Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!water!watnot!watmath!clyde!rutgers!husc6!seismo!mimsy!mark
From: mark@mimsy.UUCP
Newsgroups: comp.arch,comp.lang.c
Subject: Re: String Processing Instruction
Message-ID: <5969@mimsy.UUCP>
Date: Thu, 26-Mar-87 10:51:35 EST
Article-I.D.: mimsy.5969
Posted: Thu Mar 26 10:51:35 1987
Date-Received: Sat, 28-Mar-87 07:08:53 EST
References: <15292@amdcad.UUCP> <978@ames.UUCP>
Reply-To: mark@mimsy.UUCP (Mark Weiser)
Distribution: na
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 21
Keywords: instruction set architectures, Am29000
Xref: utgpu comp.arch:670 comp.lang.c:1342

In article <978@ames.UUCP> lamaster@pioneer.UUCP (Hugh LaMaster) writes:
>...How likely is it that a compiler [can] detect the case when it can use 
> an instruction like this and generate code automatically to use it?

This is of course a good point, and it seems that this is going to be
one of those hard cases for compiler writers.  But it opens the issue
of subroutine libraries as providing an environment which also needs
support in the instruction set.  The Unix string routines are a good
example, because they are even now often recoded in assembly for efficiency.
Various pieces of Unix kernel frequently suffer the same fate
(e.g. tty buffers).  But once one is into instructions which are good
for certain operating systems or environments, rather than certain languages,
one has narrowed one's market.  For instance, the Am29000 special string
compare is useless if strings are encoded as a length followed by the chars.  
-mark
Background: I happen to think Unix-style strings are a big market, and I love
things which make Unix faster (because there is a positive feedback cycle
which makes more Unix yet more available and desirable.).
-- 
Spoken: Mark Weiser 	ARPA:	mark@mimsy.umd.edu	Phone: +1-301-454-7817
After May 1, 1987: weiser@xerox.com