On Thu, 2006-07-13 at 19:46 +0200, Carsten Neumann wrote:
>       Hi,
> 
> the attached patch implements java.lang.StrictMath#cbrt, but I'm
> actually quite ashamed to post it: I'm not very knowledgeable about
> floating point arithmetic, so I just followed the implementation in
> fdlibm (www.netlib.org/fdlibm). Unfortunately this library does some
> weird manipulations of the bits of a double, which I simply mimicked in
> java.

That's probably okay though. Actually it's probably a lot better coding
practice to do it in Java than in C since Java at least has a strictly
defined idea of what a double is.

(A bigger problem with some VMs (*cough* GCJ) is that the internal
representation of a double isn't right. E.g. the might use the
processors extended 80-bit doubles *cough* instead of IEEE 64-bit ones)

One question though.. would it make sense to use longs instead of
int[2] for the 'word' stuff? You could use the Double.doubleFromRawLong
method to convert back to double then.

/Sven


Reply via email to