[Intel-gfx] [RESEND] drm/i915: Interactive RPS mode

[Ville Syrjälä]

On Fri, Jul 20, 2018 at 02:38:32PM +0100, Chris Wilson wrote:
> Quoting Ville Syrjälä (2018-07-20 14:32:40)
> > On Fri, Jul 20, 2018 at 02:14:11PM +0100, Chris Wilson wrote:
> > > Quoting Ville Syrjälä (2018-07-20 14:07:31)
> > > > On Fri, Jul 20, 2018 at 02:02:34PM +0100, Chris Wilson wrote:
> > > > Doing this kind of global thing from the plane hooks seems a bit
> > > > strange. How about just doing this directly from commit_tail()
> > > > etc.?
> > > 
> > > We want it upfront in prepare (so that it's set before any wait) or
> > > somewhere around there (atomic_state setup?). cleanup was chosen for the
> > > symmetry with prepare.
> > 
> > Looks like we have intel_atomic_prepare_commit() which I guess would be
> > a decent spot then. And introduce intel_atomic_cleanup_commit() to do
> > the reverse?
> 
> The only other point is I started from prepare_plane for being next to
> both the reprioritisation and the add_rps_boost_after_vblank. So that's
> quite nice.

Ok, I guess that's quite reasonable.

>  
> > Another question is what happens where there are parallel flips
> > happening? One could undo the boost from the other AFAICS. But maybe
> > we don't care enough to protect against that?
> 
> It's a counter, so the "interactive" mode remains high until all
> concurrent flips are completed.

Ah. I guess the bool in the atomic state threw me off. I suppose that
one is just an optimization to avoid calling the function more than
once?

-- 
Ville Syrjälä
Intel