Charstar

Abstract

High-performance architectures are over-provisioned with resources to extract the maximum achievable performance out of applications. Two sources of avoidable power dissipation are the leakage power from underutilized resources, along with clock power from the clock hierarchy that feeds these resources. Most reconfiguration mechanisms either focus solely on power gating execution resources alone or in addition, simply turn off the immediate clock tree segment which supplied the clock to those resources. These proposals neither attempt to gate further up the clock hierarchy nor do they involve the clock hierarchy in influencing the reconfiguration decisions. The primary contribution of CHARSTAR is optimizing reconfiguration mechanisms to become clock hierarchy aware. Resource gating decisions are cognizant of the power consumed by each node in the clock hierarchy and additionally, entire branches of the clock tree are greedily shut down whenever possible. The CHARSTAR design is further optimized for balanced spatiotemporal reconfiguration and also enables efficient joint control of resource and frequency scaling. The proposal is implemented by leveraging the inherent advantages of spatial architectures, utilizing a control mechanism driven by a lightweight offline trained neural predictor. CHARSTAR, when deployed on the CRIB tiled microarchitecture, improves processor energy efficiency by 20-25%, with efficiency improvements of roughly 2x in comparison to a naive power gating mechanism. Alternatively, it improves performance by 10-20% under varying power and energy constraints.