Minutes_2021_07_27

Numba Meeting: 2021-07-27

Attendees: Siu Kwan Lam, Graham Markall, Todd A. Anderson, Valentin Haenel, Brandon Willard, Guilherme Leobas

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

RC3
- plan for early next week
broadcast_to debugging, on PR https://github.com/numba/numba/pull/7119
- follow up meeting/debugging session after this call
Other discussions
- Removing issue category in title; e.g. "[FEATURE REQUEST] issue name" for consistency?
  - Many details to consider. postpone discusion.
  - Perhaps we simply "try to improve the title" as and when we merge, to establish a better understanding of what consititutes a good PR title, before we introduce a protocol that everyone must abide by.
- Short report on testing out Anaconda provided alternative to Binder to run live notebooks.

#7249 - Different behavior when using @njit and @cfunc with function wrappers
- surprising behavior with global not-freezing in cfunc
#7248 - [FEATURE REQUEST] Crash when no permissions to access cache directory
- needs to clarity message
#7246 - Slow prange/bad parallelization
- CUDA Target and Typed List actually have similar issues with negative indices and we are not internally consistent
- Todd suggested raising a performance warning on unsigned+signed->float cast
#7245 - [Feature Request] "Iteration" over empty tuple
- Maybe dead branch prunin can help, though it may not have enough type information to prune the empty tuple
#7239 - Working with reshape and np.matmul in Numba
#7225 - Renamed variable names may not be unique

#7247 - Solve memory leak to fix issue #7210
#7244 - Implement partial constant propagation pass
#7243 - Check for variables in parfor loop nest and init block for global variable pushing.
#7241 - Set aliasing information for inplace_binops..
#7236 - CUDA: Skip managed alloc tests on ARM
#7234 - add THREADING_LAYER_PRIORITY & NUMBA_THREADING_LAYER_PRIORITY
#7233 - Update postproc to general multiple del nodes if a variable is assigned multiple times in the same basic block
#7232 - Add conda list export.
#7229 - DO NOT MERGE: Debug for mypy issue.
#7228 - Fixes issue reported under #7217
- Warnings first #7223, fixes later
#7226 - Add support for structured types in Device Arrays
#7224 - [WIP] Improve flake8 compliance and code style (2)
#7223 - Replace assertion errors on IR assumption violation
#7222 - CUDA: Warn when debug=True and opt=True

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