Idea: I could use swimd for 1:1 solving, when target is large. If it is large enough, I could split it to pieces, solve them in parallel, and then combine them. I could use something similar to idea that is used for Hichberg. Possible problem is complexity of combining solutions together: if that part is too slow, the whole idea may not be worth it. Another thing to think about is how to exactly combine the solutions: I should examine Hichberg and start from there, there may also be some other articles that already used similar approach, I should find them and read them.