v4.1.x: gcc_builtin: fix performance regression on x86_64#8623
Merged
jsquyres merged 1 commit intoopen-mpi:v4.1.xfrom Mar 16, 2021
Merged
Conversation
in order to work around a bug in older gcc versions on x86_64, __atomic_thread_fence (__ATOMIC_SEQ_CST) was replaced with __atomic_thread_fence (__ATOMIC_ACQUIRE) based on the asumption that this did not introduce performance regressions. It was recently found that this did introduce some performance regression, mainly at scale on fat nodes. So simply use an asm memory globber to both workaround older gcc bugs and fix the performance regression. Thanks S. Biplab Raut for bringing this issue to our attention. Refs. open-mpi#8603 Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp> (cherry picked from commit d7e3f87)
ggouaillardet
force-pushed
the
topic/v4.1.x/gcc_builtin_workaround
branch
from
0ca05eb to
0c02983
Compare
jsquyres
changed the title
Member
|
bot:aws:retest |
jsquyres
approved these changes
Mar 16, 2021
Member
jsquyres
left a comment
jsquyres
left a comment
There was a problem hiding this comment.
Merge by proxy from the other release branches, and from discussion on the 16 Mar 2021 Webex.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
in order to work around a bug in older gcc versions on x86_64,
__atomic_thread_fence (__ATOMIC_SEQ_CST)
was replaced with
__atomic_thread_fence (__ATOMIC_ACQUIRE)
based on the asumption that this did not introduce performance regressions.
It was recently found that this did introduce some performance regression,
mainly at scale on fat nodes.
So simply use an asm memory globber to both workaround older gcc bugs
and fix the performance regression.
Thanks S. Biplab Raut for bringing this issue to our attention.
Refs. #8603
Signed-off-by: Gilles Gouaillardet gilles@rist.or.jp
(cherry picked from commit d7e3f87)