D. B. Skillicorn. Multiprogramming BSP programs. Department of Computing and Information Science, Queen's University, Kingston, Canada, October 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....are of limited use but when used in conjunction with sampling or other scaling techniques they are indeed very useful and powerful [13, 10] 4. 2 Parallel data mining Although parallel computing has the potential to speed up data mining, it s role in data mining is still an open question [8, 23]. The well known application of parallel data mining is in the construction of decision trees. After partitioning the dataset into disjoint subsets based on some attribute, the decision tree algorithm can be applied parallelely on each individual subset. 4.3 Distributed data mining In essence, ....

D.B. Skillicorn. Strategy for parallel data mining. Technical Report 1999-426, Queens's University, Department of Computing and Information Science, 1999.

....constructors (skeletons) The major drawbacks of BMF are that it is solely data parallel and the functions cannot be nested. Though Skillicorn does now appear to acknowledge the possibility of the nesting of functions in BMF by having the compiler flatten the nested functions into correct BMF [Ski94] However, despite it having been recognised that data parallelism is more applicable than otherwise previously thought [HS86] it is still lacking in expressibility, unless it is loosened to become multi threaded with predominantly breadth first evaluation, as is the case with F code [MSS93] ....

D. B. Skillicorn. Department of Computing and Information Science, Queen's University, Kingston, 1994. (Personal communication).

....data being supplied as a (continuous) stream rather than a single collection of data being applied to the process graph together. 4 Other Skeleton Classification Schemes 4. 1 BMF The Bird Meertens Formalism (BMF) Bir89] was originally conceived as a derivational style for functional programs [Ski92, Ski90] BMF consists of a set of theories built on a base algebra with unary and binary functions, each theory captures the behaviour of a particular class of data structures. A BMF theory begins with base types and extends them to new types using type functors, adding to the base algebra a ....

D. B. Skillicorn. Parallelism and the Bird-Meertens Formalism. Department of Computing and Information Science, Queen's University, Kingston, April 1992.

....transformations which can be applied for optimisation, or regarded as rewrite rules, while the completeness result is intended to guarantee that the language is sufficiently expressive. So BMF extends the concept of Abstract Data Types (ADTs) to that of categorical data types 1 [Ski91] Unlike ADTs, categorical data types have operations, equations relating them, and a guarantee that all of the required operations and equations are present. A theory is built from the constructors of a type using a categorical construction. However, it was discovered that BMF 2 also has an ....

D. B. Skillicorn. Practical parallel computation. Technical Report ISSN-08360227 -91-312, Department of Computing and Information Science, Queen's University, Kingston, Ontario, August 1991.

....networks. However, LogP is a slight generalisation of BSP by being asynchronous instead of barrier synchronous [NDD93] so providing more flexible program control. 2. 4 BMF The Bird Meertens Formalism (BMF) Bir89] was originally conceived as a derivational style for functional programs [Ski92b, Ski90a] BMF consists of a set of theories built on a base algebra with unary and binary functions, each theory captures the behaviour of a particular class of data structures. A BMF theory begins with base types and extends them to new types using type functors, adding to the base algebra a set ....

....basic parallel architecture classes, i.e. on the architectures in Skillicorn s classification, for it to be accepted. Skillicorn also presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91b] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91b] So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Parallelism and the Bird-Meertens Formalism. Department of Computing and Information Science, Queen's University, Kingston, April 1992.

....transformations which can be applied for optimisation, or regarded as rewrite rules, while the completeness result is intended to guarantee that the language is sufficiently expressive. So, BMF extends the concept of Abstract Data Types (ADTs) to that of catagorical data types 4 [Ski91b] Unlike ADTs, categorical data types have operations, equations relating them, and a guarantee that all of the required operations and equations are present. A theory is built from the constructors of a type using a categorical construction. 4 Categorical data types come equipped with a set of ....

....nodes of independent processor memory pairs connected by some interconnection network (c.f. Section 1.2) it is by no means certain that there is a definite progression towards a single class of parallel architecture. Skillicorn has identified four classes of parallel architecture [Ski90a, Ski90b, Ski91b] which he sees as being the basic, distinct classes: ffl SIMD 1 ffl tc 2 MIMD 3 1 Single Instruction stream, Multiple Data streams. 2 Tightly coupled, or physical bank of shared memory. 3 Multiple Instruction streams, Multiple Data streams. Chapter 3. Work Towards CLUMPS 35 ffl ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Practical parallel computation. Technical Report ISSN0836 -0227-91-312, Department of Computing and Information Science, Queen's University, Kingston, Ontario, August 1991.

....parallel computation would need to be implementable on all of these basic parallel architecture classes, for it to be accepted. Skillicorn also presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91]. A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91] So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without ....

.... also presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91]. So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without modification. It only requires recompilation. This provides for the model to be portable. For a model to be congruent it must reflect the real ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Parallelism and the Bird-Meertens Formalism. Department of Computing and Information Science, Queen's University, Kingston, April 1992.

....of [Cam94a] revising certain aspects of the unification process, providing further justification for the results, and providing more detailed illustration of previously mentioned points. There are four basic and distinct classes of parallel architecture which have been identified by Skillicorn [Ski90a, Ski90b, Ski91]: ffl SIMD 1 ffl tc 2 MIMD 3 ffl hypercuboid lc 4 MIMD ffl sparse lc MIMD where a hypercuboid connected architecture has at least logarithmic connectivity in the number of processors, while a sparsely connected architecture has constant 1 Single Instruction stream, Multiple Data streams ....

....parallel computation would need to be implementable on all of these basic parallel architecture classes, for it to be accepted. Skillicorn also presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91]. A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91] So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Practical parallel computation. Technical Report ISSN0836 -0227-91-312, Department of Computing and Information Science, Queen's University, Kingston, Ontario, August 1991.

....it. The SIMD class is traditionally programmed in a data parallel style, MIMD in a control parallel style, tc using shared variables, lc using message passing, etc. Two of his requirements for a model of parallel computation were also adopted; these are: architecture independence and congruence [Ski92a, Ski92b, Ski91]. A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91] So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without ....

....lc using message passing, etc. Two of his requirements for a model of parallel computation were also adopted; these are: architecture independence and congruence [Ski92a, Ski92b, Ski91] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91]. So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without modification. For a model to be congruent it must reflect the real costs of execution at the model level [Ski92b] That is, for a model to be ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Parallelism and the Bird-Meertens Formalism. Department of Computing and Information Science, Queen's University, Kingston, April 1992.

....computation (control methods and memory access) of all other parallel architecture classes, communication can not be so simulated efficiently due to certain architecture classes being non scalable in terms of bandwidth. 1 Introduction David Skillicorn s four classes of parallel architecture (see [Ski90a, Ski90b, Ski91]) were taken as the basic parallel architecture classes for the unification process. These architectures are: SIMD 1 , tc 2 MIMD 3 , hypercuboid 4 lc 5 MIMD and sparse 6 lc MIMD. Each of these parallel architecture classes has at least one parallel programming paradigm traditionally ....

....it. The SIMD class is traditionally programmed in a data parallel style, MIMD in a control parallel style, tc using shared variables, lc using message passing, etc. Two of his requirements for a model of parallel computation were also adopted; these are: architecture independence and congruence [Ski92a, Ski92b, Ski91]. A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91] So, the source code written in terms of a particular parallel programming paradigm is implementable on various parallel architecture classes without ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Practical parallel computation. Technical Report ISSN-0836-0227-91-312, Department of Computing and Information Science, Queen's University, Kingston, Ontario, August 1991.

....Parallel selection algorithms are useful in such practical applications as dynamic distribution of multidimensional data sets, parallel graph partitioning and parallel construction of multidimensional binary search trees. Many parallel algorithms for selection have been designed for the PRAM model [2, 3, 4, 9, 14] and for various network models including trees, meshes, hypercubes and reconfigurable architectures [6, 7, 13, 16, 22] More recently, Bader et.al. 5] implement a parallel deterministic selection algorithm on several distributed memory machines including CM 5, IBM SP 2 and INTEL Paragon. In this ....

S.G. Akl, Parallel selection in O(log log n) time using O(n= log log n) processors, Technical Report 88-221, Department of Computing and Information Science, Queen's University, Kingston, Ontario, March 1988.

....parallel case than in the sequential case. This is due to the more disparate range of parallel architectures and languages. Skillicorn presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91b] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91b] So, the source code is portable to various parallel architecture classes without modification, only requiring recompilation. For a model to be ....

.... presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91b] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91b] So, the source code is portable to various parallel architecture classes without modification, only requiring recompilation. For a model to be congruent over an architecture, the execution cost visible to the programmer should be (asymptotically) the execution cost on that architecture, ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Parallelism and the Bird-Meertens Formalism. Department of Computing and Information Science, Queen's University, Kingston, April 1992.

....case than in the sequential case. This is due to the more disparate range of parallel architectures and languages. Skillicorn presents three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91b] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91b] So, the source code is portable to various parallel architecture classes without modification, only requiring recompilation. For a model to be congruent over ....

.... three requirements for a model of parallel computation; these are given as: architecture independence, congruence and intellectual manageability [Ski92a, Ski92b, Ski91b] A model is said to be architecture independent if it is general enough to model a range of architecture types [Ski92a, Ski92b, Ski91b] So, the source code is portable to various parallel architecture classes without modification, only requiring recompilation. For a model to be congruent over an architecture, the execution cost visible to the programmer should be (asymptotically) the execution cost on that architecture, ....

[Article contains additional citation context not shown here]

D. B. Skillicorn. Practical parallel computation. Technical Report ISSN-08360227 -91-312, Department of Computing and Information Science, Queen's University, Kingston, Ontario, August 1991.

....has a worst case run time of O(N 2 ) but has an expected run time of only O(N) and is known to perform better in practice than its deterministic counterpart due to the low constant associated with the algorithm. Many parallel algorithms for selection have been designed for the PRAM model [2, 3, 4, 9, 14] and for various network models including trees, meshes, hypercubes and reconfigurable architectures [6, 7, 13, 16, 21] More recently, Bader et.al. 5] implement a parallel deterministic selection algorithm on several distributed memory machines including CM 5, IBM SP 2 and INTEL Paragon. In this ....

S.G. Akl, Parallel selection in O(log log n) time using O(n= log log n) processors, Technical Report 88-221, Department of Computing and Information Science, Queen's University, Kingston, Ontario, March 1988.

....to that region. The actual partitioning is complex, because a partition of a tree into regions is not normally itself a tree. However, with some care it is possible to get implementations of the fast parallel operations above in time complexity n=p log p for trees of size n using p processors [20]. For small p, this gives almost linear speed up over sequential implementations, while for large p it gives almost logarithmic execution times. Binary trees are easily extended to trees in which each internal node has a list of subtrees (so called Rose trees [6] Rose trees much more naturally ....

D.B. Skillicorn. Parallel implementation of tree skeletons. Technical Report 95-380, Queen's University, Department of Computing and Information Science, March 1995.

No context found.

D. B. Skillicorn. Multiprogramming BSP programs. Department of Computing and Information Science, Queen's University, Kingston, Canada, October 1996.

No context found.

D. B. Skillicorn. Multiprogramming BSP programs. Department of Computing and Information Science, Queen's University, Kingston, Canada, October 1996.

No context found.

D. B. Skillicorn. Multiprogramming BSP programs. Department of Computing and Information Science, Queen's University, Kingston, Canada, October 1996.

No context found.

D. B. Skillicorn. Multiprogramming BSP programs. Department of Computing and Information Science, Queen's University, Kingston, Canada, October 1996.

No context found.

D B Skillicorn. Parallelism and the bird-meertens formalism. Technical report, Department of Computing and Information Science, Queen's University, Kingston, Canada, April 1992.

No context found.

D B Skillicorn. Practical parallel computation: Part I. models of computation. Technical Report TR-91-313, Department of Computing and Information Science, Queen's University, Kingston, Canada, August 1991.

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