Skip to content
esc edited this page Mar 29, 2022 · 1 revision

Numba Meeting: 2022-03-22

Attendees: Andre Massella, stuart, Siu Kwan Lam, Benjamin Graham, Jim Pivarski, Todd A. Anderson, Guilherme Leobas,

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

Please refer to this calendar for the next meeting date.

0. Feature Discussion

  • Jim: New project from author of cppyy
    • auto generated C++ binding for Python
      • Wim Lavrijsen
      • Developed cppyy, which are automated bindings between C++ and Python (non-JIT), more automated than Cython/pybind11/SWIG, etc.
      • Part of ROOT to make a huge C++ codebase usable in Python (as PyROOT) since 2004. Standalone cppyy accessible on BitBucket and PyPI and conda-forge.
      • He's planning/soon will be working on a project to link it to Numba (in what capacity? not sure) A real, funded project, possibly contingent on being able to point to potential users at DOE labs. Jim has connected Wim with Lindsey Gray at Fermilab; crossing fingers, hoping this actually happens.
  • Siu: shares suggestions/proposal for Python bytecode accessibility for source reconstruction
  • Stu: BitGenerator and fastpath in C dispatcher

1. New Issues

  • #7908 - numba.typed.typeddict.Dict.get returns the default even though the key is present
  • #7909 - Delays on first operations with numba objects
  • #7912 - NUMBA_DUMP_IR generate duplicates
  • #7913 - Using a function that prints inside another AOT function causes RuntimeError: missing Environment
  • #7916 - Silently incorrect compilation involving nested functions and asserts
  • #7917 - Potential enhancements after dropping python 3.7
  • #7921 - BUG: bytes type equals goes through ufunc mechanism
  • #7922 - gcd, lcm, subtract, sign, and negative don't match numpy behavior with bool inputs
  • #7923 - Tuple of functions is not iterable / cannot be accessed by index
  • #7925 - typed.List 2x slower than reflected list (reproduced on Intel and AMD CPUs)
  • #7926 - CUDA: remove context query in launch configuration
  • #7927 - CUDA: discussion: retire the use of a separate launch configuration object, launch kernels directly!

Closed Issues

  • #7910 - Missing sys import in parfor_lowering
  • #7914 - Using a function that prints inside another AOT function causes RuntimeError: missing Environment

2. New PRs

  • #7918 - Add JIT examples to CUDA docs
  • #7919 - Disallow //= reductions in pranges.
  • #7920 - Drop pre-3.7 code path (CPU only)
  • #7924 - Retain non-modified index tuple components.

Closed PRs

  • #7911 - added sys import
  • #7915 - CUDA: Fix test checking debug info rendering.

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

Clone this wiki locally