Реалізація паралельного алгоритму сортування купою за допомогою технології C++ AMP

Abstract

The paper explores the development of a parallel heap sort algorithm using C++ AMP technology, enabling computations to be executed on a graphics processing unit (GPU). Classical sorting methods are analyzed, their limitations in parallel environments are discussed, and Heap Sort is justified as a suitable method due to its binary heap structure and the possibility of partial parallelization.A software module architecture was designed, including GPUbased heap construction and the parallel execution of the heapify procedure. The implementation uses array_view and parallel_for_each constructs of C++ AMP. Performance testing was conducted on datasets of different sizes and compared with a sequential CPU implementation. The results show that GPU acceleration improves execution time on large datasets while maintaining sorting correctness.