Re: [igt-dev] [PATCH i-g-t 2/5] tests/kms_setmode: Request the initial vbl count with RELATIVE instead of ABSOLUTE

From: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
To: Chris Wilson <chris@chris-wilson.co.uk>
Cc: igt-dev@lists.freedesktop.org, Daniel Vetter <daniel@ffwll.ch>
Subject: Re: [igt-dev] [PATCH i-g-t 2/5] tests/kms_setmode: Request the initial vbl count with RELATIVE instead of ABSOLUTE
Date: Fri, 8 Mar 2019 18:45:07 +0200	[thread overview]
Message-ID: <20190308164507.GU3888@intel.com> (raw)
In-Reply-To: <155199529401.27405.5855883981529801961@skylake-alporthouse-com>

On Thu, Mar 07, 2019 at 09:48:14PM +0000, Chris Wilson wrote:
> Quoting Ville Syrjala (2019-03-07 21:39:23)
> > From: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > 
> > Asking for the initial vblank count by specifying and absolute vblank count of 0
> > doesn't make much sense. Switch to a relative query instead.
> > 
> > It doesn't look like we care about lining up on the first vblank boundary so
> > we can just drop the NEXTONMISS flag. Only the relative timestamps of the
> > events will matter.
> > 
> > v2: Drop the NEXTONMISS (Daniel)
> 
> No, this isn't about getting the current vblank, it's about aligning the
> test to the start of the next vblank period, so that is much less of a
> chance of that vblank happening during the setup. For which NEXTONMISS
> is essential.

That was my original assesment, but then I decided it wasn't needed
after all. Can't recall how I reached to that conclusion. I'll put
it back.

Hmm. If the test did '(now - last) / num_frames' it wouldn't matter
so much if missed a vblank or two. But I suppose that might hide
some of the jitter.

> You could look at the timestamp and say if within 1ms of
> the next vblank, defer? Wasting 8ms is a nuisance but a pittance
> compared to the ~2s runtime.
> -Chris

-- 
Ville Syrjälä
Intel
_______________________________________________
igt-dev mailing list
igt-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/igt-dev