Lanczos Gamma formula (Re: AW: AW: Increasing the accuracy of Gamma for double float)



Raymond Toy wrote:
> To compute the cofficients accurately, we need to use bfloat arithmetic.
>  I set fpprec = 100, and convert the result back to float.  Thus
> float(bfloat(lanczos_coeff(607/128,15)*exp(607/128)/sqrt(2*%pi))) gives

My mistake.  The coefficients should be lanczos_coeff(607/128,14).  This
 then produces results that match the coefficients in gamma-lanczos.

Ray