Profiling using cl events seems to be broken.
I'm creating a cl_event profile event with profile_event
clEnqueueNDRangeKernel
Then calling clWaitForEvents:
clWaitForEvents(1, &profile_event);
clWaitForEvents however returns CL_EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST.
I am making sure to construct the command queue with the CL_QUEUE_PROFILING_ENABLE property.
As noted, this happens on my RX480. It works fine on a CPU device, the code also works fine on Nvidia devices. It also used to work on (earlier) AMD GPU devices.
Thanks,
Nick C.