Path: utzoo!utgpu!water!watmath!clyde!rutgers!rochester!srs!dan
From: dan@srs.UUCP (Dan Kegel)
Newsgroups: comp.arch
Subject: Re: Approaches to computer system performance
Keywords: Algorithms, assembler
Message-ID: <578@srs.UUCP>
Date: 30 Jan 88 14:14:38 GMT
References: <839@ima.ISC.COM> <2158@geac.UUCP> <604@bnr-rsc.UUCP> <180@granite.dec.com> <3081@watcgl.waterloo.edu>
Reply-To: dan@rem.UUCP (Dan Kegel)
Organization: S.R.Systems
Lines: 19

In article <3081@watcgl.waterloo.edu> tbray@watsol.waterloo.edu writes (sorta):
> Given a routine which is sucking up a lot of your resources, how much 
> can you gain by hand-coding in assembler?
> Unless you are lucky and have a deep understanding of your machine's 
> architecture *and* the implementation of that architecture, 
> probably no more than a few percentage points.
>
>Algorithms si!  Low-level hacking no!
>Tim Bray, New OED Project, University of Waterloo

Yes... but lots of people are that lucky, and sometimes the payoff is great.  
I recently rewrote a self-recursive procedure to pass all parameters in 
registers, yielding a 2x speedup.  The resulting code contains
about ten asm() statements mixed with about fifty C statements.
I almost wish it hadn't worked, though; the code is less readable,
and will have to be redone if we move to another machine.
-- 
   Dan Kegel
   srs!dan@cs.rochester.edu  dan%srs.uucp@cs.rochester.edu  rochester!srs!dan