Massively parallel simulations for disordered systems

Abstract Simulations of systems with quenched disorder are extremely demanding, suffering from the combined effect of slow relaxation and the need of performing the disorder average. As a consequence, new algorithms and improved implementations in combination with alternative and even purpose-built hardware are often instrumental for conducting meaningful studies of such systems. The ensuing demands regarding hardware availability and code complexity are substantial and sometimes prohibitive. We demonstrate how with a moderate coding effort leaving the overall structure of the simulation code unaltered as compared to a CPU implementation, very significant speed-ups can be achieved from a parallel code on GPU by mainly exploiting the trivial parallelism of the disorder samples and the near-trivial parallelism of the parallel tempering replicas. A combination of this massively parallel implementation with a careful choice of the temperature protocol for parallel tempering as well as efficient cluster updates allows us to equilibrate comparatively large systems with moderate computational resources. Graphical abstract

[1]  V. Martin-Mayor,et al.  Universality in the three-dimensional random-field Ising model. , 2013, Physical review letters.

[2]  H. Katzgraber,et al.  Population annealing: Theory and application in spin glasses. , 2015, Physical review. E, Statistical, nonlinear, and soft matter physics.

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

[4]  J. Machta,et al.  Population annealing with weighted averages: a Monte Carlo method for rough free-energy landscapes. , 2010, Physical review. E, Statistical, nonlinear, and soft matter physics.

[5]  Giorgio Parisi,et al.  Random Magnetic Fields, Supersymmetry, and Negative Dimensions , 1979 .

[6]  Lev N. Shchur,et al.  The Cluster Processor:. New Results , 1999 .

[7]  K. Hukushima,et al.  Exchange Monte Carlo Method and Application to Spin Glass Simulations , 1995, cond-mat/9512035.

[8]  K Hukushima.  Domain-wall free energy of spin-glass models: numerical method and boundary conditions. , 1999, Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics.

[9]  Martin Weigel,et al.  Performance potential for simulating spin models on GPU , 2010, J. Comput. Phys..

[10]  Giorgio Parisi,et al.  Order parameter for spin-glasses , 1983 .

[11]  Wolff,et al.  Collective Monte Carlo updating for spin systems. , 1989, Physical review letters.

[12]  Liang.  Application of cluster algorithms to spin glasses. , 1992, Physical review letters.

[13]  K. Binder,et al.  A Guide to Monte Carlo Simulations in Statistical Physics , 2000 .

[14]  C. Predescu,et al.  The incomplete beta function law for parallel tempering sampling of classical canonical systems. , 2003, The Journal of chemical physics.

[15]  Alexander K. Hartmann,et al.  Random number generators for massively parallel simulations on GPU , 2012, The European Physical Journal Special Topics.

[16]  Andrew J. Ochoa,et al.  Efficient Cluster Algorithm for Spin Glasses in Any Space Dimension. , 2015, Physical review letters.

[17]  Wenlong Wang,et al.  Comparing Monte Carlo methods for finding ground states of Ising spin glasses: Population annealing, simulated annealing, and parallel tempering. , 2014, Physical review. E, Statistical, nonlinear, and soft matter physics.

[18]  Wolfhard Janke,et al.  MULTIOVERLAP SIMULATIONS OF THE 3D EDWARDS-ANDERSON ISING SPIN GLASS , 1998 .

[19]  S. F. Schifano,et al.  Nature of the spin-glass phase at experimental length scales , 2010, 1003.2569.

[20]  Overlap distribution of the three-dimensional Ising model. , 2002, Physical review. E, Statistical, nonlinear, and soft matter physics.

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

[22]  Wolfhard Janke,et al.  Make Life Simple: Unleash the Full Power of the Parallel Tempering Algorithm , 2008 .

[23]  Martin Weigel,et al.  Estimating the density of states of frustrated spin systems , 2018, New Journal of Physics.

[24]  J Machta,et al.  Strengths and weaknesses of parallel tempering. , 2009, Physical review. E, Statistical, nonlinear, and soft matter physics.

[25]  Helmut G. Katzgraber,et al.  Universality in three-dimensional Ising spin glasses: A Monte Carlo study , 2006, cond-mat/0602212.

[26]  Raffaele Tripiccione,et al.  Thermodynamic glass transition in a spin glass without time-reversal symmetry , 2012, Proceedings of the National Academy of Sciences.

[27]  K. Binder,et al.  Spin glasses: Experimental facts, theoretical concepts, and open questions , 1986 .

[28]  Johannes Zierenberg,et al.  Massively parallel multicanonical simulations , 2017, Comput. Phys. Commun..

[29]  De Giusti,et al.  Structured Parallel Programming: patterns for efficient computation , 2015 .

[30]  Helmut G. Katzgraber,et al.  Evidence against a mean-field description of short-range spin glasses revealed through thermal boundary conditions , 2014, 1408.0438.

[31]  J. Houdayer,et al.  A cluster Monte Carlo algorithm for 2-dimensional spin glasses , 2001 .

[32]  A. Young,et al.  de Almeida–Thouless line studied using one-dimensional power-law diluted Heisenberg spin glasses , 2011, 1105.1352.

[33]  Jon Machta,et al.  GRAPHICAL REPRESENTATIONS AND CLUSTER ALGORITHMS FOR CRITICAL POINTS WITH FIELDS , 1998 .

[34]  T. Yavors’kii,et al.  Optimized GPU simulation of continuous-spin glass models , 2012, The European Physical Journal Special Topics.

[35]  Young,et al.  Numerical studies of Ising spin glasses in two, three, and four dimensions. , 1988, Physical review. B, Condensed matter.

[36]  N. Sourlas,et al.  Phase Transitions in Disordered Systems: The Example of the Random-Field Ising Model in Four Dimensions. , 2016, Physical review letters.

[37]  Matthew Scarpino.  OpenCL in Action: How to Accelerate Graphics and Computations , 2011 .

[38]  Massimo Bernaschi,et al.  Benchmarking GPU and CPU codes for Heisenberg spin glass over-relaxation , 2011, Comput. Phys. Commun..

[39]  M. Weigel.  Connected-component identification and cluster update on graphics processing units. , 2011, Physical review. E, Statistical, nonlinear, and soft matter physics.

[40]  C. Newman,et al.  The Percolation Signature of the Spin Glass Transition , 2007, 0707.0073.

[41]  Martin Weigel,et al.  Monte Carlo methods for massively parallel computers , 2017, 1709.04394.

[42]  J. Ramanujam,et al.  Parallel tempering simulation of the three-dimensional Edwards-Anderson model with compact asynchronous multispin coding on GPU , 2013, Comput. Phys. Commun..

[43]  Massimo Bernaschi,et al.  Highly optimized simulations on single- and multi-GPU systems of the 3D Ising spin glass model , 2014, Comput. Phys. Commun..

[44]  Wolfhard Janke,et al.  GPU accelerated population annealing algorithm , 2017, Comput. Phys. Commun..

[45]  田浦 健次朗.  The International Conference for High Performance Computing,Networking, Storage and Analysis (SC16) 参加報告 , 2017 .

[46]  Ignacio Rozada,et al.  Effects of setting temperatures in the parallel tempering Monte Carlo algorithm. , 2019, Physical review. E.

[47]  Paul Dupuis,et al.  On the Infinite Swapping Limit for Parallel Tempering , 2011, Multiscale Model. Simul..

[48]  Michael Bachmann,et al.  Massively parallelized replica-exchange simulations of polymers on GPUs , 2011, Comput. Phys. Commun..

[49]  M. Baity-Jesi,et al.  Phase transition in three-dimensional Heisenberg spin glasses with strong random anisotropies through a multi-GPU parallelization , 2013, 1309.1599.

[50]  Denis Navarro,et al.  Janus: An FPGA-Based System for High-Performance Scientific Computing , 2007, Computing in Science & Engineering.

[51]  Helmut G. Katzgraber,et al.  Chaos in spin glasses revealed through thermal boundary conditions , 2015, 1505.06222.

[52]  Jonathan Machta,et al.  GRAPHICAL REPRESENTATIONS AND CLUSTER ALGORITHMS II , 1998 .

[53]  T. Nattermann.  Theory of the Random Field Ising Model , 1997 .

[54]  J. Vondrák,et al.  New algorithm for the Ising problem: partition function for finite lattice graphs. , 2000, Physical review letters.

[55]  Arch D. Robison,et al.  Chapter 3 – Patterns , 2012 .

[56]  W. Janke,et al.  Approximate ground states of the random-field Potts model from graph cuts. , 2018, Physical review. E.

[57]  G. G. Stokes.  "J." , 1890, The New Yale Book of Quotations.

[58]  Wang,et al.  Replica Monte Carlo simulation of spin glasses. , 1986, Physical review letters.

[59]  G. Parisi,et al.  Aging Rate of Spin Glasses from Simulations Matches Experiments. , 2018, Physical review letters.

[60]  S. Edwards,et al.  Theory of spin glasses , 1975 .

[61]  Scott Lathrop,et al.  Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis , 2011, International Conference on High Performance Computing.

[62]  G. Parisi,et al.  The Mpemba effect in spin glasses is a persistent memory effect , 2018, Proceedings of the National Academy of Sciences.

[63]  Matthias Troyer,et al.  Feedback-optimized parallel tempering Monte Carlo , 2006, cond-mat/0602085.

[64]  A. Sokal,et al.  Generalization of the Fortuin-Kasteleyn-Swendsen-Wang representation and Monte Carlo algorithm. , 1988, Physical review. D, Particles and fields.

[65]  A. Young.  Direct determination of the probability distribution for the spin-glass order parameter , 1983 .

[66]  Koji Hukushima,et al.  Population Annealing and Its Application to a Spin Glass , 2003 .

[67]  Denis Navarro,et al.  Janus II: A new generation application-driven computer for spin-system simulations , 2013, Comput. Phys. Commun..

[68]  A. Young,et al.  Spin glasses and random fields , 1997 .