Re: [Lmuse-developer] [PATCH] muse: find the best available ALSA timer

[. T. <ter...@ro...>]

Ah, yes I forgot. My kernel is mm. And it
 does have the 1024 Hz. 
Thanks for taking time to explain. Great patch !
Tim.


--- On Tue, 3/10/09, Pedro Lopez-Cabanillas <ped...@gm...> wrote:
From: Pedro Lopez-Cabanillas <ped...@gm...>
Subject: Re: [Lmuse-developer] [PATCH] muse: find the best available ALSA timer
To: "Tim" <ter...@ro...>
Cc: lmu...@li...
Received: Tuesday, March 10, 2009, 5:21 PM

Tim wrote:
> One thing puzzles me, though:
> If the ALSA system timer only runs at 250 Hz, why then,
>  (before your patch), would muse say:
>  "AlsaTimer::setTimerTicks(): requested freq 8192 Hz too high for
timer
>   (max is 1024) freq stays at 1024 Hz"
> Was it really at 1024 Hz, or was muse wrong?
> If it was really at 250 Hz wouldn't muse have been really inaccurate?
> I never noticed any accuracy problems in that respect...
> What do you think was happening there?

It is hard to say, without looking to the system configuration where you got 
the messages. I can only propose a theory.

The ALSA system timer has a fixed resolution, determined by the kernel timer 
(wheels) frequency. Long time ago, the frequency was 100 Hz. It was changed 
in Linux 2.4 to a configurable setting of 1000 Hz by default. The default 
value changed again in 2.6 to 250 Hz (to be more friendly with laptop 
batteries or something like that.) But this is still a configurable value. If 
you can compile your own kernel for audio/MIDI usage you can change the 
default value to 1000 Hz or whatever you like. Maybe you customized the  
kernel in this way?

If that was the case, you probably applied also low-latency patches and other 
goodies to the kernel sources in addition to changing the timer frequency. In 
this case, you already could use Muse with great accuracy without needing the 
timer patch. It will be more useful to other less experienced people, using 
the stock kernels provided by standard distros.

Regards,
Pedro