Mapping of MIMO Receiver Algorithms onto Application-Specific Multi-Core Platforms

General purpose processors are inefficient for high throughput signal processing. Platforms incorporating applicationspecific cores on the other hand are more difficult to program. The Nucleus methodology enables a tool based mapping, thus, resolving this issue. In this work, the mapping of a MIMO receiver onto different platforms is demonstrated and performance data (implementation efficiency, throughput) is compared. Our studies cover both receivers with iterative and with non-iterative MIMO detection. The investigated platforms include a commercial TI-DSP-based SDR platform, the P2012 platform by STMicroelectronics and our experimental research processor core, featuring vector instructions with runtime adaptive floating-point format. The research core is also used to demonstrate how the Nucleus methodology in conjunction with adaptive number formats can be used to further reduce energy consumption.