Modified block and warp sizes for improved performance on XPU for both layernnorm and rmsnorm #661
Conversation
This file contains 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
shivam15s
reviewed
Apr 11, 2025
| f"Feature dimension {n_cols} exceeds maximum supported size of {BLOCK_SIZE}. Consider using a smaller feature dimension." | ||
| ) | ||
| if X.device.type == "xpu": # XPU-specific optimization | ||
| BLOCK_SIZE = torch.xpu.get_device_properties(X.device).max_work_group_size |
There was a problem hiding this comment.
afaik we want the block size here to be triton.next_power_of_2(n). torch.xpu.get_device_properties(X.device).max_work_group_size is supposedly giving the max possible block size on the xpu device. Better way to handle this is to change the calculate_settings function directly where you set the MAX_FUSED_SIZE according to the device.
There was a problem hiding this comment.
@shivam15s These changes are specific for the layernorm and rmsnorm.That's why i made the changes locally and this will impact other kernels.
shivam15s
reviewed
Apr 11, 2025
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Outdated suggestions cannot be applied.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.
Summary
This change is related to performance tuning on the Intel Max 1550 GPUs. By keeping the block and warp sizes the same in the forward and backward Triton kernels.
Testing Done
make testto ensure correctnessmake checkstyleto ensure code stylemake test-convergenceto ensure convergence