dpnp.atleast_3d() is very slow and have a different shape compared to numpy.atleast_3d() with 2d array.

[npolina4]

dpnp.atleast_3d() have a different shape compared to numpy.atleast_3d() with 2d array.

In [1]: import dpnp, numpy

In [2]: xnp = numpy.ones((1000, 1000))

In [3]: %time numpy.atleast_3d(xnp)
CPU times: user 17 µs, sys: 4 µs, total: 21 µs
Wall time: 24.6 µs

In [4]: x = dpnp.ones((1000, 1000))

In [6]: %time dpnp.atleast_3d(x)
CPU times: user 2min 1s, sys: 1min 4s, total: 3min 6s
Wall time: 2min 52s

And dpnp.atleast_3d() have a different shape compared to numpy.atleast_3d() with 2d array.

In [1]: import dpnp, numpy

In [2]: xnp = numpy.ones((1000, 1000))

In [3]: numpy.atleast_3d(xnp).shape
(1000, 1000, 1)

In [4]: x = dpnp.ones((1000, 1000))

In [5]: dpnp.atleast_3d(x).shape
(1, 1000, 1000)