I want to buy an ARM dev. board

[Olivier Houchard]

On Wed, May 26, 2004 at 12:49:15AM +0800, Morton Lin wrote:
> Sure, ARM940T only had Memory Protection Unit, also {I | D}-cache. But I
> thought
> it can have some similar capabilities like an MCU which had MMU. (Of course,
> for
> FreeBSD/ARM, it's a lot of work / re-write to achieve that)
> 
> IMHO, in modern embedded market, there were many MMUless MCUs used in
> valuable field. Even those MCUs which had MMU, the EmbeddedOS/RTOS went
> along with, didn't use their MMU functions. For example, uCLinux. There're
> cost and
> budget issues behind this situation. In low end application I think the
> 32bits MCU like
> the ARM7/9 series will be the main stream. In high end field, I think the
> Intel XScale
> (w/ MMU) will be the star.
> 
> So, are we going to have two directions for the FreeBSD/ARM ? Is it possible
> ? Can
> anyone tell us how much effort do we need to take and how difficult if we
> want to port
> FreeBSD/ARM to a MMUless core/platform ? Or we just move toward the Intel
> XScale
> architecture ?
> 

Well, it sounds like a huge amount of work, and I don't think FreeBSD is well
suited for that kind of applications.
Supporting MMUless cpus would require to either emulate the MMU, which 
I think would cost far to much, or to re-design the FreeBSD kernel to not rely
on the virtual memory, and it wouldn't be FreeBSD anymore :-)

Any suggestion would be welcome but I can't see any way to get FreeBSD to run
on a MMU-less cpu without a major work.

Cheers,

Olivier