Thread: Academic Question About Compilers

How would something like this:

int x;
char hello[6] = "hello";
x = atoi(hello);

be compiled? I'm asking what the compiler will do internally to compile something like this. From generation of symbols, the code to pass parameters to library functions, how the stack is involved etc.

Thanks a lot!

Maybe there is an option on your compiler to produce the assembler code?

Anyway I was interested in having a go at producing the assembler code myself
I tried it on this short program I wrote earlier.

Code:

#include <stdio.h>
#include <string.h>
#include <stdlib.h>

main(argc,argv)
	int  argc;
	char *argv[]; 
{
	system("testbat");
}

compiled with

Code:

set DJGPP=C:\DJGPP\DJGPP.ENV
set PATH=C:\DJGPP\BIN;%PATH%

gcc -S  testb.c

Which produces this:-
Which may answer some of your question (or not).

Code:

	.file	"testb.c"
	.section .text
LC0:
	.ascii "testbat\0"
.globl _main
_main:
	pushl	%ebp
	movl	%esp, %ebp
	subl	$8, %esp
	andl	$-16, %esp
	movl	$0, %eax
	addl	$15, %eax
	addl	$15, %eax
	shrl	$4, %eax
	sall	$4, %eax
	subl	%eax, %esp
	subl	$12, %esp
	pushl	$LC0
	call	_system
	addl	$16, %esp
	leave
	ret
	.ident	"GCC: (GNU) 4.0.1"

Not sure what all that is about lol,

Code:

pushl	$LC0
	call	_system

Seems to be the passing of parameter to the 'system' call, I think the other stuff
is to do with the program arguements.

Anyone know what 'leave' does?

Originally Posted by esbo

Code:

	.file	"testb.c"
	.section .text

.file is an informational directive which tells the assembler what the original source file name is. .section .text tells the assembler to place the following data and instructions into the object section called ".text."

Code:

LC0:
	.ascii "testbat\0"

LC0 is a label generated by the compiler which is used to reference the string literal "testbat\0", which you used in the source code. In other words, it is storing the string literal and giving it the name "LC0."

Code:

.globl _main
_main:

.globl tells the assembler that the symbol "_main" should be exported for linking.

Code:

	pushl	%ebp
	movl	%esp, %ebp

Standard function prolog. Save the base pointer of the parent function, and set up a new base pointer at the current stack frame.

Code:

	subl	$8, %esp
	andl	$-16, %esp
	movl	$0, %eax
	addl	$15, %eax
	addl	$15, %eax
	shrl	$4, %eax
	sall	$4, %eax
	subl	%eax, %esp
	subl	$12, %esp

A lot of manipulation of the stack pointer. This is creating space for local variables and making sure the stack is aligned properly. If you had compiled with optimization, this code should be much simpler than this.

Code:

	pushl	$LC0

Push a pointer to the string literal "testbat\0" onto the stack in preparation for the function call:

Code:

	call	_system

Call the system() function.

Code:

	addl	$16, %esp

Fix up the stack pointer, discarding the previously pushed parameters as well as the local variables.

Code:

	leave

This undoes the effect of the function prolog.

Code:

	ret

And return from main().

Code:

	.ident	"GCC: (GNU) 4.0.1"

Identify the compiler program.