[Bugfix] Remove redundant T.fill to fix precision issue

[xuchangtolearn]

No description provided.

[github-actions]

👋 Hi! Thank you for contributing to the TileLang project.

Please remember to run bash format.sh in the root directory of the project to ensure your changes are properly linted and formatted. This will help ensure your contribution passes the format check.

We appreciate you taking this step! Our team will review your contribution, and we look forward to your awesome work!

🚀

[gemini-code-assist]

Summary of Changes

Hello @xuchangtolearn, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request resolves a numerical precision bug in the Flash Attention decoding example by removing a redundant initialization step for a shared memory buffer. This change ensures more accurate computations within the attention mechanism.

Highlights

• Precision Bugfix: This pull request addresses and fixes a reported precision issue within the flash_attn_split function in the Flash Attention decoding example.
• Redundant Code Removal: An unnecessary T.fill(K_shared, 0) operation, which was contributing to the precision problem, has been removed.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request removes a line of code, T.fill(K_shared, 0), from the flash_attn_split macro.

My analysis of the surrounding code confirms that this operation is redundant. The K_shared buffer, which is initialized by this T.fill call, is used in a T.gemm operation. However, the parts of the buffer that might not be overwritten by a subsequent T.copy operation correspond to padded elements. The results of the T.gemm for these padded elements are correctly masked out and replaced with -T.infinity before they can affect the final result.

Therefore, removing this T.fill is a valid optimization that cleans up the code and may offer a minor performance benefit by eliminating an unnecessary shared memory write. The pull request title indicates this also resolves a precision issue, which is plausible given the complexities of JIT compilation and floating-point optimizations. The change is sound and improves the code.