Seismic Imaging and the Road to Petascale Capacity: RTM and the Cell /B.E. Processor

ch, on the other hand, is around the<br>corner. Seismic Imaging is definitively a field in our industry where petascale capacity is<br>needed. The question is not when this capacity will be widely available, but how. There are<br>several hardware processors and devices as a candidates for the brain of the new generation of<br>petascale supercomputers. The only thing in common is that all of them are difficult to<br>program, and the programming will be different from the programming of today’s x86<br>processor generation. We believe that among all the hardware options, the Cell /BE processor<br>has several characteristics that make it ideal for widely available petascale capacity. Besides<br>the difficulty of programing the Cell, we present a benchmark for RTM between Cell and<br>PowerPC processors. We show that using an early generation of the Cell, and a difficult<br>Kernel of a compute intensive algorithm, we may expect almost one order of magnitude of<br>performance increase.