(j3.2006) BESSEL_JN in draft of Fortran 2008

Tobias Burnus burnus
Wed Jun 18 02:58:43 EDT 2008

Van Snyder wrote:
> The reason for the transformational function is that Bessel functions
> are evaluated by recurrence, and in particular by a Miller algorith.
> When Bessel_JN is needed for n1, n1+1, ..., n2, it is far more efficient
This still does not explain why the elemental function and the 
transformational function should to have the same name.

The other question, which is difficult to answer as it depends on the 
personal need, is whether the (n1,...,n2) case happens often enough that 
it makes sense to include this feature in the core language rather than 
leaving it to an optional library function. (If one needs J_N1, ..., 
J_N2 often then one of cause wants to have a fast answer without 
starting from scratch every time.)

For my needs, I rarely need J_N and Y_N and so far I never needed 
J_N1,.... J_N2. Having spherical harmonics or an eigensolver or a sort 
function would be more useful for me. (Note: I'm not suggesting to put 
in an eigensolver into the Fortran standard; and I do like that Fortran 
2008 includes the gamma, erf and Bessel functions.)

In conclusion: I strongly oppose having a function which is, both, 
elemental and transformational. Additionally, I do not see a real need 
for the transformational function, but I do not mind if it is included 
in the standard under a different name.


More information about the J3 mailing list