[Bug bootstrap/81638] [8 Regression] AIX bootstrap failure due to Recover GOTO predictor

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81638

--- Comment #6 from Martin Liška <marxin at gcc dot gnu.org> ---
Created attachment 41878
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=41878&action=edit
Reduced test-case

Reduced test-case that contains maybe uninitialized warning from the mentioned
commit. Note that release_ssa diff in probabilities is following:

--- /tmp/good/pr81638.i.049t.release_ssa        2017-08-01 13:22:44.009392386
+0200
+++ /tmp/bad/pr81638.i.049t.release_ssa 2017-08-01 13:21:37.164092117 +0200
@@ -183,25 +183,25 @@
   <bb 6> [1.67%] [count: INV]:
   _5 = a ();
   if (_5 != 0)
-    goto <bb 10>; [67.00%] [count: INV]
+    goto <bb 10>; [51.12%] [count: INV]
   else
-    goto <bb 7>; [33.00%] [count: INV]
+    goto <bb 7>; [48.88%] [count: INV]

-  <bb 7> [0.55%] [count: INV]:
+  <bb 7> [0.82%] [count: INV]:
   _6 = a ();
   if (_6 != 0)
     goto <bb 8>; [33.00%] [count: INV]
   else
     goto <bb 10>; [67.00%] [count: INV]

-  <bb 8> [0.18%] [count: INV]:
+  <bb 8> [0.27%] [count: INV]:
   _7 = a ();
   if (_7 != 0)
     goto <bb 9>; [33.00%] [count: INV]
   else
     goto <bb 10>; [67.00%] [count: INV]

-  <bb 9> [0.06%] [count: INV]:
+  <bb 9> [0.09%] [count: INV]:
   bt ();

   <bb 10> [100.00%] [count: INV]:

which then causes the convergence. I guess it's another from set of false
positives.