[Fix][TIR] LowerThreadAllreduce with correct thread mask #15323
Merged
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
Collaborator
|
Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.
Generated by tvm-bot |
Contributor
Author
|
cc @yzh119 |
This PR fixes a bug in the LowerThreadAllreduce pass. Prior to this PR, in multi-group settings, the thread mask is not correctly set: when the reduction extent is 32, the thread mask will always be 0. This bug was not spotted because even when the mask is 0, the CUDA program still gives correct result. But in any way, having the zero mask is dangerous and should be fixed.
MasterJH5574
force-pushed
the
tvm-dev/2023-07-14-allreduce-mask-fix
branch
from
437e6ff to
c7cb4ac
Compare
yzh119
approved these changes
Jul 14, 2023
Member
yzh119
left a comment
yzh119
left a comment
There was a problem hiding this comment.
Thank @MasterJH5574 for spotting the bug!
junrushao
approved these changes
Jul 15, 2023
junrushao
pushed a commit
to junrushao/tvm
that referenced
this pull request
Jul 24, 2023
This PR fixes a bug in the LowerThreadAllreduce pass. Prior to this PR, in multi-group settings, the thread mask is not correctly set: when the reduction extent is 32, the thread mask will always be 0. This bug was not spotted because even when the mask is 0, the CUDA program still gives correct result. But in any way, having the zero mask is dangerous and should be fixed.
junrushao
pushed a commit
to junrushao/tvm
that referenced
this pull request
Jul 27, 2023
This PR fixes a bug in the LowerThreadAllreduce pass. Prior to this PR, in multi-group settings, the thread mask is not correctly set: when the reduction extent is 32, the thread mask will always be 0. This bug was not spotted because even when the mask is 0, the CUDA program still gives correct result. But in any way, having the zero mask is dangerous and should be fixed.
junrushao
pushed a commit
to junrushao/tvm
that referenced
this pull request
Jul 30, 2023
This PR fixes a bug in the LowerThreadAllreduce pass. Prior to this PR, in multi-group settings, the thread mask is not correctly set: when the reduction extent is 32, the thread mask will always be 0. This bug was not spotted because even when the mask is 0, the CUDA program still gives correct result. But in any way, having the zero mask is dangerous and should be fixed.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
This PR fixes a bug in the LowerThreadAllreduce pass.
Prior to this PR, in multi-group settings, the thread mask is not correctly set: when the reduction extent is 32, the thread mask will always be 0. This bug was not spotted because even when the mask is 0, the CUDA program still gives correct result. But in any way, having the zero mask is dangerous and should be fixed.