what tools do you use for wall clock profiling your C++ apps?

Question

I have to profile my multithreaded C++ app and find its bottlenecks. The problem is: I need to see wall clock profile. I have used oprofile and perf. No one can provide me such information.

I have used perf record -g -e sched:sched_stat_sleep <cmd> but perf record falls with SIGFPE exception. This makes me angry.

Valgrind doesn't suits me because I use fanotify_mark syscall which is not implemented in this tool.

I'm not sure google's perftools can do wall clock profiling - I haven't see any info in their documentation.

Can anyone suggest? Thank you.

Answer 1

Try PAPI http://web.eecs.utk.edu/~terpstra/using_papi/, it is an open source profiler. I normally use this profiler to measure the cache performance (e.g, misses and accesses) in some algorithms. Maybe it can be useful also for what you want.

If you have MAC, have a look on the profiler "instruments"