[Simh] PDP11 microstore

[Bob Supnik]

The PLA scheme (another invention from the fertile mind of Bill Roberts, 
architect of the LSI11 and UDA50 proto and founder of Emulex) was 
basically a microcode and gate conservation scheme. It provided enormous 
compression for the decode phase of the PDP11 and then got exploited to 
a fare-thee-well by clever microcoders like Burt Hashizume (F11) and 
Keith Henry (J11).

Sometime during the 80s, when it became clear that the process problems 
with the J11 were intractable, and it could not be shrunk for higher 
performance, I outlined a design for a single chip PDP11. Because ROM 
densities and transistor counts were so much higher by then, I felt it 
could all be done in ROM, using the usual "indirection pointers" like Rs 
and Rd for source and destination registers and a fairly simple initial 
decoding scheme to break out the interesting cases. I didn't save 
anything about it in my online archive, so I don't know what happened to 
the design notes. The idea didn't go anywhere, of course; the PDP11 was 
clearly on its last legs by then.

/Bob

On 7/1/2019 2:36 PM, simh-request at trailing-edge.com wrote:
> Message: 2 Date: Mon, 1 Jul 2019 10:50:51 -0600 From: Eric Smith 
> <spacewar at gmail.com> To: simh at trailing-edge.com Subject: Re: [Simh] 
> Which PDP-11 to choose Message-ID: 
> <CAFrGgTRM4n22yugG60SLh_xT=ttk4ifAMON9rfq5FgMYz-Lh8A at mail.gmail.com> 
> Content-Type: text/plain; charset="utf-8" <snip>
> IIRC Bob has written that no one has succeeded at building an alternate J11
> hardware implementation, e.g., in an FPGA., because the microcode is not
> entirely ROM. There is a fairly large (for the time) PLA forming part of
> the control store. The PLA could be transformed into a ROM, but IIRC it has
> _many_  inputs, so the ROM would be YUGE.
>
> Almost 20 years ago I wrote a program to translate the PLA into VHDL
> directly instantiating Xilinx 4LUT primitives to see how much resources it
> would consume in e.g. a Spartan 3 FPGA. I don't recall the numbers, but it
> didn't seem insurmountable at the time, and with today's bigger FPGAs and
> 6LUTs, it's even less of a problem.
>
> Also, I think the synthesis tools are good enough that just giving the PLA
> equations to synthesis would be fine, and my scheme of programmatically
> transforming the equations to LUT instantiations is totally unnecessary.
>
> Eric