note
clintp
<i>From the <b>Not This Way</b> department</i>, a moral of speed:
<p>
In college, for a FORTRAN class, I was assigned the 8-queens problem. At the time I was also taking Pascal and teaching myself 8086 assembler and was a bit overwhelmed with the labwork at the end of the semester.
<p>
As the instructor explained it, the intent was to solve the problem recursively. (HINT: it's almost trivial this way. Search with google for "eight queens recursion bit arithmetic" and take the first hit for a particularly clever solution...) Anyway, I realized that I misplaced the assignment and had to do it the night before it was due. So over a 300-baud dialup into a PR1ME system (running PR1MOS 7!) with a line editor I hacked up a <a href="http://www.tuxedo.org/~esr/jargon/html/entry/brute-force.html">Brute Force</a> solution to the problem.
<p>
I used a 8-digit octal counter, each digit representing a column and each digit value representing a row on the chessboard. Add one, check for collisions, repeat until overflow to 9 digits of octal. As a test I did a 4x4 board. This worked, then 8x8. This was terribly slow. I started it as a phantom process, logged off, and went to bed at 1am.
<p>
The next morning I went into the lab to collect the results. The most *interesting* part was that from 1am to 2am I used 58 minutes of CPU time to solve the problem. The prof was mildly amused as 1. I was the only person to present a correct solution for that class, 2. that I was the first person he'd had to solve it through BF&I and 3. I had the absolute slowest version he'd ever seen. Nontheless, I got full credit for the assignment.
<p>
And just think: on comparative hardware in 2002, it'd run in (calculating...) under 4 minutes! Hooray for Moore's law!
196095
196095