[benchmark] Add script for comparing benchmark results to baseline

Benchmark will fail if the average time is more than 5% higher
than the baseline.

scripts/check_benchmark_results.sh

@@ -0,0 +1,41 @@
+#!/bin/bash
+
+# Usage:
+# check_benchmark_results.sh <baseline> <results> <comparison report in txt> <comparison report in html>
+
+# Benchmark comparison script uses a hard-coded 5% threshold by default, but it causes random failures
+# Increase the threshold to 10%
+
+if [ "$(uname)" == "Darwin" ]; then
+  sed -i '' s/"if res > 0.05"/"if res > 0.1"/ $PWD/vendor/benchmark/tools/gbench/report.py
+elif [ "$(expr substr $(uname -s) 1 5)" == "Linux" ]; then
+  sed -i s/"if res > 0.05"/"if res > 0.1"/ $PWD/vendor/benchmark/tools/gbench/report.py
+fi
+
+# Run the bencmark comparison
+vendor/benchmark/tools/compare.py benchmarks $1 $2 > $3
+
+# Find cases where the average is above threshold
+tests=`fgrep mean $3 | cut -c1-85 | grep 91m`
+
+if [ -z "$tests" ]
+then
+      echo "" >> $3
+      echo "PASSED, all benchmarks within 10% threshold" >> $3
+
+      ansi2html < $3 > $4
+      exit 0
+else
+      test_names=`fgrep mean $3 | cut -c1-85 | grep 91m | cut -f1 -d' '`
+
+      echo "" >> $3
+      echo "FAILED, following benchmarks are not within 10% threshold:" >> $3
+      echo "" >> $3
+      for t in $test_names;
+      do
+        echo $t >> $3
+      done
+
+      ansi2html < $3 > $4
+      exit 1
+fi