Minutes_2021_05_11
Valentin Haenel edited this page May 17, 2021
·
1 revision
Attendees:
NOTE: All communication is subject to the Numba Code of Conduct.
-
last week stabilizing mainline
- race condition problem because
cached_property
not threadsafe
- race condition problem because
-
anything urgent for 0.54
-
Discussion on
slice
boxing (https://github.com/numba/numba/pull/6939)In [6]: @njit ...: def foo(v): ...: a = slice(1, 2, None if v else 3) ...: return a.step ...: ...: In [7]: foo(1) --------------------------------------------------------------------------- TypeError Traceback (most recent call last) <ipython-input-7-7d8d25071659> in <module> ----> 1 foo(1) TypeError: expected int64, got None In [8]: foo(2)
step
handling happened: https://github.com/numba/numba/blob/6b82cd7b508b17d9eeb48e54f22dd18c67b711a2/numba/cpython/slicing.py#L166-L174 making optional values: https://github.com/numba/numba/blob/master/numba/core/base.py#L735-L744
- #7012 - pow() producing incorrect results for negative integer exponent
-
#7006 - Call to cuLinkAddData results in UNKNOWN_CUDA_ERROR
- OS distro packaging problem unrelated to numba
- #7002 - Purely scalar inputs break njit'd functions that utilize np.stack
-
#6998 - np.linalg.solve unexpectedly returns an
F_CONTIGUOUS
array forC_CONTIGUOUS
inputs- mismatch in np implementation details on array layout
-
#6993 - Generators don't get cleaned up on
- memory leak in generator
- #7011 - WIP: Add the check of the 'noalias' attrubute before lowering 'inplace_bin' op, this fix #6949 .
- #7010 - WIP: Add ddof argument to nanvar and nanstd
- #7009 - Add writable args
- #7008 - Added support for np.random.dirichlet with all size arguments
- #7007 - Made stencil compilation not fail for arrays of conflicting types.
- *** #7005 - Make
numpy.searchsorted
match NumPy when first argument is unsorted - #7004 - Test extending the CUDA target
- #7003 - More mainline fixes
-
#7001 - Make some
numba.core
modules flake8 compliant - #6997 - CUDA: Remove catch of NotImplementedError in target.py
- *** #6996 - Add constant lowering support for
SliceType
s- Todd suggests adding test using in parfors
- #6995 - Resolve issue where nan was being assigned to int type numpy array
- #6992 - Moved typedlist iter into njit code instead of just getitem