Java null check performance

Question

I was wondering if there was any significative difference on checking if an object is null in java with direct comparison, or by using the Objects.isNull() method.

public class Test {

  public final static Long ITERATIONS = 100000000L; 

  @Test
  public void noFnCalls() {
    balong startTime = System.currentTimeMillis();
    Object x = new Object();
    Long i;
    for (i = 0L; i < ITERATIONS; i++) {
      boolean t = x == null;
    }
    long estimatedTime = System.currentTimeMillis() - startTime;
    System.out.println("noFnCalls ellapsed time: " + estimatedTime);
  }

  @Test
  public void withFnCalls() {
    long startTime = System.currentTimeMillis();
    Object x = new Object();
    Long i;
    for (i = 0L; i < ITERATIONS; i++) {
      boolean t = Objects.isNull(x);
    }
    long estimatedTime = System.currentTimeMillis() - startTime;
    System.out.println("withFnCalls ellapsed time: " + estimatedTime);
  }
}

And surprisingly, at least for me, it always takes more time to finish the "noFnCalls". I was expecting pretty much the opposite result, since it results into a method call, using stack.

This is the output: (Changes every time, obviously, but always with "noFnCalls" higher)

noFnCalls ellapsed time: 583

withFnCalls ellapsed time: 463

Why is this produced?

Writing a correct microbenchmark in Java is tricky. Have you warmed up your system? Are your results the same if you run the tests in the reverse order? — k_ssb, Jun 08 '18 at 07:02
Please see: https://stackoverflow.com/questions/504103/how-do-i-write-a-correct-micro-benchmark-in-java — lexicore, Jun 08 '18 at 07:05
Practically, it doesnt matter which of the two you use, since the performence gain is minimal. However, Objects#isNull is meant to be used as a predicate. — Glains, Jun 08 '18 at 08:02

eis · Accepted Answer · 2018-06-08T07:34:31.973

The results you see are probably due to running "noFnCalls" first, without introducing proper warmup before the test and the measurement.

I get this:

withFnCalls ellapsed time: 444
noFnCalls ellapsed time: 471
withFnCalls ellapsed time: 334
noFnCalls ellapsed time: 331
withFnCalls ellapsed time: 330
noFnCalls ellapsed time: 325
withFnCalls ellapsed time: 331
noFnCalls ellapsed time: 326
withFnCalls ellapsed time: 326
noFnCalls ellapsed time: 328

Using

import java.util.Objects;

public class Test {

  public final static Long ITERATIONS = 100000000L; 

  public static void main(String args[]) {
    withFnCalls();
    noFnCalls();
    withFnCalls();
    noFnCalls();
    withFnCalls();
    noFnCalls();
    withFnCalls();
    noFnCalls();
    withFnCalls();
    noFnCalls();
  }
  public static void noFnCalls() {
    long startTime = System.currentTimeMillis();
    Object x = new Object();
    Long i;
    for (i = 0L; i < ITERATIONS; i++) {
      boolean t = x == null;
    }
    long estimatedTime = System.currentTimeMillis() - startTime;
    System.out.println("noFnCalls ellapsed time: " + estimatedTime);
  }

  public static void withFnCalls() {
    long startTime = System.currentTimeMillis();
    Object x = new Object();
    Long i;
    for (i = 0L; i < ITERATIONS; i++) {
      boolean t = Objects.isNull(x);
    }
    long estimatedTime = System.currentTimeMillis() - startTime;
    System.out.println("withFnCalls ellapsed time: " + estimatedTime);
  }
}

and

withFnCalls ellapsed time: 3618
noFnCalls ellapsed time: 3361
withFnCalls ellapsed time: 3445
noFnCalls ellapsed time: 3278
withFnCalls ellapsed time: 3350
noFnCalls ellapsed time: 3292
withFnCalls ellapsed time: 3309
noFnCalls ellapsed time: 3262
withFnCalls ellapsed time: 3293
noFnCalls ellapsed time: 3261

If I increase to 1000000000L iterations. This was done with Java 9 64-bit server jvm, build 9+181, by Oracle, running on Windows 10 with a machine having Intel i5-2600 cpu.

Like others have said, micro-benchmarking is hard and lot of different things affect the results. You shouldn't jump on conclusions with tests like these. This kind of test does not really tell much - any differences get easily lost in the noise measuring code that are so close to each other.

Obligatory recommended thread about micro-benchmarking in java: How do I write a correct micro-benchmark in Java?.

Same result also with Java 1.8.0 x64 on Windows 10, Linux/Debian and FreeBSD. — rustyx, Jun 08 '18 at 07:41

Java null check performance

1 Answers1