Rockbox mail archiveSubject: Re: 20050729 audio buffering
Re: 20050729 audio buffering
From: Sander Sweers <sander.sweers_at_gmail.com>
Date: Fri, 29 Jul 2005 20:38:04 +0200
On 29/07/05, Pedro Vasconcelos <pbv_at_st-andrews.ac.uk> wrote:
> On Fri, 29 Jul 2005 18:09:17 +0200
> Magnus Holmgren <lear_at_algonet.se> wrote:
> > Pedro Vasconcelos wrote:
> > >>You can also note that the pcm buffer is much smaller (especially with
> > >>no crossfade), and the so called watermark limit has not been reduced
> > >
> > >
> > > Ah, I hadn't noticed that. Any clue as to why was the buffer made smaller?
> > To leave more room for the file buffering mainly. But the purpose of
> > that was to increase battery time, not reduce it...
> Sorry, but what is the "file buffering" used for? To keep next tracks
> pre-loaded and avoid hd spinup? But that it is only worth it if you load
> a playlist and stick to it, not everyone will do that. I find that I
> will listen to a few tracks, stop and switch to something diferent.
> The extra power drain of increasing the average cpu frequency will
> affect *everyone* regardless of their listening patterns...
What if there are a couple of profiles created which users can select?
The user can then choose whichever one works best for her/him. For
example a power adaptor profile which will load the file buffer all
the way and the pcm buffer will be reloaded when it is down to 75%
because power consumption is not really an issue. Or a min power
profile where the file buffer will only fill up to maximum of 15% and
the pcm buffer has to go down to 25% before it will fill it up again.
And ofcourse a default profile with sane values.
Other option might be that both buffer levels are configurable at run
time. But for example when cross fade is enable have it set a minimum
buffer level so that it would not cause problems switching tracks
(which i thought led to the buffer level changes).
The second approach would have the added benefit that you can test
different buffer levels and observe how it affects the power
consumption. If I understand correctly now this requires a change in
Received on 2005-07-29