Python Runtime

This guide walks through using Python with poly-bench, from init through run. It covers what gets created, how dependencies work, and common gotchas.

Prerequisites

You need Python installed and available on your PATH. The table below summarizes the requirements.

If Python is missing when you run init, poly-bench can offer to install it interactively during poly-bench init.

Init

When you run poly-bench init --languages python (or select Python in the interactive flow), poly-bench creates a Python project under .polybench/runtime-env/python/:

1.polybench/runtime-env/python/
2├── .venv/
3├── requirements.txt
4└── main.py

poly-bench creates a virtual environment (.venv) in .polybench/runtime-env/python/ using python -m venv .venv. All installs use the venv's pip, so dependencies stay isolated from your system Python.

Add Dependencies

To add Python packages, use the add command with the --py flag:

$poly-bench add --py numpy==1.24.0

The spec format is package==version — for example, numpy==1.24.0. Dependencies are stored in the manifest and appended to requirements.txt.

The pyright[nodejs] package is always included in requirements.txt for LSP support. Install and remove operations preserve it.

Install

Running poly-bench install (or poly-bench build) runs pip install -r requirements.txt using the venv's pip. Ensure the venv exists; if it was deleted, re-run init or create it manually with python -m venv .venv.

Build / Codegen

When you run poly-bench run, poly-bench generates Python code and executes it with the venv's Python. There is no separate compilation step — Python runs the generated code directly.

Gotchas and Troubleshooting