Ken writes:
> That's not the only problem.  IA-32 is little-endian, which means
> a painful amount of byte-swapping would go on.  That would always
> make it really slow.

Not really.  I think you're only talking about one extra cycle on loads and
stores (of more than a byte), since all the target architectures can perform
the requisite byte swaps in a single instruction these days.  Other aspects
of a "correct" emulation will add lots more overhead than the endian issues
do I think.

> The IA64 is bi-endian, I'm pretty sure, which would help a lot.

IA-64 is bi-endian (so is PA-RISC actually), but I don't know how easy it
would be to use this on a little-endian OS like 64-bit Windows for example.
The "big-endian bit" is not a privileged flag so they can't stop you from
switching your process into that mode, but you'll probably have to switch it
back before making any library calls, etc.  Certainly on HP-UX it would be
easy because HP-UX IA-64 runs big-endian.

G.

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *