What are you even talking about? You just revived 6 months old topic. We use SSE2 for good to gain a bit of performance in some places. If compiler knows that computer supports SSE2 it passes floating-point information in SSE register and generates a bit better code. Of course with infinite coder time we could get that hotspot fast with SSE1. If the link in 2.27 released topic doesn't work for you, tell us.