Proceedings of the 26th ACM international conference on Supercomputing

On behalf of the Program Committee, we are pleased to introduce the technical program of the 2012 ACM International Conference on Supercomputing (ICS), the 26th in the series. The conference is by now a well established forum to present and discuss explorations that aim at pushing the limits of performance and capacity in large-scale computing systems, while preserving power efficiency, reliability, and programmability. In this context, papers have been solicited and submitted on parallel applications, architecture, hardware, accelerators, systems software, large scale installations, data center, grid and cloud computing, reliability, power efficiency, models of computation and of programming, and theoretical foundations of performance, for terascale to exascale systems. This year's program includes 36 technical papers selected out of 161 submissions (a 22.3% acceptance rate). There were more submissions of high quality and relevance than could be accommodated in a three-day technical program. Under these constraints, a Program Committee of 70 world experts on the variety of topics of interest to the conference has worked very hard to select a high quality program, assisted by 134 external reviewers and by the Submission Chairs, Francesco Silvestri and Francesco Versaci. The task was carried over a period of two months, with two rounds of reviews--the second deepening the results of the first one-and one week of email discussion among PC members to further refine the evaluation and prepare a preliminary synthesison each submission. A total of 641 reviews were provided in the process, with each paper receiving at least 3 reviews, and half the papers receiving 5 reviews. Final deliberations were made during the Program Committee meeting, held at the University of Padova, Italy, on March 9, 2012. The meeting was attended by nearly two-thirds of the PC members; most of the other members participated via conference call. The technical program is further enriched by the two keynote addresses. Yale Patt will reflect on avenues to improve the performance of the individual core that will benefit even exascale class systems. Michael Gschwind will present the Blue Gene/Q supercomputer design, and how it addresses the memory, power, scalability, communication, and reliability "walls". The Best Paper Award will be given according to the selection made by the audience among all papers. Thus, we invite you to carefully attend all talks, and to vote according to each paper's content, relevance, and presentation quality.