Fix function naming and standardize memory copy alignment in attention kernel #70
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
Corrects function calls from copy_Mask to copy_MN throughout the kernel implementation and fixes spelling error in comment from "Golobal" to "Global". These changes ensure proper function naming consistency and improve code readability.
Changes function name from copy_Mask to copy_MN to improve code clarity and better indicate the function's role in copying matrix dimensions.
Unifies memory copy alignment to 128 bytes for both mask and bias operations in forward and backward kernel traits. Removes inconsistent 64-byte alignment for bias copy atom in backward kernel and establishes consistent 128-byte alignment across all mask and bias copy operations.
Replaces incorrect SmemCopyAtomO with proper SmemCopyAtomMask and SmemCopyAtomBias types in tiled copy operations. Ensures type safety and correct memory access patterns for mask and bias tensors in attention computation kernels.
LoserCheems
requested review from
Evanwu1125,
SNHuan,
Copilot and
wubingheng111
and removed request for
Copilot
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR improves code clarity and consistency in the attention kernel implementation by renaming a function for better semantic meaning, correcting a spelling error, and standardizing memory copy alignment parameters for mask and bias operations.
- Renamed
copy_Maskfunction tocopy_MNto better reflect its purpose of copying matrix dimensions - Fixed spelling error from "Golobal" to "Global" in a comment
- Unified memory copy alignment from mixed 64/128-bit to consistent 128-bit alignment for both mask and bias operations
Reviewed Changes
Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
| csrc/src/utils.h | Renamed function from copy_Mask to copy_MN for better semantic clarity |
| csrc/src/kernel_traits.h | Added consistent 128-bit aligned copy atoms for mask and bias operations, removed duplicate definition |
| csrc/src/flash_fwd_kernel.h | Updated function calls to use new name, fixed spelling error, and applied new copy atom types |
| ); | ||
|
|
||
| // Golobal to Shared Memory operation | ||
| // Global to Shared Memory operation |
There was a problem hiding this comment.
Fixed spelling error from 'Golobal' to 'Global' in the comment.
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.
Correct function names for clarity, fix typos, and unify memory copy alignment for mask and bias operations to enhance code readability and type safety.