Do not profile wrapper function of starpu_execute_on_{each,specific_workers}
When one calls starpu_execute_on_each_worker
or
starpu_execute_specific_workers
, it will create tasks which are
starpu_task_submit
-ed. When these tasks are getting executed, the
driver that will call wrapper_func
has already profiling probes before
and after to tell the start and stop of task execution. Thus, having
the same probes also in wrapper_func
is useless (and confusing).
A kind of call stack can be visualized like the following:
start_job_on_{cuda,hip,...}
probe start
wrapper func
probe start
func
probe end
probe end