From e7217db4a436509220b84019d4ee6b7d6c20a650 Mon Sep 17 00:00:00 2001
From: Tomas Oberhuber <tomas.oberhuber@fjfi.cvut.cz>
Date: Thu, 29 Nov 2018 14:02:03 +0100
Subject: [PATCH] Deleting legacy code.
---
.../Legacy/fast-sweeping-map/CMakeLists.txt | 22 -
.../fast-sweeping-map/MainBuildConfig.h | 64 -
.../fast-sweeping-map/fastSweepingMapConfig.h | 39 -
src/TNL/Legacy/fast-sweeping-map/main.cpp | 17 -
src/TNL/Legacy/fast-sweeping-map/main.cu | 17 -
src/TNL/Legacy/fast-sweeping-map/main.h | 88 -
.../fast-sweeping-map/tnlFastSweepingMap.h | 188 --
.../tnlFastSweepingMap2D_CUDA_v4_impl.h | 1051 ---------
.../tnlFastSweepingMap2D_impl.h | 823 -------
.../tnlFastSweepingMap_CUDA.h | 196 --
src/TNL/Legacy/fast-sweeping/CMakeLists.txt | 22 -
.../Legacy/fast-sweeping/MainBuildConfig.h | 64 -
.../Legacy/fast-sweeping/fastSweepingConfig.h | 38 -
src/TNL/Legacy/fast-sweeping/main.cpp | 17 -
src/TNL/Legacy/fast-sweeping/main.cu | 17 -
src/TNL/Legacy/fast-sweeping/main.h | 88 -
.../Legacy/fast-sweeping/tnlFastSweeping.h | 186 --
.../tnlFastSweeping2D_CUDA_impl.h | 522 -----
.../tnlFastSweeping2D_CUDA_v2_impl.h | 588 -----
.../tnlFastSweeping2D_CUDA_v3_impl.h | 920 --------
.../tnlFastSweeping2D_CUDA_v4_impl.h | 1003 ---------
.../tnlFastSweeping2D_CUDA_v5_impl.h | 697 ------
.../fast-sweeping/tnlFastSweeping2D_impl.h | 927 --------
.../tnlFastSweeping2D_openMP_impl.h | 399 ----
.../tnlFastSweeping3D_CUDA_impl.h | 961 --------
.../fast-sweeping/tnlFastSweeping3D_impl.h | 307 ---
.../fast-sweeping/tnlFastSweepingSolver.h | 36 -
.../fast-sweeping/tnlFastSweeping_CUDA.h | 194 --
.../CMakeLists.txt | 23 -
.../MainBuildConfig.h | 64 -
.../hamilton-jacobi-parallel-map/gnuplot.txt | 32 -
.../hamilton-jacobi-parallel-map/main.cpp | 17 -
.../hamilton-jacobi-parallel-map/main.cu | 17 -
.../hamilton-jacobi-parallel-map/main.h | 98 -
.../hamilton-jacobi-parallel-map/mapa_png.png | Bin 24841 -> 0 bytes
.../hamilton-jacobi-parallel-map/no-Makefile | 41 -
.../parallelMapConfig.h | 47 -
.../Legacy/hamilton-jacobi-parallel-map/run | 43 -
.../tnl-err2eoc-2.py | 141 --
.../tnlParallelMapSolver.h | 217 --
.../tnlParallelMapSolver2D_impl.h | 1315 -----------
.../hamilton-jacobi-parallel/CMakeLists.txt | 23 -
.../MainBuildConfig.h | 64 -
.../Legacy/hamilton-jacobi-parallel/main.cpp | 17 -
.../Legacy/hamilton-jacobi-parallel/main.cu | 17 -
.../Legacy/hamilton-jacobi-parallel/main.h | 142 --
.../hamilton-jacobi-parallel/no-Makefile | 41 -
.../parallelEikonalConfig.h | 46 -
src/TNL/Legacy/hamilton-jacobi-parallel/run | 64 -
.../hamilton-jacobi-parallel/tnl-err2eoc-2.py | 141 --
.../tnlParallelEikonalSolver.h | 366 ----
.../tnlParallelEikonalSolver2D_impl.h | 1928 -----------------
.../tnlParallelEikonalSolver3D_impl.h | 1706 ---------------
src/TNL/Legacy/narrow-band/CMakeLists.txt | 22 -
src/TNL/Legacy/narrow-band/MainBuildConfig.h | 64 -
src/TNL/Legacy/narrow-band/main.cpp | 17 -
src/TNL/Legacy/narrow-band/main.cu | 17 -
src/TNL/Legacy/narrow-band/main.h | 88 -
src/TNL/Legacy/narrow-band/narrowBandConfig.h | 40 -
src/TNL/Legacy/narrow-band/tnlNarrowBand.h | 186 --
.../tnlNarrowBand2D_CUDA_v4_impl.h | 1317 -----------
.../tnlNarrowBand2D_CUDA_v5_impl.h | 1313 -----------
.../Legacy/narrow-band/tnlNarrowBand2D_impl.h | 927 --------
.../narrow-band/tnlNarrowBand3D_CUDA_impl.h | 961 --------
.../Legacy/narrow-band/tnlNarrowBand3D_impl.h | 307 ---
.../Legacy/narrow-band/tnlNarrowBand_CUDA.h | 203 --
66 files changed, 21563 deletions(-)
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/CMakeLists.txt
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/MainBuildConfig.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/fastSweepingMapConfig.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/main.cpp
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/main.cu
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/main.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap_CUDA.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/CMakeLists.txt
delete mode 100644 src/TNL/Legacy/fast-sweeping/MainBuildConfig.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/fastSweepingConfig.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/main.cpp
delete mode 100644 src/TNL/Legacy/fast-sweeping/main.cu
delete mode 100644 src/TNL/Legacy/fast-sweeping/main.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_openMP_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_impl.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweepingSolver.h
delete mode 100644 src/TNL/Legacy/fast-sweeping/tnlFastSweeping_CUDA.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/CMakeLists.txt
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/MainBuildConfig.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/gnuplot.txt
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cpp
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cu
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/main.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/mapa_png.png
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/no-Makefile
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/parallelMapConfig.h
delete mode 100755 src/TNL/Legacy/hamilton-jacobi-parallel-map/run
delete mode 100755 src/TNL/Legacy/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/CMakeLists.txt
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/MainBuildConfig.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/main.cpp
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/main.cu
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/main.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/no-Makefile
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/parallelEikonalConfig.h
delete mode 100755 src/TNL/Legacy/hamilton-jacobi-parallel/run
delete mode 100755 src/TNL/Legacy/hamilton-jacobi-parallel/tnl-err2eoc-2.py
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h
delete mode 100644 src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/CMakeLists.txt
delete mode 100644 src/TNL/Legacy/narrow-band/MainBuildConfig.h
delete mode 100644 src/TNL/Legacy/narrow-band/main.cpp
delete mode 100644 src/TNL/Legacy/narrow-band/main.cu
delete mode 100644 src/TNL/Legacy/narrow-band/main.h
delete mode 100644 src/TNL/Legacy/narrow-band/narrowBandConfig.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand2D_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand3D_CUDA_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand3D_impl.h
delete mode 100644 src/TNL/Legacy/narrow-band/tnlNarrowBand_CUDA.h
diff --git a/src/TNL/Legacy/fast-sweeping-map/CMakeLists.txt b/src/TNL/Legacy/fast-sweeping-map/CMakeLists.txt
deleted file mode 100644
index 3f9db0da04..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/CMakeLists.txt
+++ /dev/null
@@ -1,22 +0,0 @@
-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 main.cu)
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(fast-sweeping-map main.cpp)
-ENDIF( BUILD_CUDA )
-target_link_libraries (fast-sweeping-map tnl )
-
-
-INSTALL( TARGETS fast-sweeping-map
- 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 ${TNL_TARGET_DATA_DIRECTORY}/examples/fast-sweeping-map )
diff --git a/src/TNL/Legacy/fast-sweeping-map/MainBuildConfig.h b/src/TNL/Legacy/fast-sweeping-map/MainBuildConfig.h
deleted file mode 100644
index ed3d686eb9..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/MainBuildConfig.h
+++ /dev/null
@@ -1,64 +0,0 @@
-/***************************************************************************
- 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() {std::cerr << "MainBuildConfig" <<std::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/Legacy/fast-sweeping-map/fastSweepingMapConfig.h b/src/TNL/Legacy/fast-sweeping-map/fastSweepingMapConfig.h
deleted file mode 100644
index 9251deca87..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/fastSweepingMapConfig.h
+++ /dev/null
@@ -1,39 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping-map/main.cpp b/src/TNL/Legacy/fast-sweeping-map/main.cpp
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/main.cpp
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping-map/main.cu b/src/TNL/Legacy/fast-sweeping-map/main.cu
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/main.cu
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping-map/main.h b/src/TNL/Legacy/fast-sweeping-map/main.h
deleted file mode 100644
index 6f23851c2e..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/main.h
+++ /dev/null
@@ -1,88 +0,0 @@
-/***************************************************************************
- 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,TNL::Devices::Host, int>, double, int> solver;
- if(!solver.init(parameters))
- {
- cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- TNL_CHECK_CUDA_DEVICE;
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver..." <<std::endl;
- solver.run();
- }
-// else if(dim == 3)
-// {
-// tnlFastSweepingMap<tnlGrid<3,double,TNL::Devices::Host, int>, double, int> solver;
-// if(!solver.init(parameters))
-// {
-// cerr << "Solver failed to initialize." <<std::endl;
-// return EXIT_FAILURE;
-// }
-// TNL_CHECK_CUDA_DEVICE;
-// std::cout << "-------------------------------------------------------------" <<std::endl;
-// std::cout << "Starting solver..." <<std::endl;
-// solver.run();
-// }
- else
- {
- std::cerr << "Unsupported number of dimensions: " << dim << "!" <<std::endl;
- return EXIT_FAILURE;
- }
-
-
- time(&stop);
- std::cout << "Solver stopped..." <<std::endl;
- std::cout <<std::endl;
- std::cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) <<std::endl;
- return EXIT_SUCCESS;
-}
-
-
diff --git a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap.h b/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap.h
deleted file mode 100644
index c568329ba2..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap.h
+++ /dev/null
@@ -1,188 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <functions/tnlMeshFunction.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h b/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h
deleted file mode 100644
index d02b8d6c5d..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_CUDA_v4_impl.h
+++ /dev/null
@@ -1,1051 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
- const String& mapFile = parameters.getParameter <String>("map");
- if(! this->map.load( mapFile ))
- cout << "Failed to load map file : " << mapFile <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,0,0, this->changed);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- cudaMemcpy(&run, this->changed,sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- cntr++;
- cout << "Finished set of sweeps #" << cntr << " " << run <<std::endl;
- }
-
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- //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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
-
- if(map_cuda[Entity.getIndex()] != 0.0)
- {
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real value = cudaDofVector2[Entity.getIndex()];
- Real im = abs(1.0/map_cuda[Entity.getIndex()]);
- Real a,b, tmp;
-
- if( i == 0 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1111(i,j);
-// else
-// setupSquare1110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1101(i,j);
-// else
-// setupSquare1100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1011(i,j);
-// else
-// setupSquare1010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1001(i,j);
-// else
-// setupSquare1000(i,j);
-// }
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0111(i,j);
-// else
-// setupSquare0110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0101(i,j);
-// else
-// setupSquare0100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0011(i,j);
-// else
-// setupSquare0010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-#endif
-
-
-
-
-#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_impl.h b/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_impl.h
deleted file mode 100644
index 4bd9e17c56..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap2D_impl.h
+++ /dev/null
@@ -1,823 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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 <<std::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" <<std::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()
-{
-
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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();
-// neighborEntities.refresh(Mesh,Entity.getIndex());
-//
-// if(dofVector[this->Entity.getIndex()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1111(i,j);
-// else
-// setupSquare1110(i,j);
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1101(i,j);
-// else
-// setupSquare1100(i,j);
-// }
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1011(i,j);
-// else
-// setupSquare1010(i,j);
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1001(i,j);
-// else
-// setupSquare1000(i,j);
-// }
-// }
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0111(i,j);
-// else
-// setupSquare0110(i,j);
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0101(i,j);
-// else
-// setupSquare0100(i,j);
-// }
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0011(i,j);
-// else
-// setupSquare0010(i,j);
-// }
-// else
-// {
-// if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0001(i,j);
-// else
-// setupSquare0000(i,j);
-// }
-// }
-// }
-//
-// }
-// }
- cout << "a" <<std::endl;
-
- //data.setLike(dofVector2.getData());
- //data=dofVector2.getData();
- //cout << data.getType() <<std::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 <<std::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)
- {
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-
- Real value = dofVector2[Entity.getIndex()];
- Real im = abs(1.0/map[Entity.getIndex()]);
- Real a,b, tmp;
-
- if( i == 0 )
- a = dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-
-
-
-
-#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap_CUDA.h b/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap_CUDA.h
deleted file mode 100644
index a23057e78c..0000000000
--- a/src/TNL/Legacy/fast-sweeping-map/tnlFastSweepingMap_CUDA.h
+++ /dev/null
@@ -1,196 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i, int* changed);
-//__global__ void runCUDA(tnlFastSweepingMap< tnlGrid< 3,double, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i);
-
-__global__ void initCUDA(tnlFastSweepingMap< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-//__global__ void initCUDA(tnlFastSweepingMap< tnlGrid< 3,double, TNL::Devices::Host, 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/Legacy/fast-sweeping/CMakeLists.txt b/src/TNL/Legacy/fast-sweeping/CMakeLists.txt
deleted file mode 100644
index 1a23d646a4..0000000000
--- a/src/TNL/Legacy/fast-sweeping/CMakeLists.txt
+++ /dev/null
@@ -1,22 +0,0 @@
-set( tnl_fast_sweeping_SOURCES
-# MainBuildConfig.h
-# tnlFastSweeping2D_impl.h
-# tnlFastSweeping.h
-# fastSweepingConfig.h
- main.cpp)
-
-
-IF( BUILD_CUDA )
- CUDA_ADD_EXECUTABLE(fast-sweeping main.cu)
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(fast-sweeping main.cpp)
-ENDIF( BUILD_CUDA )
-target_link_libraries (fast-sweeping tnl )
-
-
-INSTALL( TARGETS fast-sweeping
- 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 ${TNL_TARGET_DATA_DIRECTORY}/examples/fast-sweeping )
diff --git a/src/TNL/Legacy/fast-sweeping/MainBuildConfig.h b/src/TNL/Legacy/fast-sweeping/MainBuildConfig.h
deleted file mode 100644
index ed3d686eb9..0000000000
--- a/src/TNL/Legacy/fast-sweeping/MainBuildConfig.h
+++ /dev/null
@@ -1,64 +0,0 @@
-/***************************************************************************
- 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() {std::cerr << "MainBuildConfig" <<std::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/Legacy/fast-sweeping/fastSweepingConfig.h b/src/TNL/Legacy/fast-sweeping/fastSweepingConfig.h
deleted file mode 100644
index 3df2c1e889..0000000000
--- a/src/TNL/Legacy/fast-sweeping/fastSweepingConfig.h
+++ /dev/null
@@ -1,38 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping/main.cpp b/src/TNL/Legacy/fast-sweeping/main.cpp
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/fast-sweeping/main.cpp
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping/main.cu b/src/TNL/Legacy/fast-sweeping/main.cu
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/fast-sweeping/main.cu
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/fast-sweeping/main.h b/src/TNL/Legacy/fast-sweeping/main.h
deleted file mode 100644
index e5ac15fede..0000000000
--- a/src/TNL/Legacy/fast-sweeping/main.h
+++ /dev/null
@@ -1,88 +0,0 @@
-/***************************************************************************
- 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,TNL::Devices::Host, int>, double, int> solver;
- if(!solver.init(parameters))
- {
- cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- TNL_CHECK_CUDA_DEVICE;
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver..." <<std::endl;
- solver.run();
- }
- else if(dim == 3)
- {
- tnlFastSweeping<tnlGrid<3,double,TNL::Devices::Host, int>, double, int> solver;
- if(!solver.init(parameters))
- {
- cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- TNL_CHECK_CUDA_DEVICE;
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver..." <<std::endl;
- solver.run();
- }
- else
- {
- std::cerr << "Unsupported number of dimensions: " << dim << "!" <<std::endl;
- return EXIT_FAILURE;
- }
-
-
- time(&stop);
- std::cout << "Solver stopped..." <<std::endl;
- std::cout <<std::endl;
- std::cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) <<std::endl;
- return EXIT_SUCCESS;
-}
-
-
diff --git a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping.h
deleted file mode 100644
index 96d26db7b5..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping.h
+++ /dev/null
@@ -1,186 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <functions/tnlMeshFunction.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h
deleted file mode 100644
index bc1da169c0..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_impl.h
+++ /dev/null
@@ -1,522 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
- for(int i = 0; i < 2*n ; i++)
- {
- runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,i);
- cudaDeviceSynchronize();
- }
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- for(int i = 0; i < 2*n ; i++)
- {
- runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,i);
- cudaDeviceSynchronize();
- }
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- for(int i = 2*n - 1; i > -1; i--)
- {
- runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,3,i);
- cudaDeviceSynchronize();
- }
-
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h
deleted file mode 100644
index 3ad5b7944f..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v2_impl.h
+++ /dev/null
@@ -1,588 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-//// for(int i = 0; i < 2*n ; i++)
-// {
-// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,1,0);
-// cudaDeviceSynchronize();
-// }
-// cudaDeviceSynchronize();
-// TNL_CHECK_CUDA_DEVICE;
-//// for(int i = 0; i < 2*n ; i++)
-// {
-// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,0);
-// cudaDeviceSynchronize();
-// }
-// cudaDeviceSynchronize();
-// TNL_CHECK_CUDA_DEVICE;
-//// for(int i = 2*n - 1; i > -1; i--)
-// {
-// runCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver,3,0);
-// cudaDeviceSynchronize();
-// }
-//
-// cudaDeviceSynchronize();
-// TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h
deleted file mode 100644
index ff36d3f8e0..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v3_impl.h
+++ /dev/null
@@ -1,920 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
-// TNL_CHECK_CUDA_DEVICE;
-// for(int i = 0; i < 2*m -1; i++)
-// {
-// runCUDA<2><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,2,i);
-// cudaDeviceSynchronize();
-// }
-// cudaDeviceSynchronize();
-// TNL_CHECK_CUDA_DEVICE;
-// for(int i = 0; i < 2*m -1; i++)
-// {
-// runCUDA<4><<<numBlocks,threadsPerBlock>>>(this->cudaSolver,4,i);
-// cudaDeviceSynchronize();
-// }
-// cudaDeviceSynchronize();
-// TNL_CHECK_CUDA_DEVICE;
-// 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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h
deleted file mode 100644
index e0a9697c2e..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v4_impl.h
+++ /dev/null
@@ -1,1003 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- //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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-
- Real value = cudaDofVector2[Entity.getIndex()];
- Real a,b, tmp;
-
- if( i == 0 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1111(i,j);
- else
- setupSquare1110(i,j);
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1101(i,j);
- else
- setupSquare1100(i,j);
- }
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1011(i,j);
- else
- setupSquare1010(i,j);
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1001(i,j);
- else
- setupSquare1000(i,j);
- }
- }
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0111(i,j);
- else
- setupSquare0110(i,j);
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0101(i,j);
- else
- setupSquare0100(i,j);
- }
- }
- else
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0011(i,j);
- else
- setupSquare0010(i,j);
- }
- else
- {
- if(cudaDofVector[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-// Entity.setCoordinates(CoordinatesType(i,j));
-// Entity.refresh();
-// tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-// cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-// Entity.setCoordinates(CoordinatesType(i,j));
-// Entity.refresh();
-// tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-// cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=INT_MAX; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=INT_MAX; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=INT_MAX; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-INT_MAX; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=0.5*h; //fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=-0.5*h; //fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=0.5*h; //fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=-0.5*h; //fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=-0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=0.5*h; //fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-#endif
-
-
-
-
-#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h
deleted file mode 100644
index 1591bb6137..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_CUDA_v5_impl.h
+++ /dev/null
@@ -1,697 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/fast-sweeping/tnlFastSweeping2D_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_impl.h
deleted file mode 100644
index c4ce8fe6b2..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_impl.h
+++ /dev/null
@@ -1,927 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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" <<std::endl;
- return initGrid();
-}
-
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-bool tnlFastSweeping< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
-{
-
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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();
- neighborEntities.refresh(Mesh,Entity.getIndex());
-
- if(dofVector[this->Entity.getIndex()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1111(i,j);
- else
- setupSquare1110(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1101(i,j);
- else
- setupSquare1100(i,j);
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1011(i,j);
- else
- setupSquare1010(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1001(i,j);
- else
- setupSquare1000(i,j);
- }
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0111(i,j);
- else
- setupSquare0110(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0101(i,j);
- else
- setupSquare0100(i,j);
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0011(i,j);
- else
- setupSquare0010(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0001(i,j);
- else
- setupSquare0000(i,j);
- }
- }
- }
-
- }
- }
- cout << "a" <<std::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() <<std::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() <<std::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();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-
- Real value = dofVector2[Entity.getIndex()];
- Real a,b, tmp;
-
- if( i == 0 )
- a = dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- dofVector2[neighborEntities.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()] <<std::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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-
-
-
-
-#endif /* TNLFASTSWEEPING_IMPL_H_ */
diff --git a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_openMP_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_openMP_impl.h
deleted file mode 100644
index 54bbe931e0..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping2D_openMP_impl.h
+++ /dev/null
@@ -1,399 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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/Legacy/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h
deleted file mode 100644
index 6a5195cfe4..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_CUDA_impl.h
+++ /dev/null
@@ -1,961 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j,k));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- Real value = cudaDofVector2[Entity.getIndex()];
- Real a,b,c, tmp;
-
- if( i == 0 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0, 0 >()] );
- }
-
- if( j == 0 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1, 0 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1, 0 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1, 0 >()] );
- }
-
- if( k == 0 )
- c = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, 1 >()];
- else if( k == Mesh.getDimensions().z() - 1 )
- c = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, -1 >()];
- else
- {
- c = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/fast-sweeping/tnlFastSweeping3D_impl.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_impl.h
deleted file mode 100644
index e22de0ab85..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping3D_impl.h
+++ /dev/null
@@ -1,307 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- Real value = dofVector2[Entity.getIndex()];
- Real a,b,c, tmp;
-
- if( i == 0 )
- a = dofVector2[neighborEntities.template getEntityIndex< 1, 0, 0>()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = dofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()];
- else
- {
- a = fabsMin( dofVector2[neighborEntities.template getEntityIndex< -1, 0, 0>()],
- dofVector2[neighborEntities.template getEntityIndex< 1, 0, 0>()] );
- }
-
- if( j == 0 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, 1, 0>()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, -1, 0>()];
- else
- {
- b = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, -1, 0>()],
- dofVector2[neighborEntities.template getEntityIndex< 0, 1, 0>()] );
- }
-
- if( k == 0 )
- c = dofVector2[neighborEntities.template getEntityIndex< 0, 0, 1>()];
- else if( k == Mesh.getDimensions().z() - 1 )
- c = dofVector2[neighborEntities.template getEntityIndex< 0, 0, -1>()];
- else
- {
- c = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, 0, -1>()],
- dofVector2[neighborEntities.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/Legacy/fast-sweeping/tnlFastSweepingSolver.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweepingSolver.h
deleted file mode 100644
index fc9eb54598..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweepingSolver.h
+++ /dev/null
@@ -1,36 +0,0 @@
-/*
- * 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 Communicator,
- typename Anisotropy = tnlConstanstFunction< Mesh > >
-class tnlFastSweepingSolver : public tnlPDEProblem< Mesh,
- Communicator,
- 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/Legacy/fast-sweeping/tnlFastSweeping_CUDA.h b/src/TNL/Legacy/fast-sweeping/tnlFastSweeping_CUDA.h
deleted file mode 100644
index f531da431b..0000000000
--- a/src/TNL/Legacy/fast-sweeping/tnlFastSweeping_CUDA.h
+++ /dev/null
@@ -1,194 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i);
-__global__ void runCUDA(tnlFastSweeping< tnlGrid< 3,double, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i);
-
-__global__ void initCUDA(tnlFastSweeping< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-__global__ void initCUDA(tnlFastSweeping< tnlGrid< 3,double, TNL::Devices::Host, 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/Legacy/hamilton-jacobi-parallel-map/CMakeLists.txt b/src/TNL/Legacy/hamilton-jacobi-parallel-map/CMakeLists.txt
deleted file mode 100644
index 48382df82d..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/CMakeLists.txt
+++ /dev/null
@@ -1,23 +0,0 @@
-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 main.cu)
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(hamilton-jacobi-parallel-map main.cpp)
-ENDIF( BUILD_CUDA )
-target_link_libraries (hamilton-jacobi-parallel-map tnl )
-
-
-INSTALL( TARGETS hamilton-jacobi-parallel-map
- 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 ${TNL_TARGET_DATA_DIRECTORY}/examples/hamilton-jacobi-parallel-map )
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel-map/MainBuildConfig.h b/src/TNL/Legacy/hamilton-jacobi-parallel-map/MainBuildConfig.h
deleted file mode 100644
index ed3d686eb9..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/MainBuildConfig.h
+++ /dev/null
@@ -1,64 +0,0 @@
-/***************************************************************************
- 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() {std::cerr << "MainBuildConfig" <<std::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/Legacy/hamilton-jacobi-parallel-map/gnuplot.txt b/src/TNL/Legacy/hamilton-jacobi-parallel-map/gnuplot.txt
deleted file mode 100644
index d4ae619839..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/gnuplot.txt
+++ /dev/null
@@ -1,32 +0,0 @@
-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/Legacy/hamilton-jacobi-parallel-map/main.cpp b/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cpp
deleted file mode 100644
index b13498e173..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cpp
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel-map/main.cu b/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cu
deleted file mode 100644
index 7101976712..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.cu
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel-map/main.h b/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.h
deleted file mode 100644
index fff21c77eb..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/main.h
+++ /dev/null
@@ -1,98 +0,0 @@
-/***************************************************************************
- 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 = TNL::Devices::HostDevice;
-
- const int& dim = parameters.getParameter< int >( "dim" );
-
- if(dim == 2)
- {
-
- typedef parallelGodunovMapScheme< tnlGrid<2,double,TNL::Devices::Host, 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,TNL::Devices::Host, int>, double, int > SchemeTypeDevice;
-/*#endif*/
-
- if(device==TNL::Devices::HostDevice)
- {
- typedef TNL::Devices::Host Device;
-
-
- tnlParallelMapSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
- if(!solver.init(parameters))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::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))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::endl;
- solver.run();
- }
- }
-
-
- time(&stop);
- cout <<std::endl;
- cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) <<std::endl;
- return EXIT_SUCCESS;
-}
-
-
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel-map/mapa_png.png b/src/TNL/Legacy/hamilton-jacobi-parallel-map/mapa_png.png
deleted file mode 100644
index 668b6fe24b17b2fec486db28505b41e3beb2091a..0000000000000000000000000000000000000000
GIT binary patch
literal 0
HcmV?d00001
literal 24841
zcmX_o2RzmLAGdLE=oslFAx?GTByo~)vMH-%MD{Ez*&%y|A`~HHXD2Ir=gQ?Idy|#D
zlD+bL-2d}D=k>a;>s9$V=Q}>1_xgCHsw_`S!$LzsLPD#kaOVLD38~b{e@HU;6PZ8!
zNAN%LXEKT!NF;J(Qe}dK1Vf^DM_R-6!?Kr-HS3<&vHpk2H@E3j-d>Vb?7dEG*?&2p
zXQ%m-U{gigog&4EzTYkOg<=?E>6eBJp}Zbh_r*>&Kx`}L-+Q5HPmb?*ak73YNmv;+
z<R~`N=~qmw`6uDSdBfX3z8_DHn4#~*;F6lAHwv{!dcSL7B{z=#*0s!qdhb+^%T`h3
zXe?4m?lvwJ)O$~V3v(Iimd{h;j$R$DEGaCspZprg=Q()_7g9EoR)tY~P^8Vn6dJ)U
zK0H3&G2=M!*}&syYhcWFXL)(KFHg%^<bpruPW6g4)=A_-kV%*2yJ<pt*P!jQeaowO
zRL_DFOBiQDI};5>eCCLant`^i?(pz%_tMt%^t4uip|eP$(ceF@T%4SS?mAkh-(y=F
z8yi<vY$95oEp%T&H4WRSapaQ+nW(F)zg568qkh(U?BtjKyrft>CYMMWiES;esjaQ8
ztc)Yh2(X0pEYx@%Yzx<|)24T2cUcOsw09*ha+9Uw+z+-_j&~!D|25?==GmbhRhLrh
zwJjG-`~lyr!F?s<5!Gn|OW5l2^2eS9YoVw7aS}%hD#tTMe>3{5#J8xbLmH6&#>{d(
zPFTiQEz>U57h2X`^m^r&NRlPOr<)%YX>ga0eCT<5H{;OTM<kIhP$mtho8Qmpy&h_`
z-yJI_Cx@cEEvPjwM5c}w6cG4QTwGjQYQ`V8k668$f)f!HWeshH@5D3rE^RG3eQkbp
zgFS-WV|u-UhbU}2y4zn~H?!q_KFCi)IXZSxi0o<UaAjqsr9j(;XO2d<Rq0Auc0`O!
z!NADMc4=v8b8|DSp;cOEj5NC0l)rVu6Neri8j@CXvbA09_`BTk*|2uevC_(;--*@l
zG~yysCaY&=V=o+;wr@#jSPx0lhJSB!${+JG4Y%iM$nZiHw>i~(Dtubvc(21}BVpWB
zr+>yVEIeFBI$Ox;qF<?Q{=kxfuCDv~lx&86zx-Om(w1SJr`zMlw;5gt&GMg>JuM3h
zl+m*g7#OI{<8rjWC3(~%d4Zk1&T%HBY0~mvgQ_<TSjF(yudVIuf@M-S<~m@Do?;Zc
zP5GnmX7u$v%<Ho%8Pc_u%F*Cv3T0~%_xbB-XV)=l|GlGwo|d+~qr<4yy|id&QHJa?
zGOf$fGCsWd(Tro|SlMQmRSJ&AfGt%Om#AF+zG233uQ~Lbo}S)0KEoQPg>F5&+O6#o
zyW`=4Bi$(5fkebNhf!1h>wJ92%NCLce`4!L`H(wkRkR>W7+vsVA0Nr&4&eirkTiwJ
zcXu<G&Yimzp8iBPz3Y2>`_b6=(FJ00fmMSgx+<yx7ox1Bgb0A6yy7h{Pyf}b)Ue)L
z98Dv8>FxUZ`s{^tAt53DxYrn@NTRaZyVTUwHJ{@H!-8_fa?y}gEZ!f3Y8tdIgWZJF
zD8M3GTUr}Y;-7|NYw?muZkdjYkGEPrhhd}RVx*R4U}9owoB7kw&|q}DWn@(4^vX;p
zt7oWzH8JG#6?{k<53yU|=;W`Gj}PbWbE+pdiBlvh3*Wd=^ySMt<t#HDTZMZn+h@eQ
z_64F^YmT-@j{mM5udch|LW*_s5nYr*&e!3Xoaf+(XnvHCkT7H&V>Tq==Rt#(HK=rK
zp4$j)G6|plZvL!2w#~dmtYglxvL%A!_LJ!ypW~gDmKF<h^Pu41k&%^wc0`aPSx8!?
zK73-AgdX|4)GkY02prAx7cQ9Tlm%(L#3~M1Zx5ANNF>`C8&BMWyYuVUug%R(*t3V-
zABWlzat!}i3OqnN+uP&ucw_#!VH;(R>lEnAu<1NZ#JLTT#Lw|KrAVwA`ii2io}S0{
zavUz?%$YMr$NK~0uG3)&!geeu%7=lD2Fep}-@d(bsi?U4QIYn(no|L3+pRbyb7$v1
zYoX*tp|i4AnL@qmeVm;I<CGSis)WZqpLu#7&bBAm)$L^(7TCqj^raK2U#Ddo7Raj!
zB(#@}#7OAUqMDXns)ww@Tc*W2?nOF%4L$SfpAw6Xsr3pB64f+X<-Bw=dGEXET1|Sc
z!^`w}N3zSw^UDY_zoMcW)nh;ZtC>*7x-7BCOHr++{O9vEJqR>WW^vOM?d|OW0ReX|
zT@0BT92{h|t8rNB&+qWwF44?;cr=|oreYjp1~s{@t?ln$@8pJB%8;}liK4MeoaN=^
z&LU&^dgX9t&2F(rPz5yV^TsM*9~HeHS*deb84hC=_MUt0oz)yTD$;d?5P~w=(=#$M
za)hg*@-ungWofEYrqBC~?cA$7G=XHU2KD0|_}3#Ve@EvecB<<Ss&C%B*||4p8`TlC
z$_2+wBrzP@nwXdf%L}iQDtS+u;l+?Or=p{WhbSA}l9LFWZpx<dzboV3GuMou{=oGv
z9j03;U-3S<zdZ}QMs<hd$4BEH9v)3iO-K4pYf=m^G`K0;pR=;ELf`ONKj$;)SUGB2
zKC-gvw4n1*9;w9e5459$wR$5Qx^2d>B~a0;*Y`oryH-=f8rPztqPHD$(uo0PL*s|-
z;~ExFxHb`RsHN53nY_=>&xd;XZ0<*nWc8pt?Hg>4D(+=k1ytC=LX0A#5;80->;`IM
zsyS4HJ6bvG0I{k-Nc^$2wPjBb6BK+%>u07DgQb@3Dig@rao{9zCiiQA3Z|a(4uj;3
zqW(vRijp!VxWyO`t)sM5Ft2Y|&?;Dbv#-W}QieO)9~YmNozC~h>{dd1%-xI!IoOa#
zmHIJeIt2qIJO>Ys3hB2}G`PjxH*suq0Zan-d)Oo3Q3N)+ygqk=plr_<4DzI}&AU`*
zWo1=QldoJPOHhJ-JG-$5<%!Rz4t4uyG0)t_o~Hy~2%v~PPj(Gg?k4&`nTyFCBi3cy
zfn^7lq-fcz(RU@1`Oq}*4ivC(o3vxLT*acIqEJmKIQMtS<|RXx0=70bfikJBt*xsB
zxh)2$`zy=K)=<B&jDa$+l5nb&v$V9dq$`@NO2fm$S;D?GapYqaEA_+Y+~PA2ikwC+
z5*_i(Vg$kjflflO)=*MN$OSNmIu&K2^4$0D-&a>xVdeM?sy_Z<Z{_Z~J+6Ri(vE=(
zSfp+3<t5IH`Xgq<!}P3DpEF9D0pZCf_LcrPg@?G{q^tJMRz%d92`#Ib)dPJ$QMsU?
zKt5e9B<;KT20VCT0wb&TP(tr&NSi6Y<l!WJfkCy)#%#NJ!Wh{m0~`eS9`h16GqWqq
zD5g+ZH8lP4Q=x=Fn{#mdg3m;Uq`f1~nDWPU%)wsf8tHQr`B9mTsVveAB}1cpj}oan
z2^2N72r{I<W$Cbr&wfvOdOAXSg5aH*&cl`t^##vt9pBm>$U{sK^jiB_dvtX4^AcV$
z5<tzQ{ccyfd`lfB`ffDQObrbmezRNBug+V5B~UX-Ym7{QhB#A<Rs6jl^5wc>q<d|h
z%7pqwB0tXfmt!TI-00}DvT7fc=SrUrXvdf_2W2h{vt((<T;Sj^($RU<{D@Y#nu0-g
zeW537X=!QQmF?Y<M`%N*rrRG=pCavFZhr@T{N1{s7?h7hD96MQXS8`Vd6=#!`qN<X
zm4EzW{|>eksiLn$-}U<4NY!e}SlkAfO)#OIGN92?pnvH1%d~9GytA_F0rV2C8|}go
z5}AF}ijkfY9AeLtBz=;UIIkV|>LKJL;!Xofcy(uL@LYp(qSmXfyOHmdI8O!o_bgzM
zX^iXv%=H)~m*tMsow3+TaT-_pQ-NQ{s-GWu`{Zc&Z?XJV)ZvMdt9^`dCh+tBnP?2~
z*jdf_(6HzE7dDph)t&UNV0w<#m6g`w!|tc<;SHGcUiS9CyW}})ohUz(+`U`Srgb$W
zvfaFA0j^6LPP3pa19z7{uGgycD5Tc}9SVr4Sqk%}-TY=8>P^43P{ixk0=J)sWG+Cf
zhRY8Z;%o3#*OyKr?GB@hi;Gkop7zB;ULB{Ir5V^-w%o<<_sCO!$G>iV<R>%LdQLIp
zK}gz%ht~?AToBD<)vD4%Q`n)I*45F)r>?u|XZ7gtFcl1(vWd7~E2tQW3;FwtOENC`
z<fTO)Y^L*beEfJl!l$OBB&yv!RebUn$AT=>#YeD$n%A6c8^+&Yp9ru}QBg_H?)1`c
zS>n}cQ^-?esZoKWaIiGE&xXleQB1?NEo~{^zps<8iI&xq;)=vVE$}@4dvvg~HeTaa
z61E?SN#P2af>oCMyF0DTLs>i8XOt>Wy}acf)r!>!Vn8^EB=QjHP$<A5=01K6<icA_
z+;>t{e;*wAV@ykQb~z%kT=3%Lw2Zf1pGGy|%SZD2tkSzIQ&l@H1#H}sxI+Nv{@GoJ
zr<c{&)7Mg3)G9<U!lKn0sxTmk#c<=CoO(_9OEb&qbU#(6N33c5Tl}Hbw0pq`ht@ST
zoF&k;tV0ihrp=Ecr>HD@6O3(r_3G6Lj*;B&il{rB;o}%#9T!_(SvjeYXc|CU0LG*G
zww2}97zeFNMG}A9xdfephw0w6dO6xnK$%vhy8f6Hi-U(+K|d#0=vGRlW6Y!(XaZ%#
z*pGLoBPgUB9H6h%Z}sa5C4^yKmT`zu`!k<EPmLf`NfO%&8{ZA<XJAH$2Q*H6`}S>h
z)wW3c({X(5J>%*z2~p95-Sx+6XuT#wg(Sg`=TFH(0m6l(Jz?R@P(=q)u038ml_pcE
z|5=---^IOvvD53Va#pgK8zWt?tQtbvyn4)5faQG96H(`Duz<SxLX5AljMP!By<$xW
z>EC;vU6#Fb8`rO2ugxW<2yOC0+42iYx;cZgqLfvG3;76c>cfr=e+<&{sBgh3m@jae
z5M;vl_DmDL)(~2@TzVrO)#8AhSlYjD7$%eY8!l2};sxAumHYQ;>FD@TkA)H*5~#>V
zI)sFUk5`~S)Bl~IXQ`vA(dU&<hav%-W$X83lLEH;wGew>ZGQfX=11q**?<2032?hz
z{X;YUb>@?VcFm3$byO45A3g(6A<Y*xPFHb!&)BG_xGit-vxEiD3{Tj>7K;aeZt|6g
zc~n$UQ&Uq_#f|jmAxgb&Y?KBdG*vPa2Q9y6Ve{aKNF<7N2!}L8Vx`XtGo$`)W=W2X
zjXgwXxY4&KapH-^0PiG{Bj)xvPrpyY!SO2@vO1NZhVHWzU?5-G+CTHah&f2w?Tj?T
z?+(g<$&5qpAd`s@RvBYiwX*W^lYV7zE4d>gEDWwndKW=J;A7;N5m!JOE|&NZhu+@V
zSsJ~WjiW)%%cgflNbFm>x$XU;d@YxnMUBeuVh-|~(rMzO>w4UNp@un#l>Xw<vp=#a
z{xoPa9nGws3r8YpIQRbb#B`{^tx;TTEe>n|y#SXGXINNSAM3bSTYoBXVs~JBUOm>4
zEP0&ev!8WM>vM0<#aQCZPmi_Q2+6;M<c<xo@%U%wySWOxEIp^ggnxSN7zCLBs%CXI
z3rS0na_(nlBv7geTmSPZ$&L`tiVD(NP}49}AL+TH_6~rf6Ex|pbLY<8%ZMDZF4N6t
zkFc<{rBkL#9882(e>bChYU&XV9iNmW(bF-z;U>+%F6L@K_v6C*$4|Wi>$9kVrbsgY
z%#++}RVf~_P8WHc%26WJa_S5btC+Ot7;!YDPRtejA_CVBJ|A&L;_&xvX$DG4%28Wk
zSv8<qP6RfKs&_Pg(-oR|V?#rG!9KgeYhG&}4>?DS?S4Z&{k=R?qJWLOn-Om_m)E%j
z9pja=iGR5=*%^dMk+zk6O>o#n#p2lyD>)K{?qmRGlsKClMxl7!j8lSgV^w<l_U)3A
z62pR6Z0jnswVc|!%mq&gAE<ZMLd4?b97#v`5_V64;YC0Ixio{W@h|H6Gf;~DnY0Jo
z2(Q->5d$%;zAOdp=IKwWlJt?5LqY}fWZadVy$;K#QiAI=xNCQQdc0LYv^Yd!pTyU>
zu1^J?IyYe<dPBqetSb5fGqXK`tx#@vu`f@aL#eZ|uCj7(zAGI-yA-BVF7|+NAJ_|`
z(tV5jJ7s_WFNH{KwoPAw#5JO$I_q@?gr8XdJp^y`-AaAl<PJVw-qp1=ixM5h$S=D2
z#F_ZGxEstUuBA6l0?RH3fd6~4RO*k9)_g$CxxmUQdU!ENCN(1?!&0ECw3LG1-9}&w
zSG95Z*0mHxT%1x!nrr{6e!k|CJt3u}%KM&FOZwlX55DQIili{Ve^1^HpT3(Bi9uHC
z|CqF|*&co>m#$VYa67FohqLqo{($!AfnmWpS+%4&VIiTE6$dl3jEgoA?NZ4N$UllC
zpuW(mZBFVUx2Z8s|2XKw{S#mzAinyr65@b)cWhu4g$lxH`a{y*D{;og#9VrdY;h3l
zXtHbp=}5YwAYnV=15I$#<hpC|(DKyOr6zo(KH$t1XjEMHK4p`7LFe;_N0yIVKv7<c
z6c)Izj;3&Gc3`VYv3esdb`34-N7^-ex<VDlbNzbbp)P^#hbcca78jR(^O7PSARE<}
zMV&dNHd?1!B2I|y-xkTg4pfSnn<u9lPQMtt3YJkvPwyI<rjru+EH__eP-@PgDfn+_
z^AreGKYl0^*nW<VzHXVO3@`yy0W`q8<h&`0#kD#-pw+Zko2S+E0mBOkKT-OWf8Vy6
zUV8iQ(}G$}YRolW5Irs;9|^D|Boh5_EW=z&bA?#NNqbj2I{@0^K(|@KrU=kLcnzum
zTz+e5F<IBT&9in#>Xa#lK>^F?kCA%%ygE#T{H0;c^INF-+znT;O8r&)Of`lVGcz;L
z?4dA6{sK1nxJWzRMyXOiXinC~H~&VXF+<e_4oUBtx;o{^k3-f;@85SWZ2_8B&MNVv
ze9mU70|XVclvtwQ4^fSvO?NLIrT~c#E+g7!_t|%keNEoCo3nCo_-xLr@x}sm=v=*q
zlo3g!ykZ#I-2M9VLVC7^w`}A`4el+twkO@VQlBp7`tXWNKqFpI5Tq*B1m`8fJEB?m
z^wc`x-}KNlc;K~T8X6Ca;rcGaP4VA+@H2+2nPk_eBP0zC4JG!;v_JNe-Q!ks5)706
zm#a;1&(qt^ZY@~y&_pEho5KeaW`B&e&<L>D3D1!d4gn_#AWKkgsOZ@njOwg@0!Q*I
zQ<SeUxN~$AFou8wwdD@SHT-D6TBy3XxI4b3gooJo-KXhmET~zl)ob6T9g^nUk_Hx>
z<kHvMqOYma=Bmr2qys@I(ne7R;K!kiDJPy`l)b?c1}Gv**fkr05Ln$>raEMNEjK+s
zKM!POvTzaKKQI7=dz|G7vt^#x#nH2f$&)utU<+!XmwwaQst7<&5YKFGRrWSlHb+BO
zTbn>2Kx+a1m8h(BYl;e?%o%mz?a#{$h=q5Sse+vgPF%|j(O7B?KHY{eecs^d-FuGr
z({P_ZfBqW(;m)Pg5db)kIy*am{lXH9L5O6E-Qk4&luP%n=lF(ykb~w47r%B#?Ux(7
zP>sOozu)|&7(VLr=1JpJulQc|44*cfd10boO)*9{;g@<wLG2x!qQ1U9&@U@1&!Z@}
z+-pIxg|;31*wpl6?8bw_+sHdv0gc82Ecr&XmzhEv-vqR|t^2TR2i|)$OaY)rSomB-
zOT-m>iDV-qBjGqDsKVDRSA*X<Vi<fm&m?!mcFfUIHvRtndtzeZ*RPL;1#3g}$}uzg
zyi|+og=sVMKjz|3hshc6YMpxVcj+I#I{zanmXt0_DbYU<`6k>X*duNac8lEv;Q@Jn
zWo=E;?azxjx2V>Cia!x&{<w<4Ppe&UoRm8C>XB-L_0L*k^-pcA_l`Tr??oaK_gBi~
z&%%PbR42~uIf-~}8eLIz&lyVh^PO2@YiUFS!V7dlzi6BLk=V1eRlD~gFm)94U|0^B
zR1%r(`?Uaw+RaN+!n=D}`vQ#qG}wH({SeQLf`5Mx%h<xG9W&!8G5$MMa()PfAOl?8
zVagBuSrx~O3c44|>-X<vCFv~|50kj1wzj)i#?!K=S+aDWEh%xHK_Xy{<D>WY6SZ4N
zS{zD-JU1Vco4NeUF37EWHJ0d)0j&Ty&ILBMhPFTu2Xk_+7D~5A9M6*eX`QrxKV1RO
z2yM0^at)AMc3TCpc#yCAAZ)u4N^_v)($dm`%4jWg>h_9z?N1OMt6lQdPEB7?<VMo~
zYdCf4RLT?A)zMoRoA!9u#Bd;Mq0PlZJRKVc2M6%?*Ly6)X%r*>?0JeLYCg;SeU80i
z$Vw>@<a;>h^>uZI+|f|55RR&ail2UTFbcgkD;e4u_mO<7pb%*)zyf>#IQQ@0g>CAb
z_W{zS;T+k@>+9=_iqvrEkmg5EUv#PX;zIl}0NQUQ{-DF)A1rk*enu$BL9J$R$s1qd
z4tc;ETQal^J??hu8w@f9bL#9>!Q|q21X2o<?76c#24~hq&k2Qjl|8%JN&H)8#PFHK
z$%O%N6-@)GZ&D_kVx*>)7P+64Py%gmQ_3-3SlB`+fz`r}OFCu-NOODp{eQXg<bSQ>
z;0+}vd#kS(Qw6y{IL4wX+_$X6*n=}`{QUfG)?R1erXnM)A5T`|+}YVN(-C4uK`Ug`
z5O>xk<@&a9I6G?&s^g3!6zb!$wzuu(P;x+Y%TlJRKfLhZ*_TfD3b=ZGRh5;k`0o}H
z8*$u{GO}t9ax@5RaO**Di6uTM9fluu>Kt!&y<ov4$Usb?&C}~joKYtN-C$AkEy3xD
z31d^!JBfldk3UFBo--f~NrOrrr$p(#seHwrGwPLGI$Mj4tt~qomZTy8kFH`#kITcH
zZ@>=n$9*?10aeq&f~u-**hV8v6`i5YlN)ue#no6R=0Ht=M7C>k(xiTzKEAcUJuVH`
zxrbb73zQ*kDg>ElVc|)9W~r`C#2w-rMBu_6P+@RpwEuK&A6<4%Q^hT>uKIJn9d>mP
z^Ydua`hW_PYr-pG7_^Oy>cK-YDK%xzgc~8&0a}cmo!vA+xo2~9boBf8*CLlcn@dGv
z=iPj3&`y}Lk!-n3$sJ8j@ibB4Ey10OubGC)!sKp>h_Jl6v%S3yjnNO#Kec~@1KXF`
zKVlIk??C}iO6s>N1*DQl9T5~Tw+DhF(Cv=7Ngc{5lQAlsdroKh1(gYJ-YPP-mKB%k
zkrhKDBWSykSm5OK<Hr+?<VkZ^Ls&RC%60QWvNU;b$egK#rg(vHO^T2M6r<5Dn<59c
z!S0cFKx2xKV8?Wa`B7P06Nw)Id3tTE8mb-`GGqO)Uo(z5D><X~3X%V8c&4iQ&VB^h
z9w>?nYu|%k8*Y)<hpz8^%p5a}RAeO7j)JUn(*Tpn<@G76s6<3WluKRErS;Vbbrqv~
zgaI8b&}gbZMW464=*GcALQ;c0%C_igt_(bO@ht|48I<JbcbsWEs~y=pR;sVS8Jbp6
zTs*b0CuPPS6Hzwu#U)LH`#%V-;jHSOLB$(OEQ~IQ?*DRO#_@dbcS5-6hlq=cjJ@pD
z2efa^ZoO6b^yyQB7~HXZjw$u)chp`84T26@Iuc_v@p0dDjbOqs<Qrlpz#{IFO|EJ8
z{F>iDf4^%|@B$E~0o!L9A)XST$35H<B)n=cwmg?r>>Jgf$KaWrEyXSiXmZ9eU+snM
zd1KY2xxEt-V02V=o+S0kmXp(mJI1P#t99t7Aa!CDBeC`4d`Ow-DFTBQSX;^k>);lK
zPv24FLDPWk2&I$zdwVvMpOmke4(vr_>x#PWVHICU4E8EZ*jMRucQ^Wud6b?Cc9dn3
zeXIdpcy*Qc-yY2pi;odB3Xvb5uzaEFP1kf2@vBb3pz3A$<CL(Bg(OTgV-~e^h!L!!
z%Uz8JdxHO8smFqVlXFg&Z;|RmZ~JLm-Xz2m(CBM2*xmhrI~pp6a+W$Rn?zy*w>3u|
zd*LiS&HGMmBNZjmNEXTf=_5NNsuOMoB^l|UYfQVjM-80@?WTKeqjAcUcnmU8*`i3h
z#?a185z4@$lZpcOR+}dV>Yx&*^CRw4y3bAp%J9ZMO?iFuS1pTU4kzG*gClPo*?%r$
zUqn<Msnb{R7USa1SC4rwqs|AJ@DRbEc#!ikvZ;6Bv2CI=DhPM(k3b(+b}ki<F!`(5
zZ-$lZ{}!W$0K(r+RmC$yP4N1egnBbfdBLb)K=kI#+iLIR(r-zZGc*-KHDG>pT2_rV
zi<pc2<Xw6f_e&<rmOoC+#HCzWiaT47*se^8Y`iTIH@63b1E`H*ln7GbK1j)3U1}iY
z<C(Q%@-r)ZPlQ3Jm<jE+c6Luob%RWFjaR7Wt;w49%?ab2jIU%;JN;dyBR_7-brqT-
z<o1j$*>xpGEFZ}GViiHYwkLp_{K4)5U5ifrATA_ci8G*)B=9w7oUV9osUdCGR}(2&
zu7q~EOQ#7_{W02~SHP02Y_>xlx^ym9fUl**nOa^qy_um)AD{Uo-kkZHF<wbY2}Mso
zO)z{WOc!M0Qtjp9vOU=p4C`8y+3%Z+M8$uzp=a^Cm2%7wCI=`;@5kJ4+b@5@Gafwo
z&$!qSqqjKlCib>s6eCTT^ELCjEP;%Hd%`m4O`h;vY6?8&krkWeTLDzI1!pwTfzuV1
z0yIrGoS9xMB3{pS3XJs%BZSU^bY>~$$o&7Z0MZO!C+!7TY<|{OQDYT#d18i!hCl?d
zuczG+Pg%b3>JA(@kJ2cO`^FVdcw<Eg1QG*<mj-%Inu>fEHE#A_ZRkHcuR=!|0L}aQ
zVEWL4uP1+y37j7*s}M6#H1c{D;$thimbQSt7b44Q-P0|0Y4yH;cvQq3(K5M|&@tE7
zFwz8oG|bAuFY^@c5DaioL+uIr$7UV~&ZsA*rilG0D4?yTR*he%L())9s_3o5y=P%^
zk4uJz1*aS-=}|H|?LuGldAXt#2P+ykYVNkp$b;WPA?>?e<^K6Hk-x<O487;iwXWQ&
z?WK>zf{4z;bjsS1jf2DPNE#h8)+paEx0c34&qL%FIMKD8MDqV~BGk}8J!nI@uI>HO
z=Gg`Zdj8$wT$BpMcg7PZ{!5qThoq!!SFtxLw#3uDkBUI%=i%ck*UJB69sDBbX<gH4
zEi~*cb7->&aj#dkf;;-IsHkX{CHfy;r~vie9{kKfPz&$$0jw#AJWcj13I~Vw&LtC}
zgnZ4XdP$cpHX0aY2aXH<crX2X?+PPz(%i$J#-2PF4rPXaT9es+bpR~pjh%LrU_k!!
zs3_+AP-WrB3bbZ0efs$U%#?<`ZP3^Kk6(hPu!4|dIW2peTn-d%8Qw!R8|qwcKt$aO
z0<vl?-g%}vL62)NNZ3Vi-lSZ&lIeyi-Xr{v!L4LiefX+goC&lm;GQbU676#vE8Ro4
zo+Z@^6_Y*z_`w~$$U=>LdHIgor{1UuyFYtk_6O_JEk{!;)V|GXEMb)XzW`&f3fnrp
zd2|}ZC5^uDAK+fHA>8zP`ZCQ-2PiYJo<o(;k@xX5YG|;oH1og&(Y#XYy8d--M@yh{
z#3m(o(n(})b{2rEW?lwv@x={H9~jc{nI%KkJWNX{GOcO9R8sOOQ=A>x2)0u;6jH;W
zeay_5F-!6?g?e4>*YeAJ5|WmgnF(o<M&m+l9#Ewoif>y4YUX*Zlnz9;npUUH>*_vK
z*35fnc3$&*19jn<5_Jpz6V5!i20EM+SHp^nUcTmunu+>kYh5lI24F}o{cYxgJ)x$&
zTv$Y8cxp~93?ZSr>)(DE_@(|E7S4QKa)lpx+@iUZ4FqYkxsl!7-Of%`dC?Iq(z`v7
zw0Nt4qNOd}9nv`Emr76i0l!=pWGnwBY0*8_1C5ZjEY*$A6wM1CL*Y^6?MFZsLEkx1
zhLH?ht9Q4_D5<E({YY?IE+l@WaN3}f-@P&@FqE2+66Bk?keYh&U&->3m~AL(CsJ_y
z`}d$FgOp>C>FdLvnU!@rwG-r1cx`SvUAv`!kBNmw=0qeM_*o<oP4Z5&YF7U_=lk|q
zcu-akb)2bo^l0;s?T0{WQj|2gqk9#M2|NBD^}t8e`M~}<CW3769>a@1t3bsDFkYP+
z<|7)7@hoA0QC3_)6=cV*n4swBPLSe9_iJk&RlEW(jZI62lEkO#Xb9m@rW4wQBqb#!
zB*3|;-$+~=t$toG<|0TTaGw=LImiQ$Oyt(Bq1O)cKLUv}AuZDo2LSUetEcokb+XKa
zOezD^luvn)ChpPj;ZH1c7Z<QH%SUM2=*OzI63ld>S_OoK=j;h0U#8QBe5I7ovf4?Y
zeds<L2mj){xg{PV1Q_G7x$FKI{HgnDy`ZrDNS2fYvtLZi({-Q(1dJhTn=gfazBgb&
zVeuixfy@;^%Wk=O@c0UXO}XIoZF*f(uB-megkvSDRA2irW7uS$iCk9AQqMvMFyoK*
zcHkZ9d_7jB!gfCoLGuStqibiHE*1H))s*GJh0-Bw_gZ$W#)!?kT+jo!qje1o&LW@<
z-|n;$V7VG%)i_4|D~dwKAM+0O`q`L2>vIHbEl6|;?O^<EY;15HaD>Ti<kj!z9Us&m
zSC^L$yxPM4%Qdp%0<^}l(*0(Rp*Tox<7*&24Ot7plP`c28Ht6H^g&S9n3B|e9vpT5
z180OP8r}^L(OoR5z;osrAzUj5s}%XM&+3X|u{sjI0<|qpiH8{9J_`wgMCE~qC>7r{
z9A53j_b)3;v2}8-q8q-YJ4^A(7(W;<>Wt9mJ-4~^w~OOQjaEe!K(>IuKls=Cp^DI`
zEd>BZEG#UXxb_J>qzV}i*jk{eLYI%yY1*h6UlUGfzplFYVEXI3eppUGGW>CM`+Ws6
z*JNzg=|j@W?ZyqYwRby?4^^D43YNCqwbk)(U?e1dYOM6Fl2{kDnwP+Kk+@A;Nels(
z0CSB=L2hR=m}+F+p6&RXaNP|Nk=!oJMLPv%7MA_5=jsaUPF|mb+>EuI?=k!Xy5~Rt
zAe!(e)+$}SYjwW*RVbVL#J_+40#}rN{s7&4Q~Vu;G@ST$?<kLV4J=`DSv_M_&Zo@5
zVy0x7e>poh$06ogSYJPGNEQ(u4yfXHdEt|~WEF|Cg}i($VW6^PKFINupic`G)aPbK
zEpEAk!>hqH8jA3S%v&m_fQU#NUa4u^2-O6h(o%GYVq`DJS7<#A4N{VqABqFv3O&bX
z94`)g;fB1+8O7F8#4`d9ID`0x|8B}}P~+-gZx4Dkq=O21Mr!PYf=mVqjEwa44~IIu
z`<Ji!bXGLr!Ak>sA7FmT5XxtHmAWp7<l|$6^x)9YEdl`|Yzt0RY7S9YMJPB!kzEGL
z#$csONI)8*)^H7sVBMa&P?JHRS_Ow~1*5ed<!;aYsOvyRfHT7I3?gPnuQj55W%G-)
zU;n3z*xSERz_NIpauxWsv6rdYTfMQtiTt|a0$~k^;ek!w|8$!XVNyPlu-*z95r}K{
z1VFT!d8OJs?Gk3rXM8o*UBwcW*H%_Qc$htVN@{~`n^Aa;9HMuRnJFLJ+Lj*-_#6!w
z@SY2w&Y~$0CMSf`)6v19mBOeV_en(uWj?7jv@6w>mzTfJbgx=7#0l3n!*AU*2!(Nc
zJUD~wezY*9A5~}7{Fz;>uWu;anS}P_qh$+tjTmAY@ynx?&&CgDg^v$Ajz7%pv3w>u
zfAxK2`H*!%?Y6GZY#F6SZUuyS9zVVur&PAPTKE&v3_=S`7gUVdTJCf9tc;Yy!ORAz
zl(?D4uAA>-Ya0$mI`Y%N68t$>LWodVDfg;LIhPw!`R!{v1)o3v5gU&wjYz|p^Tz?z
zgFQcaoFY8Vq+taNnhLK2p_fnd@ID|R(L(f(c-~l6kXZcq@ng+AFsA$uR?hkTKqy8+
zR3%mQBoYZ9)BH84N#E{c{4w@y17fX~eO?PHkm5nhw*UB1w>O&rV9efLwP2upBwNeh
zY)JU0EAZYkgk^hzwGiw59#AQ*O3MsmzdXf+OwG<_j;{rCe|8`40uaR?_u`9e6W`Ap
zi2em9yZWQ;i3v;-UO5XuZm#6Vk5BfW8XXKna;Ph;1yAyEV5!!yU|`8cy2rX)CI-lI
z{eMpMnOAqj*zL+?`7TGbf_8Hr6=aTL&P@bM6`X!=OUr*}91S3QmbpN;pO%iZ`Y$vp
zm+p7Ll3@=LY``;wh2<l$0tR13x_^+P9ii4~IDjE<B~LJEP(?ShU&05OfR)E!_@ko(
z2!M;8#%1t6pqJ6S0R=lTk%FRM;>pw7IT|8x2cXbf19u8)GI3V+s(k@fEutmiZI0SG
zKQ%P?<aEJJ|MPG!U%t$(bU%?AE$b$hl&4hy7CK<!va@}OWN-9mPvpgd{Co$ttRTgL
z#nYk9mhojHE2t)c?KZ1aswRAA$)6vhYY^@;VeY@sB2zMS=uyAp!Qr*pwYX?iZL=WO
z@JAl^jfaN^<hnNHS(gMC^OtEzJqD{?&Bx4gU|2#;KT8wXbScV)6Mw&lGYZsTFyA-6
zg|Q|&|C3J%?xFepYT5<7)R5+XlD<6wbx;mEjRAPs;550!CocdAb*j{V_Usu`=nP1x
zt7Fw*Apa>?!9sonvKU?mE9K{dNZn$ESbD@55GMR_3X#iq>bsc0VVvtoURz!Lit@5a
z#9NVLN0wbYq7l#mA$F3|iD1R!QBh`l8Mn>ys8L?Bkoj7x*`Pb*&reRUxna+ioRARO
z{3wD$(zg7%72+lLy~uY8Scnw4B|!%GA@_h2lT`lRFI+t4B@~v%#zsJx@I~ytMMQ{r
zX?O#3&<)hHzlVD&UW@9s<sQbyDa5pyGg1tQ?7Yp4vIx+7`1tsmK1$=6m!}9Bs<;<T
zCZK`}48B;UB&dNh0#Vk-1til&g&$CVGtVOA)SRrXt-Za0jYAiO4AoOvG!9w9lW;d<
z|0yrYjH8>4&Go3(j}NZakB=|Nk6JZsh)FZRO3A>L43oRugomRA>k)k2jZ0rTukR$E
zYGwuo`EG*wX=Yh9wiZYrLFp7@r&pTWX0f~GD-WUP>gQ{Q5lgezLKtZRfv$5$4-XDn
z+NC=#T?~1|8yn!w?vD{;?>IaZ+uYDlH@c&S9`=jBXPn-3qOQgg1^M~;MMXzgUb#mH
zV30v3_H1kS$xivdQR2K<UQ<~3=57Xrn>`?P^8ESp`t8z!`}gnPK&c1ZL3zGS>%zTo
z48s;`=3Rgn3k#r<z`Z_7!<thh9b<+gzQC>kNq{{CN{6X~`$b&|Qcy=Q$k65odiUsq
zOte1hL$d`jSK&4iWyToXbVEYoB<KyB#C`7VCJP%~3-l~-eFsEM=7n)5*)KacxAQJN
z+XOrM%WQODeheca+X(kln+LS8m>Cv`I^qO&fR1GKxIs&RNS(wJ6-r{UVZn17oBU8}
z{r{v99NOQ;3zl6pw<4?=-ro!{R-9eM=3S!$o2GW3f-ze5e2^PB*oHnw&QYz1(i{no
zJkt9;2KqaYA_Hy{NFgK7g*(We0RInmjGUajwDqGV^0fwsOaKyM4zPvwknnD3Ao<v3
zRpG_g^r%R`&U3FZfNE?q)>J3vBq<5{y{p&{2MCERY;0=)x>Us>&SVN662hlJ53?<o
zzw|bj8FV42a}LwrIy)Jb$Ce{{dnxlZfn95tsK(?%0o~qq0cqEizsvH=5aagP3&+YD
zQ1wbnooCvjS-ja=)+Z+?K_8W7P`-r!)anJ<r<QNujQv#41^eWa_O6TU*t{_KovMo5
z>;xIQgXx7836iTzz<`G-Ha2$L=lF;i%Qw&;fuw0;37chxfO=c;a>l(_h?E2c1&!K1
zf6BzNNrsk9#kE-rLBIWBl?SBh)gAxo3NR?f#th~SSLKOZ<m~;ZCM6XW4>9%_Gl(c-
z6*+tk*P$8tHyXD`c40EC`QyA_L4}4OP}u9&0R6x2J_zuBE5U$}<{saLtX*)E$=O`t
z3*bxvng{C{O#|8IoqN(*WVtt)LNli9bYh$dY-jm!ME=+7plV5YAMWmNEtQTz1+ImR
z48kA790cDN(K78=`M+GbyT$hu3g7daj+LKqG_D4_`Xe?2SKpYOH+`BU;YD+E(`!2k
z1qKfOHB|MuIEU&ntRi~?_g6gJZ_xYlKYco^Ki)ZpXX)z6DdzR@qi%yWPme-D)*FH<
zskq?+N<o_c5_u1D5IMt5H&Iwc85%X4i#VJy0P^r@NEBWUNlU|x3HJ{$yuniEI|C{Q
z`Qs#vx#M?-t6B>9J%1b?1@&#uv(7y6f@0)LJD-D*qj!#vK@s?^8Hv@2iKruKDui8v
z<V%pXvK8{bh$IO8ukuuJUtCWqQZ+7Z`M)bSEU;kyW?Npu<B@sInAyC11Pmh#5=c?O
zzyQ+*cZkg9&<IW;67l+aEHR5Sn@+gr%eR^{g=J-BB_-qsbr4>Y)R3Ujo&enpY(>z<
zi8Hw+)Q>DJk;8xf!u-Ju>-g@sJW5{`+4?i-hEm!3lQz-Qb<9Dx*upIZ;#*CJ`7iRE
zgephJ$LHK6fapCc>K{qbZCEw)hj@4k1gD|j85RWhT-m2ytf8>(_Dh|=bE%8VK95ue
z-=L?;D%KM*O^~>8<E;{>Taulb_|@c-LlHg=?g5_p#Z7JE>SsqFpniU&+Osrx;D-tM
zifY{RdT5$Jp!eVs46k-`{f<jL*=selQJZoiH6gs*dM*X4$kyWbd3bxf41>&gplEaT
ziwg(<0gVlK_>91o{H$zFNyI`$>gwtOb@*^zUNZi}2dC;WLH&;^6IP@(fzp*hZ2Hvx
z_SIvc!?V#1_>?{-%>ZruTxfGlY^-{YM*RVZ1A`Qhh67<j3esW@#@5&J@SXRsD0V|M
zON@Q(OofiGst#I~FeSA3FU|D@_K2)WM2s2e*K>}y-asPpQf;RGGxSN=UMRGgOx6`x
z4Ty0<BRUV`Sx*hhpAT(zdHy^rGt=tyyk(G%0q~xX=1ZH#0&l9$8Z&41EU2PwCZZv0
zZe%n*IOrRwi+S0V9r~P2AyR-DY~)4&`YI}bqeIeqd$OPurQ}Te{GIcGd(Nw0Vc&mk
zllqX~Hyd+_RDN{U?3EhYxq1vB3#dSX!7~hl@Ij?86B~PxtsgN+@q~Dont5}Zor5De
zDal-iNlu<6QpA=Z_4IrbzJ8qJ)t!CB=X25T_+7`CMu8X7(bG3MS<*^=#@)U27Sdy2
z>qg(raU+^<0MYHUg1ic30vOnDH|ReLxI-klNey0VN0Rt7;d9VXki<zi8pzeptG}!A
zYXn2mbo0xC7|A*|K%D}tAEyK~>Bn&Y08Z+lsMYk3do3W_#ESCh%!RnPI2iMQ*MO1F
z(WjhGhFlPOm{I9uB3DH_SLy@IfIwH@ASUDyocNY$fQf8$KqQ%%a~OR&QN&^}SDz8s
zIHPhLQyx}kiJu3=me=RKI~i<Ovo4q1@G!MIdsJ>wGi6{2v|;El)G>f`;3@#0Zv6#S
z`_fy>j6fR%qKSv-pHiuu1py%t4W7xK4u^pjaAn244+-oM#A5yX45XKB;kwKrQG%(^
z-ESXgY`|1dQDm1YgMV{BB+YX)a-Y-46!Hzjph<iC(8~5|zcK6i^Ew6wwa?dnZY>SK
zXoK<$-g*zHvcW+(8|4@AOGvU6`NhfWJSl$aZ_;pUwL9z|Q;bRF!N2imczyoM$lxs5
zgO<W8maeYMlvCGMT2{(FFZlM)M@L6zsU(Bj&_rRPeZYx4Nrt;rorIj&Jg-!JXb7|N
zuqNP90WNM~a^ztiAV6VT!6YQkKt2tKzH8FjDmU-_9*iMr`&s-ddFwAj%!)~a(8Pa6
z;vF?Cl1z90&)y+~S7&Eu^O)S7Zu-%0Xws^COxS_1hF41Os_8}ESJMz<_ks5;MK7M*
z@!UaL0cgOMdtM)iKDYS!$vTP+FRz#Y$OJ)$vIz#UXafDmYzx>nX#&4ex4!*+2~W_r
z-0|wed+Cu>n*)hi+Jf#jI6z<v9|3Wo@|nE2`rk0JFC0ArASgp#-9)4yl>SSpbisg8
zh5aOqc7ZqDSAbYf5<Z=z-JlnJhbW~B0d!^i;^0OHPh;Z-OMyu25CxZEBsO(u&=@a8
zHe`$koCX}-A!E2701!a}(!7gaDlKPr?MKq@x}Nf-qkS`$of$zd+Xb~s@~}zr5OkSq
zfD&0Qv(hndl3taIXt^mW3Oc#&&BOr;U+SKlVK1NK{Zyipti(^4IO?PC)898C60yl4
z658Nio0Nu&YIU0YN&$kL|MZGb?qwKbfo#f;A3s2akc7E0t&Pix@)LKP9t(D>UnjRb
z9f{59Sy&wU4C;{q?*Jr<<iBVa@to>D#wSv-mH^zRBA4v~;RPZ;G=aIx^YhGPw>c}j
z_p*~z&r81vAyV`{yIMW=H{Z_d-V6c=3a}f9hw?B*v^+kUfKl54$tR#u^~5yiX~Ji*
zi}xxUPh9VHgl16PzDEJk(+6+l(w|JoJoCqxo124a0rcfI@1T^c7&~Dc4jl{S_<lAe
zAsnD}NLrGZn<I>9smmWeygSiGBK}+R;a?_xTGcKaIr-1ZSqVu=hI)FTa_O}7yE>#a
zcAYQ8Bqiz13cclW-K5l<9zD9m61MMIXSE%X`>vsrd=z`(bv%UiWz{;%M{D?s3=57I
zwd$*iibxZ+?^hA2eX|(Dr=dT`@)-riUzcVOAH?)AO%Q(IU-J;b<96KTs?olVpp0oN
z*2kf9wX|$$QylK5cS(hP95Q-@YQo&?-PfcVw55tOjWN5nB%^CB$B`meSY1uTe{U6~
zd?;=FCdpssvpXY^ISAsi(}WMsdPCkRFk%8+6(WKC%cL9-+!IMeHFXMzi;LR`upE4}
z;R95^yG!_9^x?40g^}4EZ~RHcMx5Ec4(3#B7P<tPQAUST=hnvSff@IF)YH%$;78pA
zw148HX~-IsMwF2&*Jo}2Dh+N>rSaGOoa=gO&lERTRy0k|Q~>t;H^UecW&*|h);?gh
z@Q8@#J=&hJekROpEdh;~4`zm2#W9S`xfO5pG*z-u;>-+qa1)|mH$#<!HXNVH8r^b9
zk<py`wcO=TeM==;?S(_uh2=10BXnl>_kKay(f}I;_6Q@IktE@PmbC-rwULTL%yYM}
zEDDbz0pc9d(p1dzvc9i>dPsgLwP!(}cOa)_f|If0$ClB3+2)xf{P=*&2hr!Y-rj#<
z8q4AmULg*WFd*f|KtRZ)+JN?kEQbXdk71WpDa`1gn&5^%HtKgJyFJ77rxmX<{#&m%
zAS8rVF>>p<gwR=pw03^K)vQe8kW33j6P|&4pfD?){vk-y`c+N~FggJHZClQ9U|@sy
zGtxcIU@tgD?j>;lwn@q5?)<_QgRGufE18N1vHg|;vAhNlQ>6@J-?dw!T}>~-Aiqwp
zGayK1`^`^ROgWl_sUDrP=8^IW!75sEJ&?4!)_`B`&)4%gZ1Fi1_Sq2TyFL*_aY-m)
zr90yuBqL1TZ}ToJrYn-%h&np*3Nn$Cmyf7RF(l(KJkBVyJn!ws)-qYYB9z56zvW&&
z;zbAvx{ja=CPz;ip4Hjvdh`anu<i;<SjQZB;H#h)$%S7XpFO)7OT19+njuFP@YyM_
ze33f(G$QB4TLzP|kq_>WWrgI9G(+c4%OY)<3S|y@Qf}aN$Ow4A&|^g~J`?iv)9bDn
zhL>{bz@mOQUJXe*#|u7x6lTNaOZE15#OL{e?83+A{Y_PwS5s8?9{N6mXv5z&t-46!
zOx1+&q9z0429)Y{$^R4*6=mhGoMFazT*#&G8NHgJeraC(alZ)Q_tc{8VHTk#K2M5X
zPOsl55yjS$>IE08;qw?MFMQm2Q_{YgYVVBky9LMl1+^v08702cyAb2blGwT$(D=K1
zKqV5KdsTE;3%p@5m_&?HWZ3_l)Y|;W!N$hs$&*$~fgj($*SuRMG?U-fJgH_vv8QD6
ztqe&#G#i!?j`IUK*p3k!U5=IjllQ<TANSpCsKy}q2?Wa%9n(%E5yG%A0(Ua>LtnDW
z)dO2vFu+O|9C5bV<BH!!1<LK*!j;pmBNz4(LiU&QbQ@E?jkq`M6V5UR74dlV?yyrx
zGvHy=45<(g_8eeU$QQSoDi7AfF!k%i7c!}LF0FU-564pLpLYASd3|7gM35O%h@wah
z$Nfq|?Vjb1_;|(}oHEE(R^8^SA{q-NK);7<lTLkDXFTW-Fxxo2z6;ZR-3yy}^O+}6
z(vF`$4e`uhhOC2=XX`P)@7Y9fr}BC>UyJfe{4vtD$Hd8$_ZkCJ$*_nyFnd^9ni1KN
zL64Az5*=tQc<yRy_Y*wxruz{vGwIiVEW@co6yc5gNlj$fzxS2QQ+20dd2Ot4_Qf<*
z{jh8{vukcSET0>Sd6orKI~+1w{L-=;jo&9Ei0a%(XPF>Kz*-8hKyoR)YaW!;F4a?)
zegk?&QNk&^=7dREQG58x>`fBcPZ!O#-g~|ls(}q#ABO~3XumuxlS@yJP$xyWRO*LS
zr5F|@Cnmx?7$jhwowqV>JOIDNy*3T!Ck3Ovz|M2Bco5<ds#|0?{s})L^=X;o#UN?S
z>O&fXtwY|<6__YtwY1^b0(TXle$Z0P|CFKsjbcV(>3uBV_(N=pO0<F7XUO|LkG*<I
zK-Z&%oPCGEY5-KrHr-($CndDESGaOmPz0F(LWUv0c6&U`42OqTjoB*L!VD@w_(`d5
zXmi?0=vd%>z!_UszeXMD5Bf!!C91dj`_Hp$`G{uqz`p^RD7W~mIPxS>s;AECEXdLh
zdZ?^gXz2fJ$*;FRWG(>5j?a|NBtMRl+>4q$@4&<T^aYBN{Ck^o%z&)KlX4P`O3%WJ
z;53Mx(52`P?(R&61v?-^U+=`GJsGwtWkkFs&a}-GLs$geytdPXWC=;T{-+_@ip#of
z<&ID2p(AJI_26eA@fYb0{$CaVrqymrGjOwwEMI`#1LK6h76ED!%;Uq<3xwXlGy$c|
zgt^6`lcke}?DoZTQ@<!bpSHTfo$6SfeM(9VZMrjTCdGmDA6(gHQVEd$MCJd&9|Hm<
z1V>}cfcBQi%-<!a{P^xke2}iKg;}&;xoUf^WYj3|WSF48^j6`Rc@*03s!O%JoSd#?
zPRMjcqcIS$=0{y~8;uvgT>ZhYsS|dRk9t9u`{PelCnrTjEsgVt4H_<5NSbDz^P@*;
zr)EEho@Ne$nG(<)9$wby$C3j+f}jKhCJCP#EsSx$H0$<lo@)w{_?sfP9zK;c{R;y)
zAcWom3SFu@QteJT{0A5E=imtJbXt2J)qi20h2eVBhc{OBM+IcRnS|cw)Ep64T<ZVs
zFL`yfp16CV&HMray}Ubmeod3_lamLmN};^*((6CDxj6d0ZL$<)#kpbv<9QHp2}v6+
z_WWW2uuUrL)5`dn7_)J=xkL!}M@(1fYExOrOxOemwY{XVlG!53byz7&3nnS10So`i
z5W~MWAZeG>7~i%BLOirY_(cb=QXS0@=f0TtdG>pqh0%DcQp$kT+r))LWF!_au2Cs_
z>K2@nlizZ1!WVWLqA^iv+1snjd1s`??yhN6x8J%cMFQ~&i;|%SIZk_jW;D1TW_;A)
zaQ4q!fF#J)hcj?O4-acP=0uw@%sQ+7Chn@W9?|Au`xg}<DM@al)=5sW-tIEecgDA;
z0I+EJ#d^eii$qXE!vh7L=bN#A3yHE4d$l6Sl+ehIIT)=|Q<ILwuDEzWuuBXE_HP22
z8qq%RIt!|Aa<XLFnfnYAcA^Fd@?!N1EY~k1FA*Ifs|a$?g9YugbcxD4!%qvq=&>h^
zx^K8sf&MdQdv^6f(83oZhW9Xh<hnfik0V9mLne3XXv<0s?nE(X#F^gfH$Mr3Zv;`q
zhYH3G$NBu(PKPe@v45{PrA<>%um0ch)V?nwXAn081kkVUI9BSP;#s~AiAD2RIq1KG
z6+^c7Sb{?0U8quF$SqVB`93dss~K<Y?k<9+v9Yt;v0#lS4U<#nVVYhiY$D4?K0KCk
z(vdf8Z;P<;!zvbzoF<!PQFv*>+`GMR6-`U^(GDg_F$}ObJjC4J9vLD5i=@9kh<-hi
zxuwN`2%knsXY-p&(7a58dn4Se$}OF-31JclmOd_ePt(LEC`VB`sq|pM2C=KBhGuoN
z4`h1+xlSyj7uTp1n+xTP<1NNlU{scftO%{drfe5@Y>~(+2#Th`^v}tpp^`|Vt?@8*
zi$i>7-(l$UODRt){(NpLW($)W7#O$}o;TY9Gw7hlX$SUPL1?;-WovN5Bg3b`r{4{x
zNlKNR8rPPR(%MT;!$FFH83mWps`&MS(|!}ZB&b1lF1=&42#HbK3?ZOI|IOxt-6L3y
zeM%B!0%NT}#B;s8o!O)Wx8@J7hD<>XZatcVpTLX)k?;N79te$2B4bMTs;5vWui=%l
zsn`_UT?ntducV-f#7GQx$O_YRGC6FnUBv$!9E3!Id+pyru33o1FD`PMUQy5K{rQs}
zmo2cco-0=X6Bx&e)%m*=zE$@9Pz22O3*2iBvwF4??T>t;$qV+ibYj2~WS8)C)wmC-
zFa<1x8tK-kM~8+Kv5f6K>)+~+R_ufqy04RnUG4jw5943(k%*ALPHr_dHq*aG>uY6?
zWn2)Q6d|R}<?_h^|1iiz;J%jr(<1HBv9YZ^Pd>9sV>XK33kmD858tkdT+@hjp<I-S
z&kakMLi)oYKuE(nl`(e;X3ySx$Q=z&#wx-hp=BGp8=&w$mJNs45QZTI*Jg|t_&FS8
zR1|Ibc#5W5t&kh$o8NcY#NPOqYl+Jw@Dt4etq2AFoND{Edl?UK=tnv+ChuWZ0R|6p
zr$J}{{|bMy`=2LUS>P832yKVtlRCA2x$>Msoyxv5KmWK$Q59|W{Q1t&&iGn~P%q`n
zkPeMEtf1Fe3KVHC|60UYHrBzI4((dpk^=qcw;+pbs#(Wd??qJUJ>IIk7%+Ce4c?SO
zZ^}fn0+s=BOK>d(7lJ_oPzMpG=&6oKbi?)jog7z#q&XN;TXBRW?nBYaTt_&omx3+!
za~9Qu;21E4!jvk*cS5yaOZo;9^=Ik&;{vQ=bin6f{Z$iONTd>{mZl~&3>ZieHcvb|
zaW>vP6lN^|(hMw}nC;H(beHX~>2Y4H{qDe4q2Cu{0h43@4S$0(t#4=uzeHliWry06
zQ);DrX#5)TBm82EF<ZOdT9H=``Eg7yHenD0jUXMhQ@7KihwcCqp`g2X?C`cYV6=0(
zEcZtOIznP@1vyrJP{4vF2a&}}1Fq8*6nbQ4I-p~NvJ2Vk{+Ub2Nad`Z?QNLTkUad|
z6#FMR5^HwyixoyBQCif40VBoU5^87|13a1Hc!SG%j2!i!8HVf-LjU1C16eiLbS$IW
z{^r8huVmOJBX=kk{+RblnG$|^|FqUd(FUjCQdL0>7rb+|?lBVOrA1rMgZS=`@%<8J
z^-M#G2+_5T<YP~8SJ(x(YM68N7e919IBV=I9m+}riVPSyO?Xhdl#_eUhp{1~1=4>*
zCeOq?+gvum{I(i8d>TOOrMDFzI00!F?1I1(7`-!&&T$huH%Y4NbdNv6Ae?P^$M^5>
zrrgEUsa%c!T@(I&Jt)97N26_S!|I3HtF-qDSh!jh6%{xnWJ~k-%OL6eF5>_padyQj
zE4T~tt^z_r5bb(uc#z^k7J!#YzXTb2SasL~NP6@JB3vbsJIqT?)Qx0GA3Fc&QiCiG
z*<_SIh9#QzGrTEqKs_Zo97YERwQYJ*ayjvZ<rn3RQs6yU9i<pW`&Qmenmd_DDl7zN
zuD8gl&X_YQ^~wJ!>AVB6?%OCHA)%5-l01@-m86n!XH!!4x+N=_nTb2e3`Hcela*u>
zW$!#Cgd!^>3E5lbd-VF7a{tEly{>be^AT$j{dV8Y^ndt>s}YUE5XGS+p%-g>Z4s}h
zv3RKX=ujmbD0%`uwbA1HO1+TigFJ{G(D87nUz%H#=p;p16Wi6dI!gCNoHd>lp^%ZZ
z`e{a7LL-LrYW0Yg-{J*zV%@cUcaKV>>i_m}xsb018;0xDHzJX^pCTzgld9vHuKmeG
z=ND{T++7)(JdE%UzkBx%7O@Cz>9Y*X{oQU|SG%>6Rep0II!+_sn3<W0v4rjM5eZtu
zkrUsqpz;M*I33)fVqq;|Al)_y$({9mb5oO6L5Ye2MUGjVjpQ>ecNVpxK_(HJdehlH
z*8x<cL5Jd_ZFAe%`ncwoU###rK2A20rBys=t%!ov-Hj9odLP)$ex_Ys+wC4*c-FJU
zt2CKM_TUSeiX4GPj9DfoCRlpLTwdMSWU?HPzHp&n`m-a{^<<v2Qtg~dJ-xlyUt#Zo
z3F3Hkj|jQ$Tl>yK4LpJ5eoUb+tP}nA$d?#Yx=ueIm^RYa2YTW}nZ1HXV!qtZ_`S^6
zEQ=1L?a;DHki4hQ!-$=fp*D}*Vy^n`$K$zEwMGAw=ZKd@-5?^m)!)|_TyKh($3936
zzd8iRL}K@9Sth;EmzQt9Ok-QZ2m(jV{CZbcbd9m7waYlQ!EDiD!g+RKfrv{!>}Rfb
z%4~cXng8n$U4|bw&n4@cVy#vIr_w$0oo86CSOnhzjiaIh+53fsGmT&&ux;|hsTlK6
zli82EPt?fhhH=i6@=d(Q8q-^v9eWAcBsf?}gdG0><!#16znSo&*tZ`wDT-uZ70^&(
z8bWU&s5?yE`b{w;aFw;S_}92rVU&sHLl(VR26=V5fI7*9X~cCy6Ez~9K*gVQag7|Y
zqT&g(4y$&v*|GKwS<32auX@uYo{So{;<>M0P91|)>3gL6099!=*XB{;Qet@Th`E@a
z`dgK!6g_p}HTG4#S44(aH!OqgR5ZynQ+?QxEQoyxzl-l~gD2eSUr?!_>P*z)#Z3{o
z$p~i%C?8b5``8G2*RE-s5ImXbtDFnx(~m5+^L9`XBjJF+HKX%BT4eAra=eON;1GU<
zZ=p0%ZESf>YUMj(r#q5l5FeyW)d1Z?ol$$fOFwPrH~UU{X7)y>3JN>p5E?PgDgkPr
zo=>IQ`3kuA29PXvD6uY>p{S;Y?O=*XUiZ^DYCm)OY1(VuS$(0)3SpcdqB4FLQt_Km
z%24~sBK0)T(?jk(>e74LT$<1O<=E63gRZY#V8Xr>NFbp`DcpR6#kA+;tcMgu;_g9Q
z#?yew2<)TSt68f40*&cfmoht+zTcD!iXtmyus9}EPAOzI@TzEMpqx-pO4tKNEeX=|
zt+vM=FehoP=%tgj7Jv24Zabk9Y<ud{%+(rV(dRmYk>I4bjBOc9qD@b_+8x>D(R=Vp
zjT*qjxs5-$<8LCKACqMMO8!oY(yPeo*6NN|=g2ASUUz3@WW2<Oz$@@$+sW%Z>?1{t
z#NHc-pZ{K3-6q$%UPfwb;-&CUJ7|`{;v~h)28$x!n6}T0<$xm|#YII3BYz$%WXZB-
zODx2$MVF2GadKJTNhe`%{;~%wizZQxH}njW!@~9O@z+-z+388E3%u_gp(-8lqT1nG
z(@JX_@!g|rae<aJJWUlx&(aZ1CVMEEC*$2L2leZMf_cQivvYFH3y2ZX?JN$g*iV(6
z<-g1PyS>>;z>wSFw5fUMZw+6<fu_?Sied3`T=n_H^t7vt+n=n^8zmQ#N2xr5wS$Z}
zgbL+Egv=3|K<u;WikzU{anX>gdU|?9VmH8UX2Q0fiG{TPw#(_9E;V*eVsRJN?y`JO
z+uWo^y1C{Gi$)eAh+b1QHDoRGlW)?lfs-)l((>t3$fHLNthflJ=@*w6Sc+Tan%P5D
z8CO|u>wCSu+U4GD_}fQd2+BCJQ7G_Oc=jIQvh~KKD9Xwdbi>ke6oy)q^!JYLCMGr0
z%Ltj&9A2dGK3uk<8usF2WS_r5p7{)+s<5yS9cf`<J7G+xgIqx!Bj_9bL;f+(o~;8!
zbPk*QJJLvcyOhBLH#g^`Yfh%c)zhy<PiVQC{dM&uMbKbURYq{gUL9lK&c3d$(QZrZ
z4%WZgZ+yifhd0o48=X=&c);HT<q~ZlKR<up1!KB8xM>cMdv1&yZS^l#f|;d0&y|?{
zyUTVqo8p6mr2C2k%p0h}2`$Si)pV4qtwANX$G;+xTladR!GdGkU4|lVbvLWs!Fl|9
zTfX!Orbh6rTr20}i~8L%W3~4NkPVgP^;ZUOG$Hcp=m-<%+Quesp>u+=*a4Na`FPiU
zfH)LW97&|;8{O6(m+0?HxK0t=7GA~mpQ5X?jGg+?-w@@6kRN>E#O$xz9I=G8ZzCBk
z*K1hG={O4;J7)=1fEksQ-JR|~FGF8C7NoZayXIc|s>wyS0DpgOZf<K!F<4txMzEV?
zns{J&h}vhAcmK`Wzh2sloO8ZM)*<x3viQu6<dqaDqO!k6#ubq+E2}ZUYkrSSS(7C)
zo{^NCau$bd0D2RQb2gQ+5O|sCtVi>$sJ}1tN=zuQ(@xHk0_%UMdd!87cjj&_*>AnA
z8ze;>q5Ahwmv0x-5?`Pb64qP25=GH(KZ<%S7mh4=NK_11@-i~l?{T>KjrDbjWwLh6
zr$fQ}l$Y|}9q$gxm5iAgH5HCm6%0DX%~He_?!7AHFUm3^AN?)q>F2{I{35?ZE-kFg
zbOx#=3XMh#)P?0wEex%F+FIKj*b;}}Pax3sjh~kC&yr620;^Vyv0Azo5`q^(Si7S-
z8Q;&AmX;cM{7Dshox7CzuRB|fsK+{q;|UfBNotnv?$0=td}~7I7!EvspJcn<)1&_@
zLOU?|o89ueck)I?N3_EE3FFF_;BMQd%2h0UR)(ABxXLp%zKFHkA23d-w>LFi7F@og
z8gBN(qvnE9_AY?<&|sM6i!#&o+<Xg-Eb{5oa+v>E_wJIEQP<)?2#n+E@Ijcu%KEU(
zE-G?$a(XTFv|_SG=uykaJgFQjKB$Lq-1TVccJ<Q;Mg<LtE_jMP7_oT7@i6||4w^S2
zU_e|-#W5)uHJ+JUDeksli6=N%(f6%526TSI7#M*aME`G<RpWfPi`vJb1i60F(gZQ3
z4Tr4e+rpj;oD52^{k$J)FzbDHZ&%0w$LvPVt}8C$zh+5%5!Nm)E_ks4OT&PW(&iN}
zHeVkDnqkXGTzq_m=hkLztt=-4FKMh^C<_<6q=X302?e>I?30+M*U${>WZoed8qx~A
zQ(@|zRZeMBqpREg2B9HQwuka?H2{iXDH7N6Of3A4CldcIWqt6+VZ$i6!g-fM=BLvg
zW#u45=F+v0l3!yuP;|A}Am=jot+GBb4NH?EM!EiJoy74*JA9VE463D(1W%8pZVUX+
z<>iFaL*N!Lu0WEmtdxB4v8hSdf)XB0qz9i|MAHd&po<ZFgStjW)AN5|sbdPrx1f1r
zc~?BK(DPUo7em9Xmh+vht#F@9ZY+EOn8>igZ82O`vZ0#&VweGpz8LE;Ai=>vdzOYl
zhJJ5AibguD0S#tu!x6e9f}vpX`BCRmExvm!o=z@zRuv8YG)j%#7YzZ~56Lua_Dc*l
z!7SXjZy%&xY*o`LD&kv4j<ixxl9Qnhf$52*@WH&R7a@d*VpIT9N{dL0nH+rKRX6xE
zeaC5LFs`%<$(t<g?5IxkASmo7%M$F*{=hJI+lM?^G=NzFO2;K-<z-#}e;(c+n06Iq
zMe)T9694R3lj{-1OIVCLf+&%IQdH|ld$MvR`BLh~X5Bq9p~x{V)!ux3!&NAz>Bku(
zvaL~j2g)m7m1NZckqm-Hii^ED*LV^sc5o8j8hvC_3d4hp>g41koEsPMf`XfVo2#|H
zI-biP0S7g2c!NtEPnd}s<9Vl_AHt8T&SmD=wXr0GuOAObp+@7%$x!2r-5AW<EmB+S
z0Q8l3Y={RdO59t8d#!)o41rph*#G?b0}(tuJ)PjD{P{jS0eS)zh6DZm2)BXds3-90
zSR+kbv;SgqR4T83M!v+{z)2}d$-Nm)PKwc&%`aRCXf|IPOG|(;2`oy45^_7NUDq^Z
zt>W-cn#uLK-%<E{zizO-y<JO32a&~WovZ5VPliS&b)-8F1SZd!dI$1NyYg@TV;q<a
z=~>BXBN848LBn$Rxy#DR?7tNQS%9`}C*Fe1gVL!Xe?66Cxt&)pM)CnN)M1U~V~sHh
z&*^RPYn2zsJHCE}Qx-1vU-iOV*FysjDdHuVJE?Y2iJhMFoCrinH6Cjl8^ju6#6VfW
z5=0~Yr&CGq_~+2sbLY<CZdM=@^ytwe#><sC((F9EykH%mK%qzLPuEP#$jqEqzS>VC
z5?4`NJnA6oWaKqitMPO7v~At)354_Z2Iyz^=jAJLhH0j$T)BLiK$`guYnV&f2KcPt
z{O~4V>eSKEA+SA$+k}M$<9qX$mFJ6pS9bCoPxJ(GGBa&#Z2|YBBrbWsA2MO4>NC7`
zh*5#oY16fFOGN0y)vk#)tS4{Zz70bo!cs+ceXDr#{bUE>D?E3kmQx9vW*B`eiD9mi
z51XYQ#maM(0ILU{QFsF*H3jsD9iMuiN(Xv|WGbuuR~{YQq9j6-*Jqay9WClI(dFNA
zDy9j(XRPi?1dp{P+fWmjfO3XNgsJ5<IDu4ERmsfOy{|uOHV=|d#-``FsG}@*_yMKs
zYk`4*7df4937_qe59r|a|8;Y0MvBaTGatFaf*m{I<TlMGRrJH&WYd|mZQ$P8TvOv3
zLd!TqBH&;gNChl!+_^N$$b*P=U;U5IpFf|md3S&c`Krd@Un_kKPR0f21D_cWiE}YL
zT-^Wf;cB5QWa!hrfLu3OHeFeEdt66Zxe+|C-9FS5<Rw|olL(gv%<CknJ&lOa*4Dmp
zSH7E!A?(&4cV9t{ZZ?IS87clVs5XGzjXCpf?QGwJ1Po3<ynJbCyR1JBkNMrNF@72q
z1@kpnDG6e?UScnwoQw~A<E}yX(QR+psW47VPwVUMd-*gI*;(BlhrQ@(z|2!O_yBCs
zY<KvnARppDLB_#Y2znK8sbXSE^!N3%Kcz%FQOH#L@kceDU_u>eS$gnV@rTimS<>el
z0C^(Hb1U!LMAD@7&UEc3;3WF6-<LIUdJRmD((AOe5%d}G4JE0;BcZLMgBu3<u&gg+
zJ7wBD+u*mEj)5*ot$g;s!&vJ!n~zwvw`-E8*$6OQB@(>@`PJ6xHV?!Sej1l_%aQd<
zTa)3;Kz<=T{b9?<d(%@0z&B5TP7|hb?Sq})Ptm`2Mm5HW2D6f1@xG3Lya-11wzeAQ
z>n-JHT<@1rZ|~`0DZJ;Wc`o&VF!Q6;RhVQh(aTSL5aOIVLNnv0LF+(ZGcq)!Iyr@4
zm#62}$_nY@oBM9{cw#p^KttXeaF*qYO1u?1aG+Oe*tf-XYy^2_$NX%7`NMStjhNYz
zk~bCiphWW|%kN)|gFUt_kzdQvj{&I$#Ohtbb662O!sJ4C2G!RJH9^p?FicX5?z3em
zUw4<R!pT}%s-M~M+gY3`C?b2-j8^U8U^IqxUODHt-ZjQ}5XIx4A4~6{|Km<tw=-5I
zLz)u7iZSOY%LdT5T1ITn#DZx=O@Yi_a@#8d!InUNP==wE0hR-LiKb?FQ-}b=*8%h0
zKBU+m@_5}!OObI-`%+Tn`)y}K@?<yL#ZURekG#$wyBCba7<w6Wgny#|@arpIx+MH^
z28S?sa}k%~Vq?LOb<8=V6h<K<JwW%}DkoQs`w9<ZsX<Pbk#A4|PS~5v$vC{|>d;k)
zkMA1a^DEiUkFqs2;pL9%5a|zOAw$A3S*XV8=_rB5%sN_~FBTPfXBa0^^0*{6Zs-R2
z_f98GpX2J?4NNJ+Ly8>LF#;`TMtM1_3_VCDl9GlR8o(MXEfelDxTAl_)~Q;XD#OvS
zwlY%=j}3jNJO|m=^qn7<e_)YOAVyCO6E*9*`7adW0jz9QyxlT|OZ3h(E=VIzEUzJ(
za_q@PC8dVSmB>Wb0*!PWMzDD~)Jkn!pQohnG<&B&rQve)AyzU+77p6BB+ZrYk)Rg3
z>47F}L)79GjT(_CcKS46(|~u$b95J++TDJfQNhN>#^~On_Fowb$z;J@<5MRS8Nj-<
z{BU_g*}U>!g|>;vI0f~EnUy~OcuGA`y+sNR%Fv6%UDMG~q{m_IIrmcu{eTu<-=x?h
zlkM>jbVw-PfM+!h*B68u`_Tl{DQ(&62q?9#D{wNv5p1#9YZdT*T4lmr!tb=hzQ;(t
z;Ic6LpAZzpiVqJO5=qXC*b;e8be&3RqwaP`0jj?q<s)HBVB6&<HeT~6^s5GH(i{EL
zQemccZcn+bl(;Ydu<P7sN`TDaZ=#ZR%h7Sa@J-E}T|8ix=%;GnTE)YPbn8ZfEY~4B
znJ_c>vES%C!V;FjOW25Kymh5~PhNc*SG8;s)ccoXs`~r7Qf~+o5uKnQhjs?zAa^*I
zeI9!e;4$wVU)_6RlhnRTUidN}um?FgvX{)??s~%oc^Ar;fPk5kh4<^*^(pSDN;t6K
zkhKIK98kDau4;sh>P;vAZ~@3mM|J)0klNDRjE!@=%5i$?m2I6FSQ<0D4F-aUHYy|w
zAT0c0*~h@=W}p3+UHS<~?_GY|CMvL29s_jinKMXSf)e%dBaICGn4#U#5O@2-06vR`
zpd$%PE;r2Woy1!Rpaqhx&<jcBKIcyR^pWnds0sf%)hnlb@Kr(hDKc{*=F8j#c>{sY
zi9CF4kAX6W9X+?Rtg!Is@>*w+sj9Nqev19n6rmEOy{;iO#&80l&M-VA_pSyuEg&EO
zlN!D<GBRFhSdpD%kT<9vW)jdSF|xnI|3V`jMz8qz=0|jLzg)<w4ok?=-)|h&H#A)O
z<1P|sYhrQ~PVKE72`!k~4;~kqcrZpS6t8lSngVatd&d%kevCr8x)-(`a+a2sD?frx
zy4}5t;s8)7PQVX?4oh=u!$QVSe57PlKP^32A_L%6h#4iEilW#uH7fa)u@I<$taMZe
zp%rR$2@Rt0M*f14CRp*-!f2Z_CA0M%yIci?%VZAfXEX6t$;Fy{W~>=@f<4vhOl*fW
z#06-)cLx0q0<%ZU&bM>vY;h~nLLcHko#eW60eiaDxRvK*o$BgiEyn+l-7<%Nx2cJ4
z)CTP}Zv|2yKR%<&z#KonKHq}-!W86ZP~ybZoO@JBn4VfdP|$8g?1kHIzuL&i$p8LR
zgphjgh1MCCjm@=bd`#}+$CZ_prB+9e1=WY83Da@Jo}rU{r|@>yZ8i8#V1n32Zc49-
zNHeOP`f0Dpc;KOY?cmJ%x<KHJDgI<Y{Py1nSVy7|#nMmq@UO_aL7ec&^b5yewibwL
zbYwNM@9kB%Pjz2AXKVLg!~*u+s94y5Ev`DRK1aU@jt6hVM@&k<U~Xp@k%ow|c;0X_
z{Frn8Me9a<N9f+(r<m;il2+Ut$PEYzy}Nz#^s@z{XeSzwVvUokGZW@wP_nQ{gLS46
ev(#{$)ciR%Z)1NEp4oQ%qC1onROAt{@%bM-&rk;d
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel-map/no-Makefile b/src/TNL/Legacy/hamilton-jacobi-parallel-map/no-Makefile
deleted file mode 100644
index bfdc1ef236..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/no-Makefile
+++ /dev/null
@@ -1,41 +0,0 @@
-TNL_VERSION=0.1
-TNL_INSTALL_DIR=${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/Legacy/hamilton-jacobi-parallel-map/parallelMapConfig.h b/src/TNL/Legacy/hamilton-jacobi-parallel-map/parallelMapConfig.h
deleted file mode 100644
index c07ee95aa0..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/parallelMapConfig.h
+++ /dev/null
@@ -1,47 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel-map/run b/src/TNL/Legacy/hamilton-jacobi-parallel-map/run
deleted file mode 100755
index 4844199627..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/run
+++ /dev/null
@@ -1,43 +0,0 @@
-#!/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 $dimensions
-
-
-#cp ../template.dat1 template.dat1
-#cp ../template.dat2 template.dat2
-#cp ../gplt2eps.py gplt2eps.py
-cd ..
-
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py b/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py
deleted file mode 100755
index f8cde3768e..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnl-err2eoc-2.py
+++ /dev/null
@@ -1,141 +0,0 @@
-#!/usr/bin/env python
-
-import sys, string, math
-
-arguments = sys. argv[1:]
-format = "txt"
-output_file_name = "eoc-table.txt"
-input_files = []
-verbose = 1
-size = 1.0
-
-i = 0
-while i < len( arguments ):
- if arguments[ i ] == "--format":
- format = arguments[ i + 1 ]
- i = i + 2
- continue
- if arguments[ i ] == "--output-file":
- output_file_name = arguments[ i + 1 ]
- i = i + 2
- continue
- if arguments[ i ] == "--verbose":
- verbose = float( arguments[ i + 1 ] )
- i = i +2
- continue
- if arguments[ i ] == "--size":
- size = float( arguments[ i + 1 ] )
- i = i +2
- continue
- input_files. append( arguments[ i ] )
- i = i + 1
-
-if not verbose == 0:
- print "Writing to " + output_file_name + " in " + format + "."
-
-h_list = []
-l1_norm_list = []
-l2_norm_list = []
-max_norm_list = []
-items = 0
-
-for file_name in input_files:
- if not verbose == 0:
- print "Processing file " + file_name
- file = open( file_name, "r" )
-
- l1_max = 0.0
- l_max_max = 0.0
- file.readline();
- file.readline();
- for line in file. readlines():
- data = string. split( line )
- h_list. append( size/(float(file_name[0:len(file_name)-5] ) - 1.0) )
- l1_norm_list. append( float( data[ 1 ] ) )
- l2_norm_list. append( float( data[ 2 ] ) )
- max_norm_list. append( float( data[ 3 ] ) )
- items = items + 1
- if not verbose == 0:
- print line
- file. close()
-
-h_width = 12
-err_width = 15
-file = open( output_file_name, "w" )
-if format == "latex":
- file. write( "\\begin{tabular}{|r|l|l|l|l|l|l|}\\hline\n" )
- file. write( "\\raisebox{-1ex}[0ex]{$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/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h b/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h
deleted file mode 100644
index 400e163c9d..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver.h
+++ /dev/null
@@ -1,217 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <functions/tnlMeshFunction.h>
-#include <TNL/Devices/Host.h>
-#include <mesh/tnlGrid.h>
-#include <mesh/grids/tnlGridEntity.h>
-#include <limits.h>
-#include <core/tnlDevice.h>
-
-
-#include <ctime>
-
-#ifdef HAVE_CUDA
-#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 TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorType;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorType;
- typedef tnlGrid< 2, double, TNL::Devices::Host, int > MeshType;
-#ifdef HAVE_CUDA
- typedef TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorTypeCUDA;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorTypeCUDA;
- typedef tnlGrid< 2, double, TNL::Devices::Host, 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/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h b/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h
deleted file mode 100644
index e8cbc6fc16..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel-map/tnlParallelMapSolver2D_impl.h
+++ /dev/null
@@ -1,1315 +0,0 @@
-/***************************************************************************
- 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 = TNL::Devices::HostDevice; /////////////// tnlCuda Device --- vypocet na GPU, TNL::Devices::HostDevice --- 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..." <<std::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 <<std::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 <<std::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 <<std::endl;
-
- this->tau0 = parameters.getParameter <double>("initial-tau");
- cout << "Setting initial tau to " << this->tau0 <<std::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 <<std::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 <<std::endl;
-
- cout << "Initializating scheme..." <<std::endl;
- if(!this->schemeHost.init(parameters))
- {
- cerr << "SchemeHost failed to initialize." <<std::endl;
- return false;
- }
- cout << "Scheme initialized." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- }
-#endif
-
- if(this->device == TNL::Devices::HostDevice)
- {
- 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." <<std::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 << "." <<std::endl;
- tmp[i] = runSubgrid(0, tmp[i],i,tmp_map);
- insertSubgrid(tmp[i], i);
- setSubgridValue(i, 4);
- //cout << "Computed initial SDF on subgrid " << i << "." <<std::endl;
- }
- containsCurve = false;
-
- }
- }
-#ifdef HAVE_CUDA
- else if(this->device == tnlCudaDevice)
- {
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- dim3 threadsPerBlock(this->n, this->n);
- dim3 numBlocks(this->gridCols,this->gridRows);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initRunCUDA2D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- }
-#endif
-
-
- this->currentStep = 1;
- if(this->device == TNL::Devices::HostDevice)
- 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();
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- }
-
-#endif
- cout << "Solver initialized." <<std::endl;
-
- return true;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::run()
-{
- if(this->device == TNL::Devices::HostDevice)
- {
- 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) <<std::endl;
- tmp = getSubgrid(i);
- tmp = runSubgrid(3, tmp ,i,tmp_map);
- insertSubgrid( tmp, i);
- }
- if(getBoundaryCondition(i) & 5)
- {
- //cout << "5 @ " << getBoundaryCondition(i) <<std::endl;
- tmp = getSubgrid(i);
- tmp = runSubgrid(5, tmp ,i,tmp_map);
- insertSubgrid( tmp, i);
- }
- if(getBoundaryCondition(i) & 10)
- {
- //cout << "10 @ " << getBoundaryCondition(i) <<std::endl;
- tmp = getSubgrid(i);
- tmp = runSubgrid(10, tmp ,i,tmp_map);
- insertSubgrid( tmp, i);
- }
- if(getBoundaryCondition(i) & 12)
- {
- //cout << "12 @ " << getBoundaryCondition(i) <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-
- bool* tmpb;
- cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- int i = 1;
- time_diff = 0.0;
- while (run_host || !end_cuda)
- {
- cout << "Computing at step "<< i++ <<std::endl;
- if(run_host != 0 )
- end_cuda = true;
- else
- end_cuda = false;
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- 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();
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
-
- cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
- }
- cout << "Solving time was: " << time_diff <<std::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() <<std::endl;
- cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) <<std::endl;
- cout << "Solver finished" <<std::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..." <<std::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 <<std::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 <<std::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 ) <<std::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..." <<std::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 << "." <<std::endl;
- cout << "Setting gridRows to " << this->gridRows << "." <<std::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." <<std::endl;
- if(!this->map_stretched.setSize(stretchedSize))
- cerr << "Could not allocate memory for stretched map." <<std::endl;
- if(!this->unusedCell.setSize(stretchedSize))
- cerr << "Could not allocate memory for supporting stretched grid." <<std::endl;
- int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
- cout << idealStretch <<std::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." <<std::endl;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelMapSolver<2,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
-{
- cout << "Contracting grid..." <<std::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 <<std::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" <<std::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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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();
- neighborEntities.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,neighborEntities,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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Entity.setCoordinates(Containers::StaticVector<2,int>(i,j));
- Entity.refresh();
- neighborEntities.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, neighborEntities, 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/Legacy/hamilton-jacobi-parallel/CMakeLists.txt b/src/TNL/Legacy/hamilton-jacobi-parallel/CMakeLists.txt
deleted file mode 100644
index f6a00127c7..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/CMakeLists.txt
+++ /dev/null
@@ -1,23 +0,0 @@
-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 main.cu)
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(hamilton-jacobi-parallel main.cpp)
-ENDIF( BUILD_CUDA )
-target_link_libraries (hamilton-jacobi-parallel tnl )
-
-
-INSTALL( TARGETS hamilton-jacobi-parallel
- 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 ${TNL_TARGET_DATA_DIRECTORY}/examples/hamilton-jacobi-parallel )
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel/MainBuildConfig.h b/src/TNL/Legacy/hamilton-jacobi-parallel/MainBuildConfig.h
deleted file mode 100644
index ed3d686eb9..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/MainBuildConfig.h
+++ /dev/null
@@ -1,64 +0,0 @@
-/***************************************************************************
- 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() {std::cerr << "MainBuildConfig" <<std::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/Legacy/hamilton-jacobi-parallel/main.cpp b/src/TNL/Legacy/hamilton-jacobi-parallel/main.cpp
deleted file mode 100644
index b13498e173..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/main.cpp
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel/main.cu b/src/TNL/Legacy/hamilton-jacobi-parallel/main.cu
deleted file mode 100644
index 7101976712..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/main.cu
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel/main.h b/src/TNL/Legacy/hamilton-jacobi-parallel/main.h
deleted file mode 100644
index dbaebdcebd..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/main.h
+++ /dev/null
@@ -1,142 +0,0 @@
-/***************************************************************************
- 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 = TNL::Devices::HostDevice;
-
- const int& dim = parameters.getParameter< int >( "dim" );
-
- if(dim == 2)
- {
-
- typedef parallelGodunovEikonalScheme< tnlGrid<2,double,TNL::Devices::Host, 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,TNL::Devices::Host, int>, double, int > SchemeTypeDevice;
- /*#endif*/
-
- if(device==TNL::Devices::HostDevice)
- {
- typedef TNL::Devices::Host Device;
-
-
- tnlParallelEikonalSolver<2,SchemeTypeHost,SchemeTypeDevice, Device> solver;
- if(!solver.init(parameters))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::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))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::endl;
- solver.run();
- }
- // }
- }
- else if(dim == 3)
- {
-
- typedef parallelGodunovEikonalScheme< tnlGrid<3,double,TNL::Devices::Host, 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,TNL::Devices::Host, int>, double, int > SchemeTypeDevice;
- /*#endif*/
-
- if(device==TNL::Devices::HostDevice)
- {
- typedef TNL::Devices::Host Device;
-
-
- tnlParallelEikonalSolver<3,SchemeTypeHost,SchemeTypeDevice, Device> solver;
- if(!solver.init(parameters))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::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))
- {
- std::cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver loop..." <<std::endl;
- solver.run();
- }
- // }
- }
-
- time(&stop);
- std::cout <<std::endl;
- std::cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) <<std::endl;
- return EXIT_SUCCESS;
-}
-
-
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel/no-Makefile b/src/TNL/Legacy/hamilton-jacobi-parallel/no-Makefile
deleted file mode 100644
index bfdc1ef236..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/no-Makefile
+++ /dev/null
@@ -1,41 +0,0 @@
-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/Legacy/hamilton-jacobi-parallel/parallelEikonalConfig.h b/src/TNL/Legacy/hamilton-jacobi-parallel/parallelEikonalConfig.h
deleted file mode 100644
index c27f5ebb39..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/parallelEikonalConfig.h
+++ /dev/null
@@ -1,46 +0,0 @@
-/***************************************************************************
- 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/Legacy/hamilton-jacobi-parallel/run b/src/TNL/Legacy/hamilton-jacobi-parallel/run
deleted file mode 100755
index 3aece294a9..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/run
+++ /dev/null
@@ -1,64 +0,0 @@
-#!/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 $size *.diff
-
-
diff --git a/src/TNL/Legacy/hamilton-jacobi-parallel/tnl-err2eoc-2.py b/src/TNL/Legacy/hamilton-jacobi-parallel/tnl-err2eoc-2.py
deleted file mode 100755
index f8cde3768e..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/tnl-err2eoc-2.py
+++ /dev/null
@@ -1,141 +0,0 @@
-#!/usr/bin/env python
-
-import sys, string, math
-
-arguments = sys. argv[1:]
-format = "txt"
-output_file_name = "eoc-table.txt"
-input_files = []
-verbose = 1
-size = 1.0
-
-i = 0
-while i < len( arguments ):
- if arguments[ i ] == "--format":
- format = arguments[ i + 1 ]
- i = i + 2
- continue
- if arguments[ i ] == "--output-file":
- output_file_name = arguments[ i + 1 ]
- i = i + 2
- continue
- if arguments[ i ] == "--verbose":
- verbose = float( arguments[ i + 1 ] )
- i = i +2
- continue
- if arguments[ i ] == "--size":
- size = float( arguments[ i + 1 ] )
- i = i +2
- continue
- input_files. append( arguments[ i ] )
- i = i + 1
-
-if not verbose == 0:
- print "Writing to " + output_file_name + " in " + format + "."
-
-h_list = []
-l1_norm_list = []
-l2_norm_list = []
-max_norm_list = []
-items = 0
-
-for file_name in input_files:
- if not verbose == 0:
- print "Processing file " + file_name
- file = open( file_name, "r" )
-
- l1_max = 0.0
- l_max_max = 0.0
- file.readline();
- file.readline();
- for line in file. readlines():
- data = string. split( line )
- h_list. append( size/(float(file_name[0:len(file_name)-5] ) - 1.0) )
- l1_norm_list. append( float( data[ 1 ] ) )
- l2_norm_list. append( float( data[ 2 ] ) )
- max_norm_list. append( float( data[ 3 ] ) )
- items = items + 1
- if not verbose == 0:
- print line
- file. close()
-
-h_width = 12
-err_width = 15
-file = open( output_file_name, "w" )
-if format == "latex":
- file. write( "\\begin{tabular}{|r|l|l|l|l|l|l|}\\hline\n" )
- file. write( "\\raisebox{-1ex}[0ex]{$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/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h b/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h
deleted file mode 100644
index 19cdd94935..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver.h
+++ /dev/null
@@ -1,366 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <functions/tnlMeshFunction.h>
-#include <TNL/Devices/Host.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 <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 TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorType;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorType;
- typedef tnlGrid< 2, double, TNL::Devices::Host, int > MeshType;
-#ifdef HAVE_CUDA
- typedef TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorTypeCUDA;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorTypeCUDA;
- typedef tnlGrid< 2, double, TNL::Devices::Host, 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, TNL::Devices::Host, 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 TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorType;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorType;
- typedef tnlGrid< 3, double, TNL::Devices::Host, int > MeshType;
- #ifdef HAVE_CUDA
- typedef TNL::Containers::Vector< double, TNL::Devices::Host, int > VectorTypeCUDA;
- typedef TNL::Containers::Vector< int, TNL::Devices::Host, int > IntVectorTypeCUDA;
- typedef tnlGrid< 3, double, TNL::Devices::Host, 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, TNL::Devices::Host, 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/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h b/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h
deleted file mode 100644
index 76cf49bc8a..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver2D_impl.h
+++ /dev/null
@@ -1,1928 +0,0 @@
-/***************************************************************************
- 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" <<std::endl;
- this->device = tnlCudaDevice; /////////////// tnlCuda Device --- vypocet na GPU, TNL::Devices::HostDevice --- vypocet na CPU
-
-#ifdef HAVE_CUDA
- if(this->device == tnlCudaDevice)
- {
- run_host = 1;
- }
-#endif
-
- cout << "b" <<std::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..." <<std::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 <<std::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) <<std::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 <<std::endl;
-
- this->tau0 = parameters.getParameter <double>("initial-tau");
- cout << "Setting initial tau to " << this->tau0 <<std::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 <<std::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 <<std::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 <<std::endl;
-
- cout << "Initializating scheme..." <<std::endl;
- if(!this->schemeHost.init(parameters))
- {
- cerr << "SchemeHost failed to initialize." <<std::endl;
- return false;
- }
- cout << "Scheme initialized." <<std::endl;
-
- test();
-
- VectorType* tmp = new VectorType[subgridValues.getSize()];
- bool containsCurve = false;
-
-#ifdef HAVE_CUDA
-
- if(this->device == tnlCudaDevice)
- {
- /*cout << "Testing... " <<std::endl;
- if(this->device == tnlCudaDevice)
- {
- if( !initCUDA2D(parameters, gridRows, gridCols) )
- return false;
- }*/
- //cout << "s" <<std::endl;
- cudaMalloc(&(this->cudaSolver), sizeof(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >));
- //cout << "s" <<std::endl;
- cudaMemcpy(this->cudaSolver, this,sizeof(tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int >), cudaMemcpyHostToDevice);
- //cout << "s" <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- 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();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "s " <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "s "<<std::endl;
-
- }
-#endif
-
- if(this->device == TNL::Devices::HostDevice)
- {
- for(int i = 0; i < this->subgridValues.getSize(); i++)
- {
-
- if(! tmp[i].setSize(this->n * this->n))
- cout << "Could not allocate tmp["<< i <<"] array." <<std::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 << "." <<std::endl;
- tmp[i] = runSubgrid(0, tmp[i],i);
- insertSubgrid(tmp[i], i);
- setSubgridValue(i, 4);
- //cout << "Computed initial SDF on subgrid " << i << "." <<std::endl;
- }
- containsCurve = false;
-
- }
- }
-#ifdef HAVE_CUDA
- else if(this->device == tnlCudaDevice)
- {
-// cout << "pre 1 kernel" <<std::endl;
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- dim3 threadsPerBlock(this->n, this->n);
- dim3 numBlocks(this->gridCols,this->gridRows);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initRunCUDA2D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
- cudaDeviceSynchronize();
-// cout << "post 1 kernel" <<std::endl;
-
- }
-#endif
-
-
- this->currentStep = 1;
- if(this->device == TNL::Devices::HostDevice)
- 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] <<std::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] <<std::endl;
-
- TNL_CHECK_CUDA_DEVICE;
-
- synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cout << test[0] << " " <<test[1] <<" " << test[2] << " " <<test[3] <<std::endl;
- //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
- //TNL_CHECK_CUDA_DEVICE;
- //cout << this->tmpw << " " << test[0] << " " <<test[1] << " " <<test[2] <<" " << test[3] <<std::endl;
- //free(test);
-
- }
-
-#endif
- cout << "Solver initialized." <<std::endl;
-
- return true;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::run()
-{
- if(this->device == TNL::Devices::HostDevice)
- {
-
- 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) <<std::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) <<std::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) <<std::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) <<std::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) <<std::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" <<std::endl;
- bool end_cuda = false;
- dim3 threadsPerBlock(this->n, this->n);
- dim3 numBlocks(this->gridCols,this->gridRows);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cudaMalloc(&runcuda,sizeof(bool));
- //cudaMemcpy(runcuda, &run_host, sizeof(bool), cudaMemcpyHostToDevice);
- //cout << "fn" <<std::endl;
- bool* tmpb;
- //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
- //cudaDeviceSynchronize();
- //TNL_CHECK_CUDA_DEVICE;
- cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "fn" <<std::endl;
- int i = 1;
- time_diff = 0.0;
- while (run_host || !end_cuda)
- {
- cout << "Computing at step "<< i++ <<std::endl;
- if(run_host != 0 )
- end_cuda = true;
- else
- end_cuda = false;
- //cout << "a" <<std::endl;
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- start = std::clock();
- runCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,3*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
- //cout << "a" <<std::endl;
- cudaDeviceSynchronize();
- time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
-
- //start = std::clock();
- synchronizeCUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA2D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
-
-
- //cout << "a" <<std::endl;
- //run_host = false;
- //cout << "in kernel loop" << run_host <<std::endl;
- //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
- cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
- //cout << "in kernel loop" << run_host <<std::endl;
- }
- cout << "Solving time was: " << time_diff <<std::endl;
- //cout << "b" <<std::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] <<std::endl;
-
- //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
- //cout << test[0] << test[1] << test[2] << test[3] <<std::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] <<std::endl;
- //free(test);
-
- cudaDeviceSynchronize();
- }
-#endif
- contractGrid();
- this->u0.save("u-00001.tnl");
- cout << "Maximum number of calculations on one subgrid was " << this->calculationsCount.absMax() <<std::endl;
- cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) <<std::endl;
- cout << "Solver finished" <<std::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..." <<std::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 <<std::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 <<std::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 ) <<std::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..." <<std::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 << "." <<std::endl;
- cout << "Setting gridRows to " << this->gridRows << "." <<std::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." <<std::endl;
- if(!this->unusedCell.setSize(stretchedSize))
- cerr << "Could not allocate memory for supporting stretched grid." <<std::endl;
- int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
- cout << idealStretch <<std::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 <<std::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 <<std::endl;
- k+= (i/(this->n*this->gridCols) - idealStretch +1 )* this->mesh.getDimensions().x() ;
- }
-
- //cout << "i = " << i << " : i-k = " << i-k <<std::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." <<std::endl;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelEikonalSolver<2,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
-{
- cout << "Contracting grid..." <<std::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 <<std::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<<std::endl;
- this->u0[i-k] = this->work_u[i];
- }
-
- }
-
- cout << "Grid contracted" <<std::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" <<std::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" <<std::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" <<std::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" <<std::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" <<std::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" <<std::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" <<std::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" <<std::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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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();
- neighborEntities.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,neighborEntities);
- }
- maxResidue = fu. absMax();
-
-
- if( this -> cflCondition * maxResidue != 0.0)
- currentTau = this -> cflCondition / maxResidue;
-
- /* if (maxResidue < 0.05)
- std::cout << "Max < 0.05" <<std::endl;*/
- if(currentTau > 1.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >())
- {
- //cout << currentTau << " >= " << 2.0 * this->subMesh.template getSpaceStepsProducts< 1, 0 >() <<std::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] <<std::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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Entity.setCoordinates(Containers::StaticVector<2,int>(i,j));
- Entity.refresh();
- neighborEntities.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, neighborEntities);
-
- 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..." <<std::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 ) <<std::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..." <<std::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 <<std::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) <<std::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 <<std::endl;
-
- //this->tau0_cuda = parameters.getParameter <double>("initial-tau");
- //cout << "Setting initial tau to " << this->tau0_cuda <<std::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 <<std::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 <<std::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 <<std::endl;
-
- //cout << "Initializating scheme..." <<std::endl;
- //if(!this->schemeDevice.init(parameters))
-// {
- //cerr << "Scheme failed to initialize." <<std::endl;
-// return false;
-// }
- //cout << "Scheme initialized." <<std::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/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h b/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h
deleted file mode 100644
index dc3fd54679..0000000000
--- a/src/TNL/Legacy/hamilton-jacobi-parallel/tnlParallelEikonalSolver3D_impl.h
+++ /dev/null
@@ -1,1706 +0,0 @@
-/***************************************************************************
- 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" <<std::endl;
- this->device = TNL::Devices::HostDevice; /////////////// tnlCuda Device --- vypocet na GPU, TNL::Devices::HostDevice --- vypocet na CPU
-
-#ifdef HAVE_CUDA
- if(this->device == tnlCudaDevice)
- {
- run_host = 1;
- }
-#endif
-
- cout << "b" <<std::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..." <<std::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 <<std::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) <<std::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 <<std::endl;
-
- this->tau0 = parameters.getParameter <double>("initial-tau");
- cout << "Setting initial tau to " << this->tau0 <<std::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 <<std::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 <<std::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 <<std::endl;
-
- cout << "Initializating scheme..." <<std::endl;
- if(!this->schemeHost.init(parameters))
- {
- cerr << "SchemeHost failed to initialize." <<std::endl;
- return false;
- }
- cout << "Scheme initialized." <<std::endl;
-
- test();
-
- VectorType* tmp = new VectorType[subgridValues.getSize()];
-
-
-#ifdef HAVE_CUDA
-
- if(this->device == tnlCudaDevice)
- {
- /*cout << "Testing... " <<std::endl;
- if(this->device == tnlCudaDevice)
- {
- if( !initCUDA3D(parameters, gridRows, gridCols) )
- return false;
- }*/
- //cout << "s" <<std::endl;
- cudaMalloc(&(this->cudaSolver), sizeof(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >));
- //cout << "s" <<std::endl;
- cudaMemcpy(this->cudaSolver, this,sizeof(tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int >), cudaMemcpyHostToDevice);
- //cout << "s" <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- 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();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "s " <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "s "<<std::endl;
-
- }
-#endif
-
- if(this->device == TNL::Devices::HostDevice)
- {
-#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" <<std::endl;
-
- if(! tmp[i].setSize(this->n*this->n*this->n))
- cout << "Could not allocate tmp["<< i <<"] array." <<std::endl;
-// cout << "Working on subgrid " << i <<" --- check 2" <<std::endl;
-
- tmp[i] = getSubgrid(i);
- containsCurve = false;
-// cout << "Working on subgrid " << i <<" --- check 3" <<std::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] <<std::endl;
- }
-
- }
-// cout << "Working on subgrid " << i <<" --- check 4" <<std::endl;
-
- if(containsCurve)
- {
-// cout << "Computing initial SDF on subgrid " << i << "." <<std::endl;
- tmp[i] = runSubgrid(0, tmp[i] ,i);
- insertSubgrid( tmp[i], i);
- setSubgridValue(i, 4);
-// cout << "Computed initial SDF on subgrid " << i << "." <<std::endl;
- }
- containsCurve = false;
-
- }
-// cout << "CPU: Curve found" <<std::endl;
- }
-#ifdef HAVE_CUDA
- else if(this->device == tnlCudaDevice)
- {
-// cout << "pre 1 kernel" <<std::endl;
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- dim3 threadsPerBlock(this->n, this->n, this->n);
- dim3 numBlocks(this->gridCols,this->gridRows,this->gridLevels);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initRunCUDA3D<SchemeTypeHost,SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,2*this->n*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
- cudaDeviceSynchronize();
-// cout << "post 1 kernel" <<std::endl;
-
- }
-#endif
-
-
- this->currentStep = 1;
- if(this->device == TNL::Devices::HostDevice)
- 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] <<std::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] <<std::endl;
-
- TNL_CHECK_CUDA_DEVICE;
-
- synchronizeCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cout << cudaGetErrorString(cudaDeviceSynchronize()) <<std::endl;
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cout << test[0] << " " <<test[1] <<" " << test[2] << " " <<test[3] <<std::endl;
- //cudaMemcpy(/*this->work_u.getData()*/ test, (this->tmpw), this->work_u.getSize()*sizeof(double), cudaMemcpyDeviceToHost);
- //TNL_CHECK_CUDA_DEVICE;
- //cout << this->tmpw << " " << test[0] << " " <<test[1] << " " <<test[2] <<" " << test[3] <<std::endl;
- //free(test);
-
- }
-
-#endif
- cout << "Solver initialized." <<std::endl;
-
- return true;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::run()
-{
- if(this->device == TNL::Devices::HostDevice)
- {
-
- 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) <<std::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" <<std::endl;
- bool end_cuda = false;
- dim3 threadsPerBlock(this->n, this->n, this->n);
- dim3 numBlocks(this->gridCols,this->gridRows,this->gridLevels);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cudaMalloc(&runcuda,sizeof(bool));
- //cudaMemcpy(runcuda, &run_host, sizeof(bool), cudaMemcpyHostToDevice);
- //cout << "fn" <<std::endl;
- bool* tmpb;
- //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
- //cudaDeviceSynchronize();
- //TNL_CHECK_CUDA_DEVICE;
- cudaMemcpy(&(this->run_host),this->runcuda,sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //cout << "fn" <<std::endl;
- int i = 1;
- time_diff = 0.0;
- while (run_host || !end_cuda)
- {
- cout << "Computing at step "<< i++ <<std::endl;
- if(run_host != 0 )
- end_cuda = true;
- else
- end_cuda = false;
- //cout << "a" <<std::endl;
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- start = std::clock();
- runCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock,2*this->n*this->n*this->n*sizeof(double)>>>(this->cudaSolver);
- //cout << "a" <<std::endl;
- cudaDeviceSynchronize();
- time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
-
- //start = std::clock();
- synchronizeCUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- synchronize2CUDA3D<SchemeTypeHost, SchemeTypeDevice, DeviceType><<<numBlocks,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- //time_diff += (std::clock() - start) / (double)(CLOCKS_PER_SEC);
-
-
- //cout << "a" <<std::endl;
- //run_host = false;
- //cout << "in kernel loop" << run_host <<std::endl;
- //cudaMemcpy(tmpb, &(cudaSolver->runcuda),sizeof(bool*), cudaMemcpyDeviceToHost);
- cudaMemcpy(&run_host, (this->runcuda),sizeof(int), cudaMemcpyDeviceToHost);
- //cout << "in kernel loop" << run_host <<std::endl;
- }
- cout << "Solving time was: " << time_diff <<std::endl;
- //cout << "b" <<std::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] <<std::endl;
-
- //double * test = (double*)malloc(this->work_u.getSize()*sizeof(double));
- //cout << test[0] << test[1] << test[2] << test[3] <<std::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] <<std::endl;
- //free(test);
-
- cudaDeviceSynchronize();
- }
-#endif
- contractGrid();
- this->u0.save("u-00001.tnl");
- cout << "Maximum number of calculations on one subgrid was " << this->calculationsCount.absMax() <<std::endl;
- cout << "Average number of calculations on one subgrid was " << ( (double) this->calculationsCount.sum() / (double) this->calculationsCount.getSize() ) <<std::endl;
- cout << "Solver finished" <<std::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..." <<std::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" <<std::endl;
- tmp1 = this->gridCols*this->n*((this->n-1)+j*this->n) + i + k*this->gridCols*this->n*this->gridRows*this->n;
-// cout << "b" <<std::endl;
- tmp2 = this->gridCols*this->n*((this->n)+j*this->n) + i + k*this->gridCols*this->n*this->gridRows*this->n;
-// cout << "c" <<std::endl;
- if(tmp1 > work_u.getSize())
- cout << "tmp1: " << tmp1 << " x: " << j <<" y: " << i <<" z: " << k <<std::endl;
- if(tmp2 > work_u.getSize())
- cout << "tmp2: " << tmp2 << " x: " << j <<" y: " << i <<" z: " << k <<std::endl;
- grid1 = getSubgridValue(getOwner(tmp1));
-// cout << "d" <<std::endl;
- grid2 = getSubgridValue(getOwner(tmp2));
-// cout << "e" <<std::endl;
- if(getOwner(tmp1)==getOwner(tmp2))
- cout << "i, j, k" << i << "," << j << "," << k <<std::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" <<std::endl;
- this->unusedCell[tmp2] = 0;
-// cout << "g" <<std::endl;
- if(grid2 == INT_MAX)
- {
- setSubgridValue(getOwner(tmp2), -INT_MAX);
- }
-// cout << "h" <<std::endl;
- if(! (getBoundaryCondition(getOwner(tmp2)) & 8) )
- setBoundaryCondition(getOwner(tmp2), getBoundaryCondition(getOwner(tmp2))+8);
-// cout << "i" <<std::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" <<std::endl;
- this->unusedCell[tmp1] = 0;
-// cout << "k" <<std::endl;
- if(grid1 == INT_MAX)
- {
- setSubgridValue(getOwner(tmp1), -INT_MAX);
- }
-// cout << "l" <<std::endl;
- if(! (getBoundaryCondition(getOwner(tmp1)) & 1) )
- setBoundaryCondition(getOwner(tmp1), getBoundaryCondition(getOwner(tmp1))+1);
-// cout << "m" <<std::endl;
- }
- }
- }
- }
-
-// }
-// else
-// {
-
- cout << "sync 2" <<std::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 <<std::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" <<std::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 <<std::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 ) <<std::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..." <<std::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 << "." <<std::endl;
- cout << "Setting gridRows to " << this->gridRows << "." <<std::endl;
- cout << "Setting gridLevels to " << this->gridLevels << "." <<std::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." <<std::endl;
- if(!this->unusedCell.setSize(stretchedSize))
- cerr << "Could not allocate memory for supporting stretched grid." <<std::endl;
- int idealStretch =this->mesh.getDimensions().x() + (this->mesh.getDimensions().x()-2)/(this->n-1);
- cout << idealStretch <<std::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." <<std::endl;
-}
-
-template< typename SchemeHost, typename SchemeDevice, typename Device>
-void tnlParallelEikonalSolver<3,SchemeHost, SchemeDevice, Device, double, int>::contractGrid()
-{
- cout << "Contracting grid..." <<std::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 <<std::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" <<std::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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- while( time < finalTime /*|| maxResidue > subMesh.template getSpaceStepsProducts< 1, 0, 0 >()*/)
- {
- /****
- * Compute the RHS
- */
-
- for( int i = 0; i < fu.getSize(); i ++ )
- {
-// std::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 <<std::endl;
- Entity.refresh();
-// cout << "d" <<endl;
- neighborEntities.refresh(subMesh,Entity.getIndex());
-// cout << "e" <<endl;
- fu[ i ] = schemeHost.getValue( this->subMesh, i, coords,u, time, boundaryCondition, neighborEntities );
-// std::cout << "f" <<endl;
- }
- maxResidue = fu. absMax();
-
-
- if( this -> cflCondition * maxResidue != 0.0)
- currentTau = this -> cflCondition / maxResidue;
-
- /* if (maxResidue < 0.05)
- std::cout << "Max < 0.05" <<std::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] <<std::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);
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- Entity.setCoordinates(Containers::StaticVector<3,int>(i,j,k));
- Entity.refresh();
- neighborEntities.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, neighborEntities);
- 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/Legacy/narrow-band/CMakeLists.txt b/src/TNL/Legacy/narrow-band/CMakeLists.txt
deleted file mode 100644
index 158cd20132..0000000000
--- a/src/TNL/Legacy/narrow-band/CMakeLists.txt
+++ /dev/null
@@ -1,22 +0,0 @@
-set( tnl_narrow_band_SOURCES
-# MainBuildConfig.h
-# tnlNarrowBand2D_impl.h
-# tnlNarrowBand.h
-# narrowBandConfig.h
- main.cpp)
-
-
-IF( BUILD_CUDA )
- CUDA_ADD_EXECUTABLE(narrow-band main.cu)
-ELSE( BUILD_CUDA )
- ADD_EXECUTABLE(narrow-band main.cpp)
-ENDIF( BUILD_CUDA )
-target_link_libraries (narrow-band tnl )
-
-
-INSTALL( TARGETS narrow-band
- 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 ${TNL_TARGET_DATA_DIRECTORY}/examples/narrow-band )
diff --git a/src/TNL/Legacy/narrow-band/MainBuildConfig.h b/src/TNL/Legacy/narrow-band/MainBuildConfig.h
deleted file mode 100644
index ed3d686eb9..0000000000
--- a/src/TNL/Legacy/narrow-band/MainBuildConfig.h
+++ /dev/null
@@ -1,64 +0,0 @@
-/***************************************************************************
- 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() {std::cerr << "MainBuildConfig" <<std::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/Legacy/narrow-band/main.cpp b/src/TNL/Legacy/narrow-band/main.cpp
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/narrow-band/main.cpp
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/narrow-band/main.cu b/src/TNL/Legacy/narrow-band/main.cu
deleted file mode 100644
index 8849008ff6..0000000000
--- a/src/TNL/Legacy/narrow-band/main.cu
+++ /dev/null
@@ -1,17 +0,0 @@
-/***************************************************************************
- 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/Legacy/narrow-band/main.h b/src/TNL/Legacy/narrow-band/main.h
deleted file mode 100644
index 51dbdac37c..0000000000
--- a/src/TNL/Legacy/narrow-band/main.h
+++ /dev/null
@@ -1,88 +0,0 @@
-/***************************************************************************
- 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,TNL::Devices::Host, int>, double, int> solver;
- if(!solver.init(parameters))
- {
- cerr << "Solver failed to initialize." <<std::endl;
- return EXIT_FAILURE;
- }
- TNL_CHECK_CUDA_DEVICE;
- std::cout << "-------------------------------------------------------------" <<std::endl;
- std::cout << "Starting solver..." <<std::endl;
- solver.run();
- }
-// else if(dim == 3)
-// {
-// tnlNarrowBand<tnlGrid<3,double,TNL::Devices::Host, int>, double, int> solver;
-// if(!solver.init(parameters))
-// {
-// cerr << "Solver failed to initialize." <<std::endl;
-// return EXIT_FAILURE;
-// }
-// TNL_CHECK_CUDA_DEVICE;
-// std::cout << "-------------------------------------------------------------" <<std::endl;
-// std::cout << "Starting solver..." <<std::endl;
-// solver.run();
-// }
- else
- {
- std::cerr << "Unsupported number of dimensions: " << dim << "!" <<std::endl;
- return EXIT_FAILURE;
- }
-
-
- time(&stop);
- std::cout << "Solver stopped..." <<std::endl;
- std::cout <<std::endl;
- std::cout << "Running time was: " << difftime(stop,start) << " .... " << (std::clock() - start2) / (double)(CLOCKS_PER_SEC) <<std::endl;
- return EXIT_SUCCESS;
-}
-
-
diff --git a/src/TNL/Legacy/narrow-band/narrowBandConfig.h b/src/TNL/Legacy/narrow-band/narrowBandConfig.h
deleted file mode 100644
index bab58ceac4..0000000000
--- a/src/TNL/Legacy/narrow-band/narrowBandConfig.h
+++ /dev/null
@@ -1,40 +0,0 @@
-/***************************************************************************
- 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/src/TNL/Legacy/narrow-band/tnlNarrowBand.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand.h
deleted file mode 100644
index 7d3d19bc03..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand.h
+++ /dev/null
@@ -1,186 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <functions/tnlMeshFunction.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h
deleted file mode 100644
index dff0b48c8d..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v4_impl.h
+++ /dev/null
@@ -1,1317 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-#endif
-
- int n = Mesh.getDimensions().x();
-
- dim3 threadsPerBlock2(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
- dim3 numBlocks2(statusGridSize ,statusGridSize);
- initSetupGridCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initSetupGrid2CUDA<<<numBlocks2,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
-
- /*dim3 threadsPerBlock(16, 16);
- dim3 numBlocks(n/16 + 1 ,n/16 +1);*/
- initCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
-
- cout << "Solver initialized." <<std::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!" <<std::endl;
- runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- cout << "Hi2!" <<std::endl;
- while(time < finalTime)
- {
- if(tau+time > finalTime)
- tau=finalTime-time;
-
- runNarrowBandCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver,tau);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- time += tau;
-
-
- cudaMemcpy(&reinit, this->reinitialize, sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- if(reinit != 0 /*&& time != finalTime */)
- {
- cout << time <<std::endl;
-
- initSetupGridCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initSetupGrid2CUDA<<<numBlocksNB,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- }
- }
-
- //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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real value = cudaDofVector2[Entity.getIndex()];
- Real a,b, tmp;
-
- if( i == 0 /*|| (i/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 9))*/ )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 /*|| (i/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 17))*/ )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 /*|| (j/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 3))*/ )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 /* || (j/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 5)) */)
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1111(i,j);
-// else
-// setupSquare1110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1101(i,j);
-// else
-// setupSquare1100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1011(i,j);
-// else
-// setupSquare1010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1001(i,j);
-// else
-// setupSquare1000(i,j);
-// }
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0111(i,j);
-// else
-// setupSquare0110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0101(i,j);
-// else
-// setupSquare0100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0011(i,j);
-// else
-// setupSquare0010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(solver->Mesh);
- Entity.setCoordinates(Containers::StaticVector<2,double>(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< tnlGrid< 2,double, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()];
- xf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] - value;
- }
- else if( i == solver->Mesh.getDimensions().x() - 1 || (threadIdx.x == blockDim.x - 1 && !(status & 17)) )
- {
- xb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- xf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()] - value;
- }
- else
- {
- xb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- xf = solver-> cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] - value;
- }
-
- if( j == 0 || (threadIdx.y == 0 && !(status & 3)) )
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()] ;
- yf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()] - value;
- }
- else if( j == solver->Mesh.getDimensions().y() - 1 || (threadIdx.y == blockDim.y - 1 && !(status & 5)) )
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- yf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()] - value;
- }
- else
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- yf = solver-> cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-#endif
-
-
-
-
-#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h
deleted file mode 100644
index c928104900..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_CUDA_v5_impl.h
+++ /dev/null
@@ -1,1313 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
-#endif
-
- int n = Mesh.getDimensions().x();
-
- dim3 threadsPerBlock2(NARROWBAND_SUBGRID_SIZE, NARROWBAND_SUBGRID_SIZE);
- dim3 numBlocks2(statusGridSize ,statusGridSize);
- initSetupGridCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initSetupGrid2CUDA<<<numBlocks2,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
-
- /*dim3 threadsPerBlock(16, 16);
- dim3 numBlocks(n/16 + 1 ,n/16 +1);*/
- initCUDA<<<numBlocks2,threadsPerBlock2>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
-
- cout << "Solver initialized." <<std::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!" <<std::endl;
- runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- cout << "Hi2!" <<std::endl;
- while(time < finalTime)
- {
- if(tau+time > finalTime)
- tau=finalTime-time;
-
- runNarrowBandCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver,tau);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- time += tau;
-
-
- cudaMemcpy(&reinit, this->reinitialize, sizeof(int), cudaMemcpyDeviceToHost);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- if(reinit != 0 /*&& time != finalTime */)
- {
- cout << time <<std::endl;
-
- initSetupGridCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initSetupGrid2CUDA<<<numBlocksNB,1>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- initCUDA<<<numBlocksNB,threadsPerBlockNB>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- runCUDA<<<numBlocksFS,threadsPerBlockFS>>>(this->cudaSolver,0,0);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
- }
- }
-
- //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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real value = cudaDofVector2[Entity.getIndex()];
- Real a,b, tmp;
-
- if( i == 0 /*|| (i/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 9)) */)
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 /*|| (i/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 17)) */)
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0/* || (j/NARROWBAND_SUBGRID_SIZE == 0 && !(cudaStatusVector[subgridID] & 3)) */)
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 /* || (j/NARROWBAND_SUBGRID_SIZE == NARROWBAND_SUBGRID_SIZE - 1 && !(cudaStatusVector[subgridID] & 5))*/ )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
-
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1111(i,j);
-// else
-// setupSquare1110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1101(i,j);
-// else
-// setupSquare1100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1011(i,j);
-// else
-// setupSquare1010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare1001(i,j);
-// else
-// setupSquare1000(i,j);
-// }
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0111(i,j);
-// else
-// setupSquare0110(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0101(i,j);
-// else
-// setupSquare0100(i,j);
-// }
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
-// {
-// if(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
-// setupSquare0011(i,j);
-// else
-// setupSquare0010(i,j);
-// }
-// else
-// {
-// if(cudaDofVector[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(solver->Mesh);
- Entity.setCoordinates(Containers::StaticVector<2,double>(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< tnlGrid< 2,double, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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[neighborEntities.template getEntityIndex< 1, 0 >()];
- xf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] - value;
- }
- else if( i == solver->Mesh.getDimensions().x() - 1 /*|| (threadIdx.x == blockDim.x - 1 && !(status & 17)) */)
- {
- xb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- xf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()] - value;
- }
- else
- {
- xb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- xf = solver-> cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] - value;
- }
-
- if( j == 0/* || (threadIdx.y == 0 && !(status & 3))*/ )
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()] ;
- yf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()] - value;
- }
- else if( j == solver->Mesh.getDimensions().y() - 1 /*|| (threadIdx.y == blockDim.y - 1 && !(status & 5)) */)
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- yf = solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()] - value;
- }
- else
- {
- yb = value - solver->cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- yf = solver-> cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[Entity.getIndex()]-
- cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- Real al,be, a,b,c,s;
- al=abs(cudaDofVector[Entity.getIndex()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- cudaDofVector[Entity.getIndex()]));
-
- be=abs(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- cudaDofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,cudaDofVector2[neighborEntities.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, TNL::Devices::Host, int >, 2, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
- cudaDofVector2[Entity.getIndex()]=fabsMin(cudaDofVector[Entity.getIndex()],cudaDofVector2[Entity.getIndex()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 0, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 1 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(cudaDofVector[neighborEntities.template getEntityIndex< 1, 0 >()],cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-#endif
-
-
-
-
-#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_impl.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_impl.h
deleted file mode 100644
index d42bc2a761..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand2D_impl.h
+++ /dev/null
@@ -1,927 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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" <<std::endl;
- return initGrid();
-}
-
-
-template< typename MeshReal,
- typename Device,
- typename MeshIndex,
- typename Real,
- typename Index >
-bool tnlNarrowBand< tnlGrid< 2,MeshReal, Device, MeshIndex >, Real, Index > :: initGrid()
-{
-
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(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();
- neighborEntities.refresh(Mesh,Entity.getIndex());
-
- if(dofVector[this->Entity.getIndex()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1111(i,j);
- else
- setupSquare1110(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1101(i,j);
- else
- setupSquare1100(i,j);
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1011(i,j);
- else
- setupSquare1010(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare1001(i,j);
- else
- setupSquare1000(i,j);
- }
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0111(i,j);
- else
- setupSquare0110(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0101(i,j);
- else
- setupSquare0100(i,j);
- }
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()] > 0)
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0011(i,j);
- else
- setupSquare0010(i,j);
- }
- else
- {
- if(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()] > 0)
- setupSquare0001(i,j);
- else
- setupSquare0000(i,j);
- }
- }
- }
-
- }
- }
- cout << "a" <<std::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() <<std::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() <<std::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();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 2, tnlGridEntityNoStencilStorage >,2> neighborEntities(Entity);
-
- Real value = dofVector2[Entity.getIndex()];
- Real a,b, tmp;
-
- if( i == 0 )
- a = dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()];
- else
- {
- a = fabsMin( dofVector2[neighborEntities.template getEntityIndex< -1, 0 >()],
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()] );
- }
-
- if( j == 0 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()];
- else
- {
- b = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, -1 >()],
- dofVector2[neighborEntities.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()] <<std::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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(INT_MAX,dofVector2[Entity.getIndex()]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
-// dofVector2[Entity.getIndex()]=fabsMin(-INT_MAX,dofVector2[(Entity.getIndex())]);
-// dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
-// dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-INT_MAX,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[Entity.getIndex()]-
- dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(-abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- Real al,be, a,b,c,s;
- al=abs(dofVector[Entity.getIndex()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 0 >()]-
- dofVector[Entity.getIndex()]));
-
- be=abs(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()]/
- (dofVector[neighborEntities.template getEntityIndex< 1, 1 >()]-
- dofVector[neighborEntities.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[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(-abs(a*1+b*0+c)*s,dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(abs(a*1+b*1+c)*s,dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(abs(a*0+b*1+c)*s,dofVector2[neighborEntities.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 neighborEntities = Entity.getNeighborEntities();
- dofVector2[Entity.getIndex()]=fabsMin(dofVector[Entity.getIndex()],dofVector2[(Entity.getIndex())]);
- dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 0, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 0, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 1 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 1 >()]);
- dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]=fabsMin(dofVector[neighborEntities.template getEntityIndex< 1, 0 >()],dofVector2[neighborEntities.template getEntityIndex< 1, 0 >()]);
-}
-
-
-
-
-#endif /* TNLNARROWBAND_IMPL_H_ */
diff --git a/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_CUDA_impl.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_CUDA_impl.h
deleted file mode 100644
index d362f249a7..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_CUDA_impl.h
+++ /dev/null
@@ -1,961 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- TNL_CHECK_CUDA_DEVICE;
- initCUDA<<<numBlocks,threadsPerBlock>>>(this->cudaSolver);
- cudaDeviceSynchronize();
- TNL_CHECK_CUDA_DEVICE;
-
- 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();
- TNL_CHECK_CUDA_DEVICE;
-
- 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, TNL::Devices::Host, int >, 3, tnlGridEntityNoStencilStorage > Entity(Mesh);
- Entity.setCoordinates(CoordinatesType(i,j,k));
- Entity.refresh();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- Real value = cudaDofVector2[Entity.getIndex()];
- Real a,b,c, tmp;
-
- if( i == 0 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0, 0 >()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()];
- else
- {
- a = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 1, 0, 0 >()] );
- }
-
- if( j == 0 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1, 0 >()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1, 0 >()];
- else
- {
- b = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, -1, 0 >()],
- cudaDofVector2[neighborEntities.template getEntityIndex< 0, 1, 0 >()] );
- }
-
- if( k == 0 )
- c = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, 1 >()];
- else if( k == Mesh.getDimensions().z() - 1 )
- c = cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, -1 >()];
- else
- {
- c = fabsMin( cudaDofVector2[neighborEntities.template getEntityIndex< 0, 0, -1 >()],
- cudaDofVector2[neighborEntities.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, TNL::Devices::Host, 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, TNL::Devices::Host, 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, TNL::Devices::Host, 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/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_impl.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_impl.h
deleted file mode 100644
index 6e63d527b9..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand3D_impl.h
+++ /dev/null
@@ -1,307 +0,0 @@
-/***************************************************************************
- 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 ) )
- {
- std::cerr << "I am not able to load the mesh from the file " << meshFile << "." <<std::endl;
- return false;
- }
-
-
- const String& initialCondition = parameters.getParameter <String>("initial-condition");
- if( ! dofVector.load( initialCondition ) )
- {
- std::cerr << "I am not able to load the initial condition from the file " << meshFile << "." <<std::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();
- tnlNeighborGridEntityGetter<tnlGridEntity< MeshType, 3, tnlGridEntityNoStencilStorage >,3> neighborEntities(Entity);
- Real value = dofVector2[Entity.getIndex()];
- Real a,b,c, tmp;
-
- if( i == 0 )
- a = dofVector2[neighborEntities.template getEntityIndex< 1, 0, 0>()];
- else if( i == Mesh.getDimensions().x() - 1 )
- a = dofVector2[neighborEntities.template getEntityIndex< -1, 0, 0 >()];
- else
- {
- a = fabsMin( dofVector2[neighborEntities.template getEntityIndex< -1, 0, 0>()],
- dofVector2[neighborEntities.template getEntityIndex< 1, 0, 0>()] );
- }
-
- if( j == 0 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, 1, 0>()];
- else if( j == Mesh.getDimensions().y() - 1 )
- b = dofVector2[neighborEntities.template getEntityIndex< 0, -1, 0>()];
- else
- {
- b = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, -1, 0>()],
- dofVector2[neighborEntities.template getEntityIndex< 0, 1, 0>()] );
- }
-
- if( k == 0 )
- c = dofVector2[neighborEntities.template getEntityIndex< 0, 0, 1>()];
- else if( k == Mesh.getDimensions().z() - 1 )
- c = dofVector2[neighborEntities.template getEntityIndex< 0, 0, -1>()];
- else
- {
- c = fabsMin( dofVector2[neighborEntities.template getEntityIndex< 0, 0, -1>()],
- dofVector2[neighborEntities.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/src/TNL/Legacy/narrow-band/tnlNarrowBand_CUDA.h b/src/TNL/Legacy/narrow-band/tnlNarrowBand_CUDA.h
deleted file mode 100644
index ca9b1da2cc..0000000000
--- a/src/TNL/Legacy/narrow-band/tnlNarrowBand_CUDA.h
+++ /dev/null
@@ -1,203 +0,0 @@
-/***************************************************************************
- 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 <TNL/Containers/Vector.h>
-#include <TNL/Containers/StaticVector.h>
-#include <TNL/Devices/Host.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 TNL::Containers::Vector< 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 TNL::Containers::Vector< 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, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i);
-//__global__ void runCUDA(tnlNarrowBand< tnlGrid< 3,double, TNL::Devices::Host, int >, double, int >* solver, int sweep, int i);
-
-__global__ void initCUDA(tnlNarrowBand< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-
-__global__ void initSetupGridCUDA(tnlNarrowBand< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-__global__ void initSetupGrid2CUDA(tnlNarrowBand< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-__global__ void initSetupGrid1_2CUDA(tnlNarrowBand< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver);
-__global__ void runNarrowBandCUDA(tnlNarrowBand< tnlGrid< 2,double, TNL::Devices::Host, int >, double, int >* solver, double tau);
-//__global__ void initCUDA(tnlNarrowBand< tnlGrid< 3,double, TNL::Devices::Host, int >, double, int >* solver);
-#endif
-
-
-
-#include "tnlNarrowBand2D_CUDA_v4_impl.h"
-// #include "tnlNarrowBand3D_CUDA_impl.h"
-
-#endif /* TNLNARROWBAND_H_ */
--
GitLab