Re: problem with the exponential function on the NeXT

*To*: mathgroup at yoda.physics.unc.edu*Subject*: Re: problem with the exponential function on the NeXT*From*: me at leidecker.gsfc.nasa.gov*Date*: Tue, 29 Dec 92 09:52:00 -0500

Joe Herrmann writes that he used his NeXT and Mma to evaluate Table[N[E^x],{x,-700000,-600000,5000}], and then found *wrong* results for x=-670,000 and x=-625,000. However, correct results were observed when running on a friend's Sun. I just evaluated the expression on my NeXT, and did not get the reported wrong values --- all of them looked fine. And my friend, Charles Powers, just ran the full expression on his NeXT --- all the values look fine. In particular, we both get E^(-670,000)=4.97880161*10^(-290978). (Other values supplied on request.) I am running Mma ver. 1.2 (I'll get around to upgrading someday) on a Cube with the 25 MHz 040 board, and NeXTstep 3.0; and Powers is running Mma ver. 2.1 on a NeXTstation with a 33 MHz 040, and NeXTstep 3.0. NeXTstep 2.0 had a bug in the routines that the 040's internal math co-processor used to handle the transcendentals --- perhaps Joe found something like that? Regards, Henning Leidecker