IPEX changes the return type of the API sycl_queue and sycl_device. Triton XPU backend need to align these changes.

[chengjunlu]

The IPEX changes the sycl_queue return type at commit.
https://github.com/intel-innersource/frameworks.ai.pytorch.ipex-gpu/commit/6182b5eba7b46230fe2631677dc8aba13703389e

Changes the sycl_device return type at commit.
https://github.com/intel-innersource/frameworks.ai.pytorch.ipex-gpu/commit/f65c45e75fa52209dc174a6b56e87e107c702371

Both returns the capsule of the sycl::queue and sycl::device pointer refer to the corresponding object inside IPEX internally.

The capsule names are torch.xpu.Stream.sycl_queue and torch.xpu.device.sycl_device.

The Triton XPU backend runtime need to align the new API in IPEX to support torch DL complier use cases.

[quintinwang5]

@chengjunlu we already use PyCapsule_GetPointer(cap, PyCapsule_GetName(cap)), not hardcode name.

intel-xpu-backend-for-triton/third_party/intel/backend/driver.c

Line 185 in 5a0d78f

if (!(queue = PyCapsule_GetPointer(cap, PyCapsule_GetName(cap))))

[chengjunlu]

The merged PR.
#239

[quintinwang5]

Fixed by #239