7 Mar
2011
7 Mar
'11
2:20 p.m.
... or mid-late '70s and early '80s CPUs, to be more specific. Can anyone
furnish me with a better knowledge as to which processors of that kind of time
period had hardware multiplication and division support?
http://en.wikipedia.org/wiki/Binary_multiplier mentions the 6809, but I'm
curious as to which* others had such hardware features (and when they started
moving from simple shift-add routines to more complex approaches which used
more silicon - or did that not happen until much later in the 80s?)
* mainframe, mini, micro; I'm not picky. I'm more interested in building up a
picture of how widespread hardware support was, and the various approaches
that designers used.
What do you mean by 'hardware' multiply and divide? A number of machines,
I suspect the 6809 is amongst them, had no particualrl hardware for
multiply or divide, but they did have multiply and maybe divide
instrucitons in the instruction setc. These instrucitons were implemented
by microcode using the normal registers and ALU. Does that count?
IIRC, the PDP11 got hardware multiply and divide with the 11/45 in 1972.
It was simple shift+add IIRC. The floating point boards for that machine
implemented multiply and divide suing a few speedup tricks (like
subtracinging, skipping paset a block of 1s in the multiply and then
edding, so that, for example, *15 became a *-1 + *16 if you see what I mean).
And IIRC, there wa sno multiply-specific hardare (or at least not much)
on the 11/45 CPU. Itwas mostly done in microcode.
Not all PDP11s had said instrucitons after that, though.
The EAE Extended Artihmetic Element) for the PDP11 (a Unibus card that
would wotk in any Unibus machine) and a similar add-on for the PDP8/e CPU
were essnetially hardware multiply/divide boards. The Unibus one didn't
add to the processor instruction set, rather it had its own registers
(accessed like I./O ports) that you loaded the numbers into and read the
rsult from.
The PERQ1 (4K WCS) -- 1979-- had no multiply/divide supprot, the PERQ1a --
9180 -- (and thus all later classic PERQs) had the MulStep/DivStep
hardware which let you calcu;ate one bit of the product or quotient in one
microcycle (using restoring division). However the stnadard microcode
for all classic PERQs (loaded from disk at boot-up) had multiply and
divide _machine_ code instructions.
-tony