Xref: utzoo comp.arch:17051 comp.lang.misc:5139 Newsgroups: comp.arch,comp.lang.misc Path: utzoo!henry From: henry@zoo.toronto.edu (Henry Spencer) Subject: global optimization Message-ID: <1990Jul11.164711.19378@zoo.toronto.edu> Organization: U of Toronto Zoology References: <1797@apctrc.UUCP> <1990Jul6.161158.1297@zoo.toronto.edu> <1990Jul8.230954.18881@ecn.purdue.edu> <23268@boulder.Colorado.EDU> <23285@boulder.Colorado.EDU> Date: Wed, 11 Jul 90 16:47:11 GMT In article meissner@osf.org (Michael Meissner) writes: >I've always wondered about the gains. Speaking off the cuff, I would >imagine you would get gains if you limit yourself to a Fortran style >of coding (no matter what language you use)... There are many people who will pay almost anything to get their Fortran-style programs to run a little faster. Very little of the existing software base is in the more dynamic languages like Lisp. Actually, it is likely that you will always want to write in a Fortran style, or a more sophisticated variant that a compiler can turn into Fortran style, if you want maximum speed. Deferring decisions to run time invariably has a price. >I also wonder about the computational complexity of global register >analysis... As John Mashey has commented, you should make a distinction between what you routinely do for unimportant programs and what you are willing to do for a major program that has just *got* to run faster. (Actually, he put it more dramatically: "we've got customers who will commit vile acts for another ten percent".) -- NFS is a wonderful advance: a Unix | Henry Spencer at U of Toronto Zoology filesystem with MSDOS semantics. :-( | henry@zoo.toronto.edu utzoo!henry