CLN: assorted #46181
CLN: assorted #46181
Conversation
| bint is_datetimelike, | ||
| bint skipna=True) -> None: | ||
| def group_cumprod_float64( | ||
| float64_t[:, ::1] out, |
There was a problem hiding this comment.
@realead i stumbled on your indirect_buffer repo and thought you might have some advice on further optimizing some_type[:, ::1] out in cases where out was created with np.empty(shape, dtype=whatever). Can we get extra mileage with something like view.indirect_contiguous?
(Just commented on this PR bc the diff included a line with the syntax of interest)
There was a problem hiding this comment.
in cases where
outwas created withnp.empty(shape, dtype=whatever). Can we get extra mileage with something likeview.indirect_contiguous?
IIIC: No. Numpy arrays don't support indirect layout - their layout is always direct. indirect layout would make sense e.g. for combining multiple 1D-numpy arrays into one 2D-buffer without copying the data from the original numpy arrays. I don't see how this can be applied here - as you have pointed out, out will be created by np.empty, so it has direct layout - nothing to be improved upon.
doc/source/whatsnew/vX.X.X.rstfile if fixing a bug or adding a new feature.