I’m a bit new to profiling, I’ve used gprof in the past but this is my
first time using oprofile. I’m running oprofile on the inband code to
try and better understand the m-block performance problem we are seeing.
Since Eric and I’s first short attempt at profiling a while ago showed
significant time in the kernel, I’ve enabled full kernel profiling.
I haven’t specified any specific oprofile options, just sort of started
the daemon, dumped before running the program, ran it for ~1 minute and
saved a session, then examined the single application.
I’ve hosted the results to spare the list some traffic:
So it looks like there is a lot of time spent scheduling the threads in
the kernel. We’re not concerned about optimizing the application
I can certainly help run more results to make things clearer.