Date: Thu, 23 May 2013 08:16:00 +0200 From: Luigi Rizzo <rizzo@iet.unipi.it> To: Adrian Chadd <adrian@freebsd.org> Cc: svn-src-head@freebsd.org, Luigi Rizzo <luigi@freebsd.org>, src-committers@freebsd.org, svn-src-all@freebsd.org Subject: Re: svn commit: r250911 - head/sys/kern Message-ID: <20130523061600.GB6430@onelab2.iet.unipi.it> In-Reply-To: <CAJ-Vmons5-4-nqa-yG0aLOqqzRgMuhpgY4PYD1Xgk8J4su5fog@mail.gmail.com> References: <201305221632.r4MGWIjK069433@svn.freebsd.org> <CAJ-Vmons5-4-nqa-yG0aLOqqzRgMuhpgY4PYD1Xgk8J4su5fog@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, May 22, 2013 at 04:55:51PM -0700, Adrian Chadd wrote: > Odd, this reads like one of those things that's better off being > time-capped, rather than packet-count capped. this is in fact how it works: the control loop in the polling code dynamically adjusts the actual number of packets so that the fraction of CPU cycles spent in the polling routine does not exceed the threshold (100 - kern.polling.user_frac). Working in batches avoids reading the clock and calling the poll method at every packet, which would be too expensive. cheers luigi
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20130523061600.GB6430>