While raising: broadcast under-attributed yields to the carried layout - #3002
Open
wsmoses wants to merge 1 commit into
Open
While raising: broadcast under-attributed yields to the carried layout#3002wsmoses wants to merge 1 commit into
wsmoses wants to merge 1 commit into
Conversation
A loop-carried value can be yielded with fewer attributed axes than the carried argument (a uniform chain through an index-table gather loses its lane attribution), which made the permutation matching reject the loop. Broadcast the yield up to the carried layout first, and let alignMemoryAccess report incompatibility through an ok flag instead of asserting on accesses it cannot size. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD
wsmoses
force-pushed
the
pb/yield-broadcast
branch
from
August 28, 2026 14:12
7f63d95 to
cad826a
Compare
Pangoraw
reviewed
Aug 31, 2026
Comment on lines
1537
to
1538
| auto perm = memoryEquivalentPermutation(maps.lookup(raisedYieldedIterArg), | ||
| maps.lookup(raisedIterArg)); |
Collaborator
There was a problem hiding this comment.
i would change the memoryEquivalentPermutation here to be more permissive then since it's only used for this case.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
mfem's
bilininteg_diffusion_kernels.cppaccumulators chain through index-table gathers whose raised values lose their lane attribution (a() -> ()map with a tensor type); the while raising then rejected the loop because the yield's layout could not be permutation-matched to the carried argument. Broadcast the yield up to the carried layout viaalignMemoryAccessbefore matching, and letalignMemoryAccessreport incompatibility through anokout-flag instead of asserting on accesses it cannot size.The tridiagonal golden changes form: the align at the loop tail now emits a permuting
broadcast_in_dim(dims = [0, 3, 1, 2]) instead of a transpose — equivalent IR.Part of #2968.
🤖 Generated with Claude Code
https://claude.ai/code/session_016zErYp7upmqr4NHfhod9UD