https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85501
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Last reconfirmed| |2025-10-15
Status|UNCONFIRMED |ASSIGNED
CC| |pinskia at gcc dot gnu.org
Assignee|unassigned at gcc dot gnu.org |pinskia at gcc dot
gnu.org
Ever confirmed|0 |1
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> Take:
> void f(float *myVector, float *result, int i )
> {
> for(int n = 0; n < 1024; n++)
> if (myVector[n] > 0.5){
> result[n] = 0.8f;
> }
> else {
> result[n] = 0.1f;
> }
> }
>
> This loop does get vectorized at -O3 and produces the conditional move at
> -O2 but the tree level for the scalar side has a store inside the
> conditional still.
Since GCC 11, sink1 is able to do the sinking of the store.
To do it in phiopt1/2 you need to be able to commonalize the stores which we
don't do currently because we look for exact operand_equal_p and we don't hoist
common code out earlier.
