P. Brinch-Hansen. "Model Programs for Computational Science: A Programming Methodology for Multicomputers ". Concurrency: Practice and Experience, v.5, n.5, pp.407-423, 1993.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Experience shows the existence of a set of parallel algorithmic schemes that can be applied to problems of a different nature and always repeat the same communication structure. These algorithmic schemes have been studied and characterized by different authors with different points of view [9, 10, 11 12, 13, 14]. As it is presented in [15] we characterize a parallel paradigm by a general scheme of both data and control distribution. This scheme involves particular data dependencies and communication requirements among the processes. The different implementations of the distribution scheme derived from ....

P. Brinch-Hansen. "Model Programs for Computational Science: A Programming Methodology for Multicomputers ". Concurrency: Practice and Experience, v.5, n.5, pp.407-423, 1993.

.... from e orts to do performance measurement for compiler optimization (e.g. 6] and from e orts to estimate performance statically to automate load balancing (e.g. 7] Our techniques t in well with other methodologies for dealing with applications developed for particular architectures (e.g. [8]) Archetypes frequently represent well researched patterns or abstractions; for example, the mesh archetype [9] builds on Brinch Hansen s work on parallel cellular automata in the context of multicomputers [10] In that paper, the computational complexity of parallel cellular automata is derived ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: Practice and Experience, 5(5):407-423, 1993.

....also those applied by parallelizing compilers. Programming skeletons, design patterns, and distributed objects. Our work is also in some respects complementary to work exploring the use of programming skeletons and patterns in 4 parallel computing, for example that of Cole [27] and Brinch Hansen [15], and even work exploring distributed objects, pC [12] for example. We also make use of abstractions that capture exploitable commonalities among programs, but we use these abstractions to guide a program development methodology based on program transformations. Communication libraries. Much ....

....patterns of dataflow and communication. Schmidt [66] focuses more on parallel structure, but in a different context from our work and with less emphasis on code reuse. Shaw [67] examines higher level patterns in the context of software architectures. Brinch Hansen s work on parallel structures [15] is similar in motivation to our work, but his model programs are typically more narrowly defined than our archetypes. Other work addresses lower level patterns, as for example the use of templates to develop algorithms for linear algebra in [10] Program skeletons. Much work has also been done ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: Practice and Experience, 5(5):407--423, 1993.

....application are based on the archetypes being employed in the development of that application. Our techniques fit in well with other methodologies for dealing with applications developed for particular architectures (for example, Brinch Hansen s model for programming multicomputer applications [BH93a] Archetypes frequently represent well researched patterns or abstractions; for example, the mesh archetype [Mas96] builds on Brinch Hansen s work on parallel cellular automata in the context of multicomputers [BH93b] In that paper, the computational complexity of parallel cellular automata is ....

P. Brinch Hansen. Model Programs for Computational Science: A Programming Methodology for Multicomputers. Concurrency: Practice and Experience, 5(5):407--423, 1993.

.... skeletons for functional and other programs have been developed [7, 14, 18] Algorithm templates of the more common linear algebra programs have been developed and then used in designing programs for parallel machines [4] Parallel structures have been investigated by many other researchers [8, 22]. Structuring parallel programs by means of examining dataflow patterns has also been investigated [21] Our contribution is to show that combining consideration of broadlydefined computational patterns with dataflow considerations is useful in the systematic development of efficient parallel ....

....patterns of dataflow and communication. Schmidt [33] focuses more on parallel structure, but in a different context from our work and with less emphasis on code reuse. Shaw [34] examines higher level patterns in the context of software architectures. Brinch Hansen s work on parallel structures [8] is similar in motivation to our work, but his model programs are generally more narrowly defined than our archetypes. Other work addresses lower level patterns, as for example the use of templates to develop algorithms for linear algebra in [4] Program skeletons. Much work has also been done ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: Practice and Experience, 5(5):407--423, 1993.

....1 Introduction Considerable research activity has been devoted in recent years to the programming of parallel computers. In this respect, special attention has been paid to computational science, which currently represents the main field where parallel computers can be successfully employed [1, 5, 6]. Parallelization techniques concentrate on data parallelism, a characteristic of most computation intensive applications, and produce parallel programs executing in SPMD (Single Program Multiple Data) mode [3] Although SPMD code can be less efficient than a general parallel code, it can achieve ....

P. Brinch Hansen, "Model programs for computational science: A programming methodology for multicomputers", Concurrency: practice and experience, vol. 5, no. 5, pp. 407-423, 1993.

....SPMD (single program multiple data) model [6] allowing parallel code to be generated relatively simply by (semi) automatic tools. On the other hand, several authors have recently argued that parallel programs can be classified according to their algorithmic skeleton [4] or programming paradigm [3], i.e. to the way processes making up the parallel program are created, synchronize and communicate, abstracting from the details of their sequential code. According to this view, the SPMD model can be viewed as a family of paradigms, all characterized by common synchronization and communication ....

....4 that the parallel structure of most programs can be classified according to a limited number of patterns, that can be viewed as representative of the basic ways to organize a parallel computation. This idea has been explicitly expressed in the context of parallel programming methodologies [3, 4] and parallel programming languages [5] Even though the notion of parallel structure of a program has not been well formalized yet, it can be informally defined as the way processes forming the parallel program are created, synchronize and communicate, abstracting from the details of their ....

[Article contains additional citation context not shown here]

P. Brinch Hansen, "Model programs for computational science: a programming methodology for multicomputers", Concurrency: Practice and Experience, 5(5), pp. 407--423, Aug. 1993.

....and their role in developing both sequential and parallel programs. Gamma et al. 11] address primarily the issue of patterns of computation, in the context of object oriented design. Schmidt [21] focuses more on parallel structure, but with less emphasis on code reuse. Brinch Hansen s work [5] is similar in motivation to our work, but his model programs are typically more narrowly defined. Other work addresses lower level patterns, as for example the use of templates to develop algorithms for linear algebra in [3] and the use of templates in developing formally verified software in ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: Practice and Experience, 5(5):407-- 423, 1993.

....of this approach is not to define new mechanisms for reusability but to apply the concept at the high level of the development. This means that a template may correspond to a very complex system and not only to a simple function as in the usual library approach. For example, Brinch Hansen [3] has studied several basic templates (program models) for the development of parallel scientific applications. Other examples can be seen in [7] In [9] the author discusses the problem of reusable components not at the design or programming level but at the problem level. Again, the developer ....

P. Brinch Hansen. Model programs for computational science: A programming methoddology for multicomputers. Concurrency: Practice and Experience, 5(5):407--423, August 1993.

....to define networks of process groups and perform point to point communication between processes in different groups. Channel based group to group communication as described in x5 has not yet been implemented. 9 Related work A great deal of research has been done on software reuse and templates [3, 9]. A difference between our work and much of this previous work is that our archetype includes stepwise refinement design methods, and our focus is on design reuse as much as code reuse, particularly for parallel computing targeted to different architectures. Much of our research deals with ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: Practice and Experience, 5(5):407--423, 1993.

....An intermediate approach is the support of various skeletons for parallel programming paradigms. These are less flexible than low level primitives, but more flexible than complete sub programs. The skeleton based approach usually leads to better parallel programs, because it puts parallelism first [20]. Three main types of parallelism have been identified [21] Our skeletons support these forms in a restricted manner; a process farm is supported with the sandwich, algorithmic parallelism can be exploited using a synchronous process network skeleton, and geometric parallelism is supported with a ....

P. Brinch Hansen. Model programs for computational science: A programming methodology for multicomputers. Concurrency: practice and experience, 5(5):407--423, Aug 1993.

.... literature on parallel programming techniques, though in contexts other than automatic parallelization; the parallel structure of most programs can be classified according to a limited number of patterns, that can be viewed as representative of the basic ways to organize a parallel computation [1, 2]: this notion of parallel structure of a program can be informally defined as the way processes forming the parallel program are created, synchronize and communicate, abstracting from the details of their sequential part , and has been referred to with the term Parallel Programming Paradigm, or ....

P. Brinch Hansen, "Model programs for computational science: a programming methodology for multicomputers", Concurrency: Practice and Experience, 5(5), pp. 407--423, Aug. 1993.

....incorporates the idea of archetypes, gaining ease of use at the expense of greater generality. Our techniques fit in well with other methodologies for dealing with applications developed for particular architectures (for example, Brinch Hansen s model for programming multicomputer applications [BH93a] Archetypes frequently represent well researched patterns or abstractions; for example, the mesh archetype [Mas96] builds on Brinch Hansen s work on parallel cellular automata in the context of multicomputers [BH93b] In that paper, the computational complexity of parallel cellular automata is ....

P. Brinch Hansen. Model Programs for Computational Science: A Programming Methodology for Multicomputers. Concurrency: Practice and Experience, 5(5):407--423, 1993.

....interactively and visually designed with the Monads DPV tool set. This is the job of the application programmer because every application is different. But even here Monads DPV can ease the developer s job: Although the applications are basically all different, they often have similar structures [2]. It is especially valid for the strategies of parallelization like master worker and pipeline. Therefore, the tool can provide such algorithmic skeletons [6] or templates as they are called in Monads DPV. These templates have the advantage that the application developer can simply fill the ....

P. Brinch Hansen. Model programs for computational science: A programming methololody for multicomputers. Concurrency: Practice & Experience, 5(5):407--423, Aug. 1993.

No context found.

Per Brich Hansen. Model programs for computational science : a programming methodology for multicomputers. Concurrency: Practice and Experience, Vol. 5(5), p. 407-423 (August 1993).

No context found.

Per Brinch Hansen. Model programs for computational science: A programming methololody for multicomputers. Concurrency: Practice & Experience, 5(5):407--423, August 1993.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC