On 12/18/2014 07:36 AM, Jeff Hammond wrote:

On Wed, Dec 17, 2014 at 2:00 PM, Kenneth Raffenetti <[email protected]> wrote:

...

We've got an open bug related to using the Intel Fortran compiler on OSX.

http://trac.mpich.org/projects/mpich/ticket/2024

What's missing from that bug description is that it appears that the convenience libraries (mpl, romio, etc.) are not properly embedded when linking with ifort on OSX. Your error output below is consistent with this issue. Is it possible to disable fortran in your MPICH build in this instance? That would at least confirm the source of the problem.

I don't know that you understand what you're asking. Asking a quantum chemist to disable Fortran is like asking Bret Michaels to disable bandanas and groupies with low morals!

Anyways, I can do this, and I even have a special hook so that I can compile NWChem on Mac without MPI Fortran support because I often use MPI built with LLVM.

It is rather obnoxious that a Fortran issue would break the C/C++ wrappers though.

Priority on this bug dropped after we had trouble renewing our Intel licenses for OSX. I'll poke at the source to see if we can get them active again. Ken