Virginia Tech - Wake Forest University School of Biomedical Engineering & Sciences
Computer Science Technical Reports @Virginia TechNot a member yet
997 research outputs found
Sort by
Functionally Complete Machines
Full text linkThis paper defines a functionally complete machine as a machine which is capable of evaluating every two place function over its data space. Necessary conditions on memory size for completeness are developed. These conditions are applied to System/360 as modelled by the space of bytes, the space of halfwords, and the space of words. Sufficiently large (> 64K bytes) models of System/360 are shown to be complete for the space of bytes. No models of System/360 are complete for the spaces of halfwords or words. The inequalities developed and known examples of universal decision elements suggest structures for complete machines
A Topological Model of the Data Space for a Block Structured Language
Full text linkA space X is defined, the points of which are trees which possess a unique name property. A topology is defined for X based upon partial order. The topology is shown to be reasonably natural relative to the rationals. The topological space is shown to be neither Hausdorff nor T_1. The implications of this for program convergence are discussed with examples
Generalized Structured Programming
Full text linkIn an effort to eliminate some inconveniences connected with Dijkstra's method of Structured Programming, a generalized set of basic flow graphs for structuring programs is suggested. These structures generate the set of all flow graphs that can be fully decomposed by Allen and Cocke's method of interval reduction. It will be shown that programs Composed of the proposed basic structures have most, if not all, of the Positive characteristics claimed for programs written with the classic rules of Structured Programming. Further, by extending Wirth's programming language PASCAL a set of new control constructs has been suggested that support the proposed set of flow structures
LPL: A GENERALIZED LIST PROCESSING LANGUAGE
Full text linkThe paper describes LFL, a generalized list processing language. LFL allows the user to define multiple cell structures and cell sizes at runtime, thereby allowing nonhomogeneous list structures. The paper examines the problems associated with list tracing in systems allowing multiple cell-types. Complex list tracing during garbage collection in LFL is avoided: (1) by creating a doubly-linked super list of all allocated cells and (2) by using a reference count scheme. No marking phase is required for garbage collection.
The problem of developing insertion and deletion procedures for lists with cells having multiple types of pointer structures is discussed and LFL solutions are given. LPL statements can handle singly-linked, doubly-linked, left-right-linked, and some multi-linked pointer structures automatically.
The design philosophy and the data organization for LPL are discussed in detail. Examples of the definition of cell structures are given, and all of the LPL list manipulation and creation statements arc examined and discussed
THE DYNAMIC CREATION AND MODIFICATION OF HEURISTICS IN A LEARNING PROGRAM
Full text linkPOLY FACT is a learning program that attempts to factor multivariable polynomials. The program has been successful in factoring polynomials (in simplified form) with a maximum of 84 terms, each term consisting of as many as five variables and a maximum degree of 67. The complexity of this learning task placed unusual requirements on the representation of
heuristics. By using the first-order predicate calculus notation, we enable the creation and modification of heuristics dynamically during program execution. Constraints on the creation process are implemented in a series of tables by which one can alter the flexibility given to the program. Execution of heuristics begins with a translation of the predicate calculus representation to a reverse Polish string, followed by the interpretive evaluation of the Polish string. A general procedure for developing and implementing the predicate calculus representation is suggested
