Commit 8b80f8f3 authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Implementing the SPMV benchmark.

parent fdc38e75

src/implementation/matrices/tnlEllpackMatrix_impl.h

+10 −0
Original line number Diff line number Diff line
@@ -156,6 +156,16 @@ bool tnlEllpackMatrix< Real, Device, Index >::operator != ( const tnlEllpackMatr 
   return ! ( ( *this ) == matrix );

}



/*template< typename Real,

          typename Device,

          typename Index >

   template< typename Matrix >

bool tnlEllpackMatrix< Real, Device, Index >::copyFrom( const Matrix& matrix,

                                                        const RowLengthsVector& rowLengths )

{

   return tnlMatrix< RealType, DeviceType, IndexType >::copyFrom( matrix, rowLengths );

}*/



template< typename Real,

          typename Device,

          typename Index >

src/matrices/tnlEllpackMatrix.h

+4 −0
Original line number Diff line number Diff line
@@ -63,6 +63,10 @@ class tnlEllpackMatrix : public tnlSparseMatrix< Real, Device, Index > 
   template< typename Real2, typename Device2, typename Index2 >

   bool operator != ( const tnlEllpackMatrix< Real2, Device2, Index2 >& matrix ) const;



   /*template< typename Matrix >

   bool copyFrom( const Matrix& matrix,

                  const RowLengthsVector& rowLengths );*/



#ifdef HAVE_CUDA

   __device__ __host__

#endif

tests/benchmarks/share/CMakeLists.txt

+1 −0
Original line number Diff line number Diff line
@@ -11,6 +11,7 @@ INSTALL( FILES matrix-market 
               run-tnl-benchmark-spmv

               cuda-profiler.conf

               process-cuda-profile.pl 

               tnl-log-to-html.py

               DESTINATION share/tnl-${tnlVersion}/benchmark-scripts )



INSTALL( FILES tnl-run-spmv-benchmark

tests/benchmarks/tnl-benchmark-spmv.h

+13 −3
Original line number Diff line number Diff line
@@ -206,9 +206,12 @@ bool setupBenchmark( const tnlParameterContainer& parameters ) 
#ifdef HAVE_CUDA

      typedef tnlVector< Real, tnlCuda, int > CudaVector;

      CudaVector cudaX, cudaB;

      tnlVector< int, tnlCuda, int > rowLengthsCuda;

      cudaX.setSize( csrMatrix.getColumns() );

      cudaX.setValue( 1.0 );

      cudaB.setSize( csrMatrix.getRows() );

      rowLengthsCuda.setSize( csrMatrix.getRows() );

      rowLengthsCuda = rowLengthsHost;

#endif

      benchmarkHostMatrix( csrMatrix,

                           hostX,
@@ -232,8 +235,15 @@ bool setupBenchmark( const tnlParameterContainer& parameters ) 
                           verbose,

                           logFile );

#ifdef HAVE_CUDA

      /*typedef tnlEllpackMatrix< Real, tnlCuda, int > EllpackMatrixCudaType;

      EllpackMatrixCudaType cudaEllpackMatrix( ellpackMatrix);

      typedef tnlEllpackMatrix< Real, tnlCuda, int > EllpackMatrixCudaType;

      EllpackMatrixCudaType cudaEllpackMatrix;

      cout << "Copying matrix to GPU... ";

      if( ! cudaEllpackMatrix.copyFrom( ellpackMatrix, rowLengthsCuda ) )

      {

         cerr << "I am not able to transfer the matrix on GPU." << endl;

         return false;

      }

      cout << " done." << endl;

      benchmarkHostMatrix( cudaEllpackMatrix,

                           cudaX,

                           cudaB,
@@ -242,7 +252,7 @@ bool setupBenchmark( const tnlParameterContainer& parameters ) 
                           stopTime,

                           verbose,

                           logFile );

      cudaEllpackMatrix.reset();*/

      cudaEllpackMatrix.reset();

#endif

      ellpackMatrix.reset();