[Simh] EXT : Questions regarding future simulator development

[Johnny Billquist]

On 2013-04-10 20:55, Larry Baker wrote:
> On 10 Apr 2013, at 11:22 AM, <simh-request at trailing-edge.com
> <mailto:simh-request at trailing-edge.com>> <simh-request at trailing-edge.com
> <mailto:simh-request at trailing-edge.com>> wrote:
>
>> 2> Assuming that the MASSBUS hardware devices actually supported the
>> extended configuration, you could certainly add the extra components
>> into the configuration. The question then becomes do any of the
>> operating systems that ran on the original hardware support the
>> extended configuration? If the operation systems don't support the
>> extensions, then adding the extra simulated components is a waste of
>> time. Otherwise, you just have to ask yourself "Is it worth the
>> effort?" Note that any new components added to a simulation should be
>> disabled by default for compatibility with older scripts.
>
> RXS-11M-Plus did a very nice job of supporting multiple MASSBUS
> controllers.  Not only that, it supported mixed-MASSBUS configurations
> -- tapes and disks on the same controller.  And, multi-path MASSBUS
> configurations to dual-port disk drives.  Our PDP-11/70 had several
> MASSBUS disk drives all dual-ported to two MASSBUS controllers.
>   RSX-11M-Plus issued seeks separate from data transfers.  Whichever
> controller was available could be used to issue any command to any drive
> it could reach.  As I recall, the PDP-11/74 could have even more than 4
> MASSBUS controllers, because each processor had its own MASSBUS/UNIBUS.
>   Some rather powerful disk configurations could be configured.
>   RSX-11M-Plus had all the code to make it work.  (The developers used a
> PDP-11/74 system.)

M+ supports up to 4 CPUs, 12 Unibuses, and 16 Massbuses. The 4 CPUs are 
a limitation of the IIST, unless I remember wrong. The 16 Massbuses are 
a logical extension from the four CPUs, since in the 11/74 was more or 
less a modified 11/70, and each 11/70 can have four RH70 controllers.
The 12 Unibuses are a design limitation in RSX, as you have one 
additional field in many data structures, called the Unibus run mask (or 
URM). That is a 16-bit value, and thus you have one bet per Unibus. Each 
CPU have one implied Unibus, leaving space for 12 additional Unibuses in 
the URM.

In reality, this worked by having Unibus switches (DT07), between 
segments of the Unibus. And segments could be moved from one CPU to 
another. So, the buses could be reconfigured on the fly of a running 
system. But all buses had to end up at one CPU or another sooner or 
later. And no, you could not simultaneously have one Unibus connected to 
multiple CPUs. But in case of CPU failures, you could switch the bus 
over to another CPU.

(Yes, 11/74 machines are really fun, but simh probably needs some larger 
reworking before you could do something like an 11/74.)

	Johnny