The present paper examines the actual problem of using graphics processing units (GPUs) in computing processes that are traditionally performed on central processing units (CPUs). With the development of technology and the advent of specialized architectures and libraries, GPUs have become indispensable in areas requiring intensive computing. The article examines in detail the advantages of using GPUs compared to traditional CPUs, justifying this with their ability to process in parallel and high throughput, which makes them an ideal tool for working with large amounts of data.are accidents caused by violations of rules and regulations at work sites, among them cases related to non-compliance with the rules of wearing protective helmets. The article examines methods and algorithms for recognizing protective helmets and helmets, and assesses their effectiveness.

Keywords: graphics processors, GPU, CUDA, OpenCL, cuBLAS, CL Blast, rocBLAS, parallel data processing, mathematical calculations, code optimization, memory management, machine learning, scientific research

The study presents an extensive analysis of methods for low-level optimization of the matrix multiplication algorithm for computing systems with shared memory. Based on a comparison of various approaches, including block optimization, parallel execution with OpenMP, vectorization with AVX and the use of the Intel MKL library, significant improvements in the performance of the resulting software implementations are revealed. In particular, block optimization reduces the number of cache misses, parallelism effectively uses multicore, and vectorization and Intel MKL demonstrate maximum acceleration due to more efficient software optimizations. The obtained results emphasize the importance of careful selection of optimization methods and their compliance with the architecture of the computing system in order to achieve the required performance parameters of the designed software.

Keywords: low-level optimization, block optimization, parallel execution, OpenMP, vectorization, AVX, Intel MKL, performance, benchmarking, matrix multiplication