Avoid overlapping buffer writes in matmul verification #174
Conversation
This file contains 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
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
PeterTh
approved these changes
Apr 13, 2023
psalz
approved these changes
Apr 14, 2023
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Outdated suggestions cannot be applied.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.
Déjà vu!
#151 replaces a
bool &ref-capture withbuffer<bool>withaccess::allin the matmul verification code. The verification CGF submits a distributed host task though, and when ran on multiple nodes, it produces overlapping writes to the buffer.To keep distributed verification, this PR replaces the buffer with a host object (matmul even had a comment left explaining that different processes can disagree on their exit code if part of the verification fails!).
Also, since #163 moved matmul from a
buffer<bool>tobuffer<bool, 0>, I decided to givedistr_iothe 0-dimensional verification-buffer treatment so we still have an example covering this pattern. That application is sane since verification is done in a master-node host task, which can legally do an all-write access.