[Intel-gfx] [PATCH i-g-t 6/7] gem_wsim: Busy stats balancers
Chris Wilson
chris at chris-wilson.co.uk
Tue Sep 26 11:48:08 UTC 2017
Quoting Tvrtko Ursulin (2017-09-26 12:27:33)
>
> On 25/09/2017 17:44, Chris Wilson wrote:
> > Quoting Tvrtko Ursulin (2017-09-25 16:15:01)
> >> From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
> >>
> >> Add busy and busy-avg balancers which make balancing
> >> decisions by looking at engine busyness via the i915 PMU.
> >
> > "And thus are able to make decisions on the actual instantaneous load of
> > the system, and not use metrics that lag behind by a batch or two. In
> > doing so, each client should be able to greedily maximise their own
> > usage of the system, leading to improved load balancing even in the face
> > of other uncooperative clients. On the other hand, we are only using the
> > instantaneous load without coupling in the predictive factor for
> > dispatch and execution length."
>
> Ok, thanks for the text.
>
> > Hmm, do you not want to sum busy + queued? Or at least compare! :)
>
> How to add apples and oranges? :) Queued * busy [0.0 - 1.0] ?
Just switch bases to bananas.
Hmm, I keep going back to that we do was a loadavg type of stat.
Something that tells us the current (and historical) run length. Kind of
ties into what we want to measure through QUEUED.
-Chris
More information about the Intel-gfx
mailing list