https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122901
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Last reconfirmed| |2025-11-28
Summary|gcc is not able to handle |vector lowering needs to be
|pair of vectors coming from |improved for vector_convert
|__builtin_convertvector |and others
Status|UNCONFIRMED |NEW
--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
And lowering of /1024 is not done using shifts.
```
v1_11 = MEM[(v16x8D.4678 *)_3];
v2_12 = .VEC_CONVERT (v1_11);
v2_14 = _4 * v2_12;
v2_15 = v2_14 / { 1024, 1024, 1024, 1024, 1024, 1024, 1024, 1024 };
v1_16 = .VEC_CONVERT (v2_15);
# .MEM_17 = VDEF <.MEM_22>
MEM[(v16x8D.4678 *)_3] = v1_16;
```
Looks reasonible before the lowering but then lowering does not do the right
thing.
