JMH

Java Microbenchmark Harness (JMH) is a Java toolkit for building, running, and analyzing microbenchmarks on the Java Virtual Machine. It is designed to produce accurate, repeatable measurements by controlling warmup, compilation, garbage collection, and other JVM effects, often by executing benchmarks in isolated forked JVM processes.

Benchmarks are authored as Java methods annotated with @Benchmark inside a class annotated with @State. The JMH annotation processor generates the harness code that runs the benchmarks with configurable warmup and measurement iterations, forks, and JVM arguments. Benchmarks can declare setup and teardown methods with @Setup and @TearDown, and state objects can have scoped lifetimes via @State(Scope.Thread) or @State(Scope.Benchmark).

Key features include support for multiple measurement modes (Throughput, AverageTime, SampleTime, SingleShotTime), parameterization with @Param, and grouping with @Group for related benchmarks. JMH provides a Blackhole to consume results and prevent dead code elimination. It also offers profilers and profiling options to observe GC activity, memory allocations, or CPU behavior during benchmarks.

Usage typically involves integrating with build tools such as Maven or Gradle through dedicated plugins, or running a generated benchmarks.jar produced by the annotation processor. The usual workflow is to annotate benchmark classes, build to generate the harness, and execute the resulting jar to obtain benchmark results displayed in the console and optionally exported to formats like JSON or CSV.

JMH is widely used in the Java ecosystem to obtain reliable microbenchmark measurements and to compare alternative implementations or optimizations under controlled conditions.