another independent ATAN discovery

[David G. Hough at validgh]

Article 703 in comp.arch.arithmetic:
From: broadleyaturing.ucdavis.edu (Bill Broadley)
Newsgroups: comp.sys.intel,comp.arch.arithmetic
Subject: ATAN bug
Message-ID: <3c2hdb$otpamark.ucdavis.edu>
Date: 6 Dec 94 20:26:19 GMT
Organization: University of California, Davis
Lines: 18
NNTP-Posting-Host: turing.ucdavis.edu
X-Newsreader: TIN [version 1.2 PL2]

I missed any announcment of numbers that set off the ATAN bug, but I
wrote a program to take a out=atan(tan(in)) and found the following
error above 1e-15.

in=0.8549592142878325 out=0.8549592142114351 diff=0.0000000000763974

If your interested in source it's at:
        math.ucdavis.edu:/pub/fdiv/tan.c

I'd be interested in any input on the size of that error.  It was
found by a linear search of the out=tan(atan(in)) over the range
of 0 to pi/4.

This was found in 7 hours or so.

--
Bill Broadley           Broadleyamath.ucdavis.edu           UCD Math Sys-Admin
Linux is great.         http://ucdmath.ucdavis.edu/~broadley            PGP-ok