> > > > * maximum clocks following a vblank miss (see do_rps_boost()).
> > > > */
> > > > if (!intel_state->rps_interactive) {
> > > > - intel_rps_mark_interactive(dev_priv, true);
> > > > + intel_rps_mark_interactive(&dev_priv->gt.rps, true);
> > >
> > > I wonder if we can do &plane->vma->vm->gt->rps
> >
> > agree, looks ugly, I could fix it by extracting rps. Shall I do
> > it now in a v3 or after the patch gets merged? As you can gues
> > I'd prefer doing after the patch is merged :)
>
> Mostly thinking aloud. I plan on soak testing this first; something did
> not look quite right around pm_enable/pm_disable vs gt_resume/gt_suspend
> (i.e. did not match my current expectations of where to push the gt init
> next).
yes, it's confusing and to me they look inverted in meaining :).
Do we want to have a unique resume/suspend rather than both?
For this refactoring it was just easier to keep it this way.
Andi
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
