Date: Sun, 10 Jan 1999 13:30:31 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: Charles Krug <charles AT mail DOT pentek DOT com>
cc: djgpp AT delorie DOT com
Subject: Re: Still working on c++ std <limits> file . . .
In-Reply-To: <369638F9.21FAB385@mail.pentek.com>
Message-ID: <Pine.SUN.3.91.990110133010.20678O-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp AT delorie DOT com


On Fri, 8 Jan 1999, Charles Krug wrote:

> First, libm defines functions for nan() and nanf().  <limits>
> asks for signaling_NaN and for quiet_NaN.  Are the nan's defined
> in libm signaling or quiet?

Why should you care about different types of a NaN?  Aren't these
definitions related to how the FP environment is set up by the library
startup code?  In other words, if the ``signalling NaN'' doesn't
really raise any signal, should you care about its definition?

The DJGPP library startup code (in v2.02) masks all FP exceptions, so
a NaN will never cause any signal be delivered to your programs.
Therefore, I would think you should not care about the different types
of a NaN.

> Second, do floating point numbers under djgpp have denormalized
> forms.

Please explain what do you mean by that.