[FFmpeg-devel] [PATCH] Add float support to wavpack decoder

[Michael Niedermayer]

On Tue, May 05, 2009 at 09:17:45PM +0200, Laurent Aimar wrote:
> On Tue, May 05, 2009, Reimar D?ffinger wrote:
> > On Tue, May 05, 2009 at 12:02:58AM +0200, Laurent Aimar wrote:
> > > I submit them to have your comments about them as I still dislike one part
> > > of it: I suppose IEEE 754 float storage to convert from sign, mantissa and
> > > exponent to a float value. I am open to any ideas to avoid that unless it
> > > is acceptable.
> > 
> > Of course it is possible to avoid, you can e.g. use ldexp. But it is
> > slow, so e.g. aac does not use it.
> > Or you can use av_int2flt which uses it but is even slower (though
> > possibly could be optimized to just use the union{} stuff where
> > possible, making it inline and then also using it for AAC).
>  The thing is, I need to also be able to encode non "regular" float values
> like zero+/-, infinity+/-, undernormalized, ... as wavpack allows lossless
> encoding of all float values. This makes av_int2flt/ldexp not usable.

wavpack stores audio, infinite amplitude does not exist. And +0 = -0 in
pretty much any algebra that is relevant, or to look at it from a different
point, a output amplitude of +0 sounds like -0


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

If you think the mosad wants you dead since a long time then you are either
wrong or dead since a long time.
-------------- 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/20090505/ce71d531/attachment.pgp>