[FFmpeg-devel] [PATCH] Use LCG to generate random data for dither instead of CPU-intesive LFG.

Reimar Döffinger Reimar.Doeffinger at gmx.de
Mon Apr 25 21:27:54 CEST 2011


On Mon, Apr 25, 2011 at 09:13:17PM +0200, Michael Niedermayer wrote:
> On Mon, Apr 25, 2011 at 08:42:21PM +0200, Reimar Döffinger wrote:
> > I am not able to hear any difference.
> > Differences in FATE tests:
> > AC3:
> > stddev:    4.91 PSNR: 82.50 MAXDIFF:   55 bytes:   804864/   804864
> > stddev:    0.28 PSNR:107.16 MAXDIFF:    2 bytes:  1032192/  1032192
> > 
> > EAC3:
> > stddev:   20.26 PSNR: 70.20 MAXDIFF: 1054 bytes:   884736/   884736
> > stddev:   13.27 PSNR: 73.87 MAXDIFF:  215 bytes:  2998272/  2998272
> > stddev:    5.93 PSNR: 80.87 MAXDIFF:  100 bytes:   804864/   804864
> > (no difference in fourth test)
> > ---
> >  libavcodec/ac3dec.c  |    9 +++++----
> >  libavcodec/ac3dec.h  |    2 +-
> >  libavcodec/eac3dec.c |    6 ++++--
> >  3 files changed, 10 insertions(+), 7 deletions(-)
> 
> ok, i dont understand why an LFG was used to begin with
> i also assume you didnt pick the numbers randomly but used standard
> constants

Not sure about the initial value, but the others are from lfg.h


More information about the ffmpeg-devel mailing list