Speedups of the MPI-parallelised FCFC with respect to the version without MPI on Haswell and Knights Landing CPUs with different numbers of MPI processes, measured using periodic cubic random catalogues with Ν = 5 × 10⁷ (left) and 5 × 10⁸ (right), L_box = 3 h⁻¹ Gpc, and with 200 linear s bins in [0, 200) h⁻¹ Mpc and 120 μ bins in [0, 1). The dashed lines indicate the theoretical maximum speedup. The left panel shows results without OpenMP, and the right panel presents results with the maximum available numbers of OpenMP threads. SIMD is enabled in all cases.