Computer Architecture has been the most interesting course (ECE3391). We need faster and better processors always and it is always a tradeoff. With Professor Kaeli, we could see the most recent solutions and new questions posed in the field.
The Paper review was a review of one of the papers distributed in class for the course. I did the review in one of the papers in Branch Prediction, Fast and Accurate Branch Prediction, by Brad Calder and Dirk Grunwald. Branch prediction is very important, becuase it becoming a more costlier bootleneck, as processors become wide-issue and superscalar. When you don't know what you are to execute, how can you go ahead and feed so many pipelines and execution units. The paper review blossomed into a survey and study of all the architectures in literature. This paper review will definitely help a novice get a good idea about branch prediction, and for somebody who knows, he can well get an idea of so many architectures I came across and assess them and maybe help me in knowing things better!
The project of this course was to simulate a proposed architecture and verify the results obtained for that architecture. I did the project with Sriram, and we simulated a few hybrid caches using ATOM, a tool available on the DEC ALPHA. Caches are very important that they tend to hide the lack of speed of the I/O subsystem.