About Two Months of Computing


An Addendum to Mr. Walker's
Three Years of Computing

by Tim Irvin
August 22, 1995

I first read of the “196” palindrome problem in a 1984 Scientific American, “Computer Recreations” column. Immediately intrigued, I toyed with a few half-hearted attempts to write a 196 palindrome crunching program for a brief period using the DOS-provided Basic and then Qbasic compilers. After tons of array overruns and lacking motivation to finish the job, the attempt to formulate a program to do the job was abandoned to folks who have time to do such things.

My recent job within an aircraft simulation company found me freshly exposed to software and supercomputing, on hardware platforms now the size of large stereo cartons with very capable speeds. I wondered how these machines would fare on number crunching exercises like the “196” problem. While absorbing a quick self taught tutorial in C, I imagined that I could create a program to to finally mount my personal assault on the “196” problem.

Before I set out to write such a program, I one day decided to see what the Internet could uncover regarding this mathematical curiosity. Using the “Yahoo” site search engine (keyword “palindrome”), I within seconds was directed to the formidable Fourmilab site and the host of greatly entertaining stuff on Mr. Walker's home page. I delighted to his article Three Years of Computing which included the program which had carried his own quest forward to grow a one million digit number.

Knowing that a spare Concurrent Computer Corporation Maxion model 9502 lay idle on the LAN at my simulation company, I enlisted the aid of programmer Larry Simkins to download Walker's program and file containing the huge number (why write a program when you can “borrow”), all in our “spare” time (cough, cough, sputter…).

Mr. Simkins, likewise intrigued, quickly began and concluded the task of setting up our machine with Walker's file and program. Computing was initiated at 3:15pm PDT, July 5th, 1995. After several days, working through some bugs unique to the host operating system and the C compiler, the program was running nearly continuously, excepting a daily backup and sparse foreground activity on the spare machine.

On Tuesday morning, August 22nd, the computer exited the program at the checkpoint specified. It had crunched the algorithm to add an additional million digits. We now have a two million digit number.

The speed at which the additional million digits was added is a telling improvement in performance over Walker's Sun workstation, 3 year effort, but not apart from what one would certainly expect of new technology. It is probably slow compared to what one could actually achieve using parallel processing and even better machines. The company recently added a Silicon Graphics Onyx supercomputer packed with 16 CPU's. Simkins and I have toyed with the idea of rewriting the program to use this capacity, but our network access to this new machine was strictly prohibited by our system administrator—she has more serious minded and legitimate customers to support.

The checkpoint file containing the two million digit number is yours to download if you want to continue the Quest yourself. Larry and I reset the program to exit at three million and have begun computing again.

Anyone else actively working this problem?

Tim Irvin