result_type to raise unless at least one argument is usm_ndarray

[oleksandr-pavlyk]

dpt.result_type is being changed to raise unless at least one argument is dpt.usm_ndarray.

Implementation of dpt.where was calling dpt.result_type with just dtypes, although it has the device available. Implementation of result_type was factored out into a helper function which can accommodate know device, and reused by dpt.result_type and _where_result_type function.

Test files were tweaked to avoid calling dpt.result_type with just dtypes.

Closes gh-1874 and supersedes gh-1875

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to an issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• Have you checked performance impact of proposed changes?
• Have you added documentation for your changes, if necessary?
• Have you added your changes to the changelog?
• If this PR is a work in progress, are you opening the PR as a draft?

[ndgrigorian]

@oleksandr-pavlyk
Does this change go against array API? Per spec:

arrays_and_dtypes (Union[array, dtype]) – an arbitrary number of input arrays and/or dtypes.

In theory, this should mean only dtypes is acceptable. Of course we have device-specific concerns, but, perhaps we should assume the full dtype matrix if no array is specified?

[github-actions]

Deleted rendered PR docs from intelpython.github.com/dpctl, latest should be updated shortly. 🤞

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_147 ran successfully.
Passed: 893
Failed: 2
Skipped: 119

[coveralls]

coverage: 87.681% (+0.008%) from 87.673%
when pulling 0c69706 on address-gh-1874
into 9eb8f03 on master.

[vtavana]

With these changes, a few dpnp tests fail. I am working on updating dpnp accordingly.

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_147 ran successfully.
Passed: 893
Failed: 2
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_148 ran successfully.
Passed: 892
Failed: 3
Skipped: 119

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310hdf72452_149 ran successfully.
Passed: 893
Failed: 2
Skipped: 119

[ndgrigorian]

@oleksandr-pavlyk Does this change go against array API? Per spec:

arrays_and_dtypes (Union[array, dtype]) – an arbitrary number of input arrays and/or dtypes.

In theory, this should mean only dtypes is acceptable. Of course we have device-specific concerns, but, perhaps we should assume the full dtype matrix if no array is specified?

New array API test failure introduced also seems to be a result of this problem.

[oleksandr-pavlyk]

Yes, this change goes against array API spec. Closing without merging.