v0.3.0

What's Changed

• Clean up pyproject.toml dependencies and align CI with cuda-python/numba-cuda by @leofang in #12
• Fix sporadic bitcode signature nvJitLink error in LTO path by @VijayKandiah in #14
• [Cleanups] Remove license info for removed benchmark + update contributors by @atmnp in #15
• Add array debug info by @jiel-nv in #8
• Fix recreate_with_lto to use cuda.core.ObjectCode public API by @VijayKandiah in #17
• Fix typo in CONTRIBUTORS.md by @atmnp in #20
• Update README, CONTRIBUTING, and INSTALL by @gmarkall in #18
• Add User guide by @gmarkall in #13
• Fix ExternFunction capture in overloads and linker cuda.core compat by @atmnp in #21
• [CI] Enable Python 3.11 by @atmnp in #16
• Add discriminated-union debug info for polymorphic scalar variables by @jiel-nv in #22
• Add mypy/pyright support by @ZzEeKkAa in #26
• Move experimental features into cuda.experimental submodule by @VijayKandiah in #25
• Update docs to use dependency groups (PEP 735) for dev/test install by @VijayKandiah in #30
• [Compile-Time] Re-design the extending API + dispatcher changes by @atmnp in #27
• Add docs build by @gmarkall in #28
• Add float16 vector support by @ZzEeKkAa in #32
• Fix array.view dtype resolution, typing, and lowering by @VijayKandiah in #34
• Expose variables_used linker option by @isVoid in #10
• MNT: Fix sccache summary to count CUDA sub-tool cache hits by @leofang in #29
• Support cuda-core 1.0 compatibility by @isVoid in #23
• Try to fix publish workflow by @leofang in #41
• Bump version to 0.3.0 by @leofang in #42

New Contributors

• @VijayKandiah made their first contribution in #14
• @jiel-nv made their first contribution in #8
• @gmarkall made their first contribution in #18
• @ZzEeKkAa made their first contribution in #26
• @isVoid made their first contribution in #10

Full Changelog: https://github.com/NVIDIA/numba-cuda-mlir/commits/v0.3.0