[Intel-gfx] [PATCH 8/8] drm/i915: Add drrs_interval module parameter

Daniel Vetter daniel at ffwll.ch
Mon Dec 15 06:38:10 PST 2014


On Mon, Dec 15, 2014 at 07:56:08PM +0530, Kannan, Vandana wrote:
> 
> 
> On 15-Dec-14 7:46 PM, Daniel Vetter wrote:
> >On Mon, Dec 15, 2014 at 04:25:32PM +0530, Kannan, Vandana wrote:
> >>
> >>
> >>On 15-Dec-14 3:17 PM, Daniel Vetter wrote:
> >>>On Thu, Dec 11, 2014 at 02:22:57AM +0530, Vandana Kannan wrote:
> >>>>Adding i915 module parameter for setting drrs_interval. If this param is
> >>>>set to 0, then drrs is disabled. If changed in runtime, then the new interval
> >>>>value will be considered for scheduling the next drrs work.
> >>>>drrs_interval is set to 0 by default, i.e. DRRS is disabled by default.
> >>>
> >>>Nope, please don't hide power saving features behind module options by
> >>>default. New stuff must be enabled by default, otherwise it'll bitrot and
> >>>merging to upstream is fairly useless since we still have the rebase pain
> >>>(just spread out over more people) with none of the testing.
> >>ok.. so, shall I just make the delay (drrs_interval) fixed at 1 second
> >>or let user set this delay at runtime through the same module param
> >>(excluding the disable feature if interval is 0 part) ?
> >
> >Imo we should just set an optimal value (does vbt have any hints?).
> >
> VBT does not contain a delay value..
> Based on data collected from testing so far, 1 second seems stable..
> Maybe it can go down to 800ms or so - I can test it out..
> Anything as low as 100ms just triggered too many RR switches back and forth.

There's not need imo to go to the lowest possible value, there's always a
tradeoff. 1s sounds good (psr has even worse timeout on some panels ...).
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch


More information about the Intel-gfx mailing list