Emulation of virtually addressed control flow in a multi-processor dynamic binary translator

摘要

Emulating a source application with two or more emulator processes executing on respective processors of a multi-processor platform with a shared code heap and virtual trampoline memory where each emulator process has a respective private trampoline memory. The emulation may be a dynamic binary translation which begins by fetching 602 a fragment of source instructions from emulated physical memory then generates 604 equivalent target instructions and writes 606 these to the code heap at a location indexed by its corresponding emulated physical address. Each jump instruction in the fragment is translated 608 to a jump to a slot in the virtual trampoline memory and a trap is written 612 in each private trampoline at that slot. When the trap is encountered during execution a trap handling method (figure 9) replaces the trap with a physical address in the code heap corresponding to the start of the target instruction fragment.