Inside Macintosh: PowerPC Numerics / Part 1 - The PowerPC Numerics Environment

Chapter 3 - Expression Evaluation

This chapter describes the ways in which an expression with floating-point operations can be evaluated in the PowerPC Numerics environment. The environment does not require that all floating-point operations be performed with a certain precision. Instead, it lets each implementation choose the most efficient precision to use. An implementation can dictate that all floating-point operations be performed with a given precision, or an implementation may define a method by which the best possible precision is chosen for each expression. This chapter describes the two methods that numeric implementations can use to choose a precision and compares the methods using several examples.

You should read this chapter to learn how PowerPC Numerics determines the precision of a floating-point expression.

---

Chapter Contents

About Expression Evaluation

Evaluating Expressions Without Widest Need

Evaluating Expressions With Widest Need

Comparisons of Expression Evaluation Methods