1634	 bits Are Not Enough for -digit Accuracy	 From the inequality we are likely to conclude that we can represent -digit decimal floating-point numbers accurately by -bit floating-point numbers. However we need significant bits to represent some -digit numbers accurately. In general we can show that if p q- then q significant bits are always enough for p-digit decimal accuracy. Finally we can define a compact -bit floating-point representation that will give significant bits for numbers of practical importance.
