And shout that word like Thomas Dolby!

I’m in the process of transitioning the site to its new domain name, sandsoftwaresound.net. So, here’s just a few philosophical thoughts in the meantime.

I’ve been reading “Weird Life” by David Toomey. His book is very well written — one of the best bits of science writing that I’ve read. He teaches nonfiction writing at UMass and deserves a tip o’ the hat for bringing this aspect of biology to a mass audience.

In one paragraph, he summarizes the essential elements of an experimental science. A science is a body of theory which is testable and verifiable through experimentation. Robust theory allows us to make predictions and to design new experiments. BTW, if the theory is valid, then experiments should be reproducible, too.

I like program profiling and performance analysis as a field because it is one of the few areas in computer science that exemplify experimental science. Performance events and counters give us the means to observe and measure the behavior of our programs in interesting ways.

I especially love it when I can predict an outcome such as the number of operations performed by a program. On the flip side, I especially hate it when I cannot make predictions. This is the main reason that I dislike performance events with ill-defined behavior or hardware that doesn’t provide a stable time reference. This situation is like giving an astronomer a rusty telescope with a crummy warped lens. How can an experimental scientist make observations and take measurements with a broken instrument?

These limitations aside, I would still encourage teachers and students to study performance measurement and analysis. These activities are a fun and different way to think about programs and the underlying computational engines. If enough of us band together, we even may be able to convince computer manufacturers about the need for well-defined, reliable performance monitoring and measurement!

If you’re looking for another good read, please consider Lance Fortnow’s “The Golden Ticket.” It’s an introduction to the P-NP problem and is also geared toward a broad audience. If a high school student (or anyone else!) thinks that computer science is “just programming,” please point them toward this book.