fsset@bach.lerc.nasa.gov (Scott E. Townsend) (01/16/91)
[ sorry if this is an inappropriate group, it seems the most likely one to me ] I'm trying to implement the power (x ** y) function according to the algorithm described in 'Software Manual for the Elementary Functions' by Cody & Waite. This is the book which introduced the ELEFUNT code on NETLIB. Anyway, they describe an operation REDUCE(v) with a suggested implementation [1] FLOAT(INT(16.0*V)) * 0.0625 and later require [2] |V-REDUCE(V)| < 1/16 Now my problem is how does [1] imply [2] for large V? REDUCE() is used on the y operand in x**y, but I haven't found anyplace where y is made reasonably small before REDUCE is invoked. With a y of say 1.0e100 this is a problem. I realize this is a bit obcure, but I hope someone out there has run into this before. Thanks for any & all replies. Scott -- ------------------------------------------------------------------------ Scott Townsend | Phone: 216-433-8101 NASA Lewis Research Center | Mail Stop: 5-11 Cleveland, Ohio 44135 | Email: fsset@bach.lerc.nasa.gov ------------------------------------------------------------------------