Abstract:
The paper considers software performance optimization technologies. It examines algorithm characteristics (asymptotic complexity and computational complexity) and the performance of algorithms implemented as programs. Profiling technology is used to optimize program performance. The given description of modern computing architectures shows that a processor and RAM resources cannot be fully taken into account when analyzing using asymptotic and computational complexity. Several examples demonstrate the limited possibilities of optimization using asymptotic and computational complexity. For the given examples, it is shown that using information about the capabilities of the architecture on which the program is executed allows for performance optimization without using low-level commands. Computational experiments were conducted on x86-64 (Intel Core) and ARM (Apple M1) platforms. The paper proposes a methodology for optimizing the performance of developed programs.
Keywords:asymptotic complexity, computational complexity, central processor, random access memory, performance, computing platform.
Fulltext:
PDF file (205 kB)