Microbenchmarks

Microbenchmarking is used if we want to know which methods perform better than others. A naive approach would wrap the called function inside a block of a measureTimeMillis or measureNanoTime function. These functions just calculate the difference between start and stop timestamps:

public inline fun measureTimeMillis(block: () -> Unit) : Long {
    val start = System.currentTimeMillis()
    block()
    return System.currentTimeMillis() - start
}

At a first glance, this approach gives us an idea about the performance of the code inside the block, but if we perform several experiments, it will be clear that this is the wrong approach. We've already touched on how the JVM executes and changes programming code to perform optimizations. When lines of code are executed, the JVM collects information about them. It will then try to optimize them and collect new information the next time it executes these same lines. Therefore, we have to take into account this factor and measure code after a warm-up to simulate real-world conditions.

To implement microbenchmarking correctly, you have to use the Java Microbenchmark Harness.