Xref: utzoo comp.misc:2110 comp.arch:3946
Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!claris!apple!bcase
From: bcase@Apple.COM (Brian Case)
Newsgroups: comp.misc,comp.arch
Subject: Re: Instruction Scheduling
Message-ID: <7691@apple.Apple.Com>
Date: 15 Mar 88 19:16:20 GMT
References: <12513@sgi.SGI.COM> <12560@sgi.SGI.COM> <12678@sgi.SGI.COM> <519@imagine.PAWL.RPI.EDU> <4553@aw.sei.cmu.edu>
Reply-To: bcase@apple.UUCP (Brian Case)
Organization: Apple Computer Inc, Cupertino, CA
Lines: 15
Keywords: optimization   pipeline constraints  code re-organization

In article <4553@aw.sei.cmu.edu> firth@bd.sei.cmu.edu.UUCP (Robert Firth) writes:
    [about keeping information around until the last minute to help
     reorganization and such.]

Although it doesn't reflect quit the same level of aggressiveness that
Robert talks about, the paper

    David W. Wall. Global register allocation at link-time.  _Proceedings
    of the SIGPLAN '86 Symposium on Compiler Construction._  _SIGPLAN
    Notices 21_ (7): 264-275 (July 1986).

talks about keeping information around until the very last.  I think
Robert's comments about very good compilers needing to abandon the now-
popular front-end back-end separation (so that "all you have to do is
write a new front end for a new language") will hold true.  It seems like
there is much to be gained.