Commit ba77aacf authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Debugging AdaptiveRgCSR format.

parent 169fd2ef

src/matrix/tnlAdaptiveRgCSRMatrix.h

+2 −0
Original line number Diff line number Diff line
@@ -225,6 +225,7 @@ bool tnlAdaptiveRgCSRMatrix< Real, Device, Index > :: setSize( Index new_size ) 
template< typename Real, tnlDevice Device, typename Index >

bool tnlAdaptiveRgCSRMatrix< Real, Device, Index > :: setNonzeroElements( Index elements )

{

   tnlAssert( elements !=0, );

   if( ! nonzero_elements.setSize(elements) ||  ! columns.setSize(elements) )

      return false;

   nonzero_elements.setValue( 0.0 );
@@ -249,6 +250,7 @@ template< typename Real, tnlDevice Device, typename Index > 
bool tnlAdaptiveRgCSRMatrix< Real, Device, Index > :: copyFrom( const tnlCSRMatrix< Real, tnlHost, Index >& mat )

{

	dbgFunctionName( "tnlAdaptiveRgCSRMatrix< Real, tnlHost >", "copyFrom" );

	tnlAssert( cudaBlockSize != 0, );

	if( ! this -> setSize( mat.getSize() ) )

		return false;

tests/Makefile.am

+3 −3
Original line number Diff line number Diff line
@@ -74,9 +74,9 @@ matrix_solvers_benchmark_dbg_LDADD = ../src/libtnl-0.1.la \ 
endif



#if BUILD_CUDA

matrix_formats_test_CXXFLAGS = "-DHAVE_CUDA"

sparse_matrix_benchmark_CXXFLAGS += "-DHAVE_CUDA"

tnl_benchmarks_CXXFLAGS = "-DHAVE_CUDA"

matrix_formats_test_CXXFLAGS = -DHAVE_CUDA

sparse_matrix_benchmark_CXXFLAGS += -DHAVE_CUDA

tnl_benchmarks_CXXFLAGS = -DHAVE_CUDA

#endif

tests/Makefile.in

+3 −3
Original line number Diff line number Diff line
@@ -371,7 +371,7 @@ matrix_formats_test_LDADD = ../src/libtnl-0.1.la \ 
                            ../src/matrix/libtnlmatrix-0.1.la 



sparse_matrix_benchmark_CXXFLAGS = $(OPTIMISECXXFLAGS) \

	$(CUSP_CXXFLAGS) "-DHAVE_CUDA"

	$(CUSP_CXXFLAGS) -DHAVE_CUDA

sparse_matrix_benchmark_SOURCES = $(sparse_matrix_benchmark_sources)

sparse_matrix_benchmark_LDADD = ../src/libtnl-0.1.la \

                                ../src/core/libtnlcore-0.1.la \
@@ -406,8 +406,8 @@ reorder_csr_SOURCES = ReorderCSR.cpp 




#if BUILD_CUDA

matrix_formats_test_CXXFLAGS = "-DHAVE_CUDA"

tnl_benchmarks_CXXFLAGS = "-DHAVE_CUDA"

matrix_formats_test_CXXFLAGS = -DHAVE_CUDA

tnl_benchmarks_CXXFLAGS = -DHAVE_CUDA

#endif

check_SCRIPTS = get-matrices \

                convert-matrices \

tests/run-sparse-matrix-benchmark

+1 −1
Original line number Diff line number Diff line
@@ -4,7 +4,7 @@ BASE="ftp://math.nist.gov/pub/MatrixMarket2/Harwell-Boeing/" 
                

PWD=`pwd`

IWD="$PWD"

SPARSE_MATRIX_BENCHMARK="$IWD/sparse-matrix-benchmark"

SPARSE_MATRIX_BENCHMARK="$IWD/sparse-matrix-benchmark-dbg"

STOP_TIME="1"



source ../tnl-env-variables

tests/sparse-matrix-benchmark.h

+8 −2
Original line number Diff line number Diff line
@@ -18,6 +18,8 @@ 
#ifndef SPARSEMATRIXBENCHMARK_H_

#define SPARSEMATRIXBENCHMARK_H_



#define HAVE_CUDA



#include <fstream>

#include <iomanip>

#include <matrix/tnlCSRMatrix.h>
@@ -32,6 +34,8 @@ 
#include <core/mfuncs.h>

#include <config.h>







#ifdef HAVE_CUSP

#include <cusp-test.h>

#endif
@@ -329,6 +333,7 @@ bool benchmarkMatrix( const tnlString& input_file, 
      cout << left << setw( 25 ) << "AdaptiveRow-grouped CSR " << setw( 5 ) << flush;



   tnlAdaptiveRgCSRMatrix< REAL, tnlHost > argcsr_matrix( "argcsr-matrix" );

   argcsr_matrix. setCUDABlockSize( 128 );

   if( argcsr_matrix. copyFrom( csr_matrix ) )

   {

      /*time = stop_time;
@@ -352,9 +357,9 @@ bool benchmarkMatrix( const tnlString& input_file, 
         //spmv_cuda_coa_csr_gflops[ block_iter ] = -1.0;

         //return false;

      }

      else

      else*/

         if( verbose )

            cout << right << setw( 12 ) << "OK." << endl;*/

            cout << right << setw( 12 ) << "OK." << endl;



   }

   else
@@ -373,6 +378,7 @@ bool benchmarkMatrix( const tnlString& input_file, 
      cout << left << setw( 25 ) << "AdaptiveRow-grouped CSR " << setw( 5 ) << flush;



   tnlAdaptiveRgCSRMatrix< REAL, tnlCuda > cuda_argcsr_matrix( "cuda-argcsr-matrix" );

   cuda_argcsr_matrix. setCUDABlockSize( 128 );



   if( cuda_argcsr_matrix. copyFrom( argcsr_matrix ) )

   {