Evgeni Burovski

### Description `ndimage` rotate by 180 degrees and back should give the input array back; instead it changes the bottom right corner value. ### To Reproduce ```py In [1]: import...

`ndimage.{white,black}_tophat` differs from scipy

2

### Description Results of `cupyx.scipy.ndimage` black and white tophat differ from the corresponding scipy invocations. ### To Reproduce ``` In [1]: import numpy as np In [2]: import cupy as...

### Description `ndimage.grey_erosion` for some inputs gives results different from those of scipy.ndimage ### To Reproduce ```py In [1]: import numpy as np In [2]: import cupy as xp In...

ndimage.distance_transform_edt does not accept cupy arrays for indices

1

### Description `cupyx.scipy.ndimage.distance_transform_edt` accepts numpy arrays as indices; For a cupy array it raises ``` >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> PDB post_mortem (IO-capturing turned off) >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> > /home/ev-br/.conda/envs/scipy-dev/lib/python3.11/site-packages/cupyx/scipy/ndimage/_pba_2d.py(257)_check_indices() -> raise RuntimeError("indices array has wrong...

cupyx.ndimage.affine_transfrom raises for seemingly valid inputs

1

### Description `cupyx.scipy.ndimage.affine_transfrom` raises a `RuntimeError: improper affine shape` for inputs which are OK for the `scipy.ndimage.affine_transform` ### To Reproduce ```py In [10]: import numpy as np In [11]: import...

#### Reference issue cross-ref https://github.com/scipy/scipy/pull/21012, https://github.com/scipy/scipy/pull/20772 #### What does this implement/fix? This is a POC of the layer of dispatch/delegation from `scipy.ndimage` to `cupyx.scipy.ndimage`, for a full scipy submodule. `ndimage`...

ENH: array types, signal: delegate to CuPy and JAX for correlations and convolutions

59

Towards gh-20678 #### What does this implement/fix? Delegate `signal.convolve` and its ilk to `cupyx.scipy.signal.convolve` if inputs are cupy arrays. For other array types, do the usual convert to numpy, run,...

An experiment to fish out the source of https://github.com/OpenMathLib/OpenBLAS/issues/4776 : cannot repro the error locally, let's see what the GH runners do

ENH: Decouple the benchmark runner from viewer

2

Consider a use case: I've a test suite which uses, say, `pytest-benchmark` and I'd like to be able to reuse `asv publish` to web display. A first step could be...

ENH: `asv gh-pages` custom commit message

2

Would be nice if `asv gh-pages` could grow a cmdline argument to specify a custom commit message to override the default "Generated from sources".