Minutes_2021_11_09
Valentin Haenel edited this page Nov 9, 2021
·
1 revision
Attendees: Siu Kwan Lam, Graham Markall, Jim Pivarski, Nick Riasanovsky, stuart, val, Luk, Todd Anderson, brandon willard
NOTE: All communication is subject to the Numba Code of Conduct.
- Python 3.10 update
- ~20/9000 test failures (getting close)
-
hash(float("nan"))
changed to base onid()
so it is random
- Discussion on Don't embed current thread count into parallel=True functions so that they can be cacheable. #7522
- Discussion on Flags
- these:
@njit(flags1=..., flags2, ..., opt=...)
- these:
- #7526 - Implement np.shape for nestedarray
- #7527 - Incorrect result of info locals and info args commands when debugging w/ GDB
- #7528 - Implement array reshape for CUDA
- #7530 - Reminder to postpone deprecation of eager compilation of CUDA device functions.
- #7531 - NumbaWarning: The TBB threading layer requires TBB version 2021 update 1 or later i.e., TBB_INTERFACE_VERSION >= 12010. Found TBB_INTERFACE_VERSION = 11102. The TBB threading layer is disabled. warnings.warn(problem)
- #7532 - numba-core-cpu-importing-stuff-from-numba-experimental
- #7533 - Missed optimization on typed lists
- #7534 - Include non-temporal JIT annotations to speed up memory-bound codes filling big arrays
- #7535 - Slower string manipulation performance than CPython
-
#7536 - from numba.core.typeconv import _typeconv
- Suggest to document this and all the reasons for why it can happen. The reason why people are suggesting to install 0.47.0 is because this is the last version, before we refactored everything into
core
.- this can happen if old and new versions of Numba become conflated
- this can also happen....
- Suggest to document this and all the reasons for why it can happen. The reason why people are suggesting to install 0.47.0 is because this is the last version, before we refactored everything into
- #7540 - [Docs] Jitclass support for type annotation not documented on main page
- #7541 - [Idea] A less verbose @jitclass interface with type hints
-
#7542 - [Docs] Wrong link to
np_shape
source code in docs
- #7524 - Added overload methods for Numpy's ufunc.reduce
- #7529 - Exp/3.10+new with detection
- #7537 - Add chrome tracing
-
#7538 - Fix
@overload_glue
performance regression. - #7539 - Fix str decode issue from merge #7525/#7506
- #7543 - Support for all modes in np.correlate and np.convolve
- #7544 - Exp/3.10+new with+np121
-
#7525 - Make sure
demangle()
returnsstr
type.
- Request for 0.55
-
broadcast_to
https://github.com/numba/numba/pull/7119- driven by need for
vectorize
support in aesara - Now merged
- driven by need for
-