Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!gatech!hao!husc6!mit-eddie!genrad!decvax!ucbvax!ORVILLE.ARPA!rowley
From: rowley@ORVILLE.ARPA (Karl Rowley)
Newsgroups: comp.sys.atari.st
Subject: MWC and VAX floating-point format
Message-ID: <8706171715.AA08471@orville.arpa>
Date: Wed, 17-Jun-87 13:15:24 EDT
Article-I.D.: orville.8706171715.AA08471
Posted: Wed Jun 17 13:15:24 1987
Date-Received: Sun, 21-Jun-87 10:03:22 EDT
Sender: daemon@ucbvax.BERKELEY.EDU
Distribution: world
Organization: The ARPA Internet
Lines: 23



The reason why MWC chose the VAX floating-point format has been questioned.

VAX double-precision format is identical to single-precision format, except
that the mantissa is extended.  The IEEE format does not have this property.
A floating-point library to use the VAX format is simplified by this 
similarity.  

In addition, all function arguments of type float are converted to doubles 
before they are put on the stack in MWC.  Conversion from floats to doubles 
and back is certainly simple enough in VAX format.

In other words, implementing the VAX floating-point format in software is
probably simpler than implementing the IEEE format in software.

Of course, IEEE format is superior in many ways, and doesn't have its bytes
all flipped around.



Karl Rowley
rowley@orville.nas.nasa.gov