Update module files to build gsi on Gaea-C5 (NOAA-EMC#746)

RussTreadon-NOAA · May 20, 2024 · 59d7578 · 59d7578
1 parent ebeaba1
commit 59d7578
Show file tree

Hide file tree

Showing 5 changed files with 38 additions and 54 deletions.
diff --git a/modulefiles/gsi_gaea.intel.lua b/modulefiles/gsi_gaea.intel.lua
@@ -1,42 +1,32 @@
 help([[
 ]])
 
-unload("intel")
-unload("cray-mpich")
-unload("cray-python")
-unload("darshan")
+prepend_path("MODULEPATH", "/ncrc/proj/epic/spack-stack/spack-stack-1.6.0/envs/gsi-addon-dev/install/modulefiles/Core")
 
-prepend_path("MODULEPATH", "/lustre/f2/dev/wpo/role.epic/contrib/spack-stack/spack-stack-1.4.1-c4/envs/unified-env/install/modulefiles/Core")
-prepend_path("MODULEPATH", "/lustre/f2/pdata/esrl/gsd/spack-stack/modulefiles")
-
-local stack_python_ver=os.getenv("stack_python_ver") or "3.9.12"
-local stack_intel_ver=os.getenv("stack_intel_ver") or "2022.0.2"
-local stack_cray_mpich_ver=os.getenv("stack_cray_mpich_ver") or "7.7.20"
+local stack_python_ver=os.getenv("stack_python_ver") or "3.11.6"
+local stack_intel_ver=os.getenv("stack_intel_ver") or "2023.1.0"
+local stack_cray_mpich_ver=os.getenv("stack_cray_mpich_ver") or "8.1.25"
 local cmake_ver=os.getenv("cmake_ver") or "3.23.1"
+local prod_util_ver=os.getenv("prod_util_ver") or "2.1.1"
 
 load(pathJoin("stack-intel", stack_intel_ver))
 load(pathJoin("stack-cray-mpich", stack_cray_mpich_ver))
 load(pathJoin("stack-python", stack_python_ver))
 load(pathJoin("cmake", cmake_ver))
 
 load("gsi_common")
-
-local prod_util_ver=os.getenv("prod_util_ver") or "1.2.2"
-load(pathJoin("prod-util", prod_util_ver))
-
--- Needed at runtime:
-load("alps")
+load(pathJoin("prod_util", prod_util_ver))
 
 local MKLROOT="/opt/intel/oneapi/mkl/2022.0.2/"
 prepend_path("LD_LIBRARY_PATH",pathJoin(MKLROOT,"lib/intel64"))
 pushenv("MKLROOT", MKLROOT)
 
-pushenv("GSI_BINARY_SOURCE_DIR", "/lustre/f2/dev/role.epic/contrib/GSI_data/fix/20240208")
+pushenv("GSI_BINARY_SOURCE_DIR", "/gpfs/f5/ufs-ard/world-shared/GSI_data/fix/gsi/20240208")
 
 setenv("CC","cc")
 setenv("FC","ftn")
 setenv("CXX","CC")
 pushenv("CRAYPE_LINK_TYPE","dynamic")
 
+unload("cray-libsci")
 whatis("Description: GSI environment on Gaea with Intel Compilers")
-
diff --git a/regression/regression_param.sh b/regression/regression_param.sh
@@ -26,8 +26,8 @@ case $machine in
     ;;
 	Gaea)
 	   sub_cmd="sub_gaea"
-           memnode=64
-           numcore=36
+           memnode=251
+           numcore=128
     ;;
 	wcoss2)
 	   sub_cmd="sub_wcoss2"
@@ -69,8 +69,8 @@ case $regtest in
            topts[1]="0:30:00" ; popts[1]="48/2"  ; ropts[1]="/1"
            topts[2]="0:30:00" ; popts[2]="60/3"  ; ropts[2]="/2"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:10:00" ; popts[1]="18/8/" ; ropts[1]="/1"
-           topts[2]="0:10:00" ; popts[2]="18/10/" ; ropts[2]="/2"
+           topts[1]="0:10:00" ; popts[1]="12/8/" ; ropts[1]="/1"
+           topts[2]="0:10:00" ; popts[2]="12/10/" ; ropts[2]="/2"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:10:00" ; popts[1]="12/8/" ; ropts[1]="/1"
            topts[2]="0:10:00" ; popts[2]="12/10/" ; ropts[2]="/2"
@@ -99,8 +99,8 @@ case $regtest in
            topts[1]="0:15:00" ; popts[1]="5/4/"  ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="10/4/"  ; ropts[2]="/1"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:15:00" ; popts[1]="18/1/"  ; ropts[1]="/1"
-           topts[2]="0:15:00" ; popts[2]="18/2/"  ; ropts[2]="/1"
+           topts[1]="0:15:00" ; popts[1]="64/1/"  ; ropts[1]="/1"
+           topts[2]="0:15:00" ; popts[2]="128/2/"  ; ropts[2]="/1"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:15:00" ; popts[1]="64/1/" ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="128/2/" ; ropts[2]="/1"
@@ -129,8 +129,8 @@ case $regtest in
            topts[1]="0:15:00" ; popts[1]="5/4/"  ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="10/4/"  ; ropts[2]="/1"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:15:00" ; popts[1]="18/1/"  ; ropts[1]="/1"
-           topts[2]="0:15:00" ; popts[2]="18/2/"  ; ropts[2]="/1"
+           topts[1]="0:15:00" ; popts[1]="64/1/"  ; ropts[1]="/1"
+           topts[2]="0:15:00" ; popts[2]="128/2/"  ; ropts[2]="/1"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:15:00" ; popts[1]="64/1/" ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="128/2/" ; ropts[2]="/1"
@@ -158,8 +158,8 @@ case $regtest in
            topts[1]="0:15:00" ; popts[1]="5/4/"  ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="10/4/"  ; ropts[2]="/1"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:15:00" ; popts[1]="18/1/"  ; ropts[1]="/1"
-           topts[2]="0:15:00" ; popts[2]="18/2/"  ; ropts[2]="/1"
+           topts[1]="0:15:00" ; popts[1]="64/1/"  ; ropts[1]="/1"
+           topts[2]="0:15:00" ; popts[2]="128/2/"  ; ropts[2]="/1"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:15:00" ; popts[1]="64/1/" ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="128/2/" ; ropts[2]="/1"
@@ -188,8 +188,8 @@ case $regtest in
            topts[1]="0:15:00" ; popts[1]="4/4/"  ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="6/6/"  ; ropts[2]="/1"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:15:00" ; popts[1]="4/4/"  ; ropts[1]="/1"
-           topts[2]="0:15:00" ; popts[2]="6/6/"  ; ropts[2]="/1"
+           topts[1]="0:15:00" ; popts[1]="28/1/"  ; ropts[1]="/1"
+           topts[2]="0:15:00" ; popts[2]="28/2/"  ; ropts[2]="/1"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:15:00" ; popts[1]="28/1/" ; ropts[1]="/1"
            topts[2]="0:15:00" ; popts[2]="28/2/" ; ropts[2]="/1"
@@ -218,8 +218,8 @@ case $regtest in
            topts[1]="0:30:00" ; popts[1]="6/12/"  ; ropts[1]="/1"
            topts[2]="0:30:00" ; popts[2]="8/12/"  ; ropts[2]="/1"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:30:00" ; popts[1]="8/6/"  ; ropts[1]="/1"
-           topts[2]="0:30:00" ; popts[2]="8/8/"  ; ropts[2]="/1"
+           topts[1]="0:30:00" ; popts[1]="14/8/"  ; ropts[1]="/1"
+           topts[2]="0:30:00" ; popts[2]="14/14/"  ; ropts[2]="/1"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:30:00" ; popts[1]="14/8/" ; ropts[1]="/1"
            topts[2]="0:30:00" ; popts[2]="14/14/" ; ropts[2]="/2"
@@ -248,8 +248,8 @@ case $regtest in
            topts[1]="0:10:00" ; popts[1]="12/3/" ; ropts[1]="/1"
            topts[2]="0:10:00" ; popts[2]="12/5/" ; ropts[2]="/2"
         elif [[ "$machine" = "Gaea" ]]; then
-           topts[1]="0:10:00" ; popts[1]="12/3/" ; ropts[1]="/1"
-           topts[2]="0:10:00" ; popts[2]="12/5/" ; ropts[2]="/2"
+           topts[1]="0:10:00" ; popts[1]="16/2/" ; ropts[1]="/1"
+           topts[2]="0:10:00" ; popts[2]="16/4/" ; ropts[2]="/2"
         elif [[ "$machine" = "wcoss2" ]]; then
            topts[1]="0:10:00" ; popts[1]="16/2/" ; ropts[1]="/1"
            topts[2]="0:10:00" ; popts[2]="16/4/" ; ropts[2]="/2"
@@ -315,7 +315,7 @@ elif [[ "$machine" = "Gaea" ]]; then
    export MPI_BUFS_PER_PROC=256
    export MPI_BUFS_PER_HOST=256
    export MPI_GROUP_MAX=256
-   export APRUN="srun --export=ALL --mpi=pmi2 -n \$ntasks"
+   export APRUN="srun --export=ALL -n \$ntasks"
 elif [[ "$machine" = "wcoss2" ]]; then
    export OMP_PLACES=cores
    export OMP_STACKSIZE=2G

diff --git a/regression/regression_var.sh b/regression/regression_var.sh
@@ -36,7 +36,7 @@ elif [[ -d /mnt/lfs4 || -d /jetmon || -d /mnt/lfs1 ]]; then # Jet
   export machine="Jet"
 elif [[ -d /discover ]]; then # NCCS Discover
   export machine="Discover"
-elif [[ -d /sw/gaea ]]; then # Gaea
+elif [[ -d /ncrc ]]; then # Gaea
   export machine="Gaea"
 elif [[ -d /data/prod ]]; then # S4
   export machine="S4"
@@ -52,17 +52,13 @@ echo "Running Regression Tests on '$machine'";
 case $machine in
   Gaea)
     export queue="normal"
-    export noscrub="/lustre/f2/scratch/$LOGNAME/gsi_tmp/noscrub"
-    export ptmp="/lustre/f2/scratch/$LOGNAME/gsi_tmp/ptmp"
-    export casesdir="/lustre/f2/dev/role.epic/contrib/GSI_data/CASES/regtest"
-
-    export group="global"
-    if [[ "$cmaketest" = "false" ]]; then
-      export basedir="/lustre/f2/dev/$LOGNAME/sandbox/GSI"
-    fi
+    export group="ufs-ard"
+    export noscrub="/gpfs/f5/${group}/scratch/${USER}/$LOGNAME/gsi_tmp/noscrub"
+    export ptmp="/gpfs/f5/${group}/scratch/${USER}/$LOGNAME/gsi_tmp/ptmp"
+    export casesdir="/gpfs/f5/ufs-ard/world-shared/GSI_data/CASES/regtest"
 
     export check_resource="no"
-    export accnt="nggps_emc"
+    export accnt="ufs-ard"
   ;;
   wcoss2)
       export local_or_default="${local_or_default:-/lfs/h2/emc/da/noscrub/$LOGNAME}"

diff --git a/ush/module-setup.sh b/ush/module-setup.sh
@@ -56,10 +56,8 @@ elif [[ $MACHINE_ID = gaea* ]] ; then
         # the module command fails.  Hence we actually have to source
         # /etc/profile here.
         source /etc/profile
-        __ms_source_etc_profile=yes
     fi
-
-    source /lustre/f2/dev/role.epic/contrib/Lmod_init.sh
+    module reset
 
 elif [[ $MACHINE_ID = expanse* ]]; then
     # We are on SDSC Expanse

diff --git a/ush/sub_gaea b/ush/sub_gaea
@@ -88,8 +88,8 @@ output=${output:-$jobname.out}
 myuser=$LOGNAME
 myhost=$(hostname)
 
-if [ -d /lustre/f2/scratch/$LOGNAME ]; then
-  DATA=/lustre/f2/scratch/$LOGNAME/tmp
+if [ -d /gpfs/f5/epic/scratch/${USER}/$LOGNAME ]; then
+  DATA=/gpfs/f5/epic/scratch/${USER}/$LOGNAME/tmp
 fi
 DATA=${DATA:-$ptmp/tmp}
 
@@ -110,7 +110,7 @@ echo ""
 echo "#SBATCH --output=$output"                                                    >> $cfile
 echo "#SBATCH --job-name=$jobname"                                                 >> $cfile
 echo "#SBATCH --qos=$queue"                                                        >> $cfile
-echo "#SBATCH --clusters=c4"                                                       >> $cfile
+echo "#SBATCH --clusters=c5"                                                       >> $cfile
 echo "#SBATCH --time=$timew"                                                       >> $cfile
 echo "#SBATCH --nodes=$nodes --ntasks-per-node=$procs --cpus-per-task=$threads"    >> $cfile
 echo "#SBATCH --account=$accnt"                                                    >> $cfile
@@ -121,13 +121,13 @@ echo "export ntasks=$(( $nodes * $procs ))" >> $cfile
 echo "export ppn=$procs"                    >> $cfile
 echo "export threads=$threads"              >> $cfile
 echo "export OMP_NUM_THREADS=$threads"      >> $cfile
-echo "ulimit -s unlimited"                  >> $cfile
+echo "ulimit -s unlimited" 		    >> $cfile
 
 echo "" >>$cfile
 echo ". "$(awk '{ print $1, $2, $3, $4, $5, $6, $7, $8, $9 }' $regdir/regression_var.out)  >>$cfile
 echo "" >>$cfile
 
-echo "source /lustre/f2/dev/role.epic/contrib/Lmod_init.sh"  >> $cfile
+echo "module reset" >> $cfile
 echo "module use $modulefiles"                        >> $cfile
 echo "module load gsi_gaea.intel" >> $cfile
 echo "module list"                                           >> $cfile
@@ -158,7 +158,7 @@ sbatch=${sbatch:-sbatch}
 ofile=$DATA/subout$$
 >$ofile
 chmod 777 $ofile
-$sbatch  --export=ALL $cfile >$ofile
+$sbatch  $cfile >$ofile
 rc=$?
 cat $ofile
 if [[ -w $SUBLOG ]];then