https://gcc.gnu.org/bugzilla/show_bug.cgi?id=121772
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Component|target |tree-optimization
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Alice Carlotti from comment #1)
> Is the issue that you can't use a vector constructor to directly concatenate
> two vectors in GIMPLE? That is, the following is invalid:
>
> _7 = {b_5(D), { 0, 0, 0, 0 }}
>
>
> If so, then what's the canonical way to represent vector concatenation in
> GIMPLE? VEC_PERM_EXPR?
No that is correct.
The problem is:
```
_7 = {b_5(D), { 0, 0, 0, 0 }};
_8 = .REDUC_PLUS (_7);
is not doing the correct thing:
Applying pattern match.pd:11055, gimple-match-4.cc:7859
gimple_simplified to _8 = b_5(D);
Removing dead stmt:_7 = {b_5(D), { 0, 0, 0, 0 }};
```
The match pattern:
```
/* Fold reduction of a single nonzero element constructor. */
(for reduc (IFN_REDUC_PLUS IFN_REDUC_IOR IFN_REDUC_XOR)
(simplify (reduc (CONSTRUCTOR@0))
(with { tree ctor = (TREE_CODE (@0) == SSA_NAME
? gimple_assign_rhs1 (SSA_NAME_DEF_STMT (@0)) : @0);
tree elt = ctor_single_nonzero_element (ctor); }
(if (elt
&& !HONOR_SNANS (type)
&& !HONOR_SIGNED_ZEROS (type))
{ elt; }))))
```
There is no types matching in this match pattern.
Which means it started really with r12-7326-g2ef0e75d0bbc80 .
