[FFmpeg-devel] new open source h.264 source

Michael Niedermayer michaelni
Thu Feb 21 04:28:14 CET 2008


On Wed, Feb 20, 2008 at 09:07:16PM -0500, Alexander Strange wrote:
> On Wed, Feb 20, 2008 at 5:15 PM, mark cox
> <melbournemark+ffmpeg at gmail.com> wrote:
> > On Thu, Feb 21, 2008 at 9:04 AM, mark cox
> >  <melbournemark+ffmpeg at gmail.com> wrote:
> >  > From phoronix web site http://www.phoronix.com/scan.php?page=news_item&px=NjM0MQ
> >  >
> >  >  AMD Opens Up Their Performance Library
> >  >  Posted by Michael Larabel on February 20, 2008
> >  >
> >
> >  ~/svn$ svn co https://framewave.svn.sourceforge.net/svnroot/framewave/trunk
> >  framewave
> >
> >  ~/svn/framewave/trunk/Framewave/domain/fwVideo/src$ ls -1
> >  AANIDCT.cpp
> 
> This has an SSE2 IDCT (which looks like it was just transcribed from
> x86-64 asm, so no macros or useful variable names). If it works well
> with half the xmm registers, it might be nice.
> 
> I don't think AAN is the best IDCT algorithm, though - maybe it would
> be better to port ppc/idct_altivec.c?

btw, speaking about dcts ...
there is something strange
trunk/Framewave/domain/fwJPEG/src/JPEGDCT.cpp:
#define W1 F 0.70710678118654752438 // cos(pi*4/16)
#define W2 F 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
#define W5 F 0.38268343236508977170 // cos(pi*6/16)
#define W4 F 1.30656296487637652774 // cos(pi*2/16)sqrt(2)

libavcodec/faandct.c
#define A1 0.70710678118654752438 // cos(pi*4/16)
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
#define A5 0.38268343236508977170 // cos(pi*6/16)
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)

Also googling for these constants shows only libavcodec as hit.
And the code upon which my faandct was based IIRC (jfdctfst.c)
contains only less accurate constants.

Also the actual code looks strangely similar to libavcodec/faandct.c

[...]
-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

If a bugfix only changes things apparently unrelated to the bug with no
further explanation, that is a good sign that the bugfix is wrong.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20080221/77254fd8/attachment.pgp>



More information about the ffmpeg-devel mailing list