21 Mar
2020
21 Mar
'20
12:43 a.m.
Back in the 70's, I bought an HP-67 for some ridiculous amount of money. I only
wrote one significant program on it, but I think folks here might find the idea
interesting. My X-Ray astronomy project at NRL (HEAO A1) was using a CDC 3800 for data
analysis. The machine had 48-bit words and used octal to display binary values. I
decided to write a program to convert the binary representation of a floating point number
into the numeric value.
For starters, it had to accept a pair of 24-bit values, expressed in (mock) octal. That
is, when I entered 12345, it was actually 12345(8), not 12345(10). So, the program had to
turn these values back into real integers. It then pulled out the exponent and mantissa,
did the appropriate calculations, and displayed the floating-point value. It worked, but
it wasn't all that fast. As I recall, it took about 30 seconds...
-r