MMU

[smorgan at almaden.ibm.com]

Jake Janovetz wrote:

> Isn't Java an appealing alternative in such cases?  The VM performs
> the protection by nature of the Java execution model.  If the VM can
> be written to not depend on an MMU, then the Java programs don't
> depend on an MMU, but are restricted to their own memory space by
> the VM.
>
> Then, as long as the VM is written well, it doesn't matter how well the
> Java programs are written -- they just _can't_ intrude.

Yes, Java certainly could be an appealing alternative. Having a JVM
available to us also would have simplified interaction with certain smart
tokens such as Java-based smart cards and DalSemi's Java-powered iButtons.

I think there wasn't a JVM available for RTEMS when we began the project
(July '98), or we would have used it. I'm not even sure whether there is
one available now -- is there? (I haven't followed this topic as I should
have.)

Ideally an embedded Java RTE would be very small, host-able on RTEMS with
only a relatively small ROM (and RAM) footprint and not extraordinarily
inefficient. It wouldn't need a JIT per se, but (say) a 10:1 performance
degradation over what could be done in (say) C would hurt its usefulness.

Assuming that there is a JVM available for RTEMS, how does it fare with
respect to these ideals?