[Intel-gfx] [PATCH 1/4] drm/i915: Increase PSR Idle Frame to 2.

[Daniel Vetter]

On Thu, Sep 4, 2014 at 9:16 PM, Rodrigo Vivi <rodrigo.vivi at gmail.com> wrote:
>> Hm, I don't see a pattern at all. And that sprites seem to work best
>> also looks funky. Are the results stable when you randomize them
>> (piglit can do that for you)? Can you add a residency checks in the
>> testcase (i.e. before the update and after the update) so that we know
>> it's really psr and not something else funny going on? I assume this
>> is on bdw, any difference in results on bdw?
>
> Yes, this is really strange. I couldn't find a pattern at all as well.
> Do you have any example of piglit helping randomization?
>
> The bad things with residency check is that vlv/chv doesn't have performance
> counters.
> So this test would just work on hsw/bdw/
>
> I did this work on a HSW. Test on BDW is one of tests I should do next here.

There's no shuffling in piglit unfortunately. But piglit is good for
comparing results, so I wonder whether they are stable or change. And
hsw is even more suprising since sprite tests pass more often in your
example, but sprite also has now hw invalidate on hsw. Confusing.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch