[Bug tree-optimization/109048] [13 regression] redundant mask compare generated by vectorizer.

["rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109048

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #8)
> Created attachment 54762 [details]
> heuristic
> 
> I am testing this heuristic

FAIL: gcc.dg/tree-ssa/ranger-threader-3.c scan-tree-dump-times ethread
"Register
ing.*jump thread" 1
FAIL: gcc.dg/tree-ssa/ssa-dom-thread-14.c scan-tree-dump-times dom2 "Threaded"
2

is the fallout - both are the same testcase, both can be mitigated by adjusting
the dump to scan for and making sure PROP_loop_opts_done is also provided
for functions without loops.

But it also shows a case where we optimize things significantly later
after the change.

  if ((!unsignedp || (!left && methods == OPTAB_WIDEN)))
    {
      enum optab_methods methods1 = methods;
      if (unsignedp)
        methods1 = OPTAB_MUST_WIDEN;
      expand_binop (left ? lshift_optab : rshift_arith_optab,
                           unsignedp, methods1);

used to be threaded in early threading but now would be delayed to after
loop opts.