[Simh] RTE-6/VM configuration
J. David Bryan
jdbryan at acm.org
Wed Aug 27 22:03:29 EDT 2008
On Wednesday, August 27, 2008 at 14:13, Ken Cornetet wrote:
> If memory serves, a privileged interrupt fence card was just a set of
> IO flip-flops.
There were two PIFs: the 12936A for DOS, and the 12620A for RTE. The RTE
fence was a standard interface; in fact, it was actually a "breadboard
interface," i.e., just the control, flag buffer, and flag flip-flops, plus
associated interrupt and priority logic.
The 12936A was quite a different beast. It behaved in a completely
different manner to the standard interface logic. The manual here:
http://www.bitsavers.org/pdf/hp/21xx/interfaces/12936.pdf
...details this very unusual card. It was so unusual that I had to rewrite
the SIMH HP I/O system for the next revision in order to accommodate it.
> In theory, an existing simh device could be used as a PI fence.
The LPS device (in DIAG mode) works as an RTE fence, but nothing can
substitute for the DOS fence. The problem with most SIMH devices is that
setting the control flip-flop on the card initiates an I/O operation, which
then fails if the device isn't attached to an external file. The DIAG mode
of the LPS device simulates installation of a loopback connector, so no
attachment is needed.
> I can't find anything on DVS00, though. Was that a special driver?
It was part of the 91731A Mux Subsystem for RTE. I've sent manual PDFs for
that to Al, but they're not up on Bitsavers yet.
The 91731A package supplied a physical driver (DVS00) and a logical driver
(LDV05) that mimicked the BACI driver for the 264x terminals. It required
a privileged system because the mux had no on-card buffering and would lose
characters if another device driver preempted it during character
reception.
The 12920A mux was quite an old device (early 1970s), required three I/O
slots, wasn't buffered, required a lot of CPU time for I/O, and supported a
maximum of 2400 baud. The 12792A mux was much newer (early 1980s), had an
on-board Z80 microprocessor and buffer RAM, used a single slot, supported
9600 baud, required relatively little CPU even at the highest transfer
rates, handled character editing (backspace, line delete) on the card, and
supported block mode on 264x terminals. It did not require a privileged
system.
-- Dave
More information about the Simh
mailing list