W.F. McColl. "Special purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....(R) 3.356) The bisection width is the size of the smallest edge cut of a graph which divides it into two equal parts. This graph invariant is a fundamental concept in the theory of interconnection networks of parallel computers. It influences both algorithmic design and lower bound techniques [9,12,13]. We study the problem in a more general framework. For a given graph G with vertices labeled by 1; 2; n its transposition graph T (G) is a The research of both authors was partially supported by Grant No. 95 5305 277 of Slovak Academy of Sciences Preprint submitted to Elsevier Preprint ....

McColl, W.F., Special purpose parallel computing. In Gibbons, A., Spirakis, P. (Eds.). Lectures on Parallel Computations. Cambridge University Press, Cambridge, 1993, pp. 261--311.

....Of course, an efficient algorithm for, say, a hypercube multicomputer will not necessarily perform well when run on, for example, a 2D mesh multicomputer with the same number of processors. This type of parallel computing is generally referred to as special purpose parallel computing [73]. Special purpose parallel systems are particularly appropriate in application areas where the goal of achieving the best possible performance is much more appropriate than that of achieving an architecture independent design. Examples of such areas include digital signal processing, image ....

....than that of achieving an architecture independent design. Examples of such areas include digital signal processing, image processing, computer vision, mobile robot control, particle simulation, dense matrix computations, cryptography, speech recognition, computer graphics, and game playing [73]. The following is a representative sample of special purpose parallel system in use CHAPTER 1. INTRODUCTION 10 today: 1. VLSI systems (custom VLSI chips, field programmable gate arrays) 2. Systolic architectures (application specific arrays, programmable systolic architectures) 3. Cellular ....

[Article contains additional citation context not shown here]

MCCOLL, W. F. Special Purpose Parallel Computing. In ALCOM Spring School on Parallel Computing, Cambridge International Series on Parallel Computation. Cambridge University Press. (1991).

....because the real cost of accessing memory for an algorithm depends on the total number of accesses and the pattern in which they occur. One attempt to provide some abstraction from the PRAM is the language FORK [137] A good overview of models aimed at particular architectures can be found in [145]. 5 Summary We have presented an overview of parallel programming models and languages, using a set of six criteria that an ideal model should satisfy. Four of the criteria relate to the need to be able to use the model as a target for software development. They are: ease of programming, the ....

W.F. McColl. Special purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pages 261--336. Cambridge University Press, Cambridge, 1993.

....has been made to limit those disadvantages, for example on the semi automatic derivation of systolic algorithms [18] But the application domain for architecture specific algorithms remains limited, mostly to regular and full (not sparse) numerical computations. These observations have led McColl [15] to use the phrase special purpose parallel programming (SPPP) for this style of programming. Another possibility is to write programs which are machine independent and leave scope for parallelism, but do not specify individual communications, synchronisations or other low level events. McColl ....

....[15] to use the phrase special purpose parallel programming (SPPP) for this style of programming. Another possibility is to write programs which are machine independent and leave scope for parallelism, but do not specify individual communications, synchronisations or other low level events. McColl [15] has called this general purpose parallel programming (GPPP) and promotes the idea that it is the reasonable choice for long term research and software development. GPPP allows parallel programs to become portable across all architectures, thus justifying the initial conversion work for ....

W. F. McColl, Special purpose parallel computing, in Lectures on Parallel Computation, Gibbons and Spirakis, eds., International Series on Parallel Computation, Cambridge University Press, March 1991.

....in the theory of interconnection networks of parallel computers. It influences both algorithmic design and lower bound techniques 1 The research of both authors was partially supported by Grant No. 95 5305 277 of Slovak Academy of Sciences Preprint submitted to Elsevier Preprint 26 February [10,13,14]. We study the problem in a more general framework. For a given graph G with vertices labeled by 1; 2; n its transposition graph T (G) is a Cayley graph on a permutation group with generators transpositions f(ij)g whenever ij is an edge in G (see. 9] Transposition graphs, when G is a ....

McColl, W.F., Special purpose parallel computing. In Gibbons, A., Spirakis, P. (Eds.). Lectures on Parallel Computations. Cambridge University Press, Cambridge, 1993, pp. 261--311.

....areas, their use in commercial applications, which are often non numerical in nature, has been scarce. One of the impediments in the long term commercial uptake of parallel computing has been the proliferation of differing machine architectures and corresponding programming models (e.g. see [92] for a survey) However, due to several technological and economic reasons, the various classes of parallel computers such as shared memory machines, distributed memory machines, and networks of workstations, are beginning to acquire a familiar appearance: a workstation like processor memory pair ....

....of processors, p. A number of universality results have been obtained in the parallel setting, based on the two phase randomised routing results discussed in the last section, and the BSP model embodies these results, by making no assumptions about the technology or the degree of parallelism [92, 96, 133, 134]. Several results obtained with PRAM programming, 136] quantitatively argues that, when incorporated in a practical setting, can form the basis of parallel programming languages and machine designs. Further, we wish to incorporate the multiple levels of inherent parallelism present in the ....

W. F. McColl. Special Purpose Parallel Computing. In A. M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 1993.

.... of universality results have been obtained in the parallel setting, based on the two phase randomised routing results discussed in the last section, and the BSP model embodies these results, by making no assumptions about the interconnection topology or the degree of parallelism [Val88, Val89, McC93b, McC95] Valiant[Val90b] quantitatively argues that several results obtained with PRAM programming, when incorporated in a practical setting, can form the basis of parallel programming languages and machine designs. Further, we wish to incorporate the multiple levels of inherent parallelism ....

W. F. McColl. Special Purpose Parallel Computing. In A. M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 1993.

....to execute one MERGE whereas in [9] this operation is called twice. However, we could not attain the theoretical O(h log log N) bound claimed in [9] for extract min. 4 Parallel priority queue on a BSP computer Recently a BSP implementation of the PPQ introduced in [8] was proposed in [2] see [14, 6] for details on BSP computers which can be seen as distributed memory systems in our discussion) However, this BSP PPQ uses a novel mapping between nodes and processors, which reduces the amount of inter processor communications needed to maintain the extended heap invariant. In addition [2] ....

W.F. McColl. "Special purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 1993.

No context found.

W.F. McColl. "Special purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 1993.

....acyclic graphs, where each node corresponds to some simple operation, and the arcs correspond to inputs and outputs. Let Cn denote the directed acyclic graph which has n 3 nodes v i;j;k , 0 i; j; k n, and arcs from v i;j;k to v i 1;j;k , v i;j 1;k and v i;j;k 1 where those nodes exist. In [14] it is shown that the LU decomposition of an n Theta n non singular matrix A can be computed (without pivoting) using the following set of definitions: For all 0 k i; j n, u k;k;k = a k;k;k Gamma1 l i;j;k = a i;j;k Gamma1 =u i;j;k if j = k; and l i;j Gamma1;k otherwise: u i;j;k = a ....

W F McColl. Special purpose parallel computing. In Gibbons and Spirakis [10], pages 261--336.

.... parallelism, that is parallelism across all application areas and a wide range of architectures [25] Its goals are more ambitious than most parallel programming systems which are aimed at particular kinds of applications, or work well only on particular classes of parallel architectures [26]. BSP s most fundamental properties are that: ffl It is simple to write. BSP programs look much the same as sequential programs. Only a bare minimum of extra information needs to be supplied to describe the use of parallelism. ffl It is independent of target architectures. Unlike many parallel ....

....program then becomes a matter of choosing among known algorithms for those that are optimal for the range of machine sizes envisaged for the application. For example two BSP algorithms for matrix multiplication have been developed. The first, a block parallelization of the standard n 3 algorithm [26], has (asymptotic) BSP complexity Block MM cost = n 3 =p (n 2 =p 1=2 )g p 1=2 l requiring memory at each processor of size n 2 =p. This is optimal in time and memory requirement. A more sophisticated algorithm due to McColl and Valiant [23] has BSP complexity Block and Broadcast MM ....

W.F. McColl. Special purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pages 261--336. Cambridge University Press, Cambridge, 1993.

....course, an efficient algorithm for, say, a hypercube multicomputer will not necessarily perform well when run on, for example, a 2D array multicomputer with the same number of processors. We will use the generic term special purpose to refer to this type of parallel computing. In a related paper [187], we describe a number of aspects of the work which has been done in recent years on the design, analysis, implementation and verification of special purpose parallel computing systems. The volume of published material on these topics is huge. A long, but by no means complete, bibliography is ....

....number of aspects of the work which has been done in recent years on the design, analysis, implementation and verification of special purpose parallel computing systems. The volume of published material on these topics is huge. A long, but by no means complete, bibliography is given at the end of [187]. Special purpose parallel systems are particularly appropriate in application areas where the goal of achieving the best possible performance is much more important than that of achieving an architectureindependent design. Some examples of such areas are: digital signal processing (filtering, ....

[Article contains additional citation context not shown here]

W F McColl. Special purpose parallel computing. In Gibbons and Spirakis [89].

.... parallelism, that is parallelism across all application areas and a wide range of architectures [22] Its goals are more ambitious than most parallel programming systems which are aimed at particular kinds of applications, or work well only on particular classes of parallel architectures [23]. BSP s most fundamental properties are that: ffl It is simple to write. BSP programs are much the same as sequential programs. Only a bare minimum of extra information needs to be supplied to describe the use of parallelism. ffl It is independent of target architectures. Unlike many parallel ....

....program then becomes a matter of choosing among known algorithms for those that are optimal for the range of machine sizes envisaged for the application. For example two BSP algorithms for matrix multiplication have been developed. The first, a block parallelization of the standard n 3 algorithm [23], has BSP complexity Block MM cost = n 3 =p (n 2 =p 1=2 )g p 1=2 l requiring memory at each processor of size n 2 =p. This is optimal in time and memory requirement. A more sophisticated algorithm due to McColl and Valiant [20] has BSP complexity Block and Broadcast MM cost = n 3 ....

W.F. McColl. Special purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pages 261--336. Cambridge University Press, Cambridge, 1993.

....of which physical memory module the value is held in. The algorithm designer and the programmer should not be aware of any hierarchical memory organisation based on network locality in the particular physical interconnect structure that is currently used, as in special purpose parallel computing [21]. Instead, performance of the communications network is described only in terms of its global properties, using the parameters l and g. The complexity of a superstep S in a BSP algorithm is determined as follows. Let the work w be the maximum number of local computation steps executed by any ....

W. F. McColl, "Special purpose parallel computing," in Lectures on Parallel Computation, A. M. Gibbons and P. Spirakis, Ed. Cambridge, UK: Cambridge University Press, 1993, pp. 261--336.

No context found.

W.F. McColl. "Special purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 261--336. Cambridge University Press, 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