Compiler-Intermediate Code Generation & Code Optimization

• Three Address Code
• Control Sequence
• Variable Declaration

---

Intermediate Code

Intermediate code separates the front end and the back end. It is an intermediate level between

• a high level parse tree and
• low level microprocessor instructions.

It is internal to compilers. Different compilers may have different intermediate codes.

Possible types of intermediate code:

• Syntax trees
• Static single assignment
• Three address code

Three Address Code

Each 3AC instruction consists of at most

• one operator (binary or unary) and three address (variable number)

For example,

• $x=yopz$ (binary operator)
• $x=opy$ (unary operator)
• $x=y$ (copy/ assignment)
• $gotoL$ (unconditional $goto$, $L$ is the label of the destination)
• $ifxrelationa{l}_{o}pythengotoL$ (conditional goto)

To convert a parse tree to 3AC, we need two synthesized attributes for each node in the parse tree.

Suppose we have a grammar production $E\to \cdots$

• $E.name$ stores the name of the variable containing the value of $E$.
• $E.code$ stores the sequence of 3AC that computes the value if $E$.

The function $newtemp()$ is used when the temporary variable names for intermediate results are generated.

Example

Define the attributes for the following grammar

• $S\to id=E$
• $E1\to E2+E3|E2\ast E3|-E2|(E2)|id$

$S\to id=E$

$$s.code:=E.code||id.name=E.name$$

• "||" means "followed by"
• No $S.name$ because $id=E$ is an assignment, not computing a value.

$E1\to E2+E3$

• $E1.name:=newtemp()$
• $E1.code:=E2.code||E3.code||E1.name=E2.name+E3.name$

$E1\to E2\ast E3$

• $E1.name:=newtemp()$
• $E1.code:=E2.code||E3.code||E1.name=E2.name\ast E3.name$

$E1\to -E2$

• $E1.name:=E2.name$
• $E1.code:=E2.code||E1.name=-E2.name$

$E1\to (E2.name)$

• $E1.name:=E2.name$
• $E1.code:=E2.code$

$E1\to id$

• $E1.name:=id.name$ (intrinsic)
• $E1.code:=""$ (nothing to compute)

Control Sequence

Variable Declaration

Loop

If Statement