Speed of the Math class

Question

If I am performing a lot of calculations is it best to use a Math class function or write it yourself (best in terms of speed not accuracy).
e.g. Math.pow(a-b,2) OR (a-b) * (a-b)
I am grateful for any opinions.

Answer 1

Your version would allow the compiler to use the fact that the expression is completely knowable at compilation time, providing for a trivial solution for the in-line multiplication approach.
The version below avoids thatpublic class Test {
    public static void main(String[] args) {
        long start = 0;
        long end = 0;
        long powtime = 0;
        long multtime = 0;
        double x = Double.parseDouble(args[0]);
        double y = Double.parseDouble(args[1]);
        System.out.println("Starting the pow test.");
        start = System.currentTimeMillis();
        for (int i = 0; i < 10000000; i++) {
            double value = (x - y) * (x - y);
        end = System.currentTimeMillis();
        multtime = end - start;
        System.out.println(
            "By using multiplication it took: " + ((double) multtime / 1000) + " seconds.");
        start = System.currentTimeMillis();
        for (int j = 0; j < 10000000; j++) {
            double value = Math.pow((x - y), 2);
        end = System.currentTimeMillis();
        powtime = end - start;
        System.out.println(
            "By using Math.pow() it took: " + ((double) powtime / 1000) + " seconds.");
}With this I get a 6:1 (1.272 to 0.200) ratio of Math.pow() to inline multiplication time. Still a fairly large ratio.
Playing with double, int and using increasing larger values of the exponent, the Math.pow() seems to be far slower that multiplication. Of course Math.pow() can handle complex exponents (2.3645463) and take roots while the simple multiplication approach cannot.
Chuck