On 7 Mar 2011 at 17:26, Brent Hilpert wrote:
On 2011 Mar 7, at 4:08 PM, Jules Richardson wrote:
> Richard wrote:
>> I don't think the multiply implementation in the IBM 701 was
>> shift/add microcode variety. That wouldn't have been very
>> performant for its intended purpose (scientific computing).
>
> I'll have to have a look online and see if I can find any details. I
> see that Booth's algorithm was 1951 - much earlier than I expected
> and a year before the 701. I'm not sure when Wallace trees showed
> up...
The 701 had multiply and divide instructions as did the 650. But
then, the IBM 603 could multiply and the 604 could multiply and
divide.
But I have to wonder how some of the instructions were actually
implemented in some cases. Consider the Packard Bell PB250. 300
transistors, 2500 diodes and 4 magnetostrictive delay lines for
memory. 22 bit words; bit serial ALU, but it had multiply and
divide.
It'd be fun to try to reconstruct a machine built on such a
minimalistic approach.
And let's not forget the IBM 1620 (both the I/CADET and the II did
multiplicatoin by table lookup.
--Chuck