-
cmake 3.3.0 or higher
- Debian/Ubuntu:
sudo apt-get install cmake file cmake-curses-gui - Arch Linux:
sudo pacman --sync cmake
- Debian/Ubuntu:
-
cupla
git@github.com:ComputationalRadiationPhysics/cupla.gitexport CUPLA_ROOT=<cupla_SRC_CODE_DIR>- example:
mkdir -p $HOME/srcgit clone git@github.com:ComputationalRadiationPhysics/cupla.git $HOME/src/cuplacd $HOME/src/cuplagit submodule initgit submodule updateexport CUPLA_ROOT=$HOME/src/cupla
- create build directory
mkdir -p buildCuplaExample cd buildCuplaExamplecmake $CUPLA_ROOT/example/CUDASamples/matrixMul -D<ACC_TYPE>=ON- list of supported ACC_TYPES
ALPAKA_ACC_CPU_B_SEQ_T_OMP2_ENABLEALPAKA_ACC_CPU_B_SEQ_T_THREADS_ENABLEALPAKA_ACC_GPU_CUDA_ENABLEALPAKA_ACC_CPU_B_OMP2_T_SEQ_ENABLE(only allowed in combination withCUPLA_KERNEL_OPTIandCUPLA_KERNEL_ELEM, because theblockSizemust be dim3(1,1,1)) see TuningGuide.md
- list of supported ACC_TYPES
make -j./matrixMul -wA=320 -wB=320 -hA=320 -hB=320(parameters must be a multiple of 32!)