Commit 6fc0cab5 authored by Tomáš Oberhuber's avatar Tomáš Oberhuber Committed by Jakub Klinkovský
Browse files

Fixing necessary condition for specialization of dense matrix vector multiplication.

parent 2c566727

src/Benchmarks/BLAS/dense-mv.h

+3 −1
Original line number Diff line number Diff line
@@ -105,7 +105,8 @@ benchmarkDenseMVSynthetic( Benchmark<> & benchmark, 
   };

   benchmark.time< Devices::Cuda >( reset, "GPU row", rowMajorMvCuda );



   //std::cerr << "Diff. = " << TNL::max( abs( outCudaVector2 - outCudaVector1 ) ) << std::endl;

   auto diff = TNL::max( abs( outCudaVector2 - outCudaVector1 ) );

   //std::cerr << outCudaVector1 << std::endl << outCudaVector2 << std::endl;



   rowMajorCudaMatrix.reset();

   columnMajorCudaMatrix.setDimensions( size, size );
@@ -123,6 +124,7 @@ benchmarkDenseMVSynthetic( Benchmark<> & benchmark, 
   };

   benchmark.time< Devices::Cuda >( reset, "GPU cublas", mvCublas );



   //std::cerr << "Diff. = " << diff << std::endl;

#endif

}

src/TNL/Matrices/DenseMatrixView.hpp

+2 −1
Original line number Diff line number Diff line
@@ -724,7 +724,8 @@ vectorProduct( const InVector& inVector, 
   if( end == 0 )

      end = this->getRows();



   if( std::is_same< DeviceType, Devices::Cuda >::value )

   if( std::is_same< DeviceType, Devices::Cuda >::value &&

      matrixMultiplicator == 1.0 && outVectorMultiplicator == 0.0 )

   {

#ifdef HAVE_CUDA

      if( Organization == Algorithms::Segments::ColumnMajorOrder )