On the impact of Codon compilation on energy consumption and performance of Python code

Abstract

Python programming language, known for its simplicity and rich library support, is widely used among software developers but faces performance and energy efficiency limitations. Codon compiler aims to address these issues through ahead-of-time compilation and multi-level optimizations. This research evaluates Codon's impact on performance and energy consumption by comparing it to interpreted Python and C++, and by analyzing the influence of input data size on its optimizations. The study followed a three-phase process: identifying problem classes relevant for Codon optimizations, selecting and validating experimental subjects, and conducting controlled experiments for analyzing Codon compilation impacts. Results showed that Codon can significantly improve Python's energy consumption, execution time, and memory usage. While C++ generally remained superior, Codon's performance sometimes matched or exceeded it. Codon's impact on CPU usage, unlike other metrics, varied by task and did not always show a relation to the size of input data highlighting its dependency on computational demands. This research demonstrates great potential for Codon and establishes a foundation for exploring its broader applicability across diverse domains to enhance the performance and energy efficiency of Python-based applications.