Fast algorithm for generating random bit strings and multispin coding for directed percolation
arXiv:1808.05009 · doi:10.7566/JPSJ.88.024004
Abstract
We present efficient algorithms to generate a bit string in which each bit is set with arbitrary probability. By adopting a hybrid algorithm, i.e., a finite-bit density approximation with correction techniques, we achieve 3.8 times faster random bit generation than the simple algorithm for the 32-bit case and 6.8 times faster for the 64-bit case. Employing the developed algorithm, we apply the multispin coding technique to one-dimensional bond-directed percolation. The simulations are accelerated by up to a factor of 14 compared with an optimized scalar implementation. The random bit string generation algorithm proposed here is applicable to general Monte Carlo methods.
arXiv comment9 pages, 7 figures
References in corpus (7)
- Directed percolation criticality in turbulent liquid crystals
- A Universal Transition to Turbulence in Channel Flow
- Low-density series expansions for directed percolation I: A new efficient algorithm with applications to the square lattice
- Multi-GPU Accelerated Multi-Spin Monte Carlo Simulations of the 2D Ising Model
- Order-N Cluster Monte Carlo Method for Spin Systems with Long-range Interactions
- GPU-based Swendsen-Wang multi-cluster algorithm for the simulation of two-dimensional classical spin systems
- Geometric Allocation Approaches in Markov Chain Monte Carlo