A generic energy optimization framework for heterogeneous platforms using scaling models

Amdahl's law in the context of heterogeneous many‐core systems – a survey

“…The degree of core heterogeneity covered by this model is limited, in the sense that there are only two types of cores and the scheduling favours the faster type. Gupta et al [30,103] investigated the effects of the performance of individual cores on speedup, in a heterogeneous parallel processing context. These models focus on the performance scaling of cores, via such techniques as DVFS and performance optimisation.…”

“…Secondly, the parallel part is assumed to consist of multiple phases that must be executed sequentially, phase by phase, with each phase being a parallelisable set of tasks. The model for speedup by scaling core speeds can be found in [30,Equation (5)]. In this multi-fraction model following Cassidy and Andreou [37], the execution time of each parallel phase is calculated according to a similar method to that shown in (15), which is generally correct for all schemes of a task to core allocation.…”

“…And this needs to be represented properly in any reasonable model. Similar to Gupta et al [30], Rafiev et al [38] generalised the representation of scheduling policy beyond 'favouring faster cores'. Intuitively, this may be done by enumerating the cores in the order of their scheduling priority, if these stay constant during the execution of a workload.…”

“…It is also valid for all possible workload structures, including the number of steps and the degree of parallelism of each step, and the mapping of these tasks to cores fitting arbitrary system core architectures. For instance, the speedup obtained by scaling the sequential performances of cores investigated in [30,103] is fully covered by this model. Summary: Cassidy and Andreou, questioning the original Amdahl binary division of workloads into sequential and parallel parts, extend Amdahl's law into the multi-fraction model [37].…”

A survey on energy estimation and power modeling schemes for smartphone applications

“…Coulomb counting estimates SOC based on an accumulative current drop by directly accessing the current sensor within smartphone devices. The accuracy of coulomb counting method is highly affected by numerous external and internal factors such as battery aging, temperature, and charging/discharging rate . Alternatively, terminal voltage method estimates SOC based on voltage drop because of the internal impedance of battery during its discharging.…”