FOR cool language semantics

-cs16btech11029,cs16btech11039

=========================================================================

Overview:

Used vistor design pattern for checking semantics
Created and Checked the inheritance Graph
Made 2 passes to the AST for complete type checking

Design Of Semantic Analyzer.

ABOUT VISITOR DESIGN PATTERN :

Visitor design pattern is one of the behavioral design pattern. It is used when we have to perform an operation on a group of similar kind of Objects. With the help of visitor pattern, we can move the operational logic from the objects to another class.
There are 2 interfaces in the program one is ASTnode and ASTvistor .
The Concrete Visitor class implements the methods provided by the ASTvisitor.
There are 3 classes ProgramNode ,CNode , ExpressionNode and FeatureNode which implements the methods provided by the ASTnode class.
The ASTnode interface has a virtual accept function which accepts the object of ASTvisitor.
The Program Node class overrides the accepts function. The program node has an ArrayList of CNode objects which hepls in doing the Depth first search of the Classes present in the program.
The CNode class overrides the accept function. The CNode has ArrayList of Feature Node which helps in doing the Depth first search in the features present in the Class,
At the start of the Semantic analysis an ASTvisitor Object is created which visits each ASTnode starting from the Program Node. Each Class that implements the ASTnode interface accepts this visitor.
There are 4 methods with same name as visit but it's parameters are different :

 interface ASTvisitor {
      public boolean visit(ProgramNode P);
      public boolean visit(CNode C);
      public boolean visit(FeatureNode F);
      public boolean visit(ExpressionNode E);
  }

Each of the visit function is executed depending upon which node is currently being visited.
The visit function when visiting the CNode inserts the class namespace in the Scope Table .
The visit function when visiting the ProgramNode checks the inheritance graph and wether there are multiple Main() functions or not.
The visit function when visiting the FeatureNode checks wether the features in a class are overriden or not . And it also checks if there is redefinition of a feature in the same class. (These checks happen only in the second pass when the Class namespaces for all the classes in the program is created and stored in the Scope Table.)
The visit function when visiting the ExpressionNode does the type checking by calling the getType function.
The type checking is done in the getType function which takes an AST.expression object and depending upon which class it belongs does the necessary type checking

Details of Program.

0.StartPoint

the ProgramNodeis created , then visitor Object of ASTvisitor class is created and program node accepts this visitor using the accept function.

1.Main class

Before building graph itself we will check whether Main class and whether main function is declared in side Main class , and ensure such that there exists only one defination for main() in ProgramNode constructor . Program exits here if there are multiple definations for Main classes printing appropriate error messages.
Also if there is any class that inherits from Bool, Int, String then error is printed and semantic is aborted.

2.Inheritance Graph

when AST is given to Semantic analyser it will intially form Inheritance Graph , in which root will be Object class . Inheritance graph is a Array List of type ClassNode . ClassNode class has 3 attributes self of AST.class_ class , parent of ClassNode class , isVisited of type int . self is the corresponding cool class , parent is ClassNode for the class that self inherited from . if there is no inheritance from any class then parent will be null. isVisted is used to find CYCLES in Inheritance graph . Cycles checks are done as we build graph.if the graph is well=formed we will proceed to next step of semantic analysis.
If any cycle is detected then message is printed and semantic is aborted.

3.For Each Nodes in AST

Using Vistor pattern described above , each Node of AST is visited .
For each Expression node getType function is called which does the entire type checking and also returns the expression Type.
For each feature node various checks is performed like formal parameter validation, overriden methods etc..
For each class node class scope is created in the scope table and inbuilt functions are inserted in the topmost scope.
For each Program node it's inheritance graph is constructed and validated.

4.Scope cum Symbol Table

The scope table is a HashMap with ClassScope object as key and Arraylist of Hash map as value. HashMap<ClassScope,ArrayList<HashMap<String,AST.ASTNode>>>
ClassScope is a Class which has AST.class_ , scope members.

      class ClassScope{
           public AST.class_ cl;
           public int scope;
   
           public ClassScope(AST.class_ c){
              this.cl = c;
              this.scope = 0;
          }
      }

For each class there is a ArrayList maintained . In each index of arraylist a HashMap is maintained. The indices of the Arraylist represent the scope of the program in a particular class. The HashMap has name of features as key and its ASTnode as value .
All the inbuilt functions , features of classes are stored in the outermost scope with scope value 0.
insertClass funtion is used to insert class in Scopetable . this will create a new object adds to existing hashTable.
serchTable function is used to search in Hashtable.iterates over the Hashvalues to match key if key is matched it returns value otherwise it returns null.
enterScope function for entering in a new scope .
exitScope function used for exiting Scope.
lookUpClassSpace function used for finding identifier in the class Scope .
lookUpLocal function used to identify Local Objects.
lookUpGlobal functions used to identify Global Objects.

5.Type Checking

TypeChecking : type checks have been done according to section12 in cool mannual . When type checking results in errors then its StaticType its type is assigned as Object Type and further Typechecking takes place. these checks are mostly done by calling getType() function .
Details in Typechecking
- constants type should be Int or String or Bool .
- identifiers type should be predefined , we can get them accessing ScopeTable . if an identifier is not found it will print error.
- assignment id<-expr : the only condition for assign is that Type of expr should be subtype for type of id .
- Subtype:let say Class A inherits B and B inherits C then :
  - A is subtype of A
  - A is subtype of B
  - A is subtype of C
- we assign static type of new Type as static type of expr if the expr Type is valid i.e if that Class is present otherwise we assign 'new Type' to Object.
- in other expressions such as isvoid,loop their static type are fixed like isvoid type is always Bool , loop type is always Object .
- in Arithmatic,comparison,~ operations we need all operands should be of type Int
- in let expr the type checking will be done between let.value and let.typeid and Static type of let will return the type of let expression body.
- in case expression we will check is there a possibility that type doesn't match with the provided types then a warning is issued , as it is calculated in runtime , and type of case expression will be join of all types of branch expressions . i.e least parent class for all the branches present in case .
- in case of dispatch expressions :
  - <id>(e1,e2....en) : search for method in this and its parent classes if found then its valid expression otherwise its an error
  - <id>(e1,e2...en) ~ self.<id>(e1,e2....en)
  - expr.<id>(e1,e2...en):search for method in typeOf() and its parents class if found then its valid expression otherwise its an error .
  - expr@Type.<id>(e1,e2,...en) : typeOf(expr) should be subtype for Type and id should be present in Type class , other wise its an error

Explaining testcases.

there are 5 test cases that check almost all of the semantics except semantics corresponding to SELF_TYPE .Some of the Sematic checks were
- forming correct inheritance graph
  - error if cycle is found
- no of Main classes and main() functions
  - only One Main class and it should contain main method .error otherwise.
- calling proper functions i.e Dispatches all kinds of function calls are checked in one of the test cases like
  - self@IO.out_string("bla bla ..."); => should not give error
  - self@Int.out_string("........."); => gives error
- gives error when used undeclared identifiers are used
- gives error when inherited from Bool,Int,String . etc..

Compiling Instructions.

Open semantic named folder in :

       make

Open semantic/src/java :

        ./semantic <filename>.cl

FOR cool language CodeGeneration

-cs16btech11029,cs16btech11039

==========================================================================================================

1.Overview .

1. used VisitorDesign Pattern for codegenereration 
2. parsed AST and generated LLVM IR in a Bottom UP approach . 
3 . Runtime checks were done for "Divide by Zero" , "Static Dispatch on void"

2.Design of Codegenerator

ABOUT VISITOR DESIGN PATTERN :

Visitor design pattern is one of the behavioral design pattern. It is used when we have to perform an operation on a group of similar kind of Objects. With the help of visitor pattern, we can move the operational logic from the objects to another class.
There are 2 interface in the Design one is CoolParserVisitor and Visitor. where CoolParserVisitor is provided by default other is added.
The ConcreteVisitor class implements the methods provided by the Visitor
There are 6 methods with same name as visit in Visitorclass but it's parameters are different :

interface Visitor{
	public void visit(AST.program program);
	public void visit(AST.class_ cl);
	public void visit(AST.attr attribute);
	public void visit(AST.method method);
	public void visit(AST.formal params);
	public String visit(AST.expression exp);
}

IRBuildersUtils class contains methods that are utilised in IRBuilder class
IRBuilder generates Instructions according to expression type .
Globals contains all the global variables used while creating instructions i.e hashmaps , Strings , ClassNames etc
ConcreteVisitor class takes care about creating LLVM IR by visiting the program, class and method node. First the visitor visits the program node which sets up all the default strings , functions needed , it also sets up the inheritance graph and fills the scope table .It then creates the struct of the classes and the constructors of the classes . It creates it such that the parent class's struct is created first .
It then for each class in the program it runs visit(class) which for each method visits(method).
In the visit (method) method the alloca and the store instruction for the parameters passed is generated and then it generates the instructions for the body by visiting the body.
when a AST.expr is found visit(expr) generates IR using IRBuilder.generateInstruction . This function recursively creates the needed instructions by using the helper functions present in it's parent class i.e IRBuilderUtils.
all Strings type objects in cool files are stored in global i8* pointers. The strings are obtained in the semantic analysis phase.
All the classes in cool inherit from Object class. So for each class in the cool program the struct will be like :

   %class.<class_name> = type { %class.Object, i8* }
   The first param is the Class Object and second param is a string which 
   stores the class name.

This helps in printing the type_name of the object.
All the strings that are collected by the semantic analyzer are declared as global constants at the top of the .ll file
The constructor of Object class and IO class is there in every .ll file
For division by zero checking, a function named checkDivZero is there in every .ll file for checking the error.
For checking the static dispatch on void if-else branching is used.
The semantic.java class has been modified
The scopeTable.java class has been modified

NOTE : PLEASE USE THE SEMANTIC ANALYZER THAT IS MADE BY US . ALSO PLEASE USE ALL THE FILES THAT WE HAVE GIVEN.

3.Details in CodeGeneration

GOAL : code generation from cool Annoted AST to llvm IR

1.primitive a language has same power(computational) as C++,Java.. if it has there 3 elements .

    Memory 
    Branching instructions 
    Loops / recursions

Intially ,we will Look into Memory aspect of LLVM IR and Cool

MEMORY Cool handles memory using new keyword which allocates that memory and it allows to use variables too . Cool doesn’t distingush between TYPE and a Class i.e each Class is a Type in cool Language . we can declare variables of user defined types or inbuilt classes such as “Int , String , Bool “ . LLVM:We have to see how a dataType is defined in LLVM , and convert COOL type to corresponding w in LLVM a dataType is either a pointer to some dataType , iN (N belongs to natural numbers) and a struct (user defined datatype) classes are converted to corresponding struct Eg : This class in cool
```
Class A {
s  : String <- “this” ;
i  : Int <- 3 ;
c : Bool <- true ;
Fooo : Int {
3 
}
};
```
Is converted to LLVM IR as
```
;The struct for A
%class.A = type { %class.Object, i8*, i8*, i32, i8 }
```
Here the first 2 attributes of all the Structs will be same 1st attribute is %class.Object and 2nd attribute is i8* which points to memory address that stores actual class name , 3rd attribute , 4th attribute , 5 attribute are pointers to respective members in cool Class declaration Cool doesnt contain arrays or pointers or char or structs as we see in other object oriented launguages such as c++ or java etc . But we Cool has strings as continous memory allocation To handle String we have used the same ways as clang used to store c++ strings i.e storing a i8 pointer and allocating a continous memory to it using alloca instruction in llvm . LLVM IR doesnt provide Bool as an seperate datatype we have to use i8 as default for Bool . i.e, when its true i8 stores 1 and for false we have taken 0 .Because as LLVM doesnot provide a direct complement instruction on integers but provides a xor operation which can be used to find complement when value xor,ed with -1 . member functions of all classes are global and those will not be present in struct
BRANCHING LLVM IR basically is modulue , which is a set of BasicBlocks . When there is Branching instruction that means we have to shift to a BasicBlock with out executing other BasicBlock(Instructions) LLVM IR consists 2 ways of branching Conditional branching Unconditional branching But cool provides only Conditional Branching like
```
If exp1 then exp2 else exp3 fi
```
This is converted to a Conditional Braching instruction in LLVM and use unconditional branch instruction to directly jump to next instruction after if else block:
```
...
%cmp = //i1 value that stores 1 or 0 corresponding to true or false based on exp1
br i1 %cmp , label bb1  , label bb2 
bb1 :
    //contains exp1
    br label ifend
bb2:
    //contains exp2
    br label ifend
label ifend :
...
```

LOOPS/RECURSION : LLVM provides recursion as just other Languges provides i.e we can call an llvm ir function with in same function. In case of Loops ,Cool Provides only one way to keep loop i.e

while expr loop expr1 pool

I.e when expr is true then expr1 should run else it should not run ,IN LLVM its converted as following

label condBlock : 
    %val = //conerting expr and storing its value
    %cmp = icmp i1 %val , 1 
    br  i1 %cmp bodyBlock , exitBlock 
label bodyBlock :
    …
    //converting expr1 
    br label condBlock 
label exitBlock : 
…
//has converted expressions after loop in cool

2 . How we handled Other expressions of Cool in LLVM

ID<-expr

 %val = expr value 
 ;finding the pointer of that variable i.e ID 
 %ptr = getelementptr … .
 ;Type can be found by using hashtable
 store Type  %val , %ptr

function calls only static dispatches are handled expr@type.func(params)

    ;if func is a default function provided by llvm use it directly 
    ;eg : length in substr
    ;other wise , generate that functions and using llvm 'call' inst
    %val = generateInstructions(expr)
    mangledname = getMangledname(func,type)
    call mangledname(%val);

* ***Block expressions*** 
    **{ [expr;]+ }**
    ```
        ;generate for each expr in block
    ```
* ***new TYPE*** 
    ```
        type = getTypeid(TYPE)
        %val = alloca type 
        ; if TYPE is primitive(i.e Int , Bool , String) 
        ; then %loadval =  load %val
        ; else %val 
    ```
* ***isvoid expr***
    ```
        ;if expr.type is primitive return false i.e , 0
        ;else
        %val = generateinstuctions for expr
        %ptr = getelementptr ..
        mtype = gettypeid(type,0)
        %ret = icmp ne mtype %ptr , null 
    ```
* ***Arthimatic and comparision expressions in Cool***
    ```
        ;For both arithamatic and comparision LLVM has instructions
        ;eg : in cool lets say i have an expr 2 + x where x is a variable
        ;we convert as
        %x12 = load from %x
        %summ = add i32 2 , %a
    ```
* ***~ expr***
    belongs to AST.neg class => Boolean complement
    ```
        %val = generate instructions for expr
        %rett = xor i8 %val , 1 
    ```
* ***not expr***
    belongs to AST.comp class => 1's complement
    ```
        %val = generate instructions for expr
        %rett = xor i32 %val , -1 
    ```
* ***Bool_const***
    ```
    ; if the value is true then we return "1"
    ; else return "0"
    ; these value as printed in IR 
    ```
* ***Int_const***
    ```
    ; find the value of Int and return it as String 
    ```
* ***String_const***
    ```
    
    ```
* ***ID***
    ```
    
    ```

3.How Functions are generated

The formal parameters of the function are first allocated statck space by using the alloca instruction of the LLVM IR and then the formal parameters are stored in this location using store instruction.
These formal parameters are then loaded using load instruction into a local register.
After this process the generateInstruction method in the IRBuilder class is called to generate instructions for the body of the function.
The generateInstruction method returns the register into which the returned value is stored.
This returned value type is checked with the return type of the function . If it doesn't match then it does BitCast Operation by calling genereateBitCastInst. It is done since the returned value by the generateInstruction is a subtype of the return type of the function (this is checked in the semantic phase).
Then it returns the register.

4.How default functions are handled

The in_int, in_str methods are handled using the scanf call . The instruction generated is same as that generated by Clang on compiling .cpp files.
The out_int , out_str methods are handled using the printf call. The instruction generated is same as that generated by Clang on compiling .cpp files.
The length method is handled using strlen call .
The concat method is handled using strcat call.
The abort method is handled using exit call.
The type_name method is handled by just printing the class name which is present as global string.
The substr by using strncpy function.

4.explaining testcases

arthimatic_binary.cl :

    test verifies :
        all arthimatic operations  
        < , <= , = operators
        assignment expression of primitive types
        Block expression 
        main method ,  Main Class

staticDispatch_inheritance.cl

    test verifies :
        static Dispatch
        inheritance
        assignment of non primitive types
        assignment of child classes object to Parent class Object

nestedIf_loops.cl

    test verifies : 
        nested if else conditions on various expressions 
        checks the formation of blocks in Loops and conditions

5.Compiling Instructions.

Open codegen named folder in :

       make

Open codegen/src/java :

        ./codegen <filename>.cl

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

README.md

README.md

FOR cool language semantics

-cs16btech11029,cs16btech11039

=========================================================================

contents:

1.Overview.

2.Design Of Semantic Analyzer.

3.Details of Program.

4.explaining testcases.

5.Compiling instructions.

Overview:

Design Of Semantic Analyzer.

ABOUT VISITOR DESIGN PATTERN :

Details of Program.

Explaining testcases.

Compiling Instructions.

FOR cool language CodeGeneration

-cs16btech11029,cs16btech11039

==========================================================================================================

contents:

1.Overview.

2.Design Of CodeGenerator.

3.Details in CodeGeneration.

4.explaining testcases.

5.Compiling instructions.

1.Overview .

2.Design of Codegenerator

ABOUT VISITOR DESIGN PATTERN :

NOTE : PLEASE USE THE SEMANTIC ANALYZER THAT IS MADE BY US . ALSO PLEASE USE ALL THE FILES THAT WE HAVE GIVEN.

3.Details in CodeGeneration

GOAL : code generation from cool Annoted AST to llvm IR

4.explaining testcases

5.Compiling Instructions.

Files

README.md

Latest commit

History

README.md

File metadata and controls

FOR cool language semantics

-cs16btech11029,cs16btech11039

=========================================================================

contents:

1.Overview.

2.Design Of Semantic Analyzer.

3.Details of Program.

4.explaining testcases.

5.Compiling instructions.

Overview:

Design Of Semantic Analyzer.

ABOUT VISITOR DESIGN PATTERN :

Details of Program.

Explaining testcases.

Compiling Instructions.

FOR cool language CodeGeneration

-cs16btech11029,cs16btech11039

==========================================================================================================

contents:

1.Overview.

2.Design Of CodeGenerator.

3.Details in CodeGeneration.

4.explaining testcases.

5.Compiling instructions.

1.Overview .

2.Design of Codegenerator

ABOUT VISITOR DESIGN PATTERN :

NOTE : PLEASE USE THE SEMANTIC ANALYZER THAT IS MADE BY US . ALSO PLEASE USE ALL THE FILES THAT WE HAVE GIVEN.

3.Details in CodeGeneration

GOAL : code generation from cool Annoted AST to llvm IR

4.explaining testcases

5.Compiling Instructions.