NVIDIA/numba-cuda-mlir v0.5.2

v0.5.2

Repository: NVIDIA/numba-cuda-mlir

Tag: v0.5.2

Published: 2026-09-08T21:59:48Z

Prerelease: no

Release notes:

What's Changed

• Remove CCCL dependency by @brandon-b-miller in https://github.com/NVIDIA/numba-cuda-mlir/pull/271
• Fix CodeRabbit review configuration by @atmnp in https://github.com/NVIDIA/numba-cuda-mlir/pull/295
• Reject a value-returning kernel at compile time by @jiel-nv in https://github.com/NVIDIA/numba-cuda-mlir/pull/300
• Fix error handling for grid runtime dim by @jiel-nv in https://github.com/NVIDIA/numba-cuda-mlir/pull/299
• Fix skip condition when cuobjdump can't be found. by @brandon-b-miller in https://github.com/NVIDIA/numba-cuda-mlir/pull/287
• [CR] Fix indentation on tools by @atmnp in https://github.com/NVIDIA/numba-cuda-mlir/pull/302
• Use one host independent NVPTX data layout by @brandon-b-miller in https://github.com/NVIDIA/numba-cuda-mlir/pull/289
• Preserve signedness in integer-float conversions by @atmnp in https://github.com/NVIDIA/numba-cuda-mlir/pull/294
• bump version to 0.5.2 by @atmnp in https://github.com/NVIDIA/numba-cuda-mlir/pull/306

Full Changelog: https://github.com/NVIDIA/numba-cuda-mlir/compare/v0.5.1...v0.5.2