PDP 8 Timesharing

Johnny Billquist bqt at update.uu.se
Mon Nov 30 14:09:40 CST 2015


On 2015-11-30 19:26, Rick Bensene wrote:
> Hi, all,
>
> Just figured I'd post something about my tinkering yesterday.
>
> I got an M8830 from Paul Anderson.  This is the crystal-contolled clock
> for the Omnibus PDP 8 machines.
> Yesterday, I had a chance to try it out.
>
> First, I checked the power supply pins to make sure no shorts or
> anything like that and all was good.
> A quick visual inspection showed no obvious issues.
> It was already jumpered for a 50Hz interrupt rate, so I went ahead and
> plugged it into the backplane.
> Powered the 8/e system up, and ran a few tests from the front panel to
> make sure the board was responding to its IOTs, and all seemed well.
>
> Booted up OS8 from RK05, and mounted up the multos8.rk05 drive via the
> serialdisk driver.
> Copied the MULTOS8  .SV files onto my SYS: volume, and although not
> configured exactly for my system, I figured they'd be close enough.
>
> I then stopped the serial disk server, and fired up Kermit on the laptop
> connected to the second serial port on the 8/e.  Then, I typed R MULTOS
> on the console, and it said something to the effect that I needed to set
> the date first.
> I generally don't bother setting the date at boot time, so I set the
> date to a valid date, and tried again.
>
> This time it gave a welcome message.
> I checked the accumulator, and it was counting off time as it should.  I
> checked the MQ register, and it was static, but then waited for the
> accumulator to overflow, and then the MQ incremented by 1, as it should.
>
>   I pressed CONTROL-H on the console terminal and hit RETURN, and there
> was the . OS8 prompt!
>
> I went to the laptop connected to the other serial interface, and since
> there was no MULTOS 8 password file on the SYS: device, typed CONTROL-H
> there, got the login prompt, hit RETURN, and low and behold, another .
> prompt.
>
> I played around with it for a while, and found that because of some of
> the config differences in how MULTOS8 was built on the pack image, some
> things were acting strange but in general, it definitely was timesharing
> between the two users.
> I could run concurrent things on both terminals, and the response was
> quite acceptable.
>
> I intend to make a build of MULTOS 8 to match my system's configuration,
> and tinker with it some more when I get time.
>
> Next I want to replicate the ETOS Timeshare Board (thanks to Vince and
> Jack for reverse-engineering the board and making a nice schematic!)
> I'm accumulating parts to build one on an Omnibus prototype board.
> Once I get that built, then it'll be time to try out ETOS, which uses
> the improvements in trapping IOTs and dealing with field change
> instructions that really improve timesharing performance over MULTOS 8.

Fun stuff. However, I do not expect you'll see much impovement in 
performance in ETOS compared to MULTOS.

You'll gain if you have code that do CIF n, followed by much code before 
the JMP/JMS, but that is not exactly a common pattern in most PDP-8 code.
Also, MULTOS do some clever stuff. It actually will modify in memory 
code that do polled loops for example, to not do those. Unless ETOS do 
the same, that might actually hurt ETOS more. Also, MULTOS is the only 
timesharing system I know that actually also handles the TD8E 
controller, if you happen to have one of those.
I only wish I knew where I have version 2 of MULTOS. I uploaded version 
1 a long time ago, but only at a later date realized that it wasn't the 
latest version. I *think* I should have version 2 somewhere as well, but 
my memory might be wrong also.

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol


More information about the cctalk mailing list