feat:add performance compare

[Yael-X]

The HCCL implementation is added for performance comparison, where the dispatch and combine principles of the HCCL implementation are the same as those in vllm-ascend. This script can directly compare the performance of DeepEP communication and HCCL communication used in MoE communication. One of the test data sets is as follows:

BENCHMARK REPORT (World Size: 8)
Params: Tokens=4096, Hidden=7168, TopK=8, Experts=256
Note: Dispatch times INCLUDE layout/split calculation overhead.
------------------------------------------------------------------------------------------
Operation    | DeepEP (ms)  | HCCL (ms)    | Speedup    | Saved (ms)   | Reduction
------------------------------------------------------------------------------------------
Dispatch     | 4.040        | 6.326        | 1.57x      | 2.286        | 36.1%
Combine      | 3.783        | 5.328        | 1.41x      | 1.545        | 29.0%
------------------------------------------------------------------------------------------

[gemini-code-assist]

Summary of Changes

Hello @Yael-X, 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 introduces a dedicated performance comparison script to evaluate the efficiency of DeepEP communication against HCCL communication in a Mixture-of-Experts (MoE) context. The script meticulously benchmarks the dispatch and combine operations for both implementations, providing quantitative data on their relative performance and correctness. This addition is crucial for understanding the performance implications of different communication backends in distributed MoE systems.

Highlights

• New Performance Comparison Script: A new test script, test_performance_compare.py, has been added to benchmark the performance of DeepEP communication against HCCL communication for Mixture-of-Experts (MoE) operations.
• HCCL Dispatcher Implementation: An HCCLDispatcher class is introduced, which implements dispatch and combine logic mirroring the principles used in vllm-ascend but utilizing HCCL for collective communication.
• Benchmarking and Correctness Check: The script includes comprehensive benchmarking for both dispatch and combine phases of DeepEP and HCCL, along with a correctness check using cosine similarity to validate the functional equivalence of their outputs.
• Detailed Performance Reporting: A detailed benchmark report is generated, presenting DeepEP and HCCL execution times, calculated speedup, saved time, and reduction percentages for both dispatch and combine operations.

🧠 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 adds a performance comparison script between DeepEP and HCCL for MoE communication. The new script is well-structured for benchmarking and includes correctness checks. I've found a critical issue with a constructor call that will likely cause a runtime error, along with a few medium-severity suggestions to improve code quality and maintainability, such as removing dead code, avoiding magic numbers, and improving resource management.