Comment by jzombie

Do you foresee this being faster than SIMD for things like cosine similarity? Apologies if I missed that context somewhere.