Cherry-pick ROCm fixes from PR #740 to rocm-jaxlib-v0.8.2#786
Open
srinivamd wants to merge 1 commit into
Open
Conversation
Cherry-pick 2 of 3 changes from #740 (merged to rocm-jaxlib-v0.9.2 on 2026-03-23). The third change (linalg_test tridiagonal_solve_grad skip) is already present on this branch. Changes: 1. Add dedicated ROCm lowering for scaled_matmul via lax.scaled_dot instead of the CUDA __op$block_scaled_dot custom call (upstream: jax-ml#35995) 2. Skip test_pmap on ROCm due to IndivisibleError from SPMD tiling incompatibility with 1D pmap mesh (upstream: jax-ml#35611) Fixes: ROCM-24925
Author
Cherry-pick contextThis PR cherry-picks 2 of 3 changes from #740 (merged ✅ to Changes included
Change NOT included (already present)
Sibling PRs
JiraROCM-24925 — JAX UT: 105 failures on sGpu/mGpu (P1) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Cherry-pick 2 of 3 changes from #740 (merged to
rocm-jaxlib-v0.9.2on 2026-03-23). The third change (linalg_test.pytest_tridiagonal_solve_gradskip) is already present onrocm-jaxlib-v0.8.2.Changes
jax/_src/cudnn/scaled_matmul_stablehlo.py— Add dedicated ROCm lowering forscaled_matmulvialax.scaled_dotinstead of the CUDA__op$block_scaled_dotcustom call. AMD's XLA backend does not support that custom call; routing throughlax.scaled_dotlets XLA emitkScaledDot, which ROCm can fuse via Triton or hipBLASLt. (Upstream: jax-ml/jax#35995)tests/ann_test.py— Skiptest_pmapon ROCm due toIndivisibleErrorfrom SPMD tiling incompatibility with 1D pmap mesh. (Upstream: jax-ml/jax#35611)Jira
Fixes ROCM-24925 — JAX UT: 105 failures on sGpu/mGpu
Source PR
#740 (merged ✅
dbc860f, 2026-03-23)