It compiles about 7x faster both at -O0 and -O3. My optimization is not quite on par but it is really close. Compiled code is about 5% slower.