// can move a few zeros into the integer part.
		if exp > 22 {
			f *= float64pow10[exp-22]
			exp = 22
		}
		if f > 1e15 || f < -1e15 {
			// the exponent was really too large.
			return
		}
		return f * float64pow10[exp], true
	case exp < 0 && exp >= -22: // int / 10^k
		return f / float64pow10[-exp], true
	}
	return
}