Physical Sciences and Mathematics Commons^™

Theory And Design Of Portable Parallel Programs For Heterogeneous Computing Systems And Networks, Ying-Chieh Wu

Dissertations

A recurring problem with high-performance computing is that advanced architectures generally achieve only a small fraction of their peak performance on many portions of real applications sets. The Amdahl's law corollary of this is that such architectures often spend most of their time on tasks (codes/algorithms and the data sets upon which they operate) for which they are unsuited. Heterogeneous Computing (HC) is needed in the mid 90's and beyond due to ever increasing super-speed requirements and the number of projects with these requirements. HC is defined as a special form of parallel and distributed computing that performs computations using …

Identifying And Exploiting Concurrency In Object-Based Real-Time Systems, Guohui Yu

Dissertations

The use of object-based mechanisms, i.e., abstract data types (ADTs), for constructing software systems can help to decrease development costs, increase understandability and increase maintainability. However, execution efficiency may be sacrificed due to the large number of procedure calls, and due to contention for shared ADTs in concurrent systems. Such inefficiencies are a concern in real-time applications that have stringent timing requirements. To address these issues, the potentially inefficient procedure calls are turned into a source of concurrency via asynchronous procedure calls (ARPCs), and contention for shared ADTS is reduced via ADT cloning. A framework for concurrency analysis in object-based …

Automatic Visual Recognition Using Parallel Machines, Yui-Liang Chen

Dissertations

Invariant features and quick matching algorithms are two major concerns in the area of automatic visual recognition. The former reduces the size of an established model database, and the latter shortens the computation time. This dissertation, will discussed both line invariants under perspective projection and parallel implementation of a dynamic programming technique for shape recognition. The feasibility of using parallel machines can be demonstrated through the dramatically reduced time complexity.

In this dissertation, our algorithms are implemented on the AP1000 MIMD parallel machines. For processing an object with a features, the time complexity of the proposed parallel algorithm is O(n) …

Mapping Of Portable Parallel Programs, Song Chen

Dissertations

An efficient parallel program designed for a parallel architecture includes a detailed outline of accurate assignments of concurrent computations onto processors, and data transfers onto communication links, such that the overall execution time is minimized. This process may be complex depending on the application task and the target multiprocessor architecture. Furthermore, this process is to be repeated for every different architecture even though the application task may be the same. Consequently, this has a major impact on the ever increasing cost of software development for multiprocessor systems. A remedy for this problem would be to design portable parallel programs which …

Parallel Exact Enumeration Of Self-Avoiding Walk On Cubic Lattices And Its Applications To Protein Folding Studies, Anek Vorapanya

Theses

Exact enumeration of self-avoiding walk on many lattices have been studied extensively recently. Even a short chain polymer (about 30 monomers) represented as a chain of cubic lattice sites requires a considerable amount of computer time to exhaustively search for all unique conformations. However, self-avoiding walk process can be modified such that it exhibits a high degree of independence among subprocesses. Parallel implementation of such subprocesses can reduce a great amount of enumeration time. Parallel enumeration makes longer chain enumeration possible.

Enumerating only unique conformations requires that all rotation and mirror conformations be removed. An algorithm to avoid generating such …

Concurrent Use Of Two Programming Tools For Heterogeneous Supercomputers, Javier G. Vasquez

Theses

In this thesis, a demostration of the heterogeneous use of two programming paradigms for heterogeneous computing called Cluster-M and HAsC is presented. Both paradigms can efficiently support heterogeneous networks by preserving a level of abstraction which does not include any architecture mapping details. Furthermore, they are both machine independent and hence are scalable. Unlike, almost all existing heterogeneous orchestration tools which are MIMD based, HAsC is based on the fundamental concepts of SIMD associative computing. HAsC models a heterogeneous network as a coarse grained associative computer and is designed to optimize the execution of problems with large ratios of computations …

Implementation Of An Automatic Mapping Tool For Massively Parallel Computing, Ajitha Gadangi

Theses

In this thesis, an implementation of a generic technique for fine grain mapping of portable parallel algorithms onto multiprocessor architectures is presented. The implemented mapping algorithm is a component of Cluster-M. Cluster-M is a novel parallel programming tool which facilitates the design and mapping of portable softwares onto various parallel systems. The other components of Cluster-M are the Specifications and the Representations. Using the Specifications, machine independent parallel algorithms are presented in a "clustered" fashion specifying the concurrent computations and communications at every step of the overall execution. The Representations, on the other hand, are a form of clustering the …

Methodology For Modeling High Performance Distributed And Parallel Systems, Rakesh Kushwaha

Dissertations

Performance modeling of distributed and parallel systems is of considerable importance to the high performance computing community. To achieve high performance, proper task or process assignment and data or file allocation among processing sites is essential. This dissertation describes an elegant approach to model distributed and parallel systems, which combines the optimal static solutions for data allocation with dynamic policies for task assignment. A performance-efficient system model is developed using analytical tools and techniques.

The system model is accomplished in three steps. First, the basic client-server model which allows only data transfer is evaluated. A prediction and evaluation method is …