DBL_MAX_10_EXP -- maximum int e such that 10**e is representable
sys/float_info.max_10_exp
float_info.max_10_exp