Lecturing...

[compilerbitch]

Looks like I'm going to be giving a proper for-real Cambridge undergrad lecture!

December 3rd, 10am, in the big lecture theatre in the William Gates Building, to all of the Part 1B Computer Design course (about 120 people). The subject is going to be reconfigurable computing -- i.e. using hardware compilation and FPGAs to build cut price supercomputers that have an entirely non-Von Neumann architecture. Obviously, the subject is very close to my heart, and getting to do some real lecturing as a PhD student, especially this soon after starting here, is a really big thing for me.

So, *superdupermegabounce*

Happy Sarah. This has cheered me up a lot, especially after the wobble I had at the weekend over the allergy/arthritis/not being able to go and see people thing.

Comments

[davefish.livejournal.com]

Ooh, thats really quite an exciting topic. I didn't realise that you were doing such interesting research.

[compilerbitch.livejournal.com]

Building supercomputers out of FPGAs is something I'm interested in, have been for years really. It's not actually my PhD project though -- I'm actually working on techniques that will allow very large chips to be designed and validated. This is actually, if a bit less sexy sounding, much more important. Current tools have just about hit their limit, and aren't really doing a good job of handling the bigger designs that are currently being contemplated by chip companies. It's not so much the designing as the getting right that's the problem -- to build a custom chip costs about $2 million, and if you get it wrong and need another try it will cost the same again. And again. So you have to get it right first time -- what I'm working on is some new theory that will allow tools to be created that can identify whole categories of problem by static analysis, so the chip doesn't need to be built or even simulated.

Looks interesting so far -- hopefully the first paper will go into Designing Correct Circuits in Barcelona next spring.