diff --git a/CMakeLists.txt b/CMakeLists.txt index fbd21e22c2cbb8380421db505effb1edf3839e51..0aae37de33695ded5a1ce0a68c792d21620864af 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -1054,7 +1054,10 @@ if ( WALBERLA_BUILD_WITH_CUDA ) # set ( BUILD_SHARED_LIBS ON ) set ( CUDA_ATTACH_VS_BUILD_RULE_TO_CUDA_FILE ON ) - set ( CUDA_PROPAGATE_HOST_FLAGS OFF CACHE BOOL "" ) + + if( NOT WALBERLA_CXX_COMPILER_IS_MSVC ) + set ( CUDA_PROPAGATE_HOST_FLAGS OFF CACHE BOOL "" ) + endif() if ( (NOT DEFINED CUDA_HOST_COMPILER) AND (${CMAKE_C_COMPILER} MATCHES "ccache") ) string ( STRIP "${CMAKE_C_COMPILER_ARG1}" stripped_compiler_string ) @@ -1074,11 +1077,11 @@ if ( WALBERLA_BUILD_WITH_CUDA ) # but the compiler seems to note the "isystem" not the "-I" # it is also not possible to get all system include directories - so as a workaround we at least add boost here # as system include - foreach( boostInclude ${Boost_INCLUDE_DIRS} ) + foreach( boostInclude ${Boost_INCLUDE_DIRS} AND NOT WALBERLA_CXX_COMPILER_IS_MSVC ) list( APPEND CUDA_NVCC_FLAGS "-isystem ${boostInclude}" ) endforeach() - if ( NOT "${CUDA_NVCC_FLAGS}" MATCHES "-std=" ) + if ( NOT "${CUDA_NVCC_FLAGS}" MATCHES "-std=" AND NOT WALBERLA_CXX_COMPILER_IS_MSVC ) list ( APPEND CUDA_NVCC_FLAGS "-std=c++11" ) endif () @@ -1092,7 +1095,7 @@ if ( WALBERLA_BUILD_WITH_CUDA ) set ( WALBERLA_BUILD_WITH_CUDA FALSE ) endif ( ) - set_directory_properties(PROPERTIES COMPILE_DEFINITIONS COMPILE_DEFINITIONS_SAVED_STATE) + set_directory_properties(PROPERTIES COMPILE_DEFINITIONS "${COMPILE_DEFINITIONS_SAVED_STATE}" ) endif ( ) ############################################################################################################################ diff --git a/apps/tutorials/cuda/01_GameOfLife_cuda.cpp b/apps/tutorials/cuda/01_GameOfLife_cuda.cpp index 391f4ab11687558e476b2cddac86aed133aa71ab..518acb7e16f39b0136414fd52e9b480f4ee7b11d 100644 --- a/apps/tutorials/cuda/01_GameOfLife_cuda.cpp +++ b/apps/tutorials/cuda/01_GameOfLife_cuda.cpp @@ -36,6 +36,7 @@ #include "field/AddToStorage.h" #include "field/communication/UniformMPIDatatypeInfo.h" +#include "field/vtk/VTKWriter.h" #include "geometry/initializer/ScalarFieldFromGrayScaleImage.h" #include "geometry/structured/GrayScaleImage.h" @@ -127,7 +128,7 @@ int main( int argc, char ** argv ) commScheme.addDataToCommunicate( make_shared<Packing>(gpuFieldSrcID) ); // Create Timeloop - const uint_t numberOfTimesteps = uint_t(10); // number of timesteps for non-gui runs + const uint_t numberOfTimesteps = uint_t(100); // number of timesteps for non-gui runs SweepTimeloop timeloop ( blocks, numberOfTimesteps ); // Registering the sweep @@ -136,6 +137,10 @@ int main( int argc, char ** argv ) timeloop.add() << Sweep( cuda::fieldCpyFunctor<ScalarField, GPUField >(cpuFieldID, gpuFieldDstID) ); + // Register VTK output + timeloop.addFuncAfterTimeStep( field::createVTKOutput<ScalarField>( cpuFieldID, *blocks, "game_of_life" ) ); + + // GUI output GUI gui ( timeloop, blocks, argc, argv ); gui.run(); diff --git a/tests/cuda/CudaMPI.cpp b/tests/cuda/CudaMPI.cpp index 1669a1acdab3afc377067cb6edc927558f4339dd..56d03807f25da2e62e1616d95cf52abd8d92dd9c 100644 --- a/tests/cuda/CudaMPI.cpp +++ b/tests/cuda/CudaMPI.cpp @@ -44,7 +44,7 @@ void fullFieldTransfer() Field<double,4> h_f1 ( 3, 4, 2, 42.0, field::fzyx ); Field<double,4> h_f2 ( 3, 4, 2, 27.0, field::fzyx ); - cuda::GPUField<double> d_f ( 3, 4, 2, 4, 8.0, field::fzyx ); + cuda::GPUField<double> d_f ( 3, 4, 2, 4, 0, field::fzyx ); // Transfer h_f1 from CPU to GPU d_f @@ -100,7 +100,7 @@ void blockStorageAndGui( int argc, char ** argv ) return new GPUField( s->getNumberOfXCells(*block), s->getNumberOfYCells(*block), s->getNumberOfZCells(*block), - 1 , 1.0); + 1 , 1); }, "GPU Field" ); diff --git a/tests/cuda/communication/CommTest.cpp b/tests/cuda/communication/CommTest.cpp index ada3c766f8a3854e5ae4ec2e9e761704c1bc75a6..8233ac5615a106413516627720ee563e98b4fe0a 100644 --- a/tests/cuda/communication/CommTest.cpp +++ b/tests/cuda/communication/CommTest.cpp @@ -54,7 +54,7 @@ void hostToHost() void hostToDevice() { Field<double, 1> hostField(SIZE_X, SIZE_Y, SIZE_Z, 4321.0, LAYOUT); - cuda::GPUField<double> deviceField(SIZE_X, SIZE_Y, SIZE_Z, 1, 0.0, LAYOUT); + cuda::GPUField<double> deviceField(SIZE_X, SIZE_Y, SIZE_Z, 1, 0, LAYOUT); double startTime = MPI_Wtime(); for (int i = 0; i < NUM_ITER; ++i) {