Optimizing lambda functions for SpMV in sparse matrix view!!! (fd6b28fb) · Commits · TNL / tnl-dev

src/TNL/Matrices/SparseMatrixView.hpp

+27 −2

Original line number	Diff line number	Diff line
		@@ -462,12 +462,37 @@ vectorProduct( const InVector& inVector,
		outVectorView[ row ] = outVectorMultiplicator * outVectorView[ row ] + matrixMultiplicator * value;
		}
		};
		auto keeperDirect = [=] __cuda_callable__ ( IndexType row, const ComputeRealType& value ) mutable {
		outVectorView[ row ] = value;
		};
		auto keeperMatrixMult = [=] __cuda_callable__ ( IndexType row, const ComputeRealType& value ) mutable {
		outVectorView[ row ] = matrixMultiplicator * value;
		};
		auto keeperVectorMult = [=] __cuda_callable__ ( IndexType row, const ComputeRealType& value ) mutable {
		outVectorView[ row ] = outVectorMultiplicator * outVectorView[ row ] + value;
		};

		if( lastRow == 0 )
		lastRow = this->getRows();
		if( isSymmetric() )
		this->segments.segmentsReduction( firstRow, lastRow, symmetricFetch, std::plus<>{}, keeper, ( ComputeRealType ) 0.0 );
		this->segments.segmentsReduction( firstRow, lastRow, symmetricFetch, std::plus<>{}, keeperGeneral, ( ComputeRealType ) 0.0 );
		else
		this->segments.segmentsReduction( firstRow, lastRow, fetch, std::plus<>{}, keeper, ( ComputeRealType ) 0.0 );
		{
		if( outVectorMultiplicator == 0.0 )
		{
		if( matrixMultiplicator == 1.0 )
		this->segments.segmentsReduction( firstRow, lastRow, fetch, std::plus<>{}, keeperDirect, ( ComputeRealType ) 0.0 );
		else
		this->segments.segmentsReduction( firstRow, lastRow, fetch, std::plus<>{}, keeperMatrixMult, ( ComputeRealType ) 0.0 );
		}
		else
		{
		if( matrixMultiplicator == 1.0 )
		this->segments.segmentsReduction( firstRow, lastRow, fetch, std::plus<>{}, keeperVectorMult, ( ComputeRealType ) 0.0 );
		else
		this->segments.segmentsReduction( firstRow, lastRow, fetch, std::plus<>{}, keeperGeneral, ( ComputeRealType ) 0.0 );
		}
		}
		}

		template< typename Real,