Ensemble-IR: Concise Representation for Quantum Ensemble Programs

TL;DR

Ensemble-IR is a novel intermediate representation that efficiently encodes large families of related quantum circuits, reducing redundancy and enabling scalable quantum workload management.

Contribution

It introduces a concise IR for quantum ensembles that encodes shared structure and variations, improving scalability over existing explicit enumeration methods.

Findings

Successfully applied to 18 real-world workloads

Reduces storage and computation overhead

Facilitates on-the-fly circuit recreation

Abstract

Emerging quantum applications such as error mitigation, system characterization, and hybrid protocols often require running large families of related quantum circuits. Existing intermediate representations (IRs) and frameworks such as Qiskit, QIR, MitiQ, and OpenQASM do not provide primitives to concisely express such workloads. These tools instead rely on explicit enumeration of the unique circuits within each workload. We introduce Ensemble-IR, an intermediate representation designed to concisely express these ensemble workloads. Rather than enumerating each circuit separately, Ensemble-IR encodes an entire workload through a rich, shared unified program. This program specifies common circuit structure along with symbolic operations to express points of variation - such as gate types, gate placement, parameter values, or qubit configurations. Ensemble-IR enables quantum systems to load an entire family of circuits onto a device as a single concise file, allowing contained circuits to instead be recreated on-the-fly at runtime. We demonstrate Ensemble-IR across 18 real-world workloads from prior literature, highlighting its widespread utility for scalable quantum system development.