Scalable Compilation of Quantum Circuits via Approximate Ensembles

As quantum processors scale in the number of qubits, the compilation workflow becomes paramount to the ability to run useful quantum algorithms. In our talk, we present a scalable workflow that is able to generate large ensembles of approximate compilations. Basing off prior work by Campbell and Hastings, we present theoretical results that demonstrate quadratic suppression of worst-case unitary error through ensemble sampling. By integrating these theoretical results with state-of-the-art compilation tools, we are able to create a scalable workflow leveraging numerical circuit synthesis for code generation. These approximate solutions are often shorter and less complex, which further limits the errors that arise on chip during an algorithm's execution. We show that our workflow is flexible enough to not only decrease CNOT gate count on today's NISQ computers, but extends to improve both T gate count and depth in future fault-tolerant machines.