Path: utzoo!attcan!uunet!husc6!mailrus!uflorida!novavax!proxftl!bill From: bill@proxftl.UUCP (T. William Wells) Newsgroups: comp.lang.c Subject: Re: Value of microeffiency (was: Re: Optimal ...) Summary: we care Message-ID: <416@proxftl.UUCP> Date: 4 Jul 88 02:13:05 GMT References: <163@navtech.uucp> <2775@ttrdc.UUCP> <164@navtech.uucp> <3401@rpp386.UUCP> Organization: Proximity Technology, Ft. Lauderdale Lines: 42 In article <3401@rpp386.UUCP>, jfh@rpp386.UUCP (John F. Haugh II) writes: > i don't know what value microefficiency has this week, but in general, > writing good solid algorithms is what is important. Just so you know: around here, what you are calling microefficiency makes the difference between staying in business or not. There is no argument against the idea that a good algorithm is the proper base from which to start (though there is certainly room for discussion as to what constitutes a good algorithm). But, a good algorithm is ONLY THE BEGINNING. Once you have an algorithm, you must IMPLEMENT it. BOTH are necessary. And as important. It is not either/or. Some people argue against the notion that altimately, resource consumption is irrelevant to what constitutes a good algorithm. Their "reason" is that processors get better, memory gets cheaper, disks get faster, etc. Actually, what happens is this: Consumption expands to the limits of resources. So, I might accept that the days of PC's with less than a meg have gone (and yes I know that that is BS; but, just for the sake of the argument), but that does not mean that our spelling checker can get larger. Oh no. What that means is that our customers expand their product to use the extra memory. Should our spelling checker get significantly larger, they will go to another vendor. And should it be slower than another's... Anyone who thinks that "microefficiency" is irrelevant either is not in the real world or has customers who have indefinitely deep pockets (or who just do not understand the problem, but time will cure that ignorance). > does anyone still > use profil anymore? what about size? Profilers are essential tools around here; so are any other performance measurement tools that are appropriate to the task. Including size.