[Simh] VAX 11/782 simulation [was: :Re: Pdp8 terminals]

[Johnny Billquist]

Hi.

On 2016-09-07 21:57, Anders Magnusson wrote:
> Hm, IIRC the 83x0 was as asymetric as the 782...?  It could also only
> take the interrupts at the master CPU.

Uh. No, as far as I know/remember they 83x0 were not asymmetric in the 
way the 11/782 was. As far as I know, no other VAX did ASMP at all, and 
definitely not the way the 11/782 did it. (Well, I seem to remember 
seeing something called the 11/787, which was an 11/782, but based on 
the 11/785 instead of the 11/780, hence 11/787, so the same design).

The 8300 was a dual CPU 8200. It uses VAXBI, and thus each CPU is just a 
node on that bus. So all CPUs have access to the full system. There are 
still a few things that separate the primary CPU from the secondary 
ones, but all CPUs run the full kernel, and can deal with system calls 
as well as I/O. It might be that all interrupts are routed to the 
primary CPU, but the 11/782 is way more asymmetric than that...

> ...then there was the multiprocessor MV2, that had both local and shared
> memory...

That was never a supported config from DEC, if I remember right? Was the 
shared memory some memory in the Qbus address space then? Could the uVAX 
II even deal with Qbus memory? (Uh, I bet Bob Supnik knows... :-) )

But this is similar to the 11/782, which also could have both CPU local, 
and shared memory at the same time. In the 11/782, you have special 
memory boxes that were dualported.

The 11/782 also only had peripherals connected to the primary CPU. And 
anything done in the kernel happened on the primary CPU. If you so much 
as did a system call on the slave, it would interrupt the primary CPU, 
it itself just stop processing there and then. The primary CPU would 
then schedule some process for the slave CPU, and kick it off to start 
work. So all scheduling also was done on the primary.

The slave CPU was more like a device from the primary CPU perspective.

Not sure you even had a Unibus on the slave, but anyway, you were not 
allowed to connect any peripherals on the slave CPU buses.

And to respond to Dave Hittner, yes, I believe VMS removed the ASMP 
support, thus forcing any existing 11/782 systems to be split into two 
11/780. Which is why I said that I'm not even sure where to find any 
software that would support the 11/782.

   Johnny

>
> -- Ragge
>
> Den 2016-09-07 kl. 21:15, skrev Hittner, David T (IS):
>> Wasn't the VAX 11/782 and the Asymmetric model de-supported at some
>> point during the life of VMS?
>>
>> I seem to remember hearing (at a DECUS Symposium?) that existing
>> 11/782 customers would have to separate them into two 11/780's and
>> then cluster them to upgrade to the latest version of VMS.
>>
>> Dave
>>
>>
>> -----Original Message-----
>> From: Simh [mailto:simh-bounces at trailing-edge.com] On Behalf Of Johnny
>> Billquist
>> Sent: Wednesday, September 07, 2016 2:55 PM
>> To: Ray Jewhurst
>> Cc: simh at trailing-edge.com
>> Subject: EXT :Re: [Simh] Pdp8 terminals
>>
>> Hi.
>>
>> On 2016-09-07 19:00, Ray Jewhurst wrote:
>>> My apologies. To be honest I am very ignorant about of the science of
>>> computers. I am disabled, cannot work but I am a total computer
>>> history buff. I want to learn and know and experience as much as I can
>>> and Simh is a major part of that. I  want to help where I can so
>>> please excuse my occasionally confusion.
>> No worries. You never learn if you are afraid of making mistakes.
>> However, if you are not rather experienced writing code, I think that
>> implementing something like the VAX-11/782 might be quite a task...
>> Not to mention that I don't know where you'd find a version of VMS
>> that would support the machine.
>>
>> Maybe find something a little simpler to start with? Like the talk
>> about expanding the PDP-8 to more generic support for devices than the
>> current implementation.
>>
>>     Johnny
>>
>>>
>>> On Sep 7, 2016 12:46 PM, "Johnny Billquist" <bqt at softjar.se
>>> <mailto:bqt at softjar.se>> wrote:
>>>
>>>      The 11/782 are no more asynchronous than any multiprocessor system.
>>>
>>>      The A in ASMP stands for assymetric. As in, the second processor
>>> did
>>>      not run any kernel code, but is a slave processor to the primary
>>>      processor. It gets scheduled with user-land code to run, but any
>>>      trapping to the OS means it interrupts the main processor, who do
>>>      all the work.
>>>
>>>              Johnny
>>>
>>>      On 2016-09-07 18:23, Ray Jewhurst wrote:
>>>
>>>          I think you are like I am. I would like to see every DEC
>>> simulator
>>>          possible. Right now I am doing some preliminary research
>>> into the
>>>          feasibility of a VAX 11/782 which is an asynchronous dual
>>> processor
>>>          11/780.  I will need help because I am not a real
>>> experienced coder.
>>>
>>>
>> _______________________________________________
>> Simh mailing list
>> Simh at trailing-edge.com
>> http://mailman.trailing-edge.com/mailman/listinfo/simh
>


-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol