https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83518
--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Before store-merging we have:
MEM[(int *)&arr] = { 5, 4, 3, 2 };
t_2 = arr[0];
_65 = arr[1];
_69 = arr[2];
_73 = arr[3];
arr[0] = _69;
arr[1] = _73;
arr[2] = 1;
arr[3] = t_2;
vect__2.5_38 = MEM[(int *)&arr];
and all store-merging can do with this is what it does:
MEM[(int *)&arr] = { 5, 4, 3, 2 };
t_2 = arr[0];
_65 = arr[1];
_69 = arr[2];
_73 = arr[3];
_46 = MEM[(int *)&arr + 8B];
MEM[(int *)&arr] = _46;
arr[2] = 1;
arr[3] = t_2;
vect__2.5_38 = MEM[(int *)&arr];
where the _69 and _73 sets can be DCEd later. store-merging has no framework
like FRE to do analysis what memory location contains at which point.
So we'd need another late FRE pass to handle this?
