Merge branch 'TO/matrices' into 'develop' (ac30fb61) · Commits · TNL / tnl-dev

Documentation/Examples/Matrices/CMakeLists.txt

+11 −4

Original line number	Diff line number	Diff line
		@@ -64,6 +64,11 @@ IF( BUILD_CUDA )
		${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/DenseMatrixExample_forAllRows.out
		OUTPUT DenseMatrixExample_forAllRows.out )

		CUDA_ADD_EXECUTABLE( DenseMatrixViewExample_constructor_cuda DenseMatrixViewExample_constructor.cu )
		ADD_CUSTOM_COMMAND( COMMAND DenseMatrixViewExample_constructor_cuda >
		${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/DenseMatrixViewExample_constructor.out
		OUTPUT DenseMatrixViewExample_constructor.out )

		ELSE()
		ADD_EXECUTABLE( DenseMatrixExample_Constructor_init_list DenseMatrixExample_Constructor_init_list.cpp )
		ADD_CUSTOM_COMMAND( COMMAND DenseMatrixExample_Constructor_init_list >
		@@ -130,11 +135,12 @@ ELSE()
		${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/DenseMatrixExample_forAllRows.out
		OUTPUT DenseMatrixExample_forAllRows.out )

		ENDIF()



		ADD_EXECUTABLE( DenseMatrixViewExample_constructor DenseMatrixViewExample_constructor.cpp )
		ADD_CUSTOM_COMMAND( COMMAND DenseMatrixViewExample_constructor >
		${TNL_DOCUMENTATION_OUTPUT_SNIPPETS_PATH}/DenseMatrixViewExample_constructor.out
		OUTPUT DenseMatrixViewExample_constructor.out )

		ENDIF()

		ADD_CUSTOM_TARGET( RunMatricesExamples ALL DEPENDS
		DenseMatrixExample_Constructor_init_list.out
		@@ -150,5 +156,6 @@ ADD_CUSTOM_TARGET( RunMatricesExamples ALL DEPENDS
		DenseMatrixExample_allRowsReduction.out
		DenseMatrixExample_forRows.out
		DenseMatrixExample_forAllRows.out
		DenseMatrixViewExample_constructor.out
		)

Documentation/Examples/Matrices/DenseMatrixExample_setElement.cpp

+17 −3

Original line number	Diff line number	Diff line
		#include <iostream>
		#include <TNL/Algorithms/ParallelFor.h>
		#include <TNL/Matrices/DenseMatrix.h>
		#include <TNL/Devices/Host.h>
		#include <TNL/Pointers/SharedPointer.h>
		#include <TNL/Pointers/SmartPointersRegister.h>

		template< typename Device >
		void setElements()
		{
		TNL::Matrices::DenseMatrix< double, Device > matrix( 5, 5 );
		TNL::Pointers::SharedPointer< TNL::Matrices::DenseMatrix< double, Device > > matrix( 5, 5 );
		for( int i = 0; i < 5; i++ )
		matrix.setElement( i, i, i );
		matrix->setElement( i, i, i );

		std::cout << "Matrix set from the host:" << std::endl;
		std::cout << *matrix << std::endl;

		auto f = [=] __cuda_callable__ ( int i ) mutable {
		matrix->setElement( i, i, -i );
		};
		TNL::Pointers::synchronizeSmartPointersOnDevice< Device >();
		TNL::Algorithms::ParallelFor< Device >::exec( 0, 5, f );

		std::cout << "Matrix set from its native device:" << std::endl;
		std::cout << *matrix << std::endl;

		std::cout << matrix << std::endl;
		}

		int main( int argc, char* argv[] )

Documentation/Examples/Matrices/DenseMatrixViewExample_constructor.cpp

0 → 100644

+39 −0

Original line number	Diff line number	Diff line
		#include <iostream>
		#include <TNL/Algorithms/ParallelFor.h>
		#include <TNL/Matrices/DenseMatrix.h>
		#include <TNL/Devices/Host.h>
		#include <TNL/Devices/Cuda.h>

		template< typename Device >
		void createMatrixView()
		{
		TNL::Containers::Vector< double, Device > values {
		1, 2, 3, 4,
		5, 6, 7, 8,
		9, 10, 11, 12 };

		TNL::Matrices::DenseMatrixView< double, Device, int, TNL::Containers::Segments::RowMajorOrder > matrix( 5, 5, values.getView() );

		/***
		* We need a matrix view to pass the matrix to lambda function even on CUDA device.
		*/
		/*auto matrixView = matrix.getView();
		auto f = [=] __cuda_callable__ ( int rowIdx ) mutable {
		auto row = matrixView.getRow( rowIdx );
		row.setElement( rowIdx, 10* ( rowIdx + 1 ) );
		};

		TNL::Algorithms::ParallelFor< Device >::exec( 0, matrix.getRows(), f );
		std::cout << matrix << std::endl;*/
		}

		int main( int argc, char* argv[] )
		{
		std::cout << "Creating matrix view on host: " << std::endl;
		createMatrixView< TNL::Devices::Host >();

		#ifdef HAVE_CUDA
		std::cout << "Creating matrix view on CUDA device: " << std::endl;
		createMatrixView< TNL::Devices::Cuda >();
		#endif
		}

Documentation/Examples/Matrices/DenseMatrixViewExample_constructor.cu

0 → 120000

+1 −0

Original line number	Diff line number	Diff line
		DenseMatrixViewExample_constructor.cpp
		No newline at end of file

src/TNL/Containers/Segments/BiEllpack.h

+9 −9

Original line number	Diff line number	Diff line
		@@ -22,7 +22,7 @@ namespace TNL {
		template< typename Device,
		typename Index,
		typename IndexAllocator = typename Allocators::Default< Device >::template Allocator< Index >,
		bool RowMajorOrder = std::is_same< Device, Devices::Host >::value,
		ElementsOrganization Organization = Containers::Segments::DefaultElementsOrganization< Device >::getOrganization(),
		int WarpSize = 32 >
		class BiEllpack
		{
		@@ -31,12 +31,12 @@ class BiEllpack
		using DeviceType = Device;
		using IndexType = std::remove_const_t< Index >;
		using OffsetsHolder = Containers::Vector< Index, DeviceType, IndexType, IndexAllocator >;
		static constexpr bool getRowMajorOrder() { return RowMajorOrder; }
		using ViewType = BiEllpackView< Device, Index, RowMajorOrder >;
		static constexpr bool getOrganization() { return Organization; }
		using ViewType = BiEllpackView< Device, Index, Organization >;
		template< typename Device_, typename Index_ >
		using ViewTemplate = BiEllpackView< Device_, Index_, RowMajorOrder >;
		using ConstViewType = BiEllpackView< Device, std::add_const_t< IndexType >, RowMajorOrder >;
		using SegmentViewType = BiEllpackSegmentView< IndexType, RowMajorOrder >;
		using ViewTemplate = BiEllpackView< Device_, Index_, Organization >;
		using ConstViewType = BiEllpackView< Device, std::add_const_t< IndexType >, Organization >;
		using SegmentViewType = BiEllpackSegmentView< IndexType, Organization >;

		BiEllpack() = default;

		@@ -109,8 +109,8 @@ class BiEllpack

		BiEllpack& operator=( const BiEllpack& source ) = default;

		template< typename Device_, typename Index_, typename IndexAllocator_, bool RowMajorOrder_ >
		BiEllpack& operator=( const BiEllpack< Device_, Index_, IndexAllocator_, RowMajorOrder_, WarpSize >& source );
		template< typename Device_, typename Index_, typename IndexAllocator_, ElementsOrganization Organization_ >
		BiEllpack& operator=( const BiEllpack< Device_, Index_, IndexAllocator_, Organization_, WarpSize >& source );

		void save( File& file ) const;

		@@ -163,7 +163,7 @@ class BiEllpack
		return 0;
		};

		template< typename Device_, typename Index_, typename IndexAllocator_, bool RowMajorOrder_, int WarpSize_ >
		template< typename Device_, typename Index_, typename IndexAllocator_, ElementsOrganization Organization_, int WarpSize_ >
		friend class BiEllpack;
		};