Newsgroups: comp.arch
Path: utzoo!henry
From: henry@utzoo.uucp (Henry Spencer)
Subject: Re: quest for breakthroughs (long)
Message-ID: <1989Feb22.163745.7001@utzoo.uucp>
Organization: U of Toronto Zoology
References: <740@tetons.UUCP> <76700068@p.cs.uiuc.edu> <671@oracle.oracle.com>
Date: Wed, 22 Feb 89 16:37:45 GMT

In article <671@oracle.oracle.com> csimmons@oracle.UUCP (Charles Simmons) writes:
>My conclusion is that for an elegant architechture, such as the R2000,
>combinatorial searches won't buy you much.  On less elegant
>architechtures, such as the 680x0 and 80x86, there are a sufficient
>number of special cases that a combinatorial search can find cute
>code sequences...

I haven't read the Massalin paper yet -- possibly it addresses this --
but I am compelled to wonder whether those cute code sequences are really
faster than straightforward ones, given the attention that the chip
designers usually pay to optimizing the most common (i.e. simple) cases.
There have been surprises in this area in the past.
-- 
The Earth is our mother;       |     Henry Spencer at U of Toronto Zoology
our nine months are up.        | uunet!attcan!utzoo!henry henry@zoo.toronto.edu