Re: [RFC] [Patch 7/8] statistics infrastructure - exploitation prerequisite

[Martin Peschke <mp3@de.ibm.com>]

Andrew Morton wrote:
> Martin Peschke <mp3@de.ibm.com> wrote:
>> need sched_clock for latency statistics
> 
> sched_clock() probably isn't suitable for this application.  It's a
> scheduler thing and has a number of accuracy problems.
> 
> But I thought we discussed this last time around?  Maybe not.

I have been too obsessed with the other issues to remember.

 > Maybe you've considered sched_clock()'s drawbacks and you've decided
 > they're all acceptable.

Admittedly, I didn't, but merely believed the comment above the function.

http://marc.theaimsgroup.com/?l=linux-kernel&m=114657675408686&w=2
sheds some light.

I would be happy to exploit an API that may result from that discussion.
I would plead for exporting such an API to modules. I don't see how
to implement statistics for latencies, otherwise.

Any other hints on how to replace my sched_clock() calls are welcome.
(I want to measure elapsed times in units that are understandable to
users without hardware manuals and calculator, such as milliseconds.)

Martin