Your compiler's secret life

|
You know compilers go through stages right? Generally one of those stages is to convert the programming language (a la C/C++) to the assembly language for that architecture (e.g. x86 assembly). Then the assembler turns that into machine language that can be given to that type of processor. Here's how to use gcc such that it doesn't do the assembly step and prints the assembly on the terminal


gcc -S -o - a.c

Read on for a little dissection and/or go play with it and read the man page for gcc.

Here is the C file I'm using, a.c. It just loops through the command line arguments and prints them.


#include <stdio.h>

int main( int argc, char **argv ){
        int i;
        printf("argc: %d\n",argc);
        for( i=0; i < argc; i++ ){
                printf("argv[%d]: %s\n", i, argv[i]);
        }
}

So, again. Here is the command to get the assembly on the terminal.


gcc -S -o - a.c

Let's break that down into pieces.

  • gcc is the gnu c compiler
  • -S means to stop before the assembler stage of compilation
  • -o specifies the output file. It says the next argument will be the filename to output to. Normall the compiler would send the output of the compiler to a.s. But we give the '-' character to the compiler to indicate that it is to dump the output to STDOUT. This let's us catch the output with with '| less' or '> file'.
  • a.c is the name of the source file that we are compiling.
Here are some sample outputs. One is for Linux x86 on a PII and the other from Mac OS X PPC on a G4. I won't attempt to interpret the assembly at all. That's a discussion for another day.

Linux x86 on a PII


        .file   "a.c"
        .version        "01.01"
gcc2_compiled.:
.section        .rodata
.LC0:
        .string "argc: %d\n"
.LC1:
        .string "argv[%d]: %s\n"
.text
        .align 4
.globl main
        .type    main,@function
main:
        pushl %ebp
        movl %esp,%ebp
        subl $24,%esp
        addl $-8,%esp
        movl 8(%ebp),%eax
        pushl %eax
        pushl $.LC0
        call printf
        addl $16,%esp
        movl $0,-4(%ebp)
        .p2align 4,,7
.L3:
        movl -4(%ebp),%eax
        cmpl 8(%ebp),%eax
        jl .L6
        jmp .L4
        .p2align 4,,7
.L6:
        addl $-4,%esp
        movl -4(%ebp),%eax
        leal 0(,%eax,4),%edx
        movl 12(%ebp),%eax
        movl (%eax,%edx),%edx
        pushl %edx
        movl -4(%ebp),%eax
        pushl %eax
        pushl $.LC1
        call printf
        addl $16,%esp
.L5:
        incl -4(%ebp)
        jmp .L3
        .p2align 4,,7
.L4:
.L2:
        leave
        ret
.Lfe1:
        .size    main,.Lfe1-main
        .ident  "GCC: (GNU) 2.95.4 20011002 (Debian prerelease)"

Mac OS X PPC on a G4


.data
.cstring
	.align 2
LC0:
	.ascii "argc: %d\12\0"
	.align 2
LC1:
	.ascii "argv[%d]: %s\0"
.text
	.align 2
	.globl _main
_main:
	mflr r0
	stmw r30,-8(r1)
	stw r0,8(r1)
	stwu r1,-96(r1)
	mr r30,r1
	bcl 20,31,L1$pb
L1$pb:
	mflr r31
	stw r3,120(r30)
	stw r4,124(r30)
	addis r3,r31,ha16(LC0-L1$pb)
	la r3,lo16(LC0-L1$pb)(r3)
	lwz r4,120(r30)
	bl L_printf$stub
	li r0,0
	stw r0,64(r30)
L6:
	lwz r0,64(r30)
	lwz r9,120(r30)
	cmpw cr0,r0,r9
	blt cr0,L9
	b L7
L9:
	lwz r0,64(r30)
	slwi r9,r0,2
	lwz r0,124(r30)
	add r9,r9,r0
	addis r3,r31,ha16(LC1-L1$pb)
	la r3,lo16(LC1-L1$pb)(r3)
	lwz r4,64(r30)
	lwz r5,0(r9)
	bl L_printf$stub
	lwz r9,64(r30)
	addi r0,r9,1
	stw r0,64(r30)
	b L6
L7:
	mr r3,r0
	lwz r1,0(r1)
	lwz r0,8(r1)
	mtlr r0
	lmw r30,-8(r1)
	blr
.data
.picsymbol_stub
L_printf$stub:
	.indirect_symbol _printf
	mflr r0
	bcl 20,31,L0$_printf
L0$_printf:
	mflr r11
	addis r11,r11,ha16(L_printf$lazy_ptr-L0$_printf)
	mtlr r0
	lwz r12,lo16(L_printf$lazy_ptr-L0$_printf)(r11)
	mtctr r12
	addi r11,r11,lo16(L_printf$lazy_ptr-L0$_printf)
	bctr
.data
.lazy_symbol_pointer
L_printf$lazy_ptr:
	.indirect_symbol _printf
	.long dyld_stub_binding_helper

This is a great tool as a starting place for learning some of the differences is the assemblies between different architectures. You could take the same "Hello world" program and compile it using gcc on a number of readily accessible architectures (x86, ppc).