Drum vs. Core

[Roger Holmes]

On 6 Jul, 2007, at 18:07, cctalk-request at classiccmp.org wrote:
>
> From: "Andy Holt" <andyh at andyh-rayleigh.freeserve.co.uk>
>
>>>>
> Back to the original post, does anyone know if there is a difference
> between the "operators Exec for a 1900" and the executive used by
> engineers and recorded on 7 track mag tape? Would this be of any use?
> <<<
>
> Does that imply that you - or someone you know has such a tape???

Whilst working on my 1301, I was chatting to a friend who was an ICT/ 
ICL field engineer and later manager into the Fujitsu years. I told  
him that all copies of 1900 executive have been lost and he said that  
he might have a copy tucked away but that, if he had it at all it was  
on 7 track tape and we don't have a deck to read it, but maybe  
someone somewhere has a way of reading a tape which has not been  
touched for many years.

I will let you know if he finds it, but don't hold your breath, like  
me he has a lot of buildings and does not go in some of them for long  
periods.

With the 1301 we have a copy of 'Initial Orders' on the drum, a  
numeric dump of a slightly different version from another machine and  
also (dyleine?) copies of the original manuscript. Its only two drum  
bands (so less than 800 instructions), but it is the nearest thing to  
an operating system that the 1301 had.

The 1302 Exec has not survived, nor even details of all the extra  
instructions in the machine, we know everything about a few, sketchy  
details from sales literature for others and there are probably a few  
we know nothing about. A recent documentary series "The Trap, What  
Happened to Our Dreams of Freedom" had clips of old footage, and in  
one there was a chap with a book open at a page showing what I  
immediately recognised as a 1300 instruction summary, the only  
coloured page in the 1300 programming manual. On freeze framing it  
and comparing with the 1300/1301 version, I see the arrangement is  
quite a bit different, and I think it must be a 1302 manual, but of  
course the image resolution, even on the original 16mm film would  
probably not allow it to be read, and certainly not off the video now  
on YouTube.


>
> anyhow, to answer the question - "it all depends"
>
> I've only limited knowledge of the Engineers test programs tho' I  
> think I
> have some notes somewhere.
> So some of what follows may not be precisely correct.
>
> Most test programs were normally run under operators exec and were  
> stored on
> a normal program library tape
> (these were tapes with a search program (#TAPE) at the start and  
> the rest of
> the programs in any order following.)
>
> The engineers (raw machine) test tape would have started with a  
> minimal
> exec-mode bootstrappable search program at the
> start (really only a program loader - not a full exec). This would  
> have been
> followed by the exec-mode test programs which
> were presumably in a bootstrap format. Our engineers usually ran  
> such basic
> test programs from paper tape - this may have been
> because that process relied on less of the system being operational  
> that the
> process of finding and loading one from MT.


Well, when and if he finds it, maybe what's written on the spool will  
give us a clue to if its any use or not.

>
> Of course, as always!, the system faults that one remembers were not
> detected by the test programs ...
> (On the 1905E one time odd things happened with the Limit register  
> - that
> took some locating;
> On the Honeywell dual-processor, one of the two processors started  
> giving
> some wrong results in floating-point -
> and as it was (almost) SMP this had effectively random results that  
> took
> some locating)

We've had some weird ones with Flossie too. We found a fault three  
weeks ago, all the read-modify-write instructions were randomly  
causing a subsequent instruction to fail, about once a minute. This  
is just long enough to get from the console to the oscilloscope, so  
you start a program and it seems like the machine is being a bitch  
and waiting for you to NEARLY get to find out what is wrong, but not  
quite letting you do it. At the heart of the machine is a shift  
register which is supposed to have exactly one '1' in it at all  
times. The duration of the instruction is controlled by how far along  
the register the '1' gets. With two of us working on the fault, I was  
probing various signals and eventually we traced the fault to the  
shift register when my friend shouted over that the fault had seemed  
to have disappeared. I found that the impedence/resistance of the  
scope probe on one of 25 pins on one of 4000 cards seemed to make the  
fault disappear (and reappear when removed). We think we were getting  
two '1's in the shift register, but anyway, we changed that PCB and  
all has been well since.

Roger.