On Dec 12, 2006, at 8:49 AM, John Randall wrote:
Anyone interested in the gory details of how elementary functions are
calculated in floating-point arithmetic might take a look at
fdlibm, in
particular
Sources I have found useful are Cody & Waite's "Software Manual for
the Elementary Functions". It devotes 24 pages to EXP. It assumes
that the radix B of the floating-point number system is either 10 or
2^k, where 1 <= k <= 4; and Hart et al's "Computer Approximations".
I used these when I was asked to rewrite Sharp APL's elementary
functions and matrix inverse, because Sharp wanted to sell its APL
system to Hitachi, and had to ensure to Hitachi that the Sharp code
sold had no vestiges of proprietary IBM code. We also had to write
the complex versions of all these -- which I did, with Doug Forkes'
help.
Eugene
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
