Minutes_2021_11_02

Numba Meeting: 2021-11-02

Attendees: Guilherme Leobas, Graham Markall, Nick Riasanovsky, Siu Kwan Lam, Kaustubh Chaudhari, Todd Anderson, Val, Jim Pivarski, stuart, brandon willard

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

0. Feature Discussion

• Feature request: Numpy Generator API np.random.default_rng
  • https://numpy.org/doc/stable/reference/random/generator.html
  • numpy devs said this API is designed with Numba in-mind
  • Aesara: https://github.com/aesara-devs/aesara/pull/420
• Discussion: https://github.com/numba/numba/pull/7516
• Mypyc update
  • every file I tried failed with AssertionError internal to mypyc
• Val: with-handling refactor #7498
  • ready for general review and testing
  • also look for miscompilation
• Python 3.10 and release plan
  • depends on #7498 and WIP python3.10 patch
  • will know more end of week

1. New Issues

• #7517 - Typeguard fail in mangled flags tests.
• #7518 - Segfault with caching and parallel=True.
• #7519 - GDB doesn't work with parallel=True.
• #7521 - numpy.clip accept numpy.array for a_max
• #7523 - Lowering error in parallel loop

Closed Issues

• #7512 - SIGSEGV related to intrinsic
• #7513 - Support CUDA 11.4
• #7520 - trying to pick random numbers on a simple array fails

2. New PRs

• #7515 - Make *var work for randn for parfors.
• **** #7516 - Refresh the registries in get_function early and always.
  • Related https://github.com/numba/numba/issues/7507
• #7522 - Don't embed current thread count into parallel=True functions so that they can be cacheable.

Closed PRs

• #7514 - Fixup cuda debuginfo emission for 7177

3. Next Release: Version 0.55.0/0.38.0, RC=Early Nov

• Request for 0.55
  • broadcast_to https://github.com/numba/numba/pull/7119
    • driven by need for vectorize support in aesara
    • Now merged