Implementation and Performance of a GPU-Based Monte-Carlo Framework for Determining Design Ice Load

Modern Graphics Processing Units (GPUs) with massive number of threads and many-core architecture support both graphics and general purpose computing. NVIDIA's compute unified device architecture (CUDA) takes advantage of parallel computing and utilizes the tremendous power of GPUs. The present study demonstrates a high performance computing (HPC) framework for a Monte-Carlo simulation to determine design sea ice loads which is implemented in both GPU and CPU. Results show a speedup of up to 130 times for the 4 Tesla K80 GPUs over an optimized CPU OpenMP implementation and speedup of up to 8 times for the 4 Tesla K80 over a single Tesla K80 GPU implementation. The elapsed time of the different implementations has been reduced from about 2.5 hours to 0.7 seconds.

[1]  Cliburn Chan,et al.  Understanding GPU Programming for Statistical Computation: Studies in Massively Parallel Massive Mixtures , 2010, Journal of computational and graphical statistics : a joint publication of American Statistical Association, Institute of Mathematical Statistics, Interface Foundation of North America.

[2]  Wang,et al.  Nonuniversal critical dynamics in Monte Carlo simulations. , 1987, Physical review letters.

[3]  Matthias Müller-Hannemann,et al.  Algorithm Engineering: Bridging the Gap between Algorithm Theory and Practice [outcome of a Dagstuhl Seminar] , 2010, Algorithm Engineering.

[4]  Lukasz Machura,et al.  GPU accelerated Monte Carlo simulation of Brownian motors dynamics with CUDA , 2014, Comput. Phys. Commun..

[5]  Hyesoon Kim,et al.  Memory-level and Thread-level Parallelism Aware GPU Architecture Performance Analytical Model , 2011 .

[6]  Esteban Walter Gonzalez Clua,et al.  Performance Evaluation of Optimized Implementations of Finite Difference Method for Wave Propagation Problems on GPU Architecture , 2010, 2010 22nd International Symposium on Computer Architecture and High Performance Computing Workshops.

[7]  Shadi G. Alawneh,et al.  Fast Quadratic Discriminant Analysis Using GPGPU for Sea Ice Forecasting , 2015, 2015 IEEE 17th International Conference on High Performance Computing and Communications, 2015 IEEE 7th International Symposium on Cyberspace Safety and Security, and 2015 IEEE 12th International Conference on Embedded Software and Systems.

[8]  M. N. S. Yusoff,et al.  Performance of CUDA GPU in Monte Carlo simulation of light-skin diffuse reflectance spectra , 2012, 2012 IEEE-EMBS Conference on Biomedical Engineering and Sciences.

[9]  Floriano De Rango,et al.  Proceedings of the Summer Computer Simulation Conference , 2016 .

[10]  P. D. Coddington,et al.  Analysis of random number generators using Monte Carlo simulation , 1993, cond-mat/9309017.

[11]  Pradeep Dubey,et al.  Debunking the 100X GPU vs. CPU myth: an evaluation of throughput computing on CPU and GPU , 2010, ISCA.

[12]  Steve B. Jiang,et al.  Development of a GPU-based Monte Carlo dose calculation code for coupled electron–photon transport , 2009, Physics in medicine and biology.

[13]  N. Metropolis,et al.  Equation of State Calculations by Fast Computing Machines , 1953, Resonance.

[14]  Dirk P. Kroese,et al.  Why the Monte Carlo method is so important today , 2014 .

[15]  Christopher Robert Cullinan,et al.  Computing Performance Benchmarks among CPU, GPU, and FPGA , 2012 .

[16]  Sara Ayubian,et al.  GPU-based monte-carlo simulation for a sea ice load application , 2016, SummerSim.

[17]  P. Glaskowsky.  NVIDIA ’ s Fermi : The First Complete GPU Computing Architecture , 2009 .

[18]  Mark Goldsworthy,et al.  A GPU–CUDA based direct simulation Monte Carlo algorithm for real gas flows , 2014 .

[19]  Claude Daley,et al.  Hyper-Real-Time Ice Simulation and Modeling Using GPGPU , 2015, IEEE Transactions on Computers.

[20]  Samuel H. Fuller,et al.  The Future of Computing Performance: Game Over or Next Level? , 2014 .