Date: Fri, 14 Oct 2005 14:28:08 -0700 (PDT) From: Danial Thom <danial_thom@yahoo.com> To: Kurt Jaeger <lists@complx.LF.net> Cc: freebsd-isp@freebsd.org, Drew Linsalata <drew@gothambus.com> Subject: Re: Multiport NICs - VLAN and Polling Support? Message-ID: <20051014212808.1778.qmail@web33311.mail.mud.yahoo.com> In-Reply-To: <20051014181239.GC62233@complx.LF.net>
next in thread | previous in thread | raw e-mail | index | archive | help
--- Kurt Jaeger <lists@complx.LF.net> wrote: > Hi! > > > polling is almost never a performance > advantage > > for ethernet, as virtually all modern > controllers > > have some sort of interrupt moderation > built-in. > > We had the case on fbsd 5.4p7 with SMP where > bge drivers produced > much slower throughput if we did not use > polling. We measured using > ttcp and the difference was from 2-3Mbyte/sec > to 11 Mbyte/sec. > > Yes, it used more CPU, but for that increase in > throughput, it > was worth the expense. > > So it looks a bit strange from my part of the > world to. > > -- > MfG/Best regards, Kurt Jaeger You have some other forces are work there obviously. There is no possible way that polling would yield a 5 fold increase in performance. You're still using the same cpu to do the tasks. 11MG/s is such a small number of packets that it wouldn't generate enough interrrupts to make much difference to a modern processor. Thats what happens when you use benchmarks. Benchmarks are BS. Probably one of the many "features" that FreeBSD 5.x offers. It likely it has more to do with FreeBSD 5s handling of interrupts in an SMP environment than anything having to do with polling. DT Danial __________________________________ Yahoo! Mail - PC Magazine Editors' Choice 2005 http://mail.yahoo.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051014212808.1778.qmail>