[MITgcm-support] linux_amd64_ifort+mpi_ice_nas
Holly Dail
hdail at MIT.EDU
Thu Nov 11 13:57:00 EST 2010
I've resolved this problem with the opt file ... here's what I learned:
- the new fPIC / shared-intel approach in linux_amd64_ifort+mpi_ice_nas is a more flexible way to allow compilation of large binaries than the old -mcmodel medium approach. The two are incompatible, so the last three lines of the opt file that tell the user to use -mcmodel medium are incorrect. (Should be deleted?)
- With shared libraries the appropriate modules must be loaded, but I had an error in my submission script that tried to load non-existent comp/intel/11.1.046 instead of comp-intel/11.1.046. Oops.
Holly
On Nov 10, 2010, at Nov 10 , 9:52 AM, Holly Dail wrote:
> Hello -
>
> The recent changes to linux_amd64_ifort+mpi_ice_nas have caused problems for me when running on pleiades, and I thought I'd mention them in case they affect others. Perhaps there are things I need to change in my setup to use the new optfile, but if so, I'm not sure what. For the moment I'm running fine with the checkpoint 62j version.
>
> Here is the error I get with the new optfile:
>
> MPI: r87i3n14: 0x585200004c9a426e: ./mitgcmuv: error while loading shared librar
> ies: libintlc.so.5: cannot open shared object file: No such file or directory
> MPI: r91i2n10: 0x585200004c9a426e: ./mitgcmuv: error while loading shared librar
> ies: libintlc.so.5: cannot open shared object file: No such file or directory
> MPI: could not run executable (case #4)
>
> Here is the diff on the two files ... not sure if its the fPIC or the -shared-intel that is the problem.
>
> < # $Header: /u/gcmpack/MITgcm/tools/build_options/linux_amd64_ifort+mpi_ice_nas,v 1.4 2010/07/30 06:21:33 dimitri Exp $
> < # $Name: checkpoint62j $
> ---
>> # $Header: /u/gcmpack/MITgcm/tools/build_options/linux_amd64_ifort+mpi_ice_nas,v 1.6 2010/10/26 19:32:04 heimbach Exp $
>> # $Name: $
> 17a18,19
>> CFLAGS='-fPIC'
>> LDADD='-shared-intel'
> 26c28
> < NOOPTFLAGS='-O0'
> ---
>> NOOPTFLAGS='-O0 -fPIC'
> 30c32
> < FFLAGS="$FFLAGS -w95 -W0 -WB -convert big_endian -assume byterecl -align"
> ---
>> FFLAGS="$FFLAGS -fPIC -w95 -W0 -WB -convert big_endian -assume byterecl -align"
> 32c34
> < FFLAGS="$FFLAGS -w95 -W0 -WB -convert big_endian -assume byterecl -noalign"
> ---
>> FFLAGS="$FFLAGS -fPIC -w95 -W0 -WB -convert big_endian -assume byterecl -noalign"
>
> Cheers -
> Holly
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
More information about the MITgcm-support
mailing list