saymrwulf/pytorch
1
0
Fork 0
mirror of https://github.com/saymrwulf/pytorch.git synced 2026-05-15 21:00:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Revert "Disable flaky test TestCppExtensionAOT.test_cuda_extension in… (#33404)

Browse source 
Summary:
… Windows CI (https://github.com/pytorch/pytorch/issues/33282)"

This reverts commit 5b922918d0.

Fixes https://github.com/pytorch/pytorch/issues/33270.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/33404

Differential Revision: D19972594

Pulled By: ezyang

fbshipit-source-id: c8f67536fd6e4b7135171d621ad671b1b2a21fd4
This commit is contained in:
peter 2020-02-20 09:06:26 -08:00 committed by Facebook Github Bot
parent 05fb160048
commit ffe327f7d9
2 changed files with 6 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
test/test_cpp_extensions_aot.py
View file

@ -55,7 +55,6 @@ class TestCppExtensionAOT(common.TestCase):
self.assertEqual(tensor.grad, expected_tensor_grad)
@unittest.skipIf(not TEST_CUDA, "CUDA not found")
@unittest.skipIf(IS_WINDOWS, "Flaky on Windows, see issue #33270")
def test_cuda_extension(self):
import torch_test_cpp_extension.cuda as cuda_extension

15
torch/utils/cpp_extension.py
View file

@ -454,13 +454,13 @@ class BuildExtension(build_ext, object):
include_dirs, sources,
depends, extra_postargs)
common_cflags = extra_preargs or []
common_cflags.append('/c')
cflags = []
if debug:
common_cflags.extend(self.compiler.compile_options_debug)
cflags.extend(self.compiler.compile_options_debug)
else:
common_cflags.extend(self.compiler.compile_options)
cflags.extend(self.compiler.compile_options)
common_cflags.extend(COMMON_MSVC_FLAGS)
cflags = common_cflags + pp_opts
cflags = cflags + common_cflags + pp_opts
with_cuda = any(map(_is_cuda_file, sources))
# extra_postargs can be either:
@ -477,11 +477,8 @@ class BuildExtension(build_ext, object):
if with_cuda:
cuda_cflags = []
for common_cflag in common_cflags:
if common_cflag == '/c':
cuda_cflags.append('-c')
else:
cuda_cflags.append('-Xcompiler')
cuda_cflags.append(common_cflag)
cuda_cflags.append('-Xcompiler')
cuda_cflags.append(common_cflag)
cuda_cflags.extend(pp_opts)
if isinstance(extra_postargs, dict):
cuda_post_cflags = extra_postargs['nvcc']