Promoting Green Coding in VS Code with GCPM: A Power Model for Heterogeneous Processors

Software applications have significantly improved our lives, making computer programming increasingly common and necessary in many fields. However, the energy consumption of software is often overlooked. This gap exists due to the lack of accurate, convenient, and host-independent power consumption measurement technologies for software. We address this issue by developing GreenCode-PowerMeter (GCPM), a model-based power estimation technology for computer programs. GCPM features a decision tree followed by a third degree multi-variable polynomial regression model to provide precise power consumption measurements within software development environment. Based on real power consumption data from 39 popular processors, provided by the SPEC benchmark across various CPU utilization levels, GCPM depicts a mean absolute percentile error of only 0.075%, outperforms 14 existing power models, including the TDP based model and closely aligns to Running Average Power Limit (RAPL). By integrating GCPM into the VS Code IDE as GreenCode-PowerMeasurement extension, we enable measuring software power consumption directly within the development environment. With 14 million active users of VS Code and GCPM’s high accuracy exceeding 90% compared to RAPL, requiring no sudo permission; this integration has enabled precise power analysis - simplifying the adaptation for developers towards energy optimized software development.