Messages in this thread |  | | | Subject | Re: [patch] voluntary-preempt-2.6.8.1-P0 | | From | Lee Revell <> | | Date | Sun, 15 Aug 2004 22:45:37 -0400 |
| |
On Sun, 2004-08-15 at 22:43, Ingo Molnar wrote: > * Lee Revell <rlrevell@joe-job.com> wrote: > > > http://krustophenia.net/testresults.php?dataset=2.6.8.1-P0 > > nice. (What is the difference between the left-hand and the right-hand > graphs - why is the right-hand side one 'wider'?) >
The right hand graph is logarithmically scaled on the y axis (set logscale y in gnuplot). Some of the latencies are rare enough to not show up at all on the linear scale, like the peak in the 400s.
> > The peaks on this graph should correspond directly to the length of > > the non-preemptible critical section reported by Ingo's latency > > tracer. I think the large peak around 580-600usecs is caused by the > > extract_entropy issue (which can be hit by regular processes and > > ksoftirqd), and the large peak around 80-100 by the XFree86 unmap_vmas > > issue, as the times match and these are by far the most common > > reported in latency_trace. > > just to check this theory, could you make __check_and_rekey() an empty > function? This should still produce a working random driver, albeit at > much reduced entropy. If these latencies have a relationship to the > mlockall() issue then this change should have an effect. >
Sure, will try this next.
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/
|  |