--- old/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp	2016-12-19 16:04:55.600790107 -0500
+++ new/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp	2016-12-19 16:04:55.540791749 -0500
@@ -3273,14 +3273,25 @@
   ldr(dst, Address(dst, Klass::prototype_header_offset()));
 }
 
-void MacroAssembler::store_klass(Register dst, Register src) {
-  // FIXME: Should this be a store release?  concurrent gcs assumes
-  // klass length is valid if klass field is not null.
-  if (UseCompressedClassPointers) {
-    encode_klass_not_null(src);
-    strw(src, Address(dst, oopDesc::klass_offset_in_bytes()));
+void MacroAssembler::store_klass(Register dst, Register src, Register tmp) {
+  // concurrent gcs assumes klass length is valid if klass field is not null.
+  bool is_concurrentGC = UseG1GC || UseConcMarkSweepGC;
+
+  if (is_concurrentGC) {
+    add(tmp, dst, oopDesc::klass_offset_in_bytes());
+    if (UseCompressedClassPointers) {
+      encode_klass_not_null(src);
+      stlrw(src, tmp);
+    } else {
+      stlr(src, tmp);
+    }
   } else {
-    str(src, Address(dst, oopDesc::klass_offset_in_bytes()));
+    if (UseCompressedClassPointers) {
+      encode_klass_not_null(src);
+      strw(src, Address(dst, oopDesc::klass_offset_in_bytes()));
+    } else {
+      str(src, Address(dst, oopDesc::klass_offset_in_bytes()));
+    }
   }
 }
 
@@ -3907,7 +3918,7 @@
   }
   // store klass last.  concurrent gcs assumes klass length is valid if
   // klass field is not null.
-  store_klass(top, t1);
+  store_klass(top, t1, rscratch1);
 
   mov(t1, top);
   ldr(rscratch1, Address(rthread, in_bytes(JavaThread::tlab_start_offset())));