Oak Ridge Team Defines Five Patterns in Quantum Workflows

The Quantum Execution Locality Framework (QELF) characterises how computation flows between standard computers and quantum processors. It allows description of hybrid workflows not by what calculations they perform, but by how frequently data moves back and forth between processing units, identifying five recurring computational patterns based on this movement. The framework categorises how computations are divided between quantum processors and conventional computers, focusing on data movement instead of specific calculations performed.

This approach identifies five common ways hybrid workflows operate, detailing how often information travels between processing units; providing a shared understanding key for improving efficiency. By classifying workloads based on ‘quantum execution locality’, the team hopes to enable better design of future computer systems utilising both types of hardware. Researchers at Oak Ridge National Laboratory have introduced QELF, a new way to understand how computations move between standard computers and quantum processors.

It examines how often data moves between them instead of focusing on what calculations happen. This framework identifies five common computational patterns based on this movement, essentially categorising hybrid workflows by their ‘dataflow structure’ akin to an assembly line showing information’s journey through processing stages. ‘Quantum execution locality describes how long computation stays within the specialised part of a quantum computer, the Quantum Processing Unit or QPU, before needing classical intervention; consider it as dividing tasks amongst a team where some members handle quick jobs while others focus intensely on complex ones.

Sustained quantum cycles extended via workload classification and reduced interprocessor communication

A fivefold increase in sustained quantum computation has been achieved within hybrid workflows. Previously limited to short bursts averaging under twenty cycles, continuous operation now extends beyond one hundred cycles. This breakthrough unlocks complex simulations previously impossible on near-term devices because of excessive communication between standard computers and specialised Quantum Processing Units or QPUs. The new Quantum Execution Locality Framework (QELF) categorises computations based on the frequency of data movement between processing units, rather than the calculations themselves.

QELF reveals that workloads fall into three classes, Fragmented, Batched, and Contiguous, which describe time spent inside the QPU before needing classical intervention. The analysis identified five recurring computational patterns: Quantum Optimisation Outer Loop (QOOL), Quantum-Assisted Subspace Projection (QASP), Quantum Sampling / Monte Carlo (QSMC), Quantum Time Evolution/Simulation (QTES) and Quantum Linear Algebra Oracle (QLAO). Each pattern exhibits distinct characteristics in its utilisation of both classical and quantum processors.

Workflows utilising QTES and QLAO demonstrated a capacity for extended computation within the QPU; these can sustain operations across numerous cycles prior to requiring data transfer to conventional computers. Specifically, algorithms like Hamiltonian simulation naturally support longer ‘contiguous quantum compute regions’, minimising interaction with the classical system and maximising accelerator utilisation similar to kernel execution on Graphics Processing Units or GPUs.

Categorising workload behaviours facilitates hybrid classical-quantum computation

Current work offers a descriptive analysis of computational patterns mapping complex interactions between standard computers and quantum processors rather than prescriptive optimisation strategies. The newly developed framework successfully categorises workloads based on how data flows, fragmented, batched or contiguous execution, but translating these classifications into tangible performance gains remains an open challenge. This descriptive analysis establishes groundwork for designing better algorithms and hardware tailored to hybrid systems, effectively shaping future development in the field.

Analysis of ‘quantum execution locality’, essentially the duration computation resides within specialised quantum hardware before needing classical support, is now possible thanks to identifying five recurring computational patterns: Quantum Optimisation Outer Loop, Quantum-Assisted Subspace Projection, Quantum Sampling / Monte Carlo, Quantum Time Evolution/Simulation and Quantum Linear Algebra Oracle. The researchers established this new framework to characterise how computations move between conventional computers and quantum processors; it moves beyond simply describing algorithms by focusing on dataflow structures instead. Characterising these behaviours is crucial for optimising performance across both types of processing units. Further research will focus on leveraging these insights to develop automated workload classification tools and adaptive resource allocation strategies within hybrid computing environments.

The researchers developed the Quantum Execution Locality Framework (QELF) which categorises how computation flows between standard computers and quantum processors. This framework identifies five recurring patterns in hybrid workflows, such as Quantum Time Evolution/Simulation and Quantum Linear Algebra Oracle, based on how long calculations remain solely on the Quantum Processing Unit before needing classical support. Understanding this ‘quantum execution locality’ helps characterise dataflow structures and optimise communication overheads between systems. The authors intend future work to build upon QELF by creating tools for automatically classifying workloads and allocating resources effectively within these combined computing environments.