Hybrid of genetic algorithm and local search to solve MAX-SAT problem using nVidia CUDA framework

Abstract

General Purpose computing over Graphical Processing Units (GPGPUs) is a huge shift of paradigm in parallel computing that promises a dramatic increase in performance. But GPGPUs also bring an unprecedented level of complexity in algorithmic design and software development. In this paper we describe the challenges and design choices involved in parallelizing a hybrid of Genetic Algorithm (GA) and Local Search (LS) to solve MAXimum SATisfiability (MAX-SAT) problem on a state-of-the-art nVidia Tesla GPU using nVidia Compute Unified Device Architecture (CUDA). MAX-SAT is a problem of practical importance and is often solved by employing metaheuristics based search methods like GAs and hybrid of GA with LS. Almost all the parallel GAs (pGAs) designed in the last two decades were designed for either clusters or MPPs. Unfortunately, very little research is done on the implementation of such algorithms over commodity graphics hardware. GAs in their simple form are not suitable for implementation over the Single Instruction Multiple Thread (SIMT) architecture of a GPU, same is the case with conventional LS algorithms. In this paper we explore different genetic operators that can be used for an efficient implementation of GAs over nVidia GPUs. We also design and introduce new techniques/operators for an efficient implementation of GAs and LS over such architectures. We use nVidia Tesla C1060 to perform several numerical tests and performance measurements and show that in the best case we obtain a speedup of 25x. We also discuss the effects of different optimization techniques on the overall execution time.

Automated framework for FPGA-based parallel genetic algorithms

“…The CPU code is well tuned with multi-threading techniques including Pthread and SIMD, and the code is compiled by Intel C compilier with highest optimisation level. For the maximum satisfiability problem (MAXSAT), we also compare our pGA system with a third-party GPU-based work on an nVidia Tesla C1060 [23]. We have compared our design with other FPGA-based systems qualitatively in section IV, but it is nearly impossible to compare quantitatively as they use different platforms and do not provide enough details of their execution time.…”

“…As shown in Table V, our systems have an average speedup of 24 times over the multi-core CPU. Moreover, pGA4 system gains 6.8 times speedup over the GPU-based system proposed in [23] on the nVidia Tesla C1060, when comparing the wall clock time taken to find the optimum solution.…”

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

“…They report good speedups over a CPU implementation with similar solution quality. GPU-based hybrids of GA and LS for Max-SAT was investigated in 2009 by Munawar et al in [68].…”

Parallelization Strategies for Hybrid Metaheuristics Using a Single GPU and Multi-core Resources

“…Therefore, only few research works have been investigated for local search algorithms [10][11][12]. The same goes on when dealing with hybrid metaheuristics on GPU, where there exists only few parallelization approaches [3][4][5].…”

“…To the best of our knowledge, most GPUaccelerated metaheuristics designed in the literature only exploit a single CPU core. This is typically the case for hybrid metaheuristics on GPU [3][4][5]. Thus, it might be valuable to fully utilize the other remaining CPU resources.…”