How would you profile the processor utilization of a process on an embedded system?

Answered that it could be done using procfs on an embedded linux system with kernel, but the interviewer said it was too high level and that I didn't have an OS which was odd since he said process. I then said a bare-metal program could use systick (or architecture specific implementation) or hardware timers to track WCET timing, still seemed to be an incorrect answer. I thought he might want a software approach and mentioned a few static profiling tools and schedulability calculations, but he didn't want that either. I eventually said you could mask interrupts, assert GPIO and track clock pulses but he didn't seem happy. I've thought this one over a lot but I'm not really sure what he was looking for...