HCC_AMDGPU_TARGET (for ROCm <= 4.0) or CUDA_PATH are necessary for hipcc If you are compiling with HIP, note that before running CMake you will have to If this is notĭesired, you can set USE_STATIC_OPENCL_LOADER to no. OpenCL compatible drivers need to be installed to use OpenCL. Way no local OpenCL development headers or library needs to be present and only Link with a static OpenCL ICD loader library and standard OpenCL headers. If you are compiling for OpenCL, the default setting is to download, build, and Process and will terminate with an error and a suggestion for how to remove them. CMake will detect files generated by that When building with CMake, you must NOT build the GPU library in lib/gpu The preferred GPU architecture directly included rather than having to waitįor the JIT compiler of the CUDA driver to translate it. Thus the GPU_ARCH setting is merely an optimization, to have code for To include support for all major GPU architectures supported by this toolkit. Sm_80 for Ampere (supported since CUDA 11)Ī more detailed list can be found, for example,ĬMake can detect which version of the CUDA toolkit is used and thus will try Sm_75 for Turing (supported since CUDA 10) Sm_60 or sm_61 for Pascal (supported since CUDA 8) Sm_50 or sm_52 for Maxwell (supported since CUDA 6) Sm_35 or sm_37 for Kepler (supported since CUDA 5 and until CUDA 11.x) Sm_30 for Kepler (supported since CUDA 5 and until CUDA 10.x) Sm_20 or sm_21 for Fermi (supported by CUDA 3.2 until CUDA 7.5) Sm_12 or sm_13 for GT200 (supported by CUDA 3.2 until CUDA 6.5) GPU_ARCH settings for different GPU hardware is as follows: D USE_STATIC_OPENCL_LOADER =value # downloads/includes OpenCL ICD loader library, no local OpenCL headers/libs needed # value = yes (default) or no D CUDA_MPS_SUPPORT =value # enables some tweaks required to run with active nvidia-cuda-mps daemon # value = yes or no (default) D CUDPP_OPT =value # use GPU binning on with CUDA (should be off for modern GPUs) # enables CUDA Performance Primitives, must be "no" for CUDA_MPS_SUPPORT=yes # value = yes or no (default) D HIP_USE_DEVICE_SORT =value # enables GPU sorting # value = yes (default) or no D HIP_ARCH =value # primary GPU hardware choice for GPU_API=hip # value depends on selected HIP_PLATFORM # default is 'gfx906' for HIP_PLATFORM=amd and 'sm_50' for HIP_PLATFORM=nvcc D GPU_ARCH =value # primary GPU hardware choice for GPU_API=cuda # value = sm_XX, see below # default is sm_50 D GPU_PREC =value # precision setting # value = double or mixed (default) or single D GPU_API =value # value = opencl (default) or cuda or hip Notes for saving disk space when building LAMMPS from source
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |