Benchmarking RT HAL component

Hello,
I am in need to be pointed to the right direction about real-time component benchmarking and general time measurement.

I am sure it's a quite common concern, but by simple search, I didn't find much. What I want is such: I am writing a real-time HAL component (loaded as loadrt) and would like to measure how long each instance of the main function call took. (The thread will run on the standard 1kHz.)

I am targeting the RT-PREEMPT patch with mainline kernel version.

Any idea?
Thanks.

(The only thing I found so far is ftrace.)

You can also use HALScope or hal-histogram to get a better view of the
statistics