Frans Sijstermans

High-quality and high-performance full-text document retrieval: the Parallel InfoGuide System

Describes the InfoGuide full-text document retrieval system, developed at the Philips Research Laboratories. InfoGuide combines high retrieval quality with high retrieval performance. The high quality is due to the use of the vector-space retrieval model, and the high performance originates from the implementation on the parallel POOMA machine. Implementation aspects as well as evaluation figures are presented, and it is concluded that a 200-term query can be matched against 1,500,000 documents within 3 seconds.<<ETX>>

InfoGuide: A Full-Text Document Retrieval System

This paper describes a full-text document retrieval system developed at the Philips Research Laboratories. The system is implemented on the POOMA machine, a parallel computer with a large main memory, and offers fast as well as high-quality retrieval on very large document bases.

Polyphase filtering on the TriMedia core

Lately, VLIW architectures have become popular because of their good cost-performance ratio for e.g. multimedia applications. Multimedia applications are characterized by regular signal processing and, therefore, they are apt for analysis by compilers. VLIW architectures exploit this by scheduling the instruction stream at compile time and, thus, reducing the complexity and costs of instruction issue hardware. However, sometimes we encounter signal processing algorithms that we would like to be regular and predictable but that are so only to a certain extent. Polyphase filtering is one such algorithm. It contains a regular filter part, but its input and output streams run at rates that are not correlated to each other in a simple way. Compile time analysis is, therefore, only partly possible, which poses an inherent problem for VLIW architectures. In this paper, we describe the steps that we went through to optimize the polyphase filter for a specific instance of a VLIW architecture: the Philips TriMedia processor. We show which architectural features help to make the TriMedia processor more efficient for such irregular algorithms.

Backpropagation on Distributed Memory Systems

We present a novel, parallel backpropagation algorithm. It is derived from a formal specification in a calculational manner. The design of the algorithm and its implementation on a target computer are strictly separated. We tune the algorithm for execution on a specific computer by considering the computer’s features apart rather than by considering all features together.

Template matching with a MIMD computer

The authors examine the benefits of using highly parallel, general-purpose MIMD machines for image processing. They argue that for low-level image processing the only advantages are flexibility and ease of development, which are relatively unimportant in comparison to those of more efficient special-purpose hardware. In high-level image processing, however, parallel algorithms will, in general, require computer system features that are not satisfiable by special-purpose hardware or SIMD machines. Therefore, image processing applications for which the high-level part is the most time consuming can benefit the most from MIMD machines. For an optimally efficient implementation the authors recommend a hybrid approach. Reference throughout is made to the Distributed Object Oriented Machine (DOOM), and MIMD computer intended to exploit coarse-grain parallelism.<<ETX>>