[Simh] Vax 11/780 with multiple CPUs?

[Stephen Hoffman]

All AFAIK...  And all from memory lacking cranial EDC coverage...

On May 28, 2009, David T Hittner <david.hittner at ngc.com> writes:

> It seems to me that if you want to see multiple VAX CPU support, it  
> would be better to simulate a more popular and well documented  
> system, such as the VAX 6000/7000 series.

The multiprocessors among the VAXstation 3000, VAX 6000, VAX 7000, VAX  
8000, VAX 9000 and VAX 10000 series were all symmetric multiprocessor  
(SMP) boxes.

> As you point out, the 11/784 was such a rare beast that you would  
> probably be hard-pressed to find decent documentation. Nor would you  
> likely be able to find real hardware to compare the simulated  
> behavior against.  :-)

Yes; documentation for the MA780 multiport memory is likely going to  
be fairly difficult to find, though you might find something in an old  
VAX-11/780 hardware manual.  It's the MA780 that's key here, too.

For the purposes of a "good enough" emulation, MA780 might well be  
implementable as a range of physical memory that is shared by two  
otherwise entirely separate emulations.

You'll need to scrounge up older VAX/VMS releases, and potentially  
some of the old console firmware if you can find it.  Specifically,  
release(s) somewhere between V3.0 and before V5.0.

The design of OpenVMS Alpha and specifically of OpenVMS Galaxy  
considered various of the old MA780 constructions and interfaces, too.

> Also, AFAIK, VMS dropped support for the 11/784 at one point, due to  
> no longer having a working model on which to verify behavior, just  
> like they dropped the MicroVAX I, the 11/725, the 11/782, the  
> MicroVAX 2000 etc...  :-(

Only the VAX-11/782 dual-processor was officially supported by VAX/ 
VMS, and support for that configuration and support for Asymmetric  
Multiprocessing (ASMP) in general was dropped with the advent of V5.0  
and Symmetric Multiprocessing (SMP) in VAX/VMS and (as it was later  
known) OpenVMS VAX.  The VAX-11/782 (and any hot-rodded 11/784 boxes)  
are asymmetric, and won't work on newer releases.     You had to split  
the VAX-11/782 box into two VAX-11/780 boxes and particularly  
replicate or supplement the MASSBUS or Unibus widgets to upgrade to  
V5.0 or later, though you could continue to use the MA780 as shared  
memory.

On the older VAX/VMS releases, ASMP was good for some specific  
workloads, but could see some serious thrashing on others; most kernel  
code had to be tossed over to the primary.  There was a sales tool  
around back then called "qualify" that helped determine if the  
application load was suited and suitable for ASMP; a load that was  
usually compute-bound and user-mode loads, and processes that lacked I/ 
O that could be scheduled and run on the secondary.

Support for the MicroVAX I, VAX-11/725, and other VAX and Alpha boxes  
that was discontinued was generally dropped either because of memory  
limits, or the availability of installation media devices.  This is  
not to imply we weren't (also) glad to have these boxes out of the  
engineering labs.

And if you want to do something goofy (with real hardware), the KA630,  
KA640, KA650 and KA655 (and possibly a few others of that ilk?) could  
be configured to operate as ASMP boxes (with the primary "owning" the  
Q-bus), though MicroVMS and VAX/VMS never supported those  
configurations.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20090528/2feb3e3c/attachment-0003.html>