Building CESM1 with Ifort 11.1: Hangs on shr_scam_mod.F90

[plangen@gfy_ku_dk]

Hi,

I'm trying to build CESM1 on a Linux cluster with Red Hat 4.1.2-46 and CentOS release 5.4 using the ifort (and icc) compiler version 11.1.056.

I create a new case with -compset F and -res f19_f19 and run the build script. When it tries to build the csm_share library it goes fine until it gets to shr_scam_mod.F90:

/usr/mpi/intel/openmpi-1.2.8/bin/mpif90 -c -I. -I/users/others/plangen/tools/netcdf_intel11/include -I/users/others/plangen/tools/netcdf_intel11/include -I/usr/mpi/intel/openmpi-1.2.8/include -I. -I/users/others/plangen/models/cesm1_0/scripts/Ftest02/SourceMods/src.share -I/users/others/plangen/models/cesm1_0/models/drv/shr_mct -I/users/others/plangen/models/cesm1_0/models/drv/shr -I/users/others/plangen/models/cesm1_0/models/dead_share -I/users/others/plangen/models/cesm1_0/models/csm_share/shr -I/users/others/plangen/models/cesm1_0/models/utils/timing -I/users/others/plangen/models/cesm1_0/models/utils/esmf_wrf_timemgr -I/users/gpfs1/others/plangen/models/cesm1_0/exe/Ftest02/lib/include -DMCT_INTERFACE -DHAVE_MPI -DCO2A -DLINUX -DSEQ_ -DFORTRANUNDERSCORE -DNO_R16 -DNO_SHR_VMATH -132 -fp-model precise -convert big_endian -assume byterecl -ftz -traceback -O2 /users/others/plangen/models/cesm1_0/models/csm_share/shr/shr_scam_mod.F90

... and then it just hangs there - for an hour nothing happened. Has anyone seen this before? Any suggestions?

Best regards,
Peter

 

[shaochuande2000@yahoo_com_cn]

yeah, i encountered the same problem. Also i used i ifort (and icc) compiler version 11.0.084. you should wait for some minutes，and it is OK.

 

[plangen@gfy_ku_dk]

Thanks for your answer. I did try waiting for about 45 min or an hour before I got impatient and killed it. How long did you wait? Up to an hour doesn't seem like it is going to go through...

Peter

 

[tcraig]

We are aware of this issue. It appears to be a compiler problem.

A short term fix is to edit models/csm_share/shr/shr_scam_mod.F90
and comment out any line that includes declaration or use of the
variable SDAT. This section of code is used only when scam (single
point) mode is enabled and only if the ocean model is docn with
scam turned on. The changes should look something like

line 280:
! type(shr_strdata_type) :: SDAT

line 349+:
call shr_sys_abort('shr_scam_checkSurface abort for SDAT')
! call shr_strdata_readnml(SDAT,ocn_in)
! call shr_dmodel_readgrid(SDAT%grid,SDAT%gsmap,SDAT%nxg,SDAT%nyg, &
! SDAT%domainfile, ocn_compid, ocn_mpicom, '1d', readfrac=.true., &
! scmmode=.true.,scmlon=scmlon,scmlat=scmlat)
! nfrac = mct_aVect_indexRA(SDAT%grid%data,'frac')
! if ( present(ocn_present)) ocn_present=(SDAT%grid%data%rAttr(nfrac,1).gt.0.)
! if ( present(ocn_present).and.present(ice_present)) ice_present=ocn_present
! if ( present(lnd_present)) lnd_present=(SDAT%grid%data%rAttr(nfrac,1).lt.1.)

If anyone continues to have problems, please email cesm-help at cgd dot ucar dot edu.