High performance coprocessor boards of the 80s and 90s - was Re: SGI ONYX

[Jon Elson]

On 04/20/2016 10:00 AM, Toby Thain wrote:
> On 2016-04-20 10:27 AM, Pete Turnbull wrote:
>> On 20/04/2016 14:57, Toby Thain wrote:
>>
>>> I'm changing the subject because the subject of RISC 
>>> coprocessor boards
>>> has already been interesting to me; I owned the NuBus 
>>> Levco Translink II
>>> (for Mac II family) with four TRAM slots for transputers.
>>>
>>> Also going to mention the BBC Tube coprocessor here. 
>>> Which had an ARM
>>> version, iirc.
>>
>> It did indeed - I have one.  Also a couple of 6502 
>> CoPros, a 65C102, a
>> 32016 and a pair of Z80s, which were nice in their day.
>>
>
> Nice collection. I'd forgotten about the 32016! What 
> software ran on these respective processors?
>
I got a group at work to buy a Logical Microcomputer Co. 
32016-based system, it originally came with GENIX, I think.  
They later changed to XENIX.  (I may have the order of those 
backwards.)

I cloned the whole system on wire-wrap boards, and it 
actually WORKED!

My memory was probably a bit slow, and the system was a 
DOG!!!  I was also not a Unix guru, by any means, so a lot 
of stuff was painful.  But, compared to my Z-80/S-100 
system, it was so slow as to be a real pain in the neck.

So, there were several versions of Unix derivatives for the 
32016 family.  Nat Semi also had cross compilers that ran on 
the VAX and would compile C, FORTRAN, Pascal and assembly 
source into binary that you could download onto their 
development boards.  We built a multiprocessor system that 
added on to a VAX 11/780.  But, it was too cumbersome for 
our physicists to go through all that hassle, and we had no 
debugging facility in that attached processor mode.  (I 
could have hacked some kind of error message queue to the 
system.)

Jon

Jon