A DESCRIPTION OF THE REGRESSION :
 There are regressions on doubles and ints expressions.


REPRODUCIBLE TESTCASE OR STEPS TO REPRODUCE:
import java.io.*;


public class Speed
{
    static class Evaluator
    {
       public final double eval(double i)
        { return (i * 0.123456) / (i + 1.0); }
    }
    
    Object obj; // used for isA tests
    static double r;

    public static void main(String[] args)
    { Speed sp = new Speed(args); }

    public Speed(String[] args)
    {
        obj = this;
        int ns = args.length;
        int niters = 3;
        for (int i = 0; i < ns; i++)
        {
            try
            {
                niters = Integer.parseInt(args[i]);
            } catch (Exception ex)
            {}
        }
        run(niters);
    }

    static void printTime(String mes, long time, double v)
    {
        System.out.print("    Testing " + mes + " : ");
        System.out.println( (time / 1000.0) + " s.");
        if (v > 0.0)
            System.out.println("      result = " + v);
        System.out.println();
    }

    public void run(int idx)
    {
        final int ncycle = idx;
        long start = System.currentTimeMillis();
        int k, i, n = (int) 1E7;
        for (k = 1; k <= ncycle; k++)
        {
            System.out.println();
            System.out.println(" --------------------------" + Integer.toString(k)
                    + "--------------------------------");
            System.out.println();
            System.out.println();
            
            long iter_time = 0 ;
            // Test # 2
            iter_time += testDoubleOpsViaFunctionCall((int) 1E8);
            // Test # 3
            iter_time += testInlinedDoubleOps((int) 1E8);
            // Test # 4
            iter_time += testIntOperations((int) 1E9);

        }
    }

    // double operation with function call
    long testDoubleOpsViaFunctionCall(int n)
    {
        double r = 0.0;
        Evaluator evaluator = new Evaluator();
        long start = System.currentTimeMillis();
        for (int i = 1; i <= n; i++)
            r += evaluator.eval((double) i);
        long time = System.currentTimeMillis() - start ;
        printTime (n + " double moltiplications via function call", time, r);
        return time ;
    }

    // Test # 3 -B
    long testInlinedDoubleOps(int n)
    {
        int i;
        double r = 0.0;

        long start = System.currentTimeMillis();
        for (i = 1; i <= n; i++)
            r += (i * 0.123456) / (i + 1.0);
        long time = System.currentTimeMillis() - start;
        printTime(n + " inlined double moltiplications", time, r);
        return time ;
    }

    long testIntOperations(int n)
    {
        int i, a = 0, b = 2;
        long start = System.currentTimeMillis();
        for (i = 1; i <= n; i++)
            a += a - i * 3;
        long time = System.currentTimeMillis() - start;
        printTime(n + " int moltiplications + additions", time,
                (double) a);
        return time ;
    }

}


RELEASE LAST WORKED:
5.0 Update 6

RELEASE TEST FAILS:
mustang-beta

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
Java HotSpot(TM) Server VM (build 1.5.0_06-b05, mixed mode)

After several warm-up iterations ...

    Testing 100000000 double moltiplications via function call : 1.437
      result = 1.2345597781357473E7

    Testing 100000000 inlined double moltiplications : 1.453 s.
      result = 1.2345597781357473E7

    Testing 1000000000 int moltiplications + additions : 0.969 s.
ACTUAL -

Java(TM) 2 Runtime Environment, Standard Edition (build 1.6.0-beta2-b73)
Java HotSpot(TM) Server VM (build 1.6.0-beta2-b73, mixed mode)

After several warm-up iterations ...

 Testing 100000000 double moltiplications via function call : 1.765 s.
   result = 1.2345597781357473E7

 Testing 100000000 inlined double moltiplications : 1.766 s.
   result = 1.2345597781357473E7

 Testing 1000000000 int moltiplications + additions : 2.172 s.

Release Regression From : 5.0u6
The above release value was the last known release where this 
bug was known to work. Since then there has been a regression.
Posted Date : 2006-03-21 21:48:44.0

N/A

The function IdealLoopTree::compute_profile_trip_cnt computes trip counts based on branch predictions made in the parser.  For historical reasons, the counts there are wrongly scaled by FreqCountInovocations, which is 1500.
Posted Date : 2006-03-22 19:57:31.0