Brainfunk: Yet Another Brainfuck Interpreter

Build

make
make test

Features

• C++20 interpreter and reusable execution library
• RLE optimizations for +- and ><
• Common loop optimizations: set-to-zero, multiply-with-offset, and find-zero
• C99 code generation with the same 8-bit cells and wrapping tape semantics as the interpreter
• Portable textual LLVM IR generation without linking against LLVM libraries
• ncurses TUI with pause, single-step, speed control, program I/O, and resize support
• bf, a small direct interpreter retained as a comparison implementation

Modes

bf: interpret a Brainfuck program

bfc: convert a Brainfuck program to C99

bit: print optimized bitcode

llvm: lower optimized bitcode to portable textual LLVM IR

Usage

# Run a Brainfuck program:
./brainfunk -f <source-file>

# Translate it to C:
./brainfunk -m bfc -f <source-file> -o program.c

# Print optimized bitcode:
./brainfunk -m bit -f <source-file>

# Emit portable LLVM IR:
./brainfunk -m llvm -f <source-file> -o program.ll

# Emit LLVM IR and compile it with the C runtime wrapper:
./compile-llvm-ir.sh <source-file> <executable>

# Open the visualizer (100 ms per optimized instruction):
./visualbrainfunk -f <source-file> -t 100

TUI controls:

• Space/p: pause or resume
• s: pause and execute one optimized instruction
• Up/+, Down/-: adjust the delay
• q/Esc: quit

When the Brainfuck program executes ,, the TUI displays a program-input prompt. The next byte is delivered to the program rather than interpreted as a TUI command.

The LLVM emitter writes plain text and has no build-time or runtime dependency on LLVM. llvm-ir-wrapper.c supplies the tape, main(), and byte I/O callbacks; Clang performs target selection and optimization only when the emitted IR is compiled. See doc/llvm-ir.md for the ABI and portability contract.

CPU core

The ultimate Brainfuck CPU

Credits

• brainfuck optimization strategies by Mats Linander
• compile-time-regular-expressions library by Hana Dusíková