With the following method:

  public static int vectSumOfMulAdd1(
          int[] a,
          int[] b,
          int[] c,
          int[] d) {
    int total = 0;
    for (int i = 0; i < LENGTH; i++) {
      d[i] = (int)(a[i] * b[i] + c[i]);
      total += d[i];
    }
    return total;
  }

the main loop is not vectorized because the reduction is missed by the superword optimization. If the result of the reduction is used by a control node (Return here), the reduction is wrongly excluded.

diff --git a/src/share/vm/opto/loopTransform.cpp b/src/share/vm/opto/loopTransform.cpp
--- a/src/share/vm/opto/loopTransform.cpp
+++ b/src/share/vm/opto/loopTransform.cpp
@@ -1742,7 +1742,7 @@
               // The result of the reduction must not be used in the loop
               for (DUIterator_Fast imax, i = def_node->fast_outs(imax); i < imax && ok; i++) {
                 Node* u = def_node->fast_out(i);
- if (has_ctrl(u) && !loop->is_member(get_loop(get_ctrl(u)))) {
+ if (!loop->is_member(get_loop(ctrl_or_self(u)))) {
                   continue;
                 }
                 if (u == phi) {