For the past semester I, six students and three teachers from Leiden University and the Delft University of Technology have been conducting an unusual type of seminar. Instead of the usual University seminars where everybody holds a presentation about some existing topic and also does a project, we've been doing some real research. Our goal was to look at some of the optimization challenges faced by restructuring compilers in the presence of certain kinds of datastructures (most notably, pointer structures, which a restructuring compiler really can't do anything with), and also for certain notoriously hard to optimize algorithms (the focal point has been Tarjan's algorithm for finding strongly connected components in a graph, which is nearly impossible to parallelize).
Last Tuesday, we had the last of our meetings, the formal end of the seminar. While we didn't solve the problem, we did get some nice results. And more importantly for me, this seminar has been the foundation for what will become my thesis, and I finally have a real clear picture of what I'll be doing. Specifically, I will be continuing on the pointer structure work, and currently I'm focussing on transforming a linked-list representation of a matrix algorithm into something a restructuring compiler could work with.
What that means is that if all goes well, it'll be only a few more months and then, finally, I'll get my Masters. Here's to hoping!
I've been getting a lot of work done this week, I finally feel like I'm getting out of my usual Spring slump. It feels good!
Subscribe