| Abstract |
It is our great pleasure to welcome you to PPoPP 2017, the 22nd ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming held in Austin, Texas during February 4-8, 2017, and co-located with the CGO 2017 and HPCA 2017 conferences. This year's symposium continues and reinforces the PPoPP tradition of publishing leading work on all aspects of parallel programming, including foundational and theoretical aspects, techniques, languages, compilers, runtime systems, tools, and practical experiences. Given the pervasiveness of parallel architectures in the general consumer market, PPoPP, with its interest in new parallel workloads, techniques and productivity tools for parallel programming, is becoming more relevant than ever to the computer science community. PPoPP 2017 received 132 submissions from countries all over the world. The submitted papers underwent a rigorous two-phase review process. To maintain fairness and uniform standards, the review process was double-blind, throughout. Almost all of the 132 submissions were reviewed in the first phase by four members of the combined PC and ERC. (A very small fraction received three reviews in the first phase.) All papers were assigned a discussion lead from the PC. After the first rebuttal phase and ongoing online discussions, reviewers reached a consensus to relegate half of the submissions. Their authors were subsequently notified and given the choice of withdrawing their papers. The submissions for which there was no clear consensus or that had only three reviews (very few) were retained for the second evaluation stage. In the second phase, the remaining papers received at least two additional reviews exclusively from PC members and some external specialists. After a second rebuttal period, PC and ERC members continued their online discussions and grouped papers in top, bottom and discuss categories. Finally, 35 PC members met in person over two half days from noon, November 5th through the afternoon of November 6th at the Department of Computer Science in Rice University, Houston, TX, and concluded the meeting by accepting (or conditionally accepting) a total of 29 papers. The 14 conditionally accepted papers were shepherded by volunteer PC members, and the final version was made available to all original reviewers for their approval. All in all, this process resulted in a manageable average load of 12 papers for PC members, 6 papers for ERC members, offered all authors the possibility to respond to all reviews, and helped ensure that reviewing efforts were focused on where they were needed the most. Because many quality papers could not be accommodated as regular contributions, all papers retained for the second review phase were invited to be presented as posters at the conference. As a result, 17 posters were included in the proceedings as 2 page abstracts. The posters were presented during a special two-hour late afternoon session. All authors of accepted papers were given the option of participating in a joint CGO-PPoPP Artifact Evaluation (AE) process. The AE process is intended to encourage researchers to conduct experiments in a reproducible way, to package experimental work-flows and all related materials for broad availability, and ultimately, to enable fair comparison of published techniques. This year saw a considerable increase in the amount of submitted artifacts: 27 versus 18 two years ago, almost equally split between CGO and PPoPP. The Artifact Evaluation Committee of 41 researchers and engineers spent two weeks validating and evaluating the artifacts. Each submission received at least three reviews and only eight of them fell below acceptance criteria. To help educate authors about result reproducibility, these papers were shepherded during a week's time by the AE Committee. Concurrently, the AE committee successfully tried an "open reviewing model", i.e., asked the community to publicly evaluate several artifacts already available at Github, Gila and other project hosting services. This enabled us to find additional external reviewers with access to HPC servers or proprietary benchmarks and tools. At the end of this process all submissions qualified to receive the AE seal and their authors were encouraged to submit a two page Artifact Appendix to document the process. The success of a major conference like PPoPP very much depends on the hard work of all members of the organizing committee who volunteer their time in this service. We thank all PC and ERC members for their thoughtful reviews and extensive online discussions, with special thanks to the PC members who came from all over the world to the meeting in Houston and deliberated for two half days and read additional papers overnight to produce the PPoPP 2017 program. Several of the PC members volunteered to shepherd papers, and deserve special thanks for their efforts. The AE process was lengthy and demanding, and the AE Chairs, Wonsun Ahn (for PPoPP) and Joe Devietti (for CGO) and their team did an amazing job on that Herculean task. |
