Parallel algorithms of discrete optimization in manufacturing
Parallel algorithms are a natural continuation of programming methods used for decades. After periods devoted to structural and object-oriented programming, the introduction of easily accessible parallel architectures in the form of multicore processors and GPU computing processors requires the researchers to design multithreaded algorithms to take full advantage of the hardware potential. The trend is clearly visible in the field of discrete optimization: parallel algorithms, especially metaheuristics, implemented on multiprocessor computing architectures and applied to NP-hard optimization problems allow its users not only for a significant acceleration of computations, but also constitute themselves a new class of algorithms, often obtaining super-linear speedups. Large amounts of available processors, especially in low-cost GPU architectures, allow one to solve problems of jobs scheduling of much larger sizes than the ones presented recently. Unfortunately, effective adaptation of known algorithms to be run in a concurrent environment is a difficult process, requiring the use of specific properties of definite problems and algorithms.
Autor: Wojciech Bożejko
Znajdź bibliotekę blisko siebie, i uzyskaj dostęp do ebooka w systemie IBUK Libra