頁籤選單縮合
題名 | 利用方格連接和多全通匯流排計算機來求Crout分4解法之分割演算法 |
---|---|
作者姓名(中文) | 黃竹明; 李俊德; | 書刊名 | 技術學刊 |
卷期 | 8:4 1993.12[民82.12] |
頁次 | 頁499-506 |
分類號 | 310.153 |
關鍵詞 | 方格連接和多全通匯流排計算機; 分割演算法; 軸旋操作; 通訊延遲; Crout分解法; MCCMGB; Partitioned algorithm; Pivoting; Communication delay; Crout decomposition; |
語文 | 中文(Chinese) |
中文摘要 | 本論文提出一個Crout分解法之分割演算法,另外,考慮演算法之穩定性,我們提出一個有軸旋操作的crout分解法之分割演算法。此演算法使用之架構為方格連接和多全通匯流排計算機,它改進了基本方格連接計算機之通訊延遲,為一種增強型的方格連接計算機。此分割演算法不受限於問題大小 (即資料量) ,因此,可應用到任意大小之線性方程式系統上,解決以往平行演算法中,處理單元的數量受限於目前VLSI製造技術之問題。此分割演算法之時間複雜度則為O (n3 / p2)。 |
英文摘要 | This paper proposed a partitioned algorithm of Crout Decomposition. Considering the stability of the algorithm, we proposed a partitioned algorithm of Crout Decomposition with partial pivoting. The architecture of the algorithms is MCC (Mesh-Connected Computer), with mutiple global buses (MCCMGB) - an enchanced MCC it has been designed to improved the communication delay in MCC. This partitioned approach is not restricted by problem sizes and thus can be applied to solve arbitrarily large linear systems of equations in an iterative fashion. It can also solve the problem which one can only expect limited PEs frorn the VLSI devices of the MCCMGB in early parallel algorithms. The tirne cornplexity of the partitioned algorithm for Crout Decomposition is O (n3 / p2). |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。