>>> The port which uses MIPS_64BIT_CLEAN should use another magic number
>>> for the executables. (to distinguish the executables and the shared
>>> objects are compatible with existing MIPSLE32 or not)
>> 
>> Yes.  The different magic numbers will select different emuls, which will
>> have the correct pointers for the right setregs, etc. functions.
>
> This sounds like a fine solution.  I'll flush out the genpubassym stuff
> probably today, and start reading through the emul stuff to see how that 
> works.  We want this to support the n32 programming model, too.

It would be valuable how 'NetBSD/sparc64' approachs to mixed register size
runtime environment.

Tohru Nishimura
Nara Institute of Science and Technology