projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: a4c3a64) | patch
Make GPU to CUDA transformations independent of CUDA runtime.
authorStephan Herhut <herhut@google.com>
Wed, 26 Jun 2019 12:16:11 +0000 (05:16 -0700)
committerA. Unique TensorFlower <gardener@tensorflow.org>
Wed, 26 Jun 2019 12:16:37 +0000 (05:16 -0700)
commitc72c6c390710a6e909670e2b2c0e4076cc1b89c6
tree15286a4f3e4354e14a6f382af18934ce36b68701tree | snapshot
parenta4c3a6455c404ef4e56a73be9457affb53536873commit | diff
Make GPU to CUDA transformations independent of CUDA runtime.

The actual transformation from PTX source to a CUDA binary is now factored out,
enabling compiling and testing the transformations independently of a CUDA
runtime.

MLIR has still to be built with NVPTX target support for the conversions to be
built and tested.

PiperOrigin-RevId: 255167139
mlir/CMakeLists.txt diff | blob | history
mlir/include/mlir/Conversion/GPUToCUDA/GPUToCUDAPass.h diff | blob | history
mlir/lib/Conversion/GPUToCUDA/CMakeLists.txt diff | blob | history
mlir/lib/Conversion/GPUToCUDA/ConvertKernelFuncToCubin.cpp diff | blob | history
mlir/test/Conversion/GPUToCUDA/lit.local.cfg [new file with mode: 0644] blob
mlir/test/Conversion/GPUToCUDA/lower-launch-func-to-cuda.mlir [new file with mode: 0644] blob
mlir/test/Conversion/GPUToCUDA/lower-nvvm-kernel-to-cubin.mlir [new file with mode: 0644] blob
mlir/test/lit.site.cfg.py.in diff | blob | history
mlir/tools/mlir-opt/CMakeLists.txt diff | blob | history
Domain: System / Toolchain;