IEEE IPDPS 2011
TechTalks from event: IEEE IPDPS 2011
Note 1: Only plenary sessions (keynotes, panels, and best papers) are accessible without requiring log-in. For other talks, you will need to log-in using the email you registered for IPDPS 2011. Note 2: Many of the talks (those without a thumbnail next to the their description below) are yet to be uploaded. Some of them were not recorded because of technical problems. We are working with the corresponding authors to upload the self-recorded versions here. We sincerely thank all authors for their efforts in making their videos available.
Intel Platinum Patron Night
Architecting Parallel Software: Design patterns in practice and teachingDesign patterns can systematically identify reusable elements in software engineering, and have been particularly effective in codifying practice in object-oriented software. A team of researchers centered at UC Berkeleyâ€™s Parallel Computing Laboratory continues to investigate a design pattern approach to parallel software; the effort has matured to the point that an undergraduate course was delivered on the topic in Fall 2010. This talk will briefly describe the pattern language itself, then demonstrate its application in examples from both image processing and game design.
Teaching Parallelism Using GamesAcademic institutions do not have to spend expensive multi-core hardware to support game-based courses to teach parallelism. We will discuss what teaching methodologies educators can use for integrating parallel computing curriculum inside a game engine. We will talk about the full game development process, from game design to game engineering and how parallelism is critical. We will show five game demos that mirror current trends in the industry and how educators can use in these games in the classroom. We will also show the learning outcomes, what parallelism topics are appropriate to teach students at various levels. We will demonstrate how to take games running serially and modify them to run parallel.
Starting Your Future Career at IntelIntel's Dani Napier will introduce why Intel is a great place to work-- it's challenging, has great benefits and is abundant with rewarding growth opportunities. She will expand on why parallelism is crucial to Intel's growth strategy and give an overview of the various types of jobs in which knowledge of parallel and distributed processing apply at Intel. Finally, Dani will explain the new hire development process and why Intel is the company that will help you become successful in your desired career path. Lauren Dankiewicz will discuss her background from the University of California, Berkeley. She gives an insightful and humorous commentary on the interview process at Intel, drawing similarities to dating. Lauren describes the excitement, the uncertainty, and what it takes to make the right choice! Listen to this fun and engaging real-life clip of how an intern became a full-time employee at Intel.
Opening RemarksIntel Platinum Patron Night will be held on Thursday evening, 5:30-8:30pm, in the Kuskokwim Ballroom. This will be an exciting opportunity for IPDPS attendees to network and learn about the Intel Academic Communityâ€™s free resources to support parallel computing research and teaching. Intel recruiters will share information about engineering internships and careers for recent college graduates.
25th Year IPDPS Celebration
SESSION 7: Numerical Algorithms
Automatic Library Generation for BLAS3 on GPUsHigh-performance libraries, the performance-critical building blocks for high-level applications, will assume greater importance on modern processors as they become more complex and diverse. However, automatic library generators are still immature, forcing library developers to manually tune library to meet their performance objectives. We are developing a new script-controlled compilation framework to help domain experts reduce much of the tedious and error-prone nature of manual tuning, by enabling them to leverage their expertise and reuse past optimization experiences. We focus on demonstrating improved performance and productivity obtained through using our framework to tune BLAS3 routines on three GPU platforms: up to 5.4x speedups over the CUBLAS achieved on NVIDIA GeForce 9800, 2.8x on GTX285, and 3.4x on Fermi Tesla C2050. Our results highlight the potential bene?ts of exploiting domain expertise and the relations between different routines (in terms of their algorithms and data structures).
Redesign of Higher-Level Matrix Algorithms for Multicore and Distributed Architectures and Applications in Quantum Monte Carlo SimulationA matrix operation is referred to as a hard-to-parallel matrix operation (HPMO) if it has serial bottlenecks that are hardly parallelizable. Otherwise, it is referred to as an easy-to-parallel matrix operation (EPMO). Empirical evidences showed the performance scalability of an HPMO is signi?cantly poorer than an EPMO on multicore and distributed architectures. As the result, the design of higher-level algorithms for applications, for the performance considerations on multicore and distributed architectures, should avoid the use of HPMOs as the computational kernels. In this paper, as a case study, we present an HPMO-avoiding algorithm for the Greenâ€™s function calculation in quantum Monte Carlo simulation. The original algorithm utilizes the QR-decomposition with column pivoting (QRP) as its computational kernel. QRP is an HPMO. The redesigned algorithm maintains the same simulation stability but employs the standard QR decomposition without pivoting (QR), which is an EPMO. Different implementations of the redesigned algorithm on multicore and distributed architectures are investigated. Although some implementations of the redesigned method use about a factor of three more ?oating-point operations than the original algorithm, they are about 20% faster on a quadcore system and 2.5 times faster on a 1024-CPU massively parallel processing system. The broader impact of the redesign of higher-level matrix algorithms to avoid HPMOs in other computational science applications is also discussed.
Challenges of Scaling Algebraic Multigrid across Modern Multicore ArchitecturesAlgebraic multigrid (AMG) is a popular solver for large-scale scienti?c computing and an essential component of many simulation codes. AMG has shown to be extremely ef?cient on distributed-memory architectures. However, when executed on modern multicore architectures, we face new challenges that can signi?cantly deteriorate AMGâ€™s performance. We examine its performance and scalability on three disparate multicore architectures: a cluster with four AMD Opteron Quad-core processors per node (Hera), a Cray XT5 with two AMD Opteron Hex-core processors per node (Jaguar), and an IBM BlueGene/P system with a single Quad-core processor (Intrepid). We discuss our experiences on these platforms and present results using both an MPI-only and a hybrid MPI/OpenMP model. We also discuss a set of techniques that helped to overcome the associated problems, including thread and process pinning and correct memory associations.