標題: 有效監探式協定中之主動再生過濾器架構設計
Leak Stopper: An Actively Revitalized Snoop Filter Architecture by Effective Generation Filtering
作者: 黃培哲
陳添福
Huang, Pei-Zhe
Chen, Tien-Fu
資訊科學與工程研究所
關鍵字: 快取儲存;記憶體架構;過濾器;計算機架構;Cache storage;memory architecture;filters;computer architecture
公開日期: 2016
摘要: 為了減少經由監探式協定所產生的能量消耗,過往的研究使用監探式過濾器來減少查找快取記憶體中的標籤。但是此類的過濾器有個眾所皆知的問題,它的過濾效能會隨著使用而遞減,然而有些過濾器可以趁快取記憶體中的區塊被抹除時,藉此來做部分或全部的過濾器活化。不幸地,這些重置的動作可能遠不及使用的程度。所以在本論文中,我們提出了一個新的過濾器架構,使其可以自己去主動的做活化的動作。我們藉由把讀取到快取記憶體的區塊分為好幾個世代,藉由觀察程式的行為來切換世代,被取代的世代將會清空其過濾器的所有內容,並且為了一致性會抹除該世代所讀取進來的區塊,藉此來達到主動活化的目的。
To alleviate high energy dissipation of unnecessary snooping accesses, past researches have proposed snoop filters to reduce snoop lookups. These filters have notorious problems of decreasing filtering efficiency and thus usually rely on a partial or whole filter reset by detecting block evictions. Unfortunately the reset conditions happen infrequently or unevenly. This work proposes a revitalized snoop filter (RSF), which can actively renew the destination filter by employing a generation wrapping around scheme for various reference behaviors. We further utilize a sampling mechanism for RSF to timely trigger precise filter revitalizations so that the unnecessary flushing can be minimized.
URI: http://etd.lib.nctu.edu.tw/cdrfb3/record/nctu/#GT070256100
http://hdl.handle.net/11536/142706
Appears in Collections:Thesis