Minutes_2022_02_22

Numba Meeting: 2022-02-22

Attendees: Siu Kwan Lam, Graham Markall, Jim Pivarski, Benjamin Graham, stuart, Todd A. Anderson (Intel Labs), Travis Oliphant, Val, Guilherme Leobas, Luk, Vladimir Lukianov, Ehsan Totoni, Ali

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

Please refer to this calendar for the next meeting date.

0. Feature Discussion

• Reminder on GH permission cleanup
• Tutorial section on Discourse
• Governance / vision discussion
  • Graham's vision
    • agree with lots of points by Ehsan
    • compiler toolkit
    • facilitate the use of CUDA in Python; custom logic; not primarily the standalone/sole tool to write Python programs for CUDA.
  • Governance
    • (The following are guildelines and documenting the process evolved to be. Nothing is set in stone)
    • evolved to similar rules for Core Team Membership as Pandas
      • https://github.com/pandas-dev/pandas-governance/blob/master/governance.md#core-team-membership
    • FPOC
      • triage issues
      • answer gitter
      • weekly rotation with the rest of the maintainer group
      • Cheatsheet available: https://github.com/numba/numba/wiki/Numba-FPOC-Cheatsheet
    • Maintainer vs Code owner
      • Maintainer must consider long-term health of the project
      • Code owner can be solely responsible of a single feature or a sub-section of the code base
    • Moving forward:
      • Make what we've discussed into guidelines
      • Follow through with allowing others to join as "maintainers in training"
      • Document it for reference (noting that it will be a guideline)

1. New Issues

• #7849 - CUDA: ptx attribute fails for device function
• #7852 - Support for multidimensional sparse arrays (Pydata/sparse)
  • Need to close as out of scope for Numba

Closed Issues

• #7847 - int * list isn't supported

2. New PRs

• #7844 - Add more flexible compare-and-swap
• #7850 - CUDA: Pass fastmath compiler flag down to compile_ptx and compile_device; Improve fastmath tests
• #7851 - dissallow llvm-openmp 12.* for M1

Closed PRs

• #7845 - change time.time() to time.perf_counter() in docs
• #7846 - Fix CUDA enum vectorize test on Windows
• #7848 - Support for int * list

3. Next Release: Version 0.56.0/0.39.0, RC,