Add qws support for TSUBAME4 Pegasus and Sirius

Signed-off-by: Yoshifumi Nakamura <nakamura@riken.jp>

Author: yoshifuminakamura

config/queue.csv

@@ -11,7 +11,9 @@ NQSV_AOBA_B,qsub,"-Z -v http_proxy,https_proxy,HTTP_PROXY,HTTPS_PROXY -q ${queue
 PJM_WISTERIA_O,pjsub,"-g jh260034o -L rscgrp=${queue_group},elapse=${elapse},node=${nodes} --mpi proc=${proc} --omp thread=${nthreads}"
 PJM_WISTERIA_A,pjsub,"-g jh260034a -L rscgrp=${queue_group},elapse=${elapse},node=${nodes} --mpi proc=${proc} --omp thread=${nthreads}"
 PBS_TSUKUBA,qsub,"-q ${queue_group} -l select=${nodes}:mpiprocs=${numproc_node}:ompthreads=${nthreads} -l walltime=${elapse}"
-AGE_TSUBAME4,qsub,"-l ${queue_group}=${nodes} -l h_rt=${elapse}"
+PBS_PEGASUS,qsub,"-q ${queue_group} -A CNTBENCH -l elapstim_req=${elapse} -v OMP_NUM_THREADS=${nthreads}"
+PBS_SIRIUS,qsub,"-q ${queue_group} -A CNTBENCH -W group_list=CNTBENCH -l select=${nodes}:ncpus=24:mem=124gb:ngpus=1 -l walltime=${elapse}"
+AGE_TSUBAME4,qsub,"-g jh260034 -l ${queue_group}=${nodes} -l h_rt=${elapse}"
 SLURM_CAMPHOR3,sbatch,"-p ${queue_group} -t ${elapse} --rsc p=${proc}:t=${nthreads}:c=${nthreads}:m=1G"
 NQSV_OSAKA_CPU,qsub,"-q ${queue_group} -b ${nodes} -l elapstim_req=${elapse},cpunum_job=${nthreads}"
 NQSV_OSAKA_GPU,qsub,"-q ${queue_group} -b ${nodes} -l elapstim_req=${elapse},cpunum_job=${nthreads},gpunum_job=${gpu_per_node}"

config/system.csv

@@ -19,8 +19,8 @@ AOBA_B,cross,aoba_ab_login,aoba_ab_jacamar,NQSV_AOBA_B,lx
 AOBA_S,cross,aoba_s_login,aoba_s_jacamar,NQSV_AOBA_VE,sxs
 Odyssey,cross,wisteria_login,wisteria-o_jacamar,PJM_WISTERIA_O,short-o
 Aquarius,cross,wisteria_login,wisteria-a_jacamar,PJM_WISTERIA_A,short-a
-Pegasus,cross,pegasus_login,pegasus_jacamar,PBS_TSUKUBA,regular
-Sirius,cross,sirius_login,sirius_jacamar,PBS_TSUKUBA,regular
+Pegasus,cross,pegasus_login,pegasus_jacamar,PBS_PEGASUS,gpu
+Sirius,cross,sirius_login,sirius_jacamar,PBS_SIRIUS,mcrp
 TSUBAME4,cross,tsubame4_login,tsubame4_jacamar,AGE_TSUBAME4,node_f
 Camphor3,cross,camphor3_login,camphor3_jacamar,SLURM_CAMPHOR3,jha
 SQUID_CPU,cross,squid_login,squid_jacamar,NQSV_OSAKA_CPU,SQUID

programs/qws/build.sh

@@ -89,8 +89,22 @@ case "$system" in
 	source /work/opt/local/x86_64/cores/intel/2023.0.0/mpi/latest/env/vars.sh
 	make compiler=intel arch=skylake rdma= -j8
 	;;
+    Pegasus)
+	if [[ -r /etc/profile.d/modules.sh ]]; then
+	    source /etc/profile.d/modules.sh
+	fi
+	module purge
+	module load intel/2025.3.1 intmpi/2025.3.1
+	make compiler=intel arch=skylake mpi=1 omp=1 rdma=
+	;;
+    Sirius)
+	module load aocc/5.0.0 openmpi/5.0.10/aocc5.0.0
+	make compiler=aocc arch=zen4 mpi=1 rdma= omp=1
+	;;
     TSUBAME4)
-	make -j 8 fugaku_benchmark= omp=1 compiler=openmpi-gnu arch=skylake rdma= mpi=1 powerapi= CC=mpicc CXX=mpic++
+	module load openmpi/5.0.10-gcc aocc/4.1.0
+	export OMPI_CC=clang OMPI_CXX=clang++ OMPI_FC=flang
+	make -j4 compiler=aocc arch=zen4 rdma= mpi=1 omp=1 profiler=timing AMD_MARCH=-march=znver4
 	;;
     Camphor3)
 	camphor3_modulepath="${MODULEPATH:-}"

programs/qws/list.csv

@@ -20,6 +20,8 @@ AOBA_S,yes,1,1,8,0:10:00
 AOBA_B,yes,1,1,128,0:10:00
 Odyssey,yes,1,1,12,0:10:00
 Aquarius,yes,1,1,8,0:10:00
+Pegasus,yes,1,1,96,00:10:00
+Sirius,yes,1,1,24,0:10:00
 TSUBAME4,yes,1,1,192,0:10:00
 Camphor3,yes,1,1,112,0:10:00
 FNCX,yes,1,1,1,0:10:00

programs/qws/run.sh

@@ -171,8 +171,26 @@ case "$system" in
         mpiexec -n 1 ./main 32 6 4 3 1 1 1 1 -1 -1 6 50 > CASE0
         print_results CASE0 CASE0 1 >> ../results/result
         ;;
+    Pegasus)
+        qws_numproc=$((nodes * numproc_node))
+        if [[ -r /etc/profile.d/modules.sh ]]; then
+            source /etc/profile.d/modules.sh
+        fi
+        module purge
+        module load intel/2025.3.1 intmpi/2025.3.1
+        mpirun -n ${qws_numproc} ./main 32 6 4 3 1 1 1 1 -1 -1 6 50 > CASE0
+        print_results CASE0 CASE0 ${numproc_node} >> ../results/result
+        ;;
+    Sirius)
+        qws_numproc=$((nodes * numproc_node))
+        module load aocc/5.0.0 openmpi/5.0.10/aocc5.0.0
+        mpirun -n ${qws_numproc} ./main 32 6 4 3 1 1 1 1 -1 -1 6 50 > CASE0
+        print_results CASE0 CASE0 ${numproc_node} >> ../results/result
+        ;;
     TSUBAME4)
         qws_numproc=$((nodes * numproc_node))
+        module load openmpi/5.0.10-gcc aocc/4.1.0
+        export OMPI_CC=clang OMPI_CXX=clang++ OMPI_FC=flang
         mpirun -n ${qws_numproc} ./main 32 6 4 3 1 1 1 1 -1 -1 6 50 > CASE0
         print_results CASE0 CASE0 ${numproc_node} >> ../results/result
         ;;

scripts/test_submit.sh

@@ -142,17 +142,23 @@ case "$system" in
          --mpi proc=$proc --omp thread=$nthreads \
          script.sh
     ;;
-  Pegasus|Sirius)
-    echo qsub -q $queue_group \
-         -l select=${nodes}:mpiprocs=${numproc_node}:ompthreads=${nthreads} \
+  Pegasus)
+    echo qsub -q $queue_group -A CNTBENCH \
+         -l elapstim_req=${elapse} -v OMP_NUM_THREADS=${nthreads} script.sh
+    qsub -q $queue_group -A CNTBENCH \
+         -l elapstim_req=${elapse} -v OMP_NUM_THREADS=${nthreads} script.sh
+    ;;
+  Sirius)
+    echo qsub -q $queue_group -A CNTBENCH -W group_list=CNTBENCH \
+         -l select=${nodes}:ncpus=24:mem=124gb:ngpus=1 \
          -l walltime=${elapse} script.sh
-    qsub -q $queue_group \
-         -l select=${nodes}:mpiprocs=${numproc_node}:ompthreads=${nthreads} \
+    qsub -q $queue_group -A CNTBENCH -W group_list=CNTBENCH \
+         -l select=${nodes}:ncpus=24:mem=124gb:ngpus=1 \
          -l walltime=${elapse} script.sh
     ;;
   TSUBAME4)
-    echo qsub -l ${queue_group}=${nodes} -l h_rt=${elapse} script.sh
-    qsub -l ${queue_group}=${nodes} -l h_rt=${elapse} script.sh
+    echo qsub -g jh260034 -l ${queue_group}=${nodes} -l h_rt=${elapse} script.sh
+    qsub -g jh260034 -l ${queue_group}=${nodes} -l h_rt=${elapse} script.sh
     ;;
   Camphor3)
     proc=$((nodes * numproc_node))