-
Notifications
You must be signed in to change notification settings - Fork 26.3k
Revert "Use custom CPU thread pool in async_scheduling (#12295)" #12410
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
This reverts commit 14b48a2.
facebook-github-bot
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
gchanan has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
Looks like the reverted commit broke the windows build. |
|
Windows build was broken because of unrelated to commit issue - linker falls back (because it runs out of memory) to another linker that didn't support a command line option. |
|
@peterjc123 / @yf225 have you seen the error on https://ci.pytorch.org/jenkins/job/pytorch-builds/job/pytorch-win-ws2016-cuda9-cudnn7-py3-build/23709//console before? In particular what @ilia-cher is referring to is this: is this an actual problem with the commit or do we need to change the build / build config? |
|
cc. @Yangqing |
|
I'm pretty sure it is a build script issue and not the commit, but I can accept and send a new pr and lets recheck everything |
|
leaving this open for discussion, but don't commit because phabricator diff isn't valid. |
ssnl
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If the build script can't be fixed soon, let's unbreak build first.
|
I'm a little confused by the discussion here, because the CI results seem to claim the Windows builds passed? |
|
@ezyang I guess that this PR reverts the change that may cause the aforementioned error. Let's switch the default to the x64 toolchain and then it'll be fine. |
Isn't a build error. The build will succeed, but it will take longer to link, as it's basically trying to link with the 32-bit hosted linker, which fails because the 32-bit address space is too small to be able to link successfully, so it runs the 64-bit hosted linker which then links successfully. TLDR; The x86_amd64 MSVC toolchain is 32-bit binaries cross-compiling for amd64, and if the linker runs out of memory, it starts the linker from the amd64 (aka. amd64_amd64) toolchain, which is a 64-bit binary and has enough memory to complete. |
This reverts commit 14b48a2.