UNIT-IV

Intermediate Code Generation

Compiler Design- Unit-4

 Contents
•Intermediate Code Generation
•Intermediate Languages - prefix - postfix
•Quadruple - triple - indirect triples Representation
• Syntax tree- Evaluation of expression - three-address code
•Synthesized attributes – Inherited attributes
•Intermediate languages – Declarations
•Assignment Statements
•Boolean Expressions, Case Statements
•Back patching – Procedure calls
•Code Generation
•Issues in the design of code generator
• The target machine – Runtime Storage management
•A simple Code generator
•Code Generation Algorithm
•Register and Address Descriptors
•Generating Code of Assignment Statements
•Cross Compiler – T diagrams
•Issues in Cross compilers
What is intermediate code?
During the translation of a source program into the object code for a target
machine, a compiler may generate a middle-level language code, which is
known as intermediate code
Intermediate code
The following are commonly used intermediate code representation :

▪Syntax tree
▪Postfix Notation
▪Three-Address Code
Syntax tree
•Syntax tree is more than condensed form of a parse tree.
•The operator and keyword nodes of the parse tree are moved
to their parents and a chain of single productions is replaced by
single link in syntax tree.
•The internal nodes are operators and child nodes are
operands.
• To form syntax tree put parentheses in the expression, this way
it's easy to recognize which operand should come first.
Syntax tree

•Example –
x = (a + b * c) / (a – b * c)
Postfix Notation
•The ordinary (infix) way of writing the sum of a and b is with
operator in the middle : a + b
•The postfix notation for the same expression places the operator
at the right end as ab +.
•In general, if e1 and e2 are any postfix expressions, and + is
any binary operator, the result of applying + to the values
denoted by e1 and e2 is postfix notation by e1e2 +.
•No parentheses are needed in postfix notation because the
position and arity (number of arguments) of the operators permit
only one way to decode a postfix expression.
•In postfix notation the operator follows the operand.
Postfix Notation

•Example – The postfix representation of the expression

(a – b) * (c + d) + (a – b) is

ab – cd + ab -+*.
 Three-Address Code
•A statement involving no more than three references(two for operands and
one for result) is known as three address statement.
•A sequence of three address statements is known as three address code.
Three address statement is of the form x = y op z , here x, y, z will have
address (memory location).
•Sometimes a statement might contain less than three references but it is still
called three address statement.
•For Example :a = b + c * d;
• The intermediate code generator will try to divide this expression into
sub-expressions and then generate the corresponding code.
r1 = c * d;
r2 = b + r1;
a = r2
Three-Address Code
• A three-address code has at most three address locations to
calculate the expression. A three- address code can be
represented in three forms :
o Quadruples
o Triples
o Indirect Triples
Quadruples
• Each instruction in quadruples presentation is divided into four
fields: operator, arg1, arg2, and result. The example is
represented below in quadruples format:

OP arg1 arg2 result

* c d r1
+ b r1 r2
= r2 a
Triples
•Each instruction in triples presentation has three fields : op,
arg1, and arg2.The results of respective sub-expressions are
denoted by the position of expression.

OP arg1 arg2
(0) * c d
(1) + b (0)
(2) = (1)
Indirect Triples
•This representation is an enhancement over triples
representation.
•It uses pointers instead of position to store results.
• This enables the optimizers to freely re-position the
sub-expression to produce an optimized code.

OP arg1 arg2
35 (0) * c d
36 (1) + b (0)
37 (2) = (1)
Type Expressions
Example: int[2][3]
array(2,array(3,integer))

• A basic type is a type expression

• A type name is a type expression
• A type expression can be formed by applying the array type constructor to a
number and a type expression.
• A record is a data structure with named field
• A type expression can be formed by using the type constructor → for
function types
• If s and t are type expressions, then their Cartesian product s*t is a type
expression
• Type expressions may contain variables whose values are type expressions
Type Equivalence

• They are the same basic type.

• They are formed by applying the same constructor to structurally
equivalent types.
• One is a type name that denotes the other.
Declarations
Storage Layout for Local Names
• Computing types and their widths
Storage Layout for Local Names
⚫ Syntax-directed translation of array types
Sequences of Declarations
•

• Actions at the end:

•
Fields in Records and Classes
•

•
Translation of Expressions and Statements
• We discussed how to find the types and offset of variables
• We have therefore necessary preparations to discuss about
translation to intermediate code
• We also discuss the type checking
Three-address code for expressions
Incremental Translation
Addressing Array Elements
• Layouts for a two-dimensional array:
Semantic actions for array reference
Translation of Array References

Nonterminal L has three synthesized attributes:

•L.addr
•L.array
•L.type
Conversions between primitive types in Java
Introducing type conversions into
expression evaluation
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (3)

Semantic Analysis
• meaning
Semantic Analysis computes additional information related to the
of the program once the syntactic structure is known.
• In typed languages as C, semantic analysis involves adding information to
the symbol table and performing type checking.
• The information to be computed is beyond the capabilities of standard
parsing techniques, therefore it is not regarded as syntax.
• As for Lexical and Syntax analysis, also for Semantic Analysis we need both
a Representation Formalism and an Implementation Mechanism.
•
As representation formalism this lecture illustrates what are called Syntax
Directed Translations.
 (4)

Syntax Directed Translation: Intro

• The Principle of Syntax Directed Translation states that the meaning of an

input sentence is related to its syntactic structure, i.e., to its Parse-Tree.

• By Syntax Directed Translations we indicate those formalisms for specify-

ing translations for programming language constructs guided by context-free

grammars.
– We associate Attributes to the grammar symbols representing the
language constructs.
– Values for attributes are computed by Semantic Rules associated with
grammar productions.
 (5)

Syntax Directed Translation: Intro (Cont.)

• Evaluation of Semantic Rules may:

– Generate Code;
– Insert information into the Symbol Table;
– Perform Semantic Check;
– Issue error messages;
– etc.
• There are two notations for attaching semantic rules:

1. Syntax Directed Definitions. High-level specification hiding many

implementation details (also called Attribute Grammars).
the order
2. Translation Schemes. More implementation oriented:
Indicate in which semantic rules are to be evaluated.
 (6)

Summary
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (7)

Syntax Directed Definitions

• Syntax Directed Definitions are a generalization of context-free grammars
in which:
1. Grammar symbols have an associated set of Attributes;
2. Productions are associated with Semantic Rules for computing the values of
attributes.
• Such formalism generates Annotated Parse-Trees where each node of the
tree is a record with a field for each attribute (e.g., X.a indicates the attribute
a of the grammar symbol X ).
 (8)

Syntax Directed Definitions (Cont.)

• The value of an attribute of a grammar symbol at a given parse-tree node is
defined by a semantic rule associated with the production used at that node.

• We distinguish between two kinds of attributes:

1. Synthesized Attributes. They are computed from the values of the

attributes of the children nodes.
2. Inherited Attributes. They are computed from the values of the attributes
of both the siblings and the parent nodes.
 (9)

Form of Syntax Directed Definitions

• Each production, A → α, is associated with a set of semantic rules:

b := f (c1 , c2 , . . . , ck ), where f is a function and either

1. b is a synthesized attribute of A, and c1 , c2 , . . . , ck are attributes of the

grammar symbols of the production, or
2. b is an inherited attribute of a grammar symbol in α, and c , c , . . . , c
1 2 k
are attributes of grammar symbols in α or attributes of A.
• Note. Terminal symbols are assumed to have synthesized attributes supplied
by the lexical analyzer.
• Procedure calls (e.g. print in the next slide) define values of Dummy
synthesized attributes of the non terminal on the left-hand side of the
production.
 (10)

Syntax Directed Definitions: An Example

• Example. Let us consider the Grammar for arithmetic expressions. The
Syntax Directed Definition associates to each non terminal a synthesized
attribute
called val.
SE M A N T I C RU L
PRO D U C T ION
E
→ En
L
→ E print(E.val)
E +T E.val := E .val + T .val
1 1
E → T E.val := T .val
→ ∗ := T .val ∗ F.val
T T1 F T .val 1
→ T .val := F.val
T F
→ (E) := E.val
F F.val
→ digit :=digit.lexval
F F.val
 S-Attributed Definitions (11)

Definition. An S-Attributed Definition is a Syntax

Directed Definition that uses
only synthesized attributes.

• Evaluation Order. Semantic rules in a S-Attributed

Definition can be
Definition. The annotated parse-tree
evaluated for the
by a bottom-up, or PostOrder,
input 3*5+4n is:traversal of the
parse-tree. L
• Example. 19
E.valThe
= above arithmetic
n grammar is an example
of an S-Attributed
= +15
E.val T .val = 4
= 15
T .val F.val = 4
=3 =5 digit.lexval= 4
T .val * F.val
=3 digit.lexval= 5
F.val

digit.lexval= 3
 (12)

Inherited Attributes
• Inherited Attributes are useful for expressing the dependence of a construct
on the context in which it appears.
• It is always possible to rewrite a syntax directed definition to use only
synthesized attributes, but it is often more natural to use both synthesized and
inherited attributes.
• Evaluation Order. Inherited attributes cannot be evaluated by a simple
PreOrder traversal of the parse-tree:
– Unlike synthesized attributes, the order in which the inherited attributes
of the children are computed is important!!! Indeed:
∗ Inherited attributes of the children can depend from both left and right
siblings!
 (13)

Inherited Attributes: An Example

• Example. Let us consider the syntax directed definition with both inherited
and synthesized attributes for the grammar for “type declarations”:
PRO D U C T I SE M A N T I C RU L E
ON
L.in := T .type
D→ TL
→int
T T .type :=integer
T →real T .type :=real
L → L1 , id L1 .in := L.in; addtype(id.entry, L.in)
L → id addtype(id.entry, L.in)
• The non terminal T has
keyword in the declaration.a synthesized attribute, type, determined by the

• The production D → T L is associated with the semantic rule L.in :=

T .type which set the inherited attribute L.in.
• Note: The production L → L , id distinguishes the two occurrences of L.
1
 (14)

Inherited Attributes: An Example (Cont.)

• Synthesized attributes can be evaluated by a PostOrder traversal.

• Inherited attributes that do not depend from right children can be evaluated
by a classical PreOrder traversal.
• The annotated parse-tree for the input real id , id , id is:
1 2 3
D

T .type = real L.in

, id
real L.in 3
, id
L.in 2
id
1
 Inherited Attributes: An Example (Cont.) (14)

• Synthesized attributes can be evaluated by a PostOrder traversal.

• Inherited attributes that do not depend from right children can be evaluated

by a classical PreOrder traversal.

• The annotated parse-tree for the input real id1 , id2, id3 is:
D
L.in= real
T .type = real
, id
real L.in 3
, id
L.in 2
id
1
 (14)

Inherited Attributes: An Example (Cont.)

• Synthesized attributes can be evaluated by a PostOrder traversal.

• Inherited attributes that do not depend from right children can be evaluated

by a classical PreOrder traversal.

• The annotated parse-tree for the input real id1 , id2, id3 is:
D
L.in= real
T .type = real
real , id
real L.in= 3
, id
L.in 2
id
1
 (14)

Inherited Attributes: An Example (Cont.)

• Synthesized attributes can be evaluated by a PostOrder traversal.

• Inherited attributes that do not depend from right children can be evaluated

by a classical PreOrder traversal.

• The annotated parse-tree for the input real id1 , id2, id3 is:
D
L.in= real
T .type = real
real , id
real L.in= 3
, id
L.in= real 2
id
1
• L.in is then inherited top-down the tree
by the other L-nodes.
• At each L-node the procedure addtype inserts into the symbol table the type
of the identifier.
 (15)

Summary
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (16)

Dependency Graphs
• Implementing a Syntax Directed Definition consists primarily in finding an
order for the evaluation of attributes
– Each attribute value must be available when a computation is performed.
• Dependency Graphs are the most general technique used to evaluate syntax
directed definitions with both synthesized and inherited attributes.
• A Dependency Graph shows the interdependencies among the attributes of
the various nodes of a parse-tree.
– There is a node for each attribute;
– If attribute b depends on an attribute c there is a link from the node for c
to the node for b (b ← c).

• Dependency Rule: If an attribute b depends from an attribute c, then we

need to fire the semantic rule for c first and then the semantic rule for b.
 (17)

Evaluation Order
• The evaluation order of semantic rules depends from a Topological
Sort
derived from the dependency graph.
• Topological Sort: Any ordering m , m , . . . , m such that if m → m
1 2 k i j
is a link in the dependency graph then mi < mj .

• Any topological sort of a dependency graph gives a valid order to evaluate

the semantic rules.
 (18)

Dependency Graphs: An Example

• Example. Build the dependency graph for the parse-tree of real id , id ,
1 2
id .
3
 (19)

Implementing Attribute Evaluation: General Remarks

• Attributes can be evaluated by building a dependency graph at compile-time
and then finding a topological sort.
• Disavantages

1. This method fails if the dependency graph has a cycle: We need a test for
non-circularity;
2. This method is time consuming due to the construction of the dependency
graph.
• Alternative Approach. Design the syntax directed definition in such a
way that attributes can be evaluated with a fixed order avoiding to build the
dependency graph (method followed by many compilers).
 (20)

Strongly Non-Circular Syntax Directed Definitions

• Strongly Non-Circular Syntax Directed Definitions. Formalisms for
which an attribute evaluation order can be fixed at compiler construction
time.
– They form a class that is less general than the class of non-circular
definitions.
– In the following we illustrate two kinds of strictly non-circular definitions:
S-Attributed and L-Attributed Definitions.
 (21)

Summary
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (22)

Evaluation of S-Attributed Definitions

• Synthesized Attributes can be evaluated by a bottom-up parser
as the
input is being analyzed avoiding the construction of a dependency graph.
• The parser keeps the values of the synthesized attributes in its stack.

• Whenever a reduction A → α is made, the attribute for A is computed from

the attributes of α which appear on the stack.
• Thus, a translator for an S-Attributed Definition can be simply implemented
by extending the stack of an LR-Parser.
 (23)

Extending a Parser Stack

• Extra fields are added to the stack to hold the values of synthesized attributes.

• In the simple case of just one attribute per grammar symbol the stack has two
fields: state and val
state val

Z Z.x
Y Y .x

X X.x
...
...
• The current top of the stack is indicated by the pointer top.

• Synthesized attributes are computed just before each reduction:

– Before the reduction A → X Y Z is made, the attribute for A is
computed: A.a := f (val[top], val[top − 1], val[top − 2]).
 (24)

Extending a Parser Stack: An Example expres-

• Example. Consider the S-attributed definitions for the
sions. To evaluate attributes the parser executes the following code
arithmetic
PRO D U C T I CO D E
ON
→ En print(val[top − 1])
L
→ E := + val[top − 2]
E +T val[ntop] val[top]
1
E → T
→ T ∗F := ∗ val[top − 2]
T val[ntop] val[top]
1
→ F
T
→ (E) := val[top − 1]
val[ntop]
F
→ digit
F
• The variable ntop is set
to the new top of the stack. After a reduction is done
top is set to ntop: When a reduction A → α is done with |α| = r, then
ntop = top − r + 1.
• During a shift action both the token and its value are pushed into the stack.
 (25)

Extending a Parser Stack: An Example (Cont.)

• The following Figure shows the moves made by the parser on input 3*5+4n.

– Stack states are replaced by their corresponding grammar symbol;

– Instead of the token digit the actual value is shown.
 (26)

Summary
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (27)

L-Attributed Definitions
• L-Attributed Definitions contain both synthesized and inherited attributes
but do not need to build a dependency graph to evaluate them.
• Definition. A syntax directed definition is L-Attributed if each inherited

attribute of Xj in a production A → X1 . . . Xj . . . Xn, depends only on:

1. The attributes of the symbols to the left (this is what L in L-Attributed

stands for) of X , i.e., X X . . . X , and
j 1 2 j −1
2. The inherited attributes of A.
• Theorem. Inherited attributes in L-Attributed Definitions can be computed
by a PreOrder traversal of the parse-tree.
 (28)

Evaluating L-Attributed Definitions

• L-Attributed Definitions are a class of syntax directed definitions whose
attributes can always be evaluated by single traversal of the parse-tree.
• The following procedure evaluate L-Attributed Definitions by mixing
PostOrder (synthesized) and PreOrder (inherited) traversal.
Algorithm: L-Eval(n: Node)
Input: Node of an annotated parse-tree.
Output: Attribute evaluation.
Begin
For each child m of n, from left-to-right Do
Begin
Evaluate inherited attributes of m;
L-Eval(m)
End;
Evaluate synthesized attributes of n
End.
 (29)

Summary
• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
 (30)

Translation Schemes
• Translation Schemes are more implementation oriented than syntax directed
definitions since they indicate the order in which semantic rules and attributes
are to be evaluated.
• Definition. A Translation Scheme is a context-free grammar in which

1. Attributes are associated with grammar symbols;

2. Semantic Actions are enclosed between braces {} and are inserted within
the right-hand side of productions.
• Yacc uses Translation Schemes.
 (31)

Translation Schemes (Cont.)

• Translation Schemes deal with both synthesized and inherited attributes.

• Semantic Actions are treated as terminal symbols: Annotated parse-trees

contain semantic actions as children of the node standing for the correspond- ing
production.
• Translation Schemes are useful to evaluate L-Attributed definitions at parsing
time (even if they are a general mechanism).
– An L-Attributed Syntax-Directed Definition can be turned into a
Translation Scheme.
 (32)

Translation Schemes: An Example

• Consider the Translation Scheme for the L-Attributed Definition for “type

declarations”:
D→ T {L.in := T .type} L
→
T int {T .type :=integer}
T → real {T .type :=real}

L → {L1 .in := L.in} L1 , id {addtype(id.entry, L.in)}

L → id {addtype(id.entry, L.in)}
 (33)

Translation Schemes: An Example (Cont.)

• Example (Cont). The parse-tree with semantic actions for the input

real id1, id2, id3 is:

D
{L.in := T .type}
T L

{T.type := real} {L .in := L.in} L , id {addtype(id .entry, L.in)}

real 1 1 3 3

{L .in := L .in} L , id {addtype(id .entry, L .in)}

2 1 2 2 2 1

id {addtype(id .entry, L .in)}

1 1 2

• Traversing the Parse-Tree in depth-first order (PostOrder) we can

evaluate the attributes.
 (34)

Design of Translation Schemes

• When designing a Translation Scheme we must be sure that an attribute value
is available when a semantic action is executed.
• When the semantic action involves only synthesized attributes: The
action can be put at the end of the production.
– Example. The following Production and Semantic Rule:
T →T ∗F ∗ F.val
T .val := T .val
1 1
yield the translation scheme:
T →T ∗F {T .val := T .val ∗ F.val}
1 1
 (35)

Design of Translation Schemes (Cont.)

• Rules for Implementing L-Attributed SDD’s. If we have an L-Attibuted
Syntax-Directed Definition we must enforce the following restrictions:
1. An inherited attribute for a symbol in the right-hand side of a production
must be computed in an action before the symbol;
2. A synthesized attribute for the non terminal on the left-hand side can only be
computed when all the attributes it references have been computed: The action is
usually put at the end of the production.
 (36)

Compile-Time Evaluation of Translation Schemes

• Attributes in a Translation Scheme following the above rules can be computed
at compile time similarly to the evaluation of S-Attributed Definitions.
• Main Idea. Starting from a Translation Scheme (with embedded actions) we

introduce a transformation that makes all the actions occur at the right ends
of their productions.
– For each embedded semantic action we introduce a new Marker (i.e., a
non terminal, say M ) with an empty production (M → ǫ);

– The semantic action is attached at the end of the production M → ǫ.

 (37)

Compile-Time Evaluation of Translation Schemes (Cont.)

• Example. Consider the following translation scheme:
→ aA{C.i = f (A.s)}C
S
S → bAB{C.i = f (A.s)}C
C → c{C.s = g(C.i)}
Then, we add new markers M , M with:
1 2
S → aAM C
1
S → bABM C
2
M →ǫ {M .s := f (val[top])}
1 1
M →ǫ {M .s := f (val[top − 1])}
2 2
C →c {C.s := g(val[top − 1])}

The attribute of C is the synthesized attribute of either M or

The value of C.i is always 1
inherited M : in val[top -1] when C → c is applied.
2
 (38)

Compile-Time Evaluation of Translation Schemes (Cont.)

General rules to compute translations schemes during bottom-up parsing assum-
ing an L-attributed grammar.

• For every production A → X . . . X introduce n new markers

1 n
M , . . . , M and replace the production by A → M X . . . M X .
1 n 1 1 n n
• Thus, we know the position of every synthesized and inherited attribute of

Xj and A:
1. Xj .s is stored in the val entry in the parser stack associated with Xj ;
2. Xj .i is stored in the val entry in the parser stack associated with Mj ;
3. A.i is stored in the val entry in the parser stack immediately before the
position storing M1 .
•
Remark 1. Since there is only one production for each marker a grammar

• remains LL(1) with addition of markers.

Remark 2. Adding markers to an LR(1) Grammar can introduce conflicts for
not L-Attributed SDD’s!!!
 Compile-Time Evaluation of Translation Schemes (Cont.) (39)

Example. Computing the inherited attribute X .i after reducing with M → ǫ.

j j

Mj Xj .i
top→ X X X .s
j −1 j −1 • A.i is in val[top − 2j + 2];
j −1
M Xj −1.i
j −1 • X .i is in val[top − 2j + 3];
... ... 1
• X .s is in val[top − 2j + 4];
1
X1 X1 .s
• X .i is in val[top − 2j + 5];
2
M1 X1 .i
• and so on.
(top-2j+2)→
MA A.i

(top-2j)→
 (40)

Summary

• Syntax Directed Translations

• Syntax Directed Definitions

• Implementing Syntax Directed Definitions

– Dependency Graphs
– S-Attributed Definitions
– L-Attributed Definitions
• Translation Schemes
Boolean Expression
• The translation of if-else-statements and while-statements is tied to
the translation of boolean expressions.
• In programming languages, boolean expressions are used to
• 1. Alter the flow of control.
• 2. Compute logical values.
Boolean Expression
1. Alter the flow of control.
▪ Boolean expressions are used as conditional expressions in statements that
alter the flow of control.
▪ The value of such boolean expressions is implicit in a position reached in a
program.
For example, in if (E) S, the expression E must be true if statement S is reached.
•2. Compute logical values.
▪ A boolean expression can represent true or false as values.
▪ Boolean expressions can be evaluated in analogy to arithmetic expressions
using three-address instructions with logical operators.
Boolean Expression
• The intended use of boolean expressions is determined by its
syntactic context.
• For example, an expression following the keyword if is used to alter
the flow of control, while an expression on the right side of an
assignment is used to denote a logical value.
• Such syntactic contexts can be specified in a number of ways:
• we may use two different nonterminals, use inherited attributes, or set a flag
during parsing.
• Alternatively we may build a syntax tree and invoke different procedures for
the two different uses of boolean expressions.
Boolean expressions
• Focus on the use of boolean expressions to alter the flow
of control
• Boolean expressions are composed of the boolean operators
• && - AND, I I - OR, and ! - NOT
• applied to elements that are boolean variables or relational expressions.
• Relational expressions are of the form El re1 E2, where El and E2 are arithmetic
expressions
• Grammar to generate Boolean expression:
B → B1 or B2 | B → B1 and B2 | B → not B1 | B → (B1)|B → E1 relop E2 |
B → false| B → true
Boolean expressions
• Semantic definition of the programming language determines
• whether all parts of a boolean expression must be evaluated.
• If the language definition permits (or requires) portions of a boolean
expression to go unevaluated
• the compiler can optimize the evaluation of boolean expressions by computing only
enough of an expression to determine its value.
• In an expression such as B1 I I B2, neither B1 nor B2 is necessarily
evaluated fully.
• If we determine that B1 is true, then we can conclude that the entire expression is
true without having to evaluate B2.
• Similarly, given B1&&B2, if B1 is false, then the entire expression is false.
Short-Circuit Code
• In short-circuit (or jumping) code, the boolean operators &&, I I , and !
translate into jumps.
• The operators themselves do not appear in the code
• Instead, the value of a boolean expression is represented by a position
in the code sequence.
Short-Circuit Code
• The statement

might be translated into the code

• In this translation, the Boolean expression is true if control reaches

label L2.
• If the expression is false, control goes immediately to L1, skipping L2
and the assignment x = 0.
Flow-of-Control Statements
• Translation of boolean expressions into three-address code in the
context of statements generated using the following grammar
S -> if (B) S1
S -> if ( B ) S1 else S2
S -> while ( B ) S1
• Non terminal B represents a boolean expression and nonterminal S
represents a statement
Boolean Expression
• Synthesized attributes associated with Boolean expression and Statement
• B.code ,S.code- gives the translation into three-address instructions
• Inherited attributes associated with Boolean expression and Statement
• B.true - label to which control flows if B is true
• B.false - label to which control flows if B is false
• S.next - label attached to the first 3-address statement to be executed immediately
after the code for S.
• Functions used in Syntax directed definitions
• newlabel() creates a new label each time it is called
• label(L) attaches label L to the next three-address instruction to be
generated
Syntax directed definition for P  S

• P-> S : Program P consist of statement S

• S.next is initialized to to a new label
• P.code consists of S.code followed by the new label S.next
• assign is a placeholder for assignment statements
Syntax directed definition for S  S1S2

• S1 followed by the code for S2

• The first instruction after the code for S1 is the beginning of the code for S2
• The instruction after the code for S2 is also the instruction after the code for S.
Syntax directed definition for Boolean Expression S -> if (B) S1

Syntax-directed definition

• Create a new label B.true and attach it to the first three-address instruction generated for the
statement S1
• Within B. code are jumps based on the value of B.
• If B is true, control flows to the first instruction of S1 .code, and
• If B is false, control flows to the instruction immediately following S1.code.
• By setting B.false to S.next, we ensure that control will skip the code for S1 if B evaluates to false.
 Syntax directed definition for Boolean Expression
S -> if ( B ) S1 else S2
Syntax-directed definition

• Create two labels B.true and B.false

• If B is true, B jumps to the first instruction of the code for S1
• If B is false, jumps to the first instruction of the code for S2
• Control flows from both S1 and S2 to the three-address instruction immediately following the code for S - its label
is given by the inherited attribute S.next.
• An explicit goto S.next appears after the code for S1 to skip over the code for S2.
• No goto is needed after S2, since S2. next is the same as S. next.
Boolean Expression S  while (B) S1
Syntax-directed definition

• begin – local variable that holds a new label attached to the first instruction for the while-statement, also the first
instruction for B.
• begin- is a variable rather than an attribute, because begin is local to the semantic rules for this production.
• S.next - marks the instruction that control must flow to if B is false; hence, B. false is set to be S.next
• B. true - Code for B generates a jump to this label if B is true and attached to the first instruction for S1
• goto begin - causes a jump back to the beginning of the code for the boolean expression.
• S1 .next is set to this label begin, so jumps from within S1. code can go directly to begin
Control-Flow Translation of Boolean
Expressions
• Boolean expression B is translated into three-address instructions
• B is evaluated using conditional and unconditional jumps to one of
two labels:
• B.true - if B is true
• B.fa1se - if B is false
 Control-Flow Translation of Boolean
Expressions

• If B1 is true, then B itself is true, so B1.true is the same as B.true.

• If B1 is false, then B2 must be evaluated, so make B1.false be the label of the first instruction in the code for
B2.
• The true and false exits of B2 are the same as the true and false exits of B, respectively.
Control-Flow Translation of Boolean Expressions

• If B1 is true, then B2 must be evaluated so make B1.true is assigned as the label

of the first instruction in the code for B2
• If B1 is false, then B itself is false, so B1.false is the same as B.false
• If B2 is true, then B itself is true, so B2.true is the same as B.true
• If B2 is false, then B itself is false, so B2.false is the same as B.false
Control-Flow Translation of Boolean Expressions

• No code is needed for an expression B of the form ! B1

• Interchange the true and false exits of B to get the true and false exits of B1.
Control-Flow Translation of Boolean Expressions

• Translated directly into a comparison three-address instruction with jumps to

the appropriate places
• For instance, B of the form a < b translates into
Control-Flow Translation of Boolean
Expressions

The constants true and false translate into jumps to B.true and
B.false, respectively
 Control-Flow Translation of Boolean Expressions -Example
S
Consider the
expression

S
S.nex

L1:
B S1

if B.conditon goo B.rue

goo B.false

L2: x = 0 S1.code
L1:
 B1 B2

If B1.conditon goo L2
goo L3
B1.rue = B.rue = L2
L3:If B2.conditon goo L2
B1.false = L3
goo L1
B2.rue = B.rue = L2
L2: x = 0 B2.false = B.false = L1
B.rue
L1:
B.false
 B1 B2

If x< 100 goo L2

goo L3
L3: If B2.conditon goo L2
goo L1
L2: x=0
L1:
 B1 B2

If X < 100 goo L2

goo L3
L3: If B2.conditon goo L2 B1.rue = L4
goo L1 B1.false = B.false = L1
B2.rue = B.rue = L2
L2: x=0 B2.false = B.false = L1
L1:
 B1 B2

Control-Flow Translation of the above statement is

B1.rue = L4
B1.false = B.false = L1
B2.rue = B.rue = L2
B2.false = B.false = L1

If B1.conditon goo L4
goo L1
L4: If B2.conditon goo L2
goo L1
Avoiding Redundant Gotos
if x > 200 goto L4
• Translated code of Boolean
goto L1 expressions are not
optimized ones
L4: … • Redundant gotos are noticed
 • Can be avoided using
fallthrough
ifFalse x > 200 goto L1
L4: …
(fall through)
 Using Fall Through
We now adapt the semantic rules for boolean expressions to
allow control to fall through whenever possible.
S → if (E) S1
{E.true = fall; // not newlabel; E.false =
S.next;
S1.next = S.next;
S.code = E.code || S1.code }

Similarly, the rules for if-else- and while-

statements also set E.true to fall.
 Using Fall Through Cont’d
E → E1 || E2
{E1.true = if (E.true = fall) then newlabel() else E.true; E1.false
= fall;
E2.true = E.true; E2.false = E.false;
E.code = if (E.true = fall) then E1.code ||
E2.code ||
label(E1.true)
else E1.code || E2.code }

E → E1 && E2
{E1.true = fall;
E1.false = if (E.false = fall) then newlabel() else E.false;
E2.true = E.true; E2.false = E.false;
E.code = if (E.false = fall) then E1.code ||
E2.code ||
label(E1.false)
else E1.code || E2.code }
 Using Fall Through Cont’d
E → E1 relop E2
{test = E1.addr relop E2.addr
s = if E.true != fall and E.false != fall then
gen(‘if’ test ‘goto’, E.true) || gen(‘goto’, E.false) else if (E.true
!= fall) then gen(‘if’ test ‘goto’, E.true)
else if (E.false != fall) then gen(‘ifFalse’ test ‘goto’, E.false)
else ‘’
E.code := E1.code || E2 .code || s
}
 Using Fall Through Example
if (X < 100 || x > 200 && x != y) x =
0;
=>
if x < 100 goto L2 ifFalse x > 200
goto L1 ifFalse x!=y goto L1
L2: x = 0 L1:
 Problems to be solved

• Translate the following expressions with and without fall through

• If (a==b && c==d || e==f) x==1;
• If (a==b || c==d || e==f) x==1;
• If (a==b && c==d && e==f) x==1;
• If a < b or c < d and e < f
• while a < b do
if c<d then
x=y+z
else
x=y-z
 Backpatching

• A key problem when generating code for boolean expressions and

flow-of-control statements is that of matching a jump instruction
with the target of the jump.
• For example, the translation of the boolean expression B in if ( B ) S
contains a jump, for when B is false, to the instruction following the
code for S.
• In a one-pass translation, B must be translated before S is examined.
• What then is the target of the goto that jumps over the code for S?
 Backpatching

• For the examples of the previous lectures for implementing

syntax-directed definitions, the easiest way is to use two passes.
• First syntax tree is constructed
• Then traversed in depth-first order to compute the translations given in the
definition.
• Problem in generating three address codes in a single pass for
Boolean expressions and flow of control statements is :
• We may not know the labels that control must go to at the time jump
statements are generated.
 Backpatching

• This problem is solved by generating a series of branch statements

with the targets of the jumps temporarily left unspecified.
• Each such statement will be put on a list of goto statements whose
labels will be filled in when the proper label can be determined.
• This subsequent filling of addresses for the determined labels is
called BACKPATCHING.
 Backpatching

• For implementing Backpatching

• Instructions are generated into an instruction array
• Labels act as indices to this array.
• To manipulate list of labels , three functions are used:
• makelist(i)
• merge(p1,p2) and
• backpatch(p,i)
 Backpatching

• makelist(i) : creates a new list containing only i,

an index into the array of instructions and
returns pointer to the list it has made.
• merge(i,j) – concatenates the lists pointed to by
i and j ,and returns a pointer to the
concatenated list.
• backpatch(p,i) – inserts i as the target label for
each of the statements on the list pointed to by
p.
 Backpatching

• Let’s now try to construct the translation scheme for Boolean expression.
• Let the grammar be:
B → B1 or MB2
B → B1 and MB2
B → not B1
B → (B1)
B → id1 relop id2
B → false
B → true
M→ε
 Backpatching

• What we have done is inserting a marker non-terminal M into the

grammar to cause a semantic action to pick up, at appropriate times
the index of the next instructionto be generated.
• This is done by the semantic action:
{ M.instr = nextinstr; } for the rule M → ε
 Backpatching

• Two synthesized attributes truelist and falselist of non-terminal B are

used to generate jumping code for Boolean expressions.
• B.truelist : Contains the list of all the jump statements left incomplete
to be filled by the label for the start of the code for B=true.
• B.falselist : Contains the list of all the jump statements left incomplete
to be filled by the label for the start of the code for B=false.
 Backpatching

• The variable nextinstr holds the index of the next instruction to

follow.
• This value will be backpatched onto B1.truelist in case of B → B1 and
MB2 where it contains the address of the first statement of B2.code.
• This value will be backpatched onto B1.falselist in case of B → B1 or
MB2 where it contains the address of the first statement of B2.code.
 Backpatching-Boolean expressions
• We use the following semantic actions for the above grammar :
• 1) B → B1 or M B2
backpatch(B1.falselist, M.instr)
B.truelist = merge(B1.truelist, B2.truelist)
B.falselist = B2.falselist

• If B1 is true, then B is also true, so the jumps on B1. truelist become part of B.truelist.
• If B1 is false, we must next test B2, so the target for the jumps B1.falselist must be the beginning of the code
generated for B2.
• This target is obtained using the marker nonterminal M.
• M produces, as a synthesized attribute M.instr, the index of the next instruction, just before B2 code starts
being generated
• The value M.instr will be backpatched onto the B1 .falselist (i.e., each instruction on the list B1. falselist will
receive M.instr as its target label) when we have seen the remainder of the production B -> B1 or M B2.
 Backpatching -Boolean expressions

2)B → B1 and M B2
backpatch(B1.truelist, M.instr)
B.truelist = B2.truelist
B.falselist = merge(B1.falselist, B2.falselist)

• If B1 is true, we must next test B2, so the target for the jumps B1. truelist must be the beginning of the code
generated for B2.
• This target is obtained using the marker nonterminal M.instr , the index of the next instruction, just before B2
code starts being generated
• The value M.instr will be backpatched onto the B1.truelist.
• If B1 is false, then B is also false, so the jumps on B1. falselist become part of B.falselist.
 Backpatching -Boolean expressions
Swaps the true
3) B → not B1 and false lists

B.truelist = B1.falselist
B.falselist = B1.truelist
4) B → (B1) Ignores
parenthesis
B.truelist = B1.truelist
B.falselist = B1.falselist • generates two instructions, a
conditional goto and an
5) B → id1 relop id2 unconditional one.
B.truelist = makelist(nextinstr) • Both gotos have unfilled targets
B.falselist = makelist(nextinstr+1 ) • These instructions are put on
B.truelist and B.falselist,
emit(if id1.place relop id2.place goto __ ) respectively
emit(goto ___)
 Backpatching-Boolean expressions

6) B → true
B.truelist = makelist(nextinstr)
emit(goto ___)
7) B → false
E.falselist = makelist(nextinstr)
emit(goto ___)
8) M → ε
M.instr = nextinstr;
 Backpatching-Boolean expressions Example
Expression of of the
B form B → B1 or M B2
Consider the
expression

B1
B2

B → B1 or M B2
backpatch(B1.falselist, M.instr)
B.truelist = merge(B1.truelist, B2.truelist)
B.falselist = B2.falselist
 Backpatching-Boolean expressions Example
B

B1
B2

B → id1 relop id2

B.truelist = makelist(nextinstr)
B.falselist = makelist(nextinstr+1 )
emit(if id1.place relop id2.place goto __ )
emit(goto ___)
Backpatching-Boolean expressions Example
Expression of of the
B form B → B1 or M B2

B1
B2
The marker nonterminal M in the production records the value
of nextinstr, which at this time is 102
B → B1 or M B2
backpatch(B1.falselist, M.instr)
B.truelist = merge(B1.truelist, B2.truelist)
B.falselist = B2.falselist
 Backpatching-Boolean expressions Example
B

B1 B2

B → id1 relop id2

B.truelist = makelist(nextinstr)
B.falselist = makelist(nextinstr+1 )
emit(if id1.place relop id2.place goto __ )
emit(goto ___)
Backpatching-Boolean expressions Example
B

B1 B2
The marker nonterminal M in the
production records the value of
2)B → B1 and M B2 nextinstr, which at this time is 104
backpatch(B1.truelist, M.instr)
B.truelist = B2.truelist
B.falselist = merge(B1.falselist, B2.falselist)
 Backpatching-Boolean expressions Example
B

B1
B2

B → id1 relop id2

B.truelist = makelist(nextinstr)
B.falselist = makelist(nextinstr+1 )
emit(if id1.place relop id2.place goto __ )
emit(goto ___)
 Backpatching -Boolean expressions Example
B

B1 B2

B1.truelist =102
M.instr=104
backpatch(102,104) – instruction
2)B → B1 and M B2
in 102 is filled with 104
backpatch(B1.truelist, M.instr)
B.truelist = B2.truelist
B.falselist = merge(B1.falselist, B2.falselist)
 Backpatching-Boolean expressions Example
Expression of of the
B form B → B1 or M B2

B2
B1
B1.falselist =101
M.instr=102
backpatch(101,102) – instruction
B → B1 or M B2 in 101 is filled with 102
backpatch(B1.falselist, M.instr)
B.truelist = merge(B1.truelist, B2.truelist)
B.falselist = B2.falselist
 Backpatching-Boolean expressions Example
B

The entire expression is true if and

only if the gotos of instructions 100
or 104 are reached

Expression is false if and

only if the gotos of
instructions 103 or 105 are
reached.
Backpatching-Boolean expressions Example
Annotated parse tree for
 Backpatching-Flow of control

• Backpatching can be used to translate flow-of-control statements in

one pass.
• Consider statements generated by the following grammar:

•S denotes a statement
• L denotes statement list
• B denotes boolean expression
Backpatching-Flow of control
• Boolean expressions generated by nonterminal B have two lists of
jumps
• B. truelist -Contains the list of all the jump statements left incomplete to be
filled by the label for the start of the code for B=true.
• B.falselist-Contains the list of all the jump statements left incomplete to be
filled by the label for the start of the code for B=false.
• Statements generated by nonterminals S and L have a list of unfilled
jumps
• Eventually filled by backpatching.
• S.nextlist - list of all conditional and unconditional jumps to the instruction
following the code for statement S in execution order.
• L.nextlist - list of all conditional and unconditional jumps to the instruction
following the code for statement L in execution order
Backpatching-Translation of Flow of control
statements
Backpatching-Flow of control

• In the above production

• Two occurrences of the marker nonterminal M
• Used to record the instruction numbers of the beginning of the code for B and the beginning of the code for
S1
• For the rule M → ε , semantic action is { M.instr = nextinstr; }
• After the body S1 of the while-statement is executed, control flows to the beginning.
• When the entire production is reduced to S, S1.nextlist is backpatched to make all targets on that list be
M1.instr.
• An explicit jump to the beginning of the code for B is appended after the code for S1 because control may also
"fall out the bottom."
• B.truelist is backpatched to go to the beginning of S1 by making jumps on B. truelist go to M2 . instr.
 Backpatching-Flow of control

• If control "falls out the bottom" of S1, as when S1 is an assignment

• code has to be included at the end of S1 to jump to S2
• A new nonterminal N with production N → ε is introduced for this purpose
• N has attribute N.nextlist, which will be a list consisting of the instruction
number of the jump goto - that is generated by the semantic action (7) for N.
• When B is true, jumps are backpatched to the instruction M1.instr;
• Similarly, when B is false jumps are backpatched to the beginning of the code for
S2.
• S.nextlist includes all jumps out of S1 and S2, as well as the jump generated by N.
• temp is a temporary variable that is used only for merging lists.
Backpatching-Flow of control

• Above two productions handle sequences of statements.

• In L -> L1 M S
• Instruction following the code for L1 in order of execution is the beginning of S.
• Thus the L1 .nextlist list is backpatched to the beginning of the code for S, which
is given by M. instr.
• In L -> S , L. nextlist is the same as S.nextlist
 Backpatching-Flow of control
• No new instructions are generated anywhere in these semantic
rules, except for rules (3) and (7).
• All other code is generated by the semantic actions associated with
assignment-statements and expressions.
• The flow of control causes the proper backpatching so that the
assignments and Boolean expression evaluations will connect
properly.
 Labels and Goto Statements
• Labels are stored in the symbol table (and associated with intermediate
language labels).
• Generated as soon as a jump or a declaration is met (to avoid one
additional pass)
• For “goto L”, we have to change L into the address of the
three-address code for the statement where L is attached.

• When L’s address has been found, we can do this easily with the
information in the symbol table.

• Otherwise, we have to use backpatching. We keep the list of

address to be filled in the symbol table
 Break and Continue Statements
• Break/exit: pass an additional (inherited) attribute to the translation function of
loops with the label a break/exit should jump to.
• A new label is passed when entering a new loop
• A break-statement is a jump to the first instruction after the code for the
enclosing statement S.
– keep track of the enclosing statement S
– generate an unfilled jump for the break- statement
– put the unfilled jump on S.nextlist
– backpatch when the exit place is known
Break and Continue Statements
• Continue statements :
• can be handled in a manner analogous to the break statement.
• The main difference between the two is that the target of the
generated jump is different
Translation of a switch-statement
• The "switch" or "case" statement is available in a variety of languages.
• switch-statement syntax is shown below .

• There is a selector expression E, which is to be evaluated

• Followed by n constant values Vl , V2, . . - , Vn that the expression
might take
• Has a default "value," which always matches the expression if no
other value does.
Translation of a switch-statement
• The intended translation of a switch is code to:
• 1. Evaluate the expression E.
• 2. Find the value V, in the list of cases that is the same as the value of
the expression. Recall that the default value matches the expression if
none of the values explicitly mentioned in cases does.
• 3. Execute the statement Sj associated with the value found.
Translation of a switch-statement
• Step (2) is an n-way branch, which can be implemented in one of
several ways.
• If the number of cases is small (say 10 at most)
• use a sequence of conditional jumps
• each jump tests for an individual value and transfers to the code for the
corresponding statement.
• Way to implement this sequence of conditional jumps is to create a table of
pairs
• each pair consisting of a value and a label for the corresponding
statement's code.
• The value of the expression itself, paired with the label for the default
statement is placed at the end of the table at run time.
• A simple loop generated by the compiler compares the value of the
expression with each value in the table
• if no other match is found, the last (default) entry is sure to match
Translation of a switch-statement
• If the number of values exceeds 10
• Create hash table for the values, with the labels of the various statements as
entries.
• If no entry for the value possessed by the switch expression is found, a jump
to the default statement is generated
• Special case that can be implemented even more efficiently than by
an n-way branch.
• If the values all lie in range, say min to max
• number of different values is a reasonable fraction of max - min
• then we can construct an array of max- min "buckets," where bucket j - min
-

contains the label of the statement with value j

• any bucket that would otherwise remain unfilled contains the default label.
Translation of a switch-statement
• To perform the switch, evaluate the expression to obtain the value j;
• Check that it is in the range min to max
• transfer indirectly to the table entry at offset j - min.
• For example, if the expression is of type character, a table of, say, 128 entries
(depending on the character set) may be created and transferred through
with no range testing
Syntax-Directed Translation of Switch-Statements
• A convenient translation of the switch statement is

• All tests appear at the end

• A simple code generator can
recognize the multiway branch and
generate efficient code for it
Syntax-Directed Translation of Switch-Statements
• When the keyword switch is noticed
• generate two new labels test and next, and a new temporary t.
• parse the expression E and generate code to evaluate E into t.
• After processing E, generate the jump goto test .
• When the keyword case is noticed
• create a new label Li and enter it into the symbol table.
• Each value-label pair consisting of the value V, of the case constant and Li (or
a pointer to the symbol-table entry for Li) is placed in a queue
• We process each statement case V, : S by emitting the label Li attached to
i
the code for S followed by the jump goto next.
i
Syntax-Directed Translation of
Switch-Statements
• When the end of the switch is found
• generate the code for the n-way branch.
• Read the queue of value-label pairs
• generate a sequence of three-address statements of the form shown below.

• t is the temporary holding the value of the selector expression E

• Ln, is the label for the default statement
Syntax-Directed Translation of Switch-Statements
• Another form of translation of switch case

• Compiler should emit code for each of the statements Si as it notices.

• Hence it is tedious for one-pass compiler to place the branching statements
at the beginning
• This form of translation is not effective
Procedure Calls
• Must generate the calling sequence and returning sequence.
• Can be done either at the calling or at the called procedure – depending
on machine and OS.
Procedure Calls
S → call id ( Elist )
Elist → Elist , E
Elist → E

• Calling sequence
• allocate space for activation record
• evaluate arguments
• establish environment pointers
• save status and return address
• jump to the beginning of the procedure

146
Procedure Calls …
Example

• parameters are passed by reference

• storage is statically allocated

• use param statement as place holder for the arguments

• called procedure is passed a pointer to the first parameter

• pointers to any argument can be obtained by using proper

offsets

147
Code Generation for procedure calls
• Generate three address code needed to evaluate arguments which are
expressions

• Generate a list of param three address statements

• Store arguments in a list

S → call id ( Elist )
for each item p on queue do emit('param' p)
emit('call' id.place)

Elist → Elist , E
append E.place to the end of queue

Elist → E
initialize queue to contain E.place

148
Code Generation
 Code Generation

Requirements Challenges
•Preserve semantic •Problem of
meaning of source generating
program optimal target
•Make effective program is
use of available undecidable
resources of •Many
target machine subproblems
•Code generator encountered in
itself must run code generation
efficiently are
computationally
Main Tasks of Code Generator
• Instruction selection: choosing
appropriate target-machine instructions
to implement the IR statements
• Registers allocation and assignment:
deciding what values to keep in which
registers
• Instruction ordering: deciding in what
order to schedule the execution of
instructions
Issues in the Design of a Code Generator
•Input to the Code
Generator
•The Target Program
•Instruction Selection
•Register Allocation
•Evaluation Order
Input to the Code Generator
• The input to the code generator is the intermediate representation of the
source program

Three-address representations quadruples, triples, indirect triples;

Virtual machine representations Bytecodes and stack-machine code
Linear representations Postfix notation
Graphical representations Syntax trees and DAG's

• Here it is assume that all syntactic and static semantic errors have been detected,
that the necessary type checking has taken place, and that type conversion operators
have been inserted wherever necessary.
• The code generator can therefore proceed on the assumption that its input is free of
these kinds of errors.
The Target Program
• The instruction-set architecture of the target machine has a
significant impact on the difficulty of constructing a good code
generator that produces high-quality machine code.
• The most common target-machine architectures are
• RISC (reduced instruction set computer)
• It has many registers, three-address instructions, simple addressing modes, and a
relatively simple instruction-set architecture
• CISC (complex instruction set computer),
• It has few registers, two-address instructions, a variety of addressing modes,
several
register classes, variable-length instructions, and instructions with side effects
• Stack based
• operations are done by pushing operands onto a stack and then performing the
operations on the operands at the top of the stack
The Target Program (Cont)
• To overcome the high performance penalty of interpretation, just-in-time
(JIT) Java compilers have been created.
• These JIT compilers translate bytecodes during run time to the native
hardware instruction set of the target machine
• Producing an absolute machine-language program as output has the
advantage that it can be placed in a fixed location in memory and
immediately executed
• Producing a relocatable machine-language program (often called an
object module) as output allows subprograms to be compiled separately. A
set of relocatable object modules can be linked together and loaded for
execution by a linking loader.
Instruction Selection
• The code generator must map the IR program into a code sequence
that can be executed by the target machine.
• The complexity of performing this mapping is determined by a
factors such as
• The level of the IR
• The nature of the instruction-set architecture
• The desired quality of the generated code.
Instruction Selection
• If the IR is high level, the code generator may translate each IR
statement into a sequence of machine instructions using code
templates. Such statement by-statement code generation, however,
often produces poor code that needs
• If the IR reflects some of the low-level details of the underlying
machine, then the code generator can use this information to generate
more efficient code sequences.
• The quality of the generated code is usually determined by its speed
and
size.
• On most machines, a given IR program can be implemented by many
different code sequences, with significant cost differences between the
different implementations
Register Allocation
• A key problem in code generation is deciding what values to hold in what registers.
• Registers are the fastest computational unit on the target machine, but we usually do
not have enough of them to hold all values.
• Values not held in registers need to reside in memory.
• Instructions involving register operands are invariably shorter and faster than those
involving operands in memory, so efficient utilization of registers is particularly
important.
• The use of registers is often subdivided into two sub-problems:
• Register allocation, during which we select the set of variables that will reside in registers at
each
point in the program.
• Register assignment, during which we pick the specific register that a variable will reside in.
• Finding an optimal assignment of registers to variables is difficult, even with single-
register machines.
• Mathematically, the problem is NP-complete.
• The problem is further complicated because the hardware and/or the operating system
of the target machine may require that certain register-usage conventions be observed.
Evaluation Order
• The order in which computations are performed can affect the
efficiency of the target code.
• Some computation orders require fewer registers to hold
intermediate results than others.
• However, picking a best order in the general case is a difficult NP-
complete problem.
• Initially, we shall avoid the problem by generating code for the
three- address statements in the order in which they have been
produced by the intermediate code generator.

Prepared by R I Minu
 Simple Code Generator
Content..
✔Introduction(Simple Code Generator)
✔Register and Address Descriptors
✔A Code-Generation Algorithm
✔The Function getreg
✔Generating Code for Other Types of Statements
 Introduction
• Code Generator:
✔ Code generator can be considered as the final phase of
compilation.
✔ It generates a target code for a sequence of three-address
statement.
✔ It consider each statement in turn, remembering if any of the
operands of the statement are currently in registers, and
taking advantage of that fact if possible.
 Introduction
• We assume that computed results canbe left in
registers an long as possible.
• Storing them only
a) if their register is needed for another
computation (or)
b) Just before a procedurecall, jump, or labeled
statement
• Condition (b)implies that everything must be stored just
before the end of a basic block.
 Introduction
• The reason we must do so is that, after leaving a basic block, we may be
able to go to several different blocks, or we may go to one particular block
that can be reached from several others.
• In either case, we cannot, without extra effort, assume that a datum used by
a block appears in the same register no matter how control reached that
block.
• Thus, to avoid a possible error, our simple code- generator algorithm stores
everything when moving across basic block boundaries as well as when
procedure calls are made.
 Introduction
• We canproduce reasonablecode for a three- address
statement a:=b+c
• If we generate the single instruction ADD Rj,Ri

• with cost one

• If Ri contain b but c is in a memory location, we can generate
the sequenceAdd c,Ri cost=2
(or)
Mov c,Rj cost=3
Add Rj, Ri
Register and Address Descriptors
• The code generator has to track both the registers (for
availability) and addresses (location of values) while
generating the code. For both of them, the following two
descriptors are used:
• Register descriptor
• Address descriptor
 Register descriptor
• Register descriptor is used to inform the code generator about
the availability of registers. Register descriptor keeps track of
values stored in each register. Whenever a new register is
required during code generation, this descriptor is consulted
for register availability.

– Keep rack of wha is currenly in each regiser.

– Initally all he regisers are empy
 Address descriptor
• Values of the names (identifiers) used in the program might be stored at
different locations while in execution. Address descriptors are used to keep
track of memory locations where the values of identifiers are stored.
These locations may include CPU registers, heaps, stacks, memory or a
combination of the mentioned locations.

– Keep rack of locaton where curren value of he name can be found a runtme

– The locaton migh be a regiser, sack, memory address or a se of hose

 A Code-Generation Algorithm
• Basic blocks comprise of a sequence of three- address
instructions. Code generator takes these sequence of
instructions as input.
• For each three-address statement of the form
x:=y op z
we perform the following actions
 I. Call function getReg, to decide the location of L.
II. Determine the present location (register or memory) of y by
consulting the Address Descriptor of y. If y is not presently
in register L, then generate the following instruction to
copy the value of y to L
MOV y’, L
where y’ represents the copied value of y.
III. Determine the present location of z using the same method
used in step 2 for y and generate the following instruction:
OP z’, L
where z’ represents the copied value of z.
IV. Now L contains the value of y OP z, that is intended to be
assigned to x. So, if L is a register, update its descriptor to
indicate that it contains the value of x. Update the descriptor
of x to indicate that it is stored at location L.

If y and z has no further use, they can be given back to the

system.
 The Function getreg
• Code generator uses getReg function to determine the status
of available registers and the location of name values. getReg
works as follows:
• If variable Y is already in register R, it uses that register.
• Else if some register R is available, it uses that register.
• Else if both the above options are not possible, it chooses a
register that requires minimal number of load and store
instructions.
 The Function getreg
1. If Y is in register (that holds no other values) and Y is not
live and has no next use after
X = Y op Z
then return register of Y for L.

2. Failing (1) return an empty register

3. Failing (2) if X has a next use in the block or op

requires register then get a register R, store its content into M (by
Mov R, M) and use it.

4. else select memory location X as L

 Example
• For example, the assignment d := (a - b) + (a -
c) + (a - c) might be translated into the following
three-address code sequence:
• t1 = a – b
• t2 = a – c
• t3 = t 1 + t 2
• d = t3 + t2
 Example
Stmt code reg desc addr desc
t1=a-b mov a,R0 R0 contains t1 t1 in R0
sub b,R0
t2=a-c mov a,R1 R0 contains t1 t1 in R0
sub c,R1 R1 contains t2 t2 in R1
t3=t1+t2 add R1,R0 R0 contains t3 t3 in R0
R1 contains t2 t2 in R1
d=t3+t2 add R1,R0 R0 contains d d in R0
mov R0,d d in R0 and
memory
Generating Code for Other Types
of Statements
• Indexed assignment
a:=b[i] and
a[i]:=b
• Pointer assignment
a:=*p and
*p:=a
Cross Compilation
◼ A cross compiler is a compiler capable of creating
executable code for a platform other than the one on which
the compiler is run.

◼ Cross compiler tools are used to generate executables

for embedded system or multiple platforms.

◼ It is used to compile for a platform upon which it is not

feasible to do the compiling, like microcontrollers that don't
support an operating system.
 Uses of cross compilers
• The fundamental use of a cross compiler is to separate the build
environment from target environment. This is useful in a number of
situations:
• Embedded computers where a device has extremely limited resources.
• For example, a microwave oven will have an extremely small computer to
read its touchpad and door sensor, provide output to a digital display and
speaker, and to control the machinery for cooking food.
• This computer will not be powerful enough to run a compiler, a file
system, or a development environment.
• Since debugging and testing may also require more resources than
are available on an embedded system, cross-compilation can be
less involved and less prone to errors than native compilation.
 •Compiling for multiple machines

• For example, a company may wish to support several

different versions of an operating system or to support
several different operating systems.
• By using a cross compiler, a single build environment can be
set up to compile for each of these targets.
• Compiling on a server farm.
• Similar to compiling for multiple machines, a complicated
build that involves many compile operations can be executed
across any machine that is free, regardless of its underlying
hardware or the operating system version that it is running.
 •GCC and cross compilation

• GCC, a free software collection of compilers, can be

set up to cross compile. It supports many platforms
and languages.
T-diagrams
• Different diagrams for different kinds of programs
• Visual explanation of interactions involving compilers
and interpreters
Programs
• Program P written in • Example: Sort program
language L written in Java

P sort
L Java
Machines
• Machine executing • Example: Sun workstation
language M executing SPARC machine
code

M sparc
Executing Programs
• Program implementation language must match
machine

sort
sparc

sparc
Interpreters
• Interpreter executing • Example: Lisp interpreter
language L written in running on sparc
language M

L Lisp
M sparc
Interpreting Programs
• Interpreter mediates between programming language
and machine language

sort
Lisp

Lisp
sparc

sparc
Virtual Machines
• Interpreter creates a “virtual machine”

sort sort
Lisp Lisp

Lisp
sparc Lisp

sparc
Compilers
• Compiler translating • Example: C compiler for
from source language S sparc platform
to target language T
implemented in M
S --> T C sparc
✇
M sparc
Compiling Programs
• Compiler inputs program in source language, outputs
in target language

sort sort sort

C C sparc sparc sparc
✇
sparc sparc
sparc
Compiler Correctness
• Compiled target program has the same meaning as
the source program

sort sort
C == sparc

C sparc
Java Programming Environment
• Javac: Java to Java byte code (JBC) compiler
• Java: Java Virtual Machine byte code interpreter

P P P
Java Java JBC JBC JBC
✇
M JBC
“javac” M “java” M

M
Where Do Compilers Come From?

1. Write it in machine code

L M
✇
M

A lot of work
Where Do Compilers Come From?

1. Write it in machine code

2. Write it in a lower level language and compile it
using an existing compiler

L M L M
✇ ✇
C C M M
✇
M

But where did the C compiler come

from?
Where Do Compilers Come From?

1. Write it in machine code

2. Write it in a lower level language and compile it
using an existing compiler
3. Write it in the same language that it compiles and
bootstrap

L M
✇
L
Bootstrapping a Compiler
• Write the compiler in its own language (#0)
• Write a simple native compiler (#1)
• Use compiler #1 to compile #0 to get native compiler
with more frills (#2)
• Repeat as desired
Bootstrapping a Compiler

L M L M
✇ ✇
#0 (the real L L --> M M #2 (real thing
thing) M compiled)

#1 (simple)
Bootstrapping a Compiler, Stage 2

L M L M
✇ ✇
#0 (the real L L M M #3 (compiled with
✇
thing) the real thing)
M

#2 (the real thing compiled)

Consistency test: #2 = #3 literally

Porting a Compiler
1. Rewrite back end to target new machine
L M L N
✇ ✇
C C

2. Compile on new machine

Porting a Compiler
• 1. Rewrite back end to target new machine

• 2. Compile on new machine

L N L N
✇ ✇
C C N N
✇
N
Porting a Compiler II
• Rewrite back end to target new machine
• Compile using native compiler

L N L N
✇ ✇
L L M M “cross
✇
compiler”
M
Cross Compilers
• A cross compiler compiles to a target language
different from the language of the machine it runs on
Porting a Compiler II
• Rewrite back end to target new machine
• Compile using native compiler
• Recompile using cross compiler

L N L N
✇ ✇
L L N N
✇
M