Re: [AMBER] problem (re)compiling pmemd.cuda_SPFP: Reading module netcdf at line 23 column 34: Expected integer

From: Jason Swails <jason.swails.gmail.com>
Date: Fri, 12 Jul 2013 10:21:14 -0400

On Fri, Jul 12, 2013 at 9:53 AM, Marc van der Kamp <marcvanderkamp.gmail.com
> wrote:

> Many thanks for the quick response (as always!)
> Removing $AMBERHOME/include/netcdf.mod (manually) indeed did the trick.
>
> BTW, Jason's suggestion that would allow ./configure bypass the NetCDF
> configure altogether if the existing NetCDF 'works', sounds great.
> My incompatible netcdf.mod was indeed likely due to using different
> compilers previously, but I think there may be good reasons to use
> different compilers for different executables, e.g. sander.MPI and
> pmemd.cuda.
>

This is quite difficult to support from a developer's perspective, since it
adds many degrees of freedom that will in all likelihood never get tested
pre-release. (Of course I've done the same thing myself on at least one
supercomputer...)



While I don't propose purposely trying to disallow multiple compilers for
the same installation, this should be considered an expert option. Better
solutions IMO are to 1) maintain separate Amber installations and use
something like the 'module' program manage the different installations (you
can write your own module files on clusters or install it on your own
machine:
http://modules.sourceforge.net/
) or 2) Just use the GNU compiler for everything.

All the best,
Jason

-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jul 12 2013 - 07:30:02 PDT
Custom Search