> I'll pitch in with something you can try, but no guarantee it will work.
> Because no_underscoring is a flag built into mpif90, you did NOT have to add
> it to FFLAGS or FOPTFLAGS, since it was part of the compiler. Also,
> according to Professor Case, this does bad things for amber due to the
> Fortran/C 'hybridization' of sorts. Therefore, my suspicion is that the
> presence of no_underscoring in mpif90 is what was causing your errors in the
> first place. What you can probably do is use your MPICH2 build that has the
> no_underscoring in place, but rather than using "mpif90" as your fortran
> compiler, use what is returned by "mpif90 -show" without the no_underscoring
> flag. It's a hack, but probably the only way (if it works) to avoid
> installing 2 MPICH2's.
I didn't know it's built into the compiler, then I don't know why it's still needed when I compile the other program. I think it's better to just install 2 MPICH2s.
> The other option, of course, is to install 2 MPICH2's, just keep them
> separate so you don't get any stepping on toes. mpi-selector is a nifty
> little tool that helps organize multiple MPIs on a single cluster or
> machine. It's what I have running on my machine so I can keep MPICH2 and
> OpenMPI running for both the intel and gnu compilers easily. It takes a
> little bit of setup, but it's pretty convenient when you have it up and
> working.
The mpi-selector sounds a very good idea. But when I googled "mpi-selector", I only found RPM packages. Do you know if there is a OS X package? I didn't even find the source code to compile. I will do more search to see if I can find something.
Thanks!
Chen
>
> Good luck!
> Jason
>
> On Thu, Jul 1, 2010 at 5:58 PM, case <case.biomaps.rutgers.edu> wrote:
>
>> On Thu, Jul 01, 2010, Yu Chen wrote:
>>>
>>> I added "-fno-underscoring" at the end of "FFLAGS .." and "FOPTFLAGS ..."
>> lines.
>>
>> I don't know where the idea of "no_underscoring" came from, but I'm pretty
>> sure it will do bad things to Amber. Amber has both Fortran and C code,
>> and
>> assumes in many places that the fortran compiler will add an extra
>> underscore,
>> as is the usual practice for Unix/Linux compilers. If you turn this off,
>> you should be prepared for lots of grunt work changing the code. I would
>> strongly recommend not trying to use this flag.
>>
>>>
>>>> I still recommend using the configure_openmpi script. There should be
>> no
>>>> need to manually edit the config.h file.
>>
>> Above I guess is a dead horse, I guess. I've attached a "configure_mpich2"
>> script which should do the same thing for MPICH2 that configure_openmpi
>> does
>> for OpenMPI: provide an installation of MPI where you don't have to add or
>> subtract flags, or edit config.h files, and which should work under most
>> circumstances.
>>
>> ....dac
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jul 02 2010 - 07:00:05 PDT
