Artículos de revistas
Parallel Computation in Abstract Network Machine
Author
TCHERNYKH , ANDREI
STEPANOV , ANDREI
RODRÍGUEZ DÍAZ, ANTONIO
SCHERSON , ISAAC
Institutions
Abstract
THIS PAPER DESCIBES A PARALLEL ABSTRACT NETWORK MACHINE (ANM) WHICH USES DI-STRUCTURE (DYNAMIC INCOMPLETE STRUCTURES) AND ASSOCIATIVE NETWORKS FOR REPRESENTATION OF INFORMATION. A COMPUTATIONAL PROCESS CONSISTS OF ASYNCHRONOUS LOCAL TRANSFORMATIONS OF THE NETWORKS WITH A SINGLE MECHANISM-NETWORK UNIFICATION. AL COMPUTATIONAL MECHANISMS ARE ORIENTED TOWARDS THE PROCESSING OF INCOMPLETE INFORMATION. THE ANM IS ABLE TO PERFORM PARTIAL EVALUATION WHEN GIVEN INPUT IS INCOMPLETE AND TO AUTOMATICALLY SYNTHESIZE A RESIDUAL PARALLEL PROGRAM AS A RESULT OF TRANSFORMATION. THE TECHNIQUE FOR TRANSFORMING, OPTIMIZING, SPECIALZING MULTIPURPOSE PROGRAMS IS DESCRIBED. SOME GENERAL PROBLEMS OF DECLARATIVE PARALLEL COMPUTATION WITHOUT CONCEPTS OF A SHARED MEMORY, VALUE ASSIGNMENT, SEQUENTIAL OR PARALLEL CONTROL FLOW ARE DISCUSSED THIS PAPER DESCIBES A PARALLEL ABSTRACT NETWORK MACHINE (ANM) WHICH USES DI-STRUCTURE (DYNAMIC INCOMPLETE STRUCTURES) AND ASSOCIATIVE NETWORKS FOR REPRESENTATION OF INFORMATION. A COMPUTATIONAL PROCESS CONSISTS OF ASYNCHRONOUS LOCAL TRANSFORMATIONS OF THE NETWORKS WITH A SINGLE MECHANISM-NETWORK UNIFICATION. AL COMPUTATIONAL MECHANISMS ARE ORIENTED TOWARDS THE PROCESSING OF INCOMPLETE INFORMATION. THE ANM IS ABLE TO PERFORM PARTIAL EVALUATION WHEN GIVEN INPUT IS INCOMPLETE AND TO AUTOMATICALLY SYNTHESIZE A RESIDUAL PARALLEL PROGRAM AS A RESULT OF TRANSFORMATION. THE TECHNIQUE FOR TRANSFORMING, OPTIMIZING, SPECIALZING MULTIPURPOSE PROGRAMS IS DESCRIBED. SOME GENERAL PROBLEMS OF DECLARATIVE PARALLEL COMPUTATION WITHOUT CONCEPTS OF A SHARED MEMORY, VALUE ASSIGNMENT, SEQUENTIAL OR PARALLEL CONTROL FLOW ARE DISCUSSED