Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/97727| Title: | 針對全狀態QAOA 電路模擬之混合層的效能導向最佳化研究 Performance-Driven Mixer Layer Optimization in Full-State QAOA Circuit Simulation |
| Authors: | 楊卓敏 Chuo-Min Yang |
| Advisor: | 洪士灝 Shih-Hao Hung |
| Keyword: | 高效能計算,量子近似最佳化演算法,量子電路模擬, High Performance Computing,Quantum Approximate Optimization Algorithm,Quantum Circuit Simulation, |
| Publication Year : | 2025 |
| Degree: | 碩士 |
| Abstract: | 本論文旨在解決量子近似最佳化演算法(Quantum Approximate Optimization Algorithm, QAOA)在傳統電腦上模擬的效能瓶頸。先前的研究已對Cost Layer進行優化,使得模擬的效能瓶頸轉移至Mixer Layer 並更加明顯。本論文的目的在於解決傳統電腦上模擬QAOA所面臨的這一新瓶頸。為此,本研究提出兩項核心技術:一是Cache Resident State Operation(CRSO),透過分割狀態向量以提升資料局部性與快取使用效率;二是導入「單指令多資料流」(SIMD)向量化,以加速CPU平台的算術運算效能。本研究的整合性方法取得了顯著的效能提升。評測結果顯示,在34個量子位元的模擬中,本模擬器在單一CPU節點上的執行速度相較於其他頂尖量子電路模擬器提升了最多27倍;在32個量子位元的模擬中,在單一GPU上的執行速度也提高了約13倍。此成果證明,結合考量硬體特性的演算法優化策略,能為複雜量子電路的模擬帶來顯著的效能優勢,並為相關領域的未來研究奠定穩固基礎。 This thesis tackles the remaining bottlenecks in classical Quantum Approximate Optimization Algorithm (QAOA) simulation, most notably in the mixer layer, by introducing two key techniques (1) Cache-Resident State Operation (CRSO), which partitions the state vector for better data locality and cache efficiency and (2) SIMD vectorization to accelerate CPU arithmetic. Together, these hardware-aware optimizations boost performance by up to 27x on a single-node CPU for 34-qubit circuits and 13x on a single GPU for 32-qubit circuits, laying a strong foundation for future high-performance quantum-circuit simulation research. |
| URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/97727 |
| DOI: | 10.6342/NTU202501429 |
| Fulltext Rights: | 同意授權(全球公開) |
| metadata.dc.date.embargo-lift: | 2025-07-17 |
| Appears in Collections: | 資訊網路與多媒體研究所 |
Files in This Item:
| File | Size | Format | |
|---|---|---|---|
| ntu-113-2.pdf | 1.29 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.
