Programming Models for Parallel Heterogeneous Computing

Many of the computational problems we are facing today are complex and need huge computational power to be solved. It is well-known that processors will not continue to get faster, but will get more cores instead. More cores are not only harder to utilize by an application programmer, but also challenge hardware designers. Thus various new hardware architectures are designed and evaluated in order to find the ones that will fulfill the needs of future computer systems. Prototypic configuration boards like Intels Single Chip Cloud Computer (SCC) are an attempt to deal with the ever-increasing number of cores by removing essential features of current processors like hardware cache coherency. Another approach is to accompany common general purpose CPUs with sophisticated special purpose processing units. These so called Accelerators are easier to build and very fast for specific application purposes. They are the foundation for the new trend of hybrid computer systems.