The early decades of computing were marked by limited resources. However, as we enter the twenty-first century, silicon is offering enormous computing resources on a single die and molecular-scale devices appear plausible offering a path to even greater capacities. Exploiting the capacities of these modern and future devices demands different computational models and radical shifts in the way we organize, capture, and optimize computations. A key shift is toward spatially organized computation. A natural consequence is that the dominant effects which govern our computing space change from the total number of operations and temporal locality to interconnect complexity and spatial locality. Old computational models which hide, ignore, or obfuscate communication and emphasize temporal sequences inhibit the exploitation of these modern capacities, motivating the need for new models which make communication and spatial organization more apparent.
Paper