Correlation-based temperature sensor allocation for thermal-aware mesh-based MPSoCs

As the increasing demands of high-performance application, the large power density of Multiprocessor Systems-on-Chip (MPSoC) results in serious thermal problems. To solve the thermal problems in multicore systems, in practical way, the designers usually place limited numbers of thermal sensor on the chip, which provide the temperature information for temperature regulation by using Dynamic Thermal Management (DTM). However, searching the proper locations for thermal sensor placement is an NP-hard problem. Besides, the time-varying temperature distribution of the multicore system further worsens the difficulty in finding the optimal thermal sensor allocations. In this paper, we involve the correlation-based statistical analysis to propose a thermal sensor allocation algorithm. Compared with the conventional thermal sensor allocation algorithms, the proposed method can reduce 72% average error and 23% maximum error for full-chip temperature estimation.