[Bug tree-optimization/99412] s352 benchmark of TSVC is vectorized by clang and not by gcc

[rguenth at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99412

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jan Hubicka from comment #2)
> This is also seen with zen4 comparing gcc and aocc. (about 2.3 times
> differnece)

Disabling

@@ -6877,7 +6887,7 @@ reassociate_bb (basic_block bb)
                          binary op are chosen wisely.  */
                       int len = ops.length ();
                       if (len >= 3)
                         swap_ops_for_binary_stmt (ops, len - 3, stmt);

will naturally create the reduction chain (or leave it in place) given the
current rank computation.  We do have (somewhat) robust fallback from
reduction chain to reduction (via reduction path support), so I think this
change would be OK.