, this->dofVector.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = cudaDofVector[index];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)];
+ else
+ {
+ a = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)];
+ else
+ {
+ b = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ cudaDofVector[index] = fabsMin(value, tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gid = Mesh.getCellIndex(CoordinatesType(gx,gy));
+
+ int total = blockDim.x*gridDim.x;
+
+
+
+ Real tmp = 0.0;
+ int flag = 0;
+ counter = 0;
+ tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+
+
+ if(!exactInput)
+ {
+ cudaDofVector[gid]=cudaDofVector[gid]=0.5*h*sign(cudaDofVector[gid]);
+ }
+ __threadfence();
+// printf("-----------------------------------------------------------------------------------\n");
+
+ __threadfence();
+
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1)
+ {
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1)
+ {
+
+ Index j = gy;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag=1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+ }
+
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+// printf("****************************************************************\n");
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == 0)
+ {
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ Index j = 0;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n");
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == Mesh.getDimensions().y() - 1)
+ {
+ Index i = gx;
+ Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == 0)
+ {
+ Index j = gy;
+ Index i = 0;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+// printf("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == Mesh.getDimensions().x() - 1)
+ {
+ Index j = gy;
+ Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("##################################################################################################\n");
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == 0)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+// printf("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$\n");
+ if(gx == 0 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == 0 &&
+ gy == 0)
+ {
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+
+ __threadfence();
+
+ if(flag==1)
+ cudaDofVector[gid] = tmp*3;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ Real fy = abs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ return;
+ int total = solver->Mesh.getDimensions().x();
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ int id1 = gx+gy;
+ int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ if(sweep == 1)
+// for(int i = 0; i < 2*total - 1; i++)
+ {
+ if(id1 == i)
+ {
+ solver->updateValue(gx,gy);
+ return;
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ else if(sweep == 2)
+// for(int i = 0; i < 2*total - 1; i++)
+ {
+ if(id2 == i)
+ {
+ solver->updateValue(gx,gy);
+ return;
+ }
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ else if(sweep == 3)
+// for(int i = 2*total - 2; i > -1; i--)
+ {
+ if(id1 == i)
+ {
+ solver->updateValue(gx,gy);
+ return;
+ }
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ else if(sweep == 4)
+// for(int i = 2*total - 2; i > -1; i--)
+ {
+ if(id2 == i)
+ {
+ solver->updateValue(gx,gy);
+ return;
+ }
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..e6b20eb4153b04248a3bf72b2198d923aa21d748
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h
@@ -0,0 +1,588 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_v2_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.getSpaceSteps().x();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+//
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+//
+// dofVector.save("u-00001.tnl");
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(27, 27);
+ dim3 numBlocks(1 ,1);
+
+// for(int i = 2*n - 1; i > -1; i--)
+ {
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,4,0);
+ cudaDeviceSynchronize();
+ }
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+//// for(int i = 0; i < 2*n ; i++)
+// {
+// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,0);
+// cudaDeviceSynchronize();
+// }
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+//// for(int i = 0; i < 2*n ; i++)
+// {
+// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,0);
+// cudaDeviceSynchronize();
+// }
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+//// for(int i = 2*n - 1; i > -1; i--)
+// {
+// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,3,0);
+// cudaDeviceSynchronize();
+// }
+//
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData(), cudaDofVector, this->dofVector.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = cudaDofVector[index];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)];
+ else
+ {
+ a = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)];
+ else
+ {
+ b = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ cudaDofVector[index] = fabsMin(value, tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gid = Mesh.getCellIndex(CoordinatesType(gx,gy));
+
+ int total = blockDim.x*gridDim.x;
+
+
+
+ Real tmp = 0.0;
+ int flag = 0;
+ counter = 0;
+ tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+
+
+ if(!exactInput)
+ {
+ cudaDofVector[gid]=cudaDofVector[gid]=0.5*h*sign(cudaDofVector[gid]);
+ }
+ __threadfence();
+// printf("-----------------------------------------------------------------------------------\n");
+
+ __threadfence();
+
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1)
+ {
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1)
+ {
+
+ Index j = gy;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag=1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+ }
+
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+// printf("****************************************************************\n");
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == 0)
+ {
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ Index j = 0;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n");
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == Mesh.getDimensions().y() - 1)
+ {
+ Index i = gx;
+ Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == 0)
+ {
+ Index j = gy;
+ Index i = 0;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+// printf("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == Mesh.getDimensions().x() - 1)
+ {
+ Index j = gy;
+ Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("##################################################################################################\n");
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == 0)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+// printf("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$\n");
+ if(gx == 0 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == 0 &&
+ gy == 0)
+ {
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+
+ __threadfence();
+
+ if(flag==1)
+ cudaDofVector[gid] = tmp*3;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+
+ Real tmpMin = Min(fx,abs(y));
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+{
+
+ //int gx = threadIdx.x;
+ //int gy = threadIdx.y;
+ int id1,id2;
+ int nx = solver->Mesh.getDimensions().x()+ threadIdx.x;
+ int ny = solver->Mesh.getDimensions().y()+ threadIdx.y;
+
+ int blockCount = solver->Mesh.getDimensions().x()/blockDim.x + 1;
+
+ for(int gy = threadIdx.y; gy < ny;gy+=blockDim.y)
+ {
+ for(int gx = threadIdx.x; gx < nx;gx+=blockDim.x)
+ {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && gy > -1&& gx > -1)
+ {
+ id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+
+ }
+ //gx+=blockDim.x;
+ //__syncthreads();
+ }
+ //gx = threadIdx.x;
+ //gy+=blockDim.y;
+ //__syncthreads();
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+// gx = blockDim.x*(blockCount-1) + threadIdx.x;
+// gy = threadIdx.y;
+ for(int gy = threadIdx.y; gy < ny;gy+=blockDim.y)
+ {
+ for(int gx = blockDim.x*(blockCount-1) + threadIdx.x; gx >- 1;gx-=blockDim.x)
+ {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && gy > -1&& gx > -1)
+ {
+ id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+ }
+ //gx-=blockDim.x;
+ //__syncthreads();
+ }
+ //gx = blockDim.x*(blockCount-1) + threadIdx.x;
+ //gy+=blockDim.y;
+ //__syncthreads();
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+// gx = blockDim.x*(blockCount-1) + threadIdx.x;
+// gy = blockDim.x*(blockCount-1) + threadIdx.y;
+ for(int gy = blockDim.x*(blockCount-1) +threadIdx.y; gy >- 1;gy-=blockDim.y)
+ {
+ for(int gx = blockDim.x*(blockCount-1) + threadIdx.x; gx >- 1;gx-=blockDim.x)
+ {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && gy > -1&& gx > -1)
+ {
+ id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+ }
+ //gx-=blockDim.x;
+ //__syncthreads();
+ }
+ //gx = blockDim.x*(blockCount-1) + threadIdx.x;
+ //gy-=blockDim.y;
+ //__syncthreads();
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+ //gx = threadIdx.x;
+ //gy = blockDim.x*(blockCount-1) +threadIdx.y;
+ for(int gy = blockDim.x*(blockCount-1) +threadIdx.y; gy >- 1;gy-=blockDim.y)
+ {
+ for(int gx = threadIdx.x; gx < nx;gx+=blockDim.x)
+ {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && gy > -1&& gx > -1)
+ {
+ id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+ }
+ //gx+=blockDim.x;
+ //__syncthreads();
+ }
+ //gx = threadIdx.x;
+ //gy-=blockDim.y;
+ ///__syncthreads();
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..d05b649f29d222bc96649d3d6ed92a7fa11cf14e
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h
@@ -0,0 +1,920 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_v3_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+
+
+
+__device__ double atomicSet(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong(val ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.getSpaceSteps().x();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+//
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+//
+// dofVector.save("u-00001.tnl");
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 +1 ,n/16 +1);
+ int m =n/16 +1;
+
+ for(int i = 0; i < 2*m -1; i++)
+ {
+ runCUDA<15><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,i);
+ //cudaDeviceSynchronize();
+ }
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+// for(int i = 0; i < 2*m -1; i++)
+// {
+// runCUDA<2><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,i);
+// cudaDeviceSynchronize();
+// }
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+// for(int i = 0; i < 2*m -1; i++)
+// {
+// runCUDA<4><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,4,i);
+// cudaDeviceSynchronize();
+// }
+// cudaDeviceSynchronize();
+// checkCudaDevice;
+// for(int i = 0; i < 2*m -1; i++)
+// {
+// runCUDA<8><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,8,i);
+// cudaDeviceSynchronize();
+// }
+
+
+
+
+// for(int i = 0; i < (2*m -1)/4 -1; i++)
+// {
+// runCUDA<15><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,15,i);//all
+// cudaDeviceSynchronize();
+// }
+// for(int i = (2*m -1)/4 -1; i < (2*m -1)/2 -1; i++)
+// {
+// runCUDA<5><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,5,i); //two
+// cudaDeviceSynchronize();
+// runCUDA<10><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,10,i); //two
+// cudaDeviceSynchronize();
+// }
+// for(int i = (2*m -1)/2 -1; i < (2*m -1)/2 +1; i++)
+// {
+// runCUDA<1><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,i); //separate
+// cudaDeviceSynchronize();
+// runCUDA<2><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,i); //separate
+// cudaDeviceSynchronize();
+// runCUDA<4><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,4,i); //separate
+// cudaDeviceSynchronize();
+// runCUDA<8><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,8,i); //separate
+// cudaDeviceSynchronize();
+// }
+// for(int i = (2*m -1)/2 +1; i < (2*m -1/4)*3 +1; i++)
+// {
+// runCUDA<5><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,5,i); //two
+// cudaDeviceSynchronize();
+// runCUDA<10><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,10,i); //two
+// cudaDeviceSynchronize();
+// }
+// for(int i = (2*m -1/4)*3 +1; i < 2*m -1; i++)
+// {
+// runCUDA<15><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,15,i);//all
+// cudaDeviceSynchronize();
+// }
+cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData(), cudaDofVector, this->dofVector.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = cudaDofVector[index];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)];
+ else
+ {
+ a = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)];
+ else
+ {
+ b = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ atomicSet(&cudaDofVector[index],fabsMin(value, tmp));
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gid = Mesh.getCellIndex(CoordinatesType(gx,gy));
+
+ int total = blockDim.x*gridDim.x;
+
+
+
+ Real tmp = 0.0;
+ int flag = 0;
+ counter = 0;
+ tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+
+
+ if(!exactInput)
+ {
+ cudaDofVector[gid]=cudaDofVector[gid]=0.5*h*sign(cudaDofVector[gid]);
+ }
+ __threadfence();
+// printf("-----------------------------------------------------------------------------------\n");
+
+ __threadfence();
+
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1)
+ {
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1)
+ {
+
+ Index j = gy;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag=1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+ }
+
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+// printf("****************************************************************\n");
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == 0)
+ {
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ Index j = 0;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n");
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == Mesh.getDimensions().y() - 1)
+ {
+ Index i = gx;
+ Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == 0)
+ {
+ Index j = gy;
+ Index i = 0;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+// printf("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == Mesh.getDimensions().x() - 1)
+ {
+ Index j = gy;
+ Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("##################################################################################################\n");
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == 0)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+// printf("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$\n");
+ if(gx == 0 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == 0 &&
+ gy == 0)
+ {
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+
+ __threadfence();
+
+ if(flag==1)
+ cudaDofVector[gid] = tmp*3;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+// Real fx = abs(x);
+//
+// Real tmpMin = Min(fx,abs(y));
+
+ if(abs(y) > abs(x))
+ return x;
+ else
+ return y;
+
+
+}
+
+
+template<>
+__global__ void runCUDA<1>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+{
+
+ if(blockIdx.x+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+}
+ template<>
+ __global__ void runCUDA<2>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+ if((gridDim.x - blockIdx.x - 1)+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ }
+ __syncthreads();
+ }
+
+ }
+ } /*---------------------------------------------------------------------------------------------------------------------------*/
+ template<>
+ __global__ void runCUDA<4>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+ if(blockIdx.x+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ }
+
+ template<>
+ __global__ void runCUDA<8>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+ if((gridDim.x - blockIdx.x - 1)+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+
+}
+
+
+ template<>
+ __global__ void runCUDA<5>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+
+ if(blockIdx.x+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ else if(blockIdx.x+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ }
+
+
+ template<>
+ __global__ void runCUDA<10>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+ if((gridDim.x - blockIdx.x - 1)+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+
+ else if((gridDim.x - blockIdx.x - 1)+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+
+ }
+
+
+
+ template<>
+ __global__ void runCUDA<15>(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+ {
+
+ if(blockIdx.x+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ if((gridDim.x - blockIdx.x - 1)+blockIdx.y == k)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 0; l < 2*blockDim.x - 1; l++)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ if(blockIdx.x+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id1 = threadIdx.x+threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id1 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+ if((gridDim.x - blockIdx.x - 1)+blockIdx.y == gridDim.x+gridDim.y-k-2)
+ {
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = threadIdx.y + blockDim.y*blockIdx.y;
+
+ int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+
+ for(int l = 2*blockDim.x - 2; l > -1; l--)
+ {
+ if(id2 == l)
+ {
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+ solver->updateValue(gx,gy);
+ return;
+ }
+ __syncthreads();
+ }
+
+ }
+ /*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+
+ }
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+#endif
+
+
+
+
+
+
+//__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int k)
+//{
+//
+// if(sweep==1 && blockIdx.x+blockIdx.y == k)
+// {
+// int gx = threadIdx.x + blockDim.x*blockIdx.x;
+// int gy = threadIdx.y + blockDim.y*blockIdx.y;
+//
+// int id1 = threadIdx.x+threadIdx.y;
+//
+// for(int l = 0; l < 2*blockDim.x - 1; l++)
+// {
+// if(id1 == l)
+// {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+// solver->updateValue(gx,gy);
+// }
+// __syncthreads();
+// }
+//
+// }
+// /*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// else if(sweep==2 && (gridDim.x - blockIdx.x - 1)+blockIdx.y == k)
+// {
+// int gx = threadIdx.x + blockDim.x*blockIdx.x;
+// int gy = threadIdx.y + blockDim.y*blockIdx.y;
+//
+// int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+//
+// for(int l = 0; l < 2*blockDim.x - 1; l++)
+// {
+// if(id2 == l)
+// {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+// solver->updateValue(gx,gy);
+// }
+// __syncthreads();
+// }
+//
+// }
+// /*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// else if(sweep==4 && blockIdx.x+blockIdx.y == gridDim.x+gridDim.y-k-2)
+// {
+// int gx = threadIdx.x + blockDim.x*blockIdx.x;
+// int gy = threadIdx.y + blockDim.y*blockIdx.y;
+//
+// int id1 = threadIdx.x+threadIdx.y;
+//
+// for(int l = 2*blockDim.x - 2; l > -1; l--)
+// {
+// if(id1 == l)
+// {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+// solver->updateValue(gx,gy);
+// return;
+// }
+// __syncthreads();
+// }
+//
+// }
+// /*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// else if(sweep==8 && (gridDim.x - blockIdx.x - 1)+blockIdx.y == gridDim.x+gridDim.y-k-2)
+// {
+// int gx = threadIdx.x + blockDim.x*blockIdx.x;
+// int gy = threadIdx.y + blockDim.y*blockIdx.y;
+//
+// int id2 = (blockDim.x - threadIdx.x - 1) + threadIdx.y;
+//
+// for(int l = 2*blockDim.x - 2; l > -1; l--)
+// {
+// if(id2 == l)
+// {
+// if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy /*&& gy > -1&& gx > -1*/)
+// solver->updateValue(gx,gy);
+// return;
+// }
+// __syncthreads();
+// }
+//
+// }
+// /*---------------------------------------------------------------------------------------------------------------------------*/
+//
+//
+//
+//
+//
+//}
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..230b6ce436bb53c40172eb43cf8b7e668f4bf891
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h
@@ -0,0 +1,1003 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlFastSweeping()
+:dofVector(Mesh)
+{
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ //Entity.refresh();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(1, 1024);
+ dim3 numBlocks(4,1);
+
+
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,0,0);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ //data.setLike(dofVector.getData());
+ //cudaMemcpy(data.getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaMemcpy(dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ //data.save("u-00001.tnl");
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+// cudaDofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+ atomicFabsMin(&(cudaDofVector2[Entity.getIndex()]), tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int i = threadIdx.x + blockDim.x*blockIdx.x;
+ int j = blockDim.y*blockIdx.y + threadIdx.y;
+
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ int gid = Entity.getIndex();
+
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector[gid]);
+//
+// if(abs(cudaDofVector[gid]) < 1.01*h)
+// cudaDofVector2[gid] = cudaDofVector[gid];
+
+
+
+
+
+ if(i+1 < Mesh.getDimensions().x() && j+1 < Mesh.getDimensions().y() && !exactInput)
+ {
+ if(cudaDofVector[Entity.getIndex()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1111(i,j);
+ else
+ setupSquare1110(i,j);
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1101(i,j);
+ else
+ setupSquare1100(i,j);
+ }
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1011(i,j);
+ else
+ setupSquare1010(i,j);
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1001(i,j);
+ else
+ setupSquare1000(i,j);
+ }
+ }
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0111(i,j);
+ else
+ setupSquare0110(i,j);
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0101(i,j);
+ else
+ setupSquare0100(i,j);
+ }
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0011(i,j);
+ else
+ setupSquare0010(i,j);
+ }
+ else
+ {
+ if(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0001(i,j);
+ else
+ setupSquare0000(i,j);
+ }
+ }
+ }
+
+ }
+ if(exactInput)
+ {
+ if(abs(cudaDofVector[gid]) < 1.5*h)
+ cudaDofVector2[gid] = cudaDofVector[gid];
+ }
+
+
+ return true;
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ //Real fy = abs(y);
+
+ //Real tmpMin = Min(fx,abs(y));
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+ int gx = 0;
+ int gy = threadIdx.y;
+ //if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ // return;
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ //int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ //int id1 = gx+gy;
+ //int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+ if(blockIdx.x==0)
+ {
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+
+
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+// tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+// Entity.setCoordinates(CoordinatesType(i,j));
+// Entity.refresh();
+// tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+// cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+// tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+// Entity.setCoordinates(CoordinatesType(i,j));
+// Entity.refresh();
+// tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+// cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=INT_MAX; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=INT_MAX; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-INT_MAX; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[Entity.getIndex()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-INT_MAX; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Real al,be, a,b,c,s;
+ al=abs(cudaDofVector[Entity.getIndex()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ cudaDofVector[Entity.getIndex()]));
+
+ be=abs(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b = 1.0;
+ c = -be;
+ s = h/sqrt(a*a+b*b);
+
+
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ tnlGridEntity< tnlGrid< 2,double, tnlHost, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(cudaDofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..f9f3e28f63fc7a3ad88b98980a4454035fd299b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h
@@ -0,0 +1,697 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_v5_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(assumed,val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.getSpaceSteps().x();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData(), this->dofVector.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(16, 16);
+ dim3 numBlocks(n/16 + 1 ,n/16 +1);
+
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+//
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+// for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+// for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+// {
+// for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+// {
+// updateValue(i,j);
+// }
+// }
+//
+///*---------------------------------------------------------------------------------------------------------------------------*/
+//
+//
+// dofVector.save("u-00001.tnl");
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(1, 512);
+ dim3 numBlocks(4,1);
+
+
+ runCUDA<<<numBlocks,threadsPerBlock,3*(512+1)*sizeof(double)>>>(this->cudaSolver,0,0);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData(), cudaDofVector, this->dofVector.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = cudaDofVector[index];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)];
+ else
+ {
+ a = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<-1,0>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)];
+ else
+ {
+ b = fabsMin( cudaDofVector[Mesh.template getCellNextToCell<0,-1>(index)],
+ cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+ cudaDofVector[index] = fabsMin(value, tmp);
+
+}
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, double** sharedMem, int k3)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = sharedMem[k3+1][threadIdx.y];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = sharedMem[k3][threadIdx.y];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = sharedMem[k3+2][threadIdx.y];
+ else
+ {
+ a = fabsMin( sharedMem[k3][threadIdx.y],
+ sharedMem[k3+2][threadIdx.y] );
+ }
+
+ if( j == 0 )
+ b = sharedMem[k3][threadIdx.y+1];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = sharedMem[k3+2][threadIdx.y-1];
+ else
+ {
+ b = fabsMin( sharedMem[k3][threadIdx.y+1],
+ sharedMem[k3+2][threadIdx.y-1] );
+ }
+
+
+ if(abs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+// sharedMem[k3+1][threadIdx.y] = this->fabsMin(value, tmp);
+// atomicFabsMin(&(cudaDofVector[index]), tmp);
+ cudaDofVector[index] = tmp;
+ this->fabsMin(value, tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gid = Mesh.getCellIndex(CoordinatesType(gx,gy));
+
+ int total = blockDim.x*gridDim.x;
+
+
+
+ Real tmp = 0.0;
+ int flag = 0;
+ counter = 0;
+ tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+
+
+ if(!exactInput)
+ {
+ cudaDofVector[gid]=cudaDofVector[gid]=0.5*h*sign(cudaDofVector[gid]);
+ }
+ __threadfence();
+// printf("-----------------------------------------------------------------------------------\n");
+
+ __threadfence();
+
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1)
+ {
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1)
+ {
+
+ Index j = gy;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag=1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+ }
+
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+// printf("****************************************************************\n");
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == 0)
+ {
+// printf("gx: %d, gy: %d, gid: %d \n", gx, gy,gid);
+ Index j = 0;
+ Index i = gx;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++\n");
+ if(gx > 0 && gx < Mesh.getDimensions().x()-1 && gy == Mesh.getDimensions().y() - 1)
+ {
+ Index i = gx;
+ Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == 0)
+ {
+ Index j = gy;
+ Index i = 0;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+// printf("@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@\n");
+ if(gy > 0 && gy < Mesh.getDimensions().y()-1 && gx == Mesh.getDimensions().x() - 1)
+ {
+ Index j = gy;
+ Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+// printf("##################################################################################################\n");
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == Mesh.getDimensions().x() - 1 &&
+ gy == 0)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx-1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+// printf("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$\n");
+ if(gx == 0 &&
+ gy == Mesh.getDimensions().y() - 1)
+ {
+
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy-1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+ if(gx == 0 &&
+ gy == 0)
+ {
+// tmp = sign(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))]);
+ if(cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx+1,gy))]*tmp > 0.0 &&
+ cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy+1))]*tmp > 0.0)
+
+ flag = 1;//cudaDofVector[Mesh.getCellIndex(CoordinatesType(gx,gy))] = tmp*INT_MAX;
+ }
+
+ __threadfence();
+
+ if(flag==1)
+ cudaDofVector[gid] = tmp*3;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ //Real fy = abs(y);
+
+ //Real tmpMin = Min(fx,abs(y));
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+ extern __shared__ double u[];
+ double* sharedMem[5];
+ sharedMem[0] = u;
+ sharedMem[1] = &(u[blockDim.y+1]);
+ sharedMem[2] = &(sharedMem[1][blockDim.y+1]);
+ sharedMem[3] = sharedMem[1];
+ sharedMem[4] = sharedMem[2];
+
+ int gx = 0;
+ int gy = threadIdx.y;
+ //if(solver->Mesh.getDimensions().x() <= gx || solver->Mesh.getDimensions().y() <= gy)
+ // return;
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+ //int gid = solver->Mesh.getDimensions().x() * gy + gx;
+ //int max = solver->Mesh.getDimensions().x()*solver->Mesh.getDimensions().x();
+
+ //int id1 = gx+gy;
+ //int id2 = (solver->Mesh.getDimensions().x() - gx - 1) + gy;
+
+
+ if(blockIdx.x==0)
+ {
+ if(threadIdx.y==0)
+ sharedMem[1][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(0,0))];
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ int k3=k%3;
+
+ if(threadIdx.y==0)
+ {
+ if(gx==n-1)
+ sharedMem[k3][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(0,gy+blockDim.y))];
+ else
+ sharedMem[k3][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx+1,gy))];
+ }
+// else
+// solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy-1))]=sharedMem[k3+2][threadIdx.y-1];
+
+ if(gy<n-1)
+ sharedMem[k3][threadIdx.y+1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy+1))];
+
+ solver->updateValue(gx,gy,sharedMem,k3);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+// else if(blockIdx.x==1)
+// {
+// gx=n-1;
+// gy=threadIdx.y;
+//
+// if(threadIdx.y==0)
+// sharedMem[1][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(n-1,0))];
+//
+// for(int k = 0; k < n*blockCount + blockDim.y; k++)
+// {
+// if(threadIdx.y < k+1 && gy < n)
+// {
+// int k3=k%3;
+//
+// if(threadIdx.y==0)
+// if(gx==0)
+// sharedMem[k3+2][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(n-1,gy+blockDim.y))];
+// else
+// sharedMem[k3+2][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx-1,gy))];
+// else
+// solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy-1))]=sharedMem[k3][threadIdx.y-1];
+//
+// if(gy<n-1)
+// sharedMem[k3+2][threadIdx.y+1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy+1))];
+//
+//
+// solver->updateValue(gx,gy,sharedMem,k3);
+// gx--;
+// if(gx==-1)
+// {
+// gx=n-1;
+// gy+=blockDim.y;
+// }
+// }
+//
+//
+// __syncthreads();
+// }
+// }
+// else if(blockIdx.x==2)
+// {
+// gx=0;
+// gy=n-blockDim.y-1+threadIdx.y;
+//
+// if(threadIdx.y==0)
+// sharedMem[1][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(0,n-1))];
+//
+// for(int k = 0; k < n*blockCount + blockDim.y; k++)
+// {
+// if(blockDim.y-threadIdx.y < k+1 && gy > -1)
+// {
+// int k3=k%3;
+//
+// if(threadIdx.y==blockDim.y-1)
+// if(gx==n-1)
+// sharedMem[k3][n-1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(0,gy-blockDim.y))];
+// else
+// sharedMem[k3][n-1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx+1,gy))];
+// else
+// solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy-1))]=sharedMem[k3+2][threadIdx.y-1];
+//
+// if(gy<n-1)
+// sharedMem[k3][threadIdx.y+1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy+1))];
+//
+//
+// solver->updateValue(gx,gy,sharedMem,k3);
+// gx++;
+// if(gx==n)
+// {
+// gx=0;
+// gy-=blockDim.y;
+// }
+// }
+//
+//
+// __syncthreads();
+// }
+// }
+// else if(blockIdx.x==3)
+// {
+// gx=n-1;
+// gy=n-blockDim.y-1+threadIdx.y;
+//
+// if(threadIdx.y==0)
+// sharedMem[1][0]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(n-1,n-1))];
+//
+//
+// for(int k = 0; k < n*blockCount + blockDim.y; k++)
+// {
+// if(blockDim.y-threadIdx.y < k+1 && gy > -1)
+// {
+// int k3=k%3;
+//
+// if(threadIdx.y==blockDim.y-1)
+// if(gx==n-1)
+// sharedMem[k3+2][n-1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(n-1,gy-blockDim.y))];
+// else
+// sharedMem[k3+2][n-1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx+1,gy))];
+// else
+// solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy-1))]=sharedMem[k3][threadIdx.y-1];
+//
+// if(gy<n-1)
+// sharedMem[k3+2][threadIdx.y+1]=solver->cudaDofVector[solver->Mesh.getCellIndex(Containers::StaticVector<2,int>(gx,gy+1))];
+//
+//
+// solver->updateValue(gx,gy,sharedMem,k3);
+// gx--;
+// if(gx==-1)
+// {
+// gx=n-1;
+// gy-=blockDim.y;
+// }
+// }
+//
+//
+// __syncthreads();
+// }
+// }
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy)
+ {
+ solver->initGrid();
+ }
+
+
+}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..920806c211483d1c8cc3dabd2e194704dbb560bd
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_impl.h
@@ -0,0 +1,927 @@
+/***************************************************************************
+ tnlFastSweeping2D_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: tnlFastSweeping()
+:Entity(Mesh),
+ dofVector(Mesh),
+ dofVector2(Mesh)
+{
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+ dofVector2.load(initialCondition);
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0 >();
+ Entity.refresh();
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+ cout << "a" << endl;
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ for(int i=0; i< Mesh.getDimensions().x()*Mesh.getDimensions().x();i++)
+ {
+ dofVector2[i]=INT_MAX*sign(dofVector[i]);
+ }
+
+ for(int i = 0 ; i < Mesh.getDimensions().x()-1; i++)
+ {
+ for(int j = 0 ; j < Mesh.getDimensions().x()-1; j++)
+ {
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ neighbourEntities.refresh(Mesh,Entity.getIndex());
+
+ if(dofVector[this->Entity.getIndex()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1111(i,j);
+ else
+ setupSquare1110(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1101(i,j);
+ else
+ setupSquare1100(i,j);
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1011(i,j);
+ else
+ setupSquare1010(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare1001(i,j);
+ else
+ setupSquare1000(i,j);
+ }
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0111(i,j);
+ else
+ setupSquare0110(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0101(i,j);
+ else
+ setupSquare0100(i,j);
+ }
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()] > 0)
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0011(i,j);
+ else
+ setupSquare0010(i,j);
+ }
+ else
+ {
+ if(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()] > 0)
+ setupSquare0001(i,j);
+ else
+ setupSquare0000(i,j);
+ }
+ }
+ }
+
+ }
+ }
+ cout << "a" << endl;
+
+// Real tmp = 0.0;
+// Real ax=0.5/sqrt(2.0);
+//
+// if(!exactInput)
+// {
+// for(Index i = 0; i < Mesh.getDimensions().x()*Mesh.getDimensions().y(); i++)
+// dofVector[i]=0.5*h*sign(dofVector[i]);
+// }
+//
+//
+// for(Index i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// for(Index j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+// }
+//
+//
+//
+// for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// Index j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+// {
+// Index j = Mesh.getDimensions().y() - 1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// Index i = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+// for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+// {
+// Index i = Mesh.getDimensions().x() - 1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+//
+//
+// if(tmp == 0.0)
+// {}
+// else if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+// {}
+// else
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+// }
+//
+//
+// Index i = Mesh.getDimensions().x() - 1;
+// Index j = Mesh.getDimensions().y() - 1;
+//
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// i = 0;
+// j = Mesh.getDimensions().y() -1;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+//
+//
+//
+// j = 0;
+// tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+// if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+//
+// dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+ //data.setLike(dofVector2.getData());
+ //data=dofVector2.getData();
+ //cout << data.getType() << endl;
+ dofVector2.save("u-00000.tnl");
+ //dofVector2.getData().save("u-00000.tnl");
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+// data.setLike(dofVector2.getData());
+// data = dofVector2.getData();
+// cout << data.getType() << endl;
+ dofVector2.save("u-00001.tnl");
+ //dofVector2.getData().save("u-00001.tnl");
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j)
+{
+
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ Real value = dofVector2[Entity.getIndex()];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()];
+ else
+ {
+ a = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< -1, 0 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()];
+ else
+ {
+ b = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, -1 >()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()] );
+ }
+
+
+ if(fabs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+
+ dofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+
+// if(dofVector2[Entity.getIndex()] > 1.0)
+// cout << value << " " << tmp << " " << dofVector2[Entity.getIndex()] << endl;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+{
+// this->Entity.setCoordinates(CoordinatesType(i,j));
+// this->Entity.refresh();
+// auto neighbourEntities = Entity.getNeighbourEntities();
+// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
+// dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+// dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[Entity.getIndex()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ a = be/al;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-al;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ Real al,be, a,b,c,s;
+ al=abs(dofVector[Entity.getIndex()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()]-
+ dofVector[Entity.getIndex()]));
+
+ be=abs(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]/
+ (dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()]-
+ dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()]));
+
+ a = al-be;
+ b=1.0;
+ c=-be;
+ s= h/sqrt(a*a+b*b);
+
+
+ dofVector2[Entity.getIndex()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j));
+ this->Entity.refresh();
+ auto neighbourEntities = Entity.getNeighbourEntities();
+ dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 0, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 1 >()]);
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighbourEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighbourEntities.template getEntityIndex< 1, 0 >()]);
+}
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_openMP_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_openMP_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..d766c62c528b76b61312c5c485692d6655ff8f90
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping2D_openMP_impl.h
@@ -0,0 +1,399 @@
+/***************************************************************************
+ tnlFastSweeping_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING2D_IMPL_H_
+#define TNLFASTSWEEPING2D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ h = Mesh.getSpaceSteps().x();
+
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+#ifdef HAVE_OPENMP
+// gridLock = (omp_lock_t*) malloc(sizeof(omp_lock_t)*Mesh.getDimensions().x()*Mesh.getDimensions().y());
+//
+// for(int i = 0; i < Mesh.getDimensions().x()*Mesh.getDimensions().y(); i++)
+// omp_init_lock(&gridLock[i]);
+#endif
+
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+
+ Real tmp = 0.0;
+
+ if(!exactInput)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x()*Mesh.getDimensions().y(); i++)
+ dofVector[i]=0.5*h*sign(dofVector[i]);
+ }
+
+
+ for(Index i = 1; i < Mesh.getDimensions().x()-1; i++)
+ {
+ for(Index j = 1; j < Mesh.getDimensions().y()-1; j++)
+ {
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+ }
+
+
+
+ for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+ {
+ Index j = 0;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 )
+ {}
+ else
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+ for(int i = 1; i < Mesh.getDimensions().x()-1; i++)
+ {
+ Index j = Mesh.getDimensions().y() - 1;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+ for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+ {
+ Index i = 0;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+ for(int j = 1; j < Mesh.getDimensions().y()-1; j++)
+ {
+ Index i = Mesh.getDimensions().x() - 1;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+
+
+ if(tmp == 0.0)
+ {}
+ else if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp < 0.0 ||
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp < 0.0 )
+ {}
+ else
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+ }
+
+
+ Index i = Mesh.getDimensions().x() - 1;
+ Index j = Mesh.getDimensions().y() - 1;
+
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+ if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+
+
+ j = 0;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+ if(dofVector[Mesh.getCellIndex(CoordinatesType(i-1,j))]*tmp > 0.0 &&
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+
+
+ i = 0;
+ j = Mesh.getDimensions().y() -1;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+ if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j-1))]*tmp > 0.0)
+
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+
+
+ j = 0;
+ tmp = sign(dofVector[Mesh.getCellIndex(CoordinatesType(i,j))]);
+ if(dofVector[Mesh.getCellIndex(CoordinatesType(i+1,j))]*tmp > 0.0 &&
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j+1))]*tmp > 0.0)
+
+ dofVector[Mesh.getCellIndex(CoordinatesType(i,j))] = tmp*INT_MAX;
+
+
+ dofVector.save("u-00000.tnl");
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ DofVectorType d2,d3,d4;
+ d2.setLike(dofVector);
+ d2=dofVector;
+ d3.setLike(dofVector);
+ d3=dofVector;
+ d4.setLike(dofVector);
+ d4=dofVector;
+
+
+#ifdef HAVE_OPENMP
+#pragma omp parallel sections num_threads(4)
+ {
+ {
+#endif
+
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,&dofVector);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+#ifdef HAVE_OPENMP
+ }
+#pragma omp section
+ {
+#endif
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,&d2);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+#ifdef HAVE_OPENMP
+ }
+#pragma omp section
+ {
+#endif
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j, &d3);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+#ifdef HAVE_OPENMP
+ }
+#pragma omp section
+ {
+#endif
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j, &d4);
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+#ifdef HAVE_OPENMP
+ }
+ }
+#endif
+
+
+#ifdef HAVE_OPENMP
+#pragma omp parallel for num_threads(4) schedule(dynamic)
+#endif
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ int index = Mesh.getCellIndex(CoordinatesType(i,j));
+ dofVector[index] = fabsMin(dofVector[index], d2[index]);
+ dofVector[index] = fabsMin(dofVector[index], d3[index]);
+ dofVector[index] = fabsMin(dofVector[index], d4[index]);
+ }
+ }
+
+ dofVector.save("u-00001.tnl");
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, DofVectorType* grid)
+{
+ Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+ Real value = (*grid)[index];
+ Real a,b, tmp;
+
+ if( i == 0 )
+ a = (*grid)[Mesh.template getCellNextToCell<1,0>(index)];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = (*grid)[Mesh.template getCellNextToCell<-1,0>(index)];
+ else
+ {
+ a = fabsMin( (*grid)[Mesh.template getCellNextToCell<-1,0>(index)],
+ (*grid)[Mesh.template getCellNextToCell<1,0>(index)] );
+ }
+
+ if( j == 0 )
+ b = (*grid)[Mesh.template getCellNextToCell<0,1>(index)];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = (*grid)[Mesh.template getCellNextToCell<0,-1>(index)];
+ else
+ {
+ b = fabsMin( (*grid)[Mesh.template getCellNextToCell<0,-1>(index)],
+ (*grid)[Mesh.template getCellNextToCell<0,1>(index)] );
+ }
+
+
+ if(fabs(a-b) >= h)
+ tmp = fabsMin(a,b) + sign(value)*h;
+ else
+ tmp = 0.5 * (a + b + sign(value)*sqrt(2.0 * h * h - (a - b) * (a - b) ) );
+
+#ifdef HAVE_OPENMP
+// omp_set_lock(&gridLock[index]);
+#endif
+ (*grid)[index] = fabsMin(value, tmp);
+#ifdef HAVE_OPENMP
+// omp_unset_lock(&gridLock[index]);
+#endif
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..da045676ae1f119e73cc4539732b31d5d632ca23
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h
@@ -0,0 +1,961 @@
+/***************************************************************************
+ tnlFastSweeping2D_CUDA_v4_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING3D_IMPL_H_
+#define TNLFASTSWEEPING3D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+//__device__
+//double fabsMin( double x, double y)
+//{
+// double fx = abs(x);
+//
+// if(Min(fx,abs(y)) == fx)
+// return x;
+// else
+// return y;
+//}
+//
+//__device__
+//double atomicFabsMin(double* address, double val)
+//{
+// unsigned long long int* address_as_ull =
+// (unsigned long long int*)address;
+// unsigned long long int old = *address_as_ull, assumed;
+// do {
+// assumed = old;
+// old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(assumed,val) ));
+// } while (assumed != old);
+// return __longlong_as_double(old);
+//}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ this->h = Mesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ counter = 0;
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+
+
+#ifdef HAVE_CUDA
+
+ cudaMalloc(&(cudaDofVector), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+ cudaMalloc(&(cudaDofVector2), this->dofVector.getData().getSize()*sizeof(double));
+ cudaMemcpy(cudaDofVector2, this->dofVector.getData().getData(), this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyHostToDevice);
+
+
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >), cudaMemcpyHostToDevice);
+
+#endif
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(8, 8,8);
+ dim3 numBlocks(n/8 + 1, n/8 +1, n/8 +1);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ return true;
+}
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ int n = Mesh.getDimensions().x();
+ dim3 threadsPerBlock(1, 1024);
+ dim3 numBlocks(8,1);
+
+
+ runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,0,0);
+
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ cudaMemcpy(this->dofVector.getData().getData(), cudaDofVector2, this->dofVector.getData().getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ cudaFree(cudaDofVector);
+ cudaFree(cudaDofVector2);
+ cudaFree(cudaSolver);
+ dofVector.save("u-00001.tnl");
+ cudaDeviceSynchronize();
+ return true;
+}
+
+
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, Index k)
+{
+ tnlGridEntity< tnlGrid< 3,double, tnlHost, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j,k));
+ Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ Real value = cudaDofVector2[Entity.getIndex()];
+ Real a,b,c, tmp;
+
+ if( i == 0 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0 >()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ {
+ a = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0 >()] );
+ }
+
+ if( j == 0 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0 >()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0 >()];
+ else
+ {
+ b = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0 >()] );
+ }
+
+ if( k == 0 )
+ c = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1 >()];
+ else if( k == Mesh.getDimensions().z() - 1 )
+ c = cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1 >()];
+ else
+ {
+ c = fabsMin( cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1 >()],
+ cudaDofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1 >()] );
+ }
+
+ Real hD = 3.0*h*h - 2.0*(a*a + b*b + c*c - a*b - a*c - b*c);
+
+ if(hD < 0.0)
+ tmp = fabsMin(a,fabsMin(b,c)) + sign(value)*h;
+ else
+ tmp = (1.0/3.0) * ( a + b + c + sign(value)*sqrt(hD) );
+
+ atomicFabsMin(&cudaDofVector2[Entity.getIndex()],tmp);
+
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid(int i, int j, int k)
+{
+ tnlGridEntity< tnlGrid< 3,double, tnlHost, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
+ Entity.setCoordinates(CoordinatesType(i,j,k));
+ Entity.refresh();
+ int gid = Entity.getIndex();
+
+ if(abs(cudaDofVector[gid]) < 1.0*h)
+ cudaDofVector2[gid] = 0.5*h;//cudaDofVector[gid];
+ else
+ cudaDofVector2[gid] = INT_MAX*sign(cudaDofVector[gid]);
+
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+__device__
+Real tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = abs(x);
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+
+
+}
+
+
+
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 3,double, tnlHost, int >, double, int >* solver, int sweep, int i)
+{
+
+ int gx = 0;
+ int gy = threadIdx.y;
+
+ int n = solver->Mesh.getDimensions().x();
+ int blockCount = n/blockDim.y +1;
+
+ if(blockIdx.x==0)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx = 0;
+ gy = threadIdx.y;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ __syncthreads();
+ }
+ }
+ else if(blockIdx.x==1)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==2)
+ {
+
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==3)
+ {
+ for(int gz = 0; gz < n;gz++)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+
+
+
+
+ else if(blockIdx.x==4)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx = 0;
+ gy = threadIdx.y;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==5)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=n-1;
+ gy=threadIdx.y;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy < n)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy+=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==6)
+ {
+
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=0;
+ gy=n-threadIdx.y-1;
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx++;
+ if(gx==n)
+ {
+ gx=0;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+ else if(blockIdx.x==7)
+ {
+ for(int gz = n-1; gz > -1;gz--)
+ {
+ gx=n-1;
+ gy=n-threadIdx.y-1;
+
+ for(int k = 0; k < n*blockCount + blockDim.y; k++)
+ {
+ if(threadIdx.y < k+1 && gy > -1)
+ {
+ solver->updateValue(gx,gy,gz);
+ gx--;
+ if(gx==-1)
+ {
+ gx=n-1;
+ gy-=blockDim.y;
+ }
+ }
+
+
+ __syncthreads();
+ }
+ }
+ }
+
+
+
+
+}
+
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 3,double, tnlHost, int >, double, int >* solver)
+{
+ int gx = threadIdx.x + blockDim.x*blockIdx.x;
+ int gy = blockDim.y*blockIdx.y + threadIdx.y;
+ int gz = blockDim.z*blockIdx.z + threadIdx.z;
+
+ if(solver->Mesh.getDimensions().x() > gx && solver->Mesh.getDimensions().y() > gy && solver->Mesh.getDimensions().z() > gz)
+ {
+ solver->initGrid(gx,gy,gz);
+ }
+
+
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1111( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(INT_MAX,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0000( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(-INT_MAX,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-INT_MAX,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1110( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1101( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1011( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0111( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0001( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0010( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0100( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1000( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// a = be/al;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1100( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-al;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1010( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare1001( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(cudaDofVector[index],cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//
+//
+//
+//
+//
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0011( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-al;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0101( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// Real al,be, a,b,c,s;
+// al=abs(cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,0>(index)]));
+//
+// be=abs(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]/
+// (cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)]-
+// cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)]));
+//
+// a = al-be;
+// b=1.0;
+// c=-be;
+// s= h/sqrt(a*a+b*b);
+//
+//
+// cudaDofVector2[index]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//
+//}
+//
+//template< typename MeshReal,
+// typename Device,
+// typename MeshIndex,
+// typename Real,
+// typename Index >
+//void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: setupSquare0110( Index i, Index j)
+//{
+// Index index = Mesh.getCellIndex(CoordinatesType(i,j));
+// cudaDofVector2[index]=fabsMin(cudaDofVector[index],cudaDofVector2[(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<0,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<0,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,1>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,1>(index)]);
+// cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]=fabsMin(cudaDofVector[Mesh.template getCellNextToCell<1,0>(index)],cudaDofVector2[Mesh.template getCellNextToCell<1,0>(index)]);
+//}
+#endif
+
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..dd5681e6d88b3fa666f210d0fe24f2423ac78b68
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping3D_impl.h
@@ -0,0 +1,307 @@
+/***************************************************************************
+ tnlFastSweeping2D_impl.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING3D_IMPL_H_
+#define TNLFASTSWEEPING3D_IMPL_H_
+
+#include "tnlFastSweeping.h"
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+String tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: getType()
+{
+ return String( "tnlFastSweeping< " ) +
+ MeshType::getType() + ", " +
+ ::getType< Real >() + ", " +
+ ::getType< Index >() + " >";
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: tnlFastSweeping()
+:Entity(Mesh),
+ dofVector(Mesh),
+ dofVector2(Mesh)
+{
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: init( const Config::ParameterContainer& parameters )
+{
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+
+ if( ! Mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ if( ! dofVector.load( initialCondition ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << meshFile << "." << endl;
+ return false;
+ }
+ dofVector2.load(initialCondition);
+
+ h = Mesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ Entity.refresh();
+
+ const String& exact_input = parameters.getParameter< String >( "exact-input" );
+
+ if(exact_input == "no")
+ exactInput=false;
+ else
+ exactInput=true;
+// cout << "bla "<<endl;
+ return initGrid();
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
+{
+ for(int i=0; i< Mesh.getDimensions().x()*Mesh.getDimensions().y()*Mesh.getDimensions().z();i++)
+ {
+
+ if (abs(dofVector[i]) < 1.8*h)
+ dofVector2[i]=dofVector[i];
+ else
+ dofVector2[i]=INT_MAX*sign(dofVector[i]);
+ }
+
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: run()
+{
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index k = 0; k < Mesh.getDimensions().z(); k++)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+
+
+
+
+
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = 0; j < Mesh.getDimensions().y(); j++)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = Mesh.getDimensions().x() - 1; i > -1; i--)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+ for(Index k = Mesh.getDimensions().z() -1; k > -1; k--)
+ {
+ for(Index i = 0; i < Mesh.getDimensions().x(); i++)
+ {
+ for(Index j = Mesh.getDimensions().y() - 1; j > -1; j--)
+ {
+ updateValue(i,j,k);
+ }
+ }
+ }
+
+/*---------------------------------------------------------------------------------------------------------------------------*/
+
+
+ dofVector2.save("u-00001.tnl");
+
+ cout << "bla 3"<<endl;
+ return true;
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+void tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: updateValue( Index i, Index j, Index k)
+{
+ this->Entity.setCoordinates(CoordinatesType(i,j,k));
+ this->Entity.refresh();
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ Real value = dofVector2[Entity.getIndex()];
+ Real a,b,c, tmp;
+
+ if( i == 0 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0>()];
+ else if( i == Mesh.getDimensions().x() - 1 )
+ a = dofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0 >()];
+ else
+ {
+ a = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< -1, 0, 0>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 1, 0, 0>()] );
+ }
+
+ if( j == 0 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0>()];
+ else if( j == Mesh.getDimensions().y() - 1 )
+ b = dofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0>()];
+ else
+ {
+ b = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, -1, 0>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 1, 0>()] );
+ }
+
+ if( k == 0 )
+ c = dofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1>()];
+ else if( k == Mesh.getDimensions().z() - 1 )
+ c = dofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1>()];
+ else
+ {
+ c = fabsMin( dofVector2[neighbourEntities.template getEntityIndex< 0, 0, -1>()],
+ dofVector2[neighbourEntities.template getEntityIndex< 0, 0, 1>()] );
+ }
+
+ Real hD = 3.0*h*h - 2.0*(a*a+b*b+c*c-a*b-a*c-b*c);
+
+ if(hD < 0.0)
+ tmp = fabsMin(a,fabsMin(b,c)) + sign(value)*h;
+ else
+ tmp = (1.0/3.0) * ( a + b + c + sign(value)*sqrt(hD) );
+
+
+ dofVector2[Entity.getIndex()] = fabsMin(value, tmp);
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+Real tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index > :: fabsMin( Real x, Real y)
+{
+ Real fx = fabs(x);
+ Real fy = fabs(y);
+
+ Real tmpMin = Min(fx,fy);
+
+ if(tmpMin == fx)
+ return x;
+ else
+ return y;
+
+}
+
+
+
+#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweepingSolver.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweepingSolver.h
new file mode 100644
index 0000000000000000000000000000000000000000..e5e2e1d7dfecceceb47f34bc46480daaff9975ab
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweepingSolver.h
@@ -0,0 +1,34 @@
+/*
+ * File: tnlFastSweepingSolver.h
+ * Author: oberhuber
+ *
+ * Created on July 12, 2016, 6:04 PM
+ */
+
+#pragma once
+
+#include <functions/tnlConstantFunction.h>
+#include <problems/tnlPDEProblem.h>
+
+template< typename Mesh,
+ typename Anisotropy = tnlConstanstFunction< Mesh > >
+class tnlFastSweepingSolver : public tnlPDEProblem< Mesh,
+ typename Mesh::RealType,
+ typename Mesh::DeviceType,
+ typename Mesh::IndexType >
+{
+ public:
+
+ typedef typename DifferentialOperator::RealType RealType;
+ typedef typename Mesh::DeviceType DeviceType;
+ typedef typename DifferentialOperator::IndexType IndexType;
+
+ typedef tnlMeshFunction< Mesh > MeshFunctionType;
+ typedef tnlPDEProblem< Mesh, TimeDependentProblem, RealType, DeviceType, IndexType > BaseType;
+
+ using typename BaseType::MeshType;
+ using typename BaseType::DofVectorType;
+ using typename BaseType::MeshDependentDataType;
+};
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping_CUDA.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping_CUDA.h
new file mode 100644
index 0000000000000000000000000000000000000000..310cdf3f3028eb68e71c31d821ffea1538615724
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/fast-sweeping/tnlFastSweeping_CUDA.h
@@ -0,0 +1,194 @@
+/***************************************************************************
+ tnlFastSweeping_CUDA.h - description
+ -------------------
+ begin : Oct 15 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+#ifndef TNLFASTSWEEPING_H_
+#define TNLFASTSWEEPING_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+
+#include <functions/tnlMeshFunction.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+#include <ctime>
+
+
+
+
+
+template< typename Mesh,
+ typename Real,
+ typename Index >
+class tnlFastSweeping
+{};
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ tnlFastSweeping();
+
+ __host__ static String getType();
+ __host__ bool init( const Config::ParameterContainer& parameters );
+ __host__ bool run();
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid();
+ __device__ void updateValue(const Index i, const Index j);
+ __device__ void updateValue(const Index i, const Index j, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ int counter;
+ __device__ void setupSquare1000(Index i, Index j);
+ __device__ void setupSquare1100(Index i, Index j);
+ __device__ void setupSquare1010(Index i, Index j);
+ __device__ void setupSquare1001(Index i, Index j);
+ __device__ void setupSquare1110(Index i, Index j);
+ __device__ void setupSquare1101(Index i, Index j);
+ __device__ void setupSquare1011(Index i, Index j);
+ __device__ void setupSquare1111(Index i, Index j);
+ __device__ void setupSquare0000(Index i, Index j);
+ __device__ void setupSquare0100(Index i, Index j);
+ __device__ void setupSquare0010(Index i, Index j);
+ __device__ void setupSquare0001(Index i, Index j);
+ __device__ void setupSquare0110(Index i, Index j);
+ __device__ void setupSquare0101(Index i, Index j);
+ __device__ void setupSquare0011(Index i, Index j);
+ __device__ void setupSquare0111(Index i, Index j);
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data;
+
+
+ RealType h;
+
+
+};
+
+
+
+
+
+
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >
+{
+
+public:
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+
+ __host__ static String getType();
+ __host__ bool init( const Config::ParameterContainer& parameters );
+ __host__ bool run();
+
+#ifdef HAVE_CUDA
+ __device__ bool initGrid(int i, int j, int k);
+ __device__ void updateValue(const Index i, const Index j, const Index k);
+ __device__ void updateValue(const Index i, const Index j, const Index k, double** sharedMem, const int k3);
+ __device__ Real fabsMin(const Real x, const Real y);
+
+ tnlFastSweeping< tnlGrid< 3,MeshReal, Device, MeshIndex >, Real, Index >* cudaSolver;
+ double* cudaDofVector;
+ double* cudaDofVector2;
+ int counter;
+#endif
+
+ MeshType Mesh;
+
+protected:
+
+
+
+ bool exactInput;
+
+ tnlMeshFunction<MeshType> dofVector;
+ DofVectorType data;
+
+ RealType h;
+
+
+};
+
+
+
+
+
+
+
+#ifdef HAVE_CUDA
+//template<int sweep_t>
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver, int sweep, int i);
+__global__ void runCUDA(tnlFastSweeping< tnlGrid< 3,double, tnlHost, int >, double, int >* solver, int sweep, int i);
+
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, tnlHost, int >, double, int >* solver);
+__global__ void initCUDA(tnlFastSweeping< tnlGrid< 3,double, tnlHost, int >, double, int >* solver);
+#endif
+
+/*various implementtions.... choose one*/
+//#include "tnlFastSweeping2D_CUDA_impl.h"
+//#include "tnlFastSweeping2D_CUDA_v2_impl.h"
+//#include "tnlFastSweeping2D_CUDA_v3_impl.h"
+#include "tnlFastSweeping2D_CUDA_v4_impl.h"
+//#include "tnlFastSweeping2D_CUDA_v5_impl.h"
+
+
+#include "tnlFastSweeping3D_CUDA_impl.h"
+
+#endif /* TNLFASTSWEEPING_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..bd5ed12005cdec8df667d4feae4a4e0f65dbf46e
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/CMakeLists.txt
@@ -0,0 +1,23 @@
+set( tnl_hamilton_jacobi_parallel_map_SOURCES
+# MainBuildConfig.h
+# tnlParallelMapSolver2D_impl.h
+# tnlParallelMapSolver.h
+# parallelMapConfig.h
+# main.cu
+ main.cpp)
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(hamilton-jacobi-parallel-map${debugExt} main.cu)
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(hamilton-jacobi-parallel-map${debugExt} main.cpp)
+ENDIF( BUILD_CUDA )
+target_link_libraries (hamilton-jacobi-parallel-map${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS hamilton-jacobi-parallel-map${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+#INSTALL( FILES ${tnl_hamilton_jacobi_parallel_map_SOURCES}
+# DESTINATION share/tnl-${tnlVersion}/examples/hamilton-jacobi-parallel-map )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/MainBuildConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/gnuplot.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/gnuplot.txt
new file mode 100644
index 0000000000000000000000000000000000000000..d4ae61983910a676269a23e3d992f5f46ea83a8f
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/gnuplot.txt
@@ -0,0 +1,32 @@
+tomas@tomas-linux:~/Desktop/VU_CPU_MAPA/work_dir$ gnuplot
+
+ G N U P L O T
+ Version 4.6 patchlevel 4 last modified 2013-10-02
+ Build System: Linux x86_64
+
+ Copyright (C) 1986-1993, 1998, 2004, 2007-2013
+ Thomas Williams, Colin Kelley and many others
+
+ gnuplot home: http://www.gnuplot.info
+ faq, bugs, etc: type "help FAQ"
+ immediate help: type "help" (plot window: hit 'h')
+
+Terminal type set to 'wxt'
+gnuplot> set cntrparam levels 15
+gnuplot> set cntrparam bspline
+gnuplot> set contour
+gnuplot> splot 'u-00001.gplt'
+
+gnuplot> unset surface
+gnuplot> splot 'u-00001.gplt'
+
+gnuplot> set table "test.gplt"
+gnuplot> splot 'u-00001.gplt'
+gnuplot> unset table
+
+gnuplot> set table "test2.gplt"
+gnuplot> plot 'test.gplt' index 10
+gnuplot> unset table
+
+gnuplot> plot 'test2.gplt'
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..b13498e17330fae7bb00a0bdc2abcc7a19f8e7a8
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cpp
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cu b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cu
new file mode 100644
index 0000000000000000000000000000000000000000..7101976712e153d73c5f0979b211164a36ec648d
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.cu
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cu - description
+ -------------------
+ begin : Mar 30 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.h
new file mode 100644
index 0000000000000000000000000000000000000000..55a0942f81721c1e90e670aaae18941471f6e13d
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/main.h
@@ -0,0 +1,98 @@
+/***************************************************************************
+ main.h - description
+ -------------------
+ begin : Mar 22 , 2016
+ copyright : (C) 2016 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "tnlParallelMapSolver.h"
+#include "parallelMapConfig.h"
+#include "MainBuildConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/parallelGodunovMap.h>
+#include <mesh/tnlGrid.h>
+#include <core/tnlDevice.h>
+#include <time.h>
+#include <ctime>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ time_t start;
+ time_t stop;
+ time(&start);
+ std::clock_t start2= std::clock();
+ Config::ParameterContainer parameters;
+ tnlConfigDescription configDescription;
+ parallelMapConfig< BuildConfig >::configSetup( configDescription );
+
+ if( ! parseCommandLine( argc, argv, configDescription, parameters ) )
+ return false;
+
+
+ tnlDeviceEnum device;
+ device = tnlHostDevice;
+
+ const int& dim = parameters.getParameter< int >( "dim" );
+
+ if(dim == 2)
+ {
+
+ typedef parallelGodunovMapScheme< tnlGrid<2,double,tnlHost, int>, double, int > SchemeTypeHost;
+/*#ifdef HAVE_CUDA
+ typedef parallelGodunovMapScheme< tnlGrid<2,double,tnlCuda, int>, double, int > SchemeTypeDevice;
+#endif
+#ifndef HAVE_CUDA*/
+ typedef parallelGodunovMapScheme< tnlGrid<2,double,tnlHost, int>, double, int > SchemeTypeDevice;
+/*#endif*/
+
+ if(device==tnlHostDevice)
+ {
+ typedef tnlHost Device;
+
+
+ tnlParallelMapSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ else if(device==tnlCudaDevice )
+ {
+ typedef tnlCuda Device;
+//typedef parallelGodunovMapScheme< tnlGrid<2,double,Device, int>, double, int > SchemeType;
+
+ tnlParallelMapSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ }
+
+
+ time(&stop);
+ cout << endl;
+ cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) << endl;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/mapa_png.png b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/mapa_png.png
new file mode 100644
index 0000000000000000000000000000000000000000..668b6fe24b17b2fec486db28505b41e3beb2091a
Binary files /dev/null and b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/mapa_png.png differ
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/no-Makefile b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/no-Makefile
new file mode 100644
index 0000000000000000000000000000000000000000..bfdc1ef236ca02ecfe6bc88f81d872e9524ec621
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/no-Makefile
@@ -0,0 +1,41 @@
+TNL_VERSION=0.1
+TNL_INSTALL_DIR=${HOME}/local/lib
+TNL_INCLUDE_DIR=${HOME}/local/include/tnl-${TNL_VERSION}
+
+TARGET = hamiltonJacobiParallelSolver
+#CONFIG_FILE = $(TARGET).cfg.desc
+INSTALL_DIR = ${HOME}/local
+CXX = g++
+CUDA_CXX = nvcc
+OMP_FLAGS = -DHAVE_OPENMP -fopenmp
+CXX_FLAGS = -std=gnu++0x -I$(TNL_INCLUDE_DIR) -O3 $(OMP_FLAGS) -DDEBUG
+LD_FLAGS = -L$(TNL_INSTALL_DIR) -ltnl-0.1 -lgomp
+
+SOURCES = main.cpp
+HEADERS =
+OBJECTS = main.o
+DIST = $(SOURCES) Makefile
+
+all: $(TARGET)
+clean:
+ rm -f $(OBJECTS)
+ rm -f $(TARGET)-conf.h
+
+dist: $(DIST)
+ tar zcvf $(TARGET).tgz $(DIST)
+
+install: $(TARGET)
+ cp $(TARGET) $(INSTALL_DIR)/bin
+ cp $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+uninstall: $(TARGET)
+ rm -f $(INSTALL_DIR)/bin/$(TARGET)
+ rm -f $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+$(TARGET): $(OBJECTS)
+ $(CXX) -o $(TARGET) $(OBJECTS) $(LD_FLAGS)
+
+%.o: %.cpp $(HEADERS)
+ $(CXX) -c -o $@ $(CXX_FLAGS) $<
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/parallelMapConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/parallelMapConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..c07ee95aa04bb8c7a1f3cc376aabd859ff7bc5be
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/parallelMapConfig.h
@@ -0,0 +1,47 @@
+/***************************************************************************
+ parallelMapConfig.h - description
+ -------------------
+ begin : Mar 22 , 2016
+ copyright : (C) 2016 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef HAMILTONJACOBIPARALLELMAPPROBLEMCONFIG_H_
+#define HAMILTONJACOBIPARALLELMAPPROBLEMCONFIG_H_
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class parallelMapConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Parallel Eikonal solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "hamilton-jacobi-parallel" );
+ config.addEntry < String > ( "scheme", "This defines scheme used for discretization.", "godunov" );
+ config.addEntryEnum( "godunov" );
+ config.addEntryEnum( "upwind" );
+ config.addRequiredEntry < String > ( "initial-condition", "Initial condition for solver");
+ config.addRequiredEntry < String > ( "map", "Gradient map for solver");
+ config.addEntry < String > ( "mesh", "Name of mesh.", "mesh.tnl" );
+ config.addEntry < double > ( "epsilon", "This defines epsilon for smoothening of sign().", 0.0 );
+ config.addEntry < double > ( "delta", " Allowed difference on subgrid boundaries", 0.0 );
+ config.addRequiredEntry < double > ( "stop-time", " Final time for solver");
+ config.addRequiredEntry < double > ( "initial-tau", " initial tau for solver" );
+ config.addEntry < double > ( "cfl-condition", " CFL condition", 0.0 );
+ config.addEntry < int > ( "subgrid-size", "Subgrid size.", 16 );
+ config.addRequiredEntry < int > ( "dim", "Dimension of problem.");
+ }
+};
+
+#endif /* HAMILTONJACOBIPARALLELMAPPROBLEMCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/run b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/run
new file mode 100755
index 0000000000000000000000000000000000000000..48441996274633f8d391d9b32978b05b2e4fa263
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/run
@@ -0,0 +1,43 @@
+#!/bin/bash
+
+dimensions=2
+
+size=2
+
+time=50
+
+rm -r work_dir
+mkdir work_dir
+cp mapa_png.png work_dir/mapa_png.png
+cd work_dir
+
+tnl-image-converter --image-format png\
+ --input-images mapa_png.png
+
+
+tnl-init --test-function sdf-para \
+ --x-centre 0.5 \
+ --y-centre 1.0 \
+ --offset 0.05 \
+ --output-file init.tnl \
+ --final-time 0.0 \
+ --snapshot-period 0.1
+
+hamilton-jacobi-parallel-map-dbg --initial-condition init.tnl \
+ --map mapa_png.tnl \
+ --cfl-condition 50 \
+ --mesh mesh.tnl \
+ --initial-tau 1.0e-3 \
+ --epsilon 4.0 \
+ --delta 0.0 \
+ --stop-time $time \
+ --scheme godunov \
+ --subgrid-size 8 \
+ --dim $dimensions
+
+
+#cp ../template.dat1 template.dat1
+#cp ../template.dat2 template.dat2
+#cp ../gplt2eps.py gplt2eps.py
+cd ..
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py
new file mode 100755
index 0000000000000000000000000000000000000000..f8cde3768e9b76156507e133f8bc3ecaa526fc71
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py
@@ -0,0 +1,141 @@
+#!/usr/bin/env python
+
+import sys, string, math
+
+arguments = sys. argv[1:]
+format = "txt"
+output_file_name = "eoc-table.txt"
+input_files = []
+verbose = 1
+size = 1.0
+
+i = 0
+while i < len( arguments ):
+ if arguments[ i ] == "--format":
+ format = arguments[ i + 1 ]
+ i = i + 2
+ continue
+ if arguments[ i ] == "--output-file":
+ output_file_name = arguments[ i + 1 ]
+ i = i + 2
+ continue
+ if arguments[ i ] == "--verbose":
+ verbose = float( arguments[ i + 1 ] )
+ i = i +2
+ continue
+ if arguments[ i ] == "--size":
+ size = float( arguments[ i + 1 ] )
+ i = i +2
+ continue
+ input_files. append( arguments[ i ] )
+ i = i + 1
+
+if not verbose == 0:
+ print "Writing to " + output_file_name + " in " + format + "."
+
+h_list = []
+l1_norm_list = []
+l2_norm_list = []
+max_norm_list = []
+items = 0
+
+for file_name in input_files:
+ if not verbose == 0:
+ print "Processing file " + file_name
+ file = open( file_name, "r" )
+
+ l1_max = 0.0
+ l_max_max = 0.0
+ file.readline();
+ file.readline();
+ for line in file. readlines():
+ data = string. split( line )
+ h_list. append( size/(float(file_name[0:len(file_name)-5] ) - 1.0) )
+ l1_norm_list. append( float( data[ 1 ] ) )
+ l2_norm_list. append( float( data[ 2 ] ) )
+ max_norm_list. append( float( data[ 3 ] ) )
+ items = items + 1
+ if not verbose == 0:
+ print line
+ file. close()
+
+h_width = 12
+err_width = 15
+file = open( output_file_name, "w" )
+if format == "latex":
+ file. write( "\\begin{tabular}{|r|l|l|l|l|l|l|}\\hline\n" )
+ file. write( "\\raisebox{-1ex}[0ex]{$h$}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_1\\left(\\omega_h;\\left[0,T\\right]\\right)}^{h,\\tau}$}}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_2\\left(\\omega_h;\left[0,T\\right]\\right)}^{h,\\tau}$}}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_\\infty\\left(\\omega_h;\\left[0,T\\right]\\right)}^{h,\\tau}$}}\\\\ \\cline{2-7} \n" )
+ file. write( " " + string. rjust( " ", h_width ) + "&" +
+ string. rjust( "Error", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) + "&" +
+ string. rjust( "Error", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) + "&" +
+ string. rjust( "Error.", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) +
+ "\\\\ \\hline \\hline \n")
+if format == "txt":
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ file. write( "| h | L1 Err. | L1 EOC. | L2 Err. | L2 EOC | MAX Err. | MAX EOC |\n" )
+ file. write( "+==============+================+================+================+================+================+================+\n" )
+
+
+i = 0
+while i < items:
+ if i == 0:
+ if format == "latex":
+ file. write( " " + string. ljust( str( h_list[ i ] ), h_width ) + "&" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "&"+
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "&" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "\\\\\n" )
+ if format == "txt":
+ file. write( "| " + string. ljust( str( h_list[ i ] ), h_width ) + " |" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |\n" )
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ i = i + 1;
+ continue
+ if h_list[ i ] == h_list[ i - 1 ]:
+ print "Unable to count eoc since h[ " + \
+ str( i ) + " ] = h[ " + str( i - 1 ) + \
+ " ] = " + str( h_list[ i ] ) + ". \n"
+ file. write( " eoc error: h[ " + \
+ str( i ) + " ] = h[ " + str( i - 1 ) + \
+ " ] = " + str( h_list[ i ] ) + ". \n" )
+ else:
+ h_ratio = math. log( h_list[ i ] / h_list[ i - 1 ] )
+ l1_ratio = math. log( l1_norm_list[ i ] / l1_norm_list[ i - 1 ] )
+ l2_ratio = math. log( l2_norm_list[ i ] / l2_norm_list[ i - 1 ] )
+ max_ratio = math. log( max_norm_list[ i ] / max_norm_list[ i - 1 ] )
+ if format == "latex":
+ file. write( " " + string. ljust( str( h_list[ i ] ), h_width ) + "&" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( l1_ratio / h_ratio ) + "}", err_width ) + "&" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( l2_ratio / h_ratio ) + "}", err_width ) + "&" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( max_ratio / h_ratio ) + "}", err_width ) + "\\\\\n" )
+ if format == "txt":
+ file. write( "| " + string. ljust( str( h_list[ i ] ), h_width ) + " |" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( l1_ratio / h_ratio ) + "**", err_width ) + " |" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( l2_ratio / h_ratio ) + "**", err_width ) + " |" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( max_ratio / h_ratio ) + "**", err_width ) + " |\n" )
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ i = i + 1
+
+if format == "latex":
+ file. write( "\\hline \n" )
+ file. write( "\\end{tabular} \n" )
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h
new file mode 100644
index 0000000000000000000000000000000000000000..bf0a388166c89946727b95687eca1a74764739c3
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h
@@ -0,0 +1,218 @@
+/***************************************************************************
+ tnlParallelMapSolver.h - description
+ -------------------
+ begin : Mar 22 , 2016
+ copyright : (C) 2016 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLPARALLELMAPSOLVER_H_
+#define TNLPARALLELMAPSOLVER_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlMeshFunction.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+
+
+#include <ctime>
+
+#ifdef HAVE_CUDA
+#include <cuda.h>
+#include <core/tnlCuda.h>
+#endif
+
+
+template< int Dimension,
+ typename SchemeHost,
+ typename SchemeDevice,
+ typename Device,
+ typename RealType = double,
+ typename IndexType = int >
+class tnlParallelMapSolver
+{};
+
+template<typename SchemeHost, typename SchemeDevice, typename Device>
+class tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >
+{
+public:
+
+ typedef SchemeDevice SchemeTypeDevice;
+ typedef SchemeHost SchemeTypeHost;
+ typedef Device DeviceType;
+ typedef tnlVector< double, tnlHost, int > VectorType;
+ typedef tnlVector< int, tnlHost, int > IntVectorType;
+ typedef tnlGrid< 2, double, tnlHost, int > MeshType;
+#ifdef HAVE_CUDA
+ typedef tnlVector< double, tnlHost, int > VectorTypeCUDA;
+ typedef tnlVector< int, tnlHost, int > IntVectorTypeCUDA;
+ typedef tnlGrid< 2, double, tnlHost, int > MeshTypeCUDA;
+#endif
+ tnlParallelMapSolver();
+ bool init( const Config::ParameterContainer& parameters );
+ void run();
+
+ void test();
+
+/*private:*/
+
+
+ void synchronize();
+
+ int getOwner( int i) const;
+
+ int getSubgridValue( int i ) const;
+
+ void setSubgridValue( int i, int value );
+
+ int getBoundaryCondition( int i ) const;
+
+ void setBoundaryCondition( int i, int value );
+
+ void stretchGrid();
+
+ void contractGrid();
+
+ VectorType getSubgrid( const int i ) const;
+
+ void insertSubgrid( VectorType u, const int i );
+
+ VectorType runSubgrid( int boundaryCondition, VectorType u, int subGridID,VectorType map);
+
+
+ tnlMeshFunction<MeshType> u0;
+ VectorType work_u, map_stretched, map;
+ IntVectorType subgridValues, boundaryConditions, unusedCell, calculationsCount;
+ MeshType mesh, subMesh;
+
+// tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage > Entity;
+
+ SchemeHost schemeHost;
+ SchemeDevice schemeDevice;
+ double delta, tau0, stopTime,cflCondition;
+ int gridRows, gridCols, gridLevels, currentStep, n;
+
+ std::clock_t start;
+ double time_diff;
+
+
+ tnlDeviceEnum device;
+
+ tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* getSelf()
+ {
+ return this;
+ };
+
+#ifdef HAVE_CUDA
+
+ tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver;
+
+ double* work_u_cuda;
+ double* map_stretched_cuda;
+
+ int* subgridValues_cuda;
+ int* boundaryConditions_cuda;
+ int* unusedCell_cuda;
+ int* calculationsCount_cuda;
+ double* tmpw;
+ double* tmp_map;
+
+
+ int* runcuda;
+ int run_host;
+
+
+ __device__ void getSubgridCUDA2D( const int i, tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void updateSubgridCUDA2D( const int i, tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void insertSubgridCUDA2D( double u, const int i );
+
+ __device__ void runSubgridCUDA2D( int boundaryCondition, double* u, int subGridID);
+
+ __device__ int getOwnerCUDA2D( int i) const;
+
+ __device__ int getSubgridValueCUDA2D( int i ) const;
+
+ __device__ void setSubgridValueCUDA2D( int i, int value );
+
+ __device__ int getBoundaryConditionCUDA2D( int i ) const;
+
+ __device__ void setBoundaryConditionCUDA2D( int i, int value );
+
+#endif
+
+};
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+#ifdef HAVE_CUDA
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void runCUDA2D(tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initRunCUDA2D(tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initCUDA2D( tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr, int * ptr2, int* ptr3, double* tmp_map_ptr);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronizeCUDA2D(tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronize2CUDA2D(tnlParallelMapSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+
+
+__device__
+double fabsMin( double x, double y)
+{
+ double fx = abs(x);
+
+ if(Min(fx,abs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__device__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+#endif
+
+#include "tnlParallelMapSolver2D_impl.h"
+#endif /* TNLPARALLELMAPSOLVER_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..1cb730e42d0b0b72782e61e011b6081d2eddc6ef
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h
@@ -0,0 +1,1315 @@
+/***************************************************************************
+ tnlParallelMapSolver2D_impl.h - description
+ -------------------
+ begin : Mar 22 , 2016
+ copyright : (C) 2016 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLPARALLELMAPSOLVER2D_IMPL_H_
+#define TNLPARALLELMAPSOLVER2D_IMPL_H_
+
+
+#include "tnlParallelMapSolver.h"
+#include <core/mfilename.h>
+
+
+
+
+#define MAP_SOLVER_MAX_VALUE 3
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::tnlParallelMapSolver()
+{
+ this->device = tnlHostDevice; /////////////// tnlCuda Device --- vypocet na GPU, tnlHostDevice --- vypocet na CPU
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ run_host = 1;
+ }
+#endif
+
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::test()
+{
+/*
+ for(int i =0; i < this->subgridValues.getSize(); i++ )
+ {
+ insertSubgrid(getSubgrid(i), i);
+ }
+*/
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+
+bool tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::init( const Config::ParameterContainer& parameters )
+{
+ cout << "Initializating solver..." << endl;
+ const String& meshLocation = parameters.getParameter <String>("mesh");
+ this->mesh.load( meshLocation );
+
+ this->n = parameters.getParameter <int>("subgrid-size");
+ cout << "Setting N to " << this->n << endl;
+
+ this->subMesh.setDimensions( this->n, this->n );
+ this->subMesh.setDomain( Containers::StaticVector<2,double>(0.0, 0.0),
+ Containers::StaticVector<2,double>(mesh.template getSpaceStepsProducts< 1, 0 >()*(double)(this->n), mesh.template getSpaceStepsProducts< 0, 1 >()*(double)(this->n)) );
+
+ this->subMesh.save("submesh.tnl");
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ this->u0.load( initialCondition );
+
+ /* LOAD MAP */
+ const String& mapFile = parameters.getParameter <String>("map");
+ if(! this->map.load( mapFile ))
+ cout << "Failed to load map file : " << mapFile << endl;
+
+
+ this->delta = parameters.getParameter <double>("delta");
+ this->delta *= mesh.template getSpaceStepsProducts< 1, 0 >()*mesh.template getSpaceStepsProducts< 0, 1 >();
+
+ cout << "Setting delta to " << this->delta << endl;
+
+ this->tau0 = parameters.getParameter <double>("initial-tau");
+ cout << "Setting initial tau to " << this->tau0 << endl;
+ this->stopTime = parameters.getParameter <double>("stop-time");
+
+ this->cflCondition = parameters.getParameter <double>("cfl-condition");
+ this -> cflCondition *= sqrt(mesh.template getSpaceStepsProducts< 1, 0 >()*mesh.template getSpaceStepsProducts< 0, 1 >());
+ cout << "Setting CFL to " << this->cflCondition << endl;
+
+ stretchGrid();
+ this->stopTime /= (double)(this->gridCols);
+ this->stopTime *= (1.0+1.0/((double)(this->n) - 2.0));
+ cout << "Setting stopping time to " << this->stopTime << endl;
+
+ cout << "Initializating scheme..." << endl;
+ if(!this->schemeHost.init(parameters))
+ {
+ cerr << "SchemeHost failed to initialize." << endl;
+ return false;
+ }
+ cout << "Scheme initialized." << endl;
+
+ test();
+
+ VectorType* tmp = new VectorType[subgridValues.getSize()];
+ bool containsCurve = false;
+
+#ifdef HAVE_CUDA
+
+ if(this->device == tnlCudaDevice)
+ {
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >));
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >), cudaMemcpyHostToDevice);
+
+ double** tmpdev = NULL;
+ cudaMalloc(&tmpdev, sizeof(double*));
+ cudaMalloc(&(this->tmpw), this->work_u.getSize()*sizeof(double));
+ cudaMalloc(&(this->tmp_map), this->map_stretched.getSize()*sizeof(double));
+ cudaMalloc(&(this->runcuda), sizeof(int));
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ int* tmpUC;
+ cudaMalloc(&(tmpUC), this->work_u.getSize()*sizeof(int));
+ cudaMemcpy(tmpUC, this->unusedCell.getData(), this->unusedCell.getSize()*sizeof(int), cudaMemcpyHostToDevice);
+
+ initCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<1,1>>>(this->cudaSolver, (this->tmpw), (this->runcuda),tmpUC, tmp_map);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ double* tmpu = NULL;
+ cudaMemcpy(&tmpu, tmpdev,sizeof(double*), cudaMemcpyDeviceToHost);
+ cudaMemcpy((this->tmpw), this->work_u.getData(), this->work_u.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+ cudaMemcpy((this->tmp_map), this->map_stretched.getData(), this->map_stretched.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ }
+#endif
+
+ if(this->device == tnlHostDevice)
+ {
+ VectorType tmp_map;
+ tmp_map.setSize(this->n * this->n);
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+
+ if(! tmp[i].setSize(this->n * this->n))
+ cout << "Could not allocate tmp["<< i <<"] array." << endl;
+ tmp[i] = getSubgrid(i);
+ containsCurve = false;
+
+ for(int j = 0; j < tmp[i].getSize(); j++)
+ {
+ if(tmp[i][0]*tmp[i][j] <= 0.0)
+ {
+ containsCurve = true;
+ j=tmp[i].getSize();
+ }
+
+ }
+ if(containsCurve)
+ {
+ for( int j = 0; j < tmp_map.getSize(); j++)
+ {
+ tmp_map[j] = this->map_stretched[ (i / this->gridCols) * this->n*this->n*this->gridCols
+ + (i % this->gridCols) * this->n
+ + (j/this->n) * this->n*this->gridCols
+ + (j % this->n) ];
+ }
+ //cout << "Computing initial SDF on subgrid " << i << "." << endl;
+ tmp[i] = runSubgrid(0, tmp[i],i,tmp_map);
+ insertSubgrid(tmp[i], i);
+ setSubgridValue(i, 4);
+ //cout << "Computed initial SDF on subgrid " << i << "." << endl;
+ }
+ containsCurve = false;
+
+ }
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initRunCUDA2D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ }
+#endif
+
+
+ this->currentStep = 1;
+ if(this->device == tnlHostDevice)
+ synchronize();
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+
+ synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ }
+
+#endif
+ cout << "Solver initialized." << endl;
+
+ return true;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::run()
+{
+ if(this->device == tnlHostDevice)
+ {
+ while ((this->boundaryConditions.max() > 0 )/* || !end*/)
+ {
+
+#ifdef HAVE_OPENMP
+#pragma omp parallel for num_threads(4) schedule(dynamic)
+#endif
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ if(getSubgridValue(i) != INT_MAX)
+ {
+ VectorType tmp, tmp_map;
+ tmp.setSize(this->n * this->n);
+ tmp_map.setSize(this->n * this->n);
+ for( int j = 0; j < tmp_map.getSize(); j++)
+ {
+ tmp_map[j] = this->map_stretched[ (i / this->gridCols) * this->n*this->n*this->gridCols
+ + (i % this->gridCols) * this->n
+ + (j/this->n) * this->n*this->gridCols
+ + (j % this->n) ];
+ }
+
+ if(getSubgridValue(i) == currentStep+4)
+ {
+
+ if(getBoundaryCondition(i) & 1)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 2)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(2, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 4)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(4, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 8)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(8, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ }
+ else
+ {
+
+ if(getBoundaryCondition(i) == 1)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) == 2)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(2, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) == 4)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(4, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) == 8)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(8, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ }
+
+ if(getBoundaryCondition(i) & 3)
+ {
+ //cout << "3 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(3, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ }
+ if(getBoundaryCondition(i) & 5)
+ {
+ //cout << "5 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(5, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ }
+ if(getBoundaryCondition(i) & 10)
+ {
+ //cout << "10 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(10, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ }
+ if(getBoundaryCondition(i) & 12)
+ {
+ //cout << "12 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(12, tmp ,i,tmp_map);
+ insertSubgrid( tmp, i);
+ }
+
+
+ setBoundaryCondition(i, 0);
+
+ setSubgridValue(i, getSubgridValue(i)-1);
+
+ }
+ }
+ synchronize();
+ }
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ bool end_cuda = false;
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ bool* tmpb;
+ cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+
+ int i = 1;
+ time_diff = 0.0;
+ while (run_host || !end_cuda)
+ {
+ cout << "Computing at step "<< i++ << endl;
+ if(run_host != 0 )
+ end_cuda = true;
+ else
+ end_cuda = false;
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ start = std::clock();
+ runCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+ //start = std::clock();
+ synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+ cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
+ }
+ cout << "Solving time was: " << time_diff << endl;
+
+ cudaMemcpy(this->work_u.getData()/* test*/, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+
+ cudaDeviceSynchronize();
+ }
+#endif
+ contractGrid();
+ this->u0.save("u-00001.tnl");
+ cout << "Maximum number of calculations on one subgrid was " << this->calculationsCount.absMax() << endl;
+ cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) << endl;
+ cout << "Solver finished" << endl;
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ cudaFree(this->runcuda);
+ cudaFree(this->tmpw);
+ cudaFree(this->tmp_map);
+ cudaFree(this->cudaSolver);
+ }
+#endif
+
+}
+
+//north - 1, east - 2, west - 4, south - 8
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::synchronize() //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+ cout << "Synchronizig..." << endl;
+ int tmp1, tmp2;
+ int grid1, grid2;
+
+// if(this->currentStep & 1)
+// {
+ for(int j = 0; j < this->gridRows - 1; j++)
+ {
+ for (int i = 0; i < this->gridCols*this->n; i++)
+ {
+ tmp1 = this->gridCols*this->n*((this->n-1)+j*this->n) + i;
+ tmp2 = this->gridCols*this->n*((this->n)+j*this->n) + i;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j" << i << "," << j << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 8) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+8);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 1) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+1);
+ }
+ }
+ }
+
+// }
+// else
+// {
+ for(int i = 1; i < this->gridCols; i++)
+ {
+ for (int j = 0; j < this->gridRows*this->n; j++)
+ {
+ tmp1 = this->gridCols*this->n*j + i*this->n - 1;
+ tmp2 = this->gridCols*this->n*j + i*this->n ;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j" << i << "," << j << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 4) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+4);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 2) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+2);
+ }
+ }
+ }
+// }
+
+
+ this->currentStep++;
+ int stepValue = this->currentStep + 4;
+ for (int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ if( getSubgridValue(i) == -INT_MAX )
+ setSubgridValue(i, stepValue);
+ }
+
+ cout << "Grid synchronized at step " << (this->currentStep - 1 ) << endl;
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getOwner(int i) const
+{
+
+ return (i / (this->gridCols*this->n*this->n))*this->gridCols + (i % (this->gridCols*this->n))/this->n;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValue( int i ) const
+{
+ return this->subgridValues[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValue(int i, int value)
+{
+ this->subgridValues[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryCondition( int i ) const
+{
+ return this->boundaryConditions[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryCondition(int i, int value)
+{
+ this->boundaryConditions[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::stretchGrid()
+{
+ cout << "Stretching grid..." << endl;
+
+
+ this->gridCols = ceil( ((double)(this->mesh.getDimensions().x()-1)) / ((double)(this->n-1)) );
+ this->gridRows = ceil( ((double)(this->mesh.getDimensions().y()-1)) / ((double)(this->n-1)) );
+
+
+ cout << "Setting gridCols to " << this->gridCols << "." << endl;
+ cout << "Setting gridRows to " << this->gridRows << "." << endl;
+
+ this->subgridValues.setSize(this->gridCols*this->gridRows);
+ this->subgridValues.setValue(0);
+ this->boundaryConditions.setSize(this->gridCols*this->gridRows);
+ this->boundaryConditions.setValue(0);
+ this->calculationsCount.setSize(this->gridCols*this->gridRows);
+ this->calculationsCount.setValue(0);
+
+ for(int i = 0; i < this->subgridValues.getSize(); i++ )
+ {
+ this->subgridValues[i] = INT_MAX;
+ this->boundaryConditions[i] = 0;
+ }
+
+ int stretchedSize = this->n*this->n*this->gridCols*this->gridRows;
+
+ if(!this->work_u.setSize(stretchedSize))
+ cerr << "Could not allocate memory for stretched grid." << endl;
+ if(!this->map_stretched.setSize(stretchedSize))
+ cerr << "Could not allocate memory for stretched map." << endl;
+ if(!this->unusedCell.setSize(stretchedSize))
+ cerr << "Could not allocate memory for supporting stretched grid." << endl;
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+ for(int i = 0; i < stretchedSize; i++)
+ {
+ this->unusedCell[i] = 1;
+ int diff =(this->n*this->gridCols) - idealStretch ;
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if(i%(this->n*this->gridCols) - idealStretch >= 0)
+ {
+ k+= i%(this->n*this->gridCols) - idealStretch +1 ;
+ }
+
+ if(i/(this->n*this->gridCols) - idealStretch + 1 > 0)
+ {
+ k+= (i/(this->n*this->gridCols) - idealStretch +1 )* this->mesh.getDimensions().x() ;
+ }
+
+
+ if(fabs(this->u0[i-k]) < mesh.template getSpaceStepsProducts< 1, 0 >()+mesh.template getSpaceStepsProducts< 0, 1 >() )
+ this->work_u[i] = this->u0[i-k];
+ else
+ this->work_u[i] = sign(this->u0[i-k])*MAP_SOLVER_MAX_VALUE;
+
+ this->map_stretched[i] = this->map[i-k];
+ }
+
+
+ cout << "Grid stretched." << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
+{
+ cout << "Contracting grid..." << endl;
+ int stretchedSize = this->n*this->n*this->gridCols*this->gridRows;
+
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+ for(int i = 0; i < stretchedSize; i++)
+ {
+ int diff =(this->n*this->gridCols) - idealStretch ;
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if((i%(this->n*this->gridCols) - idealStretch < 0) && (i/(this->n*this->gridCols) - idealStretch + 1 <= 0))
+ {
+ this->u0[i-k] = this->work_u[i];
+ }
+
+ }
+
+ cout << "Grid contracted" << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgrid( const int i ) const
+{
+ VectorType u;
+ u.setSize(this->n*this->n);
+
+ for( int j = 0; j < u.getSize(); j++)
+ {
+ u[j] = this->work_u[ (i / this->gridCols) * this->n*this->n*this->gridCols
+ + (i % this->gridCols) * this->n
+ + (j/this->n) * this->n*this->gridCols
+ + (j % this->n) ];
+ }
+ return u;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::insertSubgrid( VectorType u, const int i )
+{
+
+ for( int j = 0; j < this->n*this->n; j++)
+ {
+ int index = (i / this->gridCols)*this->n*this->n*this->gridCols + (i % this->gridCols)*this->n + (j/this->n)*this->n*this->gridCols + (j % this->n);
+ if( (fabs(this->work_u[index]) > fabs(u[j])) || (this->unusedCell[index] == 1) )
+ {
+ this->work_u[index] = u[j];
+ this->unusedCell[index] = 0;
+ }
+ }
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::runSubgrid( int boundaryCondition, VectorType u, int subGridID,VectorType map)
+{
+
+ VectorType fu;
+
+ fu.setLike(u);
+ fu.setValue( 0.0 );
+
+
+
+ bool tmp = false;
+ for(int i = 0; i < u.getSize(); i++)
+ {
+ if(u[0]*u[i] <= 0.0)
+ tmp=true;
+ int centerGID = (this->n*(subGridID / this->gridRows)+ (this->n >> 1))*(this->n*this->gridCols) + this->n*(subGridID % this->gridRows) + (this->n >> 1);
+ if(this->unusedCell[centerGID] == 0 || boundaryCondition == 0)
+ tmp = true;
+ }
+
+
+ double value = sign(u[0]) * u.absMax();
+
+ if(tmp)
+ {}
+
+
+ //north - 1, east - 2, west - 4, south - 8
+ else if(boundaryCondition == 4)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j = 1;j < this->n; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[i*this->n]))
+ u[i*this->n + j] = value;// u[i*this->n];
+ }
+ else if(boundaryCondition == 2)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j =0 ;j < this->n -1; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[(i+1)*this->n - 1]))
+ u[i*this->n + j] = value;// u[(i+1)*this->n - 1];
+ }
+ else if(boundaryCondition == 1)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 0;i < this->n - 1; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j + this->n*(this->n - 1)]))
+ u[i*this->n + j] = value;// u[j + this->n*(this->n - 1)];
+ }
+ else if(boundaryCondition == 8)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 1;i < this->n; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j]))
+ u[i*this->n + j] = value;// u[j];
+ }
+
+
+
+ double time = 0.0;
+ double currentTau = this->tau0;
+ double finalTime = this->stopTime;// + 3.0*(u.max() - u.min());
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+
+ double maxResidue( 1.0 );
+ tnlGridEntity<MeshType, 2, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+
+ for( int i = 0; i < u.getSize(); i ++ )
+ {
+ if(map[i] == 0.0)
+ {
+ u[i] = /*sign(u[l])**/MAP_SOLVER_MAX_VALUE;
+ }
+ }
+
+ while( time < finalTime )
+ {
+ /****
+ * Compute the RHS
+ */
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+ Entity.setCoordinates(Containers::StaticVector<2,int>(i % subMesh.getDimensions().x(),i / subMesh.getDimensions().x()));
+ Entity.refresh();
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+ if(map[i] != 0.0)
+ fu[ i ] = schemeHost.getValue( this->subMesh, i, Containers::StaticVector<2,int>(i % subMesh.getDimensions().x(),i / subMesh.getDimensions().x()), u, time, boundaryCondition,neighbourEntities,map);
+ }
+ maxResidue = fu. absMax();
+
+
+ if(maxResidue != 0.0)
+ currentTau = fabs(this -> cflCondition / maxResidue);
+
+
+ if(currentTau > 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >())
+ {
+ currentTau = 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+
+
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+
+
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+ if(map[i] != 0.0)
+ u[ i ] += currentTau * fu[ i ];
+ }
+ time += currentTau;
+
+ }
+ return u;
+}
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridCUDA2D( const int i ,tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+ int th = (blockIdx.y) * caller->n*caller->n*caller->gridCols
+ + (blockIdx.x) * caller->n
+ + threadIdx.y * caller->n*caller->gridCols
+ + threadIdx.x;
+
+ *a = caller->work_u_cuda[th];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::updateSubgridCUDA2D( const int i ,tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+ int index = (blockIdx.y) * caller->n*caller->n*caller->gridCols
+ + (blockIdx.x) * caller->n
+ + threadIdx.y * caller->n*caller->gridCols
+ + threadIdx.x;
+
+ if( (fabs(caller->work_u_cuda[index]) > fabs(*a)) || (caller->unusedCell_cuda[index] == 1) )
+ {
+ caller->work_u_cuda[index] = *a;
+ caller->unusedCell_cuda[index] = 0;
+
+ }
+
+ *a = caller->work_u_cuda[index];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::insertSubgridCUDA2D( double u, const int i )
+{
+ int index = (blockIdx.y)*this->n*this->n*this->gridCols
+ + (blockIdx.x)*this->n
+ + threadIdx.y*this->n*this->gridCols
+ + threadIdx.x;
+
+ if( (fabs(this->work_u_cuda[index]) > fabs(u)) || (this->unusedCell_cuda[index] == 1) )
+ {
+ this->work_u_cuda[index] = u;
+ this->unusedCell_cuda[index] = 0;
+
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::runSubgridCUDA2D( int boundaryCondition, double* u, int subGridID)
+{
+
+ __shared__ int tmp;
+ __shared__ double value;
+ volatile double* sharedTau = &u[blockDim.x*blockDim.y];
+ double* map_local = &u[2*blockDim.x*blockDim.y];
+
+ int i = threadIdx.x;
+ int j = threadIdx.y;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+ int gid = (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x;
+
+ /* LOAD MAP */
+ map_local[l]=this->map_stretched_cuda[gid];
+ if(map_local[l] != 0.0)
+ map_local[l] = 1.0/map_local[l];
+ /* LOADED */
+
+ bool computeFU = !((i == 0 && (boundaryCondition & 4)) or
+ (i == blockDim.x - 1 && (boundaryCondition & 2)) or
+ (j == 0 && (boundaryCondition & 8)) or
+ (j == blockDim.y - 1 && (boundaryCondition & 1)));
+
+ if(l == 0)
+ {
+ tmp = 0;
+ int centerGID = (blockDim.y*blockIdx.y + (blockDim.y>>1))*(blockDim.x*gridDim.x) + blockDim.x*blockIdx.x + (blockDim.x>>1);
+ if(this->unusedCell_cuda[centerGID] == 0 || boundaryCondition == 0)
+ tmp = 1;
+ }
+ __syncthreads();
+
+
+ if(tmp !=1)
+ {
+ if(computeFU)
+ {
+ if(boundaryCondition == 4)
+ u[l] = u[threadIdx.y * blockDim.x] ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.x);
+ else if(boundaryCondition == 2)
+ u[l] = u[threadIdx.y * blockDim.x + blockDim.x - 1] ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(this->n - 1 - threadIdx.x);
+ else if(boundaryCondition == 8)
+ u[l] = u[threadIdx.x] ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.y);
+ else if(boundaryCondition == 1)
+ u[l] = u[(blockDim.y - 1)* blockDim.x + threadIdx.x] ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(this->n - 1 - threadIdx.y);
+ }
+ }
+
+ double time = 0.0;
+ __shared__ double currentTau;
+ double cfl = this->cflCondition;
+ double fu = 0.0;
+
+ double finalTime = this->stopTime;
+ if(boundaryCondition == 0)
+ finalTime*=2.0;
+ __syncthreads();
+
+ tnlGridEntity<MeshType, 2, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Entity.setCoordinates(Containers::StaticVector<2,int>(i,j));
+ Entity.refresh();
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+
+
+ if(map_local[l] == 0.0)
+ {
+ u[l] = /*sign(u[l])**/MAP_SOLVER_MAX_VALUE;
+ computeFU = false;
+ }
+ __syncthreads();
+
+
+ while( time < finalTime )
+ {
+ sharedTau[l] = finalTime;
+
+ if(computeFU)
+ {
+ fu = schemeHost.getValueDev( this->subMesh, l, Containers::StaticVector<2,int>(i,j), u, time, boundaryCondition, neighbourEntities, map_local);
+ sharedTau[l]=abs(cfl/fu);
+ }
+
+
+
+ if(l == 0)
+ {
+ if(sharedTau[0] > 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >()) sharedTau[0] = 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ else if(l == blockDim.x*blockDim.y - 1)
+ if( time + sharedTau[l] > finalTime ) sharedTau[l] = finalTime - time;
+
+
+ if((blockDim.x == 16) && (l < 128)) sharedTau[l] = Min(sharedTau[l],sharedTau[l+128]);
+ __syncthreads();
+ if((blockDim.x == 16) && (l < 64)) sharedTau[l] = Min(sharedTau[l],sharedTau[l+64]);
+ __syncthreads();
+ if(l < 32) sharedTau[l] = Min(sharedTau[l],sharedTau[l+32]);
+ if(l < 16) sharedTau[l] = Min(sharedTau[l],sharedTau[l+16]);
+ if(l < 8) sharedTau[l] = Min(sharedTau[l],sharedTau[l+8]);
+ if(l < 4) sharedTau[l] = Min(sharedTau[l],sharedTau[l+4]);
+ if(l < 2) sharedTau[l] = Min(sharedTau[l],sharedTau[l+2]);
+ if(l < 1) currentTau = Min(sharedTau[l],sharedTau[l+1]);
+ __syncthreads();
+
+ u[l] += currentTau * fu;
+ time += currentTau;
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getOwnerCUDA2D(int i) const
+{
+
+ return ((i / (this->gridCols*this->n*this->n))*this->gridCols
+ + (i % (this->gridCols*this->n))/this->n);
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValueCUDA2D( int i ) const
+{
+ return this->subgridValues_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValueCUDA2D(int i, int value)
+{
+ this->subgridValues_cuda[i] = value;
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryConditionCUDA2D( int i ) const
+{
+ return this->boundaryConditions_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryConditionCUDA2D(int i, int value)
+{
+ this->boundaryConditions_cuda[i] = value;
+}
+
+
+
+//north - 1, east - 2, west - 4, south - 8
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void synchronizeCUDA2D(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver) //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+
+ __shared__ int boundary[4]; // north,east,west,south
+ __shared__ int subgridValue;
+ __shared__ int newSubgridValue;
+
+
+ int gid = (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x;
+ double u = cudaSolver->work_u_cuda[gid];
+ double u_cmp;
+ int subgridValue_cmp=INT_MAX;
+ int boundary_index=0;
+
+
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+ subgridValue = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x);
+ boundary[0] = 0;
+ boundary[1] = 0;
+ boundary[2] = 0;
+ boundary[3] = 0;
+ newSubgridValue = 0;
+ }
+ __syncthreads();
+
+
+
+ if( (threadIdx.x == 0 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == 0 /* && (cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.x == blockDim.x - 1 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == blockDim.y - 1 /* && (cudaSolver->currentStep & 1)*/) )
+ {
+ if(threadIdx.x == 0 && (blockIdx.x != 0)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x - 1);
+ boundary_index = 2;
+ }
+
+ if(threadIdx.x == blockDim.x - 1 && (blockIdx.x != gridDim.x - 1)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x + 1);
+ boundary_index = 1;
+ }
+
+ __threadfence();
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ u=u_cmp;
+ }
+ __threadfence();
+ if(threadIdx.y == 0 && (blockIdx.y != 0)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D((blockIdx.y - 1)*gridDim.x + blockIdx.x);
+ boundary_index = 3;
+ }
+ if(threadIdx.y == blockDim.y - 1 && (blockIdx.y != gridDim.y - 1)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D((blockIdx.y + 1)*gridDim.x + blockIdx.x);
+ boundary_index = 0;
+ }
+
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ }
+ }
+ __threadfence();
+ __syncthreads();
+
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+ if(subgridValue == INT_MAX && newSubgridValue !=0)
+ cudaSolver->setSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, -INT_MAX);
+
+ cudaSolver->setBoundaryConditionCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, boundary[0] +
+ 2 * boundary[1] +
+ 4 * boundary[2] +
+ 8 * boundary[3]);
+
+
+ if(blockIdx.x+blockIdx.y ==0)
+ {
+ cudaSolver->currentStep += 1;
+ *(cudaSolver->runcuda) = 0;
+ }
+ }
+
+}
+
+
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void synchronize2CUDA2D(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver)
+{
+
+
+ int stepValue = cudaSolver->currentStep + 4;
+ if( cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x) == -INT_MAX )
+ cudaSolver->setSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, stepValue);
+
+ atomicMax((cudaSolver->runcuda),cudaSolver->getBoundaryConditionCUDA2D(blockIdx.y*gridDim.x + blockIdx.x));
+}
+
+
+
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void initCUDA2D( tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr , int* ptr2, int* ptr3, double* tmp_map_ptr)
+{
+
+
+ cudaSolver->work_u_cuda = ptr;
+ cudaSolver->map_stretched_cuda = tmp_map_ptr;
+ cudaSolver->unusedCell_cuda = ptr3;
+ cudaSolver->subgridValues_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*sizeof(int));
+ cudaSolver->boundaryConditions_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*sizeof(int));
+ cudaSolver->runcuda = ptr2;
+ *(cudaSolver->runcuda) = 1;
+
+/* CHANGED !!!!!! from 1 to 0*/ cudaSolver->currentStep = 0;
+
+ printf("GPU memory allocated.\n");
+
+ for(int i = 0; i < cudaSolver->gridCols*cudaSolver->gridRows; i++)
+ {
+ cudaSolver->subgridValues_cuda[i] = INT_MAX;
+ cudaSolver->boundaryConditions_cuda[i] = 0;
+ }
+
+ printf("GPU memory initialized.\n");
+}
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void initRunCUDA2D(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller)
+
+{
+ extern __shared__ double u[];
+
+ int i = blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+
+ __shared__ int containsCurve;
+ if(l == 0)
+ containsCurve = 0;
+
+
+ caller->getSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+
+ if(u[0] * u[l] <= 0.0)
+ atomicMax( &containsCurve, 1);
+
+ __syncthreads();
+ if(containsCurve == 1)
+ {
+ caller->runSubgridCUDA2D(0,u,i);
+ caller->insertSubgridCUDA2D(u[l],i);
+ __syncthreads();
+ if(l == 0)
+ caller->setSubgridValueCUDA2D(i, 4);
+ }
+
+
+}
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void runCUDA2D(tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller)
+{
+ extern __shared__ double u[];
+ int i = blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+ int bound = caller->getBoundaryConditionCUDA2D(i);
+
+ if(caller->getSubgridValueCUDA2D(i) != INT_MAX && bound != 0 && caller->getSubgridValueCUDA2D(i) > 0)
+ {
+ caller->getSubgridCUDA2D(i,caller, &u[l]);
+
+
+ if(caller->getSubgridValueCUDA2D(i) == caller->currentStep+4)
+ {
+ if(bound & 1)
+ {
+ caller->runSubgridCUDA2D(1,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 2)
+ {
+ caller->runSubgridCUDA2D(2,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 4)
+ {
+ caller->runSubgridCUDA2D(4,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 8)
+ {
+ caller->runSubgridCUDA2D(8,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ }
+ else
+ {
+
+ if(bound == 1)
+ {
+ caller->runSubgridCUDA2D(1,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound == 2)
+ {
+ caller->runSubgridCUDA2D(2,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound == 4)
+ {
+ caller->runSubgridCUDA2D(4,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound == 8)
+ {
+ caller->runSubgridCUDA2D(8,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ }
+
+ if(bound & 3)
+ {
+ caller->runSubgridCUDA2D(3,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 5)
+ {
+ caller->runSubgridCUDA2D(5,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 10)
+ {
+ caller->runSubgridCUDA2D(10,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 12)
+ {
+ caller->runSubgridCUDA2D(12,u,i);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+ if(l==0)
+ {
+ caller->setBoundaryConditionCUDA2D(i, 0);
+ caller->setSubgridValueCUDA2D(i, caller->getSubgridValueCUDA2D(i) - 1 );
+ }
+
+
+ }
+
+
+
+}
+
+#endif /*HAVE_CUDA*/
+
+#endif /* TNLPARALLELMAPSOLVER2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..b65f3e6a4c8d0cc22a5e10dc46f298fcc5e0dfc0
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/CMakeLists.txt
@@ -0,0 +1,23 @@
+set( tnl_hamilton_jacobi_parallel_SOURCES
+# MainBuildConfig.h
+# tnlParallelEikonalSolver2D_impl.h
+# tnlParallelEikonalSolver3D_impl.h
+# tnlParallelEikonalSolver.h
+# parallelEikonalConfig.h
+ main.cpp)
+
+
+IF( BUILD_CUDA )
+ CUDA_ADD_EXECUTABLE(hamilton-jacobi-parallel${debugExt} main.cu)
+ELSE( BUILD_CUDA )
+ ADD_EXECUTABLE(hamilton-jacobi-parallel${debugExt} main.cpp)
+ENDIF( BUILD_CUDA )
+target_link_libraries (hamilton-jacobi-parallel${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS hamilton-jacobi-parallel${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
+#INSTALL( FILES ${tnl_hamilton_jacobi_parallel_SOURCES}
+# DESTINATION share/tnl-${tnlVersion}/examples/hamilton-jacobi-parallel )
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/MainBuildConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..b13498e17330fae7bb00a0bdc2abcc7a19f8e7a8
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cpp
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cu b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cu
new file mode 100644
index 0000000000000000000000000000000000000000..7101976712e153d73c5f0979b211164a36ec648d
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.cu
@@ -0,0 +1,17 @@
+/***************************************************************************
+ main.cu - description
+ -------------------
+ begin : Mar 30 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "main.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.h
new file mode 100644
index 0000000000000000000000000000000000000000..3a976dae469a77dc33559cf562f722913de0ef28
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/main.h
@@ -0,0 +1,142 @@
+/***************************************************************************
+ main.h - description
+ -------------------
+ begin : Mar 30 , 2015
+ copyright : (C) 2015 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "tnlParallelEikonalSolver.h"
+#include "parallelEikonalConfig.h"
+#include "MainBuildConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+#include <operators/hamilton-jacobi/godunov-eikonal/parallelGodunovEikonal.h>
+#include <mesh/tnlGrid.h>
+#include <core/tnlDevice.h>
+#include <time.h>
+#include <ctime>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ time_t start;
+ time_t stop;
+ time(&start);
+ std::clock_t start2= std::clock();
+ Config::ParameterContainer parameters;
+ tnlConfigDescription configDescription;
+ parallelEikonalConfig< BuildConfig >::configSetup( configDescription );
+
+ if( ! parseCommandLine( argc, argv, configDescription, parameters ) )
+ return false;
+
+ //if (parameters.GetParameter <String>("scheme") == "godunov")
+ //{
+ tnlDeviceEnum device;
+ device = tnlHostDevice;
+
+ const int& dim = parameters.getParameter< int >( "dim" );
+
+ if(dim == 2)
+ {
+
+ typedef parallelGodunovEikonalScheme< tnlGrid<2,double,tnlHost, int>, double, int > SchemeTypeHost;
+ /*#ifdef HAVE_CUDA
+ typedef parallelGodunovEikonalScheme< tnlGrid<2,double,tnlCuda, int>, double, int > SchemeTypeDevice;
+ #endif
+ #ifndef HAVE_CUDA*/
+ typedef parallelGodunovEikonalScheme< tnlGrid<2,double,tnlHost, int>, double, int > SchemeTypeDevice;
+ /*#endif*/
+
+ if(device==tnlHostDevice)
+ {
+ typedef tnlHost Device;
+
+
+ tnlParallelEikonalSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ else if(device==tnlCudaDevice )
+ {
+ typedef tnlCuda Device;
+ //typedef parallelGodunovEikonalScheme< tnlGrid<2,double,Device, int>, double, int > SchemeType;
+
+ tnlParallelEikonalSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ // }
+ }
+ else if(dim == 3)
+ {
+
+ typedef parallelGodunovEikonalScheme< tnlGrid<3,double,tnlHost, int>, double, int > SchemeTypeHost;
+ /*#ifdef HAVE_CUDA
+ typedef parallelGodunovEikonalScheme< tnlGrid<2,double,tnlCuda, int>, double, int > SchemeTypeDevice;
+ #endif
+ #ifndef HAVE_CUDA*/
+ typedef parallelGodunovEikonalScheme< tnlGrid<3,double,tnlHost, int>, double, int > SchemeTypeDevice;
+ /*#endif*/
+
+ if(device==tnlHostDevice)
+ {
+ typedef tnlHost Device;
+
+
+ tnlParallelEikonalSolver<3,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ else if(device==tnlCudaDevice )
+ {
+ typedef tnlCuda Device;
+ //typedef parallelGodunovEikonalScheme< tnlGrid<2,double,Device, int>, double, int > SchemeType;
+
+ tnlParallelEikonalSolver<3,SchemeTypeHost,SchemeTypeDevice, Device> solver;
+ if(!solver.init(parameters))
+ {
+ cerr << "Solver failed to initialize." << endl;
+ return EXIT_FAILURE;
+ }
+ cout << "-------------------------------------------------------------" << endl;
+ cout << "Starting solver loop..." << endl;
+ solver.run();
+ }
+ // }
+ }
+
+ time(&stop);
+ cout << endl;
+ cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) << endl;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/no-Makefile b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/no-Makefile
new file mode 100644
index 0000000000000000000000000000000000000000..bfdc1ef236ca02ecfe6bc88f81d872e9524ec621
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/no-Makefile
@@ -0,0 +1,41 @@
+TNL_VERSION=0.1
+TNL_INSTALL_DIR=${HOME}/local/lib
+TNL_INCLUDE_DIR=${HOME}/local/include/tnl-${TNL_VERSION}
+
+TARGET = hamiltonJacobiParallelSolver
+#CONFIG_FILE = $(TARGET).cfg.desc
+INSTALL_DIR = ${HOME}/local
+CXX = g++
+CUDA_CXX = nvcc
+OMP_FLAGS = -DHAVE_OPENMP -fopenmp
+CXX_FLAGS = -std=gnu++0x -I$(TNL_INCLUDE_DIR) -O3 $(OMP_FLAGS) -DDEBUG
+LD_FLAGS = -L$(TNL_INSTALL_DIR) -ltnl-0.1 -lgomp
+
+SOURCES = main.cpp
+HEADERS =
+OBJECTS = main.o
+DIST = $(SOURCES) Makefile
+
+all: $(TARGET)
+clean:
+ rm -f $(OBJECTS)
+ rm -f $(TARGET)-conf.h
+
+dist: $(DIST)
+ tar zcvf $(TARGET).tgz $(DIST)
+
+install: $(TARGET)
+ cp $(TARGET) $(INSTALL_DIR)/bin
+ cp $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+uninstall: $(TARGET)
+ rm -f $(INSTALL_DIR)/bin/$(TARGET)
+ rm -f $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+$(TARGET): $(OBJECTS)
+ $(CXX) -o $(TARGET) $(OBJECTS) $(LD_FLAGS)
+
+%.o: %.cpp $(HEADERS)
+ $(CXX) -c -o $@ $(CXX_FLAGS) $<
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/parallelEikonalConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/parallelEikonalConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..c27f5ebb39e5c4db31ed13d1a8e80b8ca8915d51
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/parallelEikonalConfig.h
@@ -0,0 +1,46 @@
+/***************************************************************************
+ parallelEikonalConfig.h - description
+ -------------------
+ begin : Oct 5, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef HAMILTONJACOBIPARALLELEIKONALPROBLEMCONFIG_H_
+#define HAMILTONJACOBIPARALLELEIKONALPROBLEMCONFIG_H_
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class parallelEikonalConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Parallel Eikonal solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "hamilton-jacobi-parallel" );
+ config.addEntry < String > ( "scheme", "This defines scheme used for discretization.", "godunov" );
+ config.addEntryEnum( "godunov" );
+ config.addEntryEnum( "upwind" );
+ config.addRequiredEntry < String > ( "initial-condition", "Initial condition for solver");
+ config.addEntry < String > ( "mesh", "Name of mesh.", "mesh.tnl" );
+ config.addEntry < double > ( "epsilon", "This defines epsilon for smoothening of sign().", 0.0 );
+ config.addEntry < double > ( "delta", " Allowed difference on subgrid boundaries", 0.0 );
+ config.addRequiredEntry < double > ( "stop-time", " Final time for solver");
+ config.addRequiredEntry < double > ( "initial-tau", " initial tau for solver" );
+ config.addEntry < double > ( "cfl-condition", " CFL condition", 0.0 );
+ config.addEntry < int > ( "subgrid-size", "Subgrid size.", 16 );
+ config.addRequiredEntry < int > ( "dim", "Dimension of problem.");
+ }
+};
+
+#endif /* HAMILTONJACOBIPARALLELEIKONALPROBLEMCONFIG_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/run b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/run
new file mode 100755
index 0000000000000000000000000000000000000000..3aece294a9c1189cd885acbe459dba20be713716
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/run
@@ -0,0 +1,64 @@
+#!/bin/bash
+
+#GRID_SIZES="0897"
+GRID_SIZES="0008 0015 0029 0057 0113 0225 0449"
+#GRID_SIZES="1793"
+
+dimensions=2
+
+size=2
+
+time=3
+
+for grid_size in $GRID_SIZES;
+
+do
+
+ rm -r grid-${grid_size}
+ mkdir grid-${grid_size}
+ cd grid-${grid_size}
+
+ tnl-grid-setup --dimensions $dimensions \
+ --origin-x -1.0 \
+ --origin-y -1.0 \
+ --origin-z -1.0 \
+ --proportions-x $size \
+ --proportions-y $size \
+ --proportions-z $size \
+ --size-x ${grid_size} \
+ --size-y ${grid_size} \
+ --size-z ${grid_size}
+
+ tnl-init --test-function sdf-para \
+ --offset 0.25 \
+ --output-file init.tnl \
+ --final-time 0.0 \
+ --snapshot-period 0.1 \
+
+
+ tnl-init --test-function sdf-para-sdf \
+ --offset 0.25 \
+ --output-file sdf.tnl \
+ --final-time 0.0 \
+ --snapshot-period 0.1
+
+ hamilton-jacobi-parallel --initial-condition init.tnl \
+ --cfl-condition 1.0e-1 \
+ --mesh mesh.tnl \
+ --initial-tau 1.0e-3 \
+ --epsilon 1.0 \
+ --delta 0.0 \
+ --stop-time $time \
+ --scheme godunov \
+ --subgrid-size 8
+
+ tnl-diff --mesh mesh.tnl --mode sequence --input-files sdf.tnl u-00001.tnl --write-difference yes --output-file ../${grid_size}.diff
+
+ cd ..
+
+done
+
+
+./tnl-err2eoc-2.py --format txt --size $size *.diff
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnl-err2eoc-2.py b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnl-err2eoc-2.py
new file mode 100755
index 0000000000000000000000000000000000000000..f8cde3768e9b76156507e133f8bc3ecaa526fc71
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnl-err2eoc-2.py
@@ -0,0 +1,141 @@
+#!/usr/bin/env python
+
+import sys, string, math
+
+arguments = sys. argv[1:]
+format = "txt"
+output_file_name = "eoc-table.txt"
+input_files = []
+verbose = 1
+size = 1.0
+
+i = 0
+while i < len( arguments ):
+ if arguments[ i ] == "--format":
+ format = arguments[ i + 1 ]
+ i = i + 2
+ continue
+ if arguments[ i ] == "--output-file":
+ output_file_name = arguments[ i + 1 ]
+ i = i + 2
+ continue
+ if arguments[ i ] == "--verbose":
+ verbose = float( arguments[ i + 1 ] )
+ i = i +2
+ continue
+ if arguments[ i ] == "--size":
+ size = float( arguments[ i + 1 ] )
+ i = i +2
+ continue
+ input_files. append( arguments[ i ] )
+ i = i + 1
+
+if not verbose == 0:
+ print "Writing to " + output_file_name + " in " + format + "."
+
+h_list = []
+l1_norm_list = []
+l2_norm_list = []
+max_norm_list = []
+items = 0
+
+for file_name in input_files:
+ if not verbose == 0:
+ print "Processing file " + file_name
+ file = open( file_name, "r" )
+
+ l1_max = 0.0
+ l_max_max = 0.0
+ file.readline();
+ file.readline();
+ for line in file. readlines():
+ data = string. split( line )
+ h_list. append( size/(float(file_name[0:len(file_name)-5] ) - 1.0) )
+ l1_norm_list. append( float( data[ 1 ] ) )
+ l2_norm_list. append( float( data[ 2 ] ) )
+ max_norm_list. append( float( data[ 3 ] ) )
+ items = items + 1
+ if not verbose == 0:
+ print line
+ file. close()
+
+h_width = 12
+err_width = 15
+file = open( output_file_name, "w" )
+if format == "latex":
+ file. write( "\\begin{tabular}{|r|l|l|l|l|l|l|}\\hline\n" )
+ file. write( "\\raisebox{-1ex}[0ex]{$h$}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_1\\left(\\omega_h;\\left[0,T\\right]\\right)}^{h,\\tau}$}}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_2\\left(\\omega_h;\left[0,T\\right]\\right)}^{h,\\tau}$}}& \n" )
+ file. write( "\\multicolumn{2}{|c|}{\\raisebox{1ex}[3.5ex]{$\\left\| \\cdot \\right\\|_{L_\\infty\\left(\\omega_h;\\left[0,T\\right]\\right)}^{h,\\tau}$}}\\\\ \\cline{2-7} \n" )
+ file. write( " " + string. rjust( " ", h_width ) + "&" +
+ string. rjust( "Error", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) + "&" +
+ string. rjust( "Error", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) + "&" +
+ string. rjust( "Error.", err_width ) + "&" +
+ string. rjust( "{\\bf EOC}", err_width ) +
+ "\\\\ \\hline \\hline \n")
+if format == "txt":
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ file. write( "| h | L1 Err. | L1 EOC. | L2 Err. | L2 EOC | MAX Err. | MAX EOC |\n" )
+ file. write( "+==============+================+================+================+================+================+================+\n" )
+
+
+i = 0
+while i < items:
+ if i == 0:
+ if format == "latex":
+ file. write( " " + string. ljust( str( h_list[ i ] ), h_width ) + "&" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "&"+
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "&" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + "&" +
+ string. rjust( " ", err_width ) + "\\\\\n" )
+ if format == "txt":
+ file. write( "| " + string. ljust( str( h_list[ i ] ), h_width ) + " |" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + " |" +
+ string. rjust( " ", err_width ) + " |\n" )
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ i = i + 1;
+ continue
+ if h_list[ i ] == h_list[ i - 1 ]:
+ print "Unable to count eoc since h[ " + \
+ str( i ) + " ] = h[ " + str( i - 1 ) + \
+ " ] = " + str( h_list[ i ] ) + ". \n"
+ file. write( " eoc error: h[ " + \
+ str( i ) + " ] = h[ " + str( i - 1 ) + \
+ " ] = " + str( h_list[ i ] ) + ". \n" )
+ else:
+ h_ratio = math. log( h_list[ i ] / h_list[ i - 1 ] )
+ l1_ratio = math. log( l1_norm_list[ i ] / l1_norm_list[ i - 1 ] )
+ l2_ratio = math. log( l2_norm_list[ i ] / l2_norm_list[ i - 1 ] )
+ max_ratio = math. log( max_norm_list[ i ] / max_norm_list[ i - 1 ] )
+ if format == "latex":
+ file. write( " " + string. ljust( str( h_list[ i ] ), h_width ) + "&" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( l1_ratio / h_ratio ) + "}", err_width ) + "&" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( l2_ratio / h_ratio ) + "}", err_width ) + "&" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + "&" +
+ string. rjust( "{\\bf " + "%.2g" % ( max_ratio / h_ratio ) + "}", err_width ) + "\\\\\n" )
+ if format == "txt":
+ file. write( "| " + string. ljust( str( h_list[ i ] ), h_width ) + " |" +
+ string. rjust( "%.2g" % l1_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( l1_ratio / h_ratio ) + "**", err_width ) + " |" +
+ string. rjust( "%.2g" % l2_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( l2_ratio / h_ratio ) + "**", err_width ) + " |" +
+ string. rjust( "%.2g" % max_norm_list[ i ], err_width ) + " |" +
+ string. rjust( "**" + "%.2g" % ( max_ratio / h_ratio ) + "**", err_width ) + " |\n" )
+ file. write( "+--------------+----------------+----------------+----------------+----------------+----------------+----------------+\n" )
+ i = i + 1
+
+if format == "latex":
+ file. write( "\\hline \n" )
+ file. write( "\\end{tabular} \n" )
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h
new file mode 100644
index 0000000000000000000000000000000000000000..f79752b7187cf9de4f58299fa9add9ce72f2c407
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h
@@ -0,0 +1,367 @@
+/***************************************************************************
+ tnlParallelEikonalSolver.h - description
+ -------------------
+ begin : Nov 28 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLPARALLELEIKONALSOLVER_H_
+#define TNLPARALLELEIKONALSOLVER_H_
+
+#include <TNL/Config/ParameterContainer.h>
+#include <core/vectors/tnlVector.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlMeshFunction.h>
+#include <core/tnlHost.h>
+#include <mesh/tnlGrid.h>
+#include <mesh/grids/tnlGridEntity.h>
+#include <limits.h>
+#include <core/tnlDevice.h>
+ #include <omp.h>
+
+
+#include <ctime>
+
+#ifdef HAVE_CUDA
+#include <cuda.h>
+#include <core/tnlCuda.h>
+#endif
+
+
+template< int Dimension,
+ typename SchemeHost,
+ typename SchemeDevice,
+ typename Device,
+ typename RealType = double,
+ typename IndexType = int >
+class tnlParallelEikonalSolver
+{};
+
+template<typename SchemeHost, typename SchemeDevice, typename Device>
+class tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >
+{
+public:
+
+ typedef SchemeDevice SchemeTypeDevice;
+ typedef SchemeHost SchemeTypeHost;
+ typedef Device DeviceType;
+ typedef tnlVector< double, tnlHost, int > VectorType;
+ typedef tnlVector< int, tnlHost, int > IntVectorType;
+ typedef tnlGrid< 2, double, tnlHost, int > MeshType;
+#ifdef HAVE_CUDA
+ typedef tnlVector< double, tnlHost, int > VectorTypeCUDA;
+ typedef tnlVector< int, tnlHost, int > IntVectorTypeCUDA;
+ typedef tnlGrid< 2, double, tnlHost, int > MeshTypeCUDA;
+#endif
+ tnlParallelEikonalSolver();
+ bool init( const Config::ParameterContainer& parameters );
+ void run();
+
+ void test();
+
+/*private:*/
+
+
+ void synchronize();
+
+ int getOwner( int i) const;
+
+ int getSubgridValue( int i ) const;
+
+ void setSubgridValue( int i, int value );
+
+ int getBoundaryCondition( int i ) const;
+
+ void setBoundaryCondition( int i, int value );
+
+ void stretchGrid();
+
+ void contractGrid();
+
+ VectorType getSubgrid( const int i ) const;
+
+ void insertSubgrid( VectorType u, const int i );
+
+ VectorType runSubgrid( int boundaryCondition, VectorType u, int subGridID);
+
+
+ tnlMeshFunction<MeshType> u0;
+ VectorType work_u;
+ IntVectorType subgridValues, boundaryConditions, unusedCell, calculationsCount;
+ MeshType mesh, subMesh;
+
+// tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage > Entity;
+
+ SchemeHost schemeHost;
+ SchemeDevice schemeDevice;
+ double delta, tau0, stopTime,cflCondition;
+ int gridRows, gridCols, gridLevels, currentStep, n;
+
+ std::clock_t start;
+ double time_diff;
+
+
+ tnlDeviceEnum device;
+
+ tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* getSelf()
+ {
+ return this;
+ };
+
+#ifdef HAVE_CUDA
+
+ tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver;
+
+ double* work_u_cuda;
+
+ int* subgridValues_cuda;
+ int*boundaryConditions_cuda;
+ int* unusedCell_cuda;
+ int* calculationsCount_cuda;
+ double* tmpw;
+ //MeshTypeCUDA mesh_cuda, subMesh_cuda;
+ //SchemeDevice scheme_cuda;
+ //double delta_cuda, tau0_cuda, stopTime_cuda,cflCondition_cuda;
+ //int gridRows_cuda, gridCols_cuda, currentStep_cuda, n_cuda;
+
+ int* runcuda;
+ int run_host;
+
+
+ __device__ void getSubgridCUDA2D( const int i, tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void updateSubgridCUDA2D( const int i, tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void insertSubgridCUDA2D( double u, const int i );
+
+ __device__ void runSubgridCUDA2D( int boundaryCondition, double* u, int subGridID);
+
+ /*__global__ void runCUDA();*/
+
+ //__device__ void synchronizeCUDA();
+
+ __device__ int getOwnerCUDA2D( int i) const;
+
+ __device__ int getSubgridValueCUDA2D( int i ) const;
+
+ __device__ void setSubgridValueCUDA2D( int i, int value );
+
+ __device__ int getBoundaryConditionCUDA2D( int i ) const;
+
+ __device__ void setBoundaryConditionCUDA2D( int i, int value );
+
+ //__device__ bool initCUDA( tnlParallelEikonalSolver<SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+ /*__global__ void initRunCUDA(tnlParallelEikonalSolver<Scheme, double, tnlHost, int >* caller);*/
+
+#endif
+
+};
+
+
+
+
+
+
+
+ template<typename SchemeHost, typename SchemeDevice, typename Device>
+ class tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >
+ {
+ public:
+
+ typedef SchemeDevice SchemeTypeDevice;
+ typedef SchemeHost SchemeTypeHost;
+ typedef Device DeviceType;
+ typedef tnlVector< double, tnlHost, int > VectorType;
+ typedef tnlVector< int, tnlHost, int > IntVectorType;
+ typedef tnlGrid< 3, double, tnlHost, int > MeshType;
+ #ifdef HAVE_CUDA
+ typedef tnlVector< double, tnlHost, int > VectorTypeCUDA;
+ typedef tnlVector< int, tnlHost, int > IntVectorTypeCUDA;
+ typedef tnlGrid< 3, double, tnlHost, int > MeshTypeCUDA;
+ #endif
+ tnlParallelEikonalSolver();
+ bool init( const Config::ParameterContainer& parameters );
+ void run();
+
+ void test();
+
+ /*private:*/
+
+
+ void synchronize();
+
+ int getOwner( int i) const;
+
+ int getSubgridValue( int i ) const;
+
+ void setSubgridValue( int i, int value );
+
+ int getBoundaryCondition( int i ) const;
+
+ void setBoundaryCondition( int i, int value );
+
+ void stretchGrid();
+
+ void contractGrid();
+
+ VectorType getSubgrid( const int i ) const;
+
+ void insertSubgrid( VectorType u, const int i );
+
+ VectorType runSubgrid( int boundaryCondition, VectorType u, int subGridID);
+
+
+ tnlMeshFunction<MeshType> u0;
+ VectorType work_u;
+ IntVectorType subgridValues, boundaryConditions, unusedCell, calculationsCount;
+ MeshType mesh, subMesh;
+ SchemeHost schemeHost;
+ SchemeDevice schemeDevice;
+ double delta, tau0, stopTime,cflCondition;
+ int gridRows, gridCols, gridLevels, currentStep, n;
+
+ std::clock_t start;
+ double time_diff;
+
+
+ tnlDeviceEnum device;
+
+ tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* getSelf()
+ {
+ return this;
+ };
+
+#ifdef HAVE_CUDA
+
+ tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver;
+
+ double* work_u_cuda;
+
+ int* subgridValues_cuda;
+ int*boundaryConditions_cuda;
+ int* unusedCell_cuda;
+ int* calculationsCount_cuda;
+ double* tmpw;
+ //MeshTypeCUDA mesh_cuda, subMesh_cuda;
+ //SchemeDevice scheme_cuda;
+ //double delta_cuda, tau0_cuda, stopTime_cuda,cflCondition_cuda;
+ //int gridRows_cuda, gridCols_cuda, currentStep_cuda, n_cuda;
+
+ int* runcuda;
+ int run_host;
+
+
+ __device__ void getSubgridCUDA3D( const int i, tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void updateSubgridCUDA3D( const int i, tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* caller, double* a);
+
+ __device__ void insertSubgridCUDA3D( double u, const int i );
+
+ __device__ void runSubgridCUDA3D( int boundaryCondition, double* u, int subGridID);
+
+ /*__global__ void runCUDA();*/
+
+ //__device__ void synchronizeCUDA();
+
+ __device__ int getOwnerCUDA3D( int i) const;
+
+ __device__ int getSubgridValueCUDA3D( int i ) const;
+
+ __device__ void setSubgridValueCUDA3D( int i, int value );
+
+ __device__ int getBoundaryConditionCUDA3D( int i ) const;
+
+ __device__ void setBoundaryConditionCUDA3D( int i, int value );
+
+ //__device__ bool initCUDA( tnlParallelEikonalSolver<SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+ /*__global__ void initRunCUDA(tnlParallelEikonalSolver<Scheme, double, tnlHost, int >* caller);*/
+
+#endif
+
+};
+
+
+
+
+
+
+#ifdef HAVE_CUDA
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void runCUDA2D(tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initRunCUDA2D(tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initCUDA2D( tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr, int * ptr2, int* ptr3);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronizeCUDA2D(tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronize2CUDA2D(tnlParallelEikonalSolver<2, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+
+
+
+
+
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void runCUDA3D(tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initRunCUDA3D(tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* caller);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void initCUDA3D( tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr, int * ptr2, int* ptr3);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronizeCUDA3D(tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__ void synchronize2CUDA3D(tnlParallelEikonalSolver<3, SchemeHost, SchemeDevice, Device, double, int >* cudaSolver);
+#endif
+
+
+#ifdef HAVE_CUDA
+__cuda_callable__
+double fabsMin( double x, double y)
+{
+ double fx = fabs(x);
+
+ if(Min(fx,fabs(y)) == fx)
+ return x;
+ else
+ return y;
+}
+
+__cuda_callable__
+double atomicFabsMin(double* address, double val)
+{
+ unsigned long long int* address_as_ull =
+ (unsigned long long int*)address;
+ unsigned long long int old = *address_as_ull, assumed;
+ do {
+ assumed = old;
+ old = atomicCAS(address_as_ull, assumed,__double_as_longlong( fabsMin(__longlong_as_double(assumed),val) ));
+ } while (assumed != old);
+ return __longlong_as_double(old);
+}
+
+#endif
+
+#include "tnlParallelEikonalSolver2D_impl.h"
+#include "tnlParallelEikonalSolver3D_impl.h"
+#endif /* TNLPARALLELEIKONALSOLVER_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..6279fbb805733172ebda8a78c68a128d0d33f92c
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h
@@ -0,0 +1,1928 @@
+/***************************************************************************
+ tnlParallelEikonalSolver2D_impl.h - description
+ -------------------
+ begin : Nov 28 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLPARALLELEIKONALSOLVER2D_IMPL_H_
+#define TNLPARALLELEIKONALSOLVER2D_IMPL_H_
+
+
+#include "tnlParallelEikonalSolver.h"
+#include <core/mfilename.h>
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::tnlParallelEikonalSolver()
+{
+ cout << "a" << endl;
+ this->device = tnlCudaDevice; /////////////// tnlCuda Device --- vypocet na GPU, tnlHostDevice --- vypocet na CPU
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ run_host = 1;
+ }
+#endif
+
+ cout << "b" << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::test()
+{
+/*
+ for(int i =0; i < this->subgridValues.getSize(); i++ )
+ {
+ insertSubgrid(getSubgrid(i), i);
+ }
+*/
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+
+bool tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::init( const Config::ParameterContainer& parameters )
+{
+ cout << "Initializating solver..." << endl;
+ const String& meshLocation = parameters.getParameter <String>("mesh");
+ this->mesh.load( meshLocation );
+
+ this->n = parameters.getParameter <int>("subgrid-size");
+ cout << "Setting N to " << this->n << endl;
+
+ this->subMesh.setDimensions( this->n, this->n );
+ this->subMesh.setDomain( Containers::StaticVector<2,double>(0.0, 0.0),
+ Containers::StaticVector<2,double>(mesh.template getSpaceStepsProducts< 1, 0 >()*(double)(this->n), mesh.template getSpaceStepsProducts< 0, 1 >()*(double)(this->n)) );
+
+ this->subMesh.save("submesh.tnl");
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ this->u0.load( initialCondition );
+
+ //cout << this->mesh.getCellCenter(0) << endl;
+
+ this->delta = parameters.getParameter <double>("delta");
+ this->delta *= mesh.template getSpaceStepsProducts< 1, 0 >()*mesh.template getSpaceStepsProducts< 0, 1 >();
+
+ cout << "Setting delta to " << this->delta << endl;
+
+ this->tau0 = parameters.getParameter <double>("initial-tau");
+ cout << "Setting initial tau to " << this->tau0 << endl;
+ this->stopTime = parameters.getParameter <double>("stop-time");
+
+ this->cflCondition = parameters.getParameter <double>("cfl-condition");
+ this -> cflCondition *= sqrt(mesh.template getSpaceStepsProducts< 1, 0 >()*mesh.template getSpaceStepsProducts< 0, 1 >());
+ cout << "Setting CFL to " << this->cflCondition << endl;
+
+ stretchGrid();
+ this->stopTime /= (double)(this->gridCols);
+ this->stopTime *= (1.0+1.0/((double)(this->n) - 2.0));
+ cout << "Setting stopping time to " << this->stopTime << endl;
+ //this->stopTime = 1.5*((double)(this->n))*parameters.getParameter <double>("stop-time")*this->mesh.template getSpaceStepsProducts< 1, 0 >();
+ //cout << "Setting stopping time to " << this->stopTime << endl;
+
+ cout << "Initializating scheme..." << endl;
+ if(!this->schemeHost.init(parameters))
+ {
+ cerr << "SchemeHost failed to initialize." << endl;
+ return false;
+ }
+ cout << "Scheme initialized." << endl;
+
+ test();
+
+ VectorType* tmp = new VectorType[subgridValues.getSize()];
+ bool containsCurve = false;
+
+#ifdef HAVE_CUDA
+
+ if(this->device == tnlCudaDevice)
+ {
+ /*cout << "Testing... " << endl;
+ if(this->device == tnlCudaDevice)
+ {
+ if( !initCUDA2D(parameters, gridRows, gridCols) )
+ return false;
+ }*/
+ //cout << "s" << endl;
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >));
+ //cout << "s" << endl;
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >), cudaMemcpyHostToDevice);
+ //cout << "s" << endl;
+ double** tmpdev = NULL;
+ cudaMalloc(&tmpdev, sizeof(double*));
+ //double* tmpw;
+ cudaMalloc(&(this->tmpw), this->work_u.getSize()*sizeof(double));
+ cudaMalloc(&(this->runcuda), sizeof(int));
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ int* tmpUC;
+ cudaMalloc(&(tmpUC), this->work_u.getSize()*sizeof(int));
+ cudaMemcpy(tmpUC, this->unusedCell.getData(), this->unusedCell.getSize()*sizeof(int), cudaMemcpyHostToDevice);
+
+ initCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<1,1>>>(this->cudaSolver, (this->tmpw), (this->runcuda),tmpUC);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "s " << endl;
+ //cudaMalloc(&(cudaSolver->work_u_cuda), this->work_u.getSize()*sizeof(double));
+ double* tmpu = NULL;
+
+ cudaMemcpy(&tmpu, tmpdev,sizeof(double*), cudaMemcpyDeviceToHost);
+ //printf("%p %p \n",tmpu,tmpw);
+ cudaMemcpy((this->tmpw), this->work_u.getData(), this->work_u.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "s "<< endl;
+
+ }
+#endif
+
+ if(this->device == tnlHostDevice)
+ {
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+
+ if(! tmp[i].setSize(this->n * this->n))
+ cout << "Could not allocate tmp["<< i <<"] array." << endl;
+ tmp[i] = getSubgrid(i);
+ containsCurve = false;
+
+ for(int j = 0; j < tmp[i].getSize(); j++)
+ {
+ if(tmp[i][0]*tmp[i][j] <= 0.0)
+ {
+ containsCurve = true;
+ j=tmp[i].getSize();
+ }
+
+ }
+ if(containsCurve)
+ {
+ //cout << "Computing initial SDF on subgrid " << i << "." << endl;
+ tmp[i] = runSubgrid(0, tmp[i],i);
+ insertSubgrid(tmp[i], i);
+ setSubgridValue(i, 4);
+ //cout << "Computed initial SDF on subgrid " << i << "." << endl;
+ }
+ containsCurve = false;
+
+ }
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+// cout << "pre 1 kernel" << endl;
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initRunCUDA2D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+// cout << "post 1 kernel" << endl;
+
+ }
+#endif
+
+
+ this->currentStep = 1;
+ if(this->device == tnlHostDevice)
+ synchronize();
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+ //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
+ //cout << test[0] <<" " << test[1] <<" " << test[2] <<" " << test[3] << endl;
+ //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->tmpw << " " << test[0] <<" " << test[1] << " " <<test[2] << " " <<test[3] << endl;
+
+ checkCudaDevice;
+
+ synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << test[0] << " " <<test[1] <<" " << test[2] << " " <<test[3] << endl;
+ //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //checkCudaDevice;
+ //cout << this->tmpw << " " << test[0] << " " <<test[1] << " " <<test[2] <<" " << test[3] << endl;
+ //free(test);
+
+ }
+
+#endif
+ cout << "Solver initialized." << endl;
+
+ return true;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::run()
+{
+ if(this->device == tnlHostDevice)
+ {
+
+ bool end = false;
+ while ((this->boundaryConditions.max() > 0 ) || !end)
+ {
+ if(this->boundaryConditions.max() == 0 )
+ end=true;
+ else
+ end=false;
+#ifdef HAVE_OPENMP
+#pragma omp parallel for num_threads(4) schedule(dynamic)
+#endif
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ if(getSubgridValue(i) != INT_MAX)
+ {
+ VectorType tmp;
+ tmp.setSize(this->n * this->n);
+ //cout << "subMesh: " << i << ", BC: " << getBoundaryCondition(i) << endl;
+
+ if(getSubgridValue(i) == currentStep+4)
+ {
+
+ if(getBoundaryCondition(i) & 1)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 2)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i);
+ insertSubgrid( tmp, 2);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 4)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(4, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 8)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(8, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ }
+
+ if( ((getBoundaryCondition(i) & 2) )|| (getBoundaryCondition(i) & 1)//)
+ /* &&(!(getBoundaryCondition(i) & 5) && !(getBoundaryCondition(i) & 10)) */)
+ {
+ //cout << "3 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i);
+ insertSubgrid( tmp, 3);
+ }
+ if( ((getBoundaryCondition(i) & 4) )|| (getBoundaryCondition(i) & 1)//)
+ /* &&(!(getBoundaryCondition(i) & 3) && !(getBoundaryCondition(i) & 12)) */)
+ {
+ //cout << "5 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(5, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( ((getBoundaryCondition(i) & 2) )|| (getBoundaryCondition(i) & 8)//)
+ /* &&(!(getBoundaryCondition(i) & 12) && !(getBoundaryCondition(i) & 3))*/ )
+ {
+ //cout << "10 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(10, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( ((getBoundaryCondition(i) & 4) )|| (getBoundaryCondition(i) & 8)//)
+ /*&&(!(getBoundaryCondition(i) & 10) && !(getBoundaryCondition(i) & 5)) */)
+ {
+ //cout << "12 @ " << getBoundaryCondition(i) << endl;
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(12, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+
+
+ /*if(getBoundaryCondition(i))
+ {
+ insertSubgrid( runSubgrid(15, getSubgrid(i),i), i);
+ }*/
+
+ setBoundaryCondition(i, 0);
+
+ setSubgridValue(i, getSubgridValue(i)-1);
+
+ }
+ }
+ synchronize();
+ }
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ //cout << "fn" << endl;
+ bool end_cuda = false;
+ dim3 threadsPerBlock(this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cudaMalloc(&runcuda,sizeof(bool));
+ //cudaMemcpy(runcuda, &run_host, sizeof(bool), cudaMemcpyHostToDevice);
+ //cout << "fn" << endl;
+ bool* tmpb;
+ //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
+ //cudaDeviceSynchronize();
+ //checkCudaDevice;
+ cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "fn" << endl;
+ int i = 1;
+ time_diff = 0.0;
+ while (run_host || !end_cuda)
+ {
+ cout << "Computing at step "<< i++ << endl;
+ if(run_host != 0 )
+ end_cuda = true;
+ else
+ end_cuda = false;
+ //cout << "a" << endl;
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ start = std::clock();
+ runCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ //cout << "a" << endl;
+ cudaDeviceSynchronize();
+ time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+ //start = std::clock();
+ synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+
+ //cout << "a" << endl;
+ //run_host = false;
+ //cout << "in kernel loop" << run_host << endl;
+ //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
+ cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
+ //cout << "in kernel loop" << run_host << endl;
+ }
+ cout << "Solving time was: " << time_diff << endl;
+ //cout << "b" << endl;
+
+ //double* tmpu;
+ //cudaMemcpy(tmpu, &(cudaSolver->work_u_cuda),sizeof(double*), cudaMemcpyHostToDevice);
+ //cudaMemcpy(this->work_u.getData(), tmpu, this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->work_u.getData()[0] << endl;
+
+ //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
+ //cout << test[0] << test[1] << test[2] << test[3] << endl;
+ cudaMemcpy(this->work_u.getData()/* test*/, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->tmpw << " " << test[0] << test[1] << test[2] << test[3] << endl;
+ //free(test);
+
+ cudaDeviceSynchronize();
+ }
+#endif
+ contractGrid();
+ this->u0.save("u-00001.tnl");
+ cout << "Maximum number of calculations on one subgrid was " << this->calculationsCount.absMax() << endl;
+ cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) << endl;
+ cout << "Solver finished" << endl;
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ cudaFree(this->runcuda);
+ cudaFree(this->tmpw);
+ cudaFree(this->cudaSolver);
+ }
+#endif
+
+}
+
+//north - 1, east - 2, west - 4, south - 8
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::synchronize() //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+ cout << "Synchronizig..." << endl;
+ int tmp1, tmp2;
+ int grid1, grid2;
+
+ if(this->currentStep & 1)
+ {
+ for(int j = 0; j < this->gridRows - 1; j++)
+ {
+ for (int i = 0; i < this->gridCols*this->n; i++)
+ {
+ tmp1 = this->gridCols*this->n*((this->n-1)+j*this->n) + i;
+ tmp2 = this->gridCols*this->n*((this->n)+j*this->n) + i;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j" << i << "," << j << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 8) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+8);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 1) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+1);
+ }
+ }
+ }
+
+ }
+ else
+ {
+ for(int i = 1; i < this->gridCols; i++)
+ {
+ for (int j = 0; j < this->gridRows*this->n; j++)
+ {
+ tmp1 = this->gridCols*this->n*j + i*this->n - 1;
+ tmp2 = this->gridCols*this->n*j + i*this->n ;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j" << i << "," << j << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 4) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+4);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 2) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+2);
+ }
+ }
+ }
+ }
+
+
+ this->currentStep++;
+ int stepValue = this->currentStep + 4;
+ for (int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ if( getSubgridValue(i) == -INT_MAX )
+ setSubgridValue(i, stepValue);
+ }
+
+ cout << "Grid synchronized at step " << (this->currentStep - 1 ) << endl;
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getOwner(int i) const
+{
+
+ return (i / (this->gridCols*this->n*this->n))*this->gridCols + (i % (this->gridCols*this->n))/this->n;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValue( int i ) const
+{
+ return this->subgridValues[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValue(int i, int value)
+{
+ this->subgridValues[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryCondition( int i ) const
+{
+ return this->boundaryConditions[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryCondition(int i, int value)
+{
+ this->boundaryConditions[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::stretchGrid()
+{
+ cout << "Stretching grid..." << endl;
+
+
+ this->gridCols = ceil( ((double)(this->mesh.getDimensions().x()-1)) / ((double)(this->n-1)) );
+ this->gridRows = ceil( ((double)(this->mesh.getDimensions().y()-1)) / ((double)(this->n-1)) );
+
+ //this->gridCols = (this->mesh.getDimensions().x()-1) / (this->n-1) ;
+ //this->gridRows = (this->mesh.getDimensions().y()-1) / (this->n-1) ;
+
+ cout << "Setting gridCols to " << this->gridCols << "." << endl;
+ cout << "Setting gridRows to " << this->gridRows << "." << endl;
+
+ this->subgridValues.setSize(this->gridCols*this->gridRows);
+ this->subgridValues.setValue(0);
+ this->boundaryConditions.setSize(this->gridCols*this->gridRows);
+ this->boundaryConditions.setValue(0);
+ this->calculationsCount.setSize(this->gridCols*this->gridRows);
+ this->calculationsCount.setValue(0);
+
+ for(int i = 0; i < this->subgridValues.getSize(); i++ )
+ {
+ this->subgridValues[i] = INT_MAX;
+ this->boundaryConditions[i] = 0;
+ }
+
+ int stretchedSize = this->n*this->n*this->gridCols*this->gridRows;
+
+ if(!this->work_u.setSize(stretchedSize))
+ cerr << "Could not allocate memory for stretched grid." << endl;
+ if(!this->unusedCell.setSize(stretchedSize))
+ cerr << "Could not allocate memory for supporting stretched grid." << endl;
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+ for(int i = 0; i < stretchedSize; i++)
+ {
+ this->unusedCell[i] = 1;
+ int diff =(this->n*this->gridCols) - idealStretch ;
+ //cout << "diff = " << diff <<endl;
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if(i%(this->n*this->gridCols) - idealStretch >= 0)
+ {
+ //cout << i%(this->n*this->gridCols) - idealStretch +1 << endl;
+ k+= i%(this->n*this->gridCols) - idealStretch +1 ;
+ }
+
+ if(i/(this->n*this->gridCols) - idealStretch + 1 > 0)
+ {
+ //cout << i/(this->n*this->gridCols) - idealStretch + 1 << endl;
+ k+= (i/(this->n*this->gridCols) - idealStretch +1 )* this->mesh.getDimensions().x() ;
+ }
+
+ //cout << "i = " << i << " : i-k = " << i-k << endl;
+ /*int j=(i % (this->n*this->gridCols)) - ( (this->mesh.getDimensions().x() - this->n)/(this->n - 1) + this->mesh.getDimensions().x() - 1)
+ + (this->n*this->gridCols - this->mesh.getDimensions().x())*(i/(this->n*this->n*this->gridCols)) ;
+
+ if(j > 0)
+ k += j;
+
+ int l = i-k - (this->u0.getSize() - 1);
+ int m = (l % this->mesh.getDimensions().x());
+
+ if(l>0)
+ k+= l + ( (l / this->mesh.getDimensions().x()) + 1 )*this->mesh.getDimensions().x() - (l % this->mesh.getDimensions().x());*/
+
+ this->work_u[i] = this->u0[i-k];
+ //cout << (i-k) <<endl;
+ }
+
+
+ cout << "Grid stretched." << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
+{
+ cout << "Contracting grid..." << endl;
+ int stretchedSize = this->n*this->n*this->gridCols*this->gridRows;
+
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+ for(int i = 0; i < stretchedSize; i++)
+ {
+ int diff =(this->n*this->gridCols) - idealStretch ;
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if((i%(this->n*this->gridCols) - idealStretch < 0) && (i/(this->n*this->gridCols) - idealStretch + 1 <= 0))
+ {
+ //cout << i <<" : " <<i-k<< endl;
+ this->u0[i-k] = this->work_u[i];
+ }
+
+ }
+
+ cout << "Grid contracted" << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgrid( const int i ) const
+{
+ VectorType u;
+ u.setSize(this->n*this->n);
+
+ for( int j = 0; j < u.getSize(); j++)
+ {
+ u[j] = this->work_u[ (i / this->gridCols) * this->n*this->n*this->gridCols
+ + (i % this->gridCols) * this->n
+ + (j/this->n) * this->n*this->gridCols
+ + (j % this->n) ];
+ }
+ return u;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::insertSubgrid( VectorType u, const int i )
+{
+
+ for( int j = 0; j < this->n*this->n; j++)
+ {
+ int index = (i / this->gridCols)*this->n*this->n*this->gridCols
+ + (i % this->gridCols)*this->n
+ + (j/this->n)*this->n*this->gridCols
+ + (j % this->n);
+ //OMP LOCK index
+ if( (fabs(this->work_u[index]) > fabs(u[j])) || (this->unusedCell[index] == 1) )
+ {
+ this->work_u[index] = u[j];
+ this->unusedCell[index] = 0;
+ }
+ //OMP UNLOCK index
+ }
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::runSubgrid( int boundaryCondition, VectorType u, int subGridID)
+{
+
+ VectorType fu;
+
+ fu.setLike(u);
+ fu.setValue( 0.0 );
+
+/*
+ * Insert Euler-Solver Here
+ */
+
+ /**/
+
+ /*for(int i = 0; i < u.getSize(); i++)
+ {
+ int x = this->subMesh.getCellCoordinates(i).x();
+ int y = this->subMesh.getCellCoordinates(i).y();
+
+ if(x == 0 && (boundaryCondition & 4) && y ==0)
+ {
+ if((u[subMesh.getCellYSuccessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 0, 1 >() > 1.0)
+ {
+ //cout << "x = 0; y = 0" << endl;
+ u[i] = u[subMesh.getCellYSuccessor( i )] - subMesh.template getSpaceStepsProducts< 0, 1 >();
+ }
+ }
+ else if(x == 0 && (boundaryCondition & 4) && y == subMesh.getDimensions().y() - 1)
+ {
+ if((u[subMesh.getCellYPredecessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 0, 1 >() > 1.0)
+ {
+ //cout << "x = 0; y = n" << endl;
+ u[i] = u[subMesh.getCellYPredecessor( i )] - subMesh.template getSpaceStepsProducts< 0, 1 >();
+ }
+ }
+
+
+ else if(x == subMesh.getDimensions().x() - 1 && (boundaryCondition & 2) && y ==0)
+ {
+ if((u[subMesh.getCellYSuccessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 0, 1 >() > 1.0)
+ {
+ //cout << "x = n; y = 0" << endl;
+ u[i] = u[subMesh.getCellYSuccessor( i )] - subMesh.template getSpaceStepsProducts< 0, 1 >();
+ }
+ }
+ else if(x == subMesh.getDimensions().x() - 1 && (boundaryCondition & 2) && y == subMesh.getDimensions().y() - 1)
+ {
+ if((u[subMesh.getCellYPredecessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 0, 1 >() > 1.0)
+ {
+ //cout << "x = n; y = n" << endl;
+ u[i] = u[subMesh.getCellYPredecessor( i )] - subMesh.template getSpaceStepsProducts< 0, 1 >();
+ }
+ }
+
+
+ else if(y == 0 && (boundaryCondition & 8) && x ==0)
+ {
+ if((u[subMesh.getCellXSuccessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 1, 0 >() > 1.0)
+ {
+ //cout << "y = 0; x = 0" << endl;
+ u[i] = u[subMesh.getCellXSuccessor( i )] - subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ }
+ else if(y == 0 && (boundaryCondition & 8) && x == subMesh.getDimensions().x() - 1)
+ {
+ if((u[subMesh.getCellXPredecessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 1, 0 >() > 1.0)
+ {
+ //cout << "y = 0; x = n" << endl;
+ u[i] = u[subMesh.getCellXPredecessor( i )] - subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ }
+
+
+ else if(y == subMesh.getDimensions().y() - 1 && (boundaryCondition & 1) && x ==0)
+ {
+ if((u[subMesh.getCellXSuccessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 1, 0 >() > 1.0) {
+ //cout << "y = n; x = 0" << endl;
+ u[i] = u[subMesh.getCellXSuccessor( i )] - subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ }
+ else if(y == subMesh.getDimensions().y() - 1 && (boundaryCondition & 1) && x == subMesh.getDimensions().x() - 1)
+ {
+ if((u[subMesh.getCellXPredecessor( i )] - u[i])/subMesh.template getSpaceStepsProducts< 1, 0 >() > 1.0)
+ {
+ //cout << "y = n; x = n" << endl;
+ u[i] = u[subMesh.getCellXPredecessor( i )] - subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ }
+ }*/
+
+ /**/
+
+
+/* bool tmp = false;
+ for(int i = 0; i < u.getSize(); i++)
+ {
+ if(u[0]*u[i] <= 0.0)
+ tmp=true;
+ }
+
+
+ if(tmp)
+ {}
+ else if(boundaryCondition == 4)
+ {
+ int i;
+ for(i = 0; i < u.getSize() - subMesh.getDimensions().x() ; i=subMesh.getCellYSuccessor(i))
+ {
+ int j;
+ for(j = i; j < subMesh.getDimensions().x() - 1; j=subMesh.getCellXSuccessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ int j;
+ for(j = i; j < subMesh.getDimensions().x() - 1; j=subMesh.getCellXSuccessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ else if(boundaryCondition == 8)
+ {
+ int i;
+ for(i = 0; i < subMesh.getDimensions().x() - 1; i=subMesh.getCellXSuccessor(i))
+ {
+ int j;
+ for(j = i; j < u.getSize() - subMesh.getDimensions().x(); j=subMesh.getCellYSuccessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ int j;
+ for(j = i; j < u.getSize() - subMesh.getDimensions().x(); j=subMesh.getCellYSuccessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+
+ }
+ else if(boundaryCondition == 2)
+ {
+ int i;
+ for(i = subMesh.getDimensions().x() - 1; i < u.getSize() - subMesh.getDimensions().x() ; i=subMesh.getCellYSuccessor(i))
+ {
+ int j;
+ for(j = i; j > (i-1)*subMesh.getDimensions().x(); j=subMesh.getCellXPredecessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ int j;
+ for(j = i; j > (i-1)*subMesh.getDimensions().x(); j=subMesh.getCellXPredecessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ else if(boundaryCondition == 1)
+ {
+ int i;
+ for(i = (subMesh.getDimensions().y() - 1)*subMesh.getDimensions().x(); i < u.getSize() - 1; i=subMesh.getCellXSuccessor(i))
+ {
+ int j;
+ for(j = i; j >=subMesh.getDimensions().x(); j=subMesh.getCellYPredecessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+ int j;
+ for(j = i; j >=subMesh.getDimensions().x(); j=subMesh.getCellYPredecessor(j))
+ {
+ u[j] = u[i];
+ }
+ u[j] = u[i];
+ }
+*/
+ /**/
+
+
+
+ bool tmp = false;
+ for(int i = 0; i < u.getSize(); i++)
+ {
+ if(u[0]*u[i] <= 0.0)
+ tmp=true;
+ int centerGID = (this->n*(subGridID / this->gridRows)+ (this->n >> 1))*(this->n*this->gridCols) + this->n*(subGridID % this->gridRows) + (this->n >> 1);
+ if(this->unusedCell[centerGID] == 0 || boundaryCondition == 0)
+ tmp = true;
+ }
+ //if(this->currentStep + 3 < getSubgridValue(subGridID))
+ //tmp = true;
+
+
+ double value = sign(u[0]) * u.absMax();
+
+ if(tmp)
+ {}
+
+
+ //north - 1, east - 2, west - 4, south - 8
+ else if(boundaryCondition == 4)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j = 1;j < this->n; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[i*this->n]))
+ u[i*this->n + j] = value;// u[i*this->n];
+ }
+ else if(boundaryCondition == 2)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j =0 ;j < this->n -1; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[(i+1)*this->n - 1]))
+ u[i*this->n + j] = value;// u[(i+1)*this->n - 1];
+ }
+ else if(boundaryCondition == 1)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 0;i < this->n - 1; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j + this->n*(this->n - 1)]))
+ u[i*this->n + j] = value;// u[j + this->n*(this->n - 1)];
+ }
+ else if(boundaryCondition == 8)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 1;i < this->n; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j]))
+ u[i*this->n + j] = value;// u[j];
+ }
+
+/*
+
+ else if(boundaryCondition == 5)
+ {
+ for(int i = 0; i < this->n - 1; i++)
+ for(int j = 1;j < this->n; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[i*this->n]))
+ u[i*this->n + j] = value;// u[i*this->n];
+ }
+ else if(boundaryCondition == 10)
+ {
+ for(int i = 1; i < this->n; i++)
+ for(int j =0 ;j < this->n -1; j++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[(i+1)*this->n - 1]))
+ u[i*this->n + j] = value;// u[(i+1)*this->n - 1];
+ }
+ else if(boundaryCondition == 3)
+ {
+ for(int j = 0; j < this->n - 1; j++)
+ for(int i = 0;i < this->n - 1; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j + this->n*(this->n - 1)]))
+ u[i*this->n + j] = value;// u[j + this->n*(this->n - 1)];
+ }
+ else if(boundaryCondition == 12)
+ {
+ for(int j = 1; j < this->n; j++)
+ for(int i = 1;i < this->n; i++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j]))
+ u[i*this->n + j] = value;// u[j];
+ }
+*/
+
+
+ /**/
+
+ /*if (u.max() > 0.0)
+ this->stopTime *=(double) this->gridCols;*/
+
+
+ double time = 0.0;
+ double currentTau = this->tau0;
+ double finalTime = this->stopTime;// + 3.0*(u.max() - u.min());
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+
+ double maxResidue( 1.0 );
+ //double lastResidue( 10000.0 );
+ tnlGridEntity<MeshType, 2, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ while( time < finalTime /*|| maxResidue > subMesh.template getSpaceStepsProducts< 1, 0 >()*/)
+ {
+ /****
+ * Compute the RHS
+ */
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+ Entity.setCoordinates(Containers::StaticVector<2,int>(i % subMesh.getDimensions().x(),i / subMesh.getDimensions().x()));
+ Entity.refresh();
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+ fu[ i ] = schemeHost.getValue( this->subMesh, i, Containers::StaticVector<2,int>(i % subMesh.getDimensions().x(),i / subMesh.getDimensions().x()), u, time, boundaryCondition,neighbourEntities);
+ }
+ maxResidue = fu. absMax();
+
+
+ if( this -> cflCondition * maxResidue != 0.0)
+ currentTau = this -> cflCondition / maxResidue;
+
+ /* if (maxResidue < 0.05)
+ cout << "Max < 0.05" << endl;*/
+ if(currentTau > 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >())
+ {
+ //cout << currentTau << " >= " << 2.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >() << endl;
+ currentTau = 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ /*if(maxResidue > lastResidue)
+ currentTau *=(1.0/10.0);*/
+
+
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+// for( int i = 0; i < fu.getSize(); i ++ )
+// {
+// //cout << "Too big RHS! i = " << i << ", fu = " << fu[i] << ", u = " << u[i] << endl;
+// if((u[i]+currentTau * fu[ i ])*u[i] < 0.0 && fu[i] != 0.0 && u[i] != 0.0 )
+// currentTau = fabs(u[i]/(2.0*fu[i]));
+//
+// }
+
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+ double add = u[i] + currentTau * fu[ i ];
+ //if( fabs(u[i]) < fabs(add) or (this->subgridValues[subGridID] == this->currentStep +4) )
+ u[ i ] = add;
+ }
+ time += currentTau;
+
+ //cout << '\r' << flush;
+ //cout << maxResidue << " " << currentTau << " @ " << time << flush;
+ //lastResidue = maxResidue;
+ }
+ //cout << "Time: " << time << ", Res: " << maxResidue <<endl;
+ /*if (u.max() > 0.0)
+ this->stopTime /=(double) this->gridCols;*/
+
+ VectorType solution;
+ solution.setLike(u);
+ for( int i = 0; i < u.getSize(); i ++ )
+ {
+ solution[i]=u[i];
+ }
+ return solution;
+}
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridCUDA2D( const int i ,tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+ //int j = threadIdx.x + threadIdx.y * blockDim.x;
+ int th = (blockIdx.y) * caller->n*caller->n*caller->gridCols
+ + (blockIdx.x) * caller->n
+ + threadIdx.y * caller->n*caller->gridCols
+ + threadIdx.x;
+ //printf("i= %d,j= %d,th= %d\n",i,j,th);
+ *a = caller->work_u_cuda[th];
+ //printf("Hi %f \n", *a);
+ //return ret;
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::updateSubgridCUDA2D( const int i ,tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+// int j = threadIdx.x + threadIdx.y * blockDim.x;
+ int index = (blockIdx.y) * caller->n*caller->n*caller->gridCols
+ + (blockIdx.x) * caller->n
+ + threadIdx.y * caller->n*caller->gridCols
+ + threadIdx.x;
+
+ if( (fabs(caller->work_u_cuda[index]) > fabs(*a)) || (caller->unusedCell_cuda[index] == 1) )
+ {
+ caller->work_u_cuda[index] = *a;
+ caller->unusedCell_cuda[index] = 0;
+
+ }
+
+ *a = caller->work_u_cuda[index];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::insertSubgridCUDA2D( double u, const int i )
+{
+
+
+// int j = threadIdx.x + threadIdx.y * blockDim.x;
+ //printf("j = %d, u = %f\n", j,u);
+
+ int index = (blockIdx.y)*this->n*this->n*this->gridCols
+ + (blockIdx.x)*this->n
+ + threadIdx.y*this->n*this->gridCols
+ + threadIdx.x;
+
+ //printf("i= %d,j= %d,index= %d\n",i,j,index);
+ if( (fabs(this->work_u_cuda[index]) > fabs(u)) || (this->unusedCell_cuda[index] == 1) )
+ {
+ this->work_u_cuda[index] = u;
+ this->unusedCell_cuda[index] = 0;
+
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::runSubgridCUDA2D( int boundaryCondition, double* u, int subGridID)
+{
+
+ __shared__ int tmp;
+ __shared__ double value;
+ //double tmpRes = 0.0;
+ volatile double* sharedTau = &u[blockDim.x*blockDim.y];
+ volatile double* absVal = &u[2*blockDim.x*blockDim.y];
+ int i = threadIdx.x;
+ int j = threadIdx.y;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+ bool computeFU = !((i == 0 && (boundaryCondition & 4)) or
+ (i == blockDim.x - 1 && (boundaryCondition & 2)) or
+ (j == 0 && (boundaryCondition & 8)) or
+ (j == blockDim.y - 1 && (boundaryCondition & 1)));
+
+ if(l == 0)
+ {
+ tmp = 0;
+ int centerGID = (blockDim.y*blockIdx.y + (blockDim.y>>1))*(blockDim.x*gridDim.x) + blockDim.x*blockIdx.x + (blockDim.x>>1);
+ if(this->unusedCell_cuda[centerGID] == 0 || boundaryCondition == 0)
+ tmp = 1;
+ }
+ __syncthreads();
+
+ /*if(!tmp && (u[0]*u[l] <= 0.0))
+ atomicMax( &tmp, 1);*/
+
+ __syncthreads();
+ if(tmp !=1)
+ {
+// if(computeFU)
+// absVal[l]=0.0;
+// else
+// absVal[l] = fabs(u[l]);
+//
+// __syncthreads();
+//
+// if((blockDim.x == 16) && (l < 128)) absVal[l] = Max(absVal[l],absVal[l+128]);
+// __syncthreads();
+// if((blockDim.x == 16) && (l < 64)) absVal[l] = Max(absVal[l],absVal[l+64]);
+// __syncthreads();
+// if(l < 32) absVal[l] = Max(absVal[l],absVal[l+32]);
+// if(l < 16) absVal[l] = Max(absVal[l],absVal[l+16]);
+// if(l < 8) absVal[l] = Max(absVal[l],absVal[l+8]);
+// if(l < 4) absVal[l] = Max(absVal[l],absVal[l+4]);
+// if(l < 2) absVal[l] = Max(absVal[l],absVal[l+2]);
+// if(l < 1) value = sign(u[0])*Max(absVal[l],absVal[l+1]);
+// __syncthreads();
+//
+// if(computeFU)
+// u[l] = value;
+ if(computeFU)
+ {
+ if(boundaryCondition == 4)
+ u[l] = u[threadIdx.y * blockDim.x] + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.x) ;//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.x+this->n);
+ else if(boundaryCondition == 2)
+ u[l] = u[threadIdx.y * blockDim.x + blockDim.x - 1] + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(this->n - 1 - threadIdx.x);//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(blockDim.x - threadIdx.x - 1+this->n);
+ else if(boundaryCondition == 8)
+ u[l] = u[threadIdx.x] + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.y) ;//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(threadIdx.y+this->n);
+ else if(boundaryCondition == 1)
+ u[l] = u[(blockDim.y - 1)* blockDim.x + threadIdx.x] + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(this->n - 1 - threadIdx.y) ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0 >()*(blockDim.y - threadIdx.y - 1 +this->n);
+ }
+ }
+
+ double time = 0.0;
+ __shared__ double currentTau;
+ double cfl = this->cflCondition;
+ double fu = 0.0;
+// if(threadIdx.x * threadIdx.y == 0)
+// {
+// currentTau = finalTime;
+// }
+ double finalTime = this->stopTime;
+ __syncthreads();
+// if( time + currentTau > finalTime ) currentTau = finalTime - time;
+
+ tnlGridEntity<MeshType, 2, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighbourEntities(Entity);
+ Entity.setCoordinates(Containers::StaticVector<2,int>(i,j));
+ Entity.refresh();
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+
+
+ while( time < finalTime )
+ {
+ if(computeFU)
+ fu = schemeHost.getValueDev( this->subMesh, l, Containers::StaticVector<2,int>(i,j)/*this->subMesh.getCellCoordinates(l)*/, u, time, boundaryCondition, neighbourEntities);
+
+ sharedTau[l]=abs(cfl/fu);
+
+ if(l == 0)
+ {
+ if(sharedTau[0] > 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >()) sharedTau[0] = 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >();
+ }
+ else if(l == blockDim.x*blockDim.y - 1)
+ if( time + sharedTau[l] > finalTime ) sharedTau[l] = finalTime - time;
+
+
+// if( (sign(u[l]+sharedTau[l]*fu) != sign(u[l])) && fu != 0.0 && fu != -0.0)
+// {
+// printf("orig: %10f", sharedTau[l]);
+// sharedTau[l]=abs(u[l]/(1.1*fu)) ;
+// printf(" new: %10f\n", sharedTau[l]);
+// }
+
+
+
+ if((blockDim.x == 16) && (l < 128)) sharedTau[l] = Min(sharedTau[l],sharedTau[l+128]);
+ __syncthreads();
+ if((blockDim.x == 16) && (l < 64)) sharedTau[l] = Min(sharedTau[l],sharedTau[l+64]);
+ __syncthreads();
+ if(l < 32) sharedTau[l] = Min(sharedTau[l],sharedTau[l+32]);
+ if(l < 16) sharedTau[l] = Min(sharedTau[l],sharedTau[l+16]);
+ if(l < 8) sharedTau[l] = Min(sharedTau[l],sharedTau[l+8]);
+ if(l < 4) sharedTau[l] = Min(sharedTau[l],sharedTau[l+4]);
+ if(l < 2) sharedTau[l] = Min(sharedTau[l],sharedTau[l+2]);
+ if(l < 1) currentTau = Min(sharedTau[l],sharedTau[l+1]);
+ __syncthreads();
+
+ u[l] += currentTau * fu;
+ time += currentTau;
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getOwnerCUDA2D(int i) const
+{
+
+ return ((i / (this->gridCols*this->n*this->n))*this->gridCols
+ + (i % (this->gridCols*this->n))/this->n);
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValueCUDA2D( int i ) const
+{
+ return this->subgridValues_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValueCUDA2D(int i, int value)
+{
+ this->subgridValues_cuda[i] = value;
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryConditionCUDA2D( int i ) const
+{
+ return this->boundaryConditions_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryConditionCUDA2D(int i, int value)
+{
+ this->boundaryConditions_cuda[i] = value;
+}
+
+
+
+//north - 1, east - 2, west - 4, south - 8
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void /*tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::*/synchronizeCUDA2D(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver) //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+
+ __shared__ int boundary[4]; // north,east,west,south
+ __shared__ int subgridValue;
+ __shared__ int newSubgridValue;
+
+
+ int gid = (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x + blockDim.x*blockIdx.x + threadIdx.x;
+ double u = cudaSolver->work_u_cuda[gid];
+ double u_cmp;
+ int subgridValue_cmp=INT_MAX;
+ int boundary_index=0;
+
+
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+ subgridValue = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x);
+ boundary[0] = 0;
+ boundary[1] = 0;
+ boundary[2] = 0;
+ boundary[3] = 0;
+ newSubgridValue = 0;
+ //printf("%d %d\n", blockDim.x, gridDim.x);
+ }
+ __syncthreads();
+
+
+
+ if( (threadIdx.x == 0 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == 0 /* && (cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.x == blockDim.x - 1 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == blockDim.y - 1 /* && (cudaSolver->currentStep & 1)*/) )
+ {
+ if(threadIdx.x == 0 && (blockIdx.x != 0)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x - 1);
+ boundary_index = 2;
+ }
+
+ if(threadIdx.x == blockDim.x - 1 && (blockIdx.x != gridDim.x - 1)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x + 1);
+ boundary_index = 1;
+ }
+
+ __threadfence();
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ u=u_cmp;
+ }
+ __threadfence();
+ if(threadIdx.y == 0 && (blockIdx.y != 0)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D((blockIdx.y - 1)*gridDim.x + blockIdx.x);
+ boundary_index = 3;
+ }
+ if(threadIdx.y == blockDim.y - 1 && (blockIdx.y != gridDim.y - 1)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA2D((blockIdx.y + 1)*gridDim.x + blockIdx.x);
+ boundary_index = 0;
+ }
+
+// __threadfence();
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ }
+ }
+ __threadfence();
+ __syncthreads();
+
+ if(threadIdx.x+threadIdx.y == 0)
+ {
+ if(subgridValue == INT_MAX && newSubgridValue !=0)
+ cudaSolver->setSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, -INT_MAX);
+
+ cudaSolver->setBoundaryConditionCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, boundary[0] +
+ 2 * boundary[1] +
+ 4 * boundary[2] +
+ 8 * boundary[3]);
+
+
+ if(blockIdx.x+blockIdx.y ==0)
+ {
+ cudaSolver->currentStep = cudaSolver->currentStep + 1;
+ *(cudaSolver->runcuda) = 0;
+ }
+//
+// int stepValue = cudaSolver->currentStep + 4;
+// if( cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x) == -INT_MAX )
+// cudaSolver->setSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, stepValue);
+//
+// atomicMax((cudaSolver->runcuda),cudaSolver->getBoundaryConditionCUDA2D(blockIdx.y*gridDim.x + blockIdx.x));
+ }
+
+
+ /*
+ //printf("I am not an empty kernel!\n");
+ //cout << "Synchronizig..." << endl;
+ int tmp1, tmp2;
+ int grid1, grid2;
+
+ if(cudaSolver->currentStep & 1)
+ {
+ //printf("I am not an empty kernel! 1\n");
+ for(int j = 0; j < cudaSolver->gridRows - 1; j++)
+ {
+ //printf("I am not an empty kernel! 3\n");
+ for (int i = 0; i < cudaSolver->gridCols*cudaSolver->n; i++)
+ {
+ tmp1 = cudaSolver->gridCols*cudaSolver->n*((cudaSolver->n-1)+j*cudaSolver->n) + i;
+ tmp2 = cudaSolver->gridCols*cudaSolver->n*((cudaSolver->n)+j*cudaSolver->n) + i;
+ grid1 = cudaSolver->getSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1));
+ grid2 = cudaSolver->getSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2));
+
+ if ((fabs(cudaSolver->work_u_cuda[tmp1]) < fabs(cudaSolver->work_u_cuda[tmp2]) - cudaSolver->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ //printf("%d %d %d %d \n",tmp1,tmp2,cudaSolver->getOwnerCUDA2D(tmp1),cudaSolver->getOwnerCUDA2D(tmp2));
+ cudaSolver->work_u_cuda[tmp2] = cudaSolver->work_u_cuda[tmp1];
+ cudaSolver->unusedCell_cuda[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ cudaSolver->setSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2), -INT_MAX);
+ }
+ if(! (cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2)) & 8) )
+ cudaSolver->setBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2), cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2))+8);
+ }
+ else if ((fabs(cudaSolver->work_u_cuda[tmp1]) > fabs(cudaSolver->work_u_cuda[tmp2]) + cudaSolver->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ //printf("%d %d %d %d \n",tmp1,tmp2,cudaSolver->getOwnerCUDA2D(tmp1),cudaSolver->getOwnerCUDA2D(tmp2));
+ cudaSolver->work_u_cuda[tmp1] = cudaSolver->work_u_cuda[tmp2];
+ cudaSolver->unusedCell_cuda[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ cudaSolver->setSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1), -INT_MAX);
+ }
+ if(! (cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1)) & 1) )
+ cudaSolver->setBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1), cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1))+1);
+ }
+ }
+ }
+
+ }
+ else
+ {
+ //printf("I am not an empty kernel! 2\n");
+ for(int i = 1; i < cudaSolver->gridCols; i++)
+ {
+ //printf("I am not an empty kernel! 4\n");
+ for (int j = 0; j < cudaSolver->gridRows*cudaSolver->n; j++)
+ {
+
+ tmp1 = cudaSolver->gridCols*cudaSolver->n*j + i*cudaSolver->n - 1;
+ tmp2 = cudaSolver->gridCols*cudaSolver->n*j + i*cudaSolver->n ;
+ grid1 = cudaSolver->getSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1));
+ grid2 = cudaSolver->getSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2));
+
+ if ((fabs(cudaSolver->work_u_cuda[tmp1]) < fabs(cudaSolver->work_u_cuda[tmp2]) - cudaSolver->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ //printf("%d %d %d %d \n",tmp1,tmp2,cudaSolver->getOwnerCUDA2D(tmp1),cudaSolver->getOwnerCUDA2D(tmp2));
+ cudaSolver->work_u_cuda[tmp2] = cudaSolver->work_u_cuda[tmp1];
+ cudaSolver->unusedCell_cuda[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ cudaSolver->setSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2), -INT_MAX);
+ }
+ if(! (cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2)) & 4) )
+ cudaSolver->setBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2), cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp2))+4);
+ }
+ else if ((fabs(cudaSolver->work_u_cuda[tmp1]) > fabs(cudaSolver->work_u_cuda[tmp2]) + cudaSolver->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ //printf("%d %d %d %d \n",tmp1,tmp2,cudaSolver->getOwnerCUDA2D(tmp1),cudaSolver->getOwnerCUDA2D(tmp2));
+ cudaSolver->work_u_cuda[tmp1] = cudaSolver->work_u_cuda[tmp2];
+ cudaSolver->unusedCell_cuda[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ cudaSolver->setSubgridValueCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1), -INT_MAX);
+ }
+ if(! (cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1)) & 2) )
+ cudaSolver->setBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1), cudaSolver->getBoundaryConditionCUDA2D(cudaSolver->getOwnerCUDA2D(tmp1))+2);
+ }
+ }
+ }
+ }
+ //printf("I am not an empty kernel! 5 cudaSolver->currentStep : %d \n", cudaSolver->currentStep);
+
+ cudaSolver->currentStep = cudaSolver->currentStep + 1;
+ int stepValue = cudaSolver->currentStep + 4;
+ for (int i = 0; i < cudaSolver->gridRows * cudaSolver->gridCols; i++)
+ {
+ if( cudaSolver->getSubgridValueCUDA2D(i) == -INT_MAX )
+ cudaSolver->setSubgridValueCUDA2D(i, stepValue);
+ }
+
+ int maxi = 0;
+ for(int q=0; q < cudaSolver->gridRows*cudaSolver->gridCols;q++)
+ {
+ //printf("%d : %d\n", q, cudaSolver->boundaryConditions_cuda[q]);
+ maxi=Max(maxi,cudaSolver->getBoundaryConditionCUDA2D(q));
+ }
+ //printf("I am not an empty kernel! %d\n", maxi);
+ *(cudaSolver->runcuda) = (maxi > 0);
+ //printf("I am not an empty kernel! 7 %d\n", cudaSolver->boundaryConditions_cuda[0]);
+ //cout << "Grid synchronized at step " << (this->currentStep - 1 ) << endl;
+*/
+}
+
+
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void synchronize2CUDA2D(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver)
+{
+// if(blockIdx.x+blockIdx.y ==0)
+// {
+// cudaSolver->currentStep = cudaSolver->currentStep + 1;
+// *(cudaSolver->runcuda) = 0;
+// }
+
+ int stepValue = cudaSolver->currentStep + 4;
+ if( cudaSolver->getSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x) == -INT_MAX )
+ cudaSolver->setSubgridValueCUDA2D(blockIdx.y*gridDim.x + blockIdx.x, stepValue);
+
+ atomicMax((cudaSolver->runcuda),cudaSolver->getBoundaryConditionCUDA2D(blockIdx.y*gridDim.x + blockIdx.x));
+}
+
+
+
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void /*tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::*/initCUDA2D( tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr , int* ptr2, int* ptr3)
+{
+ //cout << "Initializating solver..." << endl;
+ //const String& meshLocation = parameters.getParameter <String>("mesh");
+ //this->mesh_cuda.load( meshLocation );
+
+ //this->n_cuda = parameters.getParameter <int>("subgrid-size");
+ //cout << "Setting N << this->n_cuda << endl;
+
+ //this->subMesh_cuda.setDimensions( this->n_cuda, this->n_cuda );
+ //this->subMesh_cuda.setDomain( Containers::StaticVector<2,double>(0.0, 0.0),
+ //Containers::StaticVector<2,double>(this->mesh_cuda.template getSpaceStepsProducts< 1, 0 >()*(double)(this->n_cuda), this->mesh_cuda.template getSpaceStepsProducts< 0, 1 >()*(double)(this->n_cuda)) );
+
+ //this->subMesh_cuda.save("submesh.tnl");
+
+// const String& initialCondition = parameters.getParameter <String>("initial-condition");
+// this->u0.load( initialCondition );
+
+ //cout << this->mesh.getCellCenter(0) << endl;
+
+ //this->delta_cuda = parameters.getParameter <double>("delta");
+ //this->delta_cuda *= this->mesh_cuda.template getSpaceStepsProducts< 1, 0 >()*this->mesh_cuda.template getSpaceStepsProducts< 0, 1 >();
+
+ //cout << "Setting delta to " << this->delta << endl;
+
+ //this->tau0_cuda = parameters.getParameter <double>("initial-tau");
+ //cout << "Setting initial tau to " << this->tau0_cuda << endl;
+ //this->stopTime_cuda = parameters.getParameter <double>("stop-time");
+
+ //this->cflCondition_cuda = parameters.getParameter <double>("cfl-condition");
+ //this -> cflCondition_cuda *= sqrt(this->mesh_cuda.template getSpaceStepsProducts< 1, 0 >()*this->mesh_cuda.template getSpaceStepsProducts< 0, 1 >());
+ //cout << "Setting CFL to " << this->cflCondition << endl;
+////
+////
+
+// this->gridRows_cuda = gridRows;
+// this->gridCols_cuda = gridCols;
+
+ cudaSolver->work_u_cuda = ptr;//(double*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->n*cudaSolver->n*sizeof(double));
+ cudaSolver->unusedCell_cuda = ptr3;//(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->n*cudaSolver->n*sizeof(int));
+ cudaSolver->subgridValues_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*sizeof(int));
+ cudaSolver->boundaryConditions_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*sizeof(int));
+ cudaSolver->runcuda = ptr2;//(bool*)malloc(sizeof(bool));
+ *(cudaSolver->runcuda) = 1;
+ cudaSolver->currentStep = 1;
+ //cudaMemcpy(ptr,&(cudaSolver->work_u_cuda), sizeof(double*),cudaMemcpyDeviceToHost);
+ //ptr = cudaSolver->work_u_cuda;
+ printf("GPU memory allocated.\n");
+
+ for(int i = 0; i < cudaSolver->gridCols*cudaSolver->gridRows; i++)
+ {
+ cudaSolver->subgridValues_cuda[i] = INT_MAX;
+ cudaSolver->boundaryConditions_cuda[i] = 0;
+ }
+
+ /*for(long int j = 0; j < cudaSolver->n*cudaSolver->n*cudaSolver->gridCols*cudaSolver->gridRows; j++)
+ {
+ printf("%d\n",j);
+ cudaSolver->unusedCell_cuda[ j] = 1;
+ }*/
+ printf("GPU memory initialized.\n");
+
+
+ //cudaSolver->work_u_cuda[50] = 32.153438;
+////
+////
+ //stretchGrid();
+ //this->stopTime_cuda /= (double)(this->gridCols_cuda);
+ //this->stopTime_cuda *= (1.0+1.0/((double)(this->n_cuda) - 1.0));
+ //cout << "Setting stopping time to " << this->stopTime << endl;
+ //this->stopTime_cuda = 1.5*((double)(this->n_cuda))*parameters.getParameter <double>("stop-time")*this->mesh_cuda.template getSpaceStepsProducts< 1, 0 >();
+ //cout << "Setting stopping time to " << this->stopTime << endl;
+
+ //cout << "Initializating scheme..." << endl;
+ //if(!this->schemeDevice.init(parameters))
+// {
+ //cerr << "Scheme failed to initialize." << endl;
+// return false;
+// }
+ //cout << "Scheme initialized." << endl;
+
+ //test();
+
+// this->currentStep_cuda = 1;
+ //return true;
+}
+
+
+
+
+//extern __shared__ double array[];
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void /*tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::*/initRunCUDA2D(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller)
+
+{
+
+
+ extern __shared__ double u[];
+ //printf("%p\n",caller->work_u_cuda);
+
+ int i = blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+
+ __shared__ int containsCurve;
+ if(l == 0)
+ containsCurve = 0;
+
+ //double a;
+ caller->getSubgridCUDA2D(i,caller, &u[l]);
+ //printf("%f %f\n",a , u[l]);
+ //u[l] = a;
+ //printf("Hi %f \n", u[l]);
+ __syncthreads();
+ //printf("hurewrwr %f \n", u[l]);
+ if(u[0] * u[l] <= 0.0)
+ {
+ //printf("contains %d \n",i);
+ atomicMax( &containsCurve, 1);
+ }
+
+ __syncthreads();
+ //printf("hu");
+ //printf("%d : %f\n", l, u[l]);
+ if(containsCurve == 1)
+ {
+ //printf("have curve \n");
+ caller->runSubgridCUDA2D(0,u,i);
+ //printf("%d : %f\n", l, u[l]);
+ __syncthreads();
+ caller->insertSubgridCUDA2D(u[l],i);
+ __syncthreads();
+ if(l == 0)
+ caller->setSubgridValueCUDA2D(i, 4);
+ }
+
+
+}
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void /*tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::*/runCUDA2D(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >* caller)
+{
+ extern __shared__ double u[];
+ int i = blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.y * blockDim.x + threadIdx.x;
+ int bound = caller->getBoundaryConditionCUDA2D(i);
+
+ if(caller->getSubgridValueCUDA2D(i) != INT_MAX && bound != 0 && caller->getSubgridValueCUDA2D(i) > 0)
+ {
+ caller->getSubgridCUDA2D(i,caller, &u[l]);
+
+ //if(l == 0)
+ //printf("i = %d, bound = %d\n",i,caller->getSubgridValueCUDA2D(i));
+ if(caller->getSubgridValueCUDA2D(i) == caller->currentStep+4)
+ {
+ if(bound & 1)
+ {
+ caller->runSubgridCUDA2D(1,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 2 )
+ {
+ caller->runSubgridCUDA2D(2,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 4)
+ {
+ caller->runSubgridCUDA2D(4,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 8)
+ {
+ caller->runSubgridCUDA2D(8,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+
+
+
+ if( ((bound & 3 )))
+ {
+ caller->runSubgridCUDA2D(3,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 5 )))
+ {
+ caller->runSubgridCUDA2D(5,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 10 )))
+ {
+ caller->runSubgridCUDA2D(10,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( (bound & 12 ))
+ {
+ caller->runSubgridCUDA2D(12,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+
+
+
+ }
+
+
+ else
+ {
+
+
+
+
+
+
+
+
+
+ if( ((bound == 2)))
+ {
+ caller->runSubgridCUDA2D(2,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound == 1) ))
+ {
+ caller->runSubgridCUDA2D(1,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound == 8) ))
+ {
+ caller->runSubgridCUDA2D(8,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( (bound == 4))
+ {
+ caller->runSubgridCUDA2D(4,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+
+
+
+
+
+
+
+
+ if( ((bound & 3) ))
+ {
+ caller->runSubgridCUDA2D(3,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 5) ))
+ {
+ caller->runSubgridCUDA2D(5,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 10) ))
+ {
+ caller->runSubgridCUDA2D(10,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( (bound & 12) )
+ {
+ caller->runSubgridCUDA2D(12,u,i);
+ //__syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+
+
+
+
+
+
+
+
+
+
+ }
+ /*if( bound )
+ {
+ caller->runSubgridCUDA2D(15,u,i);
+ __syncthreads();
+ //caller->insertSubgridCUDA2D(u[l],i);
+ //__syncthreads();
+ //caller->getSubgridCUDA2D(i,caller, &u[l]);
+ caller->updateSubgridCUDA2D(i,caller, &u[l]);
+ __syncthreads();
+ }*/
+
+ if(l==0)
+ {
+ caller->setBoundaryConditionCUDA2D(i, 0);
+ caller->setSubgridValueCUDA2D(i, caller->getSubgridValueCUDA2D(i) - 1 );
+ }
+
+
+ }
+
+
+
+}
+
+#endif /*HAVE_CUDA*/
+
+#endif /* TNLPARALLELEIKONALSOLVER2D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..12d9003099643c923fdd6f4be18f2b07b35ebf3d
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h
@@ -0,0 +1,1706 @@
+/***************************************************************************
+ tnlParallelEikonalSolver2D_impl.h - description
+ -------------------
+ begin : Nov 28 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLPARALLELEIKONALSOLVER3D_IMPL_H_
+#define TNLPARALLELEIKONALSOLVER3D_IMPL_H_
+
+
+#include "tnlParallelEikonalSolver.h"
+#include <core/mfilename.h>
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::tnlParallelEikonalSolver()
+{
+ cout << "a" << endl;
+ this->device = tnlHostDevice; /////////////// tnlCuda Device --- vypocet na GPU, tnlHostDevice --- vypocet na CPU
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ run_host = 1;
+ }
+#endif
+
+ cout << "b" << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::test()
+{
+/*
+ for(int i =0; i < this->subgridValues.getSize(); i++ )
+ {
+ insertSubgrid(getSubgrid(i), i);
+ }
+*/
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+
+bool tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::init( const Config::ParameterContainer& parameters )
+{
+ cout << "Initializating solver..." << endl;
+ const String& meshLocation = parameters.getParameter <String>("mesh");
+ this->mesh.load( meshLocation );
+
+ this->n = parameters.getParameter <int>("subgrid-size");
+ cout << "Setting N to " << this->n << endl;
+
+ this->subMesh.setDimensions( this->n, this->n, this->n );
+ this->subMesh.setDomain( Containers::StaticVector<3,double>(0.0, 0.0, 0.0),
+ Containers::StaticVector<3,double>(mesh.template getSpaceStepsProducts< 1, 0, 0 >()*(double)(this->n), mesh.template getSpaceStepsProducts< 0, 1, 0 >()*(double)(this->n),mesh.template getSpaceStepsProducts< 0, 0, 1 >()*(double)(this->n)) );
+
+ this->subMesh.save("submesh.tnl");
+
+ const String& initialCondition = parameters.getParameter <String>("initial-condition");
+ this->u0.load( initialCondition );
+
+ //cout << this->mesh.getCellCenter(0) << endl;
+
+ this->delta = parameters.getParameter <double>("delta");
+ this->delta *= mesh.template getSpaceStepsProducts< 1, 0, 0 >()*mesh.template getSpaceStepsProducts< 0, 1, 0 >();
+
+ cout << "Setting delta to " << this->delta << endl;
+
+ this->tau0 = parameters.getParameter <double>("initial-tau");
+ cout << "Setting initial tau to " << this->tau0 << endl;
+ this->stopTime = parameters.getParameter <double>("stop-time");
+
+ this->cflCondition = parameters.getParameter <double>("cfl-condition");
+ this -> cflCondition *= sqrt(mesh.template getSpaceStepsProducts< 1, 0, 0 >()*mesh.template getSpaceStepsProducts< 0, 1, 0 >());
+ cout << "Setting CFL to " << this->cflCondition << endl;
+
+ stretchGrid();
+ this->stopTime /= (double)(this->gridCols);
+ this->stopTime *= (1.0+1.0/((double)(this->n) - 2.0));
+ cout << "Setting stopping time to " << this->stopTime << endl;
+ //this->stopTime = 1.5*((double)(this->n))*parameters.getParameter <double>("stop-time")*mesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ //cout << "Setting stopping time to " << this->stopTime << endl;
+
+ cout << "Initializating scheme..." << endl;
+ if(!this->schemeHost.init(parameters))
+ {
+ cerr << "SchemeHost failed to initialize." << endl;
+ return false;
+ }
+ cout << "Scheme initialized." << endl;
+
+ test();
+
+ VectorType* tmp = new VectorType[subgridValues.getSize()];
+
+
+#ifdef HAVE_CUDA
+
+ if(this->device == tnlCudaDevice)
+ {
+ /*cout << "Testing... " << endl;
+ if(this->device == tnlCudaDevice)
+ {
+ if( !initCUDA3D(parameters, gridRows, gridCols) )
+ return false;
+ }*/
+ //cout << "s" << endl;
+ cudaMalloc(&(this->cudaSolver), sizeof(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >));
+ //cout << "s" << endl;
+ cudaMemcpy(this->cudaSolver, this,sizeof(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >), cudaMemcpyHostToDevice);
+ //cout << "s" << endl;
+ double** tmpdev = NULL;
+ cudaMalloc(&tmpdev, sizeof(double*));
+ //double* tmpw;
+ cudaMalloc(&(this->tmpw), this->work_u.getSize()*sizeof(double));
+ cudaMalloc(&(this->runcuda), sizeof(int));
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ int* tmpUC;
+ cudaMalloc(&(tmpUC), this->work_u.getSize()*sizeof(int));
+ cudaMemcpy(tmpUC, this->unusedCell.getData(), this->unusedCell.getSize()*sizeof(int), cudaMemcpyHostToDevice);
+
+ initCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<1,1>>>(this->cudaSolver, (this->tmpw), (this->runcuda),tmpUC);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "s " << endl;
+ //cudaMalloc(&(cudaSolver->work_u_cuda), this->work_u.getSize()*sizeof(double));
+ double* tmpu = NULL;
+
+ cudaMemcpy(&tmpu, tmpdev,sizeof(double*), cudaMemcpyDeviceToHost);
+ //printf("%p %p \n",tmpu,tmpw);
+ cudaMemcpy((this->tmpw), this->work_u.getData(), this->work_u.getSize()*sizeof(double), cudaMemcpyHostToDevice);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "s "<< endl;
+
+ }
+#endif
+
+ if(this->device == tnlHostDevice)
+ {
+#ifdef HAVE_OPENMP
+#pragma omp parallel for num_threads(4) schedule(dynamic)
+#endif
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ bool containsCurve = false;
+// cout << "Working on subgrid " << i <<" --- check 1" << endl;
+
+ if(! tmp[i].setSize(this->n*this->n*this->n))
+ cout << "Could not allocate tmp["<< i <<"] array." << endl;
+// cout << "Working on subgrid " << i <<" --- check 2" << endl;
+
+ tmp[i] = getSubgrid(i);
+ containsCurve = false;
+// cout << "Working on subgrid " << i <<" --- check 3" << endl;
+
+
+ for(int j = 0; j < tmp[i].getSize(); j++)
+ {
+ if(tmp[i][0]*tmp[i][j] <= 0.0)
+ {
+ containsCurve = true;
+ j=tmp[i].getSize();
+// cout << tmp[i][0] << " " << tmp[i][j] << endl;
+ }
+
+ }
+// cout << "Working on subgrid " << i <<" --- check 4" << endl;
+
+ if(containsCurve)
+ {
+// cout << "Computing initial SDF on subgrid " << i << "." << endl;
+ tmp[i] = runSubgrid(0, tmp[i] ,i);
+ insertSubgrid( tmp[i], i);
+ setSubgridValue(i, 4);
+// cout << "Computed initial SDF on subgrid " << i << "." << endl;
+ }
+ containsCurve = false;
+
+ }
+// cout << "CPU: Curve found" << endl;
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+// cout << "pre 1 kernel" << endl;
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ dim3 threadsPerBlock(this->n, this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows,this->gridLevels);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ initRunCUDA3D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,2*this->n*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+// cout << "post 1 kernel" << endl;
+
+ }
+#endif
+
+
+ this->currentStep = 1;
+ if(this->device == tnlHostDevice)
+ synchronize();
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ dim3 threadsPerBlock(this->n, this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows,this->gridLevels);
+ //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
+ //cout << test[0] <<" " << test[1] <<" " << test[2] <<" " << test[3] << endl;
+ //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->tmpw << " " << test[0] <<" " << test[1] << " " <<test[2] << " " <<test[3] << endl;
+
+ checkCudaDevice;
+
+ synchronizeCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cout << cudaGetErrorString(cudaDeviceSynchronize()) << endl;
+ checkCudaDevice;
+ synchronize2CUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << test[0] << " " <<test[1] <<" " << test[2] << " " <<test[3] << endl;
+ //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //checkCudaDevice;
+ //cout << this->tmpw << " " << test[0] << " " <<test[1] << " " <<test[2] <<" " << test[3] << endl;
+ //free(test);
+
+ }
+
+#endif
+ cout << "Solver initialized." << endl;
+
+ return true;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::run()
+{
+ if(this->device == tnlHostDevice)
+ {
+
+ bool end = false;
+ while (/*(this->boundaryConditions.max() > 0 ) ||*/ !end)
+ {
+ if(this->boundaryConditions.max() == 0 || this->subgridValues.max() < 0)
+ end=true;
+ else
+ end=false;
+#ifdef HAVE_OPENMP
+#pragma omp parallel for num_threads(4) schedule(dynamic)
+#endif
+ for(int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ VectorType tmp;
+ tmp.setSize(this->n*this->n*this->n);
+ if(getSubgridValue(i) != INT_MAX)
+ {
+ //cout << "subMesh: " << i << ", BC: " << getBoundaryCondition(i) << endl;
+
+ if(getSubgridValue(i) == currentStep+4)
+ {
+
+ if(getBoundaryCondition(i) & 1)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(1, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 2)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(2, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 4)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(4, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 8)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(8, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 16)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(16, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ if(getBoundaryCondition(i) & 32)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(32, tmp ,i);
+ insertSubgrid( tmp, i);
+ this->calculationsCount[i]++;
+ }
+ }
+
+ if( getBoundaryCondition(i) & 19)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(19, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 21)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(21, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 26)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(26, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 28)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(28, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+
+ if( getBoundaryCondition(i) & 35)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(35, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 37)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(37, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 42)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(42, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+ if( getBoundaryCondition(i) & 44)
+ {
+ tmp = getSubgrid(i);
+ tmp = runSubgrid(44, tmp ,i);
+ insertSubgrid( tmp, i);
+ }
+
+
+ setBoundaryCondition(i, 0);
+ setSubgridValue(i, getSubgridValue(i)-1);
+
+ }
+ }
+ synchronize();
+ }
+ }
+#ifdef HAVE_CUDA
+ else if(this->device == tnlCudaDevice)
+ {
+ //cout << "fn" << endl;
+ bool end_cuda = false;
+ dim3 threadsPerBlock(this->n, this->n, this->n);
+ dim3 numBlocks(this->gridCols,this->gridRows,this->gridLevels);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cudaMalloc(&runcuda,sizeof(bool));
+ //cudaMemcpy(runcuda, &run_host, sizeof(bool), cudaMemcpyHostToDevice);
+ //cout << "fn" << endl;
+ bool* tmpb;
+ //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
+ //cudaDeviceSynchronize();
+ //checkCudaDevice;
+ cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //cout << "fn" << endl;
+ int i = 1;
+ time_diff = 0.0;
+ while (run_host || !end_cuda)
+ {
+ cout << "Computing at step "<< i++ << endl;
+ if(run_host != 0 )
+ end_cuda = true;
+ else
+ end_cuda = false;
+ //cout << "a" << endl;
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ start = std::clock();
+ runCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,2*this->n*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
+ //cout << "a" << endl;
+ cudaDeviceSynchronize();
+ time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+ //start = std::clock();
+ synchronizeCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ synchronize2CUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
+ cudaDeviceSynchronize();
+ checkCudaDevice;
+ //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
+
+
+ //cout << "a" << endl;
+ //run_host = false;
+ //cout << "in kernel loop" << run_host << endl;
+ //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
+ cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
+ //cout << "in kernel loop" << run_host << endl;
+ }
+ cout << "Solving time was: " << time_diff << endl;
+ //cout << "b" << endl;
+
+ //double* tmpu;
+ //cudaMemcpy(tmpu, &(cudaSolver->work_u_cuda),sizeof(double*), cudaMemcpyHostToDevice);
+ //cudaMemcpy(this->work_u.getData(), tmpu, this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->work_u.getData()[0] << endl;
+
+ //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
+ //cout << test[0] << test[1] << test[2] << test[3] << endl;
+ cudaMemcpy(this->work_u.getData()/* test*/, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
+ //cout << this->tmpw << " " << test[0] << test[1] << test[2] << test[3] << endl;
+ //free(test);
+
+ cudaDeviceSynchronize();
+ }
+#endif
+ contractGrid();
+ this->u0.save("u-00001.tnl");
+ cout << "Maximum number of calculations on one subgrid was " << this->calculationsCount.absMax() << endl;
+ cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) << endl;
+ cout << "Solver finished" << endl;
+
+#ifdef HAVE_CUDA
+ if(this->device == tnlCudaDevice)
+ {
+ cudaFree(this->runcuda);
+ cudaFree(this->tmpw);
+ cudaFree(this->cudaSolver);
+ }
+#endif
+
+}
+
+//north - 1, east - 2, west - 4, south - 8
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::synchronize() //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+ cout << "Synchronizig..." << endl;
+ int tmp1, tmp2;
+ int grid1, grid2;
+
+// if(this->currentStep & 1)
+// {
+ for(int j = 0; j < this->gridRows - 1; j++)
+ {
+ for (int i = 0; i < this->gridCols*this->n; i++)
+ {
+ for (int k = 0; k < this->gridLevels*this->n; k++)
+ {
+// cout << "a" << endl;
+ tmp1 = this->gridCols*this->n*((this->n-1)+j*this->n) + i + k*this->gridCols*this->n*this->gridRows*this->n;
+// cout << "b" << endl;
+ tmp2 = this->gridCols*this->n*((this->n)+j*this->n) + i + k*this->gridCols*this->n*this->gridRows*this->n;
+// cout << "c" << endl;
+ if(tmp1 > work_u.getSize())
+ cout << "tmp1: " << tmp1 << " x: " << j <<" y: " << i <<" z: " << k << endl;
+ if(tmp2 > work_u.getSize())
+ cout << "tmp2: " << tmp2 << " x: " << j <<" y: " << i <<" z: " << k << endl;
+ grid1 = getSubgridValue(getOwner(tmp1));
+// cout << "d" << endl;
+ grid2 = getSubgridValue(getOwner(tmp2));
+// cout << "e" << endl;
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j, k" << i << "," << j << "," << k << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+// cout << "f" << endl;
+ this->unusedCell[tmp2] = 0;
+// cout << "g" << endl;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+// cout << "h" << endl;
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 8) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+8);
+// cout << "i" << endl;
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+// cout << "j" << endl;
+ this->unusedCell[tmp1] = 0;
+// cout << "k" << endl;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+// cout << "l" << endl;
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 1) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+1);
+// cout << "m" << endl;
+ }
+ }
+ }
+ }
+
+// }
+// else
+// {
+
+ cout << "sync 2" << endl;
+ for(int i = 1; i < this->gridCols; i++)
+ {
+ for (int j = 0; j < this->gridRows*this->n; j++)
+ {
+ for (int k = 0; k < this->gridLevels*this->n; k++)
+ {
+ tmp1 = this->gridCols*this->n*j + i*this->n - 1 + k*this->gridCols*this->n*this->gridRows*this->n;
+ tmp2 = this->gridCols*this->n*j + i*this->n + k*this->gridCols*this->n*this->gridRows*this->n;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j, k" << i << "," << j << "," << k << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 4) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+4);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 2) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+2);
+ }
+ }
+ }
+ }
+
+ cout << "sync 3" << endl;
+
+ for(int k = 1; k < this->gridLevels; k++)
+ {
+ for (int j = 0; j < this->gridRows*this->n; j++)
+ {
+ for (int i = 0; i < this->gridCols*this->n; i++)
+ {
+ tmp1 = this->gridCols*this->n*j + i + (k*this->n-1)*this->gridCols*this->n*this->gridRows*this->n;
+ tmp2 = this->gridCols*this->n*j + i + k*this->n*this->gridCols*this->n*this->gridRows*this->n;
+ grid1 = getSubgridValue(getOwner(tmp1));
+ grid2 = getSubgridValue(getOwner(tmp2));
+ if(getOwner(tmp1)==getOwner(tmp2))
+ cout << "i, j, k" << i << "," << j << "," << k << endl;
+ if ((fabs(this->work_u[tmp1]) < fabs(this->work_u[tmp2]) - this->delta || grid2 == INT_MAX || grid2 == -INT_MAX) && (grid1 != INT_MAX && grid1 != -INT_MAX))
+ {
+ this->work_u[tmp2] = this->work_u[tmp1];
+ this->unusedCell[tmp2] = 0;
+ if(grid2 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp2), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp2)) & 32) )
+ setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+32);
+ }
+ else if ((fabs(this->work_u[tmp1]) > fabs(this->work_u[tmp2]) + this->delta || grid1 == INT_MAX || grid1 == -INT_MAX) && (grid2 != INT_MAX && grid2 != -INT_MAX))
+ {
+ this->work_u[tmp1] = this->work_u[tmp2];
+ this->unusedCell[tmp1] = 0;
+ if(grid1 == INT_MAX)
+ {
+ setSubgridValue(getOwner(tmp1), -INT_MAX);
+ }
+ if(! (getBoundaryCondition(getOwner(tmp1)) & 16) )
+ setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+16);
+ }
+ }
+ }
+ }
+// }
+
+
+
+ this->currentStep++;
+ int stepValue = this->currentStep + 4;
+ for (int i = 0; i < this->subgridValues.getSize(); i++)
+ {
+ if( getSubgridValue(i) == -INT_MAX )
+ setSubgridValue(i, stepValue);
+ }
+
+ cout << "Grid synchronized at step " << (this->currentStep - 1 ) << endl;
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getOwner(int i) const
+{
+
+ int j = i % (this->gridCols*this->gridRows*this->n*this->n);
+
+ return ( (i / (this->gridCols*this->gridRows*this->n*this->n*this->n))*this->gridCols*this->gridRows
+ + (j / (this->gridCols*this->n*this->n))*this->gridCols
+ + (j % (this->gridCols*this->n))/this->n);
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValue( int i ) const
+{
+ return this->subgridValues[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValue(int i, int value)
+{
+ this->subgridValues[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryCondition( int i ) const
+{
+ return this->boundaryConditions[i];
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryCondition(int i, int value)
+{
+ this->boundaryConditions[i] = value;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::stretchGrid()
+{
+ cout << "Stretching grid..." << endl;
+
+
+ this->gridCols = ceil( ((double)(this->mesh.getDimensions().x()-1)) / ((double)(this->n-1)) );
+ this->gridRows = ceil( ((double)(this->mesh.getDimensions().y()-1)) / ((double)(this->n-1)) );
+ this->gridLevels = ceil( ((double)(this->mesh.getDimensions().z()-1)) / ((double)(this->n-1)) );
+
+ //this->gridCols = (this->mesh.getDimensions().x()-1) / (this->n-1) ;
+ //this->gridRows = (this->mesh.getDimensions().y()-1) / (this->n-1) ;
+
+ cout << "Setting gridCols to " << this->gridCols << "." << endl;
+ cout << "Setting gridRows to " << this->gridRows << "." << endl;
+ cout << "Setting gridLevels to " << this->gridLevels << "." << endl;
+
+ this->subgridValues.setSize(this->gridCols*this->gridRows*this->gridLevels);
+ this->subgridValues.setValue(0);
+ this->boundaryConditions.setSize(this->gridCols*this->gridRows*this->gridLevels);
+ this->boundaryConditions.setValue(0);
+ this->calculationsCount.setSize(this->gridCols*this->gridRows*this->gridLevels);
+ this->calculationsCount.setValue(0);
+
+ for(int i = 0; i < this->subgridValues.getSize(); i++ )
+ {
+ this->subgridValues[i] = INT_MAX;
+ this->boundaryConditions[i] = 0;
+ }
+
+ int levelSize = this->n*this->n*this->gridCols*this->gridRows;
+ int stretchedSize = this->n*levelSize*this->gridLevels;
+
+ if(!this->work_u.setSize(stretchedSize))
+ cerr << "Could not allocate memory for stretched grid." << endl;
+ if(!this->unusedCell.setSize(stretchedSize))
+ cerr << "Could not allocate memory for supporting stretched grid." << endl;
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+
+
+
+ for(int i = 0; i < levelSize; i++)
+ {
+ int diff =(this->n*this->gridCols) - idealStretch ;
+
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if(i%(this->n*this->gridCols) - idealStretch >= 0)
+ {
+ k+= i%(this->n*this->gridCols) - idealStretch +1 ;
+ }
+
+ if(i/(this->n*this->gridCols) - idealStretch + 1 > 0)
+ {
+ k+= (i/(this->n*this->gridCols) - idealStretch +1 )* this->mesh.getDimensions().x() ;
+ }
+
+ for( int j = 0; j<this->n*this->gridLevels; j++)
+ {
+ this->unusedCell[i+j*levelSize] = 1;
+ int l = j/this->n;
+
+ if(j - idealStretch >= 0)
+ {
+ l+= j - idealStretch + 1;
+ }
+
+ this->work_u[i+j*levelSize] = this->u0[i+(j-l)*mesh.getDimensions().x()*mesh.getDimensions().y()-k];
+ }
+
+ }
+
+
+
+ cout << "Grid stretched." << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
+{
+ cout << "Contracting grid..." << endl;
+ int levelSize = this->n*this->n*this->gridCols*this->gridRows;
+ int stretchedSize = this->n*levelSize*this->gridLevels;
+
+ int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
+ cout << idealStretch << endl;
+
+
+ for(int i = 0; i < levelSize; i++)
+ {
+ int diff =(this->n*this->gridCols) - idealStretch ;
+ int k = i/this->n - i/(this->n*this->gridCols) + this->mesh.getDimensions().x()*(i/(this->n*this->n*this->gridCols)) + (i/(this->n*this->gridCols))*diff;
+
+ if((i%(this->n*this->gridCols) - idealStretch < 0) && (i/(this->n*this->gridCols) - idealStretch + 1 <= 0) )
+ {
+ for( int j = 0; j<this->n*this->gridLevels; j++)
+ {
+ int l = j/this->n;
+ if(j - idealStretch < 0)
+ this->u0[i+(j-l)*mesh.getDimensions().x()*mesh.getDimensions().y()-k] = this->work_u[i+j*levelSize];
+ }
+ }
+
+ }
+
+ cout << "Grid contracted" << endl;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getSubgrid( const int i ) const
+{
+
+ VectorType u;
+ u.setSize(this->n*this->n*this->n);
+
+ int idx, idy, idz;
+ idz = i / (gridRows*this->gridCols);
+ idy = (i % (this->gridRows*this->gridCols)) / this->gridCols;
+ idx = i % (this->gridCols);
+
+ for( int j = 0; j < this->n; j++)
+ {
+ // int index = (i / this->gridCols)*this->n*this->n*this->gridCols + (i % this->gridCols)*this->n + (j/this->n)*this->n*this->gridCols + (j % this->n);
+ for( int k = 0; k < this->n; k++)
+ {
+ for( int l = 0; l < this->n; l++)
+ {
+ int index = (idz*this->n + l) * this->n*this->n*this->gridCols*this->gridRows
+ + (idy) * this->n*this->n*this->gridCols
+ + (idx) * this->n
+ + k * this->n*this->gridCols
+ + j;
+
+ u[j + k*this->n + l*this->n*this->n] = this->work_u[ index ];
+ }
+ }
+ }
+ return u;
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::insertSubgrid( VectorType u, const int i )
+{
+ int idx, idy, idz;
+ idz = i / (this->gridRows*this->gridCols);
+ idy = (i % (this->gridRows*this->gridCols)) / this->gridCols;
+ idx = i % (this->gridCols);
+
+ for( int j = 0; j < this->n; j++)
+ {
+ // int index = (i / this->gridCols)*this->n*this->n*this->gridCols + (i % this->gridCols)*this->n + (j/this->n)*this->n*this->gridCols + (j % this->n);
+ for( int k = 0; k < this->n; k++)
+ {
+ for( int l = 0; l < this->n; l++)
+ {
+
+ int index = (idz*this->n + l) * this->n*this->n*this->gridCols*this->gridRows
+ + (idy) * this->n*this->n*this->gridCols
+ + (idx) * this->n
+ + k * this->n*this->gridCols
+ + j;
+
+ //OMP LOCK index
+// cout<< idx << " " << idy << " " << idz << " " << j << " " << k << " " << l << " " << idz << " " << unusedCell.getSize() << " " << u.getSize() << " " << index <<endl;
+ if( (fabs(this->work_u[index]) > fabs(u[j + k*this->n + l*this->n*this->n])) || (this->unusedCell[index] == 1) )
+ {
+ this->work_u[index] = u[j + k*this->n + l*this->n*this->n];
+ this->unusedCell[index] = 0;
+ }
+ //OMP UNLOCK index
+ }
+ }
+ }
+}
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+typename tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::VectorType
+tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::runSubgrid( int boundaryCondition, VectorType u, int subGridID)
+{
+
+ VectorType fu;
+
+ fu.setLike(u);
+ fu.setValue( 0.0 );
+
+
+ bool tmp = false;
+ for(int i = 0; i < u.getSize(); i++)
+ {
+ if(u[0]*u[i] <= 0.0)
+ tmp=true;
+ }
+ int idx,idy,idz;
+ idz = subGridID / (this->gridRows*this->gridCols);
+ idy = (subGridID % (this->gridRows*this->gridCols)) / this->gridCols;
+ idx = subGridID % (this->gridCols);
+ int centerGID = (this->n*idy + (this->n>>1) )*(this->n*this->gridCols) + this->n*idx + (this->n>>1)
+ + ((this->n>>1)+this->n*idz)*this->n*this->n*this->gridRows*this->gridCols;
+ if(this->unusedCell[centerGID] == 0 || boundaryCondition == 0)
+ tmp = true;
+ //if(this->currentStep + 3 < getSubgridValue(subGridID))
+ //tmp = true;
+
+
+ double value = sign(u[0]) * u.absMax();
+
+ if(tmp)
+ {}
+
+
+ //north - 1, east - 2, west - 4, south - 8
+ else if(boundaryCondition == 4)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j = 1;j < this->n; j++)
+ for(int k = 0;k < this->n; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[i*this->n]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[i*this->n];
+ }
+ else if(boundaryCondition == 2)
+ {
+ for(int i = 0; i < this->n; i++)
+ for(int j =0 ;j < this->n -1; j++)
+ for(int k = 0;k < this->n; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[(i+1)*this->n - 1]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[(i+1)*this->n - 1];
+ }
+ else if(boundaryCondition == 1)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 0;i < this->n - 1; i++)
+ for(int k = 0;k < this->n; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j + this->n*(this->n - 1)]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[j + this->n*(this->n - 1)];
+ }
+ else if(boundaryCondition == 8)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 1;i < this->n; i++)
+ for(int k = 0;k < this->n; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[j];
+ }
+ else if(boundaryCondition == 16)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 0;i < this->n ; i++)
+ for(int k = 0;k < this->n-1; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j + this->n*(this->n - 1)]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[j + this->n*(this->n - 1)];
+ }
+ else if(boundaryCondition == 32)
+ {
+ for(int j = 0; j < this->n; j++)
+ for(int i = 0;i < this->n; i++)
+ for(int k = 1;k < this->n; k++)
+ //if(fabs(u[i*this->n + j]) < fabs(u[j]))
+ u[k*this->n*this->n + i*this->n + j] = value;// u[j];
+ }
+
+
+ double time = 0.0;
+ double currentTau = this->tau0;
+ double finalTime = this->stopTime;// + 3.0*(u.max() - u.min());
+ if(boundaryCondition == 0) finalTime *= 2.0;
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+
+ double maxResidue( 1.0 );
+ //double lastResidue( 10000.0 );
+ tnlGridEntity<MeshType, 3, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ while( time < finalTime /*|| maxResidue > subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*/)
+ {
+ /****
+ * Compute the RHS
+ */
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+// cout << "i: " << i << ", time: " << time <<endl;
+ Containers::StaticVector<3,int> coords(i % subMesh.getDimensions().x(),
+ (i % (subMesh.getDimensions().x()*subMesh.getDimensions().y())) / subMesh.getDimensions().x(),
+ i / (subMesh.getDimensions().x()*subMesh.getDimensions().y()));
+// cout << "b " << i << " " << i % subMesh.getDimensions().x() << " " << (i % (subMesh.getDimensions().x()*subMesh.getDimensions().y())) << " " << (i % subMesh.getDimensions().x()*subMesh.getDimensions().y()) / subMesh.getDimensions().x() << " " << subMesh.getDimensions().x()*subMesh.getDimensions().y() << " " <<endl;
+ Entity.setCoordinates(coords);
+// cout <<"c" << coords << endl;
+ Entity.refresh();
+// cout << "d" <<endl;
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+// cout << "e" <<endl;
+ fu[ i ] = schemeHost.getValue( this->subMesh, i, coords,u, time, boundaryCondition, neighbourEntities );
+// cout << "f" <<endl;
+ }
+ maxResidue = fu. absMax();
+
+
+ if( this -> cflCondition * maxResidue != 0.0)
+ currentTau = this -> cflCondition / maxResidue;
+
+ /* if (maxResidue < 0.05)
+ cout << "Max < 0.05" << endl;*/
+ if(currentTau > 0.5 * this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >())
+ currentTau = 0.5 * this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ /*if(maxResidue > lastResidue)
+ currentTau *=(1.0/10.0);*/
+
+
+ if( time + currentTau > finalTime ) currentTau = finalTime - time;
+// for( int i = 0; i < fu.getSize(); i ++ )
+// {
+// //cout << "Too big RHS! i = " << i << ", fu = " << fu[i] << ", u = " << u[i] << endl;
+// if((u[i]+currentTau * fu[ i ])*u[i] < 0.0 && fu[i] != 0.0 && u[i] != 0.0 )
+// currentTau = fabs(u[i]/(2.0*fu[i]));
+//
+// }
+
+
+ for( int i = 0; i < fu.getSize(); i ++ )
+ {
+ double add = u[i] + currentTau * fu[ i ];
+ //if( fabs(u[i]) < fabs(add) or (this->subgridValues[subGridID] == this->currentStep +4) )
+ u[ i ] = add;
+ }
+ time += currentTau;
+
+ //cout << '\r' << flush;
+ //cout << maxResidue << " " << currentTau << " @ " << time << flush;
+ //lastResidue = maxResidue;
+ }
+ //cout << "Time: " << time << ", Res: " << maxResidue <<endl;
+ /*if (u.max() > 0.0)
+ this->stopTime /=(double) this->gridCols;*/
+
+// VectorType solution;
+// solution.setLike(u);
+// for( int i = 0; i < u.getSize(); i ++ )
+// {
+// solution[i]=u[i];
+// }
+// return solution;
+ return u;
+}
+
+
+#ifdef HAVE_CUDA
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getSubgridCUDA3D( const int i ,tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+ //int j = threadIdx.x + threadIdx.y * blockDim.x;
+// int index = (blockIdx.z*this->n + threadIdx.z) * this->n*this->n*this->gridCols*this->gridRows
+// + (blockIdx.y) * this->n*this->n*this->gridCols
+// + (blockIdx.x) * this->n
+// + threadIdx.y * this->n*this->gridCols
+// + threadIdx.x;
+
+
+ int index = blockDim.x*blockIdx.x + threadIdx.x +
+ (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x +
+ (blockDim.z*blockIdx.z + threadIdx.z)*blockDim.x*gridDim.x*blockDim.y*gridDim.y;
+
+ //printf("i= %d,j= %d,th= %d\n",i,j,th);
+ *a = caller->work_u_cuda[index];
+ //printf("Hi %f \n", *a);
+ //return ret;
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::updateSubgridCUDA3D( const int i ,tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* caller, double* a)
+{
+// int j = threadIdx.x + threadIdx.y * blockDim.x;
+// int index = (blockIdx.z*this->n + threadIdx.z) * this->n*this->n*this->gridCols*this->gridRows
+// + (blockIdx.y) * this->n*this->n*this->gridCols
+// + (blockIdx.x) * this->n
+// + threadIdx.y * this->n*this->gridCols
+// + threadIdx.x;
+
+ int index = blockDim.x*blockIdx.x + threadIdx.x +
+ (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x +
+ (blockDim.z*blockIdx.z + threadIdx.z)*blockDim.x*gridDim.x*blockDim.y*gridDim.y;
+
+ if( (fabs(caller->work_u_cuda[index]) > fabs(*a)) || (caller->unusedCell_cuda[index] == 1) )
+ {
+ caller->work_u_cuda[index] = *a;
+ caller->unusedCell_cuda[index] = 0;
+
+ }
+
+ *a = caller->work_u_cuda[index];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::insertSubgridCUDA3D( double u, const int i )
+{
+
+
+// int j = threadIdx.x + threadIdx.y * blockDim.x;
+ //printf("j = %d, u = %f\n", j,u);
+
+// int index = (blockIdx.z*this->n + threadIdx.z) * this->n*this->n*this->gridCols*this->gridRows
+// + (blockIdx.y) * this->n*this->n*this->gridCols
+// + (blockIdx.x) * this->n
+// + threadIdx.y * this->n*this->gridCols
+// + threadIdx.x;
+
+ int index = blockDim.x*blockIdx.x + threadIdx.x +
+ (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x +
+ (blockDim.z*blockIdx.z + threadIdx.z)*blockDim.x*gridDim.x*blockDim.y*gridDim.y;
+
+ //printf("i= %d,j= %d,index= %d\n",i,j,index);
+ if( (fabs(this->work_u_cuda[index]) > fabs(u)) || (this->unusedCell_cuda[index] == 1) )
+ {
+ this->work_u_cuda[index] = u;
+ this->unusedCell_cuda[index] = 0;
+
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::runSubgridCUDA3D( int boundaryCondition, double* u, int subGridID)
+{
+
+ __shared__ int tmp;
+ __shared__ double value;
+ //double tmpRes = 0.0;
+ volatile double* sharedTau = &u[blockDim.x*blockDim.y*blockDim.z];
+// volatile double* absVal = &u[2*blockDim.x*blockDim.y*blockDim.z];
+ int i = threadIdx.x;
+ int j = threadIdx.y;
+ int k = threadIdx.z;
+ int l = threadIdx.x + threadIdx.y * blockDim.x + threadIdx.z*blockDim.x*blockDim.y;
+ bool computeFU = !((i == 0 && (boundaryCondition & 4)) or
+ (i == blockDim.x - 1 && (boundaryCondition & 2)) or
+ (j == 0 && (boundaryCondition & 8)) or
+ (j == blockDim.y - 1 && (boundaryCondition & 1))or
+ (k == 0 && (boundaryCondition & 32)) or
+ (k == blockDim.z - 1 && (boundaryCondition & 16)));
+
+ if(l == 0)
+ {
+ tmp = 0;
+ int centerGID = (blockDim.y*blockIdx.y + (blockDim.y>>1) )*(blockDim.x*gridDim.x) + blockDim.x*blockIdx.x + (blockDim.x>>1)
+ + ((blockDim.z>>1)+blockDim.z*blockIdx.z)*blockDim.x*blockDim.y*gridDim.x*gridDim.y;
+ if(this->unusedCell_cuda[centerGID] == 0 || boundaryCondition == 0)
+ tmp = 1;
+ }
+ __syncthreads();
+
+
+ __syncthreads();
+ if(tmp !=1)
+ {
+// if(computeFU)
+// absVal[l]=0.0;
+// else
+// absVal[l] = fabs(u[l]);
+//
+// __syncthreads();
+//
+// if((blockDim.x == 16) && (l < 128)) absVal[l] = Max(absVal[l],absVal[l+128]);
+// __syncthreads();
+// if((blockDim.x == 16) && (l < 64)) absVal[l] = Max(absVal[l],absVal[l+64]);
+// __syncthreads();
+// if(l < 32) absVal[l] = Max(absVal[l],absVal[l+32]);
+// if(l < 16) absVal[l] = Max(absVal[l],absVal[l+16]);
+// if(l < 8) absVal[l] = Max(absVal[l],absVal[l+8]);
+// if(l < 4) absVal[l] = Max(absVal[l],absVal[l+4]);
+// if(l < 2) absVal[l] = Max(absVal[l],absVal[l+2]);
+// if(l < 1) value = sign(u[0])*Max(absVal[l],absVal[l+1]);
+// __syncthreads();
+//
+// if(computeFU)
+// u[l] = value;
+ if(computeFU)
+ {
+ tnlGridEntity<MeshType, 3, tnlGridEntityNoStencilStorage > Ent(subMesh);
+ if(boundaryCondition == 4)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(0,j,k));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(threadIdx.x) ;//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(threadIdx.x+this->n);
+ }
+ else if(boundaryCondition == 2)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(blockDim.x - 1,j,k));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(this->n - 1 - threadIdx.x);//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(blockDim.x - threadIdx.x - 1+this->n);
+ }
+ else if(boundaryCondition == 8)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(i,0,k));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 0, 1, 0 >()*(threadIdx.y) ;//+ 2*sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(threadIdx.y+this->n);
+ }
+ else if(boundaryCondition == 1)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(i,blockDim.y - 1,k));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 0, 1, 0 >()*(this->n - 1 - threadIdx.y) ;//+ sign(u[0])*this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*(blockDim.y - threadIdx.y - 1 +this->n);
+ }
+ else if(boundaryCondition == 32)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(i,j,0));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 0, 0, 1 >()*(threadIdx.z);
+ }
+ else if(boundaryCondition == 16)
+ {
+ Ent.setCoordinates(Containers::StaticVector<3,int>(i,j,blockDim.z - 1));
+ Ent.refresh();
+ u[l] = u[Ent.getIndex()];// + sign(u[0])*this->subMesh.template getSpaceStepsProducts< 0, 0, 1 >()*(this->n - 1 - threadIdx.z) ;
+ }
+ }
+ }
+
+ double time = 0.0;
+ __shared__ double currentTau;
+ double cfl = this->cflCondition;
+ double fu = 0.0;
+// if(threadIdx.x * threadIdx.y * threadIdx.z == 0)
+// {
+// currentTau = this->tau0;
+// }
+ double finalTime = this->stopTime;
+ __syncthreads();
+ if( boundaryCondition == 0 ) finalTime *= 2.0;
+
+ tnlGridEntity<MeshType, 3, tnlGridEntityNoStencilStorage > Entity(subMesh);
+ tnlNeighbourGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighbourEntities(Entity);
+ Entity.setCoordinates(Containers::StaticVector<3,int>(i,j,k));
+ Entity.refresh();
+ neighbourEntities.refresh(subMesh,Entity.getIndex());
+
+
+ while( time < finalTime )
+ {
+ sharedTau[l]=finalTime;
+
+ if(computeFU)
+ {
+ fu = schemeHost.getValueDev( this->subMesh, l, Containers::StaticVector<3,int>(i,j,k), u, time, boundaryCondition, neighbourEntities);
+ if(abs(fu) > 0.0)
+ sharedTau[l]=abs(cfl/fu);
+ }
+
+ if(l == 0)
+ {
+ if(sharedTau[0] > 0.5 * this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >()) sharedTau[0] = 0.5 * this->subMesh.template getSpaceStepsProducts< 1, 0, 0 >();
+ }
+ else if(l == blockDim.x*blockDim.y*blockDim.z - 1)
+ {
+ if( time + sharedTau[l] > finalTime ) sharedTau[l] = finalTime - time;
+ }
+
+ __syncthreads();
+ if(l < 256) sharedTau[l] = Min(sharedTau[l],sharedTau[l+256]);
+ __syncthreads();
+ if(l < 128) sharedTau[l] = Min(sharedTau[l],sharedTau[l+128]);
+ __syncthreads();
+ if(l < 64) sharedTau[l] = Min(sharedTau[l],sharedTau[l+64]);
+ __syncthreads();
+ if(l < 32) sharedTau[l] = Min(sharedTau[l],sharedTau[l+32]);
+ __syncthreads();
+ if(l < 16) sharedTau[l] = Min(sharedTau[l],sharedTau[l+16]);
+ if(l < 8) sharedTau[l] = Min(sharedTau[l],sharedTau[l+8]);
+ if(l < 4) sharedTau[l] = Min(sharedTau[l],sharedTau[l+4]);
+ if(l < 2) sharedTau[l] = Min(sharedTau[l],sharedTau[l+2]);
+ if(l < 1) currentTau = Min(sharedTau[l],sharedTau[l+1]);
+ __syncthreads();
+
+// if(abs(fu) < 10000.0)
+// printf("bla");
+ if(computeFU)
+ u[l] += currentTau * fu;
+ time += currentTau;
+ __syncthreads();
+ }
+
+
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getOwnerCUDA3D(int i) const
+{
+ int j = i % (this->gridCols*this->gridRows*this->n*this->n);
+
+ return ( (i / (this->gridCols*this->gridRows*this->n*this->n))*this->gridCols*this->gridRows
+ + (j / (this->gridCols*this->n*this->n))*this->gridCols
+ + (j % (this->gridCols*this->n))/this->n);
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getSubgridValueCUDA3D( int i ) const
+{
+ return this->subgridValues_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::setSubgridValueCUDA3D(int i, int value)
+{
+ this->subgridValues_cuda[i] = value;
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+int tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::getBoundaryConditionCUDA3D( int i ) const
+{
+ return this->boundaryConditions_cuda[i];
+}
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__device__
+void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::setBoundaryConditionCUDA3D(int i, int value)
+{
+ this->boundaryConditions_cuda[i] = value;
+}
+
+
+
+//north - 1, east - 2, west - 4, south - 8, up -16, down - 32
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void /*tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::*/synchronizeCUDA3D(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver) //needs fix ---- maybe not anymore --- but frankly: yeah, it does -- aaaa-and maybe fixed now
+{
+
+ __shared__ int boundary[6]; // north,east,west,south
+ __shared__ int subgridValue;
+ __shared__ int newSubgridValue;
+
+
+ int gid = blockDim.x*blockIdx.x + threadIdx.x +
+ (blockDim.y*blockIdx.y + threadIdx.y)*blockDim.x*gridDim.x +
+ (blockDim.z*blockIdx.z + threadIdx.z)*blockDim.x*gridDim.x*blockDim.y*gridDim.y;
+ double u = cudaSolver->work_u_cuda[gid];
+ double u_cmp;
+ int subgridValue_cmp=INT_MAX;
+ int boundary_index=0;
+
+
+ if(threadIdx.x+threadIdx.y+threadIdx.z == 0)
+ {
+ subgridValue = cudaSolver->getSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y);
+ boundary[0] = 0;
+ boundary[1] = 0;
+ boundary[2] = 0;
+ boundary[3] = 0;
+ boundary[4] = 0;
+ boundary[5] = 0;
+ newSubgridValue = 0;
+// printf("aaa z = %d, y = %d, x = %d\n",blockIdx.z,blockIdx.y,blockIdx.x);
+ }
+ __syncthreads();
+
+
+
+ if( (threadIdx.x == 0 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == 0 /* && (cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.z == 0 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.x == blockDim.x - 1 /* && !(cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.y == blockDim.y - 1 /* && (cudaSolver->currentStep & 1)*/) ||
+ (threadIdx.z == blockDim.z - 1 /* && (cudaSolver->currentStep & 1)*/) )
+ {
+ if(threadIdx.x == 0 && (blockIdx.x != 0)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y - 1);
+ boundary_index = 2;
+ }
+
+ if(threadIdx.x == blockDim.x - 1 && (blockIdx.x != gridDim.x - 1)/* && !(cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + 1];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y + 1);
+ boundary_index = 1;
+ }
+
+ __threadfence();
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ u=u_cmp;
+ }
+ __threadfence();
+ if(threadIdx.y == 0 && (blockIdx.y != 0)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D((blockIdx.y - 1)*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y);
+ boundary_index = 3;
+ }
+ if(threadIdx.y == blockDim.y - 1 && (blockIdx.y != gridDim.y - 1)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + blockDim.x*gridDim.x];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D((blockIdx.y + 1)*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y);
+ boundary_index = 0;
+ }
+
+ __threadfence();
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ u=u_cmp;
+ }
+ __threadfence();
+
+ if(threadIdx.z == 0 && (blockIdx.z != 0)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid - blockDim.x*gridDim.x*blockDim.y*gridDim.y];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + (blockIdx.z - 1)*gridDim.x*gridDim.y);
+ boundary_index = 5;
+ }
+ if(threadIdx.z == blockDim.z - 1 && (blockIdx.z != gridDim.z - 1)/* && (cudaSolver->currentStep & 1)*/)
+ {
+ u_cmp = cudaSolver->work_u_cuda[gid + blockDim.x*gridDim.x*blockDim.y*gridDim.y];
+ subgridValue_cmp = cudaSolver->getSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + (blockIdx.z + 1)*gridDim.x*gridDim.y);
+ boundary_index = 4;
+ }
+ __threadfence();
+
+ if((subgridValue == INT_MAX || fabs(u_cmp) + cudaSolver->delta < fabs(u) ) && (subgridValue_cmp != INT_MAX && subgridValue_cmp != -INT_MAX))
+ {
+ cudaSolver->unusedCell_cuda[gid] = 0;
+ atomicMax(&newSubgridValue, INT_MAX);
+ atomicMax(&boundary[boundary_index], 1);
+ cudaSolver->work_u_cuda[gid] = u_cmp;
+ }
+ __threadfence();
+
+ }
+ __syncthreads();
+
+ if(threadIdx.x+threadIdx.y+threadIdx.z == 0)
+ {
+
+ if(subgridValue == INT_MAX && newSubgridValue != 0)
+ cudaSolver->setSubgridValueCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y, -INT_MAX);
+
+ cudaSolver->setBoundaryConditionCUDA3D(blockIdx.y*gridDim.x + blockIdx.x + blockIdx.z*gridDim.x*gridDim.y, 1 * boundary[0] +
+ 2 * boundary[1] +
+ 4 * boundary[2] +
+ 8 * boundary[3] +
+ 16 * boundary[4] +
+ 32 * boundary[5] );
+ if(blockIdx.x+blockIdx.y+blockIdx.z == 0)
+ {
+ cudaSolver->currentStep = cudaSolver->currentStep + 1;
+ *(cudaSolver->runcuda) = 0;
+ }
+ }
+}
+
+
+
+template <typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void synchronize2CUDA3D(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver)
+{
+ int stepValue = cudaSolver->currentStep + 4;
+ if( cudaSolver->getSubgridValueCUDA3D(blockIdx.z*gridDim.x*gridDim.y + blockIdx.y*gridDim.x + blockIdx.x) == -INT_MAX )
+ cudaSolver->setSubgridValueCUDA3D(blockIdx.z*gridDim.x*gridDim.y + blockIdx.y*gridDim.x + blockIdx.x, stepValue);
+
+ atomicMax((cudaSolver->runcuda),cudaSolver->getBoundaryConditionCUDA3D(blockIdx.z*gridDim.x*gridDim.y + blockIdx.y*gridDim.x + blockIdx.x));
+}
+
+
+
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device>
+__global__
+void initCUDA3D( tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* cudaSolver, double* ptr , int* ptr2, int* ptr3)
+{
+
+
+ cudaSolver->work_u_cuda = ptr;//(double*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->n*cudaSolver->n*sizeof(double));
+ cudaSolver->unusedCell_cuda = ptr3;//(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->n*cudaSolver->n*sizeof(int));
+ cudaSolver->subgridValues_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->gridLevels*sizeof(int));
+ cudaSolver->boundaryConditions_cuda =(int*)malloc(cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->gridLevels*sizeof(int));
+ cudaSolver->runcuda = ptr2;//(bool*)malloc(sizeof(bool));
+ *(cudaSolver->runcuda) = 1;
+ cudaSolver->currentStep = 1;
+ //cudaMemcpy(ptr,&(cudaSolver->work_u_cuda), sizeof(double*),cudaMemcpyDeviceToHost);
+ //ptr = cudaSolver->work_u_cuda;
+ printf("GPU memory allocated.\n");
+
+ for(int i = 0; i < cudaSolver->gridCols*cudaSolver->gridRows*cudaSolver->gridLevels; i++)
+ {
+ cudaSolver->subgridValues_cuda[i] = INT_MAX;
+ cudaSolver->boundaryConditions_cuda[i] = 0;
+ }
+
+ /*for(long int j = 0; j < cudaSolver->n*cudaSolver->n*cudaSolver->gridCols*cudaSolver->gridRows; j++)
+ {
+ printf("%d\n",j);
+ cudaSolver->unusedCell_cuda[ j] = 1;
+ }*/
+ printf("GPU memory initialized.\n");
+}
+
+
+
+
+//extern __shared__ double array[];
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void initRunCUDA3D(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* caller)
+
+{
+
+
+ extern __shared__ double u[];
+
+ int i = blockIdx.z * gridDim.x * gridDim.y + blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.z * blockDim.x * blockDim.y + threadIdx.y * blockDim.x + threadIdx.x;
+
+ __shared__ int containsCurve;
+ if(l == 0)
+ {
+// printf("z = %d, y = %d, x = %d\n",blockIdx.z,blockIdx.y,blockIdx.x);
+ containsCurve = 0;
+ }
+
+ caller->getSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ if(u[0] * u[l] <= 0.0)
+ {
+ atomicMax( &containsCurve, 1);
+ }
+
+ __syncthreads();
+ if(containsCurve == 1)
+ {
+ caller->runSubgridCUDA3D(0,u,i);
+ __syncthreads();
+// caller->insertSubgridCUDA3D(u[l],i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+
+ __syncthreads();
+ if(l == 0)
+ caller->setSubgridValueCUDA3D(i, 4);
+ }
+
+
+}
+
+
+
+
+
+template< typename SchemeHost, typename SchemeDevice, typename Device >
+__global__
+void runCUDA3D(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >* caller)
+{
+ extern __shared__ double u[];
+ int i = blockIdx.z * gridDim.x * gridDim.y + blockIdx.y * gridDim.x + blockIdx.x;
+ int l = threadIdx.z * blockDim.x * blockDim.y + threadIdx.y * blockDim.x + threadIdx.x;
+ int bound = caller->getBoundaryConditionCUDA3D(i);
+
+ if(caller->getSubgridValueCUDA3D(i) != INT_MAX && bound != 0 && caller->getSubgridValueCUDA3D(i) > 0)
+ {
+ caller->getSubgridCUDA3D(i,caller, &u[l]);
+
+ //if(l == 0)
+ //printf("i = %d, bound = %d\n",i,caller->getSubgridValueCUDA3D(i));
+ if(caller->getSubgridValueCUDA3D(i) == caller->currentStep+4)
+ {
+ if(bound & 1)
+ {
+ caller->runSubgridCUDA3D(1,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 2 )
+ {
+ caller->runSubgridCUDA3D(2,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 4)
+ {
+ caller->runSubgridCUDA3D(4,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 8)
+ {
+ caller->runSubgridCUDA3D(8,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 16)
+ {
+ caller->runSubgridCUDA3D(16,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound & 32)
+ {
+ caller->runSubgridCUDA3D(32,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+ }
+ else
+ {
+ if( ((bound == 2)))
+ {
+ caller->runSubgridCUDA3D(2,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound == 1) ))
+ {
+ caller->runSubgridCUDA3D(1,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound == 8) ))
+ {
+ caller->runSubgridCUDA3D(8,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if((bound == 4))
+ {
+ caller->runSubgridCUDA3D(4,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound == 16)
+ {
+ caller->runSubgridCUDA3D(16,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if(bound == 32)
+ {
+ caller->runSubgridCUDA3D(32,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ }
+ /* 1 2 4 8 16 32 */
+
+ if( ((bound & 19 ))) /* 1 1 0 0 1 0 */
+ {
+ caller->runSubgridCUDA3D(19,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 21 ))) /* 1 0 1 0 1 0 */
+ {
+ caller->runSubgridCUDA3D(21,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 26 ))) /* 0 1 0 1 1 0 */
+ {
+ caller->runSubgridCUDA3D(26,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( (bound & 28 )) /* 0 0 1 1 1 0 */
+ {
+ caller->runSubgridCUDA3D(28,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+
+
+ if( ((bound & 35 ))) /* 1 0 1 0 0 1 */
+ {
+ caller->runSubgridCUDA3D(35,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 37 ))) /* 1 0 1 0 0 1 */
+ {
+ caller->runSubgridCUDA3D(37,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( ((bound & 42 ))) /* 0 1 0 1 0 1 */
+ {
+ caller->runSubgridCUDA3D(42,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+ if( (bound & 44 )) /* 0 0 1 1 0 1 */
+ {
+ caller->runSubgridCUDA3D(44,u,i);
+ caller->updateSubgridCUDA3D(i,caller, &u[l]);
+ __syncthreads();
+ }
+
+ if(l==0)
+ {
+ caller->setBoundaryConditionCUDA3D(i, 0);
+ caller->setSubgridValueCUDA3D(i, caller->getSubgridValueCUDA3D(i) - 1 );
+ }
+
+
+ }
+
+
+
+}
+
+#endif /*HAVE_CUDA*/
+
+#endif /* TNLPARALLELEIKONALSOLVER3D_IMPL_H_ */
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/CMakeLists.txt b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..04348b8b5baa4bdbd772b67feaf9da3b32aed53d
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/CMakeLists.txt
@@ -0,0 +1,21 @@
+set( tnl_hamilton_jacobi_SOURCES
+ hamiltonJacobiProblemSetter.h
+ hamiltonJacobiProblemSetter_impl.h
+ hamiltonJacobiProblemSolver.h
+ hamiltonJacobiProblemSolver_impl.h
+ main.cpp
+ hamiltonJacobiProblemConfig.h
+ tnl-direct-eikonal-solver.h )
+
+ADD_EXECUTABLE(tnl-hamilton-jacobi${debugExt} main.cpp)
+target_link_libraries (tnl-hamilton-jacobi${debugExt} tnl${debugExt}-${tnlVersion} )
+
+ADD_EXECUTABLE(tnl-direct-eikonal-solver${debugExt} tnl-direct-eikonal-solver.cpp )
+target_link_libraries (tnl-direct-eikonal-solver${debugExt} tnl${debugExt}-${tnlVersion} )
+
+
+INSTALL( TARGETS tnl-hamilton-jacobi${debugExt}
+ tnl-direct-eikonal-solver${debugExt}
+ RUNTIME DESTINATION bin
+ PERMISSIONS OWNER_READ OWNER_WRITE OWNER_EXECUTE GROUP_READ GROUP_EXECUTE WORLD_READ WORLD_EXECUTE )
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem.h
new file mode 100644
index 0000000000000000000000000000000000000000..a4182050ac311e21be33388ff22fded1428a88a0
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem.h
@@ -0,0 +1,97 @@
+/***************************************************************************
+ hamiltonJacobiProblem.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <problems/tnlPDEProblem.h>
+#include <solvers/preconditioners/tnlDummyPreconditioner.h>
+#include <solvers/tnlSolverMonitor.h>
+#include <core/tnlLogger.h>
+#include <core/vectors/tnlVector.h>
+#include <solvers/pde/tnlExplicitUpdater.h>
+#include <solvers/pde/tnlLinearSystemAssembler.h>
+#include <functions/tnlMeshFunction.h>
+
+template< typename Mesh,
+ typename DifferentialOperator,
+ typename BoundaryCondition,
+ typename RightHandSide>
+class HamiltonJacobiProblem : public tnlPDEProblem< Mesh,
+ TimeDependentProblem,
+ typename DifferentialOperator::RealType,
+ typename Mesh::DeviceType,
+ typename DifferentialOperator::IndexType >
+{
+ public:
+
+ typedef typename DifferentialOperator::RealType RealType;
+ typedef typename Mesh::DeviceType DeviceType;
+ typedef typename DifferentialOperator::IndexType IndexType;
+
+ typedef tnlMeshFunction< Mesh > MeshFunctionType;
+ typedef tnlPDEProblem< Mesh, TimeDependentProblem, RealType, DeviceType, IndexType > BaseType;
+
+ using typename BaseType::MeshType;
+ using typename BaseType::DofVectorType;
+ using typename BaseType::MeshDependentDataType;
+
+ static String getTypeStatic();
+
+ String getPrologHeader() const;
+
+ void writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters ) const;
+
+ bool setup( const Config::ParameterContainer& parameters );
+
+ bool setInitialCondition( const Config::ParameterContainer& parameters,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshDependentData );
+
+ bool makeSnapshot( const RealType& time,
+ const IndexType& step,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshDependentData );
+
+ IndexType getDofs( const MeshType& mesh ) const;
+
+ void bindDofs( const MeshType& mesh,
+ DofVectorType& dofs );
+
+ void getExplicitRHS( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ DofVectorType& _u,
+ DofVectorType& _fu,
+ MeshDependentDataType& meshDependentData );
+
+ protected:
+
+ MeshFunctionType solution;
+
+ tnlExplicitUpdater< Mesh, MeshFunctionType, DifferentialOperator, BoundaryCondition, RightHandSide > explicitUpdater;
+
+ DifferentialOperator differentialOperator;
+
+ BoundaryCondition boundaryCondition;
+
+ RightHandSide rightHandSide;
+};
+
+#include "HamiltonJacobiProblem_impl.h"
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..4e84c7d98966b3e35d86e263850a7df14d85ff13
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemConfig.h
@@ -0,0 +1,37 @@
+/***************************************************************************
+ hamiltonJacobiProblemConfig.h - description
+ -------------------
+ begin : Oct 5, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ email :
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <config/tnlConfigDescription.h>
+
+template< typename ConfigTag >
+class HamiltonJacobiProblemConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Hamilton-Jacobi solver settings:" );
+ config.addEntry < String > ( "problem-name", "This defines particular problem.", "hamilton-jacobi" );
+ config.addEntry < String > ( "scheme", "This defines scheme used for discretization.", "godunov" );
+ config.addEntryEnum( "godunov" );
+ config.addEntryEnum( "upwind" );
+ config.addEntry < double > ( "epsilon", "This defines epsilon for smoothening of sign().", 3.0 );
+ config.addEntry < double > ( "-value", "Constant value of RHS.", 0.0 );
+ }
+};
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter.h
new file mode 100644
index 0000000000000000000000000000000000000000..ce4209736e95d6e58e410cc2acef09c1542270b8
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter.h
@@ -0,0 +1,35 @@
+/***************************************************************************
+ hamiltonJacobiProblemSetter.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include "HamiltonJacobiProblem.h"
+
+template< typename RealType,
+ typename DeviceType,
+ typename IndexType,
+ typename MeshType,
+ typename ConfigTag,
+ typename SolverStarter >
+class HamiltonJacobiProblemSetter
+{
+ public:
+ static bool run( const Config::ParameterContainer& parameters );
+};
+
+#include "HamiltonJacobiProblemSetter_impl.h"
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..af1d427597ffe71d606bd3ca60f206470dcf2c92
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblemSetter_impl.h
@@ -0,0 +1,75 @@
+/***************************************************************************
+ hamiltonJacobiProblemSetter_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <mesh/tnlGrid.h>
+#include <functions/tnlConstantFunction.h>
+#include <operators/tnlNeumannBoundaryConditions.h>
+#include <operators/tnlDirichletBoundaryConditions.h>
+#include <operators/hamilton-jacobi/upwindEikonal.h>
+#include <operators/hamilton-jacobi/godunovEikonal.h>
+#include <operators/hamilton-jacobi/tnlEikonalOperator.h>
+
+
+template< typename RealType,
+ typename DeviceType,
+ typename IndexType,
+ typename MeshType,
+ typename ConfigTag,
+ typename SolverStarter >
+bool HamiltonJacobiProblemSetter< RealType, DeviceType, IndexType, MeshType, ConfigTag, SolverStarter > :: run( const Config::ParameterContainer& parameters )
+{
+ static const int Dimensions = MeshType::getMeshDimensions();
+
+ if( Dimensions <= 0 || Dimensions > 3 )
+ {
+ cerr << "The problem is not defined for " << Dimensions << "dimensions." << endl;
+ return false;
+ }
+ else
+ {
+ typedef Containers::StaticVector < Dimensions, RealType > Vertex;
+ typedef tnlConstantFunction< Dimensions, RealType > ConstantFunctionType;
+ typedef tnlNeumannBoundaryConditions< MeshType, ConstantFunctionType, RealType, IndexType > BoundaryConditions;
+
+ SolverStarter solverStarter;
+
+ const String& schemeName = parameters.getParameter< String >( "scheme" );
+
+ if( schemeName == "upwind" )
+ {
+ typedef upwindEikonalScheme< MeshType, RealType, IndexType > GradientNormOperator;
+ typedef tnlConstantFunction< Dimensions, RealType > RightHandSide;
+ typedef tnlEikonalOperator< GradientNormOperator, RightHandSide > Operator;
+ typedef HamiltonJacobiProblem< MeshType, Operator, BoundaryConditions, RightHandSide > Solver;
+ return solverStarter.template run< Solver >( parameters );
+ }
+ if( schemeName == "godunov" )
+ {
+ typedef godunovEikonalScheme< MeshType, RealType, IndexType > GradientNormOperator;
+ typedef tnlConstantFunction< Dimensions, RealType > RightHandSide;
+ typedef tnlEikonalOperator< GradientNormOperator, RightHandSide > Operator;
+ typedef HamiltonJacobiProblem< MeshType, Operator, BoundaryConditions, RightHandSide > Solver;
+ return solverStarter.template run< Solver >( parameters );
+ }
+ else
+ cerr << "Unknown scheme '" << schemeName << "'." << endl;
+
+
+ return false;
+ }
+}
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..9917205ead68b98de8b55d8423042a99fb0848dd
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/HamiltonJacobiProblem_impl.h
@@ -0,0 +1,199 @@
+/***************************************************************************
+ HamiltonJacobiProblem_impl.h - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <core/mfilename.h>
+#include <matrices/tnlMatrixSetter.h>
+#include <exception>
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+String HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide > :: getTypeStatic()
+{
+ return String( "hamiltonJacobiSolver< " ) + Mesh :: getTypeStatic() + " >";
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+String HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide > :: getPrologHeader() const
+{
+ return String( "Hamilton-Jacobi" );
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+void HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide > :: writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters ) const
+{
+ //logger. WriteParameter< typename String >( "Problem name:", "problem-name", parameters );
+ //logger. WriteParameter< String >( "Used scheme:", "scheme", parameters );
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+bool HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide > :: setup( const Config::ParameterContainer& parameters )
+{
+ this->boundaryCondition.getFunction().setConstant( 1.0 );
+ //this->rightHandSide.getFunction().setConstant( 0.0 );
+ //return true;
+/*
+ const String& problemName = parameters. GetParameter< String >( "problem-name" );
+
+ this->schemeTest = parameters. GetParameter< int >( "scheme-test" );
+ this->tested = false;
+
+ const String& meshFile = parameters.GetParameter< String >( "mesh" );
+ if( ! this->mesh.load( meshFile ) )
+ {
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ return false;
+ }
+
+ const IndexType& dofs = this->mesh.getDofs();
+ dofVector. setSize( dofs );
+
+ this -> u. bind( & dofVector. getData()[ 0 * dofs ], dofs );
+ this -> v. bind( & dofVector. getData()[ 1 * dofs ], dofs );
+
+*/
+ differentialOperator.getAnisotropy().setConstant( 1.0 ); //setup( parameters );
+ differentialOperator.setSmoothing( 1.0 );
+ return true;
+
+}
+
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+typename HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::IndexType
+ HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::getDofs( const MeshType& mesh ) const
+{
+ /****
+ * Set-up DOFs and supporting grid functions
+ */
+ return mesh.template getEntitiesCount< typename MeshType::Cell >();
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+void HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::
+bindDofs( const MeshType& mesh,
+ DofVectorType& dofVector )
+{
+ //const IndexType dofs = mesh.template getEntitiesCount< typename MeshType::Cell >();
+ this->solution.bind( mesh, dofVector );
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+bool
+HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::
+setInitialCondition( const Config::ParameterContainer& parameters,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshDependentData )
+{
+ this->bindDofs( mesh, dofs );
+ const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );
+ if( ! this->solution.boundLoad( initialConditionFile ) )
+ {
+ cerr << "I am not able to load the initial condition from the file " << initialConditionFile << "." << endl;
+ return false;
+ }
+ return true;
+}
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+bool
+HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::
+makeSnapshot( const RealType& time,
+ const IndexType& step,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshDependentData )
+{
+ cout << endl << "Writing output at time " << time << " step " << step << "." << endl;
+
+ String fileName;
+ FileNameBaseNumberEnding( "u-", step, 5, ".tnl", fileName );
+ if( ! this->solution.save( fileName ) )
+ return false;
+
+ return true;
+}
+
+
+template< typename Mesh,
+ typename HamiltonJacobi,
+ typename BoundaryCondition,
+ typename RightHandSide>
+void
+HamiltonJacobiProblem< Mesh,HamiltonJacobi,BoundaryCondition,RightHandSide >::
+getExplicitRHS( const RealType& time,
+ const RealType& tau,
+ const MeshType& mesh,
+ DofVectorType& _u,
+ DofVectorType& _fu,
+ MeshDependentDataType& meshDependentData )
+{
+
+ /*
+ if(!(this->schemeTest))
+ scheme.GetExplicitRHS(time, tau, _u, _fu);
+ else if(!(this->tested))
+ {
+ this->tested = true;
+ DofVectorType tmp;
+ if(tmp.setLike(_u))
+ tmp = _u;
+ scheme.GetExplicitRHS(time, tau, tmp, _u);
+
+ }
+ */
+
+ //this->bindDofs( mesh, _u );
+ MeshFunctionType u, fu;
+ u.bind( mesh, _u );
+ fu.bind( mesh, _fu );
+ explicitUpdater.template update< typename MeshType::Cell >( time,
+ mesh,
+ this->differentialOperator,
+ this->boundaryCondition,
+ this->rightHandSide,
+ u,
+ fu );
+ //fu.save( "fu.tnl" );
+ //std::cerr << "Enter." << std::endl;
+ //getchar();
+}
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/MainBuildConfig.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/MainBuildConfig.h
new file mode 100644
index 0000000000000000000000000000000000000000..1b904c0c8b1d096a72a6ee8c3cc3cd1979d164b4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/MainBuildConfig.h
@@ -0,0 +1,64 @@
+/***************************************************************************
+ MainBuildConfig.h - description
+ -------------------
+ begin : Jul 7, 2014
+ copyright : (C) 2014 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef MAINBUILDCONFIG_H_
+#define MAINBUILDCONFIG_H_
+
+#include <solvers/tnlBuildConfigTags.h>
+
+class MainBuildConfig
+{
+ public:
+
+ static void print() { cerr << "MainBuildConfig" << endl; }
+};
+
+/****
+ * Turn off support for float and long double.
+ */
+template<> struct tnlConfigTagReal< MainBuildConfig, float > { enum { enabled = false }; };
+template<> struct tnlConfigTagReal< MainBuildConfig, long double > { enum { enabled = false }; };
+
+/****
+ * Turn off support for short int and long int indexing.
+ */
+template<> struct tnlConfigTagIndex< MainBuildConfig, short int >{ enum { enabled = false }; };
+template<> struct tnlConfigTagIndex< MainBuildConfig, long int >{ enum { enabled = false }; };
+
+/****
+ * Use of tnlGrid is enabled for allowed dimensions and Real, Device and Index types.
+ */
+template< int Dimensions, typename Real, typename Device, typename Index >
+ struct tnlConfigTagMesh< MainBuildConfig, tnlGrid< Dimensions, Real, Device, Index > >
+ { enum { enabled = tnlConfigTagDimensions< MainBuildConfig, Dimensions >::enabled &&
+ tnlConfigTagReal< MainBuildConfig, Real >::enabled &&
+ tnlConfigTagDevice< MainBuildConfig, Device >::enabled &&
+ tnlConfigTagIndex< MainBuildConfig, Index >::enabled }; };
+
+/****
+ * Please, chose your preferred time discretisation here.
+ */
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlExplicitTimeDiscretisationTag >{ enum { enabled = true }; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlSemiImplicitTimeDiscretisationTag >{ enum { enabled = false}; };
+template<> struct tnlConfigTagTimeDiscretisation< MainBuildConfig, tnlImplicitTimeDiscretisationTag >{ enum { enabled = false }; };
+
+/****
+ * Only the Runge-Kutta-Merson solver is enabled by default.
+ */
+template<> struct tnlConfigTagExplicitSolver< MainBuildConfig, tnlExplicitEulerSolverTag >{ enum { enabled = false }; };
+
+#endif /* MAINBUILDCONFIG_H_ */
diff --git a/examples/mean-curvature-flow/Makefile b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/Makefile
similarity index 55%
rename from examples/mean-curvature-flow/Makefile
rename to src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/Makefile
index 71873be3fbafbf131bd8b092c7078a9dea9b167d..03620692c53d343adcb0e73060dba775223b1c7c 100644
--- a/examples/mean-curvature-flow/Makefile
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/Makefile
@@ -1,36 +1,41 @@
-TNL_VERSION=0.1
-TNL_INSTALL_DIR=${HOME}/local/lib
-TNL_INCLUDE_DIR=${HOME}/local/include/tnl-${TNL_VERSION}
-
-TARGET = mean-curvature-flow-eoc
-INSTALL_DIR = ${HOME}/local
-CXX = g++
-CUDA_CXX = nvcc
-CXX_FLAGS = -std=gnu++0x -I$(TNL_INCLUDE_DIR) -O0 -g
-LD_FLAGS = -L$(TNL_INSTALL_DIR) -ltnl-dbg-0.1
-
-SOURCES = tnl-mean-curvature-flow-eoc.cpp
-HEADERS =
-OBJECTS = tnl-mean-curvature-flow-eoc.o
-DIST = $(SOURCES) Makefile
-
-all: $(TARGET)
-clean:
- rm -f $(OBJECTS)
-
-dist: $(DIST)
- tar zcvf $(TARGET).tgz $(DIST)
-
-install: $(TARGET)
- cp $(TARGET) $(INSTALL_DIR)/bin
- cp $(INSTALL_DIR)/share/tnl-0.1/examples/mean-curvature-flow
-
-uninstall: $(TARGET)
- rm -f $(INSTALL_DIR)/bin/$(TARGET)
- rm -f $(INSTALL_DIR)/share/tnl-0.1/examples/mean-curvature-flow
-
-$(TARGET): $(OBJECTS)
- $(CXX) -o $(TARGET) $(OBJECTS) $(LD_FLAGS)
-
-%.o: %.cpp $(HEADERS)
- $(CXX) -c -o $@ $(CXX_FLAGS) $<
+TNL_VERSION=0.1
+TNL_INSTALL_DIR=${HOME}/local/lib
+TNL_INCLUDE_DIR=${HOME}/local/include/tnl-${TNL_VERSION}
+
+TARGET = hamiltonJacobiSolver
+CONFIG_FILE = $(TARGET).cfg.desc
+INSTALL_DIR = ${HOME}/local
+CXX = g++
+CUDA_CXX = nvcc
+OMP_FLAGS = -DHAVE_OPENMP -fopenmp
+CXX_FLAGS = -std=gnu++0x -I$(TNL_INCLUDE_DIR) -O3 $(OMP_FLAGS)
+LD_FLAGS = -L$(TNL_INSTALL_DIR) -ltnl-0.1 -lgomp
+
+SOURCES = main.cpp
+HEADERS =
+OBJECTS = main.o
+DIST = $(SOURCES) Makefile
+
+all: $(TARGET)
+clean:
+ rm -f $(OBJECTS)
+ rm -f $(TARGET)-conf.h
+
+dist: $(DIST)
+ tar zcvf $(TARGET).tgz $(DIST)
+
+install: $(TARGET)
+ cp $(TARGET) $(INSTALL_DIR)/bin
+ cp $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+uninstall: $(TARGET)
+ rm -f $(INSTALL_DIR)/bin/$(TARGET)
+ rm -f $(CONFIG_FILE) $(INSTALL_DIR)/share
+
+$(TARGET): $(OBJECTS)
+ $(CXX) -o $(TARGET) $(OBJECTS) $(LD_FLAGS)
+
+%.o: %.cpp $(HEADERS)
+ $(CXX) -c -o $@ $(CXX_FLAGS) $<
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/main.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/main.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..3c6af1c9e610f76d67cc903c66d55c7f07b5f4ee
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/main.cpp
@@ -0,0 +1,33 @@
+/***************************************************************************
+ main.cpp - description
+ -------------------
+ begin : Jul 8 , 2014
+ copyright : (C) 2014 by Tomas Sobotik
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#include "HamiltonJacobiProblemConfig.h"
+#include "HamiltonJacobiProblemSetter.h"
+#include <solvers/tnlSolver.h>
+#include "MainBuildConfig.h"
+#include <solvers/tnlBuildConfigTags.h>
+
+typedef MainBuildConfig BuildConfig;
+
+int main( int argc, char* argv[] )
+{
+ tnlSolver< HamiltonJacobiProblemSetter, HamiltonJacobiProblemConfig, BuildConfig > solver;
+ if( ! solver. run( argc, argv ) )
+ return EXIT_FAILURE;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cpp b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cpp
new file mode 100644
index 0000000000000000000000000000000000000000..4343837617b4c0c1416b209b6f755fcaf1476912
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cpp
@@ -0,0 +1,14 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnl-direct-eikonal-solver.cpp
+ * Author: oberhuber
+ *
+ * Created on July 13, 2016, 1:13 PM
+ */
+
+#include "tnl-direct-eikonal-solver.h"
\ No newline at end of file
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cu b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cu
new file mode 100644
index 0000000000000000000000000000000000000000..21b17dbdd425c1bb4c6a9664a1c7ba7e2df6d0c4
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.cu
@@ -0,0 +1,14 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnl-direct-eikonal-solver.cu
+ * Author: oberhuber
+ *
+ * Created on July 13, 2016, 1:13 PM
+ */
+
+#include "tnl-direct-eikonal-solver.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.h
new file mode 100644
index 0000000000000000000000000000000000000000..a36153011e3b2e8b8e5590a27e1c0bdcf895faf5
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-direct-eikonal-solver.h
@@ -0,0 +1,72 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnl-direct-eikonal-solver.h
+ * Author: oberhuber
+ *
+ * Created on July 13, 2016, 1:09 PM
+ */
+
+#pragma once
+
+#include <solvers/tnlSolver.h>
+#include <solvers/tnlFastBuildConfigTag.h>
+#include <solvers/tnlBuildConfigTags.h>
+#include <functions/tnlConstantFunction.h>
+#include <functions/tnlMeshFunction.h>
+//#include <problems/tnlHeatEquationProblem.h>
+#include <mesh/tnlGrid.h>
+#include "tnlDirectEikonalProblem.h"
+
+//typedef tnlDefaultBuildMeshConfig BuildConfig;
+typedef tnlFastBuildConfig BuildConfig;
+
+template< typename MeshConfig >
+class tnlDirectEikonalSolverConfig
+{
+ public:
+ static void configSetup( tnlConfigDescription& config )
+ {
+ config.addDelimiter( "Direct eikonal equation solver settings:" );
+ config.addRequiredEntry< String >( "input-file", "Input file." );
+ };
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename MeshType,
+ typename MeshConfig,
+ typename SolverStarter >
+class tnlDirectEikonalSolverSetter
+{
+ public:
+
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ typedef Containers::StaticVector< MeshType::meshDimensions, Real > Vertex;
+
+ static bool run( const Config::ParameterContainer& parameters )
+ {
+ enum { Dimensions = MeshType::meshDimensions };
+ typedef tnlConstantFunction< Dimensions, Real > Anisotropy;
+ typedef tnlDirectEikonalProblem< MeshType, Anisotropy > Problem;
+ SolverStarter solverStarter;
+ return solverStarter.template run< Problem >( parameters );
+ };
+};
+
+int main( int argc, char* argv[] )
+{
+ if( ! tnlSolver< tnlDirectEikonalSolverSetter, tnlDirectEikonalSolverConfig, BuildConfig >::run( argc, argv ) )
+ return EXIT_FAILURE;
+ return EXIT_SUCCESS;
+}
+
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-eikonal-equation-eoc-test b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-eikonal-equation-eoc-test
new file mode 100644
index 0000000000000000000000000000000000000000..bcc05ef15adf4f042fe813d875545b2bae09096a
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-eikonal-equation-eoc-test
@@ -0,0 +1,201 @@
+#!/bin/bash
+
+device="host"
+dimensions="1D 2D 3D"
+#dimensions="1D"
+sizes1D="16 32 64 128 256 512 1024 2048 4096"
+#sizes1D="256"
+sizes2D="16 32 64 128 256 512 1024"
+#sizes2D="8"
+sizes3D="16 32 64 128 256"
+testFunctions="paraboloid"
+snapshotPeriod=0.1
+finalTime=1.5
+solverName="tnl-hamilton-jacobi"
+#solverName="gdb --args tnl-hamilton-jacobi-dbg"
+#
+
+setupTestFunction()
+{
+ testFunction=$1
+# if test x${testFunction} = "xexp-bump";
+# then
+ origin=-1.0
+ proportions=2.0
+ amplitude=1.0
+ waveLength=0.2
+ waveLengthX=0.2
+ waveLengthY=0.2
+ waveLengthZ=0.2
+ wavesNumber=3.0
+ wavesNumberX=0.5
+ wavesNumberY=2.0
+ wavesNumberZ=3.0
+ phase=0.1
+ phaseX=0.0
+ phaseY=0.0
+ phaseZ=0.0
+ sigma=0.5
+# fi
+}
+
+setupGrid()
+{
+ dimensions=$1
+ gridSize=$2
+ tnl-grid-setup --dimensions ${dimensions} \
+ --origin-x ${origin} \
+ --origin-y ${origin} \
+ --origin-z ${origin} \
+ --proportions-x ${proportions} \
+ --proportions-y ${proportions} \
+ --proportions-z ${proportions} \
+ --size-x ${gridSize} \
+ --size-y ${gridSize} \
+ --size-z ${gridSize}
+}
+
+setInitialCondition()
+{
+ testFunction=$1
+ tnl-init --test-function ${testFunction} \
+ --output-file initial-u.tnl \
+ --amplitude ${amplitude} \
+ --wave-length ${waveLength} \
+ --wave-length-x ${waveLengthX} \
+ --wave-length-y ${waveLengthY} \
+ --wave-length-z ${waveLengthZ} \
+ --waves-number ${wavesNumber} \
+ --waves-number-x ${wavesNumberX} \
+ --waves-number-y ${wavesNumberY} \
+ --waves-number-z ${wavesNumberZ} \
+ --phase ${phase} \
+ --phase-x ${phaseX} \
+ --phase-y ${phaseY} \
+ --phase-z ${phaseZ} \
+ --sigma ${sigma}
+
+ tnl-init --test-function ${testFunction}-sdf \
+ --output-file final-u.tnl \
+ --amplitude ${amplitude} \
+ --wave-length ${waveLength} \
+ --wave-length-x ${waveLengthX} \
+ --wave-length-y ${waveLengthY} \
+ --wave-length-z ${waveLengthZ} \
+ --waves-number ${wavesNumber} \
+ --waves-number-x ${wavesNumberX} \
+ --waves-number-y ${wavesNumberY} \
+ --waves-number-z ${wavesNumberZ} \
+ --phase ${phase} \
+ --phase-x ${phaseX} \
+ --phase-y ${phaseY} \
+ --phase-z ${phaseZ} \
+ --sigma ${sigma}
+
+}
+
+solve()
+{
+ timeDiscretisation=$1
+ discreteSolver=$2
+ ${solverName} --device ${device} \
+ --mesh mesh.tnl \
+ --initial-condition initial-u.tnl \
+ --time-discretisation ${timeDiscretisation} \
+ --time-step 10 \
+ --time-step-order 2 \
+ --discrete-solver ${discreteSolver} \
+ --merson-adaptivity 1.0e-5 \
+ --min-iterations 20 \
+ --convergence-residue 1.0e-12 \
+ --snapshot-period ${snapshotPeriod} \
+ --final-time ${finalTime}
+}
+
+computeError()
+{
+ tnl-diff --mesh mesh.tnl \
+ --input-files final-u.tnl u-*.tnl \
+ --mode sequence \
+ --snapshot-period ${snapshotPeriod} \
+ --output-file errors.txt \
+ --write-difference yes
+}
+
+runTest()
+{
+ for testFunction in ${testFunctions};
+ do
+ mkdir -p ${testFunction}
+ cd ${testFunction}
+ setupTestFunction ${testFunction}
+
+ for dim in ${dimensions};
+ do
+ mkdir -p $dim
+ cd ${dim}
+ if test $dim = 1D;
+ then
+ sizes=$sizes1D
+ fi
+ if test $dim = 2D;
+ then
+ sizes=$sizes2D
+ fi
+ if test $dim = 3D;
+ then
+ sizes=$sizes3D
+ fi
+
+ lastSize=""
+ for size in $sizes;
+ do
+ mkdir -p $size
+ cd $size
+ echo ""
+ echo ""
+ echo ""
+ if test ! -f computation-done;
+ then
+ touch computation-in-progress
+ echo "========================================================================="
+ echo "=== SETTING UP THE GRID ==="
+ echo "========================================================================="
+ setupGrid $dim $size
+ echo "========================================================================="
+ echo "=== WRITING THE EXACT SOLUTION ==="
+ echo "========================================================================="
+ setInitialCondition $testFunction
+ echo "========================================================================="
+ echo "=== STARTING THE SOLVER ==="
+ echo "========================================================================="
+ solve explicit merson
+ #solve semi-implicit gmres
+ mv computation-in-progress computation-done
+ fi
+ echo "========================================================================="
+ echo "=== COMPUTING THE ERROR ==="
+ echo "========================================================================="
+ computeError
+ echo "========================================================================="
+ echo "=== COMPUTING THE EOC ==="
+ echo "========================================================================="
+ if test ! x$lastSize = x;
+ then
+ tnl-err2eoc ../$lastSize/errors.txt errors.txt
+ fi
+ echo "========================================================================="
+ echo "=== COMPUTATION DONE ==="
+ echo "========================================================================="
+ cd ..
+ lastSize=$size
+ done
+
+ cd ..
+ done
+ cd ..
+ done
+}
+
+runTest
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-fsm-eoc-test b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-fsm-eoc-test
new file mode 100755
index 0000000000000000000000000000000000000000..6d997ce230a9de4ad1deefc1ccaa1f926021a16f
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnl-run-fsm-eoc-test
@@ -0,0 +1,204 @@
+#!/bin/bash
+
+device="host"
+dimensions="2D 3D"
+dimensions="2D"
+sizes1D="16 32 64 128 256 512 1024 2048 4096"
+#sizes1D="256"
+sizes2D="16 32 64 128 256 512 1024"
+#sizes2D="8"
+sizes3D="16 32 64 128 256"
+testFunctions="paraboloid"
+snapshotPeriod=0.1
+finalTime=1.5
+solverName="tnl-direct-eikonal-solver"
+#solverName="gdb --args tnl-direct-eikonal-solver-dbg"
+#
+
+setupTestFunction()
+{
+ testFunction=$1
+# if test x${testFunction} = "xexp-bump";
+# then
+ origin=-1.0
+ proportions=2.0
+ amplitude=1.0
+ waveLength=0.2
+ waveLengthX=0.2
+ waveLengthY=0.2
+ waveLengthZ=0.2
+ wavesNumber=3.0
+ wavesNumberX=0.5
+ wavesNumberY=2.0
+ wavesNumberZ=3.0
+ phase=0.1
+ phaseX=0.0
+ phaseY=0.0
+ phaseZ=0.0
+ sigma=0.5
+ radius=0.5
+# fi
+}
+
+setupGrid()
+{
+ dimensions=$1
+ gridSize=$2
+ tnl-grid-setup --dimensions ${dimensions} \
+ --origin-x ${origin} \
+ --origin-y ${origin} \
+ --origin-z ${origin} \
+ --proportions-x ${proportions} \
+ --proportions-y ${proportions} \
+ --proportions-z ${proportions} \
+ --size-x ${gridSize} \
+ --size-y ${gridSize} \
+ --size-z ${gridSize}
+}
+
+setInitialCondition()
+{
+ testFunction=$1
+ tnl-init --test-function ${testFunction}-sdf \
+ --output-file initial-u.tnl \
+ --amplitude ${amplitude} \
+ --wave-length ${waveLength} \
+ --wave-length-x ${waveLengthX} \
+ --wave-length-y ${waveLengthY} \
+ --wave-length-z ${waveLengthZ} \
+ --waves-number ${wavesNumber} \
+ --waves-number-x ${wavesNumberX} \
+ --waves-number-y ${wavesNumberY} \
+ --waves-number-z ${wavesNumberZ} \
+ --phase ${phase} \
+ --phase-x ${phaseX} \
+ --phase-y ${phaseY} \
+ --phase-z ${phaseZ} \
+ --sigma ${sigma} \
+ --radius ${radius}
+
+ tnl-init --test-function ${testFunction}-sdf \
+ --output-file final-u.tnl \
+ --amplitude ${amplitude} \
+ --wave-length ${waveLength} \
+ --wave-length-x ${waveLengthX} \
+ --wave-length-y ${waveLengthY} \
+ --wave-length-z ${waveLengthZ} \
+ --waves-number ${wavesNumber} \
+ --waves-number-x ${wavesNumberX} \
+ --waves-number-y ${wavesNumberY} \
+ --waves-number-z ${wavesNumberZ} \
+ --phase ${phase} \
+ --phase-x ${phaseX} \
+ --phase-y ${phaseY} \
+ --phase-z ${phaseZ} \
+ --sigma ${sigma} \
+ --radius ${radius}
+
+}
+
+solve()
+{
+ timeDiscretisation=$1
+ discreteSolver=$2
+ ${solverName} --device ${device} \
+ --mesh mesh.tnl \
+ --input-file initial-u.tnl \
+ --time-discretisation ${timeDiscretisation} \
+ --time-step 10 \
+ --time-step-order 2 \
+ --discrete-solver ${discreteSolver} \
+ --merson-adaptivity 1.0e-5 \
+ --min-iterations 20 \
+ --convergence-residue 1.0e-12 \
+ --snapshot-period ${snapshotPeriod} \
+ --final-time ${finalTime}
+}
+
+computeError()
+{
+ tnl-diff --mesh mesh.tnl \
+ --input-files final-u.tnl u-*.tnl \
+ --mode sequence \
+ --snapshot-period ${snapshotPeriod} \
+ --output-file errors.txt \
+ --write-difference yes
+}
+
+runTest()
+{
+ for testFunction in ${testFunctions};
+ do
+ mkdir -p ${testFunction}
+ cd ${testFunction}
+ setupTestFunction ${testFunction}
+
+ for dim in ${dimensions};
+ do
+ mkdir -p $dim
+ cd ${dim}
+ if test $dim = 1D;
+ then
+ sizes=$sizes1D
+ fi
+ if test $dim = 2D;
+ then
+ sizes=$sizes2D
+ fi
+ if test $dim = 3D;
+ then
+ sizes=$sizes3D
+ fi
+
+ lastSize=""
+ for size in $sizes;
+ do
+ mkdir -p $size
+ cd $size
+ echo ""
+ echo ""
+ echo ""
+ if test ! -f computation-done;
+ then
+ touch computation-in-progress
+ echo "========================================================================="
+ echo "=== SETTING UP THE GRID ==="
+ echo "========================================================================="
+ setupGrid $dim $size
+ echo "========================================================================="
+ echo "=== WRITING THE EXACT SOLUTION ==="
+ echo "========================================================================="
+ setInitialCondition $testFunction
+ echo "========================================================================="
+ echo "=== STARTING THE SOLVER ==="
+ echo "========================================================================="
+ solve explicit merson
+ #solve semi-implicit gmres
+ mv computation-in-progress computation-done
+ fi
+ echo "========================================================================="
+ echo "=== COMPUTING THE ERROR ==="
+ echo "========================================================================="
+ computeError
+ echo "========================================================================="
+ echo "=== COMPUTING THE EOC ==="
+ echo "========================================================================="
+ if test ! x$lastSize = x;
+ then
+ tnl-err2eoc ../$lastSize/errors.txt errors.txt
+ fi
+ echo "========================================================================="
+ echo "=== COMPUTATION DONE ==="
+ echo "========================================================================="
+ cd ..
+ lastSize=$size
+ done
+
+ cd ..
+ done
+ cd ..
+ done
+}
+
+runTest
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase.h
new file mode 100644
index 0000000000000000000000000000000000000000..409a2d93a1d7eb719e3b4f93b6eb9b7da53f72e5
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase.h
@@ -0,0 +1,89 @@
+/*
+ * File: tnlDirectEikonalMethodsBase.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 3:17 PM
+ */
+
+#pragma once
+
+#include <mesh/tnlGrid.h>
+#include <functions/tnlMeshFunction.h>
+
+template< typename Mesh >
+class tnlDirectEikonalMethodsBase
+{
+};
+
+template< typename Real,
+ typename Device,
+ typename Index >
+class tnlDirectEikonalMethodsBase< tnlGrid< 1, Real, Device, Index > >
+{
+ public:
+
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef Device DevcieType;
+ typedef Index IndexType;
+ typedef tnlMeshFunction< MeshType > MeshFunctionType;
+ typedef tnlMeshFunction< MeshType, 1, bool > InterfaceMapType;
+
+ void initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap );
+
+ template< typename MeshEntity >
+ void updateCell( MeshFunctionType& u,
+ const MeshEntity& cell );
+
+};
+
+
+template< typename Real,
+ typename Device,
+ typename Index >
+class tnlDirectEikonalMethodsBase< tnlGrid< 2, Real, Device, Index > >
+{
+ public:
+
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef Device DevcieType;
+ typedef Index IndexType;
+ typedef tnlMeshFunction< MeshType > MeshFunctionType;
+ typedef tnlMeshFunction< MeshType, 2, bool > InterfaceMapType;
+
+ void initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap );
+
+ template< typename MeshEntity >
+ void updateCell( MeshFunctionType& u,
+ const MeshEntity& cell );
+};
+
+template< typename Real,
+ typename Device,
+ typename Index >
+class tnlDirectEikonalMethodsBase< tnlGrid< 3, Real, Device, Index > >
+{
+ public:
+
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef Device DevcieType;
+ typedef Index IndexType;
+ typedef tnlMeshFunction< MeshType > MeshFunctionType;
+ typedef tnlMeshFunction< MeshType, 3, bool > InterfaceMapType;
+
+ void initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap );
+
+ template< typename MeshEntity >
+ void updateCell( MeshFunctionType& u,
+ const MeshEntity& cell );
+};
+
+#include "tnlDirectEikonalMethodsBase_impl.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..c2983fac2c12e1b7791bd86973a712a5fcc6c948
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalMethodsBase_impl.h
@@ -0,0 +1,233 @@
+/*
+ * File: tnlDirectEikonalMethodsBase_impl.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 3:22 PM
+ */
+
+#pragma once
+
+#include <core/tnlTypeInfo.h>
+#include <functions/tnlFunctions.h>
+
+template< typename Real,
+ typename Device,
+ typename Index >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 1, Real, Device, Index > >::
+initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap )
+{
+ const MeshType& mesh = input.getMesh();
+ typedef typename MeshType::Cell Cell;
+ Cell cell( mesh );
+ for( cell.getCoordinates().x() = 1;
+ cell.getCoordinates().x() < mesh.getDimensions().x() - 1;
+ cell.getCoordinates().x() ++ )
+ {
+ cell.refresh();
+ const RealType& c = input( cell );
+ if( ! cell.isBoundaryEntity() )
+ {
+ const auto& neighbours = cell.getNeighbourEntities();
+ //const IndexType& c = cell.getIndex();
+ const IndexType e = neighbours.template getEntityIndex< 1 >();
+ const IndexType w = neighbours.template getEntityIndex< -1 >();
+
+ if( c * input[ e ] <= 0 || c * input[ w ] <= 0 )
+ {
+ output[ cell.getIndex() ] = c;
+ interfaceMap[ cell.getIndex() ] = true;
+ continue;
+ }
+ }
+ output[ cell.getIndex() ] =
+ c > 0 ? tnlTypeInfo< RealType >::getMaxValue() :
+ -tnlTypeInfo< RealType >::getMaxValue();
+ interfaceMap[ cell.getIndex() ] = false;
+ }
+}
+
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< typename MeshEntity >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 1, Real, Device, Index > >::
+updateCell( MeshFunctionType& u,
+ const MeshEntity& cell )
+{
+}
+
+
+template< typename Real,
+ typename Device,
+ typename Index >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 2, Real, Device, Index > >::
+initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap )
+{
+ const MeshType& mesh = input.getMesh();
+ typedef typename MeshType::Cell Cell;
+ Cell cell( mesh );
+ for( cell.getCoordinates().y() = 0;
+ cell.getCoordinates().y() < mesh.getDimensions().y();
+ cell.getCoordinates().y() ++ )
+ for( cell.getCoordinates().x() = 0;
+ cell.getCoordinates().x() < mesh.getDimensions().x();
+ cell.getCoordinates().x() ++ )
+ {
+ cell.refresh();
+ const RealType& c = input( cell );
+ if( ! cell.isBoundaryEntity() )
+ {
+ auto neighbours = cell.getNeighbourEntities();
+ const IndexType e = neighbours.template getEntityIndex< 1, 0 >();
+ const IndexType w = neighbours.template getEntityIndex< -1, 0 >();
+ const IndexType n = neighbours.template getEntityIndex< 0, 1 >();
+ const IndexType s = neighbours.template getEntityIndex< 0, -1 >();
+ if( c * input[ e ] <= 0 || c * input[ w ] <= 0 ||
+ c * input[ n ] <= 0 || c * input[ s ] <= 0 )
+ {
+ output[ cell.getIndex() ] = c;
+ interfaceMap[ cell.getIndex() ] = true;
+ continue;
+ }
+ }
+ output[ cell.getIndex() ] =
+ c > 0 ? tnlTypeInfo< RealType >::getMaxValue() :
+ -tnlTypeInfo< RealType >::getMaxValue();
+ interfaceMap[ cell.getIndex() ] = false;
+ }
+}
+
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< typename MeshEntity >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 2, Real, Device, Index > >::
+updateCell( MeshFunctionType& u,
+ const MeshEntity& cell )
+{
+ const auto& neighbourEntities = cell.template getNeighbourEntities< 2 >();
+ const MeshType& mesh = cell.getMesh();
+
+ const RealType& h = mesh.getSpaceSteps().x();
+ const RealType value = u( cell );
+ Real a, b, tmp;
+
+ if( cell.getCoordinates().x() == 0 )
+ a = u[ neighbourEntities.template getEntityIndex< 1, 0 >() ];
+ else if( cell.getCoordinates().x() == mesh.getDimensions().x() - 1 )
+ a = u[ neighbourEntities.template getEntityIndex< -1, 0 >() ];
+ else
+ {
+ a = ArgAbsMin( u[ neighbourEntities.template getEntityIndex< -1, 0 >() ],
+ u[ neighbourEntities.template getEntityIndex< 1, 0 >() ] );
+ }
+
+ if( cell.getCoordinates().y() == 0 )
+ b = u[ neighbourEntities.template getEntityIndex< 0, 1 >()];
+ else if( cell.getCoordinates().y() == mesh.getDimensions().y() - 1 )
+ b = u[ neighbourEntities.template getEntityIndex< 0, -1 >() ];
+ else
+ {
+ b = ArgAbsMin( u[ neighbourEntities.template getEntityIndex< 0, -1 >() ],
+ u[ neighbourEntities.template getEntityIndex< 0, 1 >() ] );
+ }
+
+ if( fabs( a ) == tnlTypeInfo< Real >::getMaxValue() &&
+ fabs( b ) == tnlTypeInfo< Real >::getMaxValue() )
+ return;
+ if( fabs( a ) == tnlTypeInfo< Real >::getMaxValue() ||
+ fabs( b ) == tnlTypeInfo< Real >::getMaxValue() ||
+ fabs( a - b ) >= h )
+ {
+ tmp = ArgAbsMin( a, b ) + sign( value ) * h;
+ /* std::cerr << "a = " << a << " b = " << b << " h = " << h
+ << " ArgAbsMin( a, b ) = " << ArgAbsMin( a, b ) << " sign( value ) = " << sign( value )
+ << " sign( value ) * h = " << sign( value ) * h
+ << " ArgAbsMin( a, b ) + sign( value ) * h = " << ArgAbsMin( a, b ) + sign( value ) * h
+ << " tmp = " << tmp << std::endl;
+ tmp = ArgAbsMin( a, b ) + sign( value ) * h;
+ tmp = ArgAbsMin( a, b ) + sign( value ) * h;
+ tmp = ArgAbsMin( a, b ) + sign( value ) * h;
+ res = ArgAbsMin( a, b ) + sign( value ) * h;
+ std::cerr << " tmp = " << tmp << std::endl;
+ std::cerr << " res = " << res << std::endl;*/
+
+ }
+ else
+ tmp = 0.5 * ( a + b + sign( value ) * sqrt( 2.0 * h * h - ( a - b ) * ( a - b ) ) );
+
+ u[ cell.getIndex() ] = ArgAbsMin( value, tmp );
+ //std::cerr << ArgAbsMin( value, tmp ) << " ";
+}
+
+
+template< typename Real,
+ typename Device,
+ typename Index >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 3, Real, Device, Index > >::
+initInterface( const MeshFunctionType& input,
+ MeshFunctionType& output,
+ InterfaceMapType& interfaceMap )
+{
+ const MeshType& mesh = input.getMesh();
+ typedef typename MeshType::Cell Cell;
+ Cell cell( mesh );
+ for( cell.getCoordinates().z() = 1;
+ cell.getCoordinates().z() < mesh.getDimensions().z() - 1;
+ cell.getCoordinates().z() ++ )
+ for( cell.getCoordinates().y() = 1;
+ cell.getCoordinates().y() < mesh.getDimensions().y() - 1;
+ cell.getCoordinates().y() ++ )
+ for( cell.getCoordinates().x() = 1;
+ cell.getCoordinates().x() < mesh.getDimensions().x() - 1;
+ cell.getCoordinates().x() ++ )
+ {
+ cell.refresh();
+ const RealType& c = input( cell );
+ if( ! cell.isBoundaryEntity() )
+ {
+ auto neighbours = cell.getNeighbourEntities();
+ //const IndexType& c = cell.getIndex();
+ const IndexType e = neighbours.template getEntityIndex< 1, 0, 0 >();
+ const IndexType w = neighbours.template getEntityIndex< -1, 0, 0 >();
+ const IndexType n = neighbours.template getEntityIndex< 0, 1, 0 >();
+ const IndexType s = neighbours.template getEntityIndex< 0, -1, 0 >();
+ const IndexType t = neighbours.template getEntityIndex< 0, 0, 1 >();
+ const IndexType b = neighbours.template getEntityIndex< 0, 0, -1 >();
+
+ if( c * input[ e ] <= 0 || c * input[ w ] <= 0 ||
+ c * input[ n ] <= 0 || c * input[ s ] <= 0 ||
+ c * input[ t ] <= 0 || c * input[ b ] <= 0 )
+ {
+ output[ cell.getIndex() ] = c;
+ interfaceMap[ cell.getIndex() ] = true;
+ continue;
+ }
+ }
+ output[ cell.getIndex() ] =
+ c > 0 ? tnlTypeInfo< RealType >::getMaxValue() :
+ -tnlTypeInfo< RealType >::getMaxValue();
+ interfaceMap[ cell.getIndex() ] = false;
+ }
+}
+
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< typename MeshEntity >
+void
+tnlDirectEikonalMethodsBase< tnlGrid< 3, Real, Device, Index > >::
+updateCell( MeshFunctionType& u,
+ const MeshEntity& cell )
+{
+
+}
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem.h
new file mode 100644
index 0000000000000000000000000000000000000000..f4baba1528165a9ca95bf18b68eaebe09d0c7848
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem.h
@@ -0,0 +1,80 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod.h
+ * Author: oberhuber
+ *
+ * Created on July 13, 2016, 1:19 PM
+ */
+
+#pragma once
+
+#include <problems/tnlPDEProblem.h>
+#include <functions/tnlMeshFunction.h>
+#include "tnlFastSweepingMethod.h"
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class tnlDirectEikonalProblem
+ : public tnlPDEProblem< Mesh,
+ TimeIndependentProblem,
+ Real,
+ typename Mesh::DeviceType,
+ Index >
+{
+ public:
+
+ typedef Real RealType;
+ typedef typename Mesh::DeviceType DeviceType;
+ typedef Index IndexType;
+ typedef tnlMeshFunction< Mesh > MeshFunctionType;
+ typedef tnlPDEProblem< Mesh, TimeIndependentProblem, RealType, DeviceType, IndexType > BaseType;
+ typedef Anisotropy AnisotropyType;
+
+ using typename BaseType::MeshType;
+ using typename BaseType::DofVectorType;
+ using typename BaseType::MeshDependentDataType;
+
+ static String getTypeStatic();
+
+ String getPrologHeader() const;
+
+ void writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters ) const;
+
+ bool writeEpilog( tnlLogger& logger );
+
+
+ bool setup( const Config::ParameterContainer& parameters );
+
+ IndexType getDofs( const MeshType& mesh ) const;
+
+ void bindDofs( const MeshType& mesh,
+ const DofVectorType& dofs );
+
+ bool setInitialData( const Config::ParameterContainer& parameters,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshdependentData );
+
+ bool solve( const MeshType& mesh,
+ DofVectorType& dosf );
+
+
+ protected:
+
+ MeshFunctionType u;
+
+ MeshFunctionType initialData;
+
+ AnisotropyType anisotropy;
+
+};
+
+#include "tnlDirectEikonalProblem_impl.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..9dcf2fb5fa40b49462bed8ced05b4ebf605c38d6
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlDirectEikonalProblem_impl.h
@@ -0,0 +1,125 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod_impl.h
+ * Author: oberhuber
+ *
+ * Created on July 13, 2016, 1:46 PM
+ */
+
+#pragma once
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+String
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+getTypeStatic()
+{
+
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+String
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+getPrologHeader() const
+{
+ return String( "Direct eikonal solver" );
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+void
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters ) const
+{
+
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+bool
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+writeEpilog( tnlLogger& logger )
+{
+ return true;
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+bool
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+setup( const Config::ParameterContainer& parameters )
+{
+ return true;
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+Index
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+getDofs( const MeshType& mesh ) const
+{
+ return mesh.template getEntitiesCount< typename MeshType::Cell >();
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+void
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+bindDofs( const MeshType& mesh,
+ const DofVectorType& dofs )
+{
+ this->u.bind( mesh, dofs );
+}
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+bool
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+setInitialData( const Config::ParameterContainer& parameters,
+ const MeshType& mesh,
+ DofVectorType& dofs,
+ MeshDependentDataType& meshdependentData )
+{
+ String inputFile = parameters.getParameter< String >( "input-file" );
+ this->initialData.setMesh( mesh );
+ if( !this->initialData.boundLoad( inputFile ) )
+ return false;
+ return true;
+}
+
+
+template< typename Mesh,
+ typename Anisotropy,
+ typename Real,
+ typename Index >
+bool
+tnlDirectEikonalProblem< Mesh, Anisotropy, Real, Index >::
+solve( const MeshType& mesh,
+ DofVectorType& dofs )
+{
+ tnlFastSweepingMethod< MeshType, AnisotropyType > fsm;
+ fsm.solve( mesh, anisotropy, initialData );
+}
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod.h
new file mode 100644
index 0000000000000000000000000000000000000000..04332b0f6c398adb36e2df4c338775575dbbcc9e
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod.h
@@ -0,0 +1,141 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 10:04 AM
+ */
+
+#pragma once
+
+#include <mesh/tnlGrid.h>
+#include <functions/tnlConstantFunction.h>
+#include "tnlDirectEikonalMethodsBase.h"
+
+template< typename Mesh,
+ typename Anisotropy = tnlConstantFunction< Mesh::getMeshDimensions(), typename Mesh::RealType > >
+class tnlFastSweepingMethod
+{
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+class tnlFastSweepingMethod< tnlGrid< 1, Real, Device, Index >, Anisotropy >
+ : public tnlDirectEikonalMethodsBase< tnlGrid< 1, Real, Device, Index > >
+{
+ static_assert( std::is_same< Device, tnlHost >::value, "The fast sweeping method works only on CPU." );
+
+ public:
+
+ typedef tnlGrid< 1, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef tnlHost DeviceType;
+ typedef Index IndexType;
+ typedef Anisotropy AnisotropyType;
+ typedef tnlDirectEikonalMethodsBase< tnlGrid< 1, Real, Device, Index > > BaseType;
+
+ using typename BaseType::InterfaceMapType;
+ using typename BaseType::MeshFunctionType;
+
+ tnlFastSweepingMethod();
+
+ const IndexType& getMaxIterations() const;
+
+ void setMaxIterations( const IndexType& maxIterations );
+
+ void solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u );
+
+
+ protected:
+
+ const IndexType maxIterations;
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+class tnlFastSweepingMethod< tnlGrid< 2, Real, Device, Index >, Anisotropy >
+ : public tnlDirectEikonalMethodsBase< tnlGrid< 2, Real, Device, Index > >
+{
+ static_assert( std::is_same< Device, tnlHost >::value, "The fast sweeping method works only on CPU." );
+
+ public:
+
+ typedef tnlGrid< 2, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef tnlHost DeviceType;
+ typedef Index IndexType;
+ typedef Anisotropy AnisotropyType;
+ typedef tnlDirectEikonalMethodsBase< tnlGrid< 2, Real, Device, Index > > BaseType;
+
+ using typename BaseType::InterfaceMapType;
+ using typename BaseType::MeshFunctionType;
+
+ tnlFastSweepingMethod();
+
+ const IndexType& getMaxIterations() const;
+
+ void setMaxIterations( const IndexType& maxIterations );
+
+ void solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u );
+
+
+ protected:
+
+ const IndexType maxIterations;
+};
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+class tnlFastSweepingMethod< tnlGrid< 3, Real, Device, Index >, Anisotropy >
+ : public tnlDirectEikonalMethodsBase< tnlGrid< 3, Real, Device, Index > >
+{
+ static_assert( std::is_same< Device, tnlHost >::value, "The fast sweeping method works only on CPU." );
+
+ public:
+
+ typedef tnlGrid< 3, Real, Device, Index > MeshType;
+ typedef Real RealType;
+ typedef tnlHost DeviceType;
+ typedef Index IndexType;
+ typedef Anisotropy AnisotropyType;
+ typedef tnlDirectEikonalMethodsBase< tnlGrid< 3, Real, Device, Index > > BaseType;
+
+ using typename BaseType::InterfaceMapType;
+ using typename BaseType::MeshFunctionType;
+
+ tnlFastSweepingMethod();
+
+ const IndexType& getMaxIterations() const;
+
+ void setMaxIterations( const IndexType& maxIterations );
+
+ void solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u );
+
+
+ protected:
+
+ const IndexType maxIterations;
+};
+
+
+
+#include "tnlFastSweepingMethod1D_impl.h"
+#include "tnlFastSweepingMethod2D_impl.h"
+#include "tnlFastSweepingMethod3D_impl.h"
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod1D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod1D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..050bc6075454c3d23685f25a6e9d607b36c88d73
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod1D_impl.h
@@ -0,0 +1,70 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod2D_impl.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 10:32 AM
+ */
+
+#pragma once
+
+#include "tnlFastSweepingMethod.h"
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+tnlFastSweepingMethod< tnlGrid< 1, Real, Device, Index >, Anisotropy >::
+tnlFastSweepingMethod()
+: maxIterations( 1 )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+const Index&
+tnlFastSweepingMethod< tnlGrid< 1, Real, Device, Index >, Anisotropy >::
+getMaxIterations() const
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 1, Real, Device, Index >, Anisotropy >::
+setMaxIterations( const IndexType& maxIterations )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 1, Real, Device, Index >, Anisotropy >::
+solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u )
+{
+ MeshFunctionType aux;
+ InterfaceMapType interfaceMap;
+ aux.setMesh( mesh );
+ interfaceMap.setMesh( mesh );
+ std::cout << "Initiating the interface cells ..." << std::endl;
+ BaseType::initInterface( u, aux, interfaceMap );
+ aux.save( "aux-ini.tnl" );
+
+}
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod2D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod2D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..cc2cfe9401c54aa7bff5e88f4c9d9ecf40f919cd
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod2D_impl.h
@@ -0,0 +1,141 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod2D_impl.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 10:32 AM
+ */
+
+#pragma once
+
+#include "tnlFastSweepingMethod.h"
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+tnlFastSweepingMethod< tnlGrid< 2, Real, Device, Index >, Anisotropy >::
+tnlFastSweepingMethod()
+: maxIterations( 1 )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+const Index&
+tnlFastSweepingMethod< tnlGrid< 2, Real, Device, Index >, Anisotropy >::
+getMaxIterations() const
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 2, Real, Device, Index >, Anisotropy >::
+setMaxIterations( const IndexType& maxIterations )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 2, Real, Device, Index >, Anisotropy >::
+solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u )
+{
+ MeshFunctionType aux;
+ InterfaceMapType interfaceMap;
+ aux.setMesh( mesh );
+ interfaceMap.setMesh( mesh );
+ std::cout << "Initiating the interface cells ..." << std::endl;
+ BaseType::initInterface( u, aux, interfaceMap );
+ //aux.save( "aux-ini.tnl" );
+
+ typename MeshType::Cell cell( mesh );
+
+ IndexType iteration( 0 );
+ while( iteration < this->maxIterations )
+ {
+
+ for( cell.getCoordinates().y() = 0;
+ cell.getCoordinates().y() < mesh.getDimensions().y();
+ cell.getCoordinates().y()++ )
+ {
+ for( cell.getCoordinates().x() = 0;
+ cell.getCoordinates().x() < mesh.getDimensions().x();
+ cell.getCoordinates().x()++ )
+ {
+ cell.refresh();
+ if( ! interfaceMap( cell ) )
+ this->updateCell( aux, cell );
+ }
+ }
+ //aux.save( "aux-1.tnl" );
+
+ for( cell.getCoordinates().y() = 0;
+ cell.getCoordinates().y() < mesh.getDimensions().y();
+ cell.getCoordinates().y()++ )
+ {
+ for( cell.getCoordinates().x() = mesh.getDimensions().x() - 1;
+ cell.getCoordinates().x() >= 0 ;
+ cell.getCoordinates().x()-- )
+ {
+ //std::cerr << "2 -> ";
+ cell.refresh();
+ if( ! interfaceMap( cell ) )
+ this->updateCell( aux, cell );
+ }
+ }
+ //aux.save( "aux-2.tnl" );
+
+ for( cell.getCoordinates().y() = mesh.getDimensions().y() - 1;
+ cell.getCoordinates().y() >= 0 ;
+ cell.getCoordinates().y()-- )
+ {
+ for( cell.getCoordinates().x() = 0;
+ cell.getCoordinates().x() < mesh.getDimensions().x();
+ cell.getCoordinates().x()++ )
+ {
+ //std::cerr << "3 -> ";
+ cell.refresh();
+ if( ! interfaceMap( cell ) )
+ this->updateCell( aux, cell );
+ }
+ }
+ //aux.save( "aux-3.tnl" );
+
+
+ for( cell.getCoordinates().y() = mesh.getDimensions().y() - 1;
+ cell.getCoordinates().y() >= 0;
+ cell.getCoordinates().y()-- )
+ {
+ for( cell.getCoordinates().x() = mesh.getDimensions().x() - 1;
+ cell.getCoordinates().x() >= 0 ;
+ cell.getCoordinates().x()-- )
+ {
+ //std::cerr << "4 -> ";
+ cell.refresh();
+ if( ! interfaceMap( cell ) )
+ this->updateCell( aux, cell );
+ }
+ }
+ //aux.save( "aux-4.tnl" );
+ iteration++;
+ }
+}
+
diff --git a/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod3D_impl.h b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod3D_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..f252537148b9968b1a84421da7ca563169525ae7
--- /dev/null
+++ b/src/TNL/Experimental/Hamilton-Jacobi/Solvers/hamilton-jacobi/tnlFastSweepingMethod3D_impl.h
@@ -0,0 +1,69 @@
+/*
+ * To change this license header, choose License Headers in Project Properties.
+ * To change this template file, choose Tools | Templates
+ * and open the template in the editor.
+ */
+
+/*
+ * File: tnlFastSweepingMethod2D_impl.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 10:32 AM
+ */
+
+#pragma once
+
+#include "tnlFastSweepingMethod.h"
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+tnlFastSweepingMethod< tnlGrid< 3, Real, Device, Index >, Anisotropy >::
+tnlFastSweepingMethod()
+: maxIterations( 1 )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+const Index&
+tnlFastSweepingMethod< tnlGrid< 3, Real, Device, Index >, Anisotropy >::
+getMaxIterations() const
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 3, Real, Device, Index >, Anisotropy >::
+setMaxIterations( const IndexType& maxIterations )
+{
+
+}
+
+template< typename Real,
+ typename Device,
+ typename Index,
+ typename Anisotropy >
+void
+tnlFastSweepingMethod< tnlGrid< 3, Real, Device, Index >, Anisotropy >::
+solve( const MeshType& mesh,
+ const AnisotropyType& anisotropy,
+ MeshFunctionType& u )
+{
+ MeshFunctionType aux;
+ InterfaceMapType interfaceMap;
+ aux.setMesh( mesh );
+ interfaceMap.setMesh( mesh );
+ std::cout << "Initiating the interface cells ..." << std::endl;
+ BaseType::initInterface( u, aux, interfaceMap );
+ aux.save( "aux-ini.tnl" );
+}
+
diff --git a/src/TNL/File.h b/src/TNL/File.h
index 9ea11d299a1d160eec562d156ab35d9c9112b8f4..1eea42eae61b67867bfeb7218b0a1801932f2c14 100644
--- a/src/TNL/File.h
+++ b/src/TNL/File.h
@@ -119,4 +119,3 @@ bool fileExists( const String& fileName );
} // namespace TNL
#include <TNL/File_impl.h>
-
diff --git a/src/TNL/Functions/Analytic/Blob.h b/src/TNL/Functions/Analytic/Blob.h
index 56c6e22b7eab001f03b45137b23ba6f2582d4798..c436254d6af11f4563a2eb49db8f33149f4bdd8e 100644
--- a/src/TNL/Functions/Analytic/Blob.h
+++ b/src/TNL/Functions/Analytic/Blob.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump.h - description
+ Blob.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/Blob_impl.h b/src/TNL/Functions/Analytic/Blob_impl.h
index c6d969d38527cfab69858a39f3d2fac3b340287e..00328818b266d1e9c40674e3701c86116391a9c4 100644
--- a/src/TNL/Functions/Analytic/Blob_impl.h
+++ b/src/TNL/Functions/Analytic/Blob_impl.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump_impl.h - description
+ Blob_impl.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/CMakeLists.txt b/src/TNL/Functions/Analytic/CMakeLists.txt
index 01c91a1542b9b1cfdbb1bae7c2e37c4ce44fdb91..744345453d919ac82f2ef2a39186774f90c1b652 100755
--- a/src/TNL/Functions/Analytic/CMakeLists.txt
+++ b/src/TNL/Functions/Analytic/CMakeLists.txt
@@ -7,13 +7,21 @@ SET( headers Blob.h
ExpBump.h
ExpBump_impl.h
Flowerpot.h
- Flowerpot_impl.h
+ Flowerpot_impl.h
+ Paraboloid.h
+ Paraboloid_impl.h
+ ParaboloidSDF.h
+ ParaboloidSDF_impl.h
PseudoSquare.h
PseudoSquare_impl.h
SinBumps.h
SinBumps_impl.h
+ SinBumpsSDF.h
+ SinBumpsSDF_impl.h
SinWave.h
SinWave_impl.h
+ SinWaveSDF.h
+ SinWaveSDF_impl.h
Twins.h
Twins_impl.h
)
diff --git a/src/TNL/Functions/Analytic/Constant.h b/src/TNL/Functions/Analytic/Constant.h
index dc12be876f6adc50d423940d1c13c194adb26563..8e20cc6d5c816149e0422e615e499390d16bcaa7 100644
--- a/src/TNL/Functions/Analytic/Constant.h
+++ b/src/TNL/Functions/Analytic/Constant.h
@@ -44,7 +44,7 @@ class Constant : public Domain< dimensions, NonspaceDomain >
int YDiffOrder,
int ZDiffOrder >
#else
- template< int XDiffOrder,
+ template< int XDiffOrder = 0,
int YDiffOrder = 0,
int ZDiffOrder = 0 >
#endif
diff --git a/src/TNL/Functions/Analytic/Paraboloid.h b/src/TNL/Functions/Analytic/Paraboloid.h
new file mode 100644
index 0000000000000000000000000000000000000000..e07963155dbcf2a8c9d54028c7e3846c1100be2f
--- /dev/null
+++ b/src/TNL/Functions/Analytic/Paraboloid.h
@@ -0,0 +1,168 @@
+/***************************************************************************
+ Paraboloid.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+#include <TNL/Functions/Domain.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions,
+ typename Real = double >
+class ParaboloidBase : public Functions::Domain< dimensions, SpaceDomain >
+{
+ public:
+
+ ParaboloidBase();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ void setXCenter( const Real& waveLength );
+
+ Real getXCenter() const;
+
+ void setYCenter( const Real& waveLength );
+
+ Real getYCenter() const;
+
+ void setZCenter( const Real& waveLength );
+
+ Real getZCenter() const;
+
+ void setCoefficient( const Real& coefficient );
+
+ Real getCoefficient() const;
+
+ void setOffset( const Real& offset );
+
+ Real getOffset() const;
+
+ protected:
+
+ Real xCenter, yCenter, zCenter, coefficient, radius;
+};
+
+template< int Dimensions, typename Real >
+class Paraboloid
+{
+};
+
+template< typename Real >
+class Paraboloid< 1, Real > : public ParaboloidBase< 1, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 1, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ return this->getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+
+};
+
+template< typename Real >
+class Paraboloid< 2, Real > : public ParaboloidBase< 2, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 2, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ return this->getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+
+};
+
+template< typename Real >
+class Paraboloid< 3, Real > : public ParaboloidBase< 3, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 3, RealType > VertexType;
+
+
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ return this->getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+
+};
+
+template< int Dimensions,
+ typename Real >
+std::ostream& operator << ( std::ostream& str, const Paraboloid< Dimensions, Real >& f )
+{
+ str << "SDF Paraboloid function: amplitude = " << f.getCoefficient()
+ << " offset = " << f.getOffset();
+ return str;
+}
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
+#include <TNL/Functions/Analytic/Paraboloid_impl.h>
+
diff --git a/src/TNL/Functions/Analytic/ParaboloidSDF.h b/src/TNL/Functions/Analytic/ParaboloidSDF.h
new file mode 100644
index 0000000000000000000000000000000000000000..4d3177e365cfd46b45d57064d39e57ee163fd74a
--- /dev/null
+++ b/src/TNL/Functions/Analytic/ParaboloidSDF.h
@@ -0,0 +1,160 @@
+/***************************************************************************
+ ParaboloidSDF.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+#include <TNL/Functions/Domain.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions,
+ typename Real = double >
+class ParaboloidSDFBase : public Functions::Domain< dimensions, SpaceDomain >
+{
+ public:
+
+ ParaboloidSDFBase();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ void setXCenter( const Real& waveLength );
+
+ Real getXCenter() const;
+
+ void setYCenter( const Real& waveLength );
+
+ Real getYCenter() const;
+
+ void setZCenter( const Real& waveLength );
+
+ Real getZCenter() const;
+
+ void setCoefficient( const Real& coefficient );
+
+ Real getCoefficient() const;
+
+ void setOffset( const Real& offset );
+
+ Real getOffset() const;
+
+ protected:
+
+ Real xCenter, yCenter, zCenter, coefficient, radius;
+};
+
+template< int Dimensions, typename Real >
+class ParaboloidSDF
+{
+};
+
+template< typename Real >
+class ParaboloidSDF< 1, Real > : public ParaboloidSDFBase< 1, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 1, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class ParaboloidSDF< 2, Real > : public ParaboloidSDFBase< 2, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 2, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class ParaboloidSDF< 3, Real > : public ParaboloidSDFBase< 3, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 3, RealType > VertexType;
+
+
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< int Dimensions,
+ typename Real >
+std::ostream& operator << ( std::ostream& str, const ParaboloidSDF< Dimensions, Real >& f )
+{
+ str << "SDF Paraboloid SDF function: amplitude = " << f.getCoefficient()
+ << " offset = " << f.getOffset();
+ return str;
+}
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
+
+#include <TNL/Functions/Analytic/ParaboloidSDF_impl.h>
+
diff --git a/src/TNL/Functions/Analytic/ParaboloidSDF_impl.h b/src/TNL/Functions/Analytic/ParaboloidSDF_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..a4c1be5a03c72943502974ae7987139b791c6c9c
--- /dev/null
+++ b/src/TNL/Functions/Analytic/ParaboloidSDF_impl.h
@@ -0,0 +1,163 @@
+/***************************************************************************
+ ParaboloidSDFSDF_impl.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Analytic/ParaboloidSDF.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions, typename Real >
+ParaboloidSDFBase< dimensions, Real >::ParaboloidSDFBase()
+: xCenter( 0 ), yCenter( 0 ), zCenter( 0 ),
+ coefficient( 1 ), radius ( 0 )
+{
+}
+
+template< int dimensions, typename Real >
+bool ParaboloidSDFBase< dimensions, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix)
+{
+ this->xCenter = parameters.getParameter< double >( "x-center" );
+ this->yCenter = parameters.getParameter< double >( "y-center" );
+ this->zCenter = parameters.getParameter< double >( "z-center" );
+ this->coefficient = parameters.getParameter< double >( "coefficient" );
+ this->radius = parameters.getParameter< double >( "radius" );
+
+ return true;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidSDFBase< dimensions, Real >::setXCenter( const Real& xCenter )
+{
+ this->xCenter = xCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidSDFBase< dimensions, Real >::getXCenter() const
+{
+ return this->xCenter;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidSDFBase< dimensions, Real >::setYCenter( const Real& yCenter )
+{
+ this->yCenter = yCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidSDFBase< dimensions, Real >::getYCenter() const
+{
+ return this->yCenter;
+}
+template< int dimensions, typename Real >
+void ParaboloidSDFBase< dimensions, Real >::setZCenter( const Real& zCenter )
+{
+ this->zCenter = zCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidSDFBase< dimensions, Real >::getZCenter() const
+{
+ return this->zCenter;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidSDFBase< dimensions, Real >::setCoefficient( const Real& amplitude )
+{
+ this->coefficient = coefficient;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidSDFBase< dimensions, Real >::getCoefficient() const
+{
+ return this->coefficient;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidSDFBase< dimensions, Real >::setOffset( const Real& offset )
+{
+ this->radius = offset;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidSDFBase< dimensions, Real >::getOffset() const
+{
+ return this->radius;
+}
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+ParaboloidSDF< 1, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ if( YDiffOrder != 0 || ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 )
+ return ::sqrt( ( x - this -> xCenter ) * ( x - this -> xCenter ) ) - this->radius;
+ if( XDiffOrder == 1 )
+ return 1.0;
+ return 0.0;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+ParaboloidSDF< 2, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ const Real& y = v.y();
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return ::sqrt ( ( x - this -> xCenter ) * ( x - this -> xCenter )
+ + ( y - this -> yCenter ) * ( y - this -> yCenter ) ) - this->radius;
+ }
+ return 0.0;
+}
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+ParaboloidSDF< 3, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ const Real& y = v.y();
+ const Real& z = v.z();
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return ::sqrt( ( x - this -> xCenter ) * ( x - this -> xCenter )
+ + ( y - this -> yCenter ) * ( y - this -> yCenter )
+ + ( z - this -> zCenter ) * ( z - this -> zCenter ) ) - this->radius;
+ }
+ return 0.0;
+}
+
+ } //namespace Analytic
+ } // namepsace Functions
+} // namespace TNL
diff --git a/src/TNL/Functions/Analytic/Paraboloid_impl.h b/src/TNL/Functions/Analytic/Paraboloid_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..24b8a7a6aa5cdbbcf4496678c2f5c4e12fc7b63e
--- /dev/null
+++ b/src/TNL/Functions/Analytic/Paraboloid_impl.h
@@ -0,0 +1,185 @@
+/***************************************************************************
+ Paraboloid_impl.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Analytic/Paraboloid.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions, typename Real >
+ParaboloidBase< dimensions, Real >::ParaboloidBase()
+: xCenter( 0 ), yCenter( 0 ), zCenter( 0 ),
+ coefficient( 1 ), radius ( 0 )
+{
+}
+
+template< int dimensions, typename Real >
+bool ParaboloidBase< dimensions, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix)
+{
+ this->xCenter = parameters.getParameter< double >( "x-center" );
+ this->yCenter = parameters.getParameter< double >( "y-center" );
+ this->zCenter = parameters.getParameter< double >( "z-center" );
+ this->coefficient = parameters.getParameter< double >( "coefficient" );
+ this->radius = parameters.getParameter< double >( "radius" );
+
+ return true;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidBase< dimensions, Real >::setXCenter( const Real& xCenter )
+{
+ this->xCenter = xCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidBase< dimensions, Real >::getXCenter() const
+{
+ return this->xCenter;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidBase< dimensions, Real >::setYCenter( const Real& yCenter )
+{
+ this->yCenter = yCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidBase< dimensions, Real >::getYCenter() const
+{
+ return this->yCenter;
+}
+template< int dimensions, typename Real >
+void ParaboloidBase< dimensions, Real >::setZCenter( const Real& zCenter )
+{
+ this->zCenter = zCenter;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidBase< dimensions, Real >::getZCenter() const
+{
+ return this->zCenter;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidBase< dimensions, Real >::setCoefficient( const Real& amplitude )
+{
+ this->coefficient = coefficient;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidBase< dimensions, Real >::getCoefficient() const
+{
+ return this->coefficient;
+}
+
+template< int dimensions, typename Real >
+void ParaboloidBase< dimensions, Real >::setOffset( const Real& offset )
+{
+ this->radius = offset;
+}
+
+template< int dimensions, typename Real >
+Real ParaboloidBase< dimensions, Real >::getOffset() const
+{
+ return this->radius;
+}
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+Paraboloid< 1, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ if( YDiffOrder != 0 || ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 )
+ return this->coefficient * ( ( x - this -> xCenter ) * ( x - this -> xCenter ) - this->radius*this->radius );
+ if( XDiffOrder == 1 )
+ return 2.0 * this->coefficient * ( x - this -> xCenter );
+ return 0.0;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+Paraboloid< 2, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ const Real& y = v.y();
+ if( ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return this->coefficient * ( ( x - this -> xCenter ) * ( x - this -> xCenter )
+ + ( y - this -> yCenter ) * ( y - this -> yCenter ) - this->radius*this->radius );
+ }
+ if( XDiffOrder == 1 && YDiffOrder == 0)
+ return 2.0 * this->coefficient * ( x - this -> xCenter );
+ if( YDiffOrder == 1 && XDiffOrder == 0)
+ return 2.0 * this->coefficient * ( y - this -> yCenter );
+ if( XDiffOrder == 2 && YDiffOrder == 0)
+ return 2.0 * this->coefficient;
+ if( YDiffOrder == 2 && XDiffOrder == 0)
+ return 2.0 * this->coefficient;
+ return 0.0;
+}
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+Paraboloid< 3, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const Real& x = v.x();
+ const Real& y = v.y();
+ const Real& z = v.z();
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return this->coefficient * ( ( x - this -> xCenter ) * ( x - this -> xCenter )
+ + ( y - this -> yCenter ) * ( y - this -> yCenter )
+ + ( z - this -> zCenter ) * ( z - this -> zCenter ) - this->radius*this->radius );
+ }
+ if( XDiffOrder == 1 && YDiffOrder == 0 && ZDiffOrder == 0)
+ return 2.0 * this->coefficient * ( x - this -> xCenter );
+ if( YDiffOrder == 1 && XDiffOrder == 0 && ZDiffOrder == 0)
+ return 2.0 * this->coefficient * ( y - this -> yCenter );
+ if( ZDiffOrder == 1 && XDiffOrder == 0 && YDiffOrder == 0)
+ return 2.0 * this->coefficient * ( z - this -> zCenter );
+ if( XDiffOrder == 2 && YDiffOrder == 0 && ZDiffOrder == 0)
+ return 2.0 * this->coefficient;
+ if( YDiffOrder == 2 && XDiffOrder == 0 && ZDiffOrder == 0)
+ return 2.0 * this->coefficient;
+ if( ZDiffOrder == 2 && XDiffOrder == 0 && YDiffOrder == 0)
+ return 2.0 * this->coefficient;
+ return 0.0;
+}
+
+ } // namespace Analytic
+ } // namedspace Functions
+} // namespace TNL
diff --git a/src/TNL/Functions/Analytic/PseudoSquare.h b/src/TNL/Functions/Analytic/PseudoSquare.h
index 26d3086f23b1fa855acec067dc0b72a5ba878e1e..7f58906597338b78588cdf6e40940bd6b3b87d0b 100644
--- a/src/TNL/Functions/Analytic/PseudoSquare.h
+++ b/src/TNL/Functions/Analytic/PseudoSquare.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump.h - description
+ PseudoSquare.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/PseudoSquare_impl.h b/src/TNL/Functions/Analytic/PseudoSquare_impl.h
index 378e11ac11c720b573a46d922b4d66cdb63ee66b..ab81b85f0b3144f4cb8507f8a83cdfd1dbedf097 100644
--- a/src/TNL/Functions/Analytic/PseudoSquare_impl.h
+++ b/src/TNL/Functions/Analytic/PseudoSquare_impl.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump_impl.h - description
+ PseudoSquare_impl.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/SDFSchemeTest.h b/src/TNL/Functions/Analytic/SDFSchemeTest.h
new file mode 100644
index 0000000000000000000000000000000000000000..d3720fcb12f2a70c0593a1eafe7eeb8d00a58b0c
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SDFSchemeTest.h
@@ -0,0 +1,108 @@
+/***************************************************************************
+ SDFSchemeTest.h - description
+ -------------------
+ begin : Nov 19, 2013
+ copyright : (C) 2013 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+#include <functions/tnlSDFSinWaveFunction.h>
+#include <functions/tnlSDFSinWaveFunctionSDF.h>
+#include <functions/tnlSDFSinBumps.h>
+#include <functions/tnlSDFSinBumpsSDF.h>
+#include <functions/tnlExpBumpFunction.h>
+#include <functions/tnlSDFParaboloid.h>
+#include <functions/tnlSDFParaboloidSDF.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< typename function, typename Real = double >
+class SDFSchemeTestBase
+{
+ public:
+
+ SDFSchemeTestBase();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+
+ function f;
+};
+
+template< typename function, int Dimensions, typename Real >
+class SDFSchemeTest
+{
+
+};
+
+template< typename function, int Dimensions, typename Real >
+class SDFSchemeTest< function, 1, Real > : public SDFSchemeTestBase< function, Real >
+{
+ public:
+
+
+ enum { Dimensions = 1 };
+ typedef Vertex VertexType;
+ typedef typename VertexType::RealType RealType;
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ RealType getValue( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+
+
+};
+
+template< typename function, int Dimensions, typename Real >
+class SDFSchemeTest< function, 2, Real > : public SDFSchemeTestBase< function, Real >
+{
+ public:
+
+
+ enum { Dimensions = 2 };
+ typedef Vertex VertexType;
+ typedef typename VertexType::RealType RealType;
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ RealType getValue( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+
+};
+
+template< typename function, int Dimensions, typename Real >
+class SDFSchemeTest< function, 3, Real > : public SDFSchemeTestBase< function, Real >
+{
+ public:
+
+
+ enum { Dimensions = 3 };
+ typedef Vertex VertexType;
+ typedef typename VertexType::RealType RealType;
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ RealType getValue( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
+#include <functions/SDFSchemeTest_impl.h>
diff --git a/src/TNL/Functions/Analytic/SDFSchemeTest_impl.h b/src/TNL/Functions/Analytic/SDFSchemeTest_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..c2be9ef3433847562134f29aff2680d410378e3b
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SDFSchemeTest_impl.h
@@ -0,0 +1,78 @@
+/***************************************************************************
+ SDFSchemeTest_impl.h - description
+ -------------------
+ begin : Nov 19, 2013
+ copyright : (C) 2013 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <functions/SDFSchemeTest.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< typename function, typename Real >
+SDFSchemeTestBase< function, Real >::SDFSchemeTestBase()
+{
+}
+
+template< typename function, typename Real >
+bool SDFSchemeTestBase< function, Real >::v( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+{
+ f.init(parameters);
+
+ return true;
+}
+
+
+
+template< typename function, int Dimensions, typename Real >
+ template< int XDiffOrder, int YDiffOrder, int ZDiffOrder >
+Real SDFSchemeTest< function, 1, Real >::getValue( const Vertex& v,
+ const Real& time = 0.0 ) const
+{
+ if( YDiffOrder != 0 || ZDiffOrder != 0 || XDiffOrder != 0 )
+ return 0.0;
+
+ return sign( this->f.getValue<0,0,0>(v))*
+ ( 1-sqrt(this->f.getValue<1,0,0>(v)*this->f.getValue<1,0,0>(v)) );
+}
+
+
+template< typename function, int Dimensions, typename Real >
+ template< int XDiffOrder, int YDiffOrder, int ZDiffOrder >
+Real SDFSchemeTest< function, 2, Real >::getValue( const Vertex& v,
+ const Real& time = 0.0 ) const
+{
+ if( YDiffOrder != 0 || ZDiffOrder != 0 || XDiffOrder != 0 )
+ return 0.0;
+
+ return sign( this->f.getValue<0,0,0>(v))*
+ ( 1-sqrt(this->f.getValue<1,0,0>(v)*this->f.getValue<1,0,0>(v) +
+ this->f.getValue<0,1,0>(v)*this->f.getValue<0,1,0>(v)) );
+}
+
+template< typename function, int Dimensions, typename Real >
+ template< int XDiffOrder, int YDiffOrder, int ZDiffOrder >
+Real SDFSchemeTest< function, 3, Real >::getValue( const Vertex& v,
+ const Real& time = 0.0 ) const
+{
+ if( YDiffOrder != 0 || ZDiffOrder != 0 || XDiffOrder != 0 )
+ return 0.0;
+
+ return sign( this->f.getValue<0,0,0>(v))*
+ ( 1.0-sqrt(this->f.getValue<1,0,0>(v)*this->f.getValue<1,0,0>(v) +
+ this->f.getValue<0,1,0>(v)*this->f.getValue<0,1,0>(v) +
+ this->f.getValue<0,0,1>(v)*this->f.getValue<0,0,1>(v)) );
+}
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
diff --git a/src/TNL/Functions/Analytic/SinBumps.h b/src/TNL/Functions/Analytic/SinBumps.h
index d3cc0434582e8b0c6df273f3063f382d1bef3ebb..ea37dec8e547f4a11b1b9c422ab6fe78b8694a20 100644
--- a/src/TNL/Functions/Analytic/SinBumps.h
+++ b/src/TNL/Functions/Analytic/SinBumps.h
@@ -8,7 +8,11 @@
/* See Copyright Notice in tnl/Copyright */
-#pragma once
+/****
+ * Tomas Sobotik
+ */
+
+#pragma once
#include <TNL/Config/ParameterContainer.h>
#include <TNL/Containers/StaticVector.h>
@@ -39,11 +43,15 @@ class SinBumpsBase : public Domain< Vertex::size, SpaceDomain >
const VertexType& getPhase() const;
+ void setWavesNumber( const VertexType& wavesNumber );
+
+ const VertexType& getWavesNumber() const;
+
protected:
RealType amplitude;
- VertexType waveLength, phase;
+ VertexType waveLength, phase, wavesNumber;
};
template< int Dimensions, typename Real >
@@ -59,7 +67,6 @@ class SinBumps< 1, Real > : public SinBumpsBase< Containers::StaticVector< 1, R
typedef Real RealType;
typedef Containers::StaticVector< 1, RealType > VertexType;
-
SinBumps();
bool setup( const Config::ParameterContainer& parameters,
@@ -91,7 +98,6 @@ class SinBumps< 2, Real > : public SinBumpsBase< Containers::StaticVector< 2, Re
typedef Real RealType;
typedef Containers::StaticVector< 2, RealType > VertexType;
-
SinBumps();
@@ -163,4 +169,4 @@ std::ostream& operator << ( std::ostream& str, const SinBumps< Dimensions, Real
} // namespace Functions
} // namespace TNL
-#include <TNL/Functions/Analytic/SinBumps_impl.h>
\ No newline at end of file
+#include <TNL/Functions/Analytic/SinBumps_impl.h>
diff --git a/src/TNL/Functions/Analytic/SinBumpsSDF.h b/src/TNL/Functions/Analytic/SinBumpsSDF.h
new file mode 100644
index 0000000000000000000000000000000000000000..e402872d27958478e5adca8a0d17e6f24e12fa2e
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SinBumpsSDF.h
@@ -0,0 +1,172 @@
+/***************************************************************************
+ SinBumpsSDF.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+#include <TNL/Functions/Domain.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+
+template< typename Vertex >
+class SinBumpsSDFBase : public Domain< Vertex::size, SpaceDomain >
+{
+ public:
+
+ typedef Vertex VertexType;
+ typedef typename Vertex::RealType RealType;
+ enum { Dimensions = VertexType::size };
+
+ void setWaveLength( const VertexType& waveLength );
+
+ const VertexType& getWaveLength() const;
+
+ void setAmplitude( const RealType& amplitude );
+
+ const RealType& getAmplitude() const;
+
+ void setPhase( const VertexType& phase );
+
+ const VertexType& getPhase() const;
+
+ void setWavesNumber( const VertexType& wavesNumber );
+
+ const VertexType& getWavesNumber() const;
+
+ protected:
+
+ RealType amplitude;
+
+ VertexType waveLength, phase, wavesNumber;
+};
+
+template< int Dimensions, typename Real >
+class SinBumpsSDF
+{
+};
+
+template< typename Real >
+class SinBumpsSDF< 1, Real > : public SinBumpsSDFBase< Containers::StaticVector< 1, Real > >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 1, RealType > VertexType;
+
+
+ SinBumpsSDF();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class SinBumpsSDF< 2, Real > : public SinBumpsSDFBase< Containers::StaticVector< 2, Real > >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 2, RealType > VertexType;
+
+
+ SinBumpsSDF();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class SinBumpsSDF< 3, Real > : public SinBumpsSDFBase< Containers::StaticVector< 3, Real > >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 3, RealType > VertexType;
+
+ SinBumpsSDF();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< int Dimensions,
+ typename Real >
+std::ostream& operator << ( std::ostream& str, const SinBumpsSDF< Dimensions, Real >& f )
+{
+ str << "SDF Sin Bumps SDF. function: amplitude = " << f.getAmplitude()
+ << " wavelength = " << f.getWaveLength()
+ << " phase = " << f.getPhase();
+ return str;
+}
+
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
+#include <TNL/Functions/Analytic/SinBumpsSDF_impl.h>
diff --git a/src/TNL/Functions/Analytic/SinBumpsSDF_impl.h b/src/TNL/Functions/Analytic/SinBumpsSDF_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..6b38ae51238dfdca01404bcbb4002a663f409211
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SinBumpsSDF_impl.h
@@ -0,0 +1,242 @@
+/***************************************************************************
+ SinBumpsSDFSDF_impl.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Analytic/SinBumpsSDF.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< typename Vertex >
+void SinBumpsSDFBase< Vertex >::setWaveLength( const Vertex& waveLength )
+{
+ this->waveLength = waveLength;
+}
+
+template< typename Vertex >
+const Vertex& SinBumpsSDFBase< Vertex >::getWaveLength() const
+{
+ return this->waveLength;
+}
+
+template< typename Vertex >
+void SinBumpsSDFBase< Vertex >::setAmplitude( const typename Vertex::RealType& amplitude )
+{
+ this->amplitude = amplitude;
+}
+
+template< typename Vertex >
+const typename Vertex::RealType& SinBumpsSDFBase< Vertex >::getAmplitude() const
+{
+ return this->amplitude;
+}
+
+template< typename Vertex >
+void SinBumpsSDFBase< Vertex >::setPhase( const Vertex& phase )
+{
+ this->phase = phase;
+}
+
+template< typename Vertex >
+const Vertex& SinBumpsSDFBase< Vertex >::getPhase() const
+{
+ return this->phase;
+}
+
+/***
+ * 1D
+ */
+
+template< typename Real >
+SinBumpsSDF< 1, Real >::SinBumpsSDF()
+{
+}
+
+template< typename Real >
+bool SinBumpsSDF< 1, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix)
+{
+ this->amplitude = parameters.getParameter< double >( prefix+"amplitude" );
+ this->waveLength.x() = parameters.getParameter< double >( prefix+"wave-length-x" );
+ while(this->waveLength.x() > 2.0*M_PI)
+ this->waveLength.x() -= 2.0*M_PI;
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
+ this->phase.x() = parameters.getParameter< double >( prefix+"phase-x" );
+ return true;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+__cuda_callable__
+Real
+SinBumpsSDF< 1, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const RealType& x = v.x();
+ RealType xp = ::fabs( x ) + sign( x ) * this->phase.x() * this->waveLength.x() / ( 2.0*M_PI );
+ if( this->wavesNumber.x() != 0.0 && xp > this->wavesNumber.x() * this->waveLength.x() )
+ return 0.0;
+ if( YDiffOrder != 0 || ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 )
+ return sign( xp - round( (2.0 * xp ) / this->waveLength.x() ) * this->waveLength.x() / 2.0 )
+ * ( xp- round((2.0 * xp)/this->waveLength.x())* this->waveLength.x()/2.0)
+ * sign( ::sin(this-> phase.x() + 2.0 * M_PI * x / this->waveLength.x()));
+ if( XDiffOrder == 1 )
+ return 1.0;
+ return 0.0;
+}
+
+/****
+ * 2D
+ */
+
+template< typename Real >
+SinBumpsSDF< 2, Real >::SinBumpsSDF()
+{
+}
+
+template< typename Real >
+bool SinBumpsSDF< 2, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->amplitude = parameters.getParameter< double >( prefix+"amplitude" );
+ this->waveLength.x() = parameters.getParameter< double >( prefix+"wave-length-x" );
+ this->waveLength.y() = parameters.getParameter< double >( prefix+"wave-length-y" );
+ while(this->waveLength.x() > 2.0*M_PI)
+ this->waveLength.x() -= 2.0*M_PI;
+ while(this->waveLength.y() > 2.0*M_PI)
+ this->waveLength.y() -= 2.0*M_PI;
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
+ this->wavesNumber.y() = ceil( parameters.getParameter< double >( prefix+"waves-number-y" ) );
+ this->phase.x() = parameters.getParameter< double >( prefix+"phase-x" );
+ this->phase.y() = parameters.getParameter< double >( prefix+"phase-y" );
+ return true;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+__cuda_callable__
+Real
+SinBumpsSDF< 2, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const RealType& x = v.x();
+ const RealType& y = v.y();
+ RealType xp = ::sqrt(x*x) + sign( x ) * this->phase.x() * this->waveLength.x() / (2.0*M_PI);
+ RealType yp = ::sqrt(y*y) + sign( y ) * this->phase.y() * this->waveLength.y() / (2.0*M_PI);
+ if( ( xp > this->wavesNumber.x()*this->waveLength.x() && this->wavesNumber.x() != 0.0 ) ||
+ ( yp > this->wavesNumber.y()*this->waveLength.y() && this->wavesNumber.y() != 0.0 ) )
+ return 0.0;
+ const RealType sx = sign(xp - round((2.0 * xp)/this->waveLength.x())* this->waveLength.x()/2.0)
+ *(xp - round((2.0 * xp)/this->waveLength.x())* this->waveLength.x()/2.0);
+ const RealType sy = sign(yp - round((2.0 * yp)/this->waveLength.y())* this->waveLength.y()/2.0)
+ *(yp - round((2.0 * yp)/this->waveLength.y())* this->waveLength.y()/2.0);
+ RealType sxy;
+ if(sx < sy)
+ sxy = sx;
+ else
+ sxy = sy;
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return sxy * sign( ::sin( this->phase.x() + 2.0 * M_PI * x / this->waveLength.x() )
+ * ::sin( this->phase.y() + 2.0 * M_PI * y / this->waveLength.y() ) );
+ }
+ return 0.0;
+}
+
+/****
+ * 3D
+ */
+
+template< typename Real >
+SinBumpsSDF< 3, Real >::SinBumpsSDF()
+{
+}
+
+template< typename Real >
+bool SinBumpsSDF< 3, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->amplitude = parameters.getParameter< double >( prefix+"amplitude" );
+ this->waveLength.x() = parameters.getParameter< double >( prefix+"wave-length-x" );
+ this->waveLength.y() = parameters.getParameter< double >( prefix+"wave-length-y" );
+ this->waveLength.z() = parameters.getParameter< double >( prefix+"wave-length-z" );
+ while(this->waveLength.x() > 2.0*M_PI)
+ this->waveLength.x() -= 2.0*M_PI;
+ while(this->waveLength.y() > 2.0*M_PI)
+ this->waveLength.y() -= 2.0*M_PI;
+ while(this->waveLength.z() > 2.0*M_PI)
+ this->waveLength.z() -= 2.0*M_PI;
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
+ this->wavesNumber.y() = ceil( parameters.getParameter< double >( prefix+"waves-number-y" ) );
+ this->wavesNumber.z() = ceil( parameters.getParameter< double >( prefix+"waves-number-z" ) );
+ this->phase.x() = parameters.getParameter< double >( prefix+"phase-x" );
+ this->phase.y() = parameters.getParameter< double >( prefix+"phase-y" );
+ this->phase.z() = parameters.getParameter< double >(prefix+"phase-z" );
+ return true;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+__cuda_callable__
+Real
+SinBumpsSDF< 3, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const RealType& x = v.x();
+ const RealType& y = v.y();
+ const RealType& z = v.z();
+ RealType xp = ::sqrt(x*x) + sign(x)*(this->phase.x())*(this->waveLength.x())/(2.0*M_PI);
+ RealType yp = ::sqrt(y*y) + sign(y)*(this->phase.y())*(this->waveLength.y())/(2.0*M_PI);
+ RealType zp = ::sqrt(z*z) + sign(z)*(this->phase.z())*(this->waveLength.z())/(2.0*M_PI);
+ if ( ( xp > this->wavesNumber.x()*this->waveLength.x() && this->wavesNumber.x() != 0.0 ) ||
+ (yp > this->wavesNumber.y()*this->waveLength.y() && this->wavesNumber.y() != 0.0 ) ||
+ (::sqrt(z*z) > this->wavesNumber.z()*this->waveLength.z() && this->wavesNumber.z() != 0.0 ) )
+ return 0.0;
+ const RealType sx = sign(xp - round((2.0 * xp)/this->waveLength.x())* this->waveLength.x()/2.0)
+ *(xp - round((2.0 * xp)/this->waveLength.x())* this->waveLength.x()/2.0);
+ const RealType sy = sign(yp - round((2.0 * yp)/this->waveLength.y())* this->waveLength.y()/2.0)
+ *(yp - round((2.0 * yp)/this->waveLength.y())* this->waveLength.y()/2.0);
+ const RealType sz = sign(zp - round((2.0 * zp)/this->waveLength.z())* this->waveLength.z()/2.0)
+ *(zp - round((2.0 * zp)/this->waveLength.z())* this->waveLength.z()/2.0);
+ RealType sxyz;
+ if(sx <= sy && sx <= sz)
+ sxyz = sx;
+ else if ( sy <= sx && sy <= sz)
+ sxyz = sy;
+ else
+ sxyz = sz;
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ return sxyz * sign( ::sin( this->phase.x() + 2.0 * M_PI * x / this->waveLength.x() )
+ * ::sin( this->phase.y() + 2.0 * M_PI * y / this->waveLength.y() )
+ * ::sin( this->phase.z() + 2.0 * M_PI * z / this->waveLength.z() ) );
+ }
+ return 0.0;
+}
+
+ } // namespace Analytic
+ } // namespace Fucntions
+} // namespace TNL
\ No newline at end of file
diff --git a/src/TNL/Functions/Analytic/SinBumps_impl.h b/src/TNL/Functions/Analytic/SinBumps_impl.h
index 8ae3eeb6abdce64ee0db3df89c5e59ada351f58b..337b7db27296a1f72ae3d874f93a6979fb4fc434 100644
--- a/src/TNL/Functions/Analytic/SinBumps_impl.h
+++ b/src/TNL/Functions/Analytic/SinBumps_impl.h
@@ -8,6 +8,9 @@
/* See Copyright Notice in tnl/Copyright */
+/****
+ * Tomas Sobotik
+ */
#pragma once
#include <TNL/Functions/Analytic/SinBumps.h>
@@ -52,10 +55,21 @@ const Vertex& SinBumpsBase< Vertex >::getPhase() const
return this->phase;
}
+template< typename Vertex >
+void SinBumpsBase< Vertex >::setWavesNumber( const Vertex& wavesNumber )
+{
+ this->wavesNumber = wavesNumber;
+}
+
+template< typename Vertex >
+const Vertex& SinBumpsBase< Vertex >::getWavesNumber() const
+{
+ return this->wavesNumber;
+}
+
/***
* 1D
*/
-
template< typename Real >
SinBumps< 1, Real >::SinBumps()
{
@@ -68,6 +82,7 @@ bool SinBumps< 1, Real >::setup( const Config::ParameterContainer& parameters,
this->amplitude = parameters.getParameter< double >( prefix + "amplitude" );
this->waveLength.x() = parameters.getParameter< double >( prefix + "wave-length-x" );
this->phase.x() = parameters.getParameter< double >( prefix + "phase-x" );
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
return true;
}
@@ -82,9 +97,14 @@ SinBumps< 1, Real >::
getPartialDerivative( const VertexType& v,
const Real& time ) const
{
- const RealType& x = v.x();
if( YDiffOrder != 0 || ZDiffOrder != 0 )
return 0.0;
+
+ const RealType& x = v.x();
+ const RealType xp = ::fabs( x ) + sign( x ) * this->phase.x() * this->waveLength.x() / (2.0*M_PI);
+ if( this->wavesNumber.x() != 0.0 && xp > this->waveLength.x() * this->wavesNumber.x() )
+ return 0.0;
+
if( XDiffOrder == 0 )
return this->amplitude * ::sin( this->phase.x() + 2.0 * M_PI * x / this->waveLength.x() );
if( XDiffOrder == 1 )
@@ -108,7 +128,6 @@ operator()( const VertexType& v,
/****
* 2D
*/
-
template< typename Real >
SinBumps< 2, Real >::SinBumps()
{
@@ -123,10 +142,11 @@ bool SinBumps< 2, Real >::setup( const Config::ParameterContainer& parameters,
this->waveLength.y() = parameters.getParameter< double >( prefix + "wave-length-y" );
this->phase.x() = parameters.getParameter< double >( prefix + "phase-x" );
this->phase.y() = parameters.getParameter< double >( prefix + "phase-y" );
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
+ this->wavesNumber.y() = ceil( parameters.getParameter< double >( prefix+"waves-number-y" ) );
return true;
}
-
template< typename Real >
template< int XDiffOrder,
int YDiffOrder,
@@ -137,10 +157,18 @@ SinBumps< 2, Real>::
getPartialDerivative( const VertexType& v,
const Real& time ) const
{
+ if( ZDiffOrder != 0 )
+ return 0.0;
+
const RealType& x = v.x();
const RealType& y = v.y();
- if( ZDiffOrder != 0 )
+ const RealType xp = ::fabs( x ) + sign( x ) * this->phase.x() * this->waveLength.x() / (2.0*M_PI);
+ const RealType yp = ::fabs( y ) + sign( y ) * this->phase.y() * this->waveLength.y() / (2.0*M_PI);
+ //std::cerr << "this->wavesNumber.x() = " << this->wavesNumber.x() << "fabs( x ) = " << fabs( x ) << " 2.0*M_PI * this->waveLength.x() * this->wavesNumber.x() = " << 2.0*M_PI * this->waveLength.x() * this->wavesNumber.x() << std::endl;
+ if( ( this->wavesNumber.x() != 0.0 && xp > this->waveLength.x() * this->wavesNumber.x() ) ||
+ ( this->wavesNumber.y() != 0.0 && yp > this->waveLength.y() * this->wavesNumber.y() ) )
return 0.0;
+
if( XDiffOrder == 0 && YDiffOrder == 0 )
return this->amplitude *
::sin( this->phase.x() + 2.0 * M_PI * x / this->waveLength.x() ) *
@@ -171,7 +199,6 @@ operator()( const VertexType& v,
/****
* 3D
*/
-
template< typename Real >
SinBumps< 3, Real >::SinBumps()
{
@@ -188,10 +215,12 @@ bool SinBumps< 3, Real >::setup( const Config::ParameterContainer& parameters,
this->phase.x() = parameters.getParameter< double >( prefix + "phase-x" );
this->phase.y() = parameters.getParameter< double >( prefix + "phase-y" );
this->phase.z() = parameters.getParameter< double >( prefix + "phase-z" );
+ this->wavesNumber.x() = ceil( parameters.getParameter< double >( prefix+"waves-number-x" ) );
+ this->wavesNumber.y() = ceil( parameters.getParameter< double >( prefix+"waves-number-y" ) );
+ this->wavesNumber.z() = ceil( parameters.getParameter< double >( prefix+"waves-number-z" ) );
return true;
}
-
template< typename Real >
template< int XDiffOrder,
int YDiffOrder,
@@ -205,6 +234,16 @@ getPartialDerivative( const VertexType& v,
const RealType& x = v.x();
const RealType& y = v.y();
const RealType& z = v.z();
+
+ const RealType xp = ::fabs( x ) + sign( x ) * this->phase.x() * this->waveLength.x() / (2.0*M_PI);
+ const RealType yp = ::fabs( y ) + sign( y ) * this->phase.y() * this->waveLength.y() / (2.0*M_PI);
+ const RealType zp = ::fabs( z ) + sign( z ) * this->phase.z() * this->waveLength.z() / (2.0*M_PI);
+
+ if( ( this->wavesNumber.x() != 0.0 && xp > this->waveLength.x() * this->wavesNumber.x() ) ||
+ ( this->wavesNumber.y() != 0.0 && yp > this->waveLength.y() * this->wavesNumber.y() ) ||
+ ( this->wavesNumber.z() != 0.0 && zp > this->waveLength.z() * this->wavesNumber.z() ) )
+ return 0.0;
+
if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0)
return this->amplitude *
::sin( this->phase.x() + 2.0 * M_PI * x / this->waveLength.x() ) *
diff --git a/src/TNL/Functions/Analytic/SinWave.h b/src/TNL/Functions/Analytic/SinWave.h
index 98c8cdddc6e8f9d0a0433fca653ade91d5d85665..24549851fa98a8817e28430f909ad99e497811f8 100644
--- a/src/TNL/Functions/Analytic/SinWave.h
+++ b/src/TNL/Functions/Analytic/SinWave.h
@@ -8,6 +8,9 @@
/* See Copyright Notice in tnl/Copyright */
+/****
+ * Tomas Sobotik
+ */
#pragma once
#include <TNL/Config/ParameterContainer.h>
@@ -24,26 +27,32 @@ class SinWaveBase : public Domain< dimensions, SpaceDomain >
{
public:
- SinWaveBase();
+ SinWaveBase();
- bool setup( const Config::ParameterContainer& parameters,
- const String& prefix = "" );
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
- void setWaveLength( const Real& waveLength );
+ void setWaveLength( const Real& waveLength );
+
+ Real getWaveLength() const;
- Real getWaveLength() const;
+ void setAmplitude( const Real& amplitude );
- void setAmplitude( const Real& amplitude );
+ Real getAmplitude() const;
- Real getAmplitude() const;
+ void setPhase( const Real& phase );
- void setPhase( const Real& phase );
+ Real getPhase() const;
- Real getPhase() const;
+ void setWavesNumber( const Real& wavesNumber );
+
+ Real getWavesNumber() const;
protected:
+
+ bool isInsideWaves( const Real& distance ) const;
- Real waveLength, amplitude, phase, wavesNumber;
+ Real waveLength, amplitude, phase, wavesNumber;
};
template< int Dimensions, typename Real >
@@ -140,7 +149,8 @@ std::ostream& operator << ( std::ostream& str, const SinWave< Dimensions, Real >
{
str << "Sin Wave. function: amplitude = " << f.getAmplitude()
<< " wavelength = " << f.getWaveLength()
- << " phase = " << f.getPhase();
+ << " phase = " << f.getPhase()
+ << " waves number = " << f.getWavesNumber();
return str;
}
diff --git a/src/TNL/Functions/Analytic/SinWaveSDF.h b/src/TNL/Functions/Analytic/SinWaveSDF.h
new file mode 100644
index 0000000000000000000000000000000000000000..29278c35878b7b5c8f5eba6630959ee80b8df75f
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SinWaveSDF.h
@@ -0,0 +1,159 @@
+/***************************************************************************
+ SinWaveSDF.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+#include <TNL/Functions/Domain.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions,
+ typename Real = double >
+class SinWaveSDFBase : public Functions::Domain< dimensions, SpaceDomain >
+{
+ public:
+
+ SinWaveSDFBase();
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ void setWaveLength( const Real& waveLength );
+
+ Real getWaveLength() const;
+
+ void setAmplitude( const Real& amplitude );
+
+ Real getAmplitude() const;
+
+ void setPhase( const Real& phase );
+
+ Real getPhase() const;
+
+ void setWavesNumber( const Real& wavesNumber );
+
+ Real getWavesNumber() const;
+
+ protected:
+
+ __cuda_callable__
+ Real sinWaveFunctionSDF( const Real& r ) const;
+
+ Real waveLength, amplitude, phase, wavesNumber;
+};
+
+template< int Dimensions, typename Real >
+class SinWaveSDF
+{
+};
+
+template< typename Real >
+class SinWaveSDF< 1, Real > : public SinWaveSDFBase< 1, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 1, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class SinWaveSDF< 2, Real > : public SinWaveSDFBase< 2, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 2, RealType > VertexType;
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< typename Real >
+class SinWaveSDF< 3, Real > : public SinWaveSDFBase< 3, Real >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< 3, RealType > VertexType;
+
+
+
+#ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+#else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+#endif
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const;
+
+};
+
+template< int Dimensions,
+ typename Real >
+std::ostream& operator << ( std::ostream& str, const SinWaveSDF< Dimensions, Real >& f )
+{
+ str << "SDF Sin Wave SDF. function: amplitude = " << f.getAmplitude()
+ << " wavelength = " << f.getWaveLength()
+ << " phase = " << f.getPhase()
+ << " # of waves = " << f.getWavesNumber();
+ return str;
+}
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
+
+#include <TNL/Functions/Analytic/SinWaveSDF_impl.h>
diff --git a/src/TNL/Functions/Analytic/SinWaveSDF_impl.h b/src/TNL/Functions/Analytic/SinWaveSDF_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..5d010df173f833e1d480c28a92886ada7c59e460
--- /dev/null
+++ b/src/TNL/Functions/Analytic/SinWaveSDF_impl.h
@@ -0,0 +1,167 @@
+/***************************************************************************
+ tnlSDFSinWaveFunctionSDFSDF_impl.h - description
+ -------------------
+ begin : Oct 13, 2014
+ copyright : (C) 2014 by Tomas Sobotik
+
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Analytic/SinWaveSDF.h>
+
+namespace TNL {
+ namespace Functions {
+ namespace Analytic {
+
+template< int dimensions, typename Real >
+SinWaveSDFBase< dimensions, Real >::SinWaveSDFBase()
+: waveLength( 1.0 ),
+ amplitude( 1.0 ),
+ phase( 0 ),
+ wavesNumber( 0 )
+{
+}
+
+template< int dimensions, typename Real >
+bool SinWaveSDFBase< dimensions, Real >::setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ this->waveLength = parameters.getParameter< double >( prefix + "wave-length" );
+ this->amplitude = parameters.getParameter< double >( prefix + "amplitude" );
+ this->phase = parameters.getParameter< double >( prefix + "phase" );
+ while(this->phase >2.0*M_PI)
+ this->phase -= 2.0*M_PI;
+ this->wavesNumber = ceil( parameters.getParameter< double >( prefix + "waves-number" ) );
+ return true;
+}
+
+template< int dimensions, typename Real >
+void SinWaveSDFBase< dimensions, Real >::setWaveLength( const Real& waveLength )
+{
+ this->waveLength = waveLength;
+}
+
+template< int dimensions, typename Real >
+Real SinWaveSDFBase< dimensions, Real >::getWaveLength() const
+{
+ return this->waveLength;
+}
+
+template< int dimensions, typename Real >
+void SinWaveSDFBase< dimensions, Real >::setAmplitude( const Real& amplitude )
+{
+ this->amplitude = amplitude;
+}
+
+template< int dimensions, typename Real >
+Real SinWaveSDFBase< dimensions, Real >::getAmplitude() const
+{
+ return this->amplitude;
+}
+
+template< int dimensions, typename Real >
+void SinWaveSDFBase< dimensions, Real >::setPhase( const Real& phase )
+{
+ this->phase = phase;
+}
+
+template< int dimensions, typename Real >
+Real SinWaveSDFBase< dimensions, Real >::getPhase() const
+{
+ return this->phase;
+}
+
+template< int dimensions, typename Real >
+void SinWaveSDFBase< dimensions, Real >::setWavesNumber( const Real& wavesNumber )
+{
+ this->wavesNumber = wavesNumber;
+}
+
+template< int dimensions, typename Real >
+Real SinWaveSDFBase< dimensions, Real >::getWavesNumber() const
+{
+ return this->wavesNumber;
+}
+
+template< int dimensions, typename Real >
+__cuda_callable__
+Real SinWaveSDFBase< dimensions, Real >::sinWaveFunctionSDF( const Real& r ) const
+{
+ if( this->wavesNumber == 0.0 || r < this->wavesNumber * this->waveLength )
+ return sign( r - round( 2.0 * r / this->waveLength ) * this->waveLength / 2.0 )
+ * ( r - round( 2.0 * r / this->waveLength ) * this->waveLength / 2.0 )
+ * sign( ::sin( 2.0 * M_PI * r / this->waveLength ) );
+ else
+ return r - this->wavesNumber * this->waveLength;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+SinWaveSDF< 1, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ if( YDiffOrder != 0 || ZDiffOrder != 0 )
+ return 0.0;
+ const RealType& x = v.x();
+ const RealType distance = ::sqrt( x * x ) + this->phase * this->waveLength / (2.0*M_PI);
+ if( XDiffOrder == 0 )
+ return this->sinWaveFunctionSDF( distance );
+ TNL_ASSERT( false, std::cerr << "TODO: implement this" );
+ return 0.0;
+}
+
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+SinWaveSDF< 2, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ if( ZDiffOrder != 0 )
+ return 0.0;
+
+ const RealType& x = v.x();
+ const RealType& y = v.y();
+ const RealType distance = ::sqrt( x * x + y * y ) + this->phase * this->waveLength / (2.0*M_PI);
+ if( XDiffOrder == 0 && YDiffOrder == 0)
+ return this->sinWaveFunctionSDF( distance );
+ TNL_ASSERT( false, std::cerr << "TODO: implement this" );
+ return 0.0;
+}
+
+template< typename Real >
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder>
+__cuda_callable__
+Real
+SinWaveSDF< 3, Real >::
+getPartialDerivative( const VertexType& v,
+ const Real& time ) const
+{
+ const RealType& x = v.x();
+ const RealType& y = v.y();
+ const RealType& z = v.z();
+ const RealType distance = ::sqrt( x * x + y * y + z * z ) + this->phase * this->waveLength / (2.0*M_PI);
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ return this->sinWaveFunctionSDF( distance );
+ TNL_ASSERT( false, std::cerr << "TODO: implement this" );
+ return 0.0;
+}
+
+ } // namespace Analytic
+ } // namespace Functions
+} // namespace TNL
diff --git a/src/TNL/Functions/Analytic/Twins.h b/src/TNL/Functions/Analytic/Twins.h
index ebdb507ac79e84681396715254ff4d4e30ff1b9e..d6627cdef82b7f28e69b202c48fe9bd1f91870ed 100644
--- a/src/TNL/Functions/Analytic/Twins.h
+++ b/src/TNL/Functions/Analytic/Twins.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump.h - description
+ Twins.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/Twins_impl.h b/src/TNL/Functions/Analytic/Twins_impl.h
index 222f9a452adf289bcc9282c75cf2014ecc23c289..2b294320fed642d15cc9b2d7355ff2aa3f7f7a02 100644
--- a/src/TNL/Functions/Analytic/Twins_impl.h
+++ b/src/TNL/Functions/Analytic/Twins_impl.h
@@ -1,5 +1,5 @@
/***************************************************************************
- ExpBump_impl.h - description
+ Twins_impl.h - description
-------------------
begin : Dec 5, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/TNL/Functions/Analytic/VectorNorm.h b/src/TNL/Functions/Analytic/VectorNorm.h
new file mode 100644
index 0000000000000000000000000000000000000000..7dd48b1a320c62d222934ed71d8481cb64ef1da1
--- /dev/null
+++ b/src/TNL/Functions/Analytic/VectorNorm.h
@@ -0,0 +1,293 @@
+/***************************************************************************
+ VectorNorm.h - description
+ -------------------
+ begin : Feb 12, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Math.h>
+#include <TNL/Assert.h>
+
+namespace TNL {
+namespace Functions {
+namespace Analytic {
+
+template< int Dimensions_,
+ typename Real >
+class VectorNormBase : public Domain< Dimensions_, SpaceDomain >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions_, RealType > VertexType;
+
+ VectorNormBase()
+ : center( 0.0 ),
+ anisotropy( 1.0 ),
+ power( 2.0 ),
+ radius( 0.0 ),
+ multiplicator( 1.0 ),
+ maxNorm( false ){};
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "center-0", "x-coordinate of the coordinates origin for the vector norm.", 0.0 );
+ config.addEntry< double >( prefix + "center-1", "y-coordinate of the coordinates origin for the vector norm.", 0.0 );
+ config.addEntry< double >( prefix + "center-2", "z-coordinate of the coordinates origin for the vector norm.", 0.0 );
+ config.addEntry< double >( prefix + "anisotropy-0", "x-coordinate of the linear anisotropy of the vector norm.", 1.0 );
+ config.addEntry< double >( prefix + "anisotropy-1", "y-coordinate of the linear anisotropy of the vector norm.", 1.0 );
+ config.addEntry< double >( prefix + "anisotropy-2", "z-coordinate of the linear anisotropy of the vector norm.", 1.0 );
+ config.addEntry< double >( prefix + "power", "The p coefficient of the L-p vector norm", 2.0 );
+ config.addEntry< double >( prefix + "radius", "Radius of the zero-th level-set.", 0.0 );
+ config.addEntry< double >( prefix + "multiplicator", "Outer multiplicator of the norm - -1.0 turns the function graph upside/down.", 1.0 );
+ config.addEntry< bool >( prefix + "max-norm", "Turn to 'true' to get maximum norm.", false );
+ }
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->power = parameters.template getParameter< double >( prefix + "power" );
+ this->maxNorm = parameters.template getParameter< bool >( prefix + "max-norm" );
+ this->radius = parameters.template getParameter< double >( prefix + "radius" );
+ this->multiplicator = parameters.template getParameter< double >( prefix + "multiplicator" );
+ return( this->center.setup( parameters, prefix + "center-") &&
+ this->anisotropy.setup( parameters, prefix + "anisotropy-" ) );
+ };
+
+ void setCenter( const VertexType& center )
+ {
+ this->center = center;
+ };
+
+ const RealType& getCenter() const
+ {
+ return this->center;
+ }
+
+ void setAnisotropy( const VertexType& anisotropy )
+ {
+ this->anisotropy = anisotropy;
+ };
+
+ const RealType& getAnisotropy() const
+ {
+ return this->anisotropy;
+ }
+
+ void setPower( const RealType& power )
+ {
+ this->power = power;
+ }
+
+ const RealType& getPower() const
+ {
+ return this->power;
+ }
+
+ void setRadius( const RealType& radius )
+ {
+ this->radius = radius;
+ }
+
+ const RealType& getRadius() const
+ {
+ return this->radius;
+ }
+
+ void setMultiplicator( const RealType& multiplicator )
+ {
+ this->multiplicator = multiplicator;
+ }
+
+ const RealType& getMultiplicator() const
+ {
+ return this->multiplicator;
+ }
+
+ void setMaxNorm( bool maxNorm )
+ {
+ this->maxNorm = maxNorm;
+ }
+
+ const RealType& getMaxNorm() const
+ {
+ return this->maxNorm;
+ }
+
+ protected:
+
+ VertexType center, anisotropy;
+
+ RealType power, radius, multiplicator;
+
+ bool maxNorm;
+};
+
+template< int Dimensions,
+ typename Real >
+class VectorNorm
+{
+};
+
+template< typename Real >
+class VectorNorm< 1, Real > : public VectorNormBase< 1, Real >
+{
+ public:
+
+ typedef VectorNormBase< 1, Real > BaseType;
+ using typename BaseType::RealType;
+ using typename BaseType::VertexType;
+
+ static String getType();
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ const RealType& x = v.x() - this->center.x();
+ if( YDiffOrder != 0 || ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 )
+ {
+ return this->multiplicator * ( TNL::abs( x ) * this->anisotropy.x() - this->radius );
+ }
+ if( XDiffOrder == 1 )
+ {
+ return this->multiplicator * TNL::sign( x ) * this->anisotropy.x();
+ }
+ return 0.0;
+ }
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const RealType& time = 0.0 ) const
+ {
+ return this->template getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+};
+
+template< typename Real >
+class VectorNorm< 2, Real > : public VectorNormBase< 2, Real >
+{
+ public:
+
+ typedef VectorNormBase< 2, Real > BaseType;
+ using typename BaseType::RealType;
+ using typename BaseType::VertexType;
+
+ static String getType();
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__ inline
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ const RealType& x = v.x() - this->center.x();
+ const RealType& y = v.y() - this->center.y();
+ if( ZDiffOrder != 0 )
+ return 0.0;
+ if( XDiffOrder == 0 && YDiffOrder == 0 )
+ {
+ if( this->maxNorm )
+ return ( TNL::max( TNL::abs( x ) * this->anisotropy.x(),
+ TNL::abs( y ) * this->anisotropy.y() ) - this->radius ) * this->multiplicator;
+ if( this->power == 1.0 )
+ return ( ( TNL::abs( x ) * this->anisotropy.x() +
+ TNL::abs( y ) * this->anisotropy.y() ) - this->radius ) * this->multiplicator;
+ if( this->power == 2.0 )
+ return ( std::sqrt( x * x * this->anisotropy.x() +
+ y * y * this->anisotropy.y() ) - this->radius ) * this->multiplicator;
+ return ( std::pow( std::pow( TNL::abs( x ), this->power ) * this->anisotropy.x() +
+ std::pow( TNL::abs( y ), this->power ) * this->anisotropy.y(), 1.0 / this-> power ) - this->radius ) * this->multiplicator;
+ }
+ TNL_ASSERT( false, std::cerr << "Not implemented yet." << std::endl );
+ return 0.0;
+ }
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ return this->template getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+};
+
+template< typename Real >
+class VectorNorm< 3, Real > : public VectorNormBase< 3, Real >
+{
+ public:
+
+ typedef VectorNormBase< 3, Real > BaseType;
+ using typename BaseType::RealType;
+ using typename BaseType::VertexType;
+
+ static String getType();
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ const RealType& x = v.x() - this->center.x();
+ const RealType& y = v.y() - this->center.y();
+ const RealType& z = v.z() - this->center.z();
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ {
+ if( this->maxNorm )
+ return ( TNL::max( TNL::max( TNL::abs( x ) * this->anisotropy.x(),
+ TNL::abs( y ) * this->anisotropy.y() ),
+ TNL::abs( z ) * this->anisotropy.z() ) - this->radius ) * this->multiplicator;
+ if( this->power == 1.0 )
+ return ( ( TNL::abs( x ) * this->anisotropy.x() +
+ TNL::abs( y ) * this->anisotropy.y() +
+ TNL::abs( z ) * this->anisotropy.z() ) - this->radius ) * this->multiplicator;
+ if( this->power == 2.0 )
+ return ( std::sqrt( x * x * this->anisotropy.x() +
+ y * y * this->anisotropy.y() +
+ z * z * this->anisotropy.z() ) - this->radius ) * this->multiplicator ;
+ return ( std::pow( std::pow( TNL::abs( x ), this->power ) * this->anisotropy.x() +
+ std::pow( TNL::abs( y ), this->power ) * this->anisotropy.y() +
+ std::pow( TNL::abs( z ), this->power ) * this->anisotropy.z(), 1.0 / this-> power ) - this->radius ) * this->multiplicator;
+ }
+ TNL_ASSERT( false, std::cerr << "Not implemented yet." << std::endl );
+ return 0.0;
+ }
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const Real& time = 0.0 ) const
+ {
+ return this->template getPartialDerivative< 0, 0, 0 >( v, time );
+ }
+};
+
+template< int Dimensions,
+ typename Real >
+std::ostream& operator << ( std::ostream& str, const VectorNorm< Dimensions, Real >& f )
+{
+ str << "VectorNorm. function: multiplicator = " << f.getMultiplicator() << " sigma = " << f.getSigma();
+ return str;
+}
+
+} // namespace Analytic
+} // namespace Functions
+} // namespace TNL
+
+
+
+
+
+
diff --git a/src/TNL/Functions/CMakeLists.txt b/src/TNL/Functions/CMakeLists.txt
index 04fe4b86805ddbdd11f0d9a73492918db7baeb5d..7727d5539ffb2a475e3c27f3b71b6ba997a23f52 100755
--- a/src/TNL/Functions/CMakeLists.txt
+++ b/src/TNL/Functions/CMakeLists.txt
@@ -14,7 +14,11 @@ SET( headers Domain.h
MeshFunctionVTKWriter_impl.h
OperatorFunction.h
TestFunction.h
- TestFunction_impl.h )
+ TestFunction_impl.h
+ VectorField.h
+ VectorFieldGnuplotWriter.h
+ VectorFieldGnuplotWriter_impl.h
+ )
SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/TNL/Functions )
set( common_SOURCES
diff --git a/src/TNL/Functions/MeshFunction.h b/src/TNL/Functions/MeshFunction.h
index a3cf8a46c307312ff0e10127da53339c48ddb3e4..35fee42d13beb047bc4d11b108edeb0ca1890f08 100644
--- a/src/TNL/Functions/MeshFunction.h
+++ b/src/TNL/Functions/MeshFunction.h
@@ -84,7 +84,6 @@ class MeshFunction :
const SharedPointer< Vector >& dataPtr,
const IndexType& offset = 0 );
-
void setMesh( const MeshPointer& meshPointer );
template< typename Device = Devices::Host >
@@ -93,6 +92,8 @@ class MeshFunction :
const MeshPointer& getMeshPointer() const;
+ __cuda_callable__ static IndexType getDofs( const MeshPointer& meshPointer );
+
__cuda_callable__ const VectorType& getData() const;
__cuda_callable__ VectorType& getData();
@@ -144,7 +145,8 @@ class MeshFunction :
bool boundLoad( File& file );
bool write( const String& fileName,
- const String& format = "vtk" ) const;
+ const String& format = "vtk",
+ const double& scale = 1.0 ) const;
using Object::save;
diff --git a/src/TNL/Functions/MeshFunctionGnuplotWriter.h b/src/TNL/Functions/MeshFunctionGnuplotWriter.h
index 887498711e8f3e4422a7cc99994cfecc2301504d..2c21a10328d4bd49bb6f3501246ce476cca83bd2 100644
--- a/src/TNL/Functions/MeshFunctionGnuplotWriter.h
+++ b/src/TNL/Functions/MeshFunctionGnuplotWriter.h
@@ -23,7 +23,8 @@ class MeshFunctionGnuplotWriter
public:
static bool write( const MeshFunction& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -41,7 +42,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 1, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -59,7 +61,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
@@ -78,7 +81,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 2, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -96,7 +100,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 1, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -114,7 +119,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
@@ -133,7 +139,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 3, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -151,7 +158,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 2, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
/***
@@ -169,7 +177,8 @@ class MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
};
} // namespace Functions
diff --git a/src/TNL/Functions/MeshFunctionGnuplotWriter_impl.h b/src/TNL/Functions/MeshFunctionGnuplotWriter_impl.h
index 341ed711a4cb4da60f3f6501a5dec0f8b736925a..9c442de639194e909ca5d3f54c491eede082f945 100644
--- a/src/TNL/Functions/MeshFunctionGnuplotWriter_impl.h
+++ b/src/TNL/Functions/MeshFunctionGnuplotWriter_impl.h
@@ -19,7 +19,8 @@ template< typename MeshFunction >
bool
MeshFunctionGnuplotWriter< MeshFunction >::
write( const MeshFunction& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
std::cerr << "Gnuplot writer for mesh functions defined on mesh type " << MeshFunction::MeshType::getType() << " is not (yet) implemented." << std::endl;
return false;
@@ -35,7 +36,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 1, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Cell entity( mesh );
@@ -46,7 +48,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
}
@@ -61,7 +63,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Vertex entity( mesh );
@@ -72,7 +75,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
}
@@ -88,7 +91,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 2, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Cell entity( mesh );
@@ -103,7 +107,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -120,7 +124,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 1, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typedef typename MeshType::Face EntityType;
@@ -139,7 +144,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -158,7 +163,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -176,7 +181,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Vertex entity( mesh );
@@ -191,7 +197,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -209,7 +215,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 3, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Cell entity( mesh );
@@ -227,7 +234,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " " << v.z() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -244,7 +251,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 2, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typedef typename MeshType::Face EntityType;
@@ -266,7 +274,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " " << v.z() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -286,7 +294,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " " << v.z() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -306,7 +314,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " " << v.z() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
@@ -324,7 +332,8 @@ template< typename MeshReal,
bool
MeshFunctionGnuplotWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
const MeshType& mesh = function.getMesh();
typename MeshType::Vertex entity( mesh );
@@ -342,7 +351,7 @@ write( const MeshFunctionType& function,
entity.refresh();
typename MeshType::VertexType v = entity.getCenter();
str << v.x() << " " << v.y() << " " << v.z() << " "
- << function.getData().getElement( entity.getIndex() ) << std::endl;
+ << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
str << std::endl;
}
diff --git a/src/TNL/Functions/MeshFunctionVTKWriter.h b/src/TNL/Functions/MeshFunctionVTKWriter.h
index 39651c27450c407639c1278f62220b82f51f240b..b4d18e4490901c2824ff922fe10f18607198e780 100644
--- a/src/TNL/Functions/MeshFunctionVTKWriter.h
+++ b/src/TNL/Functions/MeshFunctionVTKWriter.h
@@ -23,7 +23,9 @@ class MeshFunctionVTKWriter
public:
static bool write( const MeshFunction& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunction& function,
std::ostream& str ){}
};
@@ -43,7 +45,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 1, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -63,7 +67,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -83,7 +89,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 2, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -103,7 +111,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 1, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -123,7 +133,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -143,7 +155,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 3, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -163,7 +177,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 2, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -183,7 +199,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 1, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
@@ -203,7 +221,9 @@ class MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, Me
typedef Functions::MeshFunction< MeshType, 0, RealType > MeshFunctionType;
static bool write( const MeshFunctionType& function,
- std::ostream& str );
+ std::ostream& str,
+ const double& scale );
+
static void writeHeader(const MeshFunctionType& function,
std::ostream& str );
};
diff --git a/src/TNL/Functions/MeshFunctionVTKWriter_impl.h b/src/TNL/Functions/MeshFunctionVTKWriter_impl.h
index b432671764c0545c64e5c8ea241c820525f20f05..91a589bbe81c6261b0ee61882a3b03824d970b62 100644
--- a/src/TNL/Functions/MeshFunctionVTKWriter_impl.h
+++ b/src/TNL/Functions/MeshFunctionVTKWriter_impl.h
@@ -19,7 +19,8 @@ template< typename MeshFunction >
bool
MeshFunctionVTKWriter< MeshFunction >::
write( const MeshFunction& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
std::cerr << "VTK writer for mesh functions defined on mesh type " << MeshFunction::MeshType::getType() << " is not (yet) implemented." << std::endl;
return false;
@@ -54,7 +55,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 1, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -88,7 +90,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Cell entity = mesh.template getEntity< typename MeshType::Cell >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -123,7 +125,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -157,7 +160,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Vertex entity = mesh.template getEntity< typename MeshType::Vertex >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -192,7 +195,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 2, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -237,7 +241,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Cell entity = mesh.template getEntity< typename MeshType::Cell >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -272,7 +276,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 1, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
typedef typename MeshType::template MeshEntity< 0 > Vertex;
typedef typename MeshType::template MeshEntity< 1 > Face;
@@ -326,7 +331,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Face entity = mesh.template getEntity< typename MeshType::Face >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -361,7 +366,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
typedef typename MeshType::template MeshEntity< 0 > Vertex;
writeHeader(function, str);
@@ -405,7 +411,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Vertex entity = mesh.template getEntity< typename MeshType::Vertex >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -440,7 +446,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 3, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -501,7 +508,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Cell entity = mesh.template getEntity< typename MeshType::Cell >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -536,7 +543,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 2, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -620,7 +628,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Face entity = mesh.template getEntity< typename MeshType::Face >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -655,7 +663,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 1, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -733,7 +742,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Edge entity = mesh.template getEntity< typename MeshType::Edge >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
@@ -768,7 +777,8 @@ template< typename MeshReal,
bool
MeshFunctionVTKWriter< MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 0, Real > >::
write( const MeshFunctionType& function,
- std::ostream& str )
+ std::ostream& str,
+ const double& scale )
{
writeHeader(function, str);
@@ -820,7 +830,7 @@ write( const MeshFunctionType& function,
{
typename MeshType::Vertex entity = mesh.template getEntity< typename MeshType::Vertex >( i );
entity.refresh();
- str << function.getData().getElement( entity.getIndex() ) << std::endl;
+ str << scale * function.getData().getElement( entity.getIndex() ) << std::endl;
}
return true;
diff --git a/src/TNL/Functions/MeshFunction_impl.h b/src/TNL/Functions/MeshFunction_impl.h
index 53ab95680738b8ee77c7a76ca7de9e7344f3f3a8..8f365c4d88a9993ee1c28a74fc0547d7f02cb8f4 100644
--- a/src/TNL/Functions/MeshFunction_impl.h
+++ b/src/TNL/Functions/MeshFunction_impl.h
@@ -163,6 +163,7 @@ setup( const MeshPointer& meshPointer,
else
{
std::cerr << "Missing parameter " << prefix << "file." << std::endl;
+ throw(0);
return false;
}
return true;
@@ -250,6 +251,17 @@ getMeshPointer() const
return this->meshPointer;
}
+template< typename Mesh,
+ int MeshEntityDimensions,
+ typename Real >
+__cuda_callable__
+typename MeshFunction< Mesh, MeshEntityDimensions, Real >::IndexType
+MeshFunction< Mesh, MeshEntityDimensions, Real >::
+getDofs( const MeshPointer& meshPointer )
+{
+ return meshPointer->template getEntitiesCount< MeshEntityDimensions >();
+}
+
template< typename Mesh,
int MeshEntityDimensions,
typename Real >
@@ -482,19 +494,20 @@ template< typename Mesh,
bool
MeshFunction< Mesh, MeshEntityDimensions, Real >::
write( const String& fileName,
- const String& format ) const
+ const String& format,
+ const double& scale ) const
{
std::fstream file;
file.open( fileName.getString(), std::ios::out );
if( ! file )
{
- std::cerr << "Unbable to open a file " << fileName << "." << std::endl;
+ std::cerr << "Unable to open a file " << fileName << "." << std::endl;
return false;
}
if( format == "vtk" )
- return MeshFunctionVTKWriter< ThisType >::write( *this, file );
+ return MeshFunctionVTKWriter< ThisType >::write( *this, file, scale );
else if( format == "gnuplot" )
- return MeshFunctionGnuplotWriter< ThisType >::write( *this, file );
+ return MeshFunctionGnuplotWriter< ThisType >::write( *this, file, scale );
else {
std::cerr << "Unknown output format: " << format << std::endl;
return false;
diff --git a/src/TNL/Functions/OperatorFunction.h b/src/TNL/Functions/OperatorFunction.h
index 628191165d38d3ae12052091e15b01ba43e852b7..4eaf24156e0152ac8d74210cef3c3c83752f1c38 100644
--- a/src/TNL/Functions/OperatorFunction.h
+++ b/src/TNL/Functions/OperatorFunction.h
@@ -30,18 +30,19 @@ namespace Functions {
*/
template< typename Operator,
- typename MeshFunction,
+ typename Function = typename Operator::FunctionType,
typename BoundaryConditions = void,
- bool EvaluateOnFly = false >
+ bool EvaluateOnFly = false,
+ bool IsAnalytic = ( Function::getDomainType() == SpaceDomain || Function::getDomainType() == NonspaceDomain ) >
class OperatorFunction{};
/****
* Specialization for 'On the fly' evaluation with the boundary conditions does not make sense.
*/
template< typename Operator,
- typename MeshFunction,
+ typename Function,
typename BoundaryConditions >
-class OperatorFunction< Operator, MeshFunction, BoundaryConditions, true >
+class OperatorFunction< Operator, Function, BoundaryConditions, true, false >
: public Domain< Operator::getMeshDimensions(), MeshDomain >
{
};
@@ -50,21 +51,21 @@ class OperatorFunction< Operator, MeshFunction, BoundaryConditions, true >
* Specialization for 'On the fly' evaluation and no boundary conditions.
*/
template< typename Operator,
- typename MeshFunctionT >
-class OperatorFunction< Operator, MeshFunctionT, void, true >
+ typename Function >
+class OperatorFunction< Operator, Function, void, true, false >
: public Domain< Operator::getDomainDimensions(), Operator::getDomainType() >
{
public:
- static_assert( MeshFunctionT::getDomainType() == MeshDomain ||
- MeshFunctionT::getDomainType() == MeshInteriorDomain ||
- MeshFunctionT::getDomainType() == MeshBoundaryDomain,
- "Only mesh preimageFunctions may be used in the operator preimageFunction. Use ExactOperatorFunction instead of OperatorFunction." );
- static_assert( std::is_same< typename Operator::MeshType, typename MeshFunctionT::MeshType >::value,
+ static_assert( Function::getDomainType() == MeshDomain ||
+ Function::getDomainType() == MeshInteriorDomain ||
+ Function::getDomainType() == MeshBoundaryDomain,
+ "Only mesh preimageFnctions may be used in the operator preimageFunction. Use ExactOperatorFunction instead of OperatorFunction." );
+ static_assert( std::is_same< typename Operator::MeshType, typename Function::MeshType >::value,
"Both, operator and mesh preimageFunction must be defined on the same mesh." );
typedef Operator OperatorType;
- typedef MeshFunctionT FunctionType;
+ typedef Function FunctionType;
typedef typename OperatorType::MeshType MeshType;
typedef typename OperatorType::RealType RealType;
typedef typename OperatorType::DeviceType DeviceType;
@@ -90,7 +91,7 @@ class OperatorFunction< Operator, MeshFunctionT, void, true >
const MeshPointer& getMeshPointer() const
{
- tnlTNL_ASSERT( this->preimageFunction, std::cerr << "The preimage function was not set." << std::endl );
+ TNL_ASSERT( this->preimageFunction, std::cerr << "The preimage function was not set." << std::endl );
return this->preimageFunction->getMeshPointer();
};
@@ -128,17 +129,17 @@ class OperatorFunction< Operator, MeshFunctionT, void, true >
* Specialization for precomputed evaluation and no boundary conditions.
*/
template< typename Operator,
- typename PreimageFunction >
-class OperatorFunction< Operator, PreimageFunction, void, false >
+ typename Function >
+class OperatorFunction< Operator, Function, void, false, false >
: public Domain< Operator::getDomainDimensions(), Operator::getDomainType() >
{
public:
- static_assert( PreimageFunction::getDomainType() == MeshDomain ||
- PreimageFunction::getDomainType() == MeshInteriorDomain ||
- PreimageFunction::getDomainType() == MeshBoundaryDomain,
+ static_assert( Function::getDomainType() == MeshDomain ||
+ Function::getDomainType() == MeshInteriorDomain ||
+ Function::getDomainType() == MeshBoundaryDomain,
"Only mesh preimageFunctions may be used in the operator preimageFunction. Use ExactOperatorFunction instead of OperatorFunction." );
- static_assert( std::is_same< typename Operator::MeshType, typename PreimageFunction::MeshType >::value,
+ static_assert( std::is_same< typename Operator::MeshType, typename Function::MeshType >::value,
"Both, operator and mesh preimageFunction must be defined on the same mesh." );
typedef Operator OperatorType;
@@ -146,9 +147,9 @@ class OperatorFunction< Operator, PreimageFunction, void, false >
typedef typename OperatorType::RealType RealType;
typedef typename OperatorType::DeviceType DeviceType;
typedef typename OperatorType::IndexType IndexType;
- typedef PreimageFunction PreimageFunctionType;
+ typedef Function PreimageFunctionType;
typedef Functions::MeshFunction< MeshType, Operator::getImageEntitiesDimensions() > ImageFunctionType;
- typedef OperatorFunction< Operator, PreimageFunction, void, true > OperatorFunctionType;
+ typedef OperatorFunction< Operator, Function, void, true > OperatorFunctionType;
typedef typename OperatorType::ExactOperatorType ExactOperatorType;
typedef SharedPointer< MeshType, DeviceType > MeshPointer;
@@ -172,7 +173,7 @@ class OperatorFunction< Operator, PreimageFunction, void, false >
const ImageFunctionType& getImageFunction() const { return this->imageFunction; };
- void setPreimageFunction( PreimageFunction& preimageFunction )
+ void setPreimageFunction( PreimageFunctionType& preimageFunction )
{
this->preimageFunction = &preimageFunction;
this->imageFunction.setMesh( preimageFunction.getMeshPointer() );
@@ -232,18 +233,18 @@ class OperatorFunction< Operator, PreimageFunction, void, false >
* Specialization for precomputed evaluation and with boundary conditions.
*/
template< typename Operator,
- typename PreimageFunction,
+ typename Function,
typename BoundaryConditions >
-class OperatorFunction< Operator, PreimageFunction, BoundaryConditions, false >
+class OperatorFunction< Operator, Function, BoundaryConditions, false, false >
: public Domain< Operator::getMeshDimensions(), MeshDomain >
{
public:
- static_assert( PreimageFunction::getDomainType() == MeshDomain ||
- PreimageFunction::getDomainType() == MeshInteriorDomain ||
- PreimageFunction::getDomainType() == MeshBoundaryDomain,
+ static_assert( Function::getDomainType() == MeshDomain ||
+ Function::getDomainType() == MeshInteriorDomain ||
+ Function::getDomainType() == MeshBoundaryDomain,
"Only mesh preimageFunctions may be used in the operator preimageFunction. Use ExactOperatorFunction instead of OperatorFunction." );
- static_assert( std::is_same< typename Operator::MeshType, typename PreimageFunction::MeshType >::value,
+ static_assert( std::is_same< typename Operator::MeshType, typename Function::MeshType >::value,
"Both, operator and mesh preimageFunction must be defined on the same mesh." );
static_assert( std::is_same< typename BoundaryConditions::MeshType, typename Operator::MeshType >::value,
"The operator and the boundary conditions are defined on different mesh types." );
@@ -254,10 +255,10 @@ class OperatorFunction< Operator, PreimageFunction, BoundaryConditions, false >
typedef typename OperatorType::RealType RealType;
typedef typename OperatorType::DeviceType DeviceType;
typedef typename OperatorType::IndexType IndexType;
- typedef PreimageFunction PreimageFunctionType;
+ typedef Function PreimageFunctionType;
typedef Functions::MeshFunction< MeshType, Operator::getImageEntitiesDimensions() > ImageFunctionType;
typedef BoundaryConditions BoundaryConditionsType;
- typedef OperatorFunction< Operator, PreimageFunction, void, true > OperatorFunctionType;
+ typedef OperatorFunction< Operator, Function, void, true > OperatorFunctionType;
typedef typename OperatorType::ExactOperatorType ExactOperatorType;
static constexpr int getEntitiesDimensions() { return OperatorType::getImageEntitiesDimensions(); };
@@ -286,7 +287,7 @@ class OperatorFunction< Operator, PreimageFunction, BoundaryConditions, false >
const MeshPointer& getMeshPointer() const { return imageFunction.getMeshPointer(); };
- void setPreimageFunction( const PreimageFunction& preimageFunction )
+ void setPreimageFunction( const PreimageFunctionType& preimageFunction )
{
this->preimageFunction = &preimageFunction;
}
@@ -357,6 +358,67 @@ class OperatorFunction< Operator, PreimageFunction, BoundaryConditions, false >
template< typename, typename > friend class MeshFunctionEvaluator;
};
+/****
+ * Specialization for precomputed evaluation and with boundary conditions.
+ */
+template< typename Operator,
+ typename Function >
+class OperatorFunction< Operator, Function, void, false, true >
+ : public Domain< Function::getDomainDimensions(), Function::getDomainType() >
+{
+ public:
+
+ typedef Function FunctionType;
+ typedef typename FunctionType::RealType RealType;
+ typedef typename FunctionType::VertexType VertexType;
+ typedef Operator OperatorType;
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ return( this->function.setup( parameters, prefix) &&
+ this->operator_.setup( parameters, prefix ) );
+ }
+
+ __cuda_callable__
+ FunctionType& getFunction()
+ {
+ return this->function;
+ }
+
+ __cuda_callable__
+ const FunctionType& getFunction() const
+ {
+ return this->function;
+ }
+
+ __cuda_callable__
+ OperatorType& getOperator()
+ {
+ return this->operator_;
+ }
+
+ __cuda_callable__
+ const OperatorType& getOperator() const
+ {
+ return this->operator_;
+ }
+
+ __cuda_callable__
+ RealType operator()( const VertexType& v,
+ const RealType& time = 0.0 ) const
+ {
+ return this->operator_( this->function, v, time );
+ }
+
+ protected:
+
+ Function function;
+
+ Operator operator_;
+
+};
+
} // namespace Functions
} // namespace TNL
diff --git a/src/TNL/Functions/TestFunction.h b/src/TNL/Functions/TestFunction.h
index 55e50c4f9df42f064590b06a62fd8622e5ba0492..d1b78ffb89ce6ba324d455fb3292dc1c214ffdf1 100644
--- a/src/TNL/Functions/TestFunction.h
+++ b/src/TNL/Functions/TestFunction.h
@@ -26,109 +26,129 @@ class TestFunction : public Domain< FunctionDimensions, SpaceDomain >
{
protected:
- enum TestFunctions{ constant,
- expBump,
- sinBumps,
- sinWave,
- cylinder,
- flowerpot,
- twins,
- pseudoSquare,
- blob };
-
- enum TimeDependence { none,
- linear,
- quadratic,
- cosine };
+ enum TestFunctions{ constant,
+ paraboloid,
+ expBump,
+ sinBumps,
+ sinWave,
+ cylinder,
+ flowerpot,
+ twins,
+ pseudoSquare,
+ blob,
+ vectorNorm,
+ paraboloidSDF,
+ sinWaveSDF,
+ sinBumpsSDF };
+
+ enum TimeDependence { none,
+ linear,
+ quadratic,
+ cosine };
+
+ enum Operators { identity,
+ heaviside };
public:
- enum{ Dimensions = FunctionDimensions };
- typedef Real RealType;
- typedef Containers::StaticVector< Dimensions, Real > VertexType;
+ enum{ Dimensions = FunctionDimensions };
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions, Real > VertexType;
+
+ TestFunction();
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ const TestFunction& operator = ( const TestFunction& function );
+
+ #ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+ #else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ #endif
+ __cuda_callable__
+ Real getPartialDerivative( const VertexType& vertex,
+ const Real& time = 0 ) const;
+
+ __cuda_callable__
+ Real operator()( const VertexType& vertex,
+ const Real& time = 0 ) const
+ {
+ return this->getPartialDerivative< 0, 0, 0 >( vertex, time );
+ }
+
+
+ #ifdef HAVE_NOT_CXX11
+ template< int XDiffOrder,
+ int YDiffOrder,
+ int ZDiffOrder >
+ #else
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ #endif
+ __cuda_callable__
+ Real getTimeDerivative( const VertexType& vertex,
+ const Real& time = 0 ) const;
- TestFunction();
+ #ifdef HAVE_NOT_CXX11
+ template< typename Vertex >
+ __cuda_callable__
+ Real getTimeDerivative( const Vertex& vertex,
+ const Real& time = 0 ) const
+ {
+ return this->getTimeDerivative< 0, 0, 0, Vertex >( vertex, time );
+ }
+ #endif
- static void configSetup( Config::ConfigDescription& config,
- const String& prefix = "" );
+ std::ostream& print( std::ostream& str ) const;
- bool setup( const Config::ParameterContainer& parameters,
- const String& prefix = "" );
+ ~TestFunction();
- const TestFunction& operator = ( const TestFunction& function );
+ protected:
-#ifdef HAVE_NOT_CXX11
- template< int XDiffOrder,
- int YDiffOrder,
- int ZDiffOrder >
-#else
- template< int XDiffOrder = 0,
- int YDiffOrder = 0,
- int ZDiffOrder = 0 >
-#endif
- __cuda_callable__
- Real getPartialDerivative( const VertexType& vertex,
- const Real& time = 0 ) const;
+ template< typename FunctionType >
+ bool setupFunction( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ template< typename OperatorType >
+ bool setupOperator( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
- __cuda_callable__
- Real operator()( const VertexType& vertex,
- const Real& time = 0 ) const
- {
- return this->getPartialDerivative< 0, 0, 0 >( vertex, time );
- }
-
-
-#ifdef HAVE_NOT_CXX11
- template< int XDiffOrder,
- int YDiffOrder,
- int ZDiffOrder >
-#else
- template< int XDiffOrder = 0,
- int YDiffOrder = 0,
- int ZDiffOrder = 0 >
-#endif
- __cuda_callable__
- Real getTimeDerivative( const VertexType& vertex,
- const Real& time = 0 ) const;
-
-#ifdef HAVE_NOT_CXX11
- template< typename Vertex >
- __cuda_callable__
- Real getTimeDerivative( const Vertex& vertex,
- const Real& time = 0 ) const
- {
- return this->getTimeDerivative< 0, 0, 0, Vertex >( vertex, time );
- }
-#endif
-
- std::ostream& print( std::ostream& str ) const;
-
- ~TestFunction();
- protected:
+ template< typename FunctionType >
+ void deleteFunction();
- template< typename FunctionType >
- bool setupFunction( const Config::ParameterContainer& parameters,
- const String& prefix = "" );
+ template< typename OperatorType >
+ void deleteOperator();
- template< typename FunctionType >
- void deleteFunction();
+ void deleteFunctions();
- void deleteFunctions();
+ template< typename FunctionType >
+ void copyFunction( const void* function );
- template< typename FunctionType >
- void copyFunction( const void* function );
+ template< typename FunctionType >
+ std::ostream& printFunction( std::ostream& str ) const;
- template< typename FunctionType >
- std::ostream& printFunction( std::ostream& str ) const;
+ void* function;
- void* function;
+ void* operator_;
- TestFunctions functionType;
+ TestFunctions functionType;
+
+ Operators operatorType;
- TimeDependence timeDependence;
+ TimeDependence timeDependence;
- Real timeScale;
+ Real timeScale;
};
diff --git a/src/TNL/Functions/TestFunction_impl.h b/src/TNL/Functions/TestFunction_impl.h
index 506f536518d1946e33da00f749097b168bd8d09d..a7bfdf513f3e9a088dff1354c26f26d77129dbda 100644
--- a/src/TNL/Functions/TestFunction_impl.h
+++ b/src/TNL/Functions/TestFunction_impl.h
@@ -24,16 +24,29 @@
#include <TNL/Functions/Analytic/Twins.h>
#include <TNL/Functions/Analytic/Blob.h>
#include <TNL/Functions/Analytic/PseudoSquare.h>
+#include <TNL/Functions/Analytic/Paraboloid.h>
+#include <TNL/Functions/Analytic/VectorNorm.h>
+/****
+ * The signed distance test functions
+ */
+#include <TNL/Functions/Analytic/SinBumpsSDF.h>
+#include <TNL/Functions/Analytic/SinWaveSDF.h>
+#include <TNL/Functions/Analytic/ParaboloidSDF.h>
+
+#include <TNL/Operators/Analytic/Identity.h>
+#include <TNL/Operators/Analytic/Heaviside.h>
namespace TNL {
namespace Functions {
+
template< int FunctionDimensions,
typename Real,
typename Device >
TestFunction< FunctionDimensions, Real, Device >::
TestFunction()
: function( 0 ),
+ operator_( 0 ),
timeDependence( none ),
timeScale( 1.0 )
{
@@ -49,6 +62,7 @@ configSetup( Config::ConfigDescription& config,
{
config.addRequiredEntry< String >( prefix + "test-function", "Testing function." );
config.addEntryEnum( "constant" );
+ config.addEntryEnum( "paraboloid" );
config.addEntryEnum( "exp-bump" );
config.addEntryEnum( "sin-wave" );
config.addEntryEnum( "sin-bumps" );
@@ -57,6 +71,11 @@ configSetup( Config::ConfigDescription& config,
config.addEntryEnum( "twins" );
config.addEntryEnum( "pseudoSquare" );
config.addEntryEnum( "blob" );
+ config.addEntryEnum( "paraboloid-sdf" );
+ config.addEntryEnum( "sin-wave-sdf" );
+ config.addEntryEnum( "sin-bumps-sdf" );
+ config.addEntryEnum( "heaviside-of-vector-norm" );
+
config.addEntry < double >( prefix + "constant", "Value of the constant function.", 0.0 );
config.addEntry < double >( prefix + "wave-length", "Wave length of the sine based test functions.", 1.0 );
config.addEntry < double >( prefix + "wave-length-x", "Wave length of the sine based test functions.", 1.0 );
@@ -72,8 +91,14 @@ configSetup( Config::ConfigDescription& config,
config.addEntry < double >( prefix + "waves-number-y", "Cut-off for the sine based test functions.", 0.0 );
config.addEntry < double >( prefix + "waves-number-z", "Cut-off for the sine based test functions.", 0.0 );
config.addEntry < double >( prefix + "sigma", "Sigma for the exp based test functions.", 1.0 );
+ config.addEntry < double >( prefix + "radius", "Radius for paraboloids.", 1.0 );
+ config.addEntry < double >( prefix + "coefficient", "Coefficient for paraboloids.", 1.0 );
+ config.addEntry < double >( prefix + "x-center", "x-center for paraboloids.", 0.0 );
+ config.addEntry < double >( prefix + "y-center", "y-center for paraboloids.", 0.0 );
+ config.addEntry < double >( prefix + "z-center", "z-center for paraboloids.", 0.0 );
config.addEntry < double >( prefix + "diameter", "Diameter for the cylinder, flowerpot test functions.", 1.0 );
- config.addEntry < double >( prefix + "height", "Height of zero-level-set function for the blob, pseudosquare test functions.", 1.0 );
+ config.addEntry < double >( prefix + "height", "Height of zero-level-set function for the blob, pseudosquare test functions.", 1.0 );
+ Analytic::VectorNorm< 3, double >::configSetup( config, "vector-norm-" );
config.addEntry < String >( prefix + "time-dependence", "Time dependence of the test function.", "none" );
config.addEntryEnum( "none" );
config.addEntryEnum( "linear" );
@@ -113,6 +138,36 @@ setupFunction( const Config::ParameterContainer& parameters,
return true;
}
+template< int FunctionDimensions,
+ typename Real,
+ typename Device >
+ template< typename OperatorType >
+bool
+TestFunction< FunctionDimensions, Real, Device >::
+setupOperator( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ OperatorType* auxOperator = new OperatorType;
+ if( ! auxOperator->setup( parameters, prefix ) )
+ {
+ delete auxOperator;
+ return false;
+ }
+
+ if( std::is_same< Device, Devices::Host >::value )
+ {
+ this->operator_ = auxOperator;
+ }
+ if( std::is_same< Device, Devices::Cuda >::value )
+ {
+ this->operator_ = Devices::Cuda::passToDevice( *auxOperator );
+ delete auxOperator;
+ if( ! checkCudaDevice )
+ return false;
+ }
+ return true;
+}
+
template< int FunctionDimensions,
typename Real,
typename Device >
@@ -122,6 +177,7 @@ setup( const Config::ParameterContainer& parameters,
const String& prefix )
{
using namespace TNL::Functions::Analytic;
+ using namespace TNL::Operators::Analytic;
std::cout << "Test function setup ... " << std::endl;
const String& timeDependence =
parameters.getParameter< String >(
@@ -140,60 +196,141 @@ setup( const Config::ParameterContainer& parameters,
this->timeScale = parameters.getParameter< double >( prefix + "time-scale" );
const String& testFunction = parameters.getParameter< String >( prefix + "test-function" );
- std::cout << "Test function ... " << testFunction << std::endl;
+ std::cout << "Test function ... " << testFunction << std::endl;
if( testFunction == "constant" )
{
typedef Constant< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = constant;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
+ if( testFunction == "paraboloid" )
+ {
+ typedef Paraboloid< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+ functionType = paraboloid;
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
if( testFunction == "exp-bump" )
{
typedef ExpBump< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = expBump;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "sin-bumps" )
{
typedef SinBumps< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = sinBumps;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "sin-wave" )
{
typedef SinWave< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = sinWave;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "cylinder" )
{
typedef Cylinder< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = cylinder;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "flowerpot" )
{
typedef Flowerpot< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = flowerpot;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "twins" )
{
typedef Twins< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = twins;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "pseudoSquare" )
{
typedef PseudoSquare< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = pseudoSquare;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
if( testFunction == "blob" )
{
typedef Blob< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
functionType = blob;
- return setupFunction< FunctionType >( parameters );
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
+ if( testFunction == "paraboloid-sdf" )
+ {
+ typedef ParaboloidSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+ functionType = paraboloidSDF;
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
+ if( testFunction == "sin-bumps-sdf" )
+ {
+ typedef SinBumpsSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+ functionType = sinBumpsSDF;
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
+ if( testFunction == "sin-wave-sdf" )
+ {
+ typedef SinWaveSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+ functionType = sinWaveSDF;
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
+ if( testFunction == "vector-norm" )
+ {
+ typedef VectorNorm< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+ functionType = vectorNorm;
+ operatorType = identity;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
+ }
+ if( testFunction == "heaviside-of-vector-norm" )
+ {
+ typedef VectorNorm< Dimensions, Real > FunctionType;
+ typedef Heaviside< Dimensions, Real > OperatorType;
+ functionType = vectorNorm;
+ operatorType = heaviside;
+ return ( setupFunction< FunctionType >( parameters, prefix ) &&
+ setupOperator< OperatorType >( parameters, prefix ) );
}
std::cerr << "Unknown function " << testFunction << std::endl;
return false;
@@ -246,11 +383,20 @@ operator = ( const TestFunction& function )
case blob:
this->copyFunction< Blob< FunctionDimensions, Real > >( function.function );
break;
+
+ case paraboloidSDF:
+ this->copyFunction< Paraboloid< FunctionDimensions, Real > >( function.function );
+ break;
+ case sinBumpsSDF:
+ this->copyFunction< SinBumpsSDF< FunctionDimensions, Real > >( function.function );
+ break;
+ case sinWaveSDF:
+ this->copyFunction< SinWaveSDF< FunctionDimensions, Real > >( function.function );
+ break;
default:
TNL_ASSERT( false, );
break;
}
-
}
template< int FunctionDimensions,
@@ -263,9 +409,10 @@ __cuda_callable__
Real
TestFunction< FunctionDimensions, Real, Device >::
getPartialDerivative( const VertexType& vertex,
- const Real& time ) const
+ const Real& time ) const
{
using namespace TNL::Functions::Analytic;
+ using namespace TNL::Operators::Analytic;
Real scale( 1.0 );
switch( this->timeDependence )
{
@@ -286,32 +433,138 @@ getPartialDerivative( const VertexType& vertex,
switch( functionType )
{
case constant:
- return scale * ( ( Constant< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef Constant< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ case paraboloid:
+ {
+ typedef Paraboloid< Dimensions, Real > FunctionType;
+ if( operatorType == identity )
+ {
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ if( operatorType == heaviside )
+ {
+ typedef Heaviside< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ }
case expBump:
- return scale * ( ( ExpBump< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef ExpBump< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case sinBumps:
- return scale * ( ( SinBumps< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef SinBumps< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case sinWave:
- return scale * ( ( SinWave< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef SinWave< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case cylinder:
- return scale * ( ( Cylinder< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef Cylinder< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case flowerpot:
- return scale * ( ( Flowerpot< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef Flowerpot< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case twins:
- return scale * ( ( Twins< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef Twins< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case pseudoSquare:
- return scale * ( ( PseudoSquare< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef PseudoSquare< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
case blob:
- return scale * ( ( Blob< Dimensions, Real >* ) function )->
- template getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ {
+ typedef Blob< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ case vectorNorm:
+ {
+ typedef VectorNorm< Dimensions, Real > FunctionType;
+ if( operatorType == identity )
+ {
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ if( operatorType == heaviside )
+ {
+ typedef Heaviside< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ }
+ case sinBumpsSDF:
+ {
+ typedef SinBumpsSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ case sinWaveSDF:
+ {
+ typedef SinWaveSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+ case paraboloidSDF:
+ {
+ typedef ParaboloidSDF< Dimensions, Real > FunctionType;
+ typedef Identity< Dimensions, Real > OperatorType;
+
+ return scale * ( ( OperatorType* ) this->operator_ )->
+ template getPartialDerivative< FunctionType, XDiffOrder, YDiffOrder, ZDiffOrder >( * ( FunctionType*) this->function, vertex, time );
+ }
+
default:
return 0.0;
}
@@ -348,38 +601,91 @@ getTimeDerivative( const VertexType& vertex,
switch( functionType )
{
case constant:
- return scale * ( ( Constant< Dimensions, Real >* ) function )->
+ {
+ typedef Constant< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
+ getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
+ case paraboloid:
+ {
+ typedef Paraboloid< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
case expBump:
- return scale * ( ( ExpBump< Dimensions, Real >* ) function )->
+ {
+ typedef ExpBump< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
case sinBumps:
- return scale * ( ( SinBumps< Dimensions, Real >* ) function )->
+ {
+ typedef SinBumps< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
case sinWave:
- return scale * ( ( SinWave< Dimensions, Real >* ) function )->
+ {
+ typedef SinWave< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
case cylinder:
- return scale * ( ( Cylinder< Dimensions, Real >* ) function )->
+ {
+ typedef Cylinder< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
case flowerpot:
- return scale * ( ( Flowerpot< Dimensions, Real >* ) function )->
+ {
+ typedef Flowerpot< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
case twins:
- return scale * ( ( Twins< Dimensions, Real >* ) function )->
+ {
+ typedef Twins< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
case pseudoSquare:
- return scale * ( ( PseudoSquare< Dimensions, Real >* ) function )->
+ {
+ typedef PseudoSquare< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
case blob:
- return scale * ( ( Blob< Dimensions, Real >* ) function )->
+ {
+ typedef Blob< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
break;
+ }
+
+
+ case paraboloidSDF:
+ {
+ typedef ParaboloidSDF< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
+ getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
+ case sinBumpsSDF:
+ {
+ typedef SinBumpsSDF< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
+ getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
+ case sinWaveSDF:
+ {
+ typedef SinWaveSDF< Dimensions, Real > FunctionType;
+ return scale * ( ( FunctionType* ) function )->
+ getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
default:
return 0.0;
}
@@ -405,6 +711,26 @@ deleteFunction()
}
}
+template< int FunctionDimensions,
+ typename Real,
+ typename Device >
+ template< typename OperatorType >
+void
+TestFunction< FunctionDimensions, Real, Device >::
+deleteOperator()
+{
+ if( std::is_same< Device, Devices::Host >::value )
+ {
+ if( operator_ )
+ delete ( OperatorType * ) operator_;
+ }
+ if( std::is_same< Device, Devices::Cuda >::value )
+ {
+ if( operator_ )
+ Devices::Cuda::freeFromDevice( ( OperatorType * ) operator_ );
+ }
+}
+
template< int FunctionDimensions,
typename Real,
typename Device >
@@ -413,35 +739,112 @@ TestFunction< FunctionDimensions, Real, Device >::
deleteFunctions()
{
using namespace TNL::Functions::Analytic;
+ using namespace TNL::Operators::Analytic;
switch( functionType )
{
case constant:
- deleteFunction< Constant< Dimensions, Real> >();
- break;
+ {
+ typedef Constant< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
+ case paraboloid:
+ {
+ typedef Paraboloid< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ if( operatorType == identity )
+ deleteOperator< Identity< Dimensions, Real > >();
+ if( operatorType == heaviside )
+ deleteOperator< Heaviside< Dimensions, Real > >();
+ break;
+ }
case expBump:
- deleteFunction< ExpBump< Dimensions, Real> >();
+ {
+ typedef ExpBump< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case sinBumps:
- deleteFunction< SinBumps< Dimensions, Real> >();
+ {
+ typedef SinBumps< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case sinWave:
- deleteFunction< SinWave< Dimensions, Real> >();
+ {
+ typedef SinWave< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case cylinder:
- deleteFunction< Cylinder< Dimensions, Real> >();
+ {
+ typedef Cylinder< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case flowerpot:
- deleteFunction< Flowerpot< Dimensions, Real> >();
+ {
+ typedef Flowerpot< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case twins:
- deleteFunction< Twins< Dimensions, Real> >();
+ {
+ typedef Twins< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case pseudoSquare:
- deleteFunction< PseudoSquare< Dimensions, Real> >();
+ {
+ typedef PseudoSquare< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
break;
+ }
case blob:
- deleteFunction< Blob< Dimensions, Real> >();
- break;
+ {
+ typedef Blob< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
+ case vectorNorm:
+ {
+ typedef VectorNorm< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
+
+
+ case paraboloidSDF:
+ {
+ typedef ParaboloidSDF< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
+ case sinBumpsSDF:
+ {
+ typedef SinBumpsSDF< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
+ case sinWaveSDF:
+ {
+ typedef SinWaveSDF< Dimensions, Real> FunctionType;
+ deleteFunction< FunctionType >();
+ deleteOperator< Identity< Dimensions, Real > >();
+ break;
+ }
}
}
@@ -484,7 +887,6 @@ printFunction( std::ostream& str ) const
Devices::Cuda::print( f, str );
return str;
}
- return str;
}
template< int FunctionDimensions,
diff --git a/src/TNL/Functions/VectorField.h b/src/TNL/Functions/VectorField.h
new file mode 100644
index 0000000000000000000000000000000000000000..5f98b65fcc1370166c38e49760fb7836eeecfd7c
--- /dev/null
+++ b/src/TNL/Functions/VectorField.h
@@ -0,0 +1,278 @@
+/***************************************************************************
+ VectorField.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Functions/MeshFunction.h>
+#include <TNL/Functions/VectorFieldGnuplotWriter.h>
+
+namespace TNL {
+namespace Functions {
+
+template< int Size,
+ typename Function >
+class VectorField
+ : public Functions::Domain< Function::getDomainDimensions(),
+ Function::getDomainType() >
+{
+ public:
+
+ typedef Function FunctionType;
+ typedef typename FunctionType::RealType RealType;
+ typedef typename FunctionType::VertexType VertexType;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ for( int i = 0; i < Size; i++ )
+ FunctionType::configSetup( config, prefix + String( i ) + "-" );
+ }
+
+ template< typename MeshPointer >
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ for( int i = 0; i < Size; i++ )
+ if( ! vectorField[ i ].setup( parameters, prefix + String( i ) + "-" ) )
+ {
+ std::cerr << "Unable to setup " << i << "-th coordinate of the vector field." << std::endl;
+ return false;
+ }
+ return true;
+ }
+
+ __cuda_callable__
+ const FunctionType& operator[]( int i ) const
+ {
+ return this->vectorField[ i ];
+ }
+
+ __cuda_callable__
+ FunctionType& operator[]( int i )
+ {
+ return this->vectorField[ i ];
+ }
+
+ protected:
+
+ Containers::StaticArray< Size, FunctionType > vectorField;
+};
+
+
+template< int Size,
+ typename Mesh,
+ int MeshEntityDimensions,
+ typename Real >
+class VectorField< Size, MeshFunction< Mesh, MeshEntityDimensions, Real > >
+: public Functions::Domain< MeshFunction< Mesh, MeshEntityDimensions, Real >::getDomainDimensions(),
+ MeshFunction< Mesh, MeshEntityDimensions, Real >::getDomainType() >,
+ public Object
+{
+ public:
+
+ typedef Mesh MeshType;
+ typedef Real RealType;
+ typedef SharedPointer< MeshType > MeshPointer;
+ typedef MeshFunction< MeshType, MeshEntityDimensions, RealType > FunctionType;
+ typedef SharedPointer< FunctionType > FunctionPointer;
+ typedef typename MeshType::DeviceType DeviceType;
+ typedef typename MeshType::IndexType IndexType;
+ typedef VectorField< Size, MeshFunction< Mesh, MeshEntityDimensions, RealType > > ThisType;
+ typedef Containers::StaticVector< Size, RealType > VectorType;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ for( int i = 0; i < Size; i++ )
+ FunctionType::configSetup( config, prefix + String( i ) + "-" );
+ }
+
+ VectorField() {};
+
+ VectorField( const MeshPointer& meshPointer )
+ {
+ for( int i = 0; i < Size; i++ )
+ this->vectorField[ i ]->setMesh( meshPointer );
+ };
+
+ static String getType()
+ {
+ return String( "Functions::VectorField< " ) +
+ String( Size) + ", " +
+ FunctionType::getType() +
+ " >";
+ }
+
+ String getTypeVirtual() const
+ {
+ return this->getType();
+ }
+
+ static String getSerializationType()
+ {
+ return String( "Functions::VectorField< " ) +
+ String( Size) + ", " +
+ FunctionType::getSerializationType() +
+ " >";
+ }
+
+ virtual String getSerializationTypeVirtual() const
+ {
+ return this->getSerializationType();
+ }
+
+
+ void setMesh( const MeshPointer& meshPointer )
+ {
+ for( int i = 0; i < Size; i++ )
+ this->vectorField[ i ]->setMesh( meshPointer );
+ }
+
+ template< typename Device = Devices::Host >
+ __cuda_callable__
+ const MeshType& getMesh() const
+ {
+ return this->vectorField[ 0 ].template getData< Device >().template getMesh< Device >();
+ }
+
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ for( int i = 0; i < Size; i++ )
+ if( ! vectorField[ i ].setup( meshPointer, parameters, prefix + String( i ) + "-" ) )
+ {
+ std::cerr << "Unable to setup " << i << "-th coordinate of the vector field." << std::endl;
+ return false;
+ }
+ return true;
+ }
+
+ static IndexType getDofs( const MeshPointer& meshPointer )
+ {
+ return Size * FunctionType::getDofs( meshPointer );
+ }
+
+ __cuda_callable__
+ const FunctionPointer& operator[]( int i ) const
+ {
+ return this->vectorField[ i ];
+ }
+
+ __cuda_callable__
+ FunctionPointer& operator[]( int i )
+ {
+ return this->vectorField[ i ];
+ }
+
+ __cuda_callable__
+ VectorType getVector( const IndexType index ) const
+ {
+ VectorType v;
+ for( int i = 0; i < Size; i++ )
+ v[ i ] = ( *this->vectorField[ i ] )[ index ];
+ return v;
+ }
+
+ template< typename EntityType >
+ __cuda_callable__
+ VectorType getVector( const EntityType& meshEntity ) const
+ {
+ VectorType v;
+ for( int i = 0; i < Size; i++ )
+ v[ i ] = ( *this->vectorField[ i ] )( meshEntity );
+ return v;
+ }
+
+ bool save( File& file ) const
+ {
+ if( ! Object::save( file ) )
+ return false;
+ for( int i = 0; i < Size; i++ )
+ if( ! vectorField[ i ]->save( file ) )
+ return false;
+ return true;
+ }
+
+ bool load( File& file )
+ {
+ if( ! Object::load( file ) )
+ return false;
+ for( int i = 0; i < Size; i++ )
+ if( ! vectorField[ i ]->load( file ) )
+ return false;
+ return true;
+ }
+
+ bool boundLoad( File& file )
+ {
+ if( ! Object::load( file ) )
+ return false;
+ for( int i = 0; i < Size; i++ )
+ if( ! vectorField[ i ]->boundLoad( file ) )
+ return false;
+ return true;
+ }
+
+ bool write( const String& fileName,
+ const String& format = "vtk",
+ const double& scale = 1.0 ) const
+ {
+ std::fstream file;
+ file.open( fileName.getString(), std::ios::out );
+ if( ! file )
+ {
+ std::cerr << "Unable to open a file " << fileName << "." << std::endl;
+ return false;
+ }
+ if( format == "vtk" )
+ return false; //MeshFunctionVTKWriter< ThisType >::write( *this, file );
+ else if( format == "gnuplot" )
+ return VectorFieldGnuplotWriter< ThisType >::write( *this, file, scale );
+ else {
+ std::cerr << "Unknown output format: " << format << std::endl;
+ return false;
+ }
+ return true;
+ }
+
+ using Object::save;
+
+ using Object::load;
+
+ using Object::boundLoad;
+
+ protected:
+
+ Containers::StaticArray< Size, FunctionPointer > vectorField;
+
+};
+
+template< int Dimensions,
+ typename Function >
+std::ostream& operator << ( std::ostream& str, const VectorField< Dimensions, Function >& f )
+{
+ for( int i = 0; i < Dimensions; i++ )
+ {
+ str << "[ " << f[ i ] << " ]";
+ if( i < Dimensions - 1 )
+ str << ", ";
+ }
+ return str;
+}
+
+
+} //namespace Functions
+} //namepsace TNL
diff --git a/src/TNL/Functions/VectorFieldGnuplotWriter.h b/src/TNL/Functions/VectorFieldGnuplotWriter.h
new file mode 100644
index 0000000000000000000000000000000000000000..41b59d511d680568d62ed545a8f230efc43dd575
--- /dev/null
+++ b/src/TNL/Functions/VectorFieldGnuplotWriter.h
@@ -0,0 +1,195 @@
+/***************************************************************************
+ VectorFieldGnuplotWriter.h - description
+ -------------------
+ begin : Feb 16, 2017
+ copyright : (C) 2017 by oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Meshes/Grid.h>
+
+namespace TNL {
+namespace Functions {
+
+template< int, typename > class VectorField;
+
+template< typename VectorField >
+class VectorFieldGnuplotWriter
+{
+ public:
+
+ static bool write( const VectorField& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 1D grids cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 1, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 1, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 1D grids vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 0, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 0, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+
+/***
+ * 2D grids cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 2, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 2, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 2D grids faces
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 1, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 1, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 2D grids vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 0, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 0, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+
+/***
+ * 3D grids cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 3, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 3, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 3D grids faces
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 2, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 2, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+/***
+ * 3D grids vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+class VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 0, Real > > >
+{
+ public:
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Functions::VectorField< VectorFieldSize, MeshFunction< MeshType, 0, RealType > > VectorFieldType;
+
+ static bool write( const VectorFieldType& function,
+ std::ostream& str,
+ const double& scale );
+};
+
+} // namespace Functions
+} // namespace TNL
+
+#include <TNL/Functions/VectorFieldGnuplotWriter_impl.h>
diff --git a/src/TNL/Functions/VectorFieldGnuplotWriter_impl.h b/src/TNL/Functions/VectorFieldGnuplotWriter_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..6ec84a722e7897fb0d58cc226628c208000dc86e
--- /dev/null
+++ b/src/TNL/Functions/VectorFieldGnuplotWriter_impl.h
@@ -0,0 +1,394 @@
+/***************************************************************************
+ VectorFieldGnuplotWriter.h - description
+ -------------------
+ begin : Feb 16, 2017
+ copyright : (C) 2017 by oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <iostream>
+#include <TNL/Functions/VectorFieldGnuplotWriter.h>
+
+namespace TNL {
+namespace Functions {
+
+template< typename VectorField >
+bool
+VectorFieldGnuplotWriter< VectorField >::
+write( const VectorField& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ std::cerr << "Gnuplot writer for mesh vectorFields defined on mesh type " << VectorField::MeshType::getType() << " is not (yet) implemented." << std::endl;
+ return false;
+}
+
+/****
+ * 1D grid, cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 1, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Cell entity( mesh );
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ return true;
+}
+
+/****
+ * 1D grid, vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, 0, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Vertex entity( mesh );
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() <= mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ return true;
+}
+
+
+/****
+ * 2D grid, cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 2, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Cell entity( mesh );
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() < mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+/****
+ * 2D grid, faces
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 1, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typedef typename MeshType::Face EntityType;
+ typedef typename EntityType::EntityOrientationType EntityOrientation;
+ EntityType entity( mesh );
+
+ entity.setOrientation( EntityOrientation( 1.0, 0.0 ) );
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() < mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() <= mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+
+ entity.setOrientation( EntityOrientation( 0.0, 1.0 ) );
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+
+ {
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() <= mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+
+/****
+ * 2D grid, vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, 0, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Vertex entity( mesh );
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() <= mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() <= mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+
+/****
+ * 3D grid, cells
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 3, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Cell entity( mesh );
+ for( entity.getCoordinates().z() = 0;
+ entity.getCoordinates().z() < mesh.getDimensions().z();
+ entity.getCoordinates().z() ++ )
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() < mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y() << " " << v.z();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+/****
+ * 3D grid, faces
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 2, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typedef typename MeshType::Face EntityType;
+ typedef typename EntityType::EntityOrientationType EntityOrientation;
+ EntityType entity( mesh );
+
+ entity.setOrientation( EntityOrientation( 1.0, 0.0, 0.0 ) );
+ for( entity.getCoordinates().z() = 0;
+ entity.getCoordinates().z() < mesh.getDimensions().z();
+ entity.getCoordinates().z() ++ )
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() < mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() <= mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y() << " " << v.z();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+
+ entity.setOrientation( EntityOrientation( 0.0, 1.0, 0.0 ) );
+ for( entity.getCoordinates().z() = 0;
+ entity.getCoordinates().z() < mesh.getDimensions().z();
+ entity.getCoordinates().z() ++ )
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() <= mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y() << " " << v.z();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+
+ entity.setOrientation( EntityOrientation( 0.0, 0.0, 1.0 ) );
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() < mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() <= mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().z() = 0;
+ entity.getCoordinates().z() < mesh.getDimensions().z();
+ entity.getCoordinates().z() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y() << " " << v.z();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+
+/****
+ * 3D grid, vertices
+ */
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ int VectorFieldSize >
+bool
+VectorFieldGnuplotWriter< VectorField< VectorFieldSize, MeshFunction< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, 0, Real > > >::
+write( const VectorFieldType& vectorField,
+ std::ostream& str,
+ const double& scale )
+{
+ const MeshType& mesh = vectorField.getMesh();
+ typename MeshType::Vertex entity( mesh );
+ for( entity.getCoordinates().z() = 0;
+ entity.getCoordinates().z() <= mesh.getDimensions().z();
+ entity.getCoordinates().z() ++ )
+ for( entity.getCoordinates().y() = 0;
+ entity.getCoordinates().y() <= mesh.getDimensions().y();
+ entity.getCoordinates().y() ++ )
+ {
+ for( entity.getCoordinates().x() = 0;
+ entity.getCoordinates().x() <= mesh.getDimensions().x();
+ entity.getCoordinates().x() ++ )
+ {
+ entity.refresh();
+ typename MeshType::VertexType v = entity.getCenter();
+ str << v.x() << " " << v.y() << " " << v.z();
+ for( int i = 0; i < VectorFieldSize; i++ )
+ str << " " << scale * vectorField[ i ]->getData().getElement( entity.getIndex() );
+ str << std::endl;
+ }
+ str << std::endl;
+ }
+ return true;
+}
+
+} // namespace Functions
+} // namespace TNL
+
diff --git a/src/TNL/Images/PGMImage_impl.h b/src/TNL/Images/PGMImage_impl.h
index eb6d59d3c15a54993a83d277880703f6b3d3371e..8939dd71d1156a7eabd030266fd5d57519fcc31d 100644
--- a/src/TNL/Images/PGMImage_impl.h
+++ b/src/TNL/Images/PGMImage_impl.h
@@ -186,16 +186,15 @@ write( const Meshes::Grid< 2, Real, Device, Index >& grid,
unsigned char color = 255 * vector.getElement( cell.getIndex() );
if ( ! this->binary )
- {
- int color_aux = (int)color;
- this->file << color_aux;
- this->file << ' ';
- }
- else this->file << color;
- }
-
+ {
+ int color_aux = (int)color;
+ this->file << color_aux;
+ this->file << ' ';
+ }
+ else this->file << color;
+ }
if ( ! this->binary )
- this->file << '\n';
+ this->file << '\n';
}
return true;
}
diff --git a/src/TNL/Math.h b/src/TNL/Math.h
index 62ff2bb95e7aaac0b7e9686ef43019f58cff9ae0..1ff7cc923c8eed40b6676bed16e8371137b88c04 100644
--- a/src/TNL/Math.h
+++ b/src/TNL/Math.h
@@ -182,9 +182,9 @@ template< class T >
__cuda_callable__
T sign( const T& a )
{
- if( a < ( T ) 0 ) return -1;
- if( a == ( T ) 0 ) return 0;
- return 1;
+ if( a < ( T ) 0 ) return ( T ) -1;
+ if( a == ( T ) 0 ) return ( T ) 0;
+ return ( T ) 1;
};
template< typename Real >
diff --git a/src/TNL/Matrices/Dense_impl.h b/src/TNL/Matrices/Dense_impl.h
index 783e3b6035a8120c2525da5a9832dd86ecef28ae..bec9ec70df66818844d2ceea852cd8c1d304958f 100644
--- a/src/TNL/Matrices/Dense_impl.h
+++ b/src/TNL/Matrices/Dense_impl.h
@@ -153,8 +153,8 @@ template< typename Real,
typename Device,
typename Index >
bool Dense< Real, Device, Index >::setElement( const IndexType row,
- const IndexType column,
- const RealType& value )
+ const IndexType column,
+ const RealType& value )
{
this->values.setElement( this->getElementIndex( row, column ), value );
return true;
@@ -166,9 +166,9 @@ template< typename Real,
typename Index >
__cuda_callable__
bool Dense< Real, Device, Index >::addElementFast( const IndexType row,
- const IndexType column,
- const RealType& value,
- const RealType& thisElementMultiplicator )
+ const IndexType column,
+ const RealType& value,
+ const RealType& thisElementMultiplicator )
{
TNL_ASSERT( row >= 0 && row < this->getRows() &&
column >= 0 && column < this->getColumns(),
diff --git a/src/TNL/Meshes/GridDetails/Grid1D.h b/src/TNL/Meshes/GridDetails/Grid1D.h
index d2a96c4ad5bae4bf19374c90a60cfbd5d146b100..514e964a94f6e569b8757d0c72aa14a99e0b374f 100644
--- a/src/TNL/Meshes/GridDetails/Grid1D.h
+++ b/src/TNL/Meshes/GridDetails/Grid1D.h
@@ -63,30 +63,35 @@ class Grid< 1, Real, Device, Index > : public Object
void setDimensions( const CoordinatesType& dimensions );
- __cuda_callable__ inline
+ __cuda_callable__
const CoordinatesType& getDimensions() const;
void setDomain( const VertexType& origin,
const VertexType& proportions );
__cuda_callable__
- inline const VertexType& getOrigin() const;
+ const VertexType& getOrigin() const;
__cuda_callable__
- inline const VertexType& getProportions() const;
-
+ const VertexType& getProportions() const;
+
template< typename EntityType >
__cuda_callable__
- inline IndexType getEntitiesCount() const;
-
+ IndexType getEntitiesCount() const;
+
+ template< int EntiytDimensions >
+ __cuda_callable__
+ IndexType getEntitiesCount() const;
+
+
template< typename EntityType >
__cuda_callable__
- inline EntityType getEntity( const IndexType& entityIndex ) const;
-
+ EntityType getEntity( const IndexType& entityIndex ) const;
+
template< typename EntityType >
__cuda_callable__
- inline Index getEntityIndex( const EntityType& entity ) const;
-
+ Index getEntityIndex( const EntityType& entity ) const;
+
template< typename EntityType >
__cuda_callable__
RealType getEntityMeasure( const EntityType& entity ) const;
@@ -95,14 +100,14 @@ class Grid< 1, Real, Device, Index > : public Object
inline const RealType& getCellMeasure() const;
__cuda_callable__
- inline const VertexType& getSpaceSteps() const;
+ const VertexType& getSpaceSteps() const { return this->spaceSteps;}
template< int xPow >
__cuda_callable__
- inline const RealType& getSpaceStepsProducts() const;
-
+ const RealType& getSpaceStepsProducts() const;
+
__cuda_callable__
- inline RealType getSmallestSpaceStep() const;
+ RealType getSmallestSpaceStep() const;
template< typename GridFunction >
diff --git a/src/TNL/Meshes/GridDetails/Grid1D_impl.h b/src/TNL/Meshes/GridDetails/Grid1D_impl.h
index cac0ad3da8884483d4f48e91b62baba5b30b0dda..fc82381ebb27dffe37df27dd30caec473c8ffa56 100644
--- a/src/TNL/Meshes/GridDetails/Grid1D_impl.h
+++ b/src/TNL/Meshes/GridDetails/Grid1D_impl.h
@@ -168,6 +168,29 @@ getEntitiesCount() const
return -1;
}
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< int EntityDimensions >
+__cuda_callable__ inline
+Index
+Grid< 1, Real, Device, Index >::
+getEntitiesCount() const
+{
+ static_assert( EntityDimensions <= 1 &&
+ EntityDimensions >= 0, "Wrong grid entity dimensions." );
+
+ switch( EntityDimensions )
+ {
+ case 1:
+ return this->numberOfCells;
+ case 0:
+ return this->numberOfVertices;
+ }
+ return -1;
+}
+
+
template< typename Real,
typename Device,
typename Index >
@@ -221,16 +244,16 @@ getCellMeasure() const
return this->template getSpaceStepsProducts< 1 >();
}
-template< typename Real,
+/*template< typename Real,
typename Device,
typename Index >
-__cuda_callable__ inline
-const typename Grid< 1, Real, Device, Index >::VertexType&
+__cuda_callable__
+typename Grid< 1, Real, Device, Index >::VertexType
Grid< 1, Real, Device, Index >::
getSpaceSteps() const
{
return this->spaceSteps;
-}
+}*/
template< typename Real,
typename Device,
diff --git a/src/TNL/Meshes/GridDetails/Grid2D.h b/src/TNL/Meshes/GridDetails/Grid2D.h
index 6c7dfca7f61089f61c8ec2d9e5d4be995f3c9906..1a403508f7d0b69d5ec4620436df2769b63c7096 100644
--- a/src/TNL/Meshes/GridDetails/Grid2D.h
+++ b/src/TNL/Meshes/GridDetails/Grid2D.h
@@ -69,27 +69,32 @@ class Grid< 2, Real, Device, Index > : public Object
void setDimensions( const CoordinatesType& dimensions );
__cuda_callable__
- inline const CoordinatesType& getDimensions() const;
+ const CoordinatesType& getDimensions() const;
void setDomain( const VertexType& origin,
const VertexType& proportions );
__cuda_callable__
- inline const VertexType& getOrigin() const;
+ const VertexType& getOrigin() const;
__cuda_callable__
- inline const VertexType& getProportions() const;
+ const VertexType& getProportions() const;
+
+ template< int EntityDimensions >
+ __cuda_callable__
+ IndexType getEntitiesCount() const;
+
template< typename EntityType >
__cuda_callable__
- inline IndexType getEntitiesCount() const;
-
+ IndexType getEntitiesCount() const;
+
template< typename EntityType >
__cuda_callable__
- inline EntityType getEntity( const IndexType& entityIndex ) const;
-
+ EntityType getEntity( const IndexType& entityIndex ) const;
+
template< typename EntityType >
__cuda_callable__
- inline Index getEntityIndex( const EntityType& entity ) const;
+ Index getEntityIndex( const EntityType& entity ) const;
template< typename EntityType >
__cuda_callable__
@@ -99,14 +104,14 @@ class Grid< 2, Real, Device, Index > : public Object
inline const RealType& getCellMeasure() const;
__cuda_callable__
- inline const VertexType& getSpaceSteps() const;
+ const VertexType& getSpaceSteps() const { return this->spaceSteps; }
template< int xPow, int yPow >
__cuda_callable__
- inline const RealType& getSpaceStepsProducts() const;
-
+ const RealType& getSpaceStepsProducts() const;
+
__cuda_callable__
- inline RealType getSmallestSpaceStep() const;
+ RealType getSmallestSpaceStep() const;
template< typename GridFunction >
diff --git a/src/TNL/Meshes/GridDetails/Grid2D_impl.h b/src/TNL/Meshes/GridDetails/Grid2D_impl.h
index 4d0d87fb2f1d7f752c6330c94c90137b7cb3830d..f0aa205d420be259ce63cfae6a98dcd9644d25c1 100644
--- a/src/TNL/Meshes/GridDetails/Grid2D_impl.h
+++ b/src/TNL/Meshes/GridDetails/Grid2D_impl.h
@@ -196,6 +196,31 @@ const typename Grid< 2, Real, Device, Index > :: VertexType&
return this->proportions;
}
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< int EntityDimensions >
+__cuda_callable__ inline
+Index
+Grid< 2, Real, Device, Index >::
+getEntitiesCount() const
+{
+ static_assert( EntityDimensions <= 2 &&
+ EntityDimensions >= 0, "Wrong grid entity dimensions." );
+
+ switch( EntityDimensions )
+ {
+ case 2:
+ return this->numberOfCells;
+ case 1:
+ return this->numberOfFaces;
+ case 0:
+ return this->numberOfVertices;
+ }
+ return -1;
+}
+
+
template< typename Real,
typename Device,
typename Index >
@@ -274,16 +299,16 @@ getCellMeasure() const
}
-template< typename Real,
+/*template< typename Real,
typename Device,
typename Index >
-__cuda_callable__ inline
-const typename Grid< 2, Real, Device, Index >::VertexType&
+__cuda_callable__
+typename Grid< 2, Real, Device, Index >::VertexType&
Grid< 2, Real, Device, Index >::
getSpaceSteps() const
{
return this->spaceSteps;
-}
+}*/
template< typename Real,
typename Device,
diff --git a/src/TNL/Meshes/GridDetails/Grid3D.h b/src/TNL/Meshes/GridDetails/Grid3D.h
index 860b5e473528cf0de601648a77bb0bd6f0aba490..fd1e91ac0e1298b05f917a1df383db989437b11f 100644
--- a/src/TNL/Meshes/GridDetails/Grid3D.h
+++ b/src/TNL/Meshes/GridDetails/Grid3D.h
@@ -62,28 +62,32 @@ class Grid< 3, Real, Device, Index > : public Object
void setDimensions( const CoordinatesType& );
__cuda_callable__
- inline const CoordinatesType& getDimensions() const;
+ const CoordinatesType& getDimensions() const;
void setDomain( const VertexType& origin,
const VertexType& proportions );
__cuda_callable__
- inline const VertexType& getOrigin() const;
+ const VertexType& getOrigin() const;
__cuda_callable__
- inline const VertexType& getProportions() const;
+ const VertexType& getProportions() const;
template< typename EntityType >
__cuda_callable__
- inline IndexType getEntitiesCount() const;
-
+ IndexType getEntitiesCount() const;
+
+ template< int Dimensions >
+ __cuda_callable__
+ IndexType getEntitiesCount() const;
+
template< typename EntityType >
__cuda_callable__
- inline EntityType getEntity( const IndexType& entityIndex ) const;
-
+ EntityType getEntity( const IndexType& entityIndex ) const;
+
template< typename EntityType >
__cuda_callable__
- inline Index getEntityIndex( const EntityType& entity ) const;
-
+ Index getEntityIndex( const EntityType& entity ) const;
+
template< typename EntityType >
__cuda_callable__
RealType getEntityMeasure( const EntityType& entity ) const;
@@ -92,16 +96,16 @@ class Grid< 3, Real, Device, Index > : public Object
inline const RealType& getCellMeasure() const;
__cuda_callable__
- inline const VertexType& getSpaceSteps() const;
-
+ const VertexType& getSpaceSteps() const { return this->spaceSteps; };
+
template< int xPow, int yPow, int zPow >
__cuda_callable__
- inline const RealType& getSpaceStepsProducts() const;
+ const RealType& getSpaceStepsProducts() const;
__cuda_callable__
- inline RealType getSmallestSpaceStep() const;
-
+ RealType getSmallestSpaceStep() const;
+
template< typename GridFunction >
typename GridFunction::RealType getAbsMax( const GridFunction& f ) const;
diff --git a/src/TNL/Meshes/GridDetails/Grid3D_impl.h b/src/TNL/Meshes/GridDetails/Grid3D_impl.h
index d9cdaa1dfe62359328f11c369cd16d0a43d168d5..9282d1500e1ccb18d58810654dfdbc989304e73c 100644
--- a/src/TNL/Meshes/GridDetails/Grid3D_impl.h
+++ b/src/TNL/Meshes/GridDetails/Grid3D_impl.h
@@ -268,6 +268,32 @@ getEntitiesCount() const
return -1;
}
+template< typename Real,
+ typename Device,
+ typename Index >
+ template< int EntityDimensions >
+__cuda_callable__ inline
+Index
+Grid< 3, Real, Device, Index >::
+getEntitiesCount() const
+{
+ static_assert( EntityDimensions <= 3 &&
+ EntityDimensions >= 0, "Wrong grid entity dimensions." );
+
+ switch( EntityDimensions )
+ {
+ case 3:
+ return this->numberOfCells;
+ case 2:
+ return this->numberOfFaces;
+ case 1:
+ return this->numberOfEdges;
+ case 0:
+ return this->numberOfVertices;
+ }
+ return -1;
+}
+
template< typename Real,
typename Device,
typename Index >
@@ -321,16 +347,16 @@ getCellMeasure() const
return this->template getSpaceStepsProducts< 1, 1, 1 >();
}
-template< typename Real,
+/*template< typename Real,
typename Device,
typename Index >
-__cuda_callable__ inline
-const typename Grid< 3, Real, Device, Index >::VertexType&
+__cuda_callable__
+typename Grid< 3, Real, Device, Index >::VertexType&
Grid< 3, Real, Device, Index >::
getSpaceSteps() const
{
return this->spaceSteps;
-}
+}*/
template< typename Real,
typename Device,
diff --git a/src/TNL/Operators/Advection/CMakeLists.txt b/src/TNL/Operators/Advection/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..5dff0bfc3dc52d76e01bfb2b8102946c27c451dd
--- /dev/null
+++ b/src/TNL/Operators/Advection/CMakeLists.txt
@@ -0,0 +1,6 @@
+SET( headers LaxFridrichs.h )
+
+SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/TNL/Operators/Advection )
+
+
+INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/TNL/Operators/Advection )
diff --git a/src/TNL/Operators/Advection/LaxFridrichs.h b/src/TNL/Operators/Advection/LaxFridrichs.h
new file mode 100644
index 0000000000000000000000000000000000000000..9b939e184cee92bfbdef1edff999b15cdf539aee
--- /dev/null
+++ b/src/TNL/Operators/Advection/LaxFridrichs.h
@@ -0,0 +1,317 @@
+#ifndef LaxFridrichs_H
+#define LaxFridrichs_H
+
+#include <TNL/Containers/Vector.h>
+#include <TNL/Meshes/Grid.h>
+#include <TNL/Functions/VectorField.h>
+#include <TNL/SharedPointer.h>
+
+namespace TNL {
+ namespace Operators {
+ namespace Advection {
+
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType,
+ typename VelocityFunction = Functions::MeshFunction< Mesh > >
+class LaxFridrichs
+{
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename VelocityFunction >
+class LaxFridrichs< Meshes::Grid< 1, MeshReal, Device, MeshIndex >, Real, Index, VelocityFunction >
+{
+ public:
+
+ typedef Meshes::Grid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef SharedPointer< MeshType > MeshPointer;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ typedef VelocityFunction VelocityFunctionType;
+ typedef Functions::VectorField< Dimensions, VelocityFunctionType > VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType, DeviceType > VelocityFieldPointer;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "numerical-viscosity", "Value of artificial (numerical) viscosity in the Lax-Fridrichs scheme", 1.0 );
+ }
+
+ LaxFridrichs()
+ : artificialViscosity( 1.0 ) {}
+
+ LaxFridrichs( const VelocityFieldPointer& velocityField )
+ : artificialViscosity( 1.0 ), velocityField( velocityField ) {}
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->artificialViscosity = parameters.getParameter< double >( prefix + "numerical-viscosity" );
+ return true;
+ }
+
+ static String getType();
+
+ void setViscosity(const Real& artificalViscosity)
+ {
+ this->artificialViscosity = artificalViscosity;
+ }
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocityField( const VelocityFieldPointer& velocityField )
+ {
+ this->velocityField = velocityField;
+ }
+
+ const VelocityFieldPointer& getVelocityField() const
+ {
+ return this->velocityField;
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 1, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 1, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 1 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1 >();
+ typedef Functions::FunctionAdapter< MeshType, VelocityFunctionType > FunctionAdapter;
+ return ( 0.5 / this->tau ) * this->artificialViscosity * ( u[ west ]- 2.0 * u[ center ] + u[ east ] ) -
+ FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 0 ], entity, time ) * ( u[ east ] - u[ west ] ) * hxInverse * 0.5;
+ }
+
+ protected:
+
+ RealType tau;
+
+ RealType artificialViscosity;
+
+ VelocityFieldPointer velocityField;
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename VelocityFunction >
+class LaxFridrichs< Meshes::Grid< 2, MeshReal, Device, MeshIndex >, Real, Index, VelocityFunction >
+{
+ public:
+
+ typedef Meshes::Grid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef SharedPointer< MeshType > MeshPointer;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ typedef VelocityFunction VelocityFunctionType;
+ typedef Functions::VectorField< Dimensions, VelocityFunctionType > VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType, DeviceType > VelocityFieldPointer;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "numerical-viscosity", "Value of artificial (numerical) viscosity in the Lax-Fridrichs scheme", 1.0 );
+ }
+
+ LaxFridrichs()
+ : artificialViscosity( 1.0 ) {}
+
+ LaxFridrichs( const VelocityFieldPointer& velocityField )
+ : artificialViscosity( 1.0 ), velocityField( velocityField ) {}
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->artificialViscosity = parameters.getParameter< double >( prefix + "numerical-viscosity" );
+ return true;
+ }
+
+ static String getType();
+
+ void setViscosity(const Real& artificalViscosity)
+ {
+ this->artificialViscosity = artificalViscosity;
+ }
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocityField( const VelocityFieldPointer& velocityField )
+ {
+ this->velocityField = velocityField;
+ }
+
+ const VelocityFieldPointer& getVelocityField() const
+ {
+ return this->velocityField;
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 2, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 2, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 2 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1 >();
+
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1 >();
+
+ typedef Functions::FunctionAdapter< MeshType, VelocityFunctionType > FunctionAdapter;
+ return ( 0.25 / this->tau ) * this->artificialViscosity * ( u[ west ] + u[ east ] + u[ north ] + u[ south ] - 4.0 * u[ center ] ) -
+ 0.5 * ( FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 0 ], entity, time ) * ( u[ east ] - u[ west ] ) * hxInverse +
+ FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 1 ], entity, time ) * ( u[ north ] - u[ south ] ) * hyInverse );
+ }
+
+ protected:
+
+ RealType tau;
+
+ RealType artificialViscosity;
+
+ VelocityFieldPointer velocityField;
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index,
+ typename VelocityFunction >
+class LaxFridrichs< Meshes::Grid< 3, MeshReal, Device, MeshIndex >, Real, Index, VelocityFunction >
+{
+ public:
+
+ typedef Meshes::Grid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef SharedPointer< MeshType > MeshPointer;
+ static const int Dimensions = MeshType::getMeshDimensions();
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+ typedef Functions::MeshFunction< MeshType > MeshFunctionType;
+ typedef VelocityFunction VelocityFunctionType;
+ typedef Functions::VectorField< Dimensions, VelocityFunctionType > VelocityFieldType;
+ typedef SharedPointer< VelocityFieldType, DeviceType > VelocityFieldPointer;
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "numerical-viscosity", "Value of artificial (numerical) viscosity in the Lax-Fridrichs scheme", 1.0 );
+ }
+
+ LaxFridrichs()
+ : artificialViscosity( 1.0 ) {}
+
+ LaxFridrichs( const VelocityFieldPointer& velocityField )
+ : artificialViscosity( 1.0 ), velocityField( velocityField ) {}
+
+ bool setup( const MeshPointer& meshPointer,
+ const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->artificialViscosity = parameters.getParameter< double >( prefix + "numerical-viscosity" );
+ return true;
+ }
+
+ static String getType();
+
+ void setViscosity(const Real& artificalViscosity)
+ {
+ this->artificialViscosity = artificalViscosity;
+ }
+
+ void setTau(const Real& tau)
+ {
+ this->tau = tau;
+ };
+
+ void setVelocityField( const VelocityFieldPointer& velocityField )
+ {
+ this->velocityField = velocityField;
+ }
+
+ const VelocityFieldPointer& getVelocityField() const
+ {
+ return this->velocityField;
+ }
+
+ template< typename MeshFunction, typename MeshEntity >
+ __cuda_callable__
+ Real operator()( const MeshFunction& u,
+ const MeshEntity& entity,
+ const RealType& time = 0.0 ) const
+ {
+ static_assert( MeshEntity::entityDimensions == 3, "Wrong mesh entity dimensions." );
+ static_assert( MeshFunction::getEntitiesDimensions() == 3, "Wrong preimage function" );
+ const typename MeshEntity::template NeighbourEntities< 3 >& neighbourEntities = entity.getNeighbourEntities();
+
+ const RealType& hxInverse = entity.getMesh().template getSpaceStepsProducts< -1, 0, 0 >();
+ const RealType& hyInverse = entity.getMesh().template getSpaceStepsProducts< 0, -1, 0 >();
+ const RealType& hzInverse = entity.getMesh().template getSpaceStepsProducts< 0, 0, -1 >();
+ const IndexType& center = entity.getIndex();
+ const IndexType& east = neighbourEntities.template getEntityIndex< 1, 0, 0 >();
+ const IndexType& west = neighbourEntities.template getEntityIndex< -1, 0, 0 >();
+ const IndexType& north = neighbourEntities.template getEntityIndex< 0, 1, 0 >();
+ const IndexType& south = neighbourEntities.template getEntityIndex< 0, -1, 0 >();
+ const IndexType& up = neighbourEntities.template getEntityIndex< 0, 0, 1 >();
+ const IndexType& down = neighbourEntities.template getEntityIndex< 0, 0, -1 >();
+
+ typedef Functions::FunctionAdapter< MeshType, VelocityFunctionType > FunctionAdapter;
+ return ( 0.25 / this->tau ) * this->artificialViscosity * ( u[ west ] + u[ east ] + u[ north ] + u[ south ] + u[ up ] + u[ down ] - 6.0 * u[ center ] ) -
+ 0.5 * ( FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 0 ], entity, time ) * ( u[ east ] - u[ west ] ) * hxInverse +
+ FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 1 ], entity, time ) * ( u[ north ] - u[ south ] ) * hyInverse +
+ FunctionAdapter::getValue( this->velocityField.template getData< DeviceType >()[ 2 ], entity, time ) * ( u[ up ] - u[ down ] ) * hzInverse );
+ }
+
+ protected:
+
+ RealType tau;
+
+ RealType artificialViscosity;
+
+ VelocityFieldPointer velocityField;
+};
+
+ }// namespace Advection
+ } // namepsace Operators
+} // namespace TNL
+
+#endif /* LaxFridrichs_H */
diff --git a/src/TNL/Operators/Analytic/CMakeLists.txt b/src/TNL/Operators/Analytic/CMakeLists.txt
new file mode 100755
index 0000000000000000000000000000000000000000..8cc44ea6f704cdb0086ab9925e60e7c580e125ac
--- /dev/null
+++ b/src/TNL/Operators/Analytic/CMakeLists.txt
@@ -0,0 +1,11 @@
+SET( headers Sign.h
+ Heaviside.h
+ Identity.h
+ SmoothHeaviside.h
+ Shift.h
+ Rotation.h )
+
+SET( CURRENT_DIR ${CMAKE_SOURCE_DIR}/src/TNL/Operators/Analytic )
+
+
+INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/TNL/Operators/Analytic )
diff --git a/src/TNL/Operators/Analytic/Heaviside.h b/src/TNL/Operators/Analytic/Heaviside.h
new file mode 100644
index 0000000000000000000000000000000000000000..4449998015aee2936bc7599843e3c40ed3f29d2b
--- /dev/null
+++ b/src/TNL/Operators/Analytic/Heaviside.h
@@ -0,0 +1,83 @@
+/***************************************************************************
+ Heaviside.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Functions/Domain.h>
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+template< int Dimensions,
+ typename Real = double >
+class Heaviside : public Functions::Domain< Dimensions, Functions::SpaceDomain >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions,
+ RealType > VertexType;
+
+ Heaviside() : multiplicator( 1.0 ) {}
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "multiplicator", "Outer multiplicator of the Heaviside operator - -1.0 turns the function graph upside/down.", 1.0 );
+ }
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->multiplicator = parameters.getParameter< double >( prefix + "multiplicator" );
+ return true;
+ };
+
+
+ template< typename Function >
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ const RealType aux = function( vertex, time );
+ if( aux > 0.0 )
+ return 1.0;
+ return 0.0;
+ }
+
+ template< typename Function,
+ int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ return this->operator()( function, vertex, time );
+ return 0.0;
+ }
+
+ protected:
+
+ RealType multiplicator;
+
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
\ No newline at end of file
diff --git a/src/TNL/Operators/Analytic/Identity.h b/src/TNL/Operators/Analytic/Identity.h
new file mode 100644
index 0000000000000000000000000000000000000000..fdc4824480735c3fe7e780966628340202370d72
--- /dev/null
+++ b/src/TNL/Operators/Analytic/Identity.h
@@ -0,0 +1,62 @@
+/***************************************************************************
+ Identity.h - description
+ -------------------
+ begin : Dec 5, 2013
+ copyright : (C) 2013 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+template< int Dimensions, typename Real >
+class Identity : public Functions::Domain< Dimensions, Functions::SpaceDomain >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions, RealType > VertexType;
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ return true;
+ };
+
+
+ template< typename Function >
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ return function( vertex, time );
+ }
+
+ template< typename Function,
+ int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ return function.getPartialDerivative< XDiffOrder, YDiffOrder, ZDiffOrder >( vertex, time );
+ }
+
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
\ No newline at end of file
diff --git a/src/TNL/Operators/Analytic/Rotation.h b/src/TNL/Operators/Analytic/Rotation.h
new file mode 100644
index 0000000000000000000000000000000000000000..272e9ae85ca877e6bfa0caf551e055fb8aef5ae4
--- /dev/null
+++ b/src/TNL/Operators/Analytic/Rotation.h
@@ -0,0 +1,81 @@
+/***************************************************************************
+ Rotation.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+// TODO: Implement RotationXY, RotationXZ and RotationYZ in all dimensions.
+// Where it does not make sense, the operator does nothing.
+
+template< typename Real,
+ int Dimensions >
+class RotationBase
+{
+ public:
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimenions, RealType > VertexType;
+
+ RotationBase() : center( 0.0 ) {};
+
+ void setCenter( const VertexType& center )
+ {
+ this->center = center;
+ }
+
+ __cuda_callable__
+ const VertexType& getCenter() const
+ {
+ return this->center;
+ }
+
+ protected:
+
+ VertexType center;
+};
+
+template< typename Function,
+ int Dimensions = Function::getDomainDimenions() >
+class Rotation;
+
+template< typename Function, 1 >
+class Rotation: public Functions::Domain< Function::getDomainDimenions(),
+ Function::getDomainType() >
+{
+ public:
+
+ typedef typename Function::RealType RealType;
+ typedef Containers::StaticVector< Function::getDomainDimenions(),
+ RealType > VertexType;
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" ){};
+
+
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ return function( vertex, time );
+ }
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
diff --git a/src/TNL/Operators/Analytic/Shift.h b/src/TNL/Operators/Analytic/Shift.h
new file mode 100644
index 0000000000000000000000000000000000000000..d5e8580ccf338972ed196d32f911d3655477d67f
--- /dev/null
+++ b/src/TNL/Operators/Analytic/Shift.h
@@ -0,0 +1,91 @@
+/***************************************************************************
+ Shift.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+#include <TNL/Containers/StaticVector.h>
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+template< int Dimensions, typename Real >
+class Shift : public Functions::Domain< Dimensions, Functions::SpaceDomain >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions, RealType > VertexType;
+
+
+ Shift() : shift( 0.0 ) {};
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "shift-0", "x-coordinate of the shift vector.", 0.0 );
+ config.addEntry< double >( prefix + "shift-1", "y-coordinate of the shift vector.", 0.0 );
+ config.addEntry< double >( prefix + "shift-2", "z-coordinate of the shift vector.", 0.0 );
+ }
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ return shift.setup( parameters, prefix + "shift-" );
+ };
+
+
+ void setShift( const VertexType& shift )
+ {
+ this->shift = shift;
+ }
+
+ __cuda_callable__
+ const VertexType& getShift() const
+ {
+ return this->shift;
+ }
+
+ template< typename Function >
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ return function( vertex - this->shift, time );
+ }
+
+ template< typename Function,
+ int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ return this->operator()( function, vertex - this->shift, time );
+ // TODO: implement the rest
+ }
+
+
+ protected:
+
+ VertexType shift;
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
diff --git a/src/TNL/Operators/Analytic/Sign.h b/src/TNL/Operators/Analytic/Sign.h
new file mode 100644
index 0000000000000000000000000000000000000000..ea18af636270b28372c5196c0a85f88e95b6c445
--- /dev/null
+++ b/src/TNL/Operators/Analytic/Sign.h
@@ -0,0 +1,120 @@
+/***************************************************************************
+ Sign.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+template< int Dimensions,
+ typename Real >
+class Sign : public Functions::Domain< Dimensions, Functions::SpaceDomain >
+{
+ public:
+
+ typedef Real RealType;
+ typedef Containers::StaticVector< Dimensions, RealType > VertexType;
+
+ Sign()
+ : positiveValue( 1.0 ),
+ negativeValue( -1.0 ),
+ zeroValue( 0.0 ) {}
+
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ config.addEntry< double >( prefix + "positive-value", "Value returned for positive argument.", 1.0 );
+ config.addEntry< double >( prefix + "negative-value", "Value returned for negative argument.", -1.0 );
+ config.addEntry< double >( prefix + "zero-value", "Value returned for zero argument.", 0.0 );
+ }
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ this->positiveValue = parameters.getParameter< double >( prefix + "positive-value" );
+ this->negativeValue = parameters.getParameter< double >( prefix + "negative-value" );
+ this->zeroValue = parameters.getParameter< double >( prefix + "zero-value" );
+ return true;
+ };
+
+ void setPositiveValue( const RealType& value )
+ {
+ this->positiveValue = value;
+ }
+
+ const RealType& getPositiveValue() const
+ {
+ return this->positiveValue;
+ }
+
+ void setNegativeValue( const RealType& value )
+ {
+ this->negativeValue = value;
+ }
+
+ const RealType& getNegativeValue() const
+ {
+ return this->negativeValue;
+ }
+
+ void setZeroValue( const RealType& value )
+ {
+ this->zeroValue = value;
+ }
+
+ const RealType& getZeroValue() const
+ {
+ return this->zeroValue;
+ }
+
+ template< typename Function >
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ const RealType aux = function( vertex, time );
+ if( aux > 0.0 )
+ return this->positiveValue;
+ else
+ if( aux < 0.0 )
+ return this->negativeValue;
+ return this->zeroValue;
+ }
+
+ template< typename Function,
+ int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ return this->operator()( function, vertex, time );
+ return 0.0;
+ }
+
+ protected:
+
+ RealType positiveValue, negativeValue, zeroValue;
+
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
diff --git a/src/TNL/Operators/Analytic/SmoothHeaviside.h b/src/TNL/Operators/Analytic/SmoothHeaviside.h
new file mode 100644
index 0000000000000000000000000000000000000000..e3f13c4d7fe8c3db79605f4556ff21106e4ca619
--- /dev/null
+++ b/src/TNL/Operators/Analytic/SmoothHeaviside.h
@@ -0,0 +1,79 @@
+/***************************************************************************
+ SmoothHeaviside.h - description
+ -------------------
+ begin : Feb 6, 2017
+ copyright : (C) 2017 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/* See Copyright Notice in tnl/Copyright */
+
+#pragma once
+
+#include <TNL/Functions/Domain.h>
+#include <TNL/Devices/Cuda.h>
+#include <TNL/Config/ParameterContainer.h>
+
+namespace TNL {
+namespace Operators {
+namespace Analytic {
+
+
+template< typename Function >
+class SmoothHeaviside : public Functions::Domain< Function::getDomainDimenions(),
+ Function::getDomainTyep() >
+{
+ public:
+
+ typedef typename Function::RealType RealType;
+ typedef Containers::StaticVector< Function::getDomainDimenions(),
+ RealType > VertexType;
+
+ SmoothHeaviside()
+ : sharpness( 1.0 ){}
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" ){};
+
+
+ void setSharpness( const RealType& sharpness )
+ {
+ this->sharpness = sharpness;
+ }
+
+ __cuda_callable__
+ const RealType getShaprness() const
+ {
+ return this->sharpness;
+ }
+
+ __cuda_callable__
+ RealType operator()( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ const RealType aux = function( vertex, time );
+ return 1.0 / ( 1.0 + exp( -2.0 * sharpness * aux ) );
+ }
+
+ template< int XDiffOrder = 0,
+ int YDiffOrder = 0,
+ int ZDiffOrder = 0 >
+ __cuda_callable__
+ RealType getPartialDerivative( const Function& function,
+ const VertexType& vertex,
+ const RealType& time = 0 ) const
+ {
+ if( XDiffOrder == 0 && YDiffOrder == 0 && ZDiffOrder == 0 )
+ return this->operator()( function, vertex, time );
+ // TODO: implement the rest
+ }
+
+ protected:
+
+ RealType sharpness;
+};
+
+} // namespace Analytic
+} // namespace Operators
+} // namespace TNL
diff --git a/src/TNL/Operators/CMakeLists.txt b/src/TNL/Operators/CMakeLists.txt
index dcdb7d3ddaae05cffedbe0d9468002b63ca41328..63385c2a49e16c771f1c7970180ddd97ac3be5f4 100755
--- a/src/TNL/Operators/CMakeLists.txt
+++ b/src/TNL/Operators/CMakeLists.txt
@@ -1,9 +1,12 @@
+ADD_SUBDIRECTORY( Advection )
+ADD_SUBDIRECTORY( Analytic )
ADD_SUBDIRECTORY( diffusion )
ADD_SUBDIRECTORY( euler )
ADD_SUBDIRECTORY( interpolants )
ADD_SUBDIRECTORY( operator-Q )
ADD_SUBDIRECTORY( operator-curvature )
+
SET( headers DirichletBoundaryConditions.h
ExactFunctionInverseOperator.h
ExactIdentityOperator.h
@@ -23,6 +26,11 @@ if( BUILD_CUDA)
${tnl_operators_diffusion_CUDA__SOURCES}
${tnl_operators_gradient_CUDA__SOURCES}
${tnl_operators_euler_CUDA__SOURCES}
+# ${tnl_operators_godunov_CUDA__SOURCES}
+ ${tnl_operators_godunov-eikonal_CUDA__SOURCES}
+# ${tnl_operators_upwind_CUDA__SOURCES}
+# ${tnl_operators_upwind-eikonal_CUDA__SOURCES}
+ ${common_SOURCES}
${common_SOURCES}
PARENT_SCOPE )
endif()
@@ -31,6 +39,10 @@ set( tnl_operators_SOURCES
${tnl_operators_diffusion_SOURCES}
${tnl_operators_gradient_SOURCES}
${tnl_operators_euler_SOURCES}
+# ${tnl_operators_godunov_SOURCES}
+ ${tnl_operators_godunov-eikonal_SOURCES}
+# ${tnl_operators_upwind_SOURCES}
+# ${tnl_operators_upwind-eikonal_SOURCES}
${common_SOURCES}
PARENT_SCOPE )
diff --git a/src/TNL/Operators/NeumannBoundaryConditions.h b/src/TNL/Operators/NeumannBoundaryConditions.h
index 1461135eebaa135043e1f60452a178f2633b4737..f39eb64744df7ecf29823ca79e92bdb494711d53 100644
--- a/src/TNL/Operators/NeumannBoundaryConditions.h
+++ b/src/TNL/Operators/NeumannBoundaryConditions.h
@@ -49,10 +49,22 @@ class NeumannBoundaryConditionsBase
return Functions::FunctionAdapter< typename MeshPointer::ObjectType, FunctionType >::setup( this->function, meshPointer, parameters, prefix );
}
+ static void configSetup( Config::ConfigDescription& config,
+ const String& prefix = "" )
+ {
+ Function::configSetup( config, prefix );
+ };
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" )
+ {
+ return this->function.setup( parameters, prefix );
+ };
+
void setFunction( const FunctionType& function )
{
this->function = function;
- }
+ };
FunctionType& getFunction()
{
@@ -112,7 +124,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 1, MeshReal, Device, MeshIndex >,
const auto& neighbourEntities = entity.getNeighbourEntities();
const IndexType& index = entity.getIndex();
if( entity.getCoordinates().x() == 0 )
- return u[ neighbourEntities.template getEntityIndex< 1 >() ] - entity.getMesh().getSpaceSteps().x() *
+ return u[ neighbourEntities.template getEntityIndex< 1 >() ] + entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
else
return u[ neighbourEntities.template getEntityIndex< -1 >() ] + entity.getMesh().getSpaceSteps().x() *
@@ -149,7 +161,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 1, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 1 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().x() *
+ b[ index ] = entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
else
@@ -207,7 +219,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 2, MeshReal, Device, MeshIndex >,
const IndexType& index = entity.getIndex();
if( entity.getCoordinates().x() == 0 )
{
- return u[ neighbourEntities.template getEntityIndex< 1, 0 >() ] - entity.getMesh().getSpaceSteps().x() *
+ return u[ neighbourEntities.template getEntityIndex< 1, 0 >() ] + entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().x() == entity.getMesh().getDimensions().x() - 1 )
@@ -217,7 +229,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 2, MeshReal, Device, MeshIndex >,
}
if( entity.getCoordinates().y() == 0 )
{
- return u[ neighbourEntities.template getEntityIndex< 0, 1 >() ] - entity.getMesh().getSpaceSteps().y() *
+ return u[ neighbourEntities.template getEntityIndex< 0, 1 >() ] + entity.getMesh().getSpaceSteps().y() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
// The following line is commented to avoid compiler warning
@@ -256,7 +268,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 2, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 1, 0 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().x() *
+ b[ index ] = entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().x() == entity.getMesh().getDimensions().x() - 1 )
@@ -270,7 +282,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 2, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 0, 1 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().y() *
+ b[ index ] = entity.getMesh().getSpaceSteps().y() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().y() == entity.getMesh().getDimensions().y() - 1 )
@@ -326,7 +338,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
const IndexType& index = entity.getIndex();
if( entity.getCoordinates().x() == 0 )
{
- return u[ neighbourEntities.template getEntityIndex< 1, 0, 0 >() ] - entity.getMesh().getSpaceSteps().x() *
+ return u[ neighbourEntities.template getEntityIndex< 1, 0, 0 >() ] + entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().x() == entity.getMesh().getDimensions().x() - 1 )
@@ -336,7 +348,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
}
if( entity.getCoordinates().y() == 0 )
{
- return u[ neighbourEntities.template getEntityIndex< 0, 1, 0 >() ] - entity.getMesh().getSpaceSteps().y() *
+ return u[ neighbourEntities.template getEntityIndex< 0, 1, 0 >() ] + entity.getMesh().getSpaceSteps().y() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().y() == entity.getMesh().getDimensions().y() - 1 )
@@ -346,7 +358,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
}
if( entity.getCoordinates().z() == 0 )
{
- return u[ neighbourEntities.template getEntityIndex< 0, 0, 1 >() ] - entity.getMesh().getSpaceSteps().z() *
+ return u[ neighbourEntities.template getEntityIndex< 0, 0, 1 >() ] + entity.getMesh().getSpaceSteps().z() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
// The following line is commented to avoid compiler warning
@@ -386,7 +398,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 1, 0, 0 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().x() *
+ b[ index ] = entity.getMesh().getSpaceSteps().x() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().x() == entity.getMesh().getDimensions().x() - 1 )
@@ -400,7 +412,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 0, 1, 0 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().y() *
+ b[ index ] = entity.getMesh().getSpaceSteps().y() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().y() == entity.getMesh().getDimensions().y() - 1 )
@@ -414,7 +426,7 @@ class NeumannBoundaryConditions< Meshes::Grid< 3, MeshReal, Device, MeshIndex >,
{
matrixRow.setElement( 0, index, 1.0 );
matrixRow.setElement( 1, neighbourEntities.template getEntityIndex< 0, 0, 1 >(), -1.0 );
- b[ index ] = - entity.getMesh().getSpaceSteps().z() *
+ b[ index ] = entity.getMesh().getSpaceSteps().z() *
Functions::FunctionAdapter< MeshType, FunctionType >::getValue( this->function, entity, time );
}
if( entity.getCoordinates().z() == entity.getMesh().getDimensions().z() - 1 )
@@ -440,4 +452,3 @@ std::ostream& operator << ( std::ostream& str, const NeumannBoundaryConditions<
} // namespace Operators
} // namespace TNL
-
diff --git a/src/TNL/Problems/HeatEquationProblem.h b/src/TNL/Problems/HeatEquationProblem.h
index 33046f58cc847a3654057da2150504ef77447de5..59d57ff2f56e7c55ee1d09eff951cf92ee36a9ae 100644
--- a/src/TNL/Problems/HeatEquationProblem.h
+++ b/src/TNL/Problems/HeatEquationProblem.h
@@ -32,9 +32,9 @@ template< typename Mesh,
typename DifferentialOperator = Operators::LinearDiffusion< Mesh,
typename BoundaryCondition::RealType > >
class HeatEquationProblem : public PDEProblem< Mesh,
- typename DifferentialOperator::RealType,
- typename Mesh::DeviceType,
- typename DifferentialOperator::IndexType >
+ typename DifferentialOperator::RealType,
+ typename Mesh::DeviceType,
+ typename DifferentialOperator::IndexType >
{
public:
diff --git a/src/TNL/Problems/HeatEquationProblem_impl.h b/src/TNL/Problems/HeatEquationProblem_impl.h
index 6dc77992669e167ca200a820a87bf32be79d4178..e4e8818a99004a6168dbd367acb45c6c1cbd3a64 100644
--- a/src/TNL/Problems/HeatEquationProblem_impl.h
+++ b/src/TNL/Problems/HeatEquationProblem_impl.h
@@ -183,10 +183,10 @@ makeSnapshot( const RealType& time,
DofVectorPointer& dofs,
MeshDependentDataPointer& meshDependentData )
{
- std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
+ std::cout << std::endl << "Writing output at time " << time << " step " << step << "." << std::endl;
this->bindDofs( meshPointer, dofs );
- //cout << "dofs = " << dofs << endl;
+
FileName fileName;
fileName.setFileNameBase( "u-" );
fileName.setExtension( "tnl" );
diff --git a/src/TNL/Solvers/IterativeSolver_impl.h b/src/TNL/Solvers/IterativeSolver_impl.h
index 5f8c084d799c88a33f7a00c51532c9d233b810cf..da811762ae2b5d1f11f5748e37d52b4026a7f78e 100644
--- a/src/TNL/Solvers/IterativeSolver_impl.h
+++ b/src/TNL/Solvers/IterativeSolver_impl.h
@@ -111,7 +111,7 @@ bool IterativeSolver< Real, Index> :: checkNextIteration()
this->getIterations() > this->getMaxIterations() ||
( this->getResidue() > this->getDivergenceResidue() && this->getIterations() >= this->getMinIterations() ) ||
( this->getResidue() < this->getConvergenceResidue() && this->getIterations() >= this->getMinIterations() ) )
- return false;
+ return false;
return true;
}
diff --git a/src/TNL/Solvers/ODE/Euler_impl.h b/src/TNL/Solvers/ODE/Euler_impl.h
index 46706cf24c8796ed6e92681cadbfed46fddad102..6571dee4d0d2ccf3c097a0ec358e22118d64c0a5 100644
--- a/src/TNL/Solvers/ODE/Euler_impl.h
+++ b/src/TNL/Solvers/ODE/Euler_impl.h
@@ -27,8 +27,6 @@ template< typename Problem >
Euler< Problem > :: Euler()
: cflCondition( 0.0 )
{
- //timer.reset();
- //updateTimer.reset();
};
template< typename Problem >
@@ -60,13 +58,13 @@ bool Euler< Problem > :: setup( const Config::ParameterContainer& parameters,
template< typename Problem >
void Euler< Problem > :: setCFLCondition( const RealType& cfl )
{
- this->cflCondition = cfl;
+ this -> cflCondition = cfl;
}
template< typename Problem >
const typename Problem :: RealType& Euler< Problem > :: getCFLCondition() const
{
- return this->cflCondition;
+ return this -> cflCondition;
}
template< typename Problem >
@@ -94,7 +92,7 @@ bool Euler< Problem > :: solve( DofVectorPointer& u )
this->resetIterations();
this->setResidue( this->getConvergenceResidue() + 1.0 );
- this->refreshSolverMonitor();
+ this -> refreshSolverMonitor();
/****
* Start the main loop
@@ -104,13 +102,11 @@ bool Euler< Problem > :: solve( DofVectorPointer& u )
/****
* Compute the RHS
*/
- //timer.stop();
this->problem->getExplicitRHS( time, currentTau, u, k1 );
- //timer.start();
RealType lastResidue = this->getResidue();
RealType maxResidue( 0.0 );
- if( this->cflCondition != 0.0 )
+ if( this -> cflCondition != 0.0 )
{
maxResidue = k1->absMax();
if( currentTau * maxResidue > this->cflCondition )
@@ -120,16 +116,14 @@ bool Euler< Problem > :: solve( DofVectorPointer& u )
}
}
RealType newResidue( 0.0 );
- //updateTimer.start();
computeNewTimeLevel( u, currentTau, newResidue );
- //updateTimer.stop();
this->setResidue( newResidue );
/****
* When time is close to stopTime the new residue
* may be inaccurate significantly.
*/
- if( currentTau + time == this->stopTime ) this->setResidue( lastResidue );
+ if( currentTau + time == this -> stopTime ) this->setResidue( lastResidue );
time += currentTau;
if( ! this->nextIteration() )
@@ -138,25 +132,23 @@ bool Euler< Problem > :: solve( DofVectorPointer& u )
/****
* Compute the new time step.
*/
- if( time + currentTau > this->getStopTime() )
- currentTau = this->getStopTime() - time; //we don't want to keep such tau
- else this->tau = currentTau;
+ if( time + currentTau > this -> getStopTime() )
+ currentTau = this -> getStopTime() - time; //we don't want to keep such tau
+ else this -> tau = currentTau;
- this->refreshSolverMonitor();
+ this -> refreshSolverMonitor();
/****
* Check stop conditions.
*/
if( time >= this->getStopTime() ||
- ( this->getConvergenceResidue() != 0.0 && this->getResidue() < this->getConvergenceResidue() ) )
+ ( this -> getConvergenceResidue() != 0.0 && this->getResidue() < this -> getConvergenceResidue() ) )
{
- this->refreshSolverMonitor();
- //std::cerr << std::endl << "RHS Timer = " << timer.getRealTime() << std::endl;
- //std::cerr << std::endl << "Update Timer = " << updateTimer.getRealTime() << std::endl;
+ this -> refreshSolverMonitor();
return true;
}
- if( this->cflCondition != 0.0 )
+ if( this -> cflCondition != 0.0 )
{
currentTau /= 0.95;
currentTau = min( currentTau, this->getMaxTau() );
diff --git a/src/TNL/Solvers/ODE/ExplicitSolver.h b/src/TNL/Solvers/ODE/ExplicitSolver.h
index c988e4092444c659b248df0b77656aae98ec1dc8..4c1743c66471c04fa4707730d9d629e9d02f6f50 100644
--- a/src/TNL/Solvers/ODE/ExplicitSolver.h
+++ b/src/TNL/Solvers/ODE/ExplicitSolver.h
@@ -76,7 +76,7 @@ class ExplicitSolver : public IterativeSolver< typename Problem::RealType,
void setRefreshRate( const IndexType& refreshRate );
- void refreshSolverMonitor();
+ void refreshSolverMonitor( bool force = false );
protected:
diff --git a/src/TNL/Solvers/ODE/ExplicitSolver_impl.h b/src/TNL/Solvers/ODE/ExplicitSolver_impl.h
index f40e1018cf2ec4722da9f185627e829354b0d4cb..11cd7f19285e21d92ddeeb4bcff98897ab59f0d9 100644
--- a/src/TNL/Solvers/ODE/ExplicitSolver_impl.h
+++ b/src/TNL/Solvers/ODE/ExplicitSolver_impl.h
@@ -150,7 +150,7 @@ setTimer( Timer* timer )
template< class Problem >
void
ExplicitSolver< Problem >::
-refreshSolverMonitor()
+refreshSolverMonitor( bool force )
{
if( this->solverMonitor )
{
@@ -159,7 +159,7 @@ refreshSolverMonitor()
this->solverMonitor->setTimeStep( this->getTau() );
this->solverMonitor->setTime( this->getTime() );
this->solverMonitor->setRefreshRate( this->refreshRate );
- this->solverMonitor->refresh();
+ this->solverMonitor->refresh( force );
}
}
diff --git a/src/TNL/Solvers/ODE/Merson_impl.h b/src/TNL/Solvers/ODE/Merson_impl.h
index 868bc32b59ce652acd55c7f1c58c39a172478e45..13afdb4df56fb94438272c4fb4e8f94750b68b57 100644
--- a/src/TNL/Solvers/ODE/Merson_impl.h
+++ b/src/TNL/Solvers/ODE/Merson_impl.h
@@ -170,7 +170,7 @@ bool Merson< Problem > :: solve( DofVectorPointer& u )
/****
* Start the main loop
*/
- while( 1 )
+ while( this->checkNextIteration() )
{
/****
* Compute Runge-Kutta coefficients
@@ -227,10 +227,13 @@ bool Merson< Problem > :: solve( DofVectorPointer& u )
if( time >= this->getStopTime() ||
( this->getConvergenceResidue() != 0.0 && this->getResidue() < this->getConvergenceResidue() ) )
{
- this->refreshSolverMonitor();
+ this->refreshSolverMonitor( true );
return true;
}
}
+ this->refreshSolverMonitor( true );
+ return this->checkConvergence();
+
};
template< typename Problem >
diff --git a/src/TNL/Solvers/PDE/CMakeLists.txt b/src/TNL/Solvers/PDE/CMakeLists.txt
index 0f98d12a9f92b80279e810041ec9e4ba42f7bfaa..e01c13e93a3228e51c01915ad133077987112969 100755
--- a/src/TNL/Solvers/PDE/CMakeLists.txt
+++ b/src/TNL/Solvers/PDE/CMakeLists.txt
@@ -7,8 +7,10 @@ SET( headers BackwardTimeDiscretisation.h
LinearSystemAssembler.h
LinearSystemAssembler_impl.h
NoTimeDiscretisation.h
- PDESolver.h
- PDESolver_impl.h
+ TimeDependentPDESolver.h
+ TimeDependentPDESolver_impl.h
+ TimeIndependentPDESolver.h
+ TimeIndependentPDESolver_impl.h
SemiImplicitTimeStepper.h
SemiImplicitTimeStepper_impl.h )
diff --git a/src/TNL/Solvers/PDE/PDESolver.h b/src/TNL/Solvers/PDE/TimeDependentPDESolver.h
similarity index 93%
rename from src/TNL/Solvers/PDE/PDESolver.h
rename to src/TNL/Solvers/PDE/TimeDependentPDESolver.h
index a4f47ef51c8e6472c8d4807254c98db8cdc5f661..c6a1baa174b64b985e9590372eb913eda784cd47 100644
--- a/src/TNL/Solvers/PDE/PDESolver.h
+++ b/src/TNL/Solvers/PDE/TimeDependentPDESolver.h
@@ -1,5 +1,5 @@
/***************************************************************************
- PDESolver.h - description
+ TimeDependentPDESolver.h - description
-------------------
begin : Jan 15, 2013
copyright : (C) 2013 by Tomas Oberhuber
@@ -23,7 +23,7 @@ namespace PDE {
template< typename Problem,
typename TimeStepper >
-class PDESolver : public Object
+class TimeDependentPDESolver : public Object
{
public:
@@ -38,7 +38,7 @@ class PDESolver : public Object
typedef SharedPointer< DofVectorType, DeviceType > DofVectorPointer;
typedef SharedPointer< MeshDependentDataType, DeviceType > MeshDependentDataPointer;
- PDESolver();
+ TimeDependentPDESolver();
static void configSetup( Config::ConfigDescription& config,
const String& prefix = "" );
@@ -102,4 +102,4 @@ class PDESolver : public Object
} // namespace Solvers
} // namespace TNL
-#include <TNL/Solvers/PDE/PDESolver_impl.h>
+#include <TNL/Solvers/PDE/TimeDependentPDESolver_impl.h>
diff --git a/src/TNL/Solvers/PDE/PDESolver_impl.h b/src/TNL/Solvers/PDE/TimeDependentPDESolver_impl.h
similarity index 86%
rename from src/TNL/Solvers/PDE/PDESolver_impl.h
rename to src/TNL/Solvers/PDE/TimeDependentPDESolver_impl.h
index 139d8fcc95169bd2938eb2ba9397b51425b962d4..2a22201d1e54f2d75b809c5b2e914abb332bd390 100644
--- a/src/TNL/Solvers/PDE/PDESolver_impl.h
+++ b/src/TNL/Solvers/PDE/TimeDependentPDESolver_impl.h
@@ -1,5 +1,5 @@
/***************************************************************************
- PDESolver_impl.h - description
+ TimeDependentPDESolver_impl.h - description
-------------------
begin : Jan 15, 2013
copyright : (C) 2013 by Tomas Oberhuber
@@ -10,7 +10,7 @@
#pragma once
-#include "PDESolver.h"
+#include "TimeDependentPDESolver.h"
namespace TNL {
namespace Solvers {
@@ -18,8 +18,8 @@ namespace PDE {
template< typename Problem,
typename TimeStepper >
-PDESolver< Problem, TimeStepper >::
-PDESolver()
+TimeDependentPDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver()
: timeStepper( 0 ),
initialTime( 0.0 ),
finalTime( 0.0 ),
@@ -35,7 +35,7 @@ PDESolver()
template< typename Problem,
typename TimeStepper >
void
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
configSetup( Config::ConfigDescription& config,
const String& prefix )
{
@@ -50,7 +50,7 @@ configSetup( Config::ConfigDescription& config,
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setup( const Config::ParameterContainer& parameters,
const String& prefix )
{
@@ -121,7 +121,7 @@ setup( const Config::ParameterContainer& parameters,
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
writeProlog( Logger& logger,
const Config::ParameterContainer& parameters )
{
@@ -172,7 +172,7 @@ writeProlog( Logger& logger,
template< typename Problem,
typename TimeStepper >
void
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setTimeStepper( TimeStepper& timeStepper )
{
this->timeStepper = &timeStepper;
@@ -181,7 +181,7 @@ setTimeStepper( TimeStepper& timeStepper )
template< typename Problem,
typename TimeStepper >
void
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setProblem( ProblemType& problem )
{
this->problem = &problem;
@@ -190,7 +190,7 @@ setProblem( ProblemType& problem )
template< typename Problem,
typename TimeStepper >
void
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setInitialTime( const RealType& initialTime )
{
this->initialTime = initialTime;
@@ -199,7 +199,7 @@ setInitialTime( const RealType& initialTime )
template< typename Problem,
typename TimeStepper >
const typename TimeStepper :: RealType&
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
getInitialTime() const
{
return this->initialTime;
@@ -209,12 +209,12 @@ getInitialTime() const
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setFinalTime( const RealType& finalTime )
{
if( finalTime <= this->initialTime )
{
- std::cerr << "Final time for PDESolver must larger than the initial time which is now " << this->initialTime << "." << std::endl;
+ std::cerr << "Final time for TimeDependentPDESolver must larger than the initial time which is now " << this->initialTime << "." << std::endl;
return false;
}
this->finalTime = finalTime;
@@ -224,7 +224,7 @@ setFinalTime( const RealType& finalTime )
template< typename Problem,
typename TimeStepper >
const typename TimeStepper :: RealType&
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
getFinalTime() const
{
return this->finalTime;
@@ -233,12 +233,12 @@ getFinalTime() const
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setSnapshotPeriod( const RealType& period )
{
if( period <= 0 )
{
- std::cerr << "Snapshot tau for PDESolver must be positive value." << std::endl;
+ std::cerr << "Snapshot tau for TimeDependentPDESolver must be positive value." << std::endl;
return false;
}
this->snapshotPeriod = period;
@@ -248,7 +248,7 @@ setSnapshotPeriod( const RealType& period )
template< typename Problem,
typename TimeStepper >
const typename TimeStepper::RealType&
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
getSnapshotPeriod() const
{
return this->snapshotPeriod;
@@ -257,12 +257,12 @@ getSnapshotPeriod() const
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setTimeStep( const RealType& timeStep )
{
if( timeStep <= 0 )
{
- std::cerr << "The time step for PDESolver must be positive value." << std::endl;
+ std::cerr << "The time step for TimeDependentPDESolver must be positive value." << std::endl;
return false;
}
this->timeStep = timeStep;
@@ -272,7 +272,7 @@ setTimeStep( const RealType& timeStep )
template< typename Problem,
typename TimeStepper >
const typename TimeStepper::RealType&
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
getTimeStep() const
{
return this->timeStep;
@@ -281,12 +281,12 @@ getTimeStep() const
template< typename Problem,
typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
setTimeStepOrder( const RealType& timeStepOrder )
{
if( timeStepOrder < 0 )
{
- std::cerr << "The time step order for PDESolver must be zero or positive value." << std::endl;
+ std::cerr << "The time step order for TimeDependentPDESolver must be zero or positive value." << std::endl;
return false;
}
this->timeStepOrder = timeStepOrder;
@@ -296,27 +296,27 @@ setTimeStepOrder( const RealType& timeStepOrder )
template< typename Problem,
typename TimeStepper >
const typename TimeStepper::RealType&
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
getTimeStepOrder() const
{
return this->timeStepOrder;
}
template< typename Problem, typename TimeStepper >
-void PDESolver< Problem, TimeStepper > :: setIoTimer( Timer& ioTimer )
+void TimeDependentPDESolver< Problem, TimeStepper > :: setIoTimer( Timer& ioTimer )
{
this->ioTimer = &ioTimer;
}
template< typename Problem, typename TimeStepper >
-void PDESolver< Problem, TimeStepper > :: setComputeTimer( Timer& computeTimer )
+void TimeDependentPDESolver< Problem, TimeStepper > :: setComputeTimer( Timer& computeTimer )
{
this->computeTimer = &computeTimer;
}
template< typename Problem, typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
solve()
{
TNL_ASSERT( timeStepper != 0,
@@ -326,7 +326,7 @@ solve()
if( snapshotPeriod == 0 )
{
- std::cerr << "No snapshot tau was set in PDESolver." << std::endl;
+ std::cerr << "No snapshot tau was set in TimeDependentPDESolver." << std::endl;
return false;
}
RealType t( this->initialTime );
@@ -376,7 +376,7 @@ solve()
template< typename Problem, typename TimeStepper >
bool
-PDESolver< Problem, TimeStepper >::
+TimeDependentPDESolver< Problem, TimeStepper >::
writeEpilog( Logger& logger ) const
{
return ( this->timeStepper->writeEpilog( logger ) &&
diff --git a/src/TNL/Solvers/PDE/TimeIndependentPDESolver.h b/src/TNL/Solvers/PDE/TimeIndependentPDESolver.h
new file mode 100644
index 0000000000000000000000000000000000000000..41b6a47b5135fa74822804b5551f0dabbc81a3d4
--- /dev/null
+++ b/src/TNL/Solvers/PDE/TimeIndependentPDESolver.h
@@ -0,0 +1,75 @@
+/***************************************************************************
+ tnlTimeIndependentPDESolver.h - description
+ -------------------
+ begin : Jan 15, 2013
+ copyright : (C) 2013 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <core/tnlObject.h>
+#include <config/tnlConfigDescription.h>
+#include <TNL/Config/ParameterContainer.h>
+#include <solvers/tnlSolverMonitor.h>
+#include <core/tnlLogger.h>
+
+template< typename Problem >
+class tnlTimeIndependentPDESolver : public tnlObject
+{
+ public:
+
+ typedef Problem ProblemType;
+ typedef typename ProblemType::RealType RealType;
+ typedef typename ProblemType::DeviceType DeviceType;
+ typedef typename ProblemType::IndexType IndexType;
+ typedef typename ProblemType::MeshType MeshType;
+ typedef typename ProblemType::DofVectorType DofVectorType;
+ typedef typename ProblemType::MeshDependentDataType MeshDependentDataType;
+
+ tnlTimeIndependentPDESolver();
+
+ static void configSetup( tnlConfigDescription& config,
+ const String& prefix = "" );
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ bool writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters );
+
+
+ void setProblem( ProblemType& problem );
+
+ void setComputeTimer( tnlTimer& computeTimer );
+
+ void setIoTimer( tnlTimer& ioTimer );
+
+ bool solve();
+
+ bool writeEpilog( tnlLogger& logger ) const;
+
+ protected:
+
+ MeshType mesh;
+
+ DofVectorType dofs;
+
+ MeshDependentDataType meshDependentData;
+
+ ProblemType* problem;
+
+ tnlTimer *computeTimer;
+};
+
+#include <solvers/pde/tnlTimeIndependentPDESolver_impl.h>
+
diff --git a/src/TNL/Solvers/PDE/TimeIndependentPDESolver_impl.h b/src/TNL/Solvers/PDE/TimeIndependentPDESolver_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..0f27de6489ffc3ee7a4ec37d6ca1d69c766f913a
--- /dev/null
+++ b/src/TNL/Solvers/PDE/TimeIndependentPDESolver_impl.h
@@ -0,0 +1,172 @@
+/***************************************************************************
+ tnlTimeIndependentPDESolver_impl.h - description
+ -------------------
+ begin : Jan 15, 2013
+ copyright : (C) 2013 by Tomas Oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#pragma once
+
+#include <solvers/pde/tnlTimeIndependentPDESolver.h>
+
+template< typename Problem >
+tnlTimeIndependentPDESolver< Problem >::
+tnlTimeIndependentPDESolver()
+: problem( 0 ),
+ computeTimer( 0 )
+{
+}
+
+template< typename Problem >
+void
+tnlTimeIndependentPDESolver< Problem >::
+configSetup( tnlConfigDescription& config,
+ const String& prefix )
+{
+}
+
+template< typename Problem >
+bool
+tnlTimeIndependentPDESolver< Problem >::
+setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ /****
+ * Load the mesh from the mesh file
+ */
+ const String& meshFile = parameters.getParameter< String >( "mesh" );
+ cout << "Loading a mesh from the file " << meshFile << "...";
+ if( ! this->mesh.load( meshFile ) )
+ {
+ cerr << endl;
+ cerr << "I am not able to load the mesh from the file " << meshFile << "." << endl;
+ cerr << " You may create it with tools like tnl-grid-setup or tnl-mesh-convert." << endl;
+ return false;
+ }
+ cout << " [ OK ] " << endl;
+
+ /****
+ * Set DOFs (degrees of freedom)
+ */
+ TNL_ASSERT( problem->getDofs( this->mesh ) != 0, );
+ cout << "Allocating dofs ... ";
+ if( ! this->dofs.setSize( problem->getDofs( this->mesh ) ) )
+ {
+ cerr << endl;
+ cerr << "I am not able to allocate DOFs (degrees of freedom)." << endl;
+ return false;
+ }
+ cout << " [ OK ]" << endl;
+ this->dofs.setValue( 0.0 );
+ this->problem->bindDofs( this->mesh, this->dofs );
+
+ /****
+ * Set mesh dependent data
+ */
+ this->problem->setMeshDependentData( this->mesh, this->meshDependentData );
+ this->problem->bindMeshDependentData( this->mesh, this->meshDependentData );
+
+ /***
+ * Set-up the initial condition
+ */
+ cout << "Setting up the initial condition ... ";
+ typedef typename Problem :: DofVectorType DofVectorType;
+ if( ! this->problem->setInitialData( parameters, this->mesh, this->dofs, this->meshDependentData ) )
+ return false;
+ cout << " [ OK ]" << endl;
+
+
+ return true;
+}
+
+template< typename Problem >
+bool
+tnlTimeIndependentPDESolver< Problem >::
+writeProlog( tnlLogger& logger,
+ const Config::ParameterContainer& parameters )
+{
+ logger.writeHeader( problem->getPrologHeader() );
+ problem->writeProlog( logger, parameters );
+ logger.writeSeparator();
+ mesh.writeProlog( logger );
+ logger.writeSeparator();
+ const String& solverName = parameters. getParameter< String >( "discrete-solver" );
+ logger.writeParameter< String >( "Discrete solver:", "discrete-solver", parameters );
+ if( solverName == "merson" )
+ logger.writeParameter< double >( "Adaptivity:", "merson-adaptivity", parameters, 1 );
+ if( solverName == "sor" )
+ logger.writeParameter< double >( "Omega:", "sor-omega", parameters, 1 );
+ if( solverName == "gmres" )
+ logger.writeParameter< int >( "Restarting:", "gmres-restarting", parameters, 1 );
+ logger.writeParameter< double >( "Convergence residue:", "convergence-residue", parameters );
+ logger.writeParameter< double >( "Divergence residue:", "divergence-residue", parameters );
+ logger.writeParameter< int >( "Maximal number of iterations:", "max-iterations", parameters );
+ logger.writeParameter< int >( "Minimal number of iterations:", "min-iterations", parameters );
+ logger.writeSeparator();
+ logger.writeParameter< String >( "Real type:", "real-type", parameters, 0 );
+ logger.writeParameter< String >( "Index type:", "index-type", parameters, 0 );
+ logger.writeParameter< String >( "Device:", "device", parameters, 0 );
+ logger.writeSeparator();
+ logger.writeSystemInformation( parameters );
+ logger.writeSeparator();
+ logger.writeCurrentTime( "Started at:" );
+ logger.writeSeparator();
+ return true;
+}
+
+template< typename Problem >
+void
+tnlTimeIndependentPDESolver< Problem >::
+setProblem( ProblemType& problem )
+{
+ this->problem = &problem;
+}
+
+template< typename Problem >
+void tnlTimeIndependentPDESolver< Problem > :: setIoTimer( tnlTimer& ioTimer )
+{
+ // this->ioTimer = &ioTimer;
+}
+
+template< typename Problem >
+void tnlTimeIndependentPDESolver< Problem > :: setComputeTimer( tnlTimer& computeTimer )
+{
+ this->computeTimer = &computeTimer;
+}
+
+template< typename Problem >
+bool
+tnlTimeIndependentPDESolver< Problem >::
+solve()
+{
+ TNL_ASSERT( problem != 0,
+ cerr << "No problem was set in tnlPDESolver." );
+
+ this->computeTimer->reset();
+ this->computeTimer->start();
+ if( ! this->problem->solve( this->mesh, this->dofs ) )
+ {
+ this->computeTimer->stop();
+ return false;
+ }
+ this->computeTimer->stop();
+ return true;
+}
+
+template< typename Problem >
+bool
+tnlTimeIndependentPDESolver< Problem >::
+writeEpilog( tnlLogger& logger ) const
+{
+ return this->problem->writeEpilog( logger );
+}
diff --git a/src/TNL/Solvers/SolverConfig_impl.h b/src/TNL/Solvers/SolverConfig_impl.h
index eb981bb35b9ade583e8538376978443e04584e14..a191add046b7f1b4e2b317a26e4abd23ee96606d 100644
--- a/src/TNL/Solvers/SolverConfig_impl.h
+++ b/src/TNL/Solvers/SolverConfig_impl.h
@@ -14,7 +14,7 @@
#include <TNL/Solvers/BuildConfigTags.h>
#include <TNL/Solvers/DummyProblem.h>
#include <TNL/Solvers/PDE/ExplicitTimeStepper.h>
-#include <TNL/Solvers/PDE/PDESolver.h>
+#include <TNL/Solvers/PDE/TimeDependentPDESolver.h>
namespace TNL {
namespace Solvers {
@@ -79,7 +79,7 @@ bool SolverConfig< ConfigTag, ProblemConfig >::configSetup( Config::ConfigDescri
*/
config.addDelimiter( " === Time discretisation parameters ==== " );
typedef PDE::ExplicitTimeStepper< DummyProblemType, ODE::Euler > ExplicitTimeStepper;
- PDE::PDESolver< DummyProblemType, ExplicitTimeStepper >::configSetup( config );
+ PDE::TimeDependentPDESolver< DummyProblemType, ExplicitTimeStepper >::configSetup( config );
ExplicitTimeStepper::configSetup( config );
if( ConfigTagTimeDiscretisation< ConfigTag, ExplicitTimeDiscretisationTag >::enabled ||
ConfigTagTimeDiscretisation< ConfigTag, SemiImplicitTimeDiscretisationTag >::enabled ||
diff --git a/src/TNL/Solvers/SolverStarter_impl.h b/src/TNL/Solvers/SolverStarter_impl.h
index 2d988c8b4120920ca346dd7e8e205b43f64660f9..516dab97d42bf11b0aa4185bc1acf01db2ab1542 100644
--- a/src/TNL/Solvers/SolverStarter_impl.h
+++ b/src/TNL/Solvers/SolverStarter_impl.h
@@ -27,7 +27,7 @@
#include <TNL/Solvers/Linear/Preconditioners/Diagonal.h>
#include <TNL/Solvers/PDE/ExplicitTimeStepper.h>
#include <TNL/Solvers/PDE/SemiImplicitTimeStepper.h>
-#include <TNL/Solvers/PDE/PDESolver.h>
+#include <TNL/Solvers/PDE/TimeDependentPDESolver.h>
namespace TNL {
namespace Solvers {
@@ -358,7 +358,7 @@ bool SolverStarter< ConfigTag > :: runPDESolver( Problem& problem,
/****
* Set-up the PDE solver
*/
- PDE::PDESolver< Problem, TimeStepper > solver;
+ PDE::TimeDependentPDESolver< Problem, TimeStepper > solver;
solver.setProblem( problem );
solver.setTimeStepper( timeStepper );
if( ! solver.setup( parameters ) )
diff --git a/src/Tools/tnl-compile.in b/src/Tools/tnl-compile.in
index 62dd15771afa36e3ccf44468f14644929bce4405..bf196432c86c188f1f5ac7a5dcc9532af478c8f0 100644
--- a/src/Tools/tnl-compile.in
+++ b/src/Tools/tnl-compile.in
@@ -13,3 +13,4 @@ do
done
echo -I@CMAKE_INSTALL_PREFIX@/include/tnl-@tnlVersion@ ${CUDA_FLAGS} ${CXX_STD_FLAGS} ${DEBUG_FLAGS}
+
diff --git a/src/Tools/tnl-diff.h b/src/Tools/tnl-diff.h
index 73036199d0dcb70c3d32ad1fcfae7d1f4e68318e..d8ee5077cc1f0e3d3eb7e239d75e710463f81933 100644
--- a/src/Tools/tnl-diff.h
+++ b/src/Tools/tnl-diff.h
@@ -416,7 +416,7 @@ bool setElementType( const MeshPointer& meshPointer,
std::cerr << "Unable to parse object type " << elementType << "." << std::endl;
return false;
}
- if( parsedElementType[ 0 ] == "tnlStaticVector" )
+ if( parsedElementType[ 0 ] == "Containers::StaticVector" )
return setTupleType< MeshPointer >( meshPointer, inputFileName, parsedObjectType, parsedElementType, parameters );
std::cerr << "Unknown element type " << elementType << "." << std::endl;
diff --git a/src/Tools/tnl-view.h b/src/Tools/tnl-view.h
index 60c366c4aa581f4cb107d12b1357aef61e966440..4f0456aea2dc44edd4eb0f94b8237ea5dd8dfbe8 100644
--- a/src/Tools/tnl-view.h
+++ b/src/Tools/tnl-view.h
@@ -19,6 +19,7 @@
#include <TNL/Containers/MultiVector.h>
#include <TNL/Meshes/Grid.h>
#include <TNL/Functions/MeshFunction.h>
+#include <TNL/Functions/VectorField.h>
using namespace TNL;
@@ -48,6 +49,7 @@ bool writeMeshFunction( const typename MeshFunction::MeshPointer& meshPointer,
const String& inputFileName,
const Config::ParameterContainer& parameters )
{
+
MeshFunction function( meshPointer );
std::cout << "Mesh function: " << function.getType() << std::endl;
if( ! function.load( inputFileName ) )
@@ -58,6 +60,7 @@ bool writeMeshFunction( const typename MeshFunction::MeshPointer& meshPointer,
int verbose = parameters. getParameter< int >( "verbose");
String outputFormat = parameters. getParameter< String >( "output-format" );
+ double scale = parameters. getParameter< double >( "scale" );
String outputFileName;
if( ! getOutputFileName( inputFileName,
outputFormat,
@@ -66,38 +69,73 @@ bool writeMeshFunction( const typename MeshFunction::MeshPointer& meshPointer,
if( verbose )
std::cout << " writing to " << outputFileName << " ... " << std::flush;
- return function.write( outputFileName, outputFormat );
+ return function.write( outputFileName, outputFormat, scale );
}
+template< typename VectorField >
+bool writeVectorField( const typename VectorField::FunctionType::MeshPointer& meshPointer,
+ const String& inputFileName,
+ const Config::ParameterContainer& parameters )
+{
+
+ VectorField field( meshPointer );
+ std::cout << "VectorField: " << field.getType() << std::endl;
+ if( ! field.load( inputFileName ) )
+ {
+ std::cerr << "Unable to load vector field from a file " << inputFileName << "." << std::endl;
+ return false;
+ }
+
+ int verbose = parameters. getParameter< int >( "verbose");
+ String outputFormat = parameters. getParameter< String >( "output-format" );
+ double scale = parameters. getParameter< double >( "scale" );
+ String outputFileName;
+ if( ! getOutputFileName( inputFileName,
+ outputFormat,
+ outputFileName ) )
+ return false;
+ if( verbose )
+ std::cout << " writing to " << outputFileName << " ... " << std::flush;
+
+ return field.write( outputFileName, outputFormat, scale );
+}
+
+
template< typename MeshPointer,
int EntityDimensions,
- typename Real >
+ typename Real,
+ int VectorFieldSize >
bool setMeshFunctionRealType( const MeshPointer& meshPointer,
const String& inputFileName,
+ const Containers::List< String >& parsedObjectType,
const Config::ParameterContainer& parameters )
{
- return writeMeshFunction< Functions::MeshFunction< typename MeshPointer::ObjectType, EntityDimensions, Real > >( meshPointer, inputFileName, parameters );
+ if( VectorFieldSize == 0 )
+ return writeMeshFunction< Functions::MeshFunction< typename MeshPointer::ObjectType, EntityDimensions, Real > >( meshPointer, inputFileName, parameters );
+ return writeVectorField< Functions::VectorField< VectorFieldSize, Functions::MeshFunction< typename MeshPointer::ObjectType, EntityDimensions, Real > > >( meshPointer, inputFileName, parameters );
}
template< typename MeshPointer,
- int EntityDimensions >
+ int EntityDimensions,
+ int VectorFieldSize >
bool setMeshEntityType( const MeshPointer& meshPointer,
const String& inputFileName,
const Containers::List< String >& parsedObjectType,
const Config::ParameterContainer& parameters )
{
if( parsedObjectType[ 3 ] == "float" )
- return setMeshFunctionRealType< MeshPointer, EntityDimensions, float >( meshPointer, inputFileName, parameters );
+ return setMeshFunctionRealType< MeshPointer, EntityDimensions, float, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
if( parsedObjectType[ 3 ] == "double" )
- return setMeshFunctionRealType< MeshPointer, EntityDimensions, double >( meshPointer, inputFileName, parameters );
+ return setMeshFunctionRealType< MeshPointer, EntityDimensions, double, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
if( parsedObjectType[ 3 ] == "long double" )
- return setMeshFunctionRealType< MeshPointer, EntityDimensions, long double >( meshPointer, inputFileName, parameters );
+ return setMeshFunctionRealType< MeshPointer, EntityDimensions, long double, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
std::cerr << "Unsupported arithmetics " << parsedObjectType[ 3 ] << " in mesh function " << inputFileName << std::endl;
return false;
}
template< typename MeshReal,
- typename MeshIndex >
+ typename MeshIndex,
+ int VectorFieldSize >
bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 1, MeshReal, Devices::Host, MeshIndex > >& meshPointer,
const String& inputFileName,
const Containers::List< String >& parsedObjectType,
@@ -109,10 +147,10 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 1, MeshReal, De
switch( meshEntityDimensions )
{
case 0:
- return setMeshEntityType< MeshPointer, 0 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 0, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 1:
- return setMeshEntityType< MeshPointer, 1 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 1, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
default:
std::cerr << "Unsupported mesh functions entity dimensions count " << meshEntityDimensions << "." << std::endl;
@@ -121,7 +159,8 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 1, MeshReal, De
}
template< typename MeshReal,
- typename MeshIndex >
+ typename MeshIndex,
+ int VectorFieldSize >
bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 2, MeshReal, Devices::Host, MeshIndex > >& meshPointer,
const String& inputFileName,
const Containers::List< String >& parsedObjectType,
@@ -133,13 +172,13 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 2, MeshReal, De
switch( meshEntityDimensions )
{
case 0:
- return setMeshEntityType< MeshPointer, 0 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 0, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 1:
- return setMeshEntityType< MeshPointer, 1 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 1, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 2:
- return setMeshEntityType< MeshPointer, 2 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 2, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
default:
std::cerr << "Unsupported mesh functions entity dimensions count " << meshEntityDimensions << "." << std::endl;
@@ -148,7 +187,8 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 2, MeshReal, De
}
template< typename MeshReal,
- typename MeshIndex >
+ typename MeshIndex,
+ int VectorFieldSize >
bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 3, MeshReal, Devices::Host, MeshIndex > >& meshPointer,
const String& inputFileName,
const Containers::List< String >& parsedObjectType,
@@ -160,16 +200,16 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 3, MeshReal, De
switch( meshEntityDimensions )
{
case 0:
- return setMeshEntityType< MeshPointer, 0 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 0, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 1:
- return setMeshEntityType< MeshPointer, 1 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 1, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 2:
- return setMeshEntityType< MeshPointer, 2 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 2, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
case 3:
- return setMeshEntityType< MeshPointer, 3 >( meshPointer, inputFileName, parsedObjectType, parameters );
+ return setMeshEntityType< MeshPointer, 3, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
break;
default:
std::cerr << "Unsupported mesh functions entity dimensions count " << meshEntityDimensions << "." << std::endl;
@@ -177,20 +217,48 @@ bool setMeshEntityDimensions( const SharedPointer< Meshes::Grid< 3, MeshReal, De
}
}
-template< typename MeshPointer >
+template< typename MeshPointer, int VectorFieldSize = 0 >
bool setMeshFunction( const MeshPointer& meshPointer,
const String& inputFileName,
const Containers::List< String >& parsedObjectType,
const Config::ParameterContainer& parameters )
{
+ std::cerr << parsedObjectType[ 1 ] << std::endl;
if( parsedObjectType[ 1 ] != meshPointer->getSerializationType() )
{
std::cerr << "Incompatible mesh type for the mesh function " << inputFileName << "." << std::endl;
return false;
}
- return setMeshEntityDimensions( meshPointer, inputFileName, parsedObjectType, parameters );
+ typedef typename MeshPointer::ObjectType::RealType RealType;
+ typedef typename MeshPointer::ObjectType::IndexType IndexType;
+ return setMeshEntityDimensions< RealType, IndexType, VectorFieldSize >( meshPointer, inputFileName, parsedObjectType, parameters );
}
+template< typename MeshPointer >
+bool setVectorFieldSize( const MeshPointer& meshPointer,
+ const String& inputFileName,
+ Containers::List< String >& parsedObjectType,
+ const Config::ParameterContainer& parameters )
+{
+ int vectorFieldSize = atoi( parsedObjectType[ 1 ].getString() );
+ Containers::List< String > parsedMeshFunctionType;
+ if( ! parseObjectType( parsedObjectType[ 2 ], parsedMeshFunctionType ) )
+ {
+ std::cerr << "Unable to parse mesh function type " << parsedObjectType[ 2 ] << " in a vector field." << std::endl;
+ return false;
+ }
+ switch( vectorFieldSize )
+ {
+ case 1:
+ return setMeshFunction< MeshPointer, 1 >( meshPointer, inputFileName, parsedMeshFunctionType, parameters );
+ case 2:
+ return setMeshFunction< MeshPointer, 2 >( meshPointer, inputFileName, parsedMeshFunctionType, parameters );
+ case 3:
+ return setMeshFunction< MeshPointer, 3 >( meshPointer, inputFileName, parsedMeshFunctionType, parameters );
+ }
+ std::cerr << "Unsupported vector field size " << vectorFieldSize << "." << std::endl;
+ return false;
+}
template< typename MeshPointer, typename Element, typename Real, typename Index, int Dimensions >
bool convertObject( const MeshPointer& meshPointer,
@@ -380,7 +448,7 @@ bool setElementType( const MeshPointer& meshPointer,
return false;
}
if( parsedElementType[ 0 ] == "Containers::StaticVector" ||
- parsedElementType[ 0 ] == "tnlStaticVector" ) // TODO: remove deprecated type names
+ parsedElementType[ 0 ] == "Containers::StaticVector" ) // TODO: remove deprecated type names
return setTupleType< MeshPointer >( meshPointer, inputFileName, parsedObjectType, parsedElementType, parameters );
std::cerr << "Unknown element type " << elementType << "." << std::endl;
@@ -456,6 +524,8 @@ bool processFiles( const Config::ParameterContainer& parameters )
if( parsedObjectType[ 0 ] == "Functions::MeshFunction" ||
parsedObjectType[ 0 ] == "tnlMeshFunction" ) // TODO: remove deprecated type names
setMeshFunction< MeshPointer >( meshPointer, inputFiles[ i ], parsedObjectType, parameters );
+ if( parsedObjectType[ 0 ] == "Functions::VectorField" )
+ setVectorFieldSize< MeshPointer >( meshPointer, inputFiles[ i ], parsedObjectType, parameters );
if( verbose )
std::cout << "[ OK ]. " << std::endl;
}
diff --git a/src/UnitTests/StringTest.cpp b/src/UnitTests/StringTest.cpp
index 1063d139dc9781af1fafc2f8ecf34e2a6c8401fe..3a0444ed64801402a95cec5a2ba80aa739e05163 100644
--- a/src/UnitTests/StringTest.cpp
+++ b/src/UnitTests/StringTest.cpp
@@ -1,5 +1,5 @@
/***************************************************************************
- tnlStringTest.cpp - description
+ StringTest.cpp - description
-------------------
begin : Jul 22, 2013
copyright : (C) 2013 by Tomas Oberhuber
diff --git a/src/core/tnlTypeInfo.h b/src/core/tnlTypeInfo.h
new file mode 100644
index 0000000000000000000000000000000000000000..5d5282f3c4ed13bcc190446f7c1026ac28663944
--- /dev/null
+++ b/src/core/tnlTypeInfo.h
@@ -0,0 +1,39 @@
+/*
+ * File: tnlTypeInfo.h
+ * Author: oberhuber
+ *
+ * Created on July 14, 2016, 3:46 PM
+ */
+
+#pragma once
+
+#include <limits>
+
+template< typename Type >
+class tnlTypeInfo
+{
+};
+
+template<>
+class tnlTypeInfo< double >
+{
+ public:
+
+ typedef double Type;
+
+ static __cuda_callable__
+ Type getMaxValue() { return DBL_MAX; };
+};
+
+template<>
+class tnlTypeInfo< float >
+{
+ public:
+
+ typedef float Type;
+
+ static __cuda_callable__
+ Type getMaxValue() { return FLT_MAX; };
+};
+
+
diff --git a/src/functions/tnlFunctions.h b/src/functions/tnlFunctions.h
new file mode 100644
index 0000000000000000000000000000000000000000..326fa6024b37256360fb6653d195310fe6eb9a98
--- /dev/null
+++ b/src/functions/tnlFunctions.h
@@ -0,0 +1,52 @@
+/*
+ * File: tnlFunctions.h
+ * Author: oberhuber
+ *
+ * Created on July 11, 2016, 6:01 PM
+ */
+
+#pragma once
+
+#include <core/tnlCuda.h>
+
+template< typename Real >
+__cuda_callable__
+Real sign( const Real& x, const Real& smoothing = 0.0 )
+{
+ if( x > smoothing )
+ return 1.0;
+ else if( x < -smoothing )
+ return -1.0;
+ if( smoothing == 0.0 )
+ return 0.0;
+ return sin( ( M_PI * x ) / ( 2.0 * smoothing ) );
+}
+
+template< typename Real >
+__cuda_callable__
+Real positivePart( const Real& arg)
+{
+ return arg > 0.0 ? arg : 0.0;
+}
+
+template< typename Real >
+__cuda_callable__
+Real negativePart( const Real& arg)
+{
+ return arg < 0.0 ? arg : 0.0;
+}
+
+template< typename Real >
+__cuda_callable__
+Real ArgAbsMin( const Real& x, const Real& y )
+{
+ return fabs( x ) < fabs( y ) ? x : y;
+}
+
+template< typename Real >
+__cuda_callable__
+Real ArgAbsMax( const Real& x, const Real& y )
+{
+ return fabs( x ) > fabs( y ) ? x : y;
+}
+
diff --git a/src/operators/tnlDirichletBoundaryConditions_impl.h b/src/operators/tnlDirichletBoundaryConditions_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..70d5e6f7915056de349fc91616c0dfeb682c3d8e
--- /dev/null
+++ b/src/operators/tnlDirichletBoundaryConditions_impl.h
@@ -0,0 +1,144 @@
+/***************************************************************************
+ tnlDirichletBoundaryConditions_impl.h - description
+ -------------------
+ begin : Nov 17, 2014
+ copyright : (C) 2014 by oberhuber
+ email : tomas.oberhuber@fjfi.cvut.cz
+ ***************************************************************************/
+
+/***************************************************************************
+ * *
+ * This program is free software; you can redistribute it and/or modify *
+ * it under the terms of the GNU General Public License as published by *
+ * the Free Software Foundation; either version 2 of the License, or *
+ * (at your option) any later version. *
+ * *
+ ***************************************************************************/
+
+#ifndef TNLDIRICHLETBOUNDARYCONDITIONS_IMPL_H_
+#define TNLDIRICHLETBOUNDARYCONDITIONS_IMPL_H_
+
+#include <functions/tnlFunctionAdapter.h>
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+void
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+configSetup( tnlConfigDescription& config,
+ const String& prefix )
+{
+ Function::configSetup( config );
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+bool
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ return this->function.setup( parameters );
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+void
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+setFunction( const Function& function )
+{
+ this->function = function;
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+Function&
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+getFunction()
+{
+ return this->function;
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+const Function&
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+getFunction() const
+{
+ return *this->function;
+}
+
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+template< typename EntityType,
+ typename MeshFunction >
+__cuda_callable__
+const Real
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time ) const
+{
+ //static_assert( EntityType::getDimensions() == MeshEntitiesDimensions, "Wrong mesh entity dimensions." );
+ return tnlFunctionAdapter< MeshType, Function >::template getValue( this->function, entity, time );
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+ template< typename EntityType >
+__cuda_callable__
+Index
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+getLinearSystemRowLength( const MeshType& mesh,
+ const IndexType& index,
+ const EntityType& entity ) const
+{
+ return 1;
+}
+
+template< typename Mesh,
+ typename Function,
+ int MeshEntitiesDimensions,
+ typename Real,
+ typename Index >
+ template< typename Matrix,
+ typename EntityType,
+ typename MeshFunction >
+__cuda_callable__
+void
+tnlDirichletBoundaryConditions< Mesh, Function, MeshEntitiesDimensions, Real, Index >::
+updateLinearSystem( const RealType& time,
+ const MeshType& mesh,
+ const IndexType& index,
+ const EntityType& entity,
+ const MeshFunction& u,
+ DofVectorType& b,
+ Matrix& matrix ) const
+{
+ typename Matrix::MatrixRow matrixRow = matrix.getRow( index );
+ matrixRow.setElement( 0, index, 1.0 );
+ b[ index ] = tnlFunctionAdapter< MeshType, Function >::getValue( this->function, entity, time );
+}
+
+#endif /* TNLDIRICHLETBOUNDARYCONDITIONS_IMPL_H_ */
diff --git a/src/operators/tnlNeumannReflectionBoundaryConditions.h b/src/operators/tnlNeumannReflectionBoundaryConditions.h
new file mode 100644
index 0000000000000000000000000000000000000000..86fe650a3dbd39de6bcb24691d7e901043c9830e
--- /dev/null
+++ b/src/operators/tnlNeumannReflectionBoundaryConditions.h
@@ -0,0 +1,126 @@
+#pragma once
+
+#include <TNL/Containers/StaticVector.h>
+#include <core/vectors/tnlSharedVector.h>
+#include <TNL/Config/ParameterContainer.h>
+#include <functions/tnlConstantFunction.h>
+
+template< typename Mesh,
+ typename Real = typename Mesh::RealType,
+ typename Index = typename Mesh::IndexType >
+class tnlNeumannReflectionBoundaryConditions
+{
+
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNeumannReflectionBoundaryConditions< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef tnlGrid< 1, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ typedef tnlSharedVector< RealType, DeviceType, IndexType > SharedVector;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef Containers::StaticVector< 1, RealType > VertexType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+ template< typename EntityType,
+ typename MeshFunction >
+ __cuda_callable__
+ const RealType operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time = 0 ) const;
+
+ CoordinatesType tmp;
+
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNeumannReflectionBoundaryConditions< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef tnlGrid< 2, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ typedef tnlSharedVector< RealType, DeviceType, IndexType > SharedVector;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef Containers::StaticVector< 2, RealType > VertexType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+
+ template< typename EntityType,
+ typename MeshFunction >
+ __cuda_callable__
+ const RealType operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time = 0 ) const;
+
+
+ CoordinatesType tmp;
+
+
+};
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+class tnlNeumannReflectionBoundaryConditions< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >
+{
+ public:
+
+ typedef tnlGrid< 3, MeshReal, Device, MeshIndex > MeshType;
+ typedef Real RealType;
+ typedef Device DeviceType;
+ typedef Index IndexType;
+
+ typedef tnlSharedVector< RealType, DeviceType, IndexType > SharedVector;
+ typedef tnlVector< RealType, DeviceType, IndexType> DofVectorType;
+ typedef Containers::StaticVector< 3, RealType > VertexType;
+ typedef typename MeshType::CoordinatesType CoordinatesType;
+
+
+ bool setup( const Config::ParameterContainer& parameters,
+ const String& prefix = "" );
+
+
+ template< typename EntityType,
+ typename MeshFunction >
+ __cuda_callable__
+ const RealType operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time = 0 ) const;
+
+ private:
+
+ CoordinatesType tmp;
+
+
+};
+
+#include <operators/tnlNeumannReflectionBoundaryConditions_impl.h>
+
+#endif /* TNLNEUMANNREFLECTIONBOUNDARYCONDITIONS_H */
diff --git a/src/operators/tnlNeumannReflectionBoundaryConditions_impl.h b/src/operators/tnlNeumannReflectionBoundaryConditions_impl.h
new file mode 100644
index 0000000000000000000000000000000000000000..d791383520bd267beae784fe640d012ddd3147e0
--- /dev/null
+++ b/src/operators/tnlNeumannReflectionBoundaryConditions_impl.h
@@ -0,0 +1,171 @@
+#pragma once
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool
+tnlNeumannReflectionBoundaryConditions< tnlGrid< 1, MeshReal, Device, MeshIndex >, Real, Index >::
+setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ return true;
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+ template< typename EntityType,
+ typename MeshFunction >
+__cuda_callable__
+const Real
+tnlNeumannBoundaryConditions< tnlGrid< 1, MeshReal, Device, MeshIndex >, Function, Real, Index >::
+operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time ) const
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+void tnlNeumannReflectionBoundaryConditions< tnlGrid< 1,MeshReal, Device, MeshIndex >, Real, Index >::
+setBoundaryConditions( const RealType& time,
+ const MeshType& mesh,
+ const IndexType index,
+ const CoordinatesType& coordinates,
+ DofVectorType& u,
+ DofVectorType& fu )
+{
+ tmp = coordinates;
+
+ if(coordinates.x() == 0)
+ tmp.x() = 1;
+ else if(coordinates.x() == mesh. getDimensions().x() - 1)
+ tmp.x() = coordinates.x() - 2;
+
+ u[ index ] = u[mesh.getCellIndex( tmp )];
+}
+
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool
+tnlNeumannReflectionBoundaryConditions< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
+setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ return true;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+ template< typename EntityType,
+ typename MeshFunction >
+__cuda_callable__
+const Real
+tnlNeumannBoundaryConditions< tnlGrid< 1, MeshReal, Device, MeshIndex >, Function, Real, Index >::
+operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time ) const
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+void tnlNeumannReflectionBoundaryConditions< tnlGrid< 2, MeshReal, Device, MeshIndex >, Real, Index >::
+setBoundaryConditions( const RealType& time,
+ const MeshType& mesh,
+ const IndexType index,
+ const CoordinatesType& coordinates,
+ DofVectorType& u,
+ DofVectorType& fu )
+{
+
+ tmp = coordinates;
+
+ if(coordinates.x() == 0)
+ tmp.x() = coordinates.x() + 2;
+ else if(coordinates.x() == mesh. getDimensions().x() - 1)
+ tmp.x() = coordinates.x() - 2;
+
+ if(coordinates.y() == 0)
+ tmp.y() = coordinates.y() + 2;
+ else if(coordinates.y() == mesh. getDimensions().y() - 1)
+ tmp.y() = coordinates.y() - 2;
+
+ u[ index ] = u[mesh.getCellIndex( tmp )];
+}
+
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+bool
+tnlNeumannReflectionBoundaryConditions< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
+setup( const Config::ParameterContainer& parameters,
+ const String& prefix )
+{
+ return true;
+}
+
+template< typename MeshReal,
+ typename Device,
+ typename MeshIndex,
+ typename Real,
+ typename Index >
+ template< typename EntityType,
+ typename MeshFunction >
+__cuda_callable__
+const Real
+tnlNeumannBoundaryConditions< tnlGrid< 1, MeshReal, Device, MeshIndex >, Function, Real, Index >::
+operator()( const MeshFunction& u,
+ const EntityType& entity,
+ const RealType& time ) const
+
+#ifdef HAVE_CUDA
+ __device__ __host__
+#endif
+void tnlNeumannReflectionBoundaryConditions< tnlGrid< 3, MeshReal, Device, MeshIndex >, Real, Index >::
+setBoundaryConditions( const RealType& time,
+ const MeshType& mesh,
+ const IndexType index,
+ const CoordinatesType& coordinates,
+ DofVectorType& u,
+ DofVectorType& fu )
+{
+ tmp = coordinates;
+
+ if(coordinates.x() == 0)
+ tmp.x() = coordinates.x() + 2;
+ else if(coordinates.x() == mesh. getDimensions().x() - 1)
+ tmp.x() = coordinates.x() - 2;
+
+ if(coordinates.y() == 0)
+ tmp.y() = coordinates.y() + 2;
+ else if(coordinates.y() == mesh. getDimensions().y() - 1)
+ tmp.y() = coordinates.y() - 2;
+
+ if(coordinates.z() == 0)
+ tmp.z() = coordinates.z() + 2;
+ else if(coordinates.z() == mesh. getDimensions().z() - 1)
+ tmp.z() = coordinates.z() - 2;
+
+ u[ index ] = u[mesh.getCellIndex( tmp )];
+}
+
+
+
+#endif /* TNLNEUMANNREFLECTIONBOUNDARYCONDITIONS_IMPL_H */
+
diff --git a/tests/benchmarks/heat-equation-benchmark/tnlTestGrid2D.h b/tests/benchmarks/heat-equation-benchmark/tnlTestGrid2D.h
index c4d33c9b06b9942b35cd04c0c4ff069e1050c44a..16d1fe51fa458b60d0c3a648720380f9ce16eba7 100644
--- a/tests/benchmarks/heat-equation-benchmark/tnlTestGrid2D.h
+++ b/tests/benchmarks/heat-equation-benchmark/tnlTestGrid2D.h
@@ -14,7 +14,7 @@
#include <core/tnlObject.h>
#include <core/Devices::Host.h>
-#include <core/vectors/tnlStaticVector.h>
+#include <TNL/Containers/StaticVector.h>
#include <core/vectors/tnlVector.h>
template< int Dimensions,
@@ -50,8 +50,8 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
typedef Real RealType;
typedef Device DeviceType;
typedef Index IndexType;
- typedef tnlStaticVector< 2, Real > VertexType;
- typedef tnlStaticVector< 2, Index > CoordinatesType;
+ typedef Containers::StaticVector< 2, Real > VertexType;
+ typedef Containers::StaticVector< 2, Index > CoordinatesType;
typedef Meshes::Grid< 2, Real, Devices::Host, Index > HostType;
typedef Meshes::Grid< 2, Real, tnlCuda, Index > CudaType;
typedef Meshes::Grid< 2, Real, Device, Index > ThisType;
@@ -78,13 +78,13 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
Grid();
- static tnlString getType();
+ static String getType();
- tnlString getTypeVirtual() const;
+ String getTypeVirtual() const;
- static tnlString getSerializationType();
+ static String getSerializationType();
- virtual tnlString getSerializationTypeVirtual() const;
+ virtual String getSerializationTypeVirtual() const;
void setDimensions( const Index xSize, const Index ySize );
@@ -153,17 +153,17 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
//! Method for restoring the object from a file
bool load( tnlFile& file );
- bool save( const tnlString& fileName ) const;
+ bool save( const String& fileName ) const;
- bool load( const tnlString& fileName );
+ bool load( const String& fileName );
- bool writeMesh( const tnlString& fileName,
- const tnlString& format ) const;
+ bool writeMesh( const String& fileName,
+ const String& format ) const;
template< typename MeshFunction >
bool write( const MeshFunction& function,
- const tnlString& fileName,
- const tnlString& format ) const;
+ const String& fileName,
+ const String& format ) const;
void writeProlog( tnlLogger& logger );
@@ -212,19 +212,19 @@ Meshes::Grid< 2, Real, Device, Index > :: Grid()
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getType()
+String Meshes::Grid< 2, Real, Device, Index > :: getType()
{
- return tnlString( "Meshes::Grid< " ) +
- tnlString( getMeshDimensions() ) + ", " +
- tnlString( ::getType< RealType >() ) + ", " +
- tnlString( Device :: getDeviceType() ) + ", " +
- tnlString( ::getType< IndexType >() ) + " >";
+ return String( "Meshes::Grid< " ) +
+ String( getMeshDimensions() ) + ", " +
+ String( ::getType< RealType >() ) + ", " +
+ String( Device :: getDeviceType() ) + ", " +
+ String( ::getType< IndexType >() ) + " >";
}
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
+String Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
{
return this->getType();
}
@@ -232,7 +232,7 @@ tnlString Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
+String Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
{
return HostType::getType();
};
@@ -240,7 +240,7 @@ tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationTypeVirtual() const
+String Meshes::Grid< 2, Real, Device, Index > :: getSerializationTypeVirtual() const
{
return this->getSerializationType();
};
@@ -596,7 +596,7 @@ bool Meshes::Grid< 2, Real, Device, Index > :: load( tnlFile& file )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: save( const tnlString& fileName ) const
+bool Meshes::Grid< 2, Real, Device, Index > :: save( const String& fileName ) const
{
return tnlObject :: save( fileName );
};
@@ -604,7 +604,7 @@ bool Meshes::Grid< 2, Real, Device, Index > :: save( const tnlString& fileName )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: load( const tnlString& fileName )
+bool Meshes::Grid< 2, Real, Device, Index > :: load( const String& fileName )
{
return tnlObject :: load( fileName );
};
@@ -612,8 +612,8 @@ bool Meshes::Grid< 2, Real, Device, Index > :: load( const tnlString& fileName )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: writeMesh( const tnlString& fileName,
- const tnlString& format ) const
+bool Meshes::Grid< 2, Real, Device, Index > :: writeMesh( const String& fileName,
+ const String& format ) const
{
fstream file;
file. open( fileName. getString(), ios :: out );
@@ -739,8 +739,8 @@ template< typename Real,
typename Index >
template< typename MeshFunction >
bool Meshes::Grid< 2, Real, Device, Index > :: write( const MeshFunction& function,
- const tnlString& fileName,
- const tnlString& format ) const
+ const String& fileName,
+ const String& format ) const
{
if( this->template getEntitiesCount< Cell >() != function. getSize() )
{
@@ -826,8 +826,8 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
typedef Real RealType;
typedef Device DeviceType;
typedef Index IndexType;
- typedef tnlStaticVector< 2, Real > VertexType;
- typedef tnlStaticVector< 2, Index > CoordinatesType;
+ typedef Containers::StaticVector< 2, Real > VertexType;
+ typedef Containers::StaticVector< 2, Index > CoordinatesType;
typedef Meshes::Grid< 2, Real, Devices::Host, Index > HostType;
typedef Meshes::Grid< 2, Real, tnlCuda, Index > CudaType;
typedef Meshes::Grid< 2, Real, Device, Index > ThisType;
@@ -854,13 +854,13 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
Grid();
- static tnlString getType();
+ static String getType();
- tnlString getTypeVirtual() const;
+ String getTypeVirtual() const;
- static tnlString getSerializationType();
+ static String getSerializationType();
- virtual tnlString getSerializationTypeVirtual() const;
+ virtual String getSerializationTypeVirtual() const;
void setDimensions( const Index xSize, const Index ySize );
@@ -929,17 +929,17 @@ class Meshes::Grid< 2, Real, Device, Index > : public tnlObject
//! Method for restoring the object from a file
bool load( tnlFile& file );
- bool save( const tnlString& fileName ) const;
+ bool save( const String& fileName ) const;
- bool load( const tnlString& fileName );
+ bool load( const String& fileName );
- bool writeMesh( const tnlString& fileName,
- const tnlString& format ) const;
+ bool writeMesh( const String& fileName,
+ const String& format ) const;
template< typename MeshFunction >
bool write( const MeshFunction& function,
- const tnlString& fileName,
- const tnlString& format ) const;
+ const String& fileName,
+ const String& format ) const;
void writeProlog( tnlLogger& logger );
@@ -977,19 +977,19 @@ Meshes::Grid< 2, Real, Device, Index > :: Grid()
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getType()
+String Meshes::Grid< 2, Real, Device, Index > :: getType()
{
- return tnlString( "Meshes::Grid< " ) +
- tnlString( getMeshDimensions() ) + ", " +
- tnlString( ::getType< RealType >() ) + ", " +
- tnlString( Device :: getDeviceType() ) + ", " +
- tnlString( ::getType< IndexType >() ) + " >";
+ return String( "Meshes::Grid< " ) +
+ String( getMeshDimensions() ) + ", " +
+ String( ::getType< RealType >() ) + ", " +
+ String( Device :: getDeviceType() ) + ", " +
+ String( ::getType< IndexType >() ) + " >";
}
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
+String Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
{
return this->getType();
}
@@ -997,7 +997,7 @@ tnlString Meshes::Grid< 2, Real, Device, Index > :: getTypeVirtual() const
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
+String Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
{
return HostType::getType();
};
@@ -1005,7 +1005,7 @@ tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationType()
template< typename Real,
typename Device,
typename Index >
-tnlString Meshes::Grid< 2, Real, Device, Index > :: getSerializationTypeVirtual() const
+String Meshes::Grid< 2, Real, Device, Index > :: getSerializationTypeVirtual() const
{
return this->getSerializationType();
};
@@ -1361,7 +1361,7 @@ bool Meshes::Grid< 2, Real, Device, Index > :: load( tnlFile& file )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: save( const tnlString& fileName ) const
+bool Meshes::Grid< 2, Real, Device, Index > :: save( const String& fileName ) const
{
return tnlObject :: save( fileName );
};
@@ -1369,7 +1369,7 @@ bool Meshes::Grid< 2, Real, Device, Index > :: save( const tnlString& fileName )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: load( const tnlString& fileName )
+bool Meshes::Grid< 2, Real, Device, Index > :: load( const String& fileName )
{
return tnlObject :: load( fileName );
};
@@ -1377,8 +1377,8 @@ bool Meshes::Grid< 2, Real, Device, Index > :: load( const tnlString& fileName )
template< typename Real,
typename Device,
typename Index >
-bool Meshes::Grid< 2, Real, Device, Index > :: writeMesh( const tnlString& fileName,
- const tnlString& format ) const
+bool Meshes::Grid< 2, Real, Device, Index > :: writeMesh( const String& fileName,
+ const String& format ) const
{
fstream file;
file. open( fileName. getString(), ios :: out );
@@ -1504,8 +1504,8 @@ template< typename Real,
typename Index >
template< typename MeshFunction >
bool Meshes::Grid< 2, Real, Device, Index > :: write( const MeshFunction& function,
- const tnlString& fileName,
- const tnlString& format ) const
+ const String& fileName,
+ const String& format ) const
{
if( this->template getEntitiesCount< Cell >() != function. getSize() )
{
diff --git a/tests/benchmarks/heat-equation-benchmark/tnlTestGridEntity.h b/tests/benchmarks/heat-equation-benchmark/tnlTestGridEntity.h
index db0228ff1685026ca9a9a55b86ba1195fa58f4b0..8773d9254fe7673a0c69e7961744a2900265dfd5 100644
--- a/tests/benchmarks/heat-equation-benchmark/tnlTestGridEntity.h
+++ b/tests/benchmarks/heat-equation-benchmark/tnlTestGridEntity.h
@@ -53,8 +53,8 @@ class tnlTestGridEntity< Meshes::Grid< Dimensions, Real, Device, Index >, Dimens
constexpr static int getMeshDimensions() { return meshDimensions; };
- typedef tnlStaticVector< meshDimensions, IndexType > EntityOrientationType;
- typedef tnlStaticVector< meshDimensions, IndexType > EntityBasisType;
+ typedef Containers::StaticVector< meshDimensions, IndexType > EntityOrientationType;
+ typedef Containers::StaticVector< meshDimensions, IndexType > EntityBasisType;
typedef tnlTestGridEntity< GridType, entityDimensions, Config > ThisType;
//typedef tnlTestNeighbourGridEntitiesStorage< ThisType > NeighbourGridEntitiesStorageType;