Another "Selection of the 'right' data size" post

Thread Starter


Joined Apr 11, 2010
This post reminded me of an old story. I was working as a student programming assistant at Georgia Tech in the 70s, when an unusual encounter occurred. A professor came into my office, acting a little agitated, when he dropped onto my desk, a large listing on greenbar (does anyone know what I am talking about?).

Then barely suppressing anger, he stated, "My da*n programs work!" :mad:

This being the first time I had ever heard this complaint (and the last time as it has turned out), I was a little confused myself.

Turns out, he had a couple programming examples of algorithms that he used to demonstrate numerically unstable or ill-conditioned problems in numerical analysis. When he used to run these on the university's computer system - they would not find an answer to the problem.

However, we had upgraded to a new system And his programs "worked", giving the correct answer every time. The upgrade went to a 64 bit processor and the increased word length prevented the round-off errors that had caused his algorithm to fail previously.

Happily, I taught him how to define his variables as single-precision instead of using the default precision, and his programs stopped working again! :D
Last edited:


Joined May 11, 2009
It would have been funny to be at that class. Then he demonstrated the program. "And now class as we can all see, this"


Joined Apr 24, 2011
I well remember greenbar from those bad old days at college. Mind you I took my first programming class in 7th grade (1970) where we has a teletype for input/output and paper tape for "mass storage." You would hit LIST, turn on the tape punch, then hit ENTER and get a listing to use next time. 3 years later we got a huge system upgrade, HP system with a drum memory, video terminals, and each user account got 20K of storage.

Switch to freshman year of college (1975). Punch cards. You had to BUY them, then wait to punch them, then wait for someone to run them, wrap the greenbar around your deck and drop it into the OUTPUT bin. Took an hour to run the program once.

Sophomore year I was in there during the summer, no one to run "the computer" (which was actually 2 counties away) but I had permission to run my own jobs to drop my deck into the hopper and hit LOAD... as the last punch card was being spit out the printer came to life and printed my entire output.

Wow, the mainframe actually was fast.