Print

Print
Unfortunately, the real (sorry) historical reason isn't very interesting.

Basically, someone went through the then current C math library
(I forget which one) and picked out all of the intrinsics that
weren't already in Fortran and just added them to chapter 13.
The argument (sorry again) was that if Fortran is the premier
scientific programming language, then it ought to have the
same capabilities as the competition.  Presumably, the C
people did the thinking about which functions would be used
enough to be worth doing.

Doing it this way had essentially no implementation costs for
vendors.  They all supported C; all they had to do was teach
their Fortran compiler how to link up to a C routine.

There was no pressure at all from the vendors for in more
routines or extended versions of the existing C routines. 

Dick Hendrickson

On Thu, Dec 25, 2014 at 7:38 AM, Robin Vowels <[log in to unmask]> wrote:
From: "Van Snyder" <[log in to unmask]>
Sent: Tuesday, December 23, 2014 10:30 AM

Even though Don Amos had published ACM TOMS algorithm 644 in 1986, with
remarks to improve it in 1990 and 1995, there was sentiment that
extending beyond real arguments and integer orders would have resulted
in the proposal being rejected as too much work for implementers.

Shouldn't have.  Implementation would have been straightforward,
and the algorithms were published.

The real [no pun intended] question is whether they should have been included at all.
They aren't something everybody uses every day.
Seems to me to be better to include then in your program as subroutines.

Algorithm 831, to compute modified Bessel functions (K and I) for pure
imaginary order and positive real argument appeared in 2004.

Algorithm 877, to compute cylindrical Bessel functions (J, N, H(1) and
H(2)) for complex order and positive real argument appeared in 2008.


---
This email is free from viruses and malware because avast! Antivirus protection is active.
http://www.avast.com