Profiling

Learn performance measurement with JVM profiling and JMH.

What is profiling?

Profiling measures which part of a program is slow. On Kotlin/JVM the most common tool is the JMH (Java Microbenchmark Harness) library.

JMH marks a method with @Benchmark and measures it at nanosecond precision. The result is reported as ns/op (nanoseconds per operation). A good benchmark includes warmup iterations and several measurements. The profileInfo model in this lesson shows the concept with a simple example.

JMH setup

JMH is added to a project with a Gradle plugin. Benchmark methods are marked with @Benchmark; @BenchmarkMode sets the measurement mode.

Example: @Benchmark fun work(): Int { return 42 }. JMH runs this method many times, starting measurement after a warmup lap. Results can be exported as CSV or JSON. A bad benchmark (allocating inside the measured code) gives misleading numbers, so benchmark methods are kept simple.

Exercise Write profileInfo

Define fun profileInfo(): String. It should return "JMH". In main, write println(profileInfo()).

Show example solution

fun profileInfo(): String = "JMH"
fun main() {
    println(profileInfo())
}

Test yourself

What does JMH measure?
```
JMH measures small-scale performance.
```
- Disk I/O
- Memory usage
- Micro performance
- Network latency
Show answer

Micro performance — JMH measures micro performance at method level.

Complete the @Benchmark line

@Benchmark fun ____(): Int { return 42 }

Show answer

@Benchmark fun work(): Int { return 42 }

Match the JMH parts

JMH

Benchmark library
@Benchmark

Method marker
@BenchmarkMode

Measure mode
ns/op

Nanos per op

Next lesson: Allocation

We Are Enhancing Your Experience

Cookie Preferences

Strictly Necessary Cookies

Analytical Cookies