GCC Internals: built-in functions?

GCC Internals: built-in functions?

[Amittai Aviram]

On this GCC Internals page

http://en.wikibooks.org/wiki/GNU_C_Compiler_Internals/GNU_C_Compiler_Architecture_3_4

I found the following in the "GCC Initialization" section:

"GCC built-in functions are the functions that are evaluated at compile time. For example, if the size argument of a strcpy() function is a constant then GCC replaces the function call with the required number of assignments."

I was curious about this, so I tried compiling to assembly (-S) a very simple program:

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

int main(void) {

        char s1[0x10];
        char * s0 = "HELLO";
        strcpy(s1, s0);
        printf("%s %s\n", s0, s1);
        return EXIT_SUCCESS;
}

But the resulting assembly code simply calls strcpy with the two arguments, just as I would have expected had I not read the above sentence:

	movq    $.LC0, -40(%rbp)
	movq    -40(%rbp), %rdx
	leaq    -32(%rbp), %rax
	movq    %rdx, %rsi
	movq    %rax, %rdi
	call    strcpy

(Here, .LC0 labels the string "HELLO".)

So what does that sentence actually mean and what am I missing?  Thanks!

Amittai Aviram
PhD Student in Computer Science
Yale University
646 483 2639
amittai.aviram@yale.edu
http://www.amittai.com

Re: GCC Internals: built-in functions?

[Jonathan Wakely]

On 30 January 2011 20:45, Amittai Aviram wrote:
> On this GCC Internals page
>
> http://en.wikibooks.org/wiki/GNU_C_Compiler_Internals/GNU_C_Compiler_Architecture_3_4
>
> I found the following in the "GCC Initialization" section:
>
> "GCC built-in functions are the functions that are evaluated at compile time. For example, if the size argument of a strcpy() function is a constant then GCC replaces the function call with the required number of assignments."
>
> I was curious about this, so I tried compiling to assembly (-S) a very simple program:
>
> #include <stdio.h>
> #include <stdlib.h>
> #include <string.h>
>
> int main(void) {
>
>        char s1[0x10];
>        char * s0 = "HELLO";
>        strcpy(s1, s0);
>        printf("%s %s\n", s0, s1);
>        return EXIT_SUCCESS;
> }
>
> But the resulting assembly code simply calls strcpy with the two arguments, just as I would have expected had I not read the above sentence:
>
>        movq    $.LC0, -40(%rbp)
>        movq    -40(%rbp), %rdx
>        leaq    -32(%rbp), %rax
>        movq    %rdx, %rsi
>        movq    %rax, %rdi
>        call    strcpy
>
> (Here, .LC0 labels the string "HELLO".)
>
> So what does that sentence actually mean and what am I missing?  Thanks!

strcpy has no 'size' parameter, I assume it's meant to say strncpy

Re: GCC Internals: built-in functions?

[Amittai Aviram]

On Jan 30, 2011, at 3:57 PM, Jonathan Wakely wrote:

> On 30 January 2011 20:45, Amittai Aviram wrote:
>> On this GCC Internals page
>> 
>> http://en.wikibooks.org/wiki/GNU_C_Compiler_Internals/GNU_C_Compiler_Architecture_3_4
>> 
>> I found the following in the "GCC Initialization" section:
>> 
>> "GCC built-in functions are the functions that are evaluated at compile time. For example, if the size argument of a strcpy() function is a constant then GCC replaces the function call with the required number of assignments."
>> 
>> I was curious about this, so I tried compiling to assembly (-S) a very simple program:
>> 
>> #include <stdio.h>
>> #include <stdlib.h>
>> #include <string.h>
>> 
>> int main(void) {
>> 
>>        char s1[0x10];
>>        char * s0 = "HELLO";
>>        strcpy(s1, s0);
>>        printf("%s %s\n", s0, s1);
>>        return EXIT_SUCCESS;
>> }
>> 
>> But the resulting assembly code simply calls strcpy with the two arguments, just as I would have expected had I not read the above sentence:
>> 
>>        movq    $.LC0, -40(%rbp)
>>        movq    -40(%rbp), %rdx
>>        leaq    -32(%rbp), %rax
>>        movq    %rdx, %rsi
>>        movq    %rax, %rdi
>>        call    strcpy
>> 
>> (Here, .LC0 labels the string "HELLO".)
>> 
>> So what does that sentence actually mean and what am I missing?  Thanks!
> 
> strcpy has no 'size' parameter, I assume it's meant to say strncpy

I tried it with strncpy and with __builtin_strncpy, with exactly the same results, i.e., the assembly code still calls strncpy with three arguments.

Amittai Aviram
PhD Student in Computer Science
Yale University
646 483 2639
amittai.aviram@yale.edu
http://www.amittai.com

Re: GCC Internals: built-in functions?

[Jonathan Wakely]

On 30 January 2011 21:02, Amittai Aviram wrote:
>
> I tried it with strncpy and with __builtin_strncpy, with exactly the same results, i.e., the assembly code still calls strncpy with three arguments.

Did you enable optimization?

Re: GCC Internals: built-in functions?

[Amittai Aviram]

On Jan 30, 2011, at 4:07 PM, Jonathan Wakely wrote:

> On 30 January 2011 21:02, Amittai Aviram wrote:
>> 
>> I tried it with strncpy and with __builtin_strncpy, with exactly the same results, i.e., the assembly code still calls strncpy with three arguments.
> 
> Did you enable optimization?

I hadn't and that did it--even with just "-O1."  Thank you!  Very interesting!—

        movl    $1280066888, (%rsp)
        movw    $79, 4(%rsp)

The first integer in hex is 4C 4C 45 48 ( == 'L','L','E','H')
and the second (79) is 4F ( == 'O')

So, of course, it works out when assigned in little-endian order.   The second assignment also puts NULL in the right place as the higher-order bits of the integer 79.

Amittai Aviram
PhD Student in Computer Science
Yale University
646 483 2639
amittai.aviram@yale.edu
http://www.amittai.com