/**
     * simulates throwing two dice
     * @return the sum of the two die
     */
    private static int twoDiceThrows() {
        final int dieA = ThreadLocalRandom.current().nextInt(5) + 1;
        final int dieB = ThreadLocalRandom.current().nextInt(5) + 1;
        return dieA + dieB;
    }

    private static Map<Integer, Double> serialDiceRolls(long N) {
        double fraction = 1.0 / N;
        return LongStream.range(0, N)
                        .mapToObj((i) -> twoDiceThrows())
                        .collect(groupingBy(side -> side,
                                            summingDouble(n -> fraction)));
    }

    private static Map<Integer, Double> parallelDiceRolls(long N) {
        double fraction = 1.0 / N;
        return LongStream.range(0, N)
                         .parallel() // *
                         .mapToObj((i) -> twoDiceThrows())
                         .collect(groupingBy(side -> side,
                                             summingDouble(n -> fraction)));
    }

    private static long timeIt(Runnable toTime) {
        final long start = System.currentTimeMillis();
        toTime.run();
        final long stop = System.currentTimeMillis();
        return (stop - start);
    }

    @Test
    public static void main(String[] args) throws Exception {

        // warm thread locals
        parallelDiceRolls(0);
        parallelDiceRolls(0);

        final List<Long> sizesToTry = ImmutableList.of(100L, 1_000L, 10_000L, 100_000L, 1_000_000L, 10_000_000L, 100_000_000L);

        for (long n : sizesToTry) {
            final long parallelExecutionTime = timeIt(() -> parallelDiceRolls(n));
            final long serialExecutionTime = timeIt(() -> serialDiceRolls(n));
            System.out.println(String.format("For simulation size %9d, serial took %4d millis and parallel took %4d millis",
                                             n, serialExecutionTime, parallelExecutionTime));

        }

      /*
For simulation size       100, serial took    1 millis and parallel took    2 millis
For simulation size      1000, serial took    2 millis and parallel took    2 millis
For simulation size     10000, serial took    5 millis and parallel took    4 millis
For simulation size    100000, serial took    5 millis and parallel took    8 millis
For simulation size   1000000, serial took   28 millis and parallel took   40 millis
For simulation size  10000000, serial took  181 millis and parallel took   45 millis
For simulation size 100000000, serial took 1878 millis and parallel took  577 millis
      */

    }