[FFmpeg-devel] [PATCH] unscaled float 2 int conversion

[Michael Niedermayer]

On Fri, May 16, 2008 at 01:54:46PM +0200, Benjamin Larsson wrote:
> Michael Niedermayer wrote:
> > [...]
> >>>   
> >>>       
> >> What I'm saying is that I think it is irrelevant that the bias trick is 
> >> faster on a P3 then lrint because on the P3 we have sse available. Thus 
> >> on P3 we can beat the bias trick. What would be interesting is P2 and 
> >> lower. But as I don't have access to any machine like that I can't make 
> >> any relevant benchmark.
> >>     
> >
> > P3 = P2 + some SIMD
> >   
> Ok, for the sake of argument I will accept that reasoning. On P2's the 
> bias scaling thing will be faster when we have to run float code. My 
> argument is then that for those system we shouldn't even run float code, 
> we should run fixed point code if we really wanted to have the fastest 
> codecs thus we could remove the bias trick from the codecs for the 
> benefit of the code clarity. So far we don't have any fixed point code 
> that can replace the float code and I don't expect you to agree to my 
> arguments 

You forgot to attach the benchmark results of float vs. fixed decoders
on a P2/P3.


> so how should we go forward from this when we work on 
> implementing a new audio api. The codecs should output samples in their 
> native format, that is what I think most of us agree on. But what is the 
> native format for a codec outputting samples in float when running in 
> simd mode and the same when running in non simd mode ?

SAMPLE_FMT_FLT
and
SAMPLE_FMT_FLT_BIAS_385

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

While the State exists there can be no freedom; when there is freedom there
will be no State. -- Vladimir Lenin
-------------- 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/20080516/ee03095b/attachment.pgp>