踩踏次数增加导致循环持续时间更长 Download


我有一种方法，其中的逻辑需要40毫秒左右的时间

当我多次调用此方法（并行）时-每个循环持续时间需要更多时间（很多）。

例如：

1次通话-约40毫秒（平均）

5次通话-约63毫秒（平均）

10次通话-约120毫秒（平均）

100个呼叫-约1200ms（平均）

以此类推...

代码： 

int numberOfExecution = 100;
ExecutorService executor = Executors.newFixedThreadPool(numberOfExecution);
for (int i = 0; i < numberOfExecution; i++) {
    executor.execute(new Runnable() {
        @Override
        public void run() {
            MyObject obj1 = readBytesFromFile(1);
            MyObject obj2 = readBytesFromFile(2);
            MyObject obj3 = readBytesFromFile(3);
            MyObject obj4 = readBytesFromFile(4);


            final Map<Integer, MyObject> objects = new TreeMap<>();
            objects.put(0, obj1);
            objects.put(1, obj2);
            objects.put(2, obj3);
            objects.put(3, obj4);

            Foo foo = new Foo(objects);

            long startTime = System.currentTimeMillis();
            Thread.sleep(40); // sleep for 40ms to demonstrate my 40ms logic
            long endTime = System.currentTimeMillis();
            float durationMs = endTime - startTime;

            System.out.println("Foo.bar() duration: " + durationMs);
        }
    });
}

executor.shutdown();
while (!executor.isTerminated()) { }

控制台：

Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 43.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 41.0
Foo.bar() duration: 48.0
Foo.bar() duration: 62.0
Foo.bar() duration: 48.0
Foo.bar() duration: 48.0
Foo.bar() duration: 48.0
Foo.bar() duration: 48.0
Foo.bar() duration: 49.0
Foo.bar() duration: 48.0
Foo.bar() duration: 49.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 41.0
Foo.bar() duration: 42.0
Foo.bar() duration: 49.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 49.0
Foo.bar() duration: 74.0
Foo.bar() duration: 40.0
Foo.bar() duration: 70.0
Foo.bar() duration: 40.0
Foo.bar() duration: 50.0
Foo.bar() duration: 43.0
Foo.bar() duration: 40.0
Foo.bar() duration: 43.0
Foo.bar() duration: 56.0
Foo.bar() duration: 63.0
Foo.bar() duration: 45.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 43.0
Foo.bar() duration: 43.0
Foo.bar() duration: 48.0
Foo.bar() duration: 41.0
Foo.bar() duration: 58.0
Foo.bar() duration: 59.0
Foo.bar() duration: 55.0
Foo.bar() duration: 42.0
Foo.bar() duration: 40.0
Foo.bar() duration: 42.0
Foo.bar() duration: 70.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 75.0
Foo.bar() duration: 40.0
Foo.bar() duration: 43.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 67.0
Foo.bar() duration: 59.0
Foo.bar() duration: 40.0
Foo.bar() duration: 57.0
Foo.bar() duration: 42.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 43.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 41.0
Foo.bar() duration: 42.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 42.0
Foo.bar() duration: 42.0
Foo.bar() duration: 41.0
Foo.bar() duration: 45.0
Foo.bar() duration: 44.0
Foo.bar() duration: 40.0
Foo.bar() duration: 42.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 40.0
Foo.bar() duration: 45.0
Foo.bar() duration: 41.0
Foo.bar() duration: 40.0
Foo.bar() duration: 44.0
Foo.bar() duration: 41.0
Foo.bar() duration: 42.0
Foo.bar() duration: 40.0
Foo.bar() duration: 42.0
Foo.bar() duration: 41.0
Foo.bar() duration: 42.0
Foo.bar() duration: 42.0
Foo.bar() duration: 43.0
Foo.bar() duration: 43.0

在MacBook Pro i9上运行

我想念什么？

1 个答案:

答案 0 :(得分：1)

机器中的内核数是固定的。随着我们增加需要并行执行的线程数量，将在线程之间进行上下文切换。

请检查https://www.javamex.com/tutorials/threads/context_switch.shtml。

此外，线程之间可能存在资源争用。可以并行运行的最大线程数取决于计算机核心。 简而言之，仅增加线程数将不会提高代码执行性能。



Similar searches