2022 Fiscal Year Final Research Report
High Speed FPGA Simulator for Large Scale Quantum Annealing Simulations
| Project/Area Number |
19K11998
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 60100:Computational science-related
|
|---|
| Research Institution | Tohoku University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2019-04-01 – 2023-03-31
|
| Keywords | 量子アニーリングシミュレーション / FPGA / カスタムアクセラレータ / 組合せ最適化問題 |
|---|
| Outline of Final Research Achievements |
(1) Large scale quantum annealing simulation: The connections among spins (coefficients) increases exponentially with the number of spins. This increases the required memory capacity and prevents large-scale simulation. This research proposes a method to generate the coefficients efficiently, without storing the pre-generated coefficients in the memory. As a result, large memory requirement has been eliminated, and we were able to run simulations with over 200,000 spins using a single FPGA.
(2) Acceleration: Quantum Monte-Carlo method used for the simulations is very difficult to parallelize. We proposed a method to execute the computations among multiple Trotter slices in parallel while maintaining the data dependancy. As a result, we achieved over 290 times speed-up compared to CPU serial implementation.
(3) High accuracy: Since we protect the data dependency in parallel computation, the accuracy is very high. Compared to D-Wave using MQLib benchmark suit, the accuracy is over 99%.
|
| Free Research Field |
計算機アーキテクチャ
|
| Academic Significance and Societal Importance of the Research Achievements |
量子アニーリングは最適化において重要な手法であり，交通量シミュレーション，工場の作業の最適化，避難経路最適化などの様々な実用的な問題を効率的に解くことができると知られている．しかしながら，実問題は大規模であり，D-Waveなどの量子アニーラーを用いる事は難しい．本研究プロジェクトでは２０万スピン以上に全結合シミュレーションができており，複数FPGAを使う場合はさらに大規模化ができる可能性を示した．さらにCPUの２９０倍以上の高速性を９９％以上の高い計算制度で達成できた．また，GPUやマルチコアCPUを用いた高速化も提案されており，社会的な最適化問題に応用できる可能性は十分示している．
|
