Analysis of MLE-Polynomial Commitment Schemes

Welcome to our benchmarking project for various proving systems and polynomial commitment schemes.

Overview

Our project provides comprehensive analysis of the different proving systems (Basefold, Brakedown, Gemini, Zeromorph-FRI, Hyrax) by running benchmarks across different circuit sizes, and measuring key metrics such as proof generation time, proof size, and verification time.

Installation & Usage

1. Clone the Repository:

   git clone https://github.com/sec-bit/mle-pcs-benchmark
   cd mle-pcs-benchmark

2. Build the Project:

   cargo build --release

3. Running Benchmarks:

   Navigate to the plonkish directory and run:

   cargo run -p bench-cli

   Follow the prompts to select the proof system and set the desired parameters.

For more detailed benchmark instructions and technical documentation, please refer to docs/benchmark_analysis.md.

Benchmarking

Our benchmarks cover:

• Proof Generation Time: How quickly a proof can be generated.
• Proof Size: The size of the generated proofs.
• Verification Time: How fast a proof can be verified.

Visual comparison at a glance 📊

open large	open large	open large

Why log-scaled axes?

All three plots use a logarithmic Y-axis so that huge differences (e.g. Gemini’s 75 × smaller proofs vs. Brakedown’s gigabit proofs) remain visible on a single graph.

For the full CSV tables, raw measurement scripts, and methodology, see docs/bench_results.md.

Testing Framework

We employ a robust testing framework that includes:

• Repeated sampling to average out measurements.
• Two timing methods (Arkworks timer for aggregate times and Rust’s built-in timer for individual operations).
• Automated data processing to record and analyze results.

For more in-depth details on our testing framework, please see docs/benchmark_analysis.md.

Acknowledgements

This project was originally forked from hadasz/plonkish_basefold. Since the fork, we have made several improvements:

• Added comprehensive benchmarks for Polynomial Commitment Schemes (PCS).
• Collected and analyzed extensive bench-data.
• Integrated plonky3 and enabled benchmarks using P3.
• Developed a robust testing framework for accurate measurement of proof generation, proof size, and verification time.

We are grateful to the original authors for their foundational work.

---

Future Work / Ongoing Refinements

We are continuously refining our testing framework, data processing, and analysis methods. Stay tuned for further updates as we enhance the project's performance and capabilities.