The scale modifier for hexadecimal floating-point constants must have a positive value. It specifies the number of hexadecimal positions that the fractional portion of the binary representation of a floating-point constant is shifted to the right. The hexadecimal point is assumed to be fixed at the left of the leftmost position in the fractional field. When scaling is specified, it causes an unnormalized hexadecimal fraction to be assembled (unnormalized means that the leftmost positions of the fraction contain hexadecimal zeros). The magnitude of the constant is retained, because the exponent in the characteristic portion of the constant is adjusted upward accordingly. When non-zero hexadecimal positions are lost, rounding occurs in the leftmost hexadecimal position of the lost portion. The rounding is reflected in the rightmost position saved.