A universal quantum computer with a million qubits will solve a wide range of problems, but even then, offloading entire problems to a quantum circuit may not be the best use of resources.
With that in mind, companies and researchers are paying more attention to the concept of quantum-circuit cutting, which breaks down large quantum circuits into smaller fragments for execution across multiple devices, which could include quantum and classical computers.
IBM last month incorporated circuit cutting as a key element in its quantum roadmap, and has also given the technology its own name, calling it “dynamic circuits.” Dell is taking a close look at quantum circuit-cutting to plug quantum computers as accelerators into classical computing infrastructure.
“Quantum circuit cutting is the process of partitioning a large quantum circuit into smaller fragments that can be executed independently, either successively on the same device, or across multiple devices,” said Olivia Di Matteo, who is an assistant professor at the University of British Columbia’s Department of Electrical and Computer Engineering.
The quantum advantage depends on the ability to run the largest number of circuits on a complex problem. There are higher priorities for companies developing quantum computers, which include stability of quantum circuits, error correction, and topologies, but circuit-cutting is now making it to product roadmaps.
At the Supercomputing 2023 (SC23) show last month, some companies were aware of the concept, which is just emerging from research. For Ken Durazzo, vice president of research at Dell, quantum circuit-cutting will be a milestone in ensuring quantum computers mesh with existing classical computing infrastructures.
Dell has built a blueprint for a hybrid classical-quantum computing model with conventional servers hosting and managing quantum computers attached as accelerators. Dell’s current foundation of the hybrid model involves classical computers sending an entire quantum problem to a quantum circuit, and receiving back the output once the calculations are complete.
But at some point, Durazzo hopes concepts like quantum-circuit computing emerge to break down a quantum task and bifurcate its execution between a physical quantum hardware and software-based simulators on classical systems.
“We’re going to see more and more models of that interplay between classical simulation and actual physical quantum systems, where the intelligent orchestration and the ability to understand where to best run these particular functions or operations is going to become more critical,” Durazzo told HPCwireadding “We’re well positioned for that world as that emerges.”
Dell would need to improve its host processor in a classical system that can do intelligent scheduling and break down workloads between quantum circuits and simulators running on GPUs.
IBM, Nvidia, Intel and other quantum computer makers offer quantum simulators on classical computers that can replicate quantum hardware. The goal is to help companies prototype computing on quantum circuits. Nvidia’s QODA software acts as a quantum computer surrogate and runs on GPUs.
IBM’s dynamic circuits were hailed as an achievement and major milestone by Blake Johnson, a quantum engine lead, during last month’s IBM Quantum Summit.
“Dynamic circuits marry real time classical computation with quantum operations, allowing feedback and feed forward of quantum measurements to steer the course of a computation,” Johnson said at the event.
IBM’s focus for quantum-circuit is focused on more efficient use of quantum circuitry as opposed to offloading to simulators on classical computers. A lot of things can be done with dynamic circuits, Johnson said.
“Just to give one concrete use case, we know that dynamic circuits offer new opportunities to reduce circuit depth,” Johnson said. That adds more computing capability to its quantum computers.
IBM announced that it had rolled out dynamic circuits by enabling their exploration on its live quantum systems. The company’s long-term goal is to develop a universal quantum computer, which will need to support 1 million qubits, but it’s taking an alternate road to the commercialization of its quantum computers with concepts like circuit cutting.
IBM at the Summit announced the 433-qubit ‘Osprey’ quantum chip, and IBM Quantum System Two, which can support up to 4,158 qubits in a single system. Dynamic circuits may be relevant for System Two, which can be hooked up to other System Two quantum computers to expand processing capacity. Classical servers can also be attached to a System Two for applications that include machine learning and analytics. (For more on IBM’s recent quantum announcements, see our deep dive, here.)
Quantum circuit-cutting addresses limitations in current quantum systems, UBC’s Di Matteo told HPCwire.
“One advantage is that these circuit fragments use fewer qubits, which enables us to run larger computations on the devices we have today, which are relatively small,” Di Matteo said.
Another is that the fragments are generally shorter than the original circuit, so the computations suffer less from the effects of noise and limited coherence time, which is a challenge of today’s hardware, she said.
The quantum circuit-cutting procedure also has its tradeoffs, as scaling and overhead incurred by cutting are problematic.
“Moving into the future we need to develop more efficient methods that address problems such as finding optimal cut locations for problems with arbitrary structure, and minimizing the number of additional circuit executions and measurements required,” Di Matteo said.
But the idea of quantum circuit-cutting has staying power, even when a quantum computer reaches a million qubits.
“I don’t think there is a lifespan to the concept; no matter how large a quantum computer we build, we’ll always be able to dream of something bigger that will require more resources than we have at hand,” Di Matteo said.