Kévin Petit

Currently exposing a lot of the implementation.

- [ ] Successful translation of all modules that come with OpenCL-CTS - [ ] Successful translation of OpenCL-CTS sources dumped and then compiled to SPIR-V - [ ] End-to-end...

- Provide support for cl_khr_il_program and the core support.

Rework test_allocations to remove hardcoded assumptions

1

The heuristics in these tests have been changed twice recently in https://github.com/KhronosGroup/OpenCL-CTS/pull/1852 and then https://github.com/KhronosGroup/OpenCL-CTS/pull/1932. These tests need to be reworked to better adapt to device constraints (see comments in...

Strengthen testing of `CL_DEVICE_TYPE` query and use of `CL_DEVICE_TYPE*` enums in device enumeration or context creation

1

- [ ] Add or modify a test that checks that **clGetDeviceInfo** called with `CL_DEVICE_TYPE` never returns `CL_DEVICE_TYPE_DEFAULT` or `CL_DEVICE_TYPE_ALL` - [ ] Add a test that (or check that...

A few missing cl_khr_semaphore tests

2

- [ ] `CL_DEVICE_HANDLE_LIST_KHR` as a query using `clGetSemaphoreInfoKHR` - [ ] `CL_DEVICE_HANDLE_LIST_KHR` as a property passed to `clCreateSemaphoreWithPropertiesKHR` - [ ] Negative tests for the API error codes of...

cl_khr_command_buffer::info_state does not check the state after the command buffer has completed execution

3

Once a command buffer completes execution (assuming no other pending instances), its state should be transitioned back to `CL_COMMAND_BUFFER_STATE_EXECUTABLE_KHR`. The info_state test should check that.

Fix support for CL_UNORM_INT_101010_2 in kernel read write tests

3

See https://github.com/KhronosGroup/OpenCL-CTS/pull/1129#discussion_r647109443