On Wed, 20 Feb 2008 14:35:56 -0600
John Love-Jensen <eljay@xxxxxxxxx> wrote:
We have next files:
cat << EOF > full.c
#include <stdio.h>
void imprime(const char *m)
{
printf("%s\n", m);
}
int main()
{
imprime("Beautiful day");
return 0;
}
EOF
cat << EOF > a.c
void imprime(const char *m);
int main()
{
imprime("Beautiful day");
return 0;
}
EOF
cat << EOF > b.c
#include <stdio.h>
void imprime(const char *m)
{
printf("%s\n", m);
}
EOF
Now,
gcc a.c b.c -o ab
"a.c" and "b.c" built "ab"
gcc full.c -o full
"full.c" built "full"
In comparing files sizes (Target:amd64):
du -b full
8396
du -b ab
8453
Therefore, ab > full
> Hi Felipe,
>
> > Well, the problem is when I split a C source file into two files. Output
> > object size of splitted source is bigger than size of one-file source. My
> > question to mail list is Why? What happens? What section grows up?
>
> Are you including or excluding the object header block?
In object header section, ".comment" section has different size.
.comment size of full is 000000f3
.comment size of ab is 00000114
That only explains 33 bytes between "ab" and "full". In ab, ld puts
together .comment of "a.s" and "b.s".
do I could avoid this?
> What does the assembly (.s) of the two files indicate when you use a
> --save-temps?
Function main() in "a.s" has one more instruction.
main:
.LFB2:
pushq %rbp
.LCFI0:
movq %rsp, %rbp
.LCFI1:
movl $.LC0, %edi
movl $0, %eax
call imprime
movl $0, %eax
leave
ret
and in "full.s":
main:
.LFB3:
pushq %rbp
.LCFI3:
movq %rsp, %rbp
.LCFI4:
movl $.LC0, %edi
call imprime
movl $0, %eax
leave
ret
>
> What does an object dump of the single .o versus the two separated .o files
> indicate?
A different NOP padding in .text and more text in .comment
> --Eljay
>
