GPU-Based Multi-start Local Search Algorithms

Abstract

Abstract. In practice, combinatorial optimization problems are complex and computationally time-intensive. Local search algorithms are powerful heuristics which allow to significantly reduce the computation time cost of the solution exploration space. In these algorithms, the multistart model may improve the quality and the robustness of the obtained solutions. However, solving large size and time-intensive optimization problems with this model requires a large amount of computational resources. GPU computing is recently revealed as a powerful way to harness these resources. In this paper, the focus is on the multi-start model for local search algorithms on GPU. We address its re-design, implementation and associated issues related to the GPU execution context. The preliminary results demonstrate the effectiveness of the proposed approaches and their capabilities to exploit the GPU architecture.

GPU computing in discrete optimization. Part II: Survey focused on routing problems

“…To the authors' best knowledge, this is the first report of a GPU implementation of pure LS. Further research is discussed in two follow-up papers [62,61]. The authors apply LS to different instances of well-known DOPs such as the Quadratic Assignment Problem and the TSP.…”

“…In the GPU literature we have found two main approaches. Either, a GPU based parallel neighborhood evaluation of the different local searches is performed sequentially (Luong et al [61]), or, the local searches run in parallel on the GPU (O'Neil et al [70,99], Luong et al [61]). …”

“…Another idea, used e.g. in (Zhu et al [99], Luong et al [61]), is to let the LS in each thread run only for a given number of iterations and then perform restart or load balancing before continuing. Due to the many variables involved, it is impossible to state generally that the sequential parallel neighborhood evaluation is better or worse than the one thread per local search approach.…”

“…Luong et al have published several follow-up papers to [57,62,61]. In [58], they discuss how to implement LS algorithms with large-size neighborhoods on the GPU 2 , with focus on memory issues.…”

3D Skeleton Extraction Method using Potential Field on OpenCL

“…This International Conference on Computer Science and Service System (CSSS 2014) computing model, making full use of GPU's powerful processing ability and high storage bandwidth to make up for the shortage of CPU performance, has great advantage in achieving high performance by exploring potential computing ability of computer and controlling the cost. Researchers have done some studies on GPU such as Local Search [11], Large Graph algorithm [12], simulate the Mush-room Cloud [13], bringing out perfect improvements This paper centers on two important points, one is making full use of GPU's ability in high-dense data parallel computing. The other one is digging out its potential market and discussing how to map applications from CPU to GPU.…”