Final exams, part 4: Asynchronous systems

At last, finals are over and the semester is complete! The last final was not really an exam at all. It was a presentation. We all had to do a final project, and then give a talk about it. And because this is my favorite class, I decided to overachieve like a bandit.

Instead of doing a final project, I did five final projects. One of them was writing a paper. One of them was modifying a program from the University of Manchester to work properly with Quartus II. Then I made a random number generator in hardware, and some more hardware for 16-bit cyclic redundancy checking. And a simulator for sets of production rules, which can also use IRSIM to do transistor-level simulations. You might say that I went a little nuts.

The reason I did all this is because this class was actually very, very interesting. It was aimed at graduate students (and I think I'm technically enrolled as a grad student for some reason) and the material was only dumbed down slightly! This has some good points and bad points.

Let's get the bad part out of the way first: a lot of people had trouble. For example, a homework project might have been assigned a week or two in advance, but when it comes due I'm sometimes the only person to have it done. The class included practical work and reading real research papers, and I guess that can be a little intimidating.

The good part is that it's something useful and interesting that you can really sink your teeth into. Some research papers are remarkably accessible, and the practical work is at least manageable. Let me give an example.

Here's a paper on how we might be making computers in five years. It talks about a neat new way of fabricating electronics called a nanowire crossbar array, and while it's stupendously tiny it's also not suitable to the way we design processors today. Imagine electrical signals as water slowly spreading through wires, and you need to ensure that water reaches a thousand faucets at the same time, but you don't control the plumbing. How can you do it? You can't. This is similar to the problem of clock distribution: all conventional processors require that you send a really fast signal to all parts of the chip, keeping it all synchronized like a coxswain on a rowboat. And you can't do that on a crossbar array. But using asynchronous techniques, you can make a processor on it. The processor will just be very different from almost every other processor ever made. Exciting to be on the forefront of a big new thing, isn't it?

When I came here I decided that I would use some of my Copious Free Time to go overboard on something worthwhile. This is it, and I'm glad I put in the extra effort.