Physical Sciences and Mathematics Commons^™

Parallel H-V Drawings Of Binary Trees, Panagiotis T. Metaxas, Grammati E. Pantziou, Antonis Symvonis

Computer Science Technical Reports

In this paper we present a method to obtain optimal h-v and inclusion drawings in parallel. Based on parallel tree contraction, our method computes optimal (with respect to a class of cost functions of the enclosing rectangle) drawings in $O(\log^2 n)$ parallel time by using a polynomial number of EREW processors. The number of processors reduces substantially when we study minimum area drawings. The method can be extended to compute optimal inclusion layouts in the case where each leaf $l$ of the tree is represented by rectangle $l_x \times l_y$ (the dimensions of which are part of the input). For …

On-Line And Dynamic Shortest Paths Through Graph Decompositions, Hristo N. Djidjev, Grammati E. Pantziou, Christos D. Zaroliagis

Computer Science Technical Reports

We describe algorithms for finding shortest paths and distances in a planar digraph which exploit the particular topology of the input graph. An important feature of our algorithms is that they can work in a dynamic environment, where the cost of any edge can be changed or the edge can be deleted. For outerplanar digraphs, for instance, the data structures can be updated after any such change in only $O(\log n)$ time, where $n$ is the number of vertices of the digraph. We also describe the first parallel algorithms for solving the dynamic version of the shortest path problem. Our …

Parallel Max Cut Approximations, Grammati E. Pantziou, Paul G. Spirakis, Christos D. Zaroliagis

Computer Science Technical Reports

Given a graph with positive integer edge weights one may ask whether there exists an edge cut whose weight is bigger than a given number. This problem is NP-complete. We present here an approximation algorithm in NC which provides tight upper bounds to the proportion of edge cuts whose size is bigger than a given number. Our technique is based on the methods to convert randomized parallel algorithms into deterministic ones introduced by Karp and Wigderson. The basic idea of those methods is to replace an exponentially large sample space by one of polynomial size. In this work, we prove …

Wavelet Localization Of The Radon Transform, Tim Olson, Joe Destefano

Computer Science Technical Reports

In this paper we develop an algorithm which significantly reduces radiation exposure in x-ray tomography, when a local region of the body is to be imaged. The algorithm uses the properties of wavelets to essentially localize the Radon transform. The algorithm differs from previous algorithms for doing local tomography because it recovers an approximation to the original image, not the image modulo the nullspace of the local tomography operator, or the Lambda transform of the image. This is possible because we do not truly invert the interior Radon transform, but rather sample the Radon transform sparsely away from the local …

Off-Line Cursive Handwriting Recognition Using Style Parameters, Berrin A. Yanikoglu, Peter A. Sandon

Computer Science Technical Reports

We present a system for recognizing off-line cursive English text, guided in part by global characteristics of the handwriting. A new method for finding the letter boundaries, based on minimizing a heuristic cost function, is introduced. The function is evaluated at each point along the baseline of the word to find the best possible segmentation points. The algorithm tries to find all the actual letter boundaries and as few additional ones as possible. After size and slant normalizations, the segments are classified by a one hidden layer feedforward neural network. The word recognition algorithm finds the segmentation points that are …

Integrating Theory And Practice In Parallel File Systems, Thomas H. Cormen, David Kotz

Dartmouth Scholarship

Several algorithms for parallel disk systems have appeared in the literature recently, and they are asymptotically optimal in terms of the number of disk accesses. Scalable systems with parallel disks must be able to run these algorithms. We present for the first time a list of capabilities that must be provided by the system to support these optimal algorithms: control over declustering, querying about the configuration, independent I/O, and turning off parity, file caching, and prefetching. We summarize recent theoretical and empirical work that justifies the need for these capabilities. In addition, we sketch an organization for a parallel file …

Throughput Of Existing Multiprocessor File Systems (An Informal Study), David Kotz

Computer Science Technical Reports

Fast file systems are critical for high-performance scientific computing, since many scientific applications have tremendous I/O requirements. Many parallel supercomputers have only recently obtained fully parallel I/O architectures and file systems, which are necessary for scalable I/O performance. Scalability aside, I show here that many systems lack sufficient absolute performance. I do this by surveying the performance reported in the literature, summarized in an informal table.

Accurate Verification Of Five-Axis Numerically Controlled Machining, Jerome L. Quinn

Dartmouth College Ph.D Dissertations

Current automated machining systems are composed of a number of components to aid in bringing a surface from design to physical completion. Numerically controlled (NC) milling machines are used to cut parts out of stock. Programming these machines to cut a desired surface is still largely a matter of experienced human participation. Therefore, the need exists to verify that tool programs produce the desired part.

We present recent developments in the verification of NC tool programs. Many of these methods rely on approximating the stock material as vectors whose lengths reflect the amount of uncut material at any point. This …

Efficient Parallel Algorithms For Some Tree Layout Problems, J Diaz, A Gibbons, Grammati E. Pantziou, M Serna, Paul G. Spirakis, J Toran

Computer Science Technical Reports

The minimum cut and minimum length linear arrangement problems usually occur in solving wiring problems and have a lot in common with job sequencing questions. Both problems are NP-complete for general graphs and in P for trees. We present here two algorithms in NC. The first solves the minimum length linear arrangement problem for unrooted trees in $O(\log^2 n)$ time and $O(n^2 3^{\log n})$ CREW PRAM processors. The second algorithm solves the minimum cut arrangement for unrooted trees of maximum degree $d$ in $O(d \log^2 n)$ time and $O(n^2 /\log n)$ CREW PRAM processors.

Integrating Theory And Practice In Parallel File Systems, Thomas H. Cormen, David Kotz

Computer Science Technical Reports

Several algorithms for parallel disk systems have appeared in the literature recently, and they are asymptotically optimal in terms of the number of disk accesses. Scalable systems with parallel disks must be able to run these algorithms. We present a list of capabilities that must be provided by the system to support these optimal algorithms: control over declustering, querying about the configuration, independent I/O, turning off file caching and prefetching, and bypassing parity. We summarize recent theoretical and empirical work that justifies the need for these capabilities.

Videoscheme: A Programmable Video Editing System For Automation And Media Recognition, James Matthews, Peter Gloor, Fillia Makedon

Computer Science Technical Reports

The recent development of powerful, inexpensive hardware and software support had made digital video editing possible on personal computers and workstations. To date the video editing application category has been dominated by visual, easy-to-use, direct manipulation interfaces. These systems bring high-bandwidth human-computer interaction to a task formerly characterized by slow, inflexible, indirectly-operated machines. However, the direct manipulation computer interfaces are limited by their manual nature, and can not easily accommodate algorithmically- defined operations. This paper proposes a melding of the common direct manipulation interfaces with a programming language which we have enhanced to manipulate digital audio and video. The result …

Asymptotically Tight Bounds For Performing Bmmc Permutations On Parallel Disk Systems, Thomas H. Cormen, Leonard F. Wisniewski

Computer Science Technical Reports

No abstract provided.

Vector Layout In Virtual-Memory Systems For Data-Parallel Computing, Thomas H. Cormen

Computer Science Technical Reports

In a data-parallel computer with virtual memory, the way in which vectors are laid out on the disk system affects the performance of data-parallel operations. We present a general method of vector layout called banded layout, in which we divide a vector into bands of a number of consecutive vector elements laid out in column-major order, and we analyze the effect of the band size on the major classes of data-parallel operations. We find that although the best band size varies among the operations, choosing fairly small band sizes—at most a track—works well in general.

Multiprocessor File System Interfaces, David Kotz

Dartmouth Scholarship

Increasingly, file systems for multiprocessors are designed with parallel access to multiple disks, to keep I/O from becoming a serious bottleneck for parallel applications. Although file system software can transparently provide high-performance access to parallel disks, a new file system interface is needed to facilitate parallel access to a file from a parallel application. We describe the difficulties faced when using the conventional (Unix-like) interface in parallel applications, and then outline ways to extend the conventional interface to provide convenient access to the file for parallel programs, while retaining the traditional interface for programs that have no need for explicitly …

Practical Prefetching Techniques For Multiprocessor File Systems, David Kotz, Carla Schlatter Ellis

Dartmouth Scholarship

Improvements in the processing speed of multiprocessors are outpacing improvements in the speed of disk hardware. Parallel disk I/O subsystems have been proposed as one way to close the gap between processor and disk speeds. In a previous paper we showed that prefetching and caching have the potential to deliver the performance benefits of parallel file systems to parallel applications. In this paper we describe experiments with practical prefetching policies that base decisions only on on-line reference history, and that can be implemented efficiently. We also test the ability of these policies across a range of architectural parameters.