test(async_ll): probe SDMA path via MORI_ENABLE_SDMA in CI#313
Closed
jhchouuu wants to merge 4 commits into
Closed
test(async_ll): probe SDMA path via MORI_ENABLE_SDMA in CI#313jhchouuu wants to merge 4 commits into
jhchouuu wants to merge 4 commits into
Conversation
Replace MORI_DISABLE_P2P=1 (which routes async_ll over RDMA / IBGDA) with MORI_ENABLE_SDMA=1 so the test exercises the SDMA intra-node path instead. Probes whether CI can run the async_ll kernel end-to-end over SDMA.
…rdingly Switches CI BASE_IMAGE to rocm/pytorch:rocm7.2.3_ubuntu24.04_py3.12_pytorch_release_2.10.0 and renames the built CI image to rocm/mori:ci_rocm723 so the rocm version is visible in the tag.
The previously selected rocm7.2.3 base image is unavailable, so pin to rocm/pytorch:rocm7.2.2_ubuntu24.04_py3.12_pytorch_release_2.10.0 and rename the built CI image to rocm/mori:ci_rocm722 so the rocm version is visible in the tag.
Forces the intranode async_ll pytest step to run over the SDMA transport instead of the default P2P path, so the SDMA branch in symmetric_memory.cpp's signal-pointer exchange is exercised in CI.
3 tasks
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Switches
tests/python/ops/test_dispatch_combine_async_ll.pyfromMORI_DISABLE_P2P=1toMORI_ENABLE_SDMA=1so the async_ll kernel runs over the intra-node SDMA transport instead of the RDMA / IBGDA path.This is a probe PR — the goal is to see whether CI can run the async_ll kernel end-to-end over SDMA on the current hardware/setup.
Why
The previous setting (
MORI_DISABLE_P2P=1) gatesContext::InitializePossibleTransportsso same-host peers fall through to the RDMA branch. WithMORI_ENABLE_SDMA=1(and P2P NOT disabled),context.cppinstead picksTransportType::SDMAfor same-host peers and exercises the SDMA signal-pointer exchange insymmetric_memory.cpp.Test plan
tests/python/ops/test_dispatch_combine_async_ll.pyand reports pass/fail under SDMA