Refactoring of the vector operations. (5f7eb0f3) · Commits · TNL / tnl-dev

CMakeLists.txt

+1 −1

Original line number	Diff line number	Diff line
		@@ -33,7 +33,7 @@ if( WITH_TEMPLATE_EXPLICIT_INSTANTIATION STREQUAL "yes" )
		endif()

		if( WITH_CUDA STREQUAL "yes" )
		AddCompilerFlag( "-DHAVE_NOT_CXX11" )
		AddCompilerFlag( "-DHAVE_NOT_CXX11 -U_GLIBCXX_ATOMIC_BUILTINS -U_GLIBCXX_USE_INT128" )
		else()
		AddCompilerFlag( "-std=gnu++0x" )
		endif()

install

+1 −1

Original line number	Diff line number	Diff line
		@@ -2,7 +2,7 @@

		TARGET=TNL
		INSTALL_PREFIX=${HOME}/local
		WITH_CUDA=yes
		WITH_CUDA=no
		WITH_CUSPARSE=no
		CUDA_ARCHITECTURE=2.0
		TEMPLATE_EXPLICIT_INSTANTIATION=yes

src/core/vectors/CMakeLists.txt

+2 −1

Original line number	Diff line number	Diff line
		set( headers tnlVector.h
		tnlMultiVector.h
		tnlSharedVector.h )
		tnlSharedVector.h
		tnlVectorOperations.h )

		INSTALL( FILES ${headers} DESTINATION include/tnl-${tnlVersion}/core/vectors )
		No newline at end of file

src/core/vectors/tnlVectorOperations.h

0 → 100644

+207 −0

Original line number	Diff line number	Diff line
		/***************************************************************************
		tnlVectorOperations.h - description
		-------------------
		begin : Nov 8, 2012
		copyright : (C) 2012 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 TNLVECTOROPERATIONS_H_
		#define TNLVECTOROPERATIONS_H_

		#include <core/cuda/device-check.h>
		#include <core/cuda/cuda-reduction.h>
		#include <core/cuda/reduction-operations.h>
		#include <core/tnlHost.h>
		#include <core/tnlCuda.h>

		template< typename Device >
		class tnlVectorOperations{};

		template<>
		class tnlVectorOperations< tnlHost >
		{
		public:

		template< typename Vector >
		static typename Vector :: RealType getVectorMax( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorMin( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorAbsMax( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorAbsMin( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorLpNorm( const Vector& v,
		const typename Vector :: RealType& p );

		template< typename Vector >
		static typename Vector :: RealType getVectorSum( const Vector& v );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceMax( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceMin( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceAbsMax( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceAbsMin( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceLpNorm( const Vector1& v1,
		const Vector2& v2,
		const typename Vector1 :: RealType& p );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceSum( const Vector1& v1,
		const Vector2& v2 );
		template< typename Vector >
		static void vectorScalarMultiplication( Vector& v,
		const typename Vector :: RealType& alpha );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorSdot( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha );
		template< typename Vector1, typename Vector2 >
		static void vectorSaxmy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsby( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const typename Vector1 :: RealType& beta );


		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsbz( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const Vector2& z,
		const typename Vector1 :: RealType& beta );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsbzpy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const Vector2& z,
		const typename Vector1 :: RealType& beta );
		};

		template<>
		class tnlVectorOperations< tnlCuda >
		{
		public:

		template< typename Vector >
		static typename Vector :: RealType getVectorMax( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorMin( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorAbsMax( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorAbsMin( const Vector& v );

		template< typename Vector >
		static typename Vector :: RealType getVectorLpNorm( const Vector& v,
		const typename Vector :: RealType& p );

		template< typename Vector >
		static typename Vector :: RealType getVectorSum( const Vector& v );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceMax( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceMin( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceAbsMax( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceAbsMin( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceLpNorm( const Vector1& v1,
		const Vector2& v2,
		const typename Vector1 :: RealType& p );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorDifferenceSum( const Vector1& v1,
		const Vector2& v2 );
		template< typename Vector >
		static void vectorScalarMultiplication( Vector& v,
		const typename Vector :: RealType& alpha );

		template< typename Vector1, typename Vector2 >
		static typename Vector1 :: RealType getVectorSdot( const Vector1& v1,
		const Vector2& v2 );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha );
		template< typename Vector1, typename Vector2 >
		static void vectorSaxmy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsby( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const typename Vector1 :: RealType& beta );


		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsbz( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const Vector2& z,
		const typename Vector1 :: RealType& beta );

		template< typename Vector1, typename Vector2 >
		static void vectorSaxpsbzpy( Vector1& y,
		const Vector2& x,
		const typename Vector1 :: RealType& alpha,
		const Vector2& z,
		const typename Vector1 :: RealType& beta );
		};

		#include <implementation/core/vectors/tnlVectorOperationsHost_impl.h>
		#include <implementation/core/vectors/tnlVectorOperationsCuda_impl.h>

		#endif /* TNLVECTOROPERATIONS_H_ */

src/implementation/core/CMakeLists.txt

+3 −3

Original line number	Diff line number	Diff line
		@@ -19,7 +19,8 @@ set( common_SOURCES
		${CURRENT_DIR}/tnlTimerCPU.cpp
		${CURRENT_DIR}/mfilename.cpp
		${CURRENT_DIR}/mpi-supp.cpp
		${CURRENT_DIR}/tnlHost_impl.cpp )
		${CURRENT_DIR}/tnlHost_impl.cpp
		${CURRENT_DIR}/tnlCuda.cpp )

		IF( BUILD_CUDA )
		set( tnl_implementation_core_CUDA__SOURCES
		@@ -28,7 +29,6 @@ IF( BUILD_CUDA )
		${tnl_implementation_core_vectors_CUDA__SOURCES}
		${common_SOURCES}
		${CURRENT_DIR}/memory-operations_impl.cu
		${CURRENT_DIR}/tnlCuda.cu
		PARENT_SCOPE )
		ENDIF()