Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!gem.mps.ohio-state.edu!ginosko!uunet!brunix!doorknob!nrt From: nrt@cs.brown.edu (Nick Thompson) Newsgroups: comp.sys.next Subject: Re: NeXT vs Sun3 Message-ID: Date: 22 Oct 89 03:27:56 GMT References: <925@abvax.UUCP> <13246@orstcs.CS.ORST.EDU> <17434@watdragon.waterloo.edu> <191@med.Stanford.EDU> Sender: news@brunix.UUCP Organization: Brown University Department of Computer Science Lines: 19 In-reply-to: mccalpin@masig3.masig3.ocean.fsu.edu's message of 21 Oct 89 17:24:15 GMT (John D. McCalpin) writes: Interestingly enough, the GNU compiler on the NeXT typically runs floating-point intensive codes *faster than FORTRAN*. This is the only machine that I know of that can make that claim. It says very good things about the GNU C optimizer. One could interpret it to say bad things about the Absoft FORTRAN compiler, but I have found that the Absoft compiler actually optimizes pretty well --- almost as well as the Sun-3 f77 compiler with full optimization. Is gcc on the NeXT using inline 68882 instructions? If it is and the Fortran compiler is not, this could explain the speed difference. Saving a function call for every operation will do that... Nick Nick Thompson : nrt@cs.brown.edu | nrt@browncs.bitnet | uunet!brunix!nrt "Lie down on the floor and keep calm." -mantra attributed to John Dillinger