Date: Tue, 14 Nov 2000 16:13:00 +1100 (EST) From: Bruce Evans <bde@zeta.org.au> To: "Thomas D. Dean" <tomdean@ix.netcom.com> Cc: current@FreeBSD.ORG, jhb@FreeBSD.ORG Subject: Re: sio patch to use mutexes Message-ID: <Pine.BSF.4.21.0011141541020.2084-100000@besplex.bde.org> In-Reply-To: <200011140356.eAE3uNd00390@ix.netcom.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 13 Nov 2000, Thomas D. Dean wrote: > I am seeing lots of silo overflows in -current SMP. > > #uname -a > FreeBSD celebris 5.0-CURRENT FreeBSD 5.0-CURRENT #2: \ > Mon Nov 13 19:31:49 PST 2000 \ > root@celebris:/usr/src/sys/compile/CELEBRIS-SMP i386 SMPng has many pessimizations that break sio on slow machines, but I would have thought that your machine (P5 128MHz?) was fast enough. With non-SMP kernels, I have only seen silo overflows on my slowest runnable machine (486DX2/66 with 16450 at 115200 bps. A 386/20 with an 8250 at 115200 bps should work). There is just too much code in latency-critical code for slow machines to work (on the 486DX2/66, sio latency is up by a factor of 5-10; sio overheads are only up by 25-50%). I have unfinished fixes for this. > I this is not related to the recent patch for conversion to mutex. I > tried both ways. The patches give small or negative optimizations for the SMP case and not so small pessimizations for the non-SMP case. Bruce To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0011141541020.2084-100000>