Full metadata record
DC FieldValueLanguage
dc.contributor.author黃宣偉en_US
dc.contributor.authorHuang, Hsuan-Woeien_US
dc.contributor.author陳正en_US
dc.contributor.authorCheng Chenen_US
dc.date.accessioned2014-12-12T02:18:38Z-
dc.date.available2014-12-12T02:18:38Z-
dc.date.issued1997en_US
dc.identifier.urihttp://140.113.39.130/cdrfb3/record/nctu/#NT860392047en_US
dc.identifier.urihttp://hdl.handle.net/11536/62778-
dc.description.abstract共享記憶體多處理機系統最近已成為計算機系統架構設計的一個重要 趨勢。而由於計算量需求的大增,因此具有更多處理器的多處理機系統變 成為一個趨勢,而叢集多處理機系統兼具可延伸性與區域性之特性,就成 為了一個重要的發展方向。我們將設計一個叢集多處理機系統模擬評估環 境,來探討叢集多處理機系統中的各項重要議題。我們的模擬評估環境是 一種程式驅動式模擬環境,它是由MINT所提供的記憶體參考產生器與我們 所設計的記憶體子系統所組成。我們所設計的記憶體子系統模擬器,可提 供兩層式快取記憶體、區域匯流排、叢集間快取記憶體、兩層的快取記憶 體一致性協定以及連結網路的模擬。 在本論文中,我們將利用這個模 擬評估環境,來探討有關一般叢集多處理機系統設計上的議題評估、資料 預先存取技術、以及平行編譯器支援機制等重要議題設計與方法。經過模 擬評估後我們發現,當我們需要一定數量的處理器來協同處理大量計算時 ,叢集多處理機系統由於其可延伸性,因此是一個相當不錯的設計方向。 同時我們也能瞭解到,在設計叢集多處理機系統時,其叢集中處理器個數 ,以及叢集間快取記憶體的大小,都必須保持在一定的設定,以避免過大 或過小所造成的缺點。透過預先存取設計的模擬結果,我們瞭解到在叢集 系統中最好採用簡單的預先存取機制以避免交通量的負擔,如果再加上我 們所設計的叢集與結合預先存取,則更能夠減少其對系統效能的負面影響 。最後,由平行編譯器支援技術的模擬我們瞭解到,在叢集系統中提供此 方面的支援更能有效提升系統效能,因此也是一個日受重視的研究課題。 Recently, shared-memory multiprocessor systems have become one of the design trends in computer system architectures. As the increasingly need of computation, multiprocessor system with more processors on it becomes an unavoidable trend. Thus, clustering multiprocessor system has indeed played an important role due to its high scalability and data locality. We had developed a simulation and evaluation environment for clustering multiprocessor system, which aims at investigating the key design issues of its memory subsystem. Our environment is a program-driven simulator, consisting of a memory reference generator supported by MINT and a memory subsystem simulator that we have designed. The memory subsystem simulator can support several simulation modules, including two- level cache, local bus, inter-cluster cache, cache coherence protocols, and interconnection network. With the aid of this environment, we had studied major design issues of clustering multiprocessor system, including data prefetching and parallel compiler assistant techniques. Based on a great deal of evaluation results, we have found that clustering may have good scalability compared with non-clustering architectures. By the way, the size of cluster node and related issues are also investigated. Moreover, data prefetching techniques boost the performance gain in clustering multiprocessor system, especially with our clus-prefetch and combined-prefetch design. Besides, we also provide some assistant techniques for parallel compiler to improve performance of clustering multiprocessor system in some degree.zh_TW
dc.language.isozh_TWen_US
dc.subject叢集zh_TW
dc.subject多處理機zh_TW
dc.subject預先存取zh_TW
dc.subject平行編譯器zh_TW
dc.subject模擬評估環境zh_TW
dc.subject可延伸性zh_TW
dc.subjectClusteren_US
dc.subjectMultiprocessoren_US
dc.subjectPrefetchen_US
dc.subjectParallel Compileren_US
dc.subjectSimulatoren_US
dc.subjectScalabilityen_US
dc.title叢集多處理機系統中預先存取與平行編譯器支援技術之探討及其模擬評估環境的研製zh_TW
dc.titleA Study on Prefetch and Compiler Assistant Techniques for Clustering Multiprocessor System Design and Implementation of Its Simulation and Evaluation Environmenten_US
dc.typeThesisen_US
dc.contributor.department資訊科學與工程研究所zh_TW
Appears in Collections:Thesis