Skip to content
Snippets Groups Projects
Commit db3d7c20 authored by Vít Hanousek's avatar Vít Hanousek
Browse files

Merge branch 'develop' into MIC-devel

parents 5cbb4a1b b797dcb5
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 125 additions and 106 deletions
examples/heat-equation/tnl-heat-equation.h
+ 0
− 3
View file @ db3d7c20
......@@ -64,14 +64,11 @@ class heatEquationSetter
typedef Device DeviceType;
typedef Index IndexType;
typedef Containers::StaticVector< MeshType::meshDimensions, Real > Vertex;
static bool run( const Config::ParameterContainer& parameters )
{
enum { Dimensions = MeshType::meshDimensions };
typedef Operators::LinearDiffusion< MeshType, Real, Index > ApproximateOperator;
typedef Functions::Analytic::Constant< Dimensions, Real > RightHandSide;
typedef Containers::StaticVector < MeshType::meshDimensions, Real > Vertex;
String boundaryConditionsType = parameters.getParameter< String >( "boundary-conditions-type" );
if( parameters.checkParameter( "boundary-conditions-constant" ) )
......
examples/heat-equation/tnl-run-heat-equation-eoc-test
+ 82
− 70
View file @ db3d7c20
#!/bin/bash
device="host"
threadsNumbers="1 2 4 6"
dimensions="1D 2D 3D"
#dimensions="1D"
sizes1D="16 32 64 128 256 512"
......@@ -84,6 +85,7 @@ solve()
{
timeDiscretisation=$1
discreteSolver=$2
threadsNumber=$3
${solverName} --device ${device} \
--mesh mesh.tnl \
--initial-condition exact-u-00000.tnl \
......@@ -113,7 +115,10 @@ solve()
--sigma ${sigma} \
--time-dependence ${timeDependence} \
--snapshot-period ${snapshotPeriod} \
--final-time ${finalTime}
--final-time ${finalTime} \
--refresh-rate 50 \
--openmp-enabled true \
--openmp-max-threads ${threadsNumber}
}
computeError()
......@@ -128,77 +133,84 @@ computeError()
runTest()
{
for testFunction in ${testFunctions};
for threadsNumber in ${threadsNumbers};
do
mkdir -p ${testFunction}
cd ${testFunction}
setupTestFunction ${testFunction}
for dim in ${dimensions};
do
mkdir -p $dim
cd ${dim}
if test $dim = 1D;
then
sizes=$sizes1D
fi
if test $dim = 2D;
then
sizes=$sizes2D
fi
if test $dim = 3D;
then
sizes=$sizes3D
fi
lastSize=""
for size in $sizes;
do
mkdir -p $size
cd $size
echo ""
echo ""
echo ""
if test ! -f computation-done;
then
touch computation-in-progress
echo "========================================================================="
echo "=== SETTING UP THE GRID ==="
echo "========================================================================="
setupGrid $dim $size
echo "========================================================================="
echo "=== WRITING THE EXACT SOLUTION ==="
echo "========================================================================="
setInitialCondition $testFunction
echo "========================================================================="
echo "=== STARTING THE SOLVER ==="
echo "========================================================================="
#solve explicit merson
solve semi-implicit gmres
mv computation-in-progress computation-done
fi
echo "========================================================================="
echo "=== COMPUTING THE ERROR ==="
echo "========================================================================="
computeError
echo "========================================================================="
echo "=== COMPUTING THE EOC ==="
echo "========================================================================="
if test ! x$lastSize = x;
then
tnl-err2eoc ../$lastSize/errors.txt errors.txt
fi
echo "========================================================================="
echo "=== COMPUTATION DONE ==="
echo "========================================================================="
cd ..
lastSize=$size
done
mkdir -p threads-${threadsNumber}
cd threads-${threadsNumber}
cd ..
done
cd ..
done
for testFunction in ${testFunctions};
do
mkdir -p ${testFunction}
cd ${testFunction}
setupTestFunction ${testFunction}
for dim in ${dimensions};
do
mkdir -p $dim
cd ${dim}
if test $dim = 1D;
then
sizes=$sizes1D
fi
if test $dim = 2D;
then
sizes=$sizes2D
fi
if test $dim = 3D;
then
sizes=$sizes3D
fi
lastSize=""
for size in $sizes;
do
mkdir -p $size
cd $size
echo ""
echo ""
echo ""
if test ! -f computation-done;
then
touch computation-in-progress
echo "========================================================================="
echo "=== SETTING UP THE GRID ==="
echo "========================================================================="
setupGrid $dim $size
echo "========================================================================="
echo "=== WRITING THE EXACT SOLUTION ==="
echo "========================================================================="
setInitialCondition $testFunction
echo "========================================================================="
echo "=== STARTING THE SOLVER ==="
echo "========================================================================="
#solve explicit merson ${threadsNumber}
solve semi-implicit gmres ${threadsNumber}
mv computation-in-progress computation-done
fi
echo "========================================================================="
echo "=== COMPUTING THE ERROR ==="
echo "========================================================================="
computeError
echo "========================================================================="
echo "=== COMPUTING THE EOC ==="
echo "========================================================================="
if test ! x$lastSize = x;
then
tnl-err2eoc ../$lastSize/errors.txt errors.txt
fi
echo "========================================================================="
echo "=== COMPUTATION DONE ==="
echo "========================================================================="
cd ..
lastSize=$size
done
cd ..
done
cd ..
done
cd ..
done
}
runTest
......
src/TNL/Assert.h
+ 5
− 3
View file @ db3d7c20
......@@ -24,7 +24,8 @@
#ifndef NDEBUG
#ifdef HAVE_CUDA
// __CUDA_ARCH__ is defined by the compiler only for code executed on GPU
#ifdef __CUDA_ARCH__
#define TNL_ASSERT( ___tnl__assert_condition, ___tnl__assert_command ) \
if( ! ( ___tnl__assert_condition ) ) \
{ \
......@@ -35,7 +36,7 @@
\
}
#else // HAVE_CUDA
#else // __CUDA_ARCH__
#define TNL_ASSERT( ___tnl__assert_condition, ___tnl__assert_command ) \
if( ! ( ___tnl__assert_condition ) ) \
{ \
......@@ -47,7 +48,8 @@
/*___tnl__assert_command;*/ \
throw EXIT_FAILURE; \
}
#endif /* HAVE_CUDA */
#endif // __CUDA_ARCH__
#else /* #ifndef NDEBUG */
#define TNL_ASSERT( ___tnl__assert_condition, ___tnl__assert_command )
#endif /* #ifndef NDEBUG */
\ No newline at end of file
src/TNL/Containers/Algorithms/reduction-operations.h
+ 12
− 12
View file @ db3d7c20
......@@ -76,7 +76,7 @@ class tnlParallelReductionMin
const RealType* data1,
const RealType* data2 )
{
return min( current, data1[ idx ] );
return TNL::min( current, data1[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return MaxValue< ResultType>(); };
......@@ -117,7 +117,7 @@ class tnlParallelReductionMax
const RealType* data1,
const RealType* data2 )
{
return max( current, data1[ idx ] );
return TNL::max( current, data1[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return MinValue< ResultType>(); };
......@@ -241,7 +241,7 @@ class tnlParallelReductionAbsSum : public tnlParallelReductionSum< Real, Index >
const RealType* data1,
const RealType* data2 )
{
return current + ::abs( data1[ idx ] );
return current + TNL::abs( data1[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -270,7 +270,7 @@ class tnlParallelReductionAbsMin : public tnlParallelReductionMin< Real, Index >
const RealType* data1,
const RealType* data2 )
{
return min( current, abs( data1[ idx ] ) );
return TNL::min( current, TNL::abs( data1[ idx ] ) );
};
__cuda_callable__ ResultType initialValue() { return MaxValue< ResultType>(); };
......@@ -299,7 +299,7 @@ class tnlParallelReductionAbsMax : public tnlParallelReductionMax< Real, Index >
const RealType* data1,
const RealType* data2 )
{
return std::max( current, ::abs( data1[ idx ] ) );
return std::max( current, TNL::abs( data1[ idx ] ) );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -365,7 +365,7 @@ class tnlParallelReductionLpNorm : public tnlParallelReductionSum< Real, Index >
const RealType* data1,
const RealType* data2 )
{
return current + ::pow( ::abs( data1[ idx ] ), p );
return current + TNL::pow( TNL::abs( data1[ idx ] ), p );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -514,7 +514,7 @@ class tnlParallelReductionDiffMin : public tnlParallelReductionMin< Real, Index
const RealType* data1,
const RealType* data2 )
{
return min( current, data1[ idx ] - data2[ idx ] );
return TNL::min( current, data1[ idx ] - data2[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return MaxValue< ResultType>(); };
......@@ -543,7 +543,7 @@ class tnlParallelReductionDiffMax : public tnlParallelReductionMax< Real, Index
const RealType* data1,
const RealType* data2 )
{
return max( current, data1[ idx ] - data2[ idx ] );
return TNL::max( current, data1[ idx ] - data2[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -572,7 +572,7 @@ class tnlParallelReductionDiffAbsSum : public tnlParallelReductionMax< Real, Ind
const RealType* data1,
const RealType* data2 )
{
return current + abs( data1[ idx ] - data2[ idx ] );
return current + TNL::abs( data1[ idx ] - data2[ idx ] );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -601,7 +601,7 @@ class tnlParallelReductionDiffAbsMin : public tnlParallelReductionMin< Real, Ind
const RealType* data1,
const RealType* data2 )
{
return min( current, abs( data1[ idx ] - data2[ idx ] ) );
return TNL::min( current, TNL::abs( data1[ idx ] - data2[ idx ] ) );
};
__cuda_callable__ ResultType initialValue() { return MaxValue< ResultType>(); };
......@@ -630,7 +630,7 @@ class tnlParallelReductionDiffAbsMax : public tnlParallelReductionMax< Real, Ind
const RealType* data1,
const RealType* data2 )
{
return max( current, abs( data1[ idx ] - data2[ idx ] ) );
return TNL::max( current, TNL::abs( data1[ idx ] - data2[ idx ] ) );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......@@ -699,7 +699,7 @@ class tnlParallelReductionDiffLpNorm : public tnlParallelReductionSum< Real, Ind
const RealType* data1,
const RealType* data2 )
{
return current + ::pow( abs( data1[ idx ] - data2[ idx ] ), p );
return current + TNL::pow( TNL::abs( data1[ idx ] - data2[ idx ] ), p );
};
__cuda_callable__ ResultType initialValue() { return ( ResultType ) 0; };
......
src/TNL/Containers/Array_impl.h
+ 19
− 15
View file @ db3d7c20
......@@ -175,7 +175,7 @@ setSize( const Index size )
Algorithms::ArrayOperations< Device >::allocateMemory( this->allocationPointer, size );
this->data = this->allocationPointer;
this->size = size;
if( ! this->allocationPointer )
if( size > 0 && ! this->allocationPointer )
{
std::cerr << "I am not able to allocate new array with size "
<< ( double ) this->size * sizeof( ElementType ) / 1.0e9 << " GB." << std::endl;
......@@ -370,13 +370,14 @@ operator = ( const Array< Element, Device, Index >& array )
TNL_ASSERT( array. getSize() == this->getSize(),
std::cerr << "Source size: " << array. getSize() << std::endl
<< "Target size: " << this->getSize() << std::endl );
Algorithms::ArrayOperations< Device >::
template copyMemory< Element,
Element,
Index >
( this->getData(),
array. getData(),
array. getSize() );
if( this->getSize() > 0 )
Algorithms::ArrayOperations< Device >::
template copyMemory< Element,
Element,
Index >
( this->getData(),
array. getData(),
array. getSize() );
return ( *this );
};
......@@ -391,13 +392,14 @@ operator = ( const ArrayT& array )
TNL_ASSERT( array. getSize() == this->getSize(),
std::cerr << "Source size: " << array. getSize() << std::endl
<< "Target size: " << this->getSize() << std::endl );
Algorithms::ArrayOperations< Device, typename ArrayT::DeviceType >::
template copyMemory< Element,
typename ArrayT::ElementType,
typename ArrayT::IndexType >
( this->getData(),
array. getData(),
array. getSize() );
if( this->getSize() > 0 )
Algorithms::ArrayOperations< Device, typename ArrayT::DeviceType >::
template copyMemory< Element,
typename ArrayT::ElementType,
typename ArrayT::IndexType >
( this->getData(),
array. getData(),
array. getSize() );
return ( *this );
};
......@@ -411,6 +413,8 @@ operator == ( const ArrayT& array ) const
{
if( array. getSize() != this->getSize() )
return false;
if( this->getSize() == 0 )
return true;
return Algorithms::ArrayOperations< Device, typename ArrayT::DeviceType >::
template compareMemory< typename ArrayT::ElementType,
Element,
......
src/TNL/Containers/SharedArray_impl.h
+ 2
− 0
View file @ db3d7c20
......@@ -310,6 +310,7 @@ void SharedArray< Element, Device, Index > :: setValue( const Element& e )
template< typename Element,
typename Device,
typename Index >
__cuda_callable__
const Element* SharedArray< Element, Device, Index > :: getData() const
{
return this->data;
......@@ -318,6 +319,7 @@ const Element* SharedArray< Element, Device, Index > :: getData() const
template< typename Element,
typename Device,
typename Index >
__cuda_callable__
Element* SharedArray< Element, Device, Index > :: getData()
{
return this->data;
......
src/TNL/Containers/StaticArray3D_impl.h
+ 1
− 0
View file @ db3d7c20
......@@ -68,6 +68,7 @@ String StaticArray< 3, Element >::getType()
}
template< typename Element >
__cuda_callable__
inline int StaticArray< 3, Element >::getSize() const
{
return size;
......
src/TNL/Containers/StaticVector_impl.h
+ 1
− 0
View file @ db3d7c20
......@@ -182,6 +182,7 @@ StaticVector< Size, Real >::abs() const
template< int Size, typename Real, typename Scalar >
__cuda_callable__
StaticVector< Size, Real > operator * ( const Scalar& c, const StaticVector< Size, Real >& u )
{
return u * c;
......
src/TNL/Solvers/BuildConfigTags.h
+ 1
− 1
View file @ db3d7c20
......@@ -25,7 +25,7 @@
namespace TNL {
namespace Solvers {
class tnlDefaultBuildConfigTag{};
class DefaultBuildConfigTag{};
/****
* All devices are enabled by default. Those which are not available
......
src/TNL/Solvers/Solver.h
+ 1
− 1
View file @ db3d7c20
......@@ -17,7 +17,7 @@ namespace Solvers {
template< template< typename Real, typename Device, typename Index, typename MeshType, typename ConfigTag, typename SolverStarter > class ProblemSetter,
template< typename ConfTag > class ProblemConfig,
typename ConfigTag = tnlDefaultBuildConfigTag >
typename ConfigTag = DefaultBuildConfigTag >
class Solver
{
public:
......
src/TNL/Solvers/SolverConfig_impl.h
+ 1
− 1
View file @ db3d7c20
......@@ -122,7 +122,7 @@ bool SolverConfig< ConfigTag, ProblemConfig >::configSetup( Config::ConfigDescri
if( ConfigTagSemiImplicitSolver< ConfigTag, SemiImplicitSORSolverTag >::enabled )
config.addEntryEnum( "sor" );
#ifdef HAVE_UMFPACK
if( MeshConfigSemiImplicitSolver< MeshConfig, SemiImplicitUmfpackSolverTag >::enabled )
if( ConfigTagSemiImplicitSolver< ConfigTag, SemiImplicitUmfpackSolverTag >::enabled )
config.addEntryEnum( "umfpack" );
#endif
}
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment