Scheduled Downtime
On Tuesday 24 October 2023 @ 5pm MT the forums will be in read only mode in preparation for the downtime. On Wednesday 25 October 2023 @ 5am MT, this website will be down for maintenance and expected to return online later in the morning.
Normal Operations
The forums are back online with normal operations. If you notice any issues or errors related to the forums, please reach out to help@ucar.edu

porting error when building cesm.exe

L

littledddna@gmail_com

Member
Hi, all,
I am still trying to porting CESM2.1.3 to a new cluster. But I met some problems when building the executables.
The building of individual components did not error but when it is building the executable, I got errors in timing/perf_mod.F90 related to gptl functions (see attached txt file), e.g. gptlget_memusage, gptlsetoption, gptlstart and so on.

Any clues how do I over come such issue? Thanks very much
 

Attachments

  • cesm-build-log.txt
    11.3 KB · Views: 11
J

jedwards

CSEG and Liaisons
Staff member
It looks like gptl did not build correctly, check the gptl.bldlog for more information. This error:
relocation truncated to fit: R_X86_64_PC32 against symbol `seq_comm_mct_mp_gloid_' defined in COMMON section in ../../intel/mpich/debug/nothreads/

is also a problem - it indicates that something is too large for the memory model. What case and resolution are you trying to build and what versions of compilers and netcdf library are you using?
 
L

littledddna@gmail_com

Member
Thanks very much for the reply, jedwards. I checked gptl log file and did not find any error (see attached), so I dont know where do these error come from. And I am trying to build resolution f19_g17 for compset B1850_BPRP. I am using intelcompiler/15.0.1 and netcdf/4.5.0-icc-15.0.1-parallel.
 
L

littledddna@gmail_com

Member
jedwards said:
It looks like gptl did not build correctly, check the gptl.bldlog for more information. This error:
relocation truncated to fit: R_X86_64_PC32 against symbol `seq_comm_mct_mp_gloid_' defined in COMMON section in ../../intel/mpich/debug/nothreads/

is also a problem - it indicates that something is too large for the memory model. What case and resolution are you trying to build and what versions of compilers and netcdf library are you using?
Click to expand...

Thanks very much for the reply, jedwards. I checked gptl log file and did not find any error (see attached), so I dont know where do these error come from. And I am trying to build resolution f19_g17 for compset B1850_BPRP. I am using intelcompiler/15.0.1 and netcdf/4.5.0-icc-15.0.1-parallel.
 

Attachments

  • gptl-log.txt
    3.2 KB · Views: 5
J

jedwards

CSEG and Liaisons
Staff member
Intel 15.0.1 is quite old - do you have access to a newer compiler version? The relocation truncated to fit: R_X86_64_PC32 can cause all sorts of other problems, I think we need to figure that out and maybe then the gptl issue will just go away. Please try with a simple compset such as X or A and see if you get the same error.
 
L

littledddna@gmail_com

Member
jedwards said:
Intel 15.0.1 is quite old - do you have access to a newer compiler version? The relocation truncated to fit: R_X86_64_PC32 can cause all sorts of other problems, I think we need to figure that out and maybe then the gptl issue will just go away. Please try with a simple compset such as X or A and see if you get the same error.
Click to expand...
Thanks very much, I will try today :)
 
L

littledddna@gmail_com

Member
jedwards said:
Intel 15.0.1 is quite old - do you have access to a newer compiler version? The relocation truncated to fit: R_X86_64_PC32 can cause all sorts of other problems, I think we need to figure that out and maybe then the gptl issue will just go away. Please try with a simple compset such as X or A and see if you get the same error.
Click to expand...

Hi, jedwards,

I have upgraded from intel15 to intel18 and did a X composet test case. The R_X86_64_PC32 error disappeared but the gptl related errors still exit. Any suggestions where should I go? Thanks very much
 
J

jedwards

CSEG and Liaisons
Staff member
Okay - I think that you need to adjust the GPTL flags in config_compilers.xml try adding -DFORTRANUNDERSCORE
 
L

littledddna@gmail_com

Member
jedwards said:
Okay - I think that you need to adjust the GPTL flags in config_compilers.xml try adding -DFORTRANUNDERSCORE
Click to expand...
Thanks very much, jedwards,
Finally I got the cesm.exe. However, I also got the following error

Code: 
Building cesm with output to /BIGDATA1/nuist_xzhang_1/cesm_output/simple/bld/cesm.bldlog.200624-222703
ERROR: Error gathering provenance information from manage_externals.

manage_externals error message:
    ERROR:root:SVN returned invalid XML message

manage_externals output:
    Processing externals description file : Externals.cfg
    Processing externals description file : Externals_CLM.cfg
    Processing externals description file : Externals_POP.cfg
    Processing externals description file : Externals_CISM.cfg
    Processing externals description file : Externals_CAM.cfg
    Checking status of externals: clm, fates, ptclm, mosart, ww3, cime, cice, pop, cvmix, marbl, cism, source_cism, rtm, cam, clubb, carma,
    ERROR: SVN returned invalid XML message

To solve this, either:

(1) Find and fix the problem: From /BIGDATA1/nuist_xzhang_1/CESM, try to get this command to work:
    manage_externals/checkout_externals --status --verbose --no-logging

(2) If you don't need provenance information, rebuild with  provenance information

I think it is because the cluster i am using is not able to connect to hostname `svn-ccsm-models.cgd.ucar.edu' even using proxy. So after I ran ./manage_externals/checkout_externals --status --verbose --no-logging
I got ERROR: SVN returned invalid XML message.

Is provenance information critical ? Otherwise, I think I will use provenance information option
 
L

littledddna@gmail_com

Member
Because when I change the PE layout in B1850 compset (Macro.make is the same as the sucesfully build X compset), the model still gets the following error.

Code: 
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:224:(.text+0x70ec): relocation truncated to fit: R_X86_64_PC32 against symbol `grid_mp_rcalct_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(grid.o)
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:223:(.text+0x7152): relocation truncated to fit: R_X86_64_PC32 against symbol `forcing_shf_mp_shf_qsw_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(forcing_shf.o)
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:231:(.text+0xa85e): relocation truncated to fit: R_X86_64_PC32 against symbol `grid_mp_rcalct_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(grid.o)
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:231:(.text+0xa8b5): relocation truncated to fit: R_X86_64_PC32 against symbol `forcing_fields_mp_ifrac_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(forcing_fields.o)
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:235:(.text+0xbad2): relocation truncated to fit: R_X86_64_PC32 against symbol `grid_mp_rcalct_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(grid.o)
/BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/ocn/source/ocn_import_export.F90:235:(.text+0xbb29): relocation truncated to fit: R_X86_64_PC32 against symbol `forcing_fields_mp_atm_press_' defined in LARGE_COMMON section in /BIGDATA1/nuist_xzhang_1/cesm_output/new_low_1850/bld/lib//libocn.a(forcing_fields.o)
 
J

jedwards

CSEG and Liaisons
Staff member
I think that the issue is that the pelayout is too small for the memory model - try increasing the NTASKS_OCN to see if this error goes away.
Make sure that you have variables
MAX_MPITASKS_PER_NODE
MAX_TASKS_PER_NODE
set correctly for your system.
 
You must log in or register to reply here.
Top