Introduction and index of this series is here. In the previous post I explored how to make data filters a bit faster, using some trivial merging of filters, and a largely misguided attempt at using SIMD. People smarter than me pointed out that getting good SIMD performance requires a different approach. Which is kinda obvious, and another thing that is obvious is that I have very little SIMD programming experience, and thus very little intuition of what’s a good approach.
