Skip to content

Conversation

AdvancedCompiler
Copy link
Contributor

PR Category

Operator

Type of Change

Performance Optimization

Description

Issue

Progress

  • Change is properly reviewed (1 reviewer required, 2 recommended).
  • Change is responded to an issue.
  • Change is fully covered by a UT.

Performance

f1b07505af7d8ad031b084630bed4337

Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Summary of Changes

Hello @AdvancedCompiler, 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 significantly refactors the argmax operator implementation to improve its performance, particularly by introducing specialized kernels and dynamic heuristic configurations based on input tensor dimensions.

Highlights

  • Specialized Argmax Kernels: The existing argmax_kernel has been refactored into two distinct kernels: argmax_kernel_non_inner for general cases (K > 1) and argmax_kernel_inner for specific cases where K = 1.
  • Dynamic Kernel Dispatch: The main argmax function now intelligently dispatches to the appropriate kernel (argmax_kernel_non_inner or argmax_kernel_inner) based on the value of the K dimension, allowing for more optimized execution paths.
  • Enhanced Heuristic Configurations: New and more granular heuristic functions have been introduced for TILE_K, TILE_N, ONE_TILE_PER_CTA, and num_warps for both argmax_non_inner and argmax_inner kernels, enabling better tuning for performance.
  • Improved Kernel Logic: The kernel implementations have been updated to handle data types more robustly (e.g., float16/bfloat16 to float32 for intermediate calculations) and to manage loop iterations and remainders more efficiently for better performance.
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. 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.

Copy link
Contributor

@gemini-code-assist gemini-code-assist bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request introduces significant performance optimizations for the argmax operator by implementing two specialized Triton kernels: one for reductions along the inner dimension (argmax_kernel_inner) and another for non-inner dimensions (argmax_kernel_non_inner). This is a solid approach that yields impressive performance gains as shown in the benchmarks. My review includes a critical fix for accessing tensor types within a Triton kernel, a suggestion to improve numerical stability, and recommendations for enhancing code maintainability by reducing code duplication and replacing magic numbers with named constants.

StrongSpoon
StrongSpoon previously approved these changes Sep 10, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants