ИСТИНА |
Войти в систему Регистрация |
|
Интеллектуальная Система Тематического Исследования НАукометрических данных |
||
Solving multidimensional ill-posed problems has attracted wide interests and found many practical applications. However, the most modern appli- cations require processing a large amount of datum that often very dicult to perform on personal computers. In this case usual dierent methods are applied for simplication of the problem statement but these simplications degrade the accuracy of the inverted parameters. It is supposed to solve calculating dicult applications by using parallel computation that gives us an advantage the time and the accuracy. The main practical problem investigated here is an eective solving of multidimensional Fredholm integral equation of the 1st kind by using paral- lel computation. In most cases the algorithms of solving Fredholm integral equation of the 1st kind have structures which allow us to divide large prob- lem into smaller ones which are then solved "in parallel". But there is Amdahl's law, which states that if P is the proportion of a program that can be made parallel (i.e. benet from parallelization), and (1 - P) is the proportion that cannot be parallelized (remains serial), then the maximum speedup that can be achieved by using N processors is S = \frac{1}{(1-P) + \frac{P}{N - 1}} . For example, if the sequential portion of a program is 10% of the runtime, we can get no more than a 10x speed-up, regardless of how many processors are added. This puts an upper limit on the usefulness of adding more parallel execution units. Several recent results will be presented on the study of this problem. States that it is possible to construct a parallel algorithm for solving of multidimensional Fredholm integral equation of the 1st kind, the percentage of serial calculations of which will tend to zero, that proves the eciency of multiprocessor systems for solving of this problem.