mirror of
https://github.com/cp2k/cp2k.git
synced 2026-07-23 19:45:17 -04:00
118 lines
3.4 KiB
Bash
Executable file
118 lines
3.4 KiB
Bash
Executable file
#!/bin/bash -e
|
|
|
|
# author: Ole Schuett
|
|
|
|
if (($# != 1)); then
|
|
echo "Usage: test_performance.sh <PROFILE>"
|
|
exit 1
|
|
fi
|
|
|
|
PROFILE=$1
|
|
|
|
function run_benchmark {
|
|
set +e #disable error trapping
|
|
OMP_THREADS=$1
|
|
MPI_RANKS=$2
|
|
INPUT=$3
|
|
OUTPUT=$4
|
|
echo -n "Running ${INPUT} with ${OMP_THREADS} threads and ${MPI_RANKS} ranks... "
|
|
if OMP_NUM_THREADS="${OMP_THREADS}" mpiexec -np "${MPI_RANKS}" \
|
|
"/opt/cp2k/build/bin/cp2k.psmp" "${INPUT}" &> "${OUTPUT}"; then
|
|
echo "done."
|
|
else
|
|
echo -e "failed.\n\n"
|
|
tail -n 100 "${OUTPUT}"
|
|
echo -e "\nSummary: Running ${INPUT} failed."
|
|
echo -e "Status: FAILED\n"
|
|
exit 0
|
|
fi
|
|
set -e #re-enable error trapping
|
|
}
|
|
|
|
# shellcheck disable=SC1091
|
|
source /opt/cp2k-toolchain/install/setup
|
|
|
|
echo -e '\n============== CP2K Binary Flags ============='
|
|
./build/bin/cp2k.psmp --version | grep cp2kflags
|
|
|
|
echo -e '\n========== Checking Benchmark Inputs ========='
|
|
if ! ./tools/regtesting/check_inputs.py "./build/bin/cp2k.psmp" "./benchmarks/"; then
|
|
echo -e "\nSummary: Some benchmark inputs could not be parsed."
|
|
echo -e "Status: FAILED\n"
|
|
exit 0
|
|
fi
|
|
|
|
echo -e '\n========== Running Performance Test =========='
|
|
mkdir -p /workspace/artifacts
|
|
cd ./benchmarks
|
|
TIME_START=$(date +%s)
|
|
|
|
BENCHMARKS=(
|
|
"QS/H2O-64.inp"
|
|
"QS/H2O-64_nonortho.inp"
|
|
"QS_reference/w64PBE.inp"
|
|
"QS_reference/w64SCAN.inp"
|
|
"QS_single_node/H2O-hyb.inp"
|
|
"QS_single_node/GW_PBE_4benzene.inp"
|
|
"QS_single_node/RI-HFX_H2O-32.inp"
|
|
"QS_single_node/RI-MP2_ammonia.inp"
|
|
"QS_single_node/diag_cu144_broy.inp"
|
|
"QS_single_node/bench_dftb.inp"
|
|
"QS_single_node/dbcsr.inp"
|
|
"QMMM_MQAE/MQAE_single_node.inp"
|
|
)
|
|
|
|
if [[ "${PROFILE}" == "toolchain" ]]; then
|
|
echo 'Plot: name="total_timings_32omp", title="Total Timings with 32 OpenMP Threads", ylabel="time [s]"'
|
|
echo 'Plot: name="total_timings_32mpi", title="Total Timings with 32 MPI Ranks", ylabel="time [s]"'
|
|
echo ''
|
|
|
|
for INPUT in "${BENCHMARKS[@]}"; do
|
|
INPUT_BASENAME=$(basename "${INPUT}")
|
|
LABEL=${INPUT_BASENAME%.*}
|
|
OUTPUT_MPI="/workspace/artifacts/${LABEL}_32mpi.out"
|
|
OUTPUT_OMP="/workspace/artifacts/${LABEL}_32omp.out"
|
|
cd "$(dirname "${INPUT}")"
|
|
run_benchmark 1 32 "${INPUT_BASENAME}" "${OUTPUT_MPI}"
|
|
run_benchmark 32 1 "${INPUT_BASENAME}" "${OUTPUT_OMP}"
|
|
cd ..
|
|
echo ""
|
|
/opt/cp2k/plot_performance.py \
|
|
"${LABEL} with 32 OpenMP Threads" "${LABEL}" "32omp" "${OUTPUT_OMP}" \
|
|
"${LABEL} with 32 MPI Ranks" "${LABEL}" "32mpi" "${OUTPUT_MPI}"
|
|
echo ""
|
|
done
|
|
|
|
elif [[ "${PROFILE}" == "toolchain_cuda_"* ]]; then
|
|
echo 'Plot: name="total_timings_6cpu_1gpu", title="Total Timings with 6 CPU Cores and 1 GPU", ylabel="time [s]"'
|
|
echo ''
|
|
|
|
for INPUT in "${BENCHMARKS[@]}"; do
|
|
if [[ "$INPUT" == "QS_single_node/H2O-hyb.inp" ]]; then
|
|
continue # Has no gpu acceleration, yet.
|
|
fi
|
|
INPUT_BASENAME=$(basename "${INPUT}")
|
|
LABEL=${INPUT_BASENAME%.*}
|
|
OUTPUT="/workspace/artifacts/${LABEL}_6cpu_1gpu.out"
|
|
cd "$(dirname "${INPUT}")"
|
|
export CUDA_VISIBLE_DEVICES=0
|
|
run_benchmark 3 2 "${INPUT_BASENAME}" "${OUTPUT}"
|
|
cd ..
|
|
echo ""
|
|
/opt/cp2k/plot_performance.py \
|
|
"${LABEL} with 6 CPU Cores and 1 GPU" "${LABEL}" "6cpu_1gpu" "${OUTPUT}"
|
|
echo ""
|
|
done
|
|
|
|
else
|
|
echo "Unknown PROFILE: ${PROFILE}"
|
|
exit 1
|
|
fi
|
|
|
|
TIME_END=$(date +%s)
|
|
DURATION=$(printf "%i" $(((TIME_END - TIME_START) / 60)))
|
|
|
|
echo -e "\nSummary: Performance test took ${DURATION} minutes."
|
|
echo -e "Status: OK\n"
|
|
|
|
#EOF
|