ZKCuda-No-Oversubscribe

[zhiyong1997]

Benchmark of PCS Batching

The testing kernel contains a sumcheck of $2^{19}$, and the actual length is $2^{19} \times$ parallelism. The task is distributed to parallelism number of cores. The field used is BN254. There are 3 pcs commit and opening.

On 7950X3D with 16 physical cores:

No PCS Batching

Parallelism	Commit (ms)	GKR Prove(ms)	PCS Opening(ms)	Verify(ms)
4	30	262	562	109
8	35	340	628	108
16	42	571	824	115

PCS Batching

Parallelism	Commit (ms)	GKR Prove(ms)	PCS Opening(ms)	Verify(ms)
4	57	254	623	84
8	61	324	685	90
16	81	563	845	110

[gemini-code-assist]

Summary of Changes

Hello @zhiyong1997, 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 focused this pull request on enhancing performance observability within the expander_compiler by integrating timing measurements for critical operations in the deferred PCS proving system. I've also extended the test suite to include benchmarks for a new proving system configuration.

Highlights

• Performance Measurement: I've introduced Timer instances to measure the execution time of key phases within the deferred PCS proving process, specifically for commitment, kernel proving, and batch PCS opening.
• New Configuration Testing: I've added new test runs for the zkcuda_matmul function using the BN254ConfigSha2UniKZG configuration, expanding coverage for different proving system setups.

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 is currently in preview and 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 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 adds timers to the deferred PCS proving implementation to measure performance of different stages. It also adds new benchmark runs for zkcuda_matmul with a different configuration (BN254ConfigSha2UniKZG).

[gemini-code-assist]

These lines are very similar to lines 8-10. To avoid code duplication and improve maintainability, consider using a macro to generate these calls. Since const generics must be known at compile time, a loop cannot be used here, but a macro is a good alternative.

[zhiyong1997]

Multi-Process Oversubscription v.s. Limited-process Multi-SIMD

Setting: 7950X3D with 16 pyhsical cores, testing size: $2^{13}$. Field BN254.

GKR

Parallelism	Multi-Process (ms)	Multi-SIMD (ms)
16	8	8
32	25	15
64	46	25
128	91	50
256	191	81
512	510	166

PCS Opening

Parallelism	Multi-Process (ms)	Multi-SIMD (ms)
16	57	59
32	170	70
64	187	97
128	489	132
256	975	183
512	2350	296

Note that the work is exactly the same for these two cases for PCS opening. The only difference is Multi-Process may involve context switching.