Re: wrapping fortran intrinsics

From: Mary Haley (haley AT XXXXXX)
Date: Thu Apr 04 2002 - 10:40:41 MST


> .
> .
> .
> To have access to erfc in NCL all I needed to do was run WRAPIT on erfc.f
> (the simple little program given in the comments at the beginning of the
> function), then load erf.ncl into my NCL script. "y = erfc(x)" works on
> scalars and arrays of any dimension. Kudos to the NCL team - what a great
> tool!

Hi Jeff,

Thanks for sharing your comments. We always appreciate them.

> BTW: What's the reason that NCAR doesn't open the NCL source? After all,
> the binaries are given away for free.
> .
> .
> .

We plan to release it someday, but another reason we are not releasing
it right now is because it is a difficult package to build. At this
point, it would probably create a lot of grief for users trying to
build it. The difficulty partially stems from the number of external
packages that are required (netCDF, HDF, Spherepack, LAPACK, BLAS,
flex). You need to build all of these before you can build NCL. Also,
the fact that you can import shared objects into NCL adds another
level of difficulty in compilation. You have to understand how shared
objects work, and built the "ncl" executable with the appropriate
options.

As you pointed out, the shared object capabilities makes NCL
extensible without having to rebuild it from source. We hope this
gives users enough flexibility that they don't need the NCL source
code.

We always appreciate user input. If other users are longing for NCL
source code, don't hesitate to let us know (via ncl-talk). This may
push the issue through faster.

Cheers,

--Mary Haley
  haley@ucar.edu



This archive was generated by hypermail 2b29 : Thu Apr 04 2002 - 11:04:10 MST