Publications

Multicommodity and Generalized Flows: Theory and Practice

Jeffrey D. Oldham. Ph.D. Thesis.

VSIPL++ Specification

Jeffrey D. Oldham and Mark Mitchell. The VSIPL++ Library is a C++, object-oriented, template-based library for vector, signal, and image processing. The library supports both serial execution similar to the VSIP Library but also supports single-program, multiple-data distributed computation without the need for explicit parallel communication code. The 250-page serial specification is nearing adoption, while the parallel specification's version number is 0.5. Developed in conjunction with MIT Lincoln Laboratory's High Performance Embedded Computing Software Initiative (HPEC-SI) http://www.hpec-si.org/.

Advanced Linux Programming

Mark Mitchell, Jeffrey Oldham, and Alex Samuel, New Riders, 2001, ISBN 0-7357-1043-0. This book uses a tutorial approach to teach C programmers how to write applications using the GNU/Linux operating system. http://www.advancedlinuxprogramming.com/.

Advanced Software Technology in VSIPL++

Edward Rutledge, Jeffrey Oldham, and Glenn Schrader. Department of Defense (DoD) High Performance Computing Modernization Program (HPCMP) Users Group Conference 2004 (UGC2004), Williamsburg, VA, 2004 Jun 07-11. Present VSIPL++ and its expression-template technology and parallel programming model. http://www.hpcmo.hpc.mil/Htdocs/UGC/UGC04/.

VSIPL++: Parallel Performance

Mark Mitchell, Jeffrey D. Oldham, and Nathan Sidwell. Submitted to the Eighth Annual High Performance Embedded Computing Workshop, 28-30 Sep. 2004, MIT Lincoln Laboratory, Lexington, MA. http://www.ll.mit.edu/hpec/.

VSIPL++ Serial and Parallel Performance

Mark Mitchell, Jeffrey D. Oldham, and Nathan Sidwell. Seventh Annual High Performance Embedded Computing Workshop, Sep. 2003, MIT Lincoln Laboratory, Lexington, MA. Experimental results comparing the C++ VSIPL++ and C VSIPL libraries. Introduction to the parallel VSIPL++ programming model. Distribution limited to workshop attendees only. http://www.ll.mit.edu/hpec/.

Scientific Computing Using POOMA

Jeffrey D. Oldham. C/C++ Users Journal, 2002 Nov. Gentle introduction to the POOMA parallel scientific programming toolkit. Present a C++ program using POOMA to model the two-dimensional neutron diffusion process inside an atomic bomb. http://www.cuj.com/documents/s=8007/cuj0211oldham/oldham.htm.

VSIPL++: Intuitive Programming Using C++ Templates

Mark Mitchell and Jeffrey D. Oldham. Sixth Annual High Performance Embedded Computing Workshop, 24-26 Sep. 2002, MIT Lincoln Laboratory, Lexington, MA. Introduction to the VSIPL++ application-programmer interface, supporting data-parallel programming, array-like data structures separating functionality from implementation, and single-program, multiple-data parallel programming. Distribution limited to workshop attendees only. http://www.ll.mit.edu/hpec/.

POOMA: A C++ Toolkit for High-Performance Parallel Scientific Computing

Jeffrey D. Oldham. User manual for the open-source POOMA C++ toolkit for parallel scientific computation. Users include nuclear weapon physicists at Los Alamos National Laboratory. http://www.codesourcery.com/public/pooma/manual.pdf.

QMTest: A Software Testing Tool

Mark Mitchell and Jeffrey D. Oldham. Presented at the 2002 Tenth International Python Conference. QMTest We present an open-source, general-purpose, cross-platform software testing tool suitable for testing a wide variety of software, e.g., compilers, databases, graphical user interfaces, or embedded systems.

An Implementation of a Combinatorial Approximation Algorithm for Minimum-Cost Multicommodity Flows

Andrew Goldberg, Jeffrey D. Oldham, Serge Plotkin, and Cliff Stein. A conference version of much of my thesis work. Published in proceedings of Integer Programming and Combinatorial Optimization 1998. Technical report version: STAN-CS-TR-97-1600, http://theory.stanford.edu/~oldham/publications/MCMCF-TR/MCMCF-TR.ps.

Combinatorial Approximation Algorithms for Generalized Flow Problems

Jeffrey D. Oldham. Invited paper for Journal of Algorithms, 38:1 (special issue of best SODA 1999 papers), Jan. 2001, pp. 135-169. Presented at Symposium on Discrete Algorithms (SODA) 1999. Presented a Bellman-Ford-based algorithm for the single-source generalized shortest paths problem, also called the restricted generalized uncapacitated transshipment problem. Using this algorithm as a subroutine, I presented combinatorial, fully polynomial-time approximation schemes for all generalized flow problems (with nonnegative costs). Also available as http://theory.stanford.edu/~oldham/publications/generalized/sp.ps.

Parallel Algorithms for the All-Sources Generalized Shortest Paths Problem

Jeffrey D. Oldham and Vaughan Pratt. Poster at Eleventh ACM Symposium on Parallel Algorithms and Architectures Revue, 1999 Jun 29, Saint-Malo, France, http://www.cs.dartmouth.edu/SPAA/1999/revue.html. Using monotonic piecewise-linear functions and the Floyd-Warshall algorithm, we present a logarithmic-time PRAM algorithm for the all-sources generalized shortest paths problem. Available at http://theory.stanford.edu/~oldham/publications/generalized/asgsp.ps.

Accurate Approximations for Asian Options

Donald Aingworth, Rajeev Motwani, and Jeffrey D. Oldham. Presented at the Symposium on Discrete Algorithms (SODA) 2000. We presented an asymptotic fully-polynomial approximation scheme for pricing Asian options on a lattice. This is the first algorithm yielding guaranteed error bounds in polynomial time.

How to Beat Random Sampling Every Time

Donald Aingworth, Rajeev Motwani, and Jeffrey D. Oldham. We present an unbiased estimator with lower variance than regular random sampling when each random event's value and probability are known.

Means for Searching

Jeffrey D. Oldham. We generalize binary and generalized search by demonstrating a one-to-one relationship between search techniques, means, and continuous, strictly monotonic functions.

The Incremental Investment Problem: Determining Sustainable Withdrawal and Accumulation Rates

Jeffrey D. Oldham. We define the incremental investment problem is to determine the probability of reaching a specified financial goal through periodically adding or withdrawing from an investment, presenting a one-pass, matrix-based algorithm for use with historical data. We also extend to incorporate capital gains. In progress.

Managing TCP Connections Under Persistent HTTP

Edith Cohen, Haim Kaplan, and Jeffrey D. Oldham. Computer Networks, 31:11-16, 17 May 1999, pp. 1709-1723. Presented at The Eighth International World Wide Web Conference. Persistent HTTP connections, included in the draft HTTP/1.1 protocol, permit several HTTP request/response messages to be sent using the same TCP connection. We designed and experimentally evaluated policies to predict when to terminate persistent HTTP connections. Available at http://www8.org/w8-papers/5c-protocols/policies/policies.html.

Experiences Using the ParaScope Editor: An Interactive Parallel Programming Tool

Mary W. Hall, Timothy J. Harvey, Ken Kennedy, Nathaniel McIntosh, Kathryn S. McKinley, Jeffrey D. Oldham, Michael H. Paleczny, Gerald Roth. Proceedings of the Fourth ACM SIGPLAN Symposium on Principles & Practice of Parallel Programming (PPOPP). May 1993. pp. 33-43. Technical report version: CRPC-TR93297, Rice University, also available as http://softlib.rice.edu/pub/CRPC-TRs/reports/CRPC-TR93297.pdf.