CODEWITHSUNDEEP
java
yu xiaofei
yu xiaofei

Reputation: 55

Why `a[i] = b; i++` perform better than `a[i++] = b`?

I was reading the source code of DualPivotQuicksort Algorithm in java these days. I find an interesting comment in code. It says:

/*
 * Here and below we use "a[i] = b; i++;" instead
 * of "a[i++] = b;" due to performance issue.
 */
a[less] = ak;
++less;

I just don't know why. Can anybody help me?

Upvotes: 1

Views: 363

Answers (1)

user4081625
user4081625

Reputation:

I decided to compare the execution times for 2^26 operations for each option. I iterated this 4096 times and got and average for each option. 

import java.util.ArrayList;

public class TestClass
{
public static void main (String[] args)
{
    long startTime, endTime;
    final int twoToPower26 = 67108864;
    int[] a1 = new int[twoToPower26];
    int[] a2 = new int[twoToPower26];
    int i = 0;
    int b = 1;
    ArrayList<Long> optionATimes = new ArrayList<Long>();
    ArrayList<Long> optionBTimes = new ArrayList<Long>();

    for (int k = 0; k < 4096; k++)
    {
        i = 0;
        //OPTION A: "a[i] = b; i++;"
        startTime = System.currentTimeMillis();
        while (i < twoToPower26)
        {
            a1[i] = b; 
            i++;
        }
        endTime = System.currentTimeMillis();
        optionATimes.add(endTime - startTime);

        i = 0;
        //OPTION B: "a[i++] = b;"
        startTime = System.currentTimeMillis();
        while (i < twoToPower26)
        {
            a2[i++] = b; 
        }
        endTime = System.currentTimeMillis();
        optionBTimes.add(endTime - startTime);
        System.out.println(k);
    }

    double averageTimeOptionA = calculateAverage(optionATimes);
    double averageTimeOptionB = calculateAverage(optionBTimes);
    System.out.println("OPTION A average time for 2^26 operations: " + averageTimeOptionA);
    System.out.println("OPTION B average time for 2^26 operations: " + averageTimeOptionB);
}

public static double calculateAverage(ArrayList<Long> times) {
    Long sum = (long) 0;
    for (Long time : times) {
        sum += time;
    }
    return sum.doubleValue() / times.size();
}
}

The results:
OPTION A average time for 2^26 operations: 105.9140625 ms
OPTION B average time for 2^26 operations: 105.846923828125 ms

The difference in average time for each option is only about 67 nano seconds over 2^26 iterations and in this case OPTION B was slightly faster.

I feel this shows with almost certainty that there is no difference in performance. Most likely the compiler interprets each option in the same.

Upvotes: 5

Related Questions