Skip to content

[WIP][FusedDeepMoe] offset state space of FusedDeepMoe operator#243

Open
wangqiankun13 wants to merge 1 commit intosgl-project:mainfrom
wangqiankun13:dev_state_offset
Open

[WIP][FusedDeepMoe] offset state space of FusedDeepMoe operator#243
wangqiankun13 wants to merge 1 commit intosgl-project:mainfrom
wangqiankun13:dev_state_offset

Conversation

@wangqiankun13
Copy link
Contributor

[FusedDeepMoe] the state space of FusedDeepMoe operator is offset by 16 bytes, to avoid conflicts with norm operators

…16 bytes, to avoid conflicts with norm operators

Signed-off-by: wangqiankun <wangqiankun13@huawei.com>
@gemini-code-assist
Copy link
Contributor

Summary of Changes

Hello @wangqiankun13, 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 implements a critical adjustment to the memory allocation strategy for the FusedDeepMoe operator. By introducing a 16-byte offset to its state space, the changes aim to resolve and prevent memory conflicts that could arise when FusedDeepMoe interacts with other norm operators, thereby enhancing the stability and correctness of the system's operations.

Highlights

  • Memory Offset Introduction: A new macro, WINDOWS_EXP_OFFSET, has been introduced and defined with a value of 16 bytes to specify a memory offset.
  • State Space Adjustment: The state space for the FusedDeepMoe operator is now consistently offset by 16 bytes across multiple related operator kernels.
  • Conflict Resolution: This offset is applied to windowsExp memory addresses and statusDataSpaceGm initialization to prevent potential memory conflicts with norm operators.
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.

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 a 16-byte offset to the state space of the FusedDeepMoe operator to prevent memory conflicts, as described. The change is implemented by adding a WINDOWS_EXP_OFFSET macro and applying it consistently to address calculations involving localWindowsExp and windowsExp across several kernel files. The implementation appears correct and directly addresses the stated goal. I have one suggestion to improve the robustness of a modified macro definition.

Comment on lines 47 to +50
(((epRankId == rankId) \
? ((GM_ADDR)(winContext_->localWindowsExp)) \
: ((GM_ADDR)(((HcclRankRelationResV2 *)(winContext_->remoteRes[rankId].nextDevicePtr))->windowsExp))) + \
dataState * WIN_STATE_OFFSET)
dataState * WIN_STATE_OFFSET + WINDOWS_EXP_OFFSET)
Copy link
Contributor

Choose a reason for hiding this comment

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

medium

For safety, it's a good practice to wrap the entire expression of a function-like macro in parentheses. This prevents potential operator precedence issues when the macro is used within a more complex expression.

    ((((epRankId == rankId)                                                                                        \ 
          ? ((GM_ADDR)(winContext_->localWindowsExp))                                                             \ 
          : ((GM_ADDR)(((HcclRankRelationResV2 *)(winContext_->remoteRes[rankId].nextDevicePtr))->windowsExp))) + \ 
     dataState * WIN_STATE_OFFSET + WINDOWS_EXP_OFFSET))

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.

1 participant