Quoting Alexei Podtelezhnikov (2020-04-28 19:44:13)
> On Tue, Apr 28, 2020 at 7:43 AM Chris Wilson <chris@xxxxxxxxxxxxxxxxxx> wrote:
> >
> > Give a small bump for our tolerance on comparing the expected vs
> > measured clock ticks/time from 10% to 12.5% to accommodate a bad result
> > on Sandybridge that was off by 10.3%. Hopefully, that is the worst we
> > will see.
> > - if (10 * time < 9 * ktime_to_ns(dt) ||
> > - 10 * time > 11 * ktime_to_ns(dt)) {
> > + if (10 * time < 8 * ktime_to_ns(dt) ||
> > + 8 * time > 10 * ktime_to_ns(dt)) {
>
> This is actually -25%/+20% and you could have used 5:4. If your goal
> is to cover 10.3% in either direction just barely, use 8:9.
Sigh. Wasn't thinking, but we do need a bit of spare so as to get no
false positives over several thousand runs. I was expecting to have to
bump it beyond 12.5%, but now just wait and see and then reduce it
again. Otherwise, it's back to trying to find the lost time.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
