Adds note on adding extra compile args for building cuda extensions #292
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
When building CUDA extensions, we have to pass `extra_compile_args` now to avoid cuda header collisions with half operator overloading (that happens through implicit half conversions) details: pytorch/pytorch#10301 (comment).
goldsborough
suggested changes
Sep 7, 2018
There was a problem hiding this comment.
Hi, @syed-ahmed, thanks for the contribution. I feel it may be a little unfriendly for our users to have to pass these flags manually whenever building a CUDA extension with half support. We generally try very hard to hide as much compiler configuration as possible for these extensions. I will come up with a way of automatically adding these flags from within CUDAExtension.
|
Thanks @goldsborough ! I totally agree with you! I'll close this PR since you got this. Till then, people can just refer to pytorch/pytorch#10301 (comment) if they get the operator overloading errors. |
facebook-github-bot
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Sep 10, 2018
Summary: The controller you requested could not be found. found there are some issues when using comparison operators for half types when certain THC header are included. I was able to reproduce and added a test. I also fix the issue by adding the proper definitions to avoid this issue. Reported in #10301 (comment) Related: pytorch/tutorials#292 soumith fmassa Pull Request resolved: #11395 Differential Revision: D9725102 Pulled By: goldsborough fbshipit-source-id: 630425829046bbebea3409bb792a9d62c91f41ad
PenghuiCheng
pushed a commit
to PenghuiCheng/pytorch
that referenced
this pull request
Sep 11, 2018
Summary: The controller you requested could not be found. found there are some issues when using comparison operators for half types when certain THC header are included. I was able to reproduce and added a test. I also fix the issue by adding the proper definitions to avoid this issue. Reported in pytorch#10301 (comment) Related: pytorch/tutorials#292 soumith fmassa Pull Request resolved: pytorch#11395 Differential Revision: D9725102 Pulled By: goldsborough fbshipit-source-id: 630425829046bbebea3409bb792a9d62c91f41ad
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When building CUDA extensions, we have to pass
extra_compile_argsnow to avoid cuda header collisions with half operator overloading (that happens through implicit half conversions) details: pytorch/pytorch#10301 (comment).