Re: Measuring semaphore contention times with markers

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

> [...]
> > OK, understood.  Has anyone benchmarked markers in general?  

I haven't seen Mathieu produce any numbers yet.

> > Any suggestions on how to run a benchmark that will be acceptable
> > to the kernel community?

I don't know.  There are a variety linked at
http://lbs.sourceforge.net/, and I suspect anything material would be
informative.  If in doubt, I'd find a kernel subsystem that is
particularly semaphore-heavy, then find a benchmark for that
subsystem.


> I understand the idea of bench marking these markers and it is a
> good thing to do but the goal i had in mind is to show ASAP how one
> would use markers along with probes [...]  We can do the bench
> marking later.

Sure, no one suggesting delaying the reposting of this sample code.
However, the subsequent performance question is entirely predictable,
and one might as well prepare for it.

- FChE