From: Pedro Vasconcelos <pbv_at_st-andrews.ac.uk>
Date: Mon, 6 Jun 2005 23:50:13 +0100

On Mon, 06 Jun 2005 23:16:10 +0100
BlueChip <cs_bluechip_at_webtribe.net> wrote:

>
> >Second: I'm doing some more work in the Tremor optimzation and now i've
> >got it to go from 130% -> 150% real time :-). Nothing fancy, just a
> >better allocation of iram. When I'm happy that is runs well and doesn't
> >break anything I'll commit it to the cvs.
> >

>
> Out of curiosity, what % are we realistically looking at?
>
> I understand that it will be >100, and the faster it is, the more stuff can
> be done at the same time (display updates, etc) ...but what is a realistic
> figure beyond which optimisations are merely a demonstration of coding skill?
>

Sorry, I should have been more precise: 150% is the decoding speed of
the vorbis2wav codec tester for one particular test track that I have
been using (Tracy Chapman's "Talking about a revolution" encoded at
quality 6).

For comparison, the MP3 decoder can run at >300% real time in similar
conditions.

The faster the decoding the better because it will consume less battery
(i.e. by allowing us to use the cpu at a lower clock), and/or allow for
other things to be running in the background, e.g. plugins etc.

Note that Vorbis decoding will generaly be faster for lower bitrate
files and slower for higher bitrate ones. I chose a test file at medium
quality value because that is what I use, but for the insane quality 10
(~500kbps) then decoding will be slower, possibly still not real time --
haven't tested that with this new optimization.

Pedro
_______________________________________________
http://cool.haxx.se/mailman/listinfo/rockbox
Received on 2005-06-07