RFR: Fix call to is_g1_marking_load() with is_shenandoah_state_load() in ShenandoahSupport

[Roman Kennke]

The check for is_g1_marking_load() in ShenandoahSupport is bogus (and
there is no overload left that actually returns true!), and should be
is_shenandoah_state_load().

Testing: tier3_gc_shenandoah

# HG changeset patch
# User rkennke
# Date 1536700094 -7200
#      Tue Sep 11 23:08:14 2018 +0200
# Node ID aad4dc29c7caa4c87172ace60497189857225a0d
# Parent  424866bd51c0d1e99ab867c50ec14e1507bda8e2
Fix call to is_g1_marking_load() with is_shenandoah_state_load() in
ShenandoahSupport

diff --git a/src/hotspot/share/gc/shenandoah/c2/shenandoahSupport.cpp
b/src/hotspot/share/gc/shenandoah/c2/shenandoahSupport.cpp
--- a/src/hotspot/share/gc/shenandoah/c2/shenandoahSupport.cpp
+++ b/src/hotspot/share/gc/shenandoah/c2/shenandoahSupport.cpp
@@ -2669,7 +2669,7 @@
       Node* c = old_c;
       if (c != ctrl ||
           is_dominator_same_ctrl(old_c, barrier, u, phase) ||
-          u->is_g1_marking_load()) {
+          u->is_shenandoah_state_load()) {
         phase->igvn().rehash_node_delayed(u);
         int nb = u->replace_edge(ctrl, region);
         if (u->is_CFG()) {