Minutes_2022_08_30
esc edited this page Aug 31, 2022
·
1 revision
Attendees: Siu Kwan Lam, Andre Masella, Guilherme Leobas, Jim Pivarski, Kaustubh Chaudhari, Graham Markall, LI Da, Benjamin Graham, brandon willard, Val, Shannon Quinn, Todd A. Anderson, FPOC (last week): Siu
NOTE: All communication is subject to the Numba Code of Conduct.
Please refer to this calendar for the next meeting date.
- (suggest topic) - (name)
- Guilherme is now a Numba maintainer
- Release 0.56.1
- (Todd) how to undo SSA form?
- There's a pass to undo Phi node introduced by SSA pass https://github.com/numba/numba/blob/main/numba/core/typed_passes.py#L774
- Perhaps a simple pass to renaming Var to their pre SSA names.
- Python 3.11 update
- Python 3.12 "micro-ops" https://github.com/faster-cpython/ideas/issues/454
- (JimP) Other projects that interpret Python bytecode: https://github.com/rocky/python-uncompyle6 and maybe https://github.com/alexmojaki/sorcery. Also, my experiments on uncompyle6, which focuses on lambda expressions: https://github.com/diana-hep/rejig/tree/master/reinterpreted-python/rejig
- (brandon) optimization on integer to avoid wraparound
- **** #8394 - PEP684 Subinterpreter support
- #8395 - [-1] call to cuInit results in UNKNOWN_CUDA_ERROR
- #8398 - 50% performance drop from py3.8/numba 0.51.2 to py3.9/numba 0.55.1
- #8399 - please add np.min()/max() with augment:"axis"
- #8400 - On target cuda, return type of function may not match type given in signature string
- #8402 - Pretty-printed type annotation is broken
- #8403 - Strange Result for np.nanstd in njit function
- #8404 - llvmlite assertion when using deferred type as typed dict values
- #8405 - Checks for function type casting erroneously assume compilation is always possible
- #8406 - Numba 0.56.1 Checklist
llvmlite:
- #877 - llvmlite 0.39.1 Checklist
- #8387 - Add support for compute capability in IR Lowering
- #8388 - Remove more references to the pycc binary.
- #8389 - Make C++ extensions compile with correct compiler
- #8390 - Use NumPy logic for lessthan in sort to move NaNs to the back.
- #8391 - bug fixed: solving the issue in numpy 1.17
- #8392 - Release0.56.1
- #8396 - *** [WIP] Extend cache invalidation to include changes in other files (take 2)
- #8401 - Remove Cuda toolkit version check
- #8407 - Pin NumPy version to 1.18-1.24
llvmlite:
llvmlite:
- merged - #874 - Adding zlib as dependency for aarch64