IonQ : “A Single CPU” Enough for Error-Correction Decoding

IonQ : “A Single CPU” Enough for Error-Correction Decoding

IonQ claims to have developed a decoder capable of processing a quantum computer’s errors in real time using a single standard processor (CPU). Described in a research paper published on arXiv, this work targets one of the main obstacles to fault-tolerant quantum computing.

Qubits, the basic units of these machines, are extremely fragile. Light, vibrations, or other qubits can be enough to cause them to lose their quantum nature, and with it the data they carry. Quantum computers therefore cannot measure qubits directly without destroying them. Instead, they produce syndromes, signals indicating that an error has occurred, which decoders must then analyze to determine which errors are affecting the system.

According to researchers from QuEra, a company that uses neutral atoms to build its qubits, this decoding constitutes “a major computational complexity challenge, especially as the size of the quantum system increases.” “Efficient decoders must balance speed and accuracy, ensuring real-time performance without sacrificing the ability to correctly interpret complex error patterns,” they add. Decoding relies heavily on classical algorithms and historically consumes a lot of computing power, which grows with the number of qubits.

In this context, researchers at IonQ, which develops trapped-ion systems, describe a decoder that operates in the background while the quantum computer continues its calculations. In their paper, they mention the concept of a MegaQuOp, a device capable of executing millions of operations on hundreds of logical qubits, which would represent “a major milestone for the quantum computing community.” “However, real-time decoding of hundreds of logical qubits throughout a computation involving millions of logical operations remains a major challenge,” they write.

For their tests, the researchers created benchmark circuits simulating up to 408 logical qubits distributed across 88 memory blocks and “magic state factories.” These circuits executed over 31.5 million individual quantum operations at the MegaQuOp scale, with a computational time overhead as low as 0.02%.

The decoding system covers 68 LDPC memory blocks and 20 magic state factories, implemented using 11,680 physical qubits. It relies on a dual sliding-window decoder built around a beam-search decoder. Detector error models (DEMs), which describe how gate errors or noise cause certain errors in the correction circuit, are generated in real time. According to the researchers, the entire classical decoding pipeline can run on just 12 cores of a standard CPU.

This result builds on the Walking Cat architecture for trapped-ion systems, presented by the company in April in a detailed roadmap. It includes a compiler, error-correction protocols, a microarchitecture, and a fast decoder. Its central element is a “cat factory” that produces “cat states,” superpositions of distinct states, distributed throughout the system to perform logical operations. The researchers see it as a simple architecture where “logical computation requires neither fusion nor warping of qLDPC memory blocks.”

IonQ executives believe these results show that organizations do not need to grow their classical hardware exponentially as quantum computers gain logical qubits. The ability to run error decoding on a classical computer with a single CPU is, according to them, central to the company’s roadmap, which goes beyond the 256 physical qubits of its Superion system. The company announced that Florida International University is purchasing a Superion system, to be installed by the end of 2027 in a specialized building on campus.

S’abonner
Notification pour
guest

0 Commentaires
Le plus ancien
Le plus récent Le plus populaire
0
Nous aimerions avoir votre avis, veuillez laisser un commentaire.x