Minutes_2022_08_16

Numba Meeting: 2022-08-16

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

FPOC (last week): apmasell

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

Please refer to this calendar for the next meeting date.

0. Discussion

• A moment of silence for Jon Riehl (https://twitter.com/teoliphant/status/1559321437231538178) may they R.I.P. 🖤 (Val)
• Timeline for the 0.56.1 release (Val)
  • Waiting on a fixed Scipy 1.7
  • Have all the pieces in place
• How to proceed with: https://github.com/numba/numba/pull/8347 (Val)
  • The PR is too large to be merged
  • We are interested in the flake8 changes however:
    • Public CI will never pass
    • This will break many existing PRs
  • We would like them in a more piecemeal fashion
• Caching limitations (Luk)
  • show limitations on user section of documentation.
  • how to relax limitation on recognizing symbol changes from different files.
• Use meminfo in dynamic exception PR (Guilherme)
  • Is there a memory leak in the proposed code? RE: yes
  • link: https://github.com/numba/numba/pull/8134

1. New Issues

• #8327 - Support List as a value type in njit dictionaries
• #8328 - Support creating a dictionary from a zipped value
• #8340 - nan as dict key different behavior in py3.10
• #8346 - pycc compatibilty with versioneer
• #8355 - numpy.distutils is deprecated in NumPy 1.23 and issues with setuptools>=65.

Closed Issues

• #8330 - Get the optimized bytecode from IR
• #8344 - Please upgade for numpy version

2. New PRs

• #8331 - Extend docs with info on how to call C functions from Numba
• #8332 - Disallow ambiguous overload
• #8333 - Remove dtype from kwarg list during function replacement in pre-parfor pass.
• #8334 - Add dict(*iterable) constructor
• #8335 - Remove deprecated pycc script and related source.
• #8337 - Support list/sets as dict value type
• #8338 - Calculate reductions before fusion so that use of reduction vars can stop fusion.
• #8339 - Fix #8291 parfor leak of redtoset variable
• #8341 - CUDA: Support multiple outputs for Generalized Ufuncs
• #8342 - Upgrade to ubuntu-20.04 for azure pipeline CI
• #8343 - Eliminate references to type annotation in compile_ptx
• #8345 - CUDA: Fixes for NumPy 1.23
• #8347 - Use pre-commit to make Flake8 pass again
• #8348 - Add get_max_threads_per_block method to Dispatcher
• #8349 - NumPy 1.23 testing patch
• #8356 - Update setup.py, buildscripts, CI and docs to require setuptools<60.
• #8357 - Clean up the buildscripts directory.
• #8358 - Added runtime check for np.nditer's input array layouts

llvmlite:

• #870 - Disable test_libm on macOS

Closed PRs

• #8329 - Add name attribute to CUDAGUFuncVectorize plus test case
• merged - #8336 - Fix typos of "Generalized" in GUFunc-related code
• #8350 - Revert "Merge pull request #8299 from Biswa96/mingw-fix"
• #8351 - switch RTD from mamba to miniconda
• #8352 - pin setuptools to 64 for doc builds
• #8353 - Test setuptools < 65 with gpuCI
• merged - #8354 - pin setuptools to < 65 and switch from mamba to conda on RTD

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