Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!csd4.milw.wisc.edu!uxc!uxc.cso.uiuc.edu!m.cs.uiuc.edu!p.cs.uiuc.edu!gillies
From: gillies@p.cs.uiuc.edu
Newsgroups: comp.sys.mac
Subject: Re: My disillusioned self.
Message-ID: <76000365@p.cs.uiuc.edu>
Date: 8 Mar 89 16:06:00 GMT
References: <235@jarthur.Claremont.EDU>
Lines: 20
Nf-ID: #R:jarthur.Claremont.EDU:235:p.cs.uiuc.edu:76000365:000:910
Nf-From: p.cs.uiuc.edu!gillies    Mar  8 10:06:00 1989


To get a fast grahics program, it also takes a lot of hand optimization.

I remember hand-optimizing some graphics code once in a high level
interpreted language.  I wrote the routine 5 different ways, and
benchmarked each way.  Surprisingly, the version with the most
commands ran the fastest -- more than twice as fast as the others, and
seemingly, with no preemption on the timesharing system.

The 68020 is very complex because of the ideosyncracies of the cache.
This sort of experimentation is probably fruitful in 68020 assembly
language, too.

Apple also doesn't have Andy Hertzfeld any more, but luckily they're
smart enough to pay for his independent consulting projects occasionally!
(Kudos to Apple for buying quickergraf).

Don Gillies, Dept. of Computer Science, University of Illinois
1304 W. Springfield, Urbana, Ill 61801      
ARPA: gillies@cs.uiuc.edu   UUCP: {uunet,harvard}!uiucdcs!gillies