On 06/08/2018 11:03, Chris Wilson wrote:
Quoting Tvrtko Ursulin (2018-08-06 10:53:52)
On 06/08/2018 09:30, Chris Wilson wrote:
Latency is in the eye of the beholder. In the case where a client stops
and waits for the gpu, give that request chain a small priority boost
(not so that it overtakes higher priority clients, to preserve the
external ordering) so that ideally the wait completes earlier.
Going back to my comments made against the new clients boost patch -
perhaps if you make it that preemption is not triggered for within
internal priority delta maybe it is all acceptable. Not sure how much of
the positive effect you observed remains though.
That entirely defeats the point of being able to preempt within the same
user priority; so no more "fairness" wrt FQ_CODEL and no more cheating
here.
In this case postpone these tweaks until after we implement time-slicing? :)
Regards,
Tvrtko
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
