Main menu

Navigation

SE dycore initiliazation / crash on IBM Power6

1 post / 0 new
andreasb
SE dycore initiliazation / crash on IBM Power6

Dear all,

I am using CESM1 (v1.2.0) on blizzard.dkrz.de (http://www.dkrz.de/Klimarechner-en/hpc?set_language=en), an IBM Power6 machine.

I ported the relevant machine files, and a finite volume run works fine:

./create_newcase -res 4x5_4x5     -case F2000CN.4x5 -compset F_2000_CN -mach blizzard -pecount cores4

However,

./create_newcase -res ne16np4_ne16np4 -case F2000CN.ne16np4 -compset F_2000_CN -mach blizzard -pecount cores4

crashes during initialization:

end of atm logfile output:

1: Opened existing file /scratch/b/b380029/DATA/CESM1/atm/cam/inic/homme/cami_0000-01-01_ne16np4_L26_c120525.nc 62
1: Opened existing file /scratch/b/b380029/DATA/CESM1/atm/cam/topo/USGS-gtopo30_1.9x2.5_smooth500-50_ne16np4_c050602.nc 63
0:

Corefiles are produced, and dbx sais:

Segmentation fault in mpi_init.REG_3stream_store at 0x900000002f8f5c4
0x900000002f8f5c4 (REG_3stream_store+0x1e4) c8160008         lfd   fr0,0x8(r22)

Note that the same setup works fine on my own machine.

If I compile with qflttrap option, the crash occurs in prim_driver_mod:

Trace/BPT trap in Fprim_driver_mod.__prim_driver_mod_NMOD_prim_init2 at line 614 in Fprim_driver_mod.f"
  614                            elem(ie)%state%Qdp(i,j,k,q,n0_qdp)=elem(ie)%state%Q(i,j,k,q)*dp 

Has anyone got an idea what I'm doing wrong, or how to fix that? Thanks very much!

Andreas

Some more details:

-The bluefire setup files were used as starting point.

-Compiler versions: IBM/xlf13.1.0.8 and IBM/xlC11.1.0.8

-logfiles are attached

Attachment: 

Who's new

  • 1658093099@...
  • mborreggine@...
  • kabirtam@...
  • suns@...
  • liangpeng0405@...