Subject: Maxima gives incorrect results with ceiling
From: Stavros Macrakis
Date: Fri, 7 Dec 2007 14:12:48 -0500
As I said, Maxima uses bigfloats *internally* to calculate
ceiling/floor. Unfortunately, the current method of using bigfloats is
not conservative enough to avoid error in some cases. The right way to
handle this would be interval arithmetic, but no one has coded that up
yet.
-s
On 12/7/07, van Nek <van.nek at arcor.de> wrote:
>
> Yes, the following shows that there are bigfloat calculations involved,
> because fpprec
> shouldn't influence float calculations.
>
> (%i1) fpprec:24$
> (%i2) string( ceiling(factorial(32)/exp(1)) );
> (%o2) 96800425246141091510518410066788352
> (%i3) fpprec:25$
> (%i4) string( ceiling(factorial(32)/exp(1)) );
> (%o4) ceiling(263130836933693530167218012160000000*%e^-1)
>
> Can you explain, what is going on there?
>
> Volker
>
>
> Am 7 Dec 2007 um 13:35 hat Stavros Macrakis geschrieben:
>
> *internally* it uses an approximation method
> based (as it happens) on bigfloats.
>
>