Implementation of Thread Local Storage Optimization Method Based on LoongArch

A thread-local storage optimization method for the LoongArch architecture is presented in this paper. The method utilizes LoongArch instruction characteristics and incorporates techniques such as code generation and link-time optimization to simplify and remove unnecessary instructions from programs, resulting in reduced memory usage and improved program execution speed. Our method was evaluated in multi-threaded programs, and the results indicate a 4.4% improvement in IPC (insn per cycle) efficiency and a 12.9% reduction in program execution time for application scenarios such as thread pools, network programming, and database connection pools. These findings demonstrate the effectiveness of our approach in significantly enhancing program performance.