Add get_devices method to dpctl.SyclPlatform class

[ndgrigorian]

This PR proposes implementing the get_devices method to dpctl.SyclPlatform class as per SYCL 2020 spec, which specifies including a device_type filter.

Cleans up some remaining references to host devices, missed in #1028

Closes #509

• 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
For some context, this PR is setting the stage to support composite devices

It occurred to me that get_devices itself is not supported by dpctl.SyclPlatform, leading to me noticing this long-standing issue.

[github-actions]

View rendered docs @ https://intelpython.github.io/dpctl/pulls/1992/index.html

[github-actions]

Array API standard conformance tests for dpctl= ran successfully.
Passed: 895
Failed: 1
Skipped: 118

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310h93fe807_521 ran successfully.
Passed: 894
Failed: 2
Skipped: 118

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310h93fe807_521 ran successfully.
Passed: 894
Failed: 2
Skipped: 118

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310h93fe807_523 ran successfully.
Passed: 894
Failed: 2
Skipped: 118

[ndgrigorian]

@oleksandr-pavlyk
I've made the requested change, and made a change to instead give unknown when the function is provided automatic or custom as device types.

Using these device types would yield the same result as all from pure SYCL code, i.e.,

#include <sycl/sycl.hpp>
#include <iostream>

int main()
{
   sycl::platform p{};

   auto DTy = sycl::info::device_type::custom;
   size_t n_devs_all = p.get_devices(sycl::info::device_type::all).size();
   size_t n_devs_custom = p.get_devices(sycl::info::device_type::custom).size();

   std::cout << (n_devs_all == n_devs_custom) << std::endl;

   return 0;
}

will always show 1. Same if device_type::automatic is used. It seems that in the compiler, it defaults to all unless cpu, gpu, or accelerator are provided.

To spare user confusion, I think it makes more sense to return empty lists for custom and automatic.

[coveralls]

coverage: 88.176% (-0.06%) from 88.232%
when pulling 6332448 on get-devices-from-sycl-platform
into c6b00c7 on master.

[github-actions]

Array API standard conformance tests for dpctl=0.19.0dev0=py310h93fe807_541 ran successfully.
Passed: 895
Failed: 1
Skipped: 118

[github-actions]

Array API standard conformance tests for dpctl=0.20.0dev0=py310h93fe807_11 ran successfully.
Passed: 895
Failed: 2
Skipped: 125

[antonwolfy]

@ndgrigorian, I have a minor comment, but in overall LGTM

[antonwolfy]

I believe it might be helpful to place a comment somewhere (and/or cover in tests) that automatic and custom device type are handling as unknown.

[ndgrigorian]

That's a good idea