[wellylug] Java 32 Bit/ Java 64 Bit Performance

[Daniel Pittman]

"Gerald.Roehrbein" <Gerald.Roehrbein at oraforecast.com> writes:

> I've tested Java with 32/64 Bit Linux and 32/64 Bit data model.
> Results:
>
> a) 32 Bit Linux + 32 Bit Java memory model gave best performance
>    especially with memory allocation. Test program took 45 minutes.
>
> b) 64 Bit Linux + 32 Bit Java memory model gave bad performance. Test
>    program took 90 minutes.
>
> c) 64 Bit Linux + 64 Bit Java memory model gave good performance. Test
>    program took 70 minutes.
>
> The reason for the difference was the duration of some types of memory
> allocations. The application reads 70.000 records from a database into a
> java clients memory.

That must be the most amazingly badly written benchmarking program out
there; are you sure that your differences are not due to the box (or JVM)
thrashing itself to death under the memory load?

This is a reasonable summary of the issues in hard:
http://stackoverflow.com/questions/559433/benefits-drawbacks-to-running-64-bit-jvm-on-64-bit-linux-server


You really shouldn't see such a dramatic performance difference between
the two models at a low level, though, especially given that the 64-bit
kernel enables a number of features that *reduce* computational
overhead.

[...]

> My question:
> I'm interested in a complete Java benchmark for 32/64 Bit Linux.

What for?  Seriously, are you trying to benchmark the performance of
your application, of the difference between 32 and 64 bit memory models
in terms of performance, of the JIT compiler in the various JVM
implementations, or something else?