Discover articles, books and research

Suppose that a programming language involves, among other familiar ways of composing commands Ci, a "parallel" construct (C1 par C2) . One expects, when using this language, that a sequence such as x := O; y := 1; (x := 1 par (while x=0 do y := y+1)) should be guaranteed to terminate in whatever context it is executed.

The effects of bases of two-input boolean functions are characterised in terms of their impact on some questions in parallel computation. It is found that a certain set of bases (called the P-complete set) which are not necessarily complete in the classical sense, apparently makes the circuit value problem difficult, and renders extended Turing machines and…

We consider various models of parallel computers based on communication networks of sequential processors. The degree of a parallel machine is the number of communication lines connected Lo each processor; the arity is the number of these lines which a processor can actively manipulate at any given time.

Let T(n) be the time to recognize context-free languages on a parallel random access machine without write conflicts (PRAM) using a polynomial number of processors. We assume that T(n) = Q(log n).

We describe a deterministic parallel algorithm to compute algebraic expressions in log n time using n/log(n) processors on a parallel random access machine without write confilects (P-RAM) with no free preprocessing.

We prove that every outerplanar graph can be optimally edge-coloured in polylog time using a polynomial number of processors on a parallel random access machine without write conflicts (P-RAM).

We show that every Halin graph can be optimally edge-coloured or optimally vertex-coloured in polylog time using a polynomial number of processors on a parallel random access machine without write conflicts (P-RAM).

We prove that the parsing problem for bracket context-free languages can be solved in log(n) time using n/log(n) processors on a parallel random access machine without write conflicts (P-RAM).

A general method for parallelism of some dynamic programming algorithms on VLSI was presented in [6]. We present, a general method for parallelisation for the same class of problems on more powerful parallel computers.

We present parallel O(log(n))-time algorithms for optimal edge colouring of trees and Halin graphs with n processors on a a parallel random access machine without write conflicts (P-RAM).

This paper describes an abstract machine model for parallel computation that exploits a programming paradigm based upon definitions. Proposed applications to the implementation of CAD systems, and to the modelling and simulation of concurrent systems are outlined.

Programming language principles are of fundamental importance in supporting major applications on multiprocessor architectures. Many different language paradigms for general-purpose parallel computing have been proposed.

This brief set of notes outline ways in which the author's cassandra architecture can be implemented as a parallel or distributed system.

A new model of list processing is proposed which is well suited to parallel implementation. Its main primitive functions are: "concatenate", which concatenates two lists; "split", which partitions a list into two parts; and "length", which gives the number of elements in a list.

This is an introduction to distributed memory models of parallel computation. Communication networks that have been advocated for general purposes are described and the problem of parallel routing within these networks is addressed through the permutation routing paradigm.

The order in which the matrices are multiplied affects the number of operations. The best sequential algorithm for computing an optimal order of matrix multiplication runs in O (n log n) time while the best known parallel NC algorithm runs in O (log2n) time using n6/log6n processors.

A parallel graph reduction machine simulator is described. This performs combinator reduction and can simulate various different parallel reduction strategies.

This paper considers the problem of finding an optimal order of the multiplication chain of matrices. All parallel algorithms known use the dynamic programming approach and run in a polylogarithmic time using, in the best case, n6/log6n processors.

Software Performance Engineering (SPE) offers a solution to this problem by using a software execution model, in addition to the system model. The use of SPE for parallel software has several disadvantages: complex nature of the SPE system model, non re-usable model components, and lack of analytical methods to evaluate the system model.

This paper describes a massively parallel architecture for image understanding and computer vision. A key issue in these applications is the transition from iconic to more abstract symbolic data representations.