The paper considers software performance optimization technologies. The requirements for program performance, especially on mobile platforms, necessitate the use of various optimization methods, including high-level (algorithmic) and low-level (hardware-oriented) optimization. The paper considers methods for increasing program performance, taking into account the capabilities of modern computing platforms, including central processing units (CPUs) with multi-core architecture, random access memory (RAM), and cache memory systems. Attention is paid to parallelism at the instruction level (ILP), data (SIMD), and threads (multithreading). Tools for applying parallelism using compilers are considered. Several examples are given to illustrate the features of accelerating the implemented algorithms on various platforms. The x86–64 (Intel Core i9), ARM (Apple M1 Pro), and VLIW (Elbrus 2C3 and 8B) platforms are considered. The influence of the processor architecture and memory organization, including cache memory, on program performance is shown. The proposed approaches to optimizing program performance are applicable to the implementation of image processing and recognition algorithms, as well as to computational algorithms that perform cyclic information processing.
O.A. Slavin (2025) studied this question.