Bob Supnik
2018-04-12 16:27:27 UTC
The original DZ emulator was originally written by a contributor. He
chose a model whereby lines were not explicitly represented by units. In
fact, even multiple controllers were not explicitly represented. This
model was followed in some other emulators, such as the DHU/V11 and the
NOVA QTY.
When I started writing terminal multiplexors, I was working with devices
that were essentially multiple instances of the same simple interface,
such as the LT15, DL11, and so on. I chose to have a unit per line, but
still only one (SIMH-visible) controller. This allowed fine-grained
granularity over the number of lines. I carried this model over to
"real" terminal multiplexors on the IBM 7094, SDS systems, and so on.
Both models seem to work or can be made to work. Neither creates a
SIMH-visible controller per real controller.
With respect to the DZ11/DZV11, I expect it would be possible to
parameterize the "lines per controller" math and make it U/Q sensitive.
If the CPU model is changed mid-simulation, there would be significant
trauma to any running software, but users should realize there will
trouble if they change a CPU on the fly.
Unless configuration limitations prevent some family of software from
running, I don't regard the limitations as significant.
/Bob
chose a model whereby lines were not explicitly represented by units. In
fact, even multiple controllers were not explicitly represented. This
model was followed in some other emulators, such as the DHU/V11 and the
NOVA QTY.
When I started writing terminal multiplexors, I was working with devices
that were essentially multiple instances of the same simple interface,
such as the LT15, DL11, and so on. I chose to have a unit per line, but
still only one (SIMH-visible) controller. This allowed fine-grained
granularity over the number of lines. I carried this model over to
"real" terminal multiplexors on the IBM 7094, SDS systems, and so on.
Both models seem to work or can be made to work. Neither creates a
SIMH-visible controller per real controller.
With respect to the DZ11/DZV11, I expect it would be possible to
parameterize the "lines per controller" math and make it U/Q sensitive.
If the CPU model is changed mid-simulation, there would be significant
trauma to any running software, but users should realize there will
trouble if they change a CPU on the fly.
Unless configuration limitations prevent some family of software from
running, I don't regard the limitations as significant.
/Bob