DBL_MAX_EXP -- maximum int e such that radix**(e-1) is representable
sys/float_info.max_exp
float_info.max_exp