H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11:701-752, 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....prototypes [17, 11, 18] Our source language, SML, is a mature functional language with a stable formal de nition [12] For prototyping, SML provides closeness to formalisms for proof and transformation. For parallel prototyping, SML s strictness is better suited than the laziness of Haskell [9] as it results in more predictable behaviour. We focus on common low level HOFs such as map and fold which are ubiquitous in functional programs. Explicit HOF names are used as the basis for identi cation: fun map f [ map f (h: t) f h: map f t fun fold (f: a a a) b [ b fold ....

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11:701-752, 1999.

.... parallelized in a manager worker style [11] We have performed this work in preparation of a broader activity where we will compare parallel algorithms implemented in Distributed Maple with corresponding declarative versions written in para functional language Glasgow Parallel Haskell GPH [3] which has been previously used for the parallelization of computer algebra algorithms [4] We are going to use GPH as a coordination environment for scheduling computations among Maple kernels; for this purpose, we have already developed a GPH Maple interface [10] 2 Distributed Maple ....

H.W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11(12):701-751, 1999.

....a considerable agreement between the various architectures in terms of both runtimes and speedups. 4. 5 Linear equation solver This algorithm finds an exact solution of a linear system of equations of the form Ax = b, where all values are in the integer domain, and is discussed in detail by Loidl [7]. The original algorithm was coded in Haskell and has been directly translated into SML. The method is to map the problem onto multiple homomorphic images, i.e. modulo a prime number, and solve in the homomorphic domain. The final solution is generated by applying the Chinese Remainder Algorithm ....

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency --- Practice and Experience, 11:701--752, 1999.

.... a manager worker based style [8] We have performed this work in preparation of a broader activity where we will compare parallel algorithms implemented in Distributed Maple with corresponding versions written for an environment that uses the para functional language Glasgow Parallel Haskell GPH [1] to schedule computations among Maple kernels. For this purpose, we have already developed a Haskell Maple interface [7] a corresponding GPH Maple environment is under preparation. The remainder of this paper is organized as follows: in Section 2, we state the problem and describe the sequential ....

H.W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11(12):701-751, 1999.

No context found.

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11:701-752, 1999.

No context found.

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11:701-752, 1999.

No context found.

H.-W. Loidl, P. Trinder, K. Hammond, S. Junaidu, R. Morgan, and S. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency -- Practice and Experience, 11:701--752, 1999. Available from: <URL:http://www.cee.hw.ac.uk/dsg/gph/>.

....of a list in parallel to the degree specified by its argument, in this case, to normal form using the rnf strategy. parList and rnf have a straightforward implementation using par and seq. For a discussion on how to implement efficient parallel programs using GPH, the reader should refer to [12]. 2.2. Simple Parallel Search Entrance Exit Figure 1. A maze as a tree We represent a maze as a tree with exactly one path from the entrance (root) to the exit (one leaf) Figure 1. The following algorithm is a sequential solution to this search problem: search : a Bool) a Bool) ....

H.-W. Loidl, P. Trinder, K. Hammond, S. Junaidu, R. Morgan, and S. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency -- Practice and Experience, 11:701--752, 1999.

No context found.

H.-W. Loidl, and P.W. Trinder, K. Hammond, S.B. Junaidu R.G. Morgan and S.L. Peyton Jones, "Engineering Parallel Symbolic Programs in GPH", to appear in Concurrency Practice and Experience, 2000.

....description of parallelism. This eases the development of an initial parallel version of a program. But how easy is it to tune the parallel performance of such programs In previous work we have parallelized and tuned the performance of several large programs in Glasgow parallel Haskell (GpH) [1]. In most cases we managed to improve parallel performance by adding evaluation strategies. However, in a few cases simple code restructuring was necessary in order to increase task granularity or improve data locality. This repeated ad hoc program transformation motivated our search for a ....

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11(12):701-752, October 1999.

No context found.

H.-W. Loidl, P.W. Trinder, C.V. Hall, K. Hammond, S.B. Junaidu, and S.L. Peyton Jones, "Engineering Parallel Symbolic Programs in GpH", submitted to Concurrency: Practice and Experience, September 1998.

....elements of a list in parallel to the degree speci ed by its argument, in this case, to normal form using the rnf strategy. parList and rnf have a straightforward implementation using par and seq. For a discussion on how to implement ecient parallel programs using GpH, the reader should refer to [14]. 2.2 An Example Maze Search Consider a maze de ned as an area with an entrance and an exit and internally it contains walls and obstacles. There is exactly one path that leads from the entrance to the exit. A maze can be represented in many ways, and we chose a tree structure as illustrated in ....

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency { Practice and Experience, 11:701-752, 1999.

No context found.

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency -- Practice and Experience, 11:701--752, 1999.

No context found.

H.-W. Loidl, and P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan and S.L. Peyton Jones, \Engineering Parallel Symbolic Programs in GPH", Concurrency Practice and Experience, 11(12), October 1999, pp. 701-752.

....model (Sect. 2) Performance is measured for a simple test program with good parallel behaviour (Sect. 3) as well as for one larger application with irregular parallelism and complex data structures (Sect. 4) This complements our earlier research on parallelising substantial Haskell applications [1] and developing a suite of simulation and pro ling tools. 2 The GUM Runtime System GUM is the runtime system for GpH [7] a parallel variant of the Haskell lazy functional language. Being a parallel graph reduction machine [3] GUM represents an architecture independent abstract machine model ....

....to determine whether two accident reports are for the same location, and each criteria partitions the set. The problem amounts to combining several partitions of a set into a single partition, or union nd. The program comprises 1,500 lines of Haskell code and additional details are available in [1]. The parallel (GpH) version of the algorithm uses a geometric partitioning of the input data into 32 small and 8 large tiles. Evaluation strategies [6] are used to de ne the parallel evaluation over these tiles. The best sequential eciency (see Table 1) is obtained for machines based on SPARC ....

H-W. Loidl, P.W. Trinder, K. Hammond, S.B. Junaidu, R.G. Morgan, and S.L. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11(12):701-752, Oct. 1999. Available from [8].

No context found.

H. Loidl, P. Trinder, K. Hammond, S. Junaidu, R. Morgan, and S. Peyton Jones. Engineering Parallel Symbolic Programs in GPH. Concurrency | Practice and Experience, 11(12):pages 701-752, October 1999.

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