Paper Info
Title
Component-based programming techniques for coarse-grained parallelism
Abstract
Components can be used to implement coarse-grained parallelism on large parallel systems. A parallel component is a piece of parallel code that can be executed in parallel on a set of processors or cores and has a predefined interface to be coupled with other components. Depending on the internal programming and memory model, a component may consist of computation and communication phases or, alternatively, of shared memory code. The interfaces are used for data exchange. More complex parallel programs are built up using parallel components and a flexible component interaction structure. In this article, the programming with parallel components for designing efficient programs for parallel execution platforms with a distributed address space is discussed and a mechanism for the specification of parallel components with communication interfaces is presented. The execution of these components can be adapted to the architectural characteristics of multicore clusters with their specific communication structure. The approach is applied to application programs for the solution of large systems of ordinary differential equations.
Year
Venue
Keywords
2011
SpringSim (HPC)
internal programming,complex parallel program,component-based programming technique,coarse-grained parallelism,large parallel system,parallel code,specific communication structure,flexible component interaction structure,parallel component,communication phase,communication interface,parallel execution platform
Field
DocType
Volume
Address space,Parallel Extensions,Implicit parallelism,Shared memory,Computer science,Task parallelism,Parallel computing,Parallel programming model,Data parallelism,Multi-core processor,Distributed computing
Conference
43
Issue
ISSN
Citations 
2
0735-9276
0
PageRank 
References 
Authors
0.34
9
3
Name
Order
Citations
PageRank
Jörg Dümmler1375.90
Thomas Rauber241564.60
Gudula Rünger360890.35