Re: [Intel-gfx] [PATCH 07/16] drm/i915/gt: Check for a completed last request once

[Chris Wilson <chris@chris-wilson.co.uk>]

Quoting Tvrtko Ursulin (2020-11-24 17:19:23)
> 
> On 24/11/2020 11:42, Chris Wilson wrote:
> > Pull the repeated check for the last active request being completed to a
> > single spot, when deciding whether or not execlist preemption is
> > required.
> > 
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > ---
> >   drivers/gpu/drm/i915/gt/intel_lrc.c | 15 ++++-----------
> >   1 file changed, 4 insertions(+), 11 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > index cf11cbac241b..43703efb36d1 100644
> > --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> > +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> > @@ -2141,12 +2141,9 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> >        */
> >   
> >       if ((last = *active)) {
> > -             if (need_preempt(engine, last, rb)) {
> > -                     if (i915_request_completed(last)) {
> > -                             tasklet_hi_schedule(&execlists->tasklet);
> > -                             return;
> > -                     }
> > -
> > +             if (i915_request_completed(last)) {
> > +                     goto check_secondary;
> > +             } else if (need_preempt(engine, last, rb)) {
> >                       ENGINE_TRACE(engine,
> >                                    "preempting last=%llx:%lld, prio=%d, hint=%d\n",
> >                                    last->fence.context,
> > @@ -2174,11 +2171,6 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> >                       last = NULL;
> >               } else if (need_timeslice(engine, last, rb) &&
> >                          timeslice_expired(execlists, last)) {
> > -                     if (i915_request_completed(last)) {
> > -                             tasklet_hi_schedule(&execlists->tasklet);
> > -                             return;
> > -                     }
> > -
> >                       ENGINE_TRACE(engine,
> >                                    "expired last=%llx:%lld, prio=%d, hint=%d, yield?=%s\n",
> >                                    last->fence.context,
> > @@ -2214,6 +2206,7 @@ static void execlists_dequeue(struct intel_engine_cs *engine)
> >                        * we hopefully coalesce several updates into a single
> >                        * submission.
> >                        */
> > +check_secondary:
> >                       if (!list_is_last(&last->sched.link,
> >                                         &engine->active.requests)) {
> 
> Is there a tasklet_hi_schedule in here? I see set_timeslice in my checkout.

That tasklet_hi_schedule() was a mistake. It just devolves into a
busy-spinner as we wait for HW to send the next event, which turns out
not to be as instantaneous as hoped.

I recall leaving the imprint of my palm on my face when I figured out
what was causing the spike in the profile.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx