Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!sdd.hp.com!uakari.primate.wisc.edu!aplcen!haven!adm!cmcl2!lanl!jlg From: jlg@lanl.gov (Jim Giles) Newsgroups: comp.lang.misc Subject: Re: Aggressive optimization Message-ID: <5213@lanl.gov> Date: 7 Nov 90 20:28:37 GMT References: Organization: Los Alamos Natl Lab, Los Alamos, N.M. Lines: 37 From article , by pcg@cs.aber.ac.uk (Piercarlo Grandi): > On 5 Nov 90 22:15:54 GMT, jlg@lanl.gov (Jim Giles) said: > > [...] Do we want 100 thousand lines of compiler > to debug when 10 thousand would do it? No... Naturally not all those 100 > thousand are involved in the critical translation business, but the > translation business had better be as streamlined as possible. I don't know what 100 thousand lines you are referring to. The techniques I am recommending make compilers smaller and easier to write and maintain. They also, as a side-effect of how they work, carry out several of the optimizations that you are on record as opposing. I would not trust an ad-hoc compiler of the kind you seem to be recommending to get the "hello world" example correct. > [...] > More code more bugs, no question about it. It is not an assumption. > Anything that is not related to straight translation means more bugs. In other words, ad-hoc "straight" translations is better that structured application of known compiler techniques. Since most of the known techniques are off-the-shelf the compiler writer can concentrate on getting any differences specific to his own task working correctly. > [...] > But these techniques require pretty long and elaborate code. [...] Really? The smallest, quickest written Modula-2 compiler I've seen has nearly all the techniques That I've been discussing. The goal of the thing was to produce a simple compiler that concentrates of getting the best code generated for the least effort by the compiler. Don't know where this "long and elaborate code" you're talking about is hiding. You must be thinking about the esoterica that I mentioned last time (and which I specifically said I _wasn't_ talking about). J. Giles