Merge branch 'mpi-explosive' of jlk.fjfi.cvut.cz:mmg/tnl-dev into mpi-explosive (4ba7844e) · Commits · TNL / tnl-dev

build

+23 −6

Original line number	Diff line number	Diff line
		@@ -13,6 +13,7 @@ HELP="no"
		VERBOSE=""

		WITH_CLANG="no"
		WITH_MPI="yes"
		WITH_CUDA="yes"
		WITH_CUDA_ARCH="auto"
		WITH_OPENMP="yes"
		@@ -42,6 +43,7 @@ do
		--verbose ) VERBOSE="VERBOSE=1" ;;
		--help ) HELP="yes" ;;
		--with-clang=* ) WITH_CLANG="${option#*=}" ;;
		--with-mpi=* ) WITH_MPI="${option#*=}" ;;
		--with-mic=* ) WITH_MIC="${option#*=}" ;;
		--with-cuda=* ) WITH_CUDA="${option#*=}" ;;
		--with-cuda-arch=* ) WITH_CUDA_ARCH="${option#*=}";;
		@@ -76,12 +78,13 @@ then
		echo " --build-jobs=NUM Number of processes to be used for the build. It is set to the number of available CPU cores by default."
		echo " --prefix=PATH Prefix for the installation directory. ${HOME}/local by default."
		echo " --install=yes/no Enables the installation of TNL files."
		echo " --with-mic=yes/no Enable MIC (Intel Xeon Phi). 'no' by default (Intel Compiler required)."
		echo " --with-cuda=yes/no Enable CUDA. 'yes' by default (CUDA Toolkit is required)."
		echo " --with-cuda-arch=all/auto/30/35/... Choose CUDA architecture. 'auto' by default."
		echo " --with-openmp=yes/no Enable OpenMP. 'yes' by default."
		echo " --with-tests=yes/no Enable unit tests. 'yes' by default."
		echo " --with-coverage=yes/no Enable code coverage reports for unit tests. 'no' by default (lcov is required)."
		echo " --with-mpi=yes/no Enables MPI. 'no' by default (Intel Compiler required)."
		echo " --with-mic=yes/no Enables MIC (Intel Xeon Phi). 'no' by default (Intel Compiler required)."
		echo " --with-cuda=yes/no Enables CUDA. 'yes' by default (CUDA Toolkit is required)."
		echo " --with-cuda-arch=all/auto/30/35/... Chooses CUDA architecture. 'auto' by default."
		echo " --with-openmp=yes/no Enables OpenMP. 'yes' by default."
		echo " --with-tests=yes/no Enables unit tests. 'yes' by default."
		echo " --with-coverage=yes/no Enables code coverage reports for unit tests. 'no' by default (lcov is required)."
		echo " --with-examples=yes/no Compile the 'examples' directory. 'yes' by default."
		echo " --with-templates-instantiation=yes/no Precompiles some TNL templates during the build. 'no' by default."
		echo " --cmake=CMAKE Path to cmake. 'cmake' by default."
		@@ -98,6 +101,20 @@ then
		export CC=clang
		fi

		if test ${WITH_MPI} = "yes";
		then
		if ! [ -x "$(command -v mpic++)" ]; then
		echo "Warning:mpic++ is not installed on this system. MPI support is turned off."
		else
		export CXX=mpic++
		fi
		if ! [ -x "$(command -v mpicc)" ]; then
		echo "Warning: mpicc is not installed on this system."
		else
		export CC=mpicc
		fi
		fi


		echo "Configuring ${BUILD} $TARGET ..."

src/TNL/Communicators/NoDistrCommunicator.h

+3 −3

Original line number	Diff line number	Diff line
		@@ -72,13 +72,13 @@ namespace Communicators {
		static Request ISend( const T *data, int count, int dest)
		{
		return 1;
		};
		}

		template <typename T>
		static Request IRecv( const T *data, int count, int src)
		{
		return 1;
		};
		}

		static void WaitAll(Request *reqs, int length)
		{
		@@ -87,7 +87,7 @@ namespace Communicators {
		template< typename T >
		static void Bcast( T& data, int count, int root)
		{
		};
		}

		/* template< typename T >
		static void Allreduce( T& data,

src/TNL/Containers/Algorithms/ArrayOperations.h

+2 −0

Original line number	Diff line number	Diff line
		@@ -81,10 +81,12 @@ class ArrayOperations< Devices::Cuda >
		static void freeMemory( Element* data );

		template< typename Element >
		__cuda_callable__
		static void setMemoryElement( Element* data,
		const Element& value );

		template< typename Element >
		__cuda_callable__
		static Element getMemoryElement( const Element* data );

		// TODO: does not make sense for CUDA - remove?

src/TNL/Containers/Algorithms/ArrayOperationsCuda_impl.h

+10 −2

Original line number	Diff line number	Diff line
		@@ -60,24 +60,32 @@ freeMemory( Element* data )
		}

		template< typename Element >
		void
		__cuda_callable__ void
		ArrayOperations< Devices::Cuda >::
		setMemoryElement( Element* data,
		const Element& value )
		{
		TNL_ASSERT_TRUE( data, "Attempted to set data through a nullptr." );
		#ifdef __CUDAARCH__
		*data = value;
		#else
		ArrayOperations< Devices::Cuda >::setMemory( data, value, 1 );
		#endif
		}

		template< typename Element >
		Element
		__cuda_callable__ Element
		ArrayOperations< Devices::Cuda >::
		getMemoryElement( const Element* data )
		{
		TNL_ASSERT_TRUE( data, "Attempted to get data through a nullptr." );
		#ifdef __CUDAARCH__
		return *data;
		#else
		Element result;
		ArrayOperations< Devices::Host, Devices::Cuda >::copyMemory< Element, Element, int >( &result, data, 1 );
		return result;
		#endif
		}

		template< typename Element, typename Index >

src/TNL/Containers/StaticVector.h

+18 −1

Original line number	Diff line number	Diff line
		@@ -58,6 +58,10 @@ class StaticVector : public StaticArray< Size, Real >
		__cuda_callable__
		StaticVector& operator *= ( const Real& c );

		//! Division by number
		__cuda_callable__
		StaticVector& operator /= ( const Real& c );

		//! Addition operator
		__cuda_callable__
		StaticVector operator + ( const StaticVector& u ) const;
		@@ -156,6 +160,10 @@ class StaticVector< 1, Real > : public StaticArray< 1, Real >
		__cuda_callable__
		StaticVector& operator *= ( const Real& c );

		//! Division by number
		__cuda_callable__
		StaticVector& operator /= ( const Real& c );

		//! Addition operator
		__cuda_callable__
		StaticVector operator + ( const StaticVector& u ) const;
		@@ -257,6 +265,10 @@ class StaticVector< 2, Real > : public StaticArray< 2, Real >
		__cuda_callable__
		StaticVector& operator *= ( const Real& c );

		//! Division by number
		__cuda_callable__
		StaticVector& operator /= ( const Real& c );

		//! Adding operator
		__cuda_callable__
		StaticVector operator + ( const StaticVector& u ) const;
		@@ -358,6 +370,11 @@ class StaticVector< 3, Real > : public StaticArray< 3, Real >
		__cuda_callable__
		StaticVector& operator *= ( const Real& c );

		//! Division by number
		__cuda_callable__
		StaticVector& operator /= ( const Real& c );


		//! Addition operator
		__cuda_callable__
		StaticVector operator + ( const StaticVector& u ) const;