Sebastian Berg
Sebastian Berg
> I tried and failed. One is C++ the other C Should work, I think but I guess you could defer that for now. You will have to export an...
> I haven't had to run tests against that ever You could first try to run python `PYTHONMALLOC=malloc_debug`, although getting a debug python is hopefully not a lot of trouble....
> There's a segfault and the docs need updating since output is not sorted now Whoops, need to not try to iterate (I think the setup is OK though). The...
Just to note the observation: The s390x _build_ segfaulted once here. (i.e. not our test, but the compilation)
Let's give this a shot, thanks @adrinjalali, we discussed it briefly yesterday and I don't think it is helpful to try to iterate here more. There are a lot of...
One semi related thing which I would 100% support and think we should do is introduce a `lerp` (better name) ufunc, since that is in more modern C++ versions also...
@kraj I think this should also be added to `numpy/core/include/numpy/npy_endian.h`. Does the full test suit pass with this patch? @aurel32 this may be a long shot, but since you added...
Right now if they feed a cupy array and cucim doesn't support it, they would just get an error, since the scikit-image function should just fail (no auto-conversion from cupy...
FWIW, I have no strong opinion either way. Although, _if_ you allow 0-D the question is why not allow any broadcasting (except along the concatenated dimension) and it may be...
https://github.com/numpy/numpy/issues/28549 but I am not aware of old discussions (which coesn't mean they don't exist).