Skip to content
esc edited this page Aug 24, 2022 · 1 revision

Numba Meeting: 2022-08-23

Attendees: Graham Markall, Val, Andre Masella, Benjamin Graham, brandon willard, Jim Pivarski, stuart, Todd A. Anderson, Siu Kwan Lam, Kaustubh Chaudhari, Luk, Shannon Quinn, Guilherme Leobas

FPOC (last week): Graham

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

Please refer to this calendar for the next meeting date.

0. Discussion

  • (suggest topic) - (name)
  • Update on plans for Numba 0.56.1 / llvmlite 0.39.
    • main thing is numpy 1.23 support
    • there are also other bug fixes
  • Python 3.11 workload
  • PRs arising in discussion of cache behaviour on Discourse (Graham):
    • #6406: Enable caching functions arising from exec of strings
    • #6284: Prevent cache busting by the UUID of Dispatcher
  • Functions that are indirectly cached - Luk

1. New Issues

  • #8362 - Parallel =True issue with 0.56.0
  • #8363 - Update numba\cuda\cudadrv\enums.py
  • #8370 - NumbaTypeError: Cannot modify readonly array of type: readonly array(int64, 2d, C)
  • #8372 - typed.List.append not validating incompatible types correctly
  • #8373 - jit cache pickle exception on many local cuda arrays
  • #8382 - slow performance observed in np.sort() with nan arrays
  • #8383 - On Mac M1, importing '_get_thread_id' from 'numba.np.ufunc.parallel' throws an error. Downgrading to 0.55.2 fixes the problem

Closed Issues

  • #8367 - Builtin function "map" produces memory leak
  • #8369 - List & Dict can't accept array.dtype as argument.

2. New PRs

  • #8360 - Pre-commit update Flake8 and add basic pre-commit-hooks for linting
    • Note to update PR policy
  • #8361 - Enforce Flake8 E7 statement rules with help of autopep8
  • #8364 - Enforce Flake8 E502 rule with help of autopep8
  • #8365 - Enforce Flake8 E10 and E11 indentation rules
  • #8366 - Remove setuptools as runtime dependency
  • #8368 - Remove glue_lowering in random math that requires IR
  • #8371 - Added binomial distribution
  • #8374 - Don't pickle LLVM IR for CUDA code libraries
  • #8375 - *** [WIP] Extend cache invalidation to include changes in other files
  • #8376 - Fix issue 8370
  • #8377 - Add support for NumPy 1.23.
  • #8378 - bump max python version for Python 3.11
  • #8379 - Comment out all C-level changes
  • #8380 - implement new Python 3.11 opcode: RESUME
  • #8381 - Python3.11
  • #8384 - Move strace() check into tests that actually need it.
  • #8385 - DO NOT MERGE. debugging ci timeout
  • #8386 - Fix the docs for numba.get_thread_id

llvmlite:

  • #871 - Refactor native library loading
  • #872 - Fix symbol resolution race
  • #873 - Make symbol resolution lazy

Closed PRs

  • merged - #8359 - adding warnings about cache behaviour

3. Next Release: Version 0.57.0/0.40.0, RC Jan 2023

Clone this wiki locally