[AMDGPU] Handle empty-except-for-DI regions in PreRARematerialize #524
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
The existing check for this case only comes after a derefence of what can be an iterator sentinel (leading to an assert). This may not be purely NFC in that it also avoids queuing the effectively-empty region for rescheduling, but AFAICT this should be purely an optimization. Testing this seems difficult, as the high-level scheduler avoids scheduling these "empty" regions. This means a reproducer has to depend on behavior of the scheduler passes before PreRARematStage in order to craft a region which triggers the bug. Since this is a release blocker I am posting a PR now, as both Shore Shen and I have manually verified that this resolves the particular crash from SWDEV-564142 but I am still working on making a reasonable test. (cherry picked from commit 004cfea)
Collaborator
slinder1
changed the title
slinder1
changed the title
bcahoon
approved these changes
Nov 7, 2025
|
Testcase? |
|
Why is this direct to mainline? |
Collaborator
|
!PSDB |
Collaborator
ronlieb
deleted the
amd/dev/slinder1/amd-mainline/SWDEV-563725_rematerialize_debug_info
branch
Author
Originally merged into amd-staging @ #516 and I intend to go upstream, but as you note I don't have a reasonable testcase yet. The issue with building a test is that this affects a phase of the scheduler in a way that depends on the details of earlier phases. There is also a lot of IR to churn through trying to get to a reproducer, and naively running I'm still working on it, though |
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Outdated suggestions cannot be applied.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.
The existing check for this case only comes after a derefence of what can be an iterator sentinel (leading to an assert).
This may not be purely NFC in that it also avoids queuing the effectively-empty region for rescheduling, but AFAICT this should be purely an optimization.
Testing this seems difficult, as the high-level scheduler avoids scheduling these "empty" regions. This means a reproducer has to depend on behavior of the scheduler passes before PreRARematStage in order to craft a region which triggers the bug.
Since this is a release blocker I am posting a PR now, as both Shore Shen and I have manually verified that this resolves the particular crash from SWDEV-564142 but I am still working on making a reasonable test.
(cherry picked from commit 004cfea)