simple example -O3 optimizer problem?

[Dennis Luehring <dl dot soluz at gmx dot net>]

tested following code with

http://gcc.godbolt.org/
tested with
g++-4.8 (Ubuntu 4.8.1.2ubuntu1~12.04) 4.8.1
g++ (GCC) 4.9.0 20130909 (experimental)

and the result with -O3 + defined USE_ITER seems to be a little bit long

----------

static void foo(int a, int& dummy)
{
  dummy += a;
}

#define USE_ITER

int main(int argc, char** argv)
{
  //--
  //anti-optimizer
  int dummy = 0;
  int* array = (int*)&argv;
  //--

#if defined(USE_ITER)
  int* pend = &array[10];
  for(int* p = &array[0]; p < pend; ++p) foo(*p, dummy);
#else
  for(int i = 0; i < 10; ++i) foo(array[i], dummy);
#endif

  return dummy;
}

---------

with -O2
with/without define USE_ITER

result:

main:
    lea    rdx, [rsp-8]
    lea    rcx, [rsp+32]
    mov    QWORD PTR [rsp-8], rsi
    xor    eax, eax
.L3:
    add    eax, DWORD PTR [rdx]
    add    rdx, 4
    cmp    rdx, rcx
    jb    .L3
    rep; ret

with -O3

without define USE_ITER

main:
    mov    rax, rsi
    shr    rax, 32
    add    eax, esi
    add    eax, DWORD PTR [rsp]
    add    eax, DWORD PTR [rsp+4]
    add    eax, DWORD PTR [rsp+8]
    add    eax, DWORD PTR [rsp+12]
    add    eax, DWORD PTR [rsp+16]
    add    eax, DWORD PTR [rsp+20]
    add    eax, DWORD PTR [rsp+24]
    add    eax, DWORD PTR [rsp+28]
    ret

with define USE_ITER

main:
    lea    rdi, [rsp-16]
    lea    rax, [rsp+27]
    mov    QWORD PTR [rsp-16], rsi
    lea    r8, [rsp+24]
    mov    ecx, 1
    lea    rdx, [rdi+4]
    lea    rsi, [rdi+1]
    sub    rax, rdx
    mov    rdx, rdi
    shr    rax, 2
    add    rax, 1
    cmp    rsi, r8
    cmovbe    rcx, rax
    and    edx, 15
    shr    rdx, 2
    cmp    rcx, rdx
    cmovbe    rdx, rcx
    cmp    rax, 8
    ja    .L30
.L2:
    mov    rdx, rcx
.L11:
    cmp    rdx, 1
    mov    eax, DWORD PTR [rsp-16]
    je    .L13
    add    eax, DWORD PTR [rsp-12]
    cmp    rdx, 2
    je    .L14
    add    eax, DWORD PTR [rsp-8]
    cmp    rdx, 3
    je    .L15
    add    eax, DWORD PTR [rsp-4]
    cmp    rdx, 4
    je    .L16
    add    eax, DWORD PTR [rsp]
    cmp    rdx, 5
    je    .L17
    add    eax, DWORD PTR [rsp+4]
    cmp    rdx, 6
    je    .L18
    add    eax, DWORD PTR [rsp+8]
    cmp    rdx, 7
    je    .L19
    add    eax, DWORD PTR [rsp+12]
    lea    rsi, [rsp+16]
.L4:
    cmp    rcx, rdx
    je    .L23
.L3:
    sub    rcx, rdx
    mov    r9, rcx
    shr    r9, 2
    lea    r10, [0+r9*4]
    test    r10, r10
    je    .L6
    lea    rdx, [rdi+rdx*4]
    cmp    r9, 1
    movdqu    xmm0, XMMWORD PTR [rdx]
    jbe    .L7
    movdqu    xmm1, XMMWORD PTR [rdx+16]
    cmp    r9, 2
    paddd    xmm0, xmm1
    je    .L7
    movdqu    xmm1, XMMWORD PTR [rdx+32]
    paddd    xmm0, xmm1
.L7:
    movdqa    xmm2, xmm0
    lea    rsi, [rsi+r10*4]
    psrldq    xmm2, 8
    paddd    xmm0, xmm2
    movdqa    xmm3, xmm0
    psrldq    xmm3, 4
    paddd    xmm0, xmm3
    movd    edx, xmm0
    add    eax, edx
    cmp    rcx, r10
    je    .L23
.L6:
    lea    rdx, [rsi+4]
    add    eax, DWORD PTR [rsi]
    cmp    r8, rdx
    jbe    .L23
    lea    rdx, [rsi+8]
    add    eax, DWORD PTR [rsi+4]
    cmp    r8, rdx
    jbe    .L31
    add    eax, DWORD PTR [rsi+8]
    ret
.L23:
    rep; ret
.L30:
    cmp    rsi, r8
    ja    .L2
    xor    eax, eax
    test    rdx, rdx
    mov    rsi, rdi
    je    .L3
    jmp    .L11
.L31:
    ret
.L16:
    mov    rsi, rsp
    jmp    .L4
.L17:
    lea    rsi, [rsp+4]
    jmp    .L4
.L18:
    lea    rsi, [rsp+8]
    jmp    .L4
.L19:
    lea    rsi, [rsp+12]
    jmp    .L4
.L14:
    lea    rsi, [rsp-8]
    jmp    .L4
.L15:
    lea    rsi, [rsp-4]
    jmp    .L4
.L13:
    lea    rsi, [rdi+4]
    jmp    .L4