Cell broadband engine architecture as a DSP platform
Karol Szumski, Mateusz Malanowski · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009
The slowing pace of performance improvement in the commonly available processors is a cause of concern amongst many computational scientists. This combined with the ever increasing need for computational power has caused us to turn to alternative architectures in search of performance gains. Two main candidates were the Compute Unified Device Architecture (CUDA) and the Cell Broadband Engine (CELL BE) architecture. This paper focuses on the latter, outlining the architecture and basic programming paradigms, and also contains performance comparison of algorithms currently developed by our team.