-
Notifications
You must be signed in to change notification settings - Fork 24.2k
Fix recent build error on ppc64le #129736
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/129736
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 360b2b2 with merge base f0da167 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
@@ -38,6 +38,7 @@ inline void _scale_attn_mask_fusion_kernel( | |||
constexpr int64_t T2_n = 1; | |||
auto vec_scale = at::vec::VectorizedN<T1, T1_n>(val); | |||
int64_t i = 0; | |||
#if !defined(__powerpc__) |
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.
won't think make this code silently wrong though?
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.
Hi @albanD This is temporary fix.
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.
Can you make it an error in the else just to make sure we don't have silent correctness issue?
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.
Could you explain why you think this PR will make it silently wrong
?
Looks like the change will make PowerPC use only scalar operations and skip the vector operations. The end result is the PPC will just be slower for this function but still correct.
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.
Ho sorry, I didn't realized the loop below continues on the same i and not only does the tail. Sorry about that!
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.
Sounds good!
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.
Ho super cool, thanks for taking the time to do this!
@pytorchbot merge |
Merge failedReason: This PR needs a If not, please add the To add a label, you can comment to pytorchbot, for example For more information, see Details for Dev Infra teamRaised by workflow job |
@pytorchbot label "topic: not user facing" |
Hi @albanD could you please help me with merge |
@malfet @soulitzer Could you please help me with merge. |
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Merge failedReason: 1 mandatory check(s) failed. The first few are: Dig deeper by viewing the failures on hud |
@pytorchbot rebase |
@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here |
…lashAttentionKernel.cpp
Successfully rebased |
265f4af
to
360b2b2
Compare
@albanD @malfet @@soulitzer After rebase there are no CI test failure. Sorry for inconvenience can you please try merging it again. |
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
@pratiklp00 please read the instructions in #132400 and add appropriate comment if it qualifies a cherry-pick criteria |
@malfet Thank you. |
@pytorchbot cherry-pick --onto release/2.4 -c critical |
This PR will fix the recent build issue observed on ppc64le. Fixes #128130 Pull Request resolved: #129736 Approved by: https://github.com/albanD, https://github.com/malfet (cherry picked from commit 69cbf05)
Cherry picking #129736The cherry pick PR is at #133416 and it is recommended to link a critical cherry pick PR with an issue. The following tracker issues are updated: Details for Dev Infra teamRaised by workflow job |
Fix recent build error on ppc64le (#129736) This PR will fix the recent build issue observed on ppc64le. Fixes #128130 Pull Request resolved: #129736 Approved by: https://github.com/albanD, https://github.com/malfet (cherry picked from commit 69cbf05) Co-authored-by: pratiklp00 <pratikp@linux.ibm.com>
Fix recent build error on ppc64le (pytorch#129736) This PR will fix the recent build issue observed on ppc64le. Fixes pytorch#128130 Pull Request resolved: pytorch#129736 Approved by: https://github.com/albanD, https://github.com/malfet (cherry picked from commit 69cbf05) Co-authored-by: pratiklp00 <pratikp@linux.ibm.com>
This PR will fix the recent build issue observed on ppc64le.
Fixes #128130
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10