Thanks Jeff..
a quick strings on liblam shows that they are there... so I
changed the Makefile to CC = $(MIPDIR)/bin/mpicc and
F90 = $(MPIDIR)/bin/mpif77 and it compiled and linked!.... Thanks for the
suggestions.
Pete.
On Thu, 9 Sep 2004, Jeff Squyres wrote:
> On Sep 9, 2004, at 1:04 PM, Peter Schmid wrote:
>
> > # MPICH LINE
> > #MPILIB = -L$(MPIDIR)/lib -lmpichf90 -lmpich -lPEPCF90
> > # LAM LINE
> > MPILIB = -L$(MPIDIR)/lib -llamf77mpi -llam -lmpi -llammpi++
> > -llammpio
> >
> > Anyone there familar with MPICH vs LAM.. and can give me some pointers
> > to
> > pass onto my developers?
>
>
> This style of building MPI apps is (IMHO) an unfortunate throwback to
> the mid-90's when the MPI wrapper compilers were not nearly as
> reliable. These days, there are very, very few reasons to list the MPI
> libraries and whatnot directly in your makefile.
>
> 9999 times out of 10000, you can just use LAM's "mpicc", "mpiCC" (or
> "mpic++" if you're on a case-insensitive filesystem) or "mpif77"
> wrapper compilers and they take care of all of this magic for you. So
> I would leave MPILIB blank (and any other Make macros that specify
> CFLAGS, LDFLAGS, or anything like that).
>
> That *might* solve your problem. I have to admit that I'm not quite
> sure how you'd end up with a LAM library (liblam, to be specific) that
> doesn't have lam_strncpy or argvfree... Those symbols should be in
> liblam regardless of your platform, etc. Hence, my best guess at this
> point is that you're linking to the "wrong" liblam somehow, or your
> liblam is somehow missing some symbols. Do you/your developers have
> the output from configure and make when building LAM?
>
>
--
Peter Schmid
Technical Director
Logic Technology Inc.
GE Global Research Center
ITMS Engineering Systems Group
(518) 387-6903
|
