[Bug tree-optimization/114737] New: Missed optimization : fail to optimize load with select clobber

[Bug tree-optimization/114737] New: Missed optimization : fail to optimize load with select clobber

[xxs_chy at outlook dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114737

            Bug ID: 114737
           Summary: Missed optimization : fail to optimize load with
                    select clobber
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: xxs_chy at outlook dot com
  Target Milestone: ---

Godbolt link: https://godbolt.org/z/3e5sfvfKj

```
char src(void** p, char* p1, bool c) {
    char* tostore1 = p1 + 1;
    char* tostore2 = p1 + 4;
    *p = tostore1;
    *p1 = 0;

    char* tostore = (c ? tostore1 : tostore2) + 1;
    *tostore = 1;
    return *p1;
}
```

"return *p1" can be optimized into "return 0" here.

```
char tgt(void** p, char* p1, bool c) {
    char* tostore1 = p1 + 1;
    char* tostore2 = p1 + 4;
    *p = tostore1;
    *p1 = 0;

    char* tostore = (c ? tostore1 : tostore2) + 1;
    *tostore = 1;
    return 0;
}
```

[Bug tree-optimization/114737] Missed optimization : fail to optimize load with select clobber

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114737

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-04-16
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |alias, missed-optimization
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  points-to analysis does not track pointer offsets for the final
points-to sets, so tostore and p1 point to the same object.  Specifically
pointer offset gathering when setting up the ao_ref for *to_store_1
doesn't look through PHIs:

  <bb 3> :
  # PT = nonlocal null
  tostore_11 = p1_2(D) + 2;
  goto <bb 5>; [INV]

  <bb 4> :
  # PT = nonlocal null
  tostore_10 = p1_2(D) + 5;

  <bb 5> :
  # PT = nonlocal null
  # tostore_1 = PHI <tostore_11(3), tostore_10(4)>
  *tostore_1 = 1;

what in this case would help a little bit is if this were

  <bb 3> :
  goto <bb 5>; [INV]

  <bb 4> :

  <bb 5> :
  # PT = nonlocal null
  # offset_3 = PHI <2(3), 5(4)>
  # PT = nonlocal null
  tostore_1 = p1_2(D) + offset_3;
  *tostore_1 = 1;

then eventually range-info for offset_3 could be used to constrain the
ao_ref.