diff --git a/src/matrix/tnlAdaptiveRgCSRMatrix.h b/src/matrix/tnlAdaptiveRgCSRMatrix.h
index 33e312ab49b9aeb9627eab1841862864b273ad35..3b6dccb0d2aeaef2bd3c050cdd04b1f9e99c6c1a 100644
--- a/src/matrix/tnlAdaptiveRgCSRMatrix.h
+++ b/src/matrix/tnlAdaptiveRgCSRMatrix.h
@@ -497,7 +497,7 @@ bool tnlAdaptiveRgCSRMatrix< Real, Device, Index > :: copyFrom( const tnlAdaptiv
dbgFunctionName( "tnlAdaptiveRgCSRMatrix< Real, Device, Index >", "copyFrom" );
maxGroupSize = adaptiveRgCSRMatrix. maxGroupSize;
groupSizeStep = adaptiveRgCSRMatrix. groupSizeStep;
- targetNonzeroesPerGroup = adaptiveRgCSRMatrix. targetNonzeroesPerGroup;
+ desiredChunkSize = adaptiveRgCSRMatrix. desiredChunkSize;
cudaBlockSize = adaptiveRgCSRMatrix. cudaBlockSize;
lastNonzeroElement = adaptiveRgCSRMatrix. lastNonzeroElement;
numberOfGroups = adaptiveRgCSRMatrix. numberOfGroups;
diff --git a/tests/Makefile.am b/tests/Makefile.am
index 09d37c0b62f0d32456d5eb5027a064503fee71ca..cb15ba74178c0e2c7e92ffa300b7d4e8463668e7 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -18,15 +18,15 @@ matrix_solvers_benchmark_sources = matrix-solvers-benchmark.h
tnl_benchmarks_sources = tnl-benchmarks.h
#if BUILD_CUDA
-#matrix_formats_test_sources += matrix-formats-test-cuda.cu
-#sparse_matrix_benchmark_sources += sparse-matrix-benchmark-cuda.cu
-#matrix_solvers_benchmark_sources += matrix-solvers-benchmark-cuda.cu
-#tnl_benchmarks_sources += tnl-benchmarks-cuda.cu
+matrix_formats_test_sources += matrix-formats-test-cuda.cu
+sparse_matrix_benchmark_sources += sparse-matrix-benchmark-cuda.cu
+matrix_solvers_benchmark_sources += matrix-solvers-benchmark-cuda.cu
+tnl_benchmarks_sources += tnl-benchmarks-cuda.cu
#else
-matrix_formats_test_sources += matrix-formats-test.cpp
-sparse_matrix_benchmark_sources += sparse-matrix-benchmark.cpp
-matrix_solvers_benchmark_sources += matrix-solvers-benchmark.cpp
-tnl_benchmarks_sources += tnl-benchmarks.cpp
+#matrix_formats_test_sources += matrix-formats-test.cpp
+#sparse_matrix_benchmark_sources += sparse-matrix-benchmark.cpp
+#matrix_solvers_benchmark_sources += matrix-solvers-benchmark.cpp
+#tnl_benchmarks_sources += tnl-benchmarks.cpp
#endif
matrix_formats_test_SOURCES = $(matrix_formats_test_sources)
@@ -77,9 +77,9 @@ 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
diff --git a/tests/run-sparse-matrix-benchmark b/tests/run-sparse-matrix-benchmark
index 57c7da55dd88053ee8da6855ead77b6c90886d65..6567284b2c664eaa97512a2f96112b026147309a 100755
--- a/tests/run-sparse-matrix-benchmark
+++ b/tests/run-sparse-matrix-benchmark
@@ -37,7 +37,7 @@ write_header()
echo " <td rowspan=4 colspan=2 align=center>Hybrid</td>" >> $1
echo " <td colspan=44 align=center>Row-Grouped CSR</td>" >> $1
echo " <td colspan=44 align=center>Row-Grouped CSR with rows sorted decreasingly by the number of the nonzeros</td>" >> $1
- echo " <td colspan=44 align=center>Adaptive Row-Grouped CSR</td>" >> $1
+ echo " <td colspan=66 align=center>Adaptive Row-Grouped CSR</td>" >> $1
echo " </tr>" >> $1
echo " <tr>" >> $1
@@ -49,9 +49,12 @@ write_header()
echo " <td colspan=11>Group Size = 32</td>" >> $1
echo " <td colspan=11>Group Size = 64</td>" >> $1
echo " <td colspan=11>Group Size Variable</td>" >> $1 # RgCSR rows sorted decreasingly, adaptive group size
- echo " <td colspan=11>Group Size = 1</td>" >> $1 # Adaptive RgCSR
- echo " <td colspan=11>Group Size = 2</td>" >> $1
- echo " <td colspan=11>Group Size = 3</td>" >> $1
+ echo " <td colspan=11>Chunk Size = 1</td>" >> $1 # Adaptive RgCSR
+ echo " <td colspan=11>Chunk Size = 2</td>" >> $1
+ echo " <td colspan=11>Chunk Size = 4</td>" >> $1
+ echo " <td colspan=11>Chunk Size = 8</td>" >> $1
+ echo " <td colspan=11>Chunk Size = 16</td>" >> $1
+ echo " <td colspan=11>Chunk Size = 32</td>" >> $1
echo " </tr>" >> $1
echo " <tr>" >> $1
@@ -83,16 +86,27 @@ write_header()
echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
echo " <td colspan=8>GPU</td>" >> $1
- echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the group size 1
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 1
echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
echo " <td colspan=8>GPU</td>" >> $1
- echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the group size 2
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 2
echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
echo " <td colspan=8>GPU</td>" >> $1
- echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the group size 3
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 4
+ echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
+ echo " <td colspan=8>GPU</td>" >> $1
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 8
+ echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
+ echo " <td colspan=8>GPU</td>" >> $1
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 16
+ echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
+ echo " <td colspan=8>GPU</td>" >> $1
+ echo " <td rowspan=2></td>" >> $1 # Adaptive RgCSR format with the chunk size 32
echo " <td rowspan=2 colspan=2>CPU</td>" >> $1
echo " <td colspan=8>GPU</td>" >> $1
+
+
echo " </tr>" >> $1
@@ -134,18 +148,31 @@ write_header()
echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
- echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the group size = 1
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 1
echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
- echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the group size = 2
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 2
echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
- echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the group size = 3
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 4
+ echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 8
+ echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 16
+ echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
+ echo " <td colspan=2>CUDA Block Size = 32</td>" >> $1 # Adaptive RgCSR format with the chunk size = 32
echo " <td colspan=2>CUDA Block Size = 64</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 128</td>" >> $1
echo " <td colspan=2>CUDA Block Size = 256</td>" >> $1
+
echo " </tr>" >> $1
@@ -295,7 +322,7 @@ write_header()
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
- echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the group size = 16
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 1
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
echo " <td>GFlops</td>" >> $1
@@ -306,7 +333,7 @@ write_header()
echo " <td>Speed-up</td>" >> $1
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
- echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the group size = 32
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 2
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
echo " <td>GFlops</td>" >> $1
@@ -317,7 +344,40 @@ write_header()
echo " <td>Speed-up</td>" >> $1
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
- echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the group size = 64
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 4
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 8
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 16
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>GFlops</td>" >> $1
+ echo " <td>Speed-up</td>" >> $1
+ echo " <td>Artificial Zeros</td>" >> $1 # Adaptive RgCSR format with the chunk size = 32
echo " <td>GFlops</td>" >> $1
echo " <td>Speed-up</td>" >> $1
echo " <td>GFlops</td>" >> $1
diff --git a/tests/sparse-matrix-benchmark.h b/tests/sparse-matrix-benchmark.h
index 4f3a02291f8dd86cc2c3a38b08350e7d6ee0f466..9f765bde2612e77383cd24027c750161bb4706b2 100644
--- a/tests/sparse-matrix-benchmark.h
+++ b/tests/sparse-matrix-benchmark.h
@@ -370,11 +370,11 @@ bool benchmarkMatrix( const tnlString& input_file,
* Adaptive Row-Grouped CSR format
*/
- for( int groupSize = 16; groupSize < 128; groupSize *= 2 )
+ for( int desiredChunkSize = 1; desiredChunkSize < 32; desiredChunkSize *= 2 )
{
tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, tnlHost, int > hostArgCsrMatrixBenchmark;
- hostArgCsrMatrixBenchmark. setGroupSize( groupSize );
+ hostArgCsrMatrixBenchmark. setDesiredChunkSize( desiredChunkSize );
hostArgCsrMatrixBenchmark. setCudaBlockSize( 32 );
hostArgCsrMatrixBenchmark. setup( csrMatrix );
hostArgCsrMatrixBenchmark. runBenchmark( refX, refB, verbose );
@@ -398,7 +398,7 @@ bool benchmarkMatrix( const tnlString& input_file,
}
tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, tnlCuda, int > cudaArgCsrMatrixBenchmark;
- cudaArgCsrMatrixBenchmark. setGroupSize( groupSize );
+ cudaArgCsrMatrixBenchmark. setDesiredChunkSize( desiredChunkSize );
for( int cudaBlockSize = 32; cudaBlockSize <= 256; cudaBlockSize *= 2 )
{
cudaArgCsrMatrixBenchmark. setCudaBlockSize( cudaBlockSize );
@@ -433,143 +433,6 @@ bool benchmarkMatrix( const tnlString& input_file,
}
#ifdef UNDEF
-
- /***
- * Benchmark of the Adaptive Row-grouped CSR format.
- */
- if( verbose )
- cout << left << setw( 25 ) << "AdaptiveRow-grouped CSR " << setw( 5 ) << flush;
-
- tnlAdaptiveRgCSRMatrix< Real, tnlHost > argcsrMatrix( "argcsr-matrix" );
- argcsrMatrix. setCUDABlockSize( 128 );
- if( argcsrMatrix. copyFrom( csrMatrix ) )
- {
- /*time = stop_time;
- benchmarkSpMV< Real, tnlCuda >( cuda_coacsrMatrix,
- cuda_x,
- nonzero_elements,
- cuda_b,
- time,
- spmv_cuda_coa_csr_gflops[ block_iter ],
- spmv_cuda_coa_csr_iter[ block_iter ] );
-
- if( verbose )
- cout << right << setw( 12 ) << setprecision( 2 ) << time
- << right << setw( 15 ) << spmv_cuda_coa_csr_iter[ block_iter ]
- << right << setw( 12 ) << setprecision( 2 ) << spmv_cuda_coa_csr_gflops[ block_iter ];
-
- if( refB != cuda_b )
- {
- if( verbose )
- cout << right << setw( 12 ) << "FAILED." << endl;
- //spmv_cuda_coa_csr_gflops[ block_iter ] = -1.0;
- //return false;
- }
- else*/
- if( verbose )
- cout << right << setw( 12 ) << "OK." << endl;
-
- }
- else
- {
- if( verbose )
- cout << "Format transfer failed!!!" << endl;
- }
-
-
-
-#ifdef HAVE_CUDA
- /***
- * Benchmark of the Adaptive Row-grouped CSR format on the CUDA device.
- */
- if( verbose )
- cout << left << setw( 25 ) << "AdaptiveRow-grouped CSR " << setw( 5 ) << flush;
-
- tnlAdaptiveRgCSRMatrix< Real, tnlCuda > cuda_argcsrMatrix( "cuda-argcsr-matrix" );
- cuda_argcsrMatrix. setCUDABlockSize( 128 );
-
- if( cuda_argcsrMatrix. copyFrom( argcsrMatrix ) )
- {
- time = stop_time;
- benchmarkSpMV< Real, tnlCuda >( cuda_argcsrMatrix,
- cuda_x,
- nonzero_elements,
- cuda_b,
- time,
- benchmarkStatistics. spmv_cuda_arg_csr_gflops,
- benchmarkStatistics. spmv_cuda_arg_csr_iter );
-
- if( verbose )
- cout << right << setw( 12 ) << setprecision( 2 ) << time
- << right << setw( 15 ) << benchmarkStatistics. spmv_cuda_arg_csr_iter
- << right << setw( 12 ) << setprecision( 2 ) << benchmarkStatistics. spmv_cuda_arg_csr_gflops;
-
- if( refB != cuda_b )
- {
- if( verbose )
- cout << right << setw( 12 ) << "FAILED." << endl;
- //spmv_cuda_coa_csr_gflops[ block_iter ] = -1.0;
- //return false;
- }
- else
- if( verbose )
- cout << right << setw( 12 ) << "OK." << endl;
-
- }
- else
- {
- if( verbose )
- cout << "Format transfer failed!!!" << endl;
- }
-
-#endif
-
- int block_iter = 0;
- for( int block_size = 16; block_size < 512; block_size *= 2 )
- {
- /***
- * Benchmark of the Row-grouped CSR format.
- */
- if( verbose )
- cout << left << setw( 25 ) << "Row-grouped CSR " << setw( 5 ) << block_size << flush;
-
- tnlRgCSRMatrix< Real, tnlHost, int > coacsrMatrix( "coacsr-matrix", block_size );
-
- if( coacsrMatrix. copyFrom( csrMatrix ) )
- {
- benchmarkStatistics. coa_csr_artificial_zeros[ block_iter ] = 100.0 * ( double ) coacsrMatrix. getArtificialZeroElements() / ( double ) coacsrMatrix. getNonzeroElements();
-
- time = stop_time;
- benchmarkSpMV< Real, tnlHost >( coacsrMatrix,
- host_x,
- nonzero_elements,
- host_b,
- time,
- benchmarkStatistics. spmv_coacsr_gflops[ block_iter ],
- benchmarkStatistics. spmv_coacsr_iter[ block_iter ] );
- if( verbose )
- cout << right << setw( 12 ) << setprecision( 2 ) << time
- << right << setw( 15 ) << benchmarkStatistics. spmv_coacsr_iter[ block_iter ]
- << right << setw( 12 ) << setprecision( 2 ) << benchmarkStatistics. spmv_coacsr_gflops[ block_iter ] << flush;
-
-
- if( refB != host_b )
- {
- if( verbose )
- cout << right << setw( 12 ) << "FAILED." << endl;
- benchmarkStatistics. spmv_coacsr_gflops[ block_iter ] = -1.0;
- //return false;
- }
- if( verbose )
- cout << left << setw( 12 ) << " OK."
- << right << setw( 14 ) << "Artif.zeros: " << fixed << benchmarkStatistics. coacsr_artificial_zeros[ block_iter ] << "%" << endl;
-
- }
- else
- if( verbose )
- cout << "Format transfer failed!!!" << endl;
-
-
/*
* Benchmark of the Fast CSR format.
*/
diff --git a/tests/tnlSpmvBenchmarkAdaptiveRgCSRMatrix.h b/tests/tnlSpmvBenchmarkAdaptiveRgCSRMatrix.h
index 6be6e966fbe40f914987414e13a9e96013e95415..08a9fd1192ce1ca07d5d73a72e35985c0d540b7a 100644
--- a/tests/tnlSpmvBenchmarkAdaptiveRgCSRMatrix.h
+++ b/tests/tnlSpmvBenchmarkAdaptiveRgCSRMatrix.h
@@ -35,7 +35,7 @@ class tnlSpmvBenchmarkAdaptiveRgCSRMatrix : public tnlSpmvBenchmark< Real, Devic
void writeProgress() const;
- void setGroupSize( const Index groupSize );
+ void setDesiredChunkSize( const Index desiredChunkSize );
void setCudaBlockSize( const Index cudaBlockSize );
@@ -72,7 +72,7 @@ bool tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: setup( const
//tnlAssert( this -> groupSize > 0, cerr << "groupSize = " << this -> groupSize );
if( Device == tnlHost )
{
- this -> matrix. tuneFormat( groupSize, cudaBlockSize );
+ this -> matrix. tuneFormat( desiredChunkSize, cudaBlockSize );
if( ! this -> matrix. copyFrom( matrix ) )
return false;
}
@@ -80,7 +80,7 @@ bool tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: setup( const
{
#ifdef HAVE_CUDA
tnlAdaptiveRgCSRMatrix< Real, tnlHost, Index > hostMatrix( "tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: setup : hostMatrix" );
- hostMatrix. tuneFormat( groupSize, cudaBlockSize );
+ hostMatrix. tuneFormat( desiredChunkSize, cudaBlockSize );
hostMatrix. copyFrom( matrix );
if( ! this -> matrix. copyFrom( hostMatrix ) )
return false;
@@ -107,10 +107,10 @@ void tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: writeProgress
{
cout << left << setw( this -> formatColumnWidth - 15 ) << "Adap. Row-grouped CSR ";
if( Device == tnlCuda )
- cout << setw( 5 ) << this -> groupSize
+ cout << setw( 5 ) << this -> desiredChunkSize
<< setw( 10 ) << this -> cudaBlockSize;
else
- cout << setw( 15 ) << this -> groupSize;
+ cout << setw( 15 ) << this -> desiredChunkSize;
cout << right << setw( this -> timeColumnWidth ) << setprecision( 2 ) << this -> getTime()
<< right << setw( this -> iterationsColumnWidth ) << this -> getIterations()
<< right << setw( this -> gflopsColumnWidth ) << setprecision( 2 ) << this -> getGflops();
@@ -128,9 +128,9 @@ void tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: writeProgress
template< typename Real,
tnlDevice Device,
typename Index >
-void tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: setGroupSize( const Index groupSize )
+void tnlSpmvBenchmarkAdaptiveRgCSRMatrix< Real, Device, Index > :: setDesiredChunkSize( const Index desiredChunkSize )
{
- this -> groupSize = groupSize;
+ this -> desiredChunkSize = desiredChunkSize;
}
template< typename Real,
diff --git a/tests/unit-tests/Makefile.am b/tests/unit-tests/Makefile.am
index 45eab806b7d9ec9a2aa876026a5b20f156886d33..6b4741b90853283aaa3e30dc3b2e39f76ee4f0f4 100644
--- a/tests/unit-tests/Makefile.am
+++ b/tests/unit-tests/Makefile.am
@@ -28,19 +28,19 @@ headers = core/tnl-cuda-kernels.h \
check_PROGRAMS = tnl-unit-tests
-tnl_unit_tests_SOURCES = $(headers) tnl-unit-tests.cpp
-tnl_unit_tests_LDADD = ../../src/libtnl-0.1.la \
- ../../src/core/libcore-tests.la \
- ../../src/diff/libdiff-tests.la \
- ../../src/debug/libtnldebug-0.1.la
+#tnl_unit_tests_SOURCES = $(headers) tnl-unit-tests.cpp
+#tnl_unit_tests_LDADD = ../../src/libtnl-0.1.la \
+# ../../src/core/libcore-tests.la \
+# ../../src/diff/libdiff-tests.la \
+# ../../src/debug/libtnldebug-0.1.la
#if BUILD_CUDA
-#tnl_unit_tests_SOURCES = $(headers) tnl-unit-tests-cuda.cu
-#tnl_unit_tests_CXXFLAGS = "-DHAVE_CUDA"
-#tnl_unit_tests_LDADD = ../../src/libtnl-0.1.la \
-# ../../src/core/libcore-tests.la \
-# ../../src/diff/libdiff-tests.la \
-# ../../src/debug/libtnldebug-0.1.la
+tnl_unit_tests_SOURCES = $(headers) tnl-unit-tests-cuda.cu
+tnl_unit_tests_CXXFLAGS = "-DHAVE_CUDA"
+tnl_unit_tests_LDADD = ../../src/libtnl-0.1.la \
+ ../../src/core/libcore-tests.la \
+ ../../src/diff/libdiff-tests.la \
+ ../../src/debug/libtnldebug-0.1.la
#endif
diff --git a/tests/unit-tests/Makefile.in b/tests/unit-tests/Makefile.in
index ef90d0d2081ea14fe8f82d0ffd95314ff12cc0d7..31dc69edb90e85068552d34bc10fdbda20beec06 100644
--- a/tests/unit-tests/Makefile.in
+++ b/tests/unit-tests/Makefile.in
@@ -277,7 +277,7 @@ headers = core/tnl-cuda-kernels.h \
tnl_unit_tests_SOURCES = $(headers) tnl-unit-tests-cuda.cu
tnl_unit_tests_CXXFLAGS = "-DHAVE_CUDA"
tnl_unit_tests_LDADD = ../../src/libtnl-0.1.la \
- ../../src/core/libcore-tests.la \
+ ../../src/core/libcore-tests.la \
../../src/diff/libdiff-tests.la \
../../src/debug/libtnldebug-0.1.la