8000 ENH: CreateSortedStridePerm suboptimal · Issue #434 · numpy/numpy · GitHub
[go: up one dir, main page]
Skip to content

ENH: CreateSortedStridePerm suboptimal #434

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
seberg opened this issue Sep 10, 2012 · 0 comments
Closed

ENH: CreateSortedStridePerm suboptimal #434

seberg opened this issue Sep 10, 2012 · 0 comments

Comments

@seberg
Copy link
Member
seberg commented Sep 10, 2012

The CreateSortedStridePerm does not work correctly when it comes to non C ordered arrays that include a 1 dimensional axis somewhere in the middle (and are not contiguous as this is typically optimized anyways). Its a corner case, but since I noticed thought I would post it.

For these cases it will fail to sort the array and this results unnecessarily slow operations for some things. (IE: e = np.empty(400,400,400)[::2,::2,::2]; e[:,0,:] = 5) is much faster then e[:,0:1,:] = 5. Also this means that e.copy('k') is not Fortran order, which maybe it never was, but seems not what one would expect.
seberg added a commit to seberg/numpy that referenced this issue Oct 21, 2012
@seberg
BUG: Fix bug with size 1-dims in CreateSortedStridePerm
dc4e38b 
This reverts changes done to CreateSortedStridePerm in commit
9194b3a. The problem is that it would fail for 3x1x3 Fortran order
array for example. And special handleing seems unnecessary at least
after 1-dim axis not mattering for contiguous flags.
This "closes Issue numpy#434"
@seberg seberg mentioned this issue Oct 22, 2012
Merged
seberg added a commit to seberg/numpy that referenced this issue Oct 23, 2012
@seberg
BUG: Fix bug with size 1-dims in CreateSortedStridePerm
e565afb 
This changes CreateSortedStridedPerm to not use the shape
for special casing 1-dim axis. The cleanup does not seem
to be useful in most cases and the current way is buggy.
Also insert stride so that reduce with keepdims=1 will
keep contiguous arrays contiguous. "Closes Issue numpy#434"
@seberg seberg mentioned this issue Oct 23, 2012
Merged
@seberg seberg closed this as completed Oct 25, 2012
teoliphant pushed a commit to ContinuumIO/numpy that referenced this issue Nov 6, 2012
@seberg @teoliphant
BUG: Fix bug with size 1-dims in CreateSortedStridePerm
fad7026 
This reverts changes done to CreateSortedStridePerm in commit
9194b3a. The problem is that it would fail for 3x1x3 Fortran order
array for example. And special handleing seems unnecessary at least
after 1-dim axis not mattering for contiguous flags.
This "closes Issue numpy#434"
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@seberg
0