add script for sweeping throught multiple problem sizes in parthenon …

…strong scaling

utils/parthenon/do_strong_scaling_cpu_sweep.sh

@@ -0,0 +1,54 @@
+#!/bin/bash
+
+set +x
+set +e
+
+FOOTPRINT=$1
+
+#BEGIN Do not change the following for official benchmarking 
+NXB=16
+NLIM=250
+NLVL=3
+#END
+
+NXs=(64 128 160)  
+
+EXEC=./burgers-benchmark # executable
+INP=../../../benchmarks/burgers/burgers.pin
+
+# loop
+for NX in ${NXs[@]}; do 
+i=0
+IDEAL1=0
+TIMING_FILE_NAME="cpu_${NX}.csv"
+HEADER="No. Cores, Actual, Ideal"
+echo "Saving timing to ${TIMING_FILE_NAME}"
+echo "${HEADER}"
+echo "${HEADER}" > ${TIMING_FILE_NAME}
+
+for count in  8 32 56 88 112; do
+    echo "Core count = ${count}"
+    L=`echo "((${NX}^3)/(${NXB}^3)) >= ${count}" | bc -l`
+    if (($L == 0));  then
+	continue; 
+    fi
+    outfile=$(printf "strong-scale-%d-%d.out" ${NX} ${count})
+    errfile=$(printf "strong-scale-%d-%d.err" ${NX} ${count})
+    echo "saving to output file ${outfile}"
+    ARGS="${EXEC} -i ${INP} parthenon/mesh/nx1=${NX} parthenon/mesh/nx2=${NX} parthenon/mesh/nx3=${NX} parthenon/meshblock/nx1=${NXB} parthenon/meshblock/nx2=${NXB} parthenon/meshblock/nx3=${NXB} parthenon/time/nlim=${NLIM} parthenon/mesh/numlevel=${NLVL}"
+    CMD="srun -n ${count} --hint=nomultithread  -o ${outfile} -e ${errfile} ${ARGS}"
+    echo ${CMD}
+    ${CMD}
+    wait
+    zc=$(grep 'zone-cycles/wallsecond = ' ${outfile} | cut -d '=' -f 2 | xargs)
+    echo ${zc}
+    if (( ${i} == 0 )); then
+       IDEAL1=$(echo "print(\"%.7e\" % (${zc}/${count}))" | python3)
+    fi
+    IDEAL=$(echo "print(\"%.7e\" % (${count}*${IDEAL1}))" | python3)
+    OUTSTR="${count}, ${zc}, ${IDEAL}"
+    echo "${OUTSTR}"
+    echo "${OUTSTR}" >> ${TIMING_FILE_NAME}
+    i=$((${i} + 1))
+done
+done