Benchmark
cern.jet.random.engine

Class Benchmark



  • public class Benchmarkextends Object
    Benchmarks the performance of the currently provided uniform pseudo-random number generation engines.

    All distributions are obtained by using a uniform pseudo-random number generation engine. followed by a transformation to the desired distribution. Therefore, the performance of the uniform engines is crucial.

    Comparison of uniform generation engines

    Name Period

    Speed
    [# million uniform random numbers generated/sec]
    Pentium Pro 200 Mhz, JDK 1.2, NT

    MersenneTwister 219937-1 (=106001) 2.5
    Ranlux (default luxury level 3) 10171 0.4
    Ranmar 1043 1.6
    Ranecu 1018 1.5
    java.util.Random.nextFloat() ? 2.4

    Note: Methods working on the default uniform random generator are synchronized and therefore in current VM's slow (as of June '99). Methods taking as argument a uniform random generator are not synchronized and therefore much quicker. Thus, if you need a lot of random numbers, you should use the unsynchronized approach:

    Example usage:

     edu.cornell.lassp.houle.RngPack.RandomElement generator; generator = new cern.jet.random.engine.MersenneTwister(new java.util.Date()); //generator = new edu.cornell.lassp.houle.RngPack.Ranecu(new java.util.Date()); //generator = new edu.cornell.lassp.houle.RngPack.Ranmar(new java.util.Date()); //generator = new edu.cornell.lassp.houle.RngPack.Ranlux(new java.util.Date()); //generator = makeDefaultGenerator(); for (int i=1000000; --i >=0; ) {    double uniform = generator.raw();    ... } 
    See Also:
    cern.jet.random
    • Method Detail

      • benchmark

        public static void benchmark(int times)
        Benchmarks raw() for various uniform generation engines.
      • main

        public static void main(String[] args)
        Tests various methods of this class.
      • test

        public static void test(int size,        RandomEngine randomEngine)
        Prints the first size random numbers generated by the given engine.

SCaVis 1.8 © jWork.org