refactor: reuse workspace for bmm_fp8

[ttyio]

📌 Description

refactor: reuse workspace for bmm_fp8

🔍 Related Issues

🚀 Pull Request Checklist

Thank you for contributing to FlashInfer! Before we review your pull request, please make sure the following items are complete.

✅ Pre-commit Checks

• I have installed pre-commit by running pip install pre-commit (or used your preferred method).
• I have installed the hooks with pre-commit install.
• I have run the hooks manually with pre-commit run --all-files and fixed any reported issues.

If you are unsure about how to set up pre-commit, see the pre-commit documentation.

🧪 Tests

• Tests have been added or updated as needed.
• All tests are passing (unittest, etc.).

Reviewer Notes

[gemini-code-assist]

Summary of Changes

Hello @ttyio, 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!

I've refactored the bmm_fp8 function to improve memory efficiency by reusing a pre-allocated workspace buffer for cuDNN operations. This change ensures that the workspace for bmm_fp8 is managed more consistently across different backends, reducing redundant memory allocations.

Highlights

• Explicit Workspace Passing: The execute_cudnn_gemm_with_per_tensor_q_graph function now explicitly accepts a workspace tensor as an argument, allowing external control over memory allocation.
• Optimized Workspace Allocation: Within execute_cudnn_gemm_with_per_tensor_q_graph, the workspace is now only allocated if its required size exceeds DEFAULT_WORKSPACE_SIZE, enabling the reuse of a pre-existing buffer for common cases.
• Centralized Workspace Management: The bmm_fp8 function now obtains a shared workspace_buffer once and passes it to both the cudnn and cublas backend implementations, streamlining memory management for these operations.

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 refactors the workspace management for bmm_fp8 to enable buffer reuse, which is a good optimization. The changes are logical and consistent with the goal. However, I've identified a critical issue in the new workspace size validation logic that could lead to runtime errors under certain conditions. My review includes a specific code suggestion to fix this bug, making the implementation more robust.

[yzh119]

Why not separating the cache of different backends and allocate cudnn workspace inside execute_cudnn_gemm_with_per_tensor_q_graph after graph.get_workspace_size() is calculated.

[ttyio]

I noticed that all the methods in gemm.py use DEFAULT_WORKSPACE_SIZE. I assumed the design choice was because different tactics might require different workspace sizes, and using a sufficiently large default helps avoid frequent memory reallocations when the workspace changes. If that’s incorrect, I can update all the methods to request a workspace size matching the current tactic’s requirement instead. Thanks!