[Mono-devel-list] Mono performance on SciMark

[Marcus]

I looked at the SciMark performance again, and even after eliminating the 
Synchronized attribute from Random.cs, the Mone Carlo test results are slow, 
even with -O=all. I did some profiling, and the slowdown seems to be here:

2055.898 4194302    0.000  SciMark2.Random::nextDouble()
  Callers (with count) that contribute at least for 1%:
     4194302  100 % SciMark2.MonteCarlo::integrate(int)

Where

        public double nextDouble()
        {
            int k;

            k = m[i] - m[j];
            if (k < 0)
                k += m1;
            m[j] = k;

            if (i == 0)
                i = 16;
            else
                i--;

            if (j == 0)
                j = 16;
            else
                j--;

            if (haveRange)
                return left + dm1 * (double) k * width;
            else
                return dm1 * (double) k;
        }

http://math.nist.gov/scimark