X-Authentication-Warning:	acp3bf.physik.rwth-aachen.de: broeker owned process doing -bs
Date:	Mon, 20 Mar 2000 18:20:42 +0100 (MET)
From:	Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de>
X-Sender:	broeker AT acp3bf
To:	djgpp-workers AT delorie DOT com
Subject:	Re: Unnormals???
In-Reply-To:	<Pine.SUN.3.91.1000320183001.26722O@is>
Message-ID:	<Pine.LNX.4.10.10003201816170.25247-100000@acp3bf>
MIME-Version:	1.0
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	dj-admin AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

On Mon, 20 Mar 2000, Eli Zaretskii wrote:

> On Mon, 20 Mar 2000, Hans-Bernhard Broeker wrote:

> > The sign bit alone does not identify the 'real indefinite'. The mantissa
> > is also fixed. So the 'flag', if any, would be the whole 64 bits of
> > information, not just the sign bit.
> 
> Not really: if you flip the sign but leave the rest alone, the real 
> indefinite becomes a QNaN; you don't need to change the rest of the
> bits.

If you flip any mantissa bit of it, the real indefinite becomes
a QNaN; you don't have to change the sign bit.

Where's the difference that would make the sign bit special, in that
respect? The real indefinite is exactly one pattern of bits, per FP
datatype. Its only reason for existence is that Intel felt they had to
choose a single common result for all cases where a NaN was generated 'out
of the blue', i.e. with no other NaN to copy the freely variable bits from
that are allowed in an NaN. Once it has been generated, it's just an
ordinary QNaN like all the others.

Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de)
Even if all the snow were burnt, ashes would remain.

- Raw text -