Highly optimized tensor network contractions for the simulation of classically challenging quantum computations

Experimental quantum computations have progressively entered the beyond-classical regime over the last four years on the task of random circuit sampling. Simultaneously, almost exact, highly optimized tensor network contraction has been established as the leading approach for attempting the simulation of these and other experiments. In this talk, I will describe our latest strategies for the optimization of tensor network contraction in this context, which give tight estimates on the simulation computational cost of our largest experiments. In addition, I will discuss how these strategies apply to our most recent random circuit sampling experiment (arXiv:2304.11119).