SETTE with CRAY compiler

[andmirek]

New Cray compilers have a problem with reading namelists as internal files - functionality implemented in the trunk. The solution proposed in ticket #2653 works, but requires changes to the namelist.

This ticket implements the changes for MetOffice? configuration using method proposed in #2653.

Branch SETTE: svn+ssh://andmirek@forge.ipsl.jussieu.fr/ipsl/forge/projets/nemo/svn/utils/CI/sette_r14244_crayftn_877

Branch ARCH: svn+ssh://andmirek@forge.ipsl.jussieu.fr/ipsl/forge/projets/nemo/svn/utils/build/r14691_arch_crayftn_877

[andmirek]

In 14691:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[andmirek]

In 14692:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[andmirek]

In 14693:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

The lines in the code (after inserting cpp macros) are longer than 255 characters.

[andmirek]

In 14694:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

NEMO can't be build in parallel with the new compiler - set -j 1 instead -j 8.

Add adding comma in the batch script, before running the test.

[mcastril]

Hi, I just wanted to remind this issue that we are having reading internal namelists too, this time with the PGI compiler: #2653

[andmirek]

Putting all together:

svn co svn+ssh://andmirek@forge.ipsl.jussieu.fr/ipsl/forge/projets/nemo/svn/NEMO/trunk
cd trunk
svn propedit svn:externals .

a) replace arch@HEAD with r14691_arch_crayftn_877@14707
b) replace sette@14244 with sette_r14244_crayftn_877@14710

svn update

cd sette

#common
module use /data/d00/moci/modules/modules
module load GC4-PrgEnv/2020-06a-cray877/3616

export SETTE_COMPILER=XC40_METO
module load subversion-1.8/1.8.18

export XIOS_OASIS=1

export SETTE_BATCH_CMD=qsub
export SETTE_BATCH_STAT=qstat

export SETTE_FORCING_DIR=/data/d02/mandrej/SETTE_FORCING_TRUNK_2020
export SETTE_MPMD='no'

#user specific
export SETTE_NEMO_VALIDATION_DIR=/home/d02/mandrej/nemo_validation

./sette.sh

[andmirek]

In 14707:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

Instruction was updated because with this change there is no need to have a branch for sette. Advice by Andrew Coward (and Simon) to add line:

bld::tool::fc_modsearch -J

in architecture file enables parallel build with crayftn.

[andmirek]

In 14708:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[andmirek]

In 14709:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[andmirek]

In 14710:

Error: Failed to load processor CommitTicketReference

No macro or processor named 'CommitTicketReference' found

[andmirek]

Tests with AGRIF failed to build with error:

agrifpp.sh: line 67: 17327 Illegal instruction     ${MYDIR}/${NEW_CONF}/NEMOFILES/conv ${MYDIR}/${NEW_CONF}/NEMOFILES/agrif_oce.in -rm -incdir ${MYDIR}/${NEW_CONF}/NEMOFILES/inc -comdirout ${MYDIR}/${NEW_CONF}/NEMOFILES/obj -convfile ${MYFILE} > /dev/null

ORCA2_OFF_PISCES failed reproducibility test. Compiler options: -em -s real64 -s integer32 -O2 -hflex_mp=intolerant -e0 -ez -N1023

Current code is : NEMO/trunk @ r14695  ( last change @ r14686 )

SETTE validation report generated for : 

       NEMO/trunk @ r14686 (last changed revision)

       on XC40_METO arch file


!!---------------1st pass------------------!!

   !----restart----!   
WGYRE_PISCES_ST              run.stat    restartability  passed :  14686
WGYRE_PISCES_ST              tracer.stat restartability  passed :  14686
WORCA2_ICE_PISCES_ST         run.stat    restartability  passed :  14686
WORCA2_ICE_PISCES_ST         tracer.stat restartability  passed :  14686
WORCA2_OFF_PISCES_ST         tracer.stat restartability  passed :  14686
WAMM12_ST                    run.stat    restartability  passed :  14686
WORCA2_SAS_ICE_ST            run.stat    restartability  passed :  14686
WAGRIF_DEMO_ST               directory                  MISSING :  14686
WWED025_ST                   run.stat    restartability  passed :  14686
WISOMIP+_ST                  run.stat    restartability  passed :  14686
WOVERFLOW_ST                 run.stat    restartability  passed :  14686
WLOCK_EXCHANGE_ST            run.stat    restartability  passed :  14686
WVORTEX_ST                   directory                  MISSING :  14686
WICE_AGRIF_ST                directory                  MISSING :  14686
WSWG_ST                      run.stat    restartability  passed :  14686

   !----repro----!   
WGYRE_PISCES_ST              run.stat    reproducibility passed :  14686
WGYRE_PISCES_ST              tracer.stat reproducibility passed :  14686
WORCA2_ICE_PISCES_ST         run.stat    reproducibility passed :  14686
WORCA2_ICE_PISCES_ST         tracer.stat reproducibility passed :  14686
[38;5;196mWORCA2_OFF_PISCES_ST         tracer.stat reproducibility FAILED :  14686  (results are different after   time steps) [0m
WAMM12_ST                    run.stat    reproducibility passed :  14686
WORCA2_SAS_ICE_ST            run.stat    reproducibility passed :  14686
WORCA2_ICE_OBS_ST            run.stat    reproducibility passed :  14686
WAGRIF_DEMO_ST               directory                  MISSING :  14686
WWED025_ST                   run.stat    reproducibility passed :  14686
WISOMIP+_ST                  run.stat    reproducibility passed :  14686
WVORTEX_ST                   directory                  MISSING :  14686
WICE_AGRIF_ST                directory                  MISSING :  14686
WSWG_ST                      run.stat    reproducibility passed :  14686

   !----agrif check----!   
ls: cannot access /home/d02/mandrej/nemo_validation_new_cce/NEMO_VALIDATION/WAGRIF_DEMO_NOAGRIF_ST/XC40_METO/14686/: No such file or directory
WAGRIF_DEMO_NOAGRIF_ST      WAGRIF_DEMO_ST               incomplete test

   !----result comparison check----!   

 No path for comparison specified. Result are not compare with any other revision. 
 To do it please fill NEMO_VALID_REF and NEMO_REV_REF in param.cfg. 

[andmirek]

AGRIF failure was related to building this configuration on login nodes (Amy investigation). Compute nodes have slightly different architecture than login nodes and not all (only those without any optimization) builds for compute node will run on login node.

This would indicate that AGRIF builds a test program and tries to run it.