Question

Is it possible to have Groovy NOT auto-box every single long I have in my application?

The following code ...

public class SumTest {
    public static void main(String[] args) {
        long[] longs = new long[100000000];

        long nanoStart = System.nanoTime();
        long counter = 1;
        for (int i = 0; i < longs.length; i++) {
            longs[i] = counter++;
        }

        double msPop = (System.nanoTime() - nanoStart) / 1000000d;
        System.out.println("Time taken, population: " + msPop + "ms.");

        nanoStart = System.nanoTime();
        long sum = 0;
        for (int i = 0; i < longs.length; i++) {
            sum += longs[i];
        }
        double msSum = (System.nanoTime() - nanoStart) / 1000000d;
        System.out.println("Time taken, sum: " + msSum + "ms, total: " + (msPop + msSum) + "ms");
        System.out.println(" (sum: " + sum + ")");
    }
}

... exhibits vastly different run times when renamed to '.groovy' from '.java':

Java:

Time taken, population: 94.793746ms.
Time taken, sum: 65.172605ms, total: 159.966351ms
 (sum: 5000000050000000)

Groovy:

Time taken, population: 2233.995965ms.
Time taken, sum: 2203.64302ms, total: 4437.638985ms
 (sum: 5000000050000000)

.. which is a ~30x difference.

The situation is compounded when I stick the long inside an object (as is the case in my real code):

public class SumTest {
    static class Holder {
        long l;
        Holder(long l) { this.l = l; }
        long getL() { return l; }
    }

    public static void main(String[] args) {
        Holder[] longs = new Holder[100000000];

        long nanoStart = System.nanoTime();
        long counter = 1;
        for (int i = 0; i < longs.length; i++) {
            longs[i] = new Holder(counter++);
        }

        double msPop = (System.nanoTime() - nanoStart) / 1000000d;
        System.out.println("Time taken, population: " + msPop + "ms.");

        nanoStart = System.nanoTime();
        long sum = 0;
        for (int i = 0; i < longs.length; i++) {
            sum += longs[i].getL();
        }
        double msSum = (System.nanoTime() - nanoStart) / 1000000d;
        System.out.println("Time taken, sum: " + msSum + "ms, total: " + (msPop + msSum) + "ms");
        System.out.println(" (sum: " + sum + ")");
    }
}

Run times (note that I run with -Xms16384M -Xmx16384M here):

Java

Time taken, population: 1083.784927ms.
Time taken, sum: 180.518991ms, total: 1264.3039179999998ms
 (sum: 5000000050000000)

Groovy:

Time taken, population: 9816.007447ms.
Time taken, sum: 8685.506864ms, total: 18501.514311ms
 (sum: 5000000050000000)

.. which on the total is ~15 times faster, but the most important difference comes with actual use of these object (represented by the summing): ~50x.

Can this somehow be fixed? Can I coax Groovy to not autobox every single operation involving primitives, when the operations in question only concerns primitives and primitive operations?

Was it helpful?

Solution

(Wow! Strange how thoroughly formulating a question nearly instantly leads to an answer?!)

Fix: Write @CompileStatic on top of the class.

Time taken, population: 1562.978726ms.
Time taken, sum: 183.388353ms, total: 1746.367079ms
 (sum: 5000000050000000)
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
scroll top