bfcc

A Brainf*ck to native Linux x86-64 ELF64 compiler, written in NC Lang.

Compiles .bf/.b/.fck source files directly into statically-linked ELF binaries using Linux syscalls.

Usage

bfcc <file> [-o output]

The default output is a.out.

How it works

1. Tokenizes Brainf*ck source into instructions
2. Emits x86-64 machine code (using r12 as the tape pointer)
3. Wraps it in a raw ELF64 binary with a 30,000-byte BSS tape

Examples

ncc build ./bfcc.nc
chmod +x ./bfcc
./bfcc examples/mandlebrot.b -o mandlebrot
./mandlebrot

Benchmarks

nc-bfcc results on AMD Ryzen 7 5700, Linux x86-64. Best of 5 runs.

Program	Compile Time	Run Time	Binary Size
mandelbrot.b	0.04s	0.55s	19 KB
bottles.b	<0.01s	<0.001s	8 KB
hanoi.b	0.57s	0.08s	73 KB

Compiler build time: 0.10s (ncc build ./bfcc.nc). Compiler binary: 134 KB.

Comparison with Other Brainfuck Implementations

Mandelbrot runtimes collected from published benchmarks. Hardware varies across projects (noted where known), so treat cross-project comparisons as approximate. nc-bfcc numbers are from the table above.

AOT Compilers & JITs

Implementation	Language	Approach	Mandelbrot	Bottles	Notes
nc-bfcc	NC Lang	AOT → x86-64 ELF	0.55s	<0.001s	This project. Direct machine code, no libc.
Tritium	C	Interpreter + JIT (DynASM)	~0.30s	—	Fastest known BF JIT. Uses LuaJIT's DynASM.
BF-JIT	Rust	JIT (x86-64/AArch64)	~0.39s (M3) / ~1.06s (i5)	—	Hybrid AOT/JIT. 53x speedup over naive.
bfc	Rust	AOT via LLVM	—	0.007s	Industrial-grade. Full LLVM optimization pipeline.
Cranefack	Rust	Interpreter + JIT (Cranelift)	—	—	Multiple optimization levels. Built-in benchmarking.
Nayuki BF→C	Python	Transpiler → C	~0.64s (gcc -O1)	—	BF→C, then relies on gcc/clang for codegen.
Geschwindigkeitsficken	C	Transpiler → C / x86-64 asm	—	—	9 optimization strategies. Outputs C or asm.
esotope-bfc	Python	Transpiler → C	—	—	Gold standard BF→C optimizer (2009). Full constant propagation.
awib	Brainfuck	Transpiler → C/Go/Java/Rust/i386 ELF	—	—	Written in BF itself. Targets 7 languages.

Optimized Interpreters

Implementation	Language	Mandelbrot	Notes
bf-cpp	C++	~0.77s	Claims fastest non-JIT interpreter.
BrainForked	C	~0.98s	Optimized interpreter.
Tritium (interp only)	C	~1.16s	Same project, interpreter-only mode (-r).
sbfi	C	~1.14s	Simple optimized interpreter.
bffsree	C	~1.24s (M3)	Fastest pure interpreter (no JIT/ASM). "Double-instructioning" technique.
bff4	C	—	Long-standing fast interpreter. dbfi: 248ms.

Transpiler Targets (Van Heusden compilersuite)

Mandelbrot runtimes via vanheusden.com BF→X transpiler:

Target	Mandelbrot
Rust	~1.30s
x86 asm	~1.48s
C (GCC)	~1.61s
Go	~3.58s
Node.js	~6.24s
Java	~10.66s
Python (PyPy)	~8.12s
Ruby	~1m42s
Bash	~5h16m

Key Takeaways

• nc-bfcc places in Tier 1 among BF compilers, with mandelbrot at 0.55s — competitive with JIT compilers and faster than most interpreters and transpiler pipelines.
• Only Tritium's DynASM JIT (~0.30s) and BF-JIT on Apple M3 (~0.39s) are clearly faster, both using runtime code generation.
• nc-bfcc produces tiny, self-contained binaries (8–73 KB) with zero dependencies — no libc, no dynamic linking.
• Build speed is near-instant — the compiler itself builds in 0.10s and compiles mandelbrot in 0.04s. Most LLVM/Cranelift-based alternatives have multi-second or multi-minute build times for the compiler itself.
• bffsree deserves special mention as the fastest pure interpreter (no JIT, no native codegen), achieving ~1.24s on mandelbrot through aggressive bytecode-level optimization.