8325857: G1 Full GC flushes mark stats cache too early

Flush mark stats cache after reference processing because that phase can mark some objects.

Intentionally performed serially because even on not-so-current processors this phase takes < 1ms for up to 256 threads.

src/hotspot/share/gc/g1/g1FullCollector.cpp

@@ -312,6 +312,13 @@ void G1FullCollector::phase1_mark_live_objects() {
     reference_processor()->set_active_mt_degree(old_active_mt_degree);
   }
 
+  {
+    GCTraceTime(Debug, gc, phases) debug("Phase 1: Flush Mark Stats Cache", scope()->timer());
+    for (uint i = 0; i < workers(); i++) {
+      marker(i)->flush_mark_stats_cache();
+    }
+  }
+
   // Weak oops cleanup.
   {
     GCTraceTime(Debug, gc, phases) debug("Phase 1: Weak Processing", scope()->timer());

src/hotspot/share/gc/g1/g1FullGCMarkTask.cpp

@@ -56,7 +56,6 @@ void G1FullGCMarkTask::work(uint worker_id) {
 
   // Mark stack is populated, now process and drain it.
   marker->complete_marking(collector()->oop_queue_set(), collector()->array_queue_set(), &_terminator);
-  marker->flush_mark_stats_cache();
 
   // This is the point where the entire marking should have completed.
   assert(marker->oop_stack()->is_empty(), "Marking should have completed");