The four kernels in FTF have been modified to use floating-point operations instead of doubles to improve computation time without affecting the overall performance of the algorithms.
/cc @aporeba @demelian