lkml.org 
[lkml]   [2005]   [Oct]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: Latency data - 2.6.14-rc3-rt13
From
Date
On Tue, 2005-10-11 at 15:45 -0700, Mark Knecht wrote:
> On 10/11/05, Ingo Molnar <mingo@elte.hu> wrote:
> >
> > * Mark Knecht <markknecht@gmail.com> wrote:
> >
> > > > ( softirq-timer/0-3 |#0): new 3997 us maximum-latency critical section.
> > >
> > > So the root cause of this 4mS delay is the 250Hz timer. If I change
> > > the system to use the 1Khz timer then the time in this section drops,
> > > as expected, to 1mS.
> >
> > this was a bug in the critical-section-latency measurement code of x64.
> > The timer irq is the one that leaves userspace running for the longest
> > time, between two kernel calls.
> >
> > I have fixed these bugs in -rc4-rt1, could you try it? It should report
> > much lower latencies, regardless of PM settings.
> >
> > Ingo
> >
>
> Ingo,
> This test now reports much more intersting data:
>
> ( dmesg-8010 |#0): new 13 us maximum-latency critical section.
> => started at timestamp 117628604: <do_IRQ+0x29/0x50>
> => ended at timestamp 117628618: <do_IRQ+0x39/0x50>

This is the expected, correct behavior - very small maximum latency
critical sections. Do you get anything longer (say 300 usecs or more)
if you leave it running?

So far the latency tracer on my much slower system has only gone up to
123 usecs. So the bug seems to be fixed at least on i386.

Lee

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-10-12 02:05    [W:0.024 / U:1.548 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site