Minutes_2022_05_17
esc edited this page May 18, 2022
·
1 revision
Attendees: Siu Kwan Lam, Andre Masella, Todd A. Anderson, Graham Markall, stuart, Kaustubh Chaudhari, Val, Benjamin Graham, LI Da, Nicholas Riasanovsky, Guilherme Leobas, Luk,
NOTE: All communication is subject to the Numba Code of Conduct.
Please refer to this calendar for the next meeting date.
- Update on 1.22 &
main
status - Cutting a 0.55.2 -- NumPy 1.22 support and M1 conda pkgs and wheels -- coming soon (this week)!
- Timeline 0.56.0 RC1 -- needs a burndown
- LLVM 13 & 14 upgrade
- Missing SVML patches are blocking
- Unknown quantity of work, some other blockers
- #8029 - convert NamedTuple to NamedUniTuple when possible
- self service notes
- The questions was: "if I only have three hours per week, how can I contribute to Numba in a meaningful way"
- Maybe have a board with issues that you can grab, suitable for working on even w/o much Numba experience, tasks which have a known level of "done"
- Maybe need some guidelines for how to review Numba
- Add a contributing.md
- There are open issues around refactoring code and/or tests
- May be feature contributions but also refactoring and review
- Board should contain a column for PRs submitted by 1st time contributors and update the FPOC guidelines to reflect they should be added.
- Mission statement PR: almost ready, hopefully to be merged tomorrow.
- Anticipating releases and their contents:
- make it more visible when the next release will come out
- i.e. a public schedule of when releases will come out and what LLVM/CPython/NumPy will be supported
- Cadence: 6-12 months ahead (2-4 releases)
- Record what happened when
- Luk: what is the chances of this being up to date, this needs an item on the release checklist to update the schedule
-
#8034 - Feature request: support for
numba._DUFunc
in@overload
-
#8036 -
raise
andassert
do not work correctly in CUDA device functions - #8039 - np.random.randint with size
- *** #8047 - Overload not properly producing omitted types
- #8051 - Corrupted(?) return values from record arrays
- #8055 - much slower for 3d tensor
- #8059 - receive segfault when using objmode and starargs (iterating on typed-list within object mode?)
- #8060 - Support issuing warnings in nopython mode
- #8043 - Python 3.10 kwarg peephole rewrite failing in valid case (breaks mainline)
- #8045 - Windows caching tests failing with FileNotFoundError
- #8048 - TypingError: Cannot unify ListType[int64] and list(int64)<iv=None> during using list comprehension
- #8054 - much slower for 3D-tensor calculate
- #8032 - Pass null pointer args through as is
- #8033 - DO NOT MERGE: try testing np randomgen
- #8037 - CUDA self-recursion tests
- #8038 - Support for Numpy BitGenerators PR#2: Standard Distributions support
- #8040 - Support for Numpy BitGenerators PR#3: Advanced Distributions Support.
- #8041 - Support for Numpy BitGenerators PR#4: Geneator().integers() Support.
- #8042 - Support for Numpy BitGenerators PR#5: Geneator Shuffling Methods.
- #8056 - Move caching tests from test_dispatcher to test_caching
- #8057 - Fix coverage checking
- #8035 - Fix a couple of typos RE implementation
- #8044 - Make Python 3.10 kwarg peephole less restrictive
- #8046 - Fix caching test failures
- #8049 - support str(bool) syntax
- #8050 - [Testing only] combine #8044 #8046
- #8052 - Ensure pthread is linked in when building for ppc64le.
- #8053 - [TESTING ONLY] Misc/combine 8044 8046 8052
- #8058 - Fix/pin llvmlite 039