<k​ayabanerve:matrix.org> 64-input FCMPs being slower than individual FCMPs sounds wrong. They should have the same size MSM yet the former should resolve a bit faster as the MSM has complexity O(n / log(n)).