[vorbis-dev] Vorbis packet sizes

[Sebastian Gesemann]

On Thu, 5 Jun 2003, Michael Smith wrote:

> On Friday 06 June 2003 00:31, David Etherton wrote:
> [...]
> > [...]
> > For that matter, is there a way to encode a stream in such a way that it
> > will consume less CPU time to play back (probably at the expense of some
> > compression).  Thinking about it a little bit, at least part of the decode
> > process time must linear in the output sample rate, so I'm thinking of
> > doing voiceovers or walla at 24Khz or even possibly 16Khz.  I'm hoping I
> > can do music (stereo), walla (stereo), and voiceover (mono) all at the same
> > time for less than the cost of two full 48Khz stereo streams.
> 
> Right - sample rate and number of channels are the main ones. Other than that, 
> you'd really have to tweak the encoder internals, and I doubt it'd be worth 
> it.
> 
> Lower bitrates should also take less cpu to decompress, simply because there 
> are fewer codebook symbols to look up. I don't know that this is 
> significant/noticable, though.

Yes, the more huffman code bits are there, the longer it takes to decode 
them, but the real time consuming component is IMHO the iMDCT. I'm not 
aware of the performances of the libVorbis components, but in my own 
vorbis decoder implementation the iMDCT takes about 90% of the overall 
decoding time. (tested on a 112kbps-44kHz stereo sample)

(I did NOT use the naive O(n^2) approach for the iMDCT) ;)

<p>bye,
Sebastian

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.