Reading about architectures of famous early supercomputers in comp.arch
got me thinking about the computer on which I first cut my teeth:
The Danish GIER computer. In the first 9 months after I left the Danish
high school and entered University of Copenhagen, I laid hands oa no
less than three of these machines. Wikipediai (even the Danish version)
does not have much information about these machines, but a bit more
about the predecessor, the DASK, which was only ever built in one instantiation.

The description of the DASK instruction architecture sound somewhat
like the GIER, which makes me wonder: Did the GIER mostly re-implement
the DASK architecture in transistor technology (where the DASK was
tubes) ?

The GIER architecture had an unusual word architecture: Each word had 40
data bits plus two flag bits. In data words, these were called the
a and b bits, and their use was up to software. For example, one might
use them to flag the end of a row of an array and then test for that
to exit a loop. In instruction words, one was called the "s" (for short)
bit, indicating that the instruction held two 20-bit instractions
instead of a single 40-bit instruction, while the other was the "f"
bit, indicating the the opcode earlier in the word was a floating point operation. Integer and floating point operations had the same opcode,
the "f" bit told the CPU which instruction was really in play.

The Wikipedia article on DASK mentions the short/long instructions and
says it was a 40-bit machine. I tend to think of the GIER as a 42-bit
machine, since memory words are 42 bits wide, and all 42 bits travel to
the registers, though only 40 participate in arithmetic operations.

Any Danes here?

--
Lars Poulsen in Santa Barbara

--- MBSE BBS v1.1.1 (Linux-x86_64)
* Origin: A noiseless patient Spider (3:633/280.2@fidonet)

On 2025-05-22, Lars Poulsen <lars@beagle-ears.com> wrote:

Reading about architectures of famous early supercomputers in comp.arch
got me thinking about the computer on which I first cut my teeth:
The Danish GIER computer. In the first 9 months after I left the Danish
high school and entered University of Copenhagen, I laid hands oa no
less than three of these machines. Wikipediai (even the Danish version)
does not have much information about these machines, but a bit more
about the predecessor, the DASK, which was only ever built in one instantiation.

<snip>

Any Danes here?

Not me, but I did use one of the peripherals. My first job was
in a shop with a Univac 9300 (their answer to the IBM 360/20).
Cobbled onto it was a Regnecentralen RC-2000 paper tape reader.
We were a service bureau that took data from many small customers,
a lot of which submitted data on paper tape.

The RC-2000's manual mentioned it being designed to connect to a
GIER computer, which is about as close as I come to this thread's
topic. But the paper tape reader itself was an impressive machine.
It had a photocell head, a servo-driven capstan, and a 256-byte core
memory buffer. The servo motor would do its best to keep the memory
buffer half full; the faster the computer asked for data, the faster
the tape would move, up to a maximum of 2000 frames per second.

Most of the time we never got the reader to reach full speed,
since data records were typically being listed on the computer's
printer as they were being read, and the computer was too small
to support spooling. However, I did write one program that stored
data on disk, and it made the machine open up, causing tape to
fly through at 200 inches per second. This caused such a buildup
of static electricity that sparks would crash the machine. We
wound up hanging a grounded chain of paper clips by the reader
where the tape would hit it on the way out, and even then we had
to run a kettle in the machine room to get the humidity up enough
to bleed off the static charge.

Fun memories...

--
/~\ Charlie Gibbs | Growth for the sake of
\ / <cgibbs@kltpzyxm.invalid> | growth is the ideology
X I'm really at ac.dekanfrus | of the cancer cell.
/ \ if you read it the right way. | -- Edward Abbey

--- MBSE BBS v1.1.1 (Linux-x86_64)
* Origin: ---:- FTN<->UseNet Gate -:--- (3:633/280.2@fidonet)

On Thu, 22 May 2025 19:05:16 -0000 (UTC), Lars Poulsen wrote:

Reading about architectures of famous early supercomputers in comp.arch
got me thinking about the computer on which I first cut my teeth:
The Danish GIER computer. In the first 9 months after I left the Danish
high school and entered University of Copenhagen, I laid hands oa no
less than three of these machines. Wikipediai (even the Danish version)
does not have much information about these machines, but a bit more
about the predecessor, the DASK, which was only ever built in one instantiation.

The description of the DASK instruction architecture sound somewhat like
the GIER, which makes me wonder: Did the GIER mostly re-implement the
DASK architecture in transistor technology (where the DASK was tubes) ?

The GIER architecture had an unusual word architecture: Each word had 40
data bits plus two flag bits. In data words, these were called the a and
b bits, and their use was up to software. For example, one might use
them to flag the end of a row of an array and then test for that to exit
a loop. In instruction words, one was called the "s" (for short)
bit, indicating that the instruction held two 20-bit instractions
instead of a single 40-bit instruction, while the other was the "f" bit, indicating the the opcode earlier in the word was a floating point
operation. Integer and floating point operations had the same opcode,
the "f" bit told the CPU which instruction was really in play.

The Wikipedia article on DASK mentions the short/long instructions and
says it was a 40-bit machine. I tend to think of the GIER as a 42-bit machine, since memory words are 42 bits wide, and all 42 bits travel to
the registers, though only 40 participate in arithmetic operations.

Any Danes here?

Not I, but they were interesting machines. Someone (Wirth?) implemented an early version of Algol on the GIER, and Brinch-Hansen created concurrent <something-or-other, maybe Pascal> for the rc4000, and wrote an OS in it.

It takes two things to get this stuff into Wikipedia: the documentation
must be available, and there must be someone to take enough interest to
write or update an article.


--- MBSE BBS v1.1.1 (Linux-x86_64)
* Origin: A noiseless patient Spider (3:633/280.2@fidonet)

On Thu, 22 May 2025 19:05:16 -0000 (UTC), Lars Poulsen wrote:

The Wikipedia article on DASK mentions the short/long instructions and
says it was a 40-bit machine. I tend to think of the GIER as a 42-bit
machine, since memory words are 42 bits wide, and all 42 bits travel to
the registers, though only 40 participate in arithmetic operations.

Any Danes here?

On 2025-05-26, Peter Flass -- Iron Spring Software <Peter@Iron-Spring.com> wrote:

Not I, but they were interesting machines. Someone (Wirth?) implemented an early version of Algol on the GIER, and Brinch-Hansen created concurrent
<something-or-other, maybe Pascal> for the rc4000, and wrote an OS in it.

Not Wirth - Peter Naur (who was one of the editors of the Algol-60
report.

The rc4000 system programming language was also Algol.

Peter Naur was an astronomer, who had been to California to do some
observation project at Mt Palomar, and had seen and fallen in love with computers. Back home, he advocated for the establishment of a Danish
computer development lab, and eventually one was created, which became
A/S Regnecentralen. He then advocated for university reform, and for the establishment of a Computer Science department at University of
Copenhagen. When I heard they had hired him to start such a department,
I decided to enroll there after high school instead of goingto the
Technical University (DTH - now DTU). That turned out to be a bad idea;
he was fighting the established science faculty every step of the way.
The first year classes were the "gatekeeper math classes" for Maths and Physics, and the only introduction to programming class also had to
satisfy the requirements for "computations for actuaries".

It took a few years before a proper undergraduate computer science
curriculum could replace the "Mathematics with CS as a minor" that was
ther only option when I enrolled in Naur's first year. I could not hack
the math class, and dropped out after the fall semester. I would have
done better if I had had the option of taking the "math as a toolkit"
class that many other science programs there had available. The "math
for mathematicians class" had only a 50% pass rate, and many of those
passing were people that had failed it the year before.

I had been there less than a week, when I discovered that the two GIER
machines in the basement of the math department were often free in the
evening, and I started playing with the computers at class instead of
going to classes in the day.

When I realized I was going to fail out, it was time for plan B. I knew
what I wanted to do; now I had to find out what I needed to learn to get
a job where I could do that, and then figure out how and where to learn
it. In the end, I discovered that the university was building a
campus-wide academic computer center, and I got hired as a programmer
trainee. Within two years I was maintaining operating systems programs
on the Univac 1100 mainframe and writing PDP-11 code for the remote
terminal concentrators. The rest is history.

--- MBSE BBS v1.1.1 (Linux-x86_64)
* Origin: A noiseless patient Spider (3:633/280.2@fidonet)