cse240a: Paper Summaries


Paper Summary Instructions

Unless a different deliverable is listed, you should turn in a short write up for each paper containing the following:

  1. A brief summary of the paper's main ideas (2 paragraphs or so).
  2. Your favorite idea from the paper.
  3. Three questions you have about the paper.
  4. Two suggestions for future work in the area.

Your target audience is your fellow classmates.

Brevity is key, but that does not mean that they should light on content. You should strive to distill the paper to its essence. The core of most papers can easily be expressed in a couple paragraphs. The rest is necessary as well, but it is mostly just there to back up the ideas.

Reading research papers is an important skill. You will do a lot of it, so you should to do it well. Our very own Bill Griswald has some useful tips on reading engineering research papers.

Turn the summary in via the web board in the discussion forum for that paper (there will be a new one for each paper). Use your name as the subject. This is a moderated forum, and we will hold the postings until 9am on the due date. Will then make them available for the class to peruse, and no more submissions will be accepted. At that time, we will turn off moderation, and you will be able to post responses to the summaries people have posted.

Paper summaries are to be completed on your own.

Papers

Paper 1: Early Computers

Due: October 9

The first paper is not really a paper, but an Architectural Multimedia Experience(tm)! Last year Geoff Voelker, a professor in our department co-taught a class with Ed Lazowska at the University of Washington and Steve Maurer at Berkeley on the history of computing. It was extremely well done, and they are nice enough to make audio and video of the classes available on line along with the slides.

Your first "reading" assignment is to listen or watch the second lecture of the quarter. It's available here. You want "Lecture 2: Steve Maurer, Ed Lazowska: Electronic Computing 1940-1970." The slides are in the table at the top, and the audio/video is in the table at the bottom. There are two sets of slides for this lecture.

The lecture covers the early days when there were very few abstractions hiding the hardware, so it's mostly about architecture to some degree or other. As you listen/watch think about how changes in both technology and customer demand drove changes the abstractions that the systems provided.

The rest of the lectures are worthwhile as well. The professors were able to get together a very impressive list of speakers. Almost every one is a luminary who lived through and actively participated in the history they are describing. They cover a huge range of topics far beyond architecture. Some of the stories are amazing and very entertaining. Consider it part of your liberal computer science education.


Paper 2: Reducing memory latency via non-blocking and prefetching caches

Due: October 18

This is a classic paper that describes two cache optimizations: non-blocking caches and prefetching. It's available online here it here. And I've posted it to the web board.

As an aside, caches are among the most studied structures in computer architecture. Jean-Loup Baer (one of the authors of this paper) estimated that by 2000 there had been over 2000 papers publish on the subject of memory caching.

Full bibliographic information is available at the ACM Digital Library.