[MITgcm-devel] linux_ia64_ifort+mpi_altix_gfdl

[Constantinos Evangelinos]

On Tuesday 05 December 2006 16:19, Baylor Fox-Kemper wrote:

> Good.  That was a bit of a crude fix in our version.  Thanks,
>     -Baylor
>

Actually this fix still includes redundant include paths (unless someone uses 
SCSL with MITgcm that is). Please try the following and let me know how it 
works:

#!/bin/bash
#
#  Ed Hill
#
#  tested on orion.jpl.nasa.gov (Linux orion 2.4.21-sgi230r7 #1 SMP Tue Oct 7 
22:38:53 PDT 2003 ia64 unknown), 20031029
#
# adapted to GFDL's Itanium cluster
#  heimbach at mit.edu, baylor at mit.edu and ce107 at mit.edu
#
#  Will need to do module_load ifort.9.1.036 icc.9.1.044 mpt-1.14 first

CPP='/lib/cpp  -traditional -P'

# the '-convert big_endian' is crucial on this platform
# -D_BYTESWAPIO doesn't work
FFLAGS='-mp -132 -r8 -i4 -w95 -W0 -WB -convert big_endian'

FOPTIM='-O3 -align -ip'
FC=ifort
CC=icc
DEFINES='-DALLOW_USE_MPI -DALWAYS_USE_MPI -DWORDLENGTH=4'
# These paths will likely change to newer versions of ifort
MPIINCLUDEDIR=`echo $FPATH | awk -F: '{for (i=1; i<=NF; i++) {if ($i ~ /mpt/  
{printf "%s\n",$i}}}'`
MPI_HEADER_FILES='mpif.h mpiof.h mpif_parameters.h'
MPI_HEADER_FILES_INC='./mpi_headers/mpif.h ./mpi_headers/mpiof.h ./mpi_headers/mpif_parameters.h'
INCLUDES="-I/usr/local/include  -I$MPIINCLUDEDIR"
INCLUDIRS="/usr/local/include $MPIINCLUDEDIR"
LIBS='-L/usr/local/lib -lmpi'

> On Dec 5, 2006, at 4:14 PM, Alistair Adcroft wrote:
> > Baylor,
> >
> > I've changed the lines
> >
> > INCLUDES="-I/usr/local/include -I$CPATH"
> > INCLUDEDIRS="/usr/local/include $CPATH"
> >
> > to
> >
> > cpath=`echo $CPATH | sed 's/:/ -I/g'`
> > INCLUDES="-I/usr/local/include -I$cpath"
> > cpath=`echo $CPATH | sed 's/:/ /g'`
> > INCLUDEDIRS="/usr/local/include $cpath"
> >
> > so that they avoid errors like this:
> >
> >  Warning: can't find INCLUDEDIRS="/opt/mpt/1.14/include:/opt/scsl/
> > 1.5.1.0/include"
> >
> > I'll check it into linux_ia64_ifort+mpi_altix_gfdl, OK?
> >
> > A.
> >
> > Alistair Adcroft wrote:
> >> Any objections if I update "tools/build_options/linux_ia64_ifort
> >> +mpi_altix_gfdl" to work with a 9.1 ifort and mpt-1.14 .
> >> Jobs with MPT<1.14 are no longer allowed in the queues (they
> >> uniformaly replace all MPT versions with 1.14!!!) and ifort >
> >> 9.1.033 runs substantially faster on this platform.
> >>
> >> The only catch is that GFDL uses modules and you *must* load the
> >> corresponding modules into your environment to compile and run
> >> properly. In my .cshrc I have:
> >>
> >>  if (`uname -m` == "ia64") then
> >>   module purge
> >>   module load icc.9.1.042
> >>   module load ifort.9.1.033
> >>   module load mpt-1.14
> >>  endif
> >>
> >> A.
> >
> > --
> > Dr Alistair Adcroft              email: Alistair.Adcroft at noaa.gov
> > NOAA/GFDL, Princeton University               Tel: (609) 987-5073
> > Forrestal Campus, U.S. Route 1, P.O. Box 308, Princeton, NJ 08542
> >
> > _______________________________________________
> > MITgcm-devel mailing list
> > MITgcm-devel at mitgcm.org
> > http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

-- 
Dr. Constantinos Evangelinos
Department of Earth, Atmospheric and Planetary Sciences
Massachusetts Institute of Technology

-- 
Dr. Constantinos Evangelinos                    Room 54-1518, EAPS/MIT
Earth, Atmospheric and Planetary Sciences       77 Massachusetts Avenue
Massachusetts Institute of Technology           Cambridge, MA 02139
+1-617-253-5259/+1-617-253-4464 (fax)           USA