A Quantum Instruction Set Implemented on a Superconducting Quantum Processor

A quantum algorithm consists of a sequence of operations and measurements applied to a quantum processor. To date, the instruction set which defines this sequence has been provided by a classical computer and passed via control hardware to the quantum processor. Here, we demonstrate the first experimental realization of a quantum instruction set, in which a fixed sequence of classically-defined gates perform an operation that is fully determined only by a quantum input to the fixed sequence. Specifically, we implement the density matrix exponentiation algorithm, which consumes $N$ copies of the instruction state $\rho$ to approximate the operation $e^{-i \rho \theta}$ ($\theta$ an arbitrary angle). Our implementation relies on a 99.7\% fidelity controlled-phase gate between two superconducting transmon qubits. We achieve an average algorithmic fidelity $\approx 0.9$, independent of the setting of $\rho$, to circuit depth nearly 90. This new paradigm for quantum instructions has applications to resource-efficient protocols for validating entanglement spectra, principal component analysis of large quantum states, and universal quantum emulation.