It's Not Easy Being Green: On the Energy Efficiency of Programming Languages

TL;DR

This paper tackles whether programming language choice influences energy consumption by building a detailed causal model that separates language from its implementations, application-level parallelism, and hardware factors. It introduces an enhanced measurement methodology (including performance counters and corrected RAPL-based readings) and expands benchmarks to disentangle execution time from power usage. The authors show that, once core usage and memory activity are controlled, energy consumption tracks execution time rather than language or implementation, thus eliminating prior claimed energy advantages of certain languages. They also explain previously observed anomalies via four causal paths (concurrency, libraries, cross-language boundaries, and JIT warmup) and demonstrate that aggressive parallelism is often energy-efficient. The result is a practical guideline: focus on reducing execution time and using fast implementations or optimized libraries rather than selecting languages solely for energy savings, with data and tooling openly available for reproducibility.

Abstract

Does the choice of programming language affect energy consumption? Previous highly visible studies have established associations between certain programming languages and energy consumption. A causal misinterpretation of this work has led academics and industry leaders to use or support certain languages based on their claimed impact on energy consumption. This paper tackles this causal question directly: it develops a detailed causal model capturing the complex relationship between programming language choice and energy consumption. This model identifies and incorporates several critical but previously overlooked factors that affect energy usage. These factors, such as distinguishing programming languages from their implementations, the impact of the application implementations themselves, the number of active cores, and memory activity, can significantly skew energy consumption measurements if not accounted for. We show -- via empirical experiments, improved methodology, and careful examination of anomalies -- that when these factors are controlled for, notable discrepancies in prior work vanish. Our analysis suggests that the choice of programming language implementation has no significant impact on energy consumption beyond execution time.

Figures (9)

• Figure 1: The causal model (represented as a causal diagram) of the relationship between programming language and energy consumption presented in this paper (§\ref{['section:model']}). Gray arrows represent comparatively weaker relationships, as Section \ref{['section:model']} details: programming language implementation has only a minor effect on parallelism, and memory activity plays a minimal, less controllable role in energy consumption compared to CPU activity.
• Figure 2: Industry keynotes advocating for Rust over other languages due to reduced energy consumption, based on the results of Pereira et al. 10.1145/3125374.312538210.1145/3136014.3136031DBLP:journals/scp/PereiraCRRCFS21 (§\ref{['section:prior-overview']}).
• Figure 3: The simple model used as a starting point in this paper (§\ref{['section:model-starting']}). This simple model captures the relationship implied in Pereira et al., namely that the choice of programming language has a direct impact on total energy consumption.
• Figure 4: Programming languages may have multiple implementations (§\ref{['section:model-pl-implementation']}), and implementation decisions such as garbage collection or using just-in-time compilation have a more direct impact on performance. We are in fact comparing implementations of programming languages, not the languages themselves.
• Figure 5: Energy consumption is the product of power and time (§\ref{['section:model-energy-power-time']}). Splitting into these two components allows separate consideration of each factor.