Skip to content

besselK

1 message · Ben Bolker

#
[cc'ing to R-devel]

  It's moderately obscure to me too, but ...

  There are two sets of bessel code in the package, one
in C++ from Chris Bond and the other in FORTRAN from Netlib
(I think).  The FORTRAN code is the one that's giving trouble,
you might try just removing the FORTRAN code from the src
directory and trying again.

http://www.tek-tips.com/viewthread.cfm?qid=1296207&page=11

  Eventually it would be nice to get both sets of code
running on multiple platforms, for comparison's sake --
the code that's failing (machcon.f) is the code to define
machine-dependent constants -- this is a fairly standard set of
routines from NETLIB, but they're very grungy in that you're
just supposed to uncomment the constants that are relevant
to your system.  There must be a better way to do this
with configure/looking in system libraries etc. (I don't
really know how this is done in FORTRAN).

  Does anyone out there have experience in porting
code from Netlib and ending up with clean, platform-configured
code?  (I will go look at the deSolve package next, but
since this is a bit of a side project ...)

  Ben Bolker
baptiste auguie wrote: