[Simh] Diagnostics on simulators

[Bob Supnik]

It's actually pretty rate for simulators to run DEC diagnostics 
correctly, particularly from about 1975 onward.

Initially, DEC wrote 'black box' diagnostics, which tested systems from 
the functional spec inward. This is exemplified in the original T1 to 
T17 series on the PDP11, which were instruction and exception tests 
rather than logic tests for the 11/20. However, starting in the mid-70s, 
DEC switched to a 'white box' diagnostic methodology, which worked from 
detailed knowledge of a system design to try to isolate functional 
failures to the FRU (field replaceable unit level), as part of 
simplifying (some would say reducing) the skills needed by Field Service 
technicians. More and more systems and controllers were designed with 
diagnostic modes and hooks to make white box testing work better. It's 
rare for a simulator to implement all of that, or to implement it 
correctly because it was rarely documented well.

If you look at the 11/780 diagnostics, the simulator can run EVKAA, B, 
C, and E, and the MicroVAX 3900 can run HCORE (which was derived from 
those diagnostics), and that's about it.

I am not discounting the value of running diagnostics, particularly for 
older systems that used a black box methodology. Diagnostics were 
critical in debugging the IBM 1401, 1620, and 7094 simulators. The 18b 
diagnostics, and early 12b diagnostics were very useful too. However, 
the VAX was debugged with HCORE and then AXE (the architectural 
exerciser) for the instruction set, and VMS for the IO.

/Bob