[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Audio encoders optimized for altivec (I have a G4)



On Wed, 2002-04-24 at 08:13, Michel Dänzer wrote:
> On Tue, 2002-04-23 at 21:48, David Stanaway wrote:
> > 
> > I was wondering if anyone could recommend an encoder for me?
> > 
> > I use grip as a front end, and tried oggenc from vorbis-tools but I can
> > only encode at half realtime!
> > 
> > I have only used mp3 encoders before on intel systems, so maybe this is
> > just a sign that ogg is a more complex transformation than mp3 but I was
> > hoping for at least real time perfomance.
> 
> What clock speed etc. is that? I get more than realtime with a 667 Mhz
> G4. The compression you want to achieve probably also matters, I usually
> encode for 160 kBit/s average rate and leave the rest at defaults.

Actually, it varies. I can get a little better than realtime when I am
not ripping (Using cdparanoia, I get about 3x read rates from the TiBook
DVD). I have the 500 MHz processor.

There was talk back in 2000 about vectorization of parts of the encoding
engine for oggenc generically so that optimisations could be done on the
newer i386 based chips and the altivec unit was also mentioned.
(Aparently there is a C library that can be used which is optimized to
use the altivec unit), but having a look at the source code now.  It
looks like it was not implemented.

> > It doesn't really matter to me if I creat oggs or mp3s.
> > 
> > Any recommendations?
> 
> LAME is probably the best encoder for MP3s, don't know if it supports
> altivec yet though.

It came up on some searches.  I think a port of it to MacOS used the
altivec instructions.


I might go and try and find the theory behind ogg encoding and have a
hack at the source. And see if I can do anything to make it altivec
friendly as it is a shame to have thhat good and expensive part of my
system going to waste.

On that note, I just discovered ogle-altivec which runs quite nicely :).


--
David Stanaway

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: