[Instruction] Add .read modifier to cp.async.bulk.wait_group#106
Merged
yaoyaoding merged 2 commits intomainfrom Apr 2, 2026
Merged
[Instruction] Add .read modifier to cp.async.bulk.wait_group#106yaoyaoding merged 2 commits intomainfrom
yaoyaoding merged 2 commits intomainfrom
Conversation
Member
Author
|
/ok to test 58c9682 |
yaoyaoding
force-pushed
the
b200-gemm-opt
branch
from
58c9682 to
6dc4a60
Compare
In the epilogue of shared-to-global TMA stores, the default cp.async.bulk.wait_group inserts an unnecessary L1 cache invalidation. The .read variant only waits for source reads to complete, which is sufficient when reusing shared memory buffers without subsequent global loads of the TMA-written data. Changes across the full stack: - IR: add `read` field to CopyAsyncTensorWaitGroupInst - Hidet primitive: register `_read` variants emitting `.read` PTX - Emitter: pass `read` through to primitive - Builder/Lang: expose `read` param on wait_group() - Example: use read=True in matmul_v9 epilogue - mbarrier: reduce try_wait ticks from 10M to 50K to match nvjet Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: Yaoyao Ding <dingyaoyao.cs@gmail.com>
Signed-off-by: Yaoyao Ding <dingyaoyao.cs@gmail.com>
yaoyaoding
force-pushed
the
b200-gemm-opt
branch
from
6dc4a60 to
41e7cc2
Compare
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
In the epilogue of shared-to-global TMA stores, the default cp.async.bulk.wait_group inserts an unnecessary L1 cache invalidation. The .read variant only waits for source reads to complete, which is sufficient when reusing shared memory buffers without subsequent global loads of the TMA-written data.
Changes across the full stack:
readfield to CopyAsyncTensorWaitGroupInst_readvariants emitting.readPTXreadthrough to primitivereadparam on wait_group()