[SYCL][CUDA] Support device ID and UUID in the CUDA Plugin

[zjin-lcf]

This PR tries to enable the accesses to device ID and UUID with the CUDA driver APIs. Thank you for your review.

Reference
https://github.com/intel/llvm/blob/sycl/sycl/doc/extensions/supported/sycl_ext_intel_device_info.md

[zjin-lcf]

The code changes reported by lint are not consistent with the coding style in the plugin.

[abagusetty]

The code changes reported by lint are not consistent with the coding style in the plugin.

I believe the CI uses clang-format-v13 for formatting. Were you using anything other versions ?

[zjin-lcf]

No, I just looked at other codes in the plugin, and then follow that style.

[npmiller]

The linter is consistent, there's other parts of the plugins that are formatted like it is suggesting here, and all of the plugin code has gone through the linter.

It's just that the rules it uses to format are not always directly obvious, for example for this:

     sycl::detail::pi::assertion(
-        cuDeviceGetAttribute(&value,
-                             CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID,
+        cuDeviceGetAttribute(&value, CU_DEVICE_ATTRIBUTE_PCI_DEVICE_ID,
                              device->get()) == CUDA_SUCCESS);

If you look at other places that have it like this, they have enums with longer names that wouldn't fit on the first line in 80 columns.

Line breaking rules can get very tricky in C++, and that's why we should trust the linter even if it's not always the prettiest.

So could you update the code as the linter is suggesting? And in general I suggest using clang-format, you can even build it as part of DPC++ ninja clang-format.

Other than that the patch looks good to me 👍

[abagusetty]

@zjin-lcf @npmiller Do we need a API check for this. Also would you suggest to replace sycl::detail::pi::assertion to PI_CHECK_ERROR ?

[npmiller]

According to the CUDA docs it looks like this can only return an error if the parameter is nullptr, so I think this should be fine, see:

• https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__VERSION.html#group__CUDA__VERSION

[abagusetty]

LGTM!

[bader]

@zjin-lcf, thanks for contributing this!

[zjin-lcf]

I want to thank the reviewers for their comments.