Merge
authorctornqvi
Fri, 17 Apr 2015 23:55:04 +0000
changeset 30250 58f6fc816aae
parent 30248 5c6dacbd17ae (diff)
parent 30249 faf83ebbf040 (current diff)
child 30251 9b05c9cc4698
Merge
hotspot/test/sanity/WhiteBox.java
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_implementation/g1/G1CollectedHeap.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_implementation/g1/G1CollectedHeap.java	Fri Apr 17 23:55:04 2015 +0000
@@ -29,9 +29,9 @@
 import java.util.Observer;
 
 import sun.jvm.hotspot.debugger.Address;
+import sun.jvm.hotspot.gc_interface.CollectedHeap;
 import sun.jvm.hotspot.gc_interface.CollectedHeapName;
 import sun.jvm.hotspot.memory.MemRegion;
-import sun.jvm.hotspot.memory.SharedHeap;
 import sun.jvm.hotspot.memory.SpaceClosure;
 import sun.jvm.hotspot.runtime.VM;
 import sun.jvm.hotspot.runtime.VMObjectFactory;
@@ -41,7 +41,7 @@
 
 // Mirror class for G1CollectedHeap.
 
-public class G1CollectedHeap extends SharedHeap {
+public class G1CollectedHeap extends CollectedHeap {
     // HeapRegionManager _hrm;
     static private long hrmFieldOffset;
     // MemRegion _g1_reserved;
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_interface/CollectedHeap.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_interface/CollectedHeap.java	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2005, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -32,7 +32,7 @@
 import sun.jvm.hotspot.runtime.*;
 import sun.jvm.hotspot.types.*;
 
-public class CollectedHeap extends VMObject {
+public abstract class CollectedHeap extends VMObject {
   private static long         reservedFieldOffset;
 
   static {
@@ -73,9 +73,7 @@
     return reservedRegion().contains(a);
   }
 
-  public CollectedHeapName kind() {
-    return CollectedHeapName.ABSTRACT;
-  }
+  public abstract CollectedHeapName kind();
 
   public void print() { printOn(System.out); }
   public void printOn(PrintStream tty) {
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_interface/CollectedHeapName.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/gc_interface/CollectedHeapName.java	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -31,8 +31,6 @@
 
   private CollectedHeapName(String name) { this.name = name; }
 
-  public static final CollectedHeapName ABSTRACT = new CollectedHeapName("abstract");
-  public static final CollectedHeapName SHARED_HEAP = new CollectedHeapName("SharedHeap");
   public static final CollectedHeapName GEN_COLLECTED_HEAP = new CollectedHeapName("GenCollectedHeap");
   public static final CollectedHeapName G1_COLLECTED_HEAP = new CollectedHeapName("G1CollectedHeap");
   public static final CollectedHeapName PARALLEL_SCAVENGE_HEAP = new CollectedHeapName("ParallelScavengeHeap");
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/memory/GenCollectedHeap.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/memory/GenCollectedHeap.java	Fri Apr 17 23:55:04 2015 +0000
@@ -33,8 +33,7 @@
 import sun.jvm.hotspot.types.*;
 import sun.jvm.hotspot.utilities.*;
 
-public class GenCollectedHeap extends SharedHeap {
-  private static CIntegerField nGensField;
+public class GenCollectedHeap extends CollectedHeap {
   private static AddressField youngGenField;
   private static AddressField oldGenField;
 
@@ -54,7 +53,6 @@
   private static synchronized void initialize(TypeDataBase db) {
     Type type = db.lookupType("GenCollectedHeap");
 
-    nGensField = type.getCIntegerField("_n_gens");
     youngGenField = type.getAddressField("_young_gen");
     oldGenField = type.getAddressField("_old_gen");
 
@@ -70,7 +68,7 @@
   }
 
   public int nGens() {
-    return (int) nGensField.getValue(addr);
+    return 2; // Young + Old
   }
 
   public Generation getGen(int i) {
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/memory/SharedHeap.java	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,58 +0,0 @@
-/*
- * Copyright (c) 2002, 2012, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-package sun.jvm.hotspot.memory;
-
-import java.io.*;
-import java.util.*;
-
-import sun.jvm.hotspot.debugger.*;
-import sun.jvm.hotspot.gc_interface.*;
-import sun.jvm.hotspot.runtime.*;
-import sun.jvm.hotspot.types.*;
-
-public abstract class SharedHeap extends CollectedHeap {
-  private static VirtualConstructor ctor;
-
-  static {
-    VM.registerVMInitializedObserver(new Observer() {
-        public void update(Observable o, Object data) {
-          initialize(VM.getVM().getTypeDataBase());
-        }
-      });
-  }
-
-  private static synchronized void initialize(TypeDataBase db) {
-    Type type = db.lookupType("SharedHeap");
-    ctor = new VirtualConstructor(db);
-  }
-
-  public SharedHeap(Address addr) {
-    super(addr);
-  }
-
-  public CollectedHeapName kind() {
-    return CollectedHeapName.SHARED_HEAP;
-  }
-  }
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/memory/Universe.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/memory/Universe.java	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -112,11 +112,7 @@
     return "";
   }
   public CollectedHeap heap() {
-    try {
-      return (CollectedHeap) heapConstructor.instantiateWrapperFor(collectedHeapField.getValue());
-    } catch (WrongTypeException e) {
-      return new CollectedHeap(collectedHeapField.getValue());
-    }
+    return (CollectedHeap) heapConstructor.instantiateWrapperFor(collectedHeapField.getValue());
   }
 
   public static long getNarrowOopBase() {
--- a/hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/agent/src/share/classes/sun/jvm/hotspot/tools/HeapSummary.java	Fri Apr 17 23:55:04 2015 +0000
@@ -81,53 +81,48 @@
       System.out.println();
       System.out.println("Heap Usage:");
 
-      if (heap instanceof SharedHeap) {
-         SharedHeap sharedHeap = (SharedHeap) heap;
-         if (sharedHeap instanceof GenCollectedHeap) {
-            GenCollectedHeap genHeap = (GenCollectedHeap) sharedHeap;
-            for (int n = 0; n < genHeap.nGens(); n++) {
-               Generation gen = genHeap.getGen(n);
-               if (gen instanceof sun.jvm.hotspot.memory.DefNewGeneration) {
-                  System.out.println("New Generation (Eden + 1 Survivor Space):");
-                  printGen(gen);
+      if (heap instanceof GenCollectedHeap) {
+         GenCollectedHeap genHeap = (GenCollectedHeap) heap;
+         for (int n = 0; n < genHeap.nGens(); n++) {
+            Generation gen = genHeap.getGen(n);
+            if (gen instanceof sun.jvm.hotspot.memory.DefNewGeneration) {
+               System.out.println("New Generation (Eden + 1 Survivor Space):");
+               printGen(gen);
 
-                  ContiguousSpace eden = ((DefNewGeneration)gen).eden();
-                  System.out.println("Eden Space:");
-                  printSpace(eden);
+               ContiguousSpace eden = ((DefNewGeneration)gen).eden();
+               System.out.println("Eden Space:");
+               printSpace(eden);
 
-                  ContiguousSpace from = ((DefNewGeneration)gen).from();
-                  System.out.println("From Space:");
-                  printSpace(from);
+               ContiguousSpace from = ((DefNewGeneration)gen).from();
+               System.out.println("From Space:");
+               printSpace(from);
 
-                  ContiguousSpace to = ((DefNewGeneration)gen).to();
-                  System.out.println("To Space:");
-                  printSpace(to);
-               } else {
-                  System.out.println(gen.name() + ":");
-                  printGen(gen);
-               }
+               ContiguousSpace to = ((DefNewGeneration)gen).to();
+               System.out.println("To Space:");
+               printSpace(to);
+            } else {
+               System.out.println(gen.name() + ":");
+               printGen(gen);
             }
-         } else if (sharedHeap instanceof G1CollectedHeap) {
-             G1CollectedHeap g1h = (G1CollectedHeap) sharedHeap;
-             G1MonitoringSupport g1mm = g1h.g1mm();
-             long edenRegionNum = g1mm.edenRegionNum();
-             long survivorRegionNum = g1mm.survivorRegionNum();
-             HeapRegionSetBase oldSet = g1h.oldSet();
-             HeapRegionSetBase humongousSet = g1h.humongousSet();
-             long oldRegionNum = oldSet.count().length()
-                          + humongousSet.count().capacity() / HeapRegion.grainBytes();
-             printG1Space("G1 Heap:", g1h.n_regions(),
-                          g1h.used(), g1h.capacity());
-             System.out.println("G1 Young Generation:");
-             printG1Space("Eden Space:", edenRegionNum,
-                          g1mm.edenUsed(), g1mm.edenCommitted());
-             printG1Space("Survivor Space:", survivorRegionNum,
-                          g1mm.survivorUsed(), g1mm.survivorCommitted());
-             printG1Space("G1 Old Generation:", oldRegionNum,
-                          g1mm.oldUsed(), g1mm.oldCommitted());
-         } else {
-             throw new RuntimeException("unknown SharedHeap type : " + heap.getClass());
          }
+      } else if (heap instanceof G1CollectedHeap) {
+          G1CollectedHeap g1h = (G1CollectedHeap) heap;
+          G1MonitoringSupport g1mm = g1h.g1mm();
+          long edenRegionNum = g1mm.edenRegionNum();
+          long survivorRegionNum = g1mm.survivorRegionNum();
+          HeapRegionSetBase oldSet = g1h.oldSet();
+          HeapRegionSetBase humongousSet = g1h.humongousSet();
+          long oldRegionNum = oldSet.count().length()
+                       + humongousSet.count().capacity() / HeapRegion.grainBytes();
+          printG1Space("G1 Heap:", g1h.n_regions(),
+                       g1h.used(), g1h.capacity());
+          System.out.println("G1 Young Generation:");
+          printG1Space("Eden Space:", edenRegionNum,
+                       g1mm.edenUsed(), g1mm.edenCommitted());
+          printG1Space("Survivor Space:", survivorRegionNum,
+                       g1mm.survivorUsed(), g1mm.survivorCommitted());
+          printG1Space("G1 Old Generation:", oldRegionNum,
+                       g1mm.oldUsed(), g1mm.oldCommitted());
       } else if (heap instanceof ParallelScavengeHeap) {
          ParallelScavengeHeap psh = (ParallelScavengeHeap) heap;
          PSYoungGen youngGen = psh.youngGen();
--- a/hotspot/make/aix/makefiles/adlc.make	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/aix/makefiles/adlc.make	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -140,13 +140,7 @@
 # Note "+="; it is a hook so flags.make can add more flags, like -g or -DFOO.
 ADLCFLAGS += -q -T
 
-# Normally, debugging is done directly on the ad_<arch>*.cpp files.
-# But -g will put #line directives in those files pointing back to <arch>.ad.
-# Some builds of gcc 3.2 have a bug that gets tickled by the extra #line directives
-# so skip it for 3.2 and ealier.
-ifneq "$(shell expr \( $(CC_VER_MAJOR) \> 3 \) \| \( \( $(CC_VER_MAJOR) = 3 \) \& \( $(CC_VER_MINOR) \>= 3 \) \))" "0"
 ADLCFLAGS += -g
-endif
 
 ifdef LP64
 ADLCFLAGS += -D_LP64
--- a/hotspot/make/aix/makefiles/mapfile-vers-debug	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/aix/makefiles/mapfile-vers-debug	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/aix/makefiles/mapfile-vers-product	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/aix/makefiles/mapfile-vers-product	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetInheritedAccessControlContext;
--- a/hotspot/make/aix/makefiles/ppc64.make	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/aix/makefiles/ppc64.make	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 #
-# Copyright (c) 2004, 2013, Oracle and/or its affiliates. All rights reserved.
-# Copyright 2012, 2013 SAP AG. All rights reserved.
+# Copyright (c) 2004, 2015, Oracle and/or its affiliates. All rights reserved.
+# Copyright 2012, 2015 SAP AG. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -71,9 +71,6 @@
 OPT_CFLAGS/sharedRuntimeTrig.o = $(OPT_CFLAGS/NOOPT)
 OPT_CFLAGS/sharedRuntimeTrans.o = $(OPT_CFLAGS/NOOPT)
 
-# xlc 10.01 parameters for ipa compile.
-QIPA_COMPILE=$(if $(CXX_IS_V10),-qipa)
-
 # Xlc 10.1 parameters for aggressive optimization:
 # - qhot=level=1: Most aggressive loop optimizations.
 # - qignerrno: Assume errno is not modified by system calls.
@@ -88,7 +85,7 @@
 OPT_CFLAGS/synchronizer.o = $(OPT_CFLAGS) -qnoinline
 
 # Set all the xlC V10.1 options here.
-OPT_CFLAGS += $(QIPA_COMPILE) $(QV10_OPT) $(QV10_OPT_AGGRESSIVE)
+OPT_CFLAGS += $(QV10_OPT) $(QV10_OPT_AGGRESSIVE)
 
 export OBJECT_MODE=64
 
--- a/hotspot/make/aix/makefiles/xlc.make	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/aix/makefiles/xlc.make	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 #
-# Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
-# Copyright (c) 2012, 2013 SAP. All rights reserved.
+# Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
+# Copyright (c) 2012, 2015 SAP. All rights reserved.
 # DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 #
 # This code is free software; you can redistribute it and/or modify it
@@ -34,13 +34,17 @@
 
 AS  = $(CC) -c
 
-# get xlc version
-CXX_VERSION   := $(shell $(CXX) -qversion 2>&1 | sed -n 's/.*Version: \([0-9.]*\)/\1/p')
+# get xlc version which comes as VV.RR.MMMM.LLLL where 'VV' is the version,
+# 'RR' is the release, 'MMMM' is the modification and 'LLLL' is the level.
+# We only use 'VV.RR.LLLL' to avoid integer overflows in bash when comparing
+# the version numbers (some shells only support 32-bit integer compares!).
+CXX_VERSION := $(shell $(CXX) -qversion 2>&1 | \
+                   sed -n 's/.*Version: \([0-9]\{2\}\).\([0-9]\{2\}\).[0-9]\{4\}.\([0-9]\{4\}\)/\1\2\3/p')
 
 # xlc 08.00.0000.0023 and higher supports -qtune=balanced
-CXX_SUPPORTS_BALANCED_TUNING=$(shell if [ $(subst .,,$(CXX_VERSION)) -ge 080000000023 ] ; then echo "true" ; fi)
+CXX_SUPPORTS_BALANCED_TUNING := $(shell if [ $(CXX_VERSION) -ge 08000023 ] ; then echo "true" ; fi)
 # xlc 10.01 is used with aggressive optimizations to boost performance
-CXX_IS_V10=$(shell if [ $(subst .,,$(CXX_VERSION)) -ge 100100000000 ] ; then echo "true" ; fi)
+CXX_IS_V10 := $(shell if [ $(CXX_VERSION) -ge 10010000 ] ; then echo "true" ; fi)
 
 # check for precompiled headers support
 
--- a/hotspot/make/bsd/makefiles/mapfile-vers-darwin-debug	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/bsd/makefiles/mapfile-vers-darwin-debug	Fri Apr 17 23:55:04 2015 +0000
@@ -105,6 +105,7 @@
                 _JVM_GetClassTypeAnnotations
                 _JVM_GetDeclaredClasses
                 _JVM_GetDeclaringClass
+                _JVM_GetSimpleBinaryName
                 _JVM_GetEnclosingMethodInfo
                 _JVM_GetFieldIxModifiers
                 _JVM_GetFieldTypeAnnotations
--- a/hotspot/make/bsd/makefiles/mapfile-vers-darwin-product	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/bsd/makefiles/mapfile-vers-darwin-product	Fri Apr 17 23:55:04 2015 +0000
@@ -105,6 +105,7 @@
                 _JVM_GetClassTypeAnnotations
                 _JVM_GetDeclaredClasses
                 _JVM_GetDeclaringClass
+                _JVM_GetSimpleBinaryName
                 _JVM_GetEnclosingMethodInfo
                 _JVM_GetFieldIxModifiers
                 _JVM_GetFieldTypeAnnotations
--- a/hotspot/make/bsd/makefiles/mapfile-vers-debug	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/bsd/makefiles/mapfile-vers-debug	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/bsd/makefiles/mapfile-vers-product	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/bsd/makefiles/mapfile-vers-product	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/build.sh	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/build.sh	Fri Apr 17 23:55:04 2015 +0000
@@ -40,7 +40,7 @@
     exit 1
 fi
 
-if [ "${JAVA_HOME-}" = ""  -o  ! -d "${JAVA_HOME-}" -o ! -d ${JAVA_HOME-}/jre/lib/ ]; then
+if [ "${JAVA_HOME-}" = ""  -o  ! -d "${JAVA_HOME-}" ]; then
     echo "JAVA_HOME needs to be set to a valid JDK path"
     echo "JAVA_HOME: ${JAVA_HOME-}"
     exit 1
--- a/hotspot/make/linux/makefiles/mapfile-vers-debug	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/linux/makefiles/mapfile-vers-debug	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/linux/makefiles/mapfile-vers-product	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/linux/makefiles/mapfile-vers-product	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/solaris/makefiles/mapfile-vers	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/solaris/makefiles/mapfile-vers	Fri Apr 17 23:55:04 2015 +0000
@@ -19,7 +19,7 @@
 # Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 # or visit www.oracle.com if you need additional information or have any
 # questions.
-#  
+#
 #
 
 # Define public interface.
@@ -30,7 +30,7 @@
                 JNI_CreateJavaVM;
                 JNI_GetCreatedJavaVMs;
                 JNI_GetDefaultJavaVMInitArgs;
-        
+
                 # JVM
                 JVM_ActiveProcessorCount;
                 JVM_ArrayCopy;
@@ -107,6 +107,7 @@
                 JVM_GetClassTypeAnnotations;
                 JVM_GetDeclaredClasses;
                 JVM_GetDeclaringClass;
+                JVM_GetSimpleBinaryName;
                 JVM_GetEnclosingMethodInfo;
                 JVM_GetFieldIxModifiers;
                 JVM_GetFieldTypeAnnotations;
--- a/hotspot/make/solaris/makefiles/product.make	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/make/solaris/makefiles/product.make	Fri Apr 17 23:55:04 2015 +0000
@@ -37,6 +37,11 @@
 OPT_CFLAGS/ciEnv.o = $(OPT_CFLAGS) -xinline=no%__1cFciEnvbFpost_compiled_method_load_event6MpnHnmethod__v_
 endif
 
+# Need extra inlining to get oop_ps_push_contents functions to perform well enough.
+ifndef USE_GCC
+OPT_CFLAGS/psPromotionManager.o = $(OPT_CFLAGS) -W2,-Ainline:inc=1000
+endif
+
 # (OPT_CFLAGS/SLOWER is also available, to alter compilation of buggy files)
 ifeq ("${Platform_compiler}", "sparcWorks")
 
--- a/hotspot/src/cpu/aarch64/vm/aarch64.ad	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/aarch64.ad	Fri Apr 17 23:55:04 2015 +0000
@@ -1818,6 +1818,8 @@
     case INDINDEXSCALEDI2L:
     case INDINDEXSCALEDOFFSETI2LN:
     case INDINDEXSCALEDI2LN:
+    case INDINDEXOFFSETI2L:
+    case INDINDEXOFFSETI2LN:
       scale = Address::sxtw(size);
       break;
     default:
@@ -4264,6 +4266,20 @@
   %}
 %}
 
+operand indIndexOffsetI2L(iRegP reg, iRegI ireg, immLU12 off)
+%{
+  constraint(ALLOC_IN_RC(ptr_reg));
+  match(AddP (AddP reg (ConvI2L ireg)) off);
+  op_cost(INSN_COST);
+  format %{ "$reg, $ireg, $off I2L" %}
+  interface(MEMORY_INTER) %{
+    base($reg);
+    index($ireg);
+    scale(0x0);
+    disp($off);
+  %}
+%}
+
 operand indIndexScaledOffsetI2L(iRegP reg, iRegI ireg, immIScale scale, immLU12 off)
 %{
   constraint(ALLOC_IN_RC(ptr_reg));
@@ -4324,7 +4340,7 @@
 %{
   constraint(ALLOC_IN_RC(ptr_reg));
   match(AddP reg off);
-  op_cost(INSN_COST);
+  op_cost(0);
   format %{ "[$reg, $off]" %}
   interface(MEMORY_INTER) %{
     base($reg);
@@ -4394,6 +4410,21 @@
   %}
 %}
 
+operand indIndexOffsetI2LN(iRegN reg, iRegI ireg, immLU12 off)
+%{
+  predicate(Universe::narrow_oop_shift() == 0);
+  constraint(ALLOC_IN_RC(ptr_reg));
+  match(AddP (AddP (DecodeN reg) (ConvI2L ireg)) off);
+  op_cost(INSN_COST);
+  format %{ "$reg, $ireg, $off I2L\t# narrow" %}
+  interface(MEMORY_INTER) %{
+    base($reg);
+    index($ireg);
+    scale(0x0);
+    disp($off);
+  %}
+%}
+
 operand indIndexScaledOffsetI2LN(iRegN reg, iRegI ireg, immIScale scale, immLU12 off)
 %{
   predicate(Universe::narrow_oop_shift() == 0);
@@ -4656,8 +4687,8 @@
 // memory is used to define read/write location for load/store
 // instruction defs. we can turn a memory op into an Address
 
-opclass memory(indirect, indIndexScaledOffsetI,  indIndexScaledOffsetL, indIndexScaledOffsetI2L, indIndexScaled, indIndexScaledI2L, indIndex, indOffI, indOffL,
-               indirectN, indIndexScaledOffsetIN,  indIndexScaledOffsetLN, indIndexScaledOffsetI2LN, indIndexScaledN, indIndexScaledI2LN, indIndexN, indOffIN, indOffLN);
+opclass memory(indirect, indIndexScaledOffsetI, indIndexScaledOffsetL, indIndexOffsetI2L, indIndexScaledOffsetI2L, indIndexScaled, indIndexScaledI2L, indIndex, indOffI, indOffL,
+               indirectN, indIndexScaledOffsetIN, indIndexScaledOffsetLN, indIndexOffsetI2LN, indIndexScaledOffsetI2LN, indIndexScaledN, indIndexScaledI2LN, indIndexN, indOffIN, indOffLN);
 
 
 // iRegIorL2I is used for src inputs in rules for 32 bit int (I)
@@ -7523,7 +7554,7 @@
 instruct addP_reg_reg_ext(iRegPNoSp dst, iRegP src1, iRegIorL2I src2) %{
   match(Set dst (AddP src1 (ConvI2L src2)));
 
-  ins_cost(INSN_COST);
+  ins_cost(1.9 * INSN_COST);
   format %{ "add $dst, $src1, $src2, sxtw\t# ptr" %}
 
   ins_encode %{
--- a/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/assembler_aarch64.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 1997, 2011, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 2014, 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1469,7 +1469,7 @@
     f(op, 31, 29);
     f(0b11010000, 28, 21);
     f(0b000000, 15, 10);
-    rf(Rm, 16), rf(Rn, 5), rf(Rd, 0);
+    zrf(Rm, 16), zrf(Rn, 5), zrf(Rd, 0);
   }
 
   #define INSN(NAME, op)                                \
--- a/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/interp_masm_aarch64.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 2003, 2011, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 2014, 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -40,6 +40,8 @@
 
  protected:
   // Interpreter specific version of call_VM_base
+  using MacroAssembler::call_VM_leaf_base;
+
   virtual void call_VM_leaf_base(address entry_point,
                                  int number_of_arguments);
 
--- a/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014, 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2246,6 +2246,341 @@
 }
 
 /**
+ * Helpers for multiply_to_len().
+ */
+void MacroAssembler::add2_with_carry(Register final_dest_hi, Register dest_hi, Register dest_lo,
+                                     Register src1, Register src2) {
+  adds(dest_lo, dest_lo, src1);
+  adc(dest_hi, dest_hi, zr);
+  adds(dest_lo, dest_lo, src2);
+  adc(final_dest_hi, dest_hi, zr);
+}
+
+// Generate an address from (r + r1 extend offset).  "size" is the
+// size of the operand.  The result may be in rscratch2.
+Address MacroAssembler::offsetted_address(Register r, Register r1,
+                                          Address::extend ext, int offset, int size) {
+  if (offset || (ext.shift() % size != 0)) {
+    lea(rscratch2, Address(r, r1, ext));
+    return Address(rscratch2, offset);
+  } else {
+    return Address(r, r1, ext);
+  }
+}
+
+/**
+ * Multiply 64 bit by 64 bit first loop.
+ */
+void MacroAssembler::multiply_64_x_64_loop(Register x, Register xstart, Register x_xstart,
+                                           Register y, Register y_idx, Register z,
+                                           Register carry, Register product,
+                                           Register idx, Register kdx) {
+  //
+  //  jlong carry, x[], y[], z[];
+  //  for (int idx=ystart, kdx=ystart+1+xstart; idx >= 0; idx-, kdx--) {
+  //    huge_128 product = y[idx] * x[xstart] + carry;
+  //    z[kdx] = (jlong)product;
+  //    carry  = (jlong)(product >>> 64);
+  //  }
+  //  z[xstart] = carry;
+  //
+
+  Label L_first_loop, L_first_loop_exit;
+  Label L_one_x, L_one_y, L_multiply;
+
+  subsw(xstart, xstart, 1);
+  br(Assembler::MI, L_one_x);
+
+  lea(rscratch1, Address(x, xstart, Address::lsl(LogBytesPerInt)));
+  ldr(x_xstart, Address(rscratch1));
+  ror(x_xstart, x_xstart, 32); // convert big-endian to little-endian
+
+  bind(L_first_loop);
+  subsw(idx, idx, 1);
+  br(Assembler::MI, L_first_loop_exit);
+  subsw(idx, idx, 1);
+  br(Assembler::MI, L_one_y);
+  lea(rscratch1, Address(y, idx, Address::uxtw(LogBytesPerInt)));
+  ldr(y_idx, Address(rscratch1));
+  ror(y_idx, y_idx, 32); // convert big-endian to little-endian
+  bind(L_multiply);
+
+  // AArch64 has a multiply-accumulate instruction that we can't use
+  // here because it has no way to process carries, so we have to use
+  // separate add and adc instructions.  Bah.
+  umulh(rscratch1, x_xstart, y_idx); // x_xstart * y_idx -> rscratch1:product
+  mul(product, x_xstart, y_idx);
+  adds(product, product, carry);
+  adc(carry, rscratch1, zr);   // x_xstart * y_idx + carry -> carry:product
+
+  subw(kdx, kdx, 2);
+  ror(product, product, 32); // back to big-endian
+  str(product, offsetted_address(z, kdx, Address::uxtw(LogBytesPerInt), 0, BytesPerLong));
+
+  b(L_first_loop);
+
+  bind(L_one_y);
+  ldrw(y_idx, Address(y,  0));
+  b(L_multiply);
+
+  bind(L_one_x);
+  ldrw(x_xstart, Address(x,  0));
+  b(L_first_loop);
+
+  bind(L_first_loop_exit);
+}
+
+/**
+ * Multiply 128 bit by 128. Unrolled inner loop.
+ *
+ */
+void MacroAssembler::multiply_128_x_128_loop(Register y, Register z,
+                                             Register carry, Register carry2,
+                                             Register idx, Register jdx,
+                                             Register yz_idx1, Register yz_idx2,
+                                             Register tmp, Register tmp3, Register tmp4,
+                                             Register tmp6, Register product_hi) {
+
+  //   jlong carry, x[], y[], z[];
+  //   int kdx = ystart+1;
+  //   for (int idx=ystart-2; idx >= 0; idx -= 2) { // Third loop
+  //     huge_128 tmp3 = (y[idx+1] * product_hi) + z[kdx+idx+1] + carry;
+  //     jlong carry2  = (jlong)(tmp3 >>> 64);
+  //     huge_128 tmp4 = (y[idx]   * product_hi) + z[kdx+idx] + carry2;
+  //     carry  = (jlong)(tmp4 >>> 64);
+  //     z[kdx+idx+1] = (jlong)tmp3;
+  //     z[kdx+idx] = (jlong)tmp4;
+  //   }
+  //   idx += 2;
+  //   if (idx > 0) {
+  //     yz_idx1 = (y[idx] * product_hi) + z[kdx+idx] + carry;
+  //     z[kdx+idx] = (jlong)yz_idx1;
+  //     carry  = (jlong)(yz_idx1 >>> 64);
+  //   }
+  //
+
+  Label L_third_loop, L_third_loop_exit, L_post_third_loop_done;
+
+  lsrw(jdx, idx, 2);
+
+  bind(L_third_loop);
+
+  subsw(jdx, jdx, 1);
+  br(Assembler::MI, L_third_loop_exit);
+  subw(idx, idx, 4);
+
+  lea(rscratch1, Address(y, idx, Address::uxtw(LogBytesPerInt)));
+
+  ldp(yz_idx2, yz_idx1, Address(rscratch1, 0));
+
+  lea(tmp6, Address(z, idx, Address::uxtw(LogBytesPerInt)));
+
+  ror(yz_idx1, yz_idx1, 32); // convert big-endian to little-endian
+  ror(yz_idx2, yz_idx2, 32);
+
+  ldp(rscratch2, rscratch1, Address(tmp6, 0));
+
+  mul(tmp3, product_hi, yz_idx1);  //  yz_idx1 * product_hi -> tmp4:tmp3
+  umulh(tmp4, product_hi, yz_idx1);
+
+  ror(rscratch1, rscratch1, 32); // convert big-endian to little-endian
+  ror(rscratch2, rscratch2, 32);
+
+  mul(tmp, product_hi, yz_idx2);   //  yz_idx2 * product_hi -> carry2:tmp
+  umulh(carry2, product_hi, yz_idx2);
+
+  // propagate sum of both multiplications into carry:tmp4:tmp3
+  adds(tmp3, tmp3, carry);
+  adc(tmp4, tmp4, zr);
+  adds(tmp3, tmp3, rscratch1);
+  adcs(tmp4, tmp4, tmp);
+  adc(carry, carry2, zr);
+  adds(tmp4, tmp4, rscratch2);
+  adc(carry, carry, zr);
+
+  ror(tmp3, tmp3, 32); // convert little-endian to big-endian
+  ror(tmp4, tmp4, 32);
+  stp(tmp4, tmp3, Address(tmp6, 0));
+
+  b(L_third_loop);
+  bind (L_third_loop_exit);
+
+  andw (idx, idx, 0x3);
+  cbz(idx, L_post_third_loop_done);
+
+  Label L_check_1;
+  subsw(idx, idx, 2);
+  br(Assembler::MI, L_check_1);
+
+  lea(rscratch1, Address(y, idx, Address::uxtw(LogBytesPerInt)));
+  ldr(yz_idx1, Address(rscratch1, 0));
+  ror(yz_idx1, yz_idx1, 32);
+  mul(tmp3, product_hi, yz_idx1);  //  yz_idx1 * product_hi -> tmp4:tmp3
+  umulh(tmp4, product_hi, yz_idx1);
+  lea(rscratch1, Address(z, idx, Address::uxtw(LogBytesPerInt)));
+  ldr(yz_idx2, Address(rscratch1, 0));
+  ror(yz_idx2, yz_idx2, 32);
+
+  add2_with_carry(carry, tmp4, tmp3, carry, yz_idx2);
+
+  ror(tmp3, tmp3, 32);
+  str(tmp3, Address(rscratch1, 0));
+
+  bind (L_check_1);
+
+  andw (idx, idx, 0x1);
+  subsw(idx, idx, 1);
+  br(Assembler::MI, L_post_third_loop_done);
+  ldrw(tmp4, Address(y, idx, Address::uxtw(LogBytesPerInt)));
+  mul(tmp3, tmp4, product_hi);  //  tmp4 * product_hi -> carry2:tmp3
+  umulh(carry2, tmp4, product_hi);
+  ldrw(tmp4, Address(z, idx, Address::uxtw(LogBytesPerInt)));
+
+  add2_with_carry(carry2, tmp3, tmp4, carry);
+
+  strw(tmp3, Address(z, idx, Address::uxtw(LogBytesPerInt)));
+  extr(carry, carry2, tmp3, 32);
+
+  bind(L_post_third_loop_done);
+}
+
+/**
+ * Code for BigInteger::multiplyToLen() instrinsic.
+ *
+ * r0: x
+ * r1: xlen
+ * r2: y
+ * r3: ylen
+ * r4:  z
+ * r5: zlen
+ * r10: tmp1
+ * r11: tmp2
+ * r12: tmp3
+ * r13: tmp4
+ * r14: tmp5
+ * r15: tmp6
+ * r16: tmp7
+ *
+ */
+void MacroAssembler::multiply_to_len(Register x, Register xlen, Register y, Register ylen,
+                                     Register z, Register zlen,
+                                     Register tmp1, Register tmp2, Register tmp3, Register tmp4,
+                                     Register tmp5, Register tmp6, Register product_hi) {
+
+  assert_different_registers(x, xlen, y, ylen, z, zlen, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6);
+
+  const Register idx = tmp1;
+  const Register kdx = tmp2;
+  const Register xstart = tmp3;
+
+  const Register y_idx = tmp4;
+  const Register carry = tmp5;
+  const Register product  = xlen;
+  const Register x_xstart = zlen;  // reuse register
+
+  // First Loop.
+  //
+  //  final static long LONG_MASK = 0xffffffffL;
+  //  int xstart = xlen - 1;
+  //  int ystart = ylen - 1;
+  //  long carry = 0;
+  //  for (int idx=ystart, kdx=ystart+1+xstart; idx >= 0; idx-, kdx--) {
+  //    long product = (y[idx] & LONG_MASK) * (x[xstart] & LONG_MASK) + carry;
+  //    z[kdx] = (int)product;
+  //    carry = product >>> 32;
+  //  }
+  //  z[xstart] = (int)carry;
+  //
+
+  movw(idx, ylen);      // idx = ylen;
+  movw(kdx, zlen);      // kdx = xlen+ylen;
+  mov(carry, zr);       // carry = 0;
+
+  Label L_done;
+
+  movw(xstart, xlen);
+  subsw(xstart, xstart, 1);
+  br(Assembler::MI, L_done);
+
+  multiply_64_x_64_loop(x, xstart, x_xstart, y, y_idx, z, carry, product, idx, kdx);
+
+  Label L_second_loop;
+  cbzw(kdx, L_second_loop);
+
+  Label L_carry;
+  subw(kdx, kdx, 1);
+  cbzw(kdx, L_carry);
+
+  strw(carry, Address(z, kdx, Address::uxtw(LogBytesPerInt)));
+  lsr(carry, carry, 32);
+  subw(kdx, kdx, 1);
+
+  bind(L_carry);
+  strw(carry, Address(z, kdx, Address::uxtw(LogBytesPerInt)));
+
+  // Second and third (nested) loops.
+  //
+  // for (int i = xstart-1; i >= 0; i--) { // Second loop
+  //   carry = 0;
+  //   for (int jdx=ystart, k=ystart+1+i; jdx >= 0; jdx--, k--) { // Third loop
+  //     long product = (y[jdx] & LONG_MASK) * (x[i] & LONG_MASK) +
+  //                    (z[k] & LONG_MASK) + carry;
+  //     z[k] = (int)product;
+  //     carry = product >>> 32;
+  //   }
+  //   z[i] = (int)carry;
+  // }
+  //
+  // i = xlen, j = tmp1, k = tmp2, carry = tmp5, x[i] = product_hi
+
+  const Register jdx = tmp1;
+
+  bind(L_second_loop);
+  mov(carry, zr);                // carry = 0;
+  movw(jdx, ylen);               // j = ystart+1
+
+  subsw(xstart, xstart, 1);      // i = xstart-1;
+  br(Assembler::MI, L_done);
+
+  str(z, Address(pre(sp, -4 * wordSize)));
+
+  Label L_last_x;
+  lea(z, offsetted_address(z, xstart, Address::uxtw(LogBytesPerInt), 4, BytesPerInt)); // z = z + k - j
+  subsw(xstart, xstart, 1);       // i = xstart-1;
+  br(Assembler::MI, L_last_x);
+
+  lea(rscratch1, Address(x, xstart, Address::uxtw(LogBytesPerInt)));
+  ldr(product_hi, Address(rscratch1));
+  ror(product_hi, product_hi, 32);  // convert big-endian to little-endian
+
+  Label L_third_loop_prologue;
+  bind(L_third_loop_prologue);
+
+  str(ylen, Address(sp, wordSize));
+  stp(x, xstart, Address(sp, 2 * wordSize));
+  multiply_128_x_128_loop(y, z, carry, x, jdx, ylen, product,
+                          tmp2, x_xstart, tmp3, tmp4, tmp6, product_hi);
+  ldp(z, ylen, Address(post(sp, 2 * wordSize)));
+  ldp(x, xlen, Address(post(sp, 2 * wordSize)));   // copy old xstart -> xlen
+
+  addw(tmp3, xlen, 1);
+  strw(carry, Address(z, tmp3, Address::uxtw(LogBytesPerInt)));
+  subsw(tmp3, tmp3, 1);
+  br(Assembler::MI, L_done);
+
+  lsr(carry, carry, 32);
+  strw(carry, Address(z, tmp3, Address::uxtw(LogBytesPerInt)));
+  b(L_second_loop);
+
+  // Next infrequent code is moved outside loops.
+  bind(L_last_x);
+  ldrw(product_hi, Address(x,  0));
+  b(L_third_loop_prologue);
+
+  bind(L_done);
+}
+
+/**
  * Emits code to update CRC-32 with a byte value according to constants in table
  *
  * @param [in,out]crc   Register containing the crc.
--- a/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/macroAssembler_aarch64.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 2014, 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -167,9 +167,8 @@
 
   // aliases defined in AARCH64 spec
 
-
   template<class T>
-  inline void  cmpw(Register Rd, T imm)  { subsw(zr, Rd, imm); }
+  inline void cmpw(Register Rd, T imm)  { subsw(zr, Rd, imm); }
   inline void cmp(Register Rd, unsigned imm)  { subs(zr, Rd, imm); }
 
   inline void cmnw(Register Rd, unsigned imm) { addsw(zr, Rd, imm); }
@@ -1121,9 +1120,34 @@
                       Register tmp1, Register tmp2,
                       Register tmp3, Register tmp4,
                       int int_cnt1, Register result);
-
+private:
+  void add2_with_carry(Register final_dest_hi, Register dest_hi, Register dest_lo,
+                       Register src1, Register src2);
+  void add2_with_carry(Register dest_hi, Register dest_lo, Register src1, Register src2) {
+    add2_with_carry(dest_hi, dest_hi, dest_lo, src1, src2);
+  }
+  void multiply_64_x_64_loop(Register x, Register xstart, Register x_xstart,
+                             Register y, Register y_idx, Register z,
+                             Register carry, Register product,
+                             Register idx, Register kdx);
+  void multiply_128_x_128_loop(Register y, Register z,
+                               Register carry, Register carry2,
+                               Register idx, Register jdx,
+                               Register yz_idx1, Register yz_idx2,
+                               Register tmp, Register tmp3, Register tmp4,
+                               Register tmp7, Register product_hi);
+public:
+  void multiply_to_len(Register x, Register xlen, Register y, Register ylen, Register z,
+                       Register zlen, Register tmp1, Register tmp2, Register tmp3,
+                       Register tmp4, Register tmp5, Register tmp6, Register tmp7);
   // ISB may be needed because of a safepoint
   void maybe_isb() { isb(); }
+
+private:
+  // Return the effective address r + (r1 << ext) + offset.
+  // Uses rscratch2.
+  Address offsetted_address(Register r, Register r1, Address::extend ext,
+                            int offset, int size);
 };
 
 // Used by aarch64.ad to control code generation
--- a/hotspot/src/cpu/aarch64/vm/stubGenerator_aarch64.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/stubGenerator_aarch64.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 2014, 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2356,8 +2356,45 @@
     return start;
   }
 
-#undef __
-#define __ masm->
+  /**
+   *  Arguments:
+   *
+   *  Input:
+   *    c_rarg0   - x address
+   *    c_rarg1   - x length
+   *    c_rarg2   - y address
+   *    c_rarg3   - y lenth
+   *    c_rarg4   - z address
+   *    c_rarg5   - z length
+   */
+  address generate_multiplyToLen() {
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "multiplyToLen");
+
+    address start = __ pc();
+    const Register x     = r0;
+    const Register xlen  = r1;
+    const Register y     = r2;
+    const Register ylen  = r3;
+    const Register z     = r4;
+    const Register zlen  = r5;
+
+    const Register tmp1  = r10;
+    const Register tmp2  = r11;
+    const Register tmp3  = r12;
+    const Register tmp4  = r13;
+    const Register tmp5  = r14;
+    const Register tmp6  = r15;
+    const Register tmp7  = r16;
+
+    BLOCK_COMMENT("Entry:");
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+    __ multiply_to_len(x, xlen, y, ylen, z, zlen, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7);
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(lr);
+
+    return start;
+  }
 
   // Continuation point for throwing of implicit exceptions that are
   // not handled in the current activation. Fabricates an exception
@@ -2375,6 +2412,9 @@
   // otherwise assume that stack unwinding will be initiated, so
   // caller saved registers were assumed volatile in the compiler.
 
+#undef __
+#define __ masm->
+
   address generate_throw_exception(const char* name,
                                    address runtime_entry,
                                    Register arg1 = noreg,
@@ -2518,6 +2558,10 @@
     // arraycopy stubs used by compilers
     generate_arraycopy_stubs();
 
+    if (UseMultiplyToLenIntrinsic) {
+      StubRoutines::_multiplyToLen = generate_multiplyToLen();
+    }
+
 #ifndef BUILTIN_SIM
     if (UseAESIntrinsics) {
       StubRoutines::_aescrypt_encryptBlock = generate_aescrypt_encryptBlock();
--- a/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/aarch64/vm/vm_version_aarch64.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2014, Red Hat Inc. All rights reserved.
+ * Copyright (c) 2015, Red Hat Inc. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -193,6 +193,15 @@
     }
   }
 
+  // This machine allows unaligned memory accesses
+  if (FLAG_IS_DEFAULT(UseUnalignedAccesses)) {
+    FLAG_SET_DEFAULT(UseUnalignedAccesses, true);
+  }
+
+  if (FLAG_IS_DEFAULT(UseMultiplyToLenIntrinsic)) {
+    UseMultiplyToLenIntrinsic = true;
+  }
+
 #ifdef COMPILER2
   if (FLAG_IS_DEFAULT(OptoScheduling)) {
     OptoScheduling = true;
--- a/hotspot/src/cpu/ppc/vm/methodHandles_ppc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/ppc/vm/methodHandles_ppc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -466,7 +466,7 @@
                  strstr(adaptername, "linkTo") == NULL);    // static linkers don't have MH
   const char* mh_reg_name = has_mh ? "R23_method_handle" : "G23";
   tty->print_cr("MH %s %s="INTPTR_FORMAT " sp=" INTPTR_FORMAT,
-                adaptername, mh_reg_name, (intptr_t) mh, entry_sp);
+                adaptername, mh_reg_name, p2i(mh), p2i(entry_sp));
 
   if (Verbose) {
     tty->print_cr("Registers:");
--- a/hotspot/src/cpu/ppc/vm/vm_version_ppc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/ppc/vm/vm_version_ppc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2014 SAP AG. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -172,6 +172,12 @@
     FLAG_SET_DEFAULT(UseSHA512Intrinsics, false);
   }
 
+  // This machine does not allow unaligned memory accesses
+  if (UseUnalignedAccesses) {
+    if (!FLAG_IS_DEFAULT(UseUnalignedAccesses))
+      warning("Unaligned memory access is not available on this CPU");
+    FLAG_SET_DEFAULT(UseUnalignedAccesses, false);
+  }
 }
 
 void VM_Version::print_features() {
--- a/hotspot/src/cpu/sparc/vm/vm_version_sparc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/sparc/vm/vm_version_sparc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -32,7 +32,7 @@
 
 int VM_Version::_features = VM_Version::unknown_m;
 const char* VM_Version::_features_str = "";
-unsigned int VM_Version::_L2_cache_line_size = 0;
+unsigned int VM_Version::_L2_data_cache_line_size = 0;
 
 void VM_Version::initialize() {
   _features = determine_features();
@@ -356,10 +356,17 @@
     (cache_line_size > ContendedPaddingWidth))
     ContendedPaddingWidth = cache_line_size;
 
+  // This machine does not allow unaligned memory accesses
+  if (UseUnalignedAccesses) {
+    if (!FLAG_IS_DEFAULT(UseUnalignedAccesses))
+      warning("Unaligned memory access is not available on this CPU");
+    FLAG_SET_DEFAULT(UseUnalignedAccesses, false);
+  }
+
 #ifndef PRODUCT
   if (PrintMiscellaneous && Verbose) {
     tty->print_cr("L1 data cache line size: %u", L1_data_cache_line_size());
-    tty->print_cr("L2 cache line size: %u", L2_cache_line_size());
+    tty->print_cr("L2 data cache line size: %u", L2_data_cache_line_size());
     tty->print("Allocation");
     if (AllocatePrefetchStyle <= 0) {
       tty->print_cr(": no prefetching");
--- a/hotspot/src/cpu/sparc/vm/vm_version_sparc.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/sparc/vm/vm_version_sparc.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -96,8 +96,8 @@
   static int  _features;
   static const char* _features_str;
 
-  static unsigned int _L2_cache_line_size;
-  static unsigned int L2_cache_line_size() { return _L2_cache_line_size; }
+  static unsigned int _L2_data_cache_line_size;
+  static unsigned int L2_data_cache_line_size() { return _L2_data_cache_line_size; }
 
   static void print_features();
   static int  determine_features();
@@ -171,7 +171,7 @@
   static const char* cpu_features()     { return _features_str; }
 
   // default prefetch block size on sparc
-  static intx prefetch_data_size()      { return L2_cache_line_size();  }
+  static intx prefetch_data_size()      { return L2_data_cache_line_size();  }
 
   // Prefetch
   static intx prefetch_copy_interval_in_bytes() {
--- a/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -3359,6 +3359,20 @@
 
 
 // Integer vector arithmetic
+void Assembler::vphaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
+  assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
+  int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38);
+  emit_int8(0x01);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
+void Assembler::vphaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
+  assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
+  int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38);
+  emit_int8(0x02);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
 void Assembler::paddb(XMMRegister dst, XMMRegister src) {
   NOT_LP64(assert(VM_Version::supports_sse2(), ""));
   emit_simd_arith(0xFC, dst, src, VEX_SIMD_66);
@@ -3379,6 +3393,20 @@
   emit_simd_arith(0xD4, dst, src, VEX_SIMD_66);
 }
 
+void Assembler::phaddw(XMMRegister dst, XMMRegister src) {
+  NOT_LP64(assert(VM_Version::supports_sse3(), ""));
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_int8(0x01);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
+void Assembler::phaddd(XMMRegister dst, XMMRegister src) {
+  NOT_LP64(assert(VM_Version::supports_sse3(), ""));
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_int8(0x02);
+  emit_int8((unsigned char)(0xC0 | encode));
+}
+
 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
   assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
   emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256);
@@ -3804,6 +3832,17 @@
   emit_int8(0x01);
 }
 
+void Assembler::vextractf128h(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_avx(), "");
+  bool vector256 = true;
+  int encode = vex_prefix_and_encode(src, xnoreg, dst, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A);
+  emit_int8(0x19);
+  emit_int8((unsigned char)(0xC0 | encode));
+  // 0x00 - insert into lower 128 bits
+  // 0x01 - insert into upper 128 bits
+  emit_int8(0x01);
+}
+
 void Assembler::vextractf128h(Address dst, XMMRegister src) {
   assert(VM_Version::supports_avx(), "");
   InstructionMark im(this);
--- a/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/assembler_x86.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1777,6 +1777,12 @@
   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
   void vxorps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
 
+  // Add horizontal packed integers
+  void vphaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
+  void vphaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
+  void phaddw(XMMRegister dst, XMMRegister src);
+  void phaddd(XMMRegister dst, XMMRegister src);
+
   // Add packed integers
   void paddb(XMMRegister dst, XMMRegister src);
   void paddw(XMMRegister dst, XMMRegister src);
@@ -1869,6 +1875,7 @@
   // Copy low 128bit into high 128bit of YMM registers.
   void vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
   void vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
+  void vextractf128h(XMMRegister dst, XMMRegister src);
 
   // Load/store high 128bit of YMM registers which does not destroy other half.
   void vinsertf128h(XMMRegister dst, Address src);
--- a/hotspot/src/cpu/x86/vm/globalDefinitions_x86.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/globalDefinitions_x86.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -58,4 +58,9 @@
   #endif
 #endif
 
+#if defined(COMPILER2) && !defined(JAVASE_EMBEDDED)
+// Include Restricted Transactional Memory lock eliding optimization
+#define INCLUDE_RTM_OPT 1
+#endif
+
 #endif // CPU_X86_VM_GLOBALDEFINITIONS_X86_HPP
--- a/hotspot/src/cpu/x86/vm/rtmLocking.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,60 +0,0 @@
-/*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#include "precompiled.hpp"
-#include "memory/allocation.inline.hpp"
-#include "runtime/task.hpp"
-#include "runtime/rtmLocking.hpp"
-
-// One-shot PeriodicTask subclass for enabling RTM locking
-uintx RTMLockingCounters::_calculation_flag = 0;
-
-class RTMLockingCalculationTask : public PeriodicTask {
- public:
-  RTMLockingCalculationTask(size_t interval_time) : PeriodicTask(interval_time){  }
-
-  virtual void task() {
-    RTMLockingCounters::_calculation_flag = 1;
-    // Reclaim our storage and disenroll ourself
-    delete this;
-  }
-};
-
-void RTMLockingCounters::init() {
-  if (UseRTMLocking && RTMLockingCalculationDelay > 0) {
-    RTMLockingCalculationTask* task = new RTMLockingCalculationTask(RTMLockingCalculationDelay);
-    task->enroll();
-  } else {
-    _calculation_flag = 1;
-  }
-}
-
-//------------------------------print_on-------------------------------
-void RTMLockingCounters::print_on(outputStream* st) {
-  tty->print_cr("# rtm locks total (estimated): " UINTX_FORMAT, _total_count * RTMTotalCountIncrRate);
-  tty->print_cr("# rtm lock aborts  : " UINTX_FORMAT, _abort_count);
-  for (int i = 0; i < ABORT_STATUS_LIMIT; i++) {
-    tty->print_cr("# rtm lock aborts %d: " UINTX_FORMAT, i, _abortX_count[i]);
-  }
-}
--- a/hotspot/src/cpu/x86/vm/vm_version_x86.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/vm_version_x86.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -969,6 +969,11 @@
      (cache_line_size > ContendedPaddingWidth))
      ContendedPaddingWidth = cache_line_size;
 
+  // This machine allows unaligned memory accesses
+  if (FLAG_IS_DEFAULT(UseUnalignedAccesses)) {
+    FLAG_SET_DEFAULT(UseUnalignedAccesses, true);
+  }
+
 #ifndef PRODUCT
   if (PrintMiscellaneous && Verbose) {
     tty->print_cr("Logical CPUs per core: %u",
--- a/hotspot/src/cpu/x86/vm/x86.ad	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/x86.ad	Fri Apr 17 23:55:04 2015 +0000
@@ -490,7 +490,7 @@
 class NativeJump;
 
 class CallStubImpl {
- 
+
   //--------------------------------------------------------------
   //---<  Used for optimization in Compile::shorten_branches  >---
   //--------------------------------------------------------------
@@ -500,9 +500,9 @@
   static uint size_call_trampoline() {
     return 0; // no call trampolines on this platform
   }
-  
+
   // number of relocations needed by a call trampoline stub
-  static uint reloc_call_trampoline() { 
+  static uint reloc_call_trampoline() {
     return 0; // no call trampolines on this platform
   }
 };
@@ -623,6 +623,22 @@
       if ((UseSSE < 4) && (UseAVX < 1)) // only with SSE4_1 or AVX
         return false;
     break;
+    case Op_AddReductionVL:
+      if (UseAVX < 3) // only EVEX : vector connectivity becomes an issue here
+        return false;
+    case Op_AddReductionVI:
+      if (UseSSE < 3) // requires at least SSE3
+        return false;
+    case Op_MulReductionVI:
+      if (UseSSE < 4) // requires at least SSE4
+        return false;
+    case Op_AddReductionVF:
+    case Op_AddReductionVD:
+    case Op_MulReductionVF:
+    case Op_MulReductionVD:
+      if (UseSSE < 1) // requires at least SSE
+        return false;
+    break;
     case Op_CompareAndSwapL:
 #ifdef _LP64
     case Op_CompareAndSwapP:
@@ -2532,6 +2548,574 @@
   ins_pipe( fpu_reg_reg );
 %}
 
+// ====================REDUCTION ARITHMETIC=======================================
+
+instruct rsadd2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE > 2 && UseAVX == 0);
+  match(Set dst (AddReductionVI src1 src2));
+  effect(TEMP tmp2, TEMP tmp);
+  format %{ "movdqu  $tmp2,$src2\n\t"
+            "phaddd  $tmp2,$tmp2\n\t"
+            "movd    $tmp,$src1\n\t"
+            "paddd   $tmp,$tmp2\n\t"
+            "movd    $dst,$tmp\t! add reduction2I" %}
+  ins_encode %{
+    __ movdqu($tmp2$$XMMRegister, $src2$$XMMRegister);
+    __ phaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdl($tmp$$XMMRegister, $src1$$Register);
+    __ paddd($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdl($dst$$Register, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vphaddd $tmp,$src2,$src2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpaddd  $tmp2,$tmp2,$tmp\n\t"
+            "movd    $dst,$tmp2\t! add reduction2I" %}
+  ins_encode %{
+    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsadd4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE > 2 && UseAVX == 0);
+  match(Set dst (AddReductionVI src1 src2));
+  effect(TEMP tmp2, TEMP tmp);
+  format %{ "movdqu  $tmp2,$src2\n\t"
+            "phaddd  $tmp2,$tmp2\n\t"
+            "phaddd  $tmp2,$tmp2\n\t"
+            "movd    $tmp,$src1\n\t"
+            "paddd   $tmp,$tmp2\n\t"
+            "movd    $dst,$tmp\t! add reduction4I" %}
+  ins_encode %{
+    __ movdqu($tmp2$$XMMRegister, $src2$$XMMRegister);
+    __ phaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister);
+    __ phaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdl($tmp$$XMMRegister, $src1$$Register);
+    __ paddd($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdl($dst$$Register, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vphaddd $tmp,$src2,$src2\n\t"
+            "vphaddd $tmp,$tmp,$tmp2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpaddd  $tmp2,$tmp2,$tmp\n\t"
+            "movd    $dst,$tmp2\t! add reduction4I" %}
+  ins_encode %{
+    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, false);
+    __ vphaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd8I_reduction_reg(rRegI dst, rRegI src1, vecY src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vphaddd $tmp,$src2,$src2\n\t"
+            "vphaddd $tmp,$tmp,$tmp2\n\t"
+            "vextractf128  $tmp2,$tmp\n\t"
+            "vpaddd  $tmp,$tmp,$tmp2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpaddd  $tmp2,$tmp2,$tmp\n\t"
+            "movd    $dst,$tmp2\t! add reduction8I" %}
+  ins_encode %{
+    __ vphaddd($tmp$$XMMRegister, $src2$$XMMRegister, $src2$$XMMRegister, true);
+    __ vphaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, true);
+    __ vextractf128h($tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ vpaddd($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpaddd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsadd2F_reduction_reg(regF dst, regF src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (AddReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "addss   $tmp,$src2\n\t"
+            "pshufd  $tmp2,$src2,0x01\n\t"
+            "addss   $tmp,$tmp2\n\t"
+            "movdqu  $dst,$tmp\t! add reduction2F" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ addss($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ addss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdqu($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd2F_reduction_reg(regF dst, regF src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVF src1 src2));
+  effect(TEMP tmp2, TEMP tmp);
+  format %{ "vaddss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vaddss  $dst,$tmp2,$tmp\t! add reduction2F" %}
+  ins_encode %{
+    __ vaddss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vaddss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsadd4F_reduction_reg(regF dst, regF src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (AddReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "addss   $tmp,$src2\n\t"
+            "pshufd  $tmp2,$src2,0x01\n\t"
+            "addss   $tmp,$tmp2\n\t"
+            "pshufd  $tmp2,$src2,0x02\n\t"
+            "addss   $tmp,$tmp2\n\t"
+            "pshufd  $tmp2,$src2,0x03\n\t"
+            "addss   $tmp,$tmp2\n\t"
+            "movdqu  $dst,$tmp\t! add reduction4F" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ addss($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ addss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ addss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ addss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdqu($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd4F_reduction_reg(regF dst, regF src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vaddss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x02\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x03\n\t"
+            "vaddss  $dst,$tmp2,$tmp\t! add reduction4F" %}
+  ins_encode %{
+    __ vaddss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ vaddss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct radd8F_reduction_reg(regF dst, regF src1, vecY src2, regF tmp, regF tmp2, regF tmp3) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
+  format %{ "vaddss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x02\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x03\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "vextractf128  $tmp3,$src2\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp3\n\t"
+            "pshufd  $tmp,$tmp3,0x01\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$tmp3,0x02\n\t"
+            "vaddss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$tmp3,0x03\n\t"
+            "vaddss  $dst,$tmp2,$tmp\t! add reduction8F" %}
+  ins_encode %{
+    __ vaddss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ vextractf128h($tmp3$$XMMRegister, $src2$$XMMRegister);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x01);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x02);
+    __ vaddss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x03);
+    __ vaddss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsadd2D_reduction_reg(regD dst, regD src1, vecX src2, regD tmp) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (AddReductionVD src1 src2));
+  effect(TEMP tmp, TEMP dst);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "addsd   $tmp,$src2\n\t"
+            "pshufd  $dst,$src2,0xE\n\t"
+            "addsd   $dst,$tmp\t! add reduction2D" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ addsd($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($dst$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ addsd($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd2D_reduction_reg(regD dst, regD src1, vecX src2, regD tmp, regD tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVD src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vaddsd  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0xE\n\t"
+            "vaddsd  $dst,$tmp2,$tmp\t! add reduction2D" %}
+  ins_encode %{
+    __ vaddsd($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ vaddsd($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvadd4D_reduction_reg(regD dst, regD src1, vecY src2, regD tmp, regD tmp2, regD tmp3) %{
+  predicate(UseAVX > 0);
+  match(Set dst (AddReductionVD src1 src2));
+  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
+  format %{ "vaddsd  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0xE\n\t"
+            "vaddsd  $tmp2,$tmp2,$tmp\n\t"
+            "vextractf128  $tmp3,$src2\n\t"
+            "vaddsd  $tmp2,$tmp2,$tmp3\n\t"
+            "pshufd  $tmp,$tmp3,0xE\n\t"
+            "vaddsd  $dst,$tmp2,$tmp\t! add reduction4D" %}
+  ins_encode %{
+    __ vaddsd($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ vaddsd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ vextractf128h($tmp3$$XMMRegister, $src2$$XMMRegister);
+    __ vaddsd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0xE);
+    __ vaddsd($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsmul2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE > 3 && UseAVX == 0);
+  match(Set dst (MulReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "pshufd  $tmp2,$src2,0x1\n\t"
+            "pmulld  $tmp2,$src2\n\t"
+            "movd    $tmp,$src1\n\t"
+            "pmulld  $tmp2,$tmp\n\t"
+            "movd    $dst,$tmp2\t! mul reduction2I" %}
+  ins_encode %{
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
+    __ pmulld($tmp2$$XMMRegister, $src2$$XMMRegister);
+    __ movdl($tmp$$XMMRegister, $src1$$Register);
+    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul2I_reduction_reg(rRegI dst, rRegI src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "pshufd  $tmp2,$src2,0x1\n\t"
+            "vpmulld $tmp,$src2,$tmp2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpmulld $tmp2,$tmp,$tmp2\n\t"
+            "movd    $dst,$tmp2\t! mul reduction2I" %}
+  ins_encode %{
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x1);
+    __ vpmulld($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsmul4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE > 3 && UseAVX == 0);
+  match(Set dst (MulReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "pshufd  $tmp2,$src2,0xE\n\t"
+            "pmulld  $tmp2,$src2\n\t"
+            "pshufd  $tmp,$tmp2,0x1\n\t"
+            "pmulld  $tmp2,$tmp\n\t"
+            "movd    $tmp,$src1\n\t"
+            "pmulld  $tmp2,$tmp\n\t"
+            "movd    $dst,$tmp2\t! mul reduction4I" %}
+  ins_encode %{
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ pmulld($tmp2$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp2$$XMMRegister, 0x1);
+    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ movdl($tmp$$XMMRegister, $src1$$Register);
+    __ pmulld($tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul4I_reduction_reg(rRegI dst, rRegI src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "pshufd  $tmp2,$src2,0xE\n\t"
+            "vpmulld $tmp,$src2,$tmp2\n\t"
+            "pshufd  $tmp2,$tmp,0x1\n\t"
+            "vpmulld $tmp,$tmp,$tmp2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpmulld $tmp2,$tmp,$tmp2\n\t"
+            "movd    $dst,$tmp2\t! mul reduction4I" %}
+  ins_encode %{
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ vpmulld($tmp$$XMMRegister, $src2$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
+    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul8I_reduction_reg(rRegI dst, rRegI src1, vecY src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVI src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vextractf128  $tmp,$src2\n\t"
+            "vpmulld $tmp,$tmp,$src2\n\t"
+            "pshufd  $tmp2,$tmp,0xE\n\t"
+            "vpmulld $tmp,$tmp,$tmp2\n\t"
+            "pshufd  $tmp2,$tmp,0x1\n\t"
+            "vpmulld $tmp,$tmp,$tmp2\n\t"
+            "movd    $tmp2,$src1\n\t"
+            "vpmulld $tmp2,$tmp,$tmp2\n\t"
+            "movd    $dst,$tmp2\t! mul reduction8I" %}
+  ins_encode %{
+    __ vextractf128h($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $src2$$XMMRegister, false);
+    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0xE);
+    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ pshufd($tmp2$$XMMRegister, $tmp$$XMMRegister, 0x1);
+    __ vpmulld($tmp$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($tmp2$$XMMRegister, $src1$$Register);
+    __ vpmulld($tmp2$$XMMRegister, $tmp$$XMMRegister, $tmp2$$XMMRegister, false);
+    __ movdl($dst$$Register, $tmp2$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsmul2F_reduction_reg(regF dst, regF src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (MulReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "mulss   $tmp,$src2\n\t"
+            "pshufd  $tmp2,$src2,0x01\n\t"
+            "mulss   $tmp,$tmp2\n\t"
+            "movdqu  $dst,$tmp\t! add reduction2F" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ mulss($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ mulss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdqu($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul2F_reduction_reg(regF dst, regF src1, vecD src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vmulss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vmulss  $dst,$tmp2,$tmp\t! add reduction2F" %}
+  ins_encode %{
+    __ vmulss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vmulss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsmul4F_reduction_reg(regF dst, regF src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (MulReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "mulss   $tmp,$src2\n\t"
+            "pshufd  $tmp2,$src2,0x01\n\t"
+            "mulss   $tmp,$tmp2\n\t"
+            "pshufd  $tmp2,$src2,0x02\n\t"
+            "mulss   $tmp,$tmp2\n\t"
+            "pshufd  $tmp2,$src2,0x03\n\t"
+            "mulss   $tmp,$tmp2\n\t"
+            "movdqu  $dst,$tmp\t! add reduction4F" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ mulss($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ mulss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ mulss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ pshufd($tmp2$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ mulss($tmp$$XMMRegister, $tmp2$$XMMRegister);
+    __ movdqu($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul4F_reduction_reg(regF dst, regF src1, vecX src2, regF tmp, regF tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vmulss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x02\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x03\n\t"
+            "vmulss  $dst,$tmp2,$tmp\t! add reduction4F" %}
+  ins_encode %{
+    __ vmulss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ vmulss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul8F_reduction_reg(regF dst, regF src1, vecY src2, regF tmp, regF tmp2, regF tmp3) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVF src1 src2));
+  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
+  format %{ "vmulss  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0x01\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x02\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$src2,0x03\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "vextractf128  $tmp3,$src2\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp3\n\t"
+            "pshufd  $tmp,$tmp3,0x01\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$tmp3,0x02\n\t"
+            "vmulss  $tmp2,$tmp2,$tmp\n\t"
+            "pshufd  $tmp,$tmp3,0x03\n\t"
+            "vmulss  $dst,$tmp2,$tmp\t! mul reduction8F" %}
+  ins_encode %{
+    __ vmulss($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x01);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x02);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0x03);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ vextractf128h($tmp3$$XMMRegister, $src2$$XMMRegister);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x01);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x02);
+    __ vmulss($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0x03);
+    __ vmulss($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rsmul2D_reduction_reg(regD dst, regD src1, vecX src2, regD tmp) %{
+  predicate(UseSSE >= 1 && UseAVX == 0);
+  match(Set dst (MulReductionVD src1 src2));
+  effect(TEMP tmp, TEMP dst);
+  format %{ "movdqu  $tmp,$src1\n\t"
+            "mulsd   $tmp,$src2\n\t"
+            "pshufd  $dst,$src2,0xE\n\t"
+            "mulsd   $dst,$tmp\t! add reduction2D" %}
+  ins_encode %{
+    __ movdqu($tmp$$XMMRegister, $src1$$XMMRegister);
+    __ mulsd($tmp$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($dst$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ mulsd($dst$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul2D_reduction_reg(regD dst, regD src1, vecX src2, regD tmp, regD tmp2) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVD src1 src2));
+  effect(TEMP tmp, TEMP tmp2);
+  format %{ "vmulsd  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0xE\n\t"
+            "vmulsd  $dst,$tmp2,$tmp\t! mul reduction2D" %}
+  ins_encode %{
+    __ vmulsd($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ vmulsd($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct rvmul4D_reduction_reg(regD dst, regD src1, vecY src2, regD tmp, regD tmp2, regD tmp3) %{
+  predicate(UseAVX > 0);
+  match(Set dst (MulReductionVD src1 src2));
+  effect(TEMP tmp, TEMP tmp2, TEMP tmp3);
+  format %{ "vmulsd  $tmp2,$src1,$src2\n\t"
+            "pshufd  $tmp,$src2,0xE\n\t"
+            "vmulsd  $tmp2,$tmp2,$tmp\n\t"
+            "vextractf128  $tmp3,$src2\n\t"
+            "vmulsd  $tmp2,$tmp2,$tmp3\n\t"
+            "pshufd  $tmp,$tmp3,0xE\n\t"
+            "vmulsd  $dst,$tmp2,$tmp\t! mul reduction4D" %}
+  ins_encode %{
+    __ vmulsd($tmp2$$XMMRegister, $src1$$XMMRegister, $src2$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $src2$$XMMRegister, 0xE);
+    __ vmulsd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+    __ vextractf128h($tmp3$$XMMRegister, $src2$$XMMRegister);
+    __ vmulsd($tmp2$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister);
+    __ pshufd($tmp$$XMMRegister, $tmp3$$XMMRegister, 0xE);
+    __ vmulsd($dst$$XMMRegister, $tmp2$$XMMRegister, $tmp$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
 // ====================VECTOR ARITHMETIC=======================================
 
 // --------------------------------- ADD --------------------------------------
--- a/hotspot/src/cpu/x86/vm/x86_64.ad	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/x86/vm/x86_64.ad	Fri Apr 17 23:55:04 2015 +0000
@@ -3604,6 +3604,23 @@
   %}
 %}
 
+// Indirect Memory Plus Positive Index Register Plus Offset Operand
+operand indPosIndexOffset(any_RegP reg, immL32 off, rRegI idx)
+%{
+  constraint(ALLOC_IN_RC(ptr_reg));
+  predicate(n->in(2)->in(3)->as_Type()->type()->is_long()->_lo >= 0);
+  match(AddP (AddP reg (ConvI2L idx)) off);
+
+  op_cost(10);
+  format %{"[$reg + $off + $idx]" %}
+  interface(MEMORY_INTER) %{
+    base($reg);
+    index($idx);
+    scale(0x0);
+    disp($off);
+  %}
+%}
+
 // Indirect Memory Times Scale Plus Positive Index Register Plus Offset Operand
 operand indPosIndexScaleOffset(any_RegP reg, immL32 off, rRegI idx, immI2 scale)
 %{
@@ -3755,6 +3772,23 @@
   %}
 %}
 
+// Indirect Memory Times Plus Positive Index Register Plus Offset Operand
+operand indPosIndexOffsetNarrow(rRegN reg, immL32 off, rRegI idx)
+%{
+  constraint(ALLOC_IN_RC(ptr_reg));
+  predicate(Universe::narrow_oop_shift() == 0 && n->in(2)->in(3)->as_Type()->type()->is_long()->_lo >= 0);
+  match(AddP (AddP (DecodeN reg) (ConvI2L idx)) off);
+
+  op_cost(10);
+  format %{"[$reg + $off + $idx]" %}
+  interface(MEMORY_INTER) %{
+    base($reg);
+    index($idx);
+    scale(0x0);
+    disp($off);
+  %}
+%}
+
 // Indirect Memory Times Scale Plus Positive Index Register Plus Offset Operand
 operand indPosIndexScaleOffsetNarrow(rRegN reg, immL32 off, rRegI idx, immI2 scale)
 %{
@@ -3946,11 +3980,11 @@
 // case of this is memory operands.
 
 opclass memory(indirect, indOffset8, indOffset32, indIndexOffset, indIndex,
-               indIndexScale, indIndexScaleOffset, indPosIndexScaleOffset,
+               indIndexScale, indIndexScaleOffset, indPosIndexOffset, indPosIndexScaleOffset,
                indCompressedOopOffset,
                indirectNarrow, indOffset8Narrow, indOffset32Narrow,
                indIndexOffsetNarrow, indIndexNarrow, indIndexScaleNarrow,
-               indIndexScaleOffsetNarrow, indPosIndexScaleOffsetNarrow);
+               indIndexScaleOffsetNarrow, indPosIndexOffsetNarrow, indPosIndexScaleOffsetNarrow);
 
 //----------PIPELINE-----------------------------------------------------------
 // Rules which define the behavior of the target architectures pipeline.
@@ -4984,6 +5018,17 @@
   ins_pipe(ialu_reg_reg_fat);
 %}
 
+instruct leaPPosIdxOff(rRegP dst, indPosIndexOffset mem)
+%{
+  match(Set dst mem);
+
+  ins_cost(110);
+  format %{ "leaq    $dst, $mem\t# ptr posidxoff" %}
+  opcode(0x8D);
+  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
+  ins_pipe(ialu_reg_reg_fat);
+%}
+
 instruct leaPPosIdxScaleOff(rRegP dst, indPosIndexScaleOffset mem)
 %{
   match(Set dst mem);
@@ -5068,6 +5113,18 @@
   ins_pipe(ialu_reg_reg_fat);
 %}
 
+instruct leaPPosIdxOffNarrow(rRegP dst, indPosIndexOffsetNarrow mem)
+%{
+  predicate(Universe::narrow_oop_shift() == 0);
+  match(Set dst mem);
+
+  ins_cost(110);
+  format %{ "leaq    $dst, $mem\t# ptr posidxoffnarrow" %}
+  opcode(0x8D);
+  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
+  ins_pipe(ialu_reg_reg_fat);
+%}
+
 instruct leaPPosIdxScaleOffNarrow(rRegP dst, indPosIndexScaleOffsetNarrow mem)
 %{
   predicate(Universe::narrow_oop_shift() == 0);
--- a/hotspot/src/cpu/zero/vm/vm_version_zero.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/cpu/zero/vm/vm_version_zero.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
  * Copyright 2009 Red Hat, Inc.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
@@ -30,4 +30,11 @@
 #include "runtime/stubCodeGenerator.hpp"
 #include "vm_version_zero.hpp"
 
-// This file is intentionally empty
+
+void VM_Version::initialize() {
+  // This machine does not allow unaligned memory accesses
+  if (! FLAG_IS_DEFAULT(UseUnalignedAccesses)) {
+    warning("Unaligned memory access is not available on this CPU");
+    FLAG_SET_DEFAULT(UseUnalignedAccesses, false);
+  }
+}
--- a/hotspot/src/os/aix/vm/attachListener_aix.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/attachListener_aix.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 2005, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -144,6 +144,10 @@
   }
   char* next() {
     if (*_pos == '\0') {
+      if (_pos < _end) {
+        _pos += 1;
+      }
+
       return NULL;
     }
     char* res = _pos;
@@ -214,6 +218,7 @@
 
   // bind socket
   struct sockaddr_un addr;
+  memset((void *)&addr, 0, sizeof(addr));
   addr.sun_family = AF_UNIX;
   strcpy(addr.sun_path, initial_path);
   ::unlink(initial_path);
--- a/hotspot/src/os/aix/vm/globals_aix.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/globals_aix.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -31,6 +31,10 @@
 //
 #define RUNTIME_OS_FLAGS(develop, develop_pd, product, product_pd, diagnostic, notproduct) \
                                                                                     \
+  /* Use 64K pages for virtual memory (shmat). */                                   \
+  product(bool, Use64KPages, true,                                                  \
+          "Use 64K pages if available.")                                            \
+                                                                                    \
   /* If UseLargePages == true allow or deny usage of 16M pages. 16M pages are  */   \
   /* a scarce resource and there may be situations where we do not want the VM */   \
   /* to run with 16M pages. (Will fall back to 64K pages).                     */   \
@@ -55,7 +59,7 @@
 // Defines Aix-specific default values. The flags are available on all
 // platforms, but they may have different default values on other platforms.
 //
-define_pd_global(bool, UseLargePages, true);
+define_pd_global(bool, UseLargePages, false);
 define_pd_global(bool, UseLargePagesIndividualAllocation, false);
 define_pd_global(bool, UseOSErrorReporting, false);
 define_pd_global(bool, UseThreadPriorities, true) ;
--- a/hotspot/src/os/aix/vm/interfaceSupport_aix.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/interfaceSupport_aix.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -23,8 +23,8 @@
  *
  */
 
-#ifndef OS_LINUX_VM_INTERFACESUPPORT_LINUX_HPP
-#define OS_LINUX_VM_INTERFACESUPPORT_LINUX_HPP
+#ifndef OS_AIX_VM_INTERFACESUPPORT_AIX_HPP
+#define OS_AIX_VM_INTERFACESUPPORT_AIX_HPP
 
 // Contains inlined functions for class InterfaceSupport
 
@@ -32,4 +32,4 @@
   os::write_memory_serialize_page(thread);
 }
 
-#endif // OS_LINUX_VM_INTERFACESUPPORT_LINUX_HPP
+#endif // OS_AIX_VM_INTERFACESUPPORT_AIX_HPP
--- a/hotspot/src/os/aix/vm/osThread_aix.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/osThread_aix.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -47,7 +47,7 @@
 
   _startThread_lock = new Monitor(Mutex::event, "startThread_lock", true,
                                   Monitor::_safepoint_check_never);
-  assert(_startThread_lock !=NULL, "check");
+  assert(_startThread_lock != NULL, "check");
 }
 
 void OSThread::pd_destroy() {
--- a/hotspot/src/os/aix/vm/os_aix.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/os_aix.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
  * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2014 SAP AG. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -113,6 +113,10 @@
 #define RUSAGE_THREAD   (1)               /* only the calling thread */
 #endif
 
+// PPC port
+static const uintx Use64KPagesThreshold       = 1*M;
+static const uintx MaxExpectedDataSegmentSize = SIZE_4G*2;
+
 // Add missing declarations (should be in procinfo.h but isn't until AIX 6.1).
 #if !defined(_AIXVERSION_610)
 extern "C" {
@@ -168,8 +172,8 @@
     return -1; \
 }
 
-// query dimensions of the stack of the calling thread
-static void query_stack_dimensions(address* p_stack_base, size_t* p_stack_size);
+// Query dimensions of the stack of the calling thread.
+static bool query_stack_dimensions(address* p_stack_base, size_t* p_stack_size);
 
 // function to check a given stack pointer against given stack limits
 inline bool is_valid_stackpointer(stackptr_t sp, stackptr_t stack_base, size_t stack_size) {
@@ -220,9 +224,6 @@
 int       os::Aix::_on_pase = -1;
 int       os::Aix::_os_version = -1;
 int       os::Aix::_stack_page_size = -1;
-size_t    os::Aix::_shm_default_page_size = -1;
-int       os::Aix::_can_use_64K_pages = -1;
-int       os::Aix::_can_use_16M_pages = -1;
 int       os::Aix::_xpg_sus_mode = -1;
 int       os::Aix::_extshm = -1;
 int       os::Aix::_logical_cpus = -1;
@@ -239,6 +240,63 @@
 static int      SR_signum          = SIGUSR2; // Signal used to suspend/resume a thread (must be > SIGSEGV, see 4355769)
 static sigset_t SR_sigset;
 
+// This describes the state of multipage support of the underlying
+// OS. Note that this is of no interest to the outsize world and
+// therefore should not be defined in AIX class.
+//
+// AIX supports four different page sizes - 4K, 64K, 16MB, 16GB. The
+// latter two (16M "large" resp. 16G "huge" pages) require special
+// setup and are normally not available.
+//
+// AIX supports multiple page sizes per process, for:
+//  - Stack (of the primordial thread, so not relevant for us)
+//  - Data - data, bss, heap, for us also pthread stacks
+//  - Text - text code
+//  - shared memory
+//
+// Default page sizes can be set via linker options (-bdatapsize, -bstacksize, ...)
+// and via environment variable LDR_CNTRL (DATAPSIZE, STACKPSIZE, ...).
+//
+// For shared memory, page size can be set dynamically via
+// shmctl(). Different shared memory regions can have different page
+// sizes.
+//
+// More information can be found at AIBM info center:
+//   http://publib.boulder.ibm.com/infocenter/aix/v6r1/index.jsp?topic=/com.ibm.aix.prftungd/doc/prftungd/multiple_page_size_app_support.htm
+//
+static struct {
+  size_t pagesize;            // sysconf _SC_PAGESIZE (4K)
+  size_t datapsize;           // default data page size (LDR_CNTRL DATAPSIZE)
+  size_t shmpsize;            // default shared memory page size (LDR_CNTRL SHMPSIZE)
+  size_t pthr_stack_pagesize; // stack page size of pthread threads
+  size_t textpsize;           // default text page size (LDR_CNTRL STACKPSIZE)
+  bool can_use_64K_pages;     // True if we can alloc 64K pages dynamically with Sys V shm.
+  bool can_use_16M_pages;     // True if we can alloc 16M pages dynamically with Sys V shm.
+  int error;                  // Error describing if something went wrong at multipage init.
+} g_multipage_support = {
+  (size_t) -1,
+  (size_t) -1,
+  (size_t) -1,
+  (size_t) -1,
+  (size_t) -1,
+  false, false,
+  0
+};
+
+// We must not accidentally allocate memory close to the BRK - even if
+// that would work - because then we prevent the BRK segment from
+// growing which may result in a malloc OOM even though there is
+// enough memory. The problem only arises if we shmat() or mmap() at
+// a specific wish address, e.g. to place the heap in a
+// compressed-oops-friendly way.
+static bool is_close_to_brk(address a) {
+  address a1 = (address) sbrk(0);
+  if (a >= a1 && a < (a1 + MaxExpectedDataSegmentSize)) {
+    return true;
+  }
+  return false;
+}
+
 julong os::available_memory() {
   return Aix::available_memory();
 }
@@ -277,7 +335,7 @@
   }
 
   // Maximum size 32bit disclaim() accepts. (Theoretically 4GB, but I just do not trust that.)
-  const unsigned int maxDisclaimSize = 0x80000000;
+  const unsigned int maxDisclaimSize = 0x40000000;
 
   const unsigned int numFullDisclaimsNeeded = (size / maxDisclaimSize);
   const unsigned int lastDisclaimSize = (size % maxDisclaimSize);
@@ -354,138 +412,131 @@
     case SIZE_64K: return "64K";
     case SIZE_16M: return "16M";
     case SIZE_16G: return "16G";
+    case -1:       return "not set";
     default:
       assert(false, "surprise");
       return "??";
   }
 }
 
-// Retrieve information about multipage size support. Will initialize
-// Aix::_page_size, Aix::_stack_page_size, Aix::_can_use_64K_pages,
-// Aix::_can_use_16M_pages.
+// Probe OS for multipage support.
+// Will fill the global g_multipage_support structure.
 // Must be called before calling os::large_page_init().
-void os::Aix::query_multipage_support() {
-
-  guarantee(_page_size == -1 &&
-            _stack_page_size == -1 &&
-            _can_use_64K_pages == -1 &&
-            _can_use_16M_pages == -1 &&
-            g_multipage_error == -1,
+static void query_multipage_support() {
+
+  guarantee(g_multipage_support.pagesize == -1,
             "do not call twice");
 
-  _page_size = ::sysconf(_SC_PAGESIZE);
+  g_multipage_support.pagesize = ::sysconf(_SC_PAGESIZE);
 
   // This really would surprise me.
-  assert(_page_size == SIZE_4K, "surprise!");
-
+  assert(g_multipage_support.pagesize == SIZE_4K, "surprise!");
 
   // Query default data page size (default page size for C-Heap, pthread stacks and .bss).
-  // Default data page size is influenced either by linker options (-bdatapsize)
+  // Default data page size is defined either by linker options (-bdatapsize)
   // or by environment variable LDR_CNTRL (suboption DATAPSIZE). If none is given,
   // default should be 4K.
-  size_t data_page_size = SIZE_4K;
   {
-    void* p = os::malloc(SIZE_16M, mtInternal);
-    guarantee(p != NULL, "malloc failed");
-    data_page_size = os::Aix::query_pagesize(p);
-    os::free(p);
-  }
-
-  // query default shm page size (LDR_CNTRL SHMPSIZE)
+    void* p = ::malloc(SIZE_16M);
+    g_multipage_support.datapsize = os::Aix::query_pagesize(p);
+    ::free(p);
+  }
+
+  // Query default shm page size (LDR_CNTRL SHMPSIZE).
   {
     const int shmid = ::shmget(IPC_PRIVATE, 1, IPC_CREAT | S_IRUSR | S_IWUSR);
     guarantee(shmid != -1, "shmget failed");
     void* p = ::shmat(shmid, NULL, 0);
     ::shmctl(shmid, IPC_RMID, NULL);
     guarantee(p != (void*) -1, "shmat failed");
-    _shm_default_page_size = os::Aix::query_pagesize(p);
+    g_multipage_support.shmpsize = os::Aix::query_pagesize(p);
     ::shmdt(p);
   }
 
-  // before querying the stack page size, make sure we are not running as primordial
+  // Before querying the stack page size, make sure we are not running as primordial
   // thread (because primordial thread's stack may have different page size than
   // pthread thread stacks). Running a VM on the primordial thread won't work for a
-  // number of reasons so we may just as well guarantee it here
-  guarantee(!os::Aix::is_primordial_thread(), "Must not be called for primordial thread");
-
-  // query stack page size
+  // number of reasons so we may just as well guarantee it here.
+  guarantee0(!os::Aix::is_primordial_thread());
+
+  // Query pthread stack page size.
   {
     int dummy = 0;
-    _stack_page_size = os::Aix::query_pagesize(&dummy);
-    // everything else would surprise me and should be looked into
-    guarantee(_stack_page_size == SIZE_4K || _stack_page_size == SIZE_64K, "Wrong page size");
-    // also, just for completeness: pthread stacks are allocated from C heap, so
-    // stack page size should be the same as data page size
-    guarantee(_stack_page_size == data_page_size, "stack page size should be the same as data page size");
-  }
-
-  // EXTSHM is bad: among other things, it prevents setting pagesize dynamically
-  // for system V shm.
-  if (Aix::extshm()) {
-    if (Verbose) {
-      fprintf(stderr, "EXTSHM is active - will disable large page support.\n"
-                      "Please make sure EXTSHM is OFF for large page support.\n");
-    }
-    g_multipage_error = ERROR_MP_EXTSHM_ACTIVE;
-    _can_use_64K_pages = _can_use_16M_pages = 0;
+    g_multipage_support.pthr_stack_pagesize = os::Aix::query_pagesize(&dummy);
+  }
+
+  // Query default text page size (LDR_CNTRL TEXTPSIZE).
+  /* PPC port: so far unused.
+  {
+    address any_function =
+      (address) resolve_function_descriptor_to_code_pointer((address)describe_pagesize);
+    g_multipage_support.textpsize = os::Aix::query_pagesize(any_function);
+  }
+  */
+
+  // Now probe for support of 64K pages and 16M pages.
+
+  // Before OS/400 V6R1, there is no support for pages other than 4K.
+  if (os::Aix::on_pase_V5R4_or_older()) {
+    Unimplemented();
     goto query_multipage_support_end;
   }
 
-  // now check which page sizes the OS claims it supports, and of those, which actually can be used.
+  // Now check which page sizes the OS claims it supports, and of those, which actually can be used.
   {
     const int MAX_PAGE_SIZES = 4;
     psize_t sizes[MAX_PAGE_SIZES];
     const int num_psizes = ::vmgetinfo(sizes, VMINFO_GETPSIZES, MAX_PAGE_SIZES);
     if (num_psizes == -1) {
-      if (Verbose) {
-        fprintf(stderr, "vmgetinfo(VMINFO_GETPSIZES) failed (errno: %d)\n", errno);
-        fprintf(stderr, "disabling multipage support.\n");
-      }
-      g_multipage_error = ERROR_MP_VMGETINFO_FAILED;
-      _can_use_64K_pages = _can_use_16M_pages = 0;
+      trc("vmgetinfo(VMINFO_GETPSIZES) failed (errno: %d)\n", errno);
+      trc("disabling multipage support.\n");
+      g_multipage_support.error = ERROR_MP_VMGETINFO_FAILED;
       goto query_multipage_support_end;
     }
     guarantee(num_psizes > 0, "vmgetinfo(.., VMINFO_GETPSIZES, ...) failed.");
     assert(num_psizes <= MAX_PAGE_SIZES, "Surprise! more than 4 page sizes?");
-    if (Verbose) {
-      fprintf(stderr, "vmgetinfo(.., VMINFO_GETPSIZES, ...) returns %d supported page sizes: ", num_psizes);
-      for (int i = 0; i < num_psizes; i ++) {
-        fprintf(stderr, " %s ", describe_pagesize(sizes[i]));
-      }
-      fprintf(stderr, " .\n");
+    trcVerbose("vmgetinfo(.., VMINFO_GETPSIZES, ...) returns %d supported page sizes: ", num_psizes);
+    for (int i = 0; i < num_psizes; i ++) {
+      trcVerbose(" %s ", describe_pagesize(sizes[i]));
     }
 
     // Can we use 64K, 16M pages?
-    _can_use_64K_pages = 0;
-    _can_use_16M_pages = 0;
     for (int i = 0; i < num_psizes; i ++) {
-      if (sizes[i] == SIZE_64K) {
-        _can_use_64K_pages = 1;
-      } else if (sizes[i] == SIZE_16M) {
-        _can_use_16M_pages = 1;
+      const size_t pagesize = sizes[i];
+      if (pagesize != SIZE_64K && pagesize != SIZE_16M) {
+        continue;
       }
-    }
-
-    if (!_can_use_64K_pages) {
-      g_multipage_error = ERROR_MP_VMGETINFO_CLAIMS_NO_SUPPORT_FOR_64K;
-    }
-
-    // Double-check for 16M pages: Even if AIX claims to be able to use 16M pages,
-    // there must be an actual 16M page pool, and we must run with enough rights.
-    if (_can_use_16M_pages) {
-      const int shmid = ::shmget(IPC_PRIVATE, SIZE_16M, IPC_CREAT | S_IRUSR | S_IWUSR);
-      guarantee(shmid != -1, "shmget failed");
+      bool can_use = false;
+      trcVerbose("Probing support for %s pages...", describe_pagesize(pagesize));
+      const int shmid = ::shmget(IPC_PRIVATE, pagesize,
+        IPC_CREAT | S_IRUSR | S_IWUSR);
+      guarantee0(shmid != -1); // Should always work.
+      // Try to set pagesize.
       struct shmid_ds shm_buf = { 0 };
-      shm_buf.shm_pagesize = SIZE_16M;
-      const bool can_set_pagesize = ::shmctl(shmid, SHM_PAGESIZE, &shm_buf) == 0 ? true : false;
-      const int en = errno;
-      ::shmctl(shmid, IPC_RMID, NULL);
-      if (!can_set_pagesize) {
-        if (Verbose) {
-          fprintf(stderr, "Failed to allocate even one misely 16M page. shmctl failed with %d (%s).\n"
-                          "Will deactivate 16M support.\n", en, strerror(en));
+      shm_buf.shm_pagesize = pagesize;
+      if (::shmctl(shmid, SHM_PAGESIZE, &shm_buf) != 0) {
+        const int en = errno;
+        ::shmctl(shmid, IPC_RMID, NULL); // As early as possible!
+        // PPC port trcVerbose("shmctl(SHM_PAGESIZE) failed with %s",
+        // PPC port  MiscUtils::describe_errno(en));
+      } else {
+        // Attach and double check pageisze.
+        void* p = ::shmat(shmid, NULL, 0);
+        ::shmctl(shmid, IPC_RMID, NULL); // As early as possible!
+        guarantee0(p != (void*) -1); // Should always work.
+        const size_t real_pagesize = os::Aix::query_pagesize(p);
+        if (real_pagesize != pagesize) {
+          trcVerbose("real page size (0x%llX) differs.", real_pagesize);
+        } else {
+          can_use = true;
         }
-        _can_use_16M_pages = 0;
+        ::shmdt(p);
+      }
+      trcVerbose("Can use: %s", (can_use ? "yes" : "no"));
+      if (pagesize == SIZE_64K) {
+        g_multipage_support.can_use_64K_pages = can_use;
+      } else if (pagesize == SIZE_16M) {
+        g_multipage_support.can_use_16M_pages = can_use;
       }
     }
 
@@ -493,23 +544,29 @@
 
 query_multipage_support_end:
 
-  guarantee(_page_size != -1 &&
-            _stack_page_size != -1 &&
-            _can_use_64K_pages != -1 &&
-            _can_use_16M_pages != -1, "Page sizes not properly initialized");
-
-  if (_can_use_64K_pages) {
-    g_multipage_error = 0;
-  }
-
-  if (Verbose) {
-    fprintf(stderr, "Data page size (C-Heap, bss, etc): %s\n", describe_pagesize(data_page_size));
-    fprintf(stderr, "Thread stack page size (pthread): %s\n", describe_pagesize(_stack_page_size));
-    fprintf(stderr, "Default shared memory page size: %s\n", describe_pagesize(_shm_default_page_size));
-    fprintf(stderr, "Can use 64K pages dynamically with shared meory: %s\n", (_can_use_64K_pages ? "yes" :"no"));
-    fprintf(stderr, "Can use 16M pages dynamically with shared memory: %s\n", (_can_use_16M_pages ? "yes" :"no"));
-    fprintf(stderr, "Multipage error details: %d\n", g_multipage_error);
-  }
+  trcVerbose("base page size (sysconf _SC_PAGESIZE): %s\n",
+      describe_pagesize(g_multipage_support.pagesize));
+  trcVerbose("Data page size (C-Heap, bss, etc): %s\n",
+      describe_pagesize(g_multipage_support.datapsize));
+  trcVerbose("Text page size: %s\n",
+      describe_pagesize(g_multipage_support.textpsize));
+  trcVerbose("Thread stack page size (pthread): %s\n",
+      describe_pagesize(g_multipage_support.pthr_stack_pagesize));
+  trcVerbose("Default shared memory page size: %s\n",
+      describe_pagesize(g_multipage_support.shmpsize));
+  trcVerbose("Can use 64K pages dynamically with shared meory: %s\n",
+      (g_multipage_support.can_use_64K_pages ? "yes" :"no"));
+  trcVerbose("Can use 16M pages dynamically with shared memory: %s\n",
+      (g_multipage_support.can_use_16M_pages ? "yes" :"no"));
+  trcVerbose("Multipage error details: %d\n",
+      g_multipage_support.error);
+
+  // sanity checks
+  assert0(g_multipage_support.pagesize == SIZE_4K);
+  assert0(g_multipage_support.datapsize == SIZE_4K || g_multipage_support.datapsize == SIZE_64K);
+  // PPC port: so far unused.assert0(g_multipage_support.textpsize == SIZE_4K || g_multipage_support.textpsize == SIZE_64K);
+  assert0(g_multipage_support.pthr_stack_pagesize == g_multipage_support.datapsize);
+  assert0(g_multipage_support.shmpsize == SIZE_4K || g_multipage_support.shmpsize == SIZE_64K);
 
 } // end os::Aix::query_multipage_support()
 
@@ -1557,9 +1614,12 @@
 
   st->print_cr("  default page size: %s", describe_pagesize(os::vm_page_size()));
   st->print_cr("  default stack page size: %s", describe_pagesize(os::vm_page_size()));
-  st->print_cr("  default shm page size: %s", describe_pagesize(os::Aix::shm_default_page_size()));
-  st->print_cr("  can use 64K pages dynamically: %s", (os::Aix::can_use_64K_pages() ? "yes" :"no"));
-  st->print_cr("  can use 16M pages dynamically: %s", (os::Aix::can_use_16M_pages() ? "yes" :"no"));
+  st->print_cr("  Default shared memory page size:        %s",
+    describe_pagesize(g_multipage_support.shmpsize));
+  st->print_cr("  Can use 64K pages dynamically with shared meory:  %s",
+    (g_multipage_support.can_use_64K_pages ? "yes" :"no"));
+  st->print_cr("  Can use 16M pages dynamically with shared memory: %s",
+    (g_multipage_support.can_use_16M_pages ? "yes" :"no"));
   if (g_multipage_error != 0) {
     st->print_cr("  multipage error: %d", g_multipage_error);
   }
@@ -1570,6 +1630,9 @@
 
   const char* const extshm = ::getenv("EXTSHM");
   st->print_cr("  EXTSHM=%s.", extshm ? extshm : "<unset>");
+  if ( (strcmp(extshm, "on") == 0) || (strcmp(extshm, "ON") == 0) ) {
+    st->print_cr("  *** Unsupported! Please remove EXTSHM from your environment! ***");
+  }
 
   // Call os::Aix::get_meminfo() to retrieve memory statistics.
   os::Aix::meminfo_t mi;
@@ -1812,315 +1875,386 @@
 ////////////////////////////////////////////////////////////////////////////////
 // Virtual Memory
 
-// AddrRange describes an immutable address range
-//
-// This is a helper class for the 'shared memory bookkeeping' below.
-class AddrRange {
-  friend class ShmBkBlock;
-
-  char* _start;
-  size_t _size;
-
-public:
-
-  AddrRange(char* start, size_t size)
-    : _start(start), _size(size)
-  {}
-
-  AddrRange(const AddrRange& r)
-    : _start(r.start()), _size(r.size())
-  {}
-
-  char* start() const { return _start; }
-  size_t size() const { return _size; }
-  char* end() const { return _start + _size; }
-  bool is_empty() const { return _size == 0 ? true : false; }
-
-  static AddrRange empty_range() { return AddrRange(NULL, 0); }
-
-  bool contains(const char* p) const {
-    return start() <= p && end() > p;
-  }
-
-  bool contains(const AddrRange& range) const {
-    return start() <= range.start() && end() >= range.end();
-  }
-
-  bool intersects(const AddrRange& range) const {
-    return (range.start() <= start() && range.end() > start()) ||
-           (range.start() < end() && range.end() >= end()) ||
-           contains(range);
-  }
-
-  bool is_same_range(const AddrRange& range) const {
-    return start() == range.start() && size() == range.size();
-  }
-
-  // return the closest inside range consisting of whole pages
-  AddrRange find_closest_aligned_range(size_t pagesize) const {
-    if (pagesize == 0 || is_empty()) {
-      return empty_range();
+// We need to keep small simple bookkeeping for os::reserve_memory and friends.
+
+#define VMEM_MAPPED  1
+#define VMEM_SHMATED 2
+
+struct vmembk_t {
+  int type;         // 1 - mmap, 2 - shmat
+  char* addr;
+  size_t size;      // Real size, may be larger than usersize.
+  size_t pagesize;  // page size of area
+  vmembk_t* next;
+
+  bool contains_addr(char* p) const {
+    return p >= addr && p < (addr + size);
+  }
+
+  bool contains_range(char* p, size_t s) const {
+    return contains_addr(p) && contains_addr(p + s - 1);
+  }
+
+  void print_on(outputStream* os) const {
+    os->print("[" PTR_FORMAT " - " PTR_FORMAT "] (" UINTX_FORMAT
+      " bytes, %d %s pages), %s",
+      addr, addr + size - 1, size, size / pagesize, describe_pagesize(pagesize),
+      (type == VMEM_SHMATED ? "shmat" : "mmap")
+    );
+  }
+
+  // Check that range is a sub range of memory block (or equal to memory block);
+  // also check that range is fully page aligned to the page size if the block.
+  void assert_is_valid_subrange(char* p, size_t s) const {
+    if (!contains_range(p, s)) {
+      fprintf(stderr, "[" PTR_FORMAT " - " PTR_FORMAT "] is not a sub "
+              "range of [" PTR_FORMAT " - " PTR_FORMAT "].\n",
+              p, p + s - 1, addr, addr + size - 1);
+      guarantee0(false);
     }
-    char* const from = (char*)align_size_up((intptr_t)_start, pagesize);
-    char* const to = (char*)align_size_down((intptr_t)end(), pagesize);
-    if (from > to) {
-      return empty_range();
+    if (!is_aligned_to(p, pagesize) || !is_aligned_to(p + s, pagesize)) {
+      fprintf(stderr, "range [" PTR_FORMAT " - " PTR_FORMAT "] is not"
+              " aligned to pagesize (%s)\n", p, p + s);
+      guarantee0(false);
     }
-    return AddrRange(from, to - from);
   }
 };
 
-////////////////////////////////////////////////////////////////////////////
-// shared memory bookkeeping
-//
-// the os::reserve_memory() API and friends hand out different kind of memory, depending
-// on need and circumstances. Memory may be allocated with mmap() or with shmget/shmat.
-//
-// But these memory types have to be treated differently. For example, to uncommit
-// mmap-based memory, msync(MS_INVALIDATE) is needed, to uncommit shmat-based memory,
-// disclaim64() is needed.
-//
-// Therefore we need to keep track of the allocated memory segments and their
-// properties.
-
-// ShmBkBlock: base class for all blocks in the shared memory bookkeeping
-class ShmBkBlock : public CHeapObj<mtInternal> {
-
-  ShmBkBlock* _next;
-
-protected:
-
-  AddrRange _range;
-  const size_t _pagesize;
-  const bool _pinned;
-
-public:
-
-  ShmBkBlock(AddrRange range, size_t pagesize, bool pinned)
-    : _range(range), _pagesize(pagesize), _pinned(pinned) , _next(NULL) {
-
-    assert(_pagesize == SIZE_4K || _pagesize == SIZE_64K || _pagesize == SIZE_16M, "invalid page size");
-    assert(!_range.is_empty(), "invalid range");
-  }
-
-  virtual void print(outputStream* st) const {
-    st->print("0x%p ... 0x%p (%llu) - %d %s pages - %s",
-              _range.start(), _range.end(), _range.size(),
-              _range.size() / _pagesize, describe_pagesize(_pagesize),
-              _pinned ? "pinned" : "");
-  }
-
-  enum Type { MMAP, SHMAT };
-  virtual Type getType() = 0;
-
-  char* base() const { return _range.start(); }
-  size_t size() const { return _range.size(); }
-
-  void setAddrRange(AddrRange range) {
-    _range = range;
-  }
-
-  bool containsAddress(const char* p) const {
-    return _range.contains(p);
-  }
-
-  bool containsRange(const char* p, size_t size) const {
-    return _range.contains(AddrRange((char*)p, size));
-  }
-
-  bool isSameRange(const char* p, size_t size) const {
-    return _range.is_same_range(AddrRange((char*)p, size));
-  }
-
-  virtual bool disclaim(char* p, size_t size) = 0;
-  virtual bool release() = 0;
-
-  // blocks live in a list.
-  ShmBkBlock* next() const { return _next; }
-  void set_next(ShmBkBlock* blk) { _next = blk; }
-
-}; // end: ShmBkBlock
-
-
-// ShmBkMappedBlock: describes an block allocated with mmap()
-class ShmBkMappedBlock : public ShmBkBlock {
-public:
-
-  ShmBkMappedBlock(AddrRange range)
-    : ShmBkBlock(range, SIZE_4K, false) {} // mmap: always 4K, never pinned
-
-  void print(outputStream* st) const {
-    ShmBkBlock::print(st);
-    st->print_cr(" - mmap'ed");
-  }
-
-  Type getType() {
-    return MMAP;
-  }
-
-  bool disclaim(char* p, size_t size) {
-
-    AddrRange r(p, size);
-
-    guarantee(_range.contains(r), "invalid disclaim");
-
-    // only disclaim whole ranges.
-    const AddrRange r2 = r.find_closest_aligned_range(_pagesize);
-    if (r2.is_empty()) {
-      return true;
+static struct {
+  vmembk_t* first;
+  MiscUtils::CritSect cs;
+} vmem;
+
+static void vmembk_add(char* addr, size_t size, size_t pagesize, int type) {
+  vmembk_t* p = (vmembk_t*) ::malloc(sizeof(vmembk_t));
+  assert0(p);
+  if (p) {
+    MiscUtils::AutoCritSect lck(&vmem.cs);
+    p->addr = addr; p->size = size;
+    p->pagesize = pagesize;
+    p->type = type;
+    p->next = vmem.first;
+    vmem.first = p;
+  }
+}
+
+static vmembk_t* vmembk_find(char* addr) {
+  MiscUtils::AutoCritSect lck(&vmem.cs);
+  for (vmembk_t* p = vmem.first; p; p = p->next) {
+    if (p->addr <= addr && (p->addr + p->size) > addr) {
+      return p;
     }
-
-    const int rc = ::msync(r2.start(), r2.size(), MS_INVALIDATE);
-
-    if (rc != 0) {
-      warning("msync(0x%p, %llu, MS_INVALIDATE) failed (%d)\n", r2.start(), r2.size(), errno);
-    }
-
-    return rc == 0 ? true : false;
-  }
-
-  bool release() {
-    // mmap'ed blocks are released using munmap
-    if (::munmap(_range.start(), _range.size()) != 0) {
-      warning("munmap(0x%p, %llu) failed (%d)\n", _range.start(), _range.size(), errno);
-      return false;
-    }
-    return true;
-  }
-}; // end: ShmBkMappedBlock
-
-// ShmBkShmatedBlock: describes an block allocated with shmget/shmat()
-class ShmBkShmatedBlock : public ShmBkBlock {
-public:
-
-  ShmBkShmatedBlock(AddrRange range, size_t pagesize, bool pinned)
-    : ShmBkBlock(range, pagesize, pinned) {}
-
-  void print(outputStream* st) const {
-    ShmBkBlock::print(st);
-    st->print_cr(" - shmat'ed");
-  }
-
-  Type getType() {
-    return SHMAT;
-  }
-
-  bool disclaim(char* p, size_t size) {
-
-    AddrRange r(p, size);
-
-    if (_pinned) {
-      return true;
-    }
-
-    // shmat'ed blocks are disclaimed using disclaim64
-    guarantee(_range.contains(r), "invalid disclaim");
-
-    // only disclaim whole ranges.
-    const AddrRange r2 = r.find_closest_aligned_range(_pagesize);
-    if (r2.is_empty()) {
-      return true;
-    }
-
-    const bool rc = my_disclaim64(r2.start(), r2.size());
-
-    if (Verbose && !rc) {
-      warning("failed to disclaim shm %p-%p\n", r2.start(), r2.end());
-    }
-
-    return rc;
-  }
-
-  bool release() {
-    bool rc = false;
-    if (::shmdt(_range.start()) != 0) {
-      warning("shmdt(0x%p) failed (%d)\n", _range.start(), errno);
-    } else {
-      rc = true;
-    }
-    return rc;
-  }
-
-}; // end: ShmBkShmatedBlock
-
-static ShmBkBlock* g_shmbk_list = NULL;
-static volatile jint g_shmbk_table_lock = 0;
-
-// keep some usage statistics
-static struct {
-  int nodes;    // number of nodes in list
-  size_t bytes; // reserved - not committed - bytes.
-  int reserves; // how often reserve was called
-  int lookups;  // how often a lookup was made
-} g_shmbk_stats = { 0, 0, 0, 0 };
-
-// add information about a shared memory segment to the bookkeeping
-static void shmbk_register(ShmBkBlock* p_block) {
-  guarantee(p_block, "logic error");
-  p_block->set_next(g_shmbk_list);
-  g_shmbk_list = p_block;
-  g_shmbk_stats.reserves ++;
-  g_shmbk_stats.bytes += p_block->size();
-  g_shmbk_stats.nodes ++;
-}
-
-// remove information about a shared memory segment by its starting address
-static void shmbk_unregister(ShmBkBlock* p_block) {
-  ShmBkBlock* p = g_shmbk_list;
-  ShmBkBlock* prev = NULL;
-  while (p) {
-    if (p == p_block) {
-      if (prev) {
-        prev->set_next(p->next());
-      } else {
-        g_shmbk_list = p->next();
-      }
-      g_shmbk_stats.nodes --;
-      g_shmbk_stats.bytes -= p->size();
+  }
+  return NULL;
+}
+
+static void vmembk_remove(vmembk_t* p0) {
+  MiscUtils::AutoCritSect lck(&vmem.cs);
+  assert0(p0);
+  assert0(vmem.first); // List should not be empty.
+  for (vmembk_t** pp = &(vmem.first); *pp; pp = &((*pp)->next)) {
+    if (*pp == p0) {
+      *pp = p0->next;
+      ::free(p0);
       return;
     }
-    prev = p;
-    p = p->next();
-  }
-  assert(false, "should not happen");
-}
-
-// given a pointer, return shared memory bookkeeping record for the segment it points into
-// using the returned block info must happen under lock protection
-static ShmBkBlock* shmbk_find_by_containing_address(const char* addr) {
-  g_shmbk_stats.lookups ++;
-  ShmBkBlock* p = g_shmbk_list;
-  while (p) {
-    if (p->containsAddress(addr)) {
-      return p;
+  }
+  assert0(false); // Not found?
+}
+
+static void vmembk_print_on(outputStream* os) {
+  MiscUtils::AutoCritSect lck(&vmem.cs);
+  for (vmembk_t* vmi = vmem.first; vmi; vmi = vmi->next) {
+    vmi->print_on(os);
+    os->cr();
+  }
+}
+
+// Reserve and attach a section of System V memory.
+// If <requested_addr> is not NULL, function will attempt to attach the memory at the given
+// address. Failing that, it will attach the memory anywhere.
+// If <requested_addr> is NULL, function will attach the memory anywhere.
+//
+// <alignment_hint> is being ignored by this function. It is very probable however that the
+// alignment requirements are met anyway, because shmat() attaches at 256M boundaries.
+// Should this be not enogh, we can put more work into it.
+static char* reserve_shmated_memory (
+  size_t bytes,
+  char* requested_addr,
+  size_t alignment_hint) {
+
+  trcVerbose("reserve_shmated_memory " UINTX_FORMAT " bytes, wishaddress "
+    PTR_FORMAT ", alignment_hint " UINTX_FORMAT "...",
+    bytes, requested_addr, alignment_hint);
+
+  // Either give me wish address or wish alignment but not both.
+  assert0(!(requested_addr != NULL && alignment_hint != 0));
+
+  // We must prevent anyone from attaching too close to the
+  // BRK because that may cause malloc OOM.
+  if (requested_addr != NULL && is_close_to_brk((address)requested_addr)) {
+    trcVerbose("Wish address " PTR_FORMAT " is too close to the BRK segment. "
+      "Will attach anywhere.", requested_addr);
+    // Act like the OS refused to attach there.
+    requested_addr = NULL;
+  }
+
+  // For old AS/400's (V5R4 and older) we should not even be here - System V shared memory is not
+  // really supported (max size 4GB), so reserve_mmapped_memory should have been used instead.
+  if (os::Aix::on_pase_V5R4_or_older()) {
+    ShouldNotReachHere();
+  }
+
+  // Align size of shm up to 64K to avoid errors if we later try to change the page size.
+  const size_t size = align_size_up(bytes, SIZE_64K);
+
+  // Reserve the shared segment.
+  int shmid = shmget(IPC_PRIVATE, size, IPC_CREAT | S_IRUSR | S_IWUSR);
+  if (shmid == -1) {
+    trc("shmget(.., " UINTX_FORMAT ", ..) failed (errno: %d).", size, errno);
+    return NULL;
+  }
+
+  // Important note:
+  // It is very important that we, upon leaving this function, do not leave a shm segment alive.
+  // We must right after attaching it remove it from the system. System V shm segments are global and
+  // survive the process.
+  // So, from here on: Do not assert, do not return, until we have called shmctl(IPC_RMID) (A).
+
+  struct shmid_ds shmbuf;
+  memset(&shmbuf, 0, sizeof(shmbuf));
+  shmbuf.shm_pagesize = SIZE_64K;
+  if (shmctl(shmid, SHM_PAGESIZE, &shmbuf) != 0) {
+    trcVerbose("Failed to set page size (need " UINTX_FORMAT " 64K pages) - shmctl failed with %d.",
+               size / SIZE_64K, errno);
+    // I want to know if this ever happens.
+    assert(false, "failed to set page size for shmat");
+  }
+
+  // Now attach the shared segment.
+  // Note that I attach with SHM_RND - which means that the requested address is rounded down, if
+  // needed, to the next lowest segment boundary. Otherwise the attach would fail if the address
+  // were not a segment boundary.
+  char* const addr = (char*) shmat(shmid, requested_addr, SHM_RND);
+  const int errno_shmat = errno;
+
+  // (A) Right after shmat and before handing shmat errors delete the shm segment.
+  if (::shmctl(shmid, IPC_RMID, NULL) == -1) {
+    trc("shmctl(%u, IPC_RMID) failed (%d)\n", shmid, errno);
+    assert(false, "failed to remove shared memory segment!");
+  }
+
+  // Handle shmat error. If we failed to attach, just return.
+  if (addr == (char*)-1) {
+    trcVerbose("Failed to attach segment at " PTR_FORMAT " (%d).", requested_addr, errno_shmat);
+    return NULL;
+  }
+
+  // Just for info: query the real page size. In case setting the page size did not
+  // work (see above), the system may have given us something other then 4K (LDR_CNTRL).
+  const size_t real_pagesize = os::Aix::query_pagesize(addr);
+  if (real_pagesize != shmbuf.shm_pagesize) {
+    trcVerbose("pagesize is, surprisingly, %h.", real_pagesize);
+  }
+
+  if (addr) {
+    trcVerbose("shm-allocated " PTR_FORMAT " .. " PTR_FORMAT " (" UINTX_FORMAT " bytes, " UINTX_FORMAT " %s pages)",
+      addr, addr + size - 1, size, size/real_pagesize, describe_pagesize(real_pagesize));
+  } else {
+    if (requested_addr != NULL) {
+      trcVerbose("failed to shm-allocate " UINTX_FORMAT " bytes at with address " PTR_FORMAT ".", size, requested_addr);
+    } else {
+      trcVerbose("failed to shm-allocate " UINTX_FORMAT " bytes at any address.", size);
     }
-    p = p->next();
-  }
-  return NULL;
-}
-
-// dump all information about all memory segments allocated with os::reserve_memory()
-void shmbk_dump_info() {
-  tty->print_cr("-- shared mem bookkeeping (alive: %d segments, %llu bytes, "
-    "total reserves: %d total lookups: %d)",
-    g_shmbk_stats.nodes, g_shmbk_stats.bytes, g_shmbk_stats.reserves, g_shmbk_stats.lookups);
-  const ShmBkBlock* p = g_shmbk_list;
-  int i = 0;
-  while (p) {
-    p->print(tty);
-    p = p->next();
-    i ++;
-  }
-}
-
-#define LOCK_SHMBK     { ThreadCritical _LOCK_SHMBK;
-#define UNLOCK_SHMBK   }
+  }
+
+  // book-keeping
+  vmembk_add(addr, size, real_pagesize, VMEM_SHMATED);
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+
+  return addr;
+}
+
+static bool release_shmated_memory(char* addr, size_t size) {
+
+  trcVerbose("release_shmated_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
+    addr, addr + size - 1);
+
+  bool rc = false;
+
+  // TODO: is there a way to verify shm size without doing bookkeeping?
+  if (::shmdt(addr) != 0) {
+    trcVerbose("error (%d).", errno);
+  } else {
+    trcVerbose("ok.");
+    rc = true;
+  }
+  return rc;
+}
+
+static bool uncommit_shmated_memory(char* addr, size_t size) {
+  trcVerbose("uncommit_shmated_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
+    addr, addr + size - 1);
+
+  const bool rc = my_disclaim64(addr, size);
+
+  if (!rc) {
+    trcVerbose("my_disclaim64(" PTR_FORMAT ", " UINTX_FORMAT ") failed.\n", addr, size);
+    return false;
+  }
+  return true;
+}
+
+// Reserve memory via mmap.
+// If <requested_addr> is given, an attempt is made to attach at the given address.
+// Failing that, memory is allocated at any address.
+// If <alignment_hint> is given and <requested_addr> is NULL, an attempt is made to
+// allocate at an address aligned with the given alignment. Failing that, memory
+// is aligned anywhere.
+static char* reserve_mmaped_memory(size_t bytes, char* requested_addr, size_t alignment_hint) {
+  trcVerbose("reserve_mmaped_memory " UINTX_FORMAT " bytes, wishaddress " PTR_FORMAT ", "
+    "alignment_hint " UINTX_FORMAT "...",
+    bytes, requested_addr, alignment_hint);
+
+  // If a wish address is given, but not aligned to 4K page boundary, mmap will fail.
+  if (requested_addr && !is_aligned_to(requested_addr, os::vm_page_size()) != 0) {
+    trcVerbose("Wish address " PTR_FORMAT " not aligned to page boundary.", requested_addr);
+    return NULL;
+  }
+
+  // We must prevent anyone from attaching too close to the
+  // BRK because that may cause malloc OOM.
+  if (requested_addr != NULL && is_close_to_brk((address)requested_addr)) {
+    trcVerbose("Wish address " PTR_FORMAT " is too close to the BRK segment. "
+      "Will attach anywhere.", requested_addr);
+    // Act like the OS refused to attach there.
+    requested_addr = NULL;
+  }
+
+  // Specify one or the other but not both.
+  assert0(!(requested_addr != NULL && alignment_hint > 0));
+
+  // In 64K mode, we claim the global page size (os::vm_page_size())
+  // is 64K. This is one of the few points where that illusion may
+  // break, because mmap() will always return memory aligned to 4K. So
+  // we must ensure we only ever return memory aligned to 64k.
+  if (alignment_hint) {
+    alignment_hint = lcm(alignment_hint, os::vm_page_size());
+  } else {
+    alignment_hint = os::vm_page_size();
+  }
+
+  // Size shall always be a multiple of os::vm_page_size (esp. in 64K mode).
+  const size_t size = align_size_up(bytes, os::vm_page_size());
+
+  // alignment: Allocate memory large enough to include an aligned range of the right size and
+  // cut off the leading and trailing waste pages.
+  assert0(alignment_hint != 0 && is_aligned_to(alignment_hint, os::vm_page_size())); // see above
+  const size_t extra_size = size + alignment_hint;
+
+  // Note: MAP_SHARED (instead of MAP_PRIVATE) needed to be able to
+  // later use msync(MS_INVALIDATE) (see os::uncommit_memory).
+  int flags = MAP_ANONYMOUS | MAP_SHARED;
+
+  // MAP_FIXED is needed to enforce requested_addr - manpage is vague about what
+  // it means if wishaddress is given but MAP_FIXED is not set.
+  //
+  // Important! Behaviour differs depending on whether SPEC1170 mode is active or not.
+  // SPEC1170 mode active: behaviour like POSIX, MAP_FIXED will clobber existing mappings.
+  // SPEC1170 mode not active: behaviour, unlike POSIX, is that no existing mappings will
+  // get clobbered.
+  if (requested_addr != NULL) {
+    if (!os::Aix::xpg_sus_mode()) {  // not SPEC1170 Behaviour
+      flags |= MAP_FIXED;
+    }
+  }
+
+  char* addr = (char*)::mmap(requested_addr, extra_size,
+      PROT_READ|PROT_WRITE|PROT_EXEC, flags, -1, 0);
+
+  if (addr == MAP_FAILED) {
+    trcVerbose("mmap(" PTR_FORMAT ", " UINTX_FORMAT ", ..) failed (%d)", requested_addr, size, errno);
+    return NULL;
+  }
+
+  // Handle alignment.
+  char* const addr_aligned = (char *)align_ptr_up(addr, alignment_hint);
+  const size_t waste_pre = addr_aligned - addr;
+  char* const addr_aligned_end = addr_aligned + size;
+  const size_t waste_post = extra_size - waste_pre - size;
+  if (waste_pre > 0) {
+    ::munmap(addr, waste_pre);
+  }
+  if (waste_post > 0) {
+    ::munmap(addr_aligned_end, waste_post);
+  }
+  addr = addr_aligned;
+
+  if (addr) {
+    trcVerbose("mmap-allocated " PTR_FORMAT " .. " PTR_FORMAT " (" UINTX_FORMAT " bytes)",
+      addr, addr + bytes, bytes);
+  } else {
+    if (requested_addr != NULL) {
+      trcVerbose("failed to mmap-allocate " UINTX_FORMAT " bytes at wish address " PTR_FORMAT ".", bytes, requested_addr);
+    } else {
+      trcVerbose("failed to mmap-allocate " UINTX_FORMAT " bytes at any address.", bytes);
+    }
+  }
+
+  // bookkeeping
+  vmembk_add(addr, size, SIZE_4K, VMEM_MAPPED);
+
+  // Test alignment, see above.
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+
+  return addr;
+}
+
+static bool release_mmaped_memory(char* addr, size_t size) {
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+  assert0(is_aligned_to(size, os::vm_page_size()));
+
+  trcVerbose("release_mmaped_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
+    addr, addr + size - 1);
+  bool rc = false;
+
+  if (::munmap(addr, size) != 0) {
+    trcVerbose("failed (%d)\n", errno);
+    rc = false;
+  } else {
+    trcVerbose("ok.");
+    rc = true;
+  }
+
+  return rc;
+}
+
+static bool uncommit_mmaped_memory(char* addr, size_t size) {
+
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+  assert0(is_aligned_to(size, os::vm_page_size()));
+
+  trcVerbose("uncommit_mmaped_memory [" PTR_FORMAT " - " PTR_FORMAT "].",
+    addr, addr + size - 1);
+  bool rc = false;
+
+  // Uncommit mmap memory with msync MS_INVALIDATE.
+  if (::msync(addr, size, MS_INVALIDATE) != 0) {
+    trcVerbose("failed (%d)\n", errno);
+    rc = false;
+  } else {
+    trcVerbose("ok.");
+    rc = true;
+  }
+
+  return rc;
+}
 
 // End: shared memory bookkeeping
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 
 int os::vm_page_size() {
-  // Seems redundant as all get out
+  // Seems redundant as all get out.
   assert(os::Aix::page_size() != -1, "must call os::init");
   return os::Aix::page_size();
 }
@@ -2131,91 +2265,76 @@
   return os::Aix::page_size();
 }
 
-int os::Aix::commit_memory_impl(char* addr, size_t size, bool exec) {
-
-  // Commit is a noop. There is no explicit commit
-  // needed on AIX. Memory is committed when touched.
-  //
-  // Debug : check address range for validity
-#ifdef ASSERT
-  LOCK_SHMBK
-    ShmBkBlock* const block = shmbk_find_by_containing_address(addr);
-    if (!block) {
-      fprintf(stderr, "invalid pointer: " INTPTR_FORMAT "\n", addr);
-      shmbk_dump_info();
-      assert(false, "invalid pointer");
-      return false;
-    } else if (!block->containsRange(addr, size)) {
-      fprintf(stderr, "invalid range: " INTPTR_FORMAT " .. " INTPTR_FORMAT "\n", addr, addr + size);
-      shmbk_dump_info();
-      assert(false, "invalid range");
-      return false;
-    }
-  UNLOCK_SHMBK
-#endif // ASSERT
-
-  return 0;
-}
-
-bool os::pd_commit_memory(char* addr, size_t size, bool exec) {
-  return os::Aix::commit_memory_impl(addr, size, exec) == 0;
-}
+#ifdef PRODUCT
+static void warn_fail_commit_memory(char* addr, size_t size, bool exec,
+                                    int err) {
+  warning("INFO: os::commit_memory(" PTR_FORMAT ", " SIZE_FORMAT
+          ", %d) failed; error='%s' (errno=%d)", addr, size, exec,
+          strerror(err), err);
+}
+#endif
 
 void os::pd_commit_memory_or_exit(char* addr, size_t size, bool exec,
                                   const char* mesg) {
   assert(mesg != NULL, "mesg must be specified");
-  os::Aix::commit_memory_impl(addr, size, exec);
-}
-
-int os::Aix::commit_memory_impl(char* addr, size_t size,
-                                size_t alignment_hint, bool exec) {
-  return os::Aix::commit_memory_impl(addr, size, exec);
-}
-
-bool os::pd_commit_memory(char* addr, size_t size, size_t alignment_hint,
-                          bool exec) {
-  return os::Aix::commit_memory_impl(addr, size, alignment_hint, exec) == 0;
+  if (!pd_commit_memory(addr, size, exec)) {
+    // Add extra info in product mode for vm_exit_out_of_memory():
+    PRODUCT_ONLY(warn_fail_commit_memory(addr, size, exec, errno);)
+    vm_exit_out_of_memory(size, OOM_MMAP_ERROR, mesg);
+  }
+}
+
+bool os::pd_commit_memory(char* addr, size_t size, bool exec) {
+
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+  assert0(is_aligned_to(size, os::vm_page_size()));
+
+  vmembk_t* const vmi = vmembk_find(addr);
+  assert0(vmi);
+  vmi->assert_is_valid_subrange(addr, size);
+
+  trcVerbose("commit_memory [" PTR_FORMAT " - " PTR_FORMAT "].", addr, addr + size - 1);
+
+  return true;
+}
+
+bool os::pd_commit_memory(char* addr, size_t size, size_t alignment_hint, bool exec) {
+  return pd_commit_memory(addr, size, exec);
 }
 
 void os::pd_commit_memory_or_exit(char* addr, size_t size,
                                   size_t alignment_hint, bool exec,
                                   const char* mesg) {
-  os::Aix::commit_memory_impl(addr, size, alignment_hint, exec);
+  // Alignment_hint is ignored on this OS.
+  pd_commit_memory_or_exit(addr, size, exec, mesg);
 }
 
 bool os::pd_uncommit_memory(char* addr, size_t size) {
-
-  // Delegate to ShmBkBlock class which knows how to uncommit its memory.
-
-  bool rc = false;
-  LOCK_SHMBK
-    ShmBkBlock* const block = shmbk_find_by_containing_address(addr);
-    if (!block) {
-      fprintf(stderr, "invalid pointer: 0x%p.\n", addr);
-      shmbk_dump_info();
-      assert(false, "invalid pointer");
-      return false;
-    } else if (!block->containsRange(addr, size)) {
-      fprintf(stderr, "invalid range: 0x%p .. 0x%p.\n", addr, addr + size);
-      shmbk_dump_info();
-      assert(false, "invalid range");
-      return false;
-    }
-    rc = block->disclaim(addr, size);
-  UNLOCK_SHMBK
-
-  if (Verbose && !rc) {
-    warning("failed to disclaim 0x%p .. 0x%p (0x%llX bytes).", addr, addr + size, size);
-  }
-  return rc;
+  assert0(is_aligned_to(addr, os::vm_page_size()));
+  assert0(is_aligned_to(size, os::vm_page_size()));
+
+  // Dynamically do different things for mmap/shmat.
+  const vmembk_t* const vmi = vmembk_find(addr);
+  assert0(vmi);
+  vmi->assert_is_valid_subrange(addr, size);
+
+  if (vmi->type == VMEM_SHMATED) {
+    return uncommit_shmated_memory(addr, size);
+  } else {
+    return uncommit_mmaped_memory(addr, size);
+  }
 }
 
 bool os::pd_create_stack_guard_pages(char* addr, size_t size) {
-  return os::guard_memory(addr, size);
+  // Do not call this; no need to commit stack pages on AIX.
+  ShouldNotReachHere();
+  return true;
 }
 
 bool os::remove_stack_guard_pages(char* addr, size_t size) {
-  return os::unguard_memory(addr, size);
+  // Do not call this; no need to commit stack pages on AIX.
+  ShouldNotReachHere();
+  return true;
 }
 
 void os::pd_realign_memory(char *addr, size_t bytes, size_t alignment_hint) {
@@ -2258,355 +2377,75 @@
   return end;
 }
 
-// Flags for reserve_shmatted_memory:
-#define RESSHM_WISHADDR_OR_FAIL                     1
-#define RESSHM_TRY_16M_PAGES                        2
-#define RESSHM_16M_PAGES_OR_FAIL                    4
-
-// Result of reserve_shmatted_memory:
-struct shmatted_memory_info_t {
-  char* addr;
-  size_t pagesize;
-  bool pinned;
-};
-
-// Reserve a section of shmatted memory.
-// params:
-// bytes [in]: size of memory, in bytes
-// requested_addr [in]: wish address.
-//                      NULL = no wish.
-//                      If RESSHM_WISHADDR_OR_FAIL is set in flags and wish address cannot
-//                      be obtained, function will fail. Otherwise wish address is treated as hint and
-//                      another pointer is returned.
-// flags [in]:          some flags. Valid flags are:
-//                      RESSHM_WISHADDR_OR_FAIL - fail if wish address is given and cannot be obtained.
-//                      RESSHM_TRY_16M_PAGES - try to allocate from 16M page pool
-//                          (requires UseLargePages and Use16MPages)
-//                      RESSHM_16M_PAGES_OR_FAIL - if you cannot allocate from 16M page pool, fail.
-//                          Otherwise any other page size will do.
-// p_info [out] :       holds information about the created shared memory segment.
-static bool reserve_shmatted_memory(size_t bytes, char* requested_addr, int flags, shmatted_memory_info_t* p_info) {
-
-  assert(p_info, "parameter error");
-
-  // init output struct.
-  p_info->addr = NULL;
-
-  // neither should we be here for EXTSHM=ON.
-  if (os::Aix::extshm()) {
-    ShouldNotReachHere();
-  }
-
-  // extract flags. sanity checks.
-  const bool wishaddr_or_fail =
-    flags & RESSHM_WISHADDR_OR_FAIL;
-  const bool try_16M_pages =
-    flags & RESSHM_TRY_16M_PAGES;
-  const bool f16M_pages_or_fail =
-    flags & RESSHM_16M_PAGES_OR_FAIL;
-
-  // first check: if a wish address is given and it is mandatory, but not aligned to segment boundary,
-  // shmat will fail anyway, so save some cycles by failing right away
-  if (requested_addr && ((uintptr_t)requested_addr % SIZE_256M == 0)) {
-    if (wishaddr_or_fail) {
-      return false;
-    } else {
-      requested_addr = NULL;
-    }
-  }
-
-  char* addr = NULL;
-
-  // Align size of shm up to the largest possible page size, to avoid errors later on when we try to change
-  // pagesize dynamically.
-  const size_t size = align_size_up(bytes, SIZE_16M);
-
-  // reserve the shared segment
-  int shmid = shmget(IPC_PRIVATE, size, IPC_CREAT | S_IRUSR | S_IWUSR);
-  if (shmid == -1) {
-    warning("shmget(.., %lld, ..) failed (errno: %d).", size, errno);
-    return false;
-  }
-
-  // Important note:
-  // It is very important that we, upon leaving this function, do not leave a shm segment alive.
-  // We must right after attaching it remove it from the system. System V shm segments are global and
-  // survive the process.
-  // So, from here on: Do not assert. Do not return. Always do a "goto cleanup_shm".
-
-  // try forcing the page size
-  size_t pagesize = -1; // unknown so far
-
-  if (UseLargePages) {
-
-    struct shmid_ds shmbuf;
-    memset(&shmbuf, 0, sizeof(shmbuf));
-
-    // First, try to take from 16M page pool if...
-    if (os::Aix::can_use_16M_pages()  // we can ...
-        && Use16MPages                // we are not explicitly forbidden to do so (-XX:-Use16MPages)..
-        && try_16M_pages) {           // caller wants us to.
-      shmbuf.shm_pagesize = SIZE_16M;
-      if (shmctl(shmid, SHM_PAGESIZE, &shmbuf) == 0) {
-        pagesize = SIZE_16M;
-      } else {
-        warning("Failed to allocate %d 16M pages. 16M page pool might be exhausted. (shmctl failed with %d)",
-                size / SIZE_16M, errno);
-        if (f16M_pages_or_fail) {
-          goto cleanup_shm;
-        }
-      }
-    }
-
-    // Nothing yet? Try setting 64K pages. Note that I never saw this fail, but in theory it might,
-    // because the 64K page pool may also be exhausted.
-    if (pagesize == -1) {
-      shmbuf.shm_pagesize = SIZE_64K;
-      if (shmctl(shmid, SHM_PAGESIZE, &shmbuf) == 0) {
-        pagesize = SIZE_64K;
-      } else {
-        warning("Failed to allocate %d 64K pages. (shmctl failed with %d)",
-                size / SIZE_64K, errno);
-        // here I give up. leave page_size -1 - later, after attaching, we will query the
-        // real page size of the attached memory. (in theory, it may be something different
-        // from 4K if LDR_CNTRL SHM_PSIZE is set)
-      }
-    }
-  }
-
-  // sanity point
-  assert(pagesize == -1 || pagesize == SIZE_16M || pagesize == SIZE_64K, "wrong page size");
-
-  // Now attach the shared segment.
-  addr = (char*) shmat(shmid, requested_addr, 0);
-  if (addr == (char*)-1) {
-    // How to handle attach failure:
-    // If it failed for a specific wish address, tolerate this: in that case, if wish address was
-    // mandatory, fail, if not, retry anywhere.
-    // If it failed for any other reason, treat that as fatal error.
-    addr = NULL;
-    if (requested_addr) {
-      if (wishaddr_or_fail) {
-        goto cleanup_shm;
-      } else {
-        addr = (char*) shmat(shmid, NULL, 0);
-        if (addr == (char*)-1) { // fatal
-          addr = NULL;
-          warning("shmat failed (errno: %d)", errno);
-          goto cleanup_shm;
-        }
-      }
-    } else { // fatal
-      addr = NULL;
-      warning("shmat failed (errno: %d)", errno);
-      goto cleanup_shm;
-    }
-  }
-
-  // sanity point
-  assert(addr && addr != (char*) -1, "wrong address");
-
-  // after successful Attach remove the segment - right away.
-  if (::shmctl(shmid, IPC_RMID, NULL) == -1) {
-    warning("shmctl(%u, IPC_RMID) failed (%d)\n", shmid, errno);
-    guarantee(false, "failed to remove shared memory segment!");
-  }
-  shmid = -1;
-
-  // query the real page size. In case setting the page size did not work (see above), the system
-  // may have given us something other then 4K (LDR_CNTRL)
-  {
-    const size_t real_pagesize = os::Aix::query_pagesize(addr);
-    if (pagesize != -1) {
-      assert(pagesize == real_pagesize, "unexpected pagesize after shmat");
-    } else {
-      pagesize = real_pagesize;
-    }
-  }
-
-  // Now register the reserved block with internal book keeping.
-  LOCK_SHMBK
-    const bool pinned = pagesize >= SIZE_16M ? true : false;
-    ShmBkShmatedBlock* const p_block = new ShmBkShmatedBlock(AddrRange(addr, size), pagesize, pinned);
-    assert(p_block, "");
-    shmbk_register(p_block);
-  UNLOCK_SHMBK
-
-cleanup_shm:
-
-  // if we have not done so yet, remove the shared memory segment. This is very important.
-  if (shmid != -1) {
-    if (::shmctl(shmid, IPC_RMID, NULL) == -1) {
-      warning("shmctl(%u, IPC_RMID) failed (%d)\n", shmid, errno);
-      guarantee(false, "failed to remove shared memory segment!");
-    }
-    shmid = -1;
-  }
-
-  // trace
-  if (Verbose && !addr) {
-    if (requested_addr != NULL) {
-      warning("failed to shm-allocate 0x%llX bytes at wish address 0x%p.", size, requested_addr);
-    } else {
-      warning("failed to shm-allocate 0x%llX bytes at any address.", size);
-    }
-  }
-
-  // hand info to caller
-  if (addr) {
-    p_info->addr = addr;
-    p_info->pagesize = pagesize;
-    p_info->pinned = pagesize == SIZE_16M ? true : false;
-  }
-
-  // sanity test:
-  if (requested_addr && addr && wishaddr_or_fail) {
-    guarantee(addr == requested_addr, "shmat error");
-  }
-
-  // just one more test to really make sure we have no dangling shm segments.
-  guarantee(shmid == -1, "dangling shm segments");
-
-  return addr ? true : false;
-
-} // end: reserve_shmatted_memory
-
-// Reserve memory using mmap. Behaves the same as reserve_shmatted_memory():
-// will return NULL in case of an error.
-static char* reserve_mmaped_memory(size_t bytes, char* requested_addr) {
-
-  // if a wish address is given, but not aligned to 4K page boundary, mmap will fail.
-  if (requested_addr && ((uintptr_t)requested_addr % os::vm_page_size() != 0)) {
-    warning("Wish address 0x%p not aligned to page boundary.", requested_addr);
-    return NULL;
-  }
-
-  const size_t size = align_size_up(bytes, SIZE_4K);
-
-  // Note: MAP_SHARED (instead of MAP_PRIVATE) needed to be able to
-  // msync(MS_INVALIDATE) (see os::uncommit_memory)
-  int flags = MAP_ANONYMOUS | MAP_SHARED;
-
-  // MAP_FIXED is needed to enforce requested_addr - manpage is vague about what
-  // it means if wishaddress is given but MAP_FIXED is not set.
-  //
-  // Note however that this changes semantics in SPEC1170 mode insofar as MAP_FIXED
-  // clobbers the address range, which is probably not what the caller wants. That's
-  // why I assert here (again) that the SPEC1170 compat mode is off.
-  // If we want to be able to run under SPEC1170, we have to do some porting and
-  // testing.
-  if (requested_addr != NULL) {
-    assert(!os::Aix::xpg_sus_mode(), "SPEC1170 mode not allowed.");
-    flags |= MAP_FIXED;
-  }
-
-  char* addr = (char*)::mmap(requested_addr, size, PROT_READ|PROT_WRITE|PROT_EXEC, flags, -1, 0);
-
-  if (addr == MAP_FAILED) {
-    // attach failed: tolerate for specific wish addresses. Not being able to attach
-    // anywhere is a fatal error.
-    if (requested_addr == NULL) {
-      // It's ok to fail here if the machine has not enough memory.
-      warning("mmap(NULL, 0x%llX, ..) failed (%d)", size, errno);
-    }
-    addr = NULL;
-    goto cleanup_mmap;
-  }
-
-  // If we did request a specific address and that address was not available, fail.
-  if (addr && requested_addr) {
-    guarantee(addr == requested_addr, "unexpected");
-  }
-
-  // register this mmap'ed segment with book keeping
-  LOCK_SHMBK
-    ShmBkMappedBlock* const p_block = new ShmBkMappedBlock(AddrRange(addr, size));
-    assert(p_block, "");
-    shmbk_register(p_block);
-  UNLOCK_SHMBK
-
-cleanup_mmap:
-
-  // trace
-  if (Verbose) {
-    if (addr) {
-      fprintf(stderr, "mmap-allocated 0x%p .. 0x%p (0x%llX bytes)\n", addr, addr + bytes, bytes);
-    }
-    else {
-      if (requested_addr != NULL) {
-        warning("failed to mmap-allocate 0x%llX bytes at wish address 0x%p.", bytes, requested_addr);
-      } else {
-        warning("failed to mmap-allocate 0x%llX bytes at any address.", bytes);
-      }
-    }
-  }
-
-  return addr;
-
-} // end: reserve_mmaped_memory
-
 // Reserves and attaches a shared memory segment.
 // Will assert if a wish address is given and could not be obtained.
 char* os::pd_reserve_memory(size_t bytes, char* requested_addr, size_t alignment_hint) {
-  return os::attempt_reserve_memory_at(bytes, requested_addr);
+
+  // All other Unices do a mmap(MAP_FIXED) if the addr is given,
+  // thereby clobbering old mappings at that place. That is probably
+  // not intended, never used and almost certainly an error were it
+  // ever be used this way (to try attaching at a specified address
+  // without clobbering old mappings an alternate API exists,
+  // os::attempt_reserve_memory_at()).
+  // Instead of mimicking the dangerous coding of the other platforms, here I
+  // just ignore the request address (release) or assert(debug).
+  assert0(requested_addr == NULL);
+
+  // Always round to os::vm_page_size(), which may be larger than 4K.
+  bytes = align_size_up(bytes, os::vm_page_size());
+  const size_t alignment_hint0 =
+    alignment_hint ? align_size_up(alignment_hint, os::vm_page_size()) : 0;
+
+  // In 4K mode always use mmap.
+  // In 64K mode allocate small sizes with mmap, large ones with 64K shmatted.
+  if (os::vm_page_size() == SIZE_4K) {
+    return reserve_mmaped_memory(bytes, requested_addr, alignment_hint);
+  } else {
+    if (bytes >= Use64KPagesThreshold) {
+      return reserve_shmated_memory(bytes, requested_addr, alignment_hint);
+    } else {
+      return reserve_mmaped_memory(bytes, requested_addr, alignment_hint);
+    }
+  }
 }
 
 bool os::pd_release_memory(char* addr, size_t size) {
 
-  // delegate to ShmBkBlock class which knows how to uncommit its memory.
+  // Dynamically do different things for mmap/shmat.
+  vmembk_t* const vmi = vmembk_find(addr);
+  assert0(vmi);
+
+  // Always round to os::vm_page_size(), which may be larger than 4K.
+  size = align_size_up(size, os::vm_page_size());
+  addr = (char *)align_ptr_up(addr, os::vm_page_size());
 
   bool rc = false;
-  LOCK_SHMBK
-    ShmBkBlock* const block = shmbk_find_by_containing_address(addr);
-    if (!block) {
-      fprintf(stderr, "invalid pointer: 0x%p.\n", addr);
-      shmbk_dump_info();
-      assert(false, "invalid pointer");
-      return false;
+  bool remove_bookkeeping = false;
+  if (vmi->type == VMEM_SHMATED) {
+    // For shmatted memory, we do:
+    // - If user wants to release the whole range, release the memory (shmdt).
+    // - If user only wants to release a partial range, uncommit (disclaim) that
+    //   range. That way, at least, we do not use memory anymore (bust still page
+    //   table space).
+    vmi->assert_is_valid_subrange(addr, size);
+    if (addr == vmi->addr && size == vmi->size) {
+      rc = release_shmated_memory(addr, size);
+      remove_bookkeeping = true;
+    } else {
+      rc = uncommit_shmated_memory(addr, size);
     }
-    else if (!block->isSameRange(addr, size)) {
-      if (block->getType() == ShmBkBlock::MMAP) {
-        // Release only the same range or a the beginning or the end of a range.
-        if (block->base() == addr && size < block->size()) {
-          ShmBkMappedBlock* const b = new ShmBkMappedBlock(AddrRange(block->base() + size, block->size() - size));
-          assert(b, "");
-          shmbk_register(b);
-          block->setAddrRange(AddrRange(addr, size));
-        }
-        else if (addr > block->base() && addr + size == block->base() + block->size()) {
-          ShmBkMappedBlock* const b = new ShmBkMappedBlock(AddrRange(block->base(), block->size() - size));
-          assert(b, "");
-          shmbk_register(b);
-          block->setAddrRange(AddrRange(addr, size));
-        }
-        else {
-          fprintf(stderr, "invalid mmap range: 0x%p .. 0x%p.\n", addr, addr + size);
-          shmbk_dump_info();
-          assert(false, "invalid mmap range");
-          return false;
-        }
-      }
-      else {
-        // Release only the same range. No partial release allowed.
-        // Soften the requirement a bit, because the user may think he owns a smaller size
-        // than the block is due to alignment etc.
-        if (block->base() != addr || block->size() < size) {
-          fprintf(stderr, "invalid shmget range: 0x%p .. 0x%p.\n", addr, addr + size);
-          shmbk_dump_info();
-          assert(false, "invalid shmget range");
-          return false;
-        }
-      }
-    }
-    rc = block->release();
-    assert(rc, "release failed");
-    // remove block from bookkeeping
-    shmbk_unregister(block);
-    delete block;
-  UNLOCK_SHMBK
-
-  if (!rc) {
-    warning("failed to released %lu bytes at 0x%p", size, addr);
+  } else {
+    // User may unmap partial regions but region has to be fully contained.
+#ifdef ASSERT
+    vmi->assert_is_valid_subrange(addr, size);
+#endif
+    rc = release_mmaped_memory(addr, size);
+    remove_bookkeeping = true;
+  }
+
+  // update bookkeeping
+  if (rc && remove_bookkeeping) {
+    vmembk_remove(vmi);
   }
 
   return rc;
@@ -2687,46 +2526,8 @@
 
 // Enable large page support if OS allows that.
 void os::large_page_init() {
-
-  // Note: os::Aix::query_multipage_support must run first.
-
-  if (!UseLargePages) {
-    return;
-  }
-
-  if (!Aix::can_use_64K_pages()) {
-    assert(!Aix::can_use_16M_pages(), "64K is a precondition for 16M.");
-    UseLargePages = false;
-    return;
-  }
-
-  if (!Aix::can_use_16M_pages() && Use16MPages) {
-    fprintf(stderr, "Cannot use 16M pages. Please ensure that there is a 16M page pool "
-            " and that the VM runs with CAP_BYPASS_RAC_VMM and CAP_PROPAGATE capabilities.\n");
-  }
-
-  // Do not report 16M page alignment as part of os::_page_sizes if we are
-  // explicitly forbidden from using 16M pages. Doing so would increase the
-  // alignment the garbage collector calculates with, slightly increasing
-  // heap usage. We should only pay for 16M alignment if we really want to
-  // use 16M pages.
-  if (Use16MPages && Aix::can_use_16M_pages()) {
-    _large_page_size = SIZE_16M;
-    _page_sizes[0] = SIZE_16M;
-    _page_sizes[1] = SIZE_64K;
-    _page_sizes[2] = SIZE_4K;
-    _page_sizes[3] = 0;
-  } else if (Aix::can_use_64K_pages()) {
-    _large_page_size = SIZE_64K;
-    _page_sizes[0] = SIZE_64K;
-    _page_sizes[1] = SIZE_4K;
-    _page_sizes[2] = 0;
-  }
-
-  if (Verbose) {
-    ("Default large page size is 0x%llX.", _large_page_size);
-  }
-} // end: os::large_page_init()
+  return; // Nothing to do. See query_multipage_support and friends.
+}
 
 char* os::reserve_memory_special(size_t bytes, size_t alignment, char* req_addr, bool exec) {
   // "exec" is passed in but not used. Creating the shared image for
@@ -2736,7 +2537,7 @@
 }
 
 bool os::release_memory_special(char* base, size_t bytes) {
-  // detaching the SHM segment will also delete it, see reserve_memory_special()
+  // Detaching the SHM segment will also delete it, see reserve_memory_special().
   Unimplemented();
   return false;
 }
@@ -2746,40 +2547,32 @@
 }
 
 bool os::can_commit_large_page_memory() {
-  // Well, sadly we cannot commit anything at all (see comment in
-  // os::commit_memory) but we claim to so we can make use of large pages
-  return true;
+  // Does not matter, we do not support huge pages.
+  return false;
 }
 
 bool os::can_execute_large_page_memory() {
-  // We can do that
-  return true;
+  // Does not matter, we do not support huge pages.
+  return false;
 }
 
 // Reserve memory at an arbitrary address, only if that area is
 // available (and not reserved for something else).
 char* os::pd_attempt_reserve_memory_at(size_t bytes, char* requested_addr) {
-
-  bool use_mmap = false;
-
-  // mmap: smaller graining, no large page support
-  // shm: large graining (256M), large page support, limited number of shm segments
-  //
-  // Prefer mmap wherever we either do not need large page support or have OS limits
-
-  if (!UseLargePages || bytes < SIZE_16M) {
-    use_mmap = true;
-  }
-
   char* addr = NULL;
-  if (use_mmap) {
-    addr = reserve_mmaped_memory(bytes, requested_addr);
+
+  // Always round to os::vm_page_size(), which may be larger than 4K.
+  bytes = align_size_up(bytes, os::vm_page_size());
+
+  // In 4K mode always use mmap.
+  // In 64K mode allocate small sizes with mmap, large ones with 64K shmatted.
+  if (os::vm_page_size() == SIZE_4K) {
+    return reserve_mmaped_memory(bytes, requested_addr, 0);
   } else {
-    // shmat: wish address is mandatory, and do not try 16M pages here.
-    shmatted_memory_info_t info;
-    const int flags = RESSHM_WISHADDR_OR_FAIL;
-    if (reserve_shmatted_memory(bytes, requested_addr, flags, &info)) {
-      addr = info.addr;
+    if (bytes >= Use64KPagesThreshold) {
+      return reserve_shmated_memory(bytes, requested_addr, 0);
+    } else {
+      return reserve_mmaped_memory(bytes, requested_addr, 0);
     }
   }
 
@@ -3614,18 +3407,89 @@
 // This is called _before_ the most of global arguments have been parsed.
 void os::init(void) {
   // This is basic, we want to know if that ever changes.
-  // (shared memory boundary is supposed to be a 256M aligned)
+  // (Shared memory boundary is supposed to be a 256M aligned.)
   assert(SHMLBA == ((uint64_t)0x10000000ULL)/*256M*/, "unexpected");
 
   // First off, we need to know whether we run on AIX or PASE, and
   // the OS level we run on.
   os::Aix::initialize_os_info();
 
-  // Scan environment (SPEC1170 behaviour, etc)
+  // Scan environment (SPEC1170 behaviour, etc).
   os::Aix::scan_environment();
 
   // Check which pages are supported by AIX.
-  os::Aix::query_multipage_support();
+  query_multipage_support();
+
+  // Act like we only have one page size by eliminating corner cases which
+  // we did not support very well anyway.
+  // We have two input conditions:
+  // 1) Data segment page size. This is controlled by linker setting (datapsize) on the
+  //    launcher, and/or by LDR_CNTRL environment variable. The latter overrules the linker
+  //    setting.
+  //    Data segment page size is important for us because it defines the thread stack page
+  //    size, which is needed for guard page handling, stack banging etc.
+  // 2) The ability to allocate 64k pages dynamically. If this is a given, java heap can
+  //    and should be allocated with 64k pages.
+  //
+  // So, we do the following:
+  // LDR_CNTRL    can_use_64K_pages_dynamically       what we do                      remarks
+  // 4K           no                                  4K                              old systems (aix 5.2, as/400 v5r4) or new systems with AME activated
+  // 4k           yes                                 64k (treat 4k stacks as 64k)    different loader than java and standard settings
+  // 64k          no              --- AIX 5.2 ? ---
+  // 64k          yes                                 64k                             new systems and standard java loader (we set datapsize=64k when linking)
+
+  // We explicitly leave no option to change page size, because only upgrading would work,
+  // not downgrading (if stack page size is 64k you cannot pretend its 4k).
+
+  if (g_multipage_support.datapsize == SIZE_4K) {
+    // datapsize = 4K. Data segment, thread stacks are 4K paged.
+    if (g_multipage_support.can_use_64K_pages) {
+      // .. but we are able to use 64K pages dynamically.
+      // This would be typical for java launchers which are not linked
+      // with datapsize=64K (like, any other launcher but our own).
+      //
+      // In this case it would be smart to allocate the java heap with 64K
+      // to get the performance benefit, and to fake 64k pages for the
+      // data segment (when dealing with thread stacks).
+      //
+      // However, leave a possibility to downgrade to 4K, using
+      // -XX:-Use64KPages.
+      if (Use64KPages) {
+        trcVerbose("64K page mode (faked for data segment)");
+        Aix::_page_size = SIZE_64K;
+      } else {
+        trcVerbose("4K page mode (Use64KPages=off)");
+        Aix::_page_size = SIZE_4K;
+      }
+    } else {
+      // .. and not able to allocate 64k pages dynamically. Here, just
+      // fall back to 4K paged mode and use mmap for everything.
+      trcVerbose("4K page mode");
+      Aix::_page_size = SIZE_4K;
+      FLAG_SET_ERGO(bool, Use64KPages, false);
+    }
+  } else {
+    // datapsize = 64k. Data segment, thread stacks are 64k paged.
+    //   This normally means that we can allocate 64k pages dynamically.
+    //   (There is one special case where this may be false: EXTSHM=on.
+    //    but we decided to not support that mode).
+    assert0(g_multipage_support.can_use_64K_pages);
+    Aix::_page_size = SIZE_64K;
+    trcVerbose("64K page mode");
+    FLAG_SET_ERGO(bool, Use64KPages, true);
+  }
+
+  // Short-wire stack page size to base page size; if that works, we just remove
+  // that stack page size altogether.
+  Aix::_stack_page_size = Aix::_page_size;
+
+  // For now UseLargePages is just ignored.
+  FLAG_SET_ERGO(bool, UseLargePages, false);
+  _page_sizes[0] = 0;
+  _large_page_size = -1;
+
+  // debug trace
+  trcVerbose("os::vm_page_size %s\n", describe_pagesize(os::vm_page_size()));
 
   // Next, we need to initialize libo4 and libperfstat libraries.
   if (os::Aix::on_pase()) {
@@ -3643,34 +3507,6 @@
   // need libperfstat etc.
   os::Aix::initialize_system_info();
 
-  // Initialize large page support.
-  if (UseLargePages) {
-    os::large_page_init();
-    if (!UseLargePages) {
-      // initialize os::_page_sizes
-      _page_sizes[0] = Aix::page_size();
-      _page_sizes[1] = 0;
-      if (Verbose) {
-        fprintf(stderr, "Large Page initialization failed: setting UseLargePages=0.\n");
-      }
-    }
-  } else {
-    // initialize os::_page_sizes
-    _page_sizes[0] = Aix::page_size();
-    _page_sizes[1] = 0;
-  }
-
-  // debug trace
-  if (Verbose) {
-    fprintf(stderr, "os::vm_page_size 0x%llX\n", os::vm_page_size());
-    fprintf(stderr, "os::large_page_size 0x%llX\n", os::large_page_size());
-    fprintf(stderr, "os::_page_sizes = ( ");
-    for (int i = 0; _page_sizes[i]; i ++) {
-      fprintf(stderr, " %s ", describe_pagesize(_page_sizes[i]));
-    }
-    fprintf(stderr, ")\n");
-  }
-
   _initial_pid = getpid();
 
   clock_tics_per_sec = sysconf(_SC_CLK_TCK);
@@ -3683,6 +3519,15 @@
   Aix::_main_thread = pthread_self();
 
   initial_time_count = os::elapsed_counter();
+
+  // If the pagesize of the VM is greater than 8K determine the appropriate
+  // number of initial guard pages. The user can change this with the
+  // command line arguments, if needed.
+  if (vm_page_size() > (int)Aix::vm_default_page_size()) {
+    StackYellowPages = 1;
+    StackRedPages = 1;
+    StackShadowPages = round_to((StackShadowPages*Aix::vm_default_page_size()), vm_page_size()) / vm_page_size();
+  }
 }
 
 // This is called _after_ the global arguments have been parsed.
@@ -3701,7 +3546,7 @@
   const int prot  = PROT_READ;
   const int flags = MAP_PRIVATE|MAP_ANONYMOUS;
 
-  // use optimized addresses for the polling page,
+  // Use optimized addresses for the polling page,
   // e.g. map it to a special 32-bit address.
   if (OptimizePollingPageLocation) {
     // architecture-specific list of address wishes:
@@ -3723,7 +3568,7 @@
 
     // iterate over the list of address wishes:
     for (int i=0; i<address_wishes_length; i++) {
-      // try to map with current address wish.
+      // Try to map with current address wish.
       // AIX: AIX needs MAP_FIXED if we provide an address and mmap will
       // fail if the address is already mapped.
       map_address = (address) ::mmap(address_wishes[i] - (ssize_t)page_size,
@@ -3736,7 +3581,7 @@
       }
 
       if (map_address + (ssize_t)page_size == address_wishes[i]) {
-        // map succeeded and map_address is at wished address, exit loop.
+        // Map succeeded and map_address is at wished address, exit loop.
         break;
       }
 
@@ -3745,7 +3590,7 @@
         ::munmap(map_address, map_size);
         map_address = (address) MAP_FAILED;
       }
-      // map failed, continue loop.
+      // Map failed, continue loop.
     }
   } // end OptimizePollingPageLocation
 
@@ -3761,8 +3606,9 @@
     os::set_memory_serialize_page(mem_serialize_page);
 
 #ifndef PRODUCT
-    if (Verbose && PrintMiscellaneous)
+    if (Verbose && PrintMiscellaneous) {
       tty->print("[Memory Serialize Page address: " INTPTR_FORMAT "]\n", (intptr_t)mem_serialize_page);
+    }
 #endif
   }
 
@@ -3781,16 +3627,18 @@
   // Add in 2*BytesPerWord times page size to account for VM stack during
   // class initialization depending on 32 or 64 bit VM.
   os::Aix::min_stack_allowed = MAX2(os::Aix::min_stack_allowed,
-            (size_t)(StackYellowPages+StackRedPages+StackShadowPages +
-                     2*BytesPerWord COMPILER2_PRESENT(+1)) * Aix::page_size());
+            (size_t)(StackYellowPages+StackRedPages+StackShadowPages) * Aix::page_size() +
+                     (2*BytesPerWord COMPILER2_PRESENT(+1)) * Aix::vm_default_page_size());
+
+  os::Aix::min_stack_allowed = align_size_up(os::Aix::min_stack_allowed, os::Aix::page_size());
 
   size_t threadStackSizeInBytes = ThreadStackSize * K;
   if (threadStackSizeInBytes != 0 &&
       threadStackSizeInBytes < os::Aix::min_stack_allowed) {
-        tty->print_cr("\nThe stack size specified is too small, "
-                      "Specify at least %dk",
-                      os::Aix::min_stack_allowed / K);
-        return JNI_ERR;
+    tty->print_cr("\nThe stack size specified is too small, "
+                  "Specify at least %dk",
+                  os::Aix::min_stack_allowed / K);
+    return JNI_ERR;
   }
 
   // Make the stack size a multiple of the page size so that
@@ -3801,7 +3649,7 @@
   Aix::libpthread_init();
 
   if (MaxFDLimit) {
-    // set the number of file descriptors to max. print out error
+    // Set the number of file descriptors to max. print out error
     // if getrlimit/setrlimit fails but continue regardless.
     struct rlimit nbr_files;
     int status = getrlimit(RLIMIT_NOFILE, &nbr_files);
@@ -3819,12 +3667,12 @@
   }
 
   if (PerfAllowAtExitRegistration) {
-    // only register atexit functions if PerfAllowAtExitRegistration is set.
-    // atexit functions can be delayed until process exit time, which
+    // Only register atexit functions if PerfAllowAtExitRegistration is set.
+    // Atexit functions can be delayed until process exit time, which
     // can be problematic for embedded VM situations. Embedded VMs should
     // call DestroyJavaVM() to assure that VM resources are released.
 
-    // note: perfMemory_exit_helper atexit function may be removed in
+    // Note: perfMemory_exit_helper atexit function may be removed in
     // the future if the appropriate cleanup code can be added to the
     // VM_Exit VMOperation's doit method.
     if (atexit(perfMemory_exit_helper) != 0) {
@@ -4146,8 +3994,10 @@
 
   if (read_only) {
     prot = PROT_READ;
+    flags = MAP_SHARED;
   } else {
     prot = PROT_READ | PROT_WRITE;
+    flags = MAP_PRIVATE;
   }
 
   if (allow_exec) {
@@ -4158,7 +4008,12 @@
     flags |= MAP_FIXED;
   }
 
-  char* mapped_address = (char*)mmap(addr, (size_t)bytes, prot, flags,
+  // Allow anonymous mappings if 'fd' is -1.
+  if (fd == -1) {
+    flags |= MAP_ANONYMOUS;
+  }
+
+  char* mapped_address = (char*)::mmap(addr, (size_t)bytes, prot, flags,
                                      fd, file_offset);
   if (mapped_address == MAP_FAILED) {
     return NULL;
@@ -4416,7 +4271,7 @@
   if (Verbose) {
     fprintf(stderr, "EXTSHM=%s.\n", p ? p : "<unset>");
   }
-  if (p && strcmp(p, "ON") == 0) {
+  if (p && strcasecmp(p, "ON") == 0) {
     fprintf(stderr, "Unsupported setting: EXTSHM=ON. Large Page support will be disabled.\n");
     _extshm = 1;
   } else {
@@ -4477,16 +4332,13 @@
 /////////////////////////////////////////////////////////////////////////////
 // thread stack
 
-// function to query the current stack size using pthread_getthrds_np
-//
-// ! do not change anything here unless you know what you are doing !
-static void query_stack_dimensions(address* p_stack_base, size_t* p_stack_size) {
-
+// Function to query the current stack size using pthread_getthrds_np.
+static bool query_stack_dimensions(address* p_stack_base, size_t* p_stack_size) {
   // This only works when invoked on a pthread. As we agreed not to use
-  // primordial threads anyway, I assert here
+  // primordial threads anyway, I assert here.
   guarantee(!os::Aix::is_primordial_thread(), "not allowed on the primordial thread");
 
-  // information about this api can be found (a) in the pthread.h header and
+  // Information about this api can be found (a) in the pthread.h header and
   // (b) in http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.basetechref/doc/basetrf1/pthread_getthrds_np.htm
   //
   // The use of this API to find out the current stack is kind of undefined.
@@ -4497,57 +4349,72 @@
 
   pthread_t tid = pthread_self();
   struct __pthrdsinfo pinfo;
-  char dummy[1]; // we only need this to satisfy the api and to not get E
+  char dummy[1]; // We only need this to satisfy the api and to not get E.
   int dummy_size = sizeof(dummy);
 
   memset(&pinfo, 0, sizeof(pinfo));
 
-  const int rc = pthread_getthrds_np (&tid, PTHRDSINFO_QUERY_ALL, &pinfo,
-                                      sizeof(pinfo), dummy, &dummy_size);
+  const int rc = pthread_getthrds_np(&tid, PTHRDSINFO_QUERY_ALL, &pinfo,
+                                     sizeof(pinfo), dummy, &dummy_size);
 
   if (rc != 0) {
-    fprintf(stderr, "pthread_getthrds_np failed (%d)\n", rc);
-    guarantee(0, "pthread_getthrds_np failed");
-  }
-
-  guarantee(pinfo.__pi_stackend, "returned stack base invalid");
-
-  // the following can happen when invoking pthread_getthrds_np on a pthread running on a user provided stack
-  // (when handing down a stack to pthread create, see pthread_attr_setstackaddr).
+    assert0(false);
+    trcVerbose("pthread_getthrds_np failed (%d)", rc);
+    return false;
+  }
+  guarantee0(pinfo.__pi_stackend);
+
+  // The following can happen when invoking pthread_getthrds_np on a pthread running
+  // on a user provided stack (when handing down a stack to pthread create, see
+  // pthread_attr_setstackaddr).
   // Not sure what to do here - I feel inclined to forbid this use case completely.
-  guarantee(pinfo.__pi_stacksize, "returned stack size invalid");
-
-  // On AIX, stacks are not necessarily page aligned so round the base and size accordingly
+  guarantee0(pinfo.__pi_stacksize);
+
+  // Note: the pthread stack on AIX seems to look like this:
+  //
+  // ---------------------   real base ? at page border ?
+  //
+  //     pthread internal data, like ~2K, see also
+  //     http://publib.boulder.ibm.com/infocenter/pseries/v5r3/index.jsp?topic=/com.ibm.aix.prftungd/doc/prftungd/thread_supp_tun_params.htm
+  //
+  // ---------------------   __pi_stackend - not page aligned, (xxxxF890)
+  //
+  //     stack
+  //      ....
+  //
+  //     stack
+  //
+  // ---------------------   __pi_stackend  - __pi_stacksize
+  //
+  //     padding due to AIX guard pages (?) see AIXTHREAD_GUARDPAGES
+  // ---------------------   __pi_stackaddr  (page aligned if AIXTHREAD_GUARDPAGES > 0)
+  //
+  //   AIX guard pages (?)
+  //
+
+  // So, the safe thing to do is to use the area from __pi_stackend to __pi_stackaddr;
+  // __pi_stackend however is almost never page aligned.
+  //
+
   if (p_stack_base) {
-    (*p_stack_base) = (address) align_size_up((intptr_t)pinfo.__pi_stackend, os::Aix::stack_page_size());
+    (*p_stack_base) = (address) (pinfo.__pi_stackend);
   }
 
   if (p_stack_size) {
-    (*p_stack_size) = pinfo.__pi_stacksize - os::Aix::stack_page_size();
-  }
-
-#ifndef PRODUCT
-  if (Verbose) {
-    fprintf(stderr,
-            "query_stack_dimensions() -> real stack_base=" INTPTR_FORMAT ", real stack_addr=" INTPTR_FORMAT
-            ", real stack_size=" INTPTR_FORMAT
-            ", stack_base=" INTPTR_FORMAT ", stack_size=" INTPTR_FORMAT "\n",
-            (intptr_t)pinfo.__pi_stackend, (intptr_t)pinfo.__pi_stackaddr, pinfo.__pi_stacksize,
-            (intptr_t)align_size_up((intptr_t)pinfo.__pi_stackend, os::Aix::stack_page_size()),
-            pinfo.__pi_stacksize - os::Aix::stack_page_size());
-  }
-#endif
-
-} // end query_stack_dimensions
-
-// get the current stack base from the OS (actually, the pthread library)
+    (*p_stack_size) = pinfo.__pi_stackend - pinfo.__pi_stackaddr;
+  }
+
+  return true;
+}
+
+// Get the current stack base from the OS (actually, the pthread library).
 address os::current_stack_base() {
   address p;
   query_stack_dimensions(&p, 0);
   return p;
 }
 
-// get the current stack size from the OS (actually, the pthread library)
+// Get the current stack size from the OS (actually, the pthread library).
 size_t os::current_stack_size() {
   size_t s;
   query_stack_dimensions(0, &s);
--- a/hotspot/src/os/aix/vm/os_aix.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/os_aix.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1999, 2013, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2013 SAP AG. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2013, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -35,9 +35,9 @@
   friend class os;
 
   // For signal-chaining
-  // highest so far (AIX 5.2) is SIGSAK (63)
+  // highest so far (AIX 5.2 - 6.1) is SIGSAK (63)
 #define MAXSIGNUM 63
-  // length of strings included in the libperfstat structures
+  // Length of strings included in the libperfstat structures.
 #define IDENTIFIER_LENGTH 64
 
   static struct sigaction sigact[MAXSIGNUM]; // saved preinstalled sigactions
@@ -111,22 +111,6 @@
   // (should be LDR_CNTRL DATAPSIZE because stack is allocated on heap by pthread lib)
   static int _stack_page_size;
 
-  // Default shm page size. Read: what page size shared memory will be backed
-  // with if no page size was set explicitly using shmctl(SHM_PAGESIZE).
-  // Should be LDR_CNTRL SHMPSIZE.
-  static size_t _shm_default_page_size;
-
-  // True if sys V shm can be used with 64K pages dynamically.
-  // (via shmctl(.. SHM_PAGESIZE..). Should be true for AIX 53 and
-  // newer / PASE V6R1 and newer. (0 or 1, -1 if not initialized)
-  static int _can_use_64K_pages;
-
-  // True if sys V shm can be used with 16M pages dynamically.
-  // (via shmctl(.. SHM_PAGESIZE..). Only true on AIX 5.3 and
-  // newer, if the system was set up to use 16M pages and the
-  // jvm has enough user rights. (0 or 1, -1 if not initialized)
-  static int _can_use_16M_pages;
-
   static julong available_memory();
   static julong physical_memory() { return _physical_memory; }
   static void initialize_system_info();
@@ -135,10 +119,6 @@
   // one of Aix::on_pase(), Aix::os_version().
   static void initialize_os_info();
 
-  static int commit_memory_impl(char* addr, size_t bytes, bool exec);
-  static int commit_memory_impl(char* addr, size_t bytes,
-                                size_t alignment_hint, bool exec);
-
   // Scan environment for important settings which might effect the
   // VM. Trace out settings. Warn about invalid settings and/or
   // correct them.
@@ -146,10 +126,6 @@
   // Must run after os::Aix::initialue_os_info().
   static void scan_environment();
 
-  // Retrieve information about multipage size support. Will initialize
-  // _page_size, _stack_page_size, _can_use_64K_pages/_can_use_16M_pages
-  static void query_multipage_support();
-
   // Initialize libo4 (on PASE) and libperfstat (on AIX). Call this
   // before relying on functions from either lib, e.g. Aix::get_meminfo().
   static void initialize_libo4();
@@ -187,27 +163,8 @@
     return _stack_page_size;
   }
 
-  // default shm page size. Read: what page size shared memory
-  // will be backed with if no page size was set explicitly using shmctl(SHM_PAGESIZE).
-  // Should be LDR_CNTRL SHMPSIZE.
-  static int shm_default_page_size(void) {
-    assert(_shm_default_page_size != -1, "not initialized");
-    return _shm_default_page_size;
-  }
-
-  // Return true if sys V shm can be used with 64K pages dynamically
-  // (via shmctl(.. SHM_PAGESIZE..).
-  static bool can_use_64K_pages () {
-    assert(_can_use_64K_pages != -1,  "not initialized");
-    return _can_use_64K_pages == 1 ? true : false;
-  }
-
-  // Return true if sys V shm can be used with 16M pages dynamically.
-  // (via shmctl(.. SHM_PAGESIZE..).
-  static bool can_use_16M_pages () {
-    assert(_can_use_16M_pages != -1,  "not initialized");
-    return _can_use_16M_pages == 1 ? true : false;
-  }
+  // This is used to scale stack space (guard pages etc.). The name is somehow misleading.
+  static int vm_default_page_size(void ) { return 8*K; }
 
   static address   ucontext_get_pc(const ucontext_t* uc);
   static intptr_t* ucontext_get_sp(ucontext_t* uc);
@@ -269,6 +226,11 @@
     return _os_version;
   }
 
+  // Convenience method: returns true if running on PASE V5R4 or older.
+  static bool on_pase_V5R4_or_older() {
+    return on_pase() && os_version() <= 0x0504;
+  }
+
   // Convenience method: returns true if running on AIX 5.3 or older.
   static bool on_aix_53_or_older() {
     return on_aix() && os_version() <= 0x0503;
--- a/hotspot/src/os/aix/vm/os_aix.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/os_aix.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,6 +1,6 @@
 /*
- * Copyright (c) 1999, 2014, Oracle and/or its affiliates. All rights reserved.
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright (c) 1999, 2015, Oracle and/or its affiliates. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -40,7 +40,7 @@
   return pthread_getspecific((pthread_key_t)index);
 }
 
-// File names are case-sensitive on windows only
+// File names are case-sensitive on windows only.
 inline int os::file_name_strcmp(const char* s1, const char* s2) {
   return strcmp(s1, s2);
 }
@@ -53,18 +53,19 @@
   return true;
 }
 
+// Whether or not calling code should/can commit/uncommit stack pages
+// before guarding them. Answer for AIX is definitly no, because memory
+// is automatically committed on touch.
 inline bool os::allocate_stack_guard_pages() {
   assert(uses_stack_guard_pages(), "sanity check");
-  return true;
+  return false;
 }
 
-
 // On Aix, reservations are made on a page by page basis, nothing to do.
 inline void os::pd_split_reserved_memory(char *base, size_t size,
                                          size_t split, bool realloc) {
 }
 
-
 // Bang the shadow pages if they need to be touched to be mapped.
 inline void os::bang_stack_shadow_pages() {
 }
@@ -75,15 +76,13 @@
 
 inline const int os::default_file_open_flags() { return 0;}
 
-inline DIR* os::opendir(const char* dirname)
-{
+inline DIR* os::opendir(const char* dirname) {
   assert(dirname != NULL, "just checking");
   return ::opendir(dirname);
 }
 
-inline int os::readdir_buf_size(const char *path)
-{
-  // according to aix sys/limits, NAME_MAX must be retrieved at runtime. */
+inline int os::readdir_buf_size(const char *path) {
+  // According to aix sys/limits, NAME_MAX must be retrieved at runtime.
   const long my_NAME_MAX = pathconf(path, _PC_NAME_MAX);
   return my_NAME_MAX + sizeof(dirent) + 1;
 }
@@ -104,8 +103,7 @@
   return ::ftruncate64(fd, length);
 }
 
-inline struct dirent* os::readdir(DIR* dirp, dirent *dbuf)
-{
+inline struct dirent* os::readdir(DIR* dirp, dirent *dbuf) {
   dirent* p;
   int status;
   assert(dirp != NULL, "just checking");
@@ -174,11 +172,11 @@
   RESTARTABLE_RETURN_INT(::send(fd, buf, nBytes, flags));
 }
 
-inline int os::raw_send(int fd, char* buf, size_t nBytes, uint flags) {
+inline int os::raw_send(int fd, char *buf, size_t nBytes, uint flags) {
   return os::send(fd, buf, nBytes, flags);
 }
 
-inline int os::connect(int fd, struct sockaddr* him, socklen_t len) {
+inline int os::connect(int fd, struct sockaddr *him, socklen_t len) {
   RESTARTABLE_RETURN_INT(::connect(fd, him, len));
 }
 
--- a/hotspot/src/os/aix/vm/perfMemory_aix.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/perfMemory_aix.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -797,7 +797,7 @@
   // Close the directory and reset the current working directory.
   close_directory_secure_cwd(dirp, saved_cwd_fd);
 
-  FREE_C_HEAP_ARRAY(char, dbuf, mtInternal);
+  FREE_C_HEAP_ARRAY(char, dbuf);
 }
 
 // Make the user specific temporary directory. Returns true if
@@ -1164,9 +1164,9 @@
   // store file, we don't follow them when attaching either.
   //
   if (!is_directory_secure(dirname)) {
-    FREE_C_HEAP_ARRAY(char, dirname, mtInternal);
+    FREE_C_HEAP_ARRAY(char, dirname);
     if (luser != user) {
-      FREE_C_HEAP_ARRAY(char, luser, mtInternal);
+      FREE_C_HEAP_ARRAY(char, luser);
     }
     THROW_MSG(vmSymbols::java_lang_IllegalArgumentException(),
               "Process not found");
--- a/hotspot/src/os/aix/vm/porting_aix.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os/aix/vm/porting_aix.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright 2012, 2013 SAP AG. All rights reserved.
+ * Copyright 2012, 2015 SAP AG. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -22,8 +22,18 @@
  *
  */
 
+#ifndef OS_AIX_VM_PORTING_AIX_HPP
+#define OS_AIX_VM_PORTING_AIX_HPP
+
 #include <stddef.h>
 
+// PPC port only:
+#define assert0(b) assert( (b), "" )
+#define guarantee0(b) assert( (b), "" )
+template <class T1, class T2> bool is_aligned_to(T1 what, T2 alignment) {
+  return  ( ((uintx)(what)) & (((uintx)(alignment)) - 1) ) == 0 ? true : false;
+}
+
 // Header file to contain porting-relevant code which does not have a
 // home anywhere else and which can not go into os_<platform>.h because
 // that header is included inside the os class definition, hence all
@@ -79,3 +89,62 @@
       const struct tbtable** p_tb,     // [out] optional: ptr to traceback table to get further information
       char* p_errmsg, size_t errmsglen // [out] optional: user provided buffer for error messages
     );
+
+// -------------------------------------------------------------------------
+
+// A simple critical section which shall be based upon OS critical
+// sections (CRITICAL_SECTION resp. Posix Mutex) and nothing else.
+
+#include <pthread.h>
+
+namespace MiscUtils {
+  typedef pthread_mutex_t critsect_t;
+
+  inline void init_critsect(MiscUtils::critsect_t* cs) {
+    pthread_mutex_init(cs, NULL);
+  }
+  inline void free_critsect(MiscUtils::critsect_t* cs) {
+    pthread_mutex_destroy(cs);
+  }
+  inline void enter_critsect(MiscUtils::critsect_t* cs) {
+    pthread_mutex_lock(cs);
+  }
+  inline void leave_critsect(MiscUtils::critsect_t* cs) {
+    pthread_mutex_unlock(cs);
+  }
+
+  // Need to wrap this in an object because we need to dynamically initialize
+  // critical section (because of windows, where there is no way to initialize
+  // a CRITICAL_SECTION statically. On Unix, we could use
+  // PTHREAD_MUTEX_INITIALIZER)
+
+  // Note: The critical section does NOT get cleaned up in the destructor. That is
+  // by design: the CritSect class is only ever used as global objects whose
+  // lifetime spans the whole VM life; in that context we don't want the lock to
+  // be cleaned up when global C++ objects are destroyed, but to continue to work
+  // correctly right to the very end of the process life.
+  class CritSect {
+    critsect_t _cs;
+  public:
+    CritSect()        { init_critsect(&_cs); }
+    //~CritSect()       { free_critsect(&_cs); }
+    void enter()      { enter_critsect(&_cs); }
+    void leave()      { leave_critsect(&_cs); }
+  };
+
+  class AutoCritSect {
+    CritSect* const _pcsobj;
+  public:
+    AutoCritSect(CritSect* pcsobj)
+      : _pcsobj(pcsobj)
+    {
+      _pcsobj->enter();
+    }
+    ~AutoCritSect() {
+      _pcsobj->leave();
+    }
+  };
+
+}
+
+#endif // OS_AIX_VM_PORTING_AIX_HPP
--- a/hotspot/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -129,7 +129,7 @@
     bool is_inconsistent()  { return _state == INCONSISTENT; }
     void set_inconsistent() { _state = INCONSISTENT;         }
 
-    void visit(picl_nodehdl_t nodeh, const char* name) {
+    bool visit(picl_nodehdl_t nodeh, const char* name) {
       assert(!is_inconsistent(), "Precondition");
       int curr;
       if (_picl->get_int_property(nodeh, name, &curr) == PICL_SUCCESS) {
@@ -138,7 +138,9 @@
         } else if (curr != value()) { // following iterations
           set_inconsistent();
         }
+        return true;
       }
+      return false;
     }
   };
 
@@ -155,8 +157,19 @@
       if (!l1_visitor->is_inconsistent()) {
         l1_visitor->visit(nodeh, "l1-dcache-line-size");
       }
-      if (!l2_visitor->is_inconsistent()) {
-        l2_visitor->visit(nodeh, "l2-cache-line-size");
+      static const char* l2_data_cache_line_property_name = NULL;
+      // On the first visit determine the name of the l2 cache line size property and memoize it.
+      if (l2_data_cache_line_property_name == NULL) {
+        assert(!l2_visitor->is_inconsistent(), "First iteration cannot be inconsistent");
+        l2_data_cache_line_property_name = "l2-cache-line-size";
+        if (!l2_visitor->visit(nodeh, l2_data_cache_line_property_name)) {
+          l2_data_cache_line_property_name = "l2-dcache-line-size";
+          l2_visitor->visit(nodeh, l2_data_cache_line_property_name);
+        }
+      } else {
+        if (!l2_visitor->is_inconsistent()) {
+          l2_visitor->visit(nodeh, l2_data_cache_line_property_name);
+        }
       }
 
       if (l1_visitor->is_inconsistent() && l2_visitor->is_inconsistent()) {
@@ -172,13 +185,13 @@
     UniqueValueVisitor* l2_visitor() { return &_l2_visitor; }
   };
   int _L1_data_cache_line_size;
-  int _L2_cache_line_size;
+  int _L2_data_cache_line_size;
 public:
   static int visit_cpu(picl_nodehdl_t nodeh, void *state) {
     return CPUVisitor::visit(nodeh, state);
   }
 
-  PICL(bool is_fujitsu) : _L1_data_cache_line_size(0), _L2_cache_line_size(0), _dl_handle(NULL) {
+  PICL(bool is_fujitsu) : _L1_data_cache_line_size(0), _L2_data_cache_line_size(0), _dl_handle(NULL) {
     if (!open_library()) {
       return;
     }
@@ -196,7 +209,7 @@
           _L1_data_cache_line_size = cpu_visitor.l1_visitor()->value();
         }
         if (cpu_visitor.l2_visitor()->is_assigned()) {
-          _L2_cache_line_size = cpu_visitor.l2_visitor()->value();
+          _L2_data_cache_line_size = cpu_visitor.l2_visitor()->value();
         }
       }
       _picl_shutdown();
@@ -205,7 +218,7 @@
   }
 
   unsigned int L1_data_cache_line_size() const { return _L1_data_cache_line_size; }
-  unsigned int L2_cache_line_size() const      { return _L2_cache_line_size;      }
+  unsigned int L2_data_cache_line_size() const { return _L2_data_cache_line_size; }
 };
 
 
@@ -431,7 +444,7 @@
   // Figure out cache line sizes using PICL
   PICL picl((features & sparc64_family_m) != 0);
   _L1_data_cache_line_size = picl.L1_data_cache_line_size();
-  _L2_cache_line_size      = picl.L2_cache_line_size();
+  _L2_data_cache_line_size = picl.L2_data_cache_line_size();
 
   return features;
 }
--- a/hotspot/src/share/vm/adlc/adlparse.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/adlparse.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -800,6 +800,7 @@
       }
       if (strcmp(token,"reg_def")==0)          { reg_def_parse(); }
       else if (strcmp(token,"reg_class")==0)   { reg_class_parse(); }
+      else if (strcmp(token, "reg_class_dynamic") == 0) { reg_class_dynamic_parse(); }
       else if (strcmp(token,"alloc_class")==0) { alloc_class_parse(); }
       else if (strcmp(token,"#define")==0)     { preproc_define(); }
       else { parse_err(SYNERR, "bad token %s inside register block.\n", token); break; }
@@ -2323,11 +2324,12 @@
   // Debug Stuff
   if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
 
-  RegClass *reg_class = _AD._register->addRegClass(cname);
-
-  // Collect registers in class
   skipws();
   if (_curchar == '(') {
+    // A register list is defined for the register class.
+    // Collect registers into a generic RegClass register class.
+    RegClass* reg_class = _AD._register->addRegClass<RegClass>(cname);
+
     next_char();                  // Skip '('
     skipws();
     while (_curchar != ')') {
@@ -2352,12 +2354,15 @@
     }
     next_char();                  // Skip closing ')'
   } else if (_curchar == '%') {
+    // A code snippet is defined for the register class.
+    // Collect the code snippet into a CodeSnippetRegClass register class.
+    CodeSnippetRegClass* reg_class = _AD._register->addRegClass<CodeSnippetRegClass>(cname);
     char *code = find_cpp_block("reg class");
     if (code == NULL) {
       parse_err(SYNERR, "missing code declaration for reg class.\n");
       return;
     }
-    reg_class->_user_defined = code;
+    reg_class->set_code_snippet(code);
     return;
   }
 
@@ -2374,6 +2379,87 @@
   return;
 }
 
+//------------------------------reg_class_dynamic_parse------------------------
+void ADLParser::reg_class_dynamic_parse(void) {
+  char *cname; // Name of dynamic register class being defined
+
+  // Get register class name
+  skipws();
+  cname = get_ident();
+  if (cname == NULL) {
+    parse_err(SYNERR, "missing dynamic register class name after 'reg_class_dynamic'\n");
+    return;
+  }
+
+  if (_AD._adl_debug > 1) {
+    fprintf(stdout, "Dynamic Register Class: %s\n", cname);
+  }
+
+  skipws();
+  if (_curchar != '(') {
+    parse_err(SYNERR, "missing '(' at the beginning of reg_class_dynamic definition\n");
+    return;
+  }
+  next_char();
+  skipws();
+
+  // Collect two register classes and the C++ code representing the condition code used to
+  // select between the two classes into a ConditionalRegClass register class.
+  ConditionalRegClass* reg_class = _AD._register->addRegClass<ConditionalRegClass>(cname);
+  int i;
+  for (i = 0; i < 2; i++) {
+    char* name = get_ident();
+    if (name == NULL) {
+      parse_err(SYNERR, "missing class identifier inside reg_class_dynamic list.\n");
+      return;
+    }
+    RegClass* rc = _AD._register->getRegClass(name);
+    if (rc == NULL) {
+      parse_err(SEMERR, "unknown identifier %s inside reg_class_dynamic list.\n", name);
+    } else {
+      reg_class->set_rclass_at_index(i, rc);
+    }
+
+    skipws();
+    if (_curchar == ',') {
+      next_char();
+      skipws();
+    } else {
+      parse_err(SYNERR, "missing separator ',' inside reg_class_dynamic list.\n");
+    }
+  }
+
+  // Collect the condition code.
+  skipws();
+  if (_curchar == '%') {
+    char* code = find_cpp_block("reg class dynamic");
+    if (code == NULL) {
+       parse_err(SYNERR, "missing code declaration for reg_class_dynamic.\n");
+       return;
+    }
+    reg_class->set_condition_code(code);
+  } else {
+    parse_err(SYNERR, "missing %% at the beginning of code block in reg_class_dynamic definition\n");
+    return;
+  }
+
+  skipws();
+  if (_curchar != ')') {
+    parse_err(SYNERR, "missing ')' at the end of reg_class_dynamic definition\n");
+    return;
+  }
+  next_char();
+
+  skipws();
+  if (_curchar != ';') {
+    parse_err(SYNERR, "missing ';' at the end of reg_class_dynamic definition.\n");
+    return;
+  }
+  next_char();                    // Skip trailing ';'
+
+  return;
+}
+
 //------------------------------alloc_class_parse------------------------------
 void ADLParser::alloc_class_parse(void) {
   char *name;                     // Name of allocation class being defined
--- a/hotspot/src/share/vm/adlc/adlparse.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/adlparse.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -53,6 +53,8 @@
 // ***** Register Section *****
 class RegDef;
 class RegClass;
+class CodeSnippetRegClass;
+class ConditionalRegClass;
 class AllocClass;
 class ResourceForm;
 // ***** Pipeline Section *****
@@ -125,6 +127,7 @@
   // Parse components of the register section
   void reg_def_parse(void);              // Parse register definition
   void reg_class_parse(void);            // Parse register class definition
+  void reg_class_dynamic_parse(void);    // Parse dynamic register class definition
   void alloc_class_parse(void);          // Parse allocation class definition
 
   // Parse components of the definition section
--- a/hotspot/src/share/vm/adlc/archDesc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/archDesc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -908,7 +908,7 @@
 void ArchDesc::set_stack_or_reg(const char *reg_class_name) {
   if( _register ) {
     RegClass *reg_class  = _register->getRegClass(reg_class_name);
-    reg_class->_stack_or_reg = true;
+    reg_class->set_stack_version(true);
   }
 }
 
--- a/hotspot/src/share/vm/adlc/forms.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/forms.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -68,6 +68,8 @@
 class InsEncode;
 class RegDef;
 class RegClass;
+class CodeSnippetRegClass;
+class ConditionalRegClass;
 class AllocClass;
 class ResourceForm;
 class PipeClassForm;
--- a/hotspot/src/share/vm/adlc/formsopt.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/formsopt.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -47,13 +47,19 @@
 }
 
 // record a new register class
-RegClass *RegisterForm::addRegClass(const char *className) {
-  RegClass *regClass = new RegClass(className);
+template <typename T>
+T* RegisterForm::addRegClass(const char* className) {
+  T* regClass = new T(className);
   _rclasses.addName(className);
-  _regClass.Insert(className,regClass);
+  _regClass.Insert(className, regClass);
   return regClass;
 }
 
+// Explicit instantiation for all supported register classes.
+template RegClass* RegisterForm::addRegClass<RegClass>(const char* className);
+template CodeSnippetRegClass* RegisterForm::addRegClass<CodeSnippetRegClass>(const char* className);
+template ConditionalRegClass* RegisterForm::addRegClass<ConditionalRegClass>(const char* className);
+
 // record a new register class
 AllocClass *RegisterForm::addAllocClass(char *className) {
   AllocClass *allocClass = new AllocClass(className);
@@ -67,9 +73,9 @@
 void RegisterForm::addSpillRegClass() {
   // Stack slots start at the next available even register number.
   _reg_ctr = (_reg_ctr+7) & ~7;
-  const char *rc_name   = "stack_slots";
-  RegClass   *reg_class = new RegClass(rc_name);
-  reg_class->_stack_or_reg = true;
+  const char *rc_name = "stack_slots";
+  RegClass* reg_class = new RegClass(rc_name);
+  reg_class->set_stack_version(true);
   _rclasses.addName(rc_name);
   _regClass.Insert(rc_name,reg_class);
 }
@@ -224,9 +230,11 @@
 
 //------------------------------RegClass---------------------------------------
 // Construct a register class into which registers will be inserted
-RegClass::RegClass(const char *classid) : _stack_or_reg(false), _classid(classid), _regDef(cmpstr,hashstr, Form::arena),
-                                          _user_defined(NULL)
-{
+RegClass::RegClass(const char* classid) : _stack_or_reg(false), _classid(classid), _regDef(cmpstr, hashstr, Form::arena) {
+}
+
+RegClass::~RegClass() {
+  delete _classid;
 }
 
 // record a register in this class
@@ -305,6 +313,91 @@
   fprintf(fp,"--- done with entries for reg_class %s\n\n",_classid);
 }
 
+void RegClass::declare_register_masks(FILE* fp) {
+  const char* prefix = "";
+  const char* rc_name_to_upper = toUpper(_classid);
+  fprintf(fp, "extern const RegMask _%s%s_mask;\n", prefix,  rc_name_to_upper);
+  fprintf(fp, "inline const RegMask &%s%s_mask() { return _%s%s_mask; }\n", prefix, rc_name_to_upper, prefix, rc_name_to_upper);
+  if (_stack_or_reg) {
+    fprintf(fp, "extern const RegMask _%sSTACK_OR_%s_mask;\n", prefix, rc_name_to_upper);
+    fprintf(fp, "inline const RegMask &%sSTACK_OR_%s_mask() { return _%sSTACK_OR_%s_mask; }\n", prefix, rc_name_to_upper, prefix, rc_name_to_upper);
+  }
+  delete[] rc_name_to_upper;
+}
+
+void RegClass::build_register_masks(FILE* fp) {
+  int len = RegisterForm::RegMask_Size();
+  const char *prefix = "";
+  const char* rc_name_to_upper = toUpper(_classid);
+  fprintf(fp, "const RegMask _%s%s_mask(", prefix, rc_name_to_upper);
+
+  int i;
+  for(i = 0; i < len - 1; i++) {
+    fprintf(fp," 0x%x,", regs_in_word(i, false));
+  }
+  fprintf(fp," 0x%x );\n", regs_in_word(i, false));
+
+  if (_stack_or_reg) {
+    fprintf(fp, "const RegMask _%sSTACK_OR_%s_mask(", prefix, rc_name_to_upper);
+    for(i = 0; i < len - 1; i++) {
+      fprintf(fp," 0x%x,", regs_in_word(i, true));
+    }
+    fprintf(fp," 0x%x );\n", regs_in_word(i, true));
+  }
+  delete[] rc_name_to_upper;
+}
+
+//------------------------------CodeSnippetRegClass---------------------------
+CodeSnippetRegClass::CodeSnippetRegClass(const char* classid) : RegClass(classid), _code_snippet(NULL) {
+}
+
+CodeSnippetRegClass::~CodeSnippetRegClass() {
+  delete _code_snippet;
+}
+
+void CodeSnippetRegClass::declare_register_masks(FILE* fp) {
+  const char* prefix = "";
+  const char* rc_name_to_upper = toUpper(_classid);
+  fprintf(fp, "inline const RegMask &%s%s_mask() { %s }\n", prefix, rc_name_to_upper, _code_snippet);
+  delete[] rc_name_to_upper;
+}
+
+//------------------------------ConditionalRegClass---------------------------
+ConditionalRegClass::ConditionalRegClass(const char *classid) : RegClass(classid), _condition_code(NULL) {
+}
+
+ConditionalRegClass::~ConditionalRegClass() {
+  delete _condition_code;
+}
+
+void ConditionalRegClass::declare_register_masks(FILE* fp) {
+  const char* prefix = "";
+  const char* rc_name_to_upper = toUpper(_classid);
+  const char* rclass_0_to_upper = toUpper(_rclasses[0]->_classid);
+  const char* rclass_1_to_upper = toUpper(_rclasses[1]->_classid);
+  fprintf(fp, "inline const RegMask &%s%s_mask() {"
+              " return (%s) ?"
+              " %s%s_mask() :"
+              " %s%s_mask(); }\n",
+              prefix, rc_name_to_upper,
+              _condition_code,
+              prefix, rclass_0_to_upper,
+              prefix, rclass_1_to_upper);
+  if (_stack_or_reg) {
+    fprintf(fp, "inline const RegMask &%sSTACK_OR_%s_mask() {"
+                  " return (%s) ?"
+                  " %sSTACK_OR_%s_mask() :"
+                  " %sSTACK_OR_%s_mask(); }\n",
+                  prefix, rc_name_to_upper,
+                  _condition_code,
+                  prefix, rclass_0_to_upper,
+                  prefix, rclass_1_to_upper);
+  }
+  delete[] rc_name_to_upper;
+  delete[] rclass_0_to_upper;
+  delete[] rclass_1_to_upper;
+  return;
+}
 
 //------------------------------AllocClass-------------------------------------
 AllocClass::AllocClass(char *classid) : _classid(classid), _regDef(cmpstr,hashstr, Form::arena) {
--- a/hotspot/src/share/vm/adlc/formsopt.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/formsopt.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -60,6 +60,8 @@
 class InsEncode;
 class RegDef;
 class RegClass;
+class CodeSnippetRegClass;
+class ConditionalRegClass;
 class AllocClass;
 class ResourceForm;
 class PipeClassForm;
@@ -98,7 +100,8 @@
 
   void        addRegDef(char *regName, char *callingConv, char *c_conv,
                         char * idealtype, char *encoding, char* concreteName);
-  RegClass   *addRegClass(const char *className);
+  template<typename T> T* addRegClass(const char* className);
+
   AllocClass *addAllocClass(char *allocName);
   void        addSpillRegClass();
 
@@ -154,17 +157,28 @@
 };
 
 //------------------------------RegClass---------------------------------------
+// Generic register class. This register class is the internal representation
+// for the following .ad file format:
+//
+//  reg_class ptr(RAX, RBX, ...);
+//
+// where ptr is the name of the register class, RAX and RBX are registers.
+//
+// This register class allows registers to be spilled onto the stack. Spilling
+// is allowed is field _stack_or_reg is true.
 class RegClass : public Form {
 public:
   // Public Data
   const char *_classid;         // Name of class
   NameList    _regDefs;         // List of registers in class
   Dict        _regDef;          // Dictionary of registers in class
+protected:
   bool        _stack_or_reg;    // Allowed on any stack slot
-  char*       _user_defined;
 
+public:
   // Public Methods
   RegClass(const char *classid);// Constructor
+  virtual ~RegClass();
 
   void addReg(RegDef *regDef);  // Add a register to this class
 
@@ -183,6 +197,115 @@
 
   void dump();                  // Debug printer
   void output(FILE *fp);        // Write info to output files
+
+  virtual bool has_stack_version() {
+    return _stack_or_reg;
+  }
+  virtual void set_stack_version(bool flag) {
+    _stack_or_reg = flag;
+  }
+
+  virtual void declare_register_masks(FILE* fp);
+  virtual void build_register_masks(FILE* fp);
+};
+
+//------------------------------CodeSnippetRegClass----------------------------
+// Register class that has an user-defined C++ code snippet attached to it
+// to determine at runtime which register class to use. This register class is
+// the internal representation for the following .ad file format:
+//
+//  reg_class actual_dflt_reg %{
+//      if (VM_Version::has_vfp3_32()) {
+//          return DFLT_REG_mask();
+//      } else {
+//          return DFLT_LOW_REG_mask();
+//      }
+//  %}
+//
+// where DFLT_REG_mask() and DFLT_LOW_REG_mask() are the internal names of the
+// masks of register classes dflt_reg and dflt_low_reg.
+//
+// The attached code snippet can select also between more than two register classes.
+// This register class can be, however, used only if the register class is not
+// cisc-spillable (i.e., the registers of this class are not allowed on the stack,
+// which is equivalent with _stack_or_reg being false).
+class CodeSnippetRegClass : public RegClass {
+protected:
+  char* _code_snippet;
+public:
+  CodeSnippetRegClass(const char* classid);// Constructor
+  ~CodeSnippetRegClass();
+
+  void set_code_snippet(char* code) {
+    _code_snippet = code;
+  }
+  char* code_snippet() {
+    return _code_snippet;
+  }
+  void set_stack_version(bool flag) {
+    assert(false, "User defined register classes are not allowed to spill to the stack.");
+  }
+  void declare_register_masks(FILE* fp);
+  void build_register_masks(FILE* fp) {
+    // We do not need to generate register masks because we select at runtime
+    // between register masks generated for other register classes.
+    return;
+  }
+};
+
+//------------------------------ConditionalRegClass----------------------------
+// Register class that has two register classes and a runtime condition attached
+// to it. The condition is evaluated at runtime and either one of the register
+// attached register classes is selected. This register class is the internal
+// representation for the following .ad format:
+//
+//  reg_class_dynamic actual_dflt_reg(dflt_reg, low_reg,
+//                                    %{ VM_Version::has_vfp3_32() }%
+//                                    );
+//
+// This example is equivalent to the example used with the CodeSnippetRegClass
+// register class. A ConditionalRegClass works also if a register class is cisc-spillable
+// (i.e., _stack_or_reg is true), but if can select only between two register classes.
+class ConditionalRegClass : public RegClass {
+protected:
+  // reference to condition code
+  char* _condition_code;  // C++ condition code to dynamically determine which register class to use.
+
+                          // Example syntax (equivalent to previous example):
+                          //
+                          //  reg_class actual_dflt_reg(dflt_reg, low_reg,
+                          //                            %{ VM_Version::has_vfp3_32() }%
+                          //                            );
+  // reference to conditional register classes
+  RegClass* _rclasses[2]; // 0 is the register class selected if the condition code returns true
+                          // 1 is the register class selected if the condition code returns false
+public:
+  ConditionalRegClass(const char* classid);// Constructor
+  ~ConditionalRegClass();
+
+  virtual void set_stack_version(bool flag) {
+    RegClass::set_stack_version(flag);
+    assert((_rclasses[0] != NULL), "Register class NULL for condition code == true");
+    assert((_rclasses[1] != NULL), "Register class NULL for condition code == false");
+    _rclasses[0]->set_stack_version(flag);
+    _rclasses[1]->set_stack_version(flag);
+  }
+  void declare_register_masks(FILE* fp);
+  void build_register_masks(FILE* fp) {
+    // We do not need to generate register masks because we select at runtime
+    // between register masks generated for other register classes.
+    return;
+  }
+  void set_rclass_at_index(int index, RegClass* rclass) {
+    assert((0 <= index && index < 2), "Condition code can select only between two register classes");
+    _rclasses[index] = rclass;
+  }
+  void set_condition_code(char* code) {
+    _condition_code = code;
+  }
+  char* condition_code() {
+    return _condition_code;
+  }
 };
 
 //------------------------------AllocClass-------------------------------------
--- a/hotspot/src/share/vm/adlc/formssel.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/formssel.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -4043,6 +4043,13 @@
         strcmp(opType,"ReplicateL")==0 ||
         strcmp(opType,"ReplicateF")==0 ||
         strcmp(opType,"ReplicateD")==0 ||
+        strcmp(opType,"AddReductionVI")==0 ||
+        strcmp(opType,"AddReductionVL")==0 ||
+        strcmp(opType,"AddReductionVF")==0 ||
+        strcmp(opType,"AddReductionVD")==0 ||
+        strcmp(opType,"MulReductionVI")==0 ||
+        strcmp(opType,"MulReductionVF")==0 ||
+        strcmp(opType,"MulReductionVD")==0 ||
         0 /* 0 to line up columns nicely */ )
       return 1;
   }
@@ -4135,6 +4142,10 @@
     "MulVS","MulVI","MulVF","MulVD",
     "DivVF","DivVD",
     "AndV" ,"XorV" ,"OrV",
+    "AddReductionVI", "AddReductionVL",
+    "AddReductionVF", "AddReductionVD",
+    "MulReductionVI",
+    "MulReductionVF", "MulReductionVD",
     "LShiftCntV","RShiftCntV",
     "LShiftVB","LShiftVS","LShiftVI","LShiftVL",
     "RShiftVB","RShiftVS","RShiftVI","RShiftVL",
--- a/hotspot/src/share/vm/adlc/formssel.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/formssel.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -59,6 +59,8 @@
 class InsEncode;
 class RegDef;
 class RegClass;
+class CodeSnippetRegClass;
+class ConditionalRegClass;
 class AllocClass;
 class ResourceForm;
 class PipeDesc;
--- a/hotspot/src/share/vm/adlc/output_c.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/adlc/output_c.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -138,26 +138,9 @@
     fprintf(fp_hpp,"// Register masks, one for each register class.\n");
     _register->_rclasses.reset();
     for (rc_name = NULL; (rc_name = _register->_rclasses.iter()) != NULL;) {
-      const char *prefix = "";
       RegClass *reg_class = _register->getRegClass(rc_name);
       assert(reg_class, "Using an undefined register class");
-
-      const char* rc_name_to_upper = toUpper(rc_name);
-
-      if (reg_class->_user_defined == NULL) {
-        fprintf(fp_hpp, "extern const RegMask _%s%s_mask;\n", prefix,  rc_name_to_upper);
-        fprintf(fp_hpp, "inline const RegMask &%s%s_mask() { return _%s%s_mask; }\n", prefix, rc_name_to_upper, prefix, rc_name_to_upper);
-      } else {
-        fprintf(fp_hpp, "inline const RegMask &%s%s_mask() { %s }\n", prefix, rc_name_to_upper, reg_class->_user_defined);
-      }
-
-      if (reg_class->_stack_or_reg) {
-        assert(reg_class->_user_defined == NULL, "no user defined reg class here");
-        fprintf(fp_hpp, "extern const RegMask _%sSTACK_OR_%s_mask;\n", prefix, rc_name_to_upper);
-        fprintf(fp_hpp, "inline const RegMask &%sSTACK_OR_%s_mask() { return _%sSTACK_OR_%s_mask; }\n", prefix, rc_name_to_upper, prefix, rc_name_to_upper);
-      }
-      delete[] rc_name_to_upper;
-
+      reg_class->declare_register_masks(fp_hpp);
     }
   }
 }
@@ -173,35 +156,9 @@
     fprintf(fp_cpp,"// Register masks, one for each register class.\n");
     _register->_rclasses.reset();
     for (rc_name = NULL; (rc_name = _register->_rclasses.iter()) != NULL;) {
-      const char *prefix = "";
       RegClass *reg_class = _register->getRegClass(rc_name);
       assert(reg_class, "Using an undefined register class");
-
-      if (reg_class->_user_defined != NULL) {
-        continue;
-      }
-
-      int len = RegisterForm::RegMask_Size();
-      const char* rc_name_to_upper = toUpper(rc_name);
-      fprintf(fp_cpp, "const RegMask _%s%s_mask(", prefix, rc_name_to_upper);
-
-      {
-        int i;
-        for(i = 0; i < len - 1; i++) {
-          fprintf(fp_cpp," 0x%x,", reg_class->regs_in_word(i, false));
-        }
-        fprintf(fp_cpp," 0x%x );\n", reg_class->regs_in_word(i, false));
-      }
-
-      if (reg_class->_stack_or_reg) {
-        int i;
-        fprintf(fp_cpp, "const RegMask _%sSTACK_OR_%s_mask(", prefix, rc_name_to_upper);
-        for(i = 0; i < len - 1; i++) {
-          fprintf(fp_cpp," 0x%x,",reg_class->regs_in_word(i, true));
-        }
-        fprintf(fp_cpp," 0x%x );\n",reg_class->regs_in_word(i, true));
-      }
-      delete[] rc_name_to_upper;
+      reg_class->build_register_masks(fp_cpp);
     }
   }
 }
--- a/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/c1/c1_GraphBuilder.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -3462,6 +3462,24 @@
     case vmIntrinsics::_putFloat  : return append_unsafe_put_obj(callee, T_FLOAT,   false);
     case vmIntrinsics::_putDouble : return append_unsafe_put_obj(callee, T_DOUBLE,  false);
 
+    case vmIntrinsics::_getShortUnaligned  :
+      return UseUnalignedAccesses ? append_unsafe_get_obj(callee, T_SHORT,   false) : false;
+    case vmIntrinsics::_getCharUnaligned   :
+      return UseUnalignedAccesses ? append_unsafe_get_obj(callee, T_CHAR,    false) : false;
+    case vmIntrinsics::_getIntUnaligned    :
+      return UseUnalignedAccesses ? append_unsafe_get_obj(callee, T_INT,     false) : false;
+    case vmIntrinsics::_getLongUnaligned   :
+      return UseUnalignedAccesses ? append_unsafe_get_obj(callee, T_LONG,    false) : false;
+
+    case vmIntrinsics::_putShortUnaligned  :
+      return UseUnalignedAccesses ? append_unsafe_put_obj(callee, T_SHORT,   false) : false;
+    case vmIntrinsics::_putCharUnaligned   :
+      return UseUnalignedAccesses ? append_unsafe_put_obj(callee, T_CHAR,    false) : false;
+    case vmIntrinsics::_putIntUnaligned    :
+      return UseUnalignedAccesses ? append_unsafe_put_obj(callee, T_INT,     false) : false;
+    case vmIntrinsics::_putLongUnaligned   :
+      return UseUnalignedAccesses ? append_unsafe_put_obj(callee, T_LONG,    false) : false;
+
     case vmIntrinsics::_getObjectVolatile : return append_unsafe_get_obj(callee, T_OBJECT,  true);
     case vmIntrinsics::_getBooleanVolatile: return append_unsafe_get_obj(callee, T_BOOLEAN, true);
     case vmIntrinsics::_getByteVolatile   : return append_unsafe_get_obj(callee, T_BYTE,    true);
--- a/hotspot/src/share/vm/ci/ciMethod.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/ci/ciMethod.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -688,7 +688,8 @@
 // via assert_unique_concrete_method or assert_leaf_type.
 ciMethod* ciMethod::find_monomorphic_target(ciInstanceKlass* caller,
                                             ciInstanceKlass* callee_holder,
-                                            ciInstanceKlass* actual_recv) {
+                                            ciInstanceKlass* actual_recv,
+                                            bool check_access) {
   check_is_loaded();
 
   if (actual_recv->is_interface()) {
@@ -696,7 +697,7 @@
     return NULL;
   }
 
-  ciMethod* root_m = resolve_invoke(caller, actual_recv);
+  ciMethod* root_m = resolve_invoke(caller, actual_recv, check_access);
   if (root_m == NULL) {
     // Something went wrong looking up the actual receiver method.
     return NULL;
@@ -775,7 +776,7 @@
 //
 // Given a known receiver klass, find the target for the call.
 // Return NULL if the call has no target or the target is abstract.
-ciMethod* ciMethod::resolve_invoke(ciKlass* caller, ciKlass* exact_receiver) {
+ciMethod* ciMethod::resolve_invoke(ciKlass* caller, ciKlass* exact_receiver, bool check_access) {
    check_is_loaded();
    VM_ENTRY_MARK;
 
@@ -792,9 +793,9 @@
         ||
        InstanceKlass::cast(h_recv())->is_linked() && !exact_receiver->is_interface()) {
      if (holder()->is_interface()) {
-       m = LinkResolver::resolve_interface_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass);
+       m = LinkResolver::resolve_interface_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass, check_access);
      } else {
-       m = LinkResolver::resolve_virtual_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass);
+       m = LinkResolver::resolve_virtual_call_or_null(h_recv, h_resolved, h_name, h_signature, caller_klass, check_access);
      }
    }
 
--- a/hotspot/src/share/vm/ci/ciMethod.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/ci/ciMethod.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -255,11 +255,12 @@
   // its calling environment.
   ciMethod* find_monomorphic_target(ciInstanceKlass* caller,
                                     ciInstanceKlass* callee_holder,
-                                    ciInstanceKlass* actual_receiver);
+                                    ciInstanceKlass* actual_receiver,
+                                    bool check_access = true);
 
   // Given a known receiver klass, find the target for the call.
   // Return NULL if the call has no target or is abstract.
-  ciMethod* resolve_invoke(ciKlass* caller, ciKlass* exact_receiver);
+  ciMethod* resolve_invoke(ciKlass* caller, ciKlass* exact_receiver, bool check_access = true);
 
   // Find the proper vtable index to invoke this method.
   int resolve_vtable_index(ciKlass* caller, ciKlass* receiver);
--- a/hotspot/src/share/vm/classfile/vmSymbols.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/classfile/vmSymbols.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -868,9 +868,12 @@
                                                                                                                         \
   /* Custom branch frequencies profiling support for JSR292 */                                                          \
   do_class(java_lang_invoke_MethodHandleImpl,               "java/lang/invoke/MethodHandleImpl")                        \
-  do_intrinsic(_profileBoolean, java_lang_invoke_MethodHandleImpl, profileBoolean_name, profileBoolean_signature,    F_S)  \
-   do_name(     profileBoolean_name,                               "profileBoolean")                                     \
-   do_signature(profileBoolean_signature,                           "(Z[I)Z")                                            \
+  do_intrinsic(_profileBoolean, java_lang_invoke_MethodHandleImpl, profileBoolean_name, profileBoolean_signature, F_S)  \
+   do_name(     profileBoolean_name,                             "profileBoolean")                                      \
+   do_signature(profileBoolean_signature,                        "(Z[I)Z")                                              \
+  do_intrinsic(_isCompileConstant, java_lang_invoke_MethodHandleImpl, isCompileConstant_name, isCompileConstant_signature, F_S) \
+   do_name(     isCompileConstant_name,                          "isCompileConstant")                                   \
+   do_alias(    isCompileConstant_signature,                      object_boolean_signature)                             \
                                                                                                                         \
   /* unsafe memory references (there are a lot of them...) */                                                           \
   do_signature(getObject_signature,       "(Ljava/lang/Object;J)Ljava/lang/Object;")                                    \
@@ -950,6 +953,20 @@
   do_intrinsic(_putFloatVolatile,         sun_misc_Unsafe,        putFloatVolatile_name, putFloat_signature,     F_RN)  \
   do_intrinsic(_putDoubleVolatile,        sun_misc_Unsafe,        putDoubleVolatile_name, putDouble_signature,   F_RN)  \
                                                                                                                         \
+  do_name(getShortUnaligned_name,"getShortUnaligned")     do_name(putShortUnaligned_name,"putShortUnaligned")           \
+  do_name(getCharUnaligned_name,"getCharUnaligned")       do_name(putCharUnaligned_name,"putCharUnaligned")             \
+  do_name(getIntUnaligned_name,"getIntUnaligned")         do_name(putIntUnaligned_name,"putIntUnaligned")               \
+  do_name(getLongUnaligned_name,"getLongUnaligned")       do_name(putLongUnaligned_name,"putLongUnaligned")             \
+                                                                                                                        \
+  do_intrinsic(_getShortUnaligned,         sun_misc_Unsafe,        getShortUnaligned_name, getShort_signature,     F_R)  \
+  do_intrinsic(_getCharUnaligned,          sun_misc_Unsafe,        getCharUnaligned_name, getChar_signature,       F_R)  \
+  do_intrinsic(_getIntUnaligned,           sun_misc_Unsafe,        getIntUnaligned_name, getInt_signature,         F_R)  \
+  do_intrinsic(_getLongUnaligned,          sun_misc_Unsafe,        getLongUnaligned_name, getLong_signature,       F_R)  \
+  do_intrinsic(_putShortUnaligned,         sun_misc_Unsafe,        putShortUnaligned_name, putShort_signature,     F_R)  \
+  do_intrinsic(_putCharUnaligned,          sun_misc_Unsafe,        putCharUnaligned_name, putChar_signature,       F_R)  \
+  do_intrinsic(_putIntUnaligned,           sun_misc_Unsafe,        putIntUnaligned_name, putInt_signature,         F_R)  \
+  do_intrinsic(_putLongUnaligned,          sun_misc_Unsafe,        putLongUnaligned_name, putLong_signature,       F_R)  \
+                                                                                                                        \
   /* %%% these are redundant except perhaps for getAddress, but Unsafe has native methods for them */                   \
   do_signature(getByte_raw_signature,     "(J)B")                                                                       \
   do_signature(putByte_raw_signature,     "(JB)V")                                                                      \
--- a/hotspot/src/share/vm/code/dependencies.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/code/dependencies.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -845,7 +845,13 @@
     assert((uint)n <= (uint)_num_participants, "oob");
     Method* fm = _found_methods[n];
     assert(n == _num_participants || fm != NULL, "proper usage");
-    assert(fm == NULL || fm->method_holder() == _participants[n], "sanity");
+    if (fm != NULL && fm->method_holder() != _participants[n]) {
+      // Default methods from interfaces can be added to classes. In
+      // that case the holder of the method is not the class but the
+      // interface where it's defined.
+      assert(fm->is_default_method(), "sanity");
+      return NULL;
+    }
     return fm;
   }
 
--- a/hotspot/src/share/vm/code/nmethod.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/code/nmethod.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -504,7 +504,7 @@
                                             basic_lock_owner_sp_offset,
                                             basic_lock_sp_offset, oop_maps);
     NOT_PRODUCT(if (nm != NULL)  nmethod_stats.note_native_nmethod(nm));
-    if (PrintAssembly && nm != NULL) {
+    if ((PrintAssembly || CompilerOracle::should_print(method)) && nm != NULL) {
       Disassembler::decode(nm);
     }
   }
@@ -2837,11 +2837,21 @@
           st.print(")");
           return st.as_string();
         }
+        case relocInfo::runtime_call_type: {
+          stringStream st;
+          st.print("runtime_call");
+          runtime_call_Relocation* r = iter.runtime_call_reloc();
+          address dest = r->destination();
+          CodeBlob* cb = CodeCache::find_blob(dest);
+          if (cb != NULL) {
+            st.print(" %s", cb->name());
+          }
+          return st.as_string();
+        }
         case relocInfo::virtual_call_type:     return "virtual_call";
         case relocInfo::opt_virtual_call_type: return "optimized virtual_call";
         case relocInfo::static_call_type:      return "static_call";
         case relocInfo::static_stub_type:      return "static_stub";
-        case relocInfo::runtime_call_type:     return "runtime_call";
         case relocInfo::external_word_type:    return "external_word";
         case relocInfo::internal_word_type:    return "internal_word";
         case relocInfo::section_word_type:     return "section_word";
--- a/hotspot/src/share/vm/code/pcDesc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/code/pcDesc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -54,12 +54,7 @@
   for (ScopeDesc* sd = code->scope_desc_at(real_pc(code));
        sd != NULL;
        sd = sd->sender()) {
-    tty->print("  ");
-    sd->method()->print_short_name(tty);
-    tty->print("  @%d", sd->bci());
-    if (sd->should_reexecute())
-      tty->print("  reexecute=true");
-    tty->cr();
+    sd->print_on(tty);
   }
 #endif
 }
--- a/hotspot/src/share/vm/code/scopeDesc.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/code/scopeDesc.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -157,14 +157,18 @@
 #ifndef PRODUCT
 
 void ScopeDesc::print_value_on(outputStream* st) const {
-  tty->print("   ");
+  st->print("  ");
   method()->print_short_name(st);
   int lineno = method()->line_number_from_bci(bci());
   if (lineno != -1) {
-    st->print_cr("@%d (line %d)", bci(), lineno);
+    st->print("@%d (line %d)", bci(), lineno);
   } else {
-    st->print_cr("@%d", bci());
+    st->print("@%d", bci());
   }
+  if (should_reexecute()) {
+    st->print("  reexecute=true");
+  }
+  st->cr();
 }
 
 void ScopeDesc::print_on(outputStream* st) const {
@@ -174,7 +178,7 @@
 void ScopeDesc::print_on(outputStream* st, PcDesc* pd) const {
   // header
   if (pd != NULL) {
-    tty->print_cr("ScopeDesc(pc=" PTR_FORMAT " offset=%x):", pd->real_pc(_code), pd->pc_offset());
+    st->print_cr("ScopeDesc(pc=" PTR_FORMAT " offset=%x):", pd->real_pc(_code), pd->pc_offset());
   }
 
   print_value_on(st);
@@ -192,7 +196,7 @@
   // locals
   { GrowableArray<ScopeValue*>* l = ((ScopeDesc*) this)->locals();
     if (l != NULL) {
-      tty->print_cr("   Locals");
+      st->print_cr("   Locals");
       for (int index = 0; index < l->length(); index++) {
         st->print("    - l%d: ", index);
         l->at(index)->print_on(st);
@@ -205,7 +209,7 @@
     if (l != NULL) {
       st->print_cr("   Expression stack");
       for (int index = 0; index < l->length(); index++) {
-        st->print("   - @%d: ", index);
+        st->print("    - @%d: ", index);
         l->at(index)->print_on(st);
         st->cr();
       }
@@ -225,12 +229,12 @@
 
 #ifdef COMPILER2
   if (DoEscapeAnalysis && is_top() && _objects != NULL) {
-    tty->print_cr("Objects");
+    st->print_cr("   Objects");
     for (int i = 0; i < _objects->length(); i++) {
       ObjectValue* sv = (ObjectValue*) _objects->at(i);
-      tty->print(" - %d: ", sv->id());
-      sv->print_fields_on(tty);
-      tty->cr();
+      st->print("    - %d: ", sv->id());
+      sv->print_fields_on(st);
+      st->cr();
     }
   }
 #endif // COMPILER2
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,8 +25,8 @@
 #include "precompiled.hpp"
 #include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
 #include "gc_implementation/concurrentMarkSweep/freeChunk.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/freeBlockDictionary.hpp"
-#include "memory/sharedHeap.hpp"
 #include "runtime/globals.hpp"
 #include "runtime/mutex.hpp"
 #include "runtime/orderAccess.inline.hpp"
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/cmsOopClosures.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "gc_implementation/concurrentMarkSweep/cmsOopClosures.inline.hpp"
+#include "memory/iterator.inline.hpp"
+#include "memory/specialized_oop_closures.hpp"
+
+// Generate CMS specialized oop_oop_iterate functions.
+SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_CMS(ALL_KLASS_OOP_OOP_ITERATE_DEFN)
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -32,6 +32,7 @@
 #include "gc_interface/collectedHeap.inline.hpp"
 #include "memory/allocation.inline.hpp"
 #include "memory/blockOffsetTable.inline.hpp"
+#include "memory/genCollectedHeap.hpp"
 #include "memory/resourceArea.hpp"
 #include "memory/space.inline.hpp"
 #include "memory/universe.inline.hpp"
@@ -673,10 +674,10 @@
                                                  HeapWord* bottom,              \
                                                  HeapWord* top,                 \
                                                  ClosureType* cl) {             \
-   bool is_par = SharedHeap::heap()->n_par_threads() > 0;                       \
+   bool is_par = GenCollectedHeap::heap()->n_par_threads() > 0;                 \
    if (is_par) {                                                                \
-     assert(SharedHeap::heap()->n_par_threads() ==                              \
-            SharedHeap::heap()->workers()->active_workers(), "Mismatch");       \
+     assert(GenCollectedHeap::heap()->n_par_threads() ==                        \
+            GenCollectedHeap::heap()->workers()->active_workers(), "Mismatch"); \
      walk_mem_region_with_cl_par(mr, bottom, top, cl);                          \
    } else {                                                                     \
      walk_mem_region_with_cl_nopar(mr, bottom, top, cl);                        \
@@ -1907,11 +1908,11 @@
   assert(chunk->is_free() && ffc->is_free(), "Error");
   _bt.split_block((HeapWord*)chunk, chunk->size(), new_size);
   if (rem_sz < SmallForDictionary) {
-    bool is_par = (SharedHeap::heap()->n_par_threads() > 0);
+    bool is_par = (GenCollectedHeap::heap()->n_par_threads() > 0);
     if (is_par) _indexedFreeListParLocks[rem_sz]->lock();
     assert(!is_par ||
-           (SharedHeap::heap()->n_par_threads() ==
-            SharedHeap::heap()->workers()->active_workers()), "Mismatch");
+           (GenCollectedHeap::heap()->n_par_threads() ==
+            GenCollectedHeap::heap()->workers()->active_workers()), "Mismatch");
     returnChunkToFreeList(ffc);
     split(size, rem_sz);
     if (is_par) _indexedFreeListParLocks[rem_sz]->unlock();
@@ -1982,7 +1983,7 @@
 
 bool CompactibleFreeListSpace::no_allocs_since_save_marks() {
   assert(_promoInfo.tracking(), "No preceding save_marks?");
-  assert(SharedHeap::heap()->n_par_threads() == 0,
+  assert(GenCollectedHeap::heap()->n_par_threads() == 0,
          "Shouldn't be called if using parallel gc.");
   return _promoInfo.noPromotions();
 }
@@ -1991,7 +1992,7 @@
                                                                             \
 void CompactibleFreeListSpace::                                             \
 oop_since_save_marks_iterate##nv_suffix(OopClosureType* blk) {              \
-  assert(SharedHeap::heap()->n_par_threads() == 0,                          \
+  assert(GenCollectedHeap::heap()->n_par_threads() == 0,                    \
          "Shouldn't be called (yet) during parallel part of gc.");          \
   _promoInfo.promoted_oops_iterate##nv_suffix(blk);                         \
   /*                                                                        \
@@ -2442,11 +2443,10 @@
   {
     VerifyAllOopsClosure cl(_collector, this, span, past_remark,
       _collector->markBitMap());
-    CollectedHeap* ch = Universe::heap();
 
     // Iterate over all oops in the heap. Uses the _no_header version
     // since we are not interested in following the klass pointers.
-    ch->oop_iterate_no_header(&cl);
+    GenCollectedHeap::heap()->oop_iterate_no_header(&cl);
   }
 
   if (VerifyObjectStartArray) {
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -28,7 +28,7 @@
 #include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
 #include "gc_implementation/concurrentMarkSweep/promotionInfo.hpp"
 #include "memory/binaryTreeDictionary.hpp"
-#include "memory/blockOffsetTable.inline.hpp"
+#include "memory/blockOffsetTable.hpp"
 #include "memory/freeList.hpp"
 #include "memory/space.hpp"
 
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -53,6 +53,7 @@
 #include "memory/padded.hpp"
 #include "memory/referencePolicy.hpp"
 #include "memory/resourceArea.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "memory/tenuredGeneration.hpp"
 #include "oops/oop.inline.hpp"
 #include "prims/jvmtiExport.hpp"
@@ -64,6 +65,7 @@
 #include "runtime/vmThread.hpp"
 #include "services/memoryService.hpp"
 #include "services/runtimeService.hpp"
+#include "utilities/stack.inline.hpp"
 
 // statics
 CMSCollector* ConcurrentMarkSweepGeneration::_collector = NULL;
@@ -208,10 +210,6 @@
                                            use_adaptive_freelists,
                                            dictionaryChoice);
   NOT_PRODUCT(debug_cms_space = _cmsSpace;)
-  if (_cmsSpace == NULL) {
-    vm_exit_during_initialization(
-      "CompactibleFreeListSpace allocation failure");
-  }
   _cmsSpace->_gen = this;
 
   _gc_stats = new CMSGCStats();
@@ -230,14 +228,8 @@
     typedef CMSParGCThreadState* CMSParGCThreadStatePtr;
     _par_gc_thread_states =
       NEW_C_HEAP_ARRAY(CMSParGCThreadStatePtr, ParallelGCThreads, mtGC);
-    if (_par_gc_thread_states == NULL) {
-      vm_exit_during_initialization("Could not allocate par gc structs");
-    }
     for (uint i = 0; i < ParallelGCThreads; i++) {
       _par_gc_thread_states[i] = new CMSParGCThreadState(cmsSpace());
-      if (_par_gc_thread_states[i] == NULL) {
-        vm_exit_during_initialization("Could not allocate par gc structs");
-      }
     }
   } else {
     _par_gc_thread_states = NULL;
@@ -308,8 +300,6 @@
 
 AdaptiveSizePolicy* CMSCollector::size_policy() {
   GenCollectedHeap* gch = GenCollectedHeap::heap();
-  assert(gch->kind() == CollectedHeap::GenCollectedHeap,
-    "Wrong type of heap");
   return gch->gen_policy()->size_policy();
 }
 
@@ -586,11 +576,6 @@
         return;
       }
       _hash_seed = NEW_C_HEAP_ARRAY(int, num_queues, mtGC);
-      if (_hash_seed == NULL) {
-        warning("_hash_seed array allocation failure");
-        return;
-      }
-
       typedef Padded<OopTaskQueue> PaddedOopTaskQueue;
       for (i = 0; i < num_queues; i++) {
         PaddedOopTaskQueue *q = new PaddedOopTaskQueue();
@@ -633,12 +618,7 @@
     _eden_chunk_index = 0;
     _eden_chunk_capacity = (_young_gen->max_capacity()+CMSSamplingGrain)/CMSSamplingGrain;
     _eden_chunk_array = NEW_C_HEAP_ARRAY(HeapWord*, _eden_chunk_capacity, mtGC);
-    if (_eden_chunk_array == NULL) {
-      _eden_chunk_capacity = 0;
-      warning("GC/CMS: _eden_chunk_array allocation failure");
-    }
-  }
-  assert(_eden_chunk_array != NULL || _eden_chunk_capacity == 0, "Error");
+  }
 
   // Support for parallelizing survivor space rescan
   if ((CMSParallelRemarkEnabled && CMSParallelSurvivorRemarkEnabled) || CMSParallelInitialMarkEnabled) {
@@ -648,52 +628,15 @@
     _survivor_plab_array  = NEW_C_HEAP_ARRAY(ChunkArray, ParallelGCThreads, mtGC);
     _survivor_chunk_array = NEW_C_HEAP_ARRAY(HeapWord*, 2*max_plab_samples, mtGC);
     _cursor               = NEW_C_HEAP_ARRAY(size_t, ParallelGCThreads, mtGC);
-    if (_survivor_plab_array == NULL || _survivor_chunk_array == NULL
-        || _cursor == NULL) {
-      warning("Failed to allocate survivor plab/chunk array");
-      if (_survivor_plab_array  != NULL) {
-        FREE_C_HEAP_ARRAY(ChunkArray, _survivor_plab_array);
-        _survivor_plab_array = NULL;
-      }
-      if (_survivor_chunk_array != NULL) {
-        FREE_C_HEAP_ARRAY(HeapWord*, _survivor_chunk_array);
-        _survivor_chunk_array = NULL;
-      }
-      if (_cursor != NULL) {
-        FREE_C_HEAP_ARRAY(size_t, _cursor);
-        _cursor = NULL;
-      }
-    } else {
-      _survivor_chunk_capacity = 2*max_plab_samples;
-      for (uint i = 0; i < ParallelGCThreads; i++) {
-        HeapWord** vec = NEW_C_HEAP_ARRAY(HeapWord*, max_plab_samples, mtGC);
-        if (vec == NULL) {
-          warning("Failed to allocate survivor plab array");
-          for (int j = i; j > 0; j--) {
-            FREE_C_HEAP_ARRAY(HeapWord*, _survivor_plab_array[j-1].array());
-          }
-          FREE_C_HEAP_ARRAY(ChunkArray, _survivor_plab_array);
-          FREE_C_HEAP_ARRAY(HeapWord*, _survivor_chunk_array);
-          _survivor_plab_array = NULL;
-          _survivor_chunk_array = NULL;
-          _survivor_chunk_capacity = 0;
-          break;
-        } else {
-          ChunkArray* cur =
-            ::new (&_survivor_plab_array[i]) ChunkArray(vec,
-                                                        max_plab_samples);
-          assert(cur->end() == 0, "Should be 0");
-          assert(cur->array() == vec, "Should be vec");
-          assert(cur->capacity() == max_plab_samples, "Error");
-        }
-      }
-    }
-  }
-  assert(   (   _survivor_plab_array  != NULL
-             && _survivor_chunk_array != NULL)
-         || (   _survivor_chunk_capacity == 0
-             && _survivor_chunk_index == 0),
-         "Error");
+    _survivor_chunk_capacity = 2*max_plab_samples;
+    for (uint i = 0; i < ParallelGCThreads; i++) {
+      HeapWord** vec = NEW_C_HEAP_ARRAY(HeapWord*, max_plab_samples, mtGC);
+      ChunkArray* cur = ::new (&_survivor_plab_array[i]) ChunkArray(vec, max_plab_samples);
+      assert(cur->end() == 0, "Should be 0");
+      assert(cur->array() == vec, "Should be vec");
+      assert(cur->capacity() == max_plab_samples, "Error");
+    }
+  }
 
   NOT_PRODUCT(_overflow_counter = CMSMarkStackOverflowInterval;)
   _gc_counters = new CollectorCounters("CMS", 1);
@@ -1037,7 +980,7 @@
   assert_lock_strong(freelistLock());
 
 #ifndef PRODUCT
-  if (Universe::heap()->promotion_should_fail()) {
+  if (GenCollectedHeap::heap()->promotion_should_fail()) {
     return NULL;
   }
 #endif  // #ifndef PRODUCT
@@ -1114,7 +1057,7 @@
                                            oop old, markOop m,
                                            size_t word_sz) {
 #ifndef PRODUCT
-  if (Universe::heap()->promotion_should_fail()) {
+  if (GenCollectedHeap::heap()->promotion_should_fail()) {
     return NULL;
   }
 #endif  // #ifndef PRODUCT
@@ -2524,7 +2467,7 @@
   verification_mark_bm()->iterate(&vcl);
   if (vcl.failed()) {
     gclog_or_tty->print("Verification failed");
-    Universe::heap()->print_on(gclog_or_tty);
+    gch->print_on(gclog_or_tty);
     fatal("CMS: failed marking verification after remark");
   }
 }
@@ -3071,10 +3014,10 @@
       gch->set_par_threads(n_workers);
       initialize_sequential_subtasks_for_young_gen_rescan(n_workers);
       if (n_workers > 1) {
-        GenCollectedHeap::StrongRootsScope srs(gch);
+        StrongRootsScope srs;
         workers->run_task(&tsk);
       } else {
-        GenCollectedHeap::StrongRootsScope srs(gch);
+        StrongRootsScope srs;
         tsk.work(0);
       }
       gch->set_par_threads(0);
@@ -5169,11 +5112,11 @@
     // necessarily be so, since it's possible that we are doing
     // ST marking.
     ReferenceProcessorMTDiscoveryMutator mt(ref_processor(), true);
-    GenCollectedHeap::StrongRootsScope srs(gch);
+    StrongRootsScope srs;
     workers->run_task(&tsk);
   } else {
     ReferenceProcessorMTDiscoveryMutator mt(ref_processor(), false);
-    GenCollectedHeap::StrongRootsScope srs(gch);
+    StrongRootsScope srs;
     tsk.work(0);
   }
 
@@ -5241,7 +5184,7 @@
     verify_work_stacks_empty();
 
     gch->rem_set()->prepare_for_younger_refs_iterate(false); // Not parallel.
-    GenCollectedHeap::StrongRootsScope srs(gch);
+    StrongRootsScope srs;
 
     gch->gen_process_roots(_cmsGen->level(),
                            true,  // younger gens as roots
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -38,8 +38,8 @@
 #include "runtime/mutexLocker.hpp"
 #include "runtime/virtualspace.hpp"
 #include "services/memoryService.hpp"
-#include "utilities/bitMap.inline.hpp"
-#include "utilities/stack.inline.hpp"
+#include "utilities/bitMap.hpp"
+#include "utilities/stack.hpp"
 #include "utilities/taskqueue.hpp"
 #include "utilities/yieldingWorkgroup.hpp"
 
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepThread.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepThread.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -27,7 +27,7 @@
 
 #include "gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.hpp"
 #include "gc_implementation/shared/concurrentGCThread.hpp"
-#include "runtime/thread.inline.hpp"
+#include "runtime/thread.hpp"
 
 class ConcurrentMarkSweepGeneration;
 class CMSCollector;
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/promotionInfo.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/promotionInfo.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,6 +23,7 @@
  */
 
 #include "precompiled.hpp"
+#include "memory/genOopClosures.hpp"
 #include "gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp"
 #include "gc_implementation/concurrentMarkSweep/promotionInfo.hpp"
 #include "oops/markOop.inline.hpp"
--- a/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/vmCMSOperations.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/concurrentMarkSweep/vmCMSOperations.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -62,7 +62,7 @@
     HandleMark hm;
     FreelistLocker x(_collector);
     MutexLockerEx  y(_collector->bitMapLock(), Mutex::_no_safepoint_check_flag);
-    Universe::heap()->prepare_for_verify();
+    GenCollectedHeap::heap()->prepare_for_verify();
     Universe::verify();
   }
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -34,6 +34,7 @@
 #include "gc_implementation/g1/g1Log.hpp"
 #include "gc_implementation/g1/g1OopClosures.inline.hpp"
 #include "gc_implementation/g1/g1RemSet.hpp"
+#include "gc_implementation/g1/g1StringDedup.hpp"
 #include "gc_implementation/g1/heapRegion.inline.hpp"
 #include "gc_implementation/g1/heapRegionManager.inline.hpp"
 #include "gc_implementation/g1/heapRegionRemSet.hpp"
@@ -46,6 +47,7 @@
 #include "memory/genOopClosures.inline.hpp"
 #include "memory/referencePolicy.hpp"
 #include "memory/resourceArea.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/java.hpp"
@@ -115,7 +117,7 @@
 }
 
 size_t CMBitMap::compute_size(size_t heap_size) {
-  return heap_size / mark_distance();
+  return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
 }
 
 size_t CMBitMap::mark_distance() {
@@ -1325,7 +1327,7 @@
 
   if (VerifyDuringGC) {
     HandleMark hm;  // handle scope
-    Universe::heap()->prepare_for_verify();
+    g1h->prepare_for_verify();
     Universe::verify(VerifyOption_G1UsePrevMarking,
                      " VerifyDuringGC:(before)");
   }
@@ -1352,7 +1354,7 @@
     // Verify the heap w.r.t. the previous marking bitmap.
     if (VerifyDuringGC) {
       HandleMark hm;  // handle scope
-      Universe::heap()->prepare_for_verify();
+      g1h->prepare_for_verify();
       Universe::verify(VerifyOption_G1UsePrevMarking,
                        " VerifyDuringGC:(overflow)");
     }
@@ -1378,7 +1380,7 @@
 
     if (VerifyDuringGC) {
       HandleMark hm;  // handle scope
-      Universe::heap()->prepare_for_verify();
+      g1h->prepare_for_verify();
       Universe::verify(VerifyOption_G1UseNextMarking,
                        " VerifyDuringGC:(after)");
     }
@@ -1986,13 +1988,13 @@
 
   if (VerifyDuringGC) {
     HandleMark hm;  // handle scope
-    Universe::heap()->prepare_for_verify();
+    g1h->prepare_for_verify();
     Universe::verify(VerifyOption_G1UsePrevMarking,
                      " VerifyDuringGC:(before)");
   }
   g1h->check_bitmaps("Cleanup Start");
 
-  G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy();
+  G1CollectorPolicy* g1p = g1h->g1_policy();
   g1p->record_concurrent_mark_cleanup_start();
 
   double start = os::elapsedTime();
@@ -2097,7 +2099,7 @@
 
   if (VerifyDuringGC) {
     HandleMark hm;  // handle scope
-    Universe::heap()->prepare_for_verify();
+    g1h->prepare_for_verify();
     Universe::verify(VerifyOption_G1UsePrevMarking,
                      " VerifyDuringGC:(after)");
   }
@@ -2650,7 +2652,7 @@
 
   g1h->ensure_parsability(false);
 
-  G1CollectedHeap::StrongRootsScope srs(g1h);
+  StrongRootsScope srs;
   // this is remark, so we'll use up all active threads
   uint active_workers = g1h->workers()->active_workers();
   if (active_workers == 0) {
@@ -3392,22 +3394,29 @@
 }
 #endif
 
-void CMTask::scan_object(oop obj) {
+template<bool scan>
+inline void CMTask::process_grey_object(oop obj) {
+  assert(scan || obj->is_typeArray(), "Skipping scan of grey non-typeArray");
   assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
 
   if (_cm->verbose_high()) {
-    gclog_or_tty->print_cr("[%u] we're scanning object "PTR_FORMAT,
+    gclog_or_tty->print_cr("[%u] processing grey object " PTR_FORMAT,
                            _worker_id, p2i((void*) obj));
   }
 
   size_t obj_size = obj->size();
   _words_scanned += obj_size;
 
-  obj->oop_iterate(_cm_oop_closure);
+  if (scan) {
+    obj->oop_iterate(_cm_oop_closure);
+  }
   statsOnly( ++_objs_scanned );
   check_limits();
 }
 
+template void CMTask::process_grey_object<true>(oop);
+template void CMTask::process_grey_object<false>(oop);
+
 // Closure for iteration over bitmaps
 class CMBitMapClosure : public BitMapClosure {
 private:
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1100,6 +1100,12 @@
   void regular_clock_call();
   bool concurrent() { return _concurrent; }
 
+  // Test whether objAddr might have already been passed over by the
+  // mark bitmap scan, and so needs to be pushed onto the mark stack.
+  bool is_below_finger(HeapWord* objAddr, HeapWord* global_finger) const;
+
+  template<bool scan> void process_grey_object(oop obj);
+
 public:
   // It resets the task; it should be called right at the beginning of
   // a marking phase.
@@ -1152,7 +1158,7 @@
   inline void deal_with_reference(oop obj);
 
   // It scans an object and visits its children.
-  void scan_object(oop obj);
+  void scan_object(oop obj) { process_grey_object<true>(obj); }
 
   // It pushes an object on the local queue.
   inline void push(oop obj);
--- a/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/concurrentMark.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -259,14 +259,35 @@
              ++_local_pushes );
 }
 
-// This determines whether the method below will check both the local
-// and global fingers when determining whether to push on the stack a
-// gray object (value 1) or whether it will only check the global one
-// (value 0). The tradeoffs are that the former will be a bit more
-// accurate and possibly push less on the stack, but it might also be
-// a little bit slower.
+inline bool CMTask::is_below_finger(HeapWord* objAddr,
+                                    HeapWord* global_finger) const {
+  // If objAddr is above the global finger, then the mark bitmap scan
+  // will find it later, and no push is needed.  Similarly, if we have
+  // a current region and objAddr is between the local finger and the
+  // end of the current region, then no push is needed.  The tradeoff
+  // of checking both vs only checking the global finger is that the
+  // local check will be more accurate and so result in fewer pushes,
+  // but may also be a little slower.
+  if (_finger != NULL) {
+    // We have a current region.
 
-#define _CHECK_BOTH_FINGERS_      1
+    // Finger and region values are all NULL or all non-NULL.  We
+    // use _finger to check since we immediately use its value.
+    assert(_curr_region != NULL, "invariant");
+    assert(_region_limit != NULL, "invariant");
+    assert(_region_limit <= global_finger, "invariant");
+
+    // True if objAddr is less than the local finger, or is between
+    // the region limit and the global finger.
+    if (objAddr < _finger) {
+      return true;
+    } else if (objAddr < _region_limit) {
+      return false;
+    } // Else check global finger.
+  }
+  // Check global finger.
+  return objAddr < global_finger;
+}
 
 inline void CMTask::deal_with_reference(oop obj) {
   if (_cm->verbose_high()) {
@@ -297,50 +318,43 @@
           // CAS done in CMBitMap::parMark() call in the routine above.
           HeapWord* global_finger = _cm->finger();
 
-#if _CHECK_BOTH_FINGERS_
-          // we will check both the local and global fingers
-
-          if (_finger != NULL && objAddr < _finger) {
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the local finger ("PTR_FORMAT"), "
-                                     "pushing it", _worker_id, p2i(_finger));
+          // We only need to push a newly grey object on the mark
+          // stack if it is in a section of memory the mark bitmap
+          // scan has already examined.  Mark bitmap scanning
+          // maintains progress "fingers" for determining that.
+          //
+          // Notice that the global finger might be moving forward
+          // concurrently. This is not a problem. In the worst case, we
+          // mark the object while it is above the global finger and, by
+          // the time we read the global finger, it has moved forward
+          // past this object. In this case, the object will probably
+          // be visited when a task is scanning the region and will also
+          // be pushed on the stack. So, some duplicate work, but no
+          // correctness problems.
+          if (is_below_finger(objAddr, global_finger)) {
+            if (obj->is_typeArray()) {
+              // Immediately process arrays of primitive types, rather
+              // than pushing on the mark stack.  This keeps us from
+              // adding humongous objects to the mark stack that might
+              // be reclaimed before the entry is processed - see
+              // selection of candidates for eager reclaim of humongous
+              // objects.  The cost of the additional type test is
+              // mitigated by avoiding a trip through the mark stack,
+              // by only doing a bookkeeping update and avoiding the
+              // actual scan of the object - a typeArray contains no
+              // references, and the metadata is built-in.
+              process_grey_object<false>(obj);
+            } else {
+              if (_cm->verbose_high()) {
+                gclog_or_tty->print_cr("[%u] below a finger (local: " PTR_FORMAT
+                                       ", global: " PTR_FORMAT ") pushing "
+                                       PTR_FORMAT " on mark stack",
+                                       _worker_id, p2i(_finger),
+                                       p2i(global_finger), p2i(objAddr));
+              }
+              push(obj);
             }
-            push(obj);
-          } else if (_curr_region != NULL && objAddr < _region_limit) {
-            // do nothing
-          } else if (objAddr < global_finger) {
-            // Notice that the global finger might be moving forward
-            // concurrently. This is not a problem. In the worst case, we
-            // mark the object while it is above the global finger and, by
-            // the time we read the global finger, it has moved forward
-            // passed this object. In this case, the object will probably
-            // be visited when a task is scanning the region and will also
-            // be pushed on the stack. So, some duplicate work, but no
-            // correctness problems.
-
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the global finger "
-                                     "("PTR_FORMAT"), pushing it",
-                                     _worker_id, p2i(global_finger));
-            }
-            push(obj);
-          } else {
-            // do nothing
           }
-#else // _CHECK_BOTH_FINGERS_
-          // we will only check the global finger
-
-          if (objAddr < global_finger) {
-            // see long comment above
-
-            if (_cm->verbose_high()) {
-              gclog_or_tty->print_cr("[%u] below the global finger "
-                                     "("PTR_FORMAT"), pushing it",
-                                     _worker_id, p2i(global_finger));
-            }
-            push(obj);
-          }
-#endif // _CHECK_BOTH_FINGERS_
         }
       }
     }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1Allocator.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -29,6 +29,9 @@
 #include "gc_implementation/g1/g1AllocRegion.hpp"
 #include "gc_implementation/g1/g1InCSetState.hpp"
 #include "gc_implementation/shared/parGCAllocBuffer.hpp"
+#include "gc_interface/collectedHeap.hpp"
+
+class EvacuationInfo;
 
 // Base class for G1 allocators.
 class G1Allocator : public CHeapObj<mtGC> {
--- a/hotspot/src/share/vm/gc_implementation/g1/g1BlockOffsetTable.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1BlockOffsetTable.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -23,6 +23,7 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_implementation/g1/g1CollectedHeap.hpp"
 #include "gc_implementation/g1/g1BlockOffsetTable.inline.hpp"
 #include "gc_implementation/g1/heapRegion.hpp"
 #include "memory/space.hpp"
@@ -303,9 +304,9 @@
   assert(blk_start <= threshold, "blk_start should be at or before threshold");
   assert(pointer_delta(threshold, blk_start) <= N_words,
          "offset should be <= BlockOffsetSharedArray::N");
-  assert(Universe::heap()->is_in_reserved(blk_start),
+  assert(G1CollectedHeap::heap()->is_in_reserved(blk_start),
          "reference must be into the heap");
-  assert(Universe::heap()->is_in_reserved(blk_end-1),
+  assert(G1CollectedHeap::heap()->is_in_reserved(blk_end-1),
          "limit must be within the heap");
   assert(threshold == _array->_reserved.start() + index*N_words,
          "index must agree with threshold");
@@ -458,7 +459,7 @@
 }
 
 HeapWord* G1BlockOffsetArrayContigSpace::initialize_threshold_raw() {
-  assert(!Universe::heap()->is_in_reserved(_array->_offset_array),
+  assert(!G1CollectedHeap::heap()->is_in_reserved(_array->_offset_array),
          "just checking");
   _next_offset_index = _array->index_for_raw(_bottom);
   _next_offset_index++;
@@ -468,7 +469,7 @@
 }
 
 void G1BlockOffsetArrayContigSpace::zero_bottom_entry_raw() {
-  assert(!Universe::heap()->is_in_reserved(_array->_offset_array),
+  assert(!G1CollectedHeap::heap()->is_in_reserved(_array->_offset_array),
          "just checking");
   size_t bottom_index = _array->index_for_raw(_bottom);
   assert(_array->address_for_index_raw(bottom_index) == _bottom,
@@ -477,7 +478,7 @@
 }
 
 HeapWord* G1BlockOffsetArrayContigSpace::initialize_threshold() {
-  assert(!Universe::heap()->is_in_reserved(_array->_offset_array),
+  assert(!G1CollectedHeap::heap()->is_in_reserved(_array->_offset_array),
          "just checking");
   _next_offset_index = _array->index_for(_bottom);
   _next_offset_index++;
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -70,6 +70,7 @@
 #include "runtime/orderAccess.inline.hpp"
 #include "runtime/vmThread.hpp"
 #include "utilities/globalDefinitions.hpp"
+#include "utilities/stack.inline.hpp"
 
 size_t G1CollectedHeap::_humongous_object_threshold_in_words = 0;
 
@@ -1728,7 +1729,7 @@
 
 
 G1CollectedHeap::G1CollectedHeap(G1CollectorPolicy* policy_) :
-  SharedHeap(),
+  CollectedHeap(),
   _g1_policy(policy_),
   _dirty_card_queue_set(false),
   _into_cset_dirty_card_queue_set(false),
@@ -1746,7 +1747,7 @@
   _secondary_free_list("Secondary Free List", new SecondaryFreeRegionListMtSafeChecker()),
   _old_set("Old Set", false /* humongous */, new OldRegionSetMtSafeChecker()),
   _humongous_set("Master Humongous Set", true /* humongous */, new HumongousRegionSetMtSafeChecker()),
-  _humongous_is_live(),
+  _humongous_reclaim_candidates(),
   _has_humongous_reclaim_candidates(false),
   _free_regions_coming(false),
   _young_list(new YoungList(this)),
@@ -1770,6 +1771,11 @@
 
   _g1h = this;
 
+  _workers = new FlexibleWorkGang("GC Thread", ParallelGCThreads,
+                          /* are_GC_task_threads */true,
+                          /* are_ConcurrentGC_threads */false);
+  _workers->initialize_workers();
+
   _allocator = G1Allocator::create_allocator(_g1h);
   _humongous_object_threshold_in_words = HeapRegion::GrainWords / 2;
 
@@ -1797,6 +1803,26 @@
   guarantee(_task_queues != NULL, "task_queues allocation failure.");
 }
 
+G1RegionToSpaceMapper* G1CollectedHeap::create_aux_memory_mapper(const char* description,
+                                                                 size_t size,
+                                                                 size_t translation_factor) {
+  size_t preferred_page_size = os::page_size_for_region_unaligned(size, 1);
+  // Allocate a new reserved space, preferring to use large pages.
+  ReservedSpace rs(size, preferred_page_size);
+  G1RegionToSpaceMapper* result  =
+    G1RegionToSpaceMapper::create_mapper(rs,
+                                         size,
+                                         rs.alignment(),
+                                         HeapRegion::GrainBytes,
+                                         translation_factor,
+                                         mtGC);
+  if (TracePageSizes) {
+    gclog_or_tty->print_cr("G1 '%s': pg_sz=" SIZE_FORMAT " base=" PTR_FORMAT " size=" SIZE_FORMAT " alignment=" SIZE_FORMAT " reqsize=" SIZE_FORMAT,
+                           description, preferred_page_size, p2i(rs.base()), rs.size(), rs.alignment(), size);
+  }
+  return result;
+}
+
 jint G1CollectedHeap::initialize() {
   CollectedHeap::pre_initialize();
   os::enable_vtime();
@@ -1864,57 +1890,35 @@
   ReservedSpace g1_rs = heap_rs.first_part(max_byte_size);
   G1RegionToSpaceMapper* heap_storage =
     G1RegionToSpaceMapper::create_mapper(g1_rs,
+                                         g1_rs.size(),
                                          UseLargePages ? os::large_page_size() : os::vm_page_size(),
                                          HeapRegion::GrainBytes,
                                          1,
                                          mtJavaHeap);
   heap_storage->set_mapping_changed_listener(&_listener);
 
-  // Reserve space for the block offset table. We do not support automatic uncommit
-  // for the card table at this time. BOT only.
-  ReservedSpace bot_rs(G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize));
+  // Create storage for the BOT, card table, card counts table (hot card cache) and the bitmaps.
   G1RegionToSpaceMapper* bot_storage =
-    G1RegionToSpaceMapper::create_mapper(bot_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
+    create_aux_memory_mapper("Block offset table",
+                             G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
 
   ReservedSpace cardtable_rs(G1SATBCardTableLoggingModRefBS::compute_size(g1_rs.size() / HeapWordSize));
   G1RegionToSpaceMapper* cardtable_storage =
-    G1RegionToSpaceMapper::create_mapper(cardtable_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
-
-  // Reserve space for the card counts table.
-  ReservedSpace card_counts_rs(G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize));
+    create_aux_memory_mapper("Card table",
+                             G1SATBCardTableLoggingModRefBS::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
+
   G1RegionToSpaceMapper* card_counts_storage =
-    G1RegionToSpaceMapper::create_mapper(card_counts_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         G1BlockOffsetSharedArray::N_bytes,
-                                         mtGC);
-
-  // Reserve space for prev and next bitmap.
+    create_aux_memory_mapper("Card counts table",
+                             G1BlockOffsetSharedArray::compute_size(g1_rs.size() / HeapWordSize),
+                             G1BlockOffsetSharedArray::N_bytes);
+
   size_t bitmap_size = CMBitMap::compute_size(g1_rs.size());
-
-  ReservedSpace prev_bitmap_rs(ReservedSpace::allocation_align_size_up(bitmap_size));
   G1RegionToSpaceMapper* prev_bitmap_storage =
-    G1RegionToSpaceMapper::create_mapper(prev_bitmap_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         CMBitMap::mark_distance(),
-                                         mtGC);
-
-  ReservedSpace next_bitmap_rs(ReservedSpace::allocation_align_size_up(bitmap_size));
+    create_aux_memory_mapper("Prev Bitmap", bitmap_size, CMBitMap::mark_distance());
   G1RegionToSpaceMapper* next_bitmap_storage =
-    G1RegionToSpaceMapper::create_mapper(next_bitmap_rs,
-                                         os::vm_page_size(),
-                                         HeapRegion::GrainBytes,
-                                         CMBitMap::mark_distance(),
-                                         mtGC);
+    create_aux_memory_mapper("Next Bitmap", bitmap_size, CMBitMap::mark_distance());
 
   _hrm.initialize(heap_storage, prev_bitmap_storage, next_bitmap_storage, bot_storage, cardtable_storage, card_counts_storage);
   g1_barrier_set()->initialize(cardtable_storage);
@@ -1937,8 +1941,14 @@
 
   _g1h = this;
 
-  _in_cset_fast_test.initialize(_hrm.reserved().start(), _hrm.reserved().end(), HeapRegion::GrainBytes);
-  _humongous_is_live.initialize(_hrm.reserved().start(), _hrm.reserved().end(), HeapRegion::GrainBytes);
+  {
+    HeapWord* start = _hrm.reserved().start();
+    HeapWord* end = _hrm.reserved().end();
+    size_t granularity = HeapRegion::GrainBytes;
+
+    _in_cset_fast_test.initialize(start, end, granularity);
+    _humongous_reclaim_candidates.initialize(start, end, granularity);
+  }
 
   // Create the ConcurrentMark data structure and thread.
   // (Must do this late, so that "max_regions" is defined.)
@@ -2026,15 +2036,15 @@
   }
 }
 
-void G1CollectedHeap::clear_humongous_is_live_table() {
-  guarantee(G1EagerReclaimHumongousObjects, "Should only be called if true");
-  _humongous_is_live.clear();
-}
-
 size_t G1CollectedHeap::conservative_max_heap_alignment() {
   return HeapRegion::max_region_size();
 }
 
+void G1CollectedHeap::post_initialize() {
+  CollectedHeap::post_initialize();
+  ref_processing_init();
+}
+
 void G1CollectedHeap::ref_processing_init() {
   // Reference processing in G1 currently works as follows:
   //
@@ -2071,7 +2081,6 @@
   //     * Discovery is atomic - i.e. not concurrent.
   //     * Reference discovery will not need a barrier.
 
-  SharedHeap::ref_processing_init();
   MemRegion mr = reserved_region();
 
   // Concurrent Mark ref processor
@@ -2128,6 +2137,7 @@
 }
 
 #ifndef PRODUCT
+
 class CheckGCTimeStampsHRClosure : public HeapRegionClosure {
 private:
   unsigned _gc_time_stamp;
@@ -2462,11 +2472,6 @@
   }
 };
 
-void G1CollectedHeap::oop_iterate(ExtendedOopClosure* cl) {
-  IterateOopClosureRegionClosure blk(cl);
-  heap_region_iterate(&blk);
-}
-
 // Iterates an ObjectClosure over all objects within a HeapRegion.
 
 class IterateObjectClosureRegionClosure: public HeapRegionClosure {
@@ -2486,23 +2491,6 @@
   heap_region_iterate(&blk);
 }
 
-// Calls a SpaceClosure on a HeapRegion.
-
-class SpaceClosureRegionClosure: public HeapRegionClosure {
-  SpaceClosure* _cl;
-public:
-  SpaceClosureRegionClosure(SpaceClosure* cl) : _cl(cl) {}
-  bool doHeapRegion(HeapRegion* r) {
-    _cl->do_space(r);
-    return false;
-  }
-};
-
-void G1CollectedHeap::space_iterate(SpaceClosure* cl) {
-  SpaceClosureRegionClosure blk(cl);
-  heap_region_iterate(&blk);
-}
-
 void G1CollectedHeap::heap_region_iterate(HeapRegionClosure* cl) const {
   _hrm.iterate(cl);
 }
@@ -2639,23 +2627,19 @@
   return result;
 }
 
-Space* G1CollectedHeap::space_containing(const void* addr) const {
-  return heap_region_containing(addr);
-}
-
 HeapWord* G1CollectedHeap::block_start(const void* addr) const {
-  Space* sp = space_containing(addr);
-  return sp->block_start(addr);
+  HeapRegion* hr = heap_region_containing(addr);
+  return hr->block_start(addr);
 }
 
 size_t G1CollectedHeap::block_size(const HeapWord* addr) const {
-  Space* sp = space_containing(addr);
-  return sp->block_size(addr);
+  HeapRegion* hr = heap_region_containing(addr);
+  return hr->block_size(addr);
 }
 
 bool G1CollectedHeap::block_is_obj(const HeapWord* addr) const {
-  Space* sp = space_containing(addr);
-  return sp->block_is_obj(addr);
+  HeapRegion* hr = heap_region_containing(addr);
+  return hr->block_is_obj(addr);
 }
 
 bool G1CollectedHeap::supports_tlab_allocation() const {
@@ -3336,8 +3320,8 @@
 #endif // PRODUCT
 
 G1CollectedHeap* G1CollectedHeap::heap() {
-  assert(_sh->kind() == CollectedHeap::G1CollectedHeap,
-         "not a garbage-first heap");
+  assert(_g1h != NULL, "Uninitialized access to G1CollectedHeap::heap()");
+  assert(_g1h->kind() == CollectedHeap::G1CollectedHeap, "Not a G1 heap");
   return _g1h;
 }
 
@@ -3434,12 +3418,6 @@
   return g1_rem_set()->cardsScanned();
 }
 
-bool G1CollectedHeap::humongous_region_is_always_live(uint index) {
-  HeapRegion* region = region_at(index);
-  assert(region->is_starts_humongous(), "Must start a humongous object");
-  return oop(region->bottom())->is_objArray() || !region->rem_set()->is_empty();
-}
-
 class RegisterHumongousWithInCSetFastTestClosure : public HeapRegionClosure {
  private:
   size_t _total_humongous;
@@ -3447,14 +3425,59 @@
 
   DirtyCardQueue _dcq;
 
-  bool humongous_region_is_candidate(uint index) {
-    HeapRegion* region = G1CollectedHeap::heap()->region_at(index);
-    assert(region->is_starts_humongous(), "Must start a humongous object");
+  // We don't nominate objects with many remembered set entries, on
+  // the assumption that such objects are likely still live.
+  bool is_remset_small(HeapRegion* region) const {
     HeapRegionRemSet* const rset = region->rem_set();
-    bool const allow_stale_refs = G1EagerReclaimHumongousObjectsWithStaleRefs;
-    return !oop(region->bottom())->is_objArray() &&
-           ((allow_stale_refs && rset->occupancy_less_or_equal_than(G1RSetSparseRegionEntries)) ||
-            (!allow_stale_refs && rset->is_empty()));
+    return G1EagerReclaimHumongousObjectsWithStaleRefs
+      ? rset->occupancy_less_or_equal_than(G1RSetSparseRegionEntries)
+      : rset->is_empty();
+  }
+
+  bool is_typeArray_region(HeapRegion* region) const {
+    return oop(region->bottom())->is_typeArray();
+  }
+
+  bool humongous_region_is_candidate(G1CollectedHeap* heap, HeapRegion* region) const {
+    assert(region->is_starts_humongous(), "Must start a humongous object");
+
+    // Candidate selection must satisfy the following constraints
+    // while concurrent marking is in progress:
+    //
+    // * In order to maintain SATB invariants, an object must not be
+    // reclaimed if it was allocated before the start of marking and
+    // has not had its references scanned.  Such an object must have
+    // its references (including type metadata) scanned to ensure no
+    // live objects are missed by the marking process.  Objects
+    // allocated after the start of concurrent marking don't need to
+    // be scanned.
+    //
+    // * An object must not be reclaimed if it is on the concurrent
+    // mark stack.  Objects allocated after the start of concurrent
+    // marking are never pushed on the mark stack.
+    //
+    // Nominating only objects allocated after the start of concurrent
+    // marking is sufficient to meet both constraints.  This may miss
+    // some objects that satisfy the constraints, but the marking data
+    // structures don't support efficiently performing the needed
+    // additional tests or scrubbing of the mark stack.
+    //
+    // However, we presently only nominate is_typeArray() objects.
+    // A humongous object containing references induces remembered
+    // set entries on other regions.  In order to reclaim such an
+    // object, those remembered sets would need to be cleaned up.
+    //
+    // We also treat is_typeArray() objects specially, allowing them
+    // to be reclaimed even if allocated before the start of
+    // concurrent mark.  For this we rely on mark stack insertion to
+    // exclude is_typeArray() objects, preventing reclaiming an object
+    // that is in the mark stack.  We also rely on the metadata for
+    // such objects to be built-in and so ensured to be kept live.
+    // Frequent allocation and drop of large binary blobs is an
+    // important use case for eager reclaim, and this special handling
+    // may reduce needed headroom.
+
+    return is_typeArray_region(region) && is_remset_small(region);
   }
 
  public:
@@ -3470,14 +3493,17 @@
     }
     G1CollectedHeap* g1h = G1CollectedHeap::heap();
 
-    uint region_idx = r->hrm_index();
-    bool is_candidate = humongous_region_is_candidate(region_idx);
-    // Is_candidate already filters out humongous object with large remembered sets.
-    // If we have a humongous object with a few remembered sets, we simply flush these
-    // remembered set entries into the DCQS. That will result in automatic
-    // re-evaluation of their remembered set entries during the following evacuation
-    // phase.
+    bool is_candidate = humongous_region_is_candidate(g1h, r);
+    uint rindex = r->hrm_index();
+    g1h->set_humongous_reclaim_candidate(rindex, is_candidate);
     if (is_candidate) {
+      _candidate_humongous++;
+      g1h->register_humongous_region_with_cset(rindex);
+      // Is_candidate already filters out humongous object with large remembered sets.
+      // If we have a humongous object with a few remembered sets, we simply flush these
+      // remembered set entries into the DCQS. That will result in automatic
+      // re-evaluation of their remembered set entries during the following evacuation
+      // phase.
       if (!r->rem_set()->is_empty()) {
         guarantee(r->rem_set()->occupancy_less_or_equal_than(G1RSetSparseRegionEntries),
                   "Found a not-small remembered set here. This is inconsistent with previous assumptions.");
@@ -3499,8 +3525,6 @@
         r->rem_set()->clear_locked();
       }
       assert(r->rem_set()->is_empty(), "At this point any humongous candidate remembered set must be empty.");
-      g1h->register_humongous_region_with_cset(region_idx);
-      _candidate_humongous++;
     }
     _total_humongous++;
 
@@ -3520,6 +3544,7 @@
   }
   double time = os::elapsed_counter();
 
+  // Collect reclaim candidate information and register candidates with cset.
   RegisterHumongousWithInCSetFastTestClosure cl;
   heap_region_iterate(&cl);
 
@@ -3529,10 +3554,6 @@
                                                                   cl.candidate_humongous());
   _has_humongous_reclaim_candidates = cl.candidate_humongous() > 0;
 
-  if (_has_humongous_reclaim_candidates || G1TraceEagerReclaimHumongousObjects) {
-    clear_humongous_is_live_table();
-  }
-
   // Finally flush all remembered set entries to re-check into the global DCQS.
   cl.flush_rem_set_entries();
 }
@@ -5994,11 +6015,11 @@
     // required because stale remembered sets might reference locations that
     // are currently allocated into.
     uint region_idx = r->hrm_index();
-    if (g1h->humongous_is_live(region_idx) ||
-        g1h->humongous_region_is_always_live(region_idx)) {
+    if (!g1h->is_humongous_reclaim_candidate(region_idx) ||
+        !r->rem_set()->is_empty()) {
 
       if (G1TraceEagerReclaimHumongousObjects) {
-        gclog_or_tty->print_cr("Live humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d live-other %d obj array %d",
+        gclog_or_tty->print_cr("Live humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d reclaim candidate %d type array %d",
                                region_idx,
                                obj->size()*HeapWordSize,
                                r->bottom(),
@@ -6006,20 +6027,21 @@
                                r->rem_set()->occupied(),
                                r->rem_set()->strong_code_roots_list_length(),
                                next_bitmap->isMarked(r->bottom()),
-                               g1h->humongous_is_live(region_idx),
-                               obj->is_objArray()
+                               g1h->is_humongous_reclaim_candidate(region_idx),
+                               obj->is_typeArray()
                               );
       }
 
       return false;
     }
 
-    guarantee(!obj->is_objArray(),
-              err_msg("Eagerly reclaiming object arrays is not supported, but the object "PTR_FORMAT" is.",
+    guarantee(obj->is_typeArray(),
+              err_msg("Only eagerly reclaiming type arrays is supported, but the object "
+                      PTR_FORMAT " is not.",
                       r->bottom()));
 
     if (G1TraceEagerReclaimHumongousObjects) {
-      gclog_or_tty->print_cr("Dead humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d live-other %d obj array %d",
+      gclog_or_tty->print_cr("Dead humongous region %u size "SIZE_FORMAT" start "PTR_FORMAT" length "UINT32_FORMAT" with remset "SIZE_FORMAT" code roots "SIZE_FORMAT" is marked %d reclaim candidate %d type array %d",
                              region_idx,
                              obj->size()*HeapWordSize,
                              r->bottom(),
@@ -6027,8 +6049,8 @@
                              r->rem_set()->occupied(),
                              r->rem_set()->strong_code_roots_list_length(),
                              next_bitmap->isMarked(r->bottom()),
-                             g1h->humongous_is_live(region_idx),
-                             obj->is_objArray()
+                             g1h->is_humongous_reclaim_candidate(region_idx),
+                             obj->is_typeArray()
                             );
     }
     // Need to clear mark bit of the humongous object if already set.
@@ -6163,8 +6185,6 @@
 }
 
 void G1CollectedHeap::set_region_short_lived_locked(HeapRegion* hr) {
-  assert(heap_lock_held_for_gc(),
-              "the heap lock should already be held by or for this thread");
   _young_list->push_region(hr);
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -40,9 +40,9 @@
 #include "gc_implementation/g1/heapRegionSet.hpp"
 #include "gc_implementation/shared/hSpaceCounters.hpp"
 #include "gc_implementation/shared/parGCAllocBuffer.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/barrierSet.hpp"
 #include "memory/memRegion.hpp"
-#include "memory/sharedHeap.hpp"
 #include "utilities/stack.hpp"
 
 // A "G1CollectedHeap" is an implementation of a java heap for HotSpot.
@@ -76,6 +76,7 @@
 class EvacuationFailedInfo;
 class nmethod;
 class Ticks;
+class FlexibleWorkGang;
 
 typedef OverflowTaskQueue<StarTask, mtGC>         RefToScanQueue;
 typedef GenericTaskQueueSet<RefToScanQueue, mtGC> RefToScanQueueSet;
@@ -177,7 +178,7 @@
   virtual void on_commit(uint start_idx, size_t num_regions, bool zero_filled);
 };
 
-class G1CollectedHeap : public SharedHeap {
+class G1CollectedHeap : public CollectedHeap {
   friend class VM_CollectForMetadataAllocation;
   friend class VM_G1CollectForAllocation;
   friend class VM_G1CollectFull;
@@ -204,6 +205,8 @@
   // The one and only G1CollectedHeap, so static functions can find it.
   static G1CollectedHeap* _g1h;
 
+  FlexibleWorkGang* _workers;
+
   static size_t _humongous_object_threshold_in_words;
 
   // The secondary free list which contains regions that have been
@@ -217,7 +220,6 @@
   // It keeps track of the humongous regions.
   HeapRegionSet _humongous_set;
 
-  void clear_humongous_is_live_table();
   void eagerly_reclaim_humongous_regions();
 
   // The number of regions we could create by expansion.
@@ -287,22 +289,26 @@
   // Helper for monitoring and management support.
   G1MonitoringSupport* _g1mm;
 
-  // Records whether the region at the given index is kept live by roots or
-  // references from the young generation.
-  class HumongousIsLiveBiasedMappedArray : public G1BiasedMappedArray<bool> {
+  // Records whether the region at the given index is (still) a
+  // candidate for eager reclaim.  Only valid for humongous start
+  // regions; other regions have unspecified values.  Humongous start
+  // regions are initialized at start of collection pause, with
+  // candidates removed from the set as they are found reachable from
+  // roots or the young generation.
+  class HumongousReclaimCandidates : public G1BiasedMappedArray<bool> {
    protected:
     bool default_value() const { return false; }
    public:
     void clear() { G1BiasedMappedArray<bool>::clear(); }
-    void set_live(uint region) {
-      set_by_index(region, true);
+    void set_candidate(uint region, bool value) {
+      set_by_index(region, value);
     }
-    bool is_live(uint region) {
+    bool is_candidate(uint region) {
       return get_by_index(region);
     }
   };
 
-  HumongousIsLiveBiasedMappedArray _humongous_is_live;
+  HumongousReclaimCandidates _humongous_reclaim_candidates;
   // Stores whether during humongous object registration we found candidate regions.
   // If not, we can skip a few steps.
   bool _has_humongous_reclaim_candidates;
@@ -351,6 +357,12 @@
   // heap after a compaction.
   void print_hrm_post_compaction();
 
+  // Create a memory mapper for auxiliary data structures of the given size and
+  // translation factor.
+  static G1RegionToSpaceMapper* create_aux_memory_mapper(const char* description,
+                                                         size_t size,
+                                                         size_t translation_factor);
+
   double verify(bool guard, const char* msg);
   void verify_before_gc();
   void verify_after_gc();
@@ -605,6 +617,7 @@
   void enqueue_discovered_references(uint no_of_gc_workers);
 
 public:
+  FlexibleWorkGang* workers() const { return _workers; }
 
   G1Allocator* allocator() {
     return _allocator;
@@ -630,21 +643,18 @@
   inline AllocationContextStats& allocation_context_stats();
 
   // Do anything common to GC's.
-  virtual void gc_prologue(bool full);
-  virtual void gc_epilogue(bool full);
+  void gc_prologue(bool full);
+  void gc_epilogue(bool full);
 
+  // Modify the reclaim candidate set and test for presence.
+  // These are only valid for starts_humongous regions.
+  inline void set_humongous_reclaim_candidate(uint region, bool value);
+  inline bool is_humongous_reclaim_candidate(uint region);
+
+  // Remove from the reclaim candidate set.  Also remove from the
+  // collection set so that later encounters avoid the slow path.
   inline void set_humongous_is_live(oop obj);
 
-  bool humongous_is_live(uint region) {
-    return _humongous_is_live.is_live(region);
-  }
-
-  // Returns whether the given region (which must be a humongous (start) region)
-  // is to be considered conservatively live regardless of any other conditions.
-  bool humongous_region_is_always_live(uint index);
-  // Returns whether the given region (which must be a humongous (start) region)
-  // is considered a candidate for eager reclamation.
-  bool humongous_region_is_candidate(uint index);
   // Register the given region to be part of the collection set.
   inline void register_humongous_region_with_cset(uint index);
   // Register regions with humongous objects (actually on the start region) in
@@ -1000,11 +1010,14 @@
   // Return the (conservative) maximum heap alignment for any G1 heap
   static size_t conservative_max_heap_alignment();
 
+  // Does operations required after initialization has been done.
+  void post_initialize();
+
   // Initialize weak reference processing.
-  virtual void ref_processing_init();
+  void ref_processing_init();
 
   // Explicitly import set_par_threads into this scope
-  using SharedHeap::set_par_threads;
+  using CollectedHeap::set_par_threads;
   // Set _n_par_threads according to a policy TBD.
   void set_par_threads();
 
@@ -1251,10 +1264,6 @@
 
   // Iteration functions.
 
-  // Iterate over all the ref-containing fields of all objects, calling
-  // "cl.do_oop" on each.
-  virtual void oop_iterate(ExtendedOopClosure* cl);
-
   // Iterate over all objects, calling "cl.do_object" on each.
   virtual void object_iterate(ObjectClosure* cl);
 
@@ -1262,9 +1271,6 @@
     object_iterate(cl);
   }
 
-  // Iterate over all spaces in use in the heap, in ascending address order.
-  virtual void space_iterate(SpaceClosure* cl);
-
   // Iterate over heap regions, in address order, terminating the
   // iteration early if the "doHeapRegion" method returns "true".
   void heap_region_iterate(HeapRegionClosure* blk) const;
@@ -1307,10 +1313,6 @@
 
   HeapRegion* next_compaction_region(const HeapRegion* from) const;
 
-  // A CollectedHeap will contain some number of spaces.  This finds the
-  // space containing a given address, or else returns NULL.
-  virtual Space* space_containing(const void* addr) const;
-
   // Returns the HeapRegion that contains addr. addr must not be NULL.
   template <class T>
   inline HeapRegion* heap_region_containing_raw(const T addr) const;
@@ -1344,9 +1346,6 @@
   // the block is an object.
   virtual bool block_is_obj(const HeapWord* addr) const;
 
-  // Does this heap support heap inspection? (+PrintClassHistogram)
-  virtual bool supports_heap_inspection() const { return true; }
-
   // Section on thread-local allocation buffers (TLABs)
   // See CollectedHeap for semantics.
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectedHeap.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -352,20 +352,30 @@
   return is_obj_ill(obj, heap_region_containing(obj));
 }
 
+inline void G1CollectedHeap::set_humongous_reclaim_candidate(uint region, bool value) {
+  assert(_hrm.at(region)->is_starts_humongous(), "Must start a humongous object");
+  _humongous_reclaim_candidates.set_candidate(region, value);
+}
+
+inline bool G1CollectedHeap::is_humongous_reclaim_candidate(uint region) {
+  assert(_hrm.at(region)->is_starts_humongous(), "Must start a humongous object");
+  return _humongous_reclaim_candidates.is_candidate(region);
+}
+
 inline void G1CollectedHeap::set_humongous_is_live(oop obj) {
   uint region = addr_to_region((HeapWord*)obj);
-  // We not only set the "live" flag in the humongous_is_live table, but also
+  // Clear the flag in the humongous_reclaim_candidates table.  Also
   // reset the entry in the _in_cset_fast_test table so that subsequent references
   // to the same humongous object do not go into the slow path again.
   // This is racy, as multiple threads may at the same time enter here, but this
   // is benign.
-  // During collection we only ever set the "live" flag, and only ever clear the
+  // During collection we only ever clear the "candidate" flag, and only ever clear the
   // entry in the in_cset_fast_table.
   // We only ever evaluate the contents of these tables (in the VM thread) after
   // having synchronized the worker threads with the VM thread, or in the same
   // thread (i.e. within the VM thread).
-  if (!_humongous_is_live.is_live(region)) {
-    _humongous_is_live.set_live(region);
+  if (is_humongous_reclaim_candidate(region)) {
+    set_humongous_reclaim_candidate(region, false);
     _in_cset_fast_test.clear_humongous(region);
   }
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1CollectorPolicy.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1460,7 +1460,7 @@
   _max_survivor_regions = (uint) ceil(max_survivor_regions_d);
 
   _tenuring_threshold = _survivors_age_table.compute_tenuring_threshold(
-        HeapRegion::GrainWords * _max_survivor_regions);
+        HeapRegion::GrainWords * _max_survivor_regions, counters());
 }
 
 bool G1CollectorPolicy::force_initial_mark_if_outside_cycle(
--- a/hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013, 2014 Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2013, 2015 Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -263,7 +263,6 @@
   _gc_par_phases[SystemDictionaryRoots] = new WorkerDataArray<double>(max_gc_threads, "SystemDictionary Roots (ms)", true, G1Log::LevelFinest, 3);
   _gc_par_phases[CLDGRoots] = new WorkerDataArray<double>(max_gc_threads, "CLDG Roots (ms)", true, G1Log::LevelFinest, 3);
   _gc_par_phases[JVMTIRoots] = new WorkerDataArray<double>(max_gc_threads, "JVMTI Roots (ms)", true, G1Log::LevelFinest, 3);
-  _gc_par_phases[CodeCacheRoots] = new WorkerDataArray<double>(max_gc_threads, "CodeCache Roots (ms)", true, G1Log::LevelFinest, 3);
   _gc_par_phases[CMRefRoots] = new WorkerDataArray<double>(max_gc_threads, "CM RefProcessor Roots (ms)", true, G1Log::LevelFinest, 3);
   _gc_par_phases[WaitForStrongCLD] = new WorkerDataArray<double>(max_gc_threads, "Wait For Strong CLD (ms)", true, G1Log::LevelFinest, 3);
   _gc_par_phases[WeakCLDRoots] = new WorkerDataArray<double>(max_gc_threads, "Weak CLD Roots (ms)", true, G1Log::LevelFinest, 3);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1GCPhaseTimes.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2013, 2014 Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2013, 2015 Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -51,7 +51,6 @@
     SystemDictionaryRoots,
     CLDGRoots,
     JVMTIRoots,
-    CodeCacheRoots,
     CMRefRoots,
     WaitForStrongCLD,
     WeakCLDRoots,
--- a/hotspot/src/share/vm/gc_implementation/g1/g1HotCardCache.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1HotCardCache.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -29,7 +29,7 @@
 #include "gc_implementation/g1/g1CardCounts.hpp"
 #include "memory/allocation.hpp"
 #include "runtime/safepoint.hpp"
-#include "runtime/thread.inline.hpp"
+#include "runtime/thread.hpp"
 #include "utilities/globalDefinitions.hpp"
 
 class DirtyCardQueue;
@@ -123,7 +123,7 @@
   // Resets the hot card cache and discards the entries.
   void reset_hot_cache() {
     assert(SafepointSynchronize::is_at_safepoint(), "Should be at a safepoint");
-    assert(Thread::current()->is_VM_thread(), "Current thread should be the VMthread");
+    assert(Thread::current_noinline()->is_VM_thread(), "Current thread should be the VMthread");
     if (default_use_cache()) {
         reset_hot_cache_internal();
     }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1MarkSweep.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1MarkSweep.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -61,9 +61,8 @@
                                       bool clear_all_softrefs) {
   assert(SafepointSynchronize::is_at_safepoint(), "must be at a safepoint");
 
-  SharedHeap* sh = SharedHeap::heap();
 #ifdef ASSERT
-  if (sh->collector_policy()->should_clear_all_soft_refs()) {
+  if (G1CollectedHeap::heap()->collector_policy()->should_clear_all_soft_refs()) {
     assert(clear_all_softrefs, "Policy should have been checked earler");
   }
 #endif
@@ -102,11 +101,6 @@
   BiasedLocking::restore_marks();
   GenMarkSweep::deallocate_stacks();
 
-  // "free at last gc" is calculated from these.
-  // CHF: cheating for now!!!
-  //  Universe::set_heap_capacity_at_last_gc(Universe::heap()->capacity());
-  //  Universe::set_heap_used_at_last_gc(Universe::heap()->used());
-
   CodeCache::gc_epilogue();
   JvmtiExport::gc_epilogue();
 
@@ -168,12 +162,12 @@
   Klass::clean_weak_klass_links(&GenMarkSweep::is_alive);
 
   // Delete entries for dead interned string and clean up unreferenced symbols in symbol table.
-  G1CollectedHeap::heap()->unlink_string_and_symbol_table(&GenMarkSweep::is_alive);
+  g1h->unlink_string_and_symbol_table(&GenMarkSweep::is_alive);
 
   if (VerifyDuringGC) {
     HandleMark hm;  // handle scope
     COMPILER2_PRESENT(DerivedPointerTableDeactivate dpt_deact);
-    Universe::heap()->prepare_for_verify();
+    g1h->prepare_for_verify();
     // Note: we can verify only the heap here. When an object is
     // marked, the previous value of the mark word (including
     // identity hash values, ages, etc) is preserved, and the mark
@@ -187,7 +181,7 @@
     if (!VerifySilently) {
       gclog_or_tty->print(" VerifyDuringGC:(full)[Verifying ");
     }
-    Universe::heap()->verify(VerifySilently, VerifyOption_G1UseMarkWord);
+    g1h->verify(VerifySilently, VerifyOption_G1UseMarkWord);
     if (!VerifySilently) {
       gclog_or_tty->print_cr("]");
     }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1MarkSweep.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1MarkSweep.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,7 +25,7 @@
 #ifndef SHARE_VM_GC_IMPLEMENTATION_G1_G1MARKSWEEP_HPP
 #define SHARE_VM_GC_IMPLEMENTATION_G1_G1MARKSWEEP_HPP
 
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
+#include "gc_implementation/g1/g1CollectedHeap.hpp"
 #include "gc_implementation/g1/heapRegion.hpp"
 #include "memory/genMarkSweep.hpp"
 #include "memory/generation.hpp"
--- a/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,9 +23,12 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_implementation/g1/g1_specialized_oop_closures.hpp"
 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
 #include "gc_implementation/g1/g1OopClosures.inline.hpp"
 #include "gc_implementation/g1/g1ParScanThreadState.hpp"
+#include "memory/iterator.inline.hpp"
+#include "utilities/stack.inline.hpp"
 
 G1ParCopyHelper::G1ParCopyHelper(G1CollectedHeap* g1,  G1ParScanThreadState* par_scan_state) :
   G1ParClosureSuper(g1, par_scan_state), _scanned_klass(NULL),
@@ -50,3 +53,6 @@
   assert(_worker_id < MAX2((uint)ParallelGCThreads, 1u),
          err_msg("The given worker id %u must be less than the number of threads %u", _worker_id, MAX2((uint)ParallelGCThreads, 1u)));
 }
+
+// Generate G1 specialized oop_oop_iterate functions.
+SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_G1(ALL_KLASS_OOP_OOP_ITERATE_DEFN)
--- a/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1OopClosures.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -172,7 +172,7 @@
   oopDesc* o = obj;
 #endif // CHECK_UNHANDLED_OOPS
   assert((intptr_t)o % MinObjAlignmentInBytes == 0, "not oop aligned");
-  assert(Universe::heap()->is_in_reserved(obj), "must be in heap");
+  assert(_g1->is_in_reserved(obj), "must be in heap");
 #endif // ASSERT
 
   assert(_from != NULL, "from region must be non-NULL");
--- a/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -44,37 +44,45 @@
 #endif
 #include "utilities/bitMap.inline.hpp"
 
-G1PageBasedVirtualSpace::G1PageBasedVirtualSpace() : _low_boundary(NULL),
-  _high_boundary(NULL), _committed(), _page_size(0), _special(false),
+G1PageBasedVirtualSpace::G1PageBasedVirtualSpace(ReservedSpace rs, size_t used_size, size_t page_size) :
+  _low_boundary(NULL), _high_boundary(NULL), _committed(), _page_size(0), _special(false),
   _dirty(), _executable(false) {
+  initialize_with_page_size(rs, used_size, page_size);
 }
 
-bool G1PageBasedVirtualSpace::initialize_with_granularity(ReservedSpace rs, size_t page_size) {
-  if (!rs.is_reserved()) {
-    return false;  // Allocation failed.
-  }
-  assert(_low_boundary == NULL, "VirtualSpace already initialized");
-  assert(page_size > 0, "Granularity must be non-zero.");
+void G1PageBasedVirtualSpace::initialize_with_page_size(ReservedSpace rs, size_t used_size, size_t page_size) {
+  guarantee(rs.is_reserved(), "Given reserved space must have been reserved already.");
+
+  vmassert(_low_boundary == NULL, "VirtualSpace already initialized");
+  vmassert(page_size > 0, "Page size must be non-zero.");
+
+  guarantee(is_ptr_aligned(rs.base(), page_size),
+            err_msg("Reserved space base " PTR_FORMAT " is not aligned to requested page size " SIZE_FORMAT, p2i(rs.base()), page_size));
+  guarantee(is_size_aligned(used_size, os::vm_page_size()),
+            err_msg("Given used reserved space size needs to be OS page size aligned (%d bytes) but is " SIZE_FORMAT, os::vm_page_size(), used_size));
+  guarantee(used_size <= rs.size(),
+            err_msg("Used size of reserved space " SIZE_FORMAT " bytes is smaller than reservation at " SIZE_FORMAT " bytes", used_size, rs.size()));
+  guarantee(is_size_aligned(rs.size(), page_size),
+            err_msg("Expected that the virtual space is size aligned, but " SIZE_FORMAT " is not aligned to page size " SIZE_FORMAT, rs.size(), page_size));
 
   _low_boundary  = rs.base();
-  _high_boundary = _low_boundary + rs.size();
+  _high_boundary = _low_boundary + used_size;
 
   _special = rs.special();
   _executable = rs.executable();
 
   _page_size = page_size;
 
-  assert(_committed.size() == 0, "virtual space initialized more than once");
-  uintx size_in_bits = rs.size() / page_size;
-  _committed.resize(size_in_bits, /* in_resource_area */ false);
+  vmassert(_committed.size() == 0, "virtual space initialized more than once");
+  BitMap::idx_t size_in_pages = rs.size() / page_size;
+  _committed.resize(size_in_pages, /* in_resource_area */ false);
   if (_special) {
-    _dirty.resize(size_in_bits, /* in_resource_area */ false);
+    _dirty.resize(size_in_pages, /* in_resource_area */ false);
   }
 
-  return true;
+  _tail_size = used_size % _page_size;
 }
 
-
 G1PageBasedVirtualSpace::~G1PageBasedVirtualSpace() {
   release();
 }
@@ -87,12 +95,18 @@
   _special                = false;
   _executable             = false;
   _page_size              = 0;
+  _tail_size              = 0;
   _committed.resize(0, false);
   _dirty.resize(0, false);
 }
 
 size_t G1PageBasedVirtualSpace::committed_size() const {
-  return _committed.count_one_bits() * _page_size;
+  size_t result = _committed.count_one_bits() * _page_size;
+  // The last page might not be in full.
+  if (is_last_page_partial() && _committed.at(_committed.size() - 1)) {
+    result -= _page_size - _tail_size;
+  }
+  return result;
 }
 
 size_t G1PageBasedVirtualSpace::reserved_size() const {
@@ -103,65 +117,134 @@
   return reserved_size() - committed_size();
 }
 
-uintptr_t G1PageBasedVirtualSpace::addr_to_page_index(char* addr) const {
+size_t G1PageBasedVirtualSpace::addr_to_page_index(char* addr) const {
   return (addr - _low_boundary) / _page_size;
 }
 
-bool G1PageBasedVirtualSpace::is_area_committed(uintptr_t start, size_t size_in_pages) const {
-  uintptr_t end = start + size_in_pages;
-  return _committed.get_next_zero_offset(start, end) >= end;
+bool G1PageBasedVirtualSpace::is_area_committed(size_t start_page, size_t size_in_pages) const {
+  size_t end_page = start_page + size_in_pages;
+  return _committed.get_next_zero_offset(start_page, end_page) >= end_page;
 }
 
-bool G1PageBasedVirtualSpace::is_area_uncommitted(uintptr_t start, size_t size_in_pages) const {
-  uintptr_t end = start + size_in_pages;
-  return _committed.get_next_one_offset(start, end) >= end;
+bool G1PageBasedVirtualSpace::is_area_uncommitted(size_t start_page, size_t size_in_pages) const {
+  size_t end_page = start_page + size_in_pages;
+  return _committed.get_next_one_offset(start_page, end_page) >= end_page;
 }
 
-char* G1PageBasedVirtualSpace::page_start(uintptr_t index) {
+char* G1PageBasedVirtualSpace::page_start(size_t index) const {
   return _low_boundary + index * _page_size;
 }
 
-size_t G1PageBasedVirtualSpace::byte_size_for_pages(size_t num) {
-  return num * _page_size;
+bool G1PageBasedVirtualSpace::is_after_last_page(size_t index) const {
+  guarantee(index <= _committed.size(),
+            err_msg("Given boundary page " SIZE_FORMAT " is beyond managed page count " SIZE_FORMAT, index, _committed.size()));
+  return index == _committed.size();
+}
+
+void G1PageBasedVirtualSpace::commit_preferred_pages(size_t start, size_t num_pages) {
+  vmassert(num_pages > 0, "No full pages to commit");
+  vmassert(start + num_pages <= _committed.size(),
+           err_msg("Tried to commit area from page " SIZE_FORMAT " to page " SIZE_FORMAT " "
+                   "that is outside of managed space of " SIZE_FORMAT " pages",
+                   start, start + num_pages, _committed.size()));
+
+  char* start_addr = page_start(start);
+  size_t size = num_pages * _page_size;
+
+  os::commit_memory_or_exit(start_addr, size, _page_size, _executable,
+                            err_msg("Failed to commit area from " PTR_FORMAT " to " PTR_FORMAT " of length " SIZE_FORMAT ".",
+                                    p2i(start_addr), p2i(start_addr + size), size));
+}
+
+void G1PageBasedVirtualSpace::commit_tail() {
+  vmassert(_tail_size > 0, "The size of the tail area must be > 0 when reaching here");
+
+  char* const aligned_end_address = (char*)align_ptr_down(_high_boundary, _page_size);
+  os::commit_memory_or_exit(aligned_end_address, _tail_size, os::vm_page_size(), _executable,
+                            err_msg("Failed to commit tail area from " PTR_FORMAT " to " PTR_FORMAT " of length " SIZE_FORMAT ".",
+                                    p2i(aligned_end_address), p2i(_high_boundary), _tail_size));
 }
 
-bool G1PageBasedVirtualSpace::commit(uintptr_t start, size_t size_in_pages) {
+void G1PageBasedVirtualSpace::commit_internal(size_t start_page, size_t end_page) {
+  guarantee(start_page < end_page,
+            err_msg("Given start page " SIZE_FORMAT " is larger or equal to end page " SIZE_FORMAT, start_page, end_page));
+  guarantee(end_page <= _committed.size(),
+            err_msg("Given end page " SIZE_FORMAT " is beyond end of managed page amount of " SIZE_FORMAT, end_page, _committed.size()));
+
+  size_t pages = end_page - start_page;
+  bool need_to_commit_tail = is_after_last_page(end_page) && is_last_page_partial();
+
+  // If we have to commit some (partial) tail area, decrease the amount of pages to avoid
+  // committing that in the full-page commit code.
+  if (need_to_commit_tail) {
+    pages--;
+  }
+
+  if (pages > 0) {
+    commit_preferred_pages(start_page, pages);
+  }
+
+  if (need_to_commit_tail) {
+    commit_tail();
+  }
+}
+
+char* G1PageBasedVirtualSpace::bounded_end_addr(size_t end_page) const {
+  return MIN2(_high_boundary, page_start(end_page));
+}
+
+void G1PageBasedVirtualSpace::pretouch_internal(size_t start_page, size_t end_page) {
+  guarantee(start_page < end_page,
+            err_msg("Given start page " SIZE_FORMAT " is larger or equal to end page " SIZE_FORMAT, start_page, end_page));
+
+  os::pretouch_memory(page_start(start_page), bounded_end_addr(end_page));
+}
+
+bool G1PageBasedVirtualSpace::commit(size_t start_page, size_t size_in_pages) {
   // We need to make sure to commit all pages covered by the given area.
-  guarantee(is_area_uncommitted(start, size_in_pages), "Specified area is not uncommitted");
+  guarantee(is_area_uncommitted(start_page, size_in_pages), "Specified area is not uncommitted");
 
   bool zero_filled = true;
-  uintptr_t end = start + size_in_pages;
+  size_t end_page = start_page + size_in_pages;
 
   if (_special) {
     // Check for dirty pages and update zero_filled if any found.
-    if (_dirty.get_next_one_offset(start,end) < end) {
+    if (_dirty.get_next_one_offset(start_page, end_page) < end_page) {
       zero_filled = false;
-      _dirty.clear_range(start, end);
+      _dirty.clear_range(start_page, end_page);
     }
   } else {
-    os::commit_memory_or_exit(page_start(start), byte_size_for_pages(size_in_pages), _executable,
-                              err_msg("Failed to commit pages from "SIZE_FORMAT" of length "SIZE_FORMAT, start, size_in_pages));
+    commit_internal(start_page, end_page);
   }
-  _committed.set_range(start, end);
+  _committed.set_range(start_page, end_page);
 
   if (AlwaysPreTouch) {
-    os::pretouch_memory(page_start(start), page_start(end));
+    pretouch_internal(start_page, end_page);
   }
   return zero_filled;
 }
 
-void G1PageBasedVirtualSpace::uncommit(uintptr_t start, size_t size_in_pages) {
-  guarantee(is_area_committed(start, size_in_pages), "checking");
+void G1PageBasedVirtualSpace::uncommit_internal(size_t start_page, size_t end_page) {
+  guarantee(start_page < end_page,
+            err_msg("Given start page " SIZE_FORMAT " is larger or equal to end page " SIZE_FORMAT, start_page, end_page));
 
+  char* start_addr = page_start(start_page);
+  os::uncommit_memory(start_addr, pointer_delta(bounded_end_addr(end_page), start_addr, sizeof(char)));
+}
+
+void G1PageBasedVirtualSpace::uncommit(size_t start_page, size_t size_in_pages) {
+  guarantee(is_area_committed(start_page, size_in_pages), "checking");
+
+  size_t end_page = start_page + size_in_pages;
   if (_special) {
     // Mark that memory is dirty. If committed again the memory might
     // need to be cleared explicitly.
-    _dirty.set_range(start, start + size_in_pages);
+    _dirty.set_range(start_page, end_page);
   } else {
-    os::uncommit_memory(page_start(start), byte_size_for_pages(size_in_pages));
+    uncommit_internal(start_page, end_page);
   }
 
-  _committed.clear_range(start, start + size_in_pages);
+  _committed.clear_range(start_page, end_page);
 }
 
 bool G1PageBasedVirtualSpace::contains(const void* p) const {
@@ -175,7 +258,8 @@
   out->cr();
   out->print_cr(" - committed: " SIZE_FORMAT, committed_size());
   out->print_cr(" - reserved:  " SIZE_FORMAT, reserved_size());
-  out->print_cr(" - [low_b, high_b]: [" INTPTR_FORMAT ", " INTPTR_FORMAT "]",  p2i(_low_boundary), p2i(_high_boundary));
+  out->print_cr(" - preferred page size: " SIZE_FORMAT, _page_size);
+  out->print_cr(" - [low_b, high_b]: [" PTR_FORMAT ", " PTR_FORMAT "]",  p2i(_low_boundary), p2i(_high_boundary));
 }
 
 void G1PageBasedVirtualSpace::print() {
--- a/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1PageBasedVirtualSpace.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -34,6 +34,12 @@
 // granularity.
 // (De-)Allocation requests are always OS page aligned by passing a page index
 // and multiples of pages.
+// For systems that only commits of memory in a given size (always greater than
+// page size) the base address is required to be aligned to that page size.
+// The actual size requested need not be aligned to that page size, but the size
+// of the reservation passed may be rounded up to this page size. Any fragment
+// (less than the page size) of the actual size at the tail of the request will
+// be committed using OS small pages.
 // The implementation gives an error when trying to commit or uncommit pages that
 // have already been committed or uncommitted.
 class G1PageBasedVirtualSpace VALUE_OBJ_CLASS_SPEC {
@@ -43,7 +49,11 @@
   char* _low_boundary;
   char* _high_boundary;
 
-  // The commit/uncommit granularity in bytes.
+  // The size of the tail in bytes of the handled space that needs to be committed
+  // using small pages.
+  size_t _tail_size;
+
+  // The preferred page size used for commit/uncommit in bytes.
   size_t _page_size;
 
   // Bitmap used for verification of commit/uncommit operations.
@@ -62,30 +72,55 @@
   // Indicates whether the committed space should be executable.
   bool _executable;
 
+  // Helper function for committing memory. Commit the given memory range by using
+  // _page_size pages as much as possible and the remainder with small sized pages.
+  void commit_internal(size_t start_page, size_t end_page);
+  // Commit num_pages pages of _page_size size starting from start. All argument
+  // checking has been performed.
+  void commit_preferred_pages(size_t start_page, size_t end_page);
+  // Commit space at the high end of the space that needs to be committed with small
+  // sized pages.
+  void commit_tail();
+
+  // Uncommit the given memory range.
+  void uncommit_internal(size_t start_page, size_t end_page);
+
+  // Pretouch the given memory range.
+  void pretouch_internal(size_t start_page, size_t end_page);
+
   // Returns the index of the page which contains the given address.
   uintptr_t  addr_to_page_index(char* addr) const;
   // Returns the address of the given page index.
-  char*  page_start(uintptr_t index);
-  // Returns the byte size of the given number of pages.
-  size_t byte_size_for_pages(size_t num);
+  char*  page_start(size_t index) const;
+
+  // Is the given page index the last page?
+  bool is_last_page(size_t index) const { return index == (_committed.size() - 1); }
+  // Is the given page index the first after last page?
+  bool is_after_last_page(size_t index) const;
+  // Is the last page only partially covered by this space?
+  bool is_last_page_partial() const { return !is_ptr_aligned(_high_boundary, _page_size); }
+  // Returns the end address of the given page bounded by the reserved space.
+  char* bounded_end_addr(size_t end_page) const;
 
   // Returns true if the entire area is backed by committed memory.
-  bool is_area_committed(uintptr_t start, size_t size_in_pages) const;
+  bool is_area_committed(size_t start_page, size_t size_in_pages) const;
   // Returns true if the entire area is not backed by committed memory.
-  bool is_area_uncommitted(uintptr_t start, size_t size_in_pages) const;
+  bool is_area_uncommitted(size_t start_page, size_t size_in_pages) const;
 
+  void initialize_with_page_size(ReservedSpace rs, size_t used_size, size_t page_size);
  public:
 
   // Commit the given area of pages starting at start being size_in_pages large.
   // Returns true if the given area is zero filled upon completion.
-  bool commit(uintptr_t start, size_t size_in_pages);
+  bool commit(size_t start_page, size_t size_in_pages);
 
   // Uncommit the given area of pages starting at start being size_in_pages large.
-  void uncommit(uintptr_t start, size_t size_in_pages);
+  void uncommit(size_t start_page, size_t size_in_pages);
 
-  // Initialization
-  G1PageBasedVirtualSpace();
-  bool initialize_with_granularity(ReservedSpace rs, size_t page_size);
+  // Initialize the given reserved space with the given base address and the size
+  // actually used.
+  // Prefer to commit in page_size chunks.
+  G1PageBasedVirtualSpace(ReservedSpace rs, size_t used_size, size_t page_size);
 
   // Destruction
   ~G1PageBasedVirtualSpace();
--- a/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,8 +26,10 @@
 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
 #include "gc_implementation/g1/g1OopClosures.inline.hpp"
 #include "gc_implementation/g1/g1ParScanThreadState.inline.hpp"
+#include "gc_implementation/g1/g1StringDedup.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/prefetch.inline.hpp"
+#include "utilities/stack.inline.hpp"
 
 G1ParScanThreadState::G1ParScanThreadState(G1CollectedHeap* g1h, uint queue_num, ReferenceProcessor* rp)
   : _g1h(g1h),
--- a/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1ParScanThreadState.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -63,7 +63,7 @@
   assert(has_partial_array_mask(p), "invariant");
   oop from_obj = clear_partial_array_mask(p);
 
-  assert(Universe::heap()->is_in_reserved(from_obj), "must be in heap.");
+  assert(_g1h->is_in_reserved(from_obj), "must be in heap.");
   assert(from_obj->is_objArray(), "must be obj array");
   objArrayOop from_obj_array = objArrayOop(from_obj);
   // The from-space object contains the real length.
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RegionToSpaceMapper.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RegionToSpaceMapper.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -31,17 +31,16 @@
 #include "utilities/bitMap.inline.hpp"
 
 G1RegionToSpaceMapper::G1RegionToSpaceMapper(ReservedSpace rs,
-                                             size_t commit_granularity,
+                                             size_t used_size,
+                                             size_t page_size,
                                              size_t region_granularity,
                                              MemoryType type) :
-  _storage(),
-  _commit_granularity(commit_granularity),
+  _storage(rs, used_size, page_size),
   _region_granularity(region_granularity),
   _listener(NULL),
   _commit_map() {
-  guarantee(is_power_of_2(commit_granularity), "must be");
+  guarantee(is_power_of_2(page_size), "must be");
   guarantee(is_power_of_2(region_granularity), "must be");
-  _storage.initialize_with_granularity(rs, commit_granularity);
 
   MemTracker::record_virtual_memory_type((address)rs.base(), type);
 }
@@ -55,25 +54,26 @@
 
  public:
   G1RegionsLargerThanCommitSizeMapper(ReservedSpace rs,
-                                      size_t os_commit_granularity,
+                                      size_t actual_size,
+                                      size_t page_size,
                                       size_t alloc_granularity,
                                       size_t commit_factor,
                                       MemoryType type) :
-     G1RegionToSpaceMapper(rs, os_commit_granularity, alloc_granularity, type),
-    _pages_per_region(alloc_granularity / (os_commit_granularity * commit_factor)) {
+    G1RegionToSpaceMapper(rs, actual_size, page_size, alloc_granularity, type),
+    _pages_per_region(alloc_granularity / (page_size * commit_factor)) {
 
-    guarantee(alloc_granularity >= os_commit_granularity, "allocation granularity smaller than commit granularity");
+    guarantee(alloc_granularity >= page_size, "allocation granularity smaller than commit granularity");
     _commit_map.resize(rs.size() * commit_factor / alloc_granularity, /* in_resource_area */ false);
   }
 
-  virtual void commit_regions(uintptr_t start_idx, size_t num_regions) {
-    bool zero_filled = _storage.commit(start_idx * _pages_per_region, num_regions * _pages_per_region);
+  virtual void commit_regions(uint start_idx, size_t num_regions) {
+    bool zero_filled = _storage.commit((size_t)start_idx * _pages_per_region, num_regions * _pages_per_region);
     _commit_map.set_range(start_idx, start_idx + num_regions);
     fire_on_commit(start_idx, num_regions, zero_filled);
   }
 
-  virtual void uncommit_regions(uintptr_t start_idx, size_t num_regions) {
-    _storage.uncommit(start_idx * _pages_per_region, num_regions * _pages_per_region);
+  virtual void uncommit_regions(uint start_idx, size_t num_regions) {
+    _storage.uncommit((size_t)start_idx * _pages_per_region, num_regions * _pages_per_region);
     _commit_map.clear_range(start_idx, start_idx + num_regions);
   }
 };
@@ -98,22 +98,23 @@
 
  public:
   G1RegionsSmallerThanCommitSizeMapper(ReservedSpace rs,
-                                       size_t os_commit_granularity,
+                                       size_t actual_size,
+                                       size_t page_size,
                                        size_t alloc_granularity,
                                        size_t commit_factor,
                                        MemoryType type) :
-     G1RegionToSpaceMapper(rs, os_commit_granularity, alloc_granularity, type),
-    _regions_per_page((os_commit_granularity * commit_factor) / alloc_granularity), _refcounts() {
+    G1RegionToSpaceMapper(rs, actual_size, page_size, alloc_granularity, type),
+    _regions_per_page((page_size * commit_factor) / alloc_granularity), _refcounts() {
 
-    guarantee((os_commit_granularity * commit_factor) >= alloc_granularity, "allocation granularity smaller than commit granularity");
-    _refcounts.initialize((HeapWord*)rs.base(), (HeapWord*)(rs.base() + rs.size()), os_commit_granularity);
+    guarantee((page_size * commit_factor) >= alloc_granularity, "allocation granularity smaller than commit granularity");
+    _refcounts.initialize((HeapWord*)rs.base(), (HeapWord*)(rs.base() + align_size_up(rs.size(), page_size)), page_size);
     _commit_map.resize(rs.size() * commit_factor / alloc_granularity, /* in_resource_area */ false);
   }
 
-  virtual void commit_regions(uintptr_t start_idx, size_t num_regions) {
-    for (uintptr_t i = start_idx; i < start_idx + num_regions; i++) {
-      assert(!_commit_map.at(i), err_msg("Trying to commit storage at region "INTPTR_FORMAT" that is already committed", i));
-      uintptr_t idx = region_idx_to_page_idx(i);
+  virtual void commit_regions(uint start_idx, size_t num_regions) {
+    for (uint i = start_idx; i < start_idx + num_regions; i++) {
+      assert(!_commit_map.at(i), err_msg("Trying to commit storage at region %u that is already committed", i));
+      size_t idx = region_idx_to_page_idx(i);
       uint old_refcount = _refcounts.get_by_index(idx);
       bool zero_filled = false;
       if (old_refcount == 0) {
@@ -125,10 +126,10 @@
     }
   }
 
-  virtual void uncommit_regions(uintptr_t start_idx, size_t num_regions) {
-    for (uintptr_t i = start_idx; i < start_idx + num_regions; i++) {
-      assert(_commit_map.at(i), err_msg("Trying to uncommit storage at region "INTPTR_FORMAT" that is not committed", i));
-      uintptr_t idx = region_idx_to_page_idx(i);
+  virtual void uncommit_regions(uint start_idx, size_t num_regions) {
+    for (uint i = start_idx; i < start_idx + num_regions; i++) {
+      assert(_commit_map.at(i), err_msg("Trying to uncommit storage at region %u that is not committed", i));
+      size_t idx = region_idx_to_page_idx(i);
       uint old_refcount = _refcounts.get_by_index(idx);
       assert(old_refcount > 0, "must be");
       if (old_refcount == 1) {
@@ -147,14 +148,15 @@
 }
 
 G1RegionToSpaceMapper* G1RegionToSpaceMapper::create_mapper(ReservedSpace rs,
-                                                            size_t os_commit_granularity,
+                                                            size_t actual_size,
+                                                            size_t page_size,
                                                             size_t region_granularity,
                                                             size_t commit_factor,
                                                             MemoryType type) {
 
-  if (region_granularity >= (os_commit_granularity * commit_factor)) {
-    return new G1RegionsLargerThanCommitSizeMapper(rs, os_commit_granularity, region_granularity, commit_factor, type);
+  if (region_granularity >= (page_size * commit_factor)) {
+    return new G1RegionsLargerThanCommitSizeMapper(rs, actual_size, page_size, region_granularity, commit_factor, type);
   } else {
-    return new G1RegionsSmallerThanCommitSizeMapper(rs, os_commit_granularity, region_granularity, commit_factor, type);
+    return new G1RegionsSmallerThanCommitSizeMapper(rs, actual_size, page_size, region_granularity, commit_factor, type);
   }
 }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RegionToSpaceMapper.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RegionToSpaceMapper.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -46,12 +46,12 @@
  protected:
   // Backing storage.
   G1PageBasedVirtualSpace _storage;
-  size_t _commit_granularity;
+
   size_t _region_granularity;
   // Mapping management
   BitMap _commit_map;
 
-  G1RegionToSpaceMapper(ReservedSpace rs, size_t commit_granularity, size_t region_granularity, MemoryType type);
+  G1RegionToSpaceMapper(ReservedSpace rs, size_t used_size, size_t page_size, size_t region_granularity, MemoryType type);
 
   void fire_on_commit(uint start_idx, size_t num_regions, bool zero_filled);
  public:
@@ -70,16 +70,20 @@
     return _commit_map.at(idx);
   }
 
-  virtual void commit_regions(uintptr_t start_idx, size_t num_regions = 1) = 0;
-  virtual void uncommit_regions(uintptr_t start_idx, size_t num_regions = 1) = 0;
+  virtual void commit_regions(uint start_idx, size_t num_regions = 1) = 0;
+  virtual void uncommit_regions(uint start_idx, size_t num_regions = 1) = 0;
 
   // Creates an appropriate G1RegionToSpaceMapper for the given parameters.
+  // The actual space to be used within the given reservation is given by actual_size.
+  // This is because some OSes need to round up the reservation size to guarantee
+  // alignment of page_size.
   // The byte_translation_factor defines how many bytes in a region correspond to
   // a single byte in the data structure this mapper is for.
   // Eg. in the card table, this value corresponds to the size a single card
-  // table entry corresponds to.
+  // table entry corresponds to in the heap.
   static G1RegionToSpaceMapper* create_mapper(ReservedSpace rs,
-                                              size_t os_commit_granularity,
+                                              size_t actual_size,
+                                              size_t page_size,
                                               size_t region_granularity,
                                               size_t byte_translation_factor,
                                               MemoryType type);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -38,6 +38,7 @@
 #include "oops/oop.inline.hpp"
 #include "utilities/globalDefinitions.hpp"
 #include "utilities/intHisto.hpp"
+#include "utilities/stack.inline.hpp"
 
 #define CARD_REPEAT_HISTO 0
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RemSet.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -57,7 +57,7 @@
   oopDesc* o = obj;
 #endif // CHECK_UNHANDLED_OOPS
   assert((intptr_t)o % MinObjAlignmentInBytes == 0, "not oop aligned");
-  assert(Universe::heap()->is_in_reserved(obj), "must be in heap");
+  assert(_g1->is_in_reserved(obj), "must be in heap");
 #endif // ASSERT
 
   assert(from == NULL || from->is_in_reserved(p), "p is not in from");
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RootProcessor.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RootProcessor.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -116,7 +116,7 @@
 G1RootProcessor::G1RootProcessor(G1CollectedHeap* g1h) :
     _g1h(g1h),
     _process_strong_tasks(new SubTasksDone(G1RP_PS_NumElements)),
-    _srs(g1h),
+    _srs(),
     _lock(Mutex::leaf, "G1 Root Scanning barrier lock", false, Monitor::_safepoint_check_never),
     _n_workers_discovered_strong_classes(0) {}
 
@@ -253,7 +253,8 @@
 
   {
     G1GCParPhaseTimesTracker x(phase_times, G1GCPhaseTimes::ThreadRoots, worker_i);
-    Threads::possibly_parallel_oops_do(strong_roots, thread_stack_clds, strong_code);
+    bool is_par = _g1h->n_par_threads() > 0;
+    Threads::possibly_parallel_oops_do(is_par, strong_roots, thread_stack_clds, strong_code);
   }
 }
 
@@ -323,10 +324,6 @@
 void G1RootProcessor::scan_remembered_sets(G1ParPushHeapRSClosure* scan_rs,
                                            OopClosure* scan_non_heap_weak_roots,
                                            uint worker_i) {
-  G1GCPhaseTimes* phase_times = _g1h->g1_policy()->phase_times();
-  G1GCParPhaseTimesTracker x(phase_times, G1GCPhaseTimes::CodeCacheRoots, worker_i);
-
-  // Now scan the complement of the collection set.
   G1CodeBlobClosure scavenge_cs_nmethods(scan_non_heap_weak_roots);
 
   _g1h->g1_rem_set()->oops_into_collection_set_do(scan_rs, &scavenge_cs_nmethods, worker_i);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1RootProcessor.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1RootProcessor.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,7 +26,7 @@
 #define SHARE_VM_GC_IMPLEMENTATION_G1_ROOTPROCESSOR_HPP
 
 #include "memory/allocation.hpp"
-#include "memory/sharedHeap.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "runtime/mutex.hpp"
 
 class CLDClosure;
@@ -46,7 +46,7 @@
 class G1RootProcessor : public StackObj {
   G1CollectedHeap* _g1h;
   SubTasksDone* _process_strong_tasks;
-  SharedHeap::StrongRootsScope _srs;
+  StrongRootsScope _srs;
 
   // Used to implement the Thread work barrier.
   Monitor _lock;
--- a/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1SATBCardTableModRefBS.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -206,7 +206,7 @@
   if (new_val == NULL) return;
   // Otherwise, log it.
   G1SATBCardTableLoggingModRefBS* g1_bs =
-    barrier_set_cast<G1SATBCardTableLoggingModRefBS>(Universe::heap()->barrier_set());
+    barrier_set_cast<G1SATBCardTableLoggingModRefBS>(G1CollectedHeap::heap()->barrier_set());
   g1_bs->write_ref_field_work(field, new_val);
 }
 
--- a/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupQueue.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupQueue.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -24,6 +24,7 @@
 
 #include "precompiled.hpp"
 #include "classfile/javaClasses.inline.hpp"
+#include "gc_implementation/g1/g1CollectedHeap.hpp"
 #include "gc_implementation/g1/g1StringDedup.hpp"
 #include "gc_implementation/g1/g1StringDedupQueue.hpp"
 #include "memory/gcLocker.hpp"
@@ -163,7 +164,7 @@
     while (!iter.is_empty()) {
       oop obj = iter.next();
       if (obj != NULL) {
-        guarantee(Universe::heap()->is_in_reserved(obj), "Object must be on the heap");
+        guarantee(G1CollectedHeap::heap()->is_in_reserved(obj), "Object must be on the heap");
         guarantee(!obj->is_forwarded(), "Object must not be forwarded");
         guarantee(java_lang_String::is_instance(obj), "Object must be a String");
       }
--- a/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupTable.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupTable.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -27,6 +27,7 @@
 #include "classfile/javaClasses.inline.hpp"
 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
+#include "gc_implementation/g1/g1StringDedup.hpp"
 #include "gc_implementation/g1/g1StringDedupTable.hpp"
 #include "memory/gcLocker.hpp"
 #include "memory/padded.inline.hpp"
@@ -519,7 +520,7 @@
     while (*entry != NULL) {
       typeArrayOop value = (*entry)->obj();
       guarantee(value != NULL, "Object must not be NULL");
-      guarantee(Universe::heap()->is_in_reserved(value), "Object must be on the heap");
+      guarantee(G1CollectedHeap::heap()->is_in_reserved(value), "Object must be on the heap");
       guarantee(!value->is_forwarded(), "Object must not be forwarded");
       guarantee(value->is_typeArray(), "Object must be a typeArrayOop");
       unsigned int hash = hash_code(value);
--- a/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupTable.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/g1StringDedupTable.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -29,6 +29,7 @@
 #include "runtime/mutexLocker.hpp"
 
 class G1StringDedupEntryCache;
+class G1StringDedupUnlinkOrOopsDoClosure;
 
 //
 // Table entry in the deduplication hashtable. Points weakly to the
--- a/hotspot/src/share/vm/gc_implementation/g1/heapRegion.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/heapRegion.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -31,7 +31,6 @@
 #include "gc_implementation/g1/survRateGroup.hpp"
 #include "gc_implementation/shared/ageTable.hpp"
 #include "gc_implementation/shared/spaceDecorator.hpp"
-#include "memory/space.inline.hpp"
 #include "memory/watermark.hpp"
 #include "utilities/macros.hpp"
 
--- a/hotspot/src/share/vm/gc_implementation/g1/heapRegionManager.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/heapRegionManager.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -330,8 +330,12 @@
         assert(!hrclaimer->is_region_claimed(ch_index),
                "Must not have been claimed yet because claiming of humongous continuation first claims the start region");
 
-        // There's no need to actually claim the continues humongous region, but we can do it in an assert as an extra precaution.
-        assert(hrclaimer->claim_region(ch_index), "We should always be able to claim the continuesHumongous part of the humongous object");
+        // Claim the region so no other worker tries to process the region. When a worker processes a
+        // starts_humongous region it may also process the associated continues_humongous regions.
+        // The continues_humongous regions can be changed to free regions. Unless this worker claims
+        // all of these regions, other workers might try claim and process these newly free regions.
+        bool claim_result = hrclaimer->claim_region(ch_index);
+        guarantee(claim_result, "We should always be able to claim the continuesHumongous part of the humongous object");
 
         bool res2 = blk->doHeapRegion(chr);
         if (res2) {
--- a/hotspot/src/share/vm/gc_implementation/g1/heapRegionSet.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/heapRegionSet.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -419,6 +419,7 @@
   ReservedSpace bot_rs(G1BlockOffsetSharedArray::compute_size(heap.word_size()));
   G1RegionToSpaceMapper* bot_storage =
     G1RegionToSpaceMapper::create_mapper(bot_rs,
+                                         bot_rs.size(),
                                          os::vm_page_size(),
                                          HeapRegion::GrainBytes,
                                          G1BlockOffsetSharedArray::N_bytes,
--- a/hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/satbQueue.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,8 +25,8 @@
 #include "precompiled.hpp"
 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
 #include "gc_implementation/g1/satbQueue.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/allocation.inline.hpp"
-#include "memory/sharedHeap.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/mutexLocker.hpp"
 #include "runtime/thread.hpp"
--- a/hotspot/src/share/vm/gc_implementation/g1/vmStructs_g1.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/vmStructs_g1.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,8 +26,8 @@
 #define SHARE_VM_GC_IMPLEMENTATION_G1_VMSTRUCTS_G1_HPP
 
 #include "gc_implementation/g1/heapRegion.hpp"
-#include "gc_implementation/g1/heapRegionManager.inline.hpp"
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
+#include "gc_implementation/g1/heapRegionManager.hpp"
+#include "gc_implementation/g1/g1CollectedHeap.hpp"
 
 #define VM_STRUCTS_G1(nonstatic_field, static_field)                          \
                                                                               \
@@ -70,7 +70,7 @@
                                                                               \
   declare_toplevel_type(G1HeapRegionTable)                                    \
                                                                               \
-  declare_type(G1CollectedHeap, SharedHeap)                                   \
+  declare_type(G1CollectedHeap, CollectedHeap)                                \
                                                                               \
   declare_type(G1OffsetTableContigSpace, CompactibleSpace)                    \
   declare_type(HeapRegion, G1OffsetTableContigSpace)                          \
--- a/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/g1/vm_operations_g1.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -225,15 +225,10 @@
 
 void VM_CGC_Operation::doit() {
   TraceCPUTime tcpu(G1Log::finer(), true, gclog_or_tty);
-  GCTraceTime t(_printGCMessage, G1Log::fine(), true, G1CollectedHeap::heap()->gc_timer_cm(), G1CollectedHeap::heap()->concurrent_mark()->concurrent_gc_id());
-  SharedHeap* sh = SharedHeap::heap();
-  // This could go away if CollectedHeap gave access to _gc_is_active...
-  if (sh != NULL) {
-    IsGCActiveMark x;
-    _cl->do_void();
-  } else {
-    _cl->do_void();
-  }
+  G1CollectedHeap* g1h = G1CollectedHeap::heap();
+  GCTraceTime t(_printGCMessage, G1Log::fine(), true, g1h->gc_timer_cm(), g1h->concurrent_mark()->concurrent_gc_id());
+  IsGCActiveMark x;
+  _cl->do_void();
 }
 
 bool VM_CGC_Operation::doit_prologue() {
@@ -244,14 +239,12 @@
   }
 
   Heap_lock->lock();
-  SharedHeap::heap()->_thread_holds_heap_lock_for_gc = true;
   return true;
 }
 
 void VM_CGC_Operation::doit_epilogue() {
   // Note the relative order of the unlocks must match that in
   // VM_GC_Operation::doit_epilogue()
-  SharedHeap::heap()->_thread_holds_heap_lock_for_gc = false;
   Heap_lock->unlock();
   if (_needs_pll) {
     release_and_notify_pending_list_lock();
--- a/hotspot/src/share/vm/gc_implementation/parNew/parCardTableModRefBS.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parNew/parCardTableModRefBS.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,12 +23,12 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/allocation.inline.hpp"
 #include "memory/cardTableModRefBS.hpp"
 #include "memory/cardTableRS.hpp"
-#include "memory/sharedHeap.hpp"
+#include "memory/genCollectedHeap.hpp"
 #include "memory/space.inline.hpp"
-#include "memory/universe.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/java.hpp"
 #include "runtime/mutexLocker.hpp"
@@ -449,7 +449,7 @@
   // Do a dirty read here. If we pass the conditional then take the rare
   // event lock and do the read again in case some other thread had already
   // succeeded and done the resize.
-  int cur_collection = Universe::heap()->total_collections();
+  int cur_collection = GenCollectedHeap::heap()->total_collections();
   if (_last_LNC_resizing_collection[i] != cur_collection) {
     MutexLocker x(ParGCRareEvent_lock);
     if (_last_LNC_resizing_collection[i] != cur_collection) {
--- a/hotspot/src/share/vm/gc_implementation/parNew/parNewGeneration.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parNew/parNewGeneration.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -42,7 +42,7 @@
 #include "memory/generation.hpp"
 #include "memory/referencePolicy.hpp"
 #include "memory/resourceArea.hpp"
-#include "memory/sharedHeap.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "memory/space.hpp"
 #include "oops/objArrayOop.hpp"
 #include "oops/oop.inline.hpp"
@@ -53,6 +53,7 @@
 #include "runtime/thread.inline.hpp"
 #include "utilities/copy.hpp"
 #include "utilities/globalDefinitions.hpp"
+#include "utilities/stack.inline.hpp"
 #include "utilities/workgroup.hpp"
 
 #ifdef _MSC_VER
@@ -117,7 +118,7 @@
 void ParScanThreadState::scan_partial_array_and_push_remainder(oop old) {
   assert(old->is_objArray(), "must be obj array");
   assert(old->is_forwarded(), "must be forwarded");
-  assert(Universe::heap()->is_in_reserved(old), "must be in heap.");
+  assert(GenCollectedHeap::heap()->is_in_reserved(old), "must be in heap.");
   assert(!old_gen()->is_in(old), "must be in young generation.");
 
   objArrayOop obj = objArrayOop(old->forwardee());
@@ -199,9 +200,9 @@
   for (size_t i = 0; i != num_take_elems; i++) {
     oop cur = of_stack->pop();
     oop obj_to_push = cur->forwardee();
-    assert(Universe::heap()->is_in_reserved(cur), "Should be in heap");
+    assert(GenCollectedHeap::heap()->is_in_reserved(cur), "Should be in heap");
     assert(!old_gen()->is_in_reserved(cur), "Should be in young gen");
-    assert(Universe::heap()->is_in_reserved(obj_to_push), "Should be in heap");
+    assert(GenCollectedHeap::heap()->is_in_reserved(obj_to_push), "Should be in heap");
     if (should_be_partially_scanned(obj_to_push, cur)) {
       assert(arrayOop(cur)->length() == 0, "entire array remaining to be scanned");
       obj_to_push = cur;
@@ -596,8 +597,6 @@
   // and handle marks.
   ResourceMark rm;
   HandleMark hm;
-  // We would need multiple old-gen queues otherwise.
-  assert(gch->n_gens() == 2, "Par young collection currently only works with one older gen.");
 
   ParScanThreadState& par_scan_state = _state_set->thread_state(worker_id);
   assert(_state_set->is_valid(worker_id), "Should not have been called");
@@ -697,7 +696,7 @@
 
   _par_cl->do_oop_nv(p);
 
-  if (Universe::heap()->is_in_reserved(p)) {
+  if (GenCollectedHeap::heap()->is_in_reserved(p)) {
     oop obj = oopDesc::load_decode_heap_oop_not_null(p);
     _rs->write_ref_field_gc_par(p, obj);
   }
@@ -724,7 +723,7 @@
 
   _cl->do_oop_nv(p);
 
-  if (Universe::heap()->is_in_reserved(p)) {
+  if (GenCollectedHeap::heap()->is_in_reserved(p)) {
     oop obj = oopDesc::load_decode_heap_oop_not_null(p);
     _rs->write_ref_field_gc_par(p, obj);
   }
@@ -823,8 +822,6 @@
 void ParNewRefProcTaskExecutor::execute(ProcessTask& task)
 {
   GenCollectedHeap* gch = GenCollectedHeap::heap();
-  assert(gch->kind() == CollectedHeap::GenCollectedHeap,
-         "not a generational heap");
   FlexibleWorkGang* workers = gch->workers();
   assert(workers != NULL, "Need parallel worker threads.");
   _state_set.reset(workers->active_workers(), _generation.promotion_failed());
@@ -899,7 +896,7 @@
     _gc_tracer.report_promotion_failed(_promotion_failed_info);
   }
   // Reset the PromotionFailureALot counters.
-  NOT_PRODUCT(Universe::heap()->reset_promotion_should_fail();)
+  NOT_PRODUCT(gch->reset_promotion_should_fail();)
 }
 
 void ParNewGeneration::collect(bool   full,
@@ -912,8 +909,6 @@
 
   _gc_timer->register_gc_start();
 
-  assert(gch->kind() == CollectedHeap::GenCollectedHeap,
-    "not a CMS generational heap");
   AdaptiveSizePolicy* size_policy = gch->gen_policy()->size_policy();
   FlexibleWorkGang* workers = gch->workers();
   assert(workers != NULL, "Need workgang for parallel work");
@@ -922,8 +917,6 @@
                                    workers->active_workers(),
                                    Threads::number_of_non_daemon_threads());
   workers->set_active_workers(active_workers);
-  assert(gch->n_gens() == 2,
-         "Par collection currently only works with single older gen.");
   _old_gen = gch->old_gen();
 
   // If the next generation is too full to accommodate worst-case promotion
@@ -974,10 +967,10 @@
   // in the multi-threaded case, but we special-case n=1 here to get
   // repeatable measurements of the 1-thread overhead of the parallel code.
   if (n_workers > 1) {
-    GenCollectedHeap::StrongRootsScope srs(gch);
+    StrongRootsScope srs;
     workers->run_task(&tsk);
   } else {
-    GenCollectedHeap::StrongRootsScope srs(gch);
+    StrongRootsScope srs;
     tsk.work(0);
   }
   thread_state_set.reset(0 /* Bad value in debug if not reset */,
@@ -1194,7 +1187,7 @@
   } else {
     // Is in to-space; do copying ourselves.
     Copy::aligned_disjoint_words((HeapWord*)old, (HeapWord*)new_obj, sz);
-    assert(Universe::heap()->is_in_reserved(new_obj), "illegal forwarding pointer value.");
+    assert(GenCollectedHeap::heap()->is_in_reserved(new_obj), "illegal forwarding pointer value.");
     forward_ptr = old->forward_to_atomic(new_obj);
     // Restore the mark word copied above.
     new_obj->set_mark(m);
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/gc_implementation/parNew/parOopClosures.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "memory/iterator.inline.hpp"
+#include "memory/specialized_oop_closures.hpp"
+#include "gc_implementation/parNew/parOopClosures.inline.hpp"
+
+// Generate ParNew specialized oop_oop_iterate functions.
+SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_P(ALL_KLASS_OOP_OOP_ITERATE_DEFN);
--- a/hotspot/src/share/vm/gc_implementation/parNew/parOopClosures.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parNew/parOopClosures.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -70,7 +70,7 @@
 inline void ParScanClosure::do_oop_work(T* p,
                                         bool gc_barrier,
                                         bool root_scan) {
-  assert((!Universe::heap()->is_in_reserved(p) ||
+  assert((!GenCollectedHeap::heap()->is_in_reserved(p) ||
           generation()->is_in_reserved(p))
          && (generation()->level() == 0 || gc_barrier),
          "The gen must be right, and we must be doing the barrier "
@@ -82,7 +82,7 @@
 #ifndef PRODUCT
       if (_g->to()->is_in_reserved(obj)) {
         tty->print_cr("Scanning field (" PTR_FORMAT ") twice?", p2i(p));
-        GenCollectedHeap* gch =  (GenCollectedHeap*)Universe::heap();
+        GenCollectedHeap* gch = GenCollectedHeap::heap();
         Space* sp = gch->space_containing(p);
         oop obj = oop(sp->block_start(p));
         assert((HeapWord*)obj < (HeapWord*)p, "Error");
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/asPSOldGen.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/asPSOldGen.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -89,7 +89,7 @@
   assert(virtual_space()->is_aligned(gen_size_limit()), "not aligned");
   assert(gen_size_limit() >= virtual_space()->committed_size(), "bad gen size");
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   size_t result =  gen_size_limit() - virtual_space()->committed_size();
   size_t result_aligned = align_size_down(result, heap->generation_alignment());
   return result_aligned;
@@ -101,7 +101,7 @@
     return uncommitted_bytes;
   }
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t gen_alignment = heap->generation_alignment();
   PSAdaptiveSizePolicy* policy = heap->size_policy();
   const size_t working_size =
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/asPSYoungGen.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/asPSYoungGen.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -73,7 +73,7 @@
   size_t current_committed_size = virtual_space()->committed_size();
   assert((gen_size_limit() >= current_committed_size),
     "generation size limit is wrong");
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   size_t result =  gen_size_limit() - current_committed_size;
   size_t result_aligned = align_size_down(result, heap->generation_alignment());
   return result_aligned;
@@ -91,7 +91,7 @@
 
   if (eden_space()->is_empty()) {
     // Respect the minimum size for eden and for the young gen as a whole.
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     const size_t eden_alignment = heap->space_alignment();
     const size_t gen_alignment = heap->generation_alignment();
 
@@ -128,7 +128,7 @@
 // If to_space is below from_space, to_space is not considered.
 // to_space can be.
 size_t ASPSYoungGen::available_to_live() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t alignment = heap->space_alignment();
 
   // Include any space that is committed but is not in eden.
@@ -292,7 +292,7 @@
 
   assert(eden_start < from_start, "Cannot push into from_space");
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t alignment = heap->space_alignment();
   const bool maintain_minimum =
     (requested_eden_size + 2 * requested_survivor_size) <= min_gen_size();
@@ -345,8 +345,6 @@
 
     // Does the optimal to-space overlap from-space?
     if (to_start < (char*)from_space()->end()) {
-      assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
       // Calculate the minimum offset possible for from_end
       size_t from_size =
         pointer_delta(from_space()->top(), from_start, sizeof(char));
@@ -509,9 +507,7 @@
   assert(from_space()->top() == old_from_top, "from top changed!");
 
   if (PrintAdaptiveSizePolicy) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     gclog_or_tty->print("AdaptiveSizePolicy::survivor space sizes: "
                   "collection: %d "
                   "(" SIZE_FORMAT ", " SIZE_FORMAT ") -> "
@@ -542,7 +538,7 @@
   }
   MemRegion cmr((HeapWord*)virtual_space()->low(),
                 (HeapWord*)virtual_space()->high());
-  Universe::heap()->barrier_set()->resize_covered_region(cmr);
+  ParallelScavengeHeap::heap()->barrier_set()->resize_covered_region(cmr);
 
   space_invariants();
 }
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/cardTableExtension.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -76,9 +76,7 @@
 
  public:
   CheckForUnmarkedObjects() {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     _young_gen = heap->young_gen();
     _card_table = barrier_set_cast<CardTableExtension>(heap->barrier_set());
     // No point in asserting barrier set type here. Need to make CardTableExtension
@@ -325,9 +323,7 @@
 void CardTableExtension::verify_all_young_refs_imprecise() {
   CheckForUnmarkedObjects check;
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSOldGen* old_gen = heap->old_gen();
 
   old_gen->object_iterate(&check);
@@ -335,9 +331,7 @@
 
 // This should be called immediately after a scavenge, before mutators resume.
 void CardTableExtension::verify_all_young_refs_precise() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSOldGen* old_gen = heap->old_gen();
 
   CheckForPreciseMarks check(
@@ -351,7 +345,7 @@
 
 void CardTableExtension::verify_all_young_refs_precise_helper(MemRegion mr) {
   CardTableExtension* card_table =
-    barrier_set_cast<CardTableExtension>(Universe::heap()->barrier_set());
+    barrier_set_cast<CardTableExtension>(ParallelScavengeHeap::heap()->barrier_set());
 
   jbyte* bot = card_table->byte_for(mr.start());
   jbyte* top = card_table->byte_for(mr.end());
@@ -523,7 +517,7 @@
     cur_committed = new_committed;
   }
 #ifdef ASSERT
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   assert(cur_committed.start() ==
     (HeapWord*) align_size_up((uintptr_t) cur_committed.start(),
                               os::vm_page_size()),
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -78,13 +78,7 @@
 
   CardTableExtension* const barrier_set = new CardTableExtension(reserved_region());
   barrier_set->initialize();
-  _barrier_set = barrier_set;
-  oopDesc::set_bs(_barrier_set);
-  if (_barrier_set == NULL) {
-    vm_shutdown_during_initialization(
-      "Could not reserve enough space for barrier set");
-    return JNI_ENOMEM;
-  }
+  set_barrier_set(barrier_set);
 
   // Make up the generations
   // Calculate the maximum size that a generation can grow.  This
@@ -95,6 +89,7 @@
   double max_gc_pause_sec = ((double) MaxGCPauseMillis)/1000.0;
   double max_gc_minor_pause_sec = ((double) MaxGCMinorPauseMillis)/1000.0;
 
+  _psh = this;
   _gens = new AdjoiningGenerations(heap_rs, _collector_policy, generation_alignment());
 
   _old_gen = _gens->old_gen();
@@ -120,7 +115,6 @@
   // initialize the policy counters - 2 collectors, 3 generations
   _gc_policy_counters =
     new PSGCAdaptivePolicyCounters("ParScav:MSC", 2, 3, _size_policy);
-  _psh = this;
 
   // Set up the GCTaskManager
   _gc_task_manager = GCTaskManager::create(ParallelGCThreads);
@@ -176,27 +170,11 @@
 }
 
 bool ParallelScavengeHeap::is_in(const void* p) const {
-  if (young_gen()->is_in(p)) {
-    return true;
-  }
-
-  if (old_gen()->is_in(p)) {
-    return true;
-  }
-
-  return false;
+  return young_gen()->is_in(p) || old_gen()->is_in(p);
 }
 
 bool ParallelScavengeHeap::is_in_reserved(const void* p) const {
-  if (young_gen()->is_in_reserved(p)) {
-    return true;
-  }
-
-  if (old_gen()->is_in_reserved(p)) {
-    return true;
-  }
-
-  return false;
+  return young_gen()->is_in_reserved(p) || old_gen()->is_in_reserved(p);
 }
 
 bool ParallelScavengeHeap::is_scavengable(const void* addr) {
@@ -265,7 +243,7 @@
     // total_collections() value!
     {
       MutexLocker ml(Heap_lock);
-      gc_count = Universe::heap()->total_collections();
+      gc_count = total_collections();
 
       result = young_gen()->allocate(size);
       if (result != NULL) {
@@ -315,8 +293,7 @@
       // This prevents us from looping until time out on requests that can
       // not be satisfied.
       if (op.prologue_succeeded()) {
-        assert(Universe::heap()->is_in_or_null(op.result()),
-          "result not in heap");
+        assert(is_in_or_null(op.result()), "result not in heap");
 
         // If GC was locked out during VM operation then retry allocation
         // and/or stall as necessary.
@@ -426,7 +403,7 @@
 HeapWord* ParallelScavengeHeap::failed_mem_allocate(size_t size) {
   assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
   assert(Thread::current() == (Thread*)VMThread::vm_thread(), "should be in vm thread");
-  assert(!Universe::heap()->is_gc_active(), "not reentrant");
+  assert(!is_gc_active(), "not reentrant");
   assert(!Heap_lock->owned_by_self(), "this thread should not own the Heap_lock");
 
   // We assume that allocation in eden will fail unless we collect.
@@ -514,18 +491,14 @@
   {
     MutexLocker ml(Heap_lock);
     // This value is guarded by the Heap_lock
-    gc_count      = Universe::heap()->total_collections();
-    full_gc_count = Universe::heap()->total_full_collections();
+    gc_count      = total_collections();
+    full_gc_count = total_full_collections();
   }
 
   VM_ParallelGCSystemGC op(gc_count, full_gc_count, cause);
   VMThread::execute(&op);
 }
 
-void ParallelScavengeHeap::oop_iterate(ExtendedOopClosure* cl) {
-  Unimplemented();
-}
-
 void ParallelScavengeHeap::object_iterate(ObjectClosure* cl) {
   young_gen()->object_iterate(cl);
   old_gen()->object_iterate(cl);
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -32,8 +32,9 @@
 #include "gc_implementation/parallelScavenge/psYoungGen.hpp"
 #include "gc_implementation/shared/gcPolicyCounters.hpp"
 #include "gc_implementation/shared/gcWhen.hpp"
-#include "gc_interface/collectedHeap.inline.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/collectorPolicy.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "utilities/ostream.hpp"
 
 class AdjoiningGenerations;
@@ -131,9 +132,6 @@
   // the young gen.
   virtual bool is_scavengable(const void* addr);
 
-  // Does this heap support heap inspection? (+PrintClassHistogram)
-  bool supports_heap_inspection() const { return true; }
-
   size_t max_capacity() const;
 
   // Whether p is in the allocated part of the heap
@@ -201,7 +199,6 @@
   // initializing stores to an object at this address.
   virtual bool can_elide_initializing_store_barrier(oop new_obj);
 
-  void oop_iterate(ExtendedOopClosure* cl);
   void object_iterate(ObjectClosure* cl);
   void safe_object_iterate(ObjectClosure* cl) { object_iterate(cl); }
 
@@ -238,7 +235,7 @@
   void gen_mangle_unused_area() PRODUCT_RETURN;
 
   // Call these in sequential code around the processing of strong roots.
-  class ParStrongRootsScope : public MarkingCodeBlobClosure::MarkScope {
+  class ParStrongRootsScope : public MarkScope {
    public:
     ParStrongRootsScope();
     ~ParStrongRootsScope();
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/pcTasks.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/pcTasks.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -41,13 +41,14 @@
 #include "runtime/thread.hpp"
 #include "runtime/vmThread.hpp"
 #include "services/management.hpp"
+#include "utilities/stack.inline.hpp"
 
 //
 // ThreadRootsMarkingTask
 //
 
 void ThreadRootsMarkingTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   ResourceMark rm;
 
@@ -78,7 +79,7 @@
 
 
 void MarkFromRootsTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   NOT_PRODUCT(GCTraceTime tm("MarkFromRootsTask",
     PrintGCDetails && TraceParallelOldGCTasks, true, NULL, PSParallelCompact::gc_tracer()->gc_id()));
@@ -149,7 +150,7 @@
 
 void RefProcTaskProxy::do_it(GCTaskManager* manager, uint which)
 {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   NOT_PRODUCT(GCTraceTime tm("RefProcTask",
     PrintGCDetails && TraceParallelOldGCTasks, true, NULL, PSParallelCompact::gc_tracer()->gc_id()));
@@ -167,7 +168,7 @@
 
 void RefProcTaskExecutor::execute(ProcessTask& task)
 {
-  ParallelScavengeHeap* heap = PSParallelCompact::gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   uint parallel_gc_threads = heap->gc_task_manager()->workers();
   uint active_gc_threads = heap->gc_task_manager()->active_workers();
   RegionTaskQueueSet* qset = ParCompactionManager::region_array();
@@ -188,7 +189,7 @@
 
 void RefProcTaskExecutor::execute(EnqueueTask& task)
 {
-  ParallelScavengeHeap* heap = PSParallelCompact::gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   uint parallel_gc_threads = heap->gc_task_manager()->workers();
   GCTaskQueue* q = GCTaskQueue::create();
   for(uint i=0; i<parallel_gc_threads; i++) {
@@ -205,7 +206,7 @@
   _terminator(t) {}
 
 void StealMarkingTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   NOT_PRODUCT(GCTraceTime tm("StealMarkingTask",
     PrintGCDetails && TraceParallelOldGCTasks, true, NULL, PSParallelCompact::gc_tracer()->gc_id()));
@@ -237,7 +238,7 @@
   _terminator(t) {}
 
 void StealRegionCompactionTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   NOT_PRODUCT(GCTraceTime tm("StealRegionCompactionTask",
     PrintGCDetails && TraceParallelOldGCTasks, true, NULL, PSParallelCompact::gc_tracer()->gc_id()));
@@ -319,7 +320,7 @@
 }
 
 void DrainStacksCompactionTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   NOT_PRODUCT(GCTraceTime tm("DrainStacksCompactionTask",
     PrintGCDetails && TraceParallelOldGCTasks, true, NULL, PSParallelCompact::gc_tracer()->gc_id()));
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -30,7 +30,10 @@
 #include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
 #include "gc_implementation/parallelScavenge/psCompactionManager.inline.hpp"
 #include "gc_implementation/parallelScavenge/psOldGen.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
+#include "gc_implementation/parallelScavenge/psParallelCompact.inline.hpp"
+#include "memory/iterator.inline.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
 #include "oops/objArrayKlass.inline.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/atomic.inline.hpp"
@@ -57,8 +60,7 @@
     _region_stack(NULL),
     _region_stack_index((uint)max_uintx) {
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   _old_gen = heap->old_gen();
   _start_array = old_gen()->start_array();
@@ -174,6 +176,142 @@
   return _manager_array[index];
 }
 
+void InstanceKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  assert(obj != NULL, "can't follow the content of NULL object");
+
+  PSParallelCompact::follow_klass(cm, this);
+  // Only mark the header and let the scan of the meta-data mark
+  // everything else.
+
+  PSParallelCompact::MarkAndPushClosure cl(cm);
+  InstanceKlass::oop_oop_iterate_oop_maps<true>(obj, &cl);
+}
+
+void InstanceMirrorKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  InstanceKlass::oop_pc_follow_contents(obj, cm);
+
+  // Follow the klass field in the mirror.
+  Klass* klass = java_lang_Class::as_Klass(obj);
+  if (klass != NULL) {
+    // An anonymous class doesn't have its own class loader, so the call
+    // to follow_klass will mark and push its java mirror instead of the
+    // class loader. When handling the java mirror for an anonymous class
+    // we need to make sure its class loader data is claimed, this is done
+    // by calling follow_class_loader explicitly. For non-anonymous classes
+    // the call to follow_class_loader is made when the class loader itself
+    // is handled.
+    if (klass->oop_is_instance() && InstanceKlass::cast(klass)->is_anonymous()) {
+      PSParallelCompact::follow_class_loader(cm, klass->class_loader_data());
+    } else {
+      PSParallelCompact::follow_klass(cm, klass);
+    }
+  } else {
+    // If klass is NULL then this a mirror for a primitive type.
+    // We don't have to follow them, since they are handled as strong
+    // roots in Universe::oops_do.
+    assert(java_lang_Class::is_primitive(obj), "Sanity check");
+  }
+
+  PSParallelCompact::MarkAndPushClosure cl(cm);
+  oop_oop_iterate_statics<true>(obj, &cl);
+}
+
+void InstanceClassLoaderKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  InstanceKlass::oop_pc_follow_contents(obj, cm);
+
+  ClassLoaderData * const loader_data = java_lang_ClassLoader::loader_data(obj);
+  if (loader_data != NULL) {
+    PSParallelCompact::follow_class_loader(cm, loader_data);
+  }
+}
+
+template <class T>
+static void oop_pc_follow_contents_specialized(InstanceRefKlass* klass, oop obj, ParCompactionManager* cm) {
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  T heap_oop = oopDesc::load_heap_oop(referent_addr);
+  debug_only(
+    if(TraceReferenceGC && PrintGCDetails) {
+      gclog_or_tty->print_cr("InstanceRefKlass::oop_pc_follow_contents " PTR_FORMAT, p2i(obj));
+    }
+  )
+  if (!oopDesc::is_null(heap_oop)) {
+    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);
+    if (PSParallelCompact::mark_bitmap()->is_unmarked(referent) &&
+        PSParallelCompact::ref_processor()->discover_reference(obj, klass->reference_type())) {
+      // reference already enqueued, referent will be traversed later
+      klass->InstanceKlass::oop_pc_follow_contents(obj, cm);
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("       Non NULL enqueued " PTR_FORMAT, p2i(obj));
+        }
+      )
+      return;
+    } else {
+      // treat referent as normal oop
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("       Non NULL normal " PTR_FORMAT, p2i(obj));
+        }
+      )
+      PSParallelCompact::mark_and_push(cm, referent_addr);
+    }
+  }
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
+    // Treat discovered as normal oop, if ref is not "active",
+    // i.e. if next is non-NULL.
+    T  next_oop = oopDesc::load_heap_oop(next_addr);
+    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
+      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("   Process discovered as normal "
+                                 PTR_FORMAT, p2i(discovered_addr));
+        }
+      )
+      PSParallelCompact::mark_and_push(cm, discovered_addr);
+    }
+  } else {
+#ifdef ASSERT
+    // In the case of older JDKs which do not use the discovered
+    // field for the pending list, an inactive ref (next != NULL)
+    // must always have a NULL discovered field.
+    T next = oopDesc::load_heap_oop(next_addr);
+    oop discovered = java_lang_ref_Reference::discovered(obj);
+    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
+           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
+                   p2i(obj)));
+#endif
+  }
+  PSParallelCompact::mark_and_push(cm, next_addr);
+  klass->InstanceKlass::oop_pc_follow_contents(obj, cm);
+}
+
+
+void InstanceRefKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  if (UseCompressedOops) {
+    oop_pc_follow_contents_specialized<narrowOop>(this, obj, cm);
+  } else {
+    oop_pc_follow_contents_specialized<oop>(this, obj, cm);
+  }
+}
+
+void ObjArrayKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  PSParallelCompact::follow_klass(cm, this);
+
+  if (UseCompressedOops) {
+    oop_pc_follow_contents_specialized<narrowOop>(objArrayOop(obj), 0, cm);
+  } else {
+    oop_pc_follow_contents_specialized<oop>(objArrayOop(obj), 0, cm);
+  }
+}
+
+void TypeArrayKlass::oop_pc_follow_contents(oop obj, ParCompactionManager* cm) {
+  assert(obj->is_typeArray(),"must be a type array");
+  // Performance tweak: We skip iterating over the klass pointer since we
+  // know that Universe::TypeArrayKlass never moves.
+}
+
 void ParCompactionManager::follow_marking_stacks() {
   do {
     // Drain the overflow stack first, to allow stealing from the marking stack.
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psCompactionManager.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,9 +26,11 @@
 #define SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSCOMPACTIONMANAGER_INLINE_HPP
 
 #include "gc_implementation/parallelScavenge/psCompactionManager.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
-#include "oops/objArrayKlass.inline.hpp"
-#include "oops/oop.pcgc.inline.hpp"
+#include "gc_implementation/parallelScavenge/psParallelCompact.inline.hpp"
+#include "oops/objArrayOop.hpp"
+#include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
 
 void ParCompactionManager::push_objarray(oop obj, size_t index)
 {
@@ -49,16 +51,42 @@
 }
 
 inline void ParCompactionManager::follow_contents(oop obj) {
-  obj->follow_contents(this);
+  assert(PSParallelCompact::mark_bitmap()->is_marked(obj), "should be marked");
+  obj->pc_follow_contents(this);
+}
+
+template <class T>
+inline void oop_pc_follow_contents_specialized(objArrayOop obj, int index, ParCompactionManager* cm) {
+  const size_t len = size_t(obj->length());
+  const size_t beg_index = size_t(index);
+  assert(beg_index < len || len == 0, "index too large");
+
+  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
+  const size_t end_index = beg_index + stride;
+  T* const base = (T*)obj->base();
+  T* const beg = base + beg_index;
+  T* const end = base + end_index;
+
+  // Push the non-NULL elements of the next stride on the marking stack.
+  for (T* e = beg; e < end; e++) {
+    PSParallelCompact::mark_and_push<T>(cm, e);
+  }
+
+  if (end_index < len) {
+    cm->push_objarray(obj, end_index); // Push the continuation.
+  }
 }
 
 inline void ParCompactionManager::follow_contents(objArrayOop obj, int index) {
-  ObjArrayKlass* k = (ObjArrayKlass*)obj->klass();
-  k->oop_follow_contents(this, obj, index);
+  if (UseCompressedOops) {
+    oop_pc_follow_contents_specialized<narrowOop>(obj, index, this);
+  } else {
+    oop_pc_follow_contents_specialized<oop>(obj, index, this);
+  }
 }
 
 inline void ParCompactionManager::update_contents(oop obj) {
-  obj->update_contents(this);
+  obj->pc_update_contents();
 }
 
 #endif // SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSCOMPACTIONMANAGER_INLINE_HPP
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweep.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -59,7 +59,7 @@
 CollectorCounters*  PSMarkSweep::_counters = NULL;
 
 void PSMarkSweep::initialize() {
-  MemRegion mr = Universe::heap()->reserved_region();
+  MemRegion mr = ParallelScavengeHeap::heap()->reserved_region();
   _ref_processor = new ReferenceProcessor(mr);     // a vanilla ref proc
   _counters = new CollectorCounters("PSMarkSweep", 1);
 }
@@ -81,9 +81,9 @@
 void PSMarkSweep::invoke(bool maximum_heap_compaction) {
   assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
   assert(Thread::current() == (Thread*)VMThread::vm_thread(), "should be in vm thread");
-  assert(!Universe::heap()->is_gc_active(), "not reentrant");
+  assert(!ParallelScavengeHeap::heap()->is_gc_active(), "not reentrant");
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   GCCause::Cause gc_cause = heap->gc_cause();
   PSAdaptiveSizePolicy* policy = heap->size_policy();
   IsGCActiveMark mark;
@@ -110,8 +110,7 @@
     return false;
   }
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   GCCause::Cause gc_cause = heap->gc_cause();
 
   _gc_timer->register_gc_start();
@@ -487,9 +486,7 @@
 }
 
 void PSMarkSweep::allocate_stacks() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
 
   MutableSpace* to_space = young_gen->to_space();
@@ -515,8 +512,7 @@
   GCTraceTime tm("phase 1", PrintGCDetails && Verbose, true, _gc_timer, _gc_tracer->gc_id());
   trace(" 1");
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   // Need to clear claim bits before the tracing starts.
   ClassLoaderDataGraph::clear_claimed_marks();
@@ -582,9 +578,7 @@
   // phase2, phase3 and phase4, but the ValidateMarkSweep live oops
   // tracking expects us to do so. See comment under phase4.
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSOldGen* old_gen = heap->old_gen();
 
   // Begin compacting into the old gen
@@ -606,9 +600,7 @@
   GCTraceTime tm("phase 3", PrintGCDetails && Verbose, true, _gc_timer, _gc_tracer->gc_id());
   trace("3");
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
   PSOldGen* old_gen = heap->old_gen();
 
@@ -651,9 +643,7 @@
 
   // All pointers are now adjusted, move objects accordingly
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
   PSOldGen* old_gen = heap->old_gen();
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweepDecorator.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psMarkSweepDecorator.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -38,15 +38,12 @@
 
 
 void PSMarkSweepDecorator::set_destination_decorator_tenured() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   _destination_decorator = heap->old_gen()->object_mark_sweep();
 }
 
 void PSMarkSweepDecorator::advance_destination_decorator() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   assert(_destination_decorator != NULL, "Sanity");
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psOldGen.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psOldGen.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -107,20 +107,22 @@
     SpaceMangler::mangle_region(cmr);
   }
 
-  Universe::heap()->barrier_set()->resize_covered_region(cmr);
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
+  BarrierSet* bs = heap->barrier_set();
 
-  CardTableModRefBS* _ct =
-    barrier_set_cast<CardTableModRefBS>(Universe::heap()->barrier_set());
+  bs->resize_covered_region(cmr);
+
+  CardTableModRefBS* ct = barrier_set_cast<CardTableModRefBS>(bs);
 
   // Verify that the start and end of this generation is the start of a card.
   // If this wasn't true, a single card could span more than one generation,
   // which would cause problems when we commit/uncommit memory, and when we
   // clear and dirty cards.
-  guarantee(_ct->is_card_aligned(_reserved.start()), "generation must be card aligned");
-  if (_reserved.end() != Universe::heap()->reserved_region().end()) {
+  guarantee(ct->is_card_aligned(_reserved.start()), "generation must be card aligned");
+  if (_reserved.end() != heap->reserved_region().end()) {
     // Don't check at the very end of the heap as we'll assert that we're probing off
     // the end if we try.
-    guarantee(_ct->is_card_aligned(_reserved.end()), "generation must be card aligned");
+    guarantee(ct->is_card_aligned(_reserved.end()), "generation must be card aligned");
   }
 
   //
@@ -161,8 +163,7 @@
 }
 
 void PSOldGen::precompact() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   // Reset start array first.
   start_array()->reset();
@@ -197,7 +198,7 @@
 
   // Allocations in the old generation need to be reported
   if (res != NULL) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     heap->size_policy()->tenured_allocation(word_size);
   }
 
@@ -376,8 +377,7 @@
   }
 
   if (PrintAdaptiveSizePolicy) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     gclog_or_tty->print_cr("AdaptiveSizePolicy::old generation size: "
                   "collection: %d "
                   "(" SIZE_FORMAT ") -> (" SIZE_FORMAT ") ",
@@ -397,7 +397,7 @@
   size_t new_word_size = new_memregion.word_size();
 
   start_array()->set_covered_region(new_memregion);
-  Universe::heap()->barrier_set()->resize_covered_region(new_memregion);
+  ParallelScavengeHeap::heap()->barrier_set()->resize_covered_region(new_memregion);
 
   // ALWAYS do this last!!
   object_space()->initialize(new_memregion,
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -34,7 +34,7 @@
 #include "gc_implementation/parallelScavenge/psMarkSweep.hpp"
 #include "gc_implementation/parallelScavenge/psMarkSweepDecorator.hpp"
 #include "gc_implementation/parallelScavenge/psOldGen.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
+#include "gc_implementation/parallelScavenge/psParallelCompact.inline.hpp"
 #include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
 #include "gc_implementation/parallelScavenge/psScavenge.hpp"
 #include "gc_implementation/parallelScavenge/psYoungGen.hpp"
@@ -48,7 +48,10 @@
 #include "memory/gcLocker.inline.hpp"
 #include "memory/referencePolicy.hpp"
 #include "memory/referenceProcessor.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
 #include "oops/methodData.hpp"
+#include "oops/objArrayKlass.inline.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/atomic.inline.hpp"
 #include "runtime/fprofiler.hpp"
@@ -745,7 +748,7 @@
 
 HeapWord* ParallelCompactData::calc_new_pointer(HeapWord* addr) {
   assert(addr != NULL, "Should detect NULL oop earlier");
-  assert(PSParallelCompact::gc_heap()->is_in(addr), "not in heap");
+  assert(ParallelScavengeHeap::heap()->is_in(addr), "not in heap");
   assert(PSParallelCompact::mark_bitmap()->is_marked(addr), "not marked");
 
   // Region covering the object.
@@ -823,16 +826,8 @@
 PSParallelCompact::AdjustPointerClosure PSParallelCompact::_adjust_pointer_closure;
 PSParallelCompact::AdjustKlassClosure PSParallelCompact::_adjust_klass_closure;
 
-void PSParallelCompact::AdjustPointerClosure::do_oop(oop* p)       { adjust_pointer(p); }
-void PSParallelCompact::AdjustPointerClosure::do_oop(narrowOop* p) { adjust_pointer(p); }
-
 void PSParallelCompact::FollowStackClosure::do_void() { _compaction_manager->follow_marking_stacks(); }
 
-void PSParallelCompact::MarkAndPushClosure::do_oop(oop* p)       {
-  mark_and_push(_compaction_manager, p);
-}
-void PSParallelCompact::MarkAndPushClosure::do_oop(narrowOop* p) { mark_and_push(_compaction_manager, p); }
-
 void PSParallelCompact::FollowKlassClosure::do_klass(Klass* klass) {
   klass->oops_do(_mark_and_push_closure);
 }
@@ -841,9 +836,7 @@
 }
 
 void PSParallelCompact::post_initialize() {
-  ParallelScavengeHeap* heap = gc_heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   MemRegion mr = heap->reserved_region();
   _ref_processor =
     new ReferenceProcessor(mr,            // span
@@ -860,8 +853,7 @@
 }
 
 bool PSParallelCompact::initialize() {
-  ParallelScavengeHeap* heap = gc_heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   MemRegion mr = heap->reserved_region();
 
   // Was the old gen get allocated successfully?
@@ -895,7 +887,7 @@
 {
   memset(&_space_info, 0, sizeof(_space_info));
 
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
 
   _space_info[old_space_id].set_space(heap->old_gen()->object_space());
@@ -978,7 +970,7 @@
   // promotion failure does not swap spaces) because an unknown number of minor
   // collections will have swapped the spaces an unknown number of times.
   GCTraceTime tm("pre compact", print_phases(), true, &_gc_timer, _gc_tracer.gc_id());
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   _space_info[from_space_id].set_space(heap->young_gen()->from_space());
   _space_info[to_space_id].set_space(heap->young_gen()->to_space());
 
@@ -1033,7 +1025,7 @@
   MutableSpace* const from_space = _space_info[from_space_id].space();
   MutableSpace* const to_space   = _space_info[to_space_id].space();
 
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   bool eden_empty = eden_space->is_empty();
   if (!eden_empty) {
     eden_empty = absorb_live_data_from_eden(heap->size_policy(),
@@ -1971,7 +1963,7 @@
   assert(Thread::current() == (Thread*)VMThread::vm_thread(),
          "should be in vm thread");
 
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   GCCause::Cause gc_cause = heap->gc_cause();
   assert(!heap->is_gc_active(), "not reentrant");
 
@@ -1999,7 +1991,7 @@
     return false;
   }
 
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   _gc_timer.register_gc_start();
   _gc_tracer.report_gc_start(heap->gc_cause(), _gc_timer.gc_start());
@@ -2352,7 +2344,7 @@
   // Recursively traverse all live objects and mark them
   GCTraceTime tm("marking phase", print_phases(), true, &_gc_timer, _gc_tracer.gc_id());
 
-  ParallelScavengeHeap* heap = gc_heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   uint parallel_gc_threads = heap->gc_task_manager()->workers();
   uint active_gc_threads = heap->gc_task_manager()->active_workers();
   TaskQueueSetSuper* qset = ParCompactionManager::region_array();
@@ -2692,8 +2684,7 @@
   // trace("5");
   GCTraceTime tm("compaction phase", print_phases(), true, &_gc_timer, _gc_tracer.gc_id());
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSOldGen* old_gen = heap->old_gen();
   old_gen->start_array()->reset();
   uint parallel_gc_threads = heap->gc_task_manager()->workers();
@@ -2844,7 +2835,7 @@
 // heap, last_space_id is returned.  In debug mode it expects the address to be
 // in the heap and asserts such.
 PSParallelCompact::SpaceId PSParallelCompact::space_id(HeapWord* addr) {
-  assert(Universe::heap()->is_in_reserved(addr), "addr not in the heap");
+  assert(ParallelScavengeHeap::heap()->is_in_reserved(addr), "addr not in the heap");
 
   for (unsigned int id = old_space_id; id < last_space_id; ++id) {
     if (_space_info[id].space()->contains(addr)) {
@@ -3338,6 +3329,71 @@
   update_state(words);
 }
 
+void InstanceKlass::oop_pc_update_pointers(oop obj) {
+  oop_oop_iterate_oop_maps<true>(obj, PSParallelCompact::adjust_pointer_closure());
+}
+
+void InstanceMirrorKlass::oop_pc_update_pointers(oop obj) {
+  InstanceKlass::oop_pc_update_pointers(obj);
+
+  oop_oop_iterate_statics<true>(obj, PSParallelCompact::adjust_pointer_closure());
+}
+
+void InstanceClassLoaderKlass::oop_pc_update_pointers(oop obj) {
+  InstanceKlass::oop_pc_update_pointers(obj);
+}
+
+#ifdef ASSERT
+template <class T> static void trace_reference_gc(const char *s, oop obj,
+                                                  T* referent_addr,
+                                                  T* next_addr,
+                                                  T* discovered_addr) {
+  if(TraceReferenceGC && PrintGCDetails) {
+    gclog_or_tty->print_cr("%s obj " PTR_FORMAT, s, p2i(obj));
+    gclog_or_tty->print_cr("     referent_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(referent_addr),
+                           referent_addr ? p2i(oopDesc::load_decode_heap_oop(referent_addr)) : NULL);
+    gclog_or_tty->print_cr("     next_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(next_addr),
+                           next_addr ? p2i(oopDesc::load_decode_heap_oop(next_addr)) : NULL);
+    gclog_or_tty->print_cr("     discovered_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(discovered_addr),
+                           discovered_addr ? p2i(oopDesc::load_decode_heap_oop(discovered_addr)) : NULL);
+  }
+}
+#endif
+
+template <class T>
+static void oop_pc_update_pointers_specialized(oop obj) {
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  PSParallelCompact::adjust_pointer(referent_addr);
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  PSParallelCompact::adjust_pointer(next_addr);
+  T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+  PSParallelCompact::adjust_pointer(discovered_addr);
+  debug_only(trace_reference_gc("InstanceRefKlass::oop_update_ptrs", obj,
+                                referent_addr, next_addr, discovered_addr);)
+}
+
+void InstanceRefKlass::oop_pc_update_pointers(oop obj) {
+  InstanceKlass::oop_pc_update_pointers(obj);
+
+  if (UseCompressedOops) {
+    oop_pc_update_pointers_specialized<narrowOop>(obj);
+  } else {
+    oop_pc_update_pointers_specialized<oop>(obj);
+  }
+}
+
+void ObjArrayKlass::oop_pc_update_pointers(oop obj) {
+  assert(obj->is_objArray(), "obj must be obj array");
+  oop_oop_iterate_elements<true>(objArrayOop(obj), PSParallelCompact::adjust_pointer_closure());
+}
+
+void TypeArrayKlass::oop_pc_update_pointers(oop obj) {
+  assert(obj->is_typeArray(),"must be a type array");
+}
+
 ParMarkBitMapClosure::IterationStatus
 MoveAndUpdateClosure::do_addr(HeapWord* addr, size_t words) {
   assert(destination() != NULL, "sanity");
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,11 +26,12 @@
 #define SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPARALLELCOMPACT_HPP
 
 #include "gc_implementation/parallelScavenge/objectStartArray.hpp"
+#include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
 #include "gc_implementation/parallelScavenge/parMarkBitMap.hpp"
 #include "gc_implementation/parallelScavenge/psCompactionManager.hpp"
 #include "gc_implementation/shared/collectorCounters.hpp"
 #include "gc_implementation/shared/mutableSpace.hpp"
-#include "memory/sharedHeap.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "oops/oop.hpp"
 
 class ParallelScavengeHeap;
@@ -951,12 +952,14 @@
     virtual void do_void();
   };
 
-  class AdjustPointerClosure: public OopClosure {
+  class AdjustPointerClosure: public ExtendedOopClosure {
    public:
+    template <typename T> void do_oop_nv(T* p);
     virtual void do_oop(oop* p);
     virtual void do_oop(narrowOop* p);
-    // do not walk from thread stacks to the code cache on this phase
-    virtual void do_code_blob(CodeBlob* cb) const { }
+
+    // This closure provides its own oop verification code.
+    debug_only(virtual bool should_verify_oops() { return false; })
   };
 
   class AdjustKlassClosure : public KlassClosure {
@@ -1139,13 +1142,18 @@
   static void reset_millis_since_last_gc();
 
  public:
-  class MarkAndPushClosure: public OopClosure {
+  class MarkAndPushClosure: public ExtendedOopClosure {
    private:
     ParCompactionManager* _compaction_manager;
    public:
     MarkAndPushClosure(ParCompactionManager* cm) : _compaction_manager(cm) { }
+
+    template <typename T> void do_oop_nv(T* p);
     virtual void do_oop(oop* p);
     virtual void do_oop(narrowOop* p);
+
+    // This closure provides its own oop verification code.
+    debug_only(virtual bool should_verify_oops() { return false; })
   };
 
   // The one and only place to start following the classes.
@@ -1161,11 +1169,6 @@
 
   PSParallelCompact();
 
-  // Convenient accessor for Universe::heap().
-  static ParallelScavengeHeap* gc_heap() {
-    return (ParallelScavengeHeap*)Universe::heap();
-  }
-
   static void invoke(bool maximum_heap_compaction);
   static bool invoke_no_policy(bool maximum_heap_compaction);
 
@@ -1177,7 +1180,9 @@
   static bool initialize();
 
   // Closure accessors
-  static OopClosure* adjust_pointer_closure()      { return (OopClosure*)&_adjust_pointer_closure; }
+  static PSParallelCompact::AdjustPointerClosure* adjust_pointer_closure() {
+    return &_adjust_pointer_closure;
+  }
   static KlassClosure* adjust_klass_closure()      { return (KlassClosure*)&_adjust_klass_closure; }
   static BoolObjectClosure* is_alive_closure()     { return (BoolObjectClosure*)&_is_alive_closure; }
 
@@ -1333,39 +1338,6 @@
 }
 
 template <class T>
-inline void PSParallelCompact::mark_and_push(ParCompactionManager* cm, T* p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
-    if (mark_bitmap()->is_unmarked(obj) && mark_obj(obj)) {
-      cm->push(obj);
-    }
-  }
-}
-
-template <class T>
-inline void PSParallelCompact::adjust_pointer(T* p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop obj     = oopDesc::decode_heap_oop_not_null(heap_oop);
-    oop new_obj = (oop)summary_data().calc_new_pointer(obj);
-    assert(new_obj != NULL,                    // is forwarding ptr?
-           "should be forwarded");
-    // Just always do the update unconditionally?
-    if (new_obj != NULL) {
-      assert(Universe::heap()->is_in_reserved(new_obj),
-             "should be in object space");
-      oopDesc::encode_store_heap_oop_not_null(p, new_obj);
-    }
-  }
-}
-
-inline void PSParallelCompact::follow_klass(ParCompactionManager* cm, Klass* klass) {
-  oop holder = klass->klass_holder();
-  PSParallelCompact::mark_and_push(cm, &holder);
-}
-
-template <class T>
 inline void PSParallelCompact::KeepAliveClosure::do_oop_work(T* p) {
   mark_and_push(_compaction_manager, p);
 }
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psParallelCompact.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,88 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPARALLELCOMPACT_INLINE_HPP
+#define SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPARALLELCOMPACT_INLINE_HPP
+
+#include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
+#include "gc_implementation/parallelScavenge/psCompactionManager.hpp"
+#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
+#include "gc_interface/collectedHeap.hpp"
+#include "oops/klass.hpp"
+#include "oops/oop.inline.hpp"
+
+template <typename T>
+inline void PSParallelCompact::mark_and_push(ParCompactionManager* cm, T* p) {
+  T heap_oop = oopDesc::load_heap_oop(p);
+  if (!oopDesc::is_null(heap_oop)) {
+    oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
+    assert(ParallelScavengeHeap::heap()->is_in(obj), "should be in heap");
+
+    if (mark_bitmap()->is_unmarked(obj) && mark_obj(obj)) {
+      cm->push(obj);
+    }
+  }
+}
+
+template <typename T>
+inline void PSParallelCompact::MarkAndPushClosure::do_oop_nv(T* p) {
+  mark_and_push(_compaction_manager, p);
+}
+
+inline void PSParallelCompact::MarkAndPushClosure::do_oop(oop* p)       { do_oop_nv(p); }
+inline void PSParallelCompact::MarkAndPushClosure::do_oop(narrowOop* p) { do_oop_nv(p); }
+
+inline void PSParallelCompact::follow_klass(ParCompactionManager* cm, Klass* klass) {
+  oop holder = klass->klass_holder();
+  mark_and_push(cm, &holder);
+}
+
+template <class T>
+inline void PSParallelCompact::adjust_pointer(T* p) {
+  T heap_oop = oopDesc::load_heap_oop(p);
+  if (!oopDesc::is_null(heap_oop)) {
+    oop obj     = oopDesc::decode_heap_oop_not_null(heap_oop);
+    assert(ParallelScavengeHeap::heap()->is_in(obj), "should be in heap");
+
+    oop new_obj = (oop)summary_data().calc_new_pointer(obj);
+    assert(new_obj != NULL,                    // is forwarding ptr?
+           "should be forwarded");
+    // Just always do the update unconditionally?
+    if (new_obj != NULL) {
+      assert(ParallelScavengeHeap::heap()->is_in_reserved(new_obj),
+             "should be in object space");
+      oopDesc::encode_store_heap_oop_not_null(p, new_obj);
+    }
+  }
+}
+
+template <typename T>
+void PSParallelCompact::AdjustPointerClosure::do_oop_nv(T* p) {
+  adjust_pointer(p);
+}
+
+inline void PSParallelCompact::AdjustPointerClosure::do_oop(oop* p)       { do_oop_nv(p); }
+inline void PSParallelCompact::AdjustPointerClosure::do_oop(narrowOop* p) { do_oop_nv(p); }
+
+#endif // SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPARALLELCOMPACT_INLINE_HPP
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionLAB.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionLAB.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -103,7 +103,7 @@
 }
 
 bool PSPromotionLAB::unallocate_object(HeapWord* obj, size_t obj_size) {
-  assert(Universe::heap()->is_in(obj), "Object outside heap");
+  assert(ParallelScavengeHeap::heap()->is_in(obj), "Object outside heap");
 
   if (contains(obj)) {
     HeapWord* object_end = obj + obj_size;
@@ -137,9 +137,7 @@
 #ifdef ASSERT
 
 bool PSYoungPromotionLAB::lab_is_valid(MemRegion lab) {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   MutableSpace* to_space = heap->young_gen()->to_space();
   MemRegion used = to_space->used_region();
   if (used.contains(lab)) {
@@ -150,10 +148,9 @@
 }
 
 bool PSOldPromotionLAB::lab_is_valid(MemRegion lab) {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
   assert(_start_array->covered_region().contains(lab), "Sanity");
 
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSOldGen* old_gen = heap->old_gen();
   MemRegion used = old_gen->object_space()->used_region();
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionLAB.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionLAB.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2002, 2012, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -26,7 +26,7 @@
 #define SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPROMOTIONLAB_HPP
 
 #include "gc_implementation/parallelScavenge/objectStartArray.hpp"
-#include "gc_interface/collectedHeap.inline.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/allocation.hpp"
 
 //
@@ -59,7 +59,7 @@
   // The shared initialize code invokes this.
   debug_only(virtual bool lab_is_valid(MemRegion lab) { return false; });
 
-  PSPromotionLAB() : _top(NULL), _bottom(NULL), _end(NULL) { }
+  PSPromotionLAB() : _top(NULL), _bottom(NULL), _end(NULL), _state(zero_size) { }
 
  public:
   // Filling and flushing.
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionManager.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionManager.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -32,6 +32,9 @@
 #include "memory/allocation.inline.hpp"
 #include "memory/memRegion.hpp"
 #include "memory/padded.inline.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
+#include "oops/objArrayKlass.inline.hpp"
 #include "oops/oop.inline.hpp"
 #include "utilities/stack.inline.hpp"
 
@@ -41,8 +44,7 @@
 MutableSpace*                  PSPromotionManager::_young_space = NULL;
 
 void PSPromotionManager::initialize() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   _old_gen = heap->old_gen();
   _young_space = heap->young_gen()->to_space();
@@ -85,8 +87,7 @@
 }
 
 void PSPromotionManager::pre_scavenge() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   _young_space = heap->young_gen()->to_space();
 
@@ -129,7 +130,7 @@
 void
 PSPromotionManager::print_taskqueue_stats(outputStream* const out) {
   out->print_cr("== GC Tasks Stats, GC %3d",
-                Universe::heap()->total_collections());
+                ParallelScavengeHeap::heap()->total_collections());
 
   TaskQueueStats totals;
   out->print("thr "); TaskQueueStats::print_header(1, out); out->cr();
@@ -157,8 +158,7 @@
 #endif // TASKQUEUE_STATS
 
 PSPromotionManager::PSPromotionManager() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   // We set the old lab's start array.
   _old_lab.set_start_array(old_gen()->start_array());
@@ -188,8 +188,7 @@
 
   // We need to get an assert in here to make sure the labs are always flushed.
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   // Do not prefill the LAB's, save heap wastage!
   HeapWord* lab_base = young_space()->top();
@@ -210,8 +209,7 @@
   totally_drain = totally_drain || _totally_drain;
 
 #ifdef ASSERT
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   MutableSpace* to_space = heap->young_gen()->to_space();
   MutableSpace* old_space = heap->old_gen()->object_space();
 #endif /* ASSERT */
@@ -308,6 +306,118 @@
   }
 }
 
+class PushContentsClosure : public ExtendedOopClosure {
+  PSPromotionManager* _pm;
+ public:
+  PushContentsClosure(PSPromotionManager* pm) : _pm(pm) {}
+
+  template <typename T> void do_oop_nv(T* p) {
+    if (PSScavenge::should_scavenge(p)) {
+      _pm->claim_or_forward_depth(p);
+    }
+  }
+
+  virtual void do_oop(oop* p)       { do_oop_nv(p); }
+  virtual void do_oop(narrowOop* p) { do_oop_nv(p); }
+
+  // Don't use the oop verification code in the oop_oop_iterate framework.
+  debug_only(virtual bool should_verify_oops() { return false; })
+};
+
+void InstanceKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+  PushContentsClosure cl(pm);
+  oop_oop_iterate_oop_maps_reverse<true>(obj, &cl);
+}
+
+void InstanceMirrorKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+    // Note that we don't have to follow the mirror -> klass pointer, since all
+    // klasses that are dirty will be scavenged when we iterate over the
+    // ClassLoaderData objects.
+
+  InstanceKlass::oop_ps_push_contents(obj, pm);
+
+  PushContentsClosure cl(pm);
+  oop_oop_iterate_statics<true>(obj, &cl);
+}
+
+void InstanceClassLoaderKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+  InstanceKlass::oop_ps_push_contents(obj, pm);
+
+  // This is called by the young collector. It will already have taken care of
+  // all class loader data. So, we don't have to follow the class loader ->
+  // class loader data link.
+}
+
+template <class T>
+static void oop_ps_push_contents_specialized(oop obj, InstanceRefKlass *klass, PSPromotionManager* pm) {
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  if (PSScavenge::should_scavenge(referent_addr)) {
+    ReferenceProcessor* rp = PSScavenge::reference_processor();
+    if (rp->discover_reference(obj, klass->reference_type())) {
+      // reference already enqueued, referent and next will be traversed later
+      klass->InstanceKlass::oop_ps_push_contents(obj, pm);
+      return;
+    } else {
+      // treat referent as normal oop
+      pm->claim_or_forward_depth(referent_addr);
+    }
+  }
+  // Treat discovered as normal oop, if ref is not "active",
+  // i.e. if next is non-NULL.
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
+    T  next_oop = oopDesc::load_heap_oop(next_addr);
+    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
+      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("   Process discovered as normal "
+                                 PTR_FORMAT, p2i(discovered_addr));
+        }
+      )
+      if (PSScavenge::should_scavenge(discovered_addr)) {
+        pm->claim_or_forward_depth(discovered_addr);
+      }
+    }
+  } else {
+#ifdef ASSERT
+    // In the case of older JDKs which do not use the discovered
+    // field for the pending list, an inactive ref (next != NULL)
+    // must always have a NULL discovered field.
+    oop next = oopDesc::load_decode_heap_oop(next_addr);
+    oop discovered = java_lang_ref_Reference::discovered(obj);
+    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
+           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
+                   p2i(obj)));
+#endif
+  }
+
+  // Treat next as normal oop;  next is a link in the reference queue.
+  if (PSScavenge::should_scavenge(next_addr)) {
+    pm->claim_or_forward_depth(next_addr);
+  }
+  klass->InstanceKlass::oop_ps_push_contents(obj, pm);
+}
+
+void InstanceRefKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+  if (UseCompressedOops) {
+    oop_ps_push_contents_specialized<narrowOop>(obj, this, pm);
+  } else {
+    oop_ps_push_contents_specialized<oop>(obj, this, pm);
+  }
+}
+
+void ObjArrayKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+  assert(obj->is_objArray(), "obj must be obj array");
+  PushContentsClosure cl(pm);
+  oop_oop_iterate_elements<true>(objArrayOop(obj), &cl);
+}
+
+void TypeArrayKlass::oop_ps_push_contents(oop obj, PSPromotionManager* pm) {
+  assert(obj->is_typeArray(),"must be a type array");
+  ShouldNotReachHere();
+}
+
 oop PSPromotionManager::oop_promotion_failed(oop obj, markOop obj_mark) {
   assert(_old_gen_is_full || PromotionFailureALot, "Sanity");
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionManager.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psPromotionManager.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,11 +25,12 @@
 #ifndef SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPROMOTIONMANAGER_INLINE_HPP
 #define SHARE_VM_GC_IMPLEMENTATION_PARALLELSCAVENGE_PSPROMOTIONMANAGER_INLINE_HPP
 
+#include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
 #include "gc_implementation/parallelScavenge/psOldGen.hpp"
 #include "gc_implementation/parallelScavenge/psPromotionManager.hpp"
 #include "gc_implementation/parallelScavenge/psPromotionLAB.inline.hpp"
 #include "gc_implementation/parallelScavenge/psScavenge.hpp"
-#include "oops/oop.psgc.inline.hpp"
+#include "oops/oop.inline.hpp"
 
 inline PSPromotionManager* PSPromotionManager::manager_array(int index) {
   assert(_manager_array != NULL, "access of NULL manager_array");
@@ -57,9 +58,7 @@
 template <class T>
 inline void PSPromotionManager::claim_or_forward_depth(T* p) {
   assert(should_scavenge(p, true), "revisiting object?");
-  assert(Universe::heap()->kind() == CollectedHeap::ParallelScavengeHeap,
-         "Sanity");
-  assert(Universe::heap()->is_in(p), "pointer outside heap");
+  assert(ParallelScavengeHeap::heap()->is_in(p), "pointer outside heap");
 
   claim_or_forward_internal_depth(p);
 }
@@ -92,7 +91,7 @@
 }
 
 inline void PSPromotionManager::push_contents(oop obj) {
-  obj->push_contents(this);
+  obj->ps_push_contents(this);
 }
 //
 // This method is pretty bulky. It would be nice to split it up
@@ -150,7 +149,7 @@
     // Otherwise try allocating obj tenured
     if (new_obj == NULL) {
 #ifndef PRODUCT
-      if (Universe::heap()->promotion_should_fail()) {
+      if (ParallelScavengeHeap::heap()->promotion_should_fail()) {
         return oop_promotion_failed(o, test_mark);
       }
 #endif  // #ifndef PRODUCT
@@ -296,7 +295,7 @@
   // that are outside the heap. These pointers are either from roots
   // or from metadata.
   if ((!PSScavenge::is_obj_in_young((HeapWord*)p)) &&
-      Universe::heap()->is_in_reserved(p)) {
+      ParallelScavengeHeap::heap()->is_in_reserved(p)) {
     if (PSScavenge::is_obj_in_young(new_obj)) {
       PSScavenge::card_table()->inline_write_ref_field_gc(p, new_obj);
     }
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -87,8 +87,7 @@
 
 public:
   PSKeepAliveClosure(PSPromotionManager* pm) : _promotion_manager(pm) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     _to_space = heap->young_gen()->to_space();
 
     assert(_promotion_manager != NULL, "Sanity");
@@ -218,11 +217,9 @@
 bool PSScavenge::invoke() {
   assert(SafepointSynchronize::is_at_safepoint(), "should be at safepoint");
   assert(Thread::current() == (Thread*)VMThread::vm_thread(), "should be in vm thread");
-  assert(!Universe::heap()->is_gc_active(), "not reentrant");
+  assert(!ParallelScavengeHeap::heap()->is_gc_active(), "not reentrant");
 
-  ParallelScavengeHeap* const heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* const heap = ParallelScavengeHeap::heap();
   PSAdaptiveSizePolicy* policy = heap->size_policy();
   IsGCActiveMark mark;
 
@@ -273,9 +270,8 @@
     return false;
   }
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   GCCause::Cause gc_cause = heap->gc_cause();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
 
   // Check for potential problems.
   if (!should_attempt_scavenge()) {
@@ -713,9 +709,7 @@
 // unforwarding markOops. It then restores any preserved mark oops,
 // and clears the _preserved_mark_stack.
 void PSScavenge::clean_up_failed_promotion() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
 
   {
@@ -742,7 +736,7 @@
   }
 
   // Reset the PromotionFailureALot counters.
-  NOT_PRODUCT(Universe::heap()->reset_promotion_should_fail();)
+  NOT_PRODUCT(heap->reset_promotion_should_fail();)
 }
 
 // This method is called whenever an attempt to promote an object
@@ -761,8 +755,7 @@
 }
 
 bool PSScavenge::should_attempt_scavenge() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSGCAdaptivePolicyCounters* counters = heap->gc_policy_counters();
 
   if (UsePerfData) {
@@ -838,9 +831,7 @@
                                                     MaxTenuringThreshold;
   }
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   PSYoungGen* young_gen = heap->young_gen();
   PSOldGen* old_gen = heap->old_gen();
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psScavenge.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2002, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2002, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -33,7 +33,7 @@
 #include "utilities/globalDefinitions.hpp"
 
 inline void PSScavenge::save_to_space_top_before_gc() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   _to_space_top_before_gc = heap->young_gen()->to_space()->top();
 }
 
@@ -56,7 +56,7 @@
 template <class T>
 inline bool PSScavenge::should_scavenge(T* p, bool check_to_space) {
   if (check_to_space) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     return should_scavenge(p, heap->young_gen()->to_space());
   }
   return should_scavenge(p);
@@ -97,7 +97,6 @@
     ParallelScavengeHeap* psh = ParallelScavengeHeap::heap();
     assert(!psh->is_in_reserved(p), "GC barrier needed");
     if (PSScavenge::should_scavenge(p)) {
-      assert(!Universe::heap()->is_in_reserved(p), "Not from meta-data?");
       assert(PSScavenge::should_scavenge(p, true), "revisiting object?");
 
       oop o = *p;
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psTasks.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psTasks.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -47,7 +47,7 @@
 //
 
 void ScavengeRootsTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   PSPromotionManager* pm = PSPromotionManager::gc_thread_promotion_manager(which);
   PSScavengeRootsClosure roots_closure(pm);
@@ -118,7 +118,7 @@
 //
 
 void ThreadRootsTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   PSPromotionManager* pm = PSPromotionManager::gc_thread_promotion_manager(which);
   PSScavengeRootsClosure roots_closure(pm);
@@ -143,7 +143,7 @@
   _terminator(t) {}
 
 void StealTask::do_it(GCTaskManager* manager, uint which) {
-  assert(Universe::heap()->is_gc_active(), "called outside gc");
+  assert(ParallelScavengeHeap::heap()->is_gc_active(), "called outside gc");
 
   PSPromotionManager* pm =
     PSPromotionManager::gc_thread_promotion_manager(which);
@@ -181,10 +181,8 @@
 
   {
     PSPromotionManager* pm = PSPromotionManager::gc_thread_promotion_manager(which);
-
-    assert(Universe::heap()->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
     CardTableExtension* card_table =
-      barrier_set_cast<CardTableExtension>(Universe::heap()->barrier_set());
+      barrier_set_cast<CardTableExtension>(ParallelScavengeHeap::heap()->barrier_set());
 
     card_table->scavenge_contents_parallel(_gen->start_array(),
                                            _gen->object_space(),
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/psYoungGen.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/psYoungGen.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -62,7 +62,7 @@
 
   MemRegion cmr((HeapWord*)virtual_space()->low(),
                 (HeapWord*)virtual_space()->high());
-  Universe::heap()->barrier_set()->resize_covered_region(cmr);
+  ParallelScavengeHeap::heap()->barrier_set()->resize_covered_region(cmr);
 
   if (ZapUnusedHeapArea) {
     // Mangle newly committed space immediately because it
@@ -103,7 +103,7 @@
                                            _max_gen_size, _virtual_space);
 
   // Compute maximum space sizes for performance counters
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   size_t alignment = heap->space_alignment();
   size_t size = virtual_space()->reserved_size();
 
@@ -153,8 +153,7 @@
 }
 
 void PSYoungGen::compute_initial_space_boundaries() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   // Compute sizes
   size_t alignment = heap->space_alignment();
@@ -208,7 +207,7 @@
 
 #ifndef PRODUCT
 void PSYoungGen::space_invariants() {
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t alignment = heap->space_alignment();
 
   // Currently, our eden size cannot shrink to zero
@@ -494,7 +493,7 @@
   char* to_start   = (char*)to_space()->bottom();
   char* to_end     = (char*)to_space()->end();
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t alignment = heap->space_alignment();
   const bool maintain_minimum =
     (requested_eden_size + 2 * requested_survivor_size) <= min_gen_size();
@@ -546,8 +545,6 @@
 
     // Does the optimal to-space overlap from-space?
     if (to_start < (char*)from_space()->end()) {
-      assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
       // Calculate the minimum offset possible for from_end
       size_t from_size = pointer_delta(from_space()->top(), from_start, sizeof(char));
 
@@ -708,9 +705,7 @@
   assert(from_space()->top() == old_from_top, "from top changed!");
 
   if (PrintAdaptiveSizePolicy) {
-    ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-    assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "Sanity");
-
+    ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
     gclog_or_tty->print("AdaptiveSizePolicy::survivor space sizes: "
                   "collection: %d "
                   "(" SIZE_FORMAT ", " SIZE_FORMAT ") -> "
@@ -843,7 +838,7 @@
 // from-space.
 size_t PSYoungGen::available_to_live() {
   size_t delta_in_survivor = 0;
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
   const size_t space_alignment = heap->space_alignment();
   const size_t gen_alignment = heap->generation_alignment();
 
@@ -927,7 +922,7 @@
 
   MemRegion cmr((HeapWord*)virtual_space()->low(),
                 (HeapWord*)virtual_space()->high());
-  Universe::heap()->barrier_set()->resize_covered_region(cmr);
+  ParallelScavengeHeap::heap()->barrier_set()->resize_covered_region(cmr);
   space_invariants();
 }
 
--- a/hotspot/src/share/vm/gc_implementation/parallelScavenge/vmPSOperations.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/parallelScavenge/vmPSOperations.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -41,8 +41,7 @@
 void VM_ParallelGCFailedAllocation::doit() {
   SvcGCMarker sgcm(SvcGCMarker::MINOR);
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap, "must be a ParallelScavengeHeap");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   GCCauseSetter gccs(heap, _gc_cause);
   _result = heap->failed_mem_allocate(_word_size);
@@ -63,9 +62,7 @@
 void VM_ParallelGCSystemGC::doit() {
   SvcGCMarker sgcm(SvcGCMarker::FULL);
 
-  ParallelScavengeHeap* heap = (ParallelScavengeHeap*)Universe::heap();
-  assert(heap->kind() == CollectedHeap::ParallelScavengeHeap,
-    "must be a ParallelScavengeHeap");
+  ParallelScavengeHeap* heap = ParallelScavengeHeap::heap();
 
   GCCauseSetter gccs(heap, _gc_cause);
   if (_gc_cause == GCCause::_gc_locker || _gc_cause == GCCause::_wb_young_gc
--- a/hotspot/src/share/vm/gc_implementation/shared/ageTable.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/ageTable.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,9 +25,9 @@
 #include "precompiled.hpp"
 #include "gc_implementation/shared/ageTable.hpp"
 #include "gc_implementation/shared/gcPolicyCounters.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/collectorPolicy.hpp"
 #include "memory/resourceArea.hpp"
-#include "memory/sharedHeap.hpp"
 #include "runtime/atomic.inline.hpp"
 #include "utilities/copy.hpp"
 
@@ -79,7 +79,7 @@
   }
 }
 
-uint ageTable::compute_tenuring_threshold(size_t survivor_capacity) {
+uint ageTable::compute_tenuring_threshold(size_t survivor_capacity, GCPolicyCounters* gc_counters) {
   size_t desired_survivor_size = (size_t)((((double) survivor_capacity)*TargetSurvivorRatio)/100);
   uint result;
 
@@ -126,9 +126,6 @@
       age++;
     }
     if (UsePerfData) {
-      SharedHeap* sh = SharedHeap::heap();
-      CollectorPolicy* policy = sh->collector_policy();
-      GCPolicyCounters* gc_counters = policy->counters();
       gc_counters->tenuring_threshold()->set_value(result);
       gc_counters->desired_survivor_size()->set_value(
         desired_survivor_size*oopSize);
--- a/hotspot/src/share/vm/gc_implementation/shared/ageTable.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/ageTable.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -29,6 +29,8 @@
 #include "oops/oop.hpp"
 #include "runtime/perfData.hpp"
 
+class GCPolicyCounters;
+
 /* Copyright (c) 1992-2009 Oracle and/or its affiliates, and Stanford University.
    See the LICENSE file for license information. */
 
@@ -69,7 +71,7 @@
   void merge_par(ageTable* subTable);
 
   // calculate new tenuring threshold based on age information
-  uint compute_tenuring_threshold(size_t survivor_capacity);
+  uint compute_tenuring_threshold(size_t survivor_capacity, GCPolicyCounters* gc_counters);
 
  private:
   PerfVariable* _perf_sizes[table_size];
--- a/hotspot/src/share/vm/gc_implementation/shared/markSweep.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/markSweep.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -28,6 +28,8 @@
 #include "gc_implementation/shared/gcTrace.hpp"
 #include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
 #include "oops/methodData.hpp"
 #include "oops/objArrayKlass.inline.hpp"
 #include "oops/oop.inline.hpp"
@@ -55,16 +57,183 @@
 CLDToOopClosure               MarkSweep::follow_cld_closure(&mark_and_push_closure);
 CLDToOopClosure               MarkSweep::adjust_cld_closure(&adjust_pointer_closure);
 
-void MarkSweep::MarkAndPushClosure::do_oop(oop* p)       { mark_and_push(p); }
-void MarkSweep::MarkAndPushClosure::do_oop(narrowOop* p) { mark_and_push(p); }
+template <typename T>
+void MarkSweep::MarkAndPushClosure::do_oop_nv(T* p)       { mark_and_push(p); }
+void MarkSweep::MarkAndPushClosure::do_oop(oop* p)        { do_oop_nv(p); }
+void MarkSweep::MarkAndPushClosure::do_oop(narrowOop* p)  { do_oop_nv(p); }
 
 void MarkSweep::follow_class_loader(ClassLoaderData* cld) {
   MarkSweep::follow_cld_closure.do_cld(cld);
 }
 
+void InstanceKlass::oop_ms_follow_contents(oop obj) {
+  assert(obj != NULL, "can't follow the content of NULL object");
+  MarkSweep::follow_klass(this);
+
+  oop_oop_iterate_oop_maps<true>(obj, &MarkSweep::mark_and_push_closure);
+}
+
+void InstanceMirrorKlass::oop_ms_follow_contents(oop obj) {
+  InstanceKlass::oop_ms_follow_contents(obj);
+
+  // Follow the klass field in the mirror
+  Klass* klass = java_lang_Class::as_Klass(obj);
+  if (klass != NULL) {
+    // An anonymous class doesn't have its own class loader, so the call
+    // to follow_klass will mark and push its java mirror instead of the
+    // class loader. When handling the java mirror for an anonymous class
+    // we need to make sure its class loader data is claimed, this is done
+    // by calling follow_class_loader explicitly. For non-anonymous classes
+    // the call to follow_class_loader is made when the class loader itself
+    // is handled.
+    if (klass->oop_is_instance() && InstanceKlass::cast(klass)->is_anonymous()) {
+      MarkSweep::follow_class_loader(klass->class_loader_data());
+    } else {
+      MarkSweep::follow_klass(klass);
+    }
+  } else {
+    // If klass is NULL then this a mirror for a primitive type.
+    // We don't have to follow them, since they are handled as strong
+    // roots in Universe::oops_do.
+    assert(java_lang_Class::is_primitive(obj), "Sanity check");
+  }
+
+  oop_oop_iterate_statics<true>(obj, &MarkSweep::mark_and_push_closure);
+}
+
+void InstanceClassLoaderKlass::oop_ms_follow_contents(oop obj) {
+  InstanceKlass::oop_ms_follow_contents(obj);
+
+  ClassLoaderData * const loader_data = java_lang_ClassLoader::loader_data(obj);
+
+  // We must NULL check here, since the class loader
+  // can be found before the loader data has been set up.
+  if(loader_data != NULL) {
+    MarkSweep::follow_class_loader(loader_data);
+  }
+}
+
+template <class T>
+static void oop_ms_follow_contents_specialized(InstanceRefKlass* klass, oop obj) {
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  T heap_oop = oopDesc::load_heap_oop(referent_addr);
+  debug_only(
+    if(TraceReferenceGC && PrintGCDetails) {
+      gclog_or_tty->print_cr("InstanceRefKlass::oop_ms_follow_contents_specialized " PTR_FORMAT, p2i(obj));
+    }
+  )
+  if (!oopDesc::is_null(heap_oop)) {
+    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);
+    if (!referent->is_gc_marked() &&
+        MarkSweep::ref_processor()->discover_reference(obj, klass->reference_type())) {
+      // reference was discovered, referent will be traversed later
+      klass->InstanceKlass::oop_ms_follow_contents(obj);
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("       Non NULL enqueued " PTR_FORMAT, p2i(obj));
+        }
+      )
+      return;
+    } else {
+      // treat referent as normal oop
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("       Non NULL normal " PTR_FORMAT, p2i(obj));
+        }
+      )
+      MarkSweep::mark_and_push(referent_addr);
+    }
+  }
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
+    // Treat discovered as normal oop, if ref is not "active",
+    // i.e. if next is non-NULL.
+    T  next_oop = oopDesc::load_heap_oop(next_addr);
+    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
+      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("   Process discovered as normal "
+                                 PTR_FORMAT, p2i(discovered_addr));
+        }
+      )
+      MarkSweep::mark_and_push(discovered_addr);
+    }
+  } else {
+#ifdef ASSERT
+    // In the case of older JDKs which do not use the discovered
+    // field for the pending list, an inactive ref (next != NULL)
+    // must always have a NULL discovered field.
+    oop next = oopDesc::load_decode_heap_oop(next_addr);
+    oop discovered = java_lang_ref_Reference::discovered(obj);
+    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
+        err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
+            p2i(obj)));
+#endif
+  }
+  // treat next as normal oop.  next is a link in the reference queue.
+  debug_only(
+    if(TraceReferenceGC && PrintGCDetails) {
+      gclog_or_tty->print_cr("   Process next as normal " PTR_FORMAT, p2i(next_addr));
+    }
+  )
+  MarkSweep::mark_and_push(next_addr);
+  klass->InstanceKlass::oop_ms_follow_contents(obj);
+}
+
+void InstanceRefKlass::oop_ms_follow_contents(oop obj) {
+  if (UseCompressedOops) {
+    oop_ms_follow_contents_specialized<narrowOop>(this, obj);
+  } else {
+    oop_ms_follow_contents_specialized<oop>(this, obj);
+  }
+}
+
+template <class T>
+static void oop_ms_follow_contents_specialized(oop obj, int index) {
+  objArrayOop a = objArrayOop(obj);
+  const size_t len = size_t(a->length());
+  const size_t beg_index = size_t(index);
+  assert(beg_index < len || len == 0, "index too large");
+
+  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
+  const size_t end_index = beg_index + stride;
+  T* const base = (T*)a->base();
+  T* const beg = base + beg_index;
+  T* const end = base + end_index;
+
+  // Push the non-NULL elements of the next stride on the marking stack.
+  for (T* e = beg; e < end; e++) {
+    MarkSweep::mark_and_push<T>(e);
+  }
+
+  if (end_index < len) {
+    MarkSweep::push_objarray(a, end_index); // Push the continuation.
+  }
+}
+
+void ObjArrayKlass::oop_ms_follow_contents(oop obj) {
+  assert (obj->is_array(), "obj must be array");
+  MarkSweep::follow_klass(this);
+  if (UseCompressedOops) {
+    oop_ms_follow_contents_specialized<narrowOop>(obj, 0);
+  } else {
+    oop_ms_follow_contents_specialized<oop>(obj, 0);
+  }
+}
+
+void TypeArrayKlass::oop_ms_follow_contents(oop obj) {
+  assert(obj->is_typeArray(),"must be a type array");
+  // Performance tweak: We skip iterating over the klass pointer since we
+  // know that Universe::TypeArrayKlass never moves.
+}
+
 void MarkSweep::follow_array(objArrayOop array, int index) {
-  ObjArrayKlass* k = (ObjArrayKlass*)array->klass();
-  k->oop_follow_contents(array, index);
+  if (UseCompressedOops) {
+    oop_ms_follow_contents_specialized<narrowOop>(array, index);
+  } else {
+    oop_ms_follow_contents_specialized<oop>(array, index);
+  }
 }
 
 void MarkSweep::follow_stack() {
@@ -112,8 +281,10 @@
 
 MarkSweep::AdjustPointerClosure MarkSweep::adjust_pointer_closure;
 
-void MarkSweep::AdjustPointerClosure::do_oop(oop* p)       { adjust_pointer(p); }
-void MarkSweep::AdjustPointerClosure::do_oop(narrowOop* p) { adjust_pointer(p); }
+template <typename T>
+void MarkSweep::AdjustPointerClosure::do_oop_nv(T* p)      { adjust_pointer(p); }
+void MarkSweep::AdjustPointerClosure::do_oop(oop* p)       { do_oop_nv(p); }
+void MarkSweep::AdjustPointerClosure::do_oop(narrowOop* p) { do_oop_nv(p); }
 
 void MarkSweep::adjust_marks() {
   assert( _preserved_oop_stack.size() == _preserved_mark_stack.size(),
@@ -175,3 +346,84 @@
 }
 
 #endif
+
+int InstanceKlass::oop_ms_adjust_pointers(oop obj) {
+  int size = size_helper();
+  oop_oop_iterate_oop_maps<true>(obj, &MarkSweep::adjust_pointer_closure);
+  return size;
+}
+
+int InstanceMirrorKlass::oop_ms_adjust_pointers(oop obj) {
+  int size = oop_size(obj);
+  InstanceKlass::oop_ms_adjust_pointers(obj);
+
+  oop_oop_iterate_statics<true>(obj, &MarkSweep::adjust_pointer_closure);
+  return size;
+}
+
+int InstanceClassLoaderKlass::oop_ms_adjust_pointers(oop obj) {
+  return InstanceKlass::oop_ms_adjust_pointers(obj);
+}
+
+#ifdef ASSERT
+template <class T> static void trace_reference_gc(const char *s, oop obj,
+                                                  T* referent_addr,
+                                                  T* next_addr,
+                                                  T* discovered_addr) {
+  if(TraceReferenceGC && PrintGCDetails) {
+    gclog_or_tty->print_cr("%s obj " PTR_FORMAT, s, p2i(obj));
+    gclog_or_tty->print_cr("     referent_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(referent_addr),
+                           p2i(referent_addr ?
+                               (address)oopDesc::load_decode_heap_oop(referent_addr) : NULL));
+    gclog_or_tty->print_cr("     next_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(next_addr),
+                           p2i(next_addr ? (address)oopDesc::load_decode_heap_oop(next_addr) : NULL));
+    gclog_or_tty->print_cr("     discovered_addr/* " PTR_FORMAT " / "
+                           PTR_FORMAT, p2i(discovered_addr),
+                           p2i(discovered_addr ?
+                               (address)oopDesc::load_decode_heap_oop(discovered_addr) : NULL));
+  }
+}
+#endif
+
+template <class T> void static adjust_object_specialized(oop obj) {
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  MarkSweep::adjust_pointer(referent_addr);
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  MarkSweep::adjust_pointer(next_addr);
+  T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+  MarkSweep::adjust_pointer(discovered_addr);
+  debug_only(trace_reference_gc("InstanceRefKlass::oop_ms_adjust_pointers", obj,
+                                referent_addr, next_addr, discovered_addr);)
+}
+
+int InstanceRefKlass::oop_ms_adjust_pointers(oop obj) {
+  int size = size_helper();
+  InstanceKlass::oop_ms_adjust_pointers(obj);
+
+  if (UseCompressedOops) {
+    adjust_object_specialized<narrowOop>(obj);
+  } else {
+    adjust_object_specialized<oop>(obj);
+  }
+  return size;
+}
+
+int ObjArrayKlass::oop_ms_adjust_pointers(oop obj) {
+  assert(obj->is_objArray(), "obj must be obj array");
+  objArrayOop a = objArrayOop(obj);
+  // Get size before changing pointers.
+  // Don't call size() or oop_size() since that is a virtual call.
+  int size = a->object_size();
+  oop_oop_iterate_elements<true>(a, &MarkSweep::adjust_pointer_closure);
+  return size;
+}
+
+int TypeArrayKlass::oop_ms_adjust_pointers(oop obj) {
+  assert(obj->is_typeArray(), "must be a type array");
+  typeArrayOop t = typeArrayOop(obj);
+  // Performance tweak: We skip iterating over the klass pointer since we
+  // know that Universe::TypeArrayKlass never moves.
+  return t->object_size();
+}
--- a/hotspot/src/share/vm/gc_implementation/shared/markSweep.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/markSweep.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -60,8 +60,9 @@
     virtual void do_oop(narrowOop* p);
   };
 
-  class MarkAndPushClosure: public OopClosure {
+  class MarkAndPushClosure: public ExtendedOopClosure {
    public:
+    template <typename T> void do_oop_nv(T* p);
     virtual void do_oop(oop* p);
     virtual void do_oop(narrowOop* p);
   };
@@ -73,8 +74,12 @@
 
   class AdjustPointerClosure: public OopsInGenClosure {
    public:
+    template <typename T> void do_oop_nv(T* p);
     virtual void do_oop(oop* p);
     virtual void do_oop(narrowOop* p);
+
+    // This closure provides its own oop verification code.
+    debug_only(virtual bool should_verify_oops() { return false; })
   };
 
   // Used for java/lang/ref handling
--- a/hotspot/src/share/vm/gc_implementation/shared/markSweep.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/markSweep.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -28,11 +28,15 @@
 #include "gc_implementation/shared/markSweep.hpp"
 #include "gc_interface/collectedHeap.hpp"
 #include "oops/markOop.inline.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceClassLoaderKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
+#include "oops/instanceRefKlass.inline.hpp"
+#include "oops/objArrayKlass.inline.hpp"
 #include "utilities/stack.inline.hpp"
 #include "utilities/macros.hpp"
 #if INCLUDE_ALL_GCS
 #include "gc_implementation/g1/g1StringDedup.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
 #endif // INCLUDE_ALL_GCS
 
 inline void MarkSweep::mark_object(oop obj) {
@@ -59,7 +63,9 @@
 }
 
 inline void MarkSweep::follow_object(oop obj) {
-  obj->follow_contents();
+  assert(obj->is_gc_marked(), "should be marked");
+
+  obj->ms_follow_contents();
 }
 
 template <class T> inline void MarkSweep::follow_root(T* p) {
@@ -95,13 +101,15 @@
 }
 
 inline int MarkSweep::adjust_pointers(oop obj) {
-  return obj->adjust_pointers();
+  return obj->ms_adjust_pointers();
 }
 
 template <class T> inline void MarkSweep::adjust_pointer(T* p) {
   T heap_oop = oopDesc::load_heap_oop(p);
   if (!oopDesc::is_null(heap_oop)) {
     oop obj     = oopDesc::decode_heap_oop_not_null(heap_oop);
+    assert(Universe::heap()->is_in(obj), "should be in heap");
+
     oop new_obj = oop(obj->mark()->decode_pointer());
     assert(new_obj != NULL ||                         // is forwarding ptr?
            obj->mark() == markOopDesc::prototype() || // not gc marked?
--- a/hotspot/src/share/vm/gc_implementation/shared/mutableNUMASpace.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/mutableNUMASpace.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,7 +26,7 @@
 #include "precompiled.hpp"
 #include "gc_implementation/shared/mutableNUMASpace.hpp"
 #include "gc_implementation/shared/spaceDecorator.hpp"
-#include "memory/sharedHeap.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/atomic.inline.hpp"
 #include "runtime/thread.inline.hpp"
--- a/hotspot/src/share/vm/gc_implementation/shared/vmGCOperations.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/vmGCOperations.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -116,8 +116,6 @@
     _prologue_succeeded = false;
   } else {
     _prologue_succeeded = true;
-    SharedHeap* sh = SharedHeap::heap();
-    if (sh != NULL) sh->_thread_holds_heap_lock_for_gc = true;
   }
   return _prologue_succeeded;
 }
@@ -126,22 +124,11 @@
 void VM_GC_Operation::doit_epilogue() {
   assert(Thread::current()->is_Java_thread(), "just checking");
   // Release the Heap_lock first.
-  SharedHeap* sh = SharedHeap::heap();
-  if (sh != NULL) sh->_thread_holds_heap_lock_for_gc = false;
   Heap_lock->unlock();
   release_and_notify_pending_list_lock();
 }
 
-bool VM_GC_HeapInspection::doit_prologue() {
-  if (Universe::heap()->supports_heap_inspection()) {
-    return VM_GC_Operation::doit_prologue();
-  } else {
-    return false;
-  }
-}
-
 bool VM_GC_HeapInspection::skip_operation() const {
-  assert(Universe::heap()->supports_heap_inspection(), "huh?");
   return false;
 }
 
--- a/hotspot/src/share/vm/gc_implementation/shared/vmGCOperations.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_implementation/shared/vmGCOperations.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -150,7 +150,6 @@
   ~VM_GC_HeapInspection() {}
   virtual VMOp_Type type() const { return VMOp_GC_HeapInspection; }
   virtual bool skip_operation() const;
-  virtual bool doit_prologue();
   virtual void doit();
   void set_csv_format(bool value) {_csv_format = value;}
   void set_print_help(bool value) {_print_help = value;}
--- a/hotspot/src/share/vm/gc_interface/collectedHeap.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_interface/collectedHeap.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -220,6 +220,11 @@
   }
 }
 
+void CollectedHeap::set_barrier_set(BarrierSet* barrier_set) {
+  _barrier_set = barrier_set;
+  oopDesc::set_bs(_barrier_set);
+}
+
 void CollectedHeap::pre_initialize() {
   // Used for ReduceInitialCardMarks (when COMPILER2 is used);
   // otherwise remains unused.
--- a/hotspot/src/share/vm/gc_interface/collectedHeap.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_interface/collectedHeap.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -75,9 +75,8 @@
 
 //
 // CollectedHeap
-//   SharedHeap
-//     GenCollectedHeap
-//     G1CollectedHeap
+//   GenCollectedHeap
+//   G1CollectedHeap
 //   ParallelScavengeHeap
 //
 class CollectedHeap : public CHeapObj<mtInternal> {
@@ -205,7 +204,7 @@
   // In many heaps, there will be a need to perform some initialization activities
   // after the Universe is fully formed, but before general heap allocation is allowed.
   // This is the correct place to place such initialization methods.
-  virtual void post_initialize() = 0;
+  virtual void post_initialize();
 
   // Stop any onging concurrent work and prepare for exit.
   virtual void stop() {}
@@ -240,22 +239,11 @@
   }
 
   // Returns "TRUE" iff "p" points into the committed areas of the heap.
-  // Since this method can be expensive in general, we restrict its
-  // use to assertion checking only.
+  // This method can be expensive so avoid using it in performance critical
+  // code.
   virtual bool is_in(const void* p) const = 0;
 
-  bool is_in_or_null(const void* p) const {
-    return p == NULL || is_in(p);
-  }
-
-  bool is_in_place(Metadata** p) {
-    return !Universe::heap()->is_in(p);
-  }
-  bool is_in_place(oop* p) { return Universe::heap()->is_in(p); }
-  bool is_in_place(narrowOop* p) {
-    oop o = oopDesc::load_decode_heap_oop_not_null(p);
-    return Universe::heap()->is_in((const void*)o);
-  }
+  DEBUG_ONLY(bool is_in_or_null(const void* p) const { return p == NULL || is_in(p); })
 
   // Let's define some terms: a "closed" subset of a heap is one that
   //
@@ -451,9 +439,6 @@
   // remembered set.
   virtual void flush_deferred_store_barrier(JavaThread* thread);
 
-  // Does this heap support heap inspection (+PrintClassHistogram?)
-  virtual bool supports_heap_inspection() const = 0;
-
   // Perform a collection of the heap; intended for use in implementing
   // "System.gc".  This probably implies as full a collection as the
   // "CollectedHeap" supports.
@@ -470,6 +455,7 @@
 
   // Returns the barrier set for this heap
   BarrierSet* barrier_set() { return _barrier_set; }
+  void set_barrier_set(BarrierSet* barrier_set);
 
   // Returns "true" iff there is a stop-world GC in progress.  (I assume
   // that it should answer "false" for the concurrent part of a concurrent
@@ -497,12 +483,6 @@
   // Return the CollectorPolicy for the heap
   virtual CollectorPolicy* collector_policy() const = 0;
 
-  void oop_iterate_no_header(OopClosure* cl);
-
-  // Iterate over all the ref-containing fields of all objects, calling
-  // "cl.do_oop" on each.
-  virtual void oop_iterate(ExtendedOopClosure* cl) = 0;
-
   // Iterate over all objects, calling "cl.do_object" on each.
   virtual void object_iterate(ObjectClosure* cl) = 0;
 
--- a/hotspot/src/share/vm/gc_interface/collectedHeap.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/gc_interface/collectedHeap.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -236,12 +236,6 @@
   return (oop)obj;
 }
 
-inline void CollectedHeap::oop_iterate_no_header(OopClosure* cl) {
-  NoHeaderExtendedOopClosure no_header_cl(cl);
-  oop_iterate(&no_header_cl);
-}
-
-
 inline HeapWord* CollectedHeap::align_allocation_or_fail(HeapWord* addr,
                                                          HeapWord* end,
                                                          unsigned short alignment_in_bytes) {
--- a/hotspot/src/share/vm/interpreter/linkResolver.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/interpreter/linkResolver.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1406,10 +1406,11 @@
                                                  KlassHandle resolved_klass,
                                                  Symbol* name,
                                                  Symbol* signature,
-                                                 KlassHandle current_klass) {
+                                                 KlassHandle current_klass,
+                                                 bool check_access) {
   EXCEPTION_MARK;
   CallInfo info;
-  resolve_virtual_call(info, Handle(), receiver_klass, resolved_klass, name, signature, current_klass, true, false, THREAD);
+  resolve_virtual_call(info, Handle(), receiver_klass, resolved_klass, name, signature, current_klass, check_access, false, THREAD);
   if (HAS_PENDING_EXCEPTION) {
     CLEAR_PENDING_EXCEPTION;
     return methodHandle();
@@ -1422,10 +1423,11 @@
                                                  KlassHandle resolved_klass,
                                                  Symbol* name,
                                                  Symbol* signature,
-                                                 KlassHandle current_klass) {
+                                                 KlassHandle current_klass,
+                                                 bool check_access) {
   EXCEPTION_MARK;
   CallInfo info;
-  resolve_interface_call(info, Handle(), receiver_klass, resolved_klass, name, signature, current_klass, true, false, THREAD);
+  resolve_interface_call(info, Handle(), receiver_klass, resolved_klass, name, signature, current_klass, check_access, false, THREAD);
   if (HAS_PENDING_EXCEPTION) {
     CLEAR_PENDING_EXCEPTION;
     return methodHandle();
@@ -1453,10 +1455,11 @@
                                                   KlassHandle resolved_klass,
                                                   Symbol* name,
                                                   Symbol* signature,
-                                                  KlassHandle current_klass) {
+                                                  KlassHandle current_klass,
+                                                  bool check_access) {
   EXCEPTION_MARK;
   CallInfo info;
-  resolve_static_call(info, resolved_klass, name, signature, current_klass, true, false, THREAD);
+  resolve_static_call(info, resolved_klass, name, signature, current_klass, check_access, false, THREAD);
   if (HAS_PENDING_EXCEPTION) {
     CLEAR_PENDING_EXCEPTION;
     return methodHandle();
@@ -1464,11 +1467,15 @@
   return info.selected_method();
 }
 
-methodHandle LinkResolver::resolve_special_call_or_null(KlassHandle resolved_klass, Symbol* name, Symbol* signature,
-                                                        KlassHandle current_klass) {
+methodHandle LinkResolver::resolve_special_call_or_null(
+                                                  KlassHandle resolved_klass,
+                                                  Symbol* name,
+                                                  Symbol* signature,
+                                                  KlassHandle current_klass,
+                                                  bool check_access) {
   EXCEPTION_MARK;
   CallInfo info;
-  resolve_special_call(info, resolved_klass, name, signature, current_klass, true, THREAD);
+  resolve_special_call(info, resolved_klass, name, signature, current_klass, check_access, THREAD);
   if (HAS_PENDING_EXCEPTION) {
     CLEAR_PENDING_EXCEPTION;
     return methodHandle();
--- a/hotspot/src/share/vm/interpreter/linkResolver.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/interpreter/linkResolver.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -181,10 +181,10 @@
 
   // same as above for compile-time resolution; but returns null handle instead of throwing an exception on error
   // also, does not initialize klass (i.e., no side effects)
-  static methodHandle resolve_virtual_call_or_null  (KlassHandle receiver_klass, KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass);
-  static methodHandle resolve_interface_call_or_null(KlassHandle receiver_klass, KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass);
-  static methodHandle resolve_static_call_or_null   (KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass);
-  static methodHandle resolve_special_call_or_null  (KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass);
+  static methodHandle resolve_virtual_call_or_null  (KlassHandle receiver_klass, KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass, bool check_access = true);
+  static methodHandle resolve_interface_call_or_null(KlassHandle receiver_klass, KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass, bool check_access = true);
+  static methodHandle resolve_static_call_or_null   (KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass, bool check_access = true);
+  static methodHandle resolve_special_call_or_null  (KlassHandle resolved_klass, Symbol* method_name, Symbol* method_signature, KlassHandle current_klass, bool check_access = true);
   static int vtable_index_of_interface_method(KlassHandle klass, methodHandle resolved_method);
 
   // same as above for compile-time resolution; returns vtable_index if current_klass if linked
--- a/hotspot/src/share/vm/memory/cardGeneration.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/cardGeneration.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -60,7 +60,7 @@
   // which would cause problems when we commit/uncommit memory, and when we
   // clear and dirty cards.
   guarantee(_rs->is_aligned(reserved_mr.start()), "generation must be card aligned");
-  if (reserved_mr.end() != Universe::heap()->reserved_region().end()) {
+  if (reserved_mr.end() != GenCollectedHeap::heap()->reserved_region().end()) {
     // Don't check at the very end of the heap as we'll assert that we're probing off
     // the end if we try.
     guarantee(_rs->is_aligned(reserved_mr.end()), "generation must be card aligned");
@@ -78,7 +78,7 @@
        heap_word_size(_virtual_space.committed_size());
     MemRegion mr(space()->bottom(), new_word_size);
     // Expand card table
-    Universe::heap()->barrier_set()->resize_covered_region(mr);
+    GenCollectedHeap::heap()->barrier_set()->resize_covered_region(mr);
     // Expand shared block offset array
     _bts->resize(new_word_size);
 
@@ -170,7 +170,7 @@
   _bts->resize(new_word_size);
   MemRegion mr(space()->bottom(), new_word_size);
   // Shrink the card table
-  Universe::heap()->barrier_set()->resize_covered_region(mr);
+  GenCollectedHeap::heap()->barrier_set()->resize_covered_region(mr);
 
   if (Verbose && PrintGC) {
     size_t new_mem_size = _virtual_space.committed_size();
--- a/hotspot/src/share/vm/memory/cardTableModRefBS.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/cardTableModRefBS.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,10 +23,11 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/allocation.inline.hpp"
 #include "memory/cardTableModRefBS.inline.hpp"
 #include "memory/cardTableRS.hpp"
-#include "memory/sharedHeap.hpp"
+#include "memory/genCollectedHeap.hpp"
 #include "memory/space.hpp"
 #include "memory/space.inline.hpp"
 #include "memory/universe.hpp"
@@ -450,21 +451,20 @@
     // This is an example of where n_par_threads() is used instead
     // of workers()->active_workers().  n_par_threads can be set to 0 to
     // turn off parallelism.  For example when this code is called as
-    // part of verification and SharedHeap::process_roots() is being
-    // used, then n_par_threads() may have been set to 0.  active_workers
-    // is not overloaded with the meaning that it is a switch to disable
-    // parallelism and so keeps the meaning of the number of
-    // active gc workers.  If parallelism has not been shut off by
-    // setting n_par_threads to 0, then n_par_threads should be
-    // equal to active_workers.  When a different mechanism for shutting
-    // off parallelism is used, then active_workers can be used in
+    // part of verification during root processing then n_par_threads()
+    // may have been set to 0. active_workers is not overloaded with
+    // the meaning that it is a switch to disable parallelism and so keeps
+    // the meaning of the number of active gc workers. If parallelism has
+    // not been shut off by setting n_par_threads to 0, then n_par_threads
+    // should be equal to active_workers.  When a different mechanism for
+    // shutting off parallelism is used, then active_workers can be used in
     // place of n_par_threads.
-    int n_threads =  SharedHeap::heap()->n_par_threads();
+    int n_threads =  GenCollectedHeap::heap()->n_par_threads();
     bool is_par = n_threads > 0;
     if (is_par) {
 #if INCLUDE_ALL_GCS
-      assert(SharedHeap::heap()->n_par_threads() ==
-             SharedHeap::heap()->workers()->active_workers(), "Mismatch");
+      assert(GenCollectedHeap::heap()->n_par_threads() ==
+             GenCollectedHeap::heap()->workers()->active_workers(), "Mismatch");
       non_clean_card_iterate_parallel_work(sp, mr, cl, ct, n_threads);
 #else  // INCLUDE_ALL_GCS
       fatal("Parallel gc not supported here.");
--- a/hotspot/src/share/vm/memory/cardTableRS.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/cardTableRS.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2001, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -38,16 +38,18 @@
   GenRemSet(),
   _cur_youngergen_card_val(youngergenP1_card)
 {
-  guarantee(Universe::heap()->kind() == CollectedHeap::GenCollectedHeap, "sanity");
   _ct_bs = new CardTableModRefBSForCTRS(whole_heap);
   _ct_bs->initialize();
   set_bs(_ct_bs);
-  _last_cur_val_in_gen = NEW_C_HEAP_ARRAY3(jbyte, GenCollectedHeap::max_gens + 1,
+  // max_gens is really GenCollectedHeap::heap()->gen_policy()->number_of_generations()
+  // (which is always 2, young & old), but GenCollectedHeap has not been initialized yet.
+  uint max_gens = 2;
+  _last_cur_val_in_gen = NEW_C_HEAP_ARRAY3(jbyte, max_gens + 1,
                          mtGC, CURRENT_PC, AllocFailStrategy::RETURN_NULL);
   if (_last_cur_val_in_gen == NULL) {
     vm_exit_during_initialization("Could not create last_cur_val_in_gen array.");
   }
-  for (int i = 0; i < GenCollectedHeap::max_gens + 1; i++) {
+  for (uint i = 0; i < max_gens + 1; i++) {
     _last_cur_val_in_gen[i] = clean_card_val();
   }
   _ct_bs->set_CTRS(this);
@@ -167,16 +169,20 @@
     // Cannot yet substitute active_workers for n_par_threads
     // in the case where parallelism is being turned off by
     // setting n_par_threads to 0.
-    _is_par = (SharedHeap::heap()->n_par_threads() > 0);
+    _is_par = (GenCollectedHeap::heap()->n_par_threads() > 0);
     assert(!_is_par ||
-           (SharedHeap::heap()->n_par_threads() ==
-            SharedHeap::heap()->workers()->active_workers()), "Mismatch");
+           (GenCollectedHeap::heap()->n_par_threads() ==
+            GenCollectedHeap::heap()->workers()->active_workers()), "Mismatch");
 }
 
 bool ClearNoncleanCardWrapper::is_word_aligned(jbyte* entry) {
   return (((intptr_t)entry) & (BytesPerWord-1)) == 0;
 }
 
+// The regions are visited in *decreasing* address order.
+// This order aids with imprecise card marking, where a dirty
+// card may cause scanning, and summarization marking, of objects
+// that extend onto subsequent cards.
 void ClearNoncleanCardWrapper::do_MemRegion(MemRegion mr) {
   assert(mr.word_size() > 0, "Error");
   assert(_ct->is_aligned(mr.start()), "mr.start() should be card aligned");
@@ -591,10 +597,6 @@
   // At present, we only know how to verify the card table RS for
   // generational heaps.
   VerifyCTGenClosure blk(this);
-  CollectedHeap* ch = Universe::heap();
-
-  if (ch->kind() == CollectedHeap::GenCollectedHeap) {
-    GenCollectedHeap::heap()->generation_iterate(&blk, false);
-    _ct_bs->verify();
-    }
-  }
+  GenCollectedHeap::heap()->generation_iterate(&blk, false);
+  _ct_bs->verify();
+}
--- a/hotspot/src/share/vm/memory/collectorPolicy.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/collectorPolicy.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -669,7 +669,7 @@
       }
 
       // Read the gc count while the heap lock is held.
-      gc_count_before = Universe::heap()->total_collections();
+      gc_count_before = gch->total_collections();
     }
 
     VM_GenCollectForAllocation op(size, is_tlab, gc_count_before);
--- a/hotspot/src/share/vm/memory/defNewGeneration.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/defNewGeneration.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -48,6 +48,9 @@
 #include "utilities/copy.hpp"
 #include "utilities/globalDefinitions.hpp"
 #include "utilities/stack.inline.hpp"
+#if INCLUDE_ALL_GCS
+#include "gc_implementation/parNew/parOopClosures.hpp"
+#endif
 
 //
 // DefNewGeneration functions.
@@ -190,7 +193,9 @@
 {
   MemRegion cmr((HeapWord*)_virtual_space.low(),
                 (HeapWord*)_virtual_space.high());
-  Universe::heap()->barrier_set()->resize_covered_region(cmr);
+  GenCollectedHeap* gch = GenCollectedHeap::heap();
+
+  gch->barrier_set()->resize_covered_region(cmr);
 
   _eden_space = new ContiguousSpace();
   _from_space = new ContiguousSpace();
@@ -202,13 +207,13 @@
   // Compute the maximum eden and survivor space sizes. These sizes
   // are computed assuming the entire reserved space is committed.
   // These values are exported as performance counters.
-  uintx alignment = GenCollectedHeap::heap()->collector_policy()->space_alignment();
+  uintx alignment = gch->collector_policy()->space_alignment();
   uintx size = _virtual_space.reserved_size();
   _max_survivor_size = compute_survivor_size(size, alignment);
   _max_eden_size = size - (2*_max_survivor_size);
 
   // allocate the performance counters
-  GenCollectorPolicy* gcp = (GenCollectorPolicy*) GenCollectedHeap::heap()->collector_policy();
+  GenCollectorPolicy* gcp = (GenCollectorPolicy*)gch->collector_policy();
 
   // Generation counters -- generation 0, 3 subspaces
   _gen_counters = new GenerationCounters("new", 0, 3,
@@ -378,8 +383,7 @@
 
   int next_level = level() + 1;
   GenCollectedHeap* gch = GenCollectedHeap::heap();
-  assert(next_level < gch->n_gens(),
-         "DefNewGeneration cannot be an oldest gen");
+  assert(next_level == 1, "DefNewGeneration must be a young gen");
 
   Generation* old_gen = gch->old_gen();
   size_t old_size = old_gen->capacity();
@@ -431,7 +435,7 @@
                              SpaceDecorator::DontMangle);
     MemRegion cmr((HeapWord*)_virtual_space.low(),
                   (HeapWord*)_virtual_space.high());
-    Universe::heap()->barrier_set()->resize_covered_region(cmr);
+    gch->barrier_set()->resize_covered_region(cmr);
     if (Verbose && PrintGC) {
       size_t new_size_after  = _virtual_space.committed_size();
       size_t eden_size_after = eden()->capacity();
@@ -550,8 +554,9 @@
 
 void DefNewGeneration::adjust_desired_tenuring_threshold() {
   // Set the desired survivor size to half the real survivor space
+  GCPolicyCounters* gc_counters = GenCollectedHeap::heap()->collector_policy()->counters();
   _tenuring_threshold =
-    age_table()->compute_tenuring_threshold(to()->capacity()/HeapWordSize);
+    age_table()->compute_tenuring_threshold(to()->capacity()/HeapWordSize, gc_counters);
 }
 
 void DefNewGeneration::collect(bool   full,
@@ -688,7 +693,7 @@
     gc_tracer.report_promotion_failed(_promotion_failed_info);
 
     // Reset the PromotionFailureALot counters.
-    NOT_PRODUCT(Universe::heap()->reset_promotion_should_fail();)
+    NOT_PRODUCT(gch->reset_promotion_should_fail();)
   }
   if (PrintGC && !PrintGCDetails) {
     gch->print_heap_change(gch_prev_used);
--- a/hotspot/src/share/vm/memory/defNewGeneration.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/defNewGeneration.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,9 +25,9 @@
 #ifndef SHARE_VM_MEMORY_DEFNEWGENERATION_INLINE_HPP
 #define SHARE_VM_MEMORY_DEFNEWGENERATION_INLINE_HPP
 
-#include "gc_interface/collectedHeap.hpp"
 #include "memory/cardTableRS.hpp"
 #include "memory/defNewGeneration.hpp"
+#include "memory/genCollectedHeap.hpp"
 #include "memory/genOopClosures.inline.hpp"
 #include "memory/space.hpp"
 
@@ -60,7 +60,7 @@
   // We could check that p is also in an older generation, but
   // dirty cards in the youngest gen are never scanned, so the
   // extra check probably isn't worthwhile.
-  if (Universe::heap()->is_in_reserved(p)) {
+  if (GenCollectedHeap::heap()->is_in_reserved(p)) {
     oop obj = oopDesc::load_decode_heap_oop_not_null(p);
     _rs->inline_write_ref_field_gc(p, obj);
   }
@@ -84,7 +84,7 @@
   // we set a younger_gen card if we have an older->youngest
   // generation pointer.
   oop obj = oopDesc::load_decode_heap_oop_not_null(p);
-  if (((HeapWord*)obj < _boundary) && Universe::heap()->is_in_reserved(p)) {
+  if (((HeapWord*)obj < _boundary) && GenCollectedHeap::heap()->is_in_reserved(p)) {
     _rs->inline_write_ref_field_gc(p, obj);
   }
 }
--- a/hotspot/src/share/vm/memory/freeList.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/freeList.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,10 +23,10 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/freeBlockDictionary.hpp"
 #include "memory/freeList.hpp"
 #include "memory/metachunk.hpp"
-#include "memory/sharedHeap.hpp"
 #include "runtime/globals.hpp"
 #include "runtime/mutex.hpp"
 #include "runtime/vmThread.hpp"
--- a/hotspot/src/share/vm/memory/gcLocker.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/gcLocker.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,9 +23,9 @@
  */
 
 #include "precompiled.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/gcLocker.inline.hpp"
 #include "memory/resourceArea.hpp"
-#include "memory/sharedHeap.hpp"
 #include "runtime/atomic.inline.hpp"
 #include "runtime/thread.inline.hpp"
 
--- a/hotspot/src/share/vm/memory/genCollectedHeap.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/genCollectedHeap.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -39,7 +39,7 @@
 #include "memory/genOopClosures.inline.hpp"
 #include "memory/generationSpec.hpp"
 #include "memory/resourceArea.hpp"
-#include "memory/sharedHeap.hpp"
+#include "memory/strongRootsScope.hpp"
 #include "memory/space.hpp"
 #include "oops/oop.inline.hpp"
 #include "runtime/biasedLocking.hpp"
@@ -50,9 +50,10 @@
 #include "runtime/vmThread.hpp"
 #include "services/management.hpp"
 #include "services/memoryService.hpp"
+#include "utilities/macros.hpp"
+#include "utilities/stack.inline.hpp"
 #include "utilities/vmError.hpp"
 #include "utilities/workgroup.hpp"
-#include "utilities/macros.hpp"
 #if INCLUDE_ALL_GCS
 #include "gc_implementation/concurrentMarkSweep/concurrentMarkSweepThread.hpp"
 #include "gc_implementation/concurrentMarkSweep/vmCMSOperations.hpp"
@@ -78,21 +79,27 @@
 };
 
 GenCollectedHeap::GenCollectedHeap(GenCollectorPolicy *policy) :
-  SharedHeap(),
+  CollectedHeap(),
   _rem_set(NULL),
   _gen_policy(policy),
   _process_strong_tasks(new SubTasksDone(GCH_PS_NumElements)),
   _full_collections_completed(0)
 {
   assert(policy != NULL, "Sanity check");
+  if (UseConcMarkSweepGC) {
+    _workers = new FlexibleWorkGang("GC Thread", ParallelGCThreads,
+                            /* are_GC_task_threads */true,
+                            /* are_ConcurrentGC_threads */false);
+    _workers->initialize_workers();
+  } else {
+    // Serial GC does not use workers.
+    _workers = NULL;
+  }
 }
 
 jint GenCollectedHeap::initialize() {
   CollectedHeap::pre_initialize();
 
-  _n_gens = gen_policy()->number_of_generations();
-  assert(_n_gens == 2, "There is no support for more than two generations");
-
   // While there are no constraints in the GC code that HeapWordSize
   // be any particular value, there are multiple other areas in the
   // system which believe this to be true (e.g. oop->object_size in some
@@ -166,7 +173,8 @@
 }
 
 void GenCollectedHeap::post_initialize() {
-  SharedHeap::post_initialize();
+  CollectedHeap::post_initialize();
+  ref_processing_init();
   GenCollectorPolicy *policy = (GenCollectorPolicy *)collector_policy();
   guarantee(policy->is_generation_policy(), "Illegal policy type");
   assert((_young_gen->kind() == Generation::DefNew) ||
@@ -185,7 +193,6 @@
 }
 
 void GenCollectedHeap::ref_processing_init() {
-  SharedHeap::ref_processing_init();
   _young_gen->ref_processor_init();
   _old_gen->ref_processor_init();
 }
@@ -200,8 +207,7 @@
 
 // Save the "used_region" for generations level and lower.
 void GenCollectedHeap::save_used_regions(int level) {
-  assert(level >= 0, "Illegal level parameter");
-  assert(level < _n_gens, "Illegal level parameter");
+  assert(level == 0 || level == 1, "Illegal level parameter");
   if (level == 1) {
     _old_gen->save_used_region();
   }
@@ -417,7 +423,6 @@
   assert(Heap_lock->is_locked(),
          "the requesting thread should have the Heap_lock");
   guarantee(!is_gc_active(), "collection is not reentrant");
-  assert(max_level < n_gens(), "sanity check");
 
   if (GC_locker::check_active_before_gc()) {
     return; // GC is disabled (e.g. JNI GetXXXCritical operation)
@@ -435,7 +440,7 @@
   {
     FlagSetting fl(_is_gc_active, true);
 
-    bool complete = full && (max_level == (n_gens()-1));
+    bool complete = full && (max_level == 1 /* old */);
     const char* gc_cause_prefix = complete ? "Full GC" : "GC";
     TraceCPUTime tcpu(PrintGCDetails, true, gclog_or_tty);
     // The PrintGCDetails logging starts before we have incremented the GC id. We will do that later
@@ -507,7 +512,7 @@
     // Update "complete" boolean wrt what actually transpired --
     // for instance, a promotion failure could have led to
     // a whole heap collection.
-    complete = complete || (max_level_collected == n_gens() - 1);
+    complete = complete || (max_level_collected == 1 /* old */);
 
     if (complete) { // We did a "major" collection
       // FIXME: See comment at pre_full_gc_dump call
@@ -524,7 +529,7 @@
     }
 
     // Adjust generation sizes.
-    if (max_level_collected == 1) {
+    if (max_level_collected == 1 /* old */) {
       _old_gen->compute_new_size();
     }
     _young_gen->compute_new_size();
@@ -560,7 +565,8 @@
 }
 
 void GenCollectedHeap::set_par_threads(uint t) {
-  SharedHeap::set_par_threads(t);
+  assert(t == 0 || !UseSerialGC, "Cannot have parallel threads");
+  CollectedHeap::set_par_threads(t);
   set_n_termination(t);
 }
 
@@ -586,7 +592,7 @@
                                      CLDClosure* strong_cld_closure,
                                      CLDClosure* weak_cld_closure,
                                      CodeBlobClosure* code_roots) {
-  StrongRootsScope srs(this, activate_scope);
+  StrongRootsScope srs(activate_scope);
 
   // General roots.
   assert(Threads::thread_claim_parity() != 0, "must have called prologue code");
@@ -606,7 +612,8 @@
   // Only process code roots from thread stacks if we aren't visiting the entire CodeCache anyway
   CodeBlobClosure* roots_from_code_p = (so & SO_AllCodeCache) ? NULL : code_roots;
 
-  Threads::possibly_parallel_oops_do(strong_roots, roots_from_clds_p, roots_from_code_p);
+  bool is_par = n_par_threads() > 0;
+  Threads::possibly_parallel_oops_do(is_par, strong_roots, roots_from_clds_p, roots_from_code_p);
 
   if (!_process_strong_tasks->is_task_claimed(GCH_PS_Universe_oops_do)) {
     Universe::oops_do(strong_roots);
@@ -771,19 +778,19 @@
 #endif // INCLUDE_ALL_GCS
   } else if (cause == GCCause::_wb_young_gc) {
     // minor collection for WhiteBox API
-    collect(cause, 0);
+    collect(cause, 0 /* young */);
   } else {
 #ifdef ASSERT
   if (cause == GCCause::_scavenge_alot) {
     // minor collection only
-    collect(cause, 0);
+    collect(cause, 0 /* young */);
   } else {
     // Stop-the-world full collection
-    collect(cause, n_gens() - 1);
+    collect(cause, 1 /* old */);
   }
 #else
     // Stop-the-world full collection
-    collect(cause, n_gens() - 1);
+    collect(cause, 1 /* old */);
 #endif
   }
 }
@@ -798,7 +805,7 @@
 void GenCollectedHeap::collect_locked(GCCause::Cause cause) {
   // The caller has the Heap_lock
   assert(Heap_lock->owned_by_self(), "this thread should own the Heap_lock");
-  collect_locked(cause, n_gens() - 1);
+  collect_locked(cause, 1 /* old */);
 }
 
 // this is the private collection interface
@@ -854,7 +861,7 @@
 #endif // INCLUDE_ALL_GCS
 
 void GenCollectedHeap::do_full_collection(bool clear_all_soft_refs) {
-   do_full_collection(clear_all_soft_refs, _n_gens - 1);
+   do_full_collection(clear_all_soft_refs, 1 /* old */);
 }
 
 void GenCollectedHeap::do_full_collection(bool clear_all_soft_refs,
@@ -886,7 +893,7 @@
                   clear_all_soft_refs  /* clear_all_soft_refs */,
                   0                    /* size */,
                   false                /* is_tlab */,
-                  n_gens() - 1         /* max_level */);
+                  1  /* old */         /* max_level */);
   }
 }
 
@@ -899,17 +906,6 @@
 
 // Returns "TRUE" iff "p" points into the committed areas of the heap.
 bool GenCollectedHeap::is_in(const void* p) const {
-  #ifndef ASSERT
-  guarantee(VerifyBeforeGC      ||
-            VerifyDuringGC      ||
-            VerifyBeforeExit    ||
-            VerifyDuringStartup ||
-            PrintAssembly       ||
-            tty->count() != 0   ||   // already printing
-            VerifyAfterGC       ||
-    VMError::fatal_error_in_progress(), "too expensive");
-
-  #endif
   return _young_gen->is_in(p) || _old_gen->is_in(p);
 }
 
@@ -923,6 +919,11 @@
 }
 #endif
 
+void GenCollectedHeap::oop_iterate_no_header(OopClosure* cl) {
+  NoHeaderExtendedOopClosure no_header_cl(cl);
+  oop_iterate(&no_header_cl);
+}
+
 void GenCollectedHeap::oop_iterate(ExtendedOopClosure* cl) {
   _young_gen->oop_iterate(cl);
   _old_gen->oop_iterate(cl);
@@ -1092,11 +1093,6 @@
   }
 }
 
-void GenCollectedHeap::space_iterate(SpaceClosure* cl) {
-  _young_gen->space_iterate(cl, true);
-  _old_gen->space_iterate(cl, true);
-}
-
 bool GenCollectedHeap::is_maximal_no_gc() const {
   return _young_gen->is_maximal_no_gc() && _old_gen->is_maximal_no_gc();
 }
@@ -1112,9 +1108,7 @@
   return _gch;
 }
 
-
 void GenCollectedHeap::prepare_for_compaction() {
-  guarantee(_n_gens = 2, "Wrong number of generations");
   // Start by compacting into same gen.
   CompactPoint cp(_old_gen);
   _old_gen->prepare_for_compaction(&cp);
--- a/hotspot/src/share/vm/memory/genCollectedHeap.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/genCollectedHeap.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,15 +26,16 @@
 #define SHARE_VM_MEMORY_GENCOLLECTEDHEAP_HPP
 
 #include "gc_implementation/shared/adaptiveSizePolicy.hpp"
+#include "gc_interface/collectedHeap.hpp"
 #include "memory/collectorPolicy.hpp"
 #include "memory/generation.hpp"
-#include "memory/sharedHeap.hpp"
 
 class SubTasksDone;
+class FlexibleWorkGang;
 
-// A "GenCollectedHeap" is a SharedHeap that uses generational
+// A "GenCollectedHeap" is a CollectedHeap that uses generational
 // collection.  It has two generations, young and old.
-class GenCollectedHeap : public SharedHeap {
+class GenCollectedHeap : public CollectedHeap {
   friend class GenCollectorPolicy;
   friend class Generation;
   friend class DefNewGeneration;
@@ -51,10 +52,6 @@
   friend class GCCauseSetter;
   friend class VMStructs;
 public:
-  enum SomeConstants {
-    max_gens = 10
-  };
-
   friend class VM_PopulateDumpSharedSpace;
 
  protected:
@@ -62,8 +59,6 @@
   static GenCollectedHeap* _gch;
 
  private:
-  int _n_gens;
-
   Generation* _young_gen;
   Generation* _old_gen;
 
@@ -93,6 +88,8 @@
   // In block contents verification, the number of header words to skip
   NOT_PRODUCT(static size_t _skip_header_HeapWords;)
 
+  FlexibleWorkGang* _workers;
+
 protected:
   // Helper functions for allocation
   HeapWord* attempt_allocation(size_t size,
@@ -125,6 +122,8 @@
 public:
   GenCollectedHeap(GenCollectorPolicy *policy);
 
+  FlexibleWorkGang* workers() const { return _workers; }
+
   GCStats* gc_stats(int level) const;
 
   // Returns JNI_OK on success
@@ -178,9 +177,6 @@
   HeapWord** top_addr() const;
   HeapWord** end_addr() const;
 
-  // Does this heap support heap inspection? (+PrintClassHistogram)
-  virtual bool supports_heap_inspection() const { return true; }
-
   // Perform a full collection of the heap; intended for use in implementing
   // "System.gc". This implies as full a collection as the CollectedHeap
   // supports. Caller does not hold the Heap_lock on entry.
@@ -223,6 +219,7 @@
   }
 
   // Iteration functions.
+  void oop_iterate_no_header(OopClosure* cl);
   void oop_iterate(ExtendedOopClosure* cl);
   void object_iterate(ObjectClosure* cl);
   void safe_object_iterate(ObjectClosure* cl);
@@ -331,7 +328,6 @@
     _old_gen->update_gc_stats(current_level, full);
   }
 
-  // Override.
   bool no_gc_in_progress() { return !is_gc_active(); }
 
   // Override.
@@ -363,18 +359,11 @@
   // If "old_to_young" determines the order.
   void generation_iterate(GenClosure* cl, bool old_to_young);
 
-  void space_iterate(SpaceClosure* cl);
-
   // Return "true" if all generations have reached the
   // maximal committed limit that they can reach, without a garbage
   // collection.
   virtual bool is_maximal_no_gc() const;
 
-  int n_gens() const {
-    assert(_n_gens == gen_policy()->number_of_generations(), "Sanity");
-    return _n_gens;
-  }
-
   // This function returns the "GenRemSet" object that allows us to scan
   // generations in a fully generational heap.
   GenRemSet* rem_set() { return _rem_set; }
@@ -531,8 +520,8 @@
   void record_gen_tops_before_GC() PRODUCT_RETURN;
 
 protected:
-  virtual void gc_prologue(bool full);
-  virtual void gc_epilogue(bool full);
+  void gc_prologue(bool full);
+  void gc_epilogue(bool full);
 };
 
 #endif // SHARE_VM_MEMORY_GENCOLLECTEDHEAP_HPP
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/memory/genOopClosures.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,30 @@
+/* Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "memory/genOopClosures.inline.hpp"
+#include "memory/iterator.inline.hpp"
+#include "memory/specialized_oop_closures.hpp"
+
+// Generate Serial GC specialized oop_oop_iterate functions.
+SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_S(ALL_KLASS_OOP_OOP_ITERATE_DEFN)
--- a/hotspot/src/share/vm/memory/genOopClosures.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/genOopClosures.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -31,7 +31,6 @@
 #include "memory/genOopClosures.hpp"
 #include "memory/genRemSet.hpp"
 #include "memory/generation.hpp"
-#include "memory/sharedHeap.hpp"
 #include "memory/space.hpp"
 
 inline OopsInGenClosure::OopsInGenClosure(Generation* gen) :
--- a/hotspot/src/share/vm/memory/generation.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/generation.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -187,7 +187,7 @@
   assert(obj_size == (size_t)obj->size(), "bad obj_size passed in");
 
 #ifndef PRODUCT
-  if (Universe::heap()->promotion_should_fail()) {
+  if (GenCollectedHeap::heap()->promotion_should_fail()) {
     return NULL;
   }
 #endif  // #ifndef PRODUCT
--- a/hotspot/src/share/vm/memory/iterator.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/iterator.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -23,8 +23,11 @@
  */
 
 #include "precompiled.hpp"
-#include "memory/iterator.hpp"
+#include "memory/iterator.inline.hpp"
+#include "memory/universe.hpp"
 #include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
 
 void KlassToOopClosure::do_klass(Klass* k) {
   assert(_oop_closure != NULL, "Not initialized?");
@@ -61,19 +64,18 @@
   }
 }
 
-MarkingCodeBlobClosure::MarkScope::MarkScope(bool activate)
-  : _active(activate)
-{
-  if (_active)  nmethod::oops_do_marking_prologue();
-}
-
-MarkingCodeBlobClosure::MarkScope::~MarkScope() {
-  if (_active)  nmethod::oops_do_marking_epilogue();
-}
-
 void MarkingCodeBlobClosure::do_code_blob(CodeBlob* cb) {
   nmethod* nm = cb->as_nmethod_or_null();
   if (nm != NULL && !nm->test_set_oops_do_mark()) {
     do_nmethod(nm);
   }
 }
+
+// Generate the *Klass::oop_oop_iterate functions for the base class
+// of the oop closures. These versions use the virtual do_oop calls,
+// instead of the devirtualized do_oop_nv version.
+ALL_KLASS_OOP_OOP_ITERATE_DEFN(ExtendedOopClosure,  _v)
+
+// Generate the *Klass::oop_oop_iterate functions
+// for the NoHeaderExtendedOopClosure helper class.
+ALL_KLASS_OOP_OOP_ITERATE_DEFN(NoHeaderExtendedOopClosure, _nv)
--- a/hotspot/src/share/vm/memory/iterator.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/iterator.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -44,9 +44,7 @@
 class OopClosure : public Closure {
  public:
   virtual void do_oop(oop* o) = 0;
-  virtual void do_oop_v(oop* o) { do_oop(o); }
   virtual void do_oop(narrowOop* o) = 0;
-  virtual void do_oop_v(narrowOop* o) { do_oop(o); }
 };
 
 // ExtendedOopClosure adds extra code to be run during oop iterations.
@@ -74,11 +72,9 @@
   // Currently, only CMS and G1 need these.
 
   virtual bool do_metadata() { return do_metadata_nv(); }
-  bool do_metadata_v()       { return do_metadata(); }
   bool do_metadata_nv()      { return false; }
 
   virtual void do_klass(Klass* k)   { do_klass_nv(k); }
-  void do_klass_v(Klass* k)         { do_klass(k); }
   void do_klass_nv(Klass* k)        { ShouldNotReachHere(); }
 
   virtual void do_class_loader_data(ClassLoaderData* cld) { ShouldNotReachHere(); }
@@ -87,6 +83,14 @@
   // location without an intervening "major reset" (like the end of a GC).
   virtual bool idempotent() { return false; }
   virtual bool apply_to_weak_ref_discovered_field() { return false; }
+
+#ifdef ASSERT
+  // Default verification of each visited oop field.
+  template <typename T> void verify(T* p);
+
+  // Can be used by subclasses to turn off the default verification of oop fields.
+  virtual bool should_verify_oops() { return true; }
+#endif
 };
 
 // Wrapper closure only used to implement oop_iterate_no_header().
@@ -147,7 +151,6 @@
 };
 
 class CLDToKlassAndOopClosure : public CLDClosure {
-  friend class SharedHeap;
   friend class G1CollectedHeap;
  protected:
   OopClosure*   _oop_closure;
@@ -284,16 +287,6 @@
   // Called for each code blob, but at most once per unique blob.
 
   virtual void do_code_blob(CodeBlob* cb);
-
-  class MarkScope : public StackObj {
-  protected:
-    bool _active;
-  public:
-    MarkScope(bool activate = true);
-      // = { if (active) nmethod::oops_do_marking_prologue(); }
-    ~MarkScope();
-      // = { if (active) nmethod::oops_do_marking_epilogue(); }
-  };
 };
 
 // MonitorClosure is used for iterating over monitors in the monitors cache
@@ -364,16 +357,33 @@
   }
 };
 
+// The two class template specializations are used to dispatch calls
+// to the ExtendedOopClosure functions. If use_non_virtual_call is true,
+// the non-virtual versions are called (E.g. do_oop_nv), otherwise the
+// virtual versions are called (E.g. do_oop).
 
-// Helper defines for ExtendOopClosure
+template <bool use_non_virtual_call>
+class Devirtualizer {};
 
-#define if_do_metadata_checked(closure, nv_suffix)       \
-  /* Make sure the non-virtual and the virtual versions match. */     \
-  assert(closure->do_metadata##nv_suffix() == closure->do_metadata(), \
-      "Inconsistency in do_metadata");                                \
-  if (closure->do_metadata##nv_suffix())
+// Dispatches to the non-virtual functions.
+template <> class Devirtualizer<true> {
+ public:
+  template <class OopClosureType, typename T> static void do_oop(OopClosureType* closure, T* p);
+  template <class OopClosureType>             static void do_klass(OopClosureType* closure, Klass* k);
+  template <class OopClosureType>             static bool do_metadata(OopClosureType* closure);
+};
 
-#define assert_should_ignore_metadata(closure, nv_suffix)                                  \
-  assert(!closure->do_metadata##nv_suffix(), "Code to handle metadata is not implemented")
+// Dispatches to the virtual functions.
+template <> class Devirtualizer<false> {
+ public:
+  template <class OopClosureType, typename T> static void do_oop(OopClosureType* closure, T* p);
+  template <class OopClosureType>             static void do_klass(OopClosureType* closure, Klass* k);
+  template <class OopClosureType>             static bool do_metadata(OopClosureType* closure);
+};
+
+// Helper to convert the oop iterate macro suffixes into bool values that can be used by template functions.
+#define nvs_nv_to_bool true
+#define nvs_v_to_bool  false
+#define nvs_to_bool(nv_suffix) nvs##nv_suffix##_to_bool
 
 #endif // SHARE_VM_MEMORY_ITERATOR_HPP
--- a/hotspot/src/share/vm/memory/iterator.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/iterator.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -28,6 +28,12 @@
 #include "classfile/classLoaderData.hpp"
 #include "memory/iterator.hpp"
 #include "oops/klass.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.inline.hpp"
+#include "oops/instanceClassLoaderKlass.inline.hpp"
+#include "oops/instanceRefKlass.inline.hpp"
+#include "oops/objArrayKlass.inline.hpp"
+#include "oops/typeArrayKlass.inline.hpp"
 #include "utilities/debug.hpp"
 
 inline void MetadataAwareOopClosure::do_class_loader_data(ClassLoaderData* cld) {
@@ -44,4 +50,63 @@
 
 inline void MetadataAwareOopClosure::do_klass(Klass* k)       { do_klass_nv(k); }
 
+#ifdef ASSERT
+// This verification is applied to all visited oops.
+// The closures can turn is off by overriding should_verify_oops().
+template <typename T>
+void ExtendedOopClosure::verify(T* p) {
+  if (should_verify_oops()) {
+    T heap_oop = oopDesc::load_heap_oop(p);
+    if (!oopDesc::is_null(heap_oop)) {
+      oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
+      assert(Universe::heap()->is_in_closed_subset(o),
+             err_msg("should be in closed *p " PTR_FORMAT " " PTR_FORMAT, p2i(p), p2i(o)));
+    }
+  }
+}
+#endif
+
+// Implementation of the non-virtual do_oop dispatch.
+
+template <class OopClosureType, typename T>
+inline void Devirtualizer<true>::do_oop(OopClosureType* closure, T* p) {
+  debug_only(closure->verify(p));
+  closure->do_oop_nv(p);
+}
+template <class OopClosureType>
+inline void Devirtualizer<true>::do_klass(OopClosureType* closure, Klass* k) {
+  closure->do_klass_nv(k);
+}
+template <class OopClosureType>
+inline bool Devirtualizer<true>::do_metadata(OopClosureType* closure) {
+  // Make sure the non-virtual and the virtual versions match.
+  assert(closure->do_metadata_nv() == closure->do_metadata(), "Inconsistency in do_metadata");
+  return closure->do_metadata_nv();
+}
+
+// Implementation of the virtual do_oop dispatch.
+
+template <class OopClosureType, typename T>
+void Devirtualizer<false>::do_oop(OopClosureType* closure, T* p) {
+  debug_only(closure->verify(p));
+  closure->do_oop(p);
+}
+template <class OopClosureType>
+void Devirtualizer<false>::do_klass(OopClosureType* closure, Klass* k) {
+  closure->do_klass(k);
+}
+template <class OopClosureType>
+bool Devirtualizer<false>::do_metadata(OopClosureType* closure) {
+  return closure->do_metadata();
+}
+
+// The list of all "specializable" oop_oop_iterate function definitions.
+#define ALL_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)                  \
+  ALL_INSTANCE_KLASS_OOP_OOP_ITERATE_DEFN(             OopClosureType, nv_suffix)  \
+  ALL_INSTANCE_REF_KLASS_OOP_OOP_ITERATE_DEFN(         OopClosureType, nv_suffix)  \
+  ALL_INSTANCE_MIRROR_KLASS_OOP_OOP_ITERATE_DEFN(      OopClosureType, nv_suffix)  \
+  ALL_INSTANCE_CLASS_LOADER_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  ALL_OBJ_ARRAY_KLASS_OOP_OOP_ITERATE_DEFN(            OopClosureType, nv_suffix)  \
+  ALL_TYPE_ARRAY_KLASS_OOP_OOP_ITERATE_DEFN(           OopClosureType, nv_suffix)
+
 #endif // SHARE_VM_MEMORY_ITERATOR_INLINE_HPP
--- a/hotspot/src/share/vm/memory/sharedHeap.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,94 +0,0 @@
-/*
- * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#include "precompiled.hpp"
-#include "classfile/stringTable.hpp"
-#include "classfile/systemDictionary.hpp"
-#include "code/codeCache.hpp"
-#include "gc_interface/collectedHeap.inline.hpp"
-#include "memory/sharedHeap.hpp"
-#include "oops/oop.inline.hpp"
-#include "runtime/atomic.inline.hpp"
-#include "runtime/fprofiler.hpp"
-#include "runtime/java.hpp"
-#include "utilities/copy.hpp"
-#include "utilities/workgroup.hpp"
-
-SharedHeap* SharedHeap::_sh;
-
-SharedHeap::SharedHeap() :
-  CollectedHeap(),
-  _workers(NULL)
-{
-  _sh = this;  // ch is static, should be set only once.
-  if (UseConcMarkSweepGC || UseG1GC) {
-    _workers = new FlexibleWorkGang("GC Thread", ParallelGCThreads,
-                            /* are_GC_task_threads */true,
-                            /* are_ConcurrentGC_threads */false);
-    if (_workers == NULL) {
-      vm_exit_during_initialization("Failed necessary allocation.");
-    } else {
-      _workers->initialize_workers();
-    }
-  }
-}
-
-bool SharedHeap::heap_lock_held_for_gc() {
-  Thread* t = Thread::current();
-  return    Heap_lock->owned_by_self()
-         || (   (t->is_GC_task_thread() ||  t->is_VM_thread())
-             && _thread_holds_heap_lock_for_gc);
-}
-
-void SharedHeap::set_par_threads(uint t) {
-  assert(t == 0 || !UseSerialGC, "Cannot have parallel threads");
-  _n_par_threads = t;
-}
-
-SharedHeap::StrongRootsScope::StrongRootsScope(SharedHeap* heap, bool activate)
-  : MarkScope(activate), _sh(heap)
-{
-  if (_active) {
-    Threads::change_thread_claim_parity();
-    // Zero the claimed high water mark in the StringTable
-    StringTable::clear_parallel_claimed_index();
-  }
-}
-
-SharedHeap::StrongRootsScope::~StrongRootsScope() {
-  Threads::assert_all_threads_claimed();
-}
-
-void SharedHeap::set_barrier_set(BarrierSet* bs) {
-  _barrier_set = bs;
-  // Cached barrier set for fast access in oops
-  oopDesc::set_bs(bs);
-}
-
-void SharedHeap::post_initialize() {
-  CollectedHeap::post_initialize();
-  ref_processing_init();
-}
-
-void SharedHeap::ref_processing_init() {}
--- a/hotspot/src/share/vm/memory/sharedHeap.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,213 +0,0 @@
-/*
- * Copyright (c) 2000, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#ifndef SHARE_VM_MEMORY_SHAREDHEAP_HPP
-#define SHARE_VM_MEMORY_SHAREDHEAP_HPP
-
-#include "gc_interface/collectedHeap.hpp"
-#include "memory/generation.hpp"
-
-// A "SharedHeap" is an implementation of a java heap for HotSpot.  This
-// is an abstract class: there may be many different kinds of heaps.  This
-// class defines the functions that a heap must implement, and contains
-// infrastructure common to all heaps.
-
-class Generation;
-class BarrierSet;
-class GenRemSet;
-class Space;
-class SpaceClosure;
-class OopClosure;
-class OopsInGenClosure;
-class ObjectClosure;
-class SubTasksDone;
-class WorkGang;
-class FlexibleWorkGang;
-class CollectorPolicy;
-class KlassClosure;
-
-// Note on use of FlexibleWorkGang's for GC.
-// There are three places where task completion is determined.
-// In
-//    1) ParallelTaskTerminator::offer_termination() where _n_threads
-//    must be set to the correct value so that count of workers that
-//    have offered termination will exactly match the number
-//    working on the task.  Tasks such as those derived from GCTask
-//    use ParallelTaskTerminator's.  Tasks that want load balancing
-//    by work stealing use this method to gauge completion.
-//    2) SubTasksDone has a variable _n_threads that is used in
-//    all_tasks_completed() to determine completion.  all_tasks_complete()
-//    counts the number of tasks that have been done and then reset
-//    the SubTasksDone so that it can be used again.  When the number of
-//    tasks is set to the number of GC workers, then _n_threads must
-//    be set to the number of active GC workers. G1RootProcessor and
-//    GenCollectedHeap have SubTasksDone.
-//    3) SequentialSubTasksDone has an _n_threads that is used in
-//    a way similar to SubTasksDone and has the same dependency on the
-//    number of active GC workers.  CompactibleFreeListSpace and Space
-//    have SequentialSubTasksDone's.
-//
-// Examples of using SubTasksDone and SequentialSubTasksDone:
-//  G1RootProcessor and GenCollectedHeap::process_roots() use
-//  SubTasksDone* _process_strong_tasks to claim tasks for workers
-//
-//  GenCollectedHeap::gen_process_roots() calls
-//      rem_set()->younger_refs_iterate()
-//  to scan the card table and which eventually calls down into
-//  CardTableModRefBS::par_non_clean_card_iterate_work().  This method
-//  uses SequentialSubTasksDone* _pst to claim tasks.
-//  Both SubTasksDone and SequentialSubTasksDone call their method
-//  all_tasks_completed() to count the number of GC workers that have
-//  finished their work.  That logic is "when all the workers are
-//  finished the tasks are finished".
-//
-//  The pattern that appears  in the code is to set _n_threads
-//  to a value > 1 before a task that you would like executed in parallel
-//  and then to set it to 0 after that task has completed.  A value of
-//  0 is a "special" value in set_n_threads() which translates to
-//  setting _n_threads to 1.
-//
-//  Some code uses _n_termination to decide if work should be done in
-//  parallel.  The notorious possibly_parallel_oops_do() in threads.cpp
-//  is an example of such code.  Look for variable "is_par" for other
-//  examples.
-//
-//  The active_workers is not reset to 0 after a parallel phase.  It's
-//  value may be used in later phases and in one instance at least
-//  (the parallel remark) it has to be used (the parallel remark depends
-//  on the partitioning done in the previous parallel scavenge).
-
-class SharedHeap : public CollectedHeap {
-  friend class VMStructs;
-
-  friend class VM_GC_Operation;
-  friend class VM_CGC_Operation;
-
-protected:
-  // There should be only a single instance of "SharedHeap" in a program.
-  // This is enforced with the protected constructor below, which will also
-  // set the static pointer "_sh" to that instance.
-  static SharedHeap* _sh;
-
-  // If we're doing parallel GC, use this gang of threads.
-  FlexibleWorkGang* _workers;
-
-  // Full initialization is done in a concrete subtype's "initialize"
-  // function.
-  SharedHeap();
-
-  // Returns true if the calling thread holds the heap lock,
-  // or the calling thread is a par gc thread and the heap_lock is held
-  // by the vm thread doing a gc operation.
-  bool heap_lock_held_for_gc();
-  // True if the heap_lock is held by the a non-gc thread invoking a gc
-  // operation.
-  bool _thread_holds_heap_lock_for_gc;
-
-public:
-  static SharedHeap* heap() { return _sh; }
-
-  void set_barrier_set(BarrierSet* bs);
-
-  // Does operations required after initialization has been done.
-  virtual void post_initialize();
-
-  // Initialization of ("weak") reference processing support
-  virtual void ref_processing_init();
-
-  // Iteration functions.
-  void oop_iterate(ExtendedOopClosure* cl) = 0;
-
-  // Iterate over all spaces in use in the heap, in an undefined order.
-  virtual void space_iterate(SpaceClosure* cl) = 0;
-
-  // A SharedHeap will contain some number of spaces.  This finds the
-  // space whose reserved area contains the given address, or else returns
-  // NULL.
-  virtual Space* space_containing(const void* addr) const = 0;
-
-  bool no_gc_in_progress() { return !is_gc_active(); }
-
-  // Note, the below comment needs to be updated to reflect the changes
-  // introduced by JDK-8076225. This should be done as part of JDK-8076289.
-  //
-  //Some collectors will perform "process_strong_roots" in parallel.
-  // Such a call will involve claiming some fine-grained tasks, such as
-  // scanning of threads.  To make this process simpler, we provide the
-  // "strong_roots_parity()" method.  Collectors that start parallel tasks
-  // whose threads invoke "process_strong_roots" must
-  // call "change_strong_roots_parity" in sequential code starting such a
-  // task.  (This also means that a parallel thread may only call
-  // process_strong_roots once.)
-  //
-  // For calls to process_roots by sequential code, the parity is
-  // updated automatically.
-  //
-  // The idea is that objects representing fine-grained tasks, such as
-  // threads, will contain a "parity" field.  A task will is claimed in the
-  // current "process_roots" call only if its parity field is the
-  // same as the "strong_roots_parity"; task claiming is accomplished by
-  // updating the parity field to the strong_roots_parity with a CAS.
-  //
-  // If the client meats this spec, then strong_roots_parity() will have
-  // the following properties:
-  //   a) to return a different value than was returned before the last
-  //      call to change_strong_roots_parity, and
-  //   c) to never return a distinguished value (zero) with which such
-  //      task-claiming variables may be initialized, to indicate "never
-  //      claimed".
- public:
-
-  // Call these in sequential code around process_roots.
-  // strong_roots_prologue calls change_strong_roots_parity, if
-  // parallel tasks are enabled.
-  class StrongRootsScope : public MarkingCodeBlobClosure::MarkScope {
-    SharedHeap*   _sh;
-
-   public:
-    StrongRootsScope(SharedHeap* heap, bool activate = true);
-    ~StrongRootsScope();
-  };
-
- private:
-
- public:
-  FlexibleWorkGang* workers() const { return _workers; }
-
-  // The functions below are helper functions that a subclass of
-  // "SharedHeap" can use in the implementation of its virtual
-  // functions.
-
-public:
-
-  // Do anything common to GC's.
-  virtual void gc_prologue(bool full) = 0;
-  virtual void gc_epilogue(bool full) = 0;
-
-  // Sets the number of parallel threads that will be doing tasks
-  // (such as process roots) subsequently.
-  virtual void set_par_threads(uint t);
-};
-
-#endif // SHARE_VM_MEMORY_SHAREDHEAP_HPP
--- a/hotspot/src/share/vm/memory/space.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/space.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -31,6 +31,7 @@
 #include "memory/blockOffsetTable.inline.hpp"
 #include "memory/defNewGeneration.hpp"
 #include "memory/genCollectedHeap.hpp"
+#include "memory/genOopClosures.inline.hpp"
 #include "memory/space.hpp"
 #include "memory/space.inline.hpp"
 #include "memory/universe.inline.hpp"
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/memory/strongRootsScope.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,53 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "classfile/stringTable.hpp"
+#include "code/nmethod.hpp"
+#include "memory/strongRootsScope.hpp"
+#include "runtime/thread.hpp"
+
+MarkScope::MarkScope(bool activate) : _active(activate) {
+  if (_active) {
+    nmethod::oops_do_marking_prologue();
+  }
+}
+
+MarkScope::~MarkScope() {
+  if (_active) {
+    nmethod::oops_do_marking_epilogue();
+  }
+}
+
+StrongRootsScope::StrongRootsScope(bool activate) : MarkScope(activate) {
+  if (_active) {
+    Threads::change_thread_claim_parity();
+    // Zero the claimed high water mark in the StringTable
+    StringTable::clear_parallel_claimed_index();
+  }
+}
+
+StrongRootsScope::~StrongRootsScope() {
+  Threads::assert_all_threads_claimed();
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/memory/strongRootsScope.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,46 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_MEMORY_STRONGROOTSSCOPE_HPP
+#define SHARE_VM_MEMORY_STRONGROOTSSCOPE_HPP
+
+#include "memory/allocation.hpp"
+
+class MarkScope : public StackObj {
+ protected:
+  bool _active;
+ public:
+  MarkScope(bool activate = true);
+  ~MarkScope();
+};
+
+// Sets up and tears down the required state for parallel root processing.
+
+class StrongRootsScope : public MarkScope {
+ public:
+  StrongRootsScope(bool activate = true);
+  ~StrongRootsScope();
+};
+
+#endif // SHARE_VM_MEMORY_STRONGROOTSSCOPE_HPP
--- a/hotspot/src/share/vm/memory/tenuredGeneration.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/memory/tenuredGeneration.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -36,6 +36,9 @@
 #include "oops/oop.inline.hpp"
 #include "runtime/java.hpp"
 #include "utilities/macros.hpp"
+#if INCLUDE_ALL_GCS
+#include "gc_implementation/parNew/parOopClosures.hpp"
+#endif
 
 TenuredGeneration::TenuredGeneration(ReservedSpace rs,
                                      size_t initial_byte_size, int level,
--- a/hotspot/src/share/vm/oops/instanceClassLoaderKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,146 +0,0 @@
-/*
- * Copyright (c) 2011, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#include "precompiled.hpp"
-#include "classfile/javaClasses.hpp"
-#include "classfile/systemDictionary.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
-#include "gc_interface/collectedHeap.inline.hpp"
-#include "memory/genOopClosures.inline.hpp"
-#include "memory/iterator.inline.hpp"
-#include "memory/oopFactory.hpp"
-#include "memory/specialized_oop_closures.hpp"
-#include "oops/instanceKlass.hpp"
-#include "oops/instanceClassLoaderKlass.hpp"
-#include "oops/instanceMirrorKlass.hpp"
-#include "oops/instanceOop.hpp"
-#include "oops/oop.inline.hpp"
-#include "oops/symbol.hpp"
-#include "runtime/handles.inline.hpp"
-#include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/parNew/parOopClosures.inline.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
-#endif // INCLUDE_ALL_GCS
-
-// Macro to define InstanceClassLoaderKlass::oop_oop_iterate for virtual/nonvirtual for
-// all closures.  Macros calling macros above for each oop size.
-// Since ClassLoader objects have only a pointer to the loader_data, they are not
-// compressed nor does the pointer move.
-
-#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)\
-                                                                                \
-int InstanceClassLoaderKlass::                                                  \
-oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {                  \
-  /* Get size before changing pointers */                                       \
-  int size = InstanceKlass::oop_oop_iterate##nv_suffix(obj, closure);           \
-                                                                                \
-  if_do_metadata_checked(closure, nv_suffix) {                                  \
-    ClassLoaderData* cld = java_lang_ClassLoader::loader_data(obj);             \
-    /* cld can be null if we have a non-registered class loader. */             \
-    if (cld != NULL) {                                                          \
-      closure->do_class_loader_data(cld);                                       \
-    }                                                                           \
-  }                                                                             \
-                                                                                \
-  return size;                                                                  \
-}
-
-#if INCLUDE_ALL_GCS
-#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix) \
-                                                                                \
-int InstanceClassLoaderKlass::                                                  \
-oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {        \
-  /* Get size before changing pointers */                                       \
-  int size = InstanceKlass::oop_oop_iterate_backwards##nv_suffix(obj, closure); \
-  return size;                                                                  \
-}
-#endif // INCLUDE_ALL_GCS
-
-
-#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)      \
-                                                                                \
-int InstanceClassLoaderKlass::                                                  \
-oop_oop_iterate##nv_suffix##_m(oop obj,                                         \
-                               OopClosureType* closure,                         \
-                               MemRegion mr) {                                  \
-  int size = InstanceKlass::oop_oop_iterate##nv_suffix##_m(obj, closure, mr);   \
-                                                                                \
-  if_do_metadata_checked(closure, nv_suffix) {                                  \
-    if (mr.contains(obj)) {                                                     \
-      ClassLoaderData* cld = java_lang_ClassLoader::loader_data(obj);           \
-      /* cld can be null if we have a non-registered class loader. */           \
-      if (cld != NULL) {                                                        \
-        closure->do_class_loader_data(cld);                                     \
-      }                                                                         \
-    }                                                                           \
-  }                                                                             \
-                                                                                \
-  return size;                                                                  \
-}
-
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN)
-#if INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-#endif // INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN_m)
-
-void InstanceClassLoaderKlass::oop_follow_contents(oop obj) {
-  InstanceKlass::oop_follow_contents(obj);
-  ClassLoaderData * const loader_data = java_lang_ClassLoader::loader_data(obj);
-
-  // We must NULL check here, since the class loader
-  // can be found before the loader data has been set up.
-  if(loader_data != NULL) {
-    MarkSweep::follow_class_loader(loader_data);
-  }
-}
-
-#if INCLUDE_ALL_GCS
-void InstanceClassLoaderKlass::oop_follow_contents(ParCompactionManager* cm,
-        oop obj) {
-  InstanceKlass::oop_follow_contents(cm, obj);
-  ClassLoaderData * const loader_data = java_lang_ClassLoader::loader_data(obj);
-  if (loader_data != NULL) {
-    PSParallelCompact::follow_class_loader(cm, loader_data);
-  }
-}
-
-void InstanceClassLoaderKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  InstanceKlass::oop_push_contents(pm, obj);
-
-  // This is called by the young collector. It will already have taken care of
-  // all class loader data. So, we don't have to follow the class loader ->
-  // class loader data link.
-}
-
-int InstanceClassLoaderKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  InstanceKlass::oop_update_pointers(cm, obj);
-  return size_helper();
-}
-#endif // INCLUDE_ALL_GCS
-
--- a/hotspot/src/share/vm/oops/instanceClassLoaderKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceClassLoaderKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -48,34 +48,60 @@
 
   InstanceClassLoaderKlass() { assert(DumpSharedSpaces || UseSharedSpaces, "only for CDS"); }
 
-  // Iterators
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_v_m(obj, blk, mr);
-  }
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
+
+  // Oop fields (and metadata) iterators
+  //  [nv = true]  Use non-virtual calls to do_oop_nv.
+  //  [nv = false] Use virtual calls to do_oop.
+  //
+  // The InstanceClassLoaderKlass iterators also visit the CLD pointer (or mirror of anonymous klasses.)
 
-#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)                \
-  int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);                         \
+ private:
+  // Forward iteration
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate(oop obj, OopClosureType* closure);
+
+#if INCLUDE_ALL_GCS
+  // Reverse iteration
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_reverse(oop obj, OopClosureType* closure);
+#endif
+
+  // Bounded range iteration
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+ public:
+
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)   \
+  int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);                    \
   int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* blk, MemRegion mr);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceClassLoaderKlass_OOP_OOP_ITERATE_DECL)
 
 #if INCLUDE_ALL_GCS
-#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)      \
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)  \
   int oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* blk);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
 #endif // INCLUDE_ALL_GCS
 
-    // Garbage collection
-  void oop_follow_contents(oop obj);
-
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
 };
 
 #endif // SHARE_VM_OOPS_INSTANCECLASSLOADERKLASS_HPP
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/oops/instanceClassLoaderKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,110 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_OOPS_INSTANCECLASSLOADERKLASS_INLINE_HPP
+#define SHARE_VM_OOPS_INSTANCECLASSLOADERKLASS_INLINE_HPP
+
+#include "classfile/javaClasses.hpp"
+#include "oops/instanceClassLoaderKlass.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
+#include "utilities/macros.hpp"
+
+template <bool nv, class OopClosureType>
+inline int InstanceClassLoaderKlass::oop_oop_iterate(oop obj, OopClosureType* closure) {
+  int size = InstanceKlass::oop_oop_iterate<nv>(obj, closure);
+
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    ClassLoaderData* cld = java_lang_ClassLoader::loader_data(obj);
+    // cld can be null if we have a non-registered class loader.
+    if (cld != NULL) {
+      closure->do_class_loader_data(cld);
+    }
+  }
+
+  return size;
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, class OopClosureType>
+inline int InstanceClassLoaderKlass::oop_oop_iterate_reverse(oop obj, OopClosureType* closure) {
+  int size = InstanceKlass::oop_oop_iterate_reverse<nv>(obj, closure);
+
+  assert(!Devirtualizer<nv>::do_metadata(closure),
+      "Code to handle metadata is not implemented");
+
+  return size;
+}
+#endif // INCLUDE_ALL_GCS
+
+
+template <bool nv, class OopClosureType>
+inline int InstanceClassLoaderKlass::oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  int size = InstanceKlass::oop_oop_iterate_bounded<nv>(obj, closure, mr);
+
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    if (mr.contains(obj)) {
+      ClassLoaderData* cld = java_lang_ClassLoader::loader_data(obj);
+      // cld can be null if we have a non-registered class loader.
+      if (cld != NULL) {
+        closure->do_class_loader_data(cld);
+      }
+    }
+  }
+
+  return size;
+}
+
+
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)              \
+                                                                                              \
+int InstanceClassLoaderKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                               \
+}
+
+#if INCLUDE_ALL_GCS
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
+                                                                                                        \
+int InstanceClassLoaderKlass::oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate_reverse<nvs_to_bool(nv_suffix)>(obj, closure);                                 \
+}
+#else
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+
+#define InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)                              \
+                                                                                                                \
+int InstanceClassLoaderKlass::oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_bounded<nvs_to_bool(nv_suffix)>(obj, closure, mr);                                     \
+}
+
+#define ALL_INSTANCE_CLASS_LOADER_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)     \
+  InstanceClassLoaderKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)     \
+  InstanceClassLoaderKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+
+#endif // SHARE_VM_OOPS_INSTANCECLASSLOADERKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/oops/instanceKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceKlass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -28,12 +28,10 @@
 #include "classfile/verifier.hpp"
 #include "classfile/vmSymbols.hpp"
 #include "compiler/compileBroker.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
 #include "interpreter/oopMapCache.hpp"
 #include "interpreter/rewriter.hpp"
 #include "jvmtifiles/jvmti.h"
-#include "memory/genOopClosures.inline.hpp"
 #include "memory/heapInspection.hpp"
 #include "memory/iterator.inline.hpp"
 #include "memory/metadataFactory.hpp"
@@ -41,7 +39,7 @@
 #include "memory/specialized_oop_closures.hpp"
 #include "oops/fieldStreams.hpp"
 #include "oops/instanceClassLoaderKlass.hpp"
-#include "oops/instanceKlass.hpp"
+#include "oops/instanceKlass.inline.hpp"
 #include "oops/instanceMirrorKlass.hpp"
 #include "oops/instanceOop.hpp"
 #include "oops/klass.inline.hpp"
@@ -64,17 +62,6 @@
 #include "services/threadService.hpp"
 #include "utilities/dtrace.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/concurrentMarkSweep/cmsOopClosures.inline.hpp"
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
-#include "gc_implementation/g1/g1OopClosures.inline.hpp"
-#include "gc_implementation/g1/g1RemSet.inline.hpp"
-#include "gc_implementation/g1/heapRegionManager.inline.hpp"
-#include "gc_implementation/parNew/parOopClosures.inline.hpp"
-#include "gc_implementation/parallelScavenge/parallelScavengeHeap.inline.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.inline.hpp"
-#endif // INCLUDE_ALL_GCS
 #ifdef COMPILER1
 #include "c1/c1_Compiler.hpp"
 #endif
@@ -1993,288 +1980,6 @@
 }
 #endif //PRODUCT
 
-
-// Garbage collection
-
-#ifdef ASSERT
-template <class T> void assert_is_in(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in(o), "should be in heap");
-  }
-}
-template <class T> void assert_is_in_closed_subset(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in_closed_subset(o),
-           err_msg("should be in closed *p " INTPTR_FORMAT " " INTPTR_FORMAT, (address)p, (address)o));
-  }
-}
-template <class T> void assert_is_in_reserved(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in_reserved(o), "should be in reserved");
-  }
-}
-template <class T> void assert_nothing(T *p) {}
-
-#else
-template <class T> void assert_is_in(T *p) {}
-template <class T> void assert_is_in_closed_subset(T *p) {}
-template <class T> void assert_is_in_reserved(T *p) {}
-template <class T> void assert_nothing(T *p) {}
-#endif // ASSERT
-
-//
-// Macros that iterate over areas of oops which are specialized on type of
-// oop pointer either narrow or wide, depending on UseCompressedOops
-//
-// Parameters are:
-//   T         - type of oop to point to (either oop or narrowOop)
-//   start_p   - starting pointer for region to iterate over
-//   count     - number of oops or narrowOops to iterate over
-//   do_oop    - action to perform on each oop (it's arbitrary C code which
-//               makes it more efficient to put in a macro rather than making
-//               it a template function)
-//   assert_fn - assert function which is template function because performance
-//               doesn't matter when enabled.
-#define InstanceKlass_SPECIALIZED_OOP_ITERATE( \
-  T, start_p, count, do_oop,                \
-  assert_fn)                                \
-{                                           \
-  T* p         = (T*)(start_p);             \
-  T* const end = p + (count);               \
-  while (p < end) {                         \
-    (assert_fn)(p);                         \
-    do_oop;                                 \
-    ++p;                                    \
-  }                                         \
-}
-
-#define InstanceKlass_SPECIALIZED_OOP_REVERSE_ITERATE( \
-  T, start_p, count, do_oop,                \
-  assert_fn)                                \
-{                                           \
-  T* const start = (T*)(start_p);           \
-  T*       p     = start + (count);         \
-  while (start < p) {                       \
-    --p;                                    \
-    (assert_fn)(p);                         \
-    do_oop;                                 \
-  }                                         \
-}
-
-#define InstanceKlass_SPECIALIZED_BOUNDED_OOP_ITERATE( \
-  T, start_p, count, low, high,             \
-  do_oop, assert_fn)                        \
-{                                           \
-  T* const l = (T*)(low);                   \
-  T* const h = (T*)(high);                  \
-  assert(mask_bits((intptr_t)l, sizeof(T)-1) == 0 && \
-         mask_bits((intptr_t)h, sizeof(T)-1) == 0,   \
-         "bounded region must be properly aligned"); \
-  T* p       = (T*)(start_p);               \
-  T* end     = p + (count);                 \
-  if (p < l) p = l;                         \
-  if (end > h) end = h;                     \
-  while (p < end) {                         \
-    (assert_fn)(p);                         \
-    do_oop;                                 \
-    ++p;                                    \
-  }                                         \
-}
-
-
-// The following macros call specialized macros, passing either oop or
-// narrowOop as the specialization type.  These test the UseCompressedOops
-// flag.
-#define InstanceKlass_OOP_MAP_ITERATE(obj, do_oop, assert_fn)            \
-{                                                                        \
-  /* Compute oopmap block range. The common case                         \
-     is nonstatic_oop_map_size == 1. */                                  \
-  OopMapBlock* map           = start_of_nonstatic_oop_maps();            \
-  OopMapBlock* const end_map = map + nonstatic_oop_map_count();          \
-  if (UseCompressedOops) {                                               \
-    while (map < end_map) {                                              \
-      InstanceKlass_SPECIALIZED_OOP_ITERATE(narrowOop,                   \
-        obj->obj_field_addr<narrowOop>(map->offset()), map->count(),     \
-        do_oop, assert_fn)                                               \
-      ++map;                                                             \
-    }                                                                    \
-  } else {                                                               \
-    while (map < end_map) {                                              \
-      InstanceKlass_SPECIALIZED_OOP_ITERATE(oop,                         \
-        obj->obj_field_addr<oop>(map->offset()), map->count(),           \
-        do_oop, assert_fn)                                               \
-      ++map;                                                             \
-    }                                                                    \
-  }                                                                      \
-}
-
-#define InstanceKlass_OOP_MAP_REVERSE_ITERATE(obj, do_oop, assert_fn)    \
-{                                                                        \
-  OopMapBlock* const start_map = start_of_nonstatic_oop_maps();          \
-  OopMapBlock* map             = start_map + nonstatic_oop_map_count();  \
-  if (UseCompressedOops) {                                               \
-    while (start_map < map) {                                            \
-      --map;                                                             \
-      InstanceKlass_SPECIALIZED_OOP_REVERSE_ITERATE(narrowOop,           \
-        obj->obj_field_addr<narrowOop>(map->offset()), map->count(),     \
-        do_oop, assert_fn)                                               \
-    }                                                                    \
-  } else {                                                               \
-    while (start_map < map) {                                            \
-      --map;                                                             \
-      InstanceKlass_SPECIALIZED_OOP_REVERSE_ITERATE(oop,                 \
-        obj->obj_field_addr<oop>(map->offset()), map->count(),           \
-        do_oop, assert_fn)                                               \
-    }                                                                    \
-  }                                                                      \
-}
-
-#define InstanceKlass_BOUNDED_OOP_MAP_ITERATE(obj, low, high, do_oop,    \
-                                              assert_fn)                 \
-{                                                                        \
-  /* Compute oopmap block range. The common case is                      \
-     nonstatic_oop_map_size == 1, so we accept the                       \
-     usually non-existent extra overhead of examining                    \
-     all the maps. */                                                    \
-  OopMapBlock* map           = start_of_nonstatic_oop_maps();            \
-  OopMapBlock* const end_map = map + nonstatic_oop_map_count();          \
-  if (UseCompressedOops) {                                               \
-    while (map < end_map) {                                              \
-      InstanceKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(narrowOop,           \
-        obj->obj_field_addr<narrowOop>(map->offset()), map->count(),     \
-        low, high,                                                       \
-        do_oop, assert_fn)                                               \
-      ++map;                                                             \
-    }                                                                    \
-  } else {                                                               \
-    while (map < end_map) {                                              \
-      InstanceKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(oop,                 \
-        obj->obj_field_addr<oop>(map->offset()), map->count(),           \
-        low, high,                                                       \
-        do_oop, assert_fn)                                               \
-      ++map;                                                             \
-    }                                                                    \
-  }                                                                      \
-}
-
-void InstanceKlass::oop_follow_contents(oop obj) {
-  assert(obj != NULL, "can't follow the content of NULL object");
-  MarkSweep::follow_klass(obj->klass());
-  InstanceKlass_OOP_MAP_ITERATE( \
-    obj, \
-    MarkSweep::mark_and_push(p), \
-    assert_is_in_closed_subset)
-}
-
-#if INCLUDE_ALL_GCS
-void InstanceKlass::oop_follow_contents(ParCompactionManager* cm,
-                                        oop obj) {
-  assert(obj != NULL, "can't follow the content of NULL object");
-  PSParallelCompact::follow_klass(cm, obj->klass());
-  // Only mark the header and let the scan of the meta-data mark
-  // everything else.
-  InstanceKlass_OOP_MAP_ITERATE( \
-    obj, \
-    PSParallelCompact::mark_and_push(cm, p), \
-    assert_is_in)
-}
-#endif // INCLUDE_ALL_GCS
-
-// closure's do_metadata() method dictates whether the given closure should be
-// applied to the klass ptr in the object header.
-
-#define InstanceKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)        \
-                                                                             \
-int InstanceKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) { \
-  /* header */                                                          \
-  if_do_metadata_checked(closure, nv_suffix) {                          \
-    closure->do_klass##nv_suffix(obj->klass());                         \
-  }                                                                     \
-  InstanceKlass_OOP_MAP_ITERATE(                                        \
-    obj,                                                                \
-    (closure)->do_oop##nv_suffix(p),                                    \
-    assert_is_in_closed_subset)                                         \
-  return size_helper();                                                 \
-}
-
-#if INCLUDE_ALL_GCS
-#define InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix) \
-                                                                                \
-int InstanceKlass::oop_oop_iterate_backwards##nv_suffix(oop obj,                \
-                                              OopClosureType* closure) {        \
-  assert_should_ignore_metadata(closure, nv_suffix);                            \
-                                                                                \
-  /* instance variables */                                                      \
-  InstanceKlass_OOP_MAP_REVERSE_ITERATE(                                        \
-    obj,                                                                        \
-    (closure)->do_oop##nv_suffix(p),                                            \
-    assert_is_in_closed_subset)                                                 \
-   return size_helper();                                                        \
-}
-#endif // INCLUDE_ALL_GCS
-
-#define InstanceKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix) \
-                                                                        \
-int InstanceKlass::oop_oop_iterate##nv_suffix##_m(oop obj,              \
-                                                  OopClosureType* closure, \
-                                                  MemRegion mr) {          \
-  if_do_metadata_checked(closure, nv_suffix) {                           \
-    if (mr.contains(obj)) {                                              \
-      closure->do_klass##nv_suffix(obj->klass());                        \
-    }                                                                    \
-  }                                                                      \
-  InstanceKlass_BOUNDED_OOP_MAP_ITERATE(                                 \
-    obj, mr.start(), mr.end(),                                           \
-    (closure)->do_oop##nv_suffix(p),                                     \
-    assert_is_in_closed_subset)                                          \
-  return size_helper();                                                  \
-}
-
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceKlass_OOP_OOP_ITERATE_DEFN_m)
-#if INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-#endif // INCLUDE_ALL_GCS
-
-int InstanceKlass::oop_adjust_pointers(oop obj) {
-  int size = size_helper();
-  InstanceKlass_OOP_MAP_ITERATE( \
-    obj, \
-    MarkSweep::adjust_pointer(p), \
-    assert_is_in)
-  return size;
-}
-
-#if INCLUDE_ALL_GCS
-void InstanceKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  InstanceKlass_OOP_MAP_REVERSE_ITERATE( \
-    obj, \
-    if (PSScavenge::should_scavenge(p)) { \
-      pm->claim_or_forward_depth(p); \
-    }, \
-    assert_nothing )
-}
-
-int InstanceKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  int size = size_helper();
-  InstanceKlass_OOP_MAP_ITERATE( \
-    obj, \
-    PSParallelCompact::adjust_pointer(p), \
-    assert_is_in)
-  return size;
-}
-
-#endif // INCLUDE_ALL_GCS
-
 void InstanceKlass::clean_implementors_list(BoolObjectClosure* is_alive) {
   assert(class_loader_data()->is_alive(is_alive), "this klass should be live");
   if (is_interface()) {
@@ -2695,6 +2400,57 @@
   return false;
 }
 
+bool InstanceKlass::find_inner_classes_attr(instanceKlassHandle k, int* ooff, int* noff, TRAPS) {
+  constantPoolHandle i_cp(THREAD, k->constants());
+  for (InnerClassesIterator iter(k); !iter.done(); iter.next()) {
+    int ioff = iter.inner_class_info_index();
+    if (ioff != 0) {
+      // Check to see if the name matches the class we're looking for
+      // before attempting to find the class.
+      if (i_cp->klass_name_at_matches(k, ioff)) {
+        Klass* inner_klass = i_cp->klass_at(ioff, CHECK_false);
+        if (k() == inner_klass) {
+          *ooff = iter.outer_class_info_index();
+          *noff = iter.inner_name_index();
+          return true;
+        }
+      }
+    }
+  }
+  return false;
+}
+
+Klass* InstanceKlass::compute_enclosing_class_impl(instanceKlassHandle k, bool* inner_is_member, TRAPS) {
+  instanceKlassHandle outer_klass;
+  *inner_is_member = false;
+  int ooff = 0, noff = 0;
+  if (find_inner_classes_attr(k, &ooff, &noff, THREAD)) {
+    constantPoolHandle i_cp(THREAD, k->constants());
+    if (ooff != 0) {
+      Klass* ok = i_cp->klass_at(ooff, CHECK_NULL);
+      outer_klass = instanceKlassHandle(THREAD, ok);
+      *inner_is_member = true;
+    }
+    if (outer_klass.is_null()) {
+      // It may be anonymous; try for that.
+      int encl_method_class_idx = k->enclosing_method_class_index();
+      if (encl_method_class_idx != 0) {
+        Klass* ok = i_cp->klass_at(encl_method_class_idx, CHECK_NULL);
+        outer_klass = instanceKlassHandle(THREAD, ok);
+        *inner_is_member = false;
+      }
+    }
+  }
+
+  // If no inner class attribute found for this class.
+  if (outer_klass.is_null())  return NULL;
+
+  // Throws an exception if outer klass has not declared k as an inner klass
+  // We need evidence that each klass knows about the other, or else
+  // the system could allow a spoof of an inner class to gain access rights.
+  Reflection::check_for_inner_class(outer_klass, k, *inner_is_member, CHECK_NULL);
+  return outer_klass();
+}
 
 jint InstanceKlass::compute_modifier_flags(TRAPS) const {
   jint access = access_flags().as_int();
--- a/hotspot/src/share/vm/oops/instanceKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -406,13 +406,17 @@
   bool is_same_class_package(oop classloader2, Symbol* classname2);
   static bool is_same_class_package(oop class_loader1, Symbol* class_name1, oop class_loader2, Symbol* class_name2);
 
-  // find an enclosing class (defined where original code was, in jvm.cpp!)
+  // find an enclosing class
   Klass* compute_enclosing_class(bool* inner_is_member, TRAPS) {
     instanceKlassHandle self(THREAD, this);
     return compute_enclosing_class_impl(self, inner_is_member, THREAD);
   }
   static Klass* compute_enclosing_class_impl(instanceKlassHandle self,
-                                               bool* inner_is_member, TRAPS);
+                                             bool* inner_is_member, TRAPS);
+
+  // Find InnerClasses attribute for k and return outer_class_info_index & inner_name_index.
+  static bool find_inner_classes_attr(instanceKlassHandle k,
+                                      int* ooff, int* noff, TRAPS);
 
   // tell if two classes have the same enclosing class (at package level)
   bool is_same_package_member(Klass* class2, TRAPS) {
@@ -965,10 +969,6 @@
   void adjust_default_methods(InstanceKlass* holder, bool* trace_name_printed);
 #endif // INCLUDE_JVMTI
 
-  // Garbage collection
-  void oop_follow_contents(oop obj);
-  int  oop_adjust_pointers(oop obj);
-
   void clean_implementors_list(BoolObjectClosure* is_alive);
   void clean_method_data(BoolObjectClosure* is_alive);
   void clean_dependent_nmethods();
@@ -992,32 +992,108 @@
   static void notify_unload_class(InstanceKlass* ik);
   static void release_C_heap_structures(InstanceKlass* ik);
 
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
-
   // Naming
   const char* signature_name() const;
 
-  // Iterators
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
+
+  // Oop fields (and metadata) iterators
+  //  [nv = true]  Use non-virtual calls to do_oop_nv.
+  //  [nv = false] Use virtual calls to do_oop.
+  //
+  // The InstanceKlass iterators also visits the Object's klass.
+
+  // Forward iteration
+ public:
+  // Iterate over all oop fields in the oop maps.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps(oop obj, OopClosureType* closure);
+
+ protected:
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate(oop obj, OopClosureType* closure);
+
+ private:
+  // Iterate over all oop fields in the oop maps.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps_specialized(oop obj, OopClosureType* closure);
+
+  // Iterate over all oop fields in one oop map.
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_map(OopMapBlock* map, oop obj, OopClosureType* closure);
+
 
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_v_m(obj, blk, mr);
-  }
+  // Reverse iteration
+#if INCLUDE_ALL_GCS
+ public:
+  // Iterate over all oop fields in the oop maps.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps_reverse(oop obj, OopClosureType* closure);
+
+ protected:
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_reverse(oop obj, OopClosureType* closure);
+
+ private:
+  // Iterate over all oop fields in the oop maps.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps_specialized_reverse(oop obj, OopClosureType* closure);
+
+  // Iterate over all oop fields in one oop map.
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_map_reverse(OopMapBlock* map, oop obj, OopClosureType* closure);
+#endif
+
 
-#define InstanceKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)      \
-  int  oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);           \
-  int  oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* blk,        \
-                                      MemRegion mr);
+  // Bounded range iteration
+ public:
+  // Iterate over all oop fields in the oop maps.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+ protected:
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+ private:
+  // Iterate over all oop fields in the oop maps.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_maps_specialized_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Iterate over all oop fields in one oop map.
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_oop_map_bounded(OopMapBlock* map, oop obj, OopClosureType* closure, MemRegion mr);
+
+
+ public:
+
+#define InstanceKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)                   \
+  int  oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure);                    \
+  int  oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceKlass_OOP_OOP_ITERATE_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceKlass_OOP_OOP_ITERATE_DECL)
 
 #if INCLUDE_ALL_GCS
-#define InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix) \
-  int  oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* blk);
+#define InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)  \
+  int  oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/oops/instanceKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,215 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_OOPS_INSTANCEKLASS_INLINE_HPP
+#define SHARE_VM_OOPS_INSTANCEKLASS_INLINE_HPP
+
+#include "memory/iterator.hpp"
+#include "oops/instanceKlass.hpp"
+#include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
+#include "utilities/macros.hpp"
+
+// The iteration over the oops in objects is a hot path in the GC code.
+// By force inlining the following functions, we get similar GC performance
+// as the previous macro based implementation.
+#ifdef TARGET_COMPILER_visCPP
+#define INLINE __forceinline
+#else
+#define INLINE inline
+#endif
+
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_map(OopMapBlock* map, oop obj, OopClosureType* closure) {
+  T* p         = (T*)obj->obj_field_addr<T>(map->offset());
+  T* const end = p + map->count();
+
+  for (; p < end; ++p) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_map_reverse(OopMapBlock* map, oop obj, OopClosureType* closure) {
+  T* const start = (T*)obj->obj_field_addr<T>(map->offset());
+  T*       p     = start + map->count();
+
+  while (start < p) {
+    --p;
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+#endif
+
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_map_bounded(OopMapBlock* map, oop obj, OopClosureType* closure, MemRegion mr) {
+  T* p   = (T*)obj->obj_field_addr<T>(map->offset());
+  T* end = p + map->count();
+
+  T* const l   = (T*)mr.start();
+  T* const h   = (T*)mr.end();
+  assert(mask_bits((intptr_t)l, sizeof(T)-1) == 0 &&
+         mask_bits((intptr_t)h, sizeof(T)-1) == 0,
+         "bounded region must be properly aligned");
+
+  if (p < l) {
+    p = l;
+  }
+  if (end > h) {
+    end = h;
+  }
+
+  for (;p < end; ++p) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps_specialized(oop obj, OopClosureType* closure) {
+  OopMapBlock* map           = start_of_nonstatic_oop_maps();
+  OopMapBlock* const end_map = map + nonstatic_oop_map_count();
+
+  for (; map < end_map; ++map) {
+    oop_oop_iterate_oop_map<nv, T>(map, obj, closure);
+  }
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps_specialized_reverse(oop obj, OopClosureType* closure) {
+  OopMapBlock* const start_map = start_of_nonstatic_oop_maps();
+  OopMapBlock* map             = start_map + nonstatic_oop_map_count();
+
+  while (start_map < map) {
+    --map;
+    oop_oop_iterate_oop_map_reverse<nv, T>(map, obj, closure);
+  }
+}
+#endif
+
+template <bool nv, typename T, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps_specialized_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  OopMapBlock* map           = start_of_nonstatic_oop_maps();
+  OopMapBlock* const end_map = map + nonstatic_oop_map_count();
+
+  for (;map < end_map; ++map) {
+    oop_oop_iterate_oop_map_bounded<nv, T>(map, obj, closure, mr);
+  }
+}
+
+template <bool nv, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps(oop obj, OopClosureType* closure) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_oop_maps_specialized<nv, narrowOop>(obj, closure);
+  } else {
+    oop_oop_iterate_oop_maps_specialized<nv, oop>(obj, closure);
+  }
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps_reverse(oop obj, OopClosureType* closure) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_oop_maps_specialized_reverse<nv, narrowOop>(obj, closure);
+  } else {
+    oop_oop_iterate_oop_maps_specialized_reverse<nv, oop>(obj, closure);
+  }
+}
+#endif
+
+template <bool nv, class OopClosureType>
+INLINE void InstanceKlass::oop_oop_iterate_oop_maps_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_oop_maps_specialized_bounded<nv, narrowOop>(obj, closure, mr);
+  } else {
+    oop_oop_iterate_oop_maps_specialized_bounded<nv, oop>(obj, closure, mr);
+  }
+}
+
+template <bool nv, class OopClosureType>
+INLINE int InstanceKlass::oop_oop_iterate(oop obj, OopClosureType* closure) {
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    Devirtualizer<nv>::do_klass(closure, this);
+  }
+
+  oop_oop_iterate_oop_maps<nv>(obj, closure);
+
+  return size_helper();
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, class OopClosureType>
+INLINE int InstanceKlass::oop_oop_iterate_reverse(oop obj, OopClosureType* closure) {
+  assert(!Devirtualizer<nv>::do_metadata(closure),
+      "Code to handle metadata is not implemented");
+
+  oop_oop_iterate_oop_maps_reverse<nv>(obj, closure);
+
+  return size_helper();
+}
+#endif
+
+template <bool nv, class OopClosureType>
+INLINE int InstanceKlass::oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    if (mr.contains(obj)) {
+      Devirtualizer<nv>::do_klass(closure, this);
+    }
+  }
+
+  oop_oop_iterate_oop_maps_bounded<nv>(obj, closure, mr);
+
+  return size_helper();
+}
+
+#undef INLINE
+
+
+#define InstanceKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)              \
+int InstanceKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                    \
+}
+
+#if INCLUDE_ALL_GCS
+#define InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
+int InstanceKlass::oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate_reverse<nvs_to_bool(nv_suffix)>(obj, closure);                      \
+}
+#else
+#define InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+#define InstanceKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)                              \
+int InstanceKlass::oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_bounded<nvs_to_bool(nv_suffix)>(obj, closure, mr);                          \
+}
+
+#define ALL_INSTANCE_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  InstanceKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)   \
+  InstanceKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)   \
+  InstanceKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+
+#endif // SHARE_VM_OOPS_INSTANCEKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/oops/instanceMirrorKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceMirrorKlass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,9 +25,7 @@
 #include "precompiled.hpp"
 #include "classfile/javaClasses.hpp"
 #include "classfile/systemDictionary.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
-#include "memory/genOopClosures.inline.hpp"
 #include "memory/iterator.inline.hpp"
 #include "memory/oopFactory.hpp"
 #include "memory/specialized_oop_closures.hpp"
@@ -38,313 +36,9 @@
 #include "oops/symbol.hpp"
 #include "runtime/handles.inline.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/concurrentMarkSweep/cmsOopClosures.inline.hpp"
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
-#include "gc_implementation/g1/g1OopClosures.inline.hpp"
-#include "gc_implementation/g1/g1RemSet.inline.hpp"
-#include "gc_implementation/g1/heapRegionManager.inline.hpp"
-#include "gc_implementation/parNew/parOopClosures.inline.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.inline.hpp"
-#endif // INCLUDE_ALL_GCS
 
 int InstanceMirrorKlass::_offset_of_static_fields = 0;
 
-#ifdef ASSERT
-template <class T> void assert_is_in(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in(o), "should be in heap");
-  }
-}
-template <class T> void assert_is_in_closed_subset(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in_closed_subset(o), "should be in closed");
-  }
-}
-template <class T> void assert_is_in_reserved(T *p) {
-  T heap_oop = oopDesc::load_heap_oop(p);
-  if (!oopDesc::is_null(heap_oop)) {
-    oop o = oopDesc::decode_heap_oop_not_null(heap_oop);
-    assert(Universe::heap()->is_in_reserved(o), "should be in reserved");
-  }
-}
-template <class T> void assert_nothing(T *p) {}
-
-#else
-template <class T> void assert_is_in(T *p) {}
-template <class T> void assert_is_in_closed_subset(T *p) {}
-template <class T> void assert_is_in_reserved(T *p) {}
-template <class T> void assert_nothing(T *p) {}
-#endif // ASSERT
-
-#define InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE( \
-  T, start_p, count, do_oop,                         \
-  assert_fn)                                         \
-{                                                    \
-  T* p         = (T*)(start_p);                      \
-  T* const end = p + (count);                        \
-  while (p < end) {                                  \
-    (assert_fn)(p);                                  \
-    do_oop;                                          \
-    ++p;                                             \
-  }                                                  \
-}
-
-#define InstanceMirrorKlass_SPECIALIZED_BOUNDED_OOP_ITERATE( \
-  T, start_p, count, low, high,                              \
-  do_oop, assert_fn)                                         \
-{                                                            \
-  T* const l = (T*)(low);                                    \
-  T* const h = (T*)(high);                                   \
-  assert(mask_bits((intptr_t)l, sizeof(T)-1) == 0 &&         \
-         mask_bits((intptr_t)h, sizeof(T)-1) == 0,           \
-         "bounded region must be properly aligned");         \
-  T* p       = (T*)(start_p);                                \
-  T* end     = p + (count);                                  \
-  if (p < l) p = l;                                          \
-  if (end > h) end = h;                                      \
-  while (p < end) {                                          \
-    (assert_fn)(p);                                          \
-    do_oop;                                                  \
-    ++p;                                                     \
-  }                                                          \
-}
-
-
-#define InstanceMirrorKlass_OOP_ITERATE(start_p, count,    \
-                                  do_oop, assert_fn)       \
-{                                                          \
-  if (UseCompressedOops) {                                 \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE(narrowOop, \
-      start_p, count,                                      \
-      do_oop, assert_fn)                                   \
-  } else {                                                 \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE(oop,       \
-      start_p, count,                                      \
-      do_oop, assert_fn)                                   \
-  }                                                        \
-}
-
-// The following macros call specialized macros, passing either oop or
-// narrowOop as the specialization type.  These test the UseCompressedOops
-// flag.
-#define InstanceMirrorKlass_BOUNDED_OOP_ITERATE(start_p, count, low, high, \
-                                          do_oop, assert_fn)               \
-{                                                                          \
-  if (UseCompressedOops) {                                                 \
-    InstanceMirrorKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(narrowOop,         \
-      start_p, count,                                                      \
-      low, high,                                                           \
-      do_oop, assert_fn)                                                   \
-  } else {                                                                 \
-    InstanceMirrorKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(oop,               \
-      start_p, count,                                                      \
-      low, high,                                                           \
-      do_oop, assert_fn)                                                   \
-  }                                                                        \
-}
-
-
-void InstanceMirrorKlass::oop_follow_contents(oop obj) {
-  InstanceKlass::oop_follow_contents(obj);
-
-  // Follow the klass field in the mirror.
-  Klass* klass = java_lang_Class::as_Klass(obj);
-  if (klass != NULL) {
-    // An anonymous class doesn't have its own class loader, so the call
-    // to follow_klass will mark and push its java mirror instead of the
-    // class loader. When handling the java mirror for an anonymous class
-    // we need to make sure its class loader data is claimed, this is done
-    // by calling follow_class_loader explicitly. For non-anonymous classes
-    // the call to follow_class_loader is made when the class loader itself
-    // is handled.
-    if (klass->oop_is_instance() && InstanceKlass::cast(klass)->is_anonymous()) {
-      MarkSweep::follow_class_loader(klass->class_loader_data());
-    } else {
-      MarkSweep::follow_klass(klass);
-    }
-  } else {
-    // If klass is NULL then this a mirror for a primitive type.
-    // We don't have to follow them, since they are handled as strong
-    // roots in Universe::oops_do.
-    assert(java_lang_Class::is_primitive(obj), "Sanity check");
-  }
-
-  InstanceMirrorKlass_OOP_ITERATE(                                                    \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),        \
-    MarkSweep::mark_and_push(p),                                                      \
-    assert_is_in_closed_subset)
-}
-
-#if INCLUDE_ALL_GCS
-void InstanceMirrorKlass::oop_follow_contents(ParCompactionManager* cm,
-                                              oop obj) {
-  InstanceKlass::oop_follow_contents(cm, obj);
-
-  // Follow the klass field in the mirror.
-  Klass* klass = java_lang_Class::as_Klass(obj);
-  if (klass != NULL) {
-    // An anonymous class doesn't have its own class loader, so the call
-    // to follow_klass will mark and push its java mirror instead of the
-    // class loader. When handling the java mirror for an anonymous class
-    // we need to make sure its class loader data is claimed, this is done
-    // by calling follow_class_loader explicitly. For non-anonymous classes
-    // the call to follow_class_loader is made when the class loader itself
-    // is handled.
-    if (klass->oop_is_instance() && InstanceKlass::cast(klass)->is_anonymous()) {
-      PSParallelCompact::follow_class_loader(cm, klass->class_loader_data());
-    } else {
-      PSParallelCompact::follow_klass(cm, klass);
-    }
-  } else {
-    // If klass is NULL then this a mirror for a primitive type.
-    // We don't have to follow them, since they are handled as strong
-    // roots in Universe::oops_do.
-    assert(java_lang_Class::is_primitive(obj), "Sanity check");
-  }
-
-  InstanceMirrorKlass_OOP_ITERATE(                                                    \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),        \
-    PSParallelCompact::mark_and_push(cm, p),                                          \
-    assert_is_in)
-}
-#endif // INCLUDE_ALL_GCS
-
-int InstanceMirrorKlass::oop_adjust_pointers(oop obj) {
-  int size = oop_size(obj);
-  InstanceKlass::oop_adjust_pointers(obj);
-
-  InstanceMirrorKlass_OOP_ITERATE(                                                    \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),        \
-    MarkSweep::adjust_pointer(p),                                                     \
-    assert_nothing)
-  return size;
-}
-
-#define InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE_DEFN(T, nv_suffix)                \
-  InstanceMirrorKlass_OOP_ITERATE(                                                    \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),        \
-      (closure)->do_oop##nv_suffix(p),                                                \
-    assert_is_in_closed_subset)                                                       \
-  return oop_size(obj);                                                               \
-
-#define InstanceMirrorKlass_BOUNDED_SPECIALIZED_OOP_ITERATE(T, nv_suffix, mr)         \
-  InstanceMirrorKlass_BOUNDED_OOP_ITERATE(                                            \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),        \
-    mr.start(), mr.end(),                                                             \
-      (closure)->do_oop##nv_suffix(p),                                                \
-    assert_is_in_closed_subset)                                                       \
-  return oop_size(obj);                                                               \
-
-
-// Macro to define InstanceMirrorKlass::oop_oop_iterate for virtual/nonvirtual for
-// all closures.  Macros calling macros above for each oop size.
-
-#define InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)           \
-                                                                                      \
-int InstanceMirrorKlass::                                                             \
-oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {                        \
-  /* Get size before changing pointers */                                             \
-  InstanceKlass::oop_oop_iterate##nv_suffix(obj, closure);                            \
-                                                                                      \
-  if_do_metadata_checked(closure, nv_suffix) {                                        \
-    Klass* klass = java_lang_Class::as_Klass(obj);                                    \
-    /* We'll get NULL for primitive mirrors. */                                       \
-    if (klass != NULL) {                                                              \
-      closure->do_klass##nv_suffix(klass);                                            \
-    }                                                                                 \
-  }                                                                                   \
-                                                                                      \
-  if (UseCompressedOops) {                                                            \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE_DEFN(narrowOop, nv_suffix);           \
-  } else {                                                                            \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE_DEFN(oop, nv_suffix);                 \
-  }                                                                                   \
-}
-
-#if INCLUDE_ALL_GCS
-#define InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix) \
-                                                                                      \
-int InstanceMirrorKlass::                                                             \
-oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {              \
-  /* Get size before changing pointers */                                             \
-  InstanceKlass::oop_oop_iterate_backwards##nv_suffix(obj, closure);                  \
-                                                                                      \
-  if (UseCompressedOops) {                                                            \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE_DEFN(narrowOop, nv_suffix);           \
-  } else {                                                                            \
-    InstanceMirrorKlass_SPECIALIZED_OOP_ITERATE_DEFN(oop, nv_suffix);                 \
-  }                                                                                   \
-}
-#endif // INCLUDE_ALL_GCS
-
-
-#define InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)         \
-                                                                                      \
-int InstanceMirrorKlass::                                                             \
-oop_oop_iterate##nv_suffix##_m(oop obj,                                               \
-                               OopClosureType* closure,                               \
-                               MemRegion mr) {                                        \
-  InstanceKlass::oop_oop_iterate##nv_suffix##_m(obj, closure, mr);                    \
-                                                                                      \
-  if_do_metadata_checked(closure, nv_suffix) {                                        \
-    if (mr.contains(obj)) {                                                           \
-      Klass* klass = java_lang_Class::as_Klass(obj);                                  \
-      /* We'll get NULL for primitive mirrors. */                                     \
-      if (klass != NULL) {                                                            \
-        closure->do_klass##nv_suffix(klass);                                          \
-      }                                                                               \
-    }                                                                                 \
-  }                                                                                   \
-                                                                                      \
-  if (UseCompressedOops) {                                                            \
-    InstanceMirrorKlass_BOUNDED_SPECIALIZED_OOP_ITERATE(narrowOop, nv_suffix, mr);    \
-  } else {                                                                            \
-    InstanceMirrorKlass_BOUNDED_SPECIALIZED_OOP_ITERATE(oop, nv_suffix, mr);          \
-  }                                                                                   \
-}
-
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN)
-#if INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-#endif // INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN_m)
-
-#if INCLUDE_ALL_GCS
-void InstanceMirrorKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  // Note that we don't have to follow the mirror -> klass pointer, since all
-  // klasses that are dirty will be scavenged when we iterate over the
-  // ClassLoaderData objects.
-
-  InstanceKlass::oop_push_contents(pm, obj);
-  InstanceMirrorKlass_OOP_ITERATE(                                            \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),\
-    if (PSScavenge::should_scavenge(p)) {                                     \
-      pm->claim_or_forward_depth(p);                                          \
-    },                                                                        \
-    assert_nothing )
-}
-
-int InstanceMirrorKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  int size = oop_size(obj);
-  InstanceKlass::oop_update_pointers(cm, obj);
-
-  InstanceMirrorKlass_OOP_ITERATE(                                            \
-    start_of_static_fields(obj), java_lang_Class::static_oop_field_count(obj),\
-    PSParallelCompact::adjust_pointer(p),                                     \
-    assert_nothing)
-  return size;
-}
-#endif // INCLUDE_ALL_GCS
-
 int InstanceMirrorKlass::instance_size(KlassHandle k) {
   if (k() != NULL && k->oop_is_instance()) {
     return align_object_size(size_helper() + InstanceKlass::cast(k())->static_field_size());
--- a/hotspot/src/share/vm/oops/instanceMirrorKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceMirrorKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -88,19 +88,66 @@
   // allocation
   instanceOop allocate_instance(KlassHandle k, TRAPS);
 
-  // Garbage collection
-  int  oop_adjust_pointers(oop obj);
-  void oop_follow_contents(oop obj);
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
 
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
+  // Oop fields (and metadata) iterators
+  //  [nv = true]  Use non-virtual calls to do_oop_nv.
+  //  [nv = false] Use virtual calls to do_oop.
+  //
+  // The InstanceMirrorKlass iterators also visit the hidden Klass pointer.
+
+ public:
+  // Iterate over the static fields.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_statics(oop obj, OopClosureType* closure);
+
+ private:
+  // Iterate over the static fields.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_statics_specialized(oop obj, OopClosureType* closure);
 
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_v_m(obj, blk, mr);
-  }
+  // Forward iteration
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate(oop obj, OopClosureType* closure);
+
+
+  // Reverse iteration
+#if INCLUDE_ALL_GCS
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_reverse(oop obj, OopClosureType* closure);
+#endif
+
+
+  // Bounded range iteration
+  // Iterate over the oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Iterate over the static fields.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_statics_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Iterate over the static fields.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_statics_specialized_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+
+ public:
 
 #define InstanceMirrorKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)           \
   int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);                       \
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/oops/instanceMirrorKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,164 @@
+/* Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_OOPS_INSTANCEMIRRORKLASS_INLINE_HPP
+#define SHARE_VM_OOPS_INSTANCEMIRRORKLASS_INLINE_HPP
+
+#include "classfile/javaClasses.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/instanceMirrorKlass.hpp"
+#include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
+#include "utilities/macros.hpp"
+
+template <bool nv, typename T, class OopClosureType>
+void InstanceMirrorKlass::oop_oop_iterate_statics_specialized(oop obj, OopClosureType* closure) {
+  T* p         = (T*)start_of_static_fields(obj);
+  T* const end = p + java_lang_Class::static_oop_field_count(obj);
+
+  for (; p < end; ++p) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+template <bool nv, class OopClosureType>
+void InstanceMirrorKlass::oop_oop_iterate_statics(oop obj, OopClosureType* closure) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_statics_specialized<nv, narrowOop>(obj, closure);
+  } else {
+    oop_oop_iterate_statics_specialized<nv, oop>(obj, closure);
+  }
+}
+
+template <bool nv, class OopClosureType>
+int InstanceMirrorKlass::oop_oop_iterate(oop obj, OopClosureType* closure) {
+  InstanceKlass::oop_oop_iterate<nv>(obj, closure);
+
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    Klass* klass = java_lang_Class::as_Klass(obj);
+    // We'll get NULL for primitive mirrors.
+    if (klass != NULL) {
+      Devirtualizer<nv>::do_klass(closure, klass);
+    }
+  }
+
+  oop_oop_iterate_statics<nv>(obj, closure);
+
+  return oop_size(obj);
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, class OopClosureType>
+int InstanceMirrorKlass::oop_oop_iterate_reverse(oop obj, OopClosureType* closure) {
+  InstanceKlass::oop_oop_iterate_reverse<nv>(obj, closure);
+
+  InstanceMirrorKlass::oop_oop_iterate_statics<nv>(obj, closure);
+
+  return oop_size(obj);
+}
+#endif
+
+template <bool nv, typename T, class OopClosureType>
+void InstanceMirrorKlass::oop_oop_iterate_statics_specialized_bounded(oop obj,
+                                                                     OopClosureType* closure,
+                                                                     MemRegion mr) {
+  T* p   = (T*)start_of_static_fields(obj);
+  T* end = p + java_lang_Class::static_oop_field_count(obj);
+
+  T* const l   = (T*)mr.start();
+  T* const h   = (T*)mr.end();
+  assert(mask_bits((intptr_t)l, sizeof(T)-1) == 0 &&
+         mask_bits((intptr_t)h, sizeof(T)-1) == 0,
+         "bounded region must be properly aligned");
+
+  if (p < l) {
+    p = l;
+  }
+  if (end > h) {
+    end = h;
+  }
+
+  for (;p < end; ++p) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+template <bool nv, class OopClosureType>
+void InstanceMirrorKlass::oop_oop_iterate_statics_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_statics_specialized_bounded<nv, narrowOop>(obj, closure, mr);
+  } else {
+    oop_oop_iterate_statics_specialized_bounded<nv, oop>(obj, closure, mr);
+  }
+}
+
+template <bool nv, class OopClosureType>
+int InstanceMirrorKlass::oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  InstanceKlass::oop_oop_iterate_bounded<nv>(obj, closure, mr);
+
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    if (mr.contains(obj)) {
+      Klass* klass = java_lang_Class::as_Klass(obj);
+      // We'll get NULL for primitive mirrors.
+      if (klass != NULL) {
+        Devirtualizer<nv>::do_klass(closure, klass);
+      }
+    }
+  }
+
+  oop_oop_iterate_statics_bounded<nv>(obj, closure, mr);
+
+  return oop_size(obj);
+}
+
+
+#define InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)              \
+                                                                                         \
+int InstanceMirrorKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                          \
+}
+
+#if INCLUDE_ALL_GCS
+#define InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
+                                                                                                   \
+int InstanceMirrorKlass::oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate_reverse<nvs_to_bool(nv_suffix)>(obj, closure);                            \
+}
+#else
+#define InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+
+#define InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)                              \
+                                                                                                           \
+int InstanceMirrorKlass::oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_bounded<nvs_to_bool(nv_suffix)>(obj, closure, mr);                                \
+}
+
+#define ALL_INSTANCE_MIRROR_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)    \
+  InstanceMirrorKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)    \
+  InstanceMirrorKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+
+#endif // SHARE_VM_OOPS_INSTANCEMIRRORKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/oops/instanceRefKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceRefKlass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,421 +25,16 @@
 #include "precompiled.hpp"
 #include "classfile/javaClasses.hpp"
 #include "classfile/systemDictionary.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
-#include "gc_interface/collectedHeap.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
 #include "memory/genCollectedHeap.hpp"
-#include "memory/genOopClosures.inline.hpp"
 #include "memory/specialized_oop_closures.hpp"
-#include "oops/instanceRefKlass.hpp"
+#include "oops/instanceRefKlass.inline.hpp"
 #include "oops/oop.inline.hpp"
 #include "utilities/preserveException.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
-#include "gc_implementation/g1/g1OopClosures.inline.hpp"
-#include "gc_implementation/g1/g1RemSet.inline.hpp"
-#include "gc_implementation/g1/heapRegionManager.inline.hpp"
-#include "gc_implementation/parNew/parOopClosures.inline.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.inline.hpp"
-#endif // INCLUDE_ALL_GCS
 
 PRAGMA_FORMAT_MUTE_WARNINGS_FOR_GCC
 
-template <class T>
-void specialized_oop_follow_contents(InstanceRefKlass* ref, oop obj) {
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
-  T heap_oop = oopDesc::load_heap_oop(referent_addr);
-  debug_only(
-    if(TraceReferenceGC && PrintGCDetails) {
-      gclog_or_tty->print_cr("InstanceRefKlass::oop_follow_contents " INTPTR_FORMAT, (void *)obj);
-    }
-  )
-  if (!oopDesc::is_null(heap_oop)) {
-    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);
-    if (!referent->is_gc_marked() &&
-        MarkSweep::ref_processor()->discover_reference(obj, ref->reference_type())) {
-      // reference was discovered, referent will be traversed later
-      ref->InstanceKlass::oop_follow_contents(obj);
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("       Non NULL enqueued " INTPTR_FORMAT, (void *)obj);
-        }
-      )
-      return;
-    } else {
-      // treat referent as normal oop
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("       Non NULL normal " INTPTR_FORMAT, (void *)obj);
-        }
-      )
-      MarkSweep::mark_and_push(referent_addr);
-    }
-  }
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
-  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
-    // Treat discovered as normal oop, if ref is not "active",
-    // i.e. if next is non-NULL.
-    T  next_oop = oopDesc::load_heap_oop(next_addr);
-    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
-      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("   Process discovered as normal "
-                                 INTPTR_FORMAT, discovered_addr);
-        }
-      )
-      MarkSweep::mark_and_push(discovered_addr);
-    }
-  } else {
-#ifdef ASSERT
-    // In the case of older JDKs which do not use the discovered
-    // field for the pending list, an inactive ref (next != NULL)
-    // must always have a NULL discovered field.
-    oop next = oopDesc::load_decode_heap_oop(next_addr);
-    oop discovered = java_lang_ref_Reference::discovered(obj);
-    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
-           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
-                   (oopDesc*)obj));
-#endif
-  }
-  // treat next as normal oop.  next is a link in the reference queue.
-  debug_only(
-    if(TraceReferenceGC && PrintGCDetails) {
-      gclog_or_tty->print_cr("   Process next as normal " INTPTR_FORMAT, next_addr);
-    }
-  )
-  MarkSweep::mark_and_push(next_addr);
-  ref->InstanceKlass::oop_follow_contents(obj);
-}
-
-void InstanceRefKlass::oop_follow_contents(oop obj) {
-  if (UseCompressedOops) {
-    specialized_oop_follow_contents<narrowOop>(this, obj);
-  } else {
-    specialized_oop_follow_contents<oop>(this, obj);
-  }
-}
-
-#if INCLUDE_ALL_GCS
-template <class T>
-void specialized_oop_follow_contents(InstanceRefKlass* ref,
-                                     ParCompactionManager* cm,
-                                     oop obj) {
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
-  T heap_oop = oopDesc::load_heap_oop(referent_addr);
-  debug_only(
-    if(TraceReferenceGC && PrintGCDetails) {
-      gclog_or_tty->print_cr("InstanceRefKlass::oop_follow_contents " INTPTR_FORMAT, (void *)obj);
-    }
-  )
-  if (!oopDesc::is_null(heap_oop)) {
-    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);
-    if (PSParallelCompact::mark_bitmap()->is_unmarked(referent) &&
-        PSParallelCompact::ref_processor()->
-          discover_reference(obj, ref->reference_type())) {
-      // reference already enqueued, referent will be traversed later
-      ref->InstanceKlass::oop_follow_contents(cm, obj);
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("       Non NULL enqueued " INTPTR_FORMAT, (void *)obj);
-        }
-      )
-      return;
-    } else {
-      // treat referent as normal oop
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("       Non NULL normal " INTPTR_FORMAT, (void *)obj);
-        }
-      )
-      PSParallelCompact::mark_and_push(cm, referent_addr);
-    }
-  }
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
-  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
-    // Treat discovered as normal oop, if ref is not "active",
-    // i.e. if next is non-NULL.
-    T  next_oop = oopDesc::load_heap_oop(next_addr);
-    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
-      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("   Process discovered as normal "
-                                 INTPTR_FORMAT, discovered_addr);
-        }
-      )
-      PSParallelCompact::mark_and_push(cm, discovered_addr);
-    }
-  } else {
-#ifdef ASSERT
-    // In the case of older JDKs which do not use the discovered
-    // field for the pending list, an inactive ref (next != NULL)
-    // must always have a NULL discovered field.
-    T next = oopDesc::load_heap_oop(next_addr);
-    oop discovered = java_lang_ref_Reference::discovered(obj);
-    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
-           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
-                   (oopDesc*)obj));
-#endif
-  }
-  PSParallelCompact::mark_and_push(cm, next_addr);
-  ref->InstanceKlass::oop_follow_contents(cm, obj);
-}
-
-void InstanceRefKlass::oop_follow_contents(ParCompactionManager* cm,
-                                           oop obj) {
-  if (UseCompressedOops) {
-    specialized_oop_follow_contents<narrowOop>(this, cm, obj);
-  } else {
-    specialized_oop_follow_contents<oop>(this, cm, obj);
-  }
-}
-#endif // INCLUDE_ALL_GCS
-
-#ifdef ASSERT
-template <class T> void trace_reference_gc(const char *s, oop obj,
-                                           T* referent_addr,
-                                           T* next_addr,
-                                           T* discovered_addr) {
-  if(TraceReferenceGC && PrintGCDetails) {
-    gclog_or_tty->print_cr("%s obj " INTPTR_FORMAT, s, (address)obj);
-    gclog_or_tty->print_cr("     referent_addr/* " INTPTR_FORMAT " / "
-         INTPTR_FORMAT, referent_addr,
-         referent_addr ?
-           (address)oopDesc::load_decode_heap_oop(referent_addr) : NULL);
-    gclog_or_tty->print_cr("     next_addr/* " INTPTR_FORMAT " / "
-         INTPTR_FORMAT, next_addr,
-         next_addr ? (address)oopDesc::load_decode_heap_oop(next_addr) : NULL);
-    gclog_or_tty->print_cr("     discovered_addr/* " INTPTR_FORMAT " / "
-         INTPTR_FORMAT, discovered_addr,
-         discovered_addr ?
-           (address)oopDesc::load_decode_heap_oop(discovered_addr) : NULL);
-  }
-}
-#endif
-
-template <class T> void specialized_oop_adjust_pointers(InstanceRefKlass *ref, oop obj) {
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
-  MarkSweep::adjust_pointer(referent_addr);
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
-  MarkSweep::adjust_pointer(next_addr);
-  T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
-  MarkSweep::adjust_pointer(discovered_addr);
-  debug_only(trace_reference_gc("InstanceRefKlass::oop_adjust_pointers", obj,
-                                referent_addr, next_addr, discovered_addr);)
-}
-
-int InstanceRefKlass::oop_adjust_pointers(oop obj) {
-  int size = size_helper();
-  InstanceKlass::oop_adjust_pointers(obj);
-
-  if (UseCompressedOops) {
-    specialized_oop_adjust_pointers<narrowOop>(this, obj);
-  } else {
-    specialized_oop_adjust_pointers<oop>(this, obj);
-  }
-  return size;
-}
-
-#define InstanceRefKlass_SPECIALIZED_OOP_ITERATE(T, nv_suffix, contains)        \
-  T* disc_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);             \
-  if (closure->apply_to_weak_ref_discovered_field()) {                          \
-    closure->do_oop##nv_suffix(disc_addr);                                      \
-  }                                                                             \
-                                                                                \
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);           \
-  T heap_oop = oopDesc::load_heap_oop(referent_addr);                           \
-  ReferenceProcessor* rp = closure->_ref_processor;                             \
-  if (!oopDesc::is_null(heap_oop)) {                                            \
-    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);                 \
-    if (!referent->is_gc_marked() && (rp != NULL) &&                            \
-        rp->discover_reference(obj, reference_type())) {                        \
-      return size;                                                              \
-    } else if (contains(referent_addr)) {                                       \
-      /* treat referent as normal oop */                                        \
-      closure->do_oop##nv_suffix(referent_addr);                                \
-    }                                                                           \
-  }                                                                             \
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);                   \
-  if (ReferenceProcessor::pending_list_uses_discovered_field()) {               \
-    T next_oop  = oopDesc::load_heap_oop(next_addr);                            \
-    /* Treat discovered as normal oop, if ref is not "active" (next non-NULL) */\
-    if (!oopDesc::is_null(next_oop) && contains(disc_addr)) {                   \
-        /* i.e. ref is not "active" */                                          \
-      debug_only(                                                               \
-        if(TraceReferenceGC && PrintGCDetails) {                                \
-          gclog_or_tty->print_cr("   Process discovered as normal "             \
-                                 INTPTR_FORMAT, disc_addr);                     \
-        }                                                                       \
-      )                                                                         \
-      closure->do_oop##nv_suffix(disc_addr);                                    \
-    }                                                                           \
-  } else {                                                                      \
-    /* In the case of older JDKs which do not use the discovered field for  */  \
-    /* the pending list, an inactive ref (next != NULL) must always have a  */  \
-    /* NULL discovered field. */                                                \
-    debug_only(                                                                 \
-      T next_oop = oopDesc::load_heap_oop(next_addr);                           \
-      T disc_oop = oopDesc::load_heap_oop(disc_addr);                           \
-      assert(oopDesc::is_null(next_oop) || oopDesc::is_null(disc_oop),          \
-           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL" \
-                   "discovered field", (oopDesc*)obj));                                   \
-    )                                                                           \
-  }                                                                             \
-  /* treat next as normal oop */                                                \
-  if (contains(next_addr)) {                                                    \
-    closure->do_oop##nv_suffix(next_addr);                                      \
-  }                                                                             \
-  return size;                                                                  \
-
-
-template <class T> bool contains(T *t) { return true; }
-
-// Macro to define InstanceRefKlass::oop_oop_iterate for virtual/nonvirtual for
-// all closures.  Macros calling macros above for each oop size.
-
-#define InstanceRefKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)        \
-                                                                                \
-int InstanceRefKlass::                                                          \
-oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {                  \
-  /* Get size before changing pointers */                                       \
-  int size = InstanceKlass::oop_oop_iterate##nv_suffix(obj, closure);           \
-                                                                                \
-  if (UseCompressedOops) {                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(narrowOop, nv_suffix, contains);   \
-  } else {                                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(oop, nv_suffix, contains);         \
-  }                                                                             \
-}
-
-#if INCLUDE_ALL_GCS
-#define InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix) \
-                                                                                \
-int InstanceRefKlass::                                                          \
-oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {        \
-  /* Get size before changing pointers */                                       \
-  int size = InstanceKlass::oop_oop_iterate_backwards##nv_suffix(obj, closure); \
-                                                                                \
-  if (UseCompressedOops) {                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(narrowOop, nv_suffix, contains);   \
-  } else {                                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(oop, nv_suffix, contains);         \
-  }                                                                             \
-}
-#endif // INCLUDE_ALL_GCS
-
-
-#define InstanceRefKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)      \
-                                                                                \
-int InstanceRefKlass::                                                          \
-oop_oop_iterate##nv_suffix##_m(oop obj,                                         \
-                               OopClosureType* closure,                         \
-                               MemRegion mr) {                                  \
-  int size = InstanceKlass::oop_oop_iterate##nv_suffix##_m(obj, closure, mr);   \
-  if (UseCompressedOops) {                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(narrowOop, nv_suffix, mr.contains); \
-  } else {                                                                      \
-    InstanceRefKlass_SPECIALIZED_OOP_ITERATE(oop, nv_suffix, mr.contains);      \
-  }                                                                             \
-}
-
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceRefKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceRefKlass_OOP_OOP_ITERATE_DEFN)
-#if INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN)
-#endif // INCLUDE_ALL_GCS
-ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceRefKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceRefKlass_OOP_OOP_ITERATE_DEFN_m)
-
-#if INCLUDE_ALL_GCS
-template <class T>
-void specialized_oop_push_contents(InstanceRefKlass *ref,
-                                   PSPromotionManager* pm, oop obj) {
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
-  if (PSScavenge::should_scavenge(referent_addr)) {
-    ReferenceProcessor* rp = PSScavenge::reference_processor();
-    if (rp->discover_reference(obj, ref->reference_type())) {
-      // reference already enqueued, referent and next will be traversed later
-      ref->InstanceKlass::oop_push_contents(pm, obj);
-      return;
-    } else {
-      // treat referent as normal oop
-      pm->claim_or_forward_depth(referent_addr);
-    }
-  }
-  // Treat discovered as normal oop, if ref is not "active",
-  // i.e. if next is non-NULL.
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
-  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
-    T  next_oop = oopDesc::load_heap_oop(next_addr);
-    if (!oopDesc::is_null(next_oop)) { // i.e. ref is not "active"
-      T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
-      debug_only(
-        if(TraceReferenceGC && PrintGCDetails) {
-          gclog_or_tty->print_cr("   Process discovered as normal "
-                                 INTPTR_FORMAT, discovered_addr);
-        }
-      )
-      if (PSScavenge::should_scavenge(discovered_addr)) {
-        pm->claim_or_forward_depth(discovered_addr);
-      }
-    }
-  } else {
-#ifdef ASSERT
-    // In the case of older JDKs which do not use the discovered
-    // field for the pending list, an inactive ref (next != NULL)
-    // must always have a NULL discovered field.
-    oop next = oopDesc::load_decode_heap_oop(next_addr);
-    oop discovered = java_lang_ref_Reference::discovered(obj);
-    assert(oopDesc::is_null(next) || oopDesc::is_null(discovered),
-           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL discovered field",
-                   (oopDesc*)obj));
-#endif
-  }
-
-  // Treat next as normal oop;  next is a link in the reference queue.
-  if (PSScavenge::should_scavenge(next_addr)) {
-    pm->claim_or_forward_depth(next_addr);
-  }
-  ref->InstanceKlass::oop_push_contents(pm, obj);
-}
-
-void InstanceRefKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  if (UseCompressedOops) {
-    specialized_oop_push_contents<narrowOop>(this, pm, obj);
-  } else {
-    specialized_oop_push_contents<oop>(this, pm, obj);
-  }
-}
-
-template <class T>
-void specialized_oop_update_pointers(InstanceRefKlass *ref,
-                                    ParCompactionManager* cm, oop obj) {
-  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
-  PSParallelCompact::adjust_pointer(referent_addr);
-  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
-  PSParallelCompact::adjust_pointer(next_addr);
-  T* discovered_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
-  PSParallelCompact::adjust_pointer(discovered_addr);
-  debug_only(trace_reference_gc("InstanceRefKlass::oop_update_ptrs", obj,
-                                referent_addr, next_addr, discovered_addr);)
-}
-
-int InstanceRefKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  InstanceKlass::oop_update_pointers(cm, obj);
-  if (UseCompressedOops) {
-    specialized_oop_update_pointers<narrowOop>(this, cm, obj);
-  } else {
-    specialized_oop_update_pointers<oop>(this, cm, obj);
-  }
-  return size_helper();
-}
-#endif // INCLUDE_ALL_GCS
-
 void InstanceRefKlass::update_nonstatic_oop_maps(Klass* k) {
   // Clear the nonstatic oop-map entries corresponding to referent
   // and nextPending field.  They are treated specially by the
@@ -483,12 +78,6 @@
   InstanceKlass::oop_verify_on(obj, st);
   // Verify referent field
   oop referent = java_lang_ref_Reference::referent(obj);
-
-  // We should make this general to all heaps
-  GenCollectedHeap* gch = NULL;
-  if (Universe::heap()->kind() == CollectedHeap::GenCollectedHeap)
-    gch = GenCollectedHeap::heap();
-
   if (referent != NULL) {
     guarantee(referent->is_oop(), "referent field heap failed");
   }
--- a/hotspot/src/share/vm/oops/instanceRefKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/instanceRefKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -64,30 +64,71 @@
     return (InstanceRefKlass*) k;
   }
 
-  // Garbage collection
-  int  oop_adjust_pointers(oop obj);
-  void oop_follow_contents(oop obj);
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
 
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
+  // Oop fields (and metadata) iterators
+  //  [nv = true]  Use non-virtual calls to do_oop_nv.
+  //  [nv = false] Use virtual calls to do_oop.
+  //
+  // The InstanceRefKlass iterators also support reference processing.
+
+
+  // Forward iteration
+private:
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate(oop obj, OopClosureType* closure);
 
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_v_m(obj, blk, mr);
-  }
+  // Reverse iteration
+#if INCLUDE_ALL_GCS
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_reverse(oop obj, OopClosureType* closure);
+#endif // INCLUDE_ALL_GCS
+
+  // Bounded range iteration
+  // Iterate over all oop fields and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Reference processing part of the iterators.
 
-#define InstanceRefKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)                \
-  int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);                         \
-  int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* blk, MemRegion mr);
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType, class Contains>
+  inline void oop_oop_iterate_ref_processing_specialized(oop obj, OopClosureType* closure, Contains& contains);
+
+  // Only perform reference processing if the referent object is within mr.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_ref_processing_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Reference processing
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_ref_processing(oop obj, OopClosureType* closure);
+
+
+ public:
+
+#define InstanceRefKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)               \
+  int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure);                    \
+  int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceRefKlass_OOP_OOP_ITERATE_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceRefKlass_OOP_OOP_ITERATE_DECL)
 
 #if INCLUDE_ALL_GCS
-#define InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)      \
-  int oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* blk);
+#define InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)     \
+  int oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure);
 
   ALL_OOP_OOP_ITERATE_CLOSURES_1(InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/oops/instanceRefKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,187 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_OOPS_INSTANCEREFKLASS_INLINE_HPP
+#define SHARE_VM_OOPS_INSTANCEREFKLASS_INLINE_HPP
+
+#include "classfile/javaClasses.hpp"
+#include "memory/referenceProcessor.hpp"
+#include "oops/instanceRefKlass.hpp"
+#include "oops/instanceKlass.inline.hpp"
+#include "oops/oop.inline.hpp"
+#include "utilities/debug.hpp"
+#include "utilities/globalDefinitions.hpp"
+#include "utilities/macros.hpp"
+
+template <bool nv, typename T, class OopClosureType, class Contains>
+void InstanceRefKlass::oop_oop_iterate_ref_processing_specialized(oop obj, OopClosureType* closure, Contains& contains) {
+  T* disc_addr = (T*)java_lang_ref_Reference::discovered_addr(obj);
+  if (closure->apply_to_weak_ref_discovered_field()) {
+    Devirtualizer<nv>::do_oop(closure, disc_addr);
+  }
+
+  T* referent_addr = (T*)java_lang_ref_Reference::referent_addr(obj);
+  T heap_oop = oopDesc::load_heap_oop(referent_addr);
+  ReferenceProcessor* rp = closure->_ref_processor;
+  if (!oopDesc::is_null(heap_oop)) {
+    oop referent = oopDesc::decode_heap_oop_not_null(heap_oop);
+    if (!referent->is_gc_marked() && (rp != NULL) &&
+        rp->discover_reference(obj, reference_type())) {
+      return;
+    } else if (contains(referent_addr)) {
+      // treat referent as normal oop
+      Devirtualizer<nv>::do_oop(closure, referent_addr);
+    }
+  }
+  T* next_addr = (T*)java_lang_ref_Reference::next_addr(obj);
+  if (ReferenceProcessor::pending_list_uses_discovered_field()) {
+    T next_oop  = oopDesc::load_heap_oop(next_addr);
+    // Treat discovered as normal oop, if ref is not "active" (next non-NULL)
+    if (!oopDesc::is_null(next_oop) && contains(disc_addr)) {
+      // i.e. ref is not "active"
+      debug_only(
+        if(TraceReferenceGC && PrintGCDetails) {
+          gclog_or_tty->print_cr("   Process discovered as normal "
+                                 PTR_FORMAT, p2i(disc_addr));
+        }
+      )
+      Devirtualizer<nv>::do_oop(closure, disc_addr);
+    }
+  } else {
+    // In the case of older JDKs which do not use the discovered field for
+    // the pending list, an inactive ref (next != NULL) must always have a
+    // NULL discovered field.
+    debug_only(
+      T next_oop = oopDesc::load_heap_oop(next_addr);
+      T disc_oop = oopDesc::load_heap_oop(disc_addr);
+      assert(oopDesc::is_null(next_oop) || oopDesc::is_null(disc_oop),
+           err_msg("Found an inactive reference " PTR_FORMAT " with a non-NULL"
+                   "discovered field", p2i(obj)));
+    )
+  }
+  // treat next as normal oop
+  if (contains(next_addr)) {
+    Devirtualizer<nv>::do_oop(closure, next_addr);
+  }
+}
+
+class AlwaysContains {
+ public:
+  template <typename T> bool operator()(T* p) const { return true; }
+};
+
+template <bool nv, class OopClosureType>
+void InstanceRefKlass::oop_oop_iterate_ref_processing(oop obj, OopClosureType* closure) {
+  AlwaysContains always_contains;
+  if (UseCompressedOops) {
+    oop_oop_iterate_ref_processing_specialized<nv, narrowOop>(obj, closure, always_contains);
+  } else {
+    oop_oop_iterate_ref_processing_specialized<nv, oop>(obj, closure, always_contains);
+  }
+}
+
+class MrContains {
+  const MemRegion _mr;
+ public:
+  MrContains(MemRegion mr) : _mr(mr) {}
+  template <typename T> bool operator()(T* p) const { return _mr.contains(p); }
+};
+
+template <bool nv, class OopClosureType>
+void InstanceRefKlass::oop_oop_iterate_ref_processing_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  const MrContains contains(mr);
+  if (UseCompressedOops) {
+    oop_oop_iterate_ref_processing_specialized<nv, narrowOop>(obj, closure, contains);
+  } else {
+    oop_oop_iterate_ref_processing_specialized<nv, oop>(obj, closure, contains);
+  }
+}
+
+template <bool nv, class OopClosureType>
+int InstanceRefKlass::oop_oop_iterate(oop obj, OopClosureType* closure) {
+  // Get size before changing pointers
+  int size = InstanceKlass::oop_oop_iterate<nv>(obj, closure);
+
+  oop_oop_iterate_ref_processing<nv>(obj, closure);
+
+  return size;
+}
+
+#if INCLUDE_ALL_GCS
+template <bool nv, class OopClosureType>
+int InstanceRefKlass::
+oop_oop_iterate_reverse(oop obj, OopClosureType* closure) {
+  // Get size before changing pointers
+  int size = InstanceKlass::oop_oop_iterate_reverse<nv>(obj, closure);
+
+  oop_oop_iterate_ref_processing<nv>(obj, closure);
+
+  return size;
+}
+#endif // INCLUDE_ALL_GCS
+
+
+template <bool nv, class OopClosureType>
+int InstanceRefKlass::oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  // Get size before changing pointers
+  int size = InstanceKlass::oop_oop_iterate_bounded<nv>(obj, closure, mr);
+
+  oop_oop_iterate_ref_processing_bounded<nv>(obj, closure, mr);
+
+  return size;
+}
+
+// Macro to define InstanceRefKlass::oop_oop_iterate for virtual/nonvirtual for
+// all closures.  Macros calling macros above for each oop size.
+
+#define InstanceRefKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)              \
+                                                                                      \
+int InstanceRefKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                       \
+}
+
+#if INCLUDE_ALL_GCS
+#define InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
+                                                                                                \
+int InstanceRefKlass::oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate_reverse<nvs_to_bool(nv_suffix)>(obj, closure);                         \
+}
+#else
+#define InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+
+#define InstanceRefKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)                              \
+                                                                                                        \
+int InstanceRefKlass::oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_bounded<nvs_to_bool(nv_suffix)>(obj, closure, mr);                             \
+}
+
+#define ALL_INSTANCE_REF_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  InstanceRefKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)    \
+  InstanceRefKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)    \
+  InstanceRefKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+
+
+#endif // SHARE_VM_OOPS_INSTANCEREFKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/oops/klass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/klass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -27,7 +27,6 @@
 #include "classfile/dictionary.hpp"
 #include "classfile/systemDictionary.hpp"
 #include "classfile/vmSymbols.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
 #include "memory/heapInspection.hpp"
 #include "memory/metadataFactory.hpp"
@@ -43,9 +42,6 @@
 #include "utilities/stack.inline.hpp"
 #if INCLUDE_ALL_GCS
 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.hpp"
 #endif // INCLUDE_ALL_GCS
 
 void Klass::set_name(Symbol* n) {
--- a/hotspot/src/share/vm/oops/klass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/klass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,21 +25,14 @@
 #ifndef SHARE_VM_OOPS_KLASS_HPP
 #define SHARE_VM_OOPS_KLASS_HPP
 
-#include "memory/genOopClosures.hpp"
 #include "memory/iterator.hpp"
 #include "memory/memRegion.hpp"
 #include "memory/specialized_oop_closures.hpp"
-#include "oops/klassPS.hpp"
 #include "oops/metadata.hpp"
 #include "oops/oop.hpp"
 #include "trace/traceMacros.hpp"
 #include "utilities/accessFlags.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/concurrentMarkSweep/cmsOopClosures.hpp"
-#include "gc_implementation/g1/g1OopClosures.hpp"
-#include "gc_implementation/parNew/parOopClosures.hpp"
-#endif // INCLUDE_ALL_GCS
 
 //
 // A Klass provides:
@@ -61,6 +54,7 @@
 class ClassLoaderData;
 class klassVtable;
 class ParCompactionManager;
+class PSPromotionManager;
 class KlassSizeStats;
 class fieldDescriptor;
 
@@ -478,13 +472,6 @@
   //     and the package separators as '/'.
   virtual const char* signature_name() const;
 
-  // garbage collection support
-  virtual void oop_follow_contents(oop obj) = 0;
-  virtual int  oop_adjust_pointers(oop obj) = 0;
-
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS_PV
-
   // type testing operations
  protected:
   virtual bool oop_is_instance_slow()       const { return false; }
@@ -581,60 +568,35 @@
     clean_weak_klass_links(is_alive, false /* clean_alive_klasses */);
   }
 
-  // iterators
-  virtual int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) = 0;
-  virtual int oop_oop_iterate_v(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate(obj, blk);
-  }
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  virtual void oop_ms_follow_contents(oop obj) = 0;
+  virtual int  oop_ms_adjust_pointers(oop obj) = 0;
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  virtual void oop_ps_push_contents(  oop obj, PSPromotionManager* pm)   = 0;
+  // Parallel Compact
+  virtual void oop_pc_follow_contents(oop obj, ParCompactionManager* cm) = 0;
+  virtual void oop_pc_update_pointers(oop obj) = 0;
+#endif
+
+  // Iterators specialized to particular subtypes
+  // of ExtendedOopClosure, to avoid closure virtual calls.
+#define Klass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)                                      \
+  virtual int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) = 0;                    \
+  /* Iterates "closure" over all the oops in "obj" (of type "this") within "mr". */                \
+  virtual int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) = 0;
+
+  ALL_OOP_OOP_ITERATE_CLOSURES_1(Klass_OOP_OOP_ITERATE_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_2(Klass_OOP_OOP_ITERATE_DECL)
 
 #if INCLUDE_ALL_GCS
-  // In case we don't have a specialized backward scanner use forward
-  // iteration.
-  virtual int oop_oop_iterate_backwards_v(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
-#endif // INCLUDE_ALL_GCS
-
-  // Iterates "blk" over all the oops in "obj" (of type "this") within "mr".
-  // (I don't see why the _m should be required, but without it the Solaris
-  // C++ gives warning messages about overridings of the "oop_oop_iterate"
-  // defined above "hiding" this virtual function.  (DLD, 6/20/00)) */
-  virtual int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) = 0;
-  virtual int oop_oop_iterate_v_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_m(obj, blk, mr);
-  }
+#define Klass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)                    \
+  virtual int oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) = 0;
 
-  // Versions of the above iterators specialized to particular subtypes
-  // of OopClosure, to avoid closure virtual calls.
-#define Klass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)                \
-  virtual int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk) {     \
-    /* Default implementation reverts to general version. */                 \
-    return oop_oop_iterate(obj, blk);                                        \
-  }                                                                          \
-                                                                             \
-  /* Iterates "blk" over all the oops in "obj" (of type "this") within "mr". \
-     (I don't see why the _m should be required, but without it the Solaris  \
-     C++ gives warning messages about overridings of the "oop_oop_iterate"   \
-     defined above "hiding" this virtual function.  (DLD, 6/20/00)) */       \
-  virtual int oop_oop_iterate##nv_suffix##_m(oop obj,                        \
-                                             OopClosureType* blk,            \
-                                             MemRegion mr) {                 \
-    return oop_oop_iterate_m(obj, blk, mr);                                  \
-  }
-
-  SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_1(Klass_OOP_OOP_ITERATE_DECL)
-  SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_2(Klass_OOP_OOP_ITERATE_DECL)
-
-#if INCLUDE_ALL_GCS
-#define Klass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)      \
-  virtual int oop_oop_iterate_backwards##nv_suffix(oop obj,                  \
-                                                   OopClosureType* blk) {    \
-    /* Default implementation reverts to general version. */                 \
-    return oop_oop_iterate_backwards_v(obj, blk);                            \
-  }
-
-  SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_1(Klass_OOP_OOP_ITERATE_BACKWARDS_DECL)
-  SPECIALIZED_OOP_OOP_ITERATE_CLOSURES_2(Klass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_1(Klass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_2(Klass_OOP_OOP_ITERATE_BACKWARDS_DECL)
 #endif // INCLUDE_ALL_GCS
 
   virtual void array_klasses_do(void f(Klass* k)) {}
--- a/hotspot/src/share/vm/oops/klassPS.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,54 +0,0 @@
-/*
- * Copyright (c) 2007, 2013, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#ifndef SHARE_VM_OOPS_KLASSPS_HPP
-#define SHARE_VM_OOPS_KLASSPS_HPP
-
-  // Expands to Parallel Scavenge and Parallel Old declarations
-
-#include "utilities/macros.hpp"
-
-#if INCLUDE_ALL_GCS
-#define PARALLEL_GC_DECLS \
-  virtual void oop_push_contents(PSPromotionManager* pm, oop obj);          \
-  /* Parallel Old GC support                                                \
-                                                                            \
-   The 2-arg version of oop_update_pointers is for objects that are         \
-   known not to cross chunk boundaries.  The 4-arg version is for           \
-   objects that do (or may) cross chunk boundaries; it updates only those   \
-   oops that are in the region [beg_addr, end_addr).  */                    \
-  virtual void oop_follow_contents(ParCompactionManager* cm, oop obj);      \
-  virtual int  oop_update_pointers(ParCompactionManager* cm, oop obj);
-
-// Pure virtual version for klass.hpp
-#define PARALLEL_GC_DECLS_PV \
-  virtual void oop_push_contents(PSPromotionManager* pm, oop obj) = 0;      \
-  virtual void oop_follow_contents(ParCompactionManager* cm, oop obj) = 0;  \
-  virtual int  oop_update_pointers(ParCompactionManager* cm, oop obj) = 0;
-#else  // INCLUDE_ALL_GCS
-#define PARALLEL_GC_DECLS
-#define PARALLEL_GC_DECLS_PV
-#endif // INCLUDE_ALL_GCS
-
-#endif // SHARE_VM_OOPS_KLASSPS_HPP
--- a/hotspot/src/share/vm/oops/methodData.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/methodData.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -152,6 +152,7 @@
 
 void BitData::print_data_on(outputStream* st, const char* extra) const {
   print_shared(st, "BitData", extra);
+  st->cr();
 }
 
 // ==================================================================
--- a/hotspot/src/share/vm/oops/methodData.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/methodData.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -2056,7 +2056,7 @@
 
   // Whole-method sticky bits and flags
   enum {
-    _trap_hist_limit    = 21,   // decoupled from Deoptimization::Reason_LIMIT
+    _trap_hist_limit    = 22,   // decoupled from Deoptimization::Reason_LIMIT
     _trap_hist_mask     = max_jubyte,
     _extra_data_count   = 4     // extra DataLayout headers, for trap history
   }; // Public flag values
--- a/hotspot/src/share/vm/oops/objArrayKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/objArrayKlass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,9 +26,7 @@
 #include "classfile/symbolTable.hpp"
 #include "classfile/systemDictionary.hpp"
 #include "classfile/vmSymbols.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
-#include "memory/genOopClosures.inline.hpp"
 #include "memory/iterator.inline.hpp"
 #include "memory/metadataFactory.hpp"
 #include "memory/resourceArea.hpp"
@@ -45,17 +43,6 @@
 #include "runtime/orderAccess.inline.hpp"
 #include "utilities/copy.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/concurrentMarkSweep/cmsOopClosures.inline.hpp"
-#include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
-#include "gc_implementation/g1/g1OopClosures.inline.hpp"
-#include "gc_implementation/g1/g1RemSet.inline.hpp"
-#include "gc_implementation/g1/heapRegionManager.inline.hpp"
-#include "gc_implementation/parNew/parOopClosures.inline.hpp"
-#include "gc_implementation/parallelScavenge/psCompactionManager.hpp"
-#include "gc_implementation/parallelScavenge/psPromotionManager.inline.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.inline.hpp"
-#endif // INCLUDE_ALL_GCS
 
 ObjArrayKlass* ObjArrayKlass::allocate(ClassLoaderData* loader_data, int n, KlassHandle klass_handle, Symbol* name, TRAPS) {
   assert(ObjArrayKlass::header_size() <= InstanceKlass::header_size(),
@@ -410,179 +397,6 @@
   bottom_klass()->initialize(THREAD);  // dispatches to either InstanceKlass or TypeArrayKlass
 }
 
-#define ObjArrayKlass_SPECIALIZED_OOP_ITERATE(T, a, p, do_oop) \
-{                                   \
-  T* p         = (T*)(a)->base();   \
-  T* const end = p + (a)->length(); \
-  while (p < end) {                 \
-    do_oop;                         \
-    p++;                            \
-  }                                 \
-}
-
-#define ObjArrayKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(T, a, p, low, high, do_oop) \
-{                                   \
-  T* const l = (T*)(low);           \
-  T* const h = (T*)(high);          \
-  T* p       = (T*)(a)->base();     \
-  T* end     = p + (a)->length();   \
-  if (p < l) p = l;                 \
-  if (end > h) end = h;             \
-  while (p < end) {                 \
-    do_oop;                         \
-    ++p;                            \
-  }                                 \
-}
-
-#define ObjArrayKlass_OOP_ITERATE(a, p, do_oop)      \
-  if (UseCompressedOops) {                           \
-    ObjArrayKlass_SPECIALIZED_OOP_ITERATE(narrowOop, \
-      a, p, do_oop)                                  \
-  } else {                                           \
-    ObjArrayKlass_SPECIALIZED_OOP_ITERATE(oop,       \
-      a, p, do_oop)                                  \
-  }
-
-#define ObjArrayKlass_BOUNDED_OOP_ITERATE(a, p, low, high, do_oop) \
-  if (UseCompressedOops) {                                   \
-    ObjArrayKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(narrowOop, \
-      a, p, low, high, do_oop)                               \
-  } else {                                                   \
-    ObjArrayKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(oop,       \
-      a, p, low, high, do_oop)                               \
-  }
-
-void ObjArrayKlass::oop_follow_contents(oop obj) {
-  assert (obj->is_array(), "obj must be array");
-  MarkSweep::follow_klass(obj->klass());
-  if (UseCompressedOops) {
-    objarray_follow_contents<narrowOop>(obj, 0);
-  } else {
-    objarray_follow_contents<oop>(obj, 0);
-  }
-}
-
-#if INCLUDE_ALL_GCS
-void ObjArrayKlass::oop_follow_contents(ParCompactionManager* cm,
-                                        oop obj) {
-  assert(obj->is_array(), "obj must be array");
-  PSParallelCompact::follow_klass(cm, obj->klass());
-  if (UseCompressedOops) {
-    objarray_follow_contents<narrowOop>(cm, obj, 0);
-  } else {
-    objarray_follow_contents<oop>(cm, obj, 0);
-  }
-}
-#endif // INCLUDE_ALL_GCS
-
-#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)           \
-                                                                                \
-int ObjArrayKlass::oop_oop_iterate##nv_suffix(oop obj,                          \
-                                              OopClosureType* closure) {        \
-  assert (obj->is_array(), "obj must be array");                                \
-  objArrayOop a = objArrayOop(obj);                                             \
-  /* Get size before changing pointers. */                                      \
-  /* Don't call size() or oop_size() since that is a virtual call. */           \
-  int size = a->object_size();                                                  \
-  if_do_metadata_checked(closure, nv_suffix) {                                  \
-    closure->do_klass##nv_suffix(obj->klass());                                 \
-  }                                                                             \
-  ObjArrayKlass_OOP_ITERATE(a, p, (closure)->do_oop##nv_suffix(p))              \
-  return size;                                                                  \
-}
-
-#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)         \
-                                                                                \
-int ObjArrayKlass::oop_oop_iterate##nv_suffix##_m(oop obj,                      \
-                                                  OopClosureType* closure,      \
-                                                  MemRegion mr) {               \
-  assert(obj->is_array(), "obj must be array");                                 \
-  objArrayOop a  = objArrayOop(obj);                                            \
-  /* Get size before changing pointers. */                                      \
-  /* Don't call size() or oop_size() since that is a virtual call */            \
-  int size = a->object_size();                                                  \
-  if_do_metadata_checked(closure, nv_suffix) {                                  \
-    /* SSS: Do we need to pass down mr here? */                                 \
-    closure->do_klass##nv_suffix(a->klass());                                   \
-  }                                                                             \
-  ObjArrayKlass_BOUNDED_OOP_ITERATE(                                            \
-    a, p, mr.start(), mr.end(), (closure)->do_oop##nv_suffix(p))                \
-  return size;                                                                  \
-}
-
-// Like oop_oop_iterate but only iterates over a specified range and only used
-// for objArrayOops.
-#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN_r(OopClosureType, nv_suffix)         \
-                                                                                \
-int ObjArrayKlass::oop_oop_iterate_range##nv_suffix(oop obj,                    \
-                                                  OopClosureType* closure,      \
-                                                  int start, int end) {         \
-  assert(obj->is_array(), "obj must be array");                                 \
-  objArrayOop a  = objArrayOop(obj);                                            \
-  /* Get size before changing pointers. */                                      \
-  /* Don't call size() or oop_size() since that is a virtual call */            \
-  int size = a->object_size();                                                  \
-  if (UseCompressedOops) {                                                      \
-    HeapWord* low = start == 0 ? (HeapWord*)a : (HeapWord*)a->obj_at_addr<narrowOop>(start);\
-    /* this might be wierd if end needs to be aligned on HeapWord boundary */   \
-    HeapWord* high = (HeapWord*)((narrowOop*)a->base() + end);                  \
-    MemRegion mr(low, high);                                                    \
-    if_do_metadata_checked(closure, nv_suffix) {                                \
-      /* SSS: Do we need to pass down mr here? */                               \
-      closure->do_klass##nv_suffix(a->klass());                                 \
-    }                                                                           \
-    ObjArrayKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(narrowOop,                    \
-      a, p, low, high, (closure)->do_oop##nv_suffix(p))                         \
-  } else {                                                                      \
-    HeapWord* low = start == 0 ? (HeapWord*)a : (HeapWord*)a->obj_at_addr<oop>(start);  \
-    HeapWord* high = (HeapWord*)((oop*)a->base() + end);                        \
-    MemRegion mr(low, high);                                                    \
-    if_do_metadata_checked(closure, nv_suffix) {                                \
-      /* SSS: Do we need to pass down mr here? */                               \
-      closure->do_klass##nv_suffix(a->klass());                                 \
-    }                                                                           \
-    ObjArrayKlass_SPECIALIZED_BOUNDED_OOP_ITERATE(oop,                          \
-      a, p, low, high, (closure)->do_oop##nv_suffix(p))                         \
-  }                                                                             \
-  return size;                                                                  \
-}
-
-ALL_OOP_OOP_ITERATE_CLOSURES_1(ObjArrayKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(ObjArrayKlass_OOP_OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_1(ObjArrayKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(ObjArrayKlass_OOP_OOP_ITERATE_DEFN_m)
-ALL_OOP_OOP_ITERATE_CLOSURES_1(ObjArrayKlass_OOP_OOP_ITERATE_DEFN_r)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(ObjArrayKlass_OOP_OOP_ITERATE_DEFN_r)
-
-int ObjArrayKlass::oop_adjust_pointers(oop obj) {
-  assert(obj->is_objArray(), "obj must be obj array");
-  objArrayOop a = objArrayOop(obj);
-  // Get size before changing pointers.
-  // Don't call size() or oop_size() since that is a virtual call.
-  int size = a->object_size();
-  ObjArrayKlass_OOP_ITERATE(a, p, MarkSweep::adjust_pointer(p))
-  return size;
-}
-
-#if INCLUDE_ALL_GCS
-void ObjArrayKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  assert(obj->is_objArray(), "obj must be obj array");
-  ObjArrayKlass_OOP_ITERATE( \
-    objArrayOop(obj), p, \
-    if (PSScavenge::should_scavenge(p)) { \
-      pm->claim_or_forward_depth(p); \
-    })
-}
-
-int ObjArrayKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  assert (obj->is_objArray(), "obj must be obj array");
-  objArrayOop a = objArrayOop(obj);
-  int size = a->object_size();
-  ObjArrayKlass_OOP_ITERATE(a, p, PSParallelCompact::adjust_pointer(p))
-  return size;
-}
-#endif // INCLUDE_ALL_GCS
-
 // JVM support
 
 jint ObjArrayKlass::compute_modifier_flags(TRAPS) const {
--- a/hotspot/src/share/vm/oops/objArrayKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/objArrayKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,7 +26,6 @@
 #define SHARE_VM_OOPS_OBJARRAYKLASS_HPP
 
 #include "classfile/classLoaderData.hpp"
-#include "memory/specialized_oop_closures.hpp"
 #include "oops/arrayKlass.hpp"
 #include "utilities/macros.hpp"
 
@@ -103,28 +102,67 @@
   // Initialization (virtual from Klass)
   void initialize(TRAPS);
 
-  // Garbage collection
-  void oop_follow_contents(oop obj);
-  inline void oop_follow_contents(oop obj, int index);
-  template <class T> inline void objarray_follow_contents(oop obj, int index);
-
-  int  oop_adjust_pointers(oop obj);
-
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
 #if INCLUDE_ALL_GCS
-  inline void oop_follow_contents(ParCompactionManager* cm, oop obj, int index);
-  template <class T> inline void
-    objarray_follow_contents(ParCompactionManager* cm, oop obj, int index);
-#endif // INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
+
+  // Oop fields (and metadata) iterators
+  //  [nv = true]  Use non-virtual calls to do_oop_nv.
+  //  [nv = false] Use virtual calls to do_oop.
+  //
+  // The ObjArrayKlass iterators also visits the Object's klass.
+
+ private:
+
+  // Iterate over oop elements and metadata.
+  template <bool nv, typename OopClosureType>
+  inline int oop_oop_iterate(oop obj, OopClosureType* closure);
+
+  // Iterate over oop elements within mr, and metadata.
+  template <bool nv, typename OopClosureType>
+  inline int oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr);
+
+  // Iterate over oop elements with indices within [start, end), and metadata.
+  template <bool nv, class OopClosureType>
+  inline int oop_oop_iterate_range(oop obj, OopClosureType* closure, int start, int end);
 
-  // Iterators
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-    return oop_oop_iterate_v(obj, blk);
-  }
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-    return oop_oop_iterate_v_m(obj, blk, mr);
-  }
+  // Iterate over oop elements within [start, end), and metadata.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_range_specialized(objArrayOop a, OopClosureType* closure, int start, int end);
+
+ public:
+  // Iterate over all oop elements.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_elements(objArrayOop a, OopClosureType* closure);
+
+ private:
+  // Iterate over all oop elements.
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_elements_specialized(objArrayOop a, OopClosureType* closure);
+
+  // Iterate over all oop elements with indices within mr.
+  template <bool nv, class OopClosureType>
+  inline void oop_oop_iterate_elements_bounded(objArrayOop a, OopClosureType* closure, MemRegion mr);
+
+  // Iterate over oop elements within [low, high)..
+  // Specialized for [T = oop] or [T = narrowOop].
+  template <bool nv, typename T, class OopClosureType>
+  inline void oop_oop_iterate_elements_specialized_bounded(objArrayOop a, OopClosureType* closure, void* low, void* high);
+
+
+ public:
+
 #define ObjArrayKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)   \
   int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* blk);         \
   int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* blk,      \
@@ -135,6 +173,14 @@
   ALL_OOP_OOP_ITERATE_CLOSURES_1(ObjArrayKlass_OOP_OOP_ITERATE_DECL)
   ALL_OOP_OOP_ITERATE_CLOSURES_2(ObjArrayKlass_OOP_OOP_ITERATE_DECL)
 
+#if INCLUDE_ALL_GCS
+#define ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix) \
+  int  oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* blk);
+
+  ALL_OOP_OOP_ITERATE_CLOSURES_1(ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_2(ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+#endif // INCLUDE_ALL_GCS
+
   // JVM support
   jint compute_modifier_flags(TRAPS) const;
 
--- a/hotspot/src/share/vm/oops/objArrayKlass.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/objArrayKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2010, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2010, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -25,78 +25,165 @@
 #ifndef SHARE_VM_OOPS_OBJARRAYKLASS_INLINE_HPP
 #define SHARE_VM_OOPS_OBJARRAYKLASS_INLINE_HPP
 
-#include "gc_implementation/shared/markSweep.inline.hpp"
+#include "memory/memRegion.hpp"
+#include "memory/iterator.inline.hpp"
 #include "oops/objArrayKlass.hpp"
+#include "oops/objArrayOop.inline.hpp"
+#include "oops/oop.inline.hpp"
 #include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/parallelScavenge/psCompactionManager.inline.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
-#endif // INCLUDE_ALL_GCS
+
+template <bool nv, typename T, class OopClosureType>
+void ObjArrayKlass::oop_oop_iterate_elements_specialized(objArrayOop a, OopClosureType* closure) {
+  T* p         = (T*)a->base();
+  T* const end = p + a->length();
+
+  for (;p < end; p++) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+template <bool nv, typename T, class OopClosureType>
+void ObjArrayKlass::oop_oop_iterate_elements_specialized_bounded(
+    objArrayOop a, OopClosureType* closure, void* low, void* high) {
+
+  T* const l = (T*)low;
+  T* const h = (T*)high;
+
+  T* p   = (T*)a->base();
+  T* end = p + a->length();
 
-void ObjArrayKlass::oop_follow_contents(oop obj, int index) {
+  if (p < l) {
+    p = l;
+  }
+  if (end > h) {
+    end = h;
+  }
+
+  for (;p < end; ++p) {
+    Devirtualizer<nv>::do_oop(closure, p);
+  }
+}
+
+template <bool nv, class OopClosureType>
+void ObjArrayKlass::oop_oop_iterate_elements(objArrayOop a, OopClosureType* closure) {
   if (UseCompressedOops) {
-    objarray_follow_contents<narrowOop>(obj, index);
+    oop_oop_iterate_elements_specialized<nv, narrowOop>(a, closure);
   } else {
-    objarray_follow_contents<oop>(obj, index);
+    oop_oop_iterate_elements_specialized<nv, oop>(a, closure);
+  }
+}
+
+template <bool nv, class OopClosureType>
+void ObjArrayKlass::oop_oop_iterate_elements_bounded(objArrayOop a, OopClosureType* closure, MemRegion mr) {
+  if (UseCompressedOops) {
+    oop_oop_iterate_elements_specialized_bounded<nv, narrowOop>(a, closure, mr.start(), mr.end());
+  } else {
+    oop_oop_iterate_elements_specialized_bounded<nv, oop>(a, closure, mr.start(), mr.end());
   }
 }
 
-template <class T>
-void ObjArrayKlass::objarray_follow_contents(oop obj, int index) {
+template <bool nv, typename OopClosureType>
+int ObjArrayKlass::oop_oop_iterate(oop obj, OopClosureType* closure) {
+  assert (obj->is_array(), "obj must be array");
   objArrayOop a = objArrayOop(obj);
-  const size_t len = size_t(a->length());
-  const size_t beg_index = size_t(index);
-  assert(beg_index < len || len == 0, "index too large");
+
+  // Get size before changing pointers.
+  // Don't call size() or oop_size() since that is a virtual call.
+  int size = a->object_size();
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    Devirtualizer<nv>::do_klass(closure, obj->klass());
+  }
+
+  oop_oop_iterate_elements<nv>(a, closure);
 
-  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
-  const size_t end_index = beg_index + stride;
-  T* const base = (T*)a->base();
-  T* const beg = base + beg_index;
-  T* const end = base + end_index;
+  return size;
+}
 
-  // Push the non-NULL elements of the next stride on the marking stack.
-  for (T* e = beg; e < end; e++) {
-    MarkSweep::mark_and_push<T>(e);
+template <bool nv, typename OopClosureType>
+int ObjArrayKlass::oop_oop_iterate_bounded(oop obj, OopClosureType* closure, MemRegion mr) {
+  assert(obj->is_array(), "obj must be array");
+  objArrayOop a  = objArrayOop(obj);
+
+  // Get size before changing pointers.
+  // Don't call size() or oop_size() since that is a virtual call
+  int size = a->object_size();
+
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    Devirtualizer<nv>::do_klass(closure, a->klass());
   }
 
-  if (end_index < len) {
-    MarkSweep::push_objarray(a, end_index); // Push the continuation.
+  oop_oop_iterate_elements_bounded<nv>(a, closure, mr);
+
+  return size;
+}
+
+template <bool nv, typename T, class OopClosureType>
+void ObjArrayKlass::oop_oop_iterate_range_specialized(objArrayOop a, OopClosureType* closure, int start, int end) {
+  if (Devirtualizer<nv>::do_metadata(closure)) {
+    Devirtualizer<nv>::do_klass(closure, a->klass());
   }
+
+  T* low = start == 0 ? cast_from_oop<T*>(a) : a->obj_at_addr<T>(start);
+  T* high = (T*)a->base() + end;
+
+  oop_oop_iterate_elements_specialized_bounded<nv, T>(a, closure, low, high);
+}
+
+// Like oop_oop_iterate but only iterates over a specified range and only used
+// for objArrayOops.
+template <bool nv, class OopClosureType>
+int ObjArrayKlass::oop_oop_iterate_range(oop obj, OopClosureType* closure, int start, int end) {
+  assert(obj->is_array(), "obj must be array");
+  objArrayOop a  = objArrayOop(obj);
+
+  // Get size before changing pointers.
+  // Don't call size() or oop_size() since that is a virtual call
+  int size = a->object_size();
+
+  if (UseCompressedOops) {
+    oop_oop_iterate_range_specialized<nv, narrowOop>(a, closure, start, end);
+  } else {
+    oop_oop_iterate_range_specialized<nv, oop>(a, closure, start, end);
+  }
+
+  return size;
+}
+
+
+#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)              \
+                                                                                   \
+int ObjArrayKlass::oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {  \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                    \
 }
 
 #if INCLUDE_ALL_GCS
-void ObjArrayKlass::oop_follow_contents(ParCompactionManager* cm, oop obj,
-                                        int index) {
-  if (UseCompressedOops) {
-    objarray_follow_contents<narrowOop>(cm, obj, index);
-  } else {
-    objarray_follow_contents<oop>(cm, obj, index);
-  }
+#define ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
+int ObjArrayKlass::oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {  \
+  /* No reverse implementation ATM. */                                                       \
+  return oop_oop_iterate<nvs_to_bool(nv_suffix)>(obj, closure);                              \
+}
+#else
+#define ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)                              \
+                                                                                                     \
+int ObjArrayKlass::oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_bounded<nvs_to_bool(nv_suffix)>(obj, closure, mr);                          \
 }
 
-template <class T>
-void ObjArrayKlass::objarray_follow_contents(ParCompactionManager* cm, oop obj,
-                                             int index) {
-  objArrayOop a = objArrayOop(obj);
-  const size_t len = size_t(a->length());
-  const size_t beg_index = size_t(index);
-  assert(beg_index < len || len == 0, "index too large");
+#define ObjArrayKlass_OOP_OOP_ITERATE_DEFN_r(OopClosureType, nv_suffix)                                      \
+                                                                                                             \
+int ObjArrayKlass::oop_oop_iterate_range##nv_suffix(oop obj, OopClosureType* closure, int start, int end) {  \
+  return oop_oop_iterate_range<nvs_to_bool(nv_suffix)>(obj, closure, start, end);                            \
+}
+
 
-  const size_t stride = MIN2(len - beg_index, ObjArrayMarkingStride);
-  const size_t end_index = beg_index + stride;
-  T* const base = (T*)a->base();
-  T* const beg = base + beg_index;
-  T* const end = base + end_index;
+#define ALL_OBJ_ARRAY_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  ObjArrayKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)    \
+  ObjArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)    \
+  ObjArrayKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)    \
+  ObjArrayKlass_OOP_OOP_ITERATE_DEFN_r(        OopClosureType, nv_suffix)
 
-  // Push the non-NULL elements of the next stride on the marking stack.
-  for (T* e = beg; e < end; e++) {
-    PSParallelCompact::mark_and_push<T>(cm, e);
-  }
-
-  if (end_index < len) {
-    cm->push_objarray(a, end_index); // Push the continuation.
-  }
-}
-#endif // INCLUDE_ALL_GCS
 
 #endif // SHARE_VM_OOPS_OBJARRAYKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/oops/oop.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/oop.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -298,19 +298,6 @@
 
   // garbage collection
   bool is_gc_marked() const;
-  // Apply "MarkSweep::mark_and_push" to (the address of) every non-NULL
-  // reference field in "this".
-  void follow_contents(void);
-
-#if INCLUDE_ALL_GCS
-  // Parallel Scavenge
-  void push_contents(PSPromotionManager* pm);
-
-  // Parallel Old
-  void update_contents(ParCompactionManager* cm);
-
-  void follow_contents(ParCompactionManager* cm);
-#endif // INCLUDE_ALL_GCS
 
   bool is_scavengable() const;
 
@@ -334,9 +321,6 @@
   uint age() const;
   void incr_age();
 
-  // Adjust all pointers in this object to point at it's forwarded location and
-  // return the size of this oop.  This is used by the MarkSweep collector.
-  int adjust_pointers();
 
   // mark-sweep support
   void follow_body(int begin, int end);
@@ -345,6 +329,22 @@
   static BarrierSet* bs()            { return _bs; }
   static void set_bs(BarrierSet* bs) { _bs = bs; }
 
+  // Garbage Collection support
+
+  // Mark Sweep
+  void ms_follow_contents();
+  // Adjust all pointers in this object to point at it's forwarded location and
+  // return the size of this oop.  This is used by the MarkSweep collector.
+  int  ms_adjust_pointers();
+#if INCLUDE_ALL_GCS
+  // Parallel Compact
+  void pc_follow_contents(ParCompactionManager* pc);
+  void pc_update_contents();
+  // Parallel Scavenge
+  void ps_push_contents(PSPromotionManager* pm);
+#endif
+
+
   // iterators, returns size of object
 #define OOP_ITERATE_DECL(OopClosureType, nv_suffix)                      \
   int oop_iterate(OopClosureType* blk);                                  \
--- a/hotspot/src/share/vm/oops/oop.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/oop.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -26,13 +26,11 @@
 #define SHARE_VM_OOPS_OOP_INLINE_HPP
 
 #include "gc_implementation/shared/ageTable.hpp"
-#include "gc_implementation/shared/markSweep.inline.hpp"
 #include "gc_interface/collectedHeap.inline.hpp"
 #include "memory/barrierSet.inline.hpp"
 #include "memory/cardTableModRefBS.hpp"
 #include "memory/genCollectedHeap.hpp"
 #include "memory/generation.hpp"
-#include "memory/specialized_oop_closures.hpp"
 #include "oops/arrayKlass.hpp"
 #include "oops/arrayOop.hpp"
 #include "oops/klass.inline.hpp"
@@ -592,11 +590,6 @@
 }
 #endif // PRODUCT
 
-inline void oopDesc::follow_contents(void) {
-  assert (is_gc_marked(), "should be marked");
-  klass()->oop_follow_contents(this);
-}
-
 inline bool oopDesc::is_scavengable() const {
   return Universe::heap()->is_scavengable(this);
 }
@@ -706,21 +699,49 @@
   }
 }
 
-inline int oopDesc::adjust_pointers() {
+inline void oopDesc::ms_follow_contents() {
+  klass()->oop_ms_follow_contents(this);
+}
+
+inline int oopDesc::ms_adjust_pointers() {
   debug_only(int check_size = size());
-  int s = klass()->oop_adjust_pointers(this);
+  int s = klass()->oop_ms_adjust_pointers(this);
   assert(s == check_size, "should be the same");
   return s;
 }
 
-#define OOP_ITERATE_DEFN(OopClosureType, nv_suffix)                        \
-                                                                           \
-inline int oopDesc::oop_iterate(OopClosureType* blk) {                     \
-  return klass()->oop_oop_iterate##nv_suffix(this, blk);               \
-}                                                                          \
-                                                                           \
-inline int oopDesc::oop_iterate(OopClosureType* blk, MemRegion mr) {       \
-  return klass()->oop_oop_iterate##nv_suffix##_m(this, blk, mr);       \
+#if INCLUDE_ALL_GCS
+inline void oopDesc::pc_follow_contents(ParCompactionManager* cm) {
+  klass()->oop_pc_follow_contents(this, cm);
+}
+
+inline void oopDesc::pc_update_contents() {
+  Klass* k = klass();
+  if (!k->oop_is_typeArray()) {
+    // It might contain oops beyond the header, so take the virtual call.
+    k->oop_pc_update_pointers(this);
+  }
+  // Else skip it.  The TypeArrayKlass in the header never needs scavenging.
+}
+
+inline void oopDesc::ps_push_contents(PSPromotionManager* pm) {
+  Klass* k = klass();
+  if (!k->oop_is_typeArray()) {
+    // It might contain oops beyond the header, so take the virtual call.
+    k->oop_ps_push_contents(this, pm);
+  }
+  // Else skip it.  The TypeArrayKlass in the header never needs scavenging.
+}
+#endif
+
+#define OOP_ITERATE_DEFN(OopClosureType, nv_suffix)                   \
+                                                                      \
+inline int oopDesc::oop_iterate(OopClosureType* blk) {                \
+  return klass()->oop_oop_iterate##nv_suffix(this, blk);              \
+}                                                                     \
+                                                                      \
+inline int oopDesc::oop_iterate(OopClosureType* blk, MemRegion mr) {  \
+  return klass()->oop_oop_iterate##nv_suffix##_m(this, blk, mr);      \
 }
 
 
@@ -736,18 +757,21 @@
   return oop_iterate(&cl, mr);
 }
 
-ALL_OOP_OOP_ITERATE_CLOSURES_1(OOP_ITERATE_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(OOP_ITERATE_DEFN)
+#if INCLUDE_ALL_GCS
+#define OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)       \
+                                                                    \
+inline int oopDesc::oop_iterate_backwards(OopClosureType* blk) {    \
+  return klass()->oop_oop_iterate_backwards##nv_suffix(this, blk);  \
+}
+#else
+#define OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
 
-#if INCLUDE_ALL_GCS
-#define OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)              \
-                                                                           \
-inline int oopDesc::oop_iterate_backwards(OopClosureType* blk) {           \
-  return klass()->oop_oop_iterate_backwards##nv_suffix(this, blk);     \
-}
+#define ALL_OOPDESC_OOP_ITERATE(OopClosureType, nv_suffix)  \
+  OOP_ITERATE_DEFN(OopClosureType, nv_suffix)               \
+  OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
 
-ALL_OOP_OOP_ITERATE_CLOSURES_1(OOP_ITERATE_BACKWARDS_DEFN)
-ALL_OOP_OOP_ITERATE_CLOSURES_2(OOP_ITERATE_BACKWARDS_DEFN)
-#endif // INCLUDE_ALL_GCS
+ALL_OOP_OOP_ITERATE_CLOSURES_1(ALL_OOPDESC_OOP_ITERATE)
+ALL_OOP_OOP_ITERATE_CLOSURES_2(ALL_OOPDESC_OOP_ITERATE)
 
 #endif // SHARE_VM_OOPS_OOP_INLINE_HPP
--- a/hotspot/src/share/vm/oops/oop.pcgc.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,57 +0,0 @@
-/*
- * Copyright (c) 2005, 2015, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#ifndef SHARE_VM_OOPS_OOP_PCGC_INLINE_HPP
-#define SHARE_VM_OOPS_OOP_PCGC_INLINE_HPP
-
-#include "runtime/atomic.inline.hpp"
-#include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/parNew/parNewGeneration.hpp"
-#include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
-#include "gc_implementation/parallelScavenge/psCompactionManager.hpp"
-#include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.hpp"
-#endif // INCLUDE_ALL_GCS
-
-inline void oopDesc::update_contents(ParCompactionManager* cm) {
-  // The klass field must be updated before anything else
-  // can be done.
-  DEBUG_ONLY(Klass* original_klass = klass());
-
-  Klass* new_klass = klass();
-  if (!new_klass->oop_is_typeArray()) {
-    // It might contain oops beyond the header, so take the virtual call.
-    new_klass->oop_update_pointers(cm, this);
-  }
-  // Else skip it.  The TypeArrayKlass in the header never needs scavenging.
-}
-
-inline void oopDesc::follow_contents(ParCompactionManager* cm) {
-  assert (PSParallelCompact::mark_bitmap()->is_marked(this),
-    "should be marked");
-  klass()->oop_follow_contents(cm, this);
-}
-
-#endif // SHARE_VM_OOPS_OOP_PCGC_INLINE_HPP
--- a/hotspot/src/share/vm/oops/oop.psgc.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
@@ -1,45 +0,0 @@
-/*
- * Copyright (c) 2002, 2013, Oracle and/or its affiliates. All rights reserved.
- * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
- *
- * This code is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 only, as
- * published by the Free Software Foundation.
- *
- * This code is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
- * version 2 for more details (a copy is included in the LICENSE file that
- * accompanied this code).
- *
- * You should have received a copy of the GNU General Public License version
- * 2 along with this work; if not, write to the Free Software Foundation,
- * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
- *
- * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
- * or visit www.oracle.com if you need additional information or have any
- * questions.
- *
- */
-
-#ifndef SHARE_VM_OOPS_OOP_PSGC_INLINE_HPP
-#define SHARE_VM_OOPS_OOP_PSGC_INLINE_HPP
-
-#include "utilities/macros.hpp"
-#if INCLUDE_ALL_GCS
-#include "gc_implementation/parallelScavenge/parallelScavengeHeap.hpp"
-#include "gc_implementation/parallelScavenge/psScavenge.hpp"
-#endif // INCLUDE_ALL_GCS
-
-// ParallelScavengeHeap methods
-
-inline void oopDesc::push_contents(PSPromotionManager* pm) {
-  Klass* k = klass();
-  if (!k->oop_is_typeArray()) {
-    // It might contain oops beyond the header, so take the virtual call.
-    k->oop_push_contents(pm, this);
-  }
-  // Else skip it.  The TypeArrayKlass in the header never needs scavenging.
-}
-
-#endif // SHARE_VM_OOPS_OOP_PSGC_INLINE_HPP
--- a/hotspot/src/share/vm/oops/typeArrayKlass.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/typeArrayKlass.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -36,7 +36,7 @@
 #include "oops/klass.inline.hpp"
 #include "oops/objArrayKlass.hpp"
 #include "oops/oop.inline.hpp"
-#include "oops/typeArrayKlass.hpp"
+#include "oops/typeArrayKlass.inline.hpp"
 #include "oops/typeArrayOop.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/orderAccess.inline.hpp"
@@ -204,57 +204,6 @@
   return t->object_size();
 }
 
-void TypeArrayKlass::oop_follow_contents(oop obj) {
-  assert(obj->is_typeArray(),"must be a type array");
-  // Performance tweak: We skip iterating over the klass pointer since we
-  // know that Universe::TypeArrayKlass never moves.
-}
-
-#if INCLUDE_ALL_GCS
-void TypeArrayKlass::oop_follow_contents(ParCompactionManager* cm, oop obj) {
-  assert(obj->is_typeArray(),"must be a type array");
-  // Performance tweak: We skip iterating over the klass pointer since we
-  // know that Universe::TypeArrayKlass never moves.
-}
-#endif // INCLUDE_ALL_GCS
-
-int TypeArrayKlass::oop_adjust_pointers(oop obj) {
-  assert(obj->is_typeArray(),"must be a type array");
-  typeArrayOop t = typeArrayOop(obj);
-  // Performance tweak: We skip iterating over the klass pointer since we
-  // know that Universe::TypeArrayKlass never moves.
-  return t->object_size();
-}
-
-int TypeArrayKlass::oop_oop_iterate(oop obj, ExtendedOopClosure* blk) {
-  assert(obj->is_typeArray(),"must be a type array");
-  typeArrayOop t = typeArrayOop(obj);
-  // Performance tweak: We skip iterating over the klass pointer since we
-  // know that Universe::TypeArrayKlass never moves.
-  return t->object_size();
-}
-
-int TypeArrayKlass::oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr) {
-  assert(obj->is_typeArray(),"must be a type array");
-  typeArrayOop t = typeArrayOop(obj);
-  // Performance tweak: We skip iterating over the klass pointer since we
-  // know that Universe::TypeArrayKlass never moves.
-  return t->object_size();
-}
-
-#if INCLUDE_ALL_GCS
-void TypeArrayKlass::oop_push_contents(PSPromotionManager* pm, oop obj) {
-  ShouldNotReachHere();
-  assert(obj->is_typeArray(),"must be a type array");
-}
-
-int
-TypeArrayKlass::oop_update_pointers(ParCompactionManager* cm, oop obj) {
-  assert(obj->is_typeArray(),"must be a type array");
-  return typeArrayOop(obj)->object_size();
-}
-#endif // INCLUDE_ALL_GCS
-
 void TypeArrayKlass::initialize(TRAPS) {
   // Nothing to do. Having this function is handy since objArrayKlasses can be
   // initialized by calling initialize on their bottom_klass, see ObjArrayKlass::initialize
--- a/hotspot/src/share/vm/oops/typeArrayKlass.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/oops/typeArrayKlass.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -72,16 +72,46 @@
   // Copying
   void  copy_array(arrayOop s, int src_pos, arrayOop d, int dst_pos, int length, TRAPS);
 
-  // Iteration
-  int oop_oop_iterate(oop obj, ExtendedOopClosure* blk);
-  int oop_oop_iterate_m(oop obj, ExtendedOopClosure* blk, MemRegion mr);
+  // GC specific object visitors
+  //
+  // Mark Sweep
+  void oop_ms_follow_contents(oop obj);
+  int  oop_ms_adjust_pointers(oop obj);
+#if INCLUDE_ALL_GCS
+  // Parallel Scavenge
+  void oop_ps_push_contents(  oop obj, PSPromotionManager* pm);
+  // Parallel Compact
+  void oop_pc_follow_contents(oop obj, ParCompactionManager* cm);
+  void oop_pc_update_pointers(oop obj);
+#endif
+
+  // Oop iterators. Since there are no oops in TypeArrayKlasses,
+  // these functions only return the size of the object.
+
+ private:
+  // The implementation used by all oop_oop_iterate functions in TypeArrayKlasses.
+  inline int oop_oop_iterate_impl(oop obj, ExtendedOopClosure* closure);
 
-  // Garbage collection
-  void oop_follow_contents(oop obj);
-  int  oop_adjust_pointers(oop obj);
+ public:
+
+#define TypeArrayKlass_OOP_OOP_ITERATE_DECL(OopClosureType, nv_suffix)    \
+  int oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure);       \
+  int oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure,    \
+                                     MemRegion mr);                       \
+  int oop_oop_iterate_range##nv_suffix(oop obj, OopClosureType* closure,  \
+                                     int start, int end);
 
-  // Parallel Scavenge and Parallel Old
-  PARALLEL_GC_DECLS
+  ALL_OOP_OOP_ITERATE_CLOSURES_1(TypeArrayKlass_OOP_OOP_ITERATE_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_2(TypeArrayKlass_OOP_OOP_ITERATE_DECL)
+
+#if INCLUDE_ALL_GCS
+#define TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL(OopClosureType, nv_suffix)  \
+  int  oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure);
+
+  ALL_OOP_OOP_ITERATE_CLOSURES_1(TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+  ALL_OOP_OOP_ITERATE_CLOSURES_2(TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DECL)
+#endif // INCLUDE_ALL_GCS
+
 
  protected:
   // Find n'th dimensional array
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/oops/typeArrayKlass.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,73 @@
+/*
+ * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#ifndef SHARE_VM_OOPS_TYPEARRAYKLASS_INLINE_HPP
+#define SHARE_VM_OOPS_TYPEARRAYKLASS_INLINE_HPP
+
+#include "oops/oop.inline.hpp"
+#include "oops/typeArrayKlass.hpp"
+#include "oops/typeArrayOop.hpp"
+
+class ExtendedOopClosure;
+
+inline int TypeArrayKlass::oop_oop_iterate_impl(oop obj, ExtendedOopClosure* closure) {
+  assert(obj->is_typeArray(),"must be a type array");
+  typeArrayOop t = typeArrayOop(obj);
+  // Performance tweak: We skip iterating over the klass pointer since we
+  // know that Universe::TypeArrayKlass never moves.
+  return t->object_size();
+}
+
+#define TypeArrayKlass_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+                                                                        \
+int TypeArrayKlass::                                                    \
+oop_oop_iterate##nv_suffix(oop obj, OopClosureType* closure) {          \
+  return oop_oop_iterate_impl(obj, closure);                            \
+}
+
+#if INCLUDE_ALL_GCS
+#define TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)  \
+                                                                                  \
+int TypeArrayKlass::                                                              \
+oop_oop_iterate_backwards##nv_suffix(oop obj, OopClosureType* closure) {          \
+  return oop_oop_iterate_impl(obj, closure);                                      \
+}
+#else
+#define TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+#endif
+
+
+#define TypeArrayKlass_OOP_OOP_ITERATE_DEFN_m(OopClosureType, nv_suffix)          \
+                                                                                  \
+int TypeArrayKlass::                                                              \
+oop_oop_iterate##nv_suffix##_m(oop obj, OopClosureType* closure, MemRegion mr) {  \
+  return oop_oop_iterate_impl(obj, closure);                                      \
+}
+
+#define ALL_TYPE_ARRAY_KLASS_OOP_OOP_ITERATE_DEFN(OopClosureType, nv_suffix)  \
+  TypeArrayKlass_OOP_OOP_ITERATE_DEFN(          OopClosureType, nv_suffix)    \
+  TypeArrayKlass_OOP_OOP_ITERATE_DEFN_m(        OopClosureType, nv_suffix)    \
+  TypeArrayKlass_OOP_OOP_ITERATE_BACKWARDS_DEFN(OopClosureType, nv_suffix)
+
+#endif // SHARE_VM_OOPS_TYPEARRAYKLASS_INLINE_HPP
--- a/hotspot/src/share/vm/opto/c2_globals.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/c2_globals.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -324,6 +324,9 @@
   develop(bool, SuperWordRTDepCheck, false,                                 \
           "Enable runtime dependency checks.")                              \
                                                                             \
+  product(bool, SuperWordReductions, true,                                  \
+          "Enable reductions support in superword.")                        \
+                                                                            \
   notproduct(bool, TraceSuperWord, false,                                   \
           "Trace superword transforms")                                     \
                                                                             \
--- a/hotspot/src/share/vm/opto/callGenerator.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/callGenerator.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -876,7 +876,8 @@
           // Parse::do_call())
           target = C->optimize_virtual_call(caller, jvms->bci(), klass, klass,
                                             target, receiver_type, is_virtual,
-                                            call_does_dispatch, vtable_index);  // out-parameters
+                                            call_does_dispatch, vtable_index, // out-parameters
+                                            /*check_access=*/false);
           // We lack profiling at this call but type speculation may
           // provide us with a type
           speculative_receiver_type = (receiver_type != NULL) ? receiver_type->speculative_type() : NULL;
--- a/hotspot/src/share/vm/opto/castnode.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/castnode.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -216,16 +216,13 @@
   const Type *result = _type;
   if( in_type != NULL && my_type != NULL ) {
     TypePtr::PTR   in_ptr    = in_type->ptr();
-    if( in_ptr == TypePtr::Null ) {
+    if (in_ptr == TypePtr::Null) {
       result = in_type;
-    } else if( in_ptr == TypePtr::Constant ) {
-      // Casting a constant oop to an interface?
-      // (i.e., a String to a Comparable?)
-      // Then return the interface.
+    } else if (in_ptr == TypePtr::Constant) {
       const TypeOopPtr *jptr = my_type->isa_oopptr();
-      assert( jptr, "" );
-      result =  (jptr->klass()->is_interface() || !in_type->higher_equal(_type))
-      ? my_type->cast_to_ptr_type( TypePtr::NotNull )
+      assert(jptr, "");
+      result = !in_type->higher_equal(_type)
+      ? my_type->cast_to_ptr_type(TypePtr::NotNull)
       : in_type;
     } else {
       result =  my_type->cast_to_ptr_type( my_type->join_ptr(in_ptr) );
--- a/hotspot/src/share/vm/opto/cfgnode.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/cfgnode.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -525,13 +525,16 @@
       // Cut the backedge input and remove phis since no data paths left.
       // We don't cut outputs to other nodes here since we need to put them
       // on the worklist.
+      PhaseIterGVN *igvn = phase->is_IterGVN();
+      if (in(1)->outcnt() == 1) {
+        igvn->_worklist.push(in(1));
+      }
       del_req(1);
       cnt = 0;
       assert( req() == 1, "no more inputs expected" );
       uint max = outcnt();
       bool progress = true;
       Node *top = phase->C->top();
-      PhaseIterGVN *igvn = phase->is_IterGVN();
       DUIterator j;
       while(progress) {
         progress = false;
--- a/hotspot/src/share/vm/opto/cfgnode.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/cfgnode.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -263,6 +263,30 @@
   // Size is bigger to hold the probability field.  However, _prob does not
   // change the semantics so it does not appear in the hash & cmp functions.
   virtual uint size_of() const { return sizeof(*this); }
+
+private:
+  ProjNode* range_check_trap_proj(int& flip, Node*& l, Node*& r);
+  ProjNode* range_check_trap_proj() {
+    int flip_test = 0;
+    Node* l = NULL;
+    Node* r = NULL;
+    return range_check_trap_proj(flip_test, l, r);
+  }
+
+  // Helper methods for fold_compares
+  bool cmpi_folds(PhaseIterGVN* igvn);
+  bool is_ctrl_folds(Node* ctrl, PhaseIterGVN* igvn);
+  bool has_shared_region(ProjNode* proj, ProjNode*& success, ProjNode*& fail);
+  bool has_only_uncommon_traps(ProjNode* proj, ProjNode*& success, ProjNode*& fail, PhaseIterGVN* igvn);
+  static void merge_uncommon_traps(ProjNode* proj, ProjNode* success, ProjNode* fail, PhaseIterGVN* igvn);
+  static void improve_address_types(Node* l, Node* r, ProjNode* fail, PhaseIterGVN* igvn);
+  bool is_cmp_with_loadrange(ProjNode* proj);
+  bool is_null_check(ProjNode* proj, PhaseIterGVN* igvn);
+  bool is_side_effect_free_test(ProjNode* proj, PhaseIterGVN* igvn);
+  void reroute_side_effect_free_unc(ProjNode* proj, ProjNode* dom_proj, PhaseIterGVN* igvn);
+  ProjNode* uncommon_trap_proj(CallStaticJavaNode*& call) const;
+  bool fold_compares_helper(ProjNode* proj, ProjNode* success, ProjNode* fail, PhaseIterGVN* igvn);
+
 public:
 
   // Degrees of branch prediction probability by order of magnitude:
@@ -348,7 +372,7 @@
   virtual const RegMask &out_RegMask() const;
   void dominated_by(Node* prev_dom, PhaseIterGVN* igvn);
   int is_range_check(Node* &range, Node* &index, jint &offset);
-  Node* fold_compares(PhaseGVN* phase);
+  Node* fold_compares(PhaseIterGVN* phase);
   static Node* up_one_dom(Node* curr, bool linear_only = false);
 
   // Takes the type of val and filters it through the test represented
--- a/hotspot/src/share/vm/opto/classes.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/classes.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -266,9 +266,13 @@
 macro(AddVB)
 macro(AddVS)
 macro(AddVI)
+macro(AddReductionVI)
 macro(AddVL)
+macro(AddReductionVL)
 macro(AddVF)
+macro(AddReductionVF)
 macro(AddVD)
+macro(AddReductionVD)
 macro(SubVB)
 macro(SubVS)
 macro(SubVI)
@@ -277,8 +281,11 @@
 macro(SubVD)
 macro(MulVS)
 macro(MulVI)
+macro(MulReductionVI)
 macro(MulVF)
+macro(MulReductionVF)
 macro(MulVD)
+macro(MulReductionVD)
 macro(DivVF)
 macro(DivVD)
 macro(LShiftCntV)
--- a/hotspot/src/share/vm/opto/compile.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/compile.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -3049,6 +3049,15 @@
   case Op_StoreVector:
     break;
 
+  case Op_AddReductionVI:
+  case Op_AddReductionVL:
+  case Op_AddReductionVF:
+  case Op_AddReductionVD:
+  case Op_MulReductionVI:
+  case Op_MulReductionVF:
+  case Op_MulReductionVD:
+    break;
+
   case Op_PackB:
   case Op_PackS:
   case Op_PackI:
--- a/hotspot/src/share/vm/opto/compile.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/compile.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -879,9 +879,11 @@
   ciMethod* optimize_virtual_call(ciMethod* caller, int bci, ciInstanceKlass* klass,
                                   ciKlass* holder, ciMethod* callee,
                                   const TypeOopPtr* receiver_type, bool is_virtual,
-                                  bool &call_does_dispatch, int &vtable_index);
+                                  bool &call_does_dispatch, int &vtable_index,
+                                  bool check_access = true);
   ciMethod* optimize_inlining(ciMethod* caller, int bci, ciInstanceKlass* klass,
-                              ciMethod* callee, const TypeOopPtr* receiver_type);
+                              ciMethod* callee, const TypeOopPtr* receiver_type,
+                              bool check_access = true);
 
   // Report if there were too many traps at a current method and bci.
   // Report if a trap was recorded, and/or PerMethodTrapLimit was exceeded.
--- a/hotspot/src/share/vm/opto/doCall.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/doCall.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -959,13 +959,15 @@
 ciMethod* Compile::optimize_virtual_call(ciMethod* caller, int bci, ciInstanceKlass* klass,
                                          ciKlass* holder, ciMethod* callee,
                                          const TypeOopPtr* receiver_type, bool is_virtual,
-                                         bool& call_does_dispatch, int& vtable_index) {
+                                         bool& call_does_dispatch, int& vtable_index,
+                                         bool check_access) {
   // Set default values for out-parameters.
   call_does_dispatch = true;
   vtable_index       = Method::invalid_vtable_index;
 
   // Choose call strategy.
-  ciMethod* optimized_virtual_method = optimize_inlining(caller, bci, klass, callee, receiver_type);
+  ciMethod* optimized_virtual_method = optimize_inlining(caller, bci, klass, callee,
+                                                         receiver_type, check_access);
 
   // Have the call been sufficiently improved such that it is no longer a virtual?
   if (optimized_virtual_method != NULL) {
@@ -980,7 +982,8 @@
 
 // Identify possible target method and inlining style
 ciMethod* Compile::optimize_inlining(ciMethod* caller, int bci, ciInstanceKlass* klass,
-                                     ciMethod* callee, const TypeOopPtr* receiver_type) {
+                                     ciMethod* callee, const TypeOopPtr* receiver_type,
+                                     bool check_access) {
   // only use for virtual or interface calls
 
   // If it is obviously final, do not bother to call find_monomorphic_target,
@@ -1020,7 +1023,7 @@
   }
 
   ciInstanceKlass*   calling_klass = caller->holder();
-  ciMethod* cha_monomorphic_target = callee->find_monomorphic_target(calling_klass, klass, actual_receiver);
+  ciMethod* cha_monomorphic_target = callee->find_monomorphic_target(calling_klass, klass, actual_receiver, check_access);
   if (cha_monomorphic_target != NULL) {
     assert(!cha_monomorphic_target->is_abstract(), "");
     // Look at the method-receiver type.  Does it add "too much information"?
--- a/hotspot/src/share/vm/opto/graphKit.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/graphKit.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -2530,6 +2530,11 @@
 // prior to coming here.
 Node* Phase::gen_subtype_check(Node* subklass, Node* superklass, Node** ctrl, MergeMemNode* mem, PhaseGVN* gvn) {
   Compile* C = gvn->C;
+
+  if ((*ctrl)->is_top()) {
+    return C->top();
+  }
+
   // Fast check for identical types, perhaps identical constants.
   // The types can even be identical non-constants, in cases
   // involving Array.newInstance, Object.clone, etc.
@@ -2792,18 +2797,19 @@
  */
 Node* GraphKit::maybe_cast_profiled_obj(Node* obj,
                                         ciKlass* type,
-                                        bool not_null,
-                                        SafePointNode* sfpt) {
+                                        bool not_null) {
+  if (stopped()) {
+    return obj;
+  }
+
   // type == NULL if profiling tells us this object is always null
   if (type != NULL) {
     Deoptimization::DeoptReason class_reason = Deoptimization::Reason_speculate_class_check;
     Deoptimization::DeoptReason null_reason = Deoptimization::Reason_speculate_null_check;
-    ciMethod* trap_method = (sfpt == NULL) ? method() : sfpt->jvms()->method();
-    int trap_bci = (sfpt == NULL) ? bci() : sfpt->jvms()->bci();
 
     if (!too_many_traps(null_reason) && !too_many_recompiles(null_reason) &&
-        !C->too_many_traps(trap_method, trap_bci, class_reason) &&
-        !C->too_many_recompiles(trap_method, trap_bci, class_reason)) {
+        !too_many_traps(class_reason) &&
+        !too_many_recompiles(class_reason)) {
       Node* not_null_obj = NULL;
       // not_null is true if we know the object is not null and
       // there's no need for a null check
@@ -2819,12 +2825,7 @@
       ciKlass* exact_kls = type;
       Node* slow_ctl  = type_check_receiver(exact_obj, exact_kls, 1.0,
                                             &exact_obj);
-      if (sfpt != NULL) {
-        GraphKit kit(sfpt->jvms());
-        PreserveJVMState pjvms(&kit);
-        kit.set_control(slow_ctl);
-        kit.uncommon_trap_exact(class_reason, Deoptimization::Action_maybe_recompile);
-      } else {
+      {
         PreserveJVMState pjvms(this);
         set_control(slow_ctl);
         uncommon_trap_exact(class_reason, Deoptimization::Action_maybe_recompile);
--- a/hotspot/src/share/vm/opto/graphKit.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/graphKit.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -409,8 +409,7 @@
   // Cast obj to type and emit guard unless we had too many traps here already
   Node* maybe_cast_profiled_obj(Node* obj,
                                 ciKlass* type,
-                                bool not_null = false,
-                                SafePointNode* sfpt = NULL);
+                                bool not_null = false);
 
   // Cast obj to not-null on this path
   Node* cast_not_null(Node* obj, bool do_replace_in_map = true);
--- a/hotspot/src/share/vm/opto/ifnode.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/ifnode.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -25,11 +25,13 @@
 #include "precompiled.hpp"
 #include "memory/allocation.inline.hpp"
 #include "opto/addnode.hpp"
+#include "opto/castnode.hpp"
 #include "opto/cfgnode.hpp"
 #include "opto/connode.hpp"
 #include "opto/loopnode.hpp"
 #include "opto/phaseX.hpp"
 #include "opto/runtime.hpp"
+#include "opto/rootnode.hpp"
 #include "opto/subnode.hpp"
 
 // Portions of code courtesy of Clifford Click
@@ -449,62 +451,59 @@
   return new ConINode(TypeInt::ZERO);
 }
 
-//------------------------------is_range_check---------------------------------
-// Return 0 if not a range check.  Return 1 if a range check and set index and
-// offset.  Return 2 if we had to negate the test.  Index is NULL if the check
-// is versus a constant.
-int IfNode::is_range_check(Node* &range, Node* &index, jint &offset) {
+// if this IfNode follows a range check pattern return the projection
+// for the failed path
+ProjNode* IfNode::range_check_trap_proj(int& flip_test, Node*& l, Node*& r) {
   Node* b = in(1);
-  if (b == NULL || !b->is_Bool())  return 0;
+  if (b == NULL || !b->is_Bool())  return NULL;
   BoolNode* bn = b->as_Bool();
   Node* cmp = bn->in(1);
-  if (cmp == NULL)  return 0;
-  if (cmp->Opcode() != Op_CmpU)  return 0;
+  if (cmp == NULL)  return NULL;
+  if (cmp->Opcode() != Op_CmpU)  return NULL;
 
-  Node* l = cmp->in(1);
-  Node* r = cmp->in(2);
-  int flip_test = 1;
+  l = cmp->in(1);
+  r = cmp->in(2);
+  flip_test = 1;
   if (bn->_test._test == BoolTest::le) {
     l = cmp->in(2);
     r = cmp->in(1);
     flip_test = 2;
   } else if (bn->_test._test != BoolTest::lt) {
-    return 0;
+    return NULL;
   }
-  if (l->is_top())  return 0;   // Top input means dead test
-  if (r->Opcode() != Op_LoadRange)  return 0;
+  if (l->is_top())  return NULL;   // Top input means dead test
+  if (r->Opcode() != Op_LoadRange)  return NULL;
 
   // We have recognized one of these forms:
   //  Flip 1:  If (Bool[<] CmpU(l, LoadRange)) ...
   //  Flip 2:  If (Bool[<=] CmpU(LoadRange, l)) ...
 
+  ProjNode* iftrap = proj_out(flip_test == 2 ? true : false);
+  return iftrap;
+}
+
+
+//------------------------------is_range_check---------------------------------
+// Return 0 if not a range check.  Return 1 if a range check and set index and
+// offset.  Return 2 if we had to negate the test.  Index is NULL if the check
+// is versus a constant.
+int IfNode::is_range_check(Node* &range, Node* &index, jint &offset) {
+  int flip_test = 0;
+  Node* l = NULL;
+  Node* r = NULL;
+  ProjNode* iftrap = range_check_trap_proj(flip_test, l, r);
+
+  if (iftrap == NULL) {
+    return 0;
+  }
+
   // Make sure it's a real range check by requiring an uncommon trap
   // along the OOB path.  Otherwise, it's possible that the user wrote
   // something which optimized to look like a range check but behaves
   // in some other way.
-  Node* iftrap = proj_out(flip_test == 2 ? true : false);
-  bool found_trap = false;
-  if (iftrap != NULL) {
-    Node* u = iftrap->unique_ctrl_out();
-    if (u != NULL) {
-      // It could be a merge point (Region) for uncommon trap.
-      if (u->is_Region()) {
-        Node* c = u->unique_ctrl_out();
-        if (c != NULL) {
-          iftrap = u;
-          u = c;
-        }
-      }
-      if (u->in(0) == iftrap && u->is_CallStaticJava()) {
-        int req = u->as_CallStaticJava()->uncommon_trap_request();
-        if (Deoptimization::trap_request_reason(req) ==
-            Deoptimization::Reason_range_check) {
-          found_trap = true;
-        }
-      }
-    }
+  if (iftrap->is_uncommon_trap_proj(Deoptimization::Reason_range_check) == NULL) {
+    return 0;
   }
-  if (!found_trap)  return 0;   // sorry, no cigar
 
   // Look for index+offset form
   Node* ind = l;
@@ -664,11 +663,12 @@
 //------------------------------fold_compares----------------------------
 // See if a pair of CmpIs can be converted into a CmpU.  In some cases
 // the direction of this if is determined by the preceding if so it
-// can be eliminate entirely.  Given an if testing (CmpI n c) check
-// for an immediately control dependent if that is testing (CmpI n c2)
-// and has one projection leading to this if and the other projection
-// leading to a region that merges one of this ifs control
-// projections.
+// can be eliminate entirely.
+//
+// Given an if testing (CmpI n v) check for an immediately control
+// dependent if that is testing (CmpI n v2) and has one projection
+// leading to this if and the other projection leading to a region
+// that merges one of this ifs control projections.
 //
 //                   If
 //                  / |
@@ -680,79 +680,458 @@
 //            /    \  |
 //           /    Region
 //
-Node* IfNode::fold_compares(PhaseGVN* phase) {
-  if (Opcode() != Op_If) return NULL;
+// Or given an if testing (CmpI n v) check for a dominating if that is
+// testing (CmpI n v2), both having one projection leading to an
+// uncommon trap. Allow Another independent guard in between to cover
+// an explicit range check:
+// if (index < 0 || index >= array.length) {
+// which may need a null check to guard the LoadRange
+//
+//                   If
+//                  / \
+//                 /   \
+//                /     \
+//              If      unc
+//              /\
+//             /  \
+//            /    \
+//           /      unc
+//
+
+// Is the comparison for this If suitable for folding?
+bool IfNode::cmpi_folds(PhaseIterGVN* igvn) {
+  return in(1) != NULL &&
+    in(1)->is_Bool() &&
+    in(1)->in(1) != NULL &&
+    in(1)->in(1)->Opcode() == Op_CmpI &&
+    in(1)->in(1)->in(2) != NULL &&
+    in(1)->in(1)->in(2) != igvn->C->top() &&
+    (in(1)->as_Bool()->_test.is_less() ||
+     in(1)->as_Bool()->_test.is_greater());
+}
+
+// Is a dominating control suitable for folding with this if?
+bool IfNode::is_ctrl_folds(Node* ctrl, PhaseIterGVN* igvn) {
+  return ctrl != NULL &&
+    ctrl->is_Proj() &&
+    ctrl->in(0) != NULL &&
+    ctrl->in(0)->is_If() &&
+    ctrl->in(0)->outcnt() == 2 &&
+    ctrl->in(0)->as_If()->cmpi_folds(igvn) &&
+    // Must compare same value
+    ctrl->in(0)->in(1)->in(1)->in(1) != NULL &&
+    ctrl->in(0)->in(1)->in(1)->in(1) == in(1)->in(1)->in(1);
+}
+
+// Do this If and the dominating If share a region?
+bool IfNode::has_shared_region(ProjNode* proj, ProjNode*& success, ProjNode*& fail) {
+  ProjNode* otherproj = proj->other_if_proj();
+  Node* otherproj_ctrl_use = otherproj->unique_ctrl_out();
+  RegionNode* region = (otherproj_ctrl_use != NULL && otherproj_ctrl_use->is_Region()) ? otherproj_ctrl_use->as_Region() : NULL;
+  success = NULL;
+  fail = NULL;
 
+  if (otherproj->outcnt() == 1 && region != NULL && !region->has_phi()) {
+    for (int i = 0; i < 2; i++) {
+      ProjNode* proj = proj_out(i);
+      if (success == NULL && proj->outcnt() == 1 && proj->unique_out() == region) {
+        success = proj;
+      } else if (fail == NULL) {
+        fail = proj;
+      } else {
+        success = fail = NULL;
+      }
+    }
+  }
+  return success != NULL && fail != NULL;
+}
+
+// Return projection that leads to an uncommon trap if any
+ProjNode* IfNode::uncommon_trap_proj(CallStaticJavaNode*& call) const {
+  for (int i = 0; i < 2; i++) {
+    call = proj_out(i)->is_uncommon_trap_proj(Deoptimization::Reason_none);
+    if (call != NULL) {
+      return proj_out(i);
+    }
+  }
+  return NULL;
+}
+
+// Do this If and the dominating If both branch out to an uncommon trap
+bool IfNode::has_only_uncommon_traps(ProjNode* proj, ProjNode*& success, ProjNode*& fail, PhaseIterGVN* igvn) {
+  ProjNode* otherproj = proj->other_if_proj();
+  CallStaticJavaNode* dom_unc = otherproj->is_uncommon_trap_proj(Deoptimization::Reason_none);
+
+  if (otherproj->outcnt() == 1 && dom_unc != NULL) {
+    CallStaticJavaNode* unc = NULL;
+    ProjNode* unc_proj = uncommon_trap_proj(unc);
+    if (unc_proj != NULL && unc_proj->outcnt() == 1) {
+      if (dom_unc == unc) {
+        // Allow the uncommon trap to be shared through a region
+        RegionNode* r = unc->in(0)->as_Region();
+        if (r->outcnt() != 2 || r->req() != 3 || r->find_edge(otherproj) == -1 || r->find_edge(unc_proj) == -1) {
+          return false;
+        }
+        assert(r->has_phi() == NULL, "simple region shouldn't have a phi");
+      } else if (dom_unc->in(0) != otherproj || unc->in(0) != unc_proj) {
+        return false;
+      }
+      // See merge_uncommon_traps: the reason of the uncommon trap
+      // will be changed and the state of the dominating If will be
+      // used. Checked that we didn't apply this transformation in a
+      // previous compilation and it didn't cause too many traps
+      if (!igvn->C->too_many_traps(dom_unc->jvms()->method(), dom_unc->jvms()->bci(), Deoptimization::Reason_unstable_fused_if) &&
+          !igvn->C->too_many_traps(dom_unc->jvms()->method(), dom_unc->jvms()->bci(), Deoptimization::Reason_range_check)) {
+        success = unc_proj;
+        fail = unc_proj->other_if_proj();
+        return true;
+      }
+    }
+  }
+  return false;
+}
+
+// Check that the 2 CmpI can be folded into as single CmpU and proceed with the folding
+bool IfNode::fold_compares_helper(ProjNode* proj, ProjNode* success, ProjNode* fail, PhaseIterGVN* igvn) {
   Node* this_cmp = in(1)->in(1);
-  if (this_cmp != NULL && this_cmp->Opcode() == Op_CmpI &&
-      this_cmp->in(2)->is_Con() && this_cmp->in(2) != phase->C->top()) {
-    Node* ctrl = in(0);
-    BoolNode* this_bool = in(1)->as_Bool();
-    Node* n = this_cmp->in(1);
-    int hi = this_cmp->in(2)->get_int();
-    if (ctrl != NULL && ctrl->is_Proj() && ctrl->outcnt() == 1 &&
-        ctrl->in(0)->is_If() &&
-        ctrl->in(0)->outcnt() == 2 &&
-        ctrl->in(0)->in(1)->is_Bool() &&
-        ctrl->in(0)->in(1)->in(1)->Opcode() == Op_CmpI &&
-        ctrl->in(0)->in(1)->in(1)->in(2)->is_Con() &&
-        ctrl->in(0)->in(1)->in(1)->in(2) != phase->C->top() &&
-        ctrl->in(0)->in(1)->in(1)->in(1) == n) {
-      IfNode* dom_iff = ctrl->in(0)->as_If();
-      Node* otherproj = dom_iff->proj_out(!ctrl->as_Proj()->_con);
-      if (otherproj->outcnt() == 1 && otherproj->unique_out()->is_Region() &&
-          this_bool->_test._test != BoolTest::ne && this_bool->_test._test != BoolTest::eq) {
-        // Identify which proj goes to the region and which continues on
-        RegionNode* region = otherproj->unique_out()->as_Region();
-        Node* success = NULL;
-        Node* fail = NULL;
-        for (int i = 0; i < 2; i++) {
-          Node* proj = proj_out(i);
-          if (success == NULL && proj->outcnt() == 1 && proj->unique_out() == region) {
-            success = proj;
-          } else if (fail == NULL) {
-            fail = proj;
-          } else {
-            success = fail = NULL;
+  BoolNode* this_bool = in(1)->as_Bool();
+  IfNode* dom_iff = proj->in(0)->as_If();
+  BoolNode* dom_bool = dom_iff->in(1)->as_Bool();
+  Node* lo = dom_iff->in(1)->in(1)->in(2);
+  Node* hi = this_cmp->in(2);
+  Node* n = this_cmp->in(1);
+  ProjNode* otherproj = proj->other_if_proj();
+
+  const TypeInt* lo_type = IfNode::filtered_int_type(igvn, n, otherproj);
+  const TypeInt* hi_type = IfNode::filtered_int_type(igvn, n, success);
+
+  BoolTest::mask lo_test = dom_bool->_test._test;
+  BoolTest::mask hi_test = this_bool->_test._test;
+  BoolTest::mask cond = hi_test;
+
+  // Figure out which of the two tests sets the upper bound and which
+  // sets the lower bound if any.
+  if (hi_type->_lo > lo_type->_hi && hi_type->_hi == max_jint && lo_type->_lo == min_jint) {
+
+    assert((dom_bool->_test.is_less() && !proj->_con) ||
+           (dom_bool->_test.is_greater() && proj->_con), "incorrect test");
+    // this test was canonicalized
+    assert(this_bool->_test.is_less() && fail->_con, "incorrect test");
+
+    if (lo_test == BoolTest::gt || lo_test == BoolTest::le) {
+      lo = igvn->transform(new AddINode(lo, igvn->intcon(1)));
+    }
+  } else if (lo_type->_lo > hi_type->_hi && lo_type->_hi == max_jint && hi_type->_lo == min_jint) {
+    swap(lo, hi);
+    swap(lo_type, hi_type);
+    swap(lo_test, hi_test);
+
+    assert((this_bool->_test.is_less() && proj->_con) ||
+           (this_bool->_test.is_greater() && !proj->_con), "incorrect test");
+    // this test was canonicalized
+    assert(dom_bool->_test.is_less() && !fail->_con, "incorrect test");
+
+    cond = (hi_test == BoolTest::le || hi_test == BoolTest::gt) ? BoolTest::gt : BoolTest::ge;
+
+    if (lo_test == BoolTest::le) {
+      lo = igvn->transform(new AddINode(lo, igvn->intcon(1)));
+    }
+
+  } else {
+    const TypeInt* failtype  = filtered_int_type(igvn, n, proj);
+    if (failtype != NULL) {
+      const TypeInt* type2 = filtered_int_type(igvn, n, fail);
+      if (type2 != NULL) {
+        failtype = failtype->join(type2)->is_int();
+        if (failtype->_lo > failtype->_hi) {
+          // previous if determines the result of this if so
+          // replace Bool with constant
+          igvn->hash_delete(this);
+          set_req(1, igvn->intcon(success->_con));
+          return true;
+        }
+      }
+    }
+
+    lo = NULL;
+    hi = NULL;
+  }
+
+  if (lo && hi) {
+    // Merge the two compares into a single unsigned compare by building (CmpU (n - lo) (hi - lo))
+    Node* adjusted_val = igvn->transform(new SubINode(n,  lo));
+    Node* adjusted_lim = igvn->transform(new SubINode(hi, lo));
+    Node* newcmp = igvn->transform(new CmpUNode(adjusted_val, adjusted_lim));
+    Node* newbool = igvn->transform(new BoolNode(newcmp, cond));
+
+    igvn->is_IterGVN()->replace_input_of(dom_iff, 1, igvn->intcon(proj->_con));
+    igvn->hash_delete(this);
+    set_req(1, newbool);
+
+    return true;
+  }
+  return false;
+}
+
+// Merge the branches that trap for this If and the dominating If into
+// a single region that branches to the uncommon trap for the
+// dominating If
+void IfNode::merge_uncommon_traps(ProjNode* proj, ProjNode* success, ProjNode* fail, PhaseIterGVN* igvn) {
+  ProjNode* otherproj = proj->other_if_proj();
+
+  CallStaticJavaNode* unc = success->is_uncommon_trap_proj(Deoptimization::Reason_none);
+  CallStaticJavaNode* dom_unc = otherproj->is_uncommon_trap_proj(Deoptimization::Reason_none);
+
+  if (unc != dom_unc) {
+    Node* r = new RegionNode(3);
+
+    r->set_req(1, otherproj);
+    r->set_req(2, success);
+    r = igvn->transform(r);
+    assert(r->is_Region(), "can't go away");
+
+    // Make both If trap at the state of the first If: once the CmpI
+    // nodes are merged, if we trap we don't know which of the CmpI
+    // nodes would have caused the trap so we have to restart
+    // execution at the first one
+    igvn->replace_input_of(dom_unc, 0, r);
+    igvn->replace_input_of(unc, 0, igvn->C->top());
+  }
+  int trap_request = dom_unc->uncommon_trap_request();
+  Deoptimization::DeoptReason reason = Deoptimization::trap_request_reason(trap_request);
+  Deoptimization::DeoptAction action = Deoptimization::trap_request_action(trap_request);
+
+  int flip_test = 0;
+  Node* l = NULL;
+  Node* r = NULL;
+
+  if (success->in(0)->as_If()->range_check_trap_proj(flip_test, l, r) != NULL) {
+    // If this looks like a range check, change the trap to
+    // Reason_range_check so the compiler recognizes it as a range
+    // check and applies the corresponding optimizations
+    trap_request = Deoptimization::make_trap_request(Deoptimization::Reason_range_check, action);
+
+    improve_address_types(l, r, fail, igvn);
+  } else if (unc != dom_unc) {
+    // If we trap we won't know what CmpI would have caused the trap
+    // so use a special trap reason to mark this pair of CmpI nodes as
+    // bad candidate for folding. On recompilation we won't fold them
+    // and we may trap again but this time we'll know what branch
+    // traps
+    trap_request = Deoptimization::make_trap_request(Deoptimization::Reason_unstable_fused_if, action);
+  }
+  igvn->replace_input_of(dom_unc, TypeFunc::Parms, igvn->intcon(trap_request));
+}
+
+// If we are turning 2 CmpI nodes into a CmpU that follows the pattern
+// of a rangecheck on index i, on 64 bit the compares may be followed
+// by memory accesses using i as index. In that case, the CmpU tells
+// us something about the values taken by i that can help the compiler
+// (see Compile::conv_I2X_index())
+void IfNode::improve_address_types(Node* l, Node* r, ProjNode* fail, PhaseIterGVN* igvn) {
+#ifdef _LP64
+  ResourceMark rm;
+  Node_Stack stack(2);
+
+  assert(r->Opcode() == Op_LoadRange, "unexpected range check");
+  const TypeInt* array_size = igvn->type(r)->is_int();
+
+  stack.push(l, 0);
+
+  while(stack.size() > 0) {
+    Node* n = stack.node();
+    uint start = stack.index();
+
+    uint i = start;
+    for (; i < n->outcnt(); i++) {
+      Node* use = n->raw_out(i);
+      if (stack.size() == 1) {
+        if (use->Opcode() == Op_ConvI2L) {
+          const TypeLong* bounds = use->as_Type()->type()->is_long();
+          if (bounds->_lo <= array_size->_lo && bounds->_hi >= array_size->_hi &&
+              (bounds->_lo != array_size->_lo || bounds->_hi != array_size->_hi)) {
+            stack.set_index(i+1);
+            stack.push(use, 0);
+            break;
           }
         }
-        if (success != NULL && fail != NULL && !region->has_phi()) {
-          int lo = dom_iff->in(1)->in(1)->in(2)->get_int();
-          BoolNode* dom_bool = dom_iff->in(1)->as_Bool();
-          Node* dom_cmp =  dom_bool->in(1);
-          const TypeInt* failtype  = filtered_int_type(phase, n, ctrl);
-          if (failtype != NULL) {
-            const TypeInt* type2 = filtered_int_type(phase, n, fail);
-            if (type2 != NULL) {
-              failtype = failtype->join(type2)->is_int();
-            } else {
-              failtype = NULL;
-            }
-          }
+      } else if (use->is_Mem()) {
+        Node* ctrl = use->in(0);
+        for (int i = 0; i < 10 && ctrl != NULL && ctrl != fail; i++) {
+          ctrl = up_one_dom(ctrl);
+        }
+        if (ctrl == fail) {
+          Node* init_n = stack.node_at(1);
+          assert(init_n->Opcode() == Op_ConvI2L, "unexpected first node");
+          Node* new_n = igvn->C->conv_I2X_index(igvn, l, array_size);
 
-          if (failtype != NULL &&
-              dom_bool->_test._test != BoolTest::ne && dom_bool->_test._test != BoolTest::eq) {
-            int bound = failtype->_hi - failtype->_lo + 1;
-            if (failtype->_hi != max_jint && failtype->_lo != min_jint && bound > 1) {
-              // Merge the two compares into a single unsigned compare by building  (CmpU (n - lo) hi)
-              BoolTest::mask cond = fail->as_Proj()->_con ? BoolTest::lt : BoolTest::ge;
-              Node* adjusted = phase->transform(new SubINode(n, phase->intcon(failtype->_lo)));
-              Node* newcmp = phase->transform(new CmpUNode(adjusted, phase->intcon(bound)));
-              Node* newbool = phase->transform(new BoolNode(newcmp, cond));
-              phase->is_IterGVN()->replace_input_of(dom_iff, 1, phase->intcon(ctrl->as_Proj()->_con));
-              phase->hash_delete(this);
-              set_req(1, newbool);
-              return this;
-            }
-            if (failtype->_lo > failtype->_hi) {
-              // previous if determines the result of this if so
-              // replace Bool with constant
-              phase->hash_delete(this);
-              set_req(1, phase->intcon(success->as_Proj()->_con));
-              return this;
-            }
+          for (uint j = 2; j < stack.size(); j++) {
+            Node* n = stack.node_at(j);
+            Node* clone = n->clone();
+            int rep = clone->replace_edge(init_n, new_n);
+            assert(rep > 0, "can't find expected node?");
+            clone = igvn->transform(clone);
+            init_n = n;
+            new_n = clone;
+          }
+          igvn->hash_delete(use);
+          int rep = use->replace_edge(init_n, new_n);
+          assert(rep > 0, "can't find expected node?");
+          igvn->transform(use);
+          if (init_n->outcnt() == 0) {
+            igvn->_worklist.push(init_n);
           }
         }
+      } else if (use->in(0) == NULL && (igvn->type(use)->isa_long() ||
+                                        igvn->type(use)->isa_ptr())) {
+        stack.set_index(i+1);
+        stack.push(use, 0);
+        break;
+      }
+    }
+    if (i == n->outcnt()) {
+      stack.pop();
+    }
+  }
+#endif
+}
+
+bool IfNode::is_cmp_with_loadrange(ProjNode* proj) {
+  if (in(1) != NULL &&
+      in(1)->in(1) != NULL &&
+      in(1)->in(1)->in(2) != NULL) {
+    Node* other = in(1)->in(1)->in(2);
+    if (other->Opcode() == Op_LoadRange &&
+        ((other->in(0) != NULL && other->in(0) == proj) ||
+         (other->in(0) == NULL &&
+          other->in(2) != NULL &&
+          other->in(2)->is_AddP() &&
+          other->in(2)->in(1) != NULL &&
+          other->in(2)->in(1)->Opcode() == Op_CastPP &&
+          other->in(2)->in(1)->in(0) == proj))) {
+      return true;
+    }
+  }
+  return false;
+}
+
+bool IfNode::is_null_check(ProjNode* proj, PhaseIterGVN* igvn) {
+  Node* other = in(1)->in(1)->in(2);
+  if (other->in(MemNode::Address) != NULL &&
+      proj->in(0)->in(1) != NULL &&
+      proj->in(0)->in(1)->is_Bool() &&
+      proj->in(0)->in(1)->in(1) != NULL &&
+      proj->in(0)->in(1)->in(1)->Opcode() == Op_CmpP &&
+      proj->in(0)->in(1)->in(1)->in(2) != NULL &&
+      proj->in(0)->in(1)->in(1)->in(1) == other->in(MemNode::Address)->in(AddPNode::Address)->uncast() &&
+      igvn->type(proj->in(0)->in(1)->in(1)->in(2)) == TypePtr::NULL_PTR) {
+    return true;
+  }
+  return false;
+}
+
+// Check that the If that is in between the 2 integer comparisons has
+// no side effect
+bool IfNode::is_side_effect_free_test(ProjNode* proj, PhaseIterGVN* igvn) {
+  if (proj != NULL &&
+      proj->is_uncommon_trap_if_pattern(Deoptimization::Reason_none) &&
+      proj->outcnt() <= 2) {
+    if (proj->outcnt() == 1 ||
+        // Allow simple null check from LoadRange
+        (is_cmp_with_loadrange(proj) && is_null_check(proj, igvn))) {
+      CallStaticJavaNode* unc = proj->is_uncommon_trap_if_pattern(Deoptimization::Reason_none);
+      CallStaticJavaNode* dom_unc = proj->in(0)->in(0)->as_Proj()->is_uncommon_trap_if_pattern(Deoptimization::Reason_none);
+
+      // reroute_side_effect_free_unc changes the state of this
+      // uncommon trap to restart execution at the previous
+      // CmpI. Check that this change in a previous compilation didn't
+      // cause too many traps.
+      int trap_request = unc->uncommon_trap_request();
+      Deoptimization::DeoptReason reason = Deoptimization::trap_request_reason(trap_request);
+
+      if (igvn->C->too_many_traps(dom_unc->jvms()->method(), dom_unc->jvms()->bci(), reason)) {
+        return false;
+      }
+
+      return true;
+    }
+  }
+  return false;
+}
+
+// Make the If between the 2 integer comparisons trap at the state of
+// the first If: the last CmpI is the one replaced by a CmpU and the
+// first CmpI is eliminated, so the test between the 2 CmpI nodes
+// won't be guarded by the first CmpI anymore. It can trap in cases
+// where the first CmpI would have prevented it from executing: on a
+// trap, we need to restart execution at the state of the first CmpI
+void IfNode::reroute_side_effect_free_unc(ProjNode* proj, ProjNode* dom_proj, PhaseIterGVN* igvn) {
+  CallStaticJavaNode* dom_unc = dom_proj->is_uncommon_trap_if_pattern(Deoptimization::Reason_none);
+  ProjNode* otherproj = proj->other_if_proj();
+  CallStaticJavaNode* unc = proj->is_uncommon_trap_if_pattern(Deoptimization::Reason_none);
+  Node* call_proj = dom_unc->unique_ctrl_out();
+  Node* halt = call_proj->unique_ctrl_out();
+
+  Node* new_unc = dom_unc->clone();
+  call_proj = call_proj->clone();
+  halt = halt->clone();
+  Node* c = otherproj->clone();
+
+  c = igvn->transform(c);
+  new_unc->set_req(TypeFunc::Parms, unc->in(TypeFunc::Parms));
+  new_unc->set_req(0, c);
+  new_unc = igvn->transform(new_unc);
+  call_proj->set_req(0, new_unc);
+  call_proj = igvn->transform(call_proj);
+  halt->set_req(0, call_proj);
+  halt = igvn->transform(halt);
+
+  igvn->replace_node(otherproj, igvn->C->top());
+  igvn->C->root()->add_req(halt);
+}
+
+Node* IfNode::fold_compares(PhaseIterGVN* igvn) {
+  if (Opcode() != Op_If) return NULL;
+
+  if (cmpi_folds(igvn)) {
+    Node* ctrl = in(0);
+    if (is_ctrl_folds(ctrl, igvn) &&
+        ctrl->outcnt() == 1) {
+      // A integer comparison immediately dominated by another integer
+      // comparison
+      ProjNode* success = NULL;
+      ProjNode* fail = NULL;
+      ProjNode* dom_cmp = ctrl->as_Proj();
+      if (has_shared_region(dom_cmp, success, fail) &&
+          // Next call modifies graph so must be last
+          fold_compares_helper(dom_cmp, success, fail, igvn)) {
+        return this;
+      }
+      if (has_only_uncommon_traps(dom_cmp, success, fail, igvn) &&
+          // Next call modifies graph so must be last
+          fold_compares_helper(dom_cmp, success, fail, igvn)) {
+        merge_uncommon_traps(dom_cmp, success, fail, igvn);
+        return this;
+      }
+      return NULL;
+    } else if (ctrl->in(0) != NULL &&
+               ctrl->in(0)->in(0) != NULL) {
+      ProjNode* success = NULL;
+      ProjNode* fail = NULL;
+      Node* dom = ctrl->in(0)->in(0);
+      ProjNode* dom_cmp = dom->isa_Proj();
+      ProjNode* other_cmp = ctrl->isa_Proj();
+
+      // Check if it's an integer comparison dominated by another
+      // integer comparison with another test in between
+      if (is_ctrl_folds(dom, igvn) &&
+          has_only_uncommon_traps(dom_cmp, success, fail, igvn) &&
+          is_side_effect_free_test(other_cmp, igvn) &&
+          // Next call modifies graph so must be last
+          fold_compares_helper(dom_cmp, success, fail, igvn)) {
+        reroute_side_effect_free_unc(other_cmp, dom_cmp, igvn);
+        merge_uncommon_traps(dom_cmp, success, fail, igvn);
+        return this;
       }
     }
   }
@@ -1029,7 +1408,7 @@
     // Normal equivalent-test check.
     if( !dom ) return NULL;     // Dead loop?
 
-    Node* result = fold_compares(phase);
+    Node* result = fold_compares(igvn);
     if (result != NULL) {
       return result;
     }
@@ -1089,7 +1468,7 @@
   // be skipped. For example, range check predicate has two checks
   // for lower and upper bounds.
   ProjNode* unc_proj = proj_out(1 - prev_dom->as_Proj()->_con)->as_Proj();
-  if (unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_predicate))
+  if (unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_predicate) != NULL)
    prev_dom = idom;
 
   // Now walk the current IfNode's projections.
--- a/hotspot/src/share/vm/opto/library_call.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/library_call.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -262,6 +262,9 @@
   bool inline_arraycopy();
   AllocateArrayNode* tightly_coupled_allocation(Node* ptr,
                                                 RegionNode* slow_region);
+  JVMState* arraycopy_restore_alloc_state(AllocateArrayNode* alloc, int& saved_reexecute_sp);
+  void arraycopy_move_allocation_here(AllocateArrayNode* alloc, Node* dest, JVMState* saved_jvms, int saved_reexecute_sp);
+
   typedef enum { LS_xadd, LS_xchg, LS_cmpxchg } LoadStoreKind;
   bool inline_unsafe_load_store(BasicType type,  LoadStoreKind kind);
   bool inline_unsafe_ordered_store(BasicType type);
@@ -290,6 +293,7 @@
   bool inline_multiplyToLen();
 
   bool inline_profileBoolean();
+  bool inline_isCompileConstant();
 };
 
 
@@ -549,6 +553,17 @@
     if (!Matcher::match_rule_supported(Op_OverflowMulL) || !UseMathExactIntrinsics) return NULL;
     break;
 
+  case vmIntrinsics::_getShortUnaligned:
+  case vmIntrinsics::_getCharUnaligned:
+  case vmIntrinsics::_getIntUnaligned:
+  case vmIntrinsics::_getLongUnaligned:
+  case vmIntrinsics::_putShortUnaligned:
+  case vmIntrinsics::_putCharUnaligned:
+  case vmIntrinsics::_putIntUnaligned:
+  case vmIntrinsics::_putLongUnaligned:
+    if (!UseUnalignedAccesses) return NULL;
+    break;
+
  default:
     assert(id <= vmIntrinsics::LAST_COMPILER_INLINE, "caller responsibility");
     assert(id != vmIntrinsics::_Object_init && id != vmIntrinsics::_invoke, "enum out of order?");
@@ -799,6 +814,16 @@
   case vmIntrinsics::_putFloatVolatile:         return inline_unsafe_access(!is_native_ptr,  is_store, T_FLOAT,    is_volatile);
   case vmIntrinsics::_putDoubleVolatile:        return inline_unsafe_access(!is_native_ptr,  is_store, T_DOUBLE,   is_volatile);
 
+  case vmIntrinsics::_getShortUnaligned:        return inline_unsafe_access(!is_native_ptr, !is_store, T_SHORT,   !is_volatile);
+  case vmIntrinsics::_getCharUnaligned:         return inline_unsafe_access(!is_native_ptr, !is_store, T_CHAR,    !is_volatile);
+  case vmIntrinsics::_getIntUnaligned:          return inline_unsafe_access(!is_native_ptr, !is_store, T_INT,     !is_volatile);
+  case vmIntrinsics::_getLongUnaligned:         return inline_unsafe_access(!is_native_ptr, !is_store, T_LONG,    !is_volatile);
+
+  case vmIntrinsics::_putShortUnaligned:        return inline_unsafe_access(!is_native_ptr,  is_store, T_SHORT,   !is_volatile);
+  case vmIntrinsics::_putCharUnaligned:         return inline_unsafe_access(!is_native_ptr,  is_store, T_CHAR,    !is_volatile);
+  case vmIntrinsics::_putIntUnaligned:          return inline_unsafe_access(!is_native_ptr,  is_store, T_INT,     !is_volatile);
+  case vmIntrinsics::_putLongUnaligned:         return inline_unsafe_access(!is_native_ptr,  is_store, T_LONG,    !is_volatile);
+
   case vmIntrinsics::_compareAndSwapObject:     return inline_unsafe_load_store(T_OBJECT, LS_cmpxchg);
   case vmIntrinsics::_compareAndSwapInt:        return inline_unsafe_load_store(T_INT,    LS_cmpxchg);
   case vmIntrinsics::_compareAndSwapLong:       return inline_unsafe_load_store(T_LONG,   LS_cmpxchg);
@@ -900,6 +925,8 @@
 
   case vmIntrinsics::_profileBoolean:
     return inline_profileBoolean();
+  case vmIntrinsics::_isCompileConstant:
+    return inline_isCompileConstant();
 
   default:
     // If you get here, it may be that someone has added a new intrinsic
@@ -3664,6 +3691,11 @@
 //---------------------generate_array_guard_common------------------------
 Node* LibraryCallKit::generate_array_guard_common(Node* kls, RegionNode* region,
                                                   bool obj_array, bool not_array) {
+
+  if (stopped()) {
+    return NULL;
+  }
+
   // If obj_array/non_array==false/false:
   // Branch around if the given klass is in fact an array (either obj or prim).
   // If obj_array/non_array==false/true:
@@ -3911,21 +3943,23 @@
         validated = true;
       }
 
-      newcopy = new_array(klass_node, length, 0);  // no arguments to push
-
-      ArrayCopyNode* ac = ArrayCopyNode::make(this, true, original, start, newcopy, intcon(0), moved, true,
-                                              load_object_klass(original), klass_node);
-      if (!is_copyOfRange) {
-        ac->set_copyof(validated);
-      } else {
-        ac->set_copyofrange(validated);
-      }
-      Node* n = _gvn.transform(ac);
-      if (n == ac) {
-        ac->connect_outputs(this);
-      } else {
-        assert(validated, "shouldn't transform if all arguments not validated");
-        set_all_memory(n);
+      if (!stopped()) {
+        newcopy = new_array(klass_node, length, 0);  // no arguments to push
+
+        ArrayCopyNode* ac = ArrayCopyNode::make(this, true, original, start, newcopy, intcon(0), moved, true,
+                                                load_object_klass(original), klass_node);
+        if (!is_copyOfRange) {
+          ac->set_copyof(validated);
+        } else {
+          ac->set_copyofrange(validated);
+        }
+        Node* n = _gvn.transform(ac);
+        if (n == ac) {
+          ac->connect_outputs(this);
+        } else {
+          assert(validated, "shouldn't transform if all arguments not validated");
+          set_all_memory(n);
+        }
       }
     }
   } // original reexecute is set back here
@@ -4674,6 +4708,141 @@
   return true;
 }
 
+// If we have a tighly coupled allocation, the arraycopy may take care
+// of the array initialization. If one of the guards we insert between
+// the allocation and the arraycopy causes a deoptimization, an
+// unitialized array will escape the compiled method. To prevent that
+// we set the JVM state for uncommon traps between the allocation and
+// the arraycopy to the state before the allocation so, in case of
+// deoptimization, we'll reexecute the allocation and the
+// initialization.
+JVMState* LibraryCallKit::arraycopy_restore_alloc_state(AllocateArrayNode* alloc, int& saved_reexecute_sp) {
+  if (alloc != NULL) {
+    ciMethod* trap_method = alloc->jvms()->method();
+    int trap_bci = alloc->jvms()->bci();
+
+    if (!C->too_many_traps(trap_method, trap_bci, Deoptimization::Reason_intrinsic) &
+          !C->too_many_traps(trap_method, trap_bci, Deoptimization::Reason_null_check)) {
+      // Make sure there's no store between the allocation and the
+      // arraycopy otherwise visible side effects could be rexecuted
+      // in case of deoptimization and cause incorrect execution.
+      bool no_interfering_store = true;
+      Node* mem = alloc->in(TypeFunc::Memory);
+      if (mem->is_MergeMem()) {
+        for (MergeMemStream mms(merged_memory(), mem->as_MergeMem()); mms.next_non_empty2(); ) {
+          Node* n = mms.memory();
+          if (n != mms.memory2() && !(n->is_Proj() && n->in(0) == alloc->initialization())) {
+            assert(n->is_Store(), "what else?");
+            no_interfering_store = false;
+            break;
+          }
+        }
+      } else {
+        for (MergeMemStream mms(merged_memory()); mms.next_non_empty(); ) {
+          Node* n = mms.memory();
+          if (n != mem && !(n->is_Proj() && n->in(0) == alloc->initialization())) {
+            assert(n->is_Store(), "what else?");
+            no_interfering_store = false;
+            break;
+          }
+        }
+      }
+
+      if (no_interfering_store) {
+        JVMState* old_jvms = alloc->jvms()->clone_shallow(C);
+        uint size = alloc->req();
+        SafePointNode* sfpt = new SafePointNode(size, old_jvms);
+        old_jvms->set_map(sfpt);
+        for (uint i = 0; i < size; i++) {
+          sfpt->init_req(i, alloc->in(i));
+        }
+        // re-push array length for deoptimization
+        sfpt->ins_req(old_jvms->stkoff() + old_jvms->sp(), alloc->in(AllocateNode::ALength));
+        old_jvms->set_sp(old_jvms->sp()+1);
+        old_jvms->set_monoff(old_jvms->monoff()+1);
+        old_jvms->set_scloff(old_jvms->scloff()+1);
+        old_jvms->set_endoff(old_jvms->endoff()+1);
+        old_jvms->set_should_reexecute(true);
+
+        sfpt->set_i_o(map()->i_o());
+        sfpt->set_memory(map()->memory());
+        sfpt->set_control(map()->control());
+
+        JVMState* saved_jvms = jvms();
+        saved_reexecute_sp = _reexecute_sp;
+
+        set_jvms(sfpt->jvms());
+        _reexecute_sp = jvms()->sp();
+
+        return saved_jvms;
+      }
+    }
+  }
+  return NULL;
+}
+
+// In case of a deoptimization, we restart execution at the
+// allocation, allocating a new array. We would leave an uninitialized
+// array in the heap that GCs wouldn't expect. Move the allocation
+// after the traps so we don't allocate the array if we
+// deoptimize. This is possible because tightly_coupled_allocation()
+// guarantees there's no observer of the allocated array at this point
+// and the control flow is simple enough.
+void LibraryCallKit::arraycopy_move_allocation_here(AllocateArrayNode* alloc, Node* dest, JVMState* saved_jvms, int saved_reexecute_sp) {
+  if (saved_jvms != NULL && !stopped()) {
+    assert(alloc != NULL, "only with a tightly coupled allocation");
+    // restore JVM state to the state at the arraycopy
+    saved_jvms->map()->set_control(map()->control());
+    assert(saved_jvms->map()->memory() == map()->memory(), "memory state changed?");
+    assert(saved_jvms->map()->i_o() == map()->i_o(), "IO state changed?");
+    // If we've improved the types of some nodes (null check) while
+    // emitting the guards, propagate them to the current state
+    map()->replaced_nodes().apply(saved_jvms->map());
+    set_jvms(saved_jvms);
+    _reexecute_sp = saved_reexecute_sp;
+
+    // Remove the allocation from above the guards
+    CallProjections callprojs;
+    alloc->extract_projections(&callprojs, true);
+    InitializeNode* init = alloc->initialization();
+    Node* alloc_mem = alloc->in(TypeFunc::Memory);
+    C->gvn_replace_by(callprojs.fallthrough_ioproj, alloc->in(TypeFunc::I_O));
+    C->gvn_replace_by(init->proj_out(TypeFunc::Memory), alloc_mem);
+    C->gvn_replace_by(init->proj_out(TypeFunc::Control), alloc->in(0));
+
+    // move the allocation here (after the guards)
+    _gvn.hash_delete(alloc);
+    alloc->set_req(TypeFunc::Control, control());
+    alloc->set_req(TypeFunc::I_O, i_o());
+    Node *mem = reset_memory();
+    set_all_memory(mem);
+    alloc->set_req(TypeFunc::Memory, mem);
+    set_control(init->proj_out(TypeFunc::Control));
+    set_i_o(callprojs.fallthrough_ioproj);
+
+    // Update memory as done in GraphKit::set_output_for_allocation()
+    const TypeInt* length_type = _gvn.find_int_type(alloc->in(AllocateNode::ALength));
+    const TypeOopPtr* ary_type = _gvn.type(alloc->in(AllocateNode::KlassNode))->is_klassptr()->as_instance_type();
+    if (ary_type->isa_aryptr() && length_type != NULL) {
+      ary_type = ary_type->is_aryptr()->cast_to_size(length_type);
+    }
+    const TypePtr* telemref = ary_type->add_offset(Type::OffsetBot);
+    int            elemidx  = C->get_alias_index(telemref);
+    set_memory(init->proj_out(TypeFunc::Memory), Compile::AliasIdxRaw);
+    set_memory(init->proj_out(TypeFunc::Memory), elemidx);
+
+    Node* allocx = _gvn.transform(alloc);
+    assert(allocx == alloc, "where has the allocation gone?");
+    assert(dest->is_CheckCastPP(), "not an allocation result?");
+
+    _gvn.hash_delete(dest);
+    dest->set_req(0, control());
+    Node* destx = _gvn.transform(dest);
+    assert(destx == dest, "where has the allocation result gone?");
+  }
+}
+
+
 //------------------------------inline_arraycopy-----------------------
 // public static native void java.lang.System.arraycopy(Object src,  int  srcPos,
 //                                                      Object dest, int destPos,
@@ -4686,6 +4855,19 @@
   Node* dest_offset = argument(3);  // type: int
   Node* length      = argument(4);  // type: int
 
+
+  // Check for allocation before we add nodes that would confuse
+  // tightly_coupled_allocation()
+  AllocateArrayNode* alloc = tightly_coupled_allocation(dest, NULL);
+
+  int saved_reexecute_sp = -1;
+  JVMState* saved_jvms = arraycopy_restore_alloc_state(alloc, saved_reexecute_sp);
+  // See arraycopy_restore_alloc_state() comment
+  // if alloc == NULL we don't have to worry about a tightly coupled allocation so we can emit all needed guards
+  // if saved_jvms != NULL (then alloc != NULL) then we can handle guards and a tightly coupled allocation
+  // if saved_jvms == NULL and alloc != NULL, we can’t emit any guards
+  bool can_emit_guards = (alloc == NULL || saved_jvms != NULL);
+
   // The following tests must be performed
   // (1) src and dest are arrays.
   // (2) src and dest arrays must have elements of the same BasicType
@@ -4699,42 +4881,20 @@
 
   // (3) src and dest must not be null.
   // always do this here because we need the JVM state for uncommon traps
-  src  = null_check(src,  T_ARRAY);
+  Node* null_ctl = top();
+  src  = saved_jvms != NULL ? null_check_oop(src, &null_ctl, true, true) : null_check(src,  T_ARRAY);
+  assert(null_ctl->is_top(), "no null control here");
   dest = null_check(dest, T_ARRAY);
 
-  // Check for allocation before we add nodes that would confuse
-  // tightly_coupled_allocation()
-  AllocateArrayNode* alloc = tightly_coupled_allocation(dest, NULL);
-
-  ciMethod* trap_method = method();
-  int trap_bci = bci();
-  SafePointNode* sfpt = NULL;
-  if (alloc != NULL) {
-    // The JVM state for uncommon traps between the allocation and
-    // arraycopy is set to the state before the allocation: if the
-    // initialization is performed by the array copy, we don't want to
-    // go back to the interpreter with an unitialized array.
-    JVMState* old_jvms = alloc->jvms();
-    JVMState* jvms = old_jvms->clone_shallow(C);
-    uint size = alloc->req();
-    sfpt = new SafePointNode(size, jvms);
-    jvms->set_map(sfpt);
-    for (uint i = 0; i < size; i++) {
-      sfpt->init_req(i, alloc->in(i));
-    }
-    // re-push array length for deoptimization
-    sfpt->ins_req(jvms->stkoff() + jvms->sp(), alloc->in(AllocateNode::ALength));
-    jvms->set_sp(jvms->sp()+1);
-    jvms->set_monoff(jvms->monoff()+1);
-    jvms->set_scloff(jvms->scloff()+1);
-    jvms->set_endoff(jvms->endoff()+1);
-    jvms->set_should_reexecute(true);
-
-    sfpt->set_i_o(map()->i_o());
-    sfpt->set_memory(map()->memory());
-
-    trap_method = jvms->method();
-    trap_bci = jvms->bci();
+  if (!can_emit_guards) {
+    // if saved_jvms == NULL and alloc != NULL, we don't emit any
+    // guards but the arraycopy node could still take advantage of a
+    // tightly allocated allocation. tightly_coupled_allocation() is
+    // called again to make sure it takes the null check above into
+    // account: the null check is mandatory and if it caused an
+    // uncommon trap to be emitted then the allocation can't be
+    // considered tightly coupled in this context.
+    alloc = tightly_coupled_allocation(dest, NULL);
   }
 
   bool validated = false;
@@ -4753,7 +4913,7 @@
   // Is the type for dest from speculation?
   bool dest_spec = false;
 
-  if (!has_src || !has_dest) {
+  if ((!has_src || !has_dest) && can_emit_guards) {
     // We don't have sufficient type information, let's see if
     // speculative types can help. We need to have types for both src
     // and dest so that it pays off.
@@ -4782,7 +4942,7 @@
     if (could_have_src && could_have_dest) {
       // This is going to pay off so emit the required guards
       if (!has_src) {
-        src = maybe_cast_profiled_obj(src, src_k, true, sfpt);
+        src = maybe_cast_profiled_obj(src, src_k, true);
         src_type  = _gvn.type(src);
         top_src  = src_type->isa_aryptr();
         has_src = (top_src != NULL && top_src->klass() != NULL);
@@ -4798,7 +4958,7 @@
     }
   }
 
-  if (has_src && has_dest) {
+  if (has_src && has_dest && can_emit_guards) {
     BasicType src_elem  = top_src->klass()->as_array_klass()->element_type()->basic_type();
     BasicType dest_elem = top_dest->klass()->as_array_klass()->element_type()->basic_type();
     if (src_elem  == T_ARRAY)  src_elem  = T_OBJECT;
@@ -4830,7 +4990,7 @@
       if (could_have_src && could_have_dest) {
         // If we can have both exact types, emit the missing guards
         if (could_have_src && !src_spec) {
-          src = maybe_cast_profiled_obj(src, src_k, true, sfpt);
+          src = maybe_cast_profiled_obj(src, src_k, true);
         }
         if (could_have_dest && !dest_spec) {
           dest = maybe_cast_profiled_obj(dest, dest_k, true);
@@ -4839,7 +4999,16 @@
     }
   }
 
-  if (!C->too_many_traps(trap_method, trap_bci, Deoptimization::Reason_intrinsic) && !src->is_top() && !dest->is_top()) {
+  ciMethod* trap_method = method();
+  int trap_bci = bci();
+  if (saved_jvms != NULL) {
+    trap_method = alloc->jvms()->method();
+    trap_bci = alloc->jvms()->bci();
+  }
+
+  if (!C->too_many_traps(trap_method, trap_bci, Deoptimization::Reason_intrinsic) &&
+      can_emit_guards &&
+      !src->is_top() && !dest->is_top()) {
     // validate arguments: enables transformation the ArrayCopyNode
     validated = true;
 
@@ -4875,28 +5044,13 @@
     Node* not_subtype_ctrl = gen_subtype_check(src_klass, dest_klass);
 
     if (not_subtype_ctrl != top()) {
-      if (sfpt != NULL) {
-        GraphKit kit(sfpt->jvms());
-        PreserveJVMState pjvms(&kit);
-        kit.set_control(not_subtype_ctrl);
-        kit.uncommon_trap(Deoptimization::Reason_intrinsic,
-                          Deoptimization::Action_make_not_entrant);
-        assert(kit.stopped(), "Should be stopped");
-      } else {
-        PreserveJVMState pjvms(this);
-        set_control(not_subtype_ctrl);
-        uncommon_trap(Deoptimization::Reason_intrinsic,
-                      Deoptimization::Action_make_not_entrant);
-        assert(stopped(), "Should be stopped");
-      }
+      PreserveJVMState pjvms(this);
+      set_control(not_subtype_ctrl);
+      uncommon_trap(Deoptimization::Reason_intrinsic,
+                    Deoptimization::Action_make_not_entrant);
+      assert(stopped(), "Should be stopped");
     }
-    if (sfpt != NULL) {
-      GraphKit kit(sfpt->jvms());
-      kit.set_control(_gvn.transform(slow_region));
-      kit.uncommon_trap(Deoptimization::Reason_intrinsic,
-                        Deoptimization::Action_make_not_entrant);
-      assert(kit.stopped(), "Should be stopped");
-    } else {
+    {
       PreserveJVMState pjvms(this);
       set_control(_gvn.transform(slow_region));
       uncommon_trap(Deoptimization::Reason_intrinsic,
@@ -4905,6 +5059,8 @@
     }
   }
 
+  arraycopy_move_allocation_here(alloc, dest, saved_jvms, saved_reexecute_sp);
+
   if (stopped()) {
     return true;
   }
@@ -5870,12 +6026,46 @@
                           Deoptimization::Action_reinterpret);
       return true;
     }
+
+    // result is a boolean (0 or 1) and its profile (false_cnt & true_cnt)
+    // is a number of each value occurrences.
+    Node* result = argument(0);
+    if (false_cnt == 0 || true_cnt == 0) {
+      // According to profile, one value has been never seen.
+      int expected_val = (false_cnt == 0) ? 1 : 0;
+
+      Node* cmp  = _gvn.transform(new CmpINode(result, intcon(expected_val)));
+      Node* test = _gvn.transform(new BoolNode(cmp, BoolTest::eq));
+
+      IfNode* check = create_and_map_if(control(), test, PROB_ALWAYS, COUNT_UNKNOWN);
+      Node* fast_path = _gvn.transform(new IfTrueNode(check));
+      Node* slow_path = _gvn.transform(new IfFalseNode(check));
+
+      { // Slow path: uncommon trap for never seen value and then reexecute
+        // MethodHandleImpl::profileBoolean() to bump the count, so JIT knows
+        // the value has been seen at least once.
+        PreserveJVMState pjvms(this);
+        PreserveReexecuteState preexecs(this);
+        jvms()->set_should_reexecute(true);
+
+        set_control(slow_path);
+        set_i_o(i_o());
+
+        uncommon_trap_exact(Deoptimization::Reason_intrinsic,
+                            Deoptimization::Action_reinterpret);
+      }
+      // The guard for never seen value enables sharpening of the result and
+      // returning a constant. It allows to eliminate branches on the same value
+      // later on.
+      set_control(fast_path);
+      result = intcon(expected_val);
+    }
     // Stop profiling.
-    // MethodHandleImpl::profileBoolean() has profiling logic in it's bytecode.
-    // By replacing method's body with profile data (represented as ProfileBooleanNode
+    // MethodHandleImpl::profileBoolean() has profiling logic in its bytecode.
+    // By replacing method body with profile data (represented as ProfileBooleanNode
     // on IR level) we effectively disable profiling.
     // It enables full speed execution once optimized code is generated.
-    Node* profile = _gvn.transform(new ProfileBooleanNode(argument(0), false_cnt, true_cnt));
+    Node* profile = _gvn.transform(new ProfileBooleanNode(result, false_cnt, true_cnt));
     C->record_for_igvn(profile);
     set_result(profile);
     return true;
@@ -5888,3 +6078,9 @@
     return false;
   }
 }
+
+bool LibraryCallKit::inline_isCompileConstant() {
+  Node* n = argument(0);
+  set_result(n->is_Con() ? intcon(1) : intcon(0));
+  return true;
+}
--- a/hotspot/src/share/vm/opto/loopTransform.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/loopTransform.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -38,6 +38,7 @@
 #include "opto/rootnode.hpp"
 #include "opto/runtime.hpp"
 #include "opto/subnode.hpp"
+#include "opto/vectornode.hpp"
 
 //------------------------------is_loop_exit-----------------------------------
 // Given an IfNode, return the loop-exiting projection or NULL if both
@@ -1524,6 +1525,44 @@
   }
 }
 
+void PhaseIdealLoop::mark_reductions(IdealLoopTree *loop) {
+  if (SuperWordReductions == false) return;
+
+  CountedLoopNode* loop_head = loop->_head->as_CountedLoop();
+  if (loop_head->unrolled_count() > 1) {
+    return;
+  }
+
+  Node* trip_phi = loop_head->phi();
+  for (DUIterator_Fast imax, i = loop_head->fast_outs(imax); i < imax; i++) {
+    Node* phi = loop_head->fast_out(i);
+    if (phi->is_Phi() && phi->outcnt() > 0 && phi != trip_phi) {
+      // For definitions which are loop inclusive and not tripcounts.
+      Node* def_node = phi->in(LoopNode::LoopBackControl);
+
+      if (def_node != NULL) {
+        Node* n_ctrl = get_ctrl(def_node);
+        if (n_ctrl != NULL && loop->is_member(get_loop(n_ctrl))) {
+          // Now test it to see if it fits the standard pattern for a reduction operator.
+          int opc = def_node->Opcode();
+          if (opc != ReductionNode::opcode(opc, def_node->bottom_type()->basic_type())) {
+            if (!def_node->is_reduction()) { // Not marked yet
+              // To be a reduction, the arithmetic node must have the phi as input and provide a def to it
+              for (unsigned j = 1; j < def_node->req(); j++) {
+                Node* in = def_node->in(j);
+                if (in == phi) {
+                  def_node->add_flag(Node::Flag_is_reduction);
+                  break;
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+}
+
 //------------------------------dominates_backedge---------------------------------
 // Returns true if ctrl is executed on every complete iteration
 bool IdealLoopTree::dominates_backedge(Node* ctrl) {
@@ -2361,8 +2400,10 @@
     // an even number of trips).  If we are peeling, we might enable some RCE
     // and we'd rather unroll the post-RCE'd loop SO... do not unroll if
     // peeling.
-    if (should_unroll && !should_peel)
-      phase->do_unroll(this,old_new, true);
+    if (should_unroll && !should_peel) {
+      phase->mark_reductions(this);
+      phase->do_unroll(this, old_new, true);
+    }
 
     // Adjust the pre-loop limits to align the main body
     // iterations.
--- a/hotspot/src/share/vm/opto/loopnode.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/loopnode.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -872,6 +872,9 @@
   // Unroll the loop body one step - make each trip do 2 iterations.
   void do_unroll( IdealLoopTree *loop, Node_List &old_new, bool adjust_min_trip );
 
+  // Mark vector reduction candidates before loop unrolling
+  void mark_reductions( IdealLoopTree *loop );
+
   // Return true if exp is a constant times an induction var
   bool is_scaled_iv(Node* exp, Node* iv, int* p_scale);
 
--- a/hotspot/src/share/vm/opto/loopopts.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/loopopts.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -241,8 +241,8 @@
   ProjNode* dp_proj  = dp->as_Proj();
   ProjNode* unc_proj = iff->as_If()->proj_out(1 - dp_proj->_con)->as_Proj();
   if (exclude_loop_predicate &&
-      (unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_predicate) ||
-       unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_range_check))) {
+      (unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_predicate) != NULL ||
+       unc_proj->is_uncommon_trap_proj(Deoptimization::Reason_range_check) != NULL)) {
     // If this is a range check (IfNode::is_range_check), do not
     // reorder because Compile::allow_range_check_smearing might have
     // changed the check.
--- a/hotspot/src/share/vm/opto/macro.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/macro.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -2543,7 +2543,7 @@
                (bol->_test._test == BoolTest::ne), "");
         IfNode* ifn = bol->unique_out()->as_If();
         assert((ifn->outcnt() == 2) &&
-               ifn->proj_out(1)->is_uncommon_trap_proj(Deoptimization::Reason_rtm_state_change), "");
+               ifn->proj_out(1)->is_uncommon_trap_proj(Deoptimization::Reason_rtm_state_change) != NULL, "");
 #endif
         Node* repl = n->in(1);
         if (!_has_locks) {
--- a/hotspot/src/share/vm/opto/multnode.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/multnode.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -150,59 +150,67 @@
 }
 
 //-------------------------------is_uncommon_trap_proj----------------------------
-// Return true if proj is the form of "proj->[region->..]call_uct"
-bool ProjNode::is_uncommon_trap_proj(Deoptimization::DeoptReason reason) {
+// Return uncommon trap call node if proj is for "proj->[region->..]call_uct"
+// NULL otherwise
+CallStaticJavaNode* ProjNode::is_uncommon_trap_proj(Deoptimization::DeoptReason reason) {
   int path_limit = 10;
   Node* out = this;
   for (int ct = 0; ct < path_limit; ct++) {
     out = out->unique_ctrl_out();
     if (out == NULL)
-      return false;
+      return NULL;
     if (out->is_CallStaticJava()) {
-      int req = out->as_CallStaticJava()->uncommon_trap_request();
+      CallStaticJavaNode* call = out->as_CallStaticJava();
+      int req = call->uncommon_trap_request();
       if (req != 0) {
         Deoptimization::DeoptReason trap_reason = Deoptimization::trap_request_reason(req);
         if (trap_reason == reason || reason == Deoptimization::Reason_none) {
-           return true;
+          return call;
         }
       }
-      return false; // don't do further after call
+      return NULL; // don't do further after call
     }
     if (out->Opcode() != Op_Region)
-      return false;
+      return NULL;
   }
-  return false;
+  return NULL;
 }
 
 //-------------------------------is_uncommon_trap_if_pattern-------------------------
-// Return true  for "if(test)-> proj -> ...
-//                          |
-//                          V
-//                      other_proj->[region->..]call_uct"
-//
+// Return uncommon trap call node for    "if(test)-> proj -> ...
+//                                                 |
+//                                                 V
+//                                             other_proj->[region->..]call_uct"
+// NULL otherwise
 // "must_reason_predicate" means the uct reason must be Reason_predicate
-bool ProjNode::is_uncommon_trap_if_pattern(Deoptimization::DeoptReason reason) {
+CallStaticJavaNode* ProjNode::is_uncommon_trap_if_pattern(Deoptimization::DeoptReason reason) {
   Node *in0 = in(0);
-  if (!in0->is_If()) return false;
+  if (!in0->is_If()) return NULL;
   // Variation of a dead If node.
-  if (in0->outcnt() < 2)  return false;
+  if (in0->outcnt() < 2)  return NULL;
   IfNode* iff = in0->as_If();
 
   // we need "If(Conv2B(Opaque1(...)))" pattern for reason_predicate
   if (reason != Deoptimization::Reason_none) {
     if (iff->in(1)->Opcode() != Op_Conv2B ||
        iff->in(1)->in(1)->Opcode() != Op_Opaque1) {
-      return false;
+      return NULL;
     }
   }
 
   ProjNode* other_proj = iff->proj_out(1-_con);
   if (other_proj == NULL) // Should never happen, but make Parfait happy.
-      return false;
-  if (other_proj->is_uncommon_trap_proj(reason)) {
+      return NULL;
+  CallStaticJavaNode* call = other_proj->is_uncommon_trap_proj(reason);
+  if (call != NULL) {
     assert(reason == Deoptimization::Reason_none ||
            Compile::current()->is_predicate_opaq(iff->in(1)->in(1)), "should be on the list");
-    return true;
+    return call;
   }
-  return false;
+  return NULL;
 }
+
+ProjNode* ProjNode::other_if_proj() const {
+  assert(_con == 0 || _con == 1, "not an if?");
+  return in(0)->as_If()->proj_out(1-_con);
+}
--- a/hotspot/src/share/vm/opto/multnode.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/multnode.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -89,13 +89,18 @@
   virtual void dump_spec(outputStream *st) const;
 #endif
 
-  // Return true if proj is for "proj->[region->..]call_uct"
-  bool is_uncommon_trap_proj(Deoptimization::DeoptReason reason);
-  // Return true for    "if(test)-> proj -> ...
-  //                          |
-  //                          V
-  //                      other_proj->[region->..]call_uct"
-  bool is_uncommon_trap_if_pattern(Deoptimization::DeoptReason reason);
+  // Return uncommon trap call node if proj is for "proj->[region->..]call_uct"
+  // NULL otherwise
+  CallStaticJavaNode* is_uncommon_trap_proj(Deoptimization::DeoptReason reason);
+  // Return uncommon trap call node for    "if(test)-> proj -> ...
+  //                                                 |
+  //                                                 V
+  //                                             other_proj->[region->..]call_uct"
+  // NULL otherwise
+  CallStaticJavaNode* is_uncommon_trap_if_pattern(Deoptimization::DeoptReason reason);
+
+  // Return other proj node when this is a If proj node
+  ProjNode* other_if_proj() const;
 };
 
 #endif // SHARE_VM_OPTO_MULTNODE_HPP
--- a/hotspot/src/share/vm/opto/node.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/node.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -2069,7 +2069,7 @@
 
 //--------------------------unique_ctrl_out------------------------------
 // Return the unique control out if only one. Null if none or more than one.
-Node* Node::unique_ctrl_out() {
+Node* Node::unique_ctrl_out() const {
   Node* found = NULL;
   for (uint i = 0; i < outcnt(); i++) {
     Node* use = raw_out(i);
--- a/hotspot/src/share/vm/opto/node.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/node.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -673,7 +673,8 @@
     Flag_avoid_back_to_back_before   = Flag_may_be_short_branch << 1,
     Flag_avoid_back_to_back_after    = Flag_avoid_back_to_back_before << 1,
     Flag_has_call                    = Flag_avoid_back_to_back_after << 1,
-    Flag_is_expensive                = Flag_has_call << 1,
+    Flag_is_reduction                = Flag_has_call << 1,
+    Flag_is_expensive                = Flag_is_reduction << 1,
     _max_flags = (Flag_is_expensive << 1) - 1 // allow flags combination
   };
 
@@ -701,6 +702,10 @@
 
   const jushort flags() const { return _flags; }
 
+  void add_flag(jushort fl) { init_flags(fl); }
+
+  void remove_flag(jushort fl) { clear_flag(fl); }
+
   // Return a dense integer opcode number
   virtual int Opcode() const;
 
@@ -852,6 +857,10 @@
   // The node is expensive: the best control is set during loop opts
   bool is_expensive() const { return (_flags & Flag_is_expensive) != 0 && in(0) != NULL; }
 
+  // An arithmetic node which accumulates a data in a loop.
+  // It must have the loop's phi as input and provide a def to the phi.
+  bool is_reduction() const { return (_flags & Flag_is_reduction) != 0; }
+
 //----------------- Optimization
 
   // Get the worst-case Type output for this Node.
@@ -931,7 +940,7 @@
   Node* find_similar(int opc);
 
   // Return the unique control out if only one. Null if none or more than one.
-  Node* unique_ctrl_out();
+  Node* unique_ctrl_out() const;
 
 //----------------- Code Generation
 
--- a/hotspot/src/share/vm/opto/subnode.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/subnode.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -275,6 +275,8 @@
   mask commute( ) const { return mask("032147658"[_test]-'0'); }
   mask negate( ) const { return mask(_test^4); }
   bool is_canonical( ) const { return (_test == BoolTest::ne || _test == BoolTest::lt || _test == BoolTest::le || _test == BoolTest::overflow); }
+  bool is_less( )  const { return _test == BoolTest::lt || _test == BoolTest::le; }
+  bool is_greater( ) const { return _test == BoolTest::gt || _test == BoolTest::ge; }
   void dump_on(outputStream *st) const;
 };
 
--- a/hotspot/src/share/vm/opto/superword.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/superword.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -65,7 +65,8 @@
   _lpt(NULL),                             // loop tree node
   _lp(NULL),                              // LoopNode
   _bb(NULL),                              // basic block
-  _iv(NULL)                               // induction var
+  _iv(NULL),                              // induction var
+  _race_possible(false)                   // cases where SDMU is true
 {}
 
 //------------------------------transform_loop---------------------------
@@ -145,7 +146,6 @@
 void SuperWord::SLP_extract() {
 
   // Ready the block
-
   if (!construct_bb())
     return; // Exit if no interesting nodes or complex graph.
 
@@ -449,11 +449,13 @@
   int preloop_stride = pre_end->stride_con();
 
   int span = preloop_stride * p.scale_in_bytes();
-
-  // Stride one accesses are alignable.
-  if (ABS(span) == p.memory_size())
+  int mem_size = p.memory_size();
+  int offset   = p.offset_in_bytes();
+  // Stride one accesses are alignable if offset is aligned to memory operation size.
+  // Offset can be unaligned when UseUnalignedAccesses is used.
+  if (ABS(span) == mem_size && (ABS(offset) % mem_size) == 0) {
     return true;
-
+  }
   // If initial offset from start of object is computable,
   // compute alignment within the vector.
   int vw = vector_width_in_bytes(p.mem());
@@ -463,7 +465,7 @@
     if (init_nd->is_Con() && p.invar() == NULL) {
       int init = init_nd->bottom_type()->is_int()->get_con();
 
-      int init_offset = init * p.scale_in_bytes() + p.offset_in_bytes();
+      int init_offset = init * p.scale_in_bytes() + offset;
       assert(init_offset >= 0, "positive offset from object start");
 
       if (span > 0) {
@@ -640,7 +642,7 @@
   }
 
   if (isomorphic(s1, s2)) {
-    if (independent(s1, s2)) {
+    if (independent(s1, s2) || reduction(s1, s2)) {
       if (!exists_at(s1, 0) && !exists_at(s2, 1)) {
         if (!s1->is_Mem() || are_adjacent_refs(s1, s2)) {
           int s1_align = alignment(s1);
@@ -718,6 +720,28 @@
   return independent_path(shallow, deep);
 }
 
+//------------------------------reduction---------------------------
+// Is there a data path between s1 and s2 and the nodes reductions?
+bool SuperWord::reduction(Node* s1, Node* s2) {
+  bool retValue = false;
+  int d1 = depth(s1);
+  int d2 = depth(s2);
+  if (d1 + 1 == d2) {
+    if (s1->is_reduction() && s2->is_reduction()) {
+      // This is an ordered set, so s1 should define s2
+      for (DUIterator_Fast imax, i = s1->fast_outs(imax); i < imax; i++) {
+        Node* t1 = s1->fast_out(i);
+        if (t1 == s2) {
+          // both nodes are reductions and connected
+          retValue = true;
+        }
+      }
+    }
+  }
+
+  return retValue;
+}
+
 //------------------------------independent_path------------------------------
 // Helper for independent
 bool SuperWord::independent_path(Node* shallow, Node* deep, uint dp) {
@@ -761,6 +785,7 @@
 void SuperWord::extend_packlist() {
   bool changed;
   do {
+    packset_sort(_packset.length());
     changed = false;
     for (int i = 0; i < _packset.length(); i++) {
       Node_List* p = _packset.at(i);
@@ -769,6 +794,13 @@
     }
   } while (changed);
 
+  if (_race_possible) {
+    for (int i = 0; i < _packset.length(); i++) {
+      Node_List* p = _packset.at(i);
+      order_def_uses(p);
+    }
+  }
+
 #ifndef PRODUCT
   if (TraceSuperWord) {
     tty->print_cr("\nAfter extend_packlist");
@@ -825,10 +857,12 @@
 
   int align = alignment(s1);
   int savings = -1;
+  int num_s1_uses = 0;
   Node* u1 = NULL;
   Node* u2 = NULL;
   for (DUIterator_Fast imax, i = s1->fast_outs(imax); i < imax; i++) {
     Node* t1 = s1->fast_out(i);
+    num_s1_uses++;
     if (!in_bb(t1)) continue;
     for (DUIterator_Fast jmax, j = s2->fast_outs(jmax); j < jmax; j++) {
       Node* t2 = s2->fast_out(j);
@@ -845,6 +879,9 @@
       }
     }
   }
+  if (num_s1_uses > 1) {
+    _race_possible = true;
+  }
   if (savings >= 0) {
     Node_List* pair = new Node_List();
     pair->push(u1);
@@ -856,9 +893,64 @@
   return changed;
 }
 
+//------------------------------order_def_uses---------------------------
+// For extended packsets, ordinally arrange uses packset by major component
+void SuperWord::order_def_uses(Node_List* p) {
+  Node* s1 = p->at(0);
+
+  if (s1->is_Store()) return;
+
+  // reductions are always managed beforehand
+  if (s1->is_reduction()) return;
+
+  for (DUIterator_Fast imax, i = s1->fast_outs(imax); i < imax; i++) {
+    Node* t1 = s1->fast_out(i);
+
+    // Only allow operand swap on commuting operations
+    if (!t1->is_Add() && !t1->is_Mul()) {
+      break;
+    }
+
+    // Now find t1's packset
+    Node_List* p2 = NULL;
+    for (int j = 0; j < _packset.length(); j++) {
+      p2 = _packset.at(j);
+      Node* first = p2->at(0);
+      if (t1 == first) {
+        break;
+      }
+      p2 = NULL;
+    }
+    // Arrange all sub components by the major component
+    if (p2 != NULL) {
+      for (uint j = 1; j < p->size(); j++) {
+        Node* d1 = p->at(j);
+        Node* u1 = p2->at(j);
+        opnd_positions_match(s1, t1, d1, u1);
+      }
+    }
+  }
+}
+
 //---------------------------opnd_positions_match-------------------------
 // Is the use of d1 in u1 at the same operand position as d2 in u2?
 bool SuperWord::opnd_positions_match(Node* d1, Node* u1, Node* d2, Node* u2) {
+  // check reductions to see if they are marshalled to represent the reduction
+  // operator in a specified opnd
+  if (u1->is_reduction() && u2->is_reduction()) {
+    // ensure reductions have phis and reduction definitions feeding the 1st operand
+    Node* first = u1->in(2);
+    if (first->is_Phi() || first->is_reduction()) {
+      u1->swap_edges(1, 2);
+    }
+    // ensure reductions have phis and reduction definitions feeding the 1st operand
+    first = u2->in(2);
+    if (first->is_Phi() || first->is_reduction()) {
+      u2->swap_edges(1, 2);
+    }
+    return true;
+  }
+
   uint ct = u1->req();
   if (ct != u2->req()) return false;
   uint i1 = 0;
@@ -940,7 +1032,8 @@
     for (int i = 0; i < _packset.length(); i++) {
       Node_List* p1 = _packset.at(i);
       if (p1 == NULL) continue;
-      for (int j = 0; j < _packset.length(); j++) {
+      // Because of sorting we can start at i + 1
+      for (int j = i + 1; j < _packset.length(); j++) {
         Node_List* p2 = _packset.at(j);
         if (p2 == NULL) continue;
         if (i == j) continue;
@@ -1067,8 +1160,19 @@
 //------------------------------implemented---------------------------
 // Can code be generated for pack p?
 bool SuperWord::implemented(Node_List* p) {
+  bool retValue = false;
   Node* p0 = p->at(0);
-  return VectorNode::implemented(p0->Opcode(), p->size(), velt_basic_type(p0));
+  if (p0 != NULL) {
+    int opc = p0->Opcode();
+    uint size = p->size();
+    if (p0->is_reduction()) {
+      const Type *arith_type = p0->bottom_type();
+      retValue = ReductionNode::implemented(opc, size, arith_type->basic_type());
+    } else {
+      retValue = VectorNode::implemented(opc, size, velt_basic_type(p0));
+    }
+  }
+  return retValue;
 }
 
 //------------------------------same_inputs--------------------------
@@ -1102,6 +1206,18 @@
     if (!is_vector_use(p0, i))
       return false;
   }
+  // Check if reductions are connected
+  if (p0->is_reduction()) {
+    Node* second_in = p0->in(2);
+    Node_List* second_pk = my_pack(second_in);
+    if (second_pk == NULL) {
+      // Remove reduction flag if no parent pack, it is not profitable
+      p0->remove_flag(Node::Flag_is_reduction);
+      return false;
+    } else if (second_pk->size() != p->size()) {
+      return false;
+    }
+  }
   if (VectorNode::is_shift(p0)) {
     // For now, return false if shift count is vector or not scalar promotion
     // case (different shift counts) because it is not supported yet.
@@ -1123,6 +1239,9 @@
         for (uint k = 0; k < use->req(); k++) {
           Node* n = use->in(k);
           if (def == n) {
+            // reductions can be loop carried dependences
+            if (def->is_reduction() && use->is_Phi())
+              continue;
             if (!is_vector_use(use, k)) {
               return false;
             }
@@ -1407,16 +1526,33 @@
         vlen_in_bytes = vn->as_StoreVector()->memory_size();
       } else if (n->req() == 3) {
         // Promote operands to vector
-        Node* in1 = vector_opd(p, 1);
+        Node* in1 = NULL;
+        bool node_isa_reduction = n->is_reduction();
+        if (node_isa_reduction) {
+          // the input to the first reduction operation is retained
+          in1 = low_adr->in(1);
+        } else {
+          in1 = vector_opd(p, 1);
+        }
         Node* in2 = vector_opd(p, 2);
-        if (VectorNode::is_invariant_vector(in1) && (n->is_Add() || n->is_Mul())) {
+        if (VectorNode::is_invariant_vector(in1) && (node_isa_reduction == false) && (n->is_Add() || n->is_Mul())) {
           // Move invariant vector input into second position to avoid register spilling.
           Node* tmp = in1;
           in1 = in2;
           in2 = tmp;
         }
-        vn = VectorNode::make(opc, in1, in2, vlen, velt_basic_type(n));
-        vlen_in_bytes = vn->as_Vector()->length_in_bytes();
+        if (node_isa_reduction) {
+          const Type *arith_type = n->bottom_type();
+          vn = ReductionNode::make(opc, NULL, in1, in2, arith_type->basic_type());
+          if (in2->is_Load()) {
+            vlen_in_bytes = in2->as_LoadVector()->memory_size();
+          } else {
+            vlen_in_bytes = in2->as_Vector()->length_in_bytes();
+          }
+        } else {
+          vn = VectorNode::make(opc, in1, in2, vlen, velt_basic_type(n));
+          vlen_in_bytes = vn->as_Vector()->length_in_bytes();
+        }
       } else {
         ShouldNotReachHere();
       }
@@ -1556,6 +1692,8 @@
     _n_idx_list.pop();
     Node* def = use->in(idx);
 
+    if (def->is_reduction()) continue;
+
     // Insert extract operation
     _igvn.hash_delete(def);
     int def_pos = alignment(def) / data_size(def);
@@ -1576,6 +1714,7 @@
 bool SuperWord::is_vector_use(Node* use, int u_idx) {
   Node_List* u_pk = my_pack(use);
   if (u_pk == NULL) return false;
+  if (use->is_reduction()) return true;
   Node* def = use->in(u_idx);
   Node_List* d_pk = my_pack(def);
   if (d_pk == NULL) {
@@ -1613,7 +1752,7 @@
   // by the visited and post_visited sets,
   // and count number of nodes in block.
   int bb_ct = 0;
-  for (uint i = 0; i < lpt()->_body.size(); i++ ) {
+  for (uint i = 0; i < lpt()->_body.size(); i++) {
     Node *n = lpt()->_body.at(i);
     set_bb_idx(n, i); // Create a temporary map
     if (in_bb(n)) {
@@ -1674,6 +1813,7 @@
   // Do a depth first walk over out edges
   int rpo_idx = bb_ct - 1;
   int size;
+  int reduction_uses = 0;
   while ((size = _stk.length()) > 0) {
     Node* n = _stk.top(); // Leave node on stack
     if (!visited_test_set(n)) {
@@ -1685,6 +1825,14 @@
         if (in_bb(use) && !visited_test(use) &&
             // Don't go around backedge
             (!use->is_Phi() || n == entry)) {
+          if (use->is_reduction()) {
+            // First see if we can map the reduction on the given system we are on, then
+            // make a data entry operation for each reduction we see.
+            BasicType bt = use->bottom_type()->basic_type();
+            if (ReductionNode::implemented(use->Opcode(), Matcher::min_vector_size(bt), bt)) {
+              reduction_uses++;
+            }
+          }
           _stk.push(use);
         }
       }
@@ -1708,7 +1856,8 @@
     set_bb_idx(n, j);
   }
 
-  initialize_bb(); // Ensure extra info is allocated.
+  // Ensure extra info is allocated.
+  initialize_bb();
 
 #ifndef PRODUCT
   if (TraceSuperWord) {
@@ -1726,7 +1875,7 @@
   }
 #endif
   assert(rpo_idx == -1 && bb_ct == _block.length(), "all block members found");
-  return (_mem_slice_head.length() > 0) || (_data_entry.length() > 0);
+  return (_mem_slice_head.length() > 0) || (reduction_uses > 0) || (_data_entry.length() > 0);
 }
 
 //------------------------------initialize_bb---------------------------
@@ -1959,6 +2108,27 @@
   _packset.remove_at(pos);
 }
 
+void SuperWord::packset_sort(int n) {
+  // simple bubble sort so that we capitalize with O(n) when its already sorted
+  while (n != 0) {
+    bool swapped = false;
+    for (int i = 1; i < n; i++) {
+      Node_List* q_low = _packset.at(i-1);
+      Node_List* q_i = _packset.at(i);
+
+      // only swap when we find something to swap
+      if (alignment(q_low->at(0)) > alignment(q_i->at(0))) {
+        Node_List* t = q_i;
+        *(_packset.adr_at(i)) = q_low;
+        *(_packset.adr_at(i-1)) = q_i;
+        swapped = true;
+      }
+    }
+    if (swapped == false) break;
+    n--;
+  }
+}
+
 //------------------------------executed_first---------------------------
 // Return the node executed first in pack p.  Uses the RPO block list
 // to determine order.
--- a/hotspot/src/share/vm/opto/superword.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/superword.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -249,6 +249,7 @@
   LoopNode*      _lp;              // Current LoopNode
   Node*          _bb;              // Current basic block
   PhiNode*       _iv;              // Induction var
+  bool           _race_possible;   // In cases where SDMU is true
 
   // Accessors
   Arena* arena()                   { return _arena; }
@@ -337,6 +338,8 @@
   bool isomorphic(Node* s1, Node* s2);
   // Is there no data path from s1 to s2 or s2 to s1?
   bool independent(Node* s1, Node* s2);
+  // Is there a data path between s1 and s2 and both are reductions?
+  bool reduction(Node* s1, Node* s2);
   // Helper for independent
   bool independent_path(Node* shallow, Node* deep, uint dp=0);
   void set_alignment(Node* s1, Node* s2, int align);
@@ -347,6 +350,8 @@
   bool follow_use_defs(Node_List* p);
   // Extend the packset by visiting uses of nodes in pack p
   bool follow_def_uses(Node_List* p);
+  // For extended packsets, ordinally arrange uses packset by major component
+  void order_def_uses(Node_List* p);
   // Estimate the savings from executing s1 and s2 as a pack
   int est_savings(Node* s1, Node* s2);
   int adjacent_profit(Node* s1, Node* s2);
@@ -419,9 +424,12 @@
   void print_bb();
   void print_stmt(Node* s);
   char* blank(uint depth);
+
+  void packset_sort(int n);
 };
 
 
+
 //------------------------------SWPointer---------------------------
 // Information about an address for dependence checking and vector alignment
 class SWPointer VALUE_OBJ_CLASS_SPEC {
--- a/hotspot/src/share/vm/opto/type.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/type.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -4083,7 +4083,9 @@
            (tap->_klass_is_exact && !tap->klass()->is_subtype_of(klass())) ||
            // 'this' is exact and super or unrelated:
            (this->_klass_is_exact && !klass()->is_subtype_of(tap->klass())))) {
-      tary = TypeAry::make(Type::BOTTOM, tary->_size, tary->_stable);
+      if (above_centerline(ptr)) {
+        tary = TypeAry::make(Type::BOTTOM, tary->_size, tary->_stable);
+      }
       return make(NotNull, NULL, tary, lazy_klass, false, off, InstanceBot, speculative, depth);
     }
 
--- a/hotspot/src/share/vm/opto/vectornode.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/vectornode.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -250,7 +250,6 @@
   int vopc = VectorNode::opcode(opc, bt);
   // This method should not be called for unimplemented vectors.
   guarantee(vopc > 0, err_msg_res("Vector for '%s' is not implemented", NodeClassNames[opc]));
-
   switch (vopc) {
   case Op_AddVB: return new AddVBNode(n1, n2, vt);
   case Op_AddVS: return new AddVSNode(n1, n2, vt);
@@ -441,3 +440,72 @@
   return NULL;
 }
 
+int ReductionNode::opcode(int opc, BasicType bt) {
+  int vopc = opc;
+  switch (opc) {
+    case Op_AddI:
+      assert(bt == T_INT, "must be");
+      vopc = Op_AddReductionVI;
+      break;
+    case Op_AddL:
+      assert(bt == T_LONG, "must be");
+      vopc = Op_AddReductionVL;
+      break;
+    case Op_AddF:
+      assert(bt == T_FLOAT, "must be");
+      vopc = Op_AddReductionVF;
+      break;
+    case Op_AddD:
+      assert(bt == T_DOUBLE, "must be");
+      vopc = Op_AddReductionVD;
+      break;
+    case Op_MulI:
+      assert(bt == T_INT, "must be");
+      vopc = Op_MulReductionVI;
+      break;
+    case Op_MulF:
+      assert(bt == T_FLOAT, "must be");
+      vopc = Op_MulReductionVF;
+      break;
+    case Op_MulD:
+      assert(bt == T_DOUBLE, "must be");
+      vopc = Op_MulReductionVD;
+      break;
+    // TODO: add MulL for targets that support it
+    default:
+      break;
+  }
+  return vopc;
+}
+
+// Return the appropriate reduction node.
+ReductionNode* ReductionNode::make(int opc, Node *ctrl, Node* n1, Node* n2, BasicType bt) {
+
+  int vopc = opcode(opc, bt);
+
+  // This method should not be called for unimplemented vectors.
+  guarantee(vopc != opc, err_msg_res("Vector for '%s' is not implemented", NodeClassNames[opc]));
+
+  switch (vopc) {
+  case Op_AddReductionVI: return new AddReductionVINode(ctrl, n1, n2);
+  case Op_AddReductionVL: return new AddReductionVLNode(ctrl, n1, n2);
+  case Op_AddReductionVF: return new AddReductionVFNode(ctrl, n1, n2);
+  case Op_AddReductionVD: return new AddReductionVDNode(ctrl, n1, n2);
+  case Op_MulReductionVI: return new MulReductionVINode(ctrl, n1, n2);
+  case Op_MulReductionVF: return new MulReductionVFNode(ctrl, n1, n2);
+  case Op_MulReductionVD: return new MulReductionVDNode(ctrl, n1, n2);
+  }
+  fatal(err_msg_res("Missed vector creation for '%s'", NodeClassNames[vopc]));
+  return NULL;
+}
+
+bool ReductionNode::implemented(int opc, uint vlen, BasicType bt) {
+  if (is_java_primitive(bt) &&
+      (vlen > 1) && is_power_of_2(vlen) &&
+      Matcher::vector_size_supported(bt, vlen)) {
+    int vopc = ReductionNode::opcode(opc, bt);
+    return vopc != opc && Matcher::match_rule_supported(vopc);
+  }
+  return false;
+}
+
--- a/hotspot/src/share/vm/opto/vectornode.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/opto/vectornode.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -90,6 +90,37 @@
   virtual int Opcode() const;
 };
 
+//------------------------------ReductionNode------------------------------------
+// Perform reduction of a vector
+class ReductionNode : public Node {
+ public:
+  ReductionNode(Node *ctrl, Node* in1, Node* in2) : Node(ctrl, in1, in2) {}
+
+  static ReductionNode* make(int opc, Node *ctrl, Node* in1, Node* in2, BasicType bt);
+  static int  opcode(int opc, BasicType bt);
+  static bool implemented(int opc, uint vlen, BasicType bt);
+};
+
+//------------------------------AddReductionVINode--------------------------------------
+// Vector add int as a reduction
+class AddReductionVINode : public ReductionNode {
+public:
+  AddReductionVINode(Node * ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeInt::INT; }
+  virtual uint ideal_reg() const { return Op_RegI; }
+};
+
+//------------------------------AddReductionVLNode--------------------------------------
+// Vector add long as a reduction
+class AddReductionVLNode : public ReductionNode {
+public:
+  AddReductionVLNode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeLong::LONG; }
+  virtual uint ideal_reg() const { return Op_RegL; }
+};
+
 //------------------------------AddVLNode--------------------------------------
 // Vector add long
 class AddVLNode : public VectorNode {
@@ -106,6 +137,16 @@
   virtual int Opcode() const;
 };
 
+//------------------------------AddReductionVFNode--------------------------------------
+// Vector add float as a reduction
+class AddReductionVFNode : public ReductionNode {
+public:
+  AddReductionVFNode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return Type::FLOAT; }
+  virtual uint ideal_reg() const { return Op_RegF; }
+};
+
 //------------------------------AddVDNode--------------------------------------
 // Vector add double
 class AddVDNode : public VectorNode {
@@ -114,6 +155,16 @@
   virtual int Opcode() const;
 };
 
+//------------------------------AddReductionVDNode--------------------------------------
+// Vector add double as a reduction
+class AddReductionVDNode : public ReductionNode {
+public:
+  AddReductionVDNode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return Type::DOUBLE; }
+  virtual uint ideal_reg() const { return Op_RegD; }
+};
+
 //------------------------------SubVBNode--------------------------------------
 // Vector subtract byte
 class SubVBNode : public VectorNode {
@@ -178,6 +229,16 @@
   virtual int Opcode() const;
 };
 
+//------------------------------MulReductionVINode--------------------------------------
+// Vector multiply int as a reduction
+class MulReductionVINode : public ReductionNode {
+public:
+  MulReductionVINode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return TypeInt::INT; }
+  virtual uint ideal_reg() const { return Op_RegI; }
+};
+
 //------------------------------MulVFNode--------------------------------------
 // Vector multiply float
 class MulVFNode : public VectorNode {
@@ -186,6 +247,16 @@
   virtual int Opcode() const;
 };
 
+//------------------------------MulReductionVFNode--------------------------------------
+// Vector multiply float as a reduction
+class MulReductionVFNode : public ReductionNode {
+public:
+  MulReductionVFNode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return Type::FLOAT; }
+  virtual uint ideal_reg() const { return Op_RegF; }
+};
+
 //------------------------------MulVDNode--------------------------------------
 // Vector multiply double
 class MulVDNode : public VectorNode {
@@ -194,6 +265,16 @@
   virtual int Opcode() const;
 };
 
+//------------------------------MulReductionVDNode--------------------------------------
+// Vector multiply double as a reduction
+class MulReductionVDNode : public ReductionNode {
+public:
+  MulReductionVDNode(Node *ctrl, Node* in1, Node* in2) : ReductionNode(ctrl, in1, in2) {}
+  virtual int Opcode() const;
+  virtual const Type* bottom_type() const { return Type::DOUBLE; }
+  virtual uint ideal_reg() const { return Op_RegD; }
+};
+
 //------------------------------DivVFNode--------------------------------------
 // Vector divide float
 class DivVFNode : public VectorNode {
--- a/hotspot/src/share/vm/precompiled/precompiled.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/precompiled/precompiled.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -94,7 +94,6 @@
 # include "gc_implementation/shared/spaceCounters.hpp"
 # include "gc_implementation/shared/spaceDecorator.hpp"
 # include "gc_interface/collectedHeap.hpp"
-# include "gc_interface/collectedHeap.inline.hpp"
 # include "gc_interface/gcCause.hpp"
 # include "interpreter/abstractInterpreter.hpp"
 # include "interpreter/bytecode.hpp"
@@ -114,15 +113,12 @@
 # include "memory/allocation.hpp"
 # include "memory/allocation.inline.hpp"
 # include "memory/barrierSet.hpp"
-# include "memory/barrierSet.inline.hpp"
 # include "memory/blockOffsetTable.hpp"
-# include "memory/blockOffsetTable.inline.hpp"
 # include "memory/cardTableModRefBS.hpp"
 # include "memory/collectorPolicy.hpp"
 # include "memory/defNewGeneration.hpp"
 # include "memory/gcLocker.hpp"
 # include "memory/genCollectedHeap.hpp"
-# include "memory/genOopClosures.hpp"
 # include "memory/genRemSet.hpp"
 # include "memory/generation.hpp"
 # include "memory/heap.hpp"
@@ -133,10 +129,8 @@
 # include "memory/referencePolicy.hpp"
 # include "memory/referenceProcessor.hpp"
 # include "memory/resourceArea.hpp"
-# include "memory/sharedHeap.hpp"
 # include "memory/space.hpp"
 # include "memory/threadLocalAllocBuffer.hpp"
-# include "memory/threadLocalAllocBuffer.inline.hpp"
 # include "memory/universe.hpp"
 # include "memory/universe.inline.hpp"
 # include "memory/watermark.hpp"
@@ -147,7 +141,6 @@
 # include "oops/instanceOop.hpp"
 # include "oops/instanceRefKlass.hpp"
 # include "oops/klass.hpp"
-# include "oops/klassPS.hpp"
 # include "oops/klassVtable.hpp"
 # include "oops/markOop.hpp"
 # include "oops/markOop.inline.hpp"
@@ -305,7 +298,6 @@
 # include "gc_implementation/concurrentMarkSweep/promotionInfo.hpp"
 # include "gc_implementation/g1/dirtyCardQueue.hpp"
 # include "gc_implementation/g1/g1BlockOffsetTable.hpp"
-# include "gc_implementation/g1/g1BlockOffsetTable.inline.hpp"
 # include "gc_implementation/g1/g1OopClosures.hpp"
 # include "gc_implementation/g1/g1_globals.hpp"
 # include "gc_implementation/g1/ptrQueue.hpp"
@@ -319,7 +311,6 @@
 # include "gc_implementation/parallelScavenge/psGCAdaptivePolicyCounters.hpp"
 # include "gc_implementation/parallelScavenge/psGenerationCounters.hpp"
 # include "gc_implementation/parallelScavenge/psOldGen.hpp"
-# include "gc_implementation/parallelScavenge/psParallelCompact.hpp"
 # include "gc_implementation/parallelScavenge/psVirtualspace.hpp"
 # include "gc_implementation/parallelScavenge/psYoungGen.hpp"
 # include "gc_implementation/shared/gcAdaptivePolicyCounters.hpp"
--- a/hotspot/src/share/vm/prims/jvm.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/jvm.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1465,63 +1465,26 @@
 }
 JVM_END
 
-// should be in InstanceKlass.cpp, but is here for historical reasons
-Klass* InstanceKlass::compute_enclosing_class_impl(instanceKlassHandle k,
-                                                     bool* inner_is_member,
-                                                     TRAPS) {
-  Thread* thread = THREAD;
-  InnerClassesIterator iter(k);
-  if (iter.length() == 0) {
-    // No inner class info => no declaring class
+JVM_ENTRY(jstring, JVM_GetSimpleBinaryName(JNIEnv *env, jclass cls))
+{
+  oop mirror = JNIHandles::resolve_non_null(cls);
+  if (java_lang_Class::is_primitive(mirror) ||
+      !java_lang_Class::as_Klass(mirror)->oop_is_instance()) {
     return NULL;
   }
-
-  constantPoolHandle i_cp(thread, k->constants());
-
-  bool found = false;
-  Klass* ok;
-  instanceKlassHandle outer_klass;
-  *inner_is_member = false;
-
-  // Find inner_klass attribute
-  for (; !iter.done() && !found; iter.next()) {
-    int ioff = iter.inner_class_info_index();
-    int ooff = iter.outer_class_info_index();
-    int noff = iter.inner_name_index();
-    if (ioff != 0) {
-      // Check to see if the name matches the class we're looking for
-      // before attempting to find the class.
-      if (i_cp->klass_name_at_matches(k, ioff)) {
-        Klass* inner_klass = i_cp->klass_at(ioff, CHECK_NULL);
-        found = (k() == inner_klass);
-        if (found && ooff != 0) {
-          ok = i_cp->klass_at(ooff, CHECK_NULL);
-          outer_klass = instanceKlassHandle(thread, ok);
-          *inner_is_member = true;
-        }
-      }
+  instanceKlassHandle k(THREAD, InstanceKlass::cast(java_lang_Class::as_Klass(mirror)));
+  int ooff = 0, noff = 0;
+  if (InstanceKlass::find_inner_classes_attr(k, &ooff, &noff, THREAD)) {
+    if (noff != 0) {
+      constantPoolHandle i_cp(thread, k->constants());
+      Symbol* name = i_cp->symbol_at(noff);
+      Handle str = java_lang_String::create_from_symbol(name, CHECK_NULL);
+      return (jstring) JNIHandles::make_local(env, str());
     }
   }
-
-  if (found && outer_klass.is_null()) {
-    // It may be anonymous; try for that.
-    int encl_method_class_idx = k->enclosing_method_class_index();
-    if (encl_method_class_idx != 0) {
-      ok = i_cp->klass_at(encl_method_class_idx, CHECK_NULL);
-      outer_klass = instanceKlassHandle(thread, ok);
-      *inner_is_member = false;
-    }
-  }
-
-  // If no inner class attribute found for this class.
-  if (outer_klass.is_null())  return NULL;
-
-  // Throws an exception if outer klass has not declared k as an inner klass
-  // We need evidence that each klass knows about the other, or else
-  // the system could allow a spoof of an inner class to gain access rights.
-  Reflection::check_for_inner_class(outer_klass, k, *inner_is_member, CHECK_NULL);
-  return outer_klass();
+  return NULL;
 }
+JVM_END
 
 JVM_ENTRY(jstring, JVM_GetClassSignature(JNIEnv *env, jclass cls))
   assert (cls != NULL, "illegal class");
--- a/hotspot/src/share/vm/prims/jvm.h	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/jvm.h	Fri Apr 17 23:55:04 2015 +0000
@@ -426,6 +426,9 @@
 JNIEXPORT jclass JNICALL
 JVM_GetDeclaringClass(JNIEnv *env, jclass ofClass);
 
+JNIEXPORT jstring JNICALL
+JVM_GetSimpleBinaryName(JNIEnv *env, jclass ofClass);
+
 /* Generics support (JDK 1.5) */
 JNIEXPORT jstring JNICALL
 JVM_GetClassSignature(JNIEnv *env, jclass cls);
--- a/hotspot/src/share/vm/prims/methodHandles.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/methodHandles.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -994,22 +994,8 @@
 // that intrinsic (non-JNI) native methods are defined in HotSpot.
 //
 
-JVM_ENTRY(jint, MHN_getConstant(JNIEnv *env, jobject igcls, jint which)) {
-  switch (which) {
-  case MethodHandles::GC_COUNT_GWT:
-#ifdef COMPILER2
-    return true;
-#else
-    return false;
-#endif
-  }
-  return 0;
-}
-JVM_END
-
 #ifndef PRODUCT
 #define EACH_NAMED_CON(template, requirement) \
-    template(MethodHandles,GC_COUNT_GWT) \
     template(java_lang_invoke_MemberName,MN_IS_METHOD) \
     template(java_lang_invoke_MemberName,MN_IS_CONSTRUCTOR) \
     template(java_lang_invoke_MemberName,MN_IS_FIELD) \
@@ -1019,7 +1005,6 @@
     template(java_lang_invoke_MemberName,MN_SEARCH_INTERFACES) \
     template(java_lang_invoke_MemberName,MN_REFERENCE_KIND_SHIFT) \
     template(java_lang_invoke_MemberName,MN_REFERENCE_KIND_MASK) \
-    template(MethodHandles,GC_LAMBDA_SUPPORT) \
     /*end*/
 
 #define IGNORE_REQ(req_expr) /* req_expr */
@@ -1313,7 +1298,6 @@
   {CC"init",                      CC"("MEM""OBJ")V",                     FN_PTR(MHN_init_Mem)},
   {CC"expand",                    CC"("MEM")V",                          FN_PTR(MHN_expand_Mem)},
   {CC"resolve",                   CC"("MEM""CLS")"MEM,                   FN_PTR(MHN_resolve_Mem)},
-  {CC"getConstant",               CC"(I)I",                              FN_PTR(MHN_getConstant)},
   //  static native int getNamedCon(int which, Object[] name)
   {CC"getNamedCon",               CC"(I["OBJ")I",                        FN_PTR(MHN_getNamedCon)},
   //  static native int getMembers(Class<?> defc, String matchName, String matchSig,
--- a/hotspot/src/share/vm/prims/methodHandles.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/methodHandles.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2008, 2014, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2008, 2015, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -125,11 +125,6 @@
     return signature_polymorphic_name_id(klass, name) != vmIntrinsics::_none;
   }
 
-  enum {
-    // format of query to getConstant:
-    GC_COUNT_GWT = 4,
-    GC_LAMBDA_SUPPORT = 5
-  };
   static int get_named_constant(int which, Handle name_box, TRAPS);
 
 public:
--- a/hotspot/src/share/vm/prims/unsafe.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/unsafe.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -324,6 +324,24 @@
 
 #endif // not SUPPORTS_NATIVE_CX8
 
+UNSAFE_ENTRY(jboolean, Unsafe_isBigEndian0(JNIEnv *env, jobject unsafe))
+  UnsafeWrapper("Unsafe_IsBigEndian0");
+  {
+#ifdef VM_LITTLE_ENDIAN
+    return false;
+#else
+    return true;
+#endif
+  }
+UNSAFE_END
+
+UNSAFE_ENTRY(jint, Unsafe_unalignedAccess0(JNIEnv *env, jobject unsafe))
+  UnsafeWrapper("Unsafe_UnalignedAccess0");
+  {
+    return UseUnalignedAccesses;
+  }
+UNSAFE_END
+
 #define DEFINE_GETSETOOP(jboolean, Boolean) \
  \
 UNSAFE_ENTRY(jboolean, Unsafe_Get##Boolean##140(JNIEnv *env, jobject unsafe, jobject obj, jint offset)) \
@@ -1261,6 +1279,9 @@
     {CC"loadFence",          CC"()V",                    FN_PTR(Unsafe_LoadFence)},
     {CC"storeFence",         CC"()V",                    FN_PTR(Unsafe_StoreFence)},
     {CC"fullFence",          CC"()V",                    FN_PTR(Unsafe_FullFence)},
+
+    {CC"isBigEndian0",       CC"()Z",                    FN_PTR(Unsafe_isBigEndian0)},
+    {CC"unalignedAccess0",   CC"()Z",                    FN_PTR(Unsafe_unalignedAccess0)}
 };
 
 #undef CC
--- a/hotspot/src/share/vm/prims/whitebox.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/whitebox.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -89,6 +89,10 @@
   return os::vm_page_size();
 WB_END
 
+WB_ENTRY(jlong, WB_GetVMLargePageSize(JNIEnv* env, jobject o))
+  return os::large_page_size();
+WB_END
+
 class WBIsKlassAliveClosure : public KlassClosure {
     Symbol* _name;
     bool _found;
@@ -819,46 +823,9 @@
   mo.notify_all();
 WB_END
 
-void WhiteBox::sweeper_thread_entry(JavaThread* thread, TRAPS) {
-  guarantee(WhiteBoxAPI, "internal testing API :: WhiteBox has to be enabled");
-  {
-    MutexLockerEx mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
-    NMethodSweeper::_should_sweep = true;
-  }
-  NMethodSweeper::possibly_sweep();
-}
-
-JavaThread* WhiteBox::create_sweeper_thread(TRAPS) {
-  // create sweeper thread w/ custom entry -- one iteration instead of loop
-  CodeCacheSweeperThread* sweeper_thread = new CodeCacheSweeperThread();
-  sweeper_thread->set_entry_point(&WhiteBox::sweeper_thread_entry);
-
-  // create j.l.Thread object and associate it w/ sweeper thread
-  {
-    // inherit deamon property from current thread
-    bool is_daemon = java_lang_Thread::is_daemon(JavaThread::current()->threadObj());
-
-    HandleMark hm(THREAD);
-    Handle thread_group(THREAD, Universe::system_thread_group());
-    const char* name = "WB Sweeper thread";
-    sweeper_thread->allocate_threadObj(thread_group, name, is_daemon, THREAD);
-  }
-
-  {
-    MutexLocker mu(Threads_lock, THREAD);
-    Threads::add(sweeper_thread);
-  }
-  return sweeper_thread;
-}
-
-WB_ENTRY(jobject, WB_ForceNMethodSweep(JNIEnv* env, jobject o))
-  JavaThread* sweeper_thread = WhiteBox::create_sweeper_thread(Thread::current());
-  if (sweeper_thread == NULL) {
-    return NULL;
-  }
-  jobject result = JNIHandles::make_local(env, sweeper_thread->threadObj());
-  Thread::start(sweeper_thread);
-  return result;
+WB_ENTRY(void, WB_ForceNMethodSweep(JNIEnv* env, jobject o))
+  // Force a code cache sweep and block until it finished
+  NMethodSweeper::force_sweep();
 WB_END
 
 WB_ENTRY(jboolean, WB_IsInStringTable(JNIEnv* env, jobject o, jstring javaString))
@@ -1296,19 +1263,20 @@
 #define CC (char*)
 
 static JNINativeMethod methods[] = {
-  {CC"getObjectAddress",   CC"(Ljava/lang/Object;)J", (void*)&WB_GetObjectAddress  },
-  {CC"getObjectSize",      CC"(Ljava/lang/Object;)J", (void*)&WB_GetObjectSize     },
-  {CC"isObjectInOldGen",   CC"(Ljava/lang/Object;)Z", (void*)&WB_isObjectInOldGen  },
+  {CC"getObjectAddress0",   CC"(Ljava/lang/Object;)J", (void*)&WB_GetObjectAddress  },
+  {CC"getObjectSize0",      CC"(Ljava/lang/Object;)J", (void*)&WB_GetObjectSize     },
+  {CC"isObjectInOldGen0",   CC"(Ljava/lang/Object;)Z", (void*)&WB_isObjectInOldGen  },
   {CC"getHeapOopSize",     CC"()I",                   (void*)&WB_GetHeapOopSize    },
   {CC"getVMPageSize",      CC"()I",                   (void*)&WB_GetVMPageSize     },
+  {CC"getVMLargePageSize", CC"()J",                   (void*)&WB_GetVMLargePageSize},
   {CC"isClassAlive0",      CC"(Ljava/lang/String;)Z", (void*)&WB_IsClassAlive      },
-  {CC"parseCommandLine",
+  {CC"parseCommandLine0",
       CC"(Ljava/lang/String;C[Lsun/hotspot/parser/DiagnosticCommand;)[Ljava/lang/Object;",
       (void*) &WB_ParseCommandLine
   },
-  {CC"addToBootstrapClassLoaderSearch", CC"(Ljava/lang/String;)V",
+  {CC"addToBootstrapClassLoaderSearch0", CC"(Ljava/lang/String;)V",
                                                       (void*)&WB_AddToBootstrapClassLoaderSearch},
-  {CC"addToSystemClassLoaderSearch",    CC"(Ljava/lang/String;)V",
+  {CC"addToSystemClassLoaderSearch0",    CC"(Ljava/lang/String;)V",
                                                       (void*)&WB_AddToSystemClassLoaderSearch},
   {CC"getCompressedOopsMaxHeapSize", CC"()J",
       (void*)&WB_GetCompressedOopsMaxHeapSize},
@@ -1318,7 +1286,7 @@
   {CC"stressVirtualSpaceResize",CC"(JJJ)I",           (void*)&WB_StressVirtualSpaceResize},
 #if INCLUDE_ALL_GCS
   {CC"g1InConcurrentMark", CC"()Z",                   (void*)&WB_G1InConcurrentMark},
-  {CC"g1IsHumongous",      CC"(Ljava/lang/Object;)Z", (void*)&WB_G1IsHumongous     },
+  {CC"g1IsHumongous0",      CC"(Ljava/lang/Object;)Z", (void*)&WB_G1IsHumongous     },
   {CC"g1NumMaxRegions",    CC"()J",                   (void*)&WB_G1NumMaxRegions  },
   {CC"g1NumFreeRegions",   CC"()J",                   (void*)&WB_G1NumFreeRegions  },
   {CC"g1RegionSize",       CC"()I",                   (void*)&WB_G1RegionSize      },
@@ -1339,29 +1307,29 @@
 #endif // INCLUDE_NMT
   {CC"deoptimizeFrames",   CC"(Z)I",                  (void*)&WB_DeoptimizeFrames  },
   {CC"deoptimizeAll",      CC"()V",                   (void*)&WB_DeoptimizeAll     },
-  {CC"deoptimizeMethod",   CC"(Ljava/lang/reflect/Executable;Z)I",
+  {CC"deoptimizeMethod0",   CC"(Ljava/lang/reflect/Executable;Z)I",
                                                       (void*)&WB_DeoptimizeMethod  },
-  {CC"isMethodCompiled",   CC"(Ljava/lang/reflect/Executable;Z)Z",
+  {CC"isMethodCompiled0",   CC"(Ljava/lang/reflect/Executable;Z)Z",
                                                       (void*)&WB_IsMethodCompiled  },
-  {CC"isMethodCompilable", CC"(Ljava/lang/reflect/Executable;IZ)Z",
+  {CC"isMethodCompilable0", CC"(Ljava/lang/reflect/Executable;IZ)Z",
                                                       (void*)&WB_IsMethodCompilable},
-  {CC"isMethodQueuedForCompilation",
+  {CC"isMethodQueuedForCompilation0",
       CC"(Ljava/lang/reflect/Executable;)Z",          (void*)&WB_IsMethodQueuedForCompilation},
-  {CC"makeMethodNotCompilable",
+  {CC"makeMethodNotCompilable0",
       CC"(Ljava/lang/reflect/Executable;IZ)V",        (void*)&WB_MakeMethodNotCompilable},
-  {CC"testSetDontInlineMethod",
+  {CC"testSetDontInlineMethod0",
       CC"(Ljava/lang/reflect/Executable;Z)Z",         (void*)&WB_TestSetDontInlineMethod},
-  {CC"getMethodCompilationLevel",
+  {CC"getMethodCompilationLevel0",
       CC"(Ljava/lang/reflect/Executable;Z)I",         (void*)&WB_GetMethodCompilationLevel},
-  {CC"getMethodEntryBci",
+  {CC"getMethodEntryBci0",
       CC"(Ljava/lang/reflect/Executable;)I",          (void*)&WB_GetMethodEntryBci},
   {CC"getCompileQueueSize",
       CC"(I)I",                                       (void*)&WB_GetCompileQueueSize},
-  {CC"testSetForceInlineMethod",
+  {CC"testSetForceInlineMethod0",
       CC"(Ljava/lang/reflect/Executable;Z)Z",         (void*)&WB_TestSetForceInlineMethod},
-  {CC"enqueueMethodForCompilation",
+  {CC"enqueueMethodForCompilation0",
       CC"(Ljava/lang/reflect/Executable;II)Z",        (void*)&WB_EnqueueMethodForCompilation},
-  {CC"clearMethodState",
+  {CC"clearMethodState0",
       CC"(Ljava/lang/reflect/Executable;)V",          (void*)&WB_ClearMethodState},
   {CC"lockCompilation",    CC"()V",                   (void*)&WB_LockCompilation},
   {CC"unlockCompilation",  CC"()V",                   (void*)&WB_UnlockCompilation},
@@ -1400,9 +1368,9 @@
   {CC"incMetaspaceCapacityUntilGC", CC"(J)J",         (void*)&WB_IncMetaspaceCapacityUntilGC },
   {CC"metaspaceCapacityUntilGC", CC"()J",             (void*)&WB_MetaspaceCapacityUntilGC },
   {CC"getCPUFeatures",     CC"()Ljava/lang/String;",  (void*)&WB_GetCPUFeatures     },
-  {CC"getNMethod",         CC"(Ljava/lang/reflect/Executable;Z)[Ljava/lang/Object;",
+  {CC"getNMethod0",         CC"(Ljava/lang/reflect/Executable;Z)[Ljava/lang/Object;",
                                                       (void*)&WB_GetNMethod         },
-  {CC"forceNMethodSweep0", CC"()Ljava/lang/Thread;",  (void*)&WB_ForceNMethodSweep  },
+  {CC"forceNMethodSweep",  CC"()V",                   (void*)&WB_ForceNMethodSweep  },
   {CC"allocateCodeBlob",   CC"(II)J",                 (void*)&WB_AllocateCodeBlob   },
   {CC"freeCodeBlob",       CC"(J)V",                  (void*)&WB_FreeCodeBlob       },
   {CC"getCodeHeapEntries", CC"(I)[Ljava/lang/Object;",(void*)&WB_GetCodeHeapEntries },
@@ -1412,7 +1380,7 @@
   {CC"getThreadStackSize", CC"()J",                   (void*)&WB_GetThreadStackSize },
   {CC"getThreadRemainingStackSize", CC"()J",          (void*)&WB_GetThreadRemainingStackSize },
   {CC"assertMatchingSafepointCalls", CC"(ZZ)V",       (void*)&WB_AssertMatchingSafepointCalls },
-  {CC"isMonitorInflated",  CC"(Ljava/lang/Object;)Z", (void*)&WB_IsMonitorInflated  },
+  {CC"isMonitorInflated0",  CC"(Ljava/lang/Object;)Z", (void*)&WB_IsMonitorInflated  },
   {CC"forceSafepoint",     CC"()V",                   (void*)&WB_ForceSafepoint     },
   {CC"getMethodBooleanOption",
       CC"(Ljava/lang/reflect/Executable;Ljava/lang/String;)Ljava/lang/Boolean;",
--- a/hotspot/src/share/vm/prims/whitebox.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/prims/whitebox.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -70,8 +70,6 @@
     Symbol* signature_symbol);
   static const char* lookup_jstring(const char* field_name, oop object);
   static bool lookup_bool(const char* field_name, oop object);
-  static void sweeper_thread_entry(JavaThread* thread, TRAPS);
-  static JavaThread* create_sweeper_thread(TRAPS);
   static int get_blob_type(const CodeBlob* code);
   static CodeHeap* get_code_heap(int blob_type);
   static CodeBlob* allocate_code_blob(int size, int blob_type);
--- a/hotspot/src/share/vm/runtime/arguments.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/arguments.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -92,6 +92,8 @@
 bool   Arguments::_UseOnStackReplacement        = UseOnStackReplacement;
 bool   Arguments::_BackgroundCompilation        = BackgroundCompilation;
 bool   Arguments::_ClipInlining                 = ClipInlining;
+intx   Arguments::_Tier3InvokeNotifyFreqLog     = Tier3InvokeNotifyFreqLog;
+intx   Arguments::_Tier4InvocationThreshold     = Tier4InvocationThreshold;
 
 char*  Arguments::SharedArchivePath             = NULL;
 
@@ -1012,6 +1014,14 @@
   AlwaysCompileLoopMethods   = Arguments::_AlwaysCompileLoopMethods;
   UseOnStackReplacement      = Arguments::_UseOnStackReplacement;
   BackgroundCompilation      = Arguments::_BackgroundCompilation;
+  if (TieredCompilation) {
+    if (FLAG_IS_DEFAULT(Tier3InvokeNotifyFreqLog)) {
+      Tier3InvokeNotifyFreqLog = Arguments::_Tier3InvokeNotifyFreqLog;
+    }
+    if (FLAG_IS_DEFAULT(Tier4InvocationThreshold)) {
+      Tier4InvocationThreshold = Arguments::_Tier4InvocationThreshold;
+    }
+  }
 
   // Change from defaults based on mode
   switch (mode) {
@@ -2525,6 +2535,10 @@
   Arguments::_UseOnStackReplacement    = UseOnStackReplacement;
   Arguments::_ClipInlining             = ClipInlining;
   Arguments::_BackgroundCompilation    = BackgroundCompilation;
+  if (TieredCompilation) {
+    Arguments::_Tier3InvokeNotifyFreqLog = Tier3InvokeNotifyFreqLog;
+    Arguments::_Tier4InvocationThreshold = Tier4InvocationThreshold;
+  }
 
   // Setup flags for mixed which is the default
   set_mode_flags(_mixed);
@@ -3086,7 +3100,7 @@
          FLAG_SET_CMDLINE(size_t, MaxNewSize, NewSize);
       }
 
-#ifndef _ALLBSD_SOURCE  // UseLargePages is not yet supported on BSD.
+#if !defined(_ALLBSD_SOURCE) && !defined(AIX)  // UseLargePages is not yet supported on BSD and AIX.
       FLAG_SET_DEFAULT(UseLargePages, true);
 #endif
 
@@ -3145,7 +3159,8 @@
       uintx max_tenuring_thresh = 0;
       if(!parse_uintx(tail, &max_tenuring_thresh, 0)) {
         jio_fprintf(defaultStream::error_stream(),
-                    "Invalid MaxTenuringThreshold: %s\n", option->optionString);
+          "Improperly specified VM option 'MaxTenuringThreshold=%s'\n", tail);
+        return JNI_EINVAL;
       }
       FLAG_SET_CMDLINE(uintx, MaxTenuringThreshold, max_tenuring_thresh);
 
@@ -3808,7 +3823,7 @@
             hotspotrc, hotspotrc);
   }
 
-#ifdef _ALLBSD_SOURCE  // UseLargePages is not yet supported on BSD.
+#if defined(_ALLBSD_SOURCE) || defined(AIX)  // UseLargePages is not yet supported on BSD and AIX.
   UNSUPPORTED_OPTION(UseLargePages, "-XX:+UseLargePages");
 #endif
 
--- a/hotspot/src/share/vm/runtime/arguments.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/arguments.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -323,6 +323,8 @@
   static bool _BackgroundCompilation;
   static bool _ClipInlining;
   static bool _CIDynamicCompilePriority;
+  static intx _Tier3InvokeNotifyFreqLog;
+  static intx _Tier4InvocationThreshold;
 
   // Tiered
   static void set_tiered_flags();
--- a/hotspot/src/share/vm/runtime/deoptimization.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/deoptimization.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -1861,6 +1861,7 @@
   "speculate_null_check",
   "rtm_state_change",
   "unstable_if",
+  "unstable_fused_if",
   "tenured"
 };
 const char* Deoptimization::_trap_action_name[] = {
--- a/hotspot/src/share/vm/runtime/deoptimization.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/deoptimization.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -63,6 +63,7 @@
     Reason_speculate_null_check,  // saw unexpected null from type speculation
     Reason_rtm_state_change,      // rtm state change detected
     Reason_unstable_if,           // a branch predicted always false was taken
+    Reason_unstable_fused_if,     // fused two ifs that had each one untaken branch. One is now taken.
 
     // Reason_tenured is counted separately, add normal counted Reasons above.
     // Related to MethodData::_trap_hist_limit where Reason_tenured isn't included
@@ -326,6 +327,8 @@
       return Reason_null_check;
     else if (reason == Reason_unstable_if)
       return Reason_intrinsic;
+    else if (reason == Reason_unstable_fused_if)
+      return Reason_range_check;
     else
       return Reason_none;
   }
--- a/hotspot/src/share/vm/runtime/globals.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/globals.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -3909,7 +3909,10 @@
           "Enable event-based tracing")                                     \
                                                                             \
   product(bool, UseLockedTracing, false,                                    \
-          "Use locked-tracing when doing event-based tracing")
+          "Use locked-tracing when doing event-based tracing")              \
+                                                                            \
+  diagnostic(bool, UseUnalignedAccesses, false,                             \
+          "Use unaligned memory accesses in sun.misc.Unsafe")
 
 /*
  *  Macros for factoring of globals
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/src/share/vm/runtime/rtmLocking.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2014, 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "utilities/globalDefinitions.hpp"
+
+#if INCLUDE_RTM_OPT
+
+#include "memory/allocation.inline.hpp"
+#include "runtime/task.hpp"
+#include "runtime/rtmLocking.hpp"
+
+
+// One-shot PeriodicTask subclass for enabling RTM locking
+uintx RTMLockingCounters::_calculation_flag = 0;
+
+class RTMLockingCalculationTask : public PeriodicTask {
+ public:
+  RTMLockingCalculationTask(size_t interval_time) : PeriodicTask(interval_time){  }
+
+  virtual void task() {
+    RTMLockingCounters::_calculation_flag = 1;
+    // Reclaim our storage and disenroll ourself
+    delete this;
+  }
+};
+
+void RTMLockingCounters::init() {
+  if (UseRTMLocking && RTMLockingCalculationDelay > 0) {
+    RTMLockingCalculationTask* task = new RTMLockingCalculationTask(RTMLockingCalculationDelay);
+    task->enroll();
+  } else {
+    _calculation_flag = 1;
+  }
+}
+
+//------------------------------print_on-------------------------------
+void RTMLockingCounters::print_on(outputStream* st) {
+  tty->print_cr("# rtm locks total (estimated): " UINTX_FORMAT, _total_count * RTMTotalCountIncrRate);
+  tty->print_cr("# rtm lock aborts  : " UINTX_FORMAT, _abort_count);
+  for (int i = 0; i < ABORT_STATUS_LIMIT; i++) {
+    tty->print_cr("# rtm lock aborts %d: " UINTX_FORMAT, i, _abortX_count[i]);
+  }
+}
+
+#endif
--- a/hotspot/src/share/vm/runtime/sweeper.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/sweeper.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -144,6 +144,7 @@
 int      NMethodSweeper::_seen                         = 0;    // Nof. nmethod we have currently processed in current pass of CodeCache
 
 volatile bool NMethodSweeper::_should_sweep            = true; // Indicates if we should invoke the sweeper
+volatile bool NMethodSweeper::_force_sweep             = false;// Indicates if we should force a sweep
 volatile int  NMethodSweeper::_bytes_changed           = 0;    // Counts the total nmethod size if the nmethod changed from:
                                                                //   1) alive       -> not_entrant
                                                                //   2) not_entrant -> zombie
@@ -276,6 +277,23 @@
 }
 
 /**
+  * Wakes up the sweeper thread and forces a sweep. Blocks until it finished.
+  */
+void NMethodSweeper::force_sweep() {
+  ThreadBlockInVM tbivm(JavaThread::current());
+  MutexLockerEx waiter(CodeCache_lock, Mutex::_no_safepoint_check_flag);
+  // Request forced sweep
+  _force_sweep = true;
+  while (_force_sweep) {
+    // Notify sweeper that we want to force a sweep and wait for completion.
+    // In case a sweep currently takes place we timeout and try again because
+    // we want to enforce a full sweep.
+    CodeCache_lock->notify();
+    CodeCache_lock->wait(Mutex::_no_safepoint_check_flag, 1000);
+  }
+}
+
+/**
  * Handle a safepoint request
  */
 void NMethodSweeper::handle_safepoint_request() {
@@ -335,6 +353,9 @@
     }
   }
 
+  // Remember if this was a forced sweep
+  bool forced = _force_sweep;
+
   // Force stack scanning if there is only 10% free space in the code cache.
   // We force stack scanning only non-profiled code heap gets full, since critical
   // allocation go to the non-profiled heap and we must be make sure that there is
@@ -344,7 +365,7 @@
     do_stack_scanning();
   }
 
-  if (_should_sweep) {
+  if (_should_sweep || forced) {
     init_sweeper_log();
     sweep_code_cache();
   }
@@ -356,12 +377,20 @@
   _should_sweep = false;
   // If there was enough state change, 'possibly_enable_sweeper()'
   // sets '_should_sweep' to true
-   possibly_enable_sweeper();
+  possibly_enable_sweeper();
   // Reset _bytes_changed only if there was enough state change. _bytes_changed
   // can further increase by calls to 'report_state_change'.
   if (_should_sweep) {
     _bytes_changed = 0;
   }
+
+  if (forced) {
+    // Notify requester that forced sweep finished
+    assert(_force_sweep, "Should be a forced sweep");
+    MutexLockerEx mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
+    _force_sweep = false;
+    CodeCache_lock->notify();
+  }
 }
 
 void NMethodSweeper::sweep_code_cache() {
--- a/hotspot/src/share/vm/runtime/sweeper.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/sweeper.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -54,7 +54,6 @@
 //     nmethod's space is freed.
 
 class NMethodSweeper : public AllStatic {
-  friend class WhiteBox;
  private:
   enum MethodStateChange {
     None,
@@ -71,6 +70,7 @@
 
   static volatile int  _sweep_started;            // Flag to control conc sweeper
   static volatile bool _should_sweep;             // Indicates if we should invoke the sweeper
+  static volatile bool _force_sweep;              // Indicates if we should force a sweep
   static volatile int _bytes_changed;             // Counts the total nmethod size if the nmethod changed from:
                                                   //   1) alive       -> not_entrant
                                                   //   2) not_entrant -> zombie
@@ -117,6 +117,7 @@
   static void mark_active_nmethods();      // Invoked at the end of each safepoint
   static void sweeper_loop();
   static void notify(int code_blob_type);  // Possibly start the sweeper thread.
+  static void force_sweep();
 
   static int hotness_counter_reset_val();
   static void report_state_change(nmethod* nm);
--- a/hotspot/src/share/vm/runtime/thread.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/thread.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -272,6 +272,11 @@
 #endif // ASSERT
 }
 
+// Non-inlined version to be used where thread.inline.hpp shouldn't be included.
+Thread* Thread::current_noinline() {
+  return Thread::current();
+}
+
 void Thread::initialize_thread_local_storage() {
   // Note: Make sure this method only calls
   // non-blocking operations. Otherwise, it might not work
@@ -754,13 +759,9 @@
       return true;
     } else {
       guarantee(res == strong_roots_parity, "Or else what?");
-      assert(SharedHeap::heap()->workers()->active_workers() > 0,
-             "Should only fail when parallel.");
       return false;
     }
   }
-  assert(SharedHeap::heap()->workers()->active_workers() > 0,
-         "Should only fail when parallel.");
   return false;
 }
 
@@ -4056,7 +4057,7 @@
          "Not in range.");
 }
 
-#ifndef PRODUCT
+#ifdef ASSERT
 void Threads::assert_all_threads_claimed() {
   ALL_JAVA_THREADS(p) {
     const int thread_parity = p->oops_do_parity();
@@ -4064,22 +4065,9 @@
         err_msg("Thread " PTR_FORMAT " has incorrect parity %d != %d", p2i(p), thread_parity, _thread_claim_parity));
   }
 }
-#endif // PRODUCT
-
-void Threads::possibly_parallel_oops_do(OopClosure* f, CLDClosure* cld_f, CodeBlobClosure* cf) {
-  // Introduce a mechanism allowing parallel threads to claim threads as
-  // root groups.  Overhead should be small enough to use all the time,
-  // even in sequential code.
-  SharedHeap* sh = SharedHeap::heap();
-  // Cannot yet substitute active_workers for n_par_threads
-  // because of G1CollectedHeap::verify() use of
-  // SharedHeap::process_roots().  n_par_threads == 0 will
-  // turn off parallelism in process_roots while active_workers
-  // is being used for parallelism elsewhere.
-  bool is_par = sh->n_par_threads() > 0;
-  assert(!is_par ||
-         (SharedHeap::heap()->n_par_threads() ==
-         SharedHeap::heap()->workers()->active_workers()), "Mismatch");
+#endif // ASSERT
+
+void Threads::possibly_parallel_oops_do(bool is_par, OopClosure* f, CLDClosure* cld_f, CodeBlobClosure* cf) {
   int cp = Threads::thread_claim_parity();
   ALL_JAVA_THREADS(p) {
     if (p->claim_oops_do(is_par, cp)) {
--- a/hotspot/src/share/vm/runtime/thread.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/thread.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -324,6 +324,8 @@
 
   // Returns the current thread
   static inline Thread* current();
+  // ... without having to include thread.inline.hpp.
+  static Thread* current_noinline();
 
   // Common thread operations
   static void set_priority(Thread* thread, ThreadPriority priority);
@@ -1886,15 +1888,28 @@
   // Does not include JNI_VERSION_1_1
   static jboolean is_supported_jni_version(jint version);
 
+  // The "thread claim parity" provides a way for threads to be claimed
+  // by parallel worker tasks.
+  //
+  // Each thread contains a a "parity" field. A task will claim the
+  // thread only if its parity field is the same as the global parity,
+  // which is updated by calling change_thread_claim_parity().
+  //
+  // For this to work change_thread_claim_parity() needs to be called
+  // exactly once in sequential code before starting parallel tasks
+  // that should claim threads.
+  //
+  // New threads get their parity set to 0 and change_thread_claim_parity()
+  // never set the global parity to 0.
   static int thread_claim_parity() { return _thread_claim_parity; }
   static void change_thread_claim_parity();
+  static void assert_all_threads_claimed() NOT_DEBUG_RETURN;
 
-  static void assert_all_threads_claimed() PRODUCT_RETURN;
   // Apply "f->do_oop" to all root oops in all threads.
   // This version may only be called by sequential code.
   static void oops_do(OopClosure* f, CLDClosure* cld_f, CodeBlobClosure* cf);
   // This version may be called by sequential or parallel code.
-  static void possibly_parallel_oops_do(OopClosure* f, CLDClosure* cld_f, CodeBlobClosure* cf);
+  static void possibly_parallel_oops_do(bool is_par, OopClosure* f, CLDClosure* cld_f, CodeBlobClosure* cf);
   // This creates a list of GCTasks, one per thread.
   static void create_thread_roots_tasks(GCTaskQueue* q);
   // This creates a list of GCTasks, one per thread, for marking objects.
--- a/hotspot/src/share/vm/runtime/virtualspace.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/virtualspace.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -37,13 +37,22 @@
     _alignment(0), _special(false), _executable(false) {
 }
 
-ReservedSpace::ReservedSpace(size_t size) {
+ReservedSpace::ReservedSpace(size_t size, size_t preferred_page_size) {
+  bool has_preferred_page_size = preferred_page_size != 0;
   // Want to use large pages where possible and pad with small pages.
-  size_t page_size = os::page_size_for_region_unaligned(size, 1);
+  size_t page_size = has_preferred_page_size ? preferred_page_size : os::page_size_for_region_unaligned(size, 1);
   bool large_pages = page_size != (size_t)os::vm_page_size();
-  // Don't force the alignment to be large page aligned,
-  // since that will waste memory.
-  size_t alignment = os::vm_allocation_granularity();
+  size_t alignment;
+  if (large_pages && has_preferred_page_size) {
+    alignment = MAX2(page_size, (size_t)os::vm_allocation_granularity());
+    // ReservedSpace initialization requires size to be aligned to the given
+    // alignment. Align the size up.
+    size = align_size_up(size, alignment);
+  } else {
+    // Don't force the alignment to be large page aligned,
+    // since that will waste memory.
+    alignment = os::vm_allocation_granularity();
+  }
   initialize(size, alignment, large_pages, NULL, false);
 }
 
--- a/hotspot/src/share/vm/runtime/virtualspace.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/virtualspace.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -51,7 +51,11 @@
  public:
   // Constructor
   ReservedSpace();
-  ReservedSpace(size_t size);
+  // Initialize the reserved space with the given size. If preferred_page_size
+  // is set, use this as minimum page size/alignment. This may waste some space
+  // if the given size is not aligned to that value, as the reservation will be
+  // aligned up to the final alignment in this case.
+  ReservedSpace(size_t size, size_t preferred_page_size = 0);
   ReservedSpace(size_t size, size_t alignment, bool large,
                 char* requested_address = NULL);
   ReservedSpace(size_t size, size_t alignment, bool large, bool executable);
--- a/hotspot/src/share/vm/runtime/vmStructs.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/runtime/vmStructs.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -558,7 +558,6 @@
     static_field(GenCollectedHeap,             _gch,                                          GenCollectedHeap*)                     \
   nonstatic_field(GenCollectedHeap,            _young_gen,                                    Generation*)                           \
   nonstatic_field(GenCollectedHeap,            _old_gen,                                      Generation*)                           \
- nonstatic_field(GenCollectedHeap,             _n_gens,                                       int)                                   \
                                                                                                                                      \
   nonstatic_field(GenCollectorPolicy,          _young_gen_spec,                               GenerationSpec*)                       \
   nonstatic_field(GenCollectorPolicy,          _old_gen_spec,                                 GenerationSpec*)                       \
@@ -1501,8 +1500,7 @@
   /******************************************/                            \
                                                                           \
   declare_toplevel_type(CollectedHeap)                                    \
-           declare_type(SharedHeap,                   CollectedHeap)      \
-           declare_type(GenCollectedHeap,             SharedHeap)         \
+           declare_type(GenCollectedHeap,             CollectedHeap)      \
   declare_toplevel_type(Generation)                                       \
            declare_type(DefNewGeneration,             Generation)         \
            declare_type(CardGeneration,               Generation)         \
@@ -1985,13 +1983,18 @@
   declare_c2_type(PowDNode, Node)                                         \
   declare_c2_type(ReverseBytesINode, Node)                                \
   declare_c2_type(ReverseBytesLNode, Node)                                \
+  declare_c2_type(ReductionNode, Node)                                    \
   declare_c2_type(VectorNode, Node)                                       \
   declare_c2_type(AddVBNode, VectorNode)                                  \
   declare_c2_type(AddVSNode, VectorNode)                                  \
   declare_c2_type(AddVINode, VectorNode)                                  \
+  declare_c2_type(AddReductionVINode, ReductionNode)                      \
   declare_c2_type(AddVLNode, VectorNode)                                  \
+  declare_c2_type(AddReductionVLNode, ReductionNode)                      \
   declare_c2_type(AddVFNode, VectorNode)                                  \
+  declare_c2_type(AddReductionVFNode, ReductionNode)                      \
   declare_c2_type(AddVDNode, VectorNode)                                  \
+  declare_c2_type(AddReductionVDNode, ReductionNode)                      \
   declare_c2_type(SubVBNode, VectorNode)                                  \
   declare_c2_type(SubVSNode, VectorNode)                                  \
   declare_c2_type(SubVINode, VectorNode)                                  \
@@ -2000,8 +2003,11 @@
   declare_c2_type(SubVDNode, VectorNode)                                  \
   declare_c2_type(MulVSNode, VectorNode)                                  \
   declare_c2_type(MulVINode, VectorNode)                                  \
+  declare_c2_type(MulReductionVINode, ReductionNode)                      \
   declare_c2_type(MulVFNode, VectorNode)                                  \
+  declare_c2_type(MulReductionVFNode, ReductionNode)                      \
   declare_c2_type(MulVDNode, VectorNode)                                  \
+  declare_c2_type(MulReductionVDNode, ReductionNode)                      \
   declare_c2_type(DivVFNode, VectorNode)                                  \
   declare_c2_type(DivVDNode, VectorNode)                                  \
   declare_c2_type(LShiftVBNode, VectorNode)                               \
@@ -2257,8 +2263,6 @@
   declare_constant(CollectedHeap::ParallelScavengeHeap)                   \
   declare_constant(CollectedHeap::G1CollectedHeap)                        \
                                                                           \
-  declare_constant(GenCollectedHeap::max_gens)                            \
-                                                                          \
   /* constants from Generation::Name enum */                              \
                                                                           \
   declare_constant(Generation::DefNew)                                    \
@@ -2516,6 +2520,7 @@
   declare_constant(Deoptimization::Reason_speculate_null_check)           \
   declare_constant(Deoptimization::Reason_rtm_state_change)               \
   declare_constant(Deoptimization::Reason_unstable_if)                    \
+  declare_constant(Deoptimization::Reason_unstable_fused_if)              \
   declare_constant(Deoptimization::Reason_tenured)                        \
   declare_constant(Deoptimization::Reason_LIMIT)                          \
   declare_constant(Deoptimization::Reason_RECORDED_LIMIT)                 \
--- a/hotspot/src/share/vm/services/memoryManager.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/services/memoryManager.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -53,20 +53,6 @@
   volatile instanceOop _memory_mgr_obj;
 
 public:
-  enum Name {
-    Abstract,
-    CodeCache,
-    Metaspace,
-    Copy,
-    MarkSweepCompact,
-    ParNew,
-    ConcurrentMarkSweep,
-    PSScavenge,
-    PSMarkSweep,
-    G1YoungGen,
-    G1OldGen
-  };
-
   MemoryManager();
 
   int num_memory_pools() const           { return _num_pools; }
@@ -80,7 +66,6 @@
   bool is_manager(instanceHandle mh)     { return mh() == _memory_mgr_obj; }
 
   virtual instanceOop get_memory_manager_instance(TRAPS);
-  virtual MemoryManager::Name kind()     { return MemoryManager::Abstract; }
   virtual bool is_gc_memory_manager()    { return false; }
   virtual const char* name() = 0;
 
@@ -98,7 +83,6 @@
   static GCMemoryManager* get_psMarkSweep_memory_manager();
   static GCMemoryManager* get_g1YoungGen_memory_manager();
   static GCMemoryManager* get_g1OldGen_memory_manager();
-
 };
 
 class CodeCacheMemoryManager : public MemoryManager {
@@ -106,16 +90,14 @@
 public:
   CodeCacheMemoryManager() : MemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::CodeCache; }
-  const char* name()         { return "CodeCacheManager"; }
+  const char* name() { return "CodeCacheManager"; }
 };
 
 class MetaspaceMemoryManager : public MemoryManager {
 public:
   MetaspaceMemoryManager() : MemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::Metaspace; }
-  const char *name()         { return "Metaspace Manager"; }
+  const char* name() { return "Metaspace Manager"; }
 };
 
 class GCStatInfo : public ResourceObj {
@@ -202,7 +184,6 @@
 
   void set_notification_enabled(bool enabled) { _notification_enabled = enabled; }
   bool is_notification_enabled() { return _notification_enabled; }
-  virtual MemoryManager::Name kind() = 0;
 };
 
 // These subclasses of GCMemoryManager are defined to include
@@ -213,8 +194,7 @@
 public:
   CopyMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::Copy; }
-  const char* name()         { return "Copy"; }
+  const char* name() { return "Copy"; }
 };
 
 class MSCMemoryManager : public GCMemoryManager {
@@ -222,9 +202,7 @@
 public:
   MSCMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::MarkSweepCompact; }
-  const char* name()         { return "MarkSweepCompact"; }
-
+  const char* name() { return "MarkSweepCompact"; }
 };
 
 class ParNewMemoryManager : public GCMemoryManager {
@@ -232,9 +210,7 @@
 public:
   ParNewMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::ParNew; }
-  const char* name()         { return "ParNew"; }
-
+  const char* name() { return "ParNew"; }
 };
 
 class CMSMemoryManager : public GCMemoryManager {
@@ -242,9 +218,7 @@
 public:
   CMSMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::ConcurrentMarkSweep; }
-  const char* name()         { return "ConcurrentMarkSweep";}
-
+  const char* name() { return "ConcurrentMarkSweep";}
 };
 
 class PSScavengeMemoryManager : public GCMemoryManager {
@@ -252,9 +226,7 @@
 public:
   PSScavengeMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::PSScavenge; }
-  const char* name()         { return "PS Scavenge"; }
-
+  const char* name() { return "PS Scavenge"; }
 };
 
 class PSMarkSweepMemoryManager : public GCMemoryManager {
@@ -262,8 +234,7 @@
 public:
   PSMarkSweepMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::PSMarkSweep; }
-  const char* name()         { return "PS MarkSweep"; }
+  const char* name() { return "PS MarkSweep"; }
 };
 
 class G1YoungGenMemoryManager : public GCMemoryManager {
@@ -271,8 +242,7 @@
 public:
   G1YoungGenMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::G1YoungGen; }
-  const char* name()         { return "G1 Young Generation"; }
+  const char* name() { return "G1 Young Generation"; }
 };
 
 class G1OldGenMemoryManager : public GCMemoryManager {
@@ -280,8 +250,7 @@
 public:
   G1OldGenMemoryManager() : GCMemoryManager() {}
 
-  MemoryManager::Name kind() { return MemoryManager::G1OldGen; }
-  const char* name()         { return "G1 Old Generation"; }
+  const char* name() { return "G1 Old Generation"; }
 };
 
 #endif // SHARE_VM_SERVICES_MEMORYMANAGER_HPP
--- a/hotspot/src/share/vm/services/memoryService.cpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/services/memoryService.cpp	Fri Apr 17 23:55:04 2015 +0000
@@ -126,9 +126,8 @@
   CollectorPolicy* policy = heap->collector_policy();
 
   assert(policy->is_generation_policy(), "Only support two generations");
-  guarantee(heap->n_gens() == 2, "Only support two-generation heap");
-
   GenCollectorPolicy* gen_policy = policy->as_generation_policy();
+  guarantee(gen_policy->number_of_generations() == 2, "Only support two-generation heap");
   if (gen_policy != NULL) {
     Generation::Name kind = gen_policy->young_gen_spec()->name();
     switch (kind) {
--- a/hotspot/src/share/vm/utilities/globalDefinitions.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/utilities/globalDefinitions.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -407,14 +407,6 @@
 
 // Machine dependent stuff
 
-#if defined(X86) && defined(COMPILER2) && !defined(JAVASE_EMBEDDED)
-// Include Restricted Transactional Memory lock eliding optimization
-#define INCLUDE_RTM_OPT 1
-#define RTM_OPT_ONLY(code) code
-#else
-#define INCLUDE_RTM_OPT 0
-#define RTM_OPT_ONLY(code)
-#endif
 // States of Restricted Transactional Memory usage.
 enum RTMState {
   NoRTM      = 0x2, // Don't use RTM
@@ -446,6 +438,15 @@
 # include "globalDefinitions_aarch64.hpp"
 #endif
 
+#ifndef INCLUDE_RTM_OPT
+#define INCLUDE_RTM_OPT 0
+#endif
+#if INCLUDE_RTM_OPT
+#define RTM_OPT_ONLY(code) code
+#else
+#define RTM_OPT_ONLY(code)
+#endif
+
 // To assure the IRIW property on processors that are not multiple copy
 // atomic, sync instructions must be issued between volatile reads to
 // assure their ordering, instead of after volatile stores.
@@ -1345,6 +1346,13 @@
   return (intptr_t) p;
 }
 
+// swap a & b
+template<class T> static void swap(T& a, T& b) {
+  T tmp = a;
+  a = b;
+  b = tmp;
+}
+
 // Printf-style formatters for fixed- and variable-width types as pointers and
 // integers.  These are derived from the definitions in inttypes.h.  If the platform
 // doesn't provide appropriate definitions, they should be provided in
--- a/hotspot/src/share/vm/utilities/stack.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/utilities/stack.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -96,11 +96,16 @@
 public:
   friend class StackIterator<E, F>;
 
+  // Number of elements that fit in 4K bytes minus the size of two pointers
+  // (link field and malloc header).
+  static const size_t _default_segment_size =  (4096 - 2 * sizeof(E*)) / sizeof(E);
+  static size_t default_segment_size() { return _default_segment_size; }
+
   // segment_size:    number of items per segment
   // max_cache_size:  maxmium number of *segments* to cache
   // max_size:        maximum number of items allowed, rounded to a multiple of
   //                  the segment size (0 == unlimited)
-  inline Stack(size_t segment_size = default_segment_size(),
+  inline Stack(size_t segment_size = _default_segment_size,
                size_t max_cache_size = 4, size_t max_size = 0);
   inline ~Stack() { clear(true); }
 
@@ -122,8 +127,6 @@
   // clear_cache is true, also release any cached segments.
   void clear(bool clear_cache = false);
 
-  static inline size_t default_segment_size();
-
 protected:
   // Each segment includes space for _seg_size elements followed by a link
   // (pointer) to the previous segment; the space is allocated as a single block
--- a/hotspot/src/share/vm/utilities/stack.inline.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/utilities/stack.inline.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -86,14 +86,6 @@
 }
 
 template <class E, MEMFLAGS F>
-size_t Stack<E, F>::default_segment_size()
-{
-  // Number of elements that fit in 4K bytes minus the size of two pointers
-  // (link field and malloc header).
-  return (4096 - 2 * sizeof(E*)) / sizeof(E);
-}
-
-template <class E, MEMFLAGS F>
 size_t Stack<E, F>::adjust_segment_size(size_t seg_size)
 {
   const size_t elem_sz = sizeof(E);
--- a/hotspot/src/share/vm/utilities/workgroup.hpp	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/src/share/vm/utilities/workgroup.hpp	Fri Apr 17 23:55:04 2015 +0000
@@ -340,18 +340,6 @@
   }
 };
 
-// Work gangs in garbage collectors: 2009-06-10
-//
-// SharedHeap - work gang for stop-the-world parallel collection.
-//   Used by
-//     ParNewGeneration
-//     CMSParRemarkTask
-//     CMSRefProcTaskExecutor
-//     G1CollectedHeap
-//     G1ParFinalCountTask
-// ConcurrentMark
-// CMSCollector
-
 // A class that acts as a synchronisation barrier. Workers enter
 // the barrier and must wait until all other workers have entered
 // before any of them may leave.
--- a/hotspot/test/TEST.groups	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/TEST.groups	Fri Apr 17 23:55:04 2015 +0000
@@ -394,6 +394,7 @@
 hotspot_gc = \
   sanity/ExecuteInternalVMTests.java \
   gc/ \
+  -gc/g1/TestGreyReclaimedHumongousObjects.java \
   -gc/metaspace/CompressedClassSpaceSizeInJmapHeap.java
 
 hotspot_gc_closed = \
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/arraycopy/TestArrayCopyBadReexec.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8073866
+ * @summary Fix for 8064703 may also cause stores between the allocation and arraycopy to be rexecuted after a deoptimization
+ * @run main/othervm -XX:-BackgroundCompilation -XX:-UseOnStackReplacement TestArrayCopyBadReexec
+ *
+ */
+
+public class TestArrayCopyBadReexec {
+
+    static int val;
+
+    static int[] m1(int[] src, int l) {
+        if (src == null) {
+            return null;
+        }
+        int[] dest = new int[10];
+        val++;
+        try {
+            System.arraycopy(src, 0, dest, 0, l);
+        } catch (IndexOutOfBoundsException npe) {
+        }
+        return dest;
+    }
+
+    static public void main(String[] args) {
+        int[] src = new int[10];
+        int[] res = null;
+        boolean success = true;
+
+        for (int i = 0; i < 20000; i++) {
+            m1(src, 10);
+        }
+
+        int val_before = val;
+
+        m1(src, -1);
+
+        if (val - val_before != 1) {
+            System.out.println("Bad increment: " + (val - val_before));
+            throw new RuntimeException("Test failed");
+        }
+    }
+}
--- a/hotspot/test/compiler/arraycopy/TestArrayCopyNoInit.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/compiler/arraycopy/TestArrayCopyNoInit.java	Fri Apr 17 23:55:04 2015 +0000
@@ -76,7 +76,7 @@
     static TestArrayCopyNoInit[] m5(Object[] src) {
         Object tmp = src[0];
         TestArrayCopyNoInit[] dest = new TestArrayCopyNoInit[10];
-        System.arraycopy(src, 0, dest, 0, 0);
+        System.arraycopy(src, 0, dest, 0, 10);
         return dest;
     }
 
@@ -110,7 +110,7 @@
     static H[] m6(Object[] src) {
         Object tmp = src[0];
         H[] dest = new H[10];
-        System.arraycopy(src, 0, dest, 0, 0);
+        System.arraycopy(src, 0, dest, 0, 10);
         return dest;
     }
 
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/arraycopy/TestArrayCopyOfStopped.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,64 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8074676
+ * @summary after guards in Arrays.copyOf() intrinsic, control may become top
+ * @run main/othervm -XX:-BackgroundCompilation -XX:-UseOnStackReplacement TestArrayCopyOfStopped
+ *
+ */
+
+import java.util.Arrays;
+
+public class TestArrayCopyOfStopped {
+    static class A {
+    }
+
+    static class B {
+    }
+
+    static final B[] array_of_bs = new B[10];
+    static final A[] array_of_as = new A[10];
+
+    static Object[] m1_helper(Object[] array, boolean flag) {
+        if (flag) {
+            return Arrays.copyOf(array, 10, A[].class);
+        }
+        return null;
+    }
+
+    static Object[] m1(boolean flag) {
+        return m1_helper(array_of_bs, flag);
+    }
+
+    public static void main(String[] args) {
+        for (int i = 0; i < 20000; i++) {
+            m1_helper(array_of_as, (i%2) == 0);
+        }
+
+        for (int i = 0; i < 20000; i++) {
+            m1(false);
+        }
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/arraycopy/TestArrayCopyStoppedAfterGuards.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,51 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8075921
+ * @summary control becomes top after arraycopy guards and confuses tighly coupled allocation logic
+ * @run main/othervm -Xcomp -XX:CompileOnly=TestArrayCopyStoppedAfterGuards.test,System.arraycopy TestArrayCopyStoppedAfterGuards
+ *
+ */
+
+public class TestArrayCopyStoppedAfterGuards {
+
+    static void test() {
+        Object src = new Object();
+        int[] dst = new int[10];
+        System.arraycopy(src, 0, dst, 0, 10);
+    }
+
+    static public void main(String[] args) {
+        // warmup
+        Object o = new Object();
+        int[] src = new int[10];
+        int[] dst = new int[10];
+        System.arraycopy(src, 0, dst, 0, 10);
+
+        try {
+            test();
+        } catch(ArrayStoreException ase) {}
+    }
+}
--- a/hotspot/test/compiler/codecache/jmx/PoolsIndependenceTest.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/compiler/codecache/jmx/PoolsIndependenceTest.java	Fri Apr 17 23:55:04 2015 +0000
@@ -37,7 +37,6 @@
 
 /*
  * @test PoolsIndependenceTest
- * @ignore 8068385
  * @library /testlibrary /../../test/lib
  * @build PoolsIndependenceTest
  * @run main ClassFileInstaller sun.hotspot.WhiteBox
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/inlining/DefaultMethodsDependencies.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,63 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8069263
+ * @summary Deoptimization between array allocation and arraycopy may result in non initialized array
+ * @run main/othervm -XX:-BackgroundCompilation -XX:CompileOnly=DefaultMethodsDependencies::test -XX:CompileOnly=DefaultMethodsDependencies$I2::m1 DefaultMethodsDependencies
+ *
+ */
+
+public class DefaultMethodsDependencies {
+
+    interface I1 {
+        void m1();
+        // triggers processing of default methods in C1
+        default void m2() {
+        }
+    }
+
+    interface I2 extends I1 {
+        // added to C2 as default method
+        default void m1() {
+        }
+    }
+
+    static abstract class C1 implements I1 {
+    }
+
+    static class C2 extends C1 implements I2 {
+    }
+
+    static void test(C1 obj) {
+        obj.m1();
+    }
+
+    static public void main(String[] args) {
+        C2 obj = new C2();
+        for (int i = 0; i < 20000; i++) {
+            test(obj);
+        }
+    }
+}
--- a/hotspot/test/compiler/intrinsics/multiplytolen/TestMultiplyToLen.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/compiler/intrinsics/multiplytolen/TestMultiplyToLen.java	Fri Apr 17 23:55:04 2015 +0000
@@ -34,6 +34,7 @@
  *      -XX:CompileCommand=inline,java.math.BigInteger::multiply TestMultiplyToLen
  */
 
+import java.util.Arrays;
 import java.util.Random;
 import java.math.*;
 
@@ -97,12 +98,36 @@
         newsum = newsum.add(newres);
 
         if (!bytecompare(oldres,newres)) {
+          System.out.println(b1);
+          System.out.println(b2);
+          System.out.print("mismatch for:b1:" + stringify(b1) + " :b2:" + stringify(b2) + " :oldres:" + stringify(oldres) + " :newres:" + stringify(newres));
+          throw new Exception("Failed");
+        }
+      }
+
+      // Test carry propagation.  Multiple carries during bignum
+      // multiplication are rare (especially when using 64-bit
+      // arithmetic) so we have to provoke them deliberately.
+      for (int j = 4; j <= 396; j += 4) {
+        byte[] bytes = new byte[j];
+        Arrays.fill(bytes, (byte)255);
+        b1 = new BigInteger(bytes);
+        b2 = new BigInteger(bytes);
+
+        oldres = base_multiply(b1,b2);
+        newres = new_multiply(b1,b2);
+
+        oldsum = oldsum.add(oldres);
+        newsum = newsum.add(newres);
+
+        if (!bytecompare(oldres,newres)) {
           System.out.print("mismatch for:b1:" + stringify(b1) + " :b2:" + stringify(b2) + " :oldres:" + stringify(oldres) + " :newres:" + stringify(newres));
           System.out.println(b1);
           System.out.println(b2);
           throw new Exception("Failed");
         }
       }
+
       if (!bytecompare(oldsum,newsum))  {
         System.out.println("Failure: oldsum:" + stringify(oldsum) + " newsum:" + stringify(newsum));
         throw new Exception("Failed");
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/intrinsics/unsafe/HeapByteBufferTest.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,371 @@
+//
+// Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2015, Red Hat Inc. All rights reserved.
+// DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+//
+// This code is free software; you can redistribute it and/or modify it
+// under the terms of the GNU General Public License version 2 only, as
+// published by the Free Software Foundation.
+//
+// This code is distributed in the hope that it will be useful, but WITHOUT
+// ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+// FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+// version 2 for more details (a copy is included in the LICENSE file that
+// accompanied this code).
+//
+// You should have received a copy of the GNU General Public License version
+// 2 along with this work; if not, write to the Free Software Foundation,
+// Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+//
+// Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+// or visit www.oracle.com if you need additional information or have any
+// questions.
+//
+//
+
+import com.oracle.java.testlibrary.Utils;
+import static java.lang.Math.abs;
+import java.nio.ByteBuffer;
+import java.nio.ByteOrder;
+import static java.nio.ByteOrder.BIG_ENDIAN;
+import static java.nio.ByteOrder.LITTLE_ENDIAN;
+import java.util.Random;
+import java.util.Arrays;
+
+/**
+ * @test
+ * @bug 8026049
+ * @library /testlibrary
+ * @run main/othervm -XX:+UnlockDiagnosticVMOptions -XX:-UseUnalignedAccesses HeapByteBufferTest
+ * @run main/othervm HeapByteBufferTest
+ * @summary Verify that byte buffers are correctly accessed.
+ */
+
+// A wrapper for a ByteBuffer which maintains a backing array and a
+// position.  Whenever this wrapper is written the backing array and
+// the wrapped byte buffer are updated together, and whenever it is
+// read we check that the ByteBuffer and the backing array are identical.
+
+class MyByteBuffer {
+    final ByteBuffer buf;
+    final byte[] bytes;
+    int pos;
+    ByteOrder byteOrder = BIG_ENDIAN;
+
+    MyByteBuffer(ByteBuffer buf, byte[] bytes) {
+        this.buf = buf;
+        this.bytes = Arrays.copyOf(bytes, bytes.length);
+        pos = 0;
+    }
+
+    public final MyByteBuffer order(ByteOrder bo) {
+        byteOrder = bo;
+        buf.order(bo);
+        return this;
+    }
+
+    static MyByteBuffer wrap(byte[] bytes) {
+        return new MyByteBuffer(ByteBuffer.wrap(bytes), bytes);
+    }
+
+    int capacity() { return bytes.length; }
+    int position() {
+        if (buf.position() != pos)
+            throw new RuntimeException();
+        return buf.position();
+    }
+
+    byte[] array() { return buf.array(); }
+    byte[] backingArray() { return bytes; }
+
+    private static byte long7(long x) { return (byte)(x >> 56); }
+    private static byte long6(long x) { return (byte)(x >> 48); }
+    private static byte long5(long x) { return (byte)(x >> 40); }
+    private static byte long4(long x) { return (byte)(x >> 32); }
+    private static byte long3(long x) { return (byte)(x >> 24); }
+    private static byte long2(long x) { return (byte)(x >> 16); }
+    private static byte long1(long x) { return (byte)(x >>  8); }
+    private static byte long0(long x) { return (byte)(x      ); }
+
+    private static byte int3(int x) { return (byte)(x >> 24); }
+    private static byte int2(int x) { return (byte)(x >> 16); }
+    private static byte int1(int x) { return (byte)(x >>  8); }
+    private static byte int0(int x) { return (byte)(x      ); }
+
+    private static byte short1(short x) { return (byte)(x >> 8); }
+    private static byte short0(short x) { return (byte)(x     ); }
+
+    byte _get(long i) { return bytes[(int)i]; }
+    void _put(long i, byte x) { bytes[(int)i] = x; }
+
+    private void putLongX(long a, long x) {
+        if (byteOrder == BIG_ENDIAN) {
+            x = Long.reverseBytes(x);
+        }
+        _put(a + 7, long7(x));
+        _put(a + 6, long6(x));
+        _put(a + 5, long5(x));
+        _put(a + 4, long4(x));
+        _put(a + 3, long3(x));
+        _put(a + 2, long2(x));
+        _put(a + 1, long1(x));
+        _put(a    , long0(x));
+    }
+
+    private void putIntX(long a, int x) {
+        if (byteOrder == BIG_ENDIAN) {
+            x = Integer.reverseBytes(x);
+        }
+        _put(a + 3, int3(x));
+        _put(a + 2, int2(x));
+        _put(a + 1, int1(x));
+        _put(a    , int0(x));
+    }
+
+    private void putShortX(int bi, short x) {
+        if (byteOrder == BIG_ENDIAN) {
+            x = Short.reverseBytes(x);
+        }
+        _put(bi    , short0(x));
+        _put(bi + 1, short1(x));
+    }
+
+    static private int makeInt(byte b3, byte b2, byte b1, byte b0) {
+        return (((b3       ) << 24) |
+                ((b2 & 0xff) << 16) |
+                ((b1 & 0xff) <<  8) |
+                ((b0 & 0xff)      ));
+    }
+    int getIntX(long a) {
+        int x = makeInt(_get(a + 3),
+                _get(a + 2),
+                _get(a + 1),
+                _get(a));
+        if (byteOrder == BIG_ENDIAN) {
+            x = Integer.reverseBytes(x);
+        }
+        return x;
+    }
+
+    static private long makeLong(byte b7, byte b6, byte b5, byte b4,
+                                 byte b3, byte b2, byte b1, byte b0)
+    {
+        return ((((long)b7       ) << 56) |
+                (((long)b6 & 0xff) << 48) |
+                (((long)b5 & 0xff) << 40) |
+                (((long)b4 & 0xff) << 32) |
+                (((long)b3 & 0xff) << 24) |
+                (((long)b2 & 0xff) << 16) |
+                (((long)b1 & 0xff) <<  8) |
+                (((long)b0 & 0xff)      ));
+    }
+
+    long getLongX(long a) {
+        long x = makeLong(_get(a + 7),
+                _get(a + 6),
+                _get(a + 5),
+                _get(a + 4),
+                _get(a + 3),
+                _get(a + 2),
+                _get(a + 1),
+                _get(a));
+        if (byteOrder == BIG_ENDIAN) {
+            x = Long.reverseBytes(x);
+        }
+        return x;
+    }
+
+    static private short makeShort(byte b1, byte b0) {
+        return (short)((b1 << 8) | (b0 & 0xff));
+    }
+
+    short getShortX(long a) {
+        short x = makeShort(_get(a + 1),
+                            _get(a    ));
+        if (byteOrder == BIG_ENDIAN) {
+            x = Short.reverseBytes(x);
+        }
+        return x;
+    }
+
+    double getDoubleX(long a) {
+        long x = getLongX(a);
+        return Double.longBitsToDouble(x);
+    }
+
+    double getFloatX(long a) {
+        int x = getIntX(a);
+        return Float.intBitsToFloat(x);
+    }
+
+    void ck(long x, long y) {
+        if (x != y) {
+            throw new RuntimeException(" x = " + Long.toHexString(x) + ", y = " + Long.toHexString(y));
+        }
+    }
+
+    void ck(double x, double y) {
+        if (x == x && y == y && x != y) {
+            ck(x, y);
+        }
+    }
+
+    long getLong(int i) { ck(buf.getLong(i), getLongX(i)); return buf.getLong(i); }
+    int getInt(int i) { ck(buf.getInt(i), getIntX(i)); return buf.getInt(i); }
+    short getShort(int i) { ck(buf.getShort(i), getShortX(i)); return buf.getShort(i); }
+    char getChar(int i) { ck(buf.getChar(i), (char)getShortX(i)); return buf.getChar(i); }
+    double getDouble(int i) { ck(buf.getDouble(i), getDoubleX(i)); return buf.getDouble(i); }
+    float getFloat(int i) { ck(buf.getFloat(i), getFloatX(i)); return buf.getFloat(i); }
+
+    void putLong(int i, long x) { buf.putLong(i, x); putLongX(i, x); }
+    void putInt(int i, int x) { buf.putInt(i, x); putIntX(i, x); }
+    void putShort(int i, short x) { buf.putShort(i, x); putShortX(i, x); }
+    void putChar(int i, char x) { buf.putChar(i, x); putShortX(i, (short)x); }
+    void putDouble(int i, double x) { buf.putDouble(i, x); putLongX(i, Double.doubleToRawLongBits(x)); }
+    void putFloat(int i, float x) { buf.putFloat(i, x); putIntX(i, Float.floatToRawIntBits(x)); }
+
+    long getLong() { ck(buf.getLong(buf.position()), getLongX(pos)); long x = buf.getLong(); pos += 8; return x; }
+    int getInt() { ck(buf.getInt(buf.position()), getIntX(pos)); int x = buf.getInt(); pos += 4; return x; }
+    short getShort() { ck(buf.getShort(buf.position()), getShortX(pos)); short x = buf.getShort(); pos += 2; return x; }
+    char getChar() {  ck(buf.getChar(buf.position()), (char)getShortX(pos)); char x = buf.getChar(); pos += 2; return x; }
+    double getDouble() { ck(buf.getDouble(buf.position()), getDoubleX(pos)); double x = buf.getDouble(); pos += 8; return x; }
+    float getFloat() { ck(buf.getFloat(buf.position()), getFloatX(pos)); float x = buf.getFloat(); pos += 4; return x; }
+
+    void putLong(long x) { putLongX(pos, x); pos += 8; buf.putLong(x); }
+    void putInt(int x) { putIntX(pos, x); pos += 4; buf.putInt(x); }
+    void putShort(short x) { putShortX(pos, x); pos += 2; buf.putShort(x); }
+    void putChar(char x) { putShortX(pos, (short)x); pos += 2; buf.putChar(x); }
+    void putDouble(double x) { putLongX(pos, Double.doubleToRawLongBits(x)); pos += 8; buf.putDouble(x); }
+    void putFloat(float x) { putIntX(pos, Float.floatToRawIntBits(x)); pos += 4; buf.putFloat(x); }
+
+    void rewind() { pos = 0; buf.rewind(); }
+}
+
+public class HeapByteBufferTest implements Runnable {
+
+    Random random = Utils.getRandomInstance();
+    MyByteBuffer data = MyByteBuffer.wrap(new byte[1024]);
+
+    int randomOffset(Random r, MyByteBuffer buf, int size) {
+        return r.nextInt(buf.capacity() - size);
+    }
+
+    long iterations;
+
+    HeapByteBufferTest(long iterations) {
+        this.iterations = iterations;
+    }
+
+    // The core of the test.  Walk over the buffer reading and writing
+    // random data, XORing it as we go.  We can detect writes in the
+    // wrong place, writes which are too long or too short, and reads
+    // or writes of the wrong data,
+    void step(Random r) {
+        data.order((r.nextInt() & 1) != 0 ? BIG_ENDIAN : LITTLE_ENDIAN);
+
+        data.rewind();
+        while (data.position() < data.capacity())
+            data.putLong(data.getLong() ^ random.nextLong());
+
+        data.rewind();
+        while (data.position() < data.capacity())
+            data.putInt(data.getInt() ^ random.nextInt());
+
+        data.rewind();
+        while (data.position() < data.capacity())
+            data.putShort((short)(data.getShort() ^ random.nextInt()));
+
+        data.rewind();
+        while (data.position() < data.capacity())
+            data.putChar((char)(data.getChar() ^ random.nextInt()));
+
+        data.rewind();
+        while (data.position() < data.capacity()) {
+            data.putDouble(combine(data.getDouble(), random.nextLong()));
+        }
+
+        data.rewind();
+        while (data.position() < data.capacity())
+            data.putFloat(combine(data.getFloat(), random.nextInt()));
+
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 8);
+            data.putLong(offset, data.getLong(offset) ^ random.nextLong());
+        }
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 4);
+            data.putInt(offset, data.getInt(offset) ^ random.nextInt());
+        }
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 2);
+            data.putShort(offset, (short)(data.getShort(offset) ^ random.nextInt()));
+        }
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 2);
+            data.putChar(offset, (char)(data.getChar(offset) ^ random.nextInt()));
+        }
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 8);
+            data.putDouble(offset, combine(data.getDouble(offset), random.nextLong()));
+        }
+        for (int i = 0; i < 100; i++) {
+            int offset = randomOffset(r, data, 4);
+            data.putFloat(offset, combine(data.getFloat(offset), random.nextInt()));
+        }
+    }
+
+    // XOR the bit pattern of a double and a long, returning the
+    // result as a double.
+    //
+    // We convert signalling NaNs to quiet NaNs.  We need to do this
+    // because some platforms (in particular legacy 80x87) do not
+    // provide transparent conversions between integer and
+    // floating-point types even when using raw conversions but
+    // quietly convert sNaN to qNaN.  This causes spurious test
+    // failures when the template interpreter uses 80x87 and the JITs
+    // use XMM registers.
+    //
+    public double combine(double prev, long bits) {
+        bits ^= Double.doubleToRawLongBits(prev);
+        double result = Double.longBitsToDouble(bits);
+        if (Double.isNaN(result)) {
+            result = Double.longBitsToDouble(bits | 0x8000000000000l);
+        }
+        return result;
+    }
+
+    // XOR the bit pattern of a float and an int, returning the result
+    // as a float.  Convert sNaNs to qNaNs.
+    public Float combine(float prev, int bits) {
+        bits ^= Float.floatToRawIntBits(prev);
+        Float result = Float.intBitsToFloat(bits);
+        if (Float.isNaN(result)) {
+            result = Float.intBitsToFloat(bits | 0x400000);
+        }
+        return result;
+    }
+
+    public void run() {
+        for (int i = 0; i < data.capacity(); i += 8) {
+            data.putLong(i, random.nextLong());
+        }
+
+        for (int i = 0; i < iterations; i++) {
+            step(random);
+        }
+
+        if (!Arrays.equals(data.array(), data.backingArray())) {
+            throw new RuntimeException();
+        }
+    }
+
+    public static void main(String[] args) {
+        // The number of iterations is high to ensure that tiered
+        // compilation kicks in all the way up to C2.
+        long iterations = 100000;
+        if (args.length > 0)
+            iterations = Long.parseLong(args[0]);
+
+        new HeapByteBufferTest(iterations).run();
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/jsr292/MHInlineTest.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,207 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+/**
+ * @test
+ * @bug 8062280
+ * @summary C2: inlining failure due to access checks being too strict
+ * @library /testlibrary
+ * @run main/othervm MHInlineTest
+ */
+import java.lang.invoke.*;
+import com.oracle.java.testlibrary.*;
+import static com.oracle.java.testlibrary.Asserts.*;
+
+public class MHInlineTest {
+    public static void main(String[] args) throws Exception {
+        ProcessBuilder pb = ProcessTools.createJavaProcessBuilder(
+                "-XX:+IgnoreUnrecognizedVMOptions", "-showversion",
+                "-server", "-XX:-TieredCompilation", "-Xbatch",
+                "-XX:+PrintCompilation", "-XX:+UnlockDiagnosticVMOptions", "-XX:+PrintInlining",
+                "-XX:CompileCommand=dontinline,MHInlineTest::test*",
+                    "MHInlineTest$Launcher");
+
+        OutputAnalyzer analyzer = new OutputAnalyzer(pb.start());
+
+        analyzer.shouldHaveExitValue(0);
+
+        // The test is applicable only to C2 (present in Server VM).
+        if (analyzer.getStderr().contains("Server VM")) {
+            analyzer.shouldContain("MHInlineTest$B::public_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain(   "MHInlineTest$B::protected_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain(   "MHInlineTest$B::package_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain("MHInlineTest$A::package_final_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain("MHInlineTest$B::private_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain("MHInlineTest$B::private_static_x (3 bytes)   inline (hot)");
+            analyzer.shouldContain("MHInlineTest$A::package_static_x (3 bytes)   inline (hot)");
+
+            analyzer.shouldNotContain("MHInlineTest$A::protected_x (3 bytes)   virtual call");
+            analyzer.shouldNotContain("MHInlineTest$A::package_x (3 bytes)   virtual call");
+        }
+    }
+
+    static class A {
+        public static final MethodHandles.Lookup LOOKUP = MethodHandles.lookup();
+
+        public Class<?>         public_x() { return A.class; }
+        protected Class<?>   protected_x() { return A.class; }
+        Class<?>               package_x() { return A.class; }
+        final Class<?>   package_final_x() { return A.class; }
+
+        static Class<?> package_static_x() { return A.class; }
+    }
+
+    static class B extends A {
+        public static final MethodHandles.Lookup LOOKUP = MethodHandles.lookup();
+
+        @Override public    Class<?>    public_x() { return B.class; }
+        @Override protected Class<?> protected_x() { return B.class; }
+        @Override Class<?>             package_x() { return B.class; }
+
+        private   Class<?>             private_x() { return B.class; }
+        static    Class<?>      private_static_x() { return B.class; }
+    }
+
+    static final MethodHandle A_PUBLIC_X;
+    static final MethodHandle A_PROTECTED_X;
+    static final MethodHandle A_PACKAGE_X;
+    static final MethodHandle A_PACKAGE_STATIC_X;
+    static final MethodHandle A_PACKAGE_FINAL_X;
+
+    static final MethodHandle B_PRIVATE_X;
+    static final MethodHandle B_PRIVATE_STATIC_X;
+
+    static {
+        try {
+            MethodHandles.Lookup LOOKUP = MethodHandles.lookup();
+
+            A_PUBLIC_X = LOOKUP.findVirtual(
+                    A.class, "public_x", MethodType.methodType(Class.class));
+            A_PROTECTED_X = LOOKUP.findVirtual(
+                    A.class, "protected_x", MethodType.methodType(Class.class));
+            A_PACKAGE_X = LOOKUP.findVirtual(
+                    A.class, "package_x", MethodType.methodType(Class.class));
+            A_PACKAGE_FINAL_X = LOOKUP.findVirtual(
+                    A.class, "package_final_x", MethodType.methodType(Class.class));
+            A_PACKAGE_STATIC_X = LOOKUP.findStatic(
+                    A.class, "package_static_x", MethodType.methodType(Class.class));
+
+            B_PRIVATE_X = B.LOOKUP.findVirtual(
+                    B.class, "private_x", MethodType.methodType(Class.class));
+            B_PRIVATE_STATIC_X = B.LOOKUP.findStatic(
+                    B.class, "private_static_x", MethodType.methodType(Class.class));
+        } catch (Exception e) {
+            throw new Error(e);
+        }
+    }
+
+    static final A a = new B();
+
+    private static void testPublicMH() {
+        try {
+            Class<?> r = (Class<?>)A_PUBLIC_X.invokeExact(a);
+            assertEquals(r, B.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testProtectedMH() {
+        try {
+            Class<?> r = (Class<?>)A_PROTECTED_X.invokeExact(a);
+            assertEquals(r, B.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testPackageMH() {
+        try {
+            Class<?> r = (Class<?>)A_PACKAGE_X.invokeExact(a);
+            assertEquals(r, B.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testPackageFinalMH() {
+        try {
+            Class<?> r = (Class<?>)A_PACKAGE_FINAL_X.invokeExact(a);
+            assertEquals(r, A.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testPackageStaticMH() {
+        try {
+            Class<?> r = (Class<?>)A_PACKAGE_STATIC_X.invokeExact();
+            assertEquals(r, A.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testPrivateMH() {
+        try {
+            Class<?> r = (Class<?>)B_PRIVATE_X.invokeExact((B)a);
+            assertEquals(r, B.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+
+    private static void testPrivateStaticMH() {
+        try {
+            Class<?> r = (Class<?>)B_PRIVATE_STATIC_X.invokeExact();
+            assertEquals(r, B.class);
+        } catch (Throwable throwable) {
+            throw new Error(throwable);
+        }
+    }
+    static class Launcher {
+        public static void main(String[] args) throws Exception {
+            for (int i = 0; i < 20_000; i++) {
+                testPublicMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testProtectedMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testPackageMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testPackageFinalMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testPackageStaticMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testPrivateMH();
+            }
+            for (int i = 0; i < 20_000; i++) {
+                testPrivateStaticMH();
+            }
+        }
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/ProdRed_Double.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,82 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar product reduction optimizations : float test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Double
+ */
+
+public class ProdRed_Double
+{
+  public static void main(String[] args) throws Exception {
+    double[] a = new double[256*1024];
+    double[] b = new double[256*1024];
+    prodReductionInit(a,b);
+    double valid = 2000;
+    double total = 0;
+    for(int j = 0; j < 2000; j++) {
+      total = j + 1;
+      total = prodReductionImplement(a,b, total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void prodReductionInit(double[] a, double[] b)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      a[i] = i + 2;
+      b[i] = i + 1;
+    }
+  }
+
+  public static double prodReductionImplement(double[] a, double[] b, double total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      total *= a[i] - b[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/ProdRed_Float.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,82 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar product reduction optimizations : float test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Float
+ */
+
+public class ProdRed_Float
+{
+  public static void main(String[] args) throws Exception {
+    float[] a = new float[256*1024];
+    float[] b = new float[256*1024];
+    prodReductionInit(a,b);
+    float valid = 2000;
+    float total = 0;
+    for(int j = 0; j < 2000; j++) {
+      total = j + 1;
+      total = prodReductionImplement(a,b, total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void prodReductionInit(float[] a, float[] b)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      a[i] = i + 2;
+      b[i] = i + 1;
+    }
+  }
+
+  public static float prodReductionImplement(float[] a, float[] b, float total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      total *= a[i] - b[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/ProdRed_Int.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,81 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar product reduction optimizations : int test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 ProdRed_Int
+ */
+
+public class ProdRed_Int
+{
+  public static void main(String[] args) throws Exception {
+    int[] a = new int[256*1024];
+    int[] b = new int[256*1024];
+    prodReductionInit(a,b);
+    int valid = 419430401;
+    int total = 1;
+    for(int j = 0; j < 2000; j++) {
+      total = prodReductionImplement(a,b,total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void prodReductionInit(int[] a, int[] b)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      a[i] = i + 2;
+      b[i] = i + 1;
+    }
+  }
+
+  public static int prodReductionImplement(int[] a, int[] b, int total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      total *= a[i] + b[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/SumRed_Double.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,96 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar sum reduction optimizations : double test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Double
+ */
+
+public class SumRed_Double
+{
+  public static void main(String[] args) throws Exception {
+    double[] a = new double[256*1024];
+    double[] b = new double[256*1024];
+    double[] c = new double[256*1024];
+    double[] d = new double[256*1024];
+    sumReductionInit(a,b,c);
+    double total = 0;
+    double valid = 3.6028590866691944E19;
+    for(int j = 0; j < 2000; j++) {
+      total = sumReductionImplement(a,b,c,d,total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void sumReductionInit(
+    double[] a,
+    double[] b,
+    double[] c)
+  {
+    for(int j = 0; j < 1; j++)
+    {
+      for(int i = 0; i < a.length; i++)
+      {
+        a[i] = i * 1 + j;
+        b[i] = i * 1 - j;
+        c[i] = i + j;
+      }
+    }
+  }
+
+  public static double sumReductionImplement(
+    double[] a,
+    double[] b,
+    double[] c,
+    double[] d,
+    double total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      d[i]= (a[i] * b[i]) + (a[i] * c[i]) + (b[i] * c[i]);
+      total += d[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/SumRed_Float.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,96 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar sum reduction optimizations : float test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Float
+ */
+
+public class SumRed_Float
+{
+  public static void main(String[] args) throws Exception {
+    float[] a = new float[256*1024];
+    float[] b = new float[256*1024];
+    float[] c = new float[256*1024];
+    float[] d = new float[256*1024];
+    sumReductionInit(a,b,c);
+    float total = 0;
+    float valid = (float)4.611686E18;
+    for(int j = 0; j < 2000; j++) {
+      total = sumReductionImplement(a,b,c,d,total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void sumReductionInit(
+    float[] a,
+    float[] b,
+    float[] c)
+  {
+    for(int j = 0; j < 1; j++)
+    {
+      for(int i = 0; i < a.length; i++)
+      {
+        a[i] = i * 1 + j;
+        b[i] = i * 1 - j;
+        c[i] = i + j;
+      }
+    }
+  }
+
+  public static float sumReductionImplement(
+    float[] a,
+    float[] b,
+    float[] c,
+    float[] d,
+    float total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      d[i]= (a[i] * b[i]) + (a[i] * c[i]) + (b[i] * c[i]);
+      total += d[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/loopopts/superword/SumRed_Int.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,96 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8074981
+ * @summary Add C2 x86 Superword support for scalar sum reduction optimizations : int test
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=2 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=4 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=8 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ *
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:+SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ * @run main/othervm -XX:+IgnoreUnrecognizedVMOptions -XX:-SuperWordReductions -XX:LoopUnrollLimit=250 -XX:LoopMaxUnroll=16 -XX:CompileThresholdScaling=0.1 SumRed_Int
+ */
+
+public class SumRed_Int
+{
+  public static void main(String[] args) throws Exception {
+    int[] a = new int[256*1024];
+    int[] b = new int[256*1024];
+    int[] c = new int[256*1024];
+    int[] d = new int[256*1024];
+    sumReductionInit(a,b,c);
+    int total = 0;
+    int valid = 262144000;
+    for(int j = 0; j < 2000; j++) {
+      total = sumReductionImplement(a,b,c,d,total);
+    }
+    if(total == valid) {
+      System.out.println("Success");
+    } else {
+      System.out.println("Invalid sum of elements variable in total: " + total);
+      System.out.println("Expected value = " + valid);
+      throw new Exception("Failed");
+    }
+  }
+
+  public static void sumReductionInit(
+    int[] a,
+    int[] b,
+    int[] c)
+  {
+    for(int j = 0; j < 1; j++)
+    {
+      for(int i = 0; i < a.length; i++)
+      {
+        a[i] = i * 1 + j;
+        b[i] = i * 1 - j;
+        c[i] = i + j;
+      }
+    }
+  }
+
+  public static int sumReductionImplement(
+    int[] a,
+    int[] b,
+    int[] c,
+    int[] d,
+    int total)
+  {
+    for(int i = 0; i < a.length; i++)
+    {
+      d[i]= (a[i] * b[i]) + (a[i] * c[i]) + (b[i] * c[i]);
+      total += d[i];
+    }
+    return total;
+  }
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/rangechecks/TestExplicitRangeChecks.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,596 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8073480
+ * @summary explicit range checks should be recognized by C2
+ * @library /testlibrary /../../test/lib /compiler/whitebox
+ * @build  TestExplicitRangeChecks
+ * @run main ClassFileInstaller sun.hotspot.WhiteBox
+ * @run main ClassFileInstaller com.oracle.java.testlibrary.Platform
+ * @run main/othervm -ea -Xmixed -Xbootclasspath/a:. -XX:+UnlockDiagnosticVMOptions -XX:+WhiteBoxAPI
+ *                   -XX:-BackgroundCompilation -XX:-UseOnStackReplacement -XX:CompileCommand=compileonly,TestExplicitRangeChecks.test* TestExplicitRangeChecks
+ *
+ */
+
+import java.lang.annotation.*;
+import java.lang.reflect.*;
+import java.util.*;
+import sun.hotspot.WhiteBox;
+import sun.hotspot.code.NMethod;
+import com.oracle.java.testlibrary.Platform;
+import sun.misc.Unsafe;
+
+public class TestExplicitRangeChecks {
+
+    static int[] array = new int[10];
+
+    @Retention(RetentionPolicy.RUNTIME)
+    @interface Args {
+        int[] compile();
+        int[] good();
+        int[] bad();
+        boolean deoptimize() default true;
+    }
+
+    // Should be compiled as a single unsigned comparison
+    // 0 <= index < array.length
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1, 10})
+    static boolean test1_1(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // same test but so we can compile with same optimization after trap in test1_1
+    static boolean test1_2(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // Shouldn't matter whether first or second test is the one
+    // against a constants
+    // 0 <= index < array.length
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1, 10})
+    static boolean test2_1(int index, int[] array) {
+        if (index >= array.length || index < 0) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test2_2(int index, int[] array) {
+        if (index >= array.length || index < 0) {
+            return false;
+        }
+        return true;
+    }
+
+    // 0 <= index <= array.length
+    @Args(compile = {5,}, good = {0, 10}, bad = {-1, 11})
+    static boolean test3_1(int index, int[] array) {
+        if (index < 0 || index > array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test3_2(int index, int[] array) {
+        if (index < 0 || index > array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // 0 <= index <= array.length
+    @Args(compile = {5,}, good = {0, 10}, bad = {-1, 11})
+    static boolean test4_1(int index, int[] array) {
+        if (index > array.length || index < 0 ) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test4_2(int index, int[] array) {
+        if (index > array.length || index < 0) {
+            return false;
+        }
+        return true;
+    }
+
+    static int[] test5_helper(int i) {
+        return (i < 100) ? new int[10] : new int[5];
+    }
+
+    // 0 < index < array.length
+    @Args(compile = {5,}, good = {1, 9}, bad = {0, 10})
+    static boolean test5_1(int index, int[] array) {
+        array = test5_helper(index); // array.length must be not constant greater than 1
+        if (index <= 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test5_2(int index, int[] array) {
+        array = test5_helper(index); // array.length must be not constant greater than 1
+        if (index <= 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // 0 < index < array.length
+    @Args(compile = {5,}, good = {1, 9}, bad = {0, 10})
+    static boolean test6_1(int index, int[] array) {
+        array = test5_helper(index); // array.length must be not constant greater than 1
+        if (index >= array.length || index <= 0 ) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test6_2(int index, int[] array) {
+        array = test5_helper(index); // array.length must be not constant greater than 1
+        if (index >= array.length || index <= 0) {
+            return false;
+        }
+        return true;
+    }
+
+    // 0 < index <= array.length
+    @Args(compile = {5,}, good = {1, 10}, bad = {0, 11})
+    static boolean test7_1(int index, int[] array) {
+        if (index <= 0 || index > array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test7_2(int index, int[] array) {
+        if (index <= 0 || index > array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // 0 < index <= array.length
+    @Args(compile = {5,}, good = {1, 10}, bad = {0, 11})
+    static boolean test8_1(int index, int[] array) {
+        if (index > array.length || index <= 0 ) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test8_2(int index, int[] array) {
+        if (index > array.length || index <= 0) {
+            return false;
+        }
+        return true;
+    }
+
+    static int[] test9_helper1(int i) {
+        return (i < 100) ? new int[1] : new int[2];
+    }
+
+    static int[] test9_helper2(int i) {
+        return (i < 100) ? new int[10] : new int[11];
+    }
+
+    // array1.length <= index < array2.length
+    @Args(compile = {5,}, good = {1, 9}, bad = {0, 10})
+    static boolean test9_1(int index, int[] array) {
+        int[] array1 = test9_helper1(index);
+        int[] array2 = test9_helper2(index);
+        if (index < array1.length || index >= array2.length) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test9_2(int index, int[] array) {
+        int[] array1 = test9_helper1(index);
+        int[] array2 = test9_helper2(index);
+        if (index < array1.length || index >= array2.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // Previously supported pattern
+    @Args(compile = {-5,5,15}, good = {0, 9}, bad = {-1, 10}, deoptimize=false)
+    static boolean test10_1(int index, int[] array) {
+        if (index < 0 || index >= 10) {
+            return false;
+        }
+        return true;
+    }
+
+    static int[] array11 = new int[10];
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1,})
+    static boolean test11_1(int index, int[] array) {
+        if (index < 0) {
+            return false;
+        }
+        int unused = array11[index];
+        // If this one is folded with the first test then we allow
+        // array access above to proceed even for out of bound array
+        // index and the method throws an
+        // ArrayIndexOutOfBoundsException.
+        if (index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    static int[] array12 = {10, 10, 10, 10, 10, 10, 10, 10, 10, 10};
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1,})
+    static boolean test12_1(int index, int[] array) {
+        // Cannot be folded otherwise would cause incorrect array
+        // access if the array12 range check is executed before the
+        // folded test.
+        if (index < 0 || index >= array12[index]) {
+            return false;
+        }
+        return true;
+    }
+
+    // Same as test1_1 but pass null array when index < 0: shouldn't
+    // cause NPE.
+    @Args(compile = {5,}, good = {0, 9}, bad = {})
+    static boolean test13_1(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // Same as test10 but with uncommon traps
+    @Args(compile = {5}, good = {0, 9}, bad = {-1, 10})
+    static boolean test14_1(int index, int[] array) {
+        if (index < 0 || index >= 10) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test14_2(int index, int[] array) {
+        if (index < 0 || index >= 10) {
+            return false;
+        }
+        return true;
+    }
+
+    // Same as test13_1 but pass null array: null trap should be reported on first if
+    @Args(compile = {5,}, good = {0, 9}, bad = {})
+    static boolean test15_1(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        return true;
+    }
+
+    // Same as test1 but with no null check between the integer comparisons
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1, 10})
+    static boolean test16_1(int index, int[] array) {
+        int l = array.length;
+        if (index < 0 || index >= l) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test16_2(int index, int[] array) {
+        int l = array.length;
+        if (index < 0 || index >= l) {
+            return false;
+        }
+        return true;
+    }
+
+    // Same as test1 but bound check on array access should optimize
+    // out.
+    @Args(compile = {5,}, good = {0, 9}, bad = {-1, 10})
+    static boolean test17_1(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        array[index] = 0;
+        return true;
+    }
+
+    static boolean test17_2(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        array[index] = 0;
+        return true;
+    }
+
+    // Same as test1 but range check smearing should optimize
+    // 3rd range check out.
+    @Args(compile = {5,}, good = {}, bad = {})
+    static boolean test18_1(int index, int[] array) {
+        if (index < 0 || index >= array.length) {
+            return false;
+        }
+        array[index+2] = 0;
+        array[index+1] = 0;
+        return true;
+    }
+
+    static boolean test19_helper1(int index) {
+        if (index < 12) {
+            return false;
+        }
+        return true;
+    }
+
+    static boolean test19_helper2(int index) {
+        if (index > 8) {
+            return false;
+        }
+        return true;
+    }
+
+    // Second test should be optimized out
+    static boolean test19(int index, int[] array) {
+        test19_helper1(index);
+        test19_helper2(index);
+        return true;
+    }
+
+    static boolean success = true;
+
+    private static final WhiteBox WHITE_BOX = WhiteBox.getWhiteBox();
+
+    final HashMap<String,Method> tests = new HashMap<>();
+    {
+        for (Method m : this.getClass().getDeclaredMethods()) {
+            if (m.getName().matches("test[0-9]+(_[0-9])?")) {
+                assert(Modifier.isStatic(m.getModifiers())) : m;
+                tests.put(m.getName(), m);
+            }
+        }
+    }
+
+    void doTest(String name) throws Exception {
+        Method m = tests.get(name + "_1");
+
+        Args anno =  m.getAnnotation(Args.class);
+        int[] compile = anno.compile();
+        int[] good = anno.good();
+        int[] bad = anno.bad();
+        boolean deoptimize = anno.deoptimize();
+
+        // Get compiled
+        for (int i = 0; i < 20000;) {
+            for (int j = 0; j < compile.length; j++) {
+                m.invoke(null, compile[j], array);
+                i++;
+            }
+        }
+
+        if (!WHITE_BOX.isMethodCompiled(m)) {
+            System.out.println(name + "_1 not compiled");
+            success = false;
+        }
+
+        // check that good values don't trigger exception or
+        // deoptimization
+        for (int i = 0; i < good.length; i++) {
+            boolean res = (boolean)m.invoke(null, good[i], array);
+
+            if (!res) {
+                System.out.println(name + " bad result for good input " + good[i]);
+                success = false;
+            }
+            if (!WHITE_BOX.isMethodCompiled(m)) {
+                System.out.println(name + " deoptimized on valid access");
+                success = false;
+            }
+        }
+
+        // check that bad values trigger exception and deoptimization
+        for (int i = 0; i < bad.length; i++) {
+            if (i > 0 && deoptimize) {
+                m = tests.get(name + "_" + (i+1));
+                for (int k = 0; k < 20000;) {
+                    for (int j = 0; j < compile.length; j++) {
+                        m.invoke(null, compile[j], array);
+                        k++;
+                    }
+                }
+                if (!WHITE_BOX.isMethodCompiled(m)) {
+                    System.out.println(name + ("_" + (i+1)) + " not compiled");
+                    success = false;
+                }
+            }
+
+            boolean res = (boolean)m.invoke(null, bad[i], array);
+
+            if (res) {
+                System.out.println(name + " bad result for bad input " + bad[i]);
+                success = false;
+            }
+            if (Platform.isServer()) {
+                if (deoptimize && WHITE_BOX.isMethodCompiled(m)) {
+                    System.out.println(name + " not deoptimized on invalid access");
+                    success = false;
+                } else if (!deoptimize && !WHITE_BOX.isMethodCompiled(m)) {
+                    System.out.println(name + " deoptimized on invalid access");
+                    success = false;
+                }
+            }
+        }
+
+    }
+
+    private static final Unsafe UNSAFE;
+
+    static {
+        try {
+            Field unsafeField = Unsafe.class.getDeclaredField("theUnsafe");
+            unsafeField.setAccessible(true);
+            UNSAFE = (Unsafe) unsafeField.get(null);
+        }
+        catch (Exception e) {
+            throw new AssertionError(e);
+        }
+    }
+
+    // On x64, int to long conversion should optimize away in address computation
+    static int test20(int[] a) {
+        int sum = 0;
+        for (int i = 0; i < a.length; i++) {
+            sum += test20_helper(a, i);
+        }
+        return sum;
+    }
+
+    static int test20_helper(int[] a, int i) {
+        if (i < 0 || i >= a.length)
+            throw new ArrayIndexOutOfBoundsException();
+
+        long address = (((long) i) << 2) + UNSAFE.ARRAY_INT_BASE_OFFSET;
+        return UNSAFE.getInt(a, address);
+    }
+
+    static int test21(int[] a) {
+        int sum = 0;
+        for (int i = 0; i < a.length; i++) {
+            sum += test20_helper(a, i);
+        }
+        return sum;
+    }
+
+    static int test21_helper(int[] a, int i) {
+        if (i < 0 || i >= a.length)
+            throw new ArrayIndexOutOfBoundsException();
+
+        long address = (((long) i) << 2) + UNSAFE.ARRAY_INT_BASE_OFFSET;
+        return UNSAFE.getIntVolatile(a, address);
+    }
+
+    static public void main(String[] args) throws Exception {
+
+        if (WHITE_BOX.getBooleanVMFlag("BackgroundCompilation")) {
+            throw new AssertionError("Background compilation enabled");
+        }
+
+        TestExplicitRangeChecks test = new TestExplicitRangeChecks();
+
+        test.doTest("test1");
+        test.doTest("test2");
+        test.doTest("test3");
+        test.doTest("test4");
+
+        // pollute branch profile
+        for (int i = 0; i < 10000; i++) {
+            test5_helper((i%2 == 0) ? 0 : 1000);
+        }
+
+        test.doTest("test5");
+        test.doTest("test6");
+        test.doTest("test7");
+        test.doTest("test8");
+
+        // pollute branch profile
+        for (int i = 0; i < 10000; i++) {
+            test9_helper1((i%2 == 0) ? 0 : 1000);
+            test9_helper2((i%2 == 0) ? 0 : 1000);
+        }
+
+        test.doTest("test9");
+        test.doTest("test10");
+        test.doTest("test11");
+        test.doTest("test12");
+
+        test.doTest("test13");
+        {
+            Method m = test.tests.get("test13_1");
+            for (int i = 0; i < 1; i++) {
+                test13_1(-1, null);
+                if (!WHITE_BOX.isMethodCompiled(m)) {
+                    break;
+                }
+            }
+        }
+        test.doTest("test13");
+        {
+            Method m = test.tests.get("test13_1");
+            for (int i = 0; i < 10; i++) {
+                test13_1(-1, null);
+                if (!WHITE_BOX.isMethodCompiled(m)) {
+                    break;
+                }
+            }
+        }
+
+        test.doTest("test14");
+
+        test.doTest("test15");
+        {
+            Method m = test.tests.get("test15_1");
+            for (int i = 0; i < 10; i++) {
+                try {
+                    test15_1(5, null);
+                } catch(NullPointerException npe) {}
+                if (!WHITE_BOX.isMethodCompiled(m)) {
+                    break;
+                }
+            }
+        }
+        test.doTest("test15");
+        test.doTest("test16");
+        test.doTest("test17");
+        test.doTest("test18");
+
+        for (int i = 0; i < 20000; i++) {
+            test19_helper1(20);
+            test19_helper2(5);
+        }
+
+        {
+            Method m = test.tests.get("test19");
+            WHITE_BOX.enqueueMethodForCompilation(m, CompilerWhiteBoxTest.COMP_LEVEL_FULL_OPTIMIZATION);
+        }
+
+        for (int i = 0; i < 20000; i++) {
+            test20(array);
+        }
+
+        for (int i = 0; i < 20000; i++) {
+            test21(array);
+        }
+
+        if (!success) {
+            throw new RuntimeException("some tests failed");
+        }
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/compiler/types/TestMeetExactConstantArrays.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,70 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test
+ * @bug 8075587
+ * @summary meet of 2 constant arrays result in bottom
+ * @run main/othervm TestMeetExactConstantArrays
+ *
+ */
+
+public class TestMeetExactConstantArrays {
+    public abstract static class NumbersHolder {
+        public Number[] getNumbers() {
+            return null;
+        }
+    }
+
+    public static class IntegersHolder extends NumbersHolder {
+        private final static Integer integers[] = { new Integer(1) };
+
+        public Number[] getNumbers() {
+            return integers;
+        }
+    }
+
+    public static class LongsHolder extends NumbersHolder {
+        private final static Long longs[] = { new Long(1) };
+
+        public Number[] getNumbers() {
+            return longs;
+        }
+    }
+
+    public static final void loopNumbers(NumbersHolder numbersHolder) {
+        Number[] numbers = numbersHolder.getNumbers();
+        for (int i = 0; i < numbers.length; i++) {
+            numbers[i].longValue();
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        for (int i = 0; i < 10000; i++) {
+            IntegersHolder integersHolder = new IntegersHolder();
+            LongsHolder longsHolder = new LongsHolder();
+            loopNumbers(integersHolder);
+            loopNumbers(longsHolder);
+        }
+    }
+}
--- a/hotspot/test/compiler/whitebox/DeoptimizeFramesTest.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/compiler/whitebox/DeoptimizeFramesTest.java	Fri Apr 17 23:55:04 2015 +0000
@@ -32,12 +32,12 @@
  * @run main/othervm -Xbootclasspath/a:. -XX:+UnlockDiagnosticVMOptions
  *                   -XX:+WhiteBoxAPI -Xmixed
  *                   -XX:CompileCommand=compileonly,DeoptimizeFramesTest$TestCaseImpl::method
- *                   -XX:+IgnoreUnexpectedVMOptions -XX:-DeoptimizeRandom -XX:-DeoptimizeALot
+ *                   -XX:+IgnoreUnrecognizedVMOptions -XX:-DeoptimizeRandom -XX:-DeoptimizeALot
  *                   DeoptimizeFramesTest true
  * @run main/othervm -Xbootclasspath/a:. -XX:+UnlockDiagnosticVMOptions
  *                   -XX:+WhiteBoxAPI -Xmixed
  *                   -XX:CompileCommand=compileonly,DeoptimizeFramesTest$TestCaseImpl::method
- *                   -XX:+IgnoreUnexpectedVMOptions -XX:-DeoptimizeRandom -XX:-DeoptimizeALot
+ *                   -XX:+IgnoreUnrecognizedVMOptions -XX:-DeoptimizeRandom -XX:-DeoptimizeALot
  *                   DeoptimizeFramesTest false
  * @summary testing of WB::deoptimizeFrames()
  */
--- a/hotspot/test/gc/arguments/TestInitialTenuringThreshold.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/gc/arguments/TestInitialTenuringThreshold.java	Fri Apr 17 23:55:04 2015 +0000
@@ -70,6 +70,7 @@
     runWithThresholds(10, 0, true);
     runWithThresholds(9, 8, true);
     runWithThresholds(-1, 8, true);
+    runWithThresholds(0, -1, true);
     runWithThresholds(8, -1, true);
     runWithThresholds(16, 8, true);
     runWithThresholds(8, 17, true);
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/gc/arguments/TestVerifyBeforeAndAfterGCFlags.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,119 @@
+/*
+* Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+*
+* This code is free software; you can redistribute it and/or modify it
+* under the terms of the GNU General Public License version 2 only, as
+* published by the Free Software Foundation.
+*
+* This code is distributed in the hope that it will be useful, but WITHOUT
+* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+* FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+* version 2 for more details (a copy is included in the LICENSE file that
+* accompanied this code).
+*
+* You should have received a copy of the GNU General Public License version
+* 2 along with this work; if not, write to the Free Software Foundation,
+* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+*
+* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+* or visit www.oracle.com if you need additional information or have any
+* questions.
+*/
+
+/*
+ * @test TestVerifyBeforeAndAfterGCFlags
+ * @key gc
+ * @bug 8000831
+ * @summary Runs an simple application (GarbageProducer) with various
+         combinations of -XX:{+|-}Verify{After|Before}GC flags and checks that
+         output contain or doesn't contain expected patterns
+ * @modules java.management
+ * @library /testlibrary
+ * @run driver TestVerifyBeforeAndAfterGCFlags
+ */
+
+import java.util.ArrayList;
+import java.util.Collections;
+
+import com.oracle.java.testlibrary.Utils;
+import com.oracle.java.testlibrary.OutputAnalyzer;
+import com.oracle.java.testlibrary.ProcessTools;
+
+public class TestVerifyBeforeAndAfterGCFlags {
+
+    // VerifyBeforeGC:[Verifying threads heap tenured eden syms strs zone dict metaspace chunks hand C-heap code cache ]
+    public static final String VERIFY_BEFORE_GC_PATTERN = "VerifyBeforeGC:\\[Verifying\\s+([^]\\s]+\\s+)+\\]";
+    // VerifyBeforeGC: VerifyBeforeGC: VerifyBeforeGC:
+    public static final String VERIFY_BEFORE_GC_CORRUPTED_PATTERN = "VerifyBeforeGC:(?!\\[Verifying[^]]+\\])";
+
+    // VerifyAfterGC:[Verifying threads heap tenured eden syms strs zone dict metaspace chunks hand C-heap code cache ]
+    public static final String VERIFY_AFTER_GC_PATTERN = "VerifyAfterGC:\\[Verifying\\s+([^]\\s]+\\s+)+\\]";
+    // VerifyAfterGC: VerifyAfterGC: VerifyAfterGC:
+    public static final String VERIFY_AFTER_GC_CORRUPTED_PATTERN = "VerifyAfterGC:(?!\\[Verifying[^]]+\\])";
+
+    public static void main(String args[]) throws Exception {
+        String[] filteredOpts = Utils.getFilteredTestJavaOpts(
+                                    new String[] { "-Xloggc:",
+                                                   "-XX:+UseGCLogFileRotation",
+                                                   "-XX:-DisplayVMOutput",
+                                                   "VerifyBeforeGC",
+                                                   "VerifyAfterGC" });
+        testVerifyFlags(false, false, filteredOpts);
+        testVerifyFlags(true,  true,  filteredOpts);
+        testVerifyFlags(true,  false, filteredOpts);
+        testVerifyFlags(false, true,  filteredOpts);
+    }
+
+    public static void testVerifyFlags(boolean verifyBeforeGC,
+                                       boolean verifyAfterGC,
+                                       String[] opts) throws Exception {
+        ArrayList<String> vmOpts = new ArrayList<>();
+        if (opts != null && (opts.length > 0)) {
+            Collections.addAll(vmOpts, opts);
+        }
+
+        Collections.addAll(vmOpts, new String[] {
+                                       "-Xmx5m",
+                                       "-Xms5m",
+                                       "-Xmn3m",
+                                       "-XX:+UnlockDiagnosticVMOptions",
+                                       (verifyBeforeGC ? "-XX:+VerifyBeforeGC"
+                                                       : "-XX:-VerifyBeforeGC"),
+                                       (verifyAfterGC ? "-XX:+VerifyAfterGC"
+                                                      : "-XX:-VerifyAfterGC"),
+                                       GarbageProducer.class.getName() });
+        ProcessBuilder procBuilder =
+            ProcessTools.createJavaProcessBuilder(vmOpts.toArray(
+                                                   new String[vmOpts.size()]));
+        OutputAnalyzer analyzer = new OutputAnalyzer(procBuilder.start());
+
+        analyzer.shouldHaveExitValue(0);
+        analyzer.shouldNotMatch(VERIFY_BEFORE_GC_CORRUPTED_PATTERN);
+        analyzer.shouldNotMatch(VERIFY_AFTER_GC_CORRUPTED_PATTERN);
+
+        if (verifyBeforeGC) {
+            analyzer.shouldMatch(VERIFY_BEFORE_GC_PATTERN);
+        } else {
+            analyzer.shouldNotMatch(VERIFY_BEFORE_GC_PATTERN);
+        }
+
+        if (verifyAfterGC) {
+            analyzer.shouldMatch(VERIFY_AFTER_GC_PATTERN);
+        } else {
+            analyzer.shouldNotMatch(VERIFY_AFTER_GC_PATTERN);
+        }
+    }
+
+    public static class GarbageProducer {
+        static long[][] garbage = new long[10][];
+
+        public static void main(String args[]) {
+            int j = 0;
+            for(int i = 0; i<1000; i++) {
+                garbage[j] = new long[10000];
+                j = (j+1)%garbage.length;
+            }
+        }
+    }
+}
--- a/hotspot/test/gc/g1/TestGCLogMessages.java	Fri Apr 17 16:12:16 2015 -0700
+++ b/hotspot/test/gc/g1/TestGCLogMessages.java	Fri Apr 17 23:55:04 2015 +0000
@@ -66,7 +66,6 @@
         new LogMessageWithLevel("SystemDictionary Roots", Level.FINEST),
         new LogMessageWithLevel("CLDG Roots", Level.FINEST),
         new LogMessageWithLevel("JVMTI Roots", Level.FINEST),
-        new LogMessageWithLevel("CodeCache Roots", Level.FINEST),
         new LogMessageWithLevel("SATB Filtering", Level.FINEST),
         new LogMessageWithLevel("CM RefProcessor Roots", Level.FINEST),
         new LogMessageWithLevel("Wait For Strong CLD", Level.FINEST),
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/gc/g1/TestGreyReclaimedHumongousObjects.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,176 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test TestGreyReclaimedHumongousObjects.java
+ * @bug 8069367
+ * @requires vm.gc == "G1" | vm.gc == "null"
+ * @summary Test handling of marked but unscanned reclaimed humongous objects.
+ * @key gc
+ * @run main/othervm -XX:+UseG1GC -Xss32m -Xmx128m -XX:G1HeapRegionSize=1m
+ *      -XX:+UnlockExperimentalVMOptions
+ *          -XX:+G1EagerReclaimHumongousObjects
+ *          -XX:+G1EagerReclaimHumongousObjectsWithStaleRefs
+ *      TestGreyReclaimedHumongousObjects 1048576 90
+ */
+
+// This test spawns a bunch of threads, each of them rapidly
+// allocating large objects and storing them into a circular buffer
+// associated with the thread.  The circular buffer results in these
+// objects becoming dead in fairly short order.
+//
+// The situation we're trying to provoke is
+//
+// (1) A humongous object H is marked and added to the mark stack.
+//
+// (2) An evacuation pause determines H is no longer live, and
+// reclaims it.  This occurs before concurrent marking has gotten
+// around to processing the mark stack entry for H.
+//
+// (3) Concurrent marking processes the mark stack entry for H.  The
+// bug is that it would attempt to scan the now dead object.
+//
+// Unfortunately, this test is *very* sensitive to configuration.
+// Among the parameters that affect whether / how often we'll get into
+// the desired situation within a reasonable amount of time are:
+//
+// - THREAD_COUNT: The number of allocating threads.
+//
+// - OLD_COUNT: The number of objects each thread keeps.
+//
+// - MAX_MEMORY: The maximum heap size.
+//
+// - G1HeapRegionSize
+//
+// - The size of the objects being allocated.
+//
+// The parameter values specified here:
+//
+// - THREAD_COUNT = 12
+// - OLD_COUNT == 4
+// - MAX_MEMORY == 128m
+// - G1HeapRegionSize = 1m
+// - Object size = 1048576 (2 regions after header overhead and roundup)
+//
+// seems to work well at provoking the desired state fairly quickly.
+// Even relatively small perturbations may change that.  The key
+// factors seem to be keeping the heap mostly full of live objects but
+// having them become dead fairly quickly.
+
+import java.util.Date;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.ThreadFactory;
+import java.util.concurrent.TimeUnit;
+import sun.management.ManagementFactoryHelper;
+import com.sun.management.HotSpotDiagnosticMXBean;
+import com.sun.management.VMOption;
+
+public class TestGreyReclaimedHumongousObjects {
+
+    static class NamedThreadFactory implements ThreadFactory {
+       private int threadNum = 0;
+
+       @Override
+       public Thread newThread(Runnable r) {
+         return new Thread(r, THREAD_NAME + (threadNum++));
+       }
+    }
+
+    static class Runner extends Thread {
+        private final Date startDate = new Date();
+        private final int obj_size;
+        private final Object[] old_garbage;
+        private int old_index = 0;
+
+        public Runner(int obj_size) {
+            this.obj_size = obj_size;
+            old_garbage = new Object[OLD_COUNT];
+        }
+
+        private void allocate_garbage() {
+            byte[] garbage = new byte[obj_size];
+            old_garbage[Math.abs(++old_index % OLD_COUNT)] = garbage;
+        }
+
+        @Override
+        public void run() {
+            try {
+                while (!isInterrupted()) {
+                    allocate_garbage();
+                    Thread.sleep(0); // Yield, to ensure interruptable.
+                }
+            } catch (InterruptedException e) {
+                System.out.println("Aborted after "
+                                   + (new Date().getTime() - startDate.getTime())
+                                   + " ms");
+                interrupt();
+            }
+        }
+    }
+
+    public static void main(String[] args) throws Exception {
+        HotSpotDiagnosticMXBean diagnostic = ManagementFactoryHelper.getDiagnosticMXBean();
+
+        System.out.println("Max memory= " + MAX_MEMORY + " bytes");
+
+        int obj_size = 0;
+        long seconds_to_run = 0;
+        if (args.length != 2) {
+            throw new RuntimeException("Object size argument must be supplied");
+        } else {
+            obj_size = Integer.parseInt(args[0]);
+            seconds_to_run = Integer.parseInt(args[1]);
+        }
+        System.out.println("Objects size= " + obj_size + " bytes");
+        System.out.println("Seconds to run=" + seconds_to_run);
+
+        int region_size =
+            Integer.parseInt(diagnostic.getVMOption("G1HeapRegionSize").getValue());
+        if (obj_size < (region_size / 2)) {
+            throw new RuntimeException("Object size " + obj_size +
+                                       " is not humongous with region size " + region_size);
+        }
+
+        ExecutorService executor =
+            Executors.newFixedThreadPool(THREAD_COUNT, new NamedThreadFactory());
+        System.out.println("Starting " + THREAD_COUNT + " threads");
+
+        for (int i = 0; i < THREAD_COUNT; i++) {
+            executor.execute(new Runner(obj_size));
+        }
+
+        Thread.sleep(seconds_to_run * 1000);
+        executor.shutdownNow();
+
+        if (!executor.awaitTermination(10, TimeUnit.SECONDS)) {
+            System.err.println("Thread pool did not terminate after 10 seconds after shutdown");
+        }
+    }
+
+    private static final long MAX_MEMORY = Runtime.getRuntime().maxMemory();
+    private static final int OLD_COUNT = 4;
+    private static final int THREAD_COUNT = 12;
+    private static final String THREAD_NAME = "TestGreyRH-";
+}
+
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/hotspot/test/gc/g1/TestLargePageUseForAuxMemory.java	Fri Apr 17 23:55:04 2015 +0000
@@ -0,0 +1,129 @@
+/*
+ * Copyright (c) 2015, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/*
+ * @test TestLargePageUseForAuxMemory.java
+ * @bug 8058354
+ * @key gc
+ * @library /testlibrary /../../test/lib
+ * @requires (vm.gc=="G1" | vm.gc=="null")
+ * @build TestLargePageUseForAuxMemory
+ * @run main ClassFileInstaller sun.hotspot.WhiteBox
+ *                              sun.hotspot.WhiteBox$WhiteBoxPermission
+ * @summary Test that auxiliary data structures are allocated using large pages if available.
+ * @run main/othervm -Xbootclasspath/a:. -XX:+UseG1GC -XX:+WhiteBoxAPI -XX:+IgnoreUnrecognizedVMOptions -XX:+UseLargePages TestLargePageUseForAuxMemory
+ */
+
+import com.oracle.java.testlibrary.*;
+import sun.hotspot.WhiteBox;
+
+public class TestLargePageUseForAuxMemory {
+    static final int HEAP_REGION_SIZE = 4 * 1024 * 1024;
+    static long largePageSize;
+    static long smallPageSize;
+
+    static void checkSmallTables(OutputAnalyzer output, long expectedPageSize) throws Exception {
+        output.shouldContain("G1 'Block offset table': pg_sz=" + expectedPageSize);
+        output.shouldContain("G1 'Card counts table': pg_sz=" + expectedPageSize);
+    }
+
+    static void checkBitmaps(OutputAnalyzer output, long expectedPageSize) throws Exception {
+        output.shouldContain("G1 'Prev Bitmap': pg_sz=" + expectedPageSize);
+        output.shouldContain("G1 'Next Bitmap': pg_sz=" + expectedPageSize);
+    }
+
+    static void testVM(long heapsize, boolean cardsShouldUseLargePages, boolean bitmapShouldUseLargePages) throws Exception {
+        ProcessBuilder pb;
+        // Test with large page enabled.
+        pb = ProcessTools.createJavaProcessBuilder("-XX:+UseG1GC",
+                                                   "-XX:G1HeapRegionSize=" + HEAP_REGION_SIZE,
+                                                   "-Xms" + 10 * HEAP_REGION_SIZE,
+                                                   "-Xmx" + heapsize,
+                                                   "-XX:+TracePageSizes",
+                                                   "-XX:+UseLargePages",
+                                                   "-XX:+IgnoreUnrecognizedVMOptions",  // there is on ObjectAlignmentInBytes in 32 bit builds
+                                                   "-XX:ObjectAlignmentInBytes=8",
+                                                   "-version");
+
+        OutputAnalyzer output = new OutputAnalyzer(pb.start());
+        checkSmallTables(output, (cardsShouldUseLargePages ? largePageSize : smallPageSize));
+        checkBitmaps(output, (bitmapShouldUseLargePages ? largePageSize : smallPageSize));
+        output.shouldHaveExitValue(0);
+
+        // Test with large page disabled.
+        pb = ProcessTools.createJavaProcessBuilder("-XX:+UseG1GC",
+                                                   "-XX:G1HeapRegionSize=" + HEAP_REGION_SIZE,
+                                                   "-Xms" + 10 * HEAP_REGION_SIZE,
+                                                   "-Xmx" + heapsize,
+                                                   "-XX:+TracePageSizes",
+                                                   "-XX:-UseLargePages",
+                                                   "-XX:+IgnoreUnrecognizedVMOptions",  // there is on ObjectAlignmentInBytes in 32 bit builds
+                                                   "-XX:ObjectAlignmentInBytes=8",
+                                                   "-version");
+
+        output = new OutputAnalyzer(pb.start());
+        checkSmallTables(output, smallPageSize);
+        checkBitmaps(output, smallPageSize);
+        output.shouldHaveExitValue(0);
+    }
+
+    public static void main(String[] args) throws Exception {
+        if (!Platform.isDebugBuild()) {
+            System.out.println("Skip tests on non-debug builds because the required option TracePageSizes is a debug-only option.");
+            return;
+        }
+
+        WhiteBox wb = WhiteBox.getWhiteBox();
+        smallPageSize = wb.getVMPageSize();
+        largePageSize = wb.getVMLargePageSize();
+
+        if (largePageSize == 0) {
+            System.out.println("Skip tests because large page support does not seem to be available on this platform.");
+            return;
+        }
+
+        // To get large pages for the card table etc. we need at least a 1G heap (with 4k page size).
+        // 32 bit systems will have problems reserving such an amount of contiguous space, so skip the
+        // test there.
+        if (!Platform.is32bit()) {
+            // Size that a single card covers.
+            final int cardSize = 512;
+
+            final long heapSizeForCardTableUsingLargePages = largePageSize * cardSize;
+
+            testVM(heapSizeForCardTableUsingLargePages, true, true);
+            testVM(heapSizeForCardTableUsingLargePages + HEAP_REGION_SIZE, true, true);
+            testVM(heapSizeForCardTableUsingLargePages - HEAP_REGION_SIZE, false, true);
+        }
+
+        // Minimum heap requirement to get large pages for bitmaps is 128M heap. This seems okay to test
+        // everywhere.
+        final int bitmapTranslationFactor = 8 * 8; // ObjectAlignmentInBytes * BitsPerByte
+        final long heapSizeForBitmapUsingLargePages = largePageSize * bitmapTranslationFactor;
+
+        testVM(heapSizeForBitmapUsingLargePages, false, true);
+        testVM(heapSizeForBitmapUsingLargePages + HEAP_REGION_SIZE, false, true);
+        testVM(heapSizeForBitmapUsingLargePages - HEAP_REGION_SIZE, false, false);
+    }
+}
+