>From: allisonp@world.std.com

>> There is (was) at least one "coprocessor" (really, i/o chip) for 8bit micros
>> that does/did IEEE. I don't recall the name or manufacturer.
>
>There were two math processors, 8231 and 8232.  I believe 8232 was Ieee FP
>But I forget how many bits(32 I think).

Didn't AMD have an FP chip too?  I dimly remember seeing an S-100 board
(or something like that) in the Quest catalog ages ago, I'm probably totally
making up these numbers but ISTR they gave you your choice of Am9511 or Am9512.
Or something like that.  Am I insane?

Anyway I can only imagine what a joy it would be to tool out IEEE on an
8-bit micro.  I spent this past weekend writing a set of F_FLOATING routines
in 80386 assembly, at least there you've got big registers and BSR/SHLD/etc.
to help with normalizing but it was still pretty grungy (still, it was easier
than my old code which converts to/from IEEE to use the math coprocessor,
I think that's actually slower than just splitting out the sign/exp/mant
and doing it out by hand, especially for FMUL and FDIV where normalizing
is easy and there's no need to align the binary points of your operands).
Anyway, having to feel your way out one bit at a time in 16-bit pieces with
"DAD H" etc. would be really slow going!  5 msec is no surprise, especially
for division where you'd be spilling stuff in and out of the registers big
time...

John Wilson
D Bit