STL vector question

[Martin v. Loewis]

> But the real codes (check assembler instructions) differ! What I missed?

Since you did not say what assembler code you got, it is hard to tell
what you are missing. I slightly modified your code to read

#include <vector>

void foo(int*);
void foo(vector<int>&);

int main()
{
  int         a1[111];
  vector<int> a2(111);
  a1[1]=0;             // int[] access
  foo(a1);
  a2[1]=0;             // access to vector<int>[]
  foo(a2);
  
}

so that the store to a1 and a2 is not considered dead at optimization,
and also to allow finding the assembler code easier. With gcc 2.95.2,
on i586-pc-linux-gnu, using '-O2 -fno-exceptions', I get the code
fragment

	call fill_n__H3ZPiZUiZi_X01X11RCX21_X01
	addl $16,%esp               
	movl %eax,-452(%ebp)       ; this is the end of a2 construction
	movl $0,-440(%ebp)         ; a1[1] = 0
	addl $-12,%esp
	leal -444(%ebp),%eax       ; push a1 for foo
	pushl %eax
	call foo__FPi
	movl -456(%ebp),%eax       ; eax = _M_start
	addl $16,%esp              ; belongs to foo call
	movl $0,4(%eax)            ; _M_start[1] = 0
	addl $-12,%esp
	leal -456(%ebp),%eax       ; push a2 for foo
	pushl %eax
	call foo__FRt6vector2ZiZt9allocator1Zi


So the access to a2 has an additional indirection, because the vector
is a wrapper object around the array. Apart from that, I can see no
difference.

Regards,
Martin