Disclaimer: machine translated by DeepL which may contain errors.
The Rigakubu News
The Rigakubu News, July 2025.
Research Student Communicates to Faculty >
A Novel Architecture for Practical Quantum Computing
Takuo Kobori (Doctor couse, Physics)
Synge Todo (Professor, Physics)
In the quantum world, information cannot be copied.
Because of this restriction, quantum computers have had limited design freedom, and the mainstream format has been to complete all operations in-situ.
We have proposed a new design, a "load-store error-resistant quantum computer," based on the idea of "moving" quantum data.
By introducing the division of roles where the memory holds the data and the processor performs the operations, and by moving the data by load/store as necessary, we have succeeded in reducing the size of
quantum computer by about 40%, while keeping the increase in computation time to only about 3%.
Quantum computers are the next generation of computers that have the potential to solve problems at high speed that would take an enormous amount of time to execute with conventional classical computers. However, qubits are very sensitive and error-prone, and large-scale quantum computation requires advanced quantum error-correcting codes and a large number of qubits to support them.
In addition, quantum information has the fundamental restriction that it cannot be copied. This is known as the "cloning prohibition theorem" (*1). This restriction makes it difficult for quantum computers to be designed in a flexible manner based on caching and data replication, as is the case with classical computers. Until now, most quantum computers have been designed in such a way that all quantum operations are performed in the field where qubits physically exist, and there are many practical issues such as scalability and versatility. To solve these problems, we have proposed a "load-store error-tolerant quantum computer.
In this new design, the "memory area" that holds data is clearly separated from the "processor area" that performs quantum operations. This type of design has become the mainstream in classical computers due to its high practicality. The necessary data is loaded from the memory to the processor before the operation, and the results are stored back into the memory after the operation. This load/store is achieved by "moving" the information rather than copying the quantum state. In quantum error correction codes, a single logical qubit is represented by a collection of many physical qubits, and by expanding and contracting the collection like an amoeba, the quantum information can be efficiently moved.
Furthermore, in addition to the "load-store type" proposal, we have demonstrated that there is "locality" (*2) in the data accessed in quantum computation. By utilizing this property and managing the movement of quantum states, the frequency of load/store can be reduced and the overall efficiency of the computation can be increased.
Evaluation by simulation showed that the size of the quantum computer, i.e., the number of qubits required, was successfully reduced by about 40% compared to conventional designs, while the increase in computation time was only about 3%, an extremely practical result. In addition to reducing the hardware burden, this achievement greatly improves software portability and design flexibility.
In the future, along with the development of quantum hardware, this design will be applied to various quantum algorithms, which is expected to greatly advance the practical use of quantum computers.
Comparison of classical and quantum computation. In this study, we proposed a load-store type quantum computation architecture.This result was presented at The 31st IEEE International Symposium on High-Performance Computer Architecture (HPCA2025).
1 Theorem of non-cloning: A fundamental theorem in quantum computation that states that quantum manipulation to duplicate an arbitrary quantum state is impossible in principle.
2 (Memory access) locality: A bias in data access patterns when performing calculations. There are two types of locality: temporal locality, which refers to the tendency for data that has been referenced once to be accessed again in a short period of time; and spatial locality, which refers to the tendency for data held in physically close proximity to the referenced data to be referenced.