interesting Univac ad from 1956

[Don Shepherd]

I came across this ad the other day on our favorite auction site, and had to get it.

Univac ad

When I wrote FORTRAN programs for the Univac 1108 in 1974, I don't recall actually checking my denominators for zero before performing a division operation, but it's good to know the Univac would have scolded me if a zero denominator was detected. But it begs the question, what would have happened then? Apparently the computer stopped and printed out that message, presumably to the operator. What would the poor operator do upon witnessing this carelessness on the part of the programmer? Called him/her (yes, there were female programmers in those days and they were darned good) at home? Nothing? Who knows. Who cares!

My own code would probably never have encountered this situation because I don't recall doing any divide operations; our programs were oriented toward updating tape files with new information, and they were not calculation-intensive. I do teach my math students today about division by zero being "undefined," but I doubt that that fact is of much interest to them.

Most good calculators give you a "DIV by 0" today. Univac mainframes did it in 1956.

[rprosperi]

(06-05-2014 10:40 PM)Don Shepherd Wrote:  I came across this ad the other day on our favorite auction site, and had to get it.

Univac ad

When I wrote FORTRAN programs for the Univac 1108 in 1974, I don't recall actually checking my denominators for zero before performing a division operation, but it's good to know the Univac would have scolded me if a zero denominator was detected. But it begs the question, what would have happened then? Apparently the computer stopped and printed out that message, presumably to the operator. What would the poor operator do upon witnessing this carelessness on the part of the programmer? Called him/her (yes, there were female programmers in those days and they were darned good) at home? Nothing? Who knows. Who cares!

My own code would probably never have encountered this situation because I don't recall doing any divide operations; our programs were oriented toward updating tape files with new information, and they were not calculation-intensive. I do teach my math students today about division by zero being "undefined," but I doubt that that fact is of much interest to them.

Most good calculators give you a "DIV by 0" today. Univac mainframes did it in 1956.

Same year, I was learning Focal (the original one) and Fortran on a PDP-8. The math teacher warned us about the "dangers" of division by zero, and cautioned us to test for and handle it with some special message (who knows, perhaps he had seen the Univac Ad). Being annoying students, a couple of us decided that the special 'message' would be to rewind the DEC-TAPE (anyone here remember those?). That being in the realm of 'system programming' and therefoe impossible (in other words, he didn't know how either) we had to settle for ringing the bell on the teletype - (For i=1 to 100; Print chr$(7); next i) You get the idea...

So you can imgine how hard we worked to create data sets that would generate zero denominators. Much to the annoyance of the other math teachers.

Thanks for that Don.

[Dave Frederickson]

(06-05-2014 10:57 PM)rprosperi Wrote:  Same year, I was learning Focal (the original one) and Fortran on a PDP-8. The math teacher warned us about the "dangers" of division by zero, and cautioned us to test for and handle it with some special message (who knows, perhaps he had seen the Univac Ad). Being annoying students, a couple of us decided that the special 'message' would be to rewind the DEC-TAPE (anyone here remember those?).

I learned Dartmouth BASIC on a PDP-8/I in '74. The DECtapes were at the school district office.

Dave

[GeorgeOfTheJungle]

(06-05-2014 11:45 PM)Dave Frederickson Wrote:  

(06-05-2014 10:57 PM)rprosperi Wrote:  Same year, I was learning Focal (the original one) and Fortran on a PDP-8. The math teacher warned us about the "dangers" of division by zero, and cautioned us to test for and handle it with some special message (who knows, perhaps he had seen the Univac Ad). Being annoying students, a couple of us decided that the special 'message' would be to rewind the DEC-TAPE (anyone here remember those?).

I learned Dartmouth BASIC on a PDP-8/I in '74. The DECtapes were at the school district office.

Dave

In '74 I was learning (hp) BASIC on a 9830A, the "mass memory" was a 1 mega bytes hard disk in a rack the size of a refrigerator. Division by 0 was error 103 and the result was "infinite":

   

[davetheguru]

On those "old" mainframes (boys and girls) I remember if you divided by zero, it would create an exception that resulted in memory being "dumped" -- the next day, you would see a huge printout in our box (where your "output" would be, literally a mail box in the room for printouts, your card deck[program], etc.) that was your memory dump; the whole partition of RAM printed out in hex (the whole 64K if your were lucky). Then you could go thru the dump and hopefully find the problem that caused it. gdb would have been a godsend in those days!
You young whipper-snappers have it so easy these days.

[Don Shepherd]

(06-06-2014 01:22 AM)davetheguru Wrote:  the whole partition of RAM printed out in hex

Or octal on the Univac 1108, a 36-bit machine. You had a runstream card (JCL to IBM'ers) to generate a postmortem memory dump only if the program aborted, no sense in wasting all that paper otherwise. It had some trace-back features so that if you had a long listing of your FORTRAN program that showed the Sleuth code or assembler, you could find the offending source line without too much work.

Or you could just be a very careful programmer and avoid errors like this!

The relatively primitive debugging abilities of the system did encourage good programming habits, including the infamous "desk checking", which nobody does anymore I'm afraid.

[Don Shepherd]

Speaking of octal on the Univac 1108 reminds me of the decimal <--> octal conversion function of the HP-65. There were plenty of 32-bit mainframes (IBM 360) and 36-bit mainframes (Univac 1108) around in 1974 when the 65 was released, but HP chose only octal conversion, not hex. I wonder why (perhaps difficulty of representing A-F on the LEDs)? Were most of the minis of that era octal-based?

[Dave Frederickson]

(06-06-2014 12:41 AM)GeorgeOfTheJungle Wrote:  In '74 I was learning (hp) BASIC on a 9830A, the "mass memory" was a 1 mega bytes hard disk in a rack the size of a refrigerator. Division by 0 was error 103 and the result was "infinite":

The Univac I worked on in the Navy in 1976 had 32K of core memory. The RD-273 disk drive was four 12-inch, 20-surface, disk packs with hydraulic actuators. It was the size of two 'fridges. But the coolest thing was the Xynetics flatbed plotter. The "head" that held the plotter's pens had no mechanical linkage. It was driven electromagnetically.



Dave

[rprosperi]

(06-06-2014 02:05 AM)Don Shepherd Wrote:  Were most of the minis of that era octal-based?

Yes, pretty sure the PDP-8 was an octal-based machine; the PDP-11 I got to use in college was as well, IIRC, as were Data General machines (Nova, Eclipse?).

I think the Vax was the DEC machine that ushered in use of Hex, but this is all rather fuzzy; in fact it's a bit like trying to recall a dream - the harder I try to recall details, the furthur away they seem to be sliding...

[Les Bell]

(06-06-2014 03:24 AM)rprosperi Wrote:  Yes, pretty sure the PDP-8 was an octal-based machine; the PDP-11 I got to use in college was as well, IIRC, as were Data General machines (Nova, Eclipse?).

I think the Vax was the DEC machine that ushered in use of Hex, but this is all rather fuzzy; in fact it's a bit like trying to recall a dream - the harder I try to recall details, the furthur away they seem to be sliding...

The PDP-8 was a 12-bit machine, which lent itself to octal, since a word was expressed as 4 octal digits. I'm not sure about the DG machines - I used to have some Nova documentation around here somewhere, but I think it's long gone.

The Intel 8080, despite being 8-bit, was another machine that lent itself to octal; for example, a MOV instruction was always 1xy where x and y specified the destination and source locations, respectively, from this list:

B = 0
C = 1
D = 2
E = 3
H = 4
L = 5
M = 6
A = 7

It didn't take long before one could assemble 8080 machine code in one's head, although I've forgotten it all in the years since.

[rncgray]

The PDP-11 was 16 bit and the VAX 32 bit.

Richard

[rprosperi]

(06-06-2014 01:12 PM)rncgray Wrote:  The PDP-11 was 16 bit and the VAX 32 bit.

Richard

Yes, of course, but wasn't the PDP-11 assembler, et.al. Octal oriented?

[Mark Hardman]

(06-06-2014 03:24 AM)rprosperi Wrote:  

(06-06-2014 02:05 AM)Don Shepherd Wrote:  Were most of the minis of that era octal-based?

Yes, pretty sure the PDP-8 was an octal-based machine; the PDP-11 I got to use in college was as well, IIRC, as were Data General machines (Nova, Eclipse?).

Yes, the Data General minis were octal-based.

Interesting trivia, WordPerfect was originally developed and marketed for a Data General Nova 4.

I have a complete Nova 4 CPU and rack in my garage that I hope to restore to working condition someday.