NVIDIA/cutile-python

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

2137 stars · 147 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

cuTile Python is a tile-based programming language and library from NVIDIA for writing parallel kernels that run on NVIDIA GPUs, compiling to Tile IR. It integrates with CuPy/NumPy-style array workflows and requires CUDA Toolkit 13.1+ and recent NVIDIA drivers.

Use cases

• write parallel gpu kernels in python
• compile tile-based kernels for nvidia gpus
• accelerate array computations on cuda gpus
• program blackwell and ampere gpus without raw cuda c
• build custom gpu compute kernels for scientific computing

When to choose

• you need to write custom GPU kernels in Python with a tile-based model
• you target NVIDIA Ampere, Ada, or Blackwell GPUs with recent CUDA toolkits
• you want kernel authoring integrated with CuPy/NumPy workflows

When to avoid

• you need Hopper GPU support, which is not yet available
• you want a mature, long-stable toolchain - the project is new and evolving
• you cannot install CUDA Toolkit 13.1+ or driver r580+
• you need portable kernels across non-NVIDIA GPUs

Facets

library · maturity active

programming-language compiler gpu-computing sdk gpu-computing programming-languages compilers developer-tools python windows cuda tile-ir parallel-kernels kernel-programming nvidia blackwell high-performance-computing linux gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="NVIDIA/cutile-python")