Date: Sun, 26 Jul 2015 14:27:25 -0600 From: Scott Long <scott4long@yahoo.com> To: Andriy Gapon <avg@FreeBSD.org> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r284297 - in head: cddl/contrib/opensolaris/cmd/lockstat sys/kern sys/sys Message-ID: <FD6800E2-C392-4071-9CC5-D00876600F63@yahoo.com> In-Reply-To: <201506121001.t5CA1PTm089721@svn.freebsd.org> References: <201506121001.t5CA1PTm089721@svn.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Jun 12, 2015, at 4:01 AM, Andriy Gapon <avg@FreeBSD.org> wrote: > > Author: avg > Date: Fri Jun 12 10:01:24 2015 > New Revision: 284297 > URL: https://svnweb.freebsd.org/changeset/base/284297 > > Log: > several lockstat improvements > > 0. For spin events report time spent spinning, not a loop count. > While loop count is much easier and cheaper to obtain it is hard > to reason about the reported numbers, espcially for adaptive locks > where both spinning and sleeping can happen. > So, it's better to compare apples and apples. > This causes spinning to be exceptionally more expensive just by having KDTRACE_HOOKS enabled, whether or not Dtrace is actually in use. It makes it undesirable to deploy with Dtrace by default since it impacts performance. Is there a way to make the expensive collection optional, or only enable when dtrace is using the lockstat module? Also have you seen the other recent performance complaint related to this commit? Thanks, Scott
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?FD6800E2-C392-4071-9CC5-D00876600F63>