OR-parallel execution of Prolog on a multi-sequential machine

Abstract

Execution of Prolog Machine on aBased on extending the sequential execution model of Prolog to include parallel execution, we present a method for OR-parallel execution of Prolog on a multiprocessor system. The method reduces the overhead incurred by parallel processing. It allows many processing elements (PEs) to process simultaneously a common branch of a search tree, and each of these PEs creates its local environment and selects a subtree for processing without communication. The run-time overhead is small: simple and efficient operations for selecting the proper subtree. Communication is necessary only when some PEs have exhausted their search spaces and there are others still searching for solutions. The method is able to utilize most of the technology devised for sequential implementation of Prolog. It is optimized for an architecture that supports broadcast copying.

Parallel execution of prolog programs

“…4. For instance, the following models employ an environment representation technique that satisfies criteria 1 and 2 above (constant-time task creation and variable access): Versions Vectors Scheme [Hausman et al 1987], Binding Arrays Scheme [Warren 1984;1987c], Argonne-SRI Model [Warren 1987b], Manchester-Argonne Model [Warren 1987b], Delphi Model [Clocksin and Alshawi 1988], Randomized Method [Janakiram et al 1988], BC-Machine [Ali 1988], MUSE [Ali and Karlsson 1990b] (and its variations, such as stack splitting [Gupta and Pontelli 1999c], SBA · 21 [Correia et al 1997], PBA [Gupta et al 1993;), Virtual Memory Binding Arrays model [Véron et al 1993] and, Kabu-Wake Model [Masuzawa et al 1986]; while the following models employs an environment representation technique that satisfies criteria 2 and 3 above (constant-time variable access and task switch): Directory Tree Method [Ciepielewski and Haridi 1983], and Environment Closing Method [Conery 1987a]; and the following models employs an environment representation technique that satisfies criteria 1 and 3 above (constant-time task-creation and task-switch): Hashing Windows Method [Borgwardt 1984], Favored-Bindings Model , and Virtual Memory Hashing Windows model [Véron et al 1993]. Likewise, example of a model that only satisfies criterion 1 (constant time task-creation) is the Time-Stamping Model [Tinker 1988], while the example of a model that only satisfies criterion 3 (constant-time task switching) is the Variable Import Scheme [Lindstrom 1984].…”

“…This model is an evolution of the work on BC-machine by [Ali 1988]-a model where different workers concurrently start the computation of the query and automatically select different alternatives when choice points are created. The idea was already present in the Kabu Wake model [Masuzawa et al 1986].…”

“…Examples of processor-based systems are &-Prolog, Aurora, MUSE, Andorra-I, PEPSys, AO-WAM, DDAS, ACE, PBA, etc. Processor-based systems are more suited for shared memory machines, although techniques like stack-copying and stack-splitting show a high degree of locality in memory reference behavior and hence are suited for non-shared memory machines as well [Ali 1988;Ali et al 1992;Gupta and Pontelli 1999c]. As has been shown by the ACE model, MUSE's stackcopying technique can be applied to and-or parallel systems as well, so one can envisage implementing a processor-based system on a non-shared memory machine using stack-copying [Villaverde et al 2001;.…”

Paralelismo em Prolog: Conceitos e Sistemas

“…A implementação de Muse foi baseada no desenvolvimento de um sistema para paralelismo-OU chamado BC-Machine [2], focado em arquitetura de memória distribuída. A grande contribuição do sistema Muse foi o desenvolvimento do ambiente de cópia incremental, desenvolvido para minimizar o overhead de sincronização, fazendo com que cada instância dentro da mesma ramificação de busca possua uma parte consistente de dados e outra com modificações locais, e a sincronização dos dados é feita copiando apenas a diferença entre o local e global.…”

ACE: And/or-parallel copying-based execution of logic programs

“…In this approach, stack frames are explicitly copied from the stack(s) of one processor 1 to that of another whenever the latter processor needs to share a branch of the or-parallel tree of the former. In practice, by having an identical logical address space for tLaboratory for Logic, Datábase, and Advanced Programming, Dept.…”