Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uflorida!gatech!udel!rochester!pt.cs.cmu.edu!fas.ri.cmu.edu!schmitz
From: schmitz@fas.ri.cmu.edu (Donald Schmitz)
Newsgroups: comp.sys.m68k
Subject: Re: 68020 instruction scheduling
Message-ID: <5324@pt.cs.cmu.edu>
Date: 26 Jun 89 21:42:22 GMT
References: <176@uhura.cs.wisc.edu>
Organization: Carnegie-Mellon University, CS/RI
Lines: 9

>Is instruction scheduling a useful optimization on the 68020 ?

One place it can be a big win is floating point using the 68882 (note 2, not
'881).  The '882 can do a move to/from fp register and an arithmetic
operation that doesn't use that register at the same time.  I have hand
assembled code to do 4x4 matrix multiplies that takes advantage of this
feature, its about a 50% improvment over straightline C code.

Don