Confirmed that it improves perfomance of simple "for i in range(N): pass"
loop by 15% on Core2.