[Vorbis] Android App?

[Gregory Maxwell]

On Sun, Nov 1, 2009 at 3:44 PM, Tor-Einar Jarnbjo
<tor-einar at jarnbjo.name> wrote:
> Gregory Maxwell schrieb:
>>
>> I have no doubt that it's true one some, but its by no means universal.
>
> These "internet tablets" from Nokia are unfortunately the odd ones out, so
> although some mobile devices may show similar battery life, most actually
> don't. The Nokia models have a programmable DSP, which is doing both MP3 and
> Vorbis decoding with similar efficiency. Most mobile devices however have a
> dedicated, energy efficient MP3 hardware decoder, while Vorbis decoding is
> done in Software on a general purpose CPU (if Vorbis decoding is possible at
> all). So, the difference is not directly caused by the format or the format
> specific decoding complexity (I think they are quite similar for MP3 and
> Vorbis?), but that the player hardware is capable of decoding MP3 much more
> efficiently than Vorbis.

So— the original post was about android. All the android hardware is
similar to the N[89]xx, an ARM SOC of some spin or another.

The Vorbis decoder on N8xx is CPU based, it doesn't run on the DSP. No
one has bothered with a c5x port of the Vorbis decoder for it… though
it could be done now. (FWIW, there is now a Theora c64x port (c64x+ is
used in n900 and palm-pre among other things:
http://code.entropywave.com/git?p=leonora.git;a=summary)

In any case, Vorbis does require more work to decode than MP3 — it's
simply a more sophisticated format. How much that work impacts power
consumption depends on the target architecture. I expect that in the
all software cases it's more of an issue of optimization: A well tuned
vorbis implementation could likely take 1/2 or less the resources of
the reference implementation, and if you're not advertising Vorbis
support you aren't likely throwing optimization resources at it.