- 使用执行程序是关于利用CPU和/或CPU内核的，因此，如果创建一个最多利用CPU数量的线程池，则必须拥有与CPU /内核一样多的线程。
- 没错，创建新对象的成本太高了。因此，减少费用的一种方法是分批使用。如果您知道要进行的计算的种类和数量，则可以创建批处理。因此，请考虑在一个执行的任务中完成了数千次计算。您为每个线程创建批处理。一旦计算完成（java.util.concurrent.Future），就可以创建下一批。甚至可以在parralel中完成新批处理的创建（4个CPU-> 3个用于计算的线程，1个用于批处理供应的线程）。最后，您可能会获得更高的吞吐量，但会带来更高的内存需求（批次，资源调配）。
编辑：我更改了您的示例，并让它在我的小双核x200笔记本电脑上运行。
provisioned 2 batches to be executedsimpleCompuation:14computationWithObjCreation:17computationWithObjCreationAndExecutors:9
如您在源代码中所看到的，我也从批处理和执行器生命周期中删除了度量。与其他两种方法相比，这更公平。
自己查看结果…
import java.util.List;import java.util.Vector;import java.util.concurrent.ExecutorService;import java.util.concurrent.Executors;import java.util.concurrent.TimeUnit;public class ExecServicePerformance { private static int count = 100000; public static void main( String[] args ) throws InterruptedException { final int cpus = Runtime.getRuntime().availableProcessors(); final ExecutorService es = Executors.newFixedThreadPool( cpus ); final Vector< Batch > batches = new Vector< Batch >( cpus ); final int batchComputations = count / cpus; for ( int i = 0; i < cpus; i++ ) { batches.add( new Batch( batchComputations ) ); } System.out.println( "provisioned " + cpus + " batches to be executed" ); // warmup simpleCompuation(); computationWithObjCreation(); computationWithObjCreationAndExecutors( es, batches ); long start = System.currentTimeMillis(); simpleCompuation(); long stop = System.currentTimeMillis(); System.out.println( "simpleCompuation:" + ( stop - start ) ); start = System.currentTimeMillis(); computationWithObjCreation(); stop = System.currentTimeMillis(); System.out.println( "computationWithObjCreation:" + ( stop - start ) ); // Executor start = System.currentTimeMillis(); computationWithObjCreationAndExecutors( es, batches ); es.shutdown(); es.awaitTermination( 10, TimeUnit.SEConDS ); // Note: Executor#shutdown() and Executor#awaitTermination() requires // some extra time. But the result should still be clear. stop = System.currentTimeMillis(); System.out.println( "computationWithObjCreationAndExecutors:" + ( stop - start ) ); } private static void computationWithObjCreation() { for ( int i = 0; i < count; i++ ) { new Runnable() { @Override public void run() { double x = Math.random() * Math.random(); } }.run(); } } private static void simpleCompuation() { for ( int i = 0; i < count; i++ ) { double x = Math.random() * Math.random(); } } private static void computationWithObjCreationAndExecutors( ExecutorService es, List< Batch > batches ) throws InterruptedException { for ( Batch batch : batches ) { es.submit( batch ); } } private static class Batch implements Runnable { private final int computations; public Batch( final int computations ) { this.computations = computations; } @Override public void run() { int countdown = computations; while ( countdown-- > -1 ) { double x = Math.random() * Math.random(); } } }}
