Short answer: because cache misses are 10x more common and dominate the time saved from pretty much all of those features. Get your cache misses in order(which

Short answer: because cache misses are 10x more common and dominate the time saved from pretty much all of those features. Get your cache misses in order(which is easy to understand in the C model) and you're ahead of 95% of the curve(plus you get a 10-50x performance bump as a reward).