Skip to content
Valentin Haenel edited this page Mar 30, 2020 · 1 revision

Numba Meeting: 2019-12-03

Attendees: Todd, Aaron, Val, Stuart, Stan, Siu, Pearu, James

0. Feature Discussion

  • Status on stabilizing master
    • a problem with inferred dtype for np.arange
    • Much hacking has gotten us as far as we can until we can distinguish "python-like" integer types from "numpy-like" integer types
    • hacked test to accept dtype mismatch (int64 instead of int32) on win-64 only
    • will merge by end of today if everything works

1. New issues

  • #4914 - Cryptic error when compiling a function obtained via exec with bad __name__

    • not a bug
    • should give a better error message
  • #4913 - feature request, 3 argument form of pow built in

    • low priority
  • #4912 - bitarray support

  • #4908 - RuntimeError: cannot cache function 'square': no locator available for file

    • bug; pip; egg; pkgresource
  • #4906 - Generator performance

    • Refct impact perf
    • hard
  • #4905 - Numba typed list not supported with NUMBA_DISABLE_JIT

    • let typedlist fallback to list in disabled jit
    • coverage
  • **** #4903 - Parallel prange gives LoweringError

    • failing in type infer after parfor transformation
  • #4899 - Callwrapper cleaning-up too soon

  • #4898 - try excepts support

  • #4897 - Support for numpy.isnat in nopython mode needed since no w/a longer exists

    • feature request
  • #4895 - Numba talks -- scipy 2019

  • #4894 - Using string[:] data types as class attributes.

  • 4922 - prange reduction incorrect when operation happens inside a conditional block

    • need to do a proper analysis instead of simple pattern-matching

Already Closed

  • #4915 - The value of the variable does not change in the while True: loops
  • #4907 - Cuda: Using shared memory to add numbers results in overwriting
  • #4893 - Incorrect array returned from guvectorize

2. New Open PRs

  • #4911 - np.interp bugfix for float32 handling
  • **** #4910 - loop canonicalisation and type aware tuple unroller/loop body versioning passes
    • unroll tuple iteration
    • added guards to explicitly enable the passes
    • getting unittests to pass
  • #4909 - Fix some spelling typos
  • #4904 - Implement np.lcm
  • **** #4902 - [WIP] basic support for try except
    • implemented as syntactic-sugar
    • to-test:
      • check IR-inlining
  • #4901 - fix typed list equals
  • #4900 - [4899] fixed issue with PyCallWrapper and inconsistent Python/nopython ownership trees
  • **** #4896 - Specialise arange dtype on arch + python version.

Already merged/closed

  • #4892 - regex needs to accept singular form of "argument"
  • #4891 - skip the xfail test for now. Py3.8 CFG refactor seems to have changed the test case
  • #4885 - suppress spurious RuntimeWarning about ufunc sizes

4. Next Release: Version 0.47.0, RC=December 19

  • CPython 3.8

  • Requests for 0.47 (last release for the year) - jitclass performance issues - llvm 9 trial - CTK libcudadevrt.a - CI needs to take 50% of current time - Val & Stu already looking at this - also checking Azure CI config to avoid wasting compute time - Caching: - transitive dependency - other issues: i.e. function as argument, with objmode - distributing cache - Immutable list and deprecating reflected list - Switch to pytest (see above) - Using Numba to generate LLVM/NVVM IR for different targets https://github.com/numba/numba/issues/4546 - @overload for gpu

Clone this wiki locally