Scalable and targeted benchmarking of quantum computers

A quantum processor’s performance is limited by the errors it experiences. But most current performance metrics do not accurately predict what programs a particular processor can successfully run, due to complex errors that only emerge at scale. In this talk we will present scalable, holistic, and flexible benchmarking techniques, built on “mirror circuits”, that can efficiently test the capabilities of any programmable quantum computer. We will show how to create mirror circuit benchmarks based on randomized, periodic, and algorithmic circuits; we will show how mirror circuit benchmarks can be used for scalable “full stack” benchmarking; and we will demonstrate these techniques with experiments. These experimental results show that current hardware suffers from structured errors whose effect cannot be predicted from standard error metrics inferred from randomized circuits.