jiti

JITi, short for Just-In-Time interpreter, refers to a class of runtime systems that combine interpretation with on-the-fly compilation to native code. The goal is to accelerate dynamic languages by converting hot code paths into optimized machine code at runtime while preserving the flexibility of interpretation. JITi systems typically separate the execution engine into an interpreter core, a JIT compiler, and a runtime environment responsible for memory management, exceptions, and inter-language calls.

Most JITi implementations use a combination of tracing or method-based compilation. The interpreter executes until it detects frequently executed paths, collects profiling data, and triggers compilation. The JIT emits machine code optimized for the current execution profile, and the runtime switches to the compiled code. If assumptions fail, deoptimization guards revert execution to the interpreter or recompile with updated assumptions. Techniques such as inline caches, speculative inlining, and on-stack replacement are common.

Advantages include faster warm-up and improved performance on hot paths without imposing a full ahead-of-time compilation model. Limitations include increased memory usage, longer startup times, compiler complexity, and potential debugging challenges.

In practice, JITi concepts appear in many language implementations and research projects, including dynamic languages that employ interpreters with optional JIT backends. They are distinct from separate compiler pipelines like ahead-of-time compilers or full tracing JITs, though some systems blend approaches.