Hi Peter,

On Fri, 2021-07-09 at 14:39 -0400, Peter Xu wrote:
> On Fri, Jul 09, 2021 at 05:54:14PM +0200, Nicolas Saenz Julienne wrote:
> > After a long tracing session it's sometimes hard to find the moment
> > oslat started. So leave a message in the trace buffer just before the
> > main thread starts.
> > 
> > Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
> > ---
> >  src/oslat/oslat.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> > 
> > diff --git a/src/oslat/oslat.c b/src/oslat/oslat.c
> > index d35be57..20bf02b 100644
> > --- a/src/oslat/oslat.c
> > +++ b/src/oslat/oslat.c
> > @@ -400,6 +400,10 @@ static void *thread_main(void *arg)
> >  	while (g.n_threads_running != g.n_threads)
> >  		relax();
> >  
> > +	if (!g.preheat)
> > +		tracemark(false, "%s: Starting thread on CPU %d.\n",
> > +			  g.app_name, t->core_i);
> > +
> >  	frc(&t->frc_start);
> >  	doit(t);
> >  	frc(&t->frc_stop);
> > -- 
> > 2.31.1
> > 
> 
> Not sure how it would help.. but no objection either. :)

For example, I tend to enable all trace points on Isolated CPUs. It's
manageable, since once oslat runs you shouldn't be getting much action. But you
get 10s/100s MBs of noise generated while oslat is being initialized. This
makes finding the beginning of the test very annoying and time consuming.

> Acked-by: Peter Xu <peterx@redhat.com>

Thanks!