Java Programming

I'm looking for a faster implementation of BigDecimal than java.math.BigDecimal as I need the higher precision (at least MathContext.DECIMAL128) and of course stay pure Java (so no JNI).

I found a (rather old) IBM site with a IBM Version of BigDecimal being several times faster. But at least with the latest JDK 1.6, this doesn't seem to be true anymore. I also tested the eval version of ArciMath but the latest news on the homepage is alread 4.5 years old and I don't know if it's compatible with the latest Java 6.

         System.out.println("Test: BigDecimal performance");
        final long iterations = 10000000;
        
        long t = System.currentTimeMillis();
        double d = 123.456;
        for (int i = 0; i < iterations; i++) {
            final double b = d * System.currentTimeMillis() * System.currentTimeMillis();
        }
        System.out.println("double: "+(System.currentTimeMillis() - t));

        
        t = System.currentTimeMillis();
        BigDecimal bd = new BigDecimal("123.456");
        for (int i = 0; i < iterations; i++) {
            final BigDecimal b = bd.multiply(new BigDecimal(""+System.currentTimeMillis())).add(new BigDecimal(""+System.currentTimeMillis()));
        }
        System.out.println("java.math.BigDecimal: "+(System.currentTimeMillis() - t));
        
        t = System.currentTimeMillis();
        com.ibm.math.BigDecimal bd2 = new com.ibm.math.BigDecimal("123.456");
        for (int i = 0; i < iterations; i++) {
            final com.ibm.math.BigDecimal b = bd2.multiply(new com.ibm.math.BigDecimal(""+System.currentTimeMillis())).add(new com.ibm.math.BigDecimal(""+System.currentTimeMillis()));
        }
        System.out.println("com.ibm.math.BigDecimal: "+(System.currentTimeMillis() - t));
        
        t = System.currentTimeMillis();
        be.arci.math.BigDecimal bd3 = new be.arci.math.BigDecimal("123.456");
        for (int i = 0; i < iterations; i++) {
            final be.arci.math.BigDecimal b = bd3.multiply(new be.arci.math.BigDecimal(""+System.currentTimeMillis())).add(new be.arci.math.BigDecimal(""+System.currentTimeMillis()));
        }
        System.out.println("be.arci.math.BigDecimal: "+(System.currentTimeMillis() - t));

Output:

Test: BigDecimal performance
double: 1485
java.math.BigDecimal: 40967
com.ibm.math.BigDecimal: 43748
be.arci.math.BigDecimal: 33373

I read that BigDecimal got improved in JDK 1.5/1.6. But it's still so very much slower when I change from doubles to BigDecimal (with MathContect.DECIMAL128). Are there any more BigDecimal implementations that are faster than java.math.BigDecimal?