Hybrid architectures

The speed of simulations used to increase with the clock rate of processors (Moore’s law). But since a few years processor cores are no longer getting faster, but rather we are getting more and more cores. To make use of the computational power of massively parallel computers, massively parallel algorithms are required. As stream computing is becoming common, using e.g. graphical processing units (GPUs), the parallelization also needs to take hybrid architectures into account.