Re: Back End Generators

pardo@cs.washington.edu (David Keppel)
Fri, 21 Oct 1994 02:37:29 GMT

          From comp.compilers
Related articles
Back End Generators heronj@smtplink.NGC.COM (John Heron) (1994-10-12)
Re: Back End Generators heronj@smtplink.NGC.COM (John Heron) (1994-10-12)
Re: Back End Generators hbaker@netcom.com (1994-10-14)
Re: Back End Generators johnmce@world.std.com (1994-10-16)
Re: Back End Generators adrian@platon.cs.rhbnc.ac.uk (1994-10-16)
Re: Back End Generators chase@centerline.com (1994-10-21)
Re: Back End Generators pardo@cs.washington.edu (1994-10-21)
Re: Back End Generators bill@amber.ssd.csd.harris.com (1994-10-21)
Re: Back End Generators smucker@cs.wisc.edu (1994-10-21)
Re: Back End Generators Peter.Damron@Eng.Sun.COM (1994-10-18)
Re: Back End Generators hbaker@netcom.com (1994-10-28)
Re: Back End Generators anton@mips.complang.tuwien.ac.at (1994-10-24)
Re: Back End Generators davidm@Rational.COM (1994-10-25)
| List of all articles for this month |
Newsgroups: comp.compilers
From: pardo@cs.washington.edu (David Keppel)
Keywords: code, tools, bibliography
Organization: Computer Science & Engineering, U. of Washington, Seattle
References: 94-10-094 94-10-123
Date: Fri, 21 Oct 1994 02:37:29 GMT

>[Code generation via parsing.]


Well, if you stare at it for a while you can convince yourself
(maybe) that code generators in *general* use parsing, but only
some of them formalize it as that kind of a pattern-matching
problem.


Pioneered by Glanville and extended by Robert Henry, both under
the tutilage Susan Graham. See Berkeley Ph.D lists for the
theses, also a variety of papers including


%A Christopher W. Fraser
%A Robert R. Henry
%T Hard-Coding Bottom-Up Code Generation Tables to Save Time and Space
%J Software \- Practice and Experience
%V 21
%N 1
%D January 1991
%P 1-12


I vaguely recall hearing that the AT&T and Sun compilers
are based on Graham-Glanville-Henry parsing code generation.


>[Start with short sequences and gradually optimize them into
> "large" instructions.]


See:


%A Robert R. Henry
%T Code Generation by Table Lookup
%R 87-07-07
%I University of Washington Computer Science
%D 1987


The idea here is to parameterize the compiler with a basic
machine description and then compile test suites in "training"
mode. The goal is to train the compiler to recognize common
sequences and replace the commonly-appearing sequences with a
"large instruction" that can be matched directly instead of
begin rediscovered by the optimizer during each run.


;-D on ( Making a Parse of my SELF ) Pardo
--

Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.