1. -Infinity / -Infinity
- Posted by Roderick Jackson <rjackson at CSIWEB.COM> Feb 02, 1999
- 415 views
Okay... I wrote a quick test program to play with infinities and not-a-numbers = for a few minutes. My code is: atom A A =3D 1e300 print (1, A) puts (1, ' ') A =3D 1e300 * 1e300 -- this makes A =3D infinity print (1, A) puts (1, ' ') A =3D 0 - A print (1, A) puts (1, ' ') A =3D A / A print (1, A) puts (1, ' ') A =3D 0 - A print (1, A) puts (1, ' ') At "A =3D 0 - A", we have a negative infinity. I understand that. Then = "A / A" (i.e., -inf / -inf) is computed, resulting in a negative nan. = Um... is this *supposed* to be the case with infinities? I would have = thought that two negatives, even infinities, would result in a positive = number (or not-a-number). Or is my higher math education incomplete? (In case it matters, this is version 2.0 of Euphoria... has this been = addressed in the upgrade, assuming it's actually a problem?) Rod Jackson
2. Re: -Infinity / -Infinity
- Posted by Jeffrey Fielding <JJProg at CYBERBURY.NET> Feb 02, 1999
- 443 views
> At "A = 0 - A", we have a negative infinity. I understand that. Then "A / A" > (i.e., -inf / -inf) is computed, resulting in a negative nan. Um... is this > *supposed* to be the case with infinities? I would have thought that two > negatives, even infinities, would result in a positive number (or not-a-number). > Or is my higher math education incomplete? I tried it on my computer and got the same results. A/A should be 1 as long as a isn't 0 (in which case it's undefined), and it definately shouldn't be negitive. There isn't any reason why -infinity/-infinity should equal -(not a number). -- Jeffrey Fielding JJProg at cyberbury.net http://members.tripod.com/~JJProg/
3. Re: -Infinity / -Infinity
- Posted by Robert Craig <rds at EMAIL.MSN.COM> Feb 02, 1999
- 446 views
- Last edited Feb 03, 1999
Rod Jackson writes: > Then "A / A" (i.e., -inf / -inf) is computed, resulting in a negative nan I tested your example with a small C program. I got -nan using hardware (Pentium) floating-point. I guess you should complain to Intel When you are dealing with nan's, does it really matter whether you have -nan or +nan. Either way the result is "not a number", "undefined", "doesn't exist" etc. If it's not even a number, how can -1 * "not a number" have any meaning? Regards, Rob Craig Rapid Deployment Software http://members.aol.com/FilesEu/