#!/bin/bash

#

# Usage (to test): $ SMOKE=true CI_MACHINE_TYPE='skylake-2x' bash scripts/bench.bash

# For CI env: Don't set `SMOKE`

#

set -ex

if [ -v ${CI_MACHINE_TYPE} ]; then

echo "Need to set CI_MACHINE_TYPE"

exit 1

fi

# Clean-up old results

if [ "$SMOKE" = "" ]; then

rm -rf gh-pages

fi

NEXMARK_CSV_FILE='nexmark_results.csv'

NEXMARK_DRAM_CSV_FILE='dram_nexmark_results.csv'

NEXMARK_PERSISTENCE_CSV_FILE='persistence_nexmark_results.csv'

GALEN_CSV_FILE='galen_results.csv'

LDBC_CSV_FILE='ldbc_results.csv'

rm -f crates/nexmark/${NEXMARK_CSV_FILE} crates/dbsp/${GALEN_CSV_FILE} crates/dbsp/${LDBC_CSV_FILE} crates/nexmark/${NEXMARK_DRAM_CSV_FILE} crates/nexmark/${NEXMARK_PERSISTENCE_CSV_FILE}

rm -f nexmark_comment.txt

# Run nexmark benchmark

EVENT_RATE=10000000

MAX_EVENTS=100000000

GENERATORS=8

CORES=6

if [ "$SMOKE" != "" ]; then

EVENT_RATE=10000000

MAX_EVENTS=1000000

fi

cargo bench --bench nexmark -- --first-event-rate=${EVENT_RATE} --max-events=${MAX_EVENTS} --cpu-cores ${CORES} --num-event-generators ${GENERATORS} --source-buffer-size 10000 --input-batch-size 40000 --csv ${NEXMARK_CSV_FILE}

# Run galen benchmark

cargo bench --bench galen --features="with-csv" -- --workers 10 --csv ${GALEN_CSV_FILE}

# Run ldbc benchmarks

DATASET_SMALL='graph500-22'

DATASET_MEDIUM='datagen-8_4-fb'

if [ "$SMOKE" != "" ]; then

DATASET_SMALL='wiki-Talk'

DATASET_MEDIUM='kgs'

fi

cargo bench --bench ldbc-graphalytics --features="with-csv" -- bfs ${DATASET_SMALL} --threads 1 --csv ${LDBC_CSV_FILE}

cargo bench --bench ldbc-graphalytics --features="with-csv" -- bfs ${DATASET_MEDIUM} --threads 6 --csv ${LDBC_CSV_FILE}

cargo bench --bench ldbc-graphalytics --features="with-csv" -- pagerank ${DATASET_SMALL} --threads 1 --csv ${LDBC_CSV_FILE}

cargo bench --bench ldbc-graphalytics --features="with-csv" -- pagerank ${DATASET_MEDIUM} --threads 6 --csv ${LDBC_CSV_FILE}

# Run nexmark benchmark with persistence

EVENT_RATE=5000000

MAX_EVENTS=3000000

CORES=1

if [ "$SMOKE" != "" ]; then

EVENT_RATE=5000000

MAX_EVENTS=100000

fi

cargo bench --bench nexmark -- --first-event-rate=${EVENT_RATE} --max-events=${MAX_EVENTS} --cpu-cores ${CORES} --num-event-generators 6 --source-buffer-size 10000 --input-batch-size 40000 --csv ${NEXMARK_DRAM_CSV_FILE}

cargo bench --bench nexmark --features persistence -- --first-event-rate=${EVENT_RATE} --max-events=${MAX_EVENTS} --cpu-cores ${CORES} --num-event-generators 6 --source-buffer-size 10000 --input-batch-size 40000 --csv ${NEXMARK_PERSISTENCE_CSV_FILE}

# Clone repo

if [ ! -d "gh-pages" ]; then

git clone --depth 1 -b main git@github.com:gz/dbsp-benchmarks.git gh-pages

fi

pip3 install -r gh-pages/requirements.txt

# If you change this, adjust the command also in the append_csv function in utils.py:

DATE_PREFIX=`date +"%Y-%m-%d-%H-%M"`

if [ -z "${PR_COMMIT_SHA}" ]; then

# So we can run the script in non-runner environments and not pull request events

PR_COMMIT_SHA=`git rev-parse HEAD`

fi

# Add nexmark results to repo

DEPLOY_DIR="gh-pages/nexmark/${CI_MACHINE_TYPE}/${PR_COMMIT_SHA}/"

if [ -d "${DEPLOY_DIR}" ]; then

# If we already have results for this SHA (the directory exists),

# we will add the new results in a subdir

DEPLOY_DIR=${DEPLOY_DIR}${DATE_PREFIX}

fi

# Copy nexmark results

mkdir -p ${DEPLOY_DIR}

mv crates/nexmark/${NEXMARK_CSV_FILE} crates/nexmark/${NEXMARK_PERSISTENCE_CSV_FILE} crates/nexmark/${NEXMARK_DRAM_CSV_FILE} ${DEPLOY_DIR}

gzip -f ${DEPLOY_DIR}/${NEXMARK_CSV_FILE}

gzip -f ${DEPLOY_DIR}/${NEXMARK_PERSISTENCE_CSV_FILE}

gzip -f ${DEPLOY_DIR}/${NEXMARK_DRAM_CSV_FILE}

# Add galen results to repo

DEPLOY_DIR="gh-pages/galen/${CI_MACHINE_TYPE}/${PR_COMMIT_SHA}/"

if [ -d "${DEPLOY_DIR}" ]; then

# If we already have results for this SHA (the directory exists),

# we will add the new results in a subdir

DEPLOY_DIR=${DEPLOY_DIR}${DATE_PREFIX}

fi

# Copy galen results

mkdir -p ${DEPLOY_DIR}

mv crates/dbsp/${GALEN_CSV_FILE} ${DEPLOY_DIR}

gzip -f ${DEPLOY_DIR}/${GALEN_CSV_FILE}

# Add ldbc results to repo

DEPLOY_DIR="gh-pages/ldbc/${CI_MACHINE_TYPE}/${PR_COMMIT_SHA}/"

if [ -d "${DEPLOY_DIR}" ]; then

# If we already have results for this SHA (the directory exists),

# we will add the new results in a subdir

DEPLOY_DIR=${DEPLOY_DIR}${DATE_PREFIX}

fi

# Copy ldbc results

mkdir -p ${DEPLOY_DIR}

mv crates/dbsp/${LDBC_CSV_FILE} ${DEPLOY_DIR}

gzip -f ${DEPLOY_DIR}/${LDBC_CSV_FILE}

# Update CI history plots

python3 gh-pages/_scripts/ci_history.py --append --machine $CI_MACHINE_TYPE

# Push results to gh-pages repo

if [ "$SMOKE" = "" ]; then

cd gh-pages

if [ "$CI" = true ] ; then

git config user.email "no-reply@dbsp.systems"

git config user.name "dbsp-ci"

fi

git add .

git commit -a -m "Added benchmark results for $PR_COMMIT_SHA."

git push origin main

cd ..

python3 gh-pages/_scripts/compare_nexmark.py > nexmark_comment.txt

rm -rf gh-pages

git clean -f

else

python3 gh-pages/_scripts/compare_nexmark.py > nexmark_comment.txt

fi