[J3] More Bessel functions

[Keith Bierman]

On Tue, Jan 19, 2021 at 7:26 PM Steven G. Kargl via J3 <
j3 at mailman.j3-fortran.org> wrote:

> ...many sage words elided...
> > A competent vendor will provide a good one,
> True.  One can find Sun Microsystem's libm implementation
> on Netlib.,,,,

as a former member of Sun floating point group, thanks for your vote of
confidence. Competent vendors of compilers often focus on the compiler, not
in a dedicated team of numerical experts. Ideally, every compiler group
would have such experts embedded so optimization, language features and
numerical issues would be dealt with in an ideal fashion ...

Sadly, this is seldom true these days. So I'd advise against adding more
numeric routines even where there is ample demand. Focusing on making the
language suitable (both in expressiveness and optimizability) and having
portable libraries is probably the best we can expect. This has been true
for many years now (the odd vendor or two counterexamples notwithstanding).
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.j3-fortran.org/pipermail/j3/attachments/20210120/6c80f612/attachment.htm>