Impact of parallel code optimization on computer power consumption

TL;DR

This paper evaluates how different code optimization strategies and computer architectures affect power consumption in computing systems, highlighting that optimized code often peaks in power use and exploring energy profiling for parallel programs.

Contribution

It provides an empirical analysis of the impact of optimization levels and architectures on power consumption, including the potential for creating energy profiles for parallel programs.

Findings

Optimized programs tend to have peak power consumption.

Changing architecture influences power consumption patterns.

Energy profiling for parallel programs is feasible.

Abstract

The increase in performance and power of computing systems requires the wider use of program optimizations. The goal of performing optimizations is not only to reduce program runtime, but also to reduce other computer resources including power consumption. The goal of the study was to evaluate the impact of different optimization levels and various optimization strategies on power consumption. In a series of experiments, it was established that the average power consumption tends to peak for the programs with optimized source code. The articles also describes the impact of changing computer architecture on power consumption graphs. The relationships between the average and median values of power consumption by example programs are considered. The possibility of creating program energy consumption profile for a parallel program is shown.