On Tue, 12 Sep 2000, Rob Healey wrote:

> 	Are any of the mips ports on handhelds/game controllers running
> 	without an MMU?

  No.  CPUs without MMUs tend to be quite slow anyway, and even game
controllers come with a MMU now.

> 	I assume you'd need to generate totally relatively addressed
> 	object code and do some pretty fancy bus/memory signal footwork without
> 	an MMU.

  I don't know the modern theories on this, but my first thought would be
for the kernel to examine the binary and change any instruction that could
access a restricted or mmaped area or the code segment to first jump to
code that checks if it does and deals appropriately.  I suspect you could
do better then 10% runtime overhead on most programs plus substantial
overhead the first time a binary is run.  

  However, even if you did all the work to make things work like usual,
it wouldn't really be NetBSD since there would be a whole lot of MI code
that you couldn't or wouldn't want to use.

Matthew Orgass
darkstar@pgh.net