Hello,
Currently I am trying to develop some swap backend for OSv but the
performance is not ideal so I am profiling to see where the bottleneck is.
However I have encountered the following problems
1) I first used the *Timed Tracepoints, *which was described by the Wiki.
It is useful besides I have to add tracepoints in my code manually. The
problem really is that I found the *Timed Tracepoints *sometimes lost
samples. For example, I have a function A that contains a loop of function
B. In the first run, I put *Timed Tracepoints *only at the beginning and
end of function A. In the second run, I put *Timed Tracepoints *for both A
and B but I saw the number of samples of A is significantly less and total
time significantly less. I found the total number of *Tracepoint *samples
is similar. I wonder whether there is a precision limit or a limit on the
number of samples the *Timed Tracepoints *can collect.
2) I then tried to use the *Sampler Tracepoints *but *./trace summary
--timed *tells me that 60% is spent on *sched::cpu::idle *and another 39%
on *sched::cpu::do_idle *which contradicts the result from the *Timed
Tracepoints. *At least it should show the function A in the backtrace. I
wonder whether there is any guide on how to use the *Sampler *properly
besides the Wiki.
3) Finally I tried to use perf: *perf kvm
--guestvmlinux=build/release/loader.elf record -p pid *but I got this error
"Couldn't record guest kernel [0]'s reference relocation symbol.".Though
perf is woken up to write data no sample is reccorded. Is there anyone who
has also gotten this error?
If anyone has more advice on how to do perf on OSv besides the wiki, I am
also willing to listen. Thanks very much!
Best Wishes
Pan
--
You received this message because you are subscribed to the Google Groups "OSv
Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to osv-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/osv-dev/171dc58e-f401-4b46-8403-95bd0297751en%40googlegroups.com.
