Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!pyramid!prls!mips!mash From: mash@mips.UUCP (John Mashey) Newsgroups: comp.arch Subject: MIPS Performance Brief - errata Message-ID: <872@winchester.UUCP> Date: Tue, 3-Nov-87 20:37:02 EST Article-I.D.: winchest.872 Posted: Tue Nov 3 20:37:02 1987 Date-Received: Sat, 7-Nov-87 06:15:41 EST Lines: 28 Keywords: benchmarks A few errata, passed along by David Hough: >-On levels: Sun's don't correspond on MIPS's. -O1 is peephole only. >-O2 does global optimization on local variables. -O3 does global >optimization of globals and pointers. -O4 does something else that I don't >understand (doesn't seem to affect floating point). We do nothing >interprocedural at any level. (MASH: O1 corresponds pretty well, and our O2 is somewhere between Sun's O2 and O3, I think.) (MASH: I weren't sure what they did, except I knew -O4 didn't do inlining, which is why I cited the Sun -O4 number for Dhrystone.) >"Sun 4/260 w/FPA" is a misleading citation since all Sun-4's have >floating point hardware standard. So "Sun-4/260" is what to call it. >Spice 2G6 has compiled correctly at -O3 (the highest Fortran level) >for many months. One module of Spice 3b1 doesn't compile at -O3, but >that's irrelevant for Fortran. Oops! This was a real goof. I'd remembered the non-compile, and wrongly associated that with 2G6, not 3B1. Needless to say, if anybody else finds factual errors, especially in numbers for their machines, please pass them along. -- -john mashey DISCLAIMER: UUCP: {ames,decwrl.prls,pyramid}!mips!mash OR mash@mips.com DDD: 408-991-0253 or 408-720-1700, x253 USPS: MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086