[Intel-gfx] [PATCH 2/7] drm/i915: Fix computation of last_adjustment for RPS autotuning

[Chris Wilson]

On Fri, Mar 06, 2015 at 06:32:27PM +0100, Daniel Vetter wrote:
> On Fri, Mar 06, 2015 at 03:06:26PM +0000, Chris Wilson wrote:
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> 
> I might have impending w/e syndrome, but I can't see what you're fixing
> any more. Can you please augment the commit message a bit?

There are a couple of issues. First is when we jump to RPe, we then set
last_adj to a large value and then promptly double it. So we overshoot
lightish workloads (often landing in the high power domain and so
trapped). The second is that clamping could equally perturbs the last_adj
value. 
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre