Improves reduction code and removes kernels/boolean_reductions.hpp

[ndgrigorian]

This pull request further modularizes reduction code and fully moves boolean reductions (any and all) functionality into reductions.hpp.

Sequential reductions now have a submit_sequential_reduction wrapper, reducing repetitious kernel calls.

Boolean reductions now use the same basic kernels as other reductions.

Additionally, reduction type dispatching has been moved into individual reduction files, i.e., reductions/max.cpp now contains the dispatch factories, and so on for all reductions. This significantly reduces the amount of code in kernels/reductions.hpp.

• 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?
• If this PR is a work in progress, are you opening the PR as a draft?

[github-actions]

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

[coveralls]

coverage: 91.148%. remained the same
when pulling 713a177 on further-reduction-refactoring
into 7ab3731 on master.

[github-actions]

Array API standard conformance tests for dpctl=0.17.0dev0=py310h15de555_21 ran successfully.
Passed: 907
Failed: 0
Skipped: 88

[github-actions]

Array API standard conformance tests for dpctl=0.17.0dev0=py310h15de555_22 ran successfully.
Passed: 904
Failed: 3
Skipped: 88

[oleksandr-pavlyk]

New test failure:

_________________________ test_max_min_nan_propagation _________________________

    def test_max_min_nan_propagation():
        get_queue_or_skip()
    
        # float, finites
        x = dpt.arange(4, dtype="f4")
        x[0] = dpt.nan
>       assert dpt.isnan(dpt.max(x))
E       AssertionError: assert usm_ndarray(False)
E        +  where usm_ndarray(False) = <UnaryElementwiseFunc 'isnan'>(usm_ndarray(-inf, dtype=float32))
E        +    where <UnaryElementwiseFunc 'isnan'> = dpt.isnan
E        +    and   usm_ndarray(-inf, dtype=float32) = <function max at 0x7fbe36c69580>(usm_ndarray([nan,  1.,  2.,  3.], dtype=float32))
E        +      where <function max at 0x7fbe36c69580> = dpt.max

/usr/share/miniconda/envs/test_dpctl/lib/python3.11/site-packages/dpctl/tests/test_usm_ndarray_reductions.py:140: AssertionError

[ndgrigorian]

New test failure:

_________________________ test_max_min_nan_propagation _________________________

    def test_max_min_nan_propagation():
        get_queue_or_skip()
    
        # float, finites
        x = dpt.arange(4, dtype="f4")
        x[0] = dpt.nan
>       assert dpt.isnan(dpt.max(x))
E       AssertionError: assert usm_ndarray(False)
E        +  where usm_ndarray(False) = <UnaryElementwiseFunc 'isnan'>(usm_ndarray(-inf, dtype=float32))
E        +    where <UnaryElementwiseFunc 'isnan'> = dpt.isnan
E        +    and   usm_ndarray(-inf, dtype=float32) = <function max at 0x7fbe36c69580>(usm_ndarray([nan,  1.,  2.,  3.], dtype=float32))
E        +      where <function max at 0x7fbe36c69580> = dpt.max

/usr/share/miniconda/envs/test_dpctl/lib/python3.11/site-packages/dpctl/tests/test_usm_ndarray_reductions.py:140: AssertionError

This was caused by the atomic fetch_min/max operations being called for reductions which used our custom implementations.

To fix this, I added more utilities for checking if an operator is specifically a sycl op. That way if a work-around is needed for reduce_over_group in the future, but we are still using the sycl op, we can leverage atomics better.

[github-actions]

Array API standard conformance tests for dpctl=0.17.0dev0=py310h15de555_23 ran successfully.
Passed: 906
Failed: 1
Skipped: 88

[oleksandr-pavlyk]

LGMT! Thank you @ndgrigorian !