| Popis: |
Simulation experiments upon sound propagation in various media have generally a high algorithmic complexity and are unfeasible without strict limitations upon levels of approximation of the acoustical field and its environment. Specifically, the article shows that a problem of sound simulation that considers diffraction of sound can have exponential complexity over a number of diffraction points. Hence, there is a need to provide the limitations on accuracy as well as to provide high-performance implementation of such simulation. Modern high-performance implementations are all based on parallelism of computations of some form. A kind of parallelism being used is defined by a specific simulation problem and defines an approach to implement the problem algorithmically. Therefore, a general practice is to provide both hardware and software implementation of the simulation for a specific problem, and to reimplement it for any other problem which may arise. Hence, high expenses on such simulations. The paper presents an approach to mitigate these issues by building a flexible high-performance system implementing simulations of sound propagations in various media based on different mathematical models built in various knowledge domains. |
