numba-cuda 0.30.3

CUDA target for Numba

Numba CUDA Target

The CUDA target for Numba. Please visit the official documentation to get started!

Maintenance notice

Numba-CUDA is in maintenance mode. Moving forward, we intend to support only security issues and critical bug fixes through the lifetime of CUDA 13. New feature development is targeted towards Numba-CUDA-MLIR, and we recommend upgrading to Numba-CUDA-MLIR as soon as practical. For migration guidance, see Migration from Numba / Numba-CUDA.

To report security issues or critical bugs, please use the issue tracker.

To raise questions or initiate discussions, please use the Numba Discourse forum.

Installation with pip or conda

Please refer to the Installation documentation.

Installation from source

Install as an editable install:

pip install -e .

If you want to manage all run-time dependencies yourself, also pass the --no-deps flag.

Running tests

Tests must be run from the testing folder, which contains the pytest configuration and code to generate binaries used during the tests. The test binaries need to be built on the system on which the tests are run, so that they are compiled for the appropriate compute capability.

cd testing
# Optionally, build test binaries and point to their location for the test suite
make -j $(nproc)
export NUMBA_CUDA_TEST_BIN_DIR=`pwd`
# Execute tests
pytest -n auto -v --dist loadscope

Alternatively, you can use pixi to wrap all of that up for you:

# run tests against CUDA 13
pixi run -e cu13 test -n auto -v --dist loadscope

Testing should discover the numba.cuda module from the numba_cuda package. You can check where numba.cuda files are being located by running

python -c "from numba import cuda; print(cuda.__file__)"

which will show a path like:

<path to numba-cuda repo>/numba_cuda/numba/cuda/__init__.py

Contributing Guide

Review the CONTRIBUTING.md file for information on how to contribute code and issues to the project.