[tritonbench][AutoWS] Enable AutoWS matmul kernels on Hopper#1025
Open
njriasan wants to merge 1 commit intometa-pytorch:mainfrom
Open
[tritonbench][AutoWS] Enable AutoWS matmul kernels on Hopper#1025njriasan wants to merge 1 commit intometa-pytorch:mainfrom
njriasan wants to merge 1 commit intometa-pytorch:mainfrom
Conversation
Summary:
Extends the Blackwell-only TMA / persistent / descriptor matmul benchmarks
in `tritonbench` to also run on Hopper.
Changes:
1. Widen `enabled=IS_BLACKWELL` to `enabled=IS_BLACKWELL or IS_HOPPER` for
the 6 `triton_blackwell_*` benchmarks in `operator.py`. The underlying
kernels (`matmul_kernel_tma`, `matmul_kernel_tma_persistent`, and
`matmul_kernel_descriptor_persistent`) are arch-agnostic, so they run on
Hopper once the gates allow it.
2. Update `_prune_tma_persistent_configs` in `warp_spec_persistent_matmul.py`
so the `FLATTEN` requirement is computed correctly across {Hopper,
Blackwell} x {meta WS, OAI Triton WS, no WS}. `tl.range(flatten=True)` is
only supported on Blackwell with the OAI Triton WS path; on Hopper we now
require `FLATTEN=False` for both AutoWS (meta WS) and OAI Triton paths.
Differential Revision: D101840063
njriasan
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
njriasan
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
njriasan
had a problem deploying
to
docker-s3-upload
GitHub Actions
Failure
|
@njriasan has exported this pull request. If you are a Meta employee, you can view the originating Diff in D101840063. |
njriasan
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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:
Extends the Blackwell-only TMA / persistent / descriptor matmul benchmarks
in
tritonbenchto also run on Hopper.Changes:
Widen
enabled=IS_BLACKWELLtoenabled=IS_BLACKWELL or IS_HOPPERforthe 6
triton_blackwell_*benchmarks inoperator.py. The underlyingkernels (
matmul_kernel_tma,matmul_kernel_tma_persistent, andmatmul_kernel_descriptor_persistent) are arch-agnostic, so they run onHopper once the gates allow it.
Update
_prune_tma_persistent_configsinwarp_spec_persistent_matmul.pyso the
FLATTENrequirement is computed correctly across {Hopper,Blackwell} x {meta WS, OAI Triton WS, no WS}.
tl.range(flatten=True)isonly supported on Blackwell with the OAI Triton WS path; on Hopper we now
require
FLATTEN=Falsefor both AutoWS (meta WS) and OAI Triton paths.Differential Revision: D101840063