[TritonNVIDIAGPU] Add MemRead<GlobalMemory> to async TMA write ops
#6485
Conversation
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
peterbell10
approved these changes
Apr 14, 2025
Mogball
added a commit
that referenced
this pull request
Apr 16, 2025
The first PR in this revert stack is somehow changing bitwise equivalence. To unblock work, I am reverting them and relanding them with better granularity. Revert "[Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (#6487)" This reverts commit 96fc40d. Revert "[TritonNVIDIAGPU] Add `MemRead<GlobalMemory>` to async TMA write ops (#6485)" This reverts commit 31b2b23. Revert "[TritonNVIDIAGPU] Revert MMAv5 write effect on barrier (#6484)" This reverts commit f8a19d1. Revert "[Dialect] Cleanup and improve granularity of side effects (#6476)" This reverts commit f60465e.
Mogball
added a commit
that referenced
this pull request
Apr 16, 2025
Reland of #6485 All async TMA ops read the descriptor. MMAScaled is missing read effects on the scale arguments, which are always in TMEM.
Mogball
added a commit
that referenced
this pull request
Apr 17, 2025
The first PR in this revert stack is somehow changing bitwise equivalence. To unblock work, I am reverting them and relanding them with better granularity. Revert "[Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (#6487)" This reverts commit 96fc40d. Revert "[TritonNVIDIAGPU] Add `MemRead<GlobalMemory>` to async TMA write ops (#6485)" This reverts commit 31b2b23. Revert "[TritonNVIDIAGPU] Revert MMAv5 write effect on barrier (#6484)" This reverts commit f8a19d1. Revert "[Dialect] Cleanup and improve granularity of side effects (#6476)" This reverts commit f60465e.
jtang10
pushed a commit
to ROCm/triton
that referenced
this pull request
Apr 17, 2025
…ang#6476 (triton-lang#6512) The first PR in this revert stack is somehow changing bitwise equivalence. To unblock work, I am reverting them and relanding them with better granularity. Revert "[Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (triton-lang#6487)" This reverts commit 96fc40d. Revert "[TritonNVIDIAGPU] Add `MemRead<GlobalMemory>` to async TMA write ops (triton-lang#6485)" This reverts commit 31b2b23. Revert "[TritonNVIDIAGPU] Revert MMAv5 write effect on barrier (triton-lang#6484)" This reverts commit f8a19d1. Revert "[Dialect] Cleanup and improve granularity of side effects (triton-lang#6476)" This reverts commit f60465e.
njriasan
pushed a commit
to njriasan/triton
that referenced
this pull request
Apr 18, 2025
njriasan
pushed a commit
to njriasan/triton
that referenced
this pull request
Apr 18, 2025
…ang#6476 (triton-lang#6512) The first PR in this revert stack is somehow changing bitwise equivalence. To unblock work, I am reverting them and relanding them with better granularity. Revert "[Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (triton-lang#6487)" This reverts commit 96fc40d. Revert "[TritonNVIDIAGPU] Add `MemRead<GlobalMemory>` to async TMA write ops (triton-lang#6485)" This reverts commit 31b2b23. Revert "[TritonNVIDIAGPU] Revert MMAv5 write effect on barrier (triton-lang#6484)" This reverts commit f8a19d1. Revert "[Dialect] Cleanup and improve granularity of side effects (triton-lang#6476)" This reverts commit f60465e.
Mogball
added a commit
that referenced
this pull request
Apr 23, 2025
Reland of #6485 All async TMA ops read the descriptor. MMAScaled is missing read effects on the scale arguments, which are always in TMEM.
FindHao
pushed a commit
to FindHao/triton
that referenced
this pull request
Apr 30, 2025
…ang#6476 (triton-lang#6512) The first PR in this revert stack is somehow changing bitwise equivalence. To unblock work, I am reverting them and relanding them with better granularity. Revert "[Blackwell] Optimize MMA warp specialization to allow multiple consumers of MMAv5 result (triton-lang#6487)" This reverts commit 96fc40d. Revert "[TritonNVIDIAGPU] Add `MemRead<GlobalMemory>` to async TMA write ops (triton-lang#6485)" This reverts commit 31b2b23. Revert "[TritonNVIDIAGPU] Revert MMAv5 write effect on barrier (triton-lang#6484)" This reverts commit f8a19d1. Revert "[Dialect] Cleanup and improve granularity of side effects (triton-lang#6476)" This reverts commit f60465e.
FindHao
pushed a commit
to FindHao/triton
that referenced
this pull request
Apr 30, 2025
…#6518) Reland of triton-lang#6485 All async TMA ops read the descriptor. MMAScaled is missing read effects on the scale arguments, which are always in TMEM.
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.
No description provided.