Abstract
The paper offers a unified approach to use of accelerators of various types by the example of parallel global search algorithms. In the developed general scheme, the implementation elements depending on specific types of accelerators are encapsulated in a limited number of procedures, which allows to largely reduce expenses and time for software development. The offered scheme was tested on Intel Xeon Phi accelerators and general purpose GPUs. Speedup of the global optimization algorithm using accelerators compared to using CPU only is experimentally confirmed. Computational experiments are carried out using a set of a several hundred of multidimensional multiextremal problems.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
