9 Aug
2009
9 Aug
'09
9:19 p.m.
On Sun, Aug 09, 2009 at 07:28:49AM -0400, Paul Davis wrote:
On Thu, Aug 6, 2009 at 9:45 AM, David
Baron<d_baron(a)012.net.il> wrote:
I'm nit picking ... I disagree with you here, Paul, but only in one
respect, when the system is CPU bound, that is there is no free CPU time
available.
nice(1) *does* have an effect on choosing which of several competing
user-space processes will execute next, and so will assist with
scheduling processes to meet hardware-determined deadlines (ie. incoming
audio samples from a mic).
But only if there are competing user-space processes.
If the processor is idle most of the time, which shows as a load
average less than 0.5, and is the usual situation these days, nice(1)
has no effect at all.
Putting it into human ... if the system is *only* doing your audio
recording job, then nice(1) will do nothing ... except make you feel
good.
Now, if your goal is only to feel good, and not to be right about it,
use it all you like, guys. ;-)
--
James Cameron
http://quozl.linux.org.au/
I have niced these processes negative. Did not
help either (sound breakage
during other IO or Xorg activity) or the xruns. Xmms (the old one) scarcely
breaks at all at nice 0.
David - not picking on you in any way, just want to get the message
out there even more clearly: nice(1) has absolutely nothing useful to
offer any issues connected with scheduling processes to meet
hardware-determined deadlines (ie. your audio interface). People need
to get rid of this idea. Its not just "not the best way", its
completely irrelevant.