Skip to content

QUICKSTART.md

Latest commit

 

History

History
113 lines (86 loc) · 2.81 KB

QUICKSTART.md

File metadata and controls

113 lines (86 loc) · 2.81 KB
Raw

ParaCodex Quick Start Guide

Get up and running in 5 minutes. For full documentation, see README.md.

🎯 Prerequisites Checklist

  • NVIDIA GPU with CUDA support
  • Node.js 22+ and npm 9+
  • NVIDIA HPC SDK 25.7+ (nvc++ compiler)
  • Python 3.8+
  • OpenAI API key or OpenAI Pro account

⚡ Quick Setup

1. Clone and navigate

git clone https://github.com/Scientific-Computing-Lab/ParaCodex.git
cd paracodex
chmod +x setup_environment.sh install_nvidia_hpc_sdk.sh verify_environment.sh

2. Install NVIDIA HPC SDK (if not already installed)

sudo ./install_nvidia_hpc_sdk.sh
source ~/.bashrc

3. Run the environment setup

./setup_environment.sh

This installs all dependencies in one step: Codex CLI, Python packages, webapp dependencies, and copies the ParaCodex Codex skills to ~/.codex/skills/paracodex/.

4. Set up OpenAI access

Option 1: OpenAI Pro account (recommended)

codex login

Option 2: API key

export OPENAI_API_KEY="your-api-key-here"

5. Verify your environment

./verify_environment.sh

🚀 Running ParaCodex

cd pipeline/webapp
bash start.sh

Open http://localhost:5000 in your browser.

Using the web interface

  1. Source Directory — Browse to the folder with the code you want to translate
  2. Source API — The parallel API of your code (e.g. serial, cuda)
  3. Target API — The API to translate to (e.g. omp, hip)
  4. Click Start Pipeline — live logs stream in as the agent works
  5. Inspect translated code and profiling artifacts in the job view

Benchmark workdirs

Suite Source directory
Rodinia workdirs/serial_omp_rodinia_workdir/data/src/
NAS workdirs/serial_omp_nas_workdir/data/src/
HeCBench workdirs/serial_omp_hecbench_workdir/data/src/
ParEval workdirs/cuda_omp_pareval_workdir/data/src/

🔧 Common Issues

codex: command not found

npm install -g @openai/codex

node: command not found

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
source ~/.bashrc
nvm install 22

nvc++: command not found

export PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/25.7/compilers/bin:$PATH
export LD_LIBRARY_PATH=/opt/nvidia/hpc_sdk/Linux_x86_64/25.7/compilers/lib:$LD_LIBRARY_PATH

Port 5000 already in use

FLASK_RUN_PORT=8080 python pipeline/webapp/app.py

GPU not accessible

nvidia-smi
# If this fails, check NVIDIA driver installation

📚 Next Steps