[kaffe] RE: JamVM on ARM (XScale)

[Dalibor Topic]

Woloszynski, Charles wrote:
> I have heard that floating point emulation in the kernel is *much*
> slower than replacing the code with calls to a floating point library.
> I thought I saw someone talking about a floating point package that they
> used to avoid the kernel traps to get improved performance. 
> 
> I am trying to decide bewtween Kaffe and JamVM for our next Xscale
> processor project.  Any feedback on why to use over the other?

Hi Charles,

I'd suggest to try both, and to pick the one that works for you.

Kaffe's jit may offer you some performance kicks[1], otoh it may need a 
bit of work to run properly on xscale. There is a jit regression test 
suite that you could use to try to get the bugs squished, though. See 
test/internal directory for details and FAQ.debugging.

cheers,
dalibor topic

[1] In general factor 10 over the naive interpreter.