timer problems

[Frank Carmickle]

On Wed, 7 Nov 2001, Shaun Oliver wrote:

> ok I take that back it was barry that posted about the kernel timing
> issues.

I was the one who sent the message of correction.  And I will restate it
in different terms.  Alsa does not live in user space.  You insmod the
modules in to your kernel right?  If you don't have a second cpu and the
one you have is busy your shit out of luck any way you slice it.  It's all
a compromise.  You need to hear chars from your synth and at 9600 bod you
get about one char per ms.  So how long do you want to have silence in the
speech output while your waiting for the next load of chars?  Probably
none.  This is why the buffers are set the way they are.  So in order to
not have breakups in your audio you have to send more bits to the audio
hardware at the times when speakup isn't hogging the cpu to get the chars
out the port.  It's not very difficult to find the right balance.  But do
a bunch more disk io and you'll be sitting there for a while.  My machine
is very much not interactive when updatedb occurs.

The reason why alsa is better then the open sound system is that alsa uses
much less cpu load.  Thus getting more work done during those times when
speakup isn't hogging the cpu.  

-- 
     Frank Carmickle
phone:     412 761-9568
email:     frankiec at dryrose.com