Compiler

A small compiler written in C++ that translates a minimal programming language into ARM64 assembly, then assembles and links it into a native executable.

This is a learning project whose goal is to understand the entire compilation pipeline, from tokenization to code generation and assembly, before moving on to more advanced compiler concepts.

Current Features

The language currently supports a single instruction:

exit <code>;

Example (test.txt):

exit 42;

Compiling this file produces an executable that exits with status code 42.

How It Works

1. Tokenizer (src/tokenization.hpp): splits the source code into tokens (exit, an integer literal, and ;).
2. Code Generation (tokens_to_asm in src/main.cpp): walks through the token stream and generates the ARM64 assembly corresponding to exit N; (using the exit system call with exit code N).
3. Assembly and Linking: the generated .s file is assembled with as and linked with ld to produce a native executable.

Requirements

• A C++ compiler (C++17 or later)

• ARM64 cross-compilation tools:

  • aarch64-linux-gnu-as
  • aarch64-linux-gnu-ld

On Debian/Ubuntu:

sudo apt install g++ binutils-aarch64-linux-gnu

Building

g++ -std=c++17 -o comp src/main.cpp

Usage

./comp test.txt
./exec
echo $?   # Prints the exit status defined by "exit N;"

Current Limitations

• Only one instruction is supported (exit).
• No syntax error handling. Invalid input silently produces an empty program.
• ARM64 is the only supported target architecture (cross-compilation tools are required on non-ARM64 hosts).
• No Abstract Syntax Tree (AST) yet. Code generation operates directly on the token stream.

Roadmap

• Add arithmetic operations (+, -, etc.)
• Add variables
• Replace the flat token stream with a proper parser and an Abstract Syntax Tree (AST)
• Improve syntax error reporting