HW-SW integration for energy-efficient/variability-aware computing
Gasser Ayad, Andrea Acquaviva, Enrico Macii, Brahim Sahbi, Romain Lemaire · Design, Automation, and Test in Europe · 2013
Recent trends in embedded system architectures brought a rapid shift towards multicore, heterogeneous and reconfigurable platforms. This imposes a large effort for programmers to develop their applications to efficiently exploit the underlying architecture. In addition, process variability issues lead to performance and power uncertainties, impacting expected quality of service and energy efficiency of the running software. In particular, variability may lead to sub-optimal runtime task allocation.In this paper we present a holistic approach to tackle these issues exploiting high level HW/SW modeling to customize the runtime library. The customization introduces variability awareness in task allocation decisions, with the final purpose of optimizing a given objective: Execution time, power consumption, or overall energy consumption.We present a complete walkthrough, from top-level modeling down to variability-aware execution using a parallelized computational kernel running on a next generation, NoC based, heterogeneous multicore simulation platform.