Question

我有两种不同排序的实现，InsertionSort和ShellSort。

他们如下：

插入排序：

for (int pos = 0; pos < arrayToBeSorted.length; pos++) {
    for (int secondMarker = pos; secondMarker > 0; secondMarker--) {
        int currentValue = arrayToBeSorted[secondMarker];
        int valueBeingCheckedAgainst = arrayToBeSorted[secondMarker - 1];
        if (currentValue > valueBeingCheckedAgainst) {
            break;
        }
        arrayToBeSorted[secondMarker] = arrayToBeSorted[secondMarker - 1];
        arrayToBeSorted[secondMarker - 1] = currentValue;
    }
}

希尔排序：

for (int gap = a.length / a.length; gap > 0; gap = (gap / 2)) {
    for (int i = gap; i < a.length; i++) {
        int tmp = a[i];
        int j = i;
        for (; j >= gap && tmp < (a[j - gap]); j -= gap) {
            a[j] = a[j - gap];
        }
        a[j] = tmp;
    }
}

我还有10个整数数组，它们包含32000个整数。我在这些类中调用静态sortArray方法之前得到的时间。结果如下：

对于InsertionSort.sortArray：

Solving array with: 32000 elements.
Time in milliseconds:264
Time in milliseconds:271
Time in milliseconds:268
Time in milliseconds:263
Time in milliseconds:259
Time in milliseconds:257
Time in milliseconds:258
Time in milliseconds:260
Time in milliseconds:259
Time in milliseconds:261

对于ShellSort：

Solving array with: 32000 elements.
Time in milliseconds:357
Time in milliseconds:337
Time in milliseconds:167
Time in milliseconds:168
Time in milliseconds:165
Time in milliseconds:168
Time in milliseconds:167
Time in milliseconds:167
Time in milliseconds:166
Time in milliseconds:167

那么他们之间有多大差异呢？它们基本上是相同的算法？

另外，ShellSort的前两次运行可能需要更长的时间，但其余的更快？

这是128000个元素的结果，再次插入InsertionSort：

Solving array with: 128000 elements.
Time in milliseconds:4292
Time in milliseconds:4267
Time in milliseconds:4241
Time in milliseconds:4252
Time in milliseconds:4253
Time in milliseconds:4248
Time in milliseconds:4261
Time in milliseconds:4260
Time in milliseconds:4333
Time in milliseconds:4261

希尔排序：

Solving array with: 128000 elements.
Time in milliseconds:5358
Time in milliseconds:5335
Time in milliseconds:2676
Time in milliseconds:2656
Time in milliseconds:2662
Time in milliseconds:2654
Time in milliseconds:2661
Time in milliseconds:2656
Time in milliseconds:2660
Time in milliseconds:2673

我确信我传递给方法的数组是完全相同的，而且它们是随机的。

Answer 1

在插入排序中，您会变得更复杂，

for (int pos = 0; pos < arrayToBeSorted.length; pos++) {
    for (int secondMarker = pos; secondMarker > 0; secondMarker--) {
        int currentValue = arrayToBeSorted[secondMarker];
        int valueBeingCheckedAgainst = arrayToBeSorted[secondMarker - 1];
        if (currentValue > valueBeingCheckedAgainst) {
            break;
        }
        arrayToBeSorted[secondMarker] = arrayToBeSorted[secondMarker - 1];
        arrayToBeSorted[secondMarker - 1] = currentValue;
    }
}

您从内部循环中读取数组中的值，并且前一个位置的值不小，您可以向数组写入两个值。

在shell排序中，

for (int i = gap; i < a.length; i++) {
    int tmp = a[i];
    int j = i;
    for (; j >= gap && tmp < (a[j - gap]); j -= gap) {
        a[j] = a[j - gap];
    }
    a[j] = tmp;
}

你读取了一次的值，在内部循环之外，并且在内部循环体中只有一次写入，在内部循环之后只写一次值。

这样效率更高，因此shell排序更快是可以理解的。两个第一个shell排序较慢可能是因为包装

for (int gap = a.length / a.length; gap > 0; gap = (gap / 2)) {

将JIT混淆了一段时间后才注意到gap可以替换为1并且包络循环被消除。

插入排序vs Shell按间隙大小排序= 1？

1 个答案: