Repository navigation
[FlyDSL] Buffer ops refactoring - #5146
Closed
amd-nprotaso wants to merge 7 commits into
Closed
amd-nprotaso wants to merge 7 commits into
amd-nprotaso wants to merge 7 commits into
Conversation
Contributor
🏷️ CI GuideRuns automatically on every PR:
Extended tests (opt-in via labels):
PR title tags & labels: |
jli-melchior
reviewed
Sep 3, 2026
Conflict resolution: - buffer_ops.py: took main's removal of extract_base_index (it has no remaining callers in-tree) on top of this branch's rewrite. main had also dropped BufferResourceDescriptor and extract_base_index from __all__; that hunk merged cleanly. - small_m_hgemm.py / splitk_hgemm.py: accepted main's deletion. Both were replaced by kernels/gemm_a16w16_gfx950.py in ROCm#5145; this branch only touched them to thread rocdl.get_buffer_rsrc() through the raw raw_ptr_buffer_load_lds call sites, so the change dies with the files. The replacement uses fx.rocdl.make_buffer_tensor and needs no port. Also ran black on buffer_ops.py, which was not format-clean. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
zufayu
added a commit
that referenced
this pull request
Sep 7, 2026
…y looked right Pointing the FlyDSL collectors at $WORK/head was correct and insufficient: that tree was populated only inside `if grep -q "invariant-removed\|api-signature"`. On a PR deriving neither family it is an empty directory, and a collector that reads files finds nothing to read. #5207 has six buffer calls in its diff -- five bounded, one a bare `make_buffer_tensor(W_scale, max_size=True)` -- and flydslbounds returned zero. #5301, which the previous commit was verified against, deletes a guard and so derives invariant-removed. That is the only reason the fix appeared to work. A fixture cannot see this: every test here builds its own root and hands it over, so the tree is always populated. Materialising head is now unconditional; only the `evidence` call stays behind the family test. The cost is a few `git show` calls against objects already fetched at line 125. On #5207 B8 now reports exactly one candidate and it is worth asking about: `W_scale` is bound with `num_records_bytes=n_heads * self.n_per_head * 4` at line 184 of the same file, and with `max_size=True` at line 92, inside a BlockScale whose constructor already receives head and n_scale_cols. Same tensor, two treatments, one of them unbounded. Also verified silent for the right reason on #5146 (its one added buffer call is buffer_ops.py forwarding variables, not a kernel binding a tensor) and on #5282 (no buffer construction in the diff at all).
…so/flydsl_buffer_ops_refactor # Conflicts: # aiter/ops/flydsl/kernels/fmha_gfx1250/fmha_kernel.py
Author
|
Kernels will be refactor to use layout API, instead of use low level buffer operations |
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.
Replace old rocdl api ussage to new fx.