incorrect yz domain decomposition

[moguo_sun@nasa_gov]

Dear Helper,       I configure  $camcfg/configure -fc mpif90 -fc_type intel -cc mpicc -mpi_inc $MPI_INC -mpi_lib $MPI_LIB  -dyn fv -hgrid 1.9x2.5 -ntasks 8 -nthreads 16 -testand build like$camcfg/build-namelist -ntasks 128 -config ../bld/config_cache.xml
but the model generates error like this:Read in dyn_fv_inparm namelist from: atm_in
 Read in spmd_fv_inparm namelist from: atm_in
 npr_y =           32   npr_z =            4
 nprxy_x =            4   nprxy_y =           32
 npes =            8   npes_yz=          128   npes_xy =          128
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
(shr_sys_abort) ERROR: SPMD_DYN_SET : incorrect yz domain decomposition - aborting
(shr_sys_abort) WARNING: calling shr_mpi_abort() and stopping
----------and more debug info in error file:---MPI: 0x00002aaaabfae05e in waitpid () from /lib64/libpthread.so.0
MPI: (gdb) #0  0x00002aaaabfae05e in waitpid () from /lib64/libpthread.so.0
MPI: #1  0x00002aaaaaf7f39c in mpi_sgi_system (header=) at sig.c:89
MPI: #2  MPI_SGI_stacktraceback (header=) at sig.c:272
MPI: #3  0x00002aaaaaf0ac2b in print_traceback (ecode=1001) at abort.c:168
MPI: #4  0x00002aaaaaf0aeda in PMPI_Abort (comm=, errorcode=1001) at abort.c:59
MPI: #5  0x00002aaaaaf0af5d in pmpi_abort__ () from /opt/sgi/mpt/mpt-2.06/lib/libmpi.so
MPI: #6  0x000000000281fd79 in shr_mpi_mod_mp_shr_mpi_abort_ ()
MPI: #7  0x00000000028725e3 in shr_sys_mod_mp_shr_sys_abort_ ()
MPI: #8  0x000000000288bbf8 in spmd_dyn_mp_spmd_readnl_ ()
MPI: #9  0x0000000000993f80 in dyn_comp_mp_dyn_init_ ()
MPI: #10 0x0000000000cc5e08 in inital_mp_cam_initial_ ()
MPI: #11 0x00000000006ce34c in cam_comp_mp_cam_init_ ()
MPI: #12 0x0000000000664231 in atm_comp_mct_mp_atm_init_mct_ ()
MPI: #13 0x000000000074ae8c in ccsm_comp_mod_mp_ccsm_init_ ()
MPI: #14 0x000000000075164d in MAIN__ ()
MPI: #15 0x000000000040e38c in main ()
MPI: (gdb) A debugging session is active.
-------

    try to fix that problem, but could not find much help.  Appreciate if you could help. thanks.   

 

[eaton]

The -ntasks arg to build-namelist cannot be larger than the number of MPI tasks being used.  The value of -ntasks is used by build-namelist to set the variable npr_yz which determines the domain decomposition for the FV dycore.  The dycore requires the number of MPI tasks to be at least as large as the number of subdomains.  You've told build-namelist there are 128 tasks.  It translates that to a setting for npr_yz which implies 128 subdomains.  But at runtime you've only supplied 8 tasks, and so the runtime check in the FV dycore is failing.