Re: [Bug-gnubg] slow gui (strace)

[Jim Segrave]

On Thu 27 Nov 2003 (16:26 +0100), Achim Mueller wrote:
> * Joern Thyssen wrote on 27 Nov 2003:
> > For the mailing list readers:
> > 
> > Would output from gprof be more interesting, i.e., would that reveal
> > where the additional time is spend? I didn't follow the thread closely,
> > but as far as I remember Achim said that CPU time was identical, but wall
> > time was much longer (Achim, can you confirm?)! Will the wall time be
> > attributed to the correct routine in a gprof experiment?
> 
> How can I check this?

I ran gprof on Linux to see why an analysis of a match took a lot
longer under the GUI than under the CLI. gprof only counts time when
the process is running (it slows down processing, so it would be
meaningless to count wall clock time when profiling). I'll try to get
a look at the strace output - what I'd be looking for is times the
process is suspended in a syscall. Sleep calls and select calls with a
timeout value and no fd_sets (an easy way to get sub-one second
sleeps) would run up the real time without using CPU, so that's what
I'd be looking for.

-- 
Jim Segrave           address@hidden