Adding JIT-compilation to a toy interpreter with libgccjit
The source material details the process of adding Just-In-Time (JIT) compilation to a simple stack-based interpreter written in C. This toy interpreter supports a single data type, integers, and can handle functions with one parameter and recursion. The interpreter reads and executes bytecode from a .toy script file.
To implement JIT compilation, the author uses the libgccjit library, which allows generating machine code at runtime. The author creates a `toyvm_compiled_function` structure that bundles the JIT-compiled code and a gcc_jit_result pointer, which manages the code's lifetime.
The compiler generates machine code for each opcode, taking into account the stack manipulation required by the bytecode. Instead of calculating the stack depth beforehand, the compiler directly manipulates the stack using an array and a depth variable. Local variables `x` and `y` are also created for intermediate operations.
The compiler context, types, and initial block are set up at the beginning. The `add_push` and `add_pop` helper functions generate the assembly code for pushing and popping values from the stack array. The `toyvm_function_interpret` function becomes the entry point for executing the JIT-compiled code, with the interpreter loop replaced by the JIT-compiled function.
Written by urgent.news from Lobsters's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
