java8中parallelStream性能测试及结果分析

测试1

@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.NANOSECONDS)
@Warmup(iterations = 5, time = 3, timeUnit = TimeUnit.SECONDS)
@Measurement(iterations = 20, time = 3, timeUnit = TimeUnit.SECONDS)
@Fork(1)
@State(Scope.Benchmark)
public class StreamBenchTest {
 List<String> data = new ArrayList<>();
 @Setup
 public void init() {
   // prepare
   for(int i=0;i<100;i++){
     data.add(UUID.randomUUID().toString());
   }
 }
 @TearDown
 public void destory() {
   // destory
 }
 @Benchmark
 public void benchStream(){
   data.stream().forEach(e -> {
     e.getBytes();
     try {
       Thread.sleep(10);
     } catch (InterruptedException e1) {
       e1.printStackTrace();
     }
   });
 }
 @Benchmark
 public void benchParallelStream(){
   data.parallelStream().forEach(e -> {
     e.getBytes();
     try {
       Thread.sleep(10);
     } catch (InterruptedException e1) {
       e1.printStackTrace();
     }
   });
 }
 public static void main(String[] args) throws RunnerException {
   Options opt = new OptionsBuilder()
       .include(".*" +StreamBenchTest.class.getSimpleName()+ ".*")
       .forks(1)
       .build();
   new Runner(opt).run();
 }
}

parallelStream线程数

默认是Runtime.getRuntime().availableProcessors() - 1，这里为7

运行结果

# Run complete. Total time: 00:02:44
Benchmark              Mode Cnt      Score     Error Units
StreamBenchTest.benchParallelStream avgt  20  155868805.437 ± 1509175.840 ns/op
StreamBenchTest.benchStream     avgt  20 1147570372.950 ± 6138494.414 ns/op

测试2

将数据data改为30，同时sleep改为100

Benchmark              Mode Cnt      Score     Error Units
StreamBenchTest.benchParallelStream avgt  20  414230854.631 ± 725294.455 ns/op
StreamBenchTest.benchStream     avgt  20 3107250608.500 ± 4805037.628 ns/op

可以发现sleep越长，parallelStream优势越明显。

小结

parallelStream在阻塞场景下优势更明显，其线程池个数默认为
Runtime.getRuntime().availableProcessors() - 1，如果需修改则需设置-Djava.util.concurrent.ForkJoinPool.common.parallelism=8

来源：https://segmentfault.com/a/1190000012755594