Am Freitag, den 02.12.2011, 20:49 -0800 schrieb Raymond Toy:
> On 12/2/11 8:06 PM, Raymond Toy wrote:
> > On 12/2/11 4:32 PM, Stavros Macrakis wrote:
> >> Forget about the error message... the values are completely wrong for
> >> x>35 or so. Serious bug!
> >
> > Haha! I think I know how to fix this. Si can be expressed in terms of
> > the incomplete gamma function and I think that works pretty well. At
> > least for large x, it stays near %pi/2.
>
> There are quite a few problems with the incomplete gamma functions.
>
> gamma_incomplete(0, %i*100d0) -> 2.1e24*%i+2.7e23
>
> but the real answer is closer to 0.005-0.008*%i.
>
> There's also Bug 3220128 and 3220118.
>
> I know how to fix these problems, but it will take some time to update
> the code.... And the release branch is just a few days away.
>
> Ray
Hello Ray,
yes, the expansions for the function Gamma incomplete do not give
correct results for bigger values of the arguments. This is an old
problem. If you already know better expansions, I would appreciate if
you could implement the more correct expansions.
Dieter Kaiser