https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120234
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
CC| |rdapp at gcc dot gnu.org
Status|NEW |ASSIGNED
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
I have a patch to restore the vectorization GCC 15 does, but forwprop doesn't
pick up the slack. This regressed with r16-6292-ge8fd095091ebfc which
added
@@ -4011,6 +4017,9 @@ simplify_vector_constructor (gimple_stmt_iterator *gsi)
if (CONVERT_EXPR_CODE_P (conv_code)
&& (2 * TYPE_PRECISION (TREE_TYPE (TREE_TYPE (orig[0])))
== TYPE_PRECISION (TREE_TYPE (type)))
+ && orig_elem_type[0]
+ && useless_type_conversion_p (orig_elem_type[0],
+ TREE_TYPE (type))
&& mode_for_vector (as_a <scalar_mode>
(TYPE_MODE (TREE_TYPE (TREE_TYPE
(orig[0])))),
nelts * 2).exists ()
which effectively disables unpacking. I think it wants to check whether
the BIT_FIELD_REF of the vector is changing the sign of the vector element.
Thus TREE_TYPE (TREE_TYPE (orig[0])) instead of TREE_TYPE (type).
Testing the two pieces of the fix.
