Optimizing LLVM IR: Transforming Multiplication to Addition for Enhanced Execution Efficiency

Intermediate Representation (IR) optimization is essential in enhancing software performance and ensuring efficient operation within the Low-Level Virtual Machine (LLVM) project. This optimization is achieved through the development of new passes that facilitate more effective code processing. In this study, we developed a new IR optimization pass specifically for the x86_64 architecture. The pass aims to reduce execution costs by transforming multiplication operations involving constant factors into equivalent addition operations. We assessed the impact of this optimization by evaluating the number of execution cycles across various scenarios. The results indicate a significant reduction in execution costs, highlighting the potential for enhanced performance in specific application domains. This study underscores the advantages of optimization techniques within the LLVM project for the software development process. Future work will involve applying these optimizations to other architectures, such as ARM and RISC-V, and evaluating their effectiveness using not only the number of execution cycles but also other diverse metrics. This approach will enable a more comprehensive analysis of performance gains across different architectures, supporting the development of optimized solutions.