Thermal balancing policy for streaming computing on multiprocessor architectures

As feature sizes decrease, power dissipation and heat generation density exponentially increase. Thus, temperature gradients in Mul-tiprocessor Systems on Chip (MPSoCs) can seriously impact sys-tem performance and reliability. Thermal balancing policies based on task migration have been proposed to modulate power distri-bution between processing cores to achieve temperature flattening. However, in the context of MPSoC for multimedia streaming com-puting, where timeliness is critical, the impact of migration on qual-ity of service must be carefully analyzed. In this paper we present the design and implementation of a lightweight thermal balancing policy that reduces on-chip temperature gradients via task migra-tion. This policy exploits run-time temperature and load informa-tion to balance the chip temperature. Moreover, we assess the ef-fectiveness of the proposed policy for streaming computing archi-tectures using a cycle-accurate thermal-aware emulation infrastruc-ture. Our results using a real-life software defined radio multitask benchmark show that our policy achieves thermal balancing while keeping migration costs bounded. 1