THE USE OF MACHINE LEARNING METHODS FOR ADAPTING SOFTWARE TO DIFFERENT COMPUTING PLATFORMS

Abstract

Current challenges in software development arise from the diversity of hardware architectures, and adaptive compilers offer solutions to address them. This paper examines the creation of adaptive compilers that utilize machine learning methods for automatic optimization and adaptation of software across various computing platforms. Examples of adaptive compiler applications in mobile and cloud computing are presented, where they demonstrate significant improvements in software performance and efficiency. The purpose of this work is to create adaptive compilers through the integration of machine learning methods at all stages of compilation – from input code analysis to machine code generation. Scientific Novelty lies in the integration of machine learning methods into the compilation process, enabling adaptive compilers to automatically fine-tune code optimization for different computing platforms. This solution provides dynamic compilation adjustments, enhancing program performance and reducing the need for manual tuning for new architectures. Conclusions. The conducted analysis highlights the effectiveness of adaptive compilers that, through machine learning, enable automatic code optimization, thereby enhancing software performance and efficiency across different platforms. The proposed approach simplifies software adaptation, reduces development costs, and allows a focus on product functionality.