projects / platform / upstream / tvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 8935990) | patch
codegen llvm: move nvptx-specific intrinsic handling into codegen_nvptx (#5726)
authorThomas Viehmann <tv.code@beamnet.de>
Thu, 4 Jun 2020 15:46:55 +0000 (17:46 +0200)
committerGitHub <noreply@github.com>
Thu, 4 Jun 2020 15:46:55 +0000 (08:46 -0700)
commit490510d463bd760fc474f956a3098b88ebf7260a
treeb449da886a984b838cfec83c5bab65a348c160a2tree | snapshot
parent89359907d68dff2909d1fad45ae146d1e62314a6commit | diff
codegen llvm: move nvptx-specific intrinsic handling into codegen_nvptx (#5726)

See discussion in #5600.

I'm also throwing in a pointer lifetime fix for the context held by
NVPTX because otherwise topi/tests/python/test_topi_softmax.py
would sefault for me. With the test, I can also run resnet-18 on
the nvptx target in gpu_imagenet_bench.py.
src/target/llvm/codegen_llvm.cc diff | blob | history
src/target/llvm/codegen_nvptx.cc diff | blob | history
topi/python/topi/cuda/softmax.py diff | blob | history
Domain: Machine Learning / ML Framework;