Optimized GPU simulation of continuous-spin glass models

We develop a highly optimized code for simulating the Edwards-Anderson Heisenberg model on graphics processing units (GPUs). Using a number of computational tricks such as tiling, data compression and appropriate memory layouts, the simulation code combining over-relaxation, heat bath and parallel tempering moves achieves a peak performance of 0.29 ns per spin update on realistic system sizes, corresponding to a more than 150 fold speed-up over a serial CPU reference implementation. The optimized implementation is used to study the spin-glass transition in a random external magnetic field to probe the existence of a de Almeida-Thouless line in the model, for which we give benchmark results