Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!sdd.hp.com!spool.mu.edu!snorkelwacker.mit.edu!thunder.mcrcim.mcgill.edu!mouse
From: mouse@thunder.mcrcim.mcgill.edu (der Mouse)
Newsgroups: comp.lang.c
Subject: Re: how to write hash function for double
Message-ID: <1991Jun26.095958.6495@thunder.mcrcim.mcgill.edu>
Date: 26 Jun 91 09:59:58 GMT
Article-I.D.: thunder.1991Jun26.095958.6495
References: <67790003@hpcupt1.cup.hp.com>
Organization: McGill Research Centre for Intelligent Machines
Lines: 25

In article <67790003@hpcupt1.cup.hp.com>, thomasw@hpcupt1.cup.hp.com (Thomas Wang) writes:

> I want to write a hash function for C double floating point number.
> I can either convert the double to string, then hash the string
> value.  Or I can just scramble the bits contained in the double
> number.

It depends on what properties you want the hash to have, I suppose.  Do
you require that two numerically equal numbers must have the same hash
value, or do you need identical hash values only for identical
representations, instead of for identical values?

> The question is whether there can be two different bit patterns of
> double that represented the same double number?

Certainly.  On a VAX, for example, any bit-pattern with the sign and
exponent fields full of 0 bits represents an exact 0.0 value.  (There
are 2^23 - 8388608 - such bit-patterns.  And that's just for
single-float.)

					der Mouse

			old: mcgill-vision!mouse
			new: mouse@larry.mcrcim.mcgill.edu