Kaffe and scimark 2.0

[Edouard G. Parmelan]

Mark Christiaens wrote:

> We're considering to use Kaffe for a research project.  So I thought I'd
> give it a little testrun.  I downloaded the rpm on the Kaffe website and
> ran scimark 2.0 (http://math.nist.gov/scimark/) on a Celeron 500 MHz.  
> For some reason, it ran several hours without ever producing a result.
> On my Linux box at home (an old PII 233 MHz) using the IBM JDK 1.1.8, this
> benchmark runs in half a minute. 
> 
> What's wrong?  I know Kaffe isn't as fast as the commercial distributions
> but this seems a bit strange to me.  Could someone run scimark 2.0 on
> their machine and try it out?

With your report, I discover a bug with double and float on ix86 in
Kaffe.  Thanks.

I have not yet commit my fix in CVS Tree, but this is the current result
on my PII 350 MHz:

Composite Score: 8.3283354340737521
FFT (1024): 5.3105534220703872
SOR (100x100):   19.456330676365077
Monte Carlo : 1.8220261224138588
Sparse matmult (N=1000, nz=5000): 8.4418796405815346
LU (100x100): 6.6108873089378983

java.vendor: Transvirtual Technologies, Inc.
java.version: 1.0.5
os.arch: i686
os.name: Linux
os.version: 2.2.16
-- 
Edouard G. Parmelan
http://egp.free.fr