Path: utzoo!mnetor!uunet!husc6!bbn!rochester!rutgers!jvnca!hook
From: hook@jvnca.csc.org (Ed Hook)
Newsgroups: comp.lang.c
Subject: Re: Why NULL is 0
Message-ID: <224@jvnca.csc.org>
Date: 7 Apr 88 16:53:30 GMT
References: <2550049@hpisod2.HP.COM> <7412@brl-smoke.ARPA> <3351@chinet.UUCP> <10229@steinmetz.steinmetz.ge.com> <1323@PT.CS.CMU.EDU>
Reply-To: hook@jvnca.csc.org.UUCP (Ed Hook)
Organization: John Von Neumann Center, Princeton, N.J.
Lines: 37

In article <1323@PT.CS.CMU.EDU> edw@IUS1.CS.CMU.EDU (Eddie Wyatt) writes:
>Aside:
>
>> People (usually ;>) have no problem with the idea the while assigning a
>> zero to a float gives it the vaule 0.0, in most implementations the
>> float value does not have all bits set to zero.
>
>  I do believe IEEE and Vax's version of float is a bit pattern
>consisting of all zeros.  So what are these implementations that don't
>use zero bit pattern?
>

   There's at least one architecture with this characteristic. The Control
Data Cyber 205 ( and its successor, the ETA^10 ) employ a 64-bit floating
point representation in which zero appears as

                    8000 0000 0000 0000 ;

in fact, any word whose most significant nybble is an '8' is = 0.0 in this
system - the example above is the flavor of 0.0 which is produced by any
computation  whose result is zero ( and, thus, is the most familiar example 
of 0.0 ).

   I can't seem to find my copy of the IEEE Floating Point Standard & so can't
cite chapter & verse, BUT I don't believe that it requires a valid floating
point 0.0 to be represented by a bit pattern having all bits = 0. I say this
because it didn't appear to me that the 205's representation of floating point
quantities was in violation of the standard on this score, when I went through
the document recently. If I'm wrong in holding this opinion, I would appreciate
being set straight ...

Ed Hook
Control Data Corporation
John von Neumann National Supercomputer Center
Princeton, NJ

"If I said it, it's MY opinion ... at least for now ..."