Minutes_2021_08_24

Numba Meeting: 2021-08-24

Attendees:

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

numba 0.54.0 & llvmlite 0.37.0 released
0.54.1 milestone
- https://github.com/numba/numba/milestone/60
- pydata/sparse reported segfault
- wheel numpy dependency problem
continue MLIR discussion
- numba-mlir repo: https://github.com/intel/mlir-extensions
next week: continue discussion on pyomp

#7336 - Possible caching issue causing regression in PyData/Sparse in 0.54
- Previous failures of the integration-test led to CI exhaustion and the assumption that it was "broken anyway" and then the caching issue popped up.
#7328 - Possible misnomer in cuda.local.array memory documentation
#7327 - Numba does not error out when an if statement does not account for all conditions
- Feature request
#7326 - CUDA: NVVM crash with reproducer
#7325 - Issue with caching in Jupyterlab
#7324 - fast_matmul in docs example not working properly - returning wrong results
#7323 - Numba 0.54.0 Final Checklist
#7322 - A case of improving the "Segmentation fault" error
#7314 - cuda.local.array produces TypingError even though the shape argument is defined at the top of the script (i.e. globally)
#7310 - LinkedList traversing and assignment fails
#7308 - Compilation fails when creating dicts with lists as mapped values

#7330 - Cuda matmul docs
#7329 - Remove usage of the word "allocate" in reference to local memory
#7321 - Speed up import time by not importing unittest or multiprocessing until it is needed
#7318 - Defensive copy IR when doing obj looplift.
#7315 - Link ELF obj as DSO for radare2 disassembly CFG
#7311 - Temp/checkpr7294
#7309 - Update MapConstraint to handle type coercion for typed.Dict correctly.
#7307 - Prevent mutation of arrays in global tuples.
#7306 - XML test reporting in Azure

Request for 0.55
- broadcast_to https://github.com/numba/numba/pull/7119
  - driven by need for vectorize support in aesara
  - Now merged