Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!mailrus!ncar!ico!rcd From: rcd@ico.isc.com (Dick Dunn) Newsgroups: comp.arch Subject: Re: Compiler Costs Summary: one more step back Message-ID: <1990Jul12.045146.19275@ico.isc.com> Date: 12 Jul 90 04:51:46 GMT References: <1797@apctrc.UUCP> <759@bbking.KSP.Unisys.COM> Organization: Interactive Systems Corporation, Boulder, CO Lines: 21 mwm@raven.pa.dec.com (Mike (Real Amigas have keyboard garages) Meyer) writes: > rmarks@KSP.Unisys.COM (Richard Marks) writes: > Moral: Consider implementing in your favorite higher level language; > but then recode the inner loops in assembler for perfromance. > > You missed a step - instrument the code to find out where the time is > really being spent, and why. Your intuition about that is probably > wrong... I'll add one more step to precede these two: Decide whether performance really matters before you start any tuning. Mike and Richard may have assumed that this step had already been taken, but it still bears pointing out. First think whether performance is an issue, and if so, whether it will be cost-effective to fix it. Then (as Mike sez) figure out where the performance problems are...and he's right that most people miss the problem areas. Don't tune something that doesn't matter. You'll probably find that most of the time is spent in a tiny area, at which point you apply Richard's advice. -- Dick Dunn rcd@ico.isc.com -or- ico!rcd (303)449-2870 ...Never attribute to malice what can be adequately explained by stupidity.