Compiler Design

Semantic Analysis
6th Semester B.Tech. (CSE)
Course Code: 18CS1T08

Dr. Jasaswi Prasad Mohanty

School of Computer Engineering
KIIT Deemed to be University
Bhubaneswar, India
 Beyond Scanning and Parsing
 A compiler must do more than just recognize whether a sentence belongs to a
programming language grammar.
• An input program can be grammatically correct but may contain other errors
that prevent compilation.
• Lexer and parser cannot catch all program errors.
 Some language features cannot be modeled using context-free grammar (CFG)
• Whether a variable has been declared before use?
• Parameter types and numbers match in the declaration and use of a function.
• Types match on both sides of an assignment.

CD / Module-IV / Jasaswi 2 21 March 2024

 Limitations with CFGs
 CFGs deal with syntactic categories rather than specific words.
• Grammar can specify the positions in an expression where a variable name
may occur.
 Enforcing the “declare before use” rule requires knowledge that cannot be
encoded in a CFG.
• CFG cannot match one instance of a variable name with another.

CD / Module-IV / Jasaswi 3 21 March 2024

 Questions That Compiler Needs to Answer
 Has a variable been declared?
 What is the type and size of a variable?
 Is the variable scalar or an array?
 Is an array access A[i][j][k] consistent with the declaration?
 Does the name “x” correspond to a variable or a function?
 If x is a function, how many arguments does it take?
 What kind of value, if any, does a function x return?
 Are all invocations of a function consistent with the declaration?
 Track inheritance relationship.
 Ensure that classes and its methods are not multiply defined.

Finding answers to these questions is part of the semantic analysis phase.

CD / Module-IV / Jasaswi 4 21 March 2024
 Checking Dynamic Semantics
 Dynamic semantics of languages need to be checked at run time.
• Whether an overflow will occur during an arithmetic operation?
• Whether array bounds will be exceeded during execution?
• Whether recursion will exceed stack limits?
 Compilers can generate code to check dynamic semantics.

CD / Module-IV / Jasaswi 5 21 March 2024

 How does a compiler answer these questions?
 Compilers track additional information for semantic analysis.
• For example, types of variables, function parameters, and array dimensions.
Type information is stored in the symbol table or the syntax tree.
• Used not only for semantic validation but also for subsequent phases of
compilation.
• The information required may be non-local in some cases.
 Semantic analysis can be performed during parsing or in another pass that
traverses the IR produced by the parser.

CD / Module-IV / Jasaswi 6 21 March 2024

 How does a compiler answer these questions?
 Use formal methods like context-sensitive grammars.
 Use ad-hoc techniques using symbol table.
 Static semantics of PL can be specified using attribute grammars.
 Attribute grammars are extensions of context-free grammars.

CD / Module-IV / Jasaswi 7 21 March 2024

Attribute Grammar Framework
 Syntax-Directed Definition
 A syntax-directed definition (SDD) is a context-free grammar with rules and
attributes.
SDD = CFG + Semantic Rules
 A SDD specifies the values of attributes by associating semantic rules with the
grammar productions.

𝑃 → 𝑃1 + 𝑄 𝑃. 𝑣𝑎𝑙 = 𝑃1 . 𝑣𝑎𝑙 + 𝑄. 𝑣𝑎𝑙

𝑃→𝑄 𝑃. 𝑣𝑎𝑙 = 𝑄. 𝑣𝑎𝑙
 Attributes are associated with grammar symbols and Semantic rules are
associated with production.
 Attributes may be of any kind: numbers, strings, references, datatypes, tables
etc.
CD / Module-IV / Jasaswi 9 21 March 2024
 Syntax-Directed Definition
 Generalization of CFG where each grammar symbol has an associated set of
attributes.
• Let 𝐺 = (𝑇, 𝑁𝑇, 𝑆, 𝑃) be a CFG and let 𝑉 = 𝑇 ∪ 𝑁𝑇
• Every symbol 𝑋 ∈ 𝑉 is associated with a set of attributes (for e.g., denoted by
𝑋. 𝑎 and 𝑋. 𝑏)
• Each attribute takes values from a specified domain (finite or infinite), which is
its type
Typical domains of attributes are, integers, reals, characters, strings,
booleans, and structures,
• New domains can be constructed from given domains by mathematical
operations such as cross product and map
 Values of attributes are computed by semantic rules.

CD / Module-IV / Jasaswi 10 21 March 2024

 Types of Attributes
There are two kinds of attributes for nonterminals:
 A synthesized attribute for a nonterminal A at Example: 𝐴 → 𝐵𝐶𝐷 (Here A is the parent
a parse-tree node N is defined by a semantic node and B, C, D are child nodes)
rule associated with the production at N. Note 𝐴. 𝑆 = 𝐵. 𝑆
that the production must have A as its head. A
synthesized attribute at node N is defined only 𝐴. 𝑆 = 𝐶. 𝑆
in terms of attribute values at the children of N 𝐴. 𝑆 = 𝐷. 𝑆
and at N itself. Here S is the synthesized attribute.
 An inherited attribute for a nonterminal B at a Example: 𝐴 → 𝐵𝐶𝐷 (Here A is the parent
parse-tree node N is defined by a semantic rule node and B, C, D are child nodes)
associated with the production at the parent of
𝐶. 𝑖 = 𝐴. 𝑖
N. Note that the production must have B as a
symbol in its body. An inherited attribute at 𝐶. 𝑖 = 𝐵. 𝑖
node N is defined only in terms of attribute 𝐶. 𝑖 = 𝐷. 𝑖
values at N's parent, N itself, and N's siblings.
Here 𝑖 is the inherited attribute.

CD / Module-IV / Jasaswi 11 21 March 2024

 SDD of a simple desk calculator
 The following SDD is based on the grammar for arithmetic expressions with operators + and *.
 It evaluates expressions terminated by an endmarker 𝒏.
 In the SDD, each of the nonterminals has a single synthesized attribute, called 𝒗𝒂𝒍.
 The terminal 𝒅𝒊𝒈𝒊𝒕 has a synthesized attribute 𝒍𝒆𝒙𝒗𝒂𝒍, which is an integer value returned by the
lexical analyzer.

CD / Module-IV / Jasaswi 12 21 March 2024

 Evaluating an SDD
 The rules of an SDD are applied by first constructing a
parse tree and then using the rules to evaluate all of
the attributes at each of the nodes of the parse tree.
 A parse tree, showing the value(s) of its attribute(s) is
called an annotated parse tree.
 In what order do we evaluate attributes?
• SDDs do not specify any order of evaluation. Annotated parse tree for
• We must evaluate all the attributes upon which the the input string 3 * 5 + 4 n
attribute of a node depends.
• Any topological sort of dependency graph gives a
valid order in which semantic rules must be
evaluated.

CD / Module-IV / Jasaswi 13 21 March 2024

 Evaluating an SDD
 If all attributes are synthesized, (as shown in the table),
then we must evaluate the 𝒗𝒂𝒍 attributes at all of the
children of a node before we can evaluate the 𝒗𝒂𝒍 attribute
at the node itself.
• With synthesized attributes, we can evaluate attributes
in any bottom-up order, such as that of a postorder
traversal of the parse tree.
 For SDD’s with both synthesized and inherited attributes,
there is no guarantee of an order of evaluation existing.
• Consider the nonterminals 𝐴 and B, with synthesized
and inherited attributes A.s and B.i, respectively, along
with the production and rules.
• These rules are circular for which it is impossible to
evaluate either A.s at a node N or B.i at the child of N
without first evaluating the other.

CD / Module-IV / Jasaswi 14 21 March 2024

 Annotated Parse Tree: Example 1
 The annotated parse tree for the input string 3 ∗ 5 + 4 𝑛 is shown below using the
following grammar and rules.
 The values of 𝑙𝑒𝑥𝑣𝑎𝑙 are presumed supplied by the lexical analyzer.
 Each of the nodes for the nonterminals has attribute 𝑣𝑎𝑙 computed in a bottom-up order.

Grammar and Rules Annotated Parse Tree

CD / Module-IV / Jasaswi 15 21 March 2024
 Annotated Parse Tree: Example 2
 The annotated parse tree for the input string 3 ∗ 5 is shown below using the following
grammar and rules.
 Each of the nonterminals 𝑇 and 𝐹 has a synthesized attribute 𝑣𝑎𝑙.
 The terminal digit has a synthesized attribute 𝑙𝑒𝑥𝑣𝑎𝑙.
 The nonterminal 𝑇′ has two attributes: an inherited attribute 𝑖𝑛ℎ and a synthesized
attribute 𝑠𝑦𝑛.

Grammar and Rules Annotated Parse Tree

CD / Module-IV / Jasaswi 16 21 March 2024
 Annotated Parse Tree: Example 3
 Consider the following grammar
for signed binary numbers: Production Attribute Rule

+
 The attributes associated with
each grammar symbols are as
follows:

 The attribute grammar with the

rules of each production is
shown in the side table.
CD / Module-IV / Jasaswi 17 21 March 2024
 Annotated Parse Tree: Example 3
 The annotated parse tree for the input string −101 is shown below:

Production Attribute Rule

CD / Module-IV / Jasaswi 18 21 March 2024

 Dependency Graph
 If an attribute 𝑏 depends on an attribute 𝑐 then the semantic rule for 𝑏 must be
evaluated after the semantic rule for 𝑐.
 The dependencies among the nodes are depicted by a directed graph called
dependency graph.

CD / Module-IV / Jasaswi 19 21 March 2024

 Examples of Dependency Graph

Example 1 Example 2

CD / Module-IV / Jasaswi 20 21 March 2024

 Examples of Dependency Graph

Grammar and Rules Dependency Graph for float 𝑥, 𝑦, z

Example 3
CD / Module-IV / Jasaswi 21 21 March 2024
 Constructing a Dependency Graph

CD / Module-IV / Jasaswi 22 21 March 2024

 Evaluating an SDD
 In what order do we evaluate attributes?
• SDDs do not specify any order of evaluation
• We must evaluate all the attributes upon which the attribute of a node
depends
• Any topological sort of dependency graph gives a valid order in which
semantic rules must be evaluated.
 If there is any cycle in the graph, then there are no topological sorts; that is, there
is no way to evaluate the SDD on this parse tree.
 If there are no cycles, however, then there is always at least one topological sort.
 For SDD’s with both synthesized and inherited attributes, there is no guarantee
of an order of evaluation existing.

CD / Module-IV / Jasaswi 23 21 March 2024

 Examples of Order of Evaluation of attributes

 The dependency graph has

no cycles.
 Few of the topological
sorted orders in which we
may evaluate the attributes
are:
 1,2,3,4,5,6,7,8,9.
 1,3,5,2,4,6,7,8,9

CD / Module-IV / Jasaswi 24 21 March 2024

 Types of SDD
S-attributed SDD L-attributed SDD
 A SDD that uses only synthesized attributes is  A SDD that uses both synthesized and
called S-attributed SDD. inherited attributes is called L-attributed SDD.
 It is also called postfix SDD.  Each inherited attribute is restricted to inherit
 Semantic actions are always placed at right from parent or left siblings only.
end of production.  Semantic actions are placed anywhere in
 Attributes are evaluated with Bottom-up or RHS.
post-order traversal of the parse tree.  Attributes are evaluated by traversing parse
 An S attributed SDD can be implemented tree depth first, left to right parsing manner.
naturally in conjunction with an LR parser.  Example:
 Example: 𝐴 → 𝑋𝑌𝑍
𝐴 → 𝐵𝐶𝐷 Here attribute 𝑌 can only obtain its value
Here attribute 𝐴 can only obtain its value either from the parent 𝐴 or its left sibling 𝑋 but
either from 𝐴, 𝐵, 𝐶 or 𝐷. it can’t inherit from its right sibling 𝑍. Similarly,
𝑋 can only get its value from its parent 𝐴 & 𝑍
can get its value from 𝐴, 𝑋, & 𝑌 as well.
CD / Module-IV / Jasaswi 25 21 March 2024
 S-attributed SDD
Example Evaluation
 An SDD is S-attributed if every attribute is  When an SDD is S-attributed, we can
synthesized. evaluate its attributes in any bottom-up order
of the nodes of the parse tree.
 In the following SDD each attribute, 𝐿. 𝑣𝑎𝑙,
𝐸. 𝑣𝑎𝑙, 𝑇. 𝑣𝑎𝑙, and 𝐹. 𝑣𝑎𝑙 is synthesized.  It is often especially simple to evaluate the
attributes by performing a post-order traversal
Hence, the following SDD is S-attributed. of the parse tree and evaluating the attributes
at a node N when the traversal leaves N for
the last time.

 S-attributed definitions can be implemented

during bottom-up parsing, since a bottom-up
parse corresponds to a post-order traversal.
CD / Module-IV / Jasaswi 26 21 March 2024
 L-attributed SDD
 In a L-attributed SDD, each attribute must be either:  The following SDD is L-attributed
1. Synthesized, or because 𝑇 ′ . 𝑖𝑛ℎ and 𝑇1 ′ . 𝑖𝑛ℎ are
inherited attributes whereas all the
2. Inherited, but with the rules limited as follows. remaining attributes are synthesized
Suppose that there is a production 𝐴 → 𝑋1 𝑋2 … 𝑋𝑛 , attributes.
and that there is an inherited attribute 𝑋𝑖. 𝑎
computed by a rule associated with this production.
Then the rule may use only:
a) Inherited attributes associated with the head 𝐴.
b) Either inherited or synthesized attributes
associated with the occurrences of symbols
𝑋1 𝑋2 … 𝑋𝑖−1 located to the left of 𝑋𝑖.
c) Inherited or synthesized attributes associated  Any SDD containing the following
with this occurrence of 𝑋𝑖 itself, but only in such production and rules cannot be L-
a way that there are no cycles in a dependency attributed:
graph formed by the attributes of this 𝑋𝑖.

CD / Module-IV / Jasaswi 27 21 March 2024

 Notes about Inherited Attributes
 Always possible to rewrite a SDD to use only synthesized attributes.
• Inherited attributes can be simulated with synthesized attributes.
 May be more logical to use both synthesized and inherited attributes.
 Inherited attributes usually cannot be evaluated by a simple preorder traversal of
the parse tree.
• Attributes may depend on both left and right siblings!.
• Attributes that do not depend from right children can be evaluated by a
preorder traversal.

CD / Module-IV / Jasaswi 28 21 March 2024

 Bottom-up Evaluation of S-Attributed Definitions
 Suppose 𝐴 → 𝑋𝑌𝑍, and semantic rule is 𝐴. 𝑎 = 𝑓(𝑋. 𝑥, 𝑌. 𝑦, . 𝑍. 𝑧).
 Can be computed during bottom-up parsing.
• On reduction, value of new synthesized attribute 𝐴. 𝑎 is computed from the
attributes on the stack.
• Extend stack to hold values.

CD / Module-IV / Jasaswi 29 21 March 2024

 Bottom-up Evaluation of S-Attributed Definition

CD / Module-IV / Jasaswi 30 21 March 2024

 Are these SDDs S- or L-attributed?

CD / Module-IV / Jasaswi 31 21 March 2024

Relationship between S-Attributed and L-Attributed Definitions

CD / Module-IV / Jasaswi 32 21 March 2024