# inducer/pycuda

CUDA integration for Python, plus shiny features

Repository: https://github.com/inducer/pycuda
Canonical: https://ross.abutalabs.com/products/pycuda
Homepage: http://mathema.tician.de/software/pycuda
Language: Python
License: NOASSERTION
License Family: other
Topics: python, pycuda, gpu, gpu-computing, cuda, array, multidimensional-arrays, scientific-computing
Last push: 2026-07-16T16:04:07+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 92, release rhythm 47, longevity 100
- inputs: {"age_days": 5628, "days_push": 48, "days_rel": 196, "gap_med": 106.5, "n_releases_24m": 5}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2051, forks 298 (observed 2026-08-28T04:06:09.579263+00:00)

## What it is
PyCUDA is a Python wrapper giving Pythonic access to Nvidia's CUDA parallel computation API, including GPUArray multidimensional arrays and just-in-time kernel compilation. Its C++ base layer adds RAII resource cleanup, automatic error checking, and OpenGL interoperability with near-zero overhead.

## Use cases
- run CUDA kernels from Python
- do GPU-accelerated array math with numpy-like GPUArray
- compile CUDA kernels at runtime with SourceModule
- write leak-free CUDA code with automatic resource cleanup
- interoperate CUDA with OpenGL
- accelerate scientific computing workloads on Nvidia GPUs

## When to choose
- you need full access to the CUDA driver API from Python
- you want numpy-style GPU arrays and JIT kernel compilation
- you're doing scientific computing on Nvidia hardware and want mature, well-documented tooling

## When to avoid
- you need cross-vendor GPU support (use PyOpenCL instead)
- you prefer higher-level frameworks like PyTorch or CuPy for array computing
- you don't have an Nvidia GPU or CUDA toolkit installed

## Facets
- artifact type: library
- maturity: active
- function: gpu-computing, sdk, math
- domain: gpu-computing, data-science, developer-tools
- platform: python, windows
- tags: cuda, nvidia, gpu-arrays, scientific-computing, numpy, jit-compilation, opengl-interop, linux, macos, gpu

## Member repositories
- inducer/pycuda (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:09.579263+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:57:33.405148+00:00, confidence not recorded.
  - readme: https://github.com/inducer/pycuda (fetched 2026-08-28T04:06:09.579263+00:00, sha 7daa53a226a0)
  - homepage: http://mathema.tician.de/software/pycuda (fetched 2026-08-29T10:38:05.234311+00:00, sha 1e6436c2d3ed)
  - site_page: https://mathema.tician.de/aboutme (fetched 2026-08-29T10:38:05.243601+00:00, sha dc48d3674be5)
  - registry_pypi: https://pypi.org/pypi/pycuda/json (fetched 2026-08-29T10:38:05.247980+00:00, sha 887122dbc345)
  - site_page: https://documen.tician.de/pycuda (fetched 2026-08-29T10:38:05.245928+00:00, sha 9fda732be9ad)
- Data as of 2026-08-30T08:39:29.467469+00:00.