FPGAs, unobtanium and FPJ11s

[Pete Turnbull]

Roger Ivie wrote:
> On Fri, 24 Nov 2006, Richard wrote:
>>
>> Would it be possible to do the same sort of trick with the LSI-11?
>> I've got an 11/03 and it would be nice to have floating-point :-).
> 
> There were a couple of floating point options for 11/03. I haven't
> actually seen either.
> 
> The first, FIS-11, was a ROM that goes into the empty socket.
> 
> The second, FPP-11, was an add-on board that connected to the 11/03 via
> the empty socket.

I think you're confusing this with 11/23, or maybe 11/40 options.  There 
is a microcode ROM for the 11/03, which contains EIS/FIS (ie, both the 
Extended Instruction Set and the Floating Instruction Set).  It's called 
KEV11, not FIS-11.  There is a FIS option (and a separate EIS option) 
for an 11/40.  However, the FIS is not the same as other PDP-11 floating 
point instructions.  For a start, it's all stack-based (no register 
operations) and it uses a different floating point format.  Which is why 
the opcodes are different too.

There's a similarly-named option called KEF11 for an 11/23, which does 
implement the normal PDP-11 floating point instructions (in microcode). 
  It needs the MMU present, because it uses registers in the MMU; it 
doesn't implement EIS because the basic 11/23 KDF-11 chipset already has 
EIS, unlike the KD-11 chipset in the 11/03.  It doesn't implement FIS 
either, because there's no point.  There is also a quad board with a 
floating point processor which plugs into an 11/23 (or 11/24) instead of 
the KEF-11; this is called an FPF-11, and it doesn't need the MMU 
registers because it has its own.

I've never heard of an FPP-11.  There are several FP11-x boards for 
Unibus machines.

> As I understand it, the empty socket could be used for either floating
> point or the commercial instruction set. Since there's only one empty
> socket, you can't have both floating point and the CIS.

There's no CIS for an 11/03; there is a CIS option for KDF-11 machines, 
which consists of a carrier that plugs into a *pair* of microm sockets 
on an 11/23 or 11/24.  The carrier holds six chips.  There's also a CIS 
for the 11/44 (two board set).

> There's also a writable control store, WCS-11, that plugs into that
> slot.

Yes, that's a KUV-11, M8018.  I suppose if you could fit the floating 
point instruction set into 1024 microcode words, you'd almost be able to 
implement floating point -- but there would still be no registers 
available to operate on.

-- 
Pete						Peter Turnbull
						Network Manager
						University of York