From owner-freebsd-hackers Tue Apr 14 14:44:27 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id OAA21121 for freebsd-hackers-outgoing; Tue, 14 Apr 1998 14:44:27 -0700 (PDT) (envelope-from owner-freebsd-hackers@FreeBSD.ORG) Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id VAA20946; Tue, 14 Apr 1998 21:43:41 GMT (envelope-from dyson@dyson.iquest.net) Received: (from root@localhost) by dyson.iquest.net (8.8.8/8.8.8) id QAA04770; Tue, 14 Apr 1998 16:43:21 -0500 (EST) (envelope-from dyson) Message-Id: <199804142143.QAA04770@dyson.iquest.net> Subject: Re: SMP performance decrease? In-Reply-To: <199804141149.PAA19398@asteroid.svib.ru> from "Alexander B. Povolotsky" at "Apr 14, 98 03:49:35 pm" To: tarkhil@asteroid.svib.ru Date: Tue, 14 Apr 1998 16:43:21 -0500 (EST) Cc: hackers@FreeBSD.ORG, smp@FreeBSD.ORG From: "John S. Dyson" Reply-To: dyson@FreeBSD.ORG X-Mailer: ELM [version 2.4ME+ PL38 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Alexander B. Povolotsky said: > Hello! > > I've just benchmarked (by bytebench from ports) my new server, with 64M RAM > (for now; will increase soon) and 2 PII-233 processors (EliteGroup p6LX2-A > mainboard). > > Here are the results: > > Test Non-SMP SMP > arith (double) 27.5 28.1 > Drystone 2 (no reg. vars) 23.6 24.1 > Execl thoughput 90.8 46.8 <- ??? > File copy 35.1 37.7 > Pipe-based context switching 20.4 9.0 <- ??? > Shell scripts (8 concurrent) 12.8 12.2 <- ??? > > I don't understand why execl throughput and context switching is TWICE > slower in SMP mode, and I doesn't understand AT ALL why shell script > benchmark isn't twice faster. > > Wrong benchmark? Problems with something in kernel? Me being crazy? > Certain things in the SMP kernel make it actually slower, and seem slower both. If you load your machine with multiple processes, the system will be faster with SMP and two processors than with one. (Of course, it is possible to make the system slower with SMP and two processors, but that is not generally true.) Bottom line: you are running a low level benchmark, and low level benchmarks don't meassure loaded system performances very well. :-(. -- John | Never try to teach a pig to sing, dyson@freebsd.org | it just makes you look stupid, jdyson@nc.com | and it irritates the pig. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message