DSProPP 2014 - International Workshop on Multi-core/Many-Core Systems: Programming Paradigms to Practice

Massively parallel distributed systems are playing a critical role in scientific and analytics research along with significant potential for impact in Industry verticals. These systems are at the heart of Top-500 supercomputers and are the key enablers of Big Data Analytics and Exascale Supercomputing. Compared to sequential applications, the repertoire of software engineering tools and methods for developing reliable and robust parallel applications is pretty challenging. The software engineering methodologies and tools for massively parallel systems will play a critical role to achieve the sweet-spot in performance productivity trade-offs. This includes code- refactoring, automatic parallelization, software patterns, testing and debugging, performance engineering and related software engineering aspects. The purpose of this workshop is to bring together researchers and practitioners with diverse backgrounds in order to advance the state of the art in software engineering, performance engineering, and tools in all areas of massively parallel multicore/many-core systems. Key trends that further emphasize the need for focus of software engineering aspects:
Multiple parallel languages and APIs ranging from GASNet, MPI/OpenMP, HPF at one end to PGAS languages such as UPC, X10 provide the spectrum of performance and productivity trade-offs.
Run-Time scheduling is an important aspect to gain performance.
Extracting good parallel speedup for data and/or compute intensive and irregularly- structured problems without resorting to full-blown concurrent programming remains a tough challenge.
Computer systems increasingly rely on heterogeneity to achieve greater performance, scalability and energy efficiency.
Because heterogeneous systems typically comprise multiple execution contexts with different programming abstractions and runtimes, programming them remains extremely challenging. Some of the challenges in this direction have been
Unified programming model for heterogeneous systems that span diverse
Automatically distributing data-parallel portions of a program transparently execution contexts including CPUs, GPUs, FPGAs, and the cloud is an important area of research. to available computing resources, including compute clusters for distributed execution and CPU and GPU cores of individual nodes for parallel execution.
To the extent that concurrency is exposed, it poses well-known risks: unintended races, nondeterministic behavior, deadlocks, and weak memory consistency. Debugging in the context of large no. of cores is quite challenging
Performance problems are often difficult to diagnose and fix: they may stem from task granularity and scheduling choices, or from false sharing and other cache coherence effects, which are often not under direct programmer control.
The workshop will be full day event comprising of keynote talks by eminent researcher(s) in the area of many-core software systems, followed by selected technical papers presentation and a panel discussion of distinguished contributors on: “Many-Core Systems: Current Trends & Future Directions”.
The workshop primarily aims to gather researchers and practitioners addressing the main challenges and share experiences in the emerging multicore and many-core software engineering and distributed programming paradigms. This workshop aims to provide a discussion forum for people interested in programming environments, models, tools and applications specifically designed for parallel multicore and many-core distributed systems.