Subject: Re: Preparing callout(9) for a HZ-less kernel
To: None <email@example.com>
From: Mindaugas R. <rmind@NetBSD.org>
Date: 10/23/2007 18:49:02
Joerg Sonnenberger <firstname.lastname@example.org> wrote:
> > > Initial benchmarks from rmind@ have shown a small regression for the
> > > MySQL benchmark,
> > IMHO not surprising.
> I expect that to be mostly related to the nanotime call and not so much
> to the use of the rb tree. This is consistent with some additional tests
> by rmind@.
Well, RB-tree for callouts is not a point. The problematic part might be the
microtime() and similar calls, needed for HZ-less kernel. Using them in some
subsystems (eg. scheduling), cost might be noticeable.