ISCA

It is a pleasure to introduce the ISCA 2009 program representing the state-of-the-art in computer architecture research. Serving as ISCA Program Chair gave me a front row seat to observe the vitality of the research in our field, and the dedication of the members of the organizing committee, program committee and reviewers. It has been an honor to work with such an incredibly talented team.

In selecting the program committee (PC) my primary goal was to form a group of recognized experts with interests that matched the predicted mix of submission topics. Achieving balances between industrial and academic representation, and between experience and new talent was also important. This year's PC had roughly 40% membership from industry, 57% of the members had served in one of the last three ISCA PCs and 32% were serving in the ISCA PC for the first time. More significantly, over 94% of the papers had at least two reviewers rating themselves as experts in the subject area, indicating very good topic area coverage.

The selection process was heavily influenced by what I learned serving on Mark Hill's (ISCA'05) and Brad Calder's (ISCA'07) PCs. I am especially grateful to them for their generosity of advice throughout. Reviewing was double blind, with typically four PC and one non-PC review per paper. I made all the reviewing assignments except for submissions I had conflicts with, which were managed by Josep Torrellas -- in total we had 1028 reviews. Authors were invited to read the reviews and to submit a short rebuttal, after which reviewers were then asked to re-score all papers. Re-scoring after rebuttals affected over 70% of all submissions, and caused five eventually accepted papers to improve their final rank by more than thirty positions. I believe that the combination of rebuttals and re-scoring improved the quality of our evaluation process significantly.

Recognizing that sometimes the external reviewer is the most knowledgeable for a given article, I invited them to participate in the rebuttals and re-scoring phases, and asked the PC member leading the discussion for each paper to be the advocate for the external reviewer. I also used a Review Committee, an idea introduced by Steve Blackburn at ISMM'08, which consisted of experts recruited before the submissions deadline who agreed to review between four and six submissions. Besides greatly simplifying the assignment of non-PC reviews, the Review Committee gave me more opportunities for expert reviewer overlap across similar papers, which helped process calibration.

In the PC meeting we discussed papers which had a majority of scores in the accept range, plus every paper that any PC member felt that we should discuss. I explicitly did not set a maximum number of papers to accept and did not publish the list of accepted papers until after the meeting was finished. The PC worked with no breaks and through lunch from 8:00am until 5:30pm, showing remarkable stamina, focus and professionalism. I am pleased that we have been able to continue the recent ISCA trend of a larger program, with 43 papers accepted out of 210 submissions (12 out of 45 among submissions with PC authors), and I encourage future committees to continue to strive for a large and more diverse program.