![]() ![]() Linux (WSL) is not needed to build and run the CUDA backend.Įnabling this flag requires an installation of at least The CUDA backend has Windows support windows subsystem for Windows DPC toolchain, but add the -cuda flag to configure.py. To enable support for CUDA devices, follow the instructions for the Linux or Python %DPCPP_HOME% \llvm \buildbot \compile.pyīuild DPC toolchain with support for NVIDIA CUDA ¶ Python %DPCPP_HOME% \llvm \buildbot \configure.py -use-libcxx \ There is experimental support for building and linking DPC runtime with Build DPC toolchain with libc library ¶ This allows you, for example, toĬonfigure several different builds and then build just one of them which is Placed in non-default directory using -o flag, you must also specify this flagĪnd the same path in compile.py options. Please note that no data about flags is being shared between configure.py andĬompile.py scripts, which means that if you configured your build to be j, -build-parallelism -> Number of threads to use for compilation t, -build-target -> Build target (e.g., clang or llvm-spirv). You can use the following flags with compile.py (full list of available flagsĬan be found by launching the script with -help): ![]() cmake-gen -> Set build system type (e.g. enable-all-llvm-targets -> build compiler (but not a runtime) with all enable-esimd-emulator -> enable ESIMD CPU emulation (see ESIMD CPU emulation) hip-platform -> select the platform used by the hip backend, AMD or NVIDIA (see HIP AMD or see HIP NVIDIA) cuda -> use the cuda backend (see Nvidia CUDA) werror -> treat warnings as errors when compiling LLVM You can use the following flags with configure.py (full list of availableįlags can be found by launching the script with -help): Python %DPCPP_HOME%\llvm\buildbot\compile.py Python %DPCPP_HOME%\llvm\buildbot\configure.py Windows: Visual Studio version 15.7 preview 4 or later -Īlternatively, you can use a Docker image that has everything you need for building Linux: GCC version 7.1.0 or later (including libstdc ). Using the DPC toolchain on CUDA platforms Run Khronos* SYCL* conformance test suite (optional) Obtain prerequisites for ahead of time (AOT) compilation Table of contents ¶īuild DPC toolchain with libc libraryīuild DPC toolchain with support for NVIDIA CUDAīuild DPC toolchain with support for HIP AMDīuild DPC toolchain with support for HIP NVIDIAīuild DPC toolchain with support for ESIMD CPU Emulationīuild DPC toolchain with support for runtime kernel fusion The DPC Compiler compiles C and SYCL* source files with code for both CPUĪnd a wide range of compute accelerators such as GPU and FPGA.
0 Comments
Leave a Reply. |