The benchmarking from pandiyan90

benchmarking's Introduction

Java benchmarks with JMH

This project contains next benchmarks:

Benchmark for demonstration how JVM handle problems with false sharing with @Contended annotation
Benchmark for demonstration how branch prediction behaves on the example of operation with sorted and unsorted array - useful discussion on stackoverflow

Prerequisites

Java 8+
Gradle

FalseSharingBenchmark

The -XX:-RestrictContended flag is significant.

The output will look like this:

Benchmark                                      Mode  Cnt          Score         Error  Units
ContendedBenchmarks.padded                    thrpt   30  236434399.252 ± 9209297.004  ops/s
ContendedBenchmarks.padded:updatePaddedA      thrpt   30  114409851.688 ± 4803042.336  ops/s
ContendedBenchmarks.padded:updatePaddedB      thrpt   30  122024547.564 ± 5108447.473  ops/s
ContendedBenchmarks.unpadded                  thrpt   30   63941093.638 ± 2478065.296  ops/s
ContendedBenchmarks.unpadded:updateUnpaddedA  thrpt   30   32050855.915 ± 1371581.512  ops/s
ContendedBenchmarks.unpadded:updateUnpaddedB  thrpt   30   31890237.723 ± 1344691.282  ops/s

The default benchmark mode is Throughput and in this case higher value is better

BranchPredictionBenchmark

The output will look like this:

Benchmark                           Mode  Cnt  Score   Error  Units
BranchPredictionBenchmark.sorted    avgt   25  4.289 ± 0.754  ns/op
BranchPredictionBenchmark.unsorted  avgt   25  9.742 ± 0.466  ns/op

Benchmark mode for BranchPredictionBenchmark is AverageTime and in this case lower value is better

For more details about modes in JMH see specification

Recommend Projects