Commit 819b346a authored by Tomáš Oberhuber's avatar Tomáš Oberhuber
Browse files

Merge branch 'develop' of mmg-gitlab.fjfi.cvut.cz:tnl/tnl-dev into develop

parents 034769de fdd2bfe7

.gitlab-ci.yml

+2 −0
Original line number Diff line number Diff line
@@ -21,6 +21,7 @@ stages: 
    WITH_MPI: "no"

    WITH_TESTS: "yes"

    WITH_COVERAGE: "no"

    WITH_DOC: "no"

    # these are built only in the "full" config

    WITH_BENCHMARKS: "no"

    WITH_EXAMPLES: "no"
@@ -55,6 +56,7 @@ stages: 
                -DWITH_CUDA_ARCH=${WITH_CUDA_ARCH}

                -DWITH_MIC=${WITH_MIC}

                -DWITH_TESTS=${WITH_TESTS}

                -DWITH_DOC=${WITH_DOC}

                -DWITH_COVERAGE=${WITH_COVERAGE}

                -DWITH_BENCHMARKS=${WITH_BENCHMARKS}

                -DWITH_EXAMPLES=${WITH_EXAMPLES}

CMakeLists.txt

+13 −6
Original line number Diff line number Diff line
@@ -7,7 +7,9 @@ 
#

###############################################################################



cmake_minimum_required( VERSION 3.5.1 )

# cmake 3.12.2 is required due to compatibility with CUDA 10

# (see the issue reported here: https://github.com/clab/dynet/issues/1457 )

cmake_minimum_required( VERSION 3.12.2 )



project( tnl )
@@ -62,8 +64,8 @@ endif() 


# check if the compiler is good enough

if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")

   # GCC 5.0 is the first release with full C++11 support (due to libstdc++)

   # https://gcc.gnu.org/gcc-5/changes.html

   # GCC 5.0 is the first release with full C++11 support (due to libstdc++) as

   # well as full C++14 support: https://gcc.gnu.org/gcc-5/changes.html

   if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS "5.0")

      message(FATAL_ERROR "Insufficient GCC version")

   endif()
@@ -74,8 +76,13 @@ elseif(CMAKE_CXX_COMPILER_ID STREQUAL "Clang") 
   endif()

endif()



# set C++ standard

set( CMAKE_CXX_STANDARD 14 )

set( CMAKE_CXX_STANDARD_REQUIRED ON )

set( CMAKE_CXX_EXTENSIONS OFF )



# set Debug/Release options

set( CMAKE_CXX_FLAGS "-std=c++11 -pthread -Wall -Wno-unused-local-typedefs -Wno-unused-variable" )

set( CMAKE_CXX_FLAGS "-pthread -Wall -Wno-unused-local-typedefs -Wno-unused-variable" )

set( CMAKE_CXX_FLAGS_DEBUG "-g" )

set( CMAKE_CXX_FLAGS_RELEASE "-O3 -march=native -mtune=native -DNDEBUG" )

#set( CMAKE_CXX_FLAGS_RELEASE "-O3 -march=native -mtune=native -DNDEBUG -ftree-vectorizer-verbose=1 -ftree-vectorize -fopt-info-vec-missed -funroll-loops" )
@@ -147,7 +154,7 @@ endif() 
# Check for CUDA

#

if( ${WITH_CUDA} )

    find_package( CUDA )

    find_package( CUDA 9.0 )

    if( CUDA_FOUND )

        set( BUILD_CUDA TRUE)

        set(CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE OFF)
@@ -186,7 +193,7 @@ if( ${WITH_CUDA} ) 
        # disable false compiler warnings

        #   reference for the -Xcudafe --diag_suppress and --display_error_number flags: https://stackoverflow.com/a/54142937

        #   incomplete list of tokens: http://www.ssl.berkeley.edu/~jimm/grizzly_docs/SSL/opt/intel/cc/9.0/lib/locale/en_US/mcpcom.msg

        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ; -Wno-deprecated-gpu-targets --expt-relaxed-constexpr --expt-extended-lambda -Xcudafe "\"--diag_suppress=code_is_unreachable --diag_suppress=implicit_return_from_non_void_function --diag_suppress=2906 --display_error_number\"")

        set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS} ; -Wno-deprecated-gpu-targets --expt-relaxed-constexpr --expt-extended-lambda -Xcudafe "\"--diag_suppress=code_is_unreachable --diag_suppress=implicit_return_from_non_void_function --diag_suppress=unsigned_compare_with_zero --diag_suppress=2906 --diag_suppress=2913 --display_error_number\"")

        set( ALL_CUDA_ARCHS -gencode arch=compute_20,code=sm_20

                            -gencode arch=compute_30,code=sm_30

                            -gencode arch=compute_32,code=sm_32

Documentation/Doxyfile

+2 −1
Original line number Diff line number Diff line
@@ -2107,7 +2107,8 @@ INCLUDE_FILE_PATTERNS = 
# recursively expanded use the := operator instead of the = operator.

# This tag requires that the tag ENABLE_PREPROCESSING is set to YES.



PREDEFINED             =

PREDEFINED             = HAVE_MPI=1

                         HAVE_CUDA=1



# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this

# tag can be used to specify a list of macro names that should be expanded. The

src/Benchmarks/DistSpMV/tnl-benchmark-distributed-spmv.h

+4 −6
Original line number Diff line number Diff line
@@ -163,9 +163,8 @@ struct SpmvBenchmark 
   {

      MatrixType matrix;

      VectorType vector;

      if( ! matrix.load( parameters.getParameter< String >( "input-matrix" ) ) ||

          ! vector.load( parameters.getParameter< String >( "input-vector" ) ) )

          return false;

      matrix.load( parameters.getParameter< String >( "input-matrix" ) );

      File( parameters.getParameter< String >( "input-vector" ), std::ios_base::in ) >> vector;



      typename MatrixType::CompressedRowLengthsVector rowLengths;

      matrix.getCompressedRowLengths( rowLengths );
@@ -267,8 +266,7 @@ struct SpmvBenchmark 
      distributedMatrix.vectorProduct( distributedVector, distributedY );

      const int rank = CommunicatorType::GetRank( distributedMatrix.getCommunicationGroup() );

      const int nproc = CommunicatorType::GetSize( distributedMatrix.getCommunicationGroup() );

      VectorType subY( y,

                       Partitioner::getOffset( matrix.getRows(), rank, nproc ),

      typename VectorType::ViewType subY( &y[ Partitioner::getOffset( matrix.getRows(), rank, nproc ) ],

                                          Partitioner::getSizeForRank( matrix.getRows(), rank, nproc ) );

      TNL_ASSERT_EQ( distributedY.getLocalVectorView(), subY, "WRONG RESULT !!!" );

#endif

src/Benchmarks/HeatEquation/HeatEquationBenchmarkProblem_impl.h

+9 −4
Original line number Diff line number Diff line
@@ -144,7 +144,11 @@ setInitialCondition( const Config::ParameterContainer& parameters, 
{

   const String& initialConditionFile = parameters.getParameter< String >( "initial-condition" );

   Functions::MeshFunction< Mesh > u( this->getMesh(), dofsPointer );

   if( ! u.boundLoad( initialConditionFile ) )

   try

   {

      u.boundLoad( initialConditionFile );

   }

   catch(...)

   {

      std::cerr << "I am not able to load the initial condition from the file " << initialConditionFile << "." << std::endl;

      return false;
@@ -200,8 +204,7 @@ makeSnapshot( const RealType& time, 
   fileName.setIndex( step );



   //FileNameBaseNumberEnding( "u-", step, 5, ".tnl", fileName );

   if( ! u.save( fileName.getFileName() ) )

      return false;

   u.save( fileName.getFileName() );

   return true;

}
@@ -595,11 +598,13 @@ getExplicitUpdate( const RealType& time, 
            userData.real_u = uDofs->getData();

            userData.real_fu = fuDofs->getData();

#endif                        

            /*

            const IndexType gridXSize = mesh->getDimensions().x();

            const IndexType gridYSize = mesh->getDimensions().y();

            dim3 cudaBlockSize( 16, 16 );

            dim3 cudaGridSize( gridXSize / 16 + ( gridXSize % 16 != 0 ),

                               gridYSize / 16 + ( gridYSize % 16 != 0 ) );

            */



            TNL::Devices::Cuda::synchronizeDevice();

            int cudaErr;