John writes:
> The one point in Wirt's proposal with which I will disagree at
> this time is the notion that the emulation should be developed for
> the 32-bit Intel 80x86 chip, and only later moved to IA-64.

The problem is the relative "size" of the architectures.  IA-32 is "smaller"
in terms of registers and other resources than PA-RISC is, and this will
have an effect on emulator design and performance.  The big advantage of
IA-32 is that it is ubiquitous and if you want to get something into a lot
of people's hands quickly and cheaply, then IA-32 is the way to go.

IA-64 on the other hand is basically a superset of PA-RISC, with lots more
registers and a very compatible virtual memory system that would make a high
performance emulator quite practical.

So if it were me, I would probably develop a pure C language emulator first
for simplicity and universal compatibility (i.e. IA-32) but would not expect
the performance to be stellar (if it would run as fast as a 918 on a 2GHz
IA-32 box I would be quite pleased).  The next step might be an IA-64
version with hand-coded instruction emulation, which ought to perform at a
more significant fraction of the native clock speed.

G.

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