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

Numba Meeting: 2020-01-06

Attendees: Stuart, Ehsan, Graham, Aaron, Pearu, Todd

0. Feature Discussion

  • 0.48 plans
    • any leftover PRs from 0.47 and anything in flight before 0.49 refactor
    • thread masking
      • plan to get in early 0.49
    • first-class function type
      • split PR
        • merge cfunc support ASAP
        • Pearu continues to work on njit
  • CUDA CI
    • discussion of possibilty of public CI for GPU code

1. New issues

  • #5037 - Update CI configurations to reflect minimum requirements
  • ** #5036 - Review minimum requirements and version pinning
  • ** #5035 - llvmlite 0.31 breaks numba 0.46
    • pin llvmlite in setup.py of numba
  • #5034 - NotDefinedError
    • Stop asking for new issues in backtraces (not getting enough info from most users to do anything about the reported bugs, which are often user-code related)
  • ** #5033 - SIMD vectorization failure with nditer
  • **** #5031 - parallel computations may ignore conditions
    • Todd has a fix and is testing it
  • #5030 - How do you raise an exception from the unboxing context?
  • ** #5028 - Feature request: line profiler/performance profiler
  • ** #5026 - Typed List sort() struggles key function optional
  • #5025 - Provide documentation, docstrings, tests, examples for dynamic shared memory
  • #5022 - Unhelpful AssertionError when bad arguments passed to cuda.to_device
  • #5017 - cuda.to_device() cannot copy zero-element data

triaged last week:

  • #5009 - UnsupportedParforsError
  • #5005 - @jit is not working for big integers
  • #5003 - Memory leak when allocating arrays in function in parallel loop
  • #5002 - Error model does not effect complex number
  • #5001 - njit(parallel=True) issue
  • #5000 - Closure inliner not handling keyword args correctly
  • #4989 - lldb support
  • #4988 - bug: range accepts float
  • #4985 - Catch returning a non-None value from a jitclass __init__
  • #4979 - Regression finding libNVVM in 0.46.0

Already Closed

  • #5032 - LLVM Assembly output gets truncated
  • #5020 - Numba 0.47.0 final checklist
  • #5015 - semantic constant prop needs to run before static* rewrites
  • #5013 - new bug in numba lib
  • #5011 - literally doesn't work in overload_method
  • #5008 - numba.typed.List does not work with elements of type typing.NamedTuple, but works with collections.namedtuple
  • #5006 - @jit is not working
  • #5004 - Unexplained Issue with mutable list as default argument
  • #4999 - try command support
  • #4996 - [Bug Report] Compile error using prange and ravel
  • #4992 - Numba 0.47.0rc1 Checklist
  • #4986 - error while processing raw.h5ad
  • #4980 - Numba compilation error, calling function from tkinter
  • #4976 - pip install numba fails on windows with python 3.8 https://github.com/numba/numba/issues/2918

2. New Open PRs

  • #5029 - Install optional dependencies for Python 3.8 tests
  • #5027 - Add Python 3.8 classifier to setup.py
  • #5023 - Don't launch ForAll kernels with 0 elements (Fixes #5017)
  • #5016 - Fix various issues in CUDA library search (Fixes #4979)
  • ** #5014 - Enable use of records and bools for shared memory, remove ddt, add additional transpose tests
  • #5012 - Fix IndexError when accessing the "-1" element of dummyarray

\\\\\\triaged last week:

  • #4984 - Update deprecation warning class for reflected list
    • on hold
  • #4982 - Update docs to include disable_reflected_list switch
    • on hold

Already merged/closed

  • #5024 - Propagate semantic constants ahead of static rewrites.
  • #5021 - Update change log with fixes from 0.47.0 RC period.
  • #5019 - skip pycc test on Python 3.8 + macOS because of distutils issue
  • #5018 - Fix #5011.
  • #5010 - merge string prs
  • #5007 - test_consuming_strides: Keep dev array alive
  • #4998 - Fix more spelling typos
  • #4997 - State that CUDA Toolkit 8.0 required in docs
  • #4995 - Change log update for 0.47.0
  • #4994 - Remove xfail for test which has since had underlying issue fixed.
  • #4993 - Update deprecations for 0.47
  • #4991 - Continuation of #4588
  • #4990 - Implement sort on typed list
  • #4987 - Add is_internal property to all Type classes.
  • #4981 - Add ind_to_const to enable fewer equivalence classes.
  • #4978 - Fix overload_method problem with stararg
  • #4977 - Implement sort/sorted with key.
  • #4983 - Add ExtensionType mixin to permit all-type consuming overloads.

4. Next Release: Version 0.48.0, RC=Jan 16

  • Requests for 0.48
    • TBD
Clone this wiki locally