Text preview for : TR440_usenet.txt part of AEG TR440 usenet AEG tr440 TR440_usenet.txt
Back to : TR440_usenet.txt | Home
From: Prof Karl Kleine
Newsgroups: alt.folklore.computers
Subject: Re: Blinkenlights
Date: 11 Aug 2001 23:09:55 GMT
Organization: Fachhochschule Jena, Germany
Lines: 47
Message-ID: <[email protected]>
References:
NNTP-Posting-Host: hoare.gw.fh-jena.de
X-Trace: beta.szi.fh-jena.de 997571395 9683 194.94.38.12 (11 Aug 2001 23:09:55 GMT)
X-Complaints-To: [email protected]
NNTP-Posting-Date: 11 Aug 2001 23:09:55 GMT
User-Agent: tin/1.4.2-20000205 ("Possession") (UNIX) (Linux/2.2.16 (i686))
Sergej Roytman wrote:
> A naive question. What did the blinkenlights on the old iron that had
> them, show? I had always assumed that they were tied to the machine's
> data- and address busses (with appropriate buffering), but then the
> things would be useless much of the time. On the other hand, the BLs
> [........ deleted .....]
Story #2: crash analysis by Polaroid photo for Telefunken TR440
As a student of electrical engineering in the begin of the 70ies at
Ruhr-Universitaet Bochum, Germany, I had a student job in one of the
workshops of the university. As I also had some aquaintance with
the people and the machines at the computing center I got the job
of constructing and building a very special tripod for a Polaroid
camera: Upon a machine crash (or upon seeing a very peculiar pattern)
the operator would grab the camera with the attached tripod, hook
the rubber feet into the corners of the blinkenlights console
(everything was made to measure and fixed) and take a shnapshot.
Yes, that was a real snapshot of the machine state, with instruction
register, instruction address, the user registers, flags, etc.
After that, reboot.
The trick was to have a camera position slightly off center, as
to eliminated / minimize reflections of various lights in the room,
and at the same time get a good rendering of the on/off state of
the blinkenlights. The second issue was that the whole camera /
tripod assembly was fixed and that it lay in some corner of the
room in a cabinet, no adjustments to be made, just pressing the
shutter release for the operator and taking out the Polaroid.
The TR440 will be unknown to most readers here. It was a German
mainframe for scientific calculation, of which about 40 to 60
machines were built und mostly used in universities and research
facilities under heavy grants from the German government to have
a national computing force. Some of it's software was actually
rather advanced for the late 60ies / early 70ies, and I only found
some of the facilities again a decade later with VAX/VMS. The
machine itself is forgotten today; I found just one picture of
it on the web, but no technical information online. I only have
a minimal collection of printed material like the instruction
set summary. That branch of Telefunken was later bought by Nixdorf
and that in turn by Siemens. So there's nothing left :-(.
__________________________________________________________________
Prof. Karl Kleine http://www.fh-jena.de/~kleine
Fachhochschule Jena [email protected]
PF 100314, D-07703 Jena, Germany +49-3641-205-502 [fax -503]
--
From: Prof Karl Kleine
Newsgroups: alt.folklore.computers
Subject: Re: Multipass Algol compiler, (D. Gries)
Date: 13 Nov 2001 22:29:36 GMT
Organization: Fachhochschule Jena, Germany
Lines: 43
Message-ID: <[email protected]>
References: <[email protected]> <[email protected]> <[email protected]> <[email protected]> <[email protected]> <[email protected]> <[email protected]> <[email protected]>
NNTP-Posting-Host: hoare.gw.fh-jena.de
X-Trace: beta.szi.fh-jena.de 1005690576 18834 194.94.38.12 (13 Nov 2001 22:29:36 GMT)
X-Complaints-To: [email protected]
NNTP-Posting-Date: 13 Nov 2001 22:29:36 GMT
User-Agent: tin/1.4.2-20000205 ("Possession") (UNIX) (Linux/2.2.16 (i686))
Steve O'Hara-Smith wrote:
[.........]
> Eeek, I hadn't realised it was quite *that* chaotic. Now I'll
> stick my neck out and guess that some of the popular 'strange' sizes
> were used for several different reasons in several different ways (like
> 45 bits could be 9+12+12+12 and 9+36 float *or* sign and 11 BCD but perhaps
> not often on the same machine).
The German Telefunken TR440 mainframe, built in the 70ies (+-), had a
payload wordlength of 48 bits, plus 2 bits typecode, plus 2 bits error
detection/correction (Dreierprobe). Total wordlength thus 52 bits.
It was an accumulator machine with 3 more 48 bit registers, and
a somewhat weird index register structure, which I do no recall
in detail right now. Have to look at the few remnant docs at home.
The possible type codes were
- 48 bit int
- 48 bit float (have to check for mantissa / exponent layout)
- instruction (two 24 bit instructions in a word)
- other, which usually meant 6 8 bit bytes.
Addresses were word addresses.
AFAIK all of these beasts are out of commission now and there is nearly
nothing left w.r.t. documentation. The TR440 was an attempt to start a German
national computer line for scientific computing. It was a project heavily
subsidized by grants to universities to buy one of these (instead of a CDC
or IBM). One of those days I will scan what I still have (Grosse Befehlsliste,
an A4 format booklet with all the instructions and their effects described
in a brief format.) and a few other such things, and make that available.
Karl Kleine
PS: Back to the title of the tread: All compilers from the 60ies and 70ies
had a significant number of passes. That is, production quality compilers.
There were of course a number of compile-load-and-go systems that tried
to minimize that, mainly for student job batches, where each compilation
was just a few cards (typically less than a page of text, and definitely
less than say 20 pages of listing, which roughly translates to <1000 lines.)
There are a number of techniques which I consider more or less forgotten
these days, as new people in the field (euphemism for 'the youngsters')
simple do not have to cope with tight memory budgets.\
--
From: [email protected] (Bernd Oppolzer)
Newsgroups: bit.listserv.ibm-main
Subject: Re: How Long have you worked with MF's ? (poll) - OT ?
Date: 17 Apr 2002 15:07:52 -0700
Organization: None
Lines: 39
Message-ID: <02041800064100.00597@opp1>
References: <003201c1e657$11abcea0$0900a8c0@transnote>
Reply-To: IBM Mainframe Discussion List
NNTP-Posting-Host: opp1
Mime-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: 8bit
NNTP-Posting-Date: 17 Apr 2002 22:07:52 GMT
For me, it would be interesting, if machines other than IBM's count also.
I worked with a machine called Telefunken TR 440 since 1977 and switched to
IBM's in 1985.
The TR 440 was a machine with a main storage consisting of 192 k words of 48
bits. It hat four 48 bit registers and some 24 bit registers. It also had a
very rich instruction set including stack instructions. It was build by
Telefunken AG in Germany in the late 60s/early 70s, but only 50 boxes were
shipped. The operating system had more features than any other system
at that time, including IBM's.
There were compilers for FORTRAN, ALGOL, COBOL, PL/1, PASCAL and other
languages. The PL/1 compiler was a port from the Multics PL/1 compiler,
see the Multics websites. It had very good optimization features.
I remember that the PL/1 compiler computed the result of the following loop
completely at compile time; at run time only the result was written.
SUM = 0;
DO I = 1 TO 100000;
SUM = SUM + I;
END;
PUT SKIP LIST (SUM);
A real "optimizing compiler".
Unfortunatly, the machine had no offspring, because it was compatible to
nothing else in all aspects.
Would you consider this a mainframe ? In my opinion, it was.
Regards
Bernd
--
From: Reynir Stef