--Con Kolivas <kernel@xxxxxxxxxxx> wrote (on Tuesday, July 13, 2004 00:27:27 +1000):
Martin J. Bligh wrote:
Because of the recent discussion about latency in the kernel I asked William Lee Irwin III to help create some instrumentation to determine where in the kernel there were still sustained periods of non-preemptible code. He hacked together this simple patch which times periods according to the preempt count. Hopefully we can use this patch in the advice of Linus to avoid the "mental masturbation" at guessing where latency is and track down real problem areas.
Is this much different from Rick's schedstat's work, which was itself based
on some earlier patches by Bill? I'd hate to end up with two sets of patches,
and schedstats seemed pretty comprehensive to me. He's on vacation, but his
stuff is here, if you want to take a look:
http://eaglet.rain.com/rick/linux/schedstats/
No I remember his work and this is tackling it via a different area if I recall correctly. He was looking at scheduler latencies as opposed to non-preemptible kernel code.
Fair enough ... is it worth adding it to the same harness though? He had lots
of nice analysis tools set up to do comparisons and graphing, etc.
Attachment:
signature.asc
Description: OpenPGP digital signature