A VNS with Parallel Evaluation of Solutions for the Inverse Lighting Problem
Lighting design is a key issue in architectural design. The Inverse Lighting Problem (ILP) is an optimization problem that arises in lighting design and consist in finding the best configuration of lights that meets a set of goals that designers would like to achieve. In this paper, we present three different VNS that evaluate several solutions in parallel, improving the performance of a traditional VNS that has already been proposed for solving the ILP. These methods exploit the block matrix multiplication algorithms in order to increase the computational intensity of the algorithm and are specially well suited for parallel computation in GPUs architectures. The experimental analysis performed in two CPU/GPU hardware platforms for two scenarios with different complexity shows that the proposed methods provide fast results and are able to allow the interactive lighting design.
KeywordsInverse Lighting Problems Graphics processing unit Variable neighborhood search CUDA GPGPU
Authors acknowledge partial support from PEDECIBA – Uruguay and project ANII FSE_1_2014_1_102344.
- 3.Fernández, E.: Low-rank radiosity. In: Proceedings of the IV Iberoamerican Symposium in Computer Graphics, pp. 55–62 (2009)Google Scholar
- 5.Kirk, D.B., Wen-mei, W.H.: Programming Massively Parallel Processors: A Hands-on Approach. Newnes (2012)Google Scholar
- 8.Benner, P., Ezzatti, P., Quintana-Ortí, E., Remón, A.: On the impact of optimization on the time-power-energy balance of dense linear algebra factorizations. In: Aversa, R., Kołodziej, J., Zhang, J., Amato, F., Fortino, G. (eds.) ICA3PP 2013. LNCS, vol. 8286, pp. 3–10. Springer, Cham (2013). doi: 10.1007/978-3-319-03889-6_1CrossRefGoogle Scholar
- 15.Luong, T.V.: Parallel metaheuristics on GPU. Ph.D. thesis, INRIA Lille (2011)Google Scholar