Thread: Clearing up some Java speed myths

This dosn't prove speed, considering that the application is already running the first test before the second that means there could be left over memory and other leaks.

Originally Posted by Java`

Memory usage doesn't affect speed.. If you don't believe me, switch it all backwards.

If you're swapping, it most definitely will result in major speed loss.

Originally Posted by Java`

Heres a few tests to see which of the two things is faster.

Myth: For-each loops are faster then regular For loops

Code:

	public static void runForLoopBenchMarks() {

		double time;
		long start;
		int tests = 10;

		int[] testArray = new int[tests];

		start = System.nanoTime();

		for(int a = 0; a < testArray.length; a++)
		for(int b = 0; b < testArray.length; b++)
		for(int c = 0; c < testArray.length; c++)
		for(int d = 0; d < testArray.length; d++)
		for(int e = 0; e < testArray.length; e++);

		time = System.nanoTime() - start;
		System.out.println("Time for test 1: " + time);


		start = System.nanoTime();

		for(int a : testArray)
		for(int b : testArray)
		for(int c : testArray)
		for(int d : testArray)
		for(int e : testArray);

		time = System.nanoTime() - start;
		System.out.println("Time for test 2: " + time);
	}

Code:

Time for test 1: 1507376.0
Time for test 2: 158769.0

Result: Standard for loops are faster then for-each loops.


Myth: Using a variable is the same speed as using a number

Code:

	public static void runForLoopBenchMarks() {

		double time;
		long start;
		int tests = 10;

		int[] testArray = new int[tests];

		start = System.nanoTime();

		for(int a = 0; a < 10; a++)
		for(int b = 0; b < 10; b++)
		for(int c = 0; c < 10; c++)
		for(int d = 0; d < 10; d++)
		for(int e = 0; e < 10; e++);

		time = System.nanoTime() - start;
		System.out.println("Time for test 1: " + time);


		start = System.nanoTime();

		for(int a = 0; a < testArray.length; a++)
		for(int b = 0; b < testArray.length; b++)
		for(int c = 0; c < testArray.length; c++)
		for(int d = 0; d < testArray.length; d++)
		for(int e = 0; e < testArray.length; e++);

		time = System.nanoTime() - start;
		System.out.println("Time for test 2: " + time);
	}

Code:

Time for test 1: 1207958.0
Time for test 2: 168002.0

Result: Using an actual number in place of a variable will make the code faster


Myth: Switch statements are the fastest way to check values (Below are sevaral ways compared, but note the array[] index way would be faster in the case of there being more possible number values)

Code:

	public static void runIntegerBenchMarks() throws Exception {
		Test t = new Test();
		double time;
		long start, tests = 1000;
		int arg = 3;


		start = System.nanoTime();
		for(int i = 0; i < tests; i++) {

			if(arg == 1) {
				t.choice1();

			} else if(arg == 2) {
				t.choice2();

			} else if(arg == 3) {
				t.choice3();
			}
		}
		time = System.nanoTime() - start;
		System.out.println("Time for test 1: " + time);
		System.out.println("time for one   : " + (double)(time / tests) + "\n\n");


		start = System.nanoTime();
		for(int i = 0; i < tests; i++) {

			switch(arg) {
				case 1: t.choice1(); break;
				case 2: t.choice2(); break;
				case 3: t.choice3(); break;
			}
		}
		time = System.nanoTime() - start;
		System.out.println("Time for test 2: " + time);
		System.out.println("time for one   : " + (double)(time / tests) + "\n\n");


		start = System.nanoTime();
		for(int i = 0; i < tests; i++) {
			t.getClass().getMethod("choice" + arg).invoke(t);
		}
		time = System.nanoTime() - start;
		System.out.println("Time for test 3: " + time);
		System.out.println("time for one   : " + (double)(time / tests) + "\n\n");

		Option[] options = new Option[4];

		options[1] = new One();
		options[2] = new Two();
		options[3] = new Three();

		start = System.nanoTime();
		for(int i = 0; i < tests; i++) {
			options[arg].choice();
		}

		time = System.nanoTime() - start;
		System.out.println("Time for test 4: " + time);
		System.out.println("time for one   : " + (double)(time / tests) + "\n\n");

	}

Code:

Time for test 1: 48826.0
time for one   : 48.826

Time for test 2: 35589.0
time for one   : 35.589

Time for test 3: 2.0072777E7
time for one   : 20072.777

Time for test 4: 40004.0
time for one   : 40.004

Result: Switch statements are the fastest way for situations with very small possible choices, but otherwise array[] indexes would be the fastest hands down, because the time takes never increases with possible awnsers.


Myth: List.get(int) method is slower then using an array

Code:

	public static void testListGet() {

		double time;
		long start;
		int tests = 10;

		List<Object> list = new ArrayList<Object>();
		Object[] arry = new Object[100];

		for(int i = 0; i < arry.length; i++) {
			list.add(new Object());
			arry[i] = new Object();
		}

		start = System.nanoTime();

		for(int i = 0; i < tests; i++) {
			Object o = list.get(list.size()-2);
		}

		time = System.nanoTime() - start;
		System.out.println("Time for test 1: " + time);

		start = System.nanoTime();

		for(int i = 0; i < tests; i++) {
			Object o = arry[arry.length-2];
		}

		time = System.nanoTime() - start;
		System.out.println("Time for test 2: " + time);
	}

Code:

Time for test 1: 7839.0
Time for test 2: 1744.0

Result: Using an array is a LOT faster than using a list






Some misc stuff:

Code:

interface Option {

	public void choice();
}

class One implements Option {

	@Override
	public void choice() {}
}

class Two implements Option {

	@Override
	public void choice() {}
}

class Three implements Option {

	@Override
	public void choice() {}
}

class Test {

	public void choice1() {}
	public void choice2() {}
	public void choice3() {}
}

I will add more to this later.

looks like someone goes to java forums alot and copy and paste