Thermal-aware task mapping for reconfigurable channel decoding

Abstract

Low-Density Parity-Check (LDPC) and Turbo codes are the most popular channel coding for wire-line and wireless communication systems. The reason is that their error-correcting abilities are much closer to the Shannon limit. In real design, many channel decoders are designed with the reconfigurable architectures to support multiple communication systems. In the reconfigurable architectures, only parts of the function units works in runtime, and the task mapping become very important. The unfair task mapping may result in the unbalanced thermal distribution and increase the hotspots in the channel decoder chip. In this work, we propose three thermal-aware task mapping schemes for a reconfigurable radix-4 soft-input soft-output (SISO) kernel of channel decoding. By the proposed thermal-aware task mapping, the peak temperature can reduce up to 4.8°C in our experiments.