gcc编译器使用puts切换printf调用

Question

I have compiled a simple C program under my ubuntu:

#include<stdio.h>

int main()
{
        printf("hello world\n");
        return 0;
}

then I checked the disassemble of the executable using gdb and I saw the following code:

   0x000000000000063a <+0>:     push   %rbp
   0x000000000000063b <+1>:     mov    %rsp,%rbp
   0x000000000000063e <+4>:     lea    0x9f(%rip),%rdi        # 0x6e4
   0x0000000000000645 <+11>:    callq  0x510 <puts@plt>
   0x000000000000064a <+16>:    mov    $0x0,%eax
   0x000000000000064f <+21>:    pop    %rbp
   0x0000000000000650 <+22>:    retq

Here The compiler just swapped printf call with puts call. When I tried this with printf, calling it with a string that doesn't has new line appended to it, the call remained a call to printf in the disassemble.

My question is what is the mechanism of the compiler that enables such a thing? Does he read the string to be sent to printf and because printf is a libc function the compiler can decide this optimisation since he knows libc?

Answer 1

Yes. GCC reads and verifies the format string in order to issue compilation warnings about a wrong format string specifier (for example if you use %d but the argument is an unsigned int , it prints a warning.) Since it parses the string at compile-time, it can also know if it's possible or not to replace the printf() with a puts() call as an optimization ( puts() is faster because it doesn't have to parse the string.)

Answer 2

Some functions are "better known" by the compiler because they are in the most common use. printf one of such a functions. Compiler even analyzes the format string and warns if the parameters have wrong types. You can also use this feature when you write similar functions by letting (by the apriopriate attributes) the compiler that your functions uses printf style format string. And because it knows how this function works it can optimize the calls to it. There are many other functions like this. For example memcpy family ones, many math etc etc