Newsgroups: | comp.compilers |
From: | elliottm@csulb.edu (Mike Elliott) |
Keywords: | assembler, performance |
Organization: | Cal State Long Beach |
References: | 93-10-114 93-10-131 |
Date: | Fri, 29 Oct 1993 07:28:57 GMT |
Mark_Prince@gec-epl.co.uk writes:
I don't think the point is really about assemblers Vs compilers. Many
elequent statements have been made about this debate in this group but we
all already know those answers. Maybe the point to be made is that there
are people out there who are willing to make contentious statements in
high-profile magazines which then lead to 1) such a debate 2) un-informed
people who take anything on paper as gospel (not those in the know :-)
Could articles of this type lead to customers saying "What are the odds of
you coding my new super-ultra-complex system purely in assembly language ?"
Precisely. I think we all know that in microcosm human beings can
out-perform compilers. I specified a "real-world application" rather than
a tight inner loop or a subroutine or two in order to address the
macroscopic problem. The magazine article talked about entire DOS
applications, presumably something on the order of the size of recent
versions of WordPerfect or Lotus 1-2-3.
Actually I had a reference in mind, but I'm not sure it ever existed -- I
just heard about it (a Folk Reference, perhaps?); I've never actually
tracked it down, but . . .
Long ago and far away, when I was a graduate student at Edinburgh
University, I heard that the operating system EMAS (Edinburgh Multi-Access
System) was re-written in the systems programming language IMP from its
predecessor in assembly code. A study of code size difference was
performed, and the results were that the OS had changed very little at all
in overall bulk. Some sections differed by five to ten percent --
sometimes in favor of assembly, sometimes in favor of IMP. Overall it was
basically a wash.
I feel I can't cite this, because I don't know for sure that such a study
was actually done, or if it was, that these are the results. Scholarship
demands more rigor than just vague memories -- besides, I had just been
introduced to the astonishing variety of ales, stouts, lagers, bitters and
whiskies available in Scotland -- so vague memories just don't cut it.
That's the sort of reference I was seeking and I'm disappointed no one has
yet mentioned anything! If such research has not actually been done, then
it seems like it ought to be!
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.