9dfbf896a8
When diabling 64-bit fastmath using the flag NO_TFM_64BIT, the sizes of fp_digit and fp_word get smaller. Using them in math with an int gives incorrect values. Changed the fp_cmp_mag_ct to return a fp_digit since its return value is used with an fp_digit. Compare its result against a FP_LT cast as a fp_digit.