Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site sequent.UUCP
Path: utzoo!linus!security!genrad!grkermit!masscomp!clyde!floyd!vax135!cornell!uw-beaver!tektronix!ogcvax!sequent!tdl
From: tdl@sequent.UUCP
Newsgroups: net.arch
Subject: Re: Registers vs. cache
Message-ID: <306@sequent.UUCP>
Date: Tue, 13-Dec-83 11:12:24 EST
Article-I.D.: sequent.306
Posted: Tue Dec 13 11:12:24 1983
Date-Received: Sat, 17-Dec-83 02:26:38 EST
Organization: Sequent Computer Systems, Portland
Lines: 12

The work by Johnson that was mentioned in a previous article is
apparently still around BTL. There is an article by Ditzel and
McLellan called "Register Allocation for Free: The C Machine 
Stack Cache" that was presented at the 1982 conference on Architectural
Support for Programming Languages and Operating Systems (ASPLOS).
They cover most of the same arguments against caches that have been
presented here (expensive to dual-port, large addresses rather than
short register addresses, etc.) and then describe how they overcome
these problems. They come up with a solution that has all of the benefits
of registers without any of the problems.
Tom Lovett - Sequent Computer Systems, Inc.
{pur-ee | ogcvax} !sequent!tdl