v0.6.0

Highlights

• Support Numba 0.66.x - 0.67.x, built against LLVM 22 (llvmlite 0.48 - 0.49)
• Build the pass library and the OpenMP host and device runtimes with LLVM 22.1.8
• Fix offloading on Blackwell GPUs with CUDA 13 (#71) and on Jetson AGX Thor, sm_110 (#68)
• Drop support for Numba 0.62 - 0.65; use PyOMP 0.5.2 with those versions

Thanks to @awennersteen for the LLVM 22 port in #75, which this release builds on.

What's Changed

• Support numba 0.66-0.67 with LLVM 22 by @ggeorgakoudis in #78

Full Changelog: v0.5.2...v0.6.0