[Bug tree-optimization/114701] New: Missed optimization of loop invariant

[Bug tree-optimization/114701] New: Missed optimization of loop invariant

[652023330028 at smail dot nju.edu.cn]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114701

            Bug ID: 114701
           Summary: Missed optimization of loop invariant
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: 652023330028 at smail dot nju.edu.cn
  Target Milestone: ---

Hello, we noticed that maybe there is a missed optimization of loop invariant.

Loop invariant: 100/(b+d).

https://godbolt.org/z/jKs5qohMs

int a, b;
void func() {
  int d = 0;
  for (int i = 0; i < 1000; i++) {
    b += 1;
    d += -1;
    a += 100 / (b + d);
  }
}

GCC -O3 -fwrapv:
func():
        mov     r9d, DWORD PTR b[rip]
        mov     edi, DWORD PTR a[rip]
        xor     ecx, ecx
        lea     r8d, [r9+1]
.L2:
        mov     esi, r8d
        mov     eax, 100
        xor     edx, edx
        sub     esi, ecx
        sub     ecx, 1
        add     esi, ecx
        idiv    esi
        add     edi, eax
        cmp     ecx, -1000
        jne     .L2
        add     r9d, 1000
        mov     DWORD PTR a[rip], edi
        mov     DWORD PTR b[rip], r9d
        ret

Expected code (Clang -O3 -fwrapv -fno-tree-vectorize):
func():                               # @func()
        mov     ecx, dword ptr [rip + b]
        mov     esi, dword ptr [rip + a]
        mov     eax, 100
        xor     edx, edx
        idiv    ecx
        add     eax, eax
        mov     edx, 1000
        lea     edi, [rax + rax]
.LBB0_1:                                # =>This Inner Loop Header: Depth=1
        add     esi, eax
        add     esi, edi
        add     esi, edi
        add     edx, -10
        jne     .LBB0_1
        add     ecx, 1000
        mov     dword ptr [rip + b], ecx
        mov     dword ptr [rip + a], esi
        ret

Thank you very much for your time and effort! We look forward to hearing from
you.

[Bug tree-optimization/114701] Missed optimization of loop invariant

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114701

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-04-12
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |missed-optimization
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
There were similar bugs reported already, if not exact dups (b + d being
invariant).  As with the others this depends on store motion.  As with
the others the SCCP pass could compute the overall increment to 'a'.
SCEV would be required to detect that b + d is invariant and to hoist that.