[FFmpeg-devel] [PATCH] Is gray8 a paletted format?

Stefano Sabatini stefano.sabatini-lala
Thu Feb 25 00:34:57 CET 2010


On date Wednesday 2010-02-24 02:29:28 +0100, Michael Niedermayer encoded:
> On Wed, Feb 24, 2010 at 02:00:38AM +0100, Stefano Sabatini wrote:
> > On date Tuesday 2010-02-23 01:30:55 +0100, Michael Niedermayer encoded:
> > > On Tue, Feb 23, 2010 at 01:05:48AM +0100, Stefano Sabatini wrote:
> > > > On date Monday 2010-02-22 21:19:24 +0100, Michael Niedermayer encoded:
> > > > > On Tue, Feb 16, 2010 at 01:18:15AM +0100, Stefano Sabatini wrote:
> > > > > > On date Monday 2010-02-01 20:44:36 +0100, Stefano Sabatini encoded:
> > > > > [...]
> > > > > > > Conversion in some cases is ~10% slower.
> > > > > [...]
> > > > > > Ping.
> > > > > 
> > > > > 10% slower -> rejected
> > > > 
> > > > I elaborated more the results, so that you can have a more precise
> > > > view.
> > > > 
> > > > If this is not acceptable then I believe we should also skip to
> > > > allocate place for palette in the pictures for the gray format.
> > > 
> > > until we know why there is a large difference in benchmark scores, no change
> > > to this code will happen. At least not intentionally
> > 
> > I repeated the tests many times, refinining the test procedure
> > (e.g. I'm creating the result files with:
> > sh swscale-test-gray8.sh >/dev/null 2>tests/swscale.test.gray-pal
> > in order to avoid fluctuations due to console printing).
> > 
> > The new result shows no significant difference between gray8 = PAL and
> > gray8 = NOPAL, outstanding values (>= 0.02/0.03) changes at each
> > test.
> > 
> > I'm attaching the result files and the script used for generating it.
> 
> good, so as you now refined your benchmarking procedure, you can start
> working on h264 optimizations :)

Uh too much stuff on my todo and too few time.

And BTW is find to apply the patch?

Regards.
-- 
FFmpeg = Fantastic and Formidable Moronic Power Evanescent Gnome



More information about the ffmpeg-devel mailing list