Why is my mergesort slower than my quicksort?

They're both the same time complexity but when I run my merge sort on a randomly generated linked list with 100,000 entries:

public LinkedList<Integer> linkedListSort(LinkedList<Integer> list) {
    if (list.size() <= 1) return list;
    LinkedList<Integer> left = new LinkedList<Integer>();
    LinkedList<Integer> right = new LinkedList<Integer>();
    int middle = list.size()/2;
    for (int i = 0; i < middle; i++) {
        left.add((int)list.get(i)); steps++;
    }
    for (int i = middle; i < list.size(); i++) {
        right.add((int)list.get(i)); steps++;
    }
    left = linkedListSort(left);
    right = linkedListSort(right);
    return merge(left, right);
}

public LinkedList<Integer> merge(LinkedList<Integer> left, LinkedList<Integer> right) {
    LinkedList<Integer> result = new LinkedList<Integer>();
    while (!(left.isEmpty()) && !(right.isEmpty())) {
        steps++;
        if ((int)left.peekFirst() <= (int)right.peekFirst()) {
            result.add(left.poll());
        } else {
            result.add(right.poll());
        }
    }
    while (!(left.isEmpty())) {result.add(left.poll()); steps++;}
    while (!(right.isEmpty())) {result.add(right.poll()); steps++;}
    return result;
}

It's a lot slower than my quick sort which is:

public String arraySort(int[] array, int startIndex, int endIndex, int steps) {
    int leftIndex = startIndex;
    int rightIndex = endIndex;
    int pivot = array[(leftIndex + rightIndex) / 2];
    while (leftIndex <= rightIndex) {
        steps++;
        //search for an element with a higher value than the pivot, lower than it
        while (array[leftIndex] < pivot) {steps++; leftIndex++;}
        //search for an element with a lower value than the pivot, higher than it
        while (array[rightIndex] > pivot) {steps++; rightIndex--;}
        //check the left index hasn't overtaken the right index
        if (leftIndex <= rightIndex) {
            //swap the elements
            int holder = array[leftIndex];
            array[leftIndex] = array[rightIndex]; 
            array[rightIndex] = holder;
            leftIndex++; rightIndex--;
        }
    }
    if (leftIndex < endIndex) arraySort(array, leftIndex, endIndex, steps); 
    if (rightIndex > startIndex) arraySort(array, startIndex, rightIndex, steps);
    return "Quicksort on an unsorted array took " + steps + " steps.";
}

What's the reason for this? Is my quicksort/mergesort not what it should be or is it that mergesort performs badly on a linked list with a large amount of random numbers? Or something else?

Thanks!


You've implemented a version of quicksort that's doing everything "in-place" while your mergesort copies the content of left/right upon every recursive call (and same thing with merge() ). That's probably the major reason for the differences.

Second, like Luiggi mentioned in the comments above - how do you do your benchmarking ? do you get a proper JVM warmup? do you run enough cycles and take average ? proper benchmarking with the JVM can be tricky: if you don't have experience with it, better look for a micro-benchmarking framework and use it!

链接地址: http://www.djcxy.com/p/31584.html

上一篇: 用于排序零件Mergesort的QuickSort?

下一篇: 为什么我的mergesort比我的quicksort慢?