W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....up and distribute them onto the processors in such a way that it becomes feasible to exploit locality by e ecting parallel processing of user requests, each upon a subset of the data. As opposed to shared memory models, 1 this distributed memory model provides the bene t of better scalability [6]. However, this introduces new problems related to the communication and synchronization of processors and their load balance. This paper describes strategies to overcome these problems in the context of the parallelization of sux arrays [3] We propose strategies for reduction of inter processors ....

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337-391. Cambridge University Press, 1993.

....data locality. The model, however, still suffers from inefficiency when implemented on mainstream architectures which rely on software synchronisation. 1.1. 4 The Bulk Synchronous Parallel Model McColl and Valiant propose the Bulk Synchronous Parallel (BSP) model of parallel computation[Val90, McC92, McC93a] to provide a simple, unified framework for the design and programming of all kinds of general purpose parallel systems [McC94] The model parameterises the architectures using four parameters, namely p, s, l and g. p is the number of processors, s is a measure of their speed, l is a ....

....skeleton One suitable cost model would be the CLUMPS computational cost model[Cam94] The CLUMPS cost model encapsulates the performance of the underlying machine and is particularly suited to architectures which can be recursively partitioned. Alternatively, the BSP cost model could be adopted[McC92] The BSP cost model does not permit the use of localised communication synchronisation costs, nor does it permit the partitioning of the machine, but only the BSP language prevents sub group synchronisation. The BSP cost model could therefore be 152 adopted using the sub group synchronisation ....

W F McColl. General purpose parallel computing. In Gibbons and Spirakis [GS92]. 226

....of the BSP cost model. 1. Introduction One of the major aims currently focusing the efforts of the parallel computing community is the development of a realistic base for the design and programming of general purpose parallel computers. The advent of the bulk synchronous parallel (BSP) model [10, 8] has offered a solution to this problem, providing an underlying framework for the devising of both scalable parallel architectures and portable parallel software. However, when proposing BSP programming as the standard approach to portable parallel code design, one must provide effective methods ....

....scheduling of uniform dependence loops of type I and II, respectively, are devised and analysed in terms of the BSP cost model. A final section including a short summary and further work directions concludes the paper. 2. The BSP programming and cost model A bulk synchronous parallel computer [10, 8] consists of a set of processor memory pairs, a communication network providing uniformly efficient non local memory access, and a mechanism for efficient barrier synchronisation of all processors or of a subset of processors. No specialised broadcasting or combining facilities are available. The ....

W. F. McColl. General purpose parallel computing. In A. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge Univ. Press, 1993.

....parallel computation the way the von Neumann model has served sequential computing for the last fifty years. The introduction of the Bulk Synchronous Parallel (BSP) model of computation by L.G. Valiant addresses these limitations. A BSP computer as introduced in [36, 37, 38] and also examined in [10, 22, 23, 24, 39] consists of three parts: i) a collection of processor memory components, ii) a communication network that can deliver messages point to point among the components, and (iii) facilities for global synchronization, in barrier style, of all or subsets of the components. A BSP computer is ....

W. F. McColl. General purpose parallel computing. In Lectures on parallel computation, (A. Gibbons and P. Spyrakis, eds.), Cambridge University Press, 1993.

....achieved in sequential computing can hardly catch up with people s rapidly increasing desires for computing powers. A straightforward approach to solve this problem is through parallelism. The pursuit of higher performance from computers has aroused the introduction of parallel computing [74] in many areas, particularly in scienti c and engineering applications. One of the major concerns in the parallel computing community is the development of a realistic foundation for the design and programming in parallel computers as the Von Neumann model does in sequential computing. The advent ....

W. F. McColl. `General Purpose Parallel Computing'. Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pp. 337-391, Cambridge University Press, Cambridge, 1993.

....many of the previous limitations, BSP is widely regarded as a bridging model for parallel computing [18,17] In the BSP model no assumptions are made about the underlying technology or the degree of parallelism. The BSP model thus aims to provide a general purpose parallel computing platform [12,13,17,18]. A Bulk Synchronous Parallel Machine (BSPM) provides an abstraction of any real parallel machine; a BSPM has three constituent parts: 1) A number of processor memory components (processors) 2) An interconnection network which delivers messages in a point to point manner between the ....

W. F. McColl, General purpose parallel computing, in: A. M. Gibbons and P. Spirakis, eds., Lectures on Parallel Computation, Cambridge International Series on Parallel Computation 337-391 (Cambridge University Press, 1993).

....to expose the maximum possible computational parallelism in a given task. Thus, the PRAM provides a measure of the ideal parallel time complexity. Many modifications to the PRAM have been proposed that attempt to bring it closer to practical parallel computers. Goodrich [Goo93] and McColl [McC93] survey the PRAM model and its extensions. A brief overview of machine characteristics that have been the focus of efforts to improve the PRAM is given below. 1. Memory Access. The LPRAM (Local memory PRAM) ACS90] augments the CREW PRAM by associating with each processor an unlimited amount of ....

....With these applications, we evaluate the utility of BSP in terms of portability, efficiency, and predictability on an SGI Challenge and an Intel Paragon. The claim that both efficiency and portability can be achieved by using the BSP model is supported by both theoretical and experimental results [Val90a, GLR99, KS99, McC93, Val90b, Val93, WG98] . However, other general purpose models, such as LogP [CKP96] make similar claims. LogP models the performance of point to point messages with three parameters representing software overhead, network latency, and communication bandwidth. Under LogP, the programmer is not constrained by a ....

W. F. McColl. "General Purpose Parallel Computing." In A. M. Gibbons and P. Spirakis, editors, Lectures in Parallel Computation, Proceedings

....and (ii) fully asynchronous systems (synchronization is realized pairwise whenever communication is required) Processor components in the BSP model advance jointly through the program, with the required remote communication occurring between supersteps. The BSP computer as described in [18, 39, 40, 41, 50, 51, 52] consists of the following three components. 1 (1) A collection of p processor memory components numbered 0, p 1. It is assumed that each processing unit consists of a sequential processor with a block of local memory. 2) A communication network that can deliver messages point to ....

W. F. McColl. General purpose parallel computing. In Lectures on parallel computation, (A. Gibbons and P. Spirakis, eds.), Cambridge University Press, 1993.

....intended purpose was to run such applications. The performance of a particular parallel program depends on both its computational, synchronization and communication requirements. 2 Sorting on the BSP model: An overview A significant amount of work has been devoted in the study of the BSP model [61, 62, 63, 50, 51, 52, 44, 7] and the analysis [7, 10, 11, 19, 20, 34] and design of BSP algorithms [8, 9, 21, 22, 23, 24, 25, 50, 51, 52] The problem of parallel sorting n keys [48] has drawn considerable attention. First approaches were sorting networks introduced by Batcher [5] Since then, a wealth of literature ....

....on both its computational, synchronization and communication requirements. 2 Sorting on the BSP model: An overview A significant amount of work has been devoted in the study of the BSP model [61, 62, 63, 50, 51, 52, 44, 7] and the analysis [7, 10, 11, 19, 20, 34] and design of BSP algorithms [8, 9, 21, 22, 23, 24, 25, 50, 51, 52]. The problem of parallel sorting n keys [48] has drawn considerable attention. First approaches were sorting networks introduced by Batcher [5] Since then, a wealth of literature concerning parallel sorting has been published [47, 48] The first sorting network to achieve, however, the optimal ....

W. F. McColl. General purpose parallel computing. In Lectures on parallel computation, (A. Gibbons and P. Spirakis, eds.), Cambridge University Press, 1993.

....measure of how many hops a message may have to endure. Degree the maximal number of edges incident to a node. The lower the degree the easier it is to physically realize the design in hardware. There are many different proposals for topologies for parallel architectures, among them are (from [McC91] Topology Degree Diameter 1D array (ring) 2 p=2 Shuffle exchange 3 2 log p Cube connected cycles 3 (5=2) log p 2D mesh of trees 3 2 log p 3D mesh of trees 3 2 log p 2D array (toroidal) 4 p p Butterfly (wrapped) 4 2 log p de Bruin 4 log p 3D array (toroidal) 6 3 p p Pyramid ....

William F. McColl. General purpose parallel computing. In A. M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Spring School on Parallel Computation, pages 337--391. ALCOM, Cambridge University Press, 1991.

.... distributed memory MIMD systems based on commodity processor boards and off the shelf interconnects, called clusters [1] This picture may change with the widespread deployment of optical computing devices or other new technology, but at present almost all deployed systems are of these two kinds [24]. The second requirement for effective parallel computing is an abstract machine or parallel programming model, forming the bridge between the hardware and the level at which software is written. All computers consume instructions much faster than they can be written. An abstract machine is only ....

W. F. McColl. General purpose parallel computing. In A. M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pages 337--391. Cambridge University Press, 1993. 32

....approaches, the new parallelization strategy described in the current paper succeeds in automatically generating parallel code which is scalable, portable, and whose performance is analytically predictable. This is achieved by employing the bulk synchronous parallel (BSP) model of computation [11, 9] as a target platform for the parallelization and scheduling of generic loop nests. The BSP model of computation provides a unifying framework for the development of both scalable parallel architectures and portable parallel software. A BSP computer comprises a set of processor memory pairs, a ....

.... w gh; 1) where w denotes the maximum number of local operations executed by any processor during superstep S, and h is the maximum number of words received or sent by any processor during superstep S. Following the successful design and implementation of many BSP algorithms and applications [9, 11], recent research work studied the scheduling of perfect loop nests on BSP computers [5, 7, 10] The current paper extends this work by introducing a scheme for the BSP parallelization of generic, untightly nested loops whose loop bounds and referenced array subscripts are affine functions of the ....

W.F. McColl. General purpose parallel computing. In A. Gibbons and P. Spirakis, eds., Lectures on Parallel Computation, pages 337--391. Cambridge Univ. Press, 1993.

....an unrestricted communication model. There will probably never be a model that satisfies all potential users of parallelism. However, models that satisfy many of the requirements above are good candidates for general purpose parallelism, the application of parallelism across wide problem domains [144]. 4 Overview of Models We now turn to assessing existing models according to the criteria outlined in the previous section. Most of these models were not developed with the ambitious goal of general purpose parallelism, so it is not a criticism to say that some of them fail to meet al..l of the ....

....of these pieces on processors and the way in which they communicate does not have to be described so explicitly. 4.3.1 Static. The only examples in this class are those that renounce locality, which ensures that placement does not matter to performance. Bulk synchronous parallelism (BSP) [143, 144, 146, 196 198] is a model in which interconnection network properties are captured by a few architectural parameters. A BSP abstract machine consists of a collection of p abstract processors, each with local memory, connected by an interconnection network whose only properties of interest are the time to do a ....

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, Cambridge International Series on Parallel Computation, pages 337--391. Cambridge University Press, Cambridge, 1993. 50

....to remain general, or to address the problem for a specific parallel architecture if they aimed at being practical. Clearly, both options have their disadvantages. This paper avoids limiting to one of the two directions by proposing the bulk synchronous parallel (BSP) programming and cost model [11, 8] as a target platform for the derivation of scalable dag schedules for general purpose parallel computers. The paper is organised as follows. In Sect. 2, the BSP model is briefly described. Then, in Sect. 3, a new technique for the BSP scheduling of normalised uniform dags, i.e. of dags that ....

....in Sect. 4. A final section including a short summary and further work directions concludes the paper. 2 The Bulk Synchronous Parallel Model The existence of a standard model is the only way to fully impose parallel computing as a viable alternative to sequential computing. The BSP model [11, 8] provides such a unifying framework for the design and programming of general purpose parallel computers. A BSP computer consists of a set of processormemory pairs, a communication network for point to point message delivery, and a mechanism for efficient barrier synchronisation of all processors ....

[Article contains additional citation context not shown here]

W.F. McColl, General purpose parallel computing. In: A.M. Gibbons, P. Spirakis (eds.), Lectures on Parallel Computation, Cambridge Univ. Press, 1993, 337-391.

No context found.

W.F. McColl. "General purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W F McColl. General purpose parallel computing. In A M Gibbons and P Spirakis, editors, Lectures on Parallel Computation. Proc. 1991 ALCOM Spring School on Parallel Computation, pages 337--391. Cambridge University Press, 93.

No context found.

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. "General purpose parallel computing ". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. "General purpose parallel computing". In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337--391. Cambridge University Press, 1993.

No context found.

W.F. McColl. General purpose parallel computing. In A.M. Gibbons and P. Spirakis, editors, Lectures on Parallel Computation, pages 337-391. Cambridge University Press, 1993.

No context found.

W. F. McColl. General purpose parallel computing. In A. Gibbons and P. Spirakis, editors, Lectures on parallel computation, volume 4 of Cambridge International Series on Parallel Computation, chapter 14, pages 337--391. Cambridge University Press, 1993.

No context found.

W. F. McColl. General purpose parallel computing. In A. Gibbons and P. Spirakis, editors, Lectures on parallel computation, volume 4 of Cambridge International Series on Parallel Computation, chapter 14, pages 337--391. Cambridge University Press, 1993.

First 50 documents  Next 50

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