More and more applications have software implementations in order to cope with the cohabitation of several emerging standards and the fast evolution of consumer products. We show how to implement efficiently digital communications applications into DSPs with the help of the convolution co-processors, which are described, and how co-processors are useful to empower DSP performance at a small cost. ADSL application is taken as an example for broadband communications.

The efficiency of dynamic dispatch is a major impediment to the adoption of multimethods in object-oriented languages. In this article, we propose a simple multimethod dispatch scheme based on compressed dispatch tables. This scheme is applicable to any object-oriented language using a method precedence order that satisfies a specific monotonous property (e.g., as Cecil and Dylan) and guarantees that dynamic dispatch is performed in constant time, the latter being a major requirement for some languages and applications. We provide efficient algorithms to build the dispatch tables, provide their worst-case complexity, and demonstrate the effectiveness of our scheme by real measurements performed on two large object-oriented applications. Finally, we provide a detailed comparison of our technique with other existing techniques.

We present a parallel divide-and-conquer approach for detecting line segments on coarse grained architectures in general, and a particular solution on a hyper-pyramid structure. We use a method which extract directly the segments from the image without quantification such as in the classical Hough transform. The proposed solution is implemented on an environment on top of PVM.

We present in this paper four approaches to parallelizing the Hough transform on a hierarchical structure called hyper-pyramid. We show the flexibility of the structure to distribute the data and the computations according to the structure of the image and the space quantification of Hough space.