Design of the Kernel Language for the Parallel Inference Machine

Ueda, Kazunori; Chikayama, Takashi

doi:10.1093/comjnl/33.6.494

Cited by 79 publications

(23 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We regard this separation of concerns as the most important achievement of KL1 and its parallel implementation [29]. In other words, by using logical variables as communication channels we had achieved 100% network transparency within system-area networks (SAN).…”

Section: Goal Of This Papermentioning

confidence: 99%

“…Relational Language [7], the first concrete proposal of a concurrent logic language, was followed by a succession of proposals, namely Concurrent Prolog [20], PARLOG [8] and Guarded Horn Clauses (GHC) [27]. KL1 [29], the Kernel Language of the Fifth Generation Computer Systems (FGCS) project [22], was designed based on GHC by featuring (among others) mapping constructs for concurrent processes. To be precise, KL1 is based on Flat GHC [28], a subset of GHC that restricts guard goals to calls to test predicates.…”

Section: Concurrency and Logic Programmingmentioning

confidence: 99%

See 1 more Smart Citation

A Pure Meta-interpreter for Flat GHC, a Concurrent Constraint Language

Ueda

2002

Computational Logic: Logic Programming and Beyond

View full text Add to dashboard Cite

Abstract. This paper discusses the construction of a meta-interpreter of Flat GHC, one of the simplest and earliest concurrent constraint languages. Meta-interpretation has a long history in logic programming, and has been applied extensively to building programming systems, adding functionalities, modifying operational semantics and evaluation strategies, and so on. Our objective, in contrast, is to design the pair of (i) a representation of programs suitable for code mobility and (ii) a pure interpreter (or virtual machine) of the represented code, bearing networked applications of concurrent constraint programming in mind. This is more challenging than it might seem; indeed, meta-interpreters of many programming languages achieved their objectives by adding small primitives into the languages and exploiting their functionalities. A metainterpreter in a pure, simple concurrent language is useful because it is fully amenable to theoretical support including partial evaluation. After a number of trials and errors, we have arrived at treecode, a groundterm representation of Flat GHC programs that can be easily interpreted, transmitted over the network, and converted back to the original syntax. The paper describes how the interpreter works, where the subtleties lie, and what its design implies. It also describes how the interpreter, given the treecode of a program, is partially evaluated to the original program by the unfold/fold transformation system for Flat GHC.

show abstract

Section: Goal Of This Papermentioning

confidence: 99%

Section: Concurrency and Logic Programmingmentioning

confidence: 99%

A Pure Meta-interpreter for Flat GHC, a Concurrent Constraint Language

Ueda

2002

Computational Logic: Logic Programming and Beyond

View full text Add to dashboard Cite

show abstract

“…When the author proposed Guarded Horn Clauses (GHC) [36] [37] as a simplification of Concurrent Prolog and PARLOG [8], the principal design constraint was to retain channel mobility and evolving process structures [32], because GHC was supposed to be the basis of KL1 [39], a language in which to describe operating systems of the Parallel Inference Machines as well as various knowledge-based systems. 4.…”

Section: Introduction -Constraint-based Concurrencymentioning

confidence: 99%

Resource-Passing Concurrent Programming

Ueda

2001

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. The use of types to deal with access capabilities of program entities is becoming increasingly popular. In concurrent logic programming, the first attempt was made in Moded Flat GHC in 1990, which gave polarity structures (modes) to every variable occurrence and every predicate argument. Strong moding turned out to play fundamental rôles in programming, implementation and the in-depth understanding of constraint-based concurrent computation. The moding principle guarantees that each variable is written only once and encourages capability-conscious programming. Furthermore, it gives less generic modes to programs that discard or duplicate data, thus providing the view of "data as resources." A simple linearity system built upon the mode system distinguishes variables read only once from those read possibly many times, enabling compile-time garbage collection. Compared to linear types studied in other programming paradigms, the primary issue in constraint-based concurrency has been to deal with logical variables and highly non-strict data structures they induce. In this paper, we put our resource-consciousness one step forward and consider a class of 'ecological' programs which recycle or return all the resources given to them while allowing concurrent reading of data structures via controlled aliasing. This completely recyclic subset enforces us to think more about resources, but the resulting programs enjoy high symmetry which we believe has more than aesthetic implications to our programming practice in general. The type system supporting recyclic concurrent programming gives a [−1, +1] capability to each occurrence of variable and function symbols (constructors), where positive/negative values mean read/write capabilities, respectively, and fractions mean non-exclusive read/write paths. The capabilities are intended to be statically checked or reconstructed so that one can tell the polarity and exclusiveness of each piece of information handled by concurrent processes. The capability type system refines and integrates the mode system and the linearity system for Moded Flat GHC. Its arithmetic formulation contributes to the simplicity. The execution of a recyclic program proceeds so that every variable has zero-sum capability and the resources (i.e., constructors weighted by their capabilities) a process absorbs match the resources it emits. Constructors accessed by a process with an exclusive read capability can be reused for other purposes. The first half of this paper is devoted to a tutorial introduction to constraint-based concurrency in the hope that it will encourage cross-fertilization of different concurrency formalisms.

show abstract

“…The immediate goal of this research was to develop a fast theorem-proving system on the PIM, by effectively utilizing KL1 languages [24] and logic programming techniques. We intended this system to be an "advanced general-purpose inference engine" that can facilitate the development of intelligent knowledge programming software in KL1.…”

Section: Introductionmentioning

confidence: 99%

MGTP: A model generation theorem prover — Its advanced features and applications —

Hasegawa

Fujita

Koshimura

1997

Lecture Notes in Computer Science

View full text Add to dashboard Cite

This paper outlines a parallel model-generation based theoremproving system MGTP that we have been developing, focusing on the recent developments including novel techniques for efficient proof-search and successful applications. We have developed CMGTP (Constraint MGTP) to deal with constraint satisfaction problems. By attaining the constraint propagation with negative atoms, CMGTP makes it possible to reduce search spaces by orders of magnitude compared to the original MGTP. To enhance the ability to prune search spaces, we have developed a new method called non-Horn magic sets (NHM) and incorporated its relevancy testing function into the folding-up (FU) method proposed by Letz. The NHM method suppresses useless model generation with clauses irrelevant to the goal. The FU method avoids generating duplicated subproofs after case-splitting. With these methods we can eliminate two major kinds of redundancies in model-generation based theorem provers. We have studied several applications in AI such as negation as failure, abductive reasoning and modal logic systems, through extensive use of MGTP. These studies share a basic common idea, that is, to use MGTP as a meta-programming system. We can build various reasoning systems on MGTP by writing the specific inference rules for each system in MGTP input clauses. Furthermore, we are now working on other applications such as machine learning with MGTP and heuristic proof-search based on the genetic algorithm.

show abstract

Design of the Kernel Language for the Parallel Inference Machine

Cited by 79 publications

References 0 publications

A Pure Meta-interpreter for Flat GHC, a Concurrent Constraint Language

A Pure Meta-interpreter for Flat GHC, a Concurrent Constraint Language

Resource-Passing Concurrent Programming

MGTP: A model generation theorem prover — Its advanced features and applications —

Contact Info

Product

Resources

About