This is the simulator for UniZK, an accelerator for Zero-Knowledge Proof with unified hardware and flexible kernel mapping.
The codes of the following third-party libraries are sourced from:
- ecdsa: https://github.com/0xPolygonZero/plonky2-ecdsa/tree/bdb6504bca250db1548cdcce2407d4e334990c33
- imgcrop: https://github.com/ChickenLover/plonky2-zkedit/tree/d20567361bd405716f6736a88dc263978f267d62
- plonky2: https://github.com/0xPolygonZero/plonky2/tree/30b47998262642be54da5acf03dfca31af4d93f7
- plonky2v0.1: https://github.com/polymerdao/plonky2/tree/4cb0b48df1d227d5461a4c28ed025aaea64e2e62
- proto-neural-zkp: https://github.com/worldcoin/proto-neural-zkp/tree/b2b514ac0857fd5e1cb5da9399fcd6020b1730e3
- ramsim: RamSim is an enhanced version of Ramulator2, adding support for simulating computation latency and data dependencies. Most of its source code is derived from https://github.com/CMU-SAFARI/ramulator2/tree/b7c70275f04126c647edb989270cc429776955d1.
- sha256: https://github.com/polymerdao/plonky2-sha256/tree/06d128e78ed8d29b21d58294b069e852c1866f8d
- sha256-starky: https://github.com/tumberger/plonky2/tree/474ed82c385b446f89bc18e648b5ad7d5d94ce06
We appreciate the contributions of these open-source authors.
- A recent Linux distribution (recommended: Ubuntu 22.04)
- rust 1.80
- g++ 11.4
- cmake 3.22
The dependency.sh script will install the necessary tools: g++, cmake, and Rust. To run it, use the following command:
./dependency.sh
If you want to run UniZK in a Docker container, the docker.sh script will automatically create and start a new container and configure the requirements. Use this command:
./docker.sh
The files of the project will be copied into /UniZK within the container.
rustup override set nightly
cd thirdparty/ramsim
mkdir build
cd build
cmake ..
make -j
./run_cpu_test.sh
./run_plonky2_test.sh
./run_starky_test.sh
A log file with the name of the application, such as “sha256.log”, will appear in the folder with the simulation results.
Trace files can take up a lot of disk space. Ensure you have enough space when testing large applications and remember to clean up the files after testing!
The GPU performance can be evaluated using the GPU implementation of Plonky2 (available at https://github.com/sideprotocol/plonky2-gpu/tree/d6430874ae76e6cfc45d995fa3cf8b84fad70404) with our CPU-based application code located in the examples directory.