In the pursuit of cryptographic efficiency, we introduce plonky2-gpu — a GPU-accelerated version of the Plonky2 project. This iteration leverages the CUDA framework and represents a comprehensive reengineering of the original Plonky2 codebase. It specifically focuses on optimizing three key computations: Fast Fourier Transform (FFT), Merkle tree construction, and polynomial manipulation.
Hardware Requirements:
- CPU: 8 cores
- RAM: 16GB
- GPU: NVIDIA 2080 Ti
- GPU RAM: 12GB
- CUDA Version: 12+
Examples:
Plonky2-25519: Discover the enhanced performance, specially optimized for ed25519 signatures. This accelerated implementation has notably reduced the proving time for ed25519 signatures from 45 seconds to just 5 seconds. Experience the optimized performance specifically tailored for ed25519 signatures.
DISCLAIMER: This implementation is currently in its early stages and is not deemed production-ready. Use for experimentation and exploration purposes only.