Minutes_2024_02_13
esc edited this page Feb 20, 2024
·
1 revision
Attendees: FPOC (last week): Kaustubh FPOC (incoming):
NOTE: All communication is subject to the Numba Code of Conduct.
Please refer to this calendar for the next meeting date.
- 0.59.0rc1 milestone close out: https://github.com/numba/numba/milestone/73
- revive benchmarking suite. relates to https://github.com/numba/numba/pull/9428
- Need a policy on adding new stuff to deprecated features. xref #8143 which is adding fastmath to cc.export (pycc is pending deprecation).
- Team has decided on the policy to not to add any new features/functionality to things that have already been deprecated.
- Discussion on office hours at different time zones
- During Office Hours
- Jim P presented https://github.com/jpivarski-talks/2024-02-13-numba-usage-stats
- numba#9424 - Support math.nextafter in nopython mode
- numba#9426 - Update changelog script to account for missing author info on github.
-
numba#9427 -
searchsorted
with dtypedatetime64
stopped working innjit
mode going from 0.58.1 to 0.59 - numba#9429 - GPU target doesn't emit compile events
- numba#9431 - explicit specialization does not respect caching
- numba#9432 - Bug: (regression) cooperative grid sync w/ cache=True - working in 0.58.1, not working in 0.59.0
- numba#9433 - np.median raises AssertionError for empty arrays while numpy returns nan
- numba#9434 - Integrate ZLUDA for AMD CUDA
-
numba#9435 -
nextafter
(via bothmath
andnumpy
) missing for CUDA
-
numba#9428 - Optimize
range()
to enable more auto-vectorization - numba#9430 - Type system implementation #2: Added builtins support for the new type system.
- merged - numba#9425 - Post release for 0.59.0
- merged - llvmlite#1032 - v0.42 Post release
(last numba: 9435; llvmlite 1032)
2024-gantt: TBD 2023-gantt: https://github.com/numba/numba/issues/8971