A fast and frugal Gaussian boson sampling emulator

Classical algorithms have successfully reproduced expectation values of observables for certain quantum circuits using off-the-shelf computing resources; however, matching the performance of the most advanced quantum devices on sampling problems typically incurs extreme memory and computational costs, restricting such simulations to only a handful of supercomputers worldwide. In this work, we demonstrate a classical simulation outperforming Gaussian boson sampling experiments of 100 modes on established benchmark tests using a single central processing unit (CPU) or graphics processing unit (GPU). Being embarrassingly parallelizable, a small number of CPUs or GPUs allows us to match previous sampling rates that required more than 100 GPUs. We believe that algorithmic and implementation improvements have the potential to generalize our tools to photocounting and pseudo-photon number-resolving scenarios and larger system sizes. Last, most of the innovations in our tools remain valid for generic probability distributions over binary variables, rendering them potentially applicable to the simulation of qubit-based sampling problems and creating classical surrogates for classical-quantum algorithms.