Hi, MWE below I have a use case for intersection of lists of sorted integer vectors with a vector of sorted integers i.e. intersect!(vec :: Vector{Int},vectorOfVectors::Vector{Vector{Int}}) because the vectorOfVectors can be quite large in and of itself, and the fact that I cant always “edit” m…

[image] DNF: so a 6x speed difference seems roughly expected, since avx2 works on eight 32bit integers in parallel. FWIW, the fast Base code doesn’t use SIMD either. Back in 2018 when I last looked at that in detail, the core loop of the Base code was intended to run in ~1.7 cycles per selec…