Shorten splash attention kernel name to avoid 255 char limit

[patrick-toulme]

Fix for #34117

[gemini-code-assist]

Summary of Changes

Hello @patrick-toulme, 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 addresses an issue where Splash Attention kernel names could exceed the 255-character filename limit. The changes introduce a more concise naming convention by abbreviating certain flags and utilizing an MD5 hash for the block_metadata, ensuring that generated kernel names remain within acceptable length constraints while maintaining uniqueness.

Highlights

• Kernel Name Shortening: The get_kernel_name function has been updated to generate shorter names for Splash Attention kernels to avoid exceeding filename length limits.
• Metadata Hashing: block_metadata is now hashed using MD5 to produce an 8-character identifier, significantly reducing the length of the kernel name while maintaining uniqueness.
• Suffix Abbreviation: Suffixes like _residuals and _segmented have been abbreviated to _res and _seg respectively to further shorten the generated kernel names.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

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 by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

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 pull request 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. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

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 shortens the kernel name for SplashAttention to avoid exceeding the 255-character filename limit. It achieves this by hashing the block metadata instead of including it as a long string in the name. While the approach is sound, I've identified a potential issue with the chosen hash function and its truncation, which could lead to hash collisions and silent correctness bugs. My review includes a suggestion to use a more secure hash function to mitigate this risk.

[rdyro]

This looks like a good idea! Can you take a look at https://github.com/openxla/tokamax/blob/7f078e4545f00ecdac6e7d560c31e09e69f49c5d/tokamax/_src/ops/experimental/tpu/splash_attention/splash_attention_kernel.py#L100 - this is a fork of this version with the kernel name shortened.

The block information is still useful to have, but it's possible to move it to the metadata field instead: https://github.com/openxla/tokamax/blob/7f078e4545f00ecdac6e7d560c31e09e69f49c5d/tokamax/_src/ops/experimental/tpu/splash_attention/splash_attention_kernel.py#L997

What do you think of adopting a similar strategy here?