Re: Gcj binaries performance (arm-linux)

> > I'm quite surprised of this result, I thought that a binary compiled with gcj would be much faster than a java virtual machine.
>
> So would I.
>
> > I've also tried to pass some options (e.g. -march=, -mtune, -mcpu...) when compiling with gcj but results are not changing.
> >
> > Any idea to improve performance?
>
> Without seeing your code, it's impossible to say.  I don't want
> to speculate.
>
> Andrew.
>


Hi Andrew,
thanks for your interest. Here is the code:

public final class MyPanel {


  private static final int NUM_ELEMENTS = 1000;
  public static final int NUM_LOOP = 100;

  public MyPanel(){}

  public static void main(String args[]) {
    MyPanel mp = new MyPanel();
    mp.sortArray();
  }

  private void sortArray() {
    System.out.println("Test \"Bubble Sort\" (Num loop: " + NUM_LOOP + ")");
    int[] array = new int[NUM_ELEMENTS];
    int l = 0, i = 0;

    long sysTime = System.currentTimeMillis();

    for (l = 0; l < NUM_LOOP; l++) {
      for (i = 0; i < NUM_ELEMENTS; i++) {
        array[i] = NUM_ELEMENTS - i;
      }
      BubbleSort(array, NUM_ELEMENTS);
      if(l == 0)
        System.out.println("First loop time: " + (System.currentTimeMillis() - sysTime));
    }
    System.out.println("Total time elapsed: " + (System.currentTimeMillis() - sysTime));
  }

  private void BubbleSort(int array[], int elemN) {
    int i, tmp;
    int alto = elemN;

    while (alto > 0) {
      for (i = 0; i < alto - 1; i++) {
        if (array[i] > array[i + 1]) {
          tmp = array[i];
          array[i] = array[i + 1];
          array[i + 1] = tmp;
        }
      }
      alto--;
    }
  }

}