cuPQC

Authors: Lakshya Chopra, Satyam Dubey & Aditya Koranga

Published: December 08, 2024

Note: This documentation is a replica of the README available at the github/ngkore/cuPQC. Please refer to the original repository for the most up-to-date information.

Prerequisites

To utilize cuPQC, the following environment requirements must be met:

System Requirements

  • CPU: x86_64 architecture

  • GPU: NVIDIA GPU with architecture 70, 75, 80, 86, 89, or 90

  • Recommended Hardware: NVIDIA H100 (Development tested on RTX A4000)

Software Requirements

  • CUDA Toolkit: Version 12.4 or newer

  • Host Compiler: C++17 support required (g++/gcc 11 or newer)

  • Build Tools: CMake 3.20+ (optional)

CUDA Installation (Ubuntu 22.04)

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin
sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600
wget https://developer.download.nvidia.com/compute/cuda/12.6.3/local_installers/cuda-repo-ubuntu2204-12-6-local_12.6.3-560.35.05-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu2204-12-6-local_12.6.3-560.35.05-1_amd64.deb
sudo cp /var/cuda-repo-ubuntu2204-12-6-local/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-6

Post-Installation Setup

Add the following to your .bashrc file to make changes permanent:

export PATH=/usr/local/cuda-12.6/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-12.6/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

Verify the installation:

nvcc --version

Resources

cuPQC benchmarking video: https://youtu.be/mdnXKroR1wo?si=yiTMTUy1MIzl20_7

Build Instructions

  1. Clone the repository and navigate to the examples directory:

cd cuPQC/examples/

  1. Build using the provided Makefile:

make

Alternatively, build manually using nvcc:

nvcc -dlto -arch=native -std=c++17 -O3 -L../lib/ -lcupqc -lcuhash -o <binary_name> <file_name.cu> -I../include/ -I../include/cupqc

Note: Ensure the -arch flag matches your GPU’s compute capability. If programs fail to execute, delete the binary and rebuild.

Benchmarking

Tests were performed on an NVIDIA GH200.

ML-KEM 512 Performance

Build and run the benchmark:

nvcc -dlto -arch=native -std=c++17 -O3 -L../lib/ -lcupqc -o v4_max_bench v4_max_bench.cu -I../include/ -I../include/cupqc
./v4_max_bench

Results:

  • Keygen: ~20 million ops/sec

  • Encapsulation: ~18.5 million ops/sec

  • Decapsulation: ~18 million ops/sec

Additional Benchmarks

  • ML-KEM 768: ./mlkem768_bench

  • ML-KEM 1024: ./mlkem1024_bench

  • ML-DSA 44: ./bench_mldsa

  • ML-DSA 65: ./mldsa65_bench

  • ML-DSA 87: ./mldsa87_bench