Large-Scale Compute-Intensive Constrained Optimization Problems: GPGPU-Based Approach

Abstract

The large-scale compute-intensive optimization problems exist in various engineering and scientific problems and applications. Such problems contain large number of design variables which needs to optimize to obtain optimum solutions. The large-scale compute-intensive problems solved using variety of nature-inspired techniques. The DEVIIC is one of such approach found in literature, developed using VIIC technique. This paper presents design and implementation of GPGPU-based DEVIIC algorithm to address large-scale compute-intensive benchmark optimization problems, based on master–slave strategy. The proposed approach is evaluated using 12 large-scale benchmark functions found in literature. The obtained results of proposed approach compared with results found in literature, implemented existing sequential DEVIIC algorithm and proposed GPGPU-based approach. The proposed approach gives comparatively better results than results found in Sayed et al. (Inf Sci 316:457–486, 2015) [1] for functions F1, F5 to F9, and F11, and fails to obtain better results for functions F2 to F4 and F10. As the proposed approach is to develop GPGPU-based algorithm, the speedup is computed. The proposed approach significantly reduces the execution time required to obtain the best solution. The proposed approach is 23 to 35 times faster than its sequential counterpart.