Minutes_2023_11_21

Numba Meeting: 2023-11-21

Attendees: FPOC (last week): FPOC (incoming):

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

Please refer to this calendar for the next meeting date.

py312 PR is now Ready for Review: https://github.com/numba/numba/pull/9246
Github fixes needed:
- Force push should invalidate the review ✅
- Only people with the review-bit should be able to review ✅
Test the AArch64 relocation overflow fix if you can:
- llvmlite PR #1009: Fix relocation overflows by implementing preallocation in the memory manager
Discuss if we should build ppc64le for 3.12 via conda-forge
- NVIDIA tests with dev packages for ppc64le
- Decision: don't try this for RC1, revist when 3.12 support for other platforms stabilizes.
https://github.com/esc/numba-dev-files/blob/master/Makefile

numba#9281 - Add option to nb.njit to suppress performance warning
numba#9282 - ufunc.reduce error if axis is negative
** numba#9283 - Identityless reduction fails to compile when axis is a tuple
numba#9285 - bug: identity of objects in typed lists not preserved
- is fixable using back-references
numba#9286 - Bug: using constructor in jitclass causes copy list of instances to fail
numba#9288 - Python abort on any call of guvectorize with parallel=True
- This is because the workqueue threading layer is not thread safe, use tbb instead
- Some OpenMPs work, others don't, thread safety is not part of the API
numba#9289 - [Python 3.12] [Task] Profiler test failures due to Python 3.12 bytecode changes.
numba#9290 - [Python 3.12] [Task] Parfor test failures due to Python 3.12 bytecode changes.
numba#9291 - [Task] Tests failures from Python 3.11 in continuation with Python 3.12
numba#9293 - Using CUDA shared memory does not appear to work.
numba#9294 - Return structured array as part of tuple fails.
- This is potentially a duplicate array
llvmlite#1015 - Many tests fail

numba#9284 - core dump when returning a list of a deferred type (even if instantiated?)
numba#9287 - numba rewrites data instead of error raising
llvmlite#1011 - Python 3.12

numba#9280 - [WIP] Advanced Indexing #3: Added support for multiple multidimensional Indices
numba#9292 - CUDA: Switch cooperative groups to use overloads
llvmlite#1009 - Fix relocation overflows by implementing preallocation in the memory manager

(last numba: 9294; llvmlite 1015)