i agree with Peter, WRT the '181 you can certainly find them in small qualities in the after market, I may have a handful of them myself.   One thing you are missing is a small barrel shifter, which the '181 does not support, but for a 8 bit shift it should not be terrible.   Vaxen and such built their independently, but it is worth while IMO, as it allows in a single clock tick to perform pretty substantial arithmetic, although because it's a sq law the cost in die area is huge. If you look at the 68K for instance, the big area in the middle everyone sees in the pictures is the shifter (same for the Alpha chips).

Clem

On Thu, Aug 10, 2017 at 4:16 AM, Peter Jeremy <peter@rulingia.com> wrote:
On 2017-Aug-08 12:52:36 +1000, Warren Toomey <wkt@tuhs.org> wrote:
>Hi all, sorry for this off-topic posting.

Well, just promise to port Unix to it and it won't be off-topic.

>Anyway, I thought it would be fun to do something similar but with
>a bigger address space and as few chips as I could get away with.

Well, using a FPGA or microcontroller will reduce the chip count further.

You suggest that you won't use a 74LS181 because they are no longer made.
Whilst that's true, they are still readily available as NOS.  Overall, I
would suggest that using a 74LS181 (or 4) is a cleaner solution than
(ab)using a couple of EPROMs.  (The other option would be a PAL or GAL but
I suspect finding and programming them would be more difficult).

--
Peter Jeremy