Re: Double precision Floor bug?

Neither .59 nor .01 have exact double precision representation, so these

are rounded (at read time) to

.59 -> 5314247560297185/9007199254740992

.01 -> 5764607523034235/576460752303423488

everything is working correctly, it's just not working on the numbers you

thought it was working on.

On Fri, Jun 1, 2012 at 1:30 PM, Guy Footring <Guy@footring.net> wrote:

I noticed this at work in LispWorks 5.1, but I get the same result in LispWorks 6.1 when I just tried it at home:

(let ((x 0.59d0)
(d 0.01d0))
(multiple-value-bind (div rem) (floor x d)
(format t "floor(~A ~A) => ~A ~A~%" x d div rem)
(multiple-value-bind (div2 rem2) (floor rem d)
(format t "floor(~A ~A) => ~A ~A~%" rem d div2 rem2))))

floor(0.59D0 0.01D0) => 58 0.010000000000000009D0
floor(0.010000000000000009D0 0.01D0) => 1 8.673617379884036D-18

I would have expected the first result to be 59 0.000000000000000009D0

At first I thought it might some kind of floating-point rounding/precision issue, but I don't think that would explain the second value returned by floor being greater than the second argument given to floor.

Regards
Guy Footring

_______________________________________________
Lisp Hug - the mailing list for LispWorks users
lisp-hug@lispworks.com
http://www.lispworks.com/support/lisp-hug.html

--
~jrm