Parallel and Distributed Computer Graphics


Marta Fairén
Xavier Pueyo


Computer Graphics has always been a voracious consumer of computing power. Its applications usually exploit special purpose hardware and are limited by the power of the economically feasible hardware. It is only natural then that, when parallel computing became more accessible people started to study the best ways to exploit parallel computation to expand the horizons of the field.

Of all the costs involved, the biggest is often the rendering, specially when it involves some degree of realism. It will not come as a surprise then that a large proportion of the papers in this special issue involve the parallel implementation of realistic rendering algorithms. Three papers (the first by González, Moyano, Molina and Verdú; the second by Reinhard, Chalmers and Jansen; and the third by Mancini and Pacalet) address the parallel computation of ray traced images. The first of these papers surveys the use of parallelism in the computation of ray traced images. The second exploits coherence between rays in the ray tracing of complex scenes, too large to fit complete in each of the processors. The authors also discuss the case where diffuse inter-reflections are considered. Finally, the third paper in this group proposes a new architecture for the parallelization of ray traced images, that is centered on the scene graph, their simulations show potential for high efficiency of this architecture.

Another paper by Paul, Cavin and Alonso exposes a load-balancing strategy for the parallel computation of radiosity form factors, and its implementation.

Next, two papers address problems in volume rendering. The first, by Köse and Chalmers, discusses possible setups for this problem, together with results, albeit these are measured on a somewhat outdated platform. The second, by Chaussumier and Desprez, addresses the problem of dynamic load balancing in an implementation of the Shear-Warp algorithm, that they implement on a cluster of PCs.

Finally three more papers cover other aspects of graphics amenable to parallelization. MacIntyre, discusses a strategy for the parallelization of generic, interactive, graphics applications among heterogeneous workstations. Qureshi and Hatanaka evaluate strategies for scheduling tasks in Image processing problems, and compare their performances. Finally, Mollá and Vivó concentrate on a low level problem: the anti-aliased rasterization of lines.

All in all, the papers in this issue sample different aspects of computer graphics that are amenable to parallel processing. We hope these will be useful not only to researchers in the field of computer graphics, but will also attract the interest of others to contribute in this exciting field.

Guest Editors
Marta Fairén and Xavier Pueyo.


Introduction to the Special Issue