Why is alignof(max_align_t) 16?

Question

I'm running Debian on an x86_64 Intel processor.

gcc (Debian 8.3.0) compiles the following program

#include <stdio.h>
#include <stdalign.h>
#include <stddef.h>

int main(){
  printf("%zd\n",alignof(max_align_t));
}

and outputs

What datatype requires a 16 byte alignment?

`__int128`. Also, it can perfectly be any arbitrary value AFAIK. — Oppen, Apr 30 '20 at 20:08
`max_align_t` is guaranteed to be "at least as large" as the largest type, not "exactly as large". So it could be 16 even if the largest datatype is only 8. — Agent_L, Apr 30 '20 at 20:27

score 3 · Accepted Answer · answered Apr 30 '20 at 22:19

3

On x86_64, _Alignof(long double)==16.

answered Apr 30 '20 at 22:19

unless [`-m96bit-long-double`](https://gcc.gnu.org/onlinedocs/gcc/x86-Options.html) is specified – phuclv May 01 '20 at 00:24
1

@phuclv: That's contrary to the ABI and produces code that can't link with normal ABI-conforming x86_64 code. – R.. GitHub STOP HELPING ICE May 01 '20 at 00:48

1 Answers1