MULTI - a LISP based multiprocessing system

Abstract

A package of LISP functions, collectively called MULTI, which extends LISP 1.5 to multiprogremming is presented.MULTI defines the notion of a process within a LISP implementation using function invocation as the only control primitive.A process is an executable entity consisting of a process template and a set of register values.The process template defines the operations the process carries out. Process environments are saved in what can be viewed as function call instances, i.e. LISP forms which have the name of a process template in functional position and the register values following it. The flexibility of this simple conceptualization of processes is demonstrated by several examples which use MULTI to implement recursion, backtracking, generators, agendas and AND/OR graph searching. The implementation of MULTI does not assume that the host LISP system provides any data or control environment saving mechanisms such as FUNARG or INTERLISP's spaghetti stack. Thus, MULTI is portable to other LISP systems.is OR, could be proved by trying to show each of the P.'s in "parallel" and terminating the proofs of al~ the other disJuncts as soon as any one is proved.This "chronological" approach to disjunction is similar to the OR of Friedman and Wise[9] and the EITHER of Baker and Hewitt[2] which returns true as soon as any disJunct evaluates to true. This paper presents a general overview of MULTI and gives several examples.The examples are intended to show the flexibility of the system and to demonstrate the utility of the notion of a process in programming.We are not proposing MULTI as a progr~mmlng language but rather explaining in detail a system that we have been using for several years.Our current implementation of MULTI has been influenced by practicality, i.e. how to interface with ALISP [16] and how to write, debug and maintain programs which use MULTI.Debugging facilities available to a MULTI user in ALISP are detailed in a separate report [21]. General DescriDtion of MULTI

The Glair Cognitive Architecture

“…Reasoning is performed by an active connection graph (ACS) (McKay and Shapiro 1980;1981). Viewing the SNePS knowledge base as a propositional graph, every proposition-denoting term can be considered to be a node with arcs pointing to its arguments.…”

A knowledge engineering approach to natural language understanding

“…As each word of an input string is read by the system, the network representation of the string is extended and relevant rules stored in the SNePS network are triggered. All applicable rules are started in parallel by Processes of our MULTI package [8], are suspended if not all their antecedents are satisfied, and are resumed if more antecedents are satisfied as the string proceeds. The SNePS bidirectional inference capability [6] focuses attention towards the active parsing processes and cuts down the fan out of pure forward or backward chaining.…”

MultiMLton: A multicore-aware runtime for standard ML

“…Functional languages have long been viewed as particularly well-suited for expressing concurrent computation, and there has been much work over the years on building parallel implementations (McKay & Shapiro, 1980;Goldman & Gabriel, 1988;Miller, 1988;Feeley & Miller, 1990;Raymond, 2000) of these languages. This is because (a) firstclass procedures and expression-oriented syntax make it convenient to express concurrent threads of control, (b) the availability of advanced control structures like continuations make it straightforward to define and experiment with different kinds of user-level thread schedulers and concurrency structures, and (c) the lack of pervasive side-effects simplifies reasoning, facilitates optimizations, and reduces synchronization costs.…”