on Multithreaded Architectures and Applications
May 25, 2012
be held in conjunction with the
Multithreading (MT) programming and execution models, as
well as hybrid programming with accelerated architectures, are now part of the
high-end and mainstream computing scene. This trend has been driven by the need
to increase processor utilization and deal with the memory-processor speed gap.
Recent and upcoming examples architectures that fit this profile are Cray's XK6
and XMT, IBM Cyclops, and several SMT processors from Sun (Victoria Falls), IBM
(Power7), or Intel, as well as heterogeneous clusters with accelerators from
AMD (ATI), NVIDIA, and Intel. The underlying rationale to increase processor
utilization is a varying mix of new metrics that take performance improvements
as well as better power and cost budgeting into account. Yet, it remains a
challenge to identify and productively program applications for these
architectures with a resulting substantial performance improvement.
The MTAAP 2012 workshop is a full-day meeting to be held at the IPDPS 2012 focusing on Multithreading Architectures and Applications. This workshop intends to identify applications that are amenable to MT and the MT programming and execution models as well as the underlying architectures on which they can thrive. The workshop seeks to explore programming frameworks in the form of languages and libraries, compilers, analysis and debugging tools to increase the programming productivity. Topics of interest, of both theoretical and practical significance, include but are not limited to:
08:30 - 08:45 MTAAP 2012 Welcome
08:45 - 10:15 Runtime and Scheduling
08:45: Resilience to Various Failures for Read-mostly In-memory Data Structures
Larry Kaplan (Cray Inc., USA); Preston Briggs (University of Washington, USA); Miles Ohlrich (Isilon/EMC, USA); Will Leslie (Independent Contractor, USA)
09:15: Scheduling OR-parallelism in YapOr and ThOr on Multi-Core Machines
Ines Dutra (University of Porto, Portugal); Ricardo Rocha (University of Porto, Portugal); Vitor Costa (Universidade do Porto, Portugal); Fernando Silva (Universidade do Porto, Portugal); Joao Santos (CRACS INESC-Porto LA and University of Porto, Portugal)
09:45: A discussion in favor of Dynamic Scheduling for regular applications in Many-core Architectures
Elkin Garcia (University of Delaware, USA); Daniel A Orozco (University of Delaware, USA); Robert S Pavel (University of Delaware, USA); Guang Gao (University of Delaware, USA)
10:15 - 10:45 Morning Break
10:45 - 11:45 Keynote: Semantic Technologies and Complex Networks in Support of Climate Knowledge Discovery
11:45 - 13:30 Lunch
13:30 - 15:00 Algorithms and Applications
13:30: An approach to parallelize Kruskal's algorithm using Helper Threads
Anastasios Katsigiannis (National Technical University of Athens, Greece); Nikos Anastopoulos (National Technical University of Athens, Greece); Konstantinos Nikas (National Technical University of Athens & Computing Systems Laboratory (CSLAB), Greece); Nectarios Koziris (National Technical University of Athens, Greece)
14:00: Merge Path - Parallel Merging Made Simple
Saher Odeh (Technion, Israel); Oded Green (Georgia Institute of Technology & School of Computational Sciecne and Engineering, USA); Zahi Mwassi (Technion, Israel); Oz Shmueli (Technion, Israel); Yitzhak Birk (Technion, Israel)
14:30: Scalable Multi-threaded Community Detection in Social Networks
Jason Riedy (Georgia Institute of Technology, USA); David A. Bader (Georgia Institute of Technology, USA); Henning Meyerhenke (Karlsruhe Institute of Technology, Germany)
15:00 - 15:30 Afternoon Break
15:30 - 17:00 Architecture
15:30: An Early Evaluation of the Scalability of Graph Algorithms on the Intel MIC Architecture
Erik Saule (The Ohio State University, USA); Umit V. Catalyurek (The Ohio State University, USA)
16:00: PMU-guided Priority Adjustment to Guarantee Thread Performance on IBM POWER SMT Processor
Zhengyu He (Georgia Institute of Technology, USA); Bo Hong (Georgia Institute of Technology, USA)
16:30: Design Trade-offs among VLIW SIMD and Multi-core schemes
Yaohua Wang (National University of Defence Technology, P.R. China); Shuming Chen (National University of Defence Technology, P.R. China); Kai Zhang (National University of Defence Technology, P.R. China); Hu Chen (National University of Defence Technology, P.R. China); Xiaowen Chen (National University of Defence Technology, P.R. China)
Semantic Technologies and Complex Networks in Support of Climate Knowledge Discovery
Director, Marketing and Business Development
Historically the most data-intensive problems have been in the HPC domain, including areas such as seismic processing, climate simulation, intelligence, and bioinformatics. Modern science is characterized as both compute and data-intensive, and multi-disciplinary and multi-institutional. Data volumes and complexity are defying standard approaches to interpretation and driving new methods to support deeper analytics and knowledge discovery. The terminology Big Data refers to this deluge and generally implies new methods of data organization and new analytical tools. The leading technology for linked data analytics is semantic Web 3.0 technology. With semantic web technology, data sets are represented in the form of graph networks that capture the linkages between data objects. Such graph networks are highly extensible, and thus are ideally suited for both structured and unstructured data. When data is cast in this form, analytical approaches can go beyond search to support inference rules and reasoning, and is thus able to uncover deeper relationships inherent in the data.
In climate science, model-generated and observational data represent one of the largest repositories of scientific data in any discipline. Geoscientists gather data faster than they can be interpreted. The Climate Knowledge Discovery effort is a community initiative to both educate climate researchers about the potential of using knowledge discovery tools and to conduct research into ways and means of applying graph-theoretic techniques to multi-disciplinary climate model data. Tools that employ a combination of high-performance analytics, with algorithms motivated by network science, nonlinear dynamics and statistics, as well as data mining and machine learning, could provide unique insights into challenging features of the Earth system, including extreme events and chaotic regimes. Complex networks have been identified as one very promising solution. By representing the climate system as networks, the understanding of observed climate phenomena, complex relationships in the global climate system, and anticipation of the consequences of climate change can be improved. Networks constructed from climate data have been shown to detect natural changes in the climate system. There is also the potential to enhance regional climate predictions over land by exploiting atmospheric teleconnections. Such "climate networks" could be as large as millions or billions of nodes. Investigating data at this massive scale will require advanced parallel or multithreaded computing technologies and open-source semantic software stack supporting direct analytical queries and knowledge synthesis. The breakthroughs needed to address these challenges will come from collaborative efforts involving several disciplines, including end-user scientists, computer and computational scientists, computing engineers, and mathematicians.
Submitted manuscripts may follow the IEEE conference style: not exceed 12 single-spaced double-column pages using 10-point size font on 8.5x11 inch pages, including figures, tables, and references. Alternatively, authors can submit up to 15 single-spaced pages using 12-point size font on 8.5x11 inch pages, including figures, tables, and references. Authors may submit additional material as an appendix to their submission, but there is no guarantee that this material will influence the review process. Manuscripts must be submitted electronically and in PDF format. Submissions will be judged on correctness, originality, technical strength, significance, quality of presentation, and interest and relevance to the workshop attendees. Submitted papers may not have appeared in or be under consideration for another workshop, conference, or journal.
MTAAP Paper submission: http://edas.info/N11893
The proceedings of this workshop will be published together with the proceedings of other IPDPS 2012 workshops by the IEEE Computer Society Press. Accepted papers will have a page limit of 10 pages, and authors can purchase an additional 2 pages, for a total of 12 pages maximum.
Information and papers from the earlier MTAAP workshops are available:
For more information on MTAAP or if you have any questions please contact the workshop chair at firstname.lastname@example.org.
This website is hosted by the HPC Lab at Ohio State University.