[SYCL][Doc] Add test plan for sycl_ext_oneapi_device_image_backend_content extension #16698
Conversation
|
|
||
| The unit tests should be launched on every supported device configuration we have. | ||
|
|
||
| The end-to-end tests, which consist of checking the interoperability with Level Zero |
There was a problem hiding this comment.
Is there going to be test coverage for other backends, like cuda and hip?
There was a problem hiding this comment.
The unit tests will be run on all configurations but the end to end tests will only run on Level Zero and OpenCL because there does not seem to be a well-defined way to retrieve a kernel from a device image executable for other backends.
There was a problem hiding this comment.
According to the plan unit tests will only do some basic API checks but not verify that obtained result is correct.
Is it possible to use module management APIs https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__MODULE.html (like cuModuleLoad*) to check the result?
There was a problem hiding this comment.
Yes, that can be used and I should definitely add that to the test plan. For example, the test should use cuModuleLoadData to load the binary data into a module and verify that CUDA_SUCCESS is returned. We cannot go any further unfortunately as there is no portable way to retrieve our kernel from the module as is the case, for example, with Level Zero(see here). For HIP, the format of the data is not even defined so I don't think there is much to verify there.
This PR adds the test plan for sycl_ext_oneapi_device_image_backend_content extension.