在LnQ架構下實作ARM到x86-64的執行碼轉換

Abstract

本篇論文是基於LnQ架構上去設計並實作出一個ARM到LLVM的前端來建造一個ARM到x86-64的動態執行碼轉換，同時也去驗證LnQ的可移植性。LnQ是由LLVM及QEMU組合而成的，LnQ選擇用LLVM IRs來取代QEMU原本的micro operations。因此，LnQ的架構因為有著現存優化器和後端產生程式碼，所以目標是成為一個可以提供高效能以及重新移植能力的動態二進制轉換器。我們在LnQ架構上進行ARM到x86-64執行碼轉換的實驗，我們使用工業界標準的SPEC CINT 2006來驗證這轉譯器的成效。實驗結果指出，ARM執行時間比QEMU的執行時間平均快上2.08倍。

This thesis is to design and implement a frontend from ARM to LLVM base on the LnQ framework to provide an ARM to x86-64 dynamic binary translation and also prove the retargetability of LnQ. LnQ, which combines LLVM and QEMU, use LLVM IRs to take place of the original micro operations in QEMU. Therefore, the goal of LnQ framework is to enable the process of building high performance and retargetable dynamic binary translator since LnQ has existing optimizers and code generation back-ends. We conduct experiments of dynamic binary translation from ARM to x86-64 with the LnQ framework with the industry SPEC CINT 2006 benchmarks. The results shows that the average speedup is 2.08X faster than QEMU.