validating - Microbenchmarking

最新推荐文章于 2023-04-11 09:15:33 发布

原创最新推荐文章于 2023-04-11 09:15:33 发布 · 385 阅读

0 ·

CC 4.0 BY-SA版权

Thinking in Java 专栏收录该内容

155 篇文章

订阅专栏

本文通过Java代码基准测试，对比了不同数组设置方法的运行效率。实验显示，在使用共享资源的情况下，并行处理并不总是更快。

Benchmarking means timing pieces of code or algorithms to see which runs faster.

// onjava/Timer.java
// (c)2017 MindView LLC: see Copyright.txt
// We make no guarantees that this code is fit for any purpose.
// Visit http://OnJava8.com for more book information.
package onjava;

import static java.util.concurrent.TimeUnit.*;

public class Timer {
  private long start = System.nanoTime();

  public long duration() {
    return NANOSECONDS.toMillis(System.nanoTime() - start);
  }

  public static long duration(Runnable test) {
    Timer timer = new Timer();
    test.run();
    return timer.duration();
  }
}

// validating/BadMicroBenchmark.java
// (c)2017 MindView LLC: see Copyright.txt
// We make no guarantees that this code is fit for any purpose.
// Visit http://OnJava8.com for more book information.
// {ExcludeFromTravisCI}

import java.util.*;
import onjava.Timer;

public class BadMicroBenchmark {
  static final int SIZE = 250_000_000; // 250 million

  public static void main(String[] args) {
    try { // For machines with insufficient memory
      long[] la = new long[SIZE];
      System.out.println("setAll: " + Timer.duration(() -> Arrays.setAll(la, n -> n))); // Arrays.setAll since 1.8
      System.out.println(
          "parallelSetAll: " + Timer.duration(() -> Arrays.parallelSetAll(la, n -> n))); // Arrays.parallelSetAll since 1.8
    } catch (OutOfMemoryError e) {
      System.out.println("Insufficient memory");
      System.exit(0);
    }
  }
}
/* My Output:
setAll: 217
parallelSetAll: 204
*/

when depends on a common resource, the parallel version can end up being much slower, as the separate contend for that resource

// validating/BadMicroBenchmark2.java
// (c)2017 MindView LLC: see Copyright.txt
// We make no guarantees that this code is fit for any purpose.
// Visit http://OnJava8.com for more book information.
// Relying on a common resource

import java.util.*;
import onjava.Timer;

public class BadMicroBenchmark2 {
  // SIZE reduced to make it run faster:
  static final int SIZE = 5_000_000;

  public static void main(String[] args) {
    long[] la = new long[SIZE];
    Random r = new Random();
    System.out.println(
        "parallelSetAll: " + Timer.duration(() -> Arrays.parallelSetAll(la, n -> r.nextLong())));
    System.out.println("setAll: " + Timer.duration(() -> Arrays.setAll(la, n -> r.nextLong())));
    SplittableRandom sr = new SplittableRandom(); // since 1.8, is designed for parallel algorithms
// This algorithm was inspired by the "DotMix" algorithm
    System.out.println(
        "parallelSetAll: " + Timer.duration(() -> Arrays.parallelSetAll(la, n -> sr.nextLong())));
    System.out.println("setAll: " + Timer.duration(() -> Arrays.setAll(la, n -> sr.nextLong())));
  }
}
/* My Output:
parallelSetAll: 946
setAll: 106
parallelSetAll: 73
setAll: 16
*/

references:

1. On Java 8 - Bruce Eckel

2. https://github.com/wangbingfeng/OnJava8-Examples/blob/master/onjava/Timer.java

3. http://hg.openjdk.java.net/jdk8/jdk8/jdk/file/687fd7c7986d/src/share/classes/java/lang/System.java

4. http://hg.openjdk.java.net/jdk8/jdk8/jdk/file/687fd7c7986d/src/share/classes/java/util/concurrent/TimeUnit.java

5. https://github.com/wangbingfeng/OnJava8-Examples/blob/master/validating/BadMicroBenchmark.java

6. https://github.com/wangbingfeng/OnJava8-Examples/blob/master/validating/BadMicroBenchmark2.java

7. http://hg.openjdk.java.net/jdk8/jdk8/jdk/file/687fd7c7986d/src/share/classes/java/util/SplittableRandom.java