CABSim: A cycle-accurate array processor simulation environment for digital radio astronomy

Gigahertz-frequency phased arrays that are integral to next-generation radio astronomy instruments, in particular the Square Kilometre Array (SKA) radio telescope, pose a significant signal processing challenge. We argue that the development of high-performance signal processing systems is critical, not only for the particular application to future radio astronomy instrumentation, but also to the entire field of ultra-wideband phased arrays in the gigahertz bandwidth range. To this end, we have developed a cycle-accurate simulator environment and programming language for a novel, massively multicore array processor, and prototyped on it representative digital frequency domain algorithms. The results of this analysis reveal beamforming as a low compute-to-I/O processing task, best suited to high-I/O-bandwidth, streaming signal processing systems.