src/hotspot/share/opto/loopopts.cpp

@@ -1409,17 +1409,25 @@
         if( n_loop->is_member( u_loop ) ) break; // Found use in inner loop
         if( u->Opcode() == Op_Opaque1 ) break; // Found loop limit, bugfix for 4677003
       }
       bool did_break = (i < imax);  // Did we break out of the previous loop?
       if (!did_break && n->outcnt() > 1) { // All uses in outer loops!
-        Node *late_load_ctrl = NULL;
+        Node* late_load_ctrl = NULL;
+        Node* outer_strip_mined_loop_exit = NULL;
         if (n->is_Load()) {
           // If n is a load, get and save the result from get_late_ctrl(),
           // to be later used in calculating the control for n's clones.
           clear_dom_lca_tags();
           late_load_ctrl = get_late_ctrl(n, n_ctrl);
+          if (n_loop->head()->is_Loop() && n_loop->head()->as_Loop()->is_strip_mined() && !n_loop->head()->is_OuterStripMinedLoop()
+              && n_loop->tail()->in(0) == late_load_ctrl->in(0)) {
+             // late_load_ctrl is a loop exit of an inner loop of an outer strip mined loop.
+             // Use the outer strip mined loop exit instead later if it dominates x_ctrl.
+             outer_strip_mined_loop_exit = n_loop->_parent->head()->as_OuterStripMinedLoop()->outer_loop_exit();
         }
+        }
+
         // If n is a load, and the late control is the same as the current
         // control, then the cloning of n is a pointless exercise, because
         // GVN will ensure that we end up where we started.
         if (!n->is_Load() || late_load_ctrl != n_ctrl) {
           for (DUIterator_Last jmin, j = n->last_outs(jmin); j >= jmin; ) {

@@ -1455,14 +1463,21 @@
             if (n->is_Load()) {
               // For loads, add a control edge to a CFG node outside of the loop
               // to force them to not combine and return back inside the loop
               // during GVN optimization (4641526).
               //
-              // Because we are setting the actual control input, factor in
-              // the result from get_late_ctrl() so we respect any
-              // anti-dependences. (6233005).
+              // Because we are setting the actual control input, factor in the result from get_late_ctrl() so we respect
+              // any anti-dependences (6233005). If late_load_ctrl is an inner loop exit inside an outer strip mined loop
+              // whose exit dominates the latest possible placement for x, then use this outer strip mined loop exit
+              // instead of the inner loop exit (= late_load_ctrl) to move the load completely out of the loop.
+              // This still respects any anti-dependencies in get_late_ctrl().
+              if (outer_strip_mined_loop_exit != NULL && is_dominator(outer_strip_mined_loop_exit, get_late_ctrl(x, late_load_ctrl))) {
+                assert(dom_lca(outer_strip_mined_loop_exit, x_ctrl) == outer_strip_mined_loop_exit, "must be the same result");
+                x_ctrl = outer_strip_mined_loop_exit;
+              } else {
               x_ctrl = dom_lca(late_load_ctrl, x_ctrl);
+              }
 
               // Don't allow the control input to be a CFG splitting node.
               // Such nodes should only have ProjNodes as outs, e.g. IfNode
               // should only have IfTrueNode and IfFalseNode (4985384).
               x_ctrl = find_non_split_ctrl(x_ctrl);