01/24: Please submit a project proposal by Feb 5. You can submit earlier and/or discuss with me your ideas. The schedule lists details of what to include in the proposal.
01/24: Please volunteer to present your papers! If you haven't volunteered yet, then send me the dates that you prefer, either Feb 5 or Feb 12.
01/07: I put PDF of my introduction slides on the schedule.
Computer systems are facing many challenges today. In computer architecture, one of the most difficult current challenges relates to memory. After more than 40 years of success, the end is possibly near! It is uncertain how and whether DRAM can be scaled beyond 22nm to be reliable, power efficient and performance effective. Consequently, this has set off a race to redesign the memory hierarchy to either accommodate DRAM's problems, or replace DRAM altogether. The seminar will investigate these new memory architectures to understand their effectiveness for different workloads and execution scenarios. We will read and thoroughly discuss recent literature on the topic from leading conferences and journals. Students will present the papers and direct the discussion. In addition, students will undertake significant course projects to explore ideas related to the new memory architectures.
This course is a "graduate seminar." It is not a lecture course. The goal of this seminar is very different from a lecture course. In this seminar, we want to explore a new topic, possibly discovering new research directions and opportunities. While the seminar will have a certain broad set of topics, it will also be flexible enough to accomodate interests of seminar participants and new ideas that we develop at the semester progresses. To be successful, this seminar will require active and regular participation by everyone. You should also be highly motivated to contribute and explore new ideas.
We will read and discuss a number of papers, averaging about 2 to 3 per week. Student participants will be responsible for presenting these papers. The course will also give students the opportunity to use a software simulator and experiment system, OCCAM, to implement new techniques for memory. (That is, a project is required in the seminar.)
1. It's a new and exciting topic!
2. You'll get your hands dirty and get to play with cool technology!
3. You love systems programming and low level software/hardware development!
4. You have lots of ideas and can't wait to try them out!
5. You want to get involved in research right away!
6. You love a challenge and doing something creative!
7. You want to learn how to present and critically read research papers!
8. You want to see and learn how the research process works!
9. You want to learn how to write and prepare a paper for publication!
10. You reversed engineered a portion of Microsoft Windows as a kid (and then
switched to Linux)!
1. Paper presentations. Each student will do 2 to 4 presentations (depending on number of participants). The presentations will be done in PowerPoint and should be approximately 20 minutes in length (10-12 slides).
2. Seminar project. Everybody taking the course for a letter grade will do a significant project. Some topics will be suggested, but you are welcome to develop your own project.
3. Seminar participation. It is expected that everybody, regardless of letter grade or audit grading option, will attend all seminar meetings (unless sick or have a previously excused commitment) and contribute to the seminar. It's critical to read every paper beforehand.
4. Paper evaluations. We will use a "program committee" style to discuss papers; this means everybody will have to write and submit a paper review before we meet to discuss the paper.
More details about the specific requirements of the seminar will be given at the first meeting.
Grades will be determined based on three criteria:
1. Presentations - 30% of grade.
2. Project - 40% of grade.
3. Participation - 30% of grade (participation is based on #3 and #4 above).
To do well in the seminar, you need to do well in all three criteria. There are no exams or quizzes.
Semester schedule (subject to change)
We will use this review form in preparing our evaluations of the papers that we read. The review form is due via e-mail one day before a paper is to be discussed (i.e., 24 hrs in advance -- not 12 hrs in advance at mid-night the night before). All students must write a review and submit the form by the deadline. Come to the seminar prepared to talk about your review.
A set of possible projects will be described in class. You can also propose your own projects.