[Bug middle-end/123175] [14/15/16 Regression] Wrong folding of VEC_PERM_EXPR

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123175

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 63092
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=63092&action=edit
New vec_perm_indices CTOR

STH like this?  I think tree_to_vec_perm_builder need always succeed for
a VEC_PERM_EXPR that's in the IL(?)