Path: utzoo!mnetor!uunet!cbmvax!daveh From: daveh@cbmvax.UUCP (Dave Haynie) Newsgroups: comp.arch Subject: Re: Tasting of Dhrystone 2.0 Results Message-ID: <3505@cbmvax.UUCP> Date: 24 Mar 88 16:39:03 GMT References: <4076@vdsvax.steinmetz.ge.com> Organization: Commodore Technology, West Chester, PA Lines: 34 in article <4076@vdsvax.steinmetz.ge.com>, barnett@vdsvax.steinmetz.ge.com (Bruce G. Barnett) says: > |Not knowing what a 68881 chip does (last time I touched a Sun, it > |was a Sun/2, sigh), I can only suggest that if it has any bearing > |on the results that submitters please include this information. > The 68881 chip is an optional floating point chip for the low end Sun 3/50. > Silly me. This has little impact on Dhrystones, correct? > I still don't understand how one Sun 3/50 can be 40% faster than another. Other than adding to context switch time a tad, the 68881 shouldn't be involved at all. If the same exact machine is turning in a 40% difference, I'd suspect that some kind of unusal load kicks in every now and then, or maybe there's a problem with the real-time response of the timer function call. If these are different machines, I'd suspect that some of them are running a few things in the background that are eating into your Dhrystone time. The best way to run it would be with all but the normal, "everyone-needs-these", processes shut down. Or perhaps the sample size is too small for the machines we're testing these days. With [a version of] the old Dhrystone on one of the machines I've been working on, I kept getting somewhere between 15 and 16 seconds for the standard number of passes. Boosting the sample size made the results more repeatable. I haven't tried the new code yet, perhaps this has already been done? > Bruce G. Barnett > uunet!steinmetz!barnett -- Dave Haynie "The B2000 Guy" Commodore-Amiga "The Crew That Never Rests" {ihnp4|uunet|rutgers}!cbmvax!daveh PLINK: D-DAVE H BIX: hazy "I can't relax, 'cause I'm a Boinger!"