Question

我使用我的Merge sort和Radix排序实现了一百万个随机正数，长度约为20位。合并排序显着，几乎是Radix排序的6倍。

我理解Radix排序的时间复杂度也取决于整数的位数，但我的合并实现在所有输入大小上击败了我的Radix实现。

我正在使用我自己的队列类，它在我的基数排序中具有常量时间push（）和pop（）。我在合并排序中使用数组。这与此有关吗？

public static void RadixSort(long arr[]) {
            //Using 10 queues for each digit from 0-9.
    Queue q[] = new Queue[10];

    for (int i = 0; i < 10; i++)
        q[i] = new Queue();

    boolean allNumbersNotBucketed = true;
    long divisor = 1;

    while (allNumbersNotBucketed) {
        allNumbersNotBucketed = false;

        for (int i = 0; i < arr.length; i++) {
            long digit = (arr[i] / divisor) % 10;


            q[(int) digit].enqueue(arr[i]);
                            //Put number into appropriate queue.

            if(digit > 0) allNumbersNotBucketed = true;
        }
        int pos = 0;
        divisor *= 10;
        for (int i = 0; i < 10; i++) 
            while (!q[i].isEmpty())
                      arr[pos++] = q[i].dequeue();
            //Put queue contents back into array
    }
}

这是合并排序

public static void mergeSort(long[] a) {
    long[] tmp = new long[a.length];
    mergeSort(a, tmp, 0, a.length - 1);
}

private static void mergeSort(long[] a, long[] tmp, int left, int right) {
    if (left < right) {
        int center = (left + right) / 2;
        mergeSort(a, tmp, left, center);    //Divide 0 to middle
        mergeSort(a, tmp, center + 1, right); // Divide middle to center
        merge(a, tmp, left, center + 1, right); //Merge sorted lists
    }
}

private static void merge(long[] a, long[] tmp, int left, int right,
        int rightEnd) {
    long leftEnd = right - 1;
    int k = left;
    long num = rightEnd - left + 1;

    //Put smallest element into tmp while both lists
    //are non empty.
    while (left <= leftEnd && right <= rightEnd)
        if (a[left] < a[right])
            tmp[k++] = a[left++];
        else
            tmp[k++] = a[right++];

    while (left <= leftEnd)
        // Copy rest of first half
        tmp[k++] = a[left++];

    while (right <= rightEnd)
        // Copy rest of right half
        tmp[k++] = a[right++];

    // Copy tmp back
    for (long i = 0; i < num; i++, rightEnd--)
        a[rightEnd] = tmp[rightEnd];
}

修改我愚蠢地使用LinkedList样式队列。我将其更改为使用本机数组，现在合并排序的速度仅为之前的6倍。即使对于只有10位数的数字，合并排序仍然更快。我猜BigO常数在这里发挥作用。对push（）和pop（）的数百万次函数调用也可以归咎于此。

mergesort运行速度比基数排序快

0 个答案: