Execution time of the scalar and vectorised versions of FCFC on different CPUs (bars) for the full pair-counting procedure with 200 linear s bins in [0, 200) h⁻¹ Mpc and 120 μ bins in [0,1), run upon periodic random samples in a cubic box with a side length of 3 h⁻¹ Mpc. The purple lines indicate the speedups of the SIMD-parallelised versions with respect to the scalar counterparts. AVX2* indicates AVX2, but without the gather instructions.