In these systems the developers must de-sign several experiments for workload characterization observingthe architectural implications when using different combinationsof computational resources such as number of GPU, number of cores for processing, number of cores for administration of GPU,number of MPI processes and thread afﬁnity policy. It shouldalso engage factors as the clock frequency and memory usageas well select the combination of computational resources thatincreases the performance and minimizes the power consumption.This research proposes an integrated energy-aware scheme calledefﬁciently energetic acceleration (EEA) for large-scale scientiﬁcapplications running on heterogeneous architectures. This papershows the use of a monitoring tool with two components calledenerGyPU and enerGyPhi to recording EEA control factors inruntime on two environments: one cluster with multicore andaccelerator nodes (2-CPU/8-GPU) and one server with multiplecores and one coprocessor (2-CPU/1-MIC). These monitors allowto analyze multiple testing results under different parametercombinations to observe the EEA control factors that determinethe energy efﬁciency.
—Energy efﬁciency, Energy-aware EEA schemeenerGyPU, enerGyPhi, Power capping technique, Performanceevaluation.