Steve Harris wrote:
>
> On Tue, May 15, 2001 at 02:51:24PM -0400, Paul Davis wrote:
> > in real-life, there is (period_duration-cpu_time_for_period) after the
> > completion of a "proc" cycle during which the processor is free to run
> > other threads.
>
> Actually, its not as bad as it seems, because I forgot to run the tests as
> root, so there were plenty of processes intrupting. Too many infact ;)
>
> I will rerun the graph script if Abramo changes the program so that it
> sleeps for the right amount of time.
>
> I've attached my working version of hte latest test program and the very
> dodgy perl script I use to make the graph data.
I've changed ctx to be more easily used with your script, note that I've
added some arguments and moved mode to the top. I've attached the new
version to previous message.
I'd like you plot the difference and difference/workers for workers = 4
and workers = 8. If I'm not wrong the difference/workers should never be
greater than about 2.5%.
To run faster I suggest you to not repeat output 5 times: 1 times is
enough. However run it in SCHED_RR. You may also reduce periods to 100.
--
Abramo Bagnara mailto:[EMAIL PROTECTED]
Opera Unica Phone: +39.546.656023
Via Emilia Interna, 140
48014 Castel Bolognese (RA) - Italy
ALSA project http://www.alsa-project.org
It sounds good!
