N -Graphs: A Topology for Parallel Divide-and-Conquer on Transputer Networks
نویسندگان
چکیده
A parallel implementation of a divide-and-conquer template (skeleton) is derived systematically from its functional speciication. The implementation makes use of a new processor topology for divide-and-conquer, called N-graph, which suits transputer networks well: there are not more than 4 links per processor, overlapping of computations and communication within a processor is exploited, the processor network is of an arbitrary xed size, the load is balanced and all communications between processors are local. A parallel mergesort algorithm is used to illustrate the derivation process and the distributed parallel target program. Experimental results on a 64-node transputer network are reported.
منابع مشابه
Formal Derivation and Implementation of Divide-and-Conquer on a Transputer Network
This paper considers parallel program development based on functional mutually recursive speciications. The development yields a communication structure linking an arbitrary xed number of processors and an SPMD program executable on the structure. There are two steps in the development process: rst, a parallel functional implementation is obtained through formal transformations in the Bird-Meer...
متن کاملOptimality of HLF for scheduling divide-and-conquer UET task graphs on identical parallel processors
The problem of scheduling a set of n unit execution time (UET) tasks subject to precedence constraints on m identical parallel processors is known to be N P -hard in the strong sense. However, polynomial time algorithms exist for some classes of precedence graphs. In this paper, we consider a class of divide-and-conquer graphs that naturally models the execution of the recursive control abstrac...
متن کاملNode-Degree Requirements for Time-Optimal Execution of a Class of Parallel Algorithms
In this paper, we consider the time-optimal execution of a class of parallel algorithms based on the principle of divide-and-conquer. We use a computational model of this class of algorithms which incorporates both communication and the associated computation overheads that are unavoidable in a divide-and-conquer methodology. We consider the time-optimal implementation of this class of parallel...
متن کاملGeneralizing the Implementation of an Optimal Parallel Recursive Algorithm for Triangular Matrix Inversion
This paper describes a generalization of a study on an implementation of a parallel divide and conquer algorithm for triangular matrix inversion [3]. Indeed, given an original (lower) triangular matrix of size n=m2 (m, k ≥ 1) and an available number of processors p power of 2 (<n), we designed a strong cost optimal parallel algorithm i.e. whose efficiency (resp. speedup) is equal to 1 (resp. p)...
متن کاملFree Vibration Analysis of Repetitive Structures using Decomposition, and Divide-Conquer Methods
This paper consists of three sections. In the first section an efficient method is used for decomposition of the canonical matrices associated with repetitive structures. to this end, cylindrical coordinate system, as well as a special numbering scheme were employed. In the second section, divide and conquer method have been used for eigensolution of these structures, where the matrices are in ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2007