Gradle Configuration

JMH can be configured to be run as a gradle task.

• Upstream source

Plugin Configuration

Outside of the jmh plugin a core language plugin will be needed. This tutorial foucses on Kotlin but other languages can be used.

Gradle Version Wrapper

You can see the latest versions for the respective libraries on the upstream source link.

build.gradle.kts

plugins {
    alias(libs.plugins.jmh)
    alias(libs.plugins.kotlin)
}

Configuring JMH

The JMH plugin can be configured via a closure of jmh {}.

val HOME = System.getenv("HOME")

jmh {
    // Number of warmup iterations to do.
    warmupIterations.set(2)
    // Number of measurement iterations to do.
    iterations.set(2)
    // Number of subprocesses to spin up
    fork.set(2)
    // Strategy to apply when encountring duplicate classes during creation of the fat jar (i.e. while executing jmhJar task)
    duplicateClassesStrategy.set(DuplicatesStrategy.EXCLUDE)
    // Required if the jar gets rather large.
    zip64.set(true)
    // Going lower than ms is beneficial, higher is only beneficial for longer running methods.
    timeUnit.set("ms")
    // Benchmark mode. Available modes are: [Throughput/thrpt, AverageTime/avgt, SampleTime/sample, SingleShotTime/ss, All/all]
    benchmarkMode.set(mutableListOf("all"))
    // Result format to save as, json is easiest to parse.
    resultFormat.set("json")
    humanOutputFile
    // Explained in the profiler docuument.
    profilers.set(mutableListOf(
        "stack",
        "async:libPath=${HOME}/.libs/async-profiler/build/libasyncProfiler.so;output=flamegraph;dir=flamecharts"
    ))
    jmhVersion.set("1.36")
}

Source Files

The JMH plugin is expecting source code for the benchmarks to be stored at:

./src/jmh/main

Benchmarking Modes

There are several ways to benchmark and generate results, they are availalbe via jmh.benchmarkMode.

Name	Description
Throughput	Measures the number of operations per second, meaning the number of times per second your benchmark method could be executed.
Average Time	Measures the average time it takes for the benchmark method to execute (a single execution).
Sample Time	Measures how long time it takes for the benchmark method to execute, including max, min time etc.
Single Shot Time	Measures how long time a single benchmark method execution takes to run. This is good to test how it performs under a cold start (no JVM warm up).
All	Measures all of the above.