An interview with Alan Chalmers - Chair of third Eurographics Workshop on Parallel Graphics & Visualization

Primeur: What can be found there?
AC: State-of-the-art research papers from around the world on all aspects of parallel graphics and visualization will be presented and discussed in detail. It is one of the primary objectives of this Eurographics workshop series to provide a friendly atmosphere in which ideas can be discussed and analysed to the benefit of all participants.

Primeur: Is there any relationship with the Parallel Visualization and Graphics Symposium that is organized in the USA
AC: Yes - a close relationship. This Eurographics workshop, held every even year, directly complements the IEEE Symposium held on odd years. This close collaboration between these two events ensures there is an international venue, one year in Europe and the next in the USA for parallel graphics and visualization.

Primeur: Could you give us a separate definition for graphics and visualization? What are the reasons for considering them as different fields?
AC: In broad terms, computer graphics is concerned with producing realistic images which accurately model environments including the propagation of light within those environments. Examples include lighting design for architects planning a building, realistically reconstructing archaeological sites etc.

Visualization, on the other hand, is concerned with displaying results of numerical simulations in a manner which makes it easy for the user to comprehend those results. Examples include modelling the airflow over an aircraft wing, displaying the sediment flow in a river estuary etc.

Primeur: Which are the contributions of parallelism to these fields?
AC: Both computer graphics and visualization are very computationally intensive. The solution of complex models on a serial machine can take many hours, even days. Parallel processing offers the potential for solving these problems in reasonable times.

Primeur: Clusters of PCs are becoming more and more powerful and cheap. How can this affect to the visualization and graphics fields?
AC: Clusters of powerful PCs are providing an exciting, affordable alternative to the traditional very expensive parallel machines. Parallel processing is expected to become even more popular with graphics and visualization practioners enabling them to consider even more complex models in reasonable times.

However, care must be taken. The communication infrastructures of these clusters are still relatively slow and this can inhibit the number of PCs that can be considered in any parallel implementation. More work is needed to improve the performance of communication between PCs and also to develop techniques which can reduce latency by, for example, overlapping communication with computation.

Primeur: Which is the best option: distributed or parallel calculation with generic hardware, only graphic processors, or a hybrid system using a pipeline of generic and graphic processors?
AC: This really depends on the application being considered. Dedicated graphics processors, such as games playstations etc, are being increasingly coupled together to provide excellent parallel performance for certain essential computations. However, parallel processing is not simply about computing, be it the graphics or visualization algorithm. Significant effort has to be devoted to ensuring that the computational effort is distributed evenly to the processors to avoid processor idle. Furthermore, many graphic and visualization problems have very large data requirements, too large to allow the data to simply be replicated at each processor. Some form of data management will thus be necessary to ensure the appropriate piece of data is at the right place within the system when required.

Such Task and Data management is better suited to generic processors, including some interesting work with Field Programmable Gate Arrays (FPGAs).

Also, with parallel processing, as mentioned earlier, the communication infrastructure can never be ignored. Processors, be they generic or dedicated, can not simply be "coupled together", great care has to be taken to ensure that the latency in getting messages between the processors is minimised.

Primeur: What is the role of parallelism in a VR Centre?
AC: Only parallelism offers the potential to provide realistic results in reasonable time to any Virtual Reality device. Any VR Centre which wants to consider realistic environments must, therefore, embrace parallelism.I think we are still a little way from providing very realistic results in real time, but the on-going research in parallel graphics and visualization, such as will be presented at the Eurographics Workshop, is striving towards this goal: "Realism in Real-Time".

Primeur: What do you think about the research in Europe and Spain in visualization and graphics?
AC: Much of the leading research in parallel graphics and visualization has come for groups in Europe, such as those in Spain, including the one in Girona who are hosting this years workshop. Increasingly close collaboration between researchers in Europe and elsewhere in the world is accelerating the speed at which the difficulties of implementing these algorithms in parallel are being overcome, offering the exciting possibilities of investigation even more complex models. The 3rd Eurographics Workshop on Parallel Graphics and Visualization is one of the leading events this year where researchers and users can meet, discuss ideas, and build these close collaborations.

More information about this workshop at http://iiia.udg.es/egwpgv00