[TritonGEN] Use the sub-group-size of the module instead of hard code number of 16 in block load. #4764
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
chengjunlu
force-pushed
the
chengjun/tritongen_block_load_subgroupsize_32
branch
from
d865d15 to
56255da
Compare
… number of 16 in block load. Signed-off-by: Lu,Chengjun <[email protected]>
| @@ -238,7 +240,8 @@ verify2DBlockLoadHWRestriction(TritonGEN::Matrix2DBlockLoadOp op) { | |||
| VectorType resTy = op.getRes().getType(); | |||
| unsigned resElemTySize = resTy.getElementType().getIntOrFloatBitWidth(); | |||
| unsigned resSize = resTy.getNumElements() * resElemTySize; | |||
| constexpr unsigned subgroupSize = 16; | |||
| unsigned subgroupSize = triton::gpu::TritonGPUDialect::getThreadsPerWarp( | |||
There was a problem hiding this comment.
I am not sure we want to add a TritonGPU dependency to TritonGEN dialect.
Two alternative solutions I could think of are
- revert f0eed6f, and use the SPV target env.
- Assuming subgroupSize to be either 16 or 32, and only emit error if both are not satisfied.
There was a problem hiding this comment.
I see that we have already added the dependency in a earlier PR, so we can continue this approach, and reconsider the alternatives after this PR if preferred.
There was a problem hiding this comment.
Thanks. We can clean up the dependency later in cleaning up.
whitneywhtsang
approved these changes
Jul 24, 2025
|
Let's land this change first. |
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Outdated suggestions cannot be applied.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.
[TritonGEN] Use the sub-group-size of the module instead of hard code number of 16 in block load.