Date: Fri, 28 Oct 2005 19:50:57 +0200 From: "Poul-Henning Kamp" <phk@phk.freebsd.dk> To: Robert Watson <rwatson@FreeBSD.org> Cc: David Xu <davidxu@freebsd.org>, current@freebsd.org Subject: Re: Timers and timing, was: MySQL Performance 6.0rc1 Message-ID: <33946.1130521857@critter.freebsd.dk> In-Reply-To: Your message of "Fri, 28 Oct 2005 18:33:19 BST." <20051028183034.U3405@fledge.watson.org>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <20051028183034.U3405@fledge.watson.org>, Robert Watson writes:
>If we remove the call to nanotime() in the context switch, we may want to
>add a callout that calls nanotime() once each tick? Or maybe
>automatically in the callout handler, so that any code running in a
>callout can use getnanotime() without having to worry about accuracy
>(much).
The timecounters already update the once every N hardclock ticks (search
for "tc_ticktock" and this happens before callouts are called.
The reason for the N divisor is to not needlessly burn CPU cycles on
systems with very high HZ.
For Hz up to and in cluding 1000, N=1 so every tick updates the timestamps
which get{bin,nano,micro}time() copies.
For Hz higher than 1000, N is set to attempt to update the timestamps
once per millisecond:
tc_tick = (hz + 500) / 1000;
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?33946.1130521857>