Paul Koning wrote:
On Oct 12,
2013, at 12:34 PM, John Wilson <wilson at dbit.com> wrote:
On Sat,
Oct 12, 2013 at 09:22:57AM +0200, Holm Tiffe wrote:
The computers are looking like ther american
originals but if you are
looking deeper, they don't have much in common.
Most of the software is compatible, ok.
[...]
About the 50 vs. 60Hz: that will be no problem but
you will have for sure
Daytime Clock that will run away...
Nitpick: doesn't the Pro use an M146818 RTC as the clock?
So if the Russian box is SW compatible, at least the line freq
won't be a SW problem. (It'll be equally annoying to everyone by
running at 64 Hz instead of either 50 or 60.)
I'm having trouble finding the relevant code to confirm, but that sounds right
-- 64 Hz clock interrupt rate is familiar.
If you check the PRO 350 code for RT-11, in the file
PI.MAC, you will notice that there are a few lines of code
which are used to precede the Normal Clock Interrupt
Handler Routine. It clearly counts out 16 (decimal)
clock ticks and discards the last one each time, then
resets to count back to 16 (decimal) again and begins
to count all over again.
If my mental arithmetic serves correctly, that means that
the 64 Hz clock interrupt is converted to 60 clock ticks.
Since the number of lines of code is so small (less than
a dozen), I expect that most other operating systems
probably perform the same conversion for the PRO 350
rather than forcing additional calculations to keep track
of the time in a 64 Hz system as opposed to a 60 Hz
system which is the usual clock tick rate in a 60 cycle
environment.
I was considering using the same sort of concept for
RT-11, in particular (obviously), and using a clock rate
of 1000 Hz, or in the case of a real DEC M8190 system
800 Hz, which can be selected in addition to 60 Hz and
50 Hz. If I added an additional EMT request, the time
could be requested with an accuracy of 1 msec or at least
close to 1 msec even for a DEC M8190 if an 800 Hz
clock rate is selected.
The only question would be if the ticks are taking place
that often, would any of them be missed? Does anyone
have any experience with the PDP-11 as to how long
any other interrupts seize the CPU and could any of them
last for more than 1 msecs? In general, I would not
expect to try this on even a real DEC PDP-11/93, but
perhaps under the Ersatz-11 emulator where I have had
much actual experience on current CPUs such as the new
Intel hardware. In these cases, I observe that RT-11 code
runs about 100 times faster than a PDP-11/93, so I would
expect that with a clock of 1000 Hz vs 60 Hz, the code
should be able to keep up. Is there any interest in such a
feature?
Jerome Fine
