Minutes_2024_04_16

Numba Meeting: 2024-04-16

Attendees: FPOC (last week): Siu FPOC (incoming):

NOTE: All communication is subject to the Numba Code of Conduct.

Please refer to this calendar for the next meeting date.

For 0.60, should we deprecate the error_model kwarg, not really needed post np2 support.
- what should happen to error-model in CUDA which is tied up with debug=True?
  - may need to set what's the defaut scalar type is
  - deprecate it now(?)
NP2 update
- rewritten 90% of builtin ufuncs to rely on libm
- rewriting operators in terms of protocols (e.g. lhs.__add__(rhs) otherwise rhs.__radd__(lhs))
- need concept of "closed"/"protected"/"final" overload (i.e. user cannot override or extend)
0.60 release status:
- NP2 binary compatiability test all passed
- TODOs: https://github.com/numba/numba/milestone/77

numba#9528 - CUDA - Inline literal in (...) statements
numba#9529 - Wrong list result with parallel=True despite no resizing and no cross-thread access of the list

llvmlite#1037 - [Draft] Add APIs to run individual passes, combination of passes and Default optimization pipelines with new pass manager
- remember to deprecate old passmanager API for LLVM17
llvmlite#1038 - Test #986 on CI with LLVM 15

(last numba: 9529; llvmlite 1038)