Combining multi-core and GPU computing for solving combinatorial optimization problems

Chakroun, Imen; Melab, Nouredine; Mezmaz, Mohand; Tuyttens, Daniel

doi:10.1016/j.jpdc.2013.07.023

Cited by 39 publications

(53 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [13], a Master-Slave approach is adopted and only an experimental scale of 2 GPUs is reported, which is clearly not sufficient to express the power of modern compute platforms (sub-optimal speed-up are actually reported). In [17], the authors experienced multi-core pool-based B&B with parallel bounding inside the GPUs devices. Only shared memory threads are studied and small parallel scales (up to 6 threads) are considered there-in.…”

Section: Parallel Bandb With Gpusmentioning

confidence: 99%

“…Only shared memory threads are studied and small parallel scales (up to 6 threads) are considered there-in. In [14,15], the master-slave approach of [42] is combined with a GPU-guided implementations similar to [17] in an attempt to tackle large scale environments. The focus there-in is rather on feasibility and implementations issues; but unfortunately, scalability and performance optimality were not addressed in a comprehensive manner.…”

Section: Parallel Bandb With Gpusmentioning

confidence: 99%

See 1 more Smart Citation

Parallel Branch-and-Bound in multi-core multi-CPU multi-GPU heterogeneous environments

Derbel

2016

Future Generation Computer Systems

View full text Add to dashboard Cite

We investigate the design of parallel B&B in large scale heterogeneous compute environments where processing units can be composed of a mixture of multiple shared memory cores, multiple distributed CPUs and multiple GPUs devices. We describe two approaches addressing the critical issue of how to map B&B workload with the different levels of parallelism exposed by the target compute platform. We also contribute a throughout large scale experimental study which allows us to derive a comprehensive and fair analysis of the proposed approaches under different system configurations using up to 16 GPUs and up to 512 distributed cores. Our results shed more light on the main challenges one has to face when tackling B&B algorithms while describing efficient techniques to address them. In particular, we are able to obtain linear speed-ups at moderate scales where adaptive load balancing among the heterogeneous compute resources is shown to have a significant impact on performance. At the largest scales, intranode parallelism and hybrid decentralized load balancing is shown to have a crucial importance in order to alleviate locking issues among shared memory threads and to scale the distributed resources while optimizing communication costs and minimizing idle times.

show abstract

Section: Parallel Bandb With Gpusmentioning

confidence: 99%

Section: Parallel Bandb With Gpusmentioning

confidence: 99%

Parallel Branch-and-Bound in multi-core multi-CPU multi-GPU heterogeneous environments

Derbel

2016

Future Generation Computer Systems

View full text Add to dashboard Cite

show abstract

“…Researchers have proven that revisiting classic algorithms and adding parallelism to them is a very effective strategy for solving COPs. Such is the work in [32] where the authors revisited the design and implementation of Branch-and-Bound algorithms for solving large COPs on GPU-enhanced multicore machines, and in [33] a high-performance GPU implementation of the 2-opt and 3-opt local search algorithm was presented. Other metaheuristics that were enhanced using GPU are Ant Colony Optimization [34,35] and Genetic Algorithms [36,37].…”

Section: Related Workmentioning

confidence: 99%

Reducing the Size of Combinatorial Optimization Problems Using the Operator Vaccine by Fuzzy Selector with Adaptive Heuristics

Montiel

Delgadillo²

2015

Mathematical Problems in Engineering

View full text Add to dashboard Cite

Nowadays, solving optimally combinatorial problems is an open problem. Determining the best arrangement of elements proves being a very complex task that becomes critical when the problem size increases. Researchers have proposed various algorithms for solving Combinatorial Optimization Problems (COPs) that take into account the scalability; however, issues are still presented with larger COPs concerning hardware limitations such as memory and CPU speed. It has been shown that the Reduce-OptimizeExpand (ROE) method can solve COPs faster with the same resources; in this methodology, the reduction step is the most important procedure since inappropriate reductions, applied to the problem, will produce suboptimal results on the subsequent stages. In this work, an algorithm to improve the reduction step is proposed. It is based on a fuzzy inference system to classify portions of the problem and remove them, allowing COPs solving algorithms to utilize better the hardware resources by dealing with smaller problem sizes, and the use of metadata and adaptive heuristics. The Travelling Salesman Problem has been used as a case of study; instances that range from 343 to 3056 cities were used to prove that the fuzzy logic approach produces a higher percentage of successful reductions.

show abstract

“…Many parallel models for metaheuristics have been proposed to solve MOCO problems efficiently, and they have been evaluated on a wide range of academic and real-world MOCO problems in different domains [7,39]. Furthermore, parallelization of exact combinatorial-optimization methods, such as Branch and Bound [28] and Dynamic Programming [4], has been studied and implemented in multi-core environments [9]; however, it has been rarely addressed in the context of multi-objective optimization [39]. The only work we are aware is from Dhaenens et al [13], who parallelized the exact solving of MOCO problems by geometrically splitting the search space into cubes and evaluated their algorithm on one case study; but their parallelization is not able to scale well up to only 10 processors.…”

Section: Related Workmentioning

confidence: 99%

Scaling exact multi-objective combinatorial optimization by parallelization

Guo

Zulkoski

Olaechea

et al. 2014

Proceedings of the 29th ACM/IEEE International Conference on Automated Software Engineering

View full text Add to dashboard Cite

Multi-Objective Combinatorial Optimization (MOCO) is fundamental to the development and optimization of software systems. We propose five novel parallel algorithms for solving MOCO problems exactly and efficiently. Our algorithms rely on off-the-shelf solvers to search for exact Pareto-optimal solutions, and they parallelize the search via collaborative communication, divide-and-conquer, or both. We demonstrate the feasibility and performance of our algorithms by experiments on three case studies of software-system designs. A key finding is that one algorithm, which we call FS-GIA, achieves substantial (even super-linear) speedups that scale well up to 64 cores. Furthermore, we analyze the performance bottlenecks and opportunities of our parallel algorithms, which facilitates further research on exact, parallel MOCO.

show abstract

Combining multi-core and GPU computing for solving combinatorial optimization problems

Cited by 39 publications

References 9 publications

Parallel Branch-and-Bound in multi-core multi-CPU multi-GPU heterogeneous environments

Parallel Branch-and-Bound in multi-core multi-CPU multi-GPU heterogeneous environments

Reducing the Size of Combinatorial Optimization Problems Using the Operator Vaccine by Fuzzy Selector with Adaptive Heuristics

Scaling exact multi-objective combinatorial optimization by parallelization

Contact Info

Product

Resources

About