Program
MUSEPAT program
The MUSEPAT 2013 will take one and half days, Aug 19th full day and the morining of Aug 20th.
| Monday 19 | Tuesday 20 | |
| 9:30 - 11:00 |
Opening Session Keynote 1: "Conflict-free Data Types" |
Keynote 2: "Understand your parallel program challenges and potential before you write it" |
| 11:00 - 11:30 |
Coffee Break |
Coffee Break |
| 11:30 - 13:00 |
Session 1: Performance Analysis and Algorithms Self-timed Scheduling and Execution of Nonlinear Pipelines with Parallel Stages MVA-based Probabilistic Model of Shared Memory with Round Robin Arbiter for Predicting Performance With Heterogeneous Workload MHS2 : A Map-Reduce heuristic-driven minimal hitting set search algorithm |
Session 3: Programming Models and Optimization Handling Parallelism in a Concurrency Model On the Relevance of Total-Order Broadcast Implementations in Replicated Software Transactional Memories How to Cancel a Task |
| 13:00-14:30 |
Lunch |
End of MUSEPAT and Lunch (included) |
| 14:30 - 16:00 |
Session 2: Testing and Debugging Automatically Repairing Concurrency Bugs with ARC A Modular Approach to Model-Based Testing of Concurrent Programs A Dynamic Approach to Isolating Erroneous Event Patterns in Concurrent Program Executions |
|
| 16:00 - 16:30 |
Coffee Break |
|
| 16:30 - 18:00 |
Free Minds Corner |
Keynote Speakers
Nuno Preguiça, Universidade Nova de Lisboa, Portugal
Title: Conflict-free Data Types
Abstract: Replicating shared data is a fundamental mechanism for achieving fault-tolerance, availability and performance in large-scale distributed systems. Eventual consistency sidesteps the (foreground) synchronisation bottleneck, but leads to solutions that are ad-hoc, error-prone, and difficult to prove correct. In this talk, I will introduce Conflict-Free Replicated Data Types (CRDTs), a simple and theoretically sound solution for replicating shared data. A CRDT is a data types that demonstrates some simple properties: monotonicity in a semi-lattice and/or commutativity of concurrent operations. Any CRDT provably converges, provided all replicas eventually receive all operations. A CRDT requires no synchronisation: an update can execute immediately in a replica, with synchronisation being performed asynchronously. I will present how CRDTs are being used in practice in large-scale distributed systems and discuss how they can be used in other setting, such as modern multicore systems.
Bio: Nuno Preguiça has received a PhD degree from Universidade Nova de Lisboa (2003). Since 2003 he is Assistant Professor at UNL and a Research at CITI. In 2011 he spent his sabbatical leave at Inria. His primary research interests have been on data management, focusing on solution for groupware, mobile computing and cloud computing. His current focus is on solutions for cloud computing and multicore systems. He has received a Google Research Award in 2009 for his work on solutions for cloud data management based on CRDTs.
Zakhar A. Matveev, Intel
Title: Understand your parallel program challenges and potential before you write it
Abstract: In today's world, parallel multi-core and many-core hardware is everywhere, but parallel programming is still considered challenging by software developers and architects. We present a novel modeling methodology which helps overcome these challenges and understand trade-offs in the creation of effective parallel applications. The modeling approach assists in identifying potential data races and explore the parallel performance limits before the application is made parallel. It is realized in the Intel® Advisor XE toolset. Close correspondence between a serial program and an embedded parallel execution model makes this approach beneficial in educational settings. The approach enables a gentle introduction of students to the concepts of parallel computing.
Bio: Zakhar A. Matveev is a product architect at Intel® Software and Services Group, Intel Russia. Zakhar received his PhD in Computer Science from the Nizhny Novgorod State University of Architecture and Civil Engineering (2011). From the Nizhny Novgorod State University named after N.I. Lobachevsky he received M.S. degree in Mathematical Modeling and Computer Science (2005) and Bachelor degree in Math and Computer Science (2003). His professional interests are in the areas of parallel programming, software performance optimization, user interface design and usability.
Subscribe