Recompilers

Recompilers are software systems that translate code from one instruction set architecture (ISA) to another, typically to run software on hardware different from the original target. They may operate on binaries or on intermediate representations, and they are used to improve portability, performance, or both. Recompilation can occur offline, producing a retargeted executable, or online, as code is executed.

There are two main modes: static recompilation and dynamic recompilation. Static recompilers translate code before execution and produce a new binary for the target architecture. Dynamic recompilers, also known as dynamic binary translators or dynarecs, translate blocks of guest code on demand at runtime, storing generated host code in a cache and dispatching from it. Dynamic recompilers must preserve program semantics while optimally mapping registers, memory, and control flow, and they often incorporate profiling and optimization techniques.

Applications include emulation of legacy hardware, virtualization, and performance enhancement for interpreters or virtual machines. Recompilers can approach native execution speed by removing interpretation overhead and enabling host-specific optimizations, though they introduce complexity in translating low-level behavior such as memory models, exceptions, and self-modifying code.

Key challenges include ensuring correctness across complex architectures, handling dynamic code generation, memory management, and synchronization in multi-threaded environments, and balancing translation time, memory usage, and runtime overhead. Quality recompilers provide accurate translation, robust optimization, and adaptability to multiple target architectures.

See also: dynamic binary translation, just-in-time compilation, native code generation, retargetable compilers.