CS39003 - Compilers Laboratory - Assignment - 6 - Target Code Generator - tinyC - Solved

1           Preamble – tinyC
The Lexical Grammar (Assignment 3) and the Phase Structure Grammar (Assignment 4) for tinyC have already been defined as subsets of the C language specification from the International Standard ISO/IEC 9899:1999 (E). Finally, three address code (TAC) structure and a further subset of tinyC has been specified (Assignment 5) for translating the input tinyC program to TAC quad array, a supporting symbol table, and other auxiliary data structures.

In this assignment you will write a target code translator from the TAC quad array (with the supporting symbol table, and other auxiliary data structures) to the assembly language of x86-64. The translation is now machine-specific and your generated assembly code would be translated with the gcc assembler to produce the final executable codes for the tinyC program.

2           Scope of Target Translation
•   For simplicity restrict tinyC further:

1.    Skip shift and bit operators.

2.    Support only void, int, and char types. Skip double type.

3.    Support only one–dimensional arrays.

4.    Support only void, int, char, void*, int*, and char* types for returns types of functions.

5.    No type conversion to be supported.

•   For I/O, provide a library (as created in Assignment 2) using in-line assembly language program of x86-64 along with syscall for gcc assembler.:

–    int printStr(char *) – prints a string of characters. The parameter is terminated by ‘\0’. The return value is the number of characters printed.

–    int printInt(int n) – prints the integer value of n (no newline). It returns the number of characters printed.

–    int readInt(int *eP) – reads an integer (signed) and returns it. The parameter is for error (ERR = 1, OK = 0).

The header file myl.h of the library will be as follows:

#ifndef _MYL_H

#define _MYL_H

#define ERR 1 #define OK 0 int printStr(char *); int printInt(int);

int readInt(int *eP); // *eP is for error, if the input is not an integer

#endif

3           Design of the Translator
The steps for target code generation were outlined in Target Code Generation lecture presentations. In this assignment, however, you do not need to deal with any machine-independent or machine-specific optimization. Hence the translation comprises the following major steps only:

Memory Binding This deals with the design of the allocation schema of variables (including parameters and constants) that associates each variable to the respective address expression or register. This needs to handle the following:

•   Handle local variables, parameters, and return value for a function. These are automatic and reside in the Activation Record (AR) of the function. Various design schema for AR are possible based on the calling sequence protocol. A sample AR design could be as follows:

Offset
Stack Item
Responsibility
–ve
Saved Registers
Callee Saves & Restores
–ve
Callee Local Data
Callee defines and uses
0
Base Pointer of Caller
Callee Saves & Restores
Return Address
Saved by call, used by ret
+ve
Return Value
Callee writes, Caller reads
+ve
Parameters
Caller writes, Callee reads
Activation Record Structure with Management Protocol

–    Offset’s in the AR are with respect to the Base Pointer of Callee.

–    Return Value can alternatively be returned through a register (like accumulator).

–    The AR will be populated from the Symbol Table of the function.

–    Symbol Tables of nested blocks will be flattened and its variables allocated within the Symbol Table (and hence the AR) of the function where there occur in. Necessary name mangling will be performed to to take care of same lexical name for different variables in different nested scopes.

•   Handle global variables (note that local static variables are not allowed in tinyC) as static and generate allocations in static area. This will be populated from global symbol table (ST.gbl).

•   Generate Constants from Table of Constants – handle string constants as assembler symbols in DATA SEGMENT and integer constants as parts of target code (TEXT SEGMENT)

•   Register Allocations & Assignment: Create memory binding for variables in registers:

–    After a load / store the variable on the activation record and the register have identical values

–    Registers can be used to store temporary computed values

–    Register allocations are often used to pass int or pointer parameters

–    Register allocations are often used to return int or pointer values

Note: Refer to Run-Time Environment lecture presentations for details and examples on memory binding.

Code Translation This deals with the translation of 3–Address quad’s to x86-64 assembly code. This needs to handle:

•   Generation of Function Prologue – few lines of code at the beginning of a function, which prepare the stack and registers for use within the function.

•   Generate Function Epilogue – appears at the end of the function, and restores the stack and registers to the state they were in before the function was called.

•   Map 3–Address Code to Assembly – to translate the function body do:

–    Choose optimized assembly instructions for every expression, assignment and control quad.

–    Use algebraic simplification & reduction of strength for choice of assembly instructions from a quad.

–    Use Machine Idioms (like inc for i++ or ++i in place of add reg, 1).

Note: Refer to Target Code Generation lecture presentations for details.

Target Code Integrate all the above code into an Assembly File for gcc assembler.

4           The Assignment
1.    Write a target code (x86-64) translator from the 3-Address quad’s generated from the flex and bison specifications of tinyC (with restrictions as mentioned in Section 2). Assume that the input tinyC file is lexically, syntactically, and semantically correct. Hence no error handling and / or recovery is expected.

2.    Prepare a Makefile to compile and test the project.

3.    Prepare test input files ass6 roll test<number>.c to test the target code translation and generate the translation output in ass6 roll <number>.asm.

4.    Name your files as follows:

File
Naming
Flex Specification
ass6 roll.l
Bison Specification
ass6 roll.y
Data Structures (Class Definitions) and

Global Function Prototypes
ass6 roll translator.h
Data Structures, Function Implementations and 3–Address Translator
ass6 roll translator.cxx
Target Translator and x86-64 Translator main()
ass6 roll target translator.cxx
Test Inputs
ass6 roll test<number>.c
3–Address Test Outputs
ass6 roll quads<number>.out
Test Outputs
ass6 roll <number>.asm
5.    Prepare a tar-archive with the name ass6 roll.tar containing all the files and upload to Moodle.

5           Credits
                                 Design of Memory Binding:                                                                                 15 + 5 + 5 + 5 + 10 = 40

Handling of Activation Records

Handling of Nested Symbol Tables

Handling of Static Memory & Binding

Handling of Constants

Handling of Register Allocation & Assignment

                                 Design of Code Translation:                                                                                                 5 + 5 + 10 = 20

Handling of Prologue

Handling of Epilogue

Handling of Function Body

                                 Design of Target Code Management:                                                                                                         10

Integration of translated codes into an assembly file

                                 Design of Test files and correctness of outputs:                                                                  10 + 10 = 20

Test at least 5 i/p files covering all rules

Shortcoming and / or bugs, if any, should be highlighted

                                  Integrated interface of the tinyC Compiler:                                                                                             10