Re: [AMBER-Developers] mpif77 & Fortran 90

From: Lachele Foley <lfoley.ccrc.uga.edu>
Date: Wed, 24 Feb 2010 15:45:08 -0500

> I just think it's possible that a symbolic link that
> makes gfortran pretend to be g77 may have unanticipated
> consequences... but maybe not.

My system at home installed that way by default, which makes me think maybe not. Also, the install of the updated gcc/gfortran, in which I installed everything, didn't include a g77 at all, which implies that gfortran is supposed to be g77. But, neither did that install set a symlink, which implies that the gcc folks might not have been comfortable doing that for some reason -- maybe simply to emphasize that g77 is no longer needed.

However, this is ia64, which isn't quite like other systems. I'll try rebuilding the mpi to be safe. It hadn't occurred to me that something like this was even possible when I first compiled it.


:-) Lachele
--
B. Lachele Foley, PhD '92,'02
Assistant Research Scientist
Complex Carbohydrate Research Center, UGA
706-542-0263
lfoley.ccrc.uga.edu
----- Original Message -----
From: Jason Swails
[mailto:jason.swails.gmail.com]
To: AMBER Developers Mailing List
[mailto:amber-developers.ambermd.org]
Sent: Wed, 24 Feb 2010 15:29:15
-0500
Subject: Re: [AMBER-Developers] mpif77 & Fortran 90
> On Wed, Feb 24, 2010 at 3:19 PM, Lachele Foley <lfoley.ccrc.uga.edu> wrote:
> > I see.  I think I can fix this with a symlink.
> >
> > I had to install an updated gfortran a while back to get certain bits of
> amber to install at all, so I can't just reset my path or everything else
> will break.  And...  see below.
> >
> > First off, mpif77 -showme is actually mpif77 -show on my system.  Because
> it can't ever be easy... Sigh.
> >
> > byron:installer>mpif77 -show
> > g77 -L/opt/gm/lib64 -L/opt/gm/lib
> -I/scratch2/programs/mpich2-1.2.1/include
> -L/scratch2/programs/mpich2-1.2.1/lib -L/scratch2/programs/mpich2-1.2.1/lib
> -lmpich -lopa -lpthread -lrt
> >
> > byron:installer>mpif90 -show
> > gfortran -L/opt/gm/lib64 -L/opt/gm/lib
> -I/scratch2/programs/mpich2-1.2.1/include
> -I/scratch2/programs/mpich2-1.2.1/include
> -L/scratch2/programs/mpich2-1.2.1/lib -L/scratch2/programs/mpich2-1.2.1/lib
> -lmpichf90 -lmpichf90 -lmpich -lopa -lpthread -lrt
> >
> > However, and this is important:
> >
> > byron:installer>which g77
> > /usr/bin/g77
> > byron:installer>which gfortran
> > /usr/local/bin/gfortran
> > byron:installer>ls /usr/bin/gfortran
> > /usr/bin/gfortran
> > byron:installer>ls /usr/local/bin/g77
> > ls: /usr/local/bin/g77: No such file or directory
> >
> > So, I'll retry with a normal config file and a symlink from
> /usr/local/bin/g77 to /usr/local/bin/gfortran.
> >
> > At home, that's the default:
> >
> > lf.macha ~> ls -l /usr/bin/g77
> > lrwxrwxrwx 1 root root 8 2009-06-08 07:45 /usr/bin/g77 -> gfortran
> 
> Is there a reason you can't rebuild your MPI?  I was running into a
> similar problem earlier trying to configure mpich2 with the intel
> compilers.  Just running ./configure; make; make install caused the
> configure script to pick ifort for the f90 compiler, but it picked
> gfortran as the f77 compiler since ifort is technically just f90 (so
> mpif77 failed for amber since it was using the intel flags :)).  I got
> around this by setting env. variables F77=ifort, F90=ifort, FC=ifort,
> CC=icc (to cover all my bases, FC was probably unnecessary).  Is this
> unnecessary?  I just think it's possible that a symbolic link that
> makes gfortran pretend to be g77 may have unanticipated
> consequences... but maybe not.
> 
> >
> > Thanks!
> >
> > :-) Lachele
> > --
> > B. Lachele Foley, PhD '92,'02
> > Assistant Research Scientist
> > Complex Carbohydrate Research Center, UGA
> > 706-542-0263
> > lfoley.ccrc.uga.edu
> >
> >
> > ----- Original Message -----
> > From: case
> > [mailto:case.biomaps.rutgers.edu]
> > To: AMBER Developers Mailing List
> > [mailto:amber-developers.ambermd.org]
> > Sent: Wed, 24 Feb 2010 14:45:24
> > -0500
> > Subject: Re: [AMBER-Developers] mpif77 & Fortran 90
> >
> >
> >> On Wed, Feb 24, 2010, Lachele Foley wrote:
> >>
> >> > Trying to make parallel.  Get many complaints like below.  Fixes, so
> >> > far anyway, if I change config.h auto-generated FC=mpif77 to FC=mpif90.
> >> > (ia64 if relevant) Not sure where to fix or if it will cause trouble
> >> > elsewhere.
> >>
> >> Which version of mpi?  What does "mpif77 -showme" report?  What about
> >> "mpif90 -showme"?  How did you configure the MPI?
> >>
> >> Here's the deal.  There are two things that might require f90 support:
> >> One is just to compile ordinary (non-mpi) code; somehow you seem to have
> >> an mpif77 that doesn't call an f90 compiler; for me, I get something like
> >> this:
> >>
> >> poincare% mpif77 -showme
> >> gfortran -I/home/case/amber11/include -L/home/case/amber11/lib -lmpi_f77
> >> -lmpi
> >> -lopen-rte -lopen-pal -lutil
> >>
> >> which is fine.
> >>
> >> The second aspect is whether or not you need f90 bindings to MPI itself.
> >> Amber doesn't use these, and we can set things like "--disable-mpi-f90"
> when
> >> configuring openmpi.  Thus, I get things like this:
> >>
> >> poincare% mpif90 -showme
> >>
> --------------------------------------------------------------------------
> >> Unfortunately, this installation of Open MPI was not compiled with
> >> Fortran 90 support.  As such, the mpif90 compiler is non-functional.
> >>
> --------------------------------------------------------------------------
> >>
> >>
> >> So, we can't just replace mpif77 with mpif90 in config.h, because that
> would
> >> make things fail for most folk.  We(you) need to figure out what
> compiler
> >> your mpif77 is calling, and why.  Then we need to update the
> installation
> >> instructions to tell people how to avoid this problem.
> >>
> >> ...dac
> >>
> >>
> >> _______________________________________________
> >> AMBER-Developers mailing list
> >> AMBER-Developers.ambermd.org
> >> http://lists.ambermd.org/mailman/listinfo/amber-developers
> >>
> >
> > _______________________________________________
> > AMBER-Developers mailing list
> > AMBER-Developers.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber-developers
> >
> 
> 
> 
> -- 
> ---------------------------------------
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> 
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
> 
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Feb 24 2010 - 13:00:02 PST
Custom Search