完整後設資料紀錄
DC 欄位 | 值 | 語言 |
---|---|---|
dc.contributor.author | 李柏舉 | en_US |
dc.contributor.author | Lee, Bo-Jyu | en_US |
dc.contributor.author | 徐慰中 | en_US |
dc.contributor.author | Hsu, Wei-Chung | en_US |
dc.date.accessioned | 2014-12-12T01:59:20Z | - |
dc.date.available | 2014-12-12T01:59:20Z | - |
dc.date.issued | 2012 | en_US |
dc.identifier.uri | http://140.113.39.130/cdrfb3/record/nctu/#GT079955582 | en_US |
dc.identifier.uri | http://hdl.handle.net/11536/50492 | - |
dc.description.abstract | 本篇論文是基於HQEMU系統模擬器架構上去設計並實作出一個真正能產生出前端為Intel SSE的指令集到真正後端硬體上的動態執行碼轉換。 HQEMU是由LLVM及QEMU組合而成的,HQEMU針對不同的程式行為來決定要使用原本的QEMU TCG IRs這個轉換進程或者使用LLVM IRs來取代QEMU本來的TCG IRs,結合QEMU快翻的精神與HQEMU做大量優化的特性。我們修改HQEMU的動態二元引擎始能產生真正的SIMD指令,並針對這種指令新增一個優化的選項叫做向量型態的狀態對應來提升轉換的效能。我們使用工業界標準的SPEC 2006 CFP要驗證修改這個轉換器後得到的成效。實驗結果指出,修改後的執行時間比原本HQEMU的時間平均快上1.35倍。 | zh_TW |
dc.description.abstract | This thesis is to enhance the DBT engine of the HQEMU system emulator so that it can efficiently translate SIMD instructions from the target architecture into the SIMD instructions in the host machine. In the process of augmenting the DBT engine with SIMD instruction code generation capability, we also propose an optimization, called vector type state mapping for eliminate redundant SIMD load/store instructions. With the enhancement and associated optimization, we have observed 35% of speed up over the original HQEMU when emulating the SPEC 2006 CFP benchmarks in x86-32 binary on the x86-64 host. | en_US |
dc.language.iso | en_US | en_US |
dc.subject | 動態二元翻譯 | zh_TW |
dc.subject | DBT | en_US |
dc.subject | QEMU | en_US |
dc.subject | SIMD | en_US |
dc.title | 在HQEMU系統模擬器的動態二元翻譯引擎上產生SIMD指令 | zh_TW |
dc.title | SIMD Instruction Generation in the DBT Engine | en_US |
dc.type | Thesis | en_US |
dc.contributor.department | 資訊科學與工程研究所 | zh_TW |
顯示於類別: | 畢業論文 |