Minutes_2020_10_06
stuartarchibald edited this page Oct 7, 2020
·
2 revisions
Attendees: Todd, Guilherme, Val, Siu, Graham, Stuart, Juan
-
Burndown:
- NumPy 1.19
- skip ediff1d test.
- numpy changed behavior for some edge cases
- adjusting numba code to match is too big and complicated
- there's also a bug in numpy.
- add to CI
- skip ediff1d test.
- LLVM 10 build 2, with SVML patch
- fixes a SVML code-generation bug against AVX512
- llvmlite enable PIC https://github.com/numba/llvmlite/pull/627
- for linux & hopefully BSD
- Refprune LLVM pass being reviewed
- cleaning up PR but review welcome
- Check with Todd:
- https://github.com/numba/numba/pull/6224
-
https://github.com/numba/numba/pull/5892
- stuart: move commit series to new PR
- NumPy 1.19
-
Next release plan
-
maybe a patch release for Py3.9
- should just be one new opcode
-
0.53.0rc1 plan to tag in Jan 2021
-
Ask:
- librosa
- RAPIDS
- datashader
- umap
- poliastro
- clifford
- AwkwardArray
- pydata-sparse
- stumpy
- Bodo
-
Protocol:
- 3 Weeks timebox
- open an issue on issue tracker and ping major contributors
- Build and test any projects that we don't hear from ourselves
- If we don't encounter any issues -> cut the final
- else: fix until not broken
-
-
Community meeting next week?!
- type annotation
- numba-"extras"
-
#6293 - Typer does not propagate return type to all return variables
- patch 6294
- #6288 - Strange behavior with the mean() function
- #6301 - Implement np.logspace and np.geomspace
- #6302 - Implement np.gradient
- #6291 - Missing 'same_kind' case in ufunc_find_matching_loop?
- #6300 - Un-skip tests that were skipped because of #4026.
- #6299 - Fix flake8 violations in numba.cuda.simulator
- #6298 - Fix flake8 violations in numba.cuda.tests.cudadrv
- #6297 - Fix flake8 violations in numba.cuda.tests.cudapy module
- #6296 - Fix flake8 violations in numba.cuda module
- **** #6295 - Enhance inspect_cfg
- #6294 - Make return type propagate to all return variables
- #6292 - Document support for structured array attribute access
- #6290 - CUDA: Add support for complex power
- #6289 - Refactor atomic test cases to reduce code duplication
- #6303 - Merge cuda flake8 prs
- Requests for 0.52
- Fast(er)
typed.List
/typed.Dict
? Doesn't have to be in 0.52, next 3-5 months is fine.
- Fast(er)