Online task scheduler in 3D-MCPs with TADVA

Abstract

Hotspots occur frequently in three-dimensional (3D) multi-core processors (3D-MCPs), and they may adversely impact both the reliability and lifetime of a system. The authors present dynamic-voltage-assignment (DVA) strategies that reduce hotspots in and optimise the performance of 3D-MCPs by pre-emptively selecting voltages among low-power and high-performance operating modes. The proposed DVA strategies can be employed in online, thermally constrained task schedulers. Three DVA strategies, random DVA, thermal-aware DVA (TADVA) and TADVA2.0, are proposed to reduce the temperature increase in 3D-MCPs by pre-emptively and dynamically estimating the optimum VAs for all cores in the processor during runtime once the chip begins operating. In particular, TADVA2.0 uses the temperature-variation rates of the cores and takes into account two important thermal behaviours of 3D-MCPs that can effectively limit the temperature increase in 3D-MCPs. Experimental results indicate that, when compared with two previous online thermally constrained task schedulers, the proposed task scheduler with their novel DVA strategy can reduce hotspot occurrences by approximate to 60% and improve throughput by approximate to 8%. These results indicate that the proposed TADVA2.0 strategy is an effective technique for suppressing hotspot occurrences and optimising throughput for 3D-MCPs subject to thermal constraints.