Building high-quality concurrent software is a significant challenge. Issues such as performance, scalability, extensibility, and correctness, form a complex web of design decisions that software developers must navigate. In a number of well-understood software domains, reduction in the complexity of sequential software design has been enabled by the availability of collections of existing software components. In this paper, we present a component that provides similar benefits to developers of concurrent Ada-95 software that uses a replicated workers paradigm. Our strategy is to factor out concurrent aspects of a computation; this allows users to focus on the computational and data elements of an application. Our componentry supplies a skeleton in which to embed such elements that reduces the complexity and time required for software development, enables modular reasoning about correctness properties of concurrent software, and enables control of software performance characteristics. ...

ions in Stock Languages CIS TR 96-9 Matthew B. Dwyer Matthew J. Craig Department of Computing and Information Sciences Kansas State University 234 Nichols Hall Manhattan, KS 66506 (913) 532-6350 dwyer@cis.ksu.edu ABSTRACT Abstraction is a fundamental concept in software engineering. In this paper, we describe a class of abstractions that can simplify the construction of concurrent software, we call them coordination abstractions. Coordination abstractions capture communication, synchronization and topological information about concurrent computations. In the same way that a data abstraction hides the details of a type's implementation, a coordination abstraction hides the details of the coordination activities that are inherent in a concurrent computation. Coordination abstractions are different than other, more well known abstractions, but we contend that features in existing languages are sufficient for building useful, efficient coordination abstractions. We evaluate this hypoth...

this paper we discuss the particle-decomposition parallelization of a ######## ## #### code targetted to shared memory parallel architectures, following the parallelizing compilation approach, in particular utilizing the OpenMP language extensions. 1