Mergejdk8-b63

97 files changed, 7119 insertions, 1689 deletions

diff --git a/.hgtags b/.hgtags
index 4f9539a26..e4f8bc77a 100644
--- a/.hgtags
+++ b/.hgtags
@@ -288,3 +288,4 @@ b261523fe66c40a02968f0aa7e73602491bb3386 hs25-b05
 4547dc71db765276e027b0c2780b724bae0a07d3 jdk8-b61
 d0337c31c8be7716369b4e7c3bd5f352983c6a06 hs25-b06
 dccd40de8db1fa96f186e6179907818d75320440 jdk8-b62
+dc16fe422c535ecd4e9f80fb814a1bb9704da6f5 hs25-b07
diff --git a/agent/src/share/classes/sun/jvm/hotspot/interpreter/Bytecodes.java b/agent/src/share/classes/sun/jvm/hotspot/interpreter/Bytecodes.java
index 1ba20f79f..3f09ee30c 100644
--- a/agent/src/share/classes/sun/jvm/hotspot/interpreter/Bytecodes.java
+++ b/agent/src/share/classes/sun/jvm/hotspot/interpreter/Bytecodes.java
@@ -272,9 +272,10 @@ public class Bytecodes {
   public static final int _fast_aldc            = 229;
   public static final int _fast_aldc_w          = 230;
   public static final int _return_register_finalizer = 231;
-  public static final int _shouldnotreachhere   = 232; // For debugging
+  public static final int _invokehandle         = 232;
+  public static final int _shouldnotreachhere   = 233; // For debugging
 
-  public static final int number_of_codes       = 233;
+  public static final int number_of_codes       = 234;
 
   // Flag bits derived from format strings, can_trap, can_rewrite, etc.:
   // semantic flags:
@@ -787,20 +788,22 @@ public class Bytecodes {
     def(_fast_aaccess_0      , "fast_aaccess_0"      , "b_JJ" , null    , BasicType.getTObject() ,  1, true , _aload_0        );
     def(_fast_faccess_0      , "fast_faccess_0"      , "b_JJ" , null    , BasicType.getTObject() ,  1, true , _aload_0        );
 
-    def(_fast_iload          , "fast_iload"          , "bi"   , null    , BasicType.getTInt()    ,  1, false, _iload);
-    def(_fast_iload2         , "fast_iload2"         , "bi_i" , null    , BasicType.getTInt()    ,  2, false, _iload);
-    def(_fast_icaload        , "fast_icaload"        , "bi_"  , null    , BasicType.getTInt()    ,  0, false, _iload);
+    def(_fast_iload          , "fast_iload"          , "bi"   , null    , BasicType.getTInt()    ,  1, false, _iload          );
+    def(_fast_iload2         , "fast_iload2"         , "bi_i" , null    , BasicType.getTInt()    ,  2, false, _iload          );
+    def(_fast_icaload        , "fast_icaload"        , "bi_"  , null    , BasicType.getTInt()    ,  0, false, _iload          );
 
     // Faster method invocation.
-    def(_fast_invokevfinal   , "fast_invokevfinal"   , "bJJ"  , null    , BasicType.getTIllegal(), -1, true, _invokevirtual);
+    def(_fast_invokevfinal   , "fast_invokevfinal"   , "bJJ"  , null    , BasicType.getTIllegal(), -1, true, _invokevirtual   );
 
     def(_fast_linearswitch   , "fast_linearswitch"   , ""     , null    , BasicType.getTVoid()   , -1, false, _lookupswitch   );
     def(_fast_binaryswitch   , "fast_binaryswitch"   , ""     , null    , BasicType.getTVoid()   , -1, false, _lookupswitch   );
+    def(_fast_aldc           , "fast_aldc"           , "bj"   , null    , BasicType.getTObject(),   1, true,  _ldc            );
+    def(_fast_aldc_w         , "fast_aldc_w"         , "bJJ"  , null    , BasicType.getTObject(),   1, true,  _ldc_w          );
 
     def(_return_register_finalizer, "return_register_finalizer", "b"    , null    , BasicType.getTVoid()   , 0, true, _return );
 
-    def(_fast_aldc           , "fast_aldc"           , "bj"   , null    , BasicType.getTObject(),   1, true,  _ldc   );
-    def(_fast_aldc_w         , "fast_aldc_w"         , "bJJ"  , null    , BasicType.getTObject(),   1, true,  _ldc_w );
+    // special handling of signature-polymorphic methods
+    def(_invokehandle        , "invokehandle"        , "bJJ"  , null    , BasicType.getTIllegal(), -1, true, _invokevirtual   );
 
     def(_shouldnotreachhere  , "_shouldnotreachhere" , "b"    , null    , BasicType.getTVoid()   ,  0, false);
 
diff --git a/agent/src/share/classes/sun/jvm/hotspot/runtime/Bytes.java b/agent/src/share/classes/sun/jvm/hotspot/runtime/Bytes.java
index a8df44184..b0cc278b8 100644
--- a/agent/src/share/classes/sun/jvm/hotspot/runtime/Bytes.java
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/Bytes.java
@@ -30,24 +30,10 @@ import sun.jvm.hotspot.utilities.PlatformInfo;
 /** Encapsulates some byte-swapping operations defined in the VM */
 
 public class Bytes {
-  // swap if client platform is different from server's.
   private boolean swap;
 
   public Bytes(MachineDescription machDesc) {
-    String cpu = PlatformInfo.getCPU();
-    if (cpu.equals("sparc")) {
-      if (machDesc.isBigEndian()) {
-        swap = false;
-      } else {
-        swap = true;
-      }
-    } else { // intel
-      if (machDesc.isBigEndian()) {
-        swap = true;
-      } else {
-        swap = false;
-      }
-    }
+    swap = !machDesc.isBigEndian();
   }
 
   /** Should only swap if the hardware's underlying byte order is
diff --git a/agent/src/share/classes/sun/jvm/hotspot/tools/jcore/ByteCodeRewriter.java b/agent/src/share/classes/sun/jvm/hotspot/tools/jcore/ByteCodeRewriter.java
index 2a3fa0236..40dc912fd 100644
--- a/agent/src/share/classes/sun/jvm/hotspot/tools/jcore/ByteCodeRewriter.java
+++ b/agent/src/share/classes/sun/jvm/hotspot/tools/jcore/ByteCodeRewriter.java
@@ -29,6 +29,11 @@ import sun.jvm.hotspot.interpreter.*;
 import sun.jvm.hotspot.utilities.*;
 import sun.jvm.hotspot.debugger.*;
 import sun.jvm.hotspot.runtime.*;
+import java.security.AccessController;
+import java.security.PrivilegedAction;
+import java.security.AccessControlContext;
+import java.security.PrivilegedExceptionAction;
+import java.security.PrivilegedActionException;
 
 public class ByteCodeRewriter
 {
@@ -38,8 +43,20 @@ public class ByteCodeRewriter
     private byte[] code;
     private Bytes  bytes;
 
-    public static final boolean DEBUG = false;
     private static final int jintSize = 4;
+    public static final boolean DEBUG;
+
+    static {
+        String debug =  (String) AccessController.doPrivileged(
+            new PrivilegedAction() {
+                public Object run() {
+                    return System.getProperty("sun.jvm.hotspot.tools.jcore.ByteCodeRewriter.DEBUG");
+                }
+            }
+        );
+        DEBUG = (debug != null ? debug.equalsIgnoreCase("true") : false);
+    }
+
 
     protected void debugMessage(String message) {
         System.out.println(message);
@@ -54,6 +71,18 @@ public class ByteCodeRewriter
 
     }
 
+    protected short getConstantPoolIndexFromRefMap(int rawcode, int bci) {
+        int refIndex;
+        String fmt = Bytecodes.format(rawcode);
+        switch (fmt.length()) {
+            case 2: refIndex = 0xFF & method.getBytecodeByteArg(bci); break;
+            case 3: refIndex = 0xFFFF & bytes.swapShort(method.getBytecodeShortArg(bci)); break;
+            default: throw new IllegalArgumentException();
+        }
+
+        return (short)cpool.objectToCPIndex(refIndex);
+     }
+
     protected short getConstantPoolIndex(int rawcode, int bci) {
        // get ConstantPool index from ConstantPoolCacheIndex at given bci
        String fmt = Bytecodes.format(rawcode);
@@ -95,6 +124,12 @@ public class ByteCodeRewriter
         int hotspotcode = Bytecodes._illegal;
         int len = 0;
 
+        if (DEBUG) {
+            String msg = method.getMethodHolder().getName().asString() + "." +
+                         method.getName().asString() +
+                         method.getSignature().asString();
+            debugMessage(msg);
+        }
         for (int bci = 0; bci < code.length;) {
             hotspotcode = Bytecodes.codeAt(method, bci);
             bytecode = Bytecodes.javaCode(hotspotcode);
@@ -133,15 +168,15 @@ public class ByteCodeRewriter
 
                 case Bytecodes._ldc_w:
                     if (hotspotcode != bytecode) {
-                        // fast_aldc_w puts constant in CP cache
-                        cpoolIndex = getConstantPoolIndex(hotspotcode, bci + 1);
+                        // fast_aldc_w puts constant in reference map
+                        cpoolIndex = getConstantPoolIndexFromRefMap(hotspotcode, bci + 1);
                         writeShort(code, bci + 1, cpoolIndex);
                     }
                     break;
                 case Bytecodes._ldc:
                     if (hotspotcode != bytecode) {
-                        // fast_aldc puts constant in CP cache
-                        cpoolIndex = getConstantPoolIndex(hotspotcode, bci + 1);
+                        // fast_aldc puts constant in reference map
+                        cpoolIndex = getConstantPoolIndexFromRefMap(hotspotcode, bci + 1);
                         code[bci + 1] = (byte)(cpoolIndex);
                     }
                     break;
diff --git a/make/excludeSrc.make b/make/excludeSrc.make
index 00a1e3a70..721aea5c7 100644
--- a/make/excludeSrc.make
+++ b/make/excludeSrc.make
@@ -79,10 +79,10 @@ ifeq ($(INCLUDE_ALTERNATE_GCS), false)
       CXXFLAGS += -DSERIALGC
       CFLAGS += -DSERIALGC
       Src_Files_EXCLUDE += \
-	binaryTreeDictionary.cpp cmsAdaptiveSizePolicy.cpp cmsCollectorPolicy.cpp \
+	cmsAdaptiveSizePolicy.cpp cmsCollectorPolicy.cpp \
 	cmsGCAdaptivePolicyCounters.cpp cmsLockVerifier.cpp cmsPermGen.cpp compactibleFreeListSpace.cpp \
-	concurrentMarkSweepGeneration.cpp concurrentMarkSweepThread.cpp freeBlockDictionary.cpp \
-	freeChunk.cpp freeList.cpp promotionInfo.cpp vmCMSOperations.cpp collectionSetChooser.cpp \
+	concurrentMarkSweepGeneration.cpp concurrentMarkSweepThread.cpp \
+	freeChunk.cpp adaptiveFreeList.cpp promotionInfo.cpp vmCMSOperations.cpp collectionSetChooser.cpp \
 	concurrentG1Refine.cpp concurrentG1RefineThread.cpp concurrentMark.cpp concurrentMarkThread.cpp \
 	dirtyCardQueue.cpp g1AllocRegion.cpp g1BlockOffsetTable.cpp g1CollectedHeap.cpp g1GCPhaseTimes.cpp \
 	g1CollectorPolicy.cpp g1ErgoVerbose.cpp g1_globals.cpp g1HRPrinter.cpp g1MarkSweep.cpp \
diff --git a/make/hotspot_version b/make/hotspot_version
index 14577a128..1f72e227a 100644
--- a/make/hotspot_version
+++ b/make/hotspot_version
@@ -35,7 +35,7 @@ HOTSPOT_VM_COPYRIGHT=Copyright 2012
 
 HS_MAJOR_VER=25
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=06
+HS_BUILD_NUMBER=07
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=8
diff --git a/src/cpu/x86/vm/assembler_x86.cpp b/src/cpu/x86/vm/assembler_x86.cpp
index 378c1f8c6..6b9677d30 100644
--- a/src/cpu/x86/vm/assembler_x86.cpp
+++ b/src/cpu/x86/vm/assembler_x86.cpp
@@ -1007,6 +1007,67 @@ void Assembler::addss(XMMRegister dst, Address src) {
   emit_simd_arith(0x58, dst, src, VEX_SIMD_F3);
 }
 
+void Assembler::aesdec(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_aes(), "");
+  InstructionMark im(this);
+  simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xde);
+  emit_operand(dst, src);
+}
+
+void Assembler::aesdec(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_aes(), "");
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xde);
+  emit_byte(0xC0 | encode);
+}
+
+void Assembler::aesdeclast(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_aes(), "");
+  InstructionMark im(this);
+  simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdf);
+  emit_operand(dst, src);
+}
+
+void Assembler::aesdeclast(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_aes(), "");
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdf);
+  emit_byte(0xC0 | encode);
+}
+
+void Assembler::aesenc(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_aes(), "");
+  InstructionMark im(this);
+  simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdc);
+  emit_operand(dst, src);
+}
+
+void Assembler::aesenc(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_aes(), "");
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdc);
+  emit_byte(0xC0 | encode);
+}
+
+void Assembler::aesenclast(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_aes(), "");
+  InstructionMark im(this);
+  simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdd);
+  emit_operand(dst, src);
+}
+
+void Assembler::aesenclast(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_aes(), "");
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0xdd);
+  emit_byte(0xC0 | encode);
+}
+
+
 void Assembler::andl(Address dst, int32_t imm32) {
   InstructionMark im(this);
   prefix(dst);
@@ -2307,6 +2368,22 @@ void Assembler::prefix(Prefix p) {
   a_byte(p);
 }
 
+void Assembler::pshufb(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_ssse3(), "");
+  int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0x00);
+  emit_byte(0xC0 | encode);
+}
+
+void Assembler::pshufb(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_ssse3(), "");
+  assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
+  InstructionMark im(this);
+  simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
+  emit_byte(0x00);
+  emit_operand(dst, src);
+}
+
 void Assembler::pshufd(XMMRegister dst, XMMRegister src, int mode) {
   assert(isByte(mode), "invalid value");
   NOT_LP64(assert(VM_Version::supports_sse2(), ""));
@@ -8067,6 +8144,15 @@ void MacroAssembler::movptr(Address dst, Register src) {
   LP64_ONLY(movq(dst, src)) NOT_LP64(movl(dst, src));
 }
 
+void MacroAssembler::movdqu(XMMRegister dst, AddressLiteral src) {
+  if (reachable(src)) {
+    Assembler::movdqu(dst, as_Address(src));
+  } else {
+    lea(rscratch1, src);
+    Assembler::movdqu(dst, Address(rscratch1, 0));
+  }
+}
+
 void MacroAssembler::movsd(XMMRegister dst, AddressLiteral src) {
   if (reachable(src)) {
     Assembler::movsd(dst, as_Address(src));
@@ -8357,6 +8443,17 @@ void MacroAssembler::xorps(XMMRegister dst, AddressLiteral src) {
   }
 }
 
+void MacroAssembler::pshufb(XMMRegister dst, AddressLiteral src) {
+  // Used in sign-bit flipping with aligned address.
+  assert((UseAVX > 0) || (((intptr_t)src.target() & 15) == 0), "SSE mode requires address alignment 16 bytes");
+  if (reachable(src)) {
+    Assembler::pshufb(dst, as_Address(src));
+  } else {
+    lea(rscratch1, src);
+    Assembler::pshufb(dst, Address(rscratch1, 0));
+  }
+}
+
 // AVX 3-operands instructions
 
 void MacroAssembler::vaddsd(XMMRegister dst, XMMRegister nds, AddressLiteral src) {
diff --git a/src/cpu/x86/vm/assembler_x86.hpp b/src/cpu/x86/vm/assembler_x86.hpp
index c936e13f5..8a9bbaf42 100644
--- a/src/cpu/x86/vm/assembler_x86.hpp
+++ b/src/cpu/x86/vm/assembler_x86.hpp
@@ -875,6 +875,17 @@ private:
   void addss(XMMRegister dst, Address src);
   void addss(XMMRegister dst, XMMRegister src);
 
+  // AES instructions
+  void aesdec(XMMRegister dst, Address src);
+  void aesdec(XMMRegister dst, XMMRegister src);
+  void aesdeclast(XMMRegister dst, Address src);
+  void aesdeclast(XMMRegister dst, XMMRegister src);
+  void aesenc(XMMRegister dst, Address src);
+  void aesenc(XMMRegister dst, XMMRegister src);
+  void aesenclast(XMMRegister dst, Address src);
+  void aesenclast(XMMRegister dst, XMMRegister src);
+
+
   void andl(Address  dst, int32_t imm32);
   void andl(Register dst, int32_t imm32);
   void andl(Register dst, Address src);
@@ -1424,6 +1435,10 @@ private:
   void prefetcht2(Address src);
   void prefetchw(Address src);
 
+  // Shuffle Bytes
+  void pshufb(XMMRegister dst, XMMRegister src);
+  void pshufb(XMMRegister dst, Address src);
+
   // Shuffle Packed Doublewords
   void pshufd(XMMRegister dst, XMMRegister src, int mode);
   void pshufd(XMMRegister dst, Address src,     int mode);
@@ -2611,6 +2626,12 @@ public:
   void divss(XMMRegister dst, Address src)        { Assembler::divss(dst, src); }
   void divss(XMMRegister dst, AddressLiteral src);
 
+  // Move Unaligned Double Quadword
+  void movdqu(Address     dst, XMMRegister src)   { Assembler::movdqu(dst, src); }
+  void movdqu(XMMRegister dst, Address src)       { Assembler::movdqu(dst, src); }
+  void movdqu(XMMRegister dst, XMMRegister src)   { Assembler::movdqu(dst, src); }
+  void movdqu(XMMRegister dst, AddressLiteral src);
+
   void movsd(XMMRegister dst, XMMRegister src) { Assembler::movsd(dst, src); }
   void movsd(Address dst, XMMRegister src)     { Assembler::movsd(dst, src); }
   void movsd(XMMRegister dst, Address src)     { Assembler::movsd(dst, src); }
@@ -2658,6 +2679,10 @@ public:
   void xorps(XMMRegister dst, Address src)     { Assembler::xorps(dst, src); }
   void xorps(XMMRegister dst, AddressLiteral src);
 
+  // Shuffle Bytes
+  void pshufb(XMMRegister dst, XMMRegister src) { Assembler::pshufb(dst, src); }
+  void pshufb(XMMRegister dst, Address src)     { Assembler::pshufb(dst, src); }
+  void pshufb(XMMRegister dst, AddressLiteral src);
   // AVX 3-operands instructions
 
   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { Assembler::vaddsd(dst, nds, src); }
diff --git a/src/cpu/x86/vm/stubGenerator_x86_32.cpp b/src/cpu/x86/vm/stubGenerator_x86_32.cpp
index f149fde83..d8b61e0b2 100644
--- a/src/cpu/x86/vm/stubGenerator_x86_32.cpp
+++ b/src/cpu/x86/vm/stubGenerator_x86_32.cpp
@@ -2137,6 +2137,529 @@ class StubGenerator: public StubCodeGenerator {
     }
   }
 
+  // AES intrinsic stubs
+  enum {AESBlockSize = 16};
+
+  address generate_key_shuffle_mask() {
+    __ align(16);
+    StubCodeMark mark(this, "StubRoutines", "key_shuffle_mask");
+    address start = __ pc();
+    __ emit_data(0x00010203, relocInfo::none, 0 );
+    __ emit_data(0x04050607, relocInfo::none, 0 );
+    __ emit_data(0x08090a0b, relocInfo::none, 0 );
+    __ emit_data(0x0c0d0e0f, relocInfo::none, 0 );
+    return start;
+  }
+
+  // Utility routine for loading a 128-bit key word in little endian format
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void load_key(XMMRegister xmmdst, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    __ movdqu(xmmdst, Address(key, offset));
+    if (xmm_shuf_mask != NULL) {
+      __ pshufb(xmmdst, xmm_shuf_mask);
+    } else {
+      __ pshufb(xmmdst, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    }
+  }
+
+  // aesenc using specified key+offset
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void aes_enc_key(XMMRegister xmmdst, XMMRegister xmmtmp, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    load_key(xmmtmp, key, offset, xmm_shuf_mask);
+    __ aesenc(xmmdst, xmmtmp);
+  }
+
+  // aesdec using specified key+offset
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void aes_dec_key(XMMRegister xmmdst, XMMRegister xmmtmp, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    load_key(xmmtmp, key, offset, xmm_shuf_mask);
+    __ aesdec(xmmdst, xmmtmp);
+  }
+
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //
+  address generate_aescrypt_encryptBlock() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "aescrypt_encryptBlock");
+    Label L_doLast;
+    address start = __ pc();
+
+    const Register from        = rsi;      // source array address
+    const Register to          = rdx;      // destination array address
+    const Register key         = rcx;      // key array address
+    const Register keylen      = rax;
+    const Address  from_param(rbp, 8+0);
+    const Address  to_param  (rbp, 8+4);
+    const Address  key_param (rbp, 8+8);
+
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    const XMMRegister xmm_key_shuf_mask = xmm2;
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+    __ push(rsi);
+    __ movptr(from , from_param);
+    __ movptr(to   , to_param);
+    __ movptr(key  , key_param);
+
+    __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    // keylen = # of 32-bit words, convert to 128-bit words
+    __ shrl(keylen, 2);
+    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
+
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    __ movdqu(xmm_result, Address(from, 0));  // get 16 bytes of input
+
+    // For encryption, the java expanded key ordering is just what we need
+
+    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
+    __ pxor(xmm_result, xmm_temp);
+    for (int offset = 0x10; offset <= 0x90; offset += 0x10) {
+      aes_enc_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
+    }
+    load_key  (xmm_temp, key, 0xa0, xmm_key_shuf_mask);
+    __ cmpl(keylen, 0);
+    __ jcc(Assembler::equal, L_doLast);
+    __ aesenc(xmm_result, xmm_temp);                   // only in 192 and 256 bit keys
+    aes_enc_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
+    load_key(xmm_temp, key, 0xc0, xmm_key_shuf_mask);
+    __ subl(keylen, 2);
+    __ jcc(Assembler::equal, L_doLast);
+    __ aesenc(xmm_result, xmm_temp);                   // only in 256 bit keys
+    aes_enc_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
+    load_key(xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+
+    __ BIND(L_doLast);
+    __ aesenclast(xmm_result, xmm_temp);
+    __ movdqu(Address(to, 0), xmm_result);        // store the result
+    __ xorptr(rax, rax); // return 0
+    __ pop(rsi);
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+    return start;
+  }
+
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //
+  address generate_aescrypt_decryptBlock() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "aescrypt_decryptBlock");
+    Label L_doLast;
+    address start = __ pc();
+
+    const Register from        = rsi;      // source array address
+    const Register to          = rdx;      // destination array address
+    const Register key         = rcx;      // key array address
+    const Register keylen      = rax;
+    const Address  from_param(rbp, 8+0);
+    const Address  to_param  (rbp, 8+4);
+    const Address  key_param (rbp, 8+8);
+
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    const XMMRegister xmm_key_shuf_mask = xmm2;
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+    __ push(rsi);
+    __ movptr(from , from_param);
+    __ movptr(to   , to_param);
+    __ movptr(key  , key_param);
+
+    __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    // keylen = # of 32-bit words, convert to 128-bit words
+    __ shrl(keylen, 2);
+    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
+
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    __ movdqu(xmm_result, Address(from, 0));
+
+    // for decryption java expanded key ordering is rotated one position from what we want
+    // so we start from 0x10 here and hit 0x00 last
+    // we don't know if the key is aligned, hence not using load-execute form
+    load_key(xmm_temp, key, 0x10, xmm_key_shuf_mask);
+    __ pxor  (xmm_result, xmm_temp);
+    for (int offset = 0x20; offset <= 0xa0; offset += 0x10) {
+      aes_dec_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
+    }
+    __ cmpl(keylen, 0);
+    __ jcc(Assembler::equal, L_doLast);
+    // only in 192 and 256 bit keys
+    aes_dec_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xc0, xmm_key_shuf_mask);
+    __ subl(keylen, 2);
+    __ jcc(Assembler::equal, L_doLast);
+    // only in 256 bit keys
+    aes_dec_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+
+    __ BIND(L_doLast);
+    // for decryption the aesdeclast operation is always on key+0x00
+    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
+    __ aesdeclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, 0), xmm_result);  // store the result
+
+    __ xorptr(rax, rax); // return 0
+    __ pop(rsi);
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+    return start;
+  }
+
+  void handleSOERegisters(bool saving) {
+    const int saveFrameSizeInBytes = 4 * wordSize;
+    const Address saved_rbx     (rbp, -3 * wordSize);
+    const Address saved_rsi     (rbp, -2 * wordSize);
+    const Address saved_rdi     (rbp, -1 * wordSize);
+
+    if (saving) {
+      __ subptr(rsp, saveFrameSizeInBytes);
+      __ movptr(saved_rsi, rsi);
+      __ movptr(saved_rdi, rdi);
+      __ movptr(saved_rbx, rbx);
+    } else {
+      // restoring
+      __ movptr(rsi, saved_rsi);
+      __ movptr(rdi, saved_rdi);
+      __ movptr(rbx, saved_rbx);
+    }
+  }
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //   c_rarg3   - r vector byte array address
+  //   c_rarg4   - input length
+  //
+  address generate_cipherBlockChaining_encryptAESCrypt() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_encryptAESCrypt");
+    address start = __ pc();
+
+    Label L_exit, L_key_192_256, L_key_256, L_loopTop_128, L_loopTop_192, L_loopTop_256;
+    const Register from        = rsi;      // source array address
+    const Register to          = rdx;      // destination array address
+    const Register key         = rcx;      // key array address
+    const Register rvec        = rdi;      // r byte array initialized from initvector array address
+                                           // and left with the results of the last encryption block
+    const Register len_reg     = rbx;      // src len (must be multiple of blocksize 16)
+    const Register pos         = rax;
+
+    // xmm register assignments for the loops below
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    // first 6 keys preloaded into xmm2-xmm7
+    const int XMM_REG_NUM_KEY_FIRST = 2;
+    const int XMM_REG_NUM_KEY_LAST  = 7;
+    const XMMRegister xmm_key0   = as_XMMRegister(XMM_REG_NUM_KEY_FIRST);
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+    handleSOERegisters(true /*saving*/);
+
+    // load registers from incoming parameters
+    const Address  from_param(rbp, 8+0);
+    const Address  to_param  (rbp, 8+4);
+    const Address  key_param (rbp, 8+8);
+    const Address  rvec_param (rbp, 8+12);
+    const Address  len_param  (rbp, 8+16);
+    __ movptr(from , from_param);
+    __ movptr(to   , to_param);
+    __ movptr(key  , key_param);
+    __ movptr(rvec , rvec_param);
+    __ movptr(len_reg , len_param);
+
+    const XMMRegister xmm_key_shuf_mask = xmm_temp;  // used temporarily to swap key bytes up front
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    // load up xmm regs 2 thru 7 with keys 0-5
+    for (int rnum = XMM_REG_NUM_KEY_FIRST, offset = 0x00; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      load_key(as_XMMRegister(rnum), key, offset, xmm_key_shuf_mask);
+      offset += 0x10;
+    }
+
+    __ movdqu(xmm_result, Address(rvec, 0x00));   // initialize xmm_result with r vec
+
+    // now split to different paths depending on the keylen (len in ints of AESCrypt.KLE array (52=192, or 60=256))
+    __ movl(rax, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    __ cmpl(rax, 44);
+    __ jcc(Assembler::notEqual, L_key_192_256);
+
+    // 128 bit code follows here
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_128);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);                                // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);                                // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = 0x60; key_offset <= 0x90; key_offset += 0x10) {
+      aes_enc_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0xa0);
+    __ aesenclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_128);
+
+    __ BIND(L_exit);
+    __ movdqu(Address(rvec, 0), xmm_result);     // final value of r stored in rvec of CipherBlockChaining object
+
+    handleSOERegisters(false /*restoring*/);
+    __ movl(rax, 0);                             // return 0 (why?)
+    __ leave();                                  // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+  __ BIND(L_key_192_256);
+  // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
+    __ cmpl(rax, 52);
+    __ jcc(Assembler::notEqual, L_key_256);
+
+    // 192-bit code follows here (could be changed to use more xmm registers)
+    __ movptr(pos, 0);
+  __ align(OptoLoopAlignment);
+  __ BIND(L_loopTop_192);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);                                // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);                                // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = 0x60; key_offset <= 0xb0; key_offset += 0x10) {
+      aes_enc_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0xc0);
+    __ aesenclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);   // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_192);
+    __ jmp(L_exit);
+
+  __ BIND(L_key_256);
+    // 256-bit code follows here (could be changed to use more xmm registers)
+    __ movptr(pos, 0);
+  __ align(OptoLoopAlignment);
+  __ BIND(L_loopTop_256);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);                                // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);                                // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = 0x60; key_offset <= 0xd0; key_offset += 0x10) {
+      aes_enc_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0xe0);
+    __ aesenclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);   // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_256);
+    __ jmp(L_exit);
+
+    return start;
+  }
+
+
+  // CBC AES Decryption.
+  // In 32-bit stub, because of lack of registers we do not try to parallelize 4 blocks at a time.
+  //
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //   c_rarg3   - r vector byte array address
+  //   c_rarg4   - input length
+  //
+
+  address generate_cipherBlockChaining_decryptAESCrypt() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_decryptAESCrypt");
+    address start = __ pc();
+
+    Label L_exit, L_key_192_256, L_key_256;
+    Label L_singleBlock_loopTop_128;
+    Label L_singleBlock_loopTop_192, L_singleBlock_loopTop_256;
+    const Register from        = rsi;      // source array address
+    const Register to          = rdx;      // destination array address
+    const Register key         = rcx;      // key array address
+    const Register rvec        = rdi;      // r byte array initialized from initvector array address
+                                           // and left with the results of the last encryption block
+    const Register len_reg     = rbx;      // src len (must be multiple of blocksize 16)
+    const Register pos         = rax;
+
+    // xmm register assignments for the loops below
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    // first 6 keys preloaded into xmm2-xmm7
+    const int XMM_REG_NUM_KEY_FIRST = 2;
+    const int XMM_REG_NUM_KEY_LAST  = 7;
+    const int FIRST_NON_REG_KEY_offset = 0x70;
+    const XMMRegister xmm_key_first   = as_XMMRegister(XMM_REG_NUM_KEY_FIRST);
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+    handleSOERegisters(true /*saving*/);
+
+    // load registers from incoming parameters
+    const Address  from_param(rbp, 8+0);
+    const Address  to_param  (rbp, 8+4);
+    const Address  key_param (rbp, 8+8);
+    const Address  rvec_param (rbp, 8+12);
+    const Address  len_param  (rbp, 8+16);
+    __ movptr(from , from_param);
+    __ movptr(to   , to_param);
+    __ movptr(key  , key_param);
+    __ movptr(rvec , rvec_param);
+    __ movptr(len_reg , len_param);
+
+    // the java expanded key ordering is rotated one position from what we want
+    // so we start from 0x10 here and hit 0x00 last
+    const XMMRegister xmm_key_shuf_mask = xmm1;  // used temporarily to swap key bytes up front
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    // load up xmm regs 2 thru 6 with first 5 keys
+    for (int rnum = XMM_REG_NUM_KEY_FIRST, offset = 0x10; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      load_key(as_XMMRegister(rnum), key, offset, xmm_key_shuf_mask);
+      offset += 0x10;
+    }
+
+    // inside here, use the rvec register to point to previous block cipher
+    // with which we xor at the end of each newly decrypted block
+    const Register  prev_block_cipher_ptr = rvec;
+
+    // now split to different paths depending on the keylen (len in ints of AESCrypt.KLE array (52=192, or 60=256))
+    __ movl(rax, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    __ cmpl(rax, 44);
+    __ jcc(Assembler::notEqual, L_key_192_256);
+
+
+    // 128-bit code follows here, parallelized
+    __ movptr(pos, 0);
+  __ align(OptoLoopAlignment);
+  __ BIND(L_singleBlock_loopTop_128);
+    __ cmpptr(len_reg, 0);           // any blocks left??
+    __ jcc(Assembler::equal, L_exit);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ pxor  (xmm_result, xmm_key_first);                             // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = FIRST_NON_REG_KEY_offset; key_offset <= 0xa0; key_offset += 0x10) {   // 128-bit runs up to key offset a0
+      aes_dec_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0x00);                                     // final key is stored in java expanded array at offset 0
+    __ aesdeclast(xmm_result, xmm_temp);
+    __ movdqu(xmm_temp, Address(prev_block_cipher_ptr, 0x00));
+    __ pxor  (xmm_result, xmm_temp);                                  // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ lea(prev_block_cipher_ptr, Address(from, pos, Address::times_1, 0));     // set up new ptr
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jmp(L_singleBlock_loopTop_128);
+
+
+    __ BIND(L_exit);
+    __ movdqu(xmm_temp, Address(prev_block_cipher_ptr, 0x00));
+    __ movptr(rvec , rvec_param);                                     // restore this since used in loop
+    __ movdqu(Address(rvec, 0), xmm_temp);                            // final value of r stored in rvec of CipherBlockChaining object
+    handleSOERegisters(false /*restoring*/);
+    __ movl(rax, 0);                                                  // return 0 (why?)
+    __ leave();                                                       // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+
+    __ BIND(L_key_192_256);
+    // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
+    __ cmpl(rax, 52);
+    __ jcc(Assembler::notEqual, L_key_256);
+
+    // 192-bit code follows here (could be optimized to use parallelism)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_192);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ pxor  (xmm_result, xmm_key_first);                             // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = FIRST_NON_REG_KEY_offset; key_offset <= 0xc0; key_offset += 0x10) {   // 192-bit runs up to key offset c0
+      aes_dec_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0x00);                                     // final key is stored in java expanded array at offset 0
+    __ aesdeclast(xmm_result, xmm_temp);
+    __ movdqu(xmm_temp, Address(prev_block_cipher_ptr, 0x00));
+    __ pxor  (xmm_result, xmm_temp);                                  // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ lea(prev_block_cipher_ptr, Address(from, pos, Address::times_1, 0));     // set up new ptr
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual,L_singleBlock_loopTop_192);
+    __ jmp(L_exit);
+
+    __ BIND(L_key_256);
+    // 256-bit code follows here (could be optimized to use parallelism)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_256);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ pxor  (xmm_result, xmm_key_first);                             // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    for (int key_offset = FIRST_NON_REG_KEY_offset; key_offset <= 0xe0; key_offset += 0x10) {   // 256-bit runs up to key offset e0
+      aes_dec_key(xmm_result, xmm_temp, key, key_offset);
+    }
+    load_key(xmm_temp, key, 0x00);                                     // final key is stored in java expanded array at offset 0
+    __ aesdeclast(xmm_result, xmm_temp);
+    __ movdqu(xmm_temp, Address(prev_block_cipher_ptr, 0x00));
+    __ pxor  (xmm_result, xmm_temp);                                  // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ lea(prev_block_cipher_ptr, Address(from, pos, Address::times_1, 0));     // set up new ptr
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual,L_singleBlock_loopTop_256);
+    __ jmp(L_exit);
+
+    return start;
+  }
+
+
  public:
   // Information about frame layout at time of blocking runtime call.
   // Note that we only have to preserve callee-saved registers since
@@ -2332,6 +2855,16 @@ class StubGenerator: public StubCodeGenerator {
     generate_arraycopy_stubs();
 
     generate_math_stubs();
+
+    // don't bother generating these AES intrinsic stubs unless global flag is set
+    if (UseAESIntrinsics) {
+      StubRoutines::x86::_key_shuffle_mask_addr = generate_key_shuffle_mask();  // might be needed by the others
+
+      StubRoutines::_aescrypt_encryptBlock = generate_aescrypt_encryptBlock();
+      StubRoutines::_aescrypt_decryptBlock = generate_aescrypt_decryptBlock();
+      StubRoutines::_cipherBlockChaining_encryptAESCrypt = generate_cipherBlockChaining_encryptAESCrypt();
+      StubRoutines::_cipherBlockChaining_decryptAESCrypt = generate_cipherBlockChaining_decryptAESCrypt();
+    }
   }
 
 
diff --git a/src/cpu/x86/vm/stubGenerator_x86_64.cpp b/src/cpu/x86/vm/stubGenerator_x86_64.cpp
index 8ae595a56..3e223387c 100644
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp
@@ -2941,6 +2941,548 @@ class StubGenerator: public StubCodeGenerator {
     }
   }
 
+  // AES intrinsic stubs
+  enum {AESBlockSize = 16};
+
+  address generate_key_shuffle_mask() {
+    __ align(16);
+    StubCodeMark mark(this, "StubRoutines", "key_shuffle_mask");
+    address start = __ pc();
+    __ emit_data64( 0x0405060700010203, relocInfo::none );
+    __ emit_data64( 0x0c0d0e0f08090a0b, relocInfo::none );
+    return start;
+  }
+
+  // Utility routine for loading a 128-bit key word in little endian format
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void load_key(XMMRegister xmmdst, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    __ movdqu(xmmdst, Address(key, offset));
+    if (xmm_shuf_mask != NULL) {
+      __ pshufb(xmmdst, xmm_shuf_mask);
+    } else {
+      __ pshufb(xmmdst, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    }
+  }
+
+  // aesenc using specified key+offset
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void aes_enc_key(XMMRegister xmmdst, XMMRegister xmmtmp, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    load_key(xmmtmp, key, offset, xmm_shuf_mask);
+    __ aesenc(xmmdst, xmmtmp);
+  }
+
+  // aesdec using specified key+offset
+  // can optionally specify that the shuffle mask is already in an xmmregister
+  void aes_dec_key(XMMRegister xmmdst, XMMRegister xmmtmp, Register key, int offset, XMMRegister xmm_shuf_mask=NULL) {
+    load_key(xmmtmp, key, offset, xmm_shuf_mask);
+    __ aesdec(xmmdst, xmmtmp);
+  }
+
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //
+  address generate_aescrypt_encryptBlock() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "aescrypt_encryptBlock");
+    Label L_doLast;
+    address start = __ pc();
+
+    const Register from        = c_rarg0;  // source array address
+    const Register to          = c_rarg1;  // destination array address
+    const Register key         = c_rarg2;  // key array address
+    const Register keylen      = rax;
+
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    const XMMRegister xmm_key_shuf_mask = xmm2;
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+
+    __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    // keylen = # of 32-bit words, convert to 128-bit words
+    __ shrl(keylen, 2);
+    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
+
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    __ movdqu(xmm_result, Address(from, 0));  // get 16 bytes of input
+
+    // For encryption, the java expanded key ordering is just what we need
+    // we don't know if the key is aligned, hence not using load-execute form
+
+    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
+    __ pxor(xmm_result, xmm_temp);
+    for (int offset = 0x10; offset <= 0x90; offset += 0x10) {
+      aes_enc_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
+    }
+    load_key  (xmm_temp, key, 0xa0, xmm_key_shuf_mask);
+    __ cmpl(keylen, 0);
+    __ jcc(Assembler::equal, L_doLast);
+    __ aesenc(xmm_result, xmm_temp);                   // only in 192 and 256 bit keys
+    aes_enc_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
+    load_key(xmm_temp, key, 0xc0, xmm_key_shuf_mask);
+    __ subl(keylen, 2);
+    __ jcc(Assembler::equal, L_doLast);
+    __ aesenc(xmm_result, xmm_temp);                   // only in 256 bit keys
+    aes_enc_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
+    load_key(xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+
+    __ BIND(L_doLast);
+    __ aesenclast(xmm_result, xmm_temp);
+    __ movdqu(Address(to, 0), xmm_result);        // store the result
+    __ xorptr(rax, rax); // return 0
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+    return start;
+  }
+
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //
+  address generate_aescrypt_decryptBlock() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "aescrypt_decryptBlock");
+    Label L_doLast;
+    address start = __ pc();
+
+    const Register from        = c_rarg0;  // source array address
+    const Register to          = c_rarg1;  // destination array address
+    const Register key         = c_rarg2;  // key array address
+    const Register keylen      = rax;
+
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    const XMMRegister xmm_key_shuf_mask = xmm2;
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+
+    __ movl(keylen, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    // keylen = # of 32-bit words, convert to 128-bit words
+    __ shrl(keylen, 2);
+    __ subl(keylen, 11);   // every key has at least 11 128-bit words, some have more
+
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    __ movdqu(xmm_result, Address(from, 0));
+
+    // for decryption java expanded key ordering is rotated one position from what we want
+    // so we start from 0x10 here and hit 0x00 last
+    // we don't know if the key is aligned, hence not using load-execute form
+    load_key(xmm_temp, key, 0x10, xmm_key_shuf_mask);
+    __ pxor  (xmm_result, xmm_temp);
+    for (int offset = 0x20; offset <= 0xa0; offset += 0x10) {
+      aes_dec_key(xmm_result, xmm_temp, key, offset, xmm_key_shuf_mask);
+    }
+    __ cmpl(keylen, 0);
+    __ jcc(Assembler::equal, L_doLast);
+    // only in 192 and 256 bit keys
+    aes_dec_key(xmm_result, xmm_temp, key, 0xb0, xmm_key_shuf_mask);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xc0, xmm_key_shuf_mask);
+    __ subl(keylen, 2);
+    __ jcc(Assembler::equal, L_doLast);
+    // only in 256 bit keys
+    aes_dec_key(xmm_result, xmm_temp, key, 0xd0, xmm_key_shuf_mask);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xe0, xmm_key_shuf_mask);
+
+    __ BIND(L_doLast);
+    // for decryption the aesdeclast operation is always on key+0x00
+    load_key(xmm_temp, key, 0x00, xmm_key_shuf_mask);
+    __ aesdeclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, 0), xmm_result);  // store the result
+
+    __ xorptr(rax, rax); // return 0
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+    return start;
+  }
+
+
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //   c_rarg3   - r vector byte array address
+  //   c_rarg4   - input length
+  //
+  address generate_cipherBlockChaining_encryptAESCrypt() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_encryptAESCrypt");
+    address start = __ pc();
+
+    Label L_exit, L_key_192_256, L_key_256, L_loopTop_128, L_loopTop_192, L_loopTop_256;
+    const Register from        = c_rarg0;  // source array address
+    const Register to          = c_rarg1;  // destination array address
+    const Register key         = c_rarg2;  // key array address
+    const Register rvec        = c_rarg3;  // r byte array initialized from initvector array address
+                                           // and left with the results of the last encryption block
+#ifndef _WIN64
+    const Register len_reg     = c_rarg4;  // src len (must be multiple of blocksize 16)
+#else
+    const Address  len_mem(rsp, 6 * wordSize);  // length is on stack on Win64
+    const Register len_reg     = r10;      // pick the first volatile windows register
+#endif
+    const Register pos         = rax;
+
+    // xmm register assignments for the loops below
+    const XMMRegister xmm_result = xmm0;
+    const XMMRegister xmm_temp   = xmm1;
+    // keys 0-10 preloaded into xmm2-xmm12
+    const int XMM_REG_NUM_KEY_FIRST = 2;
+    const int XMM_REG_NUM_KEY_LAST  = 12;
+    const XMMRegister xmm_key0   = as_XMMRegister(XMM_REG_NUM_KEY_FIRST);
+    const XMMRegister xmm_key10  = as_XMMRegister(XMM_REG_NUM_KEY_LAST);
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+
+#ifdef _WIN64
+    // on win64, fill len_reg from stack position
+    __ movl(len_reg, len_mem);
+    // save the xmm registers which must be preserved 6-12
+    __ subptr(rsp, -rsp_after_call_off * wordSize);
+    for (int i = 6; i <= XMM_REG_NUM_KEY_LAST; i++) {
+      __ movdqu(xmm_save(i), as_XMMRegister(i));
+    }
+#endif
+
+    const XMMRegister xmm_key_shuf_mask = xmm_temp;  // used temporarily to swap key bytes up front
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    // load up xmm regs 2 thru 12 with key 0x00 - 0xa0
+    for (int rnum = XMM_REG_NUM_KEY_FIRST, offset = 0x00; rnum <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      load_key(as_XMMRegister(rnum), key, offset, xmm_key_shuf_mask);
+      offset += 0x10;
+    }
+
+    __ movdqu(xmm_result, Address(rvec, 0x00));   // initialize xmm_result with r vec
+
+    // now split to different paths depending on the keylen (len in ints of AESCrypt.KLE array (52=192, or 60=256))
+    __ movl(rax, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    __ cmpl(rax, 44);
+    __ jcc(Assembler::notEqual, L_key_192_256);
+
+    // 128 bit code follows here
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_128);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);               // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);               // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum <= XMM_REG_NUM_KEY_LAST - 1; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    __ aesenclast(xmm_result, xmm_key10);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_128);
+
+    __ BIND(L_exit);
+    __ movdqu(Address(rvec, 0), xmm_result);     // final value of r stored in rvec of CipherBlockChaining object
+
+#ifdef _WIN64
+    // restore xmm regs belonging to calling function
+    for (int i = 6; i <= XMM_REG_NUM_KEY_LAST; i++) {
+      __ movdqu(as_XMMRegister(i), xmm_save(i));
+    }
+#endif
+    __ movl(rax, 0); // return 0 (why?)
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+    __ BIND(L_key_192_256);
+    // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
+    __ cmpl(rax, 52);
+    __ jcc(Assembler::notEqual, L_key_256);
+
+    // 192-bit code follows here (could be changed to use more xmm registers)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_192);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);               // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);               // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    aes_enc_key(xmm_result, xmm_temp, key, 0xb0);
+    load_key(xmm_temp, key, 0xc0);
+    __ aesenclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_192);
+    __ jmp(L_exit);
+
+    __ BIND(L_key_256);
+    // 256-bit code follows here (could be changed to use more xmm registers)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_loopTop_256);
+    __ movdqu(xmm_temp, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of input
+    __ pxor  (xmm_result, xmm_temp);               // xor with the current r vector
+
+    __ pxor  (xmm_result, xmm_key0);               // do the aes rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      __ aesenc(xmm_result, as_XMMRegister(rnum));
+    }
+    aes_enc_key(xmm_result, xmm_temp, key, 0xb0);
+    aes_enc_key(xmm_result, xmm_temp, key, 0xc0);
+    aes_enc_key(xmm_result, xmm_temp, key, 0xd0);
+    load_key(xmm_temp, key, 0xe0);
+    __ aesenclast(xmm_result, xmm_temp);
+
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual, L_loopTop_256);
+    __ jmp(L_exit);
+
+    return start;
+  }
+
+
+
+  // This is a version of CBC/AES Decrypt which does 4 blocks in a loop at a time
+  // to hide instruction latency
+  //
+  // Arguments:
+  //
+  // Inputs:
+  //   c_rarg0   - source byte array address
+  //   c_rarg1   - destination byte array address
+  //   c_rarg2   - K (key) in little endian int array
+  //   c_rarg3   - r vector byte array address
+  //   c_rarg4   - input length
+  //
+
+  address generate_cipherBlockChaining_decryptAESCrypt_Parallel() {
+    assert(UseAES && (UseAVX > 0), "need AES instructions and misaligned SSE support");
+    __ align(CodeEntryAlignment);
+    StubCodeMark mark(this, "StubRoutines", "cipherBlockChaining_decryptAESCrypt");
+    address start = __ pc();
+
+    Label L_exit, L_key_192_256, L_key_256;
+    Label L_singleBlock_loopTop_128, L_multiBlock_loopTop_128;
+    Label L_singleBlock_loopTop_192, L_singleBlock_loopTop_256;
+    const Register from        = c_rarg0;  // source array address
+    const Register to          = c_rarg1;  // destination array address
+    const Register key         = c_rarg2;  // key array address
+    const Register rvec        = c_rarg3;  // r byte array initialized from initvector array address
+                                           // and left with the results of the last encryption block
+#ifndef _WIN64
+    const Register len_reg     = c_rarg4;  // src len (must be multiple of blocksize 16)
+#else
+    const Address  len_mem(rsp, 6 * wordSize);  // length is on stack on Win64
+    const Register len_reg     = r10;      // pick the first volatile windows register
+#endif
+    const Register pos         = rax;
+
+    // xmm register assignments for the loops below
+    const XMMRegister xmm_result = xmm0;
+    // keys 0-10 preloaded into xmm2-xmm12
+    const int XMM_REG_NUM_KEY_FIRST = 5;
+    const int XMM_REG_NUM_KEY_LAST  = 15;
+    const XMMRegister xmm_key_first   = as_XMMRegister(XMM_REG_NUM_KEY_FIRST);
+    const XMMRegister xmm_key_last  = as_XMMRegister(XMM_REG_NUM_KEY_LAST);
+
+    __ enter(); // required for proper stackwalking of RuntimeStub frame
+
+#ifdef _WIN64
+    // on win64, fill len_reg from stack position
+    __ movl(len_reg, len_mem);
+    // save the xmm registers which must be preserved 6-15
+    __ subptr(rsp, -rsp_after_call_off * wordSize);
+    for (int i = 6; i <= XMM_REG_NUM_KEY_LAST; i++) {
+      __ movdqu(xmm_save(i), as_XMMRegister(i));
+    }
+#endif
+    // the java expanded key ordering is rotated one position from what we want
+    // so we start from 0x10 here and hit 0x00 last
+    const XMMRegister xmm_key_shuf_mask = xmm1;  // used temporarily to swap key bytes up front
+    __ movdqu(xmm_key_shuf_mask, ExternalAddress(StubRoutines::x86::key_shuffle_mask_addr()));
+    // load up xmm regs 5 thru 15 with key 0x10 - 0xa0 - 0x00
+    for (int rnum = XMM_REG_NUM_KEY_FIRST, offset = 0x10; rnum <= XMM_REG_NUM_KEY_LAST; rnum++) {
+      if (rnum == XMM_REG_NUM_KEY_LAST) offset = 0x00;
+      load_key(as_XMMRegister(rnum), key, offset, xmm_key_shuf_mask);
+      offset += 0x10;
+    }
+
+    const XMMRegister xmm_prev_block_cipher = xmm1;  // holds cipher of previous block
+    // registers holding the four results in the parallelized loop
+    const XMMRegister xmm_result0 = xmm0;
+    const XMMRegister xmm_result1 = xmm2;
+    const XMMRegister xmm_result2 = xmm3;
+    const XMMRegister xmm_result3 = xmm4;
+
+    __ movdqu(xmm_prev_block_cipher, Address(rvec, 0x00));   // initialize with initial rvec
+
+    // now split to different paths depending on the keylen (len in ints of AESCrypt.KLE array (52=192, or 60=256))
+    __ movl(rax, Address(key, arrayOopDesc::length_offset_in_bytes() - arrayOopDesc::base_offset_in_bytes(T_INT)));
+    __ cmpl(rax, 44);
+    __ jcc(Assembler::notEqual, L_key_192_256);
+
+
+    // 128-bit code follows here, parallelized
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_multiBlock_loopTop_128);
+    __ cmpptr(len_reg, 4*AESBlockSize);           // see if at least 4 blocks left
+    __ jcc(Assembler::less, L_singleBlock_loopTop_128);
+
+    __ movdqu(xmm_result0, Address(from, pos, Address::times_1, 0*AESBlockSize));   // get next 4 blocks into xmmresult registers
+    __ movdqu(xmm_result1, Address(from, pos, Address::times_1, 1*AESBlockSize));
+    __ movdqu(xmm_result2, Address(from, pos, Address::times_1, 2*AESBlockSize));
+    __ movdqu(xmm_result3, Address(from, pos, Address::times_1, 3*AESBlockSize));
+
+#define DoFour(opc, src_reg)                    \
+    __ opc(xmm_result0, src_reg);               \
+    __ opc(xmm_result1, src_reg);               \
+    __ opc(xmm_result2, src_reg);               \
+    __ opc(xmm_result3, src_reg);
+
+    DoFour(pxor, xmm_key_first);
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST - 1; rnum++) {
+      DoFour(aesdec, as_XMMRegister(rnum));
+    }
+    DoFour(aesdeclast, xmm_key_last);
+    // for each result, xor with the r vector of previous cipher block
+    __ pxor(xmm_result0, xmm_prev_block_cipher);
+    __ movdqu(xmm_prev_block_cipher, Address(from, pos, Address::times_1, 0*AESBlockSize));
+    __ pxor(xmm_result1, xmm_prev_block_cipher);
+    __ movdqu(xmm_prev_block_cipher, Address(from, pos, Address::times_1, 1*AESBlockSize));
+    __ pxor(xmm_result2, xmm_prev_block_cipher);
+    __ movdqu(xmm_prev_block_cipher, Address(from, pos, Address::times_1, 2*AESBlockSize));
+    __ pxor(xmm_result3, xmm_prev_block_cipher);
+    __ movdqu(xmm_prev_block_cipher, Address(from, pos, Address::times_1, 3*AESBlockSize));   // this will carry over to next set of blocks
+
+    __ movdqu(Address(to, pos, Address::times_1, 0*AESBlockSize), xmm_result0);     // store 4 results into the next 64 bytes of output
+    __ movdqu(Address(to, pos, Address::times_1, 1*AESBlockSize), xmm_result1);
+    __ movdqu(Address(to, pos, Address::times_1, 2*AESBlockSize), xmm_result2);
+    __ movdqu(Address(to, pos, Address::times_1, 3*AESBlockSize), xmm_result3);
+
+    __ addptr(pos, 4*AESBlockSize);
+    __ subptr(len_reg, 4*AESBlockSize);
+    __ jmp(L_multiBlock_loopTop_128);
+
+    // registers used in the non-parallelized loops
+    const XMMRegister xmm_prev_block_cipher_save = xmm2;
+    const XMMRegister xmm_temp   = xmm3;
+
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_128);
+    __ cmpptr(len_reg, 0);           // any blocks left??
+    __ jcc(Assembler::equal, L_exit);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ movdqa(xmm_prev_block_cipher_save, xmm_result);              // save for next r vector
+    __ pxor  (xmm_result, xmm_key_first);               // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum  <= XMM_REG_NUM_KEY_LAST - 1; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    __ aesdeclast(xmm_result, xmm_key_last);
+    __ pxor  (xmm_result, xmm_prev_block_cipher);               // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ movdqa(xmm_prev_block_cipher, xmm_prev_block_cipher_save);              // set up next r vector with cipher input from this block
+
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jmp(L_singleBlock_loopTop_128);
+
+
+    __ BIND(L_exit);
+    __ movdqu(Address(rvec, 0), xmm_prev_block_cipher);     // final value of r stored in rvec of CipherBlockChaining object
+#ifdef _WIN64
+    // restore regs belonging to calling function
+    for (int i = 6; i <= XMM_REG_NUM_KEY_LAST; i++) {
+      __ movdqu(as_XMMRegister(i), xmm_save(i));
+    }
+#endif
+    __ movl(rax, 0); // return 0 (why?)
+    __ leave(); // required for proper stackwalking of RuntimeStub frame
+    __ ret(0);
+
+
+    __ BIND(L_key_192_256);
+    // here rax = len in ints of AESCrypt.KLE array (52=192, or 60=256)
+    __ cmpl(rax, 52);
+    __ jcc(Assembler::notEqual, L_key_256);
+
+    // 192-bit code follows here (could be optimized to use parallelism)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_192);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ movdqa(xmm_prev_block_cipher_save, xmm_result);              // save for next r vector
+    __ pxor  (xmm_result, xmm_key_first);               // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum <= XMM_REG_NUM_KEY_LAST - 1; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    aes_dec_key(xmm_result, xmm_temp, key, 0xb0);     // 192-bit key goes up to c0
+    aes_dec_key(xmm_result, xmm_temp, key, 0xc0);
+    __ aesdeclast(xmm_result, xmm_key_last);                    // xmm15 always came from key+0
+    __ pxor  (xmm_result, xmm_prev_block_cipher);               // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ movdqa(xmm_prev_block_cipher, xmm_prev_block_cipher_save);              // set up next r vector with cipher input from this block
+
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual,L_singleBlock_loopTop_192);
+    __ jmp(L_exit);
+
+    __ BIND(L_key_256);
+    // 256-bit code follows here (could be optimized to use parallelism)
+    __ movptr(pos, 0);
+    __ align(OptoLoopAlignment);
+    __ BIND(L_singleBlock_loopTop_256);
+    __ movdqu(xmm_result, Address(from, pos, Address::times_1, 0));   // get next 16 bytes of cipher input
+    __ movdqa(xmm_prev_block_cipher_save, xmm_result);              // save for next r vector
+    __ pxor  (xmm_result, xmm_key_first);               // do the aes dec rounds
+    for (int rnum = XMM_REG_NUM_KEY_FIRST + 1; rnum <= XMM_REG_NUM_KEY_LAST - 1; rnum++) {
+      __ aesdec(xmm_result, as_XMMRegister(rnum));
+    }
+    aes_dec_key(xmm_result, xmm_temp, key, 0xb0);     // 256-bit key goes up to e0
+    aes_dec_key(xmm_result, xmm_temp, key, 0xc0);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xd0);
+    aes_dec_key(xmm_result, xmm_temp, key, 0xe0);
+    __ aesdeclast(xmm_result, xmm_key_last);             // xmm15 came from key+0
+    __ pxor  (xmm_result, xmm_prev_block_cipher);               // xor with the current r vector
+    __ movdqu(Address(to, pos, Address::times_1, 0), xmm_result);     // store into the next 16 bytes of output
+    // no need to store r to memory until we exit
+    __ movdqa(xmm_prev_block_cipher, xmm_prev_block_cipher_save);              // set up next r vector with cipher input from this block
+
+    __ addptr(pos, AESBlockSize);
+    __ subptr(len_reg, AESBlockSize);
+    __ jcc(Assembler::notEqual,L_singleBlock_loopTop_256);
+    __ jmp(L_exit);
+
+    return start;
+  }
+
+
+
 #undef __
 #define __ masm->
 
@@ -3135,6 +3677,16 @@ class StubGenerator: public StubCodeGenerator {
     generate_arraycopy_stubs();
 
     generate_math_stubs();
+
+    // don't bother generating these AES intrinsic stubs unless global flag is set
+    if (UseAESIntrinsics) {
+      StubRoutines::x86::_key_shuffle_mask_addr = generate_key_shuffle_mask();  // needed by the others
+
+      StubRoutines::_aescrypt_encryptBlock = generate_aescrypt_encryptBlock();
+      StubRoutines::_aescrypt_decryptBlock = generate_aescrypt_decryptBlock();
+      StubRoutines::_cipherBlockChaining_encryptAESCrypt = generate_cipherBlockChaining_encryptAESCrypt();
+      StubRoutines::_cipherBlockChaining_decryptAESCrypt = generate_cipherBlockChaining_decryptAESCrypt_Parallel();
+    }
   }
 
  public:
diff --git a/src/cpu/x86/vm/stubRoutines_x86_32.cpp b/src/cpu/x86/vm/stubRoutines_x86_32.cpp
index 6ec4121b9..cfd4f33a6 100644
--- a/src/cpu/x86/vm/stubRoutines_x86_32.cpp
+++ b/src/cpu/x86/vm/stubRoutines_x86_32.cpp
@@ -44,3 +44,4 @@
 
 address StubRoutines::x86::_verify_mxcsr_entry         = NULL;
 address StubRoutines::x86::_verify_fpu_cntrl_wrd_entry = NULL;
+address StubRoutines::x86::_key_shuffle_mask_addr = NULL;
diff --git a/src/cpu/x86/vm/stubRoutines_x86_32.hpp b/src/cpu/x86/vm/stubRoutines_x86_32.hpp
index 64767c8ad..d53124fc6 100644
--- a/src/cpu/x86/vm/stubRoutines_x86_32.hpp
+++ b/src/cpu/x86/vm/stubRoutines_x86_32.hpp
@@ -41,10 +41,14 @@ class x86 {
  private:
   static address _verify_mxcsr_entry;
   static address _verify_fpu_cntrl_wrd_entry;
+  // shuffle mask for fixing up 128-bit words consisting of big-endian 32-bit integers
+  static address _key_shuffle_mask_addr;
 
  public:
   static address verify_mxcsr_entry()                        { return _verify_mxcsr_entry; }
   static address verify_fpu_cntrl_wrd_entry()                { return _verify_fpu_cntrl_wrd_entry; }
+  static address key_shuffle_mask_addr()                     { return _key_shuffle_mask_addr; }
+
 };
 
   static bool    returns_to_call_stub(address return_pc)     { return return_pc == _call_stub_return_address; }
diff --git a/src/cpu/x86/vm/stubRoutines_x86_64.cpp b/src/cpu/x86/vm/stubRoutines_x86_64.cpp
index 084bbf8fb..cf8ec5d7b 100644
--- a/src/cpu/x86/vm/stubRoutines_x86_64.cpp
+++ b/src/cpu/x86/vm/stubRoutines_x86_64.cpp
@@ -56,3 +56,4 @@ address StubRoutines::x86::_float_sign_flip = NULL;
 address StubRoutines::x86::_double_sign_mask = NULL;
 address StubRoutines::x86::_double_sign_flip = NULL;
 address StubRoutines::x86::_mxcsr_std = NULL;
+address StubRoutines::x86::_key_shuffle_mask_addr = NULL;
diff --git a/src/cpu/x86/vm/stubRoutines_x86_64.hpp b/src/cpu/x86/vm/stubRoutines_x86_64.hpp
index 9b9cede4f..c3efeecb7 100644
--- a/src/cpu/x86/vm/stubRoutines_x86_64.hpp
+++ b/src/cpu/x86/vm/stubRoutines_x86_64.hpp
@@ -54,6 +54,8 @@ class x86 {
   static address _double_sign_mask;
   static address _double_sign_flip;
   static address _mxcsr_std;
+  // shuffle mask for fixing up 128-bit words consisting of big-endian 32-bit integers
+  static address _key_shuffle_mask_addr;
 
  public:
 
@@ -116,6 +118,9 @@ class x86 {
   {
     return _mxcsr_std;
   }
+
+  static address key_shuffle_mask_addr()                     { return _key_shuffle_mask_addr; }
+
 };
 
 #endif // CPU_X86_VM_STUBROUTINES_X86_64_HPP
diff --git a/src/cpu/x86/vm/vm_version_x86.cpp b/src/cpu/x86/vm/vm_version_x86.cpp
index bf7b3c213..182b0ab1a 100644
--- a/src/cpu/x86/vm/vm_version_x86.cpp
+++ b/src/cpu/x86/vm/vm_version_x86.cpp
@@ -419,13 +419,16 @@ void VM_Version::get_processor_features() {
   if (UseAVX < 1)
     _cpuFeatures &= ~CPU_AVX;
 
+  if (!UseAES && !FLAG_IS_DEFAULT(UseAES))
+    _cpuFeatures &= ~CPU_AES;
+
   if (logical_processors_per_package() == 1) {
     // HT processor could be installed on a system which doesn't support HT.
     _cpuFeatures &= ~CPU_HT;
   }
 
   char buf[256];
-  jio_snprintf(buf, sizeof(buf), "(%u cores per cpu, %u threads per core) family %d model %d stepping %d%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s",
+  jio_snprintf(buf, sizeof(buf), "(%u cores per cpu, %u threads per core) family %d model %d stepping %d%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s",
                cores_per_cpu(), threads_per_core(),
                cpu_family(), _model, _stepping,
                (supports_cmov() ? ", cmov" : ""),
@@ -441,6 +444,7 @@ void VM_Version::get_processor_features() {
                (supports_popcnt() ? ", popcnt" : ""),
                (supports_avx()    ? ", avx" : ""),
                (supports_avx2()   ? ", avx2" : ""),
+               (supports_aes()    ? ", aes" : ""),
                (supports_mmx_ext() ? ", mmxext" : ""),
                (supports_3dnow_prefetch() ? ", 3dnowpref" : ""),
                (supports_lzcnt()   ? ", lzcnt": ""),
@@ -472,6 +476,29 @@ void VM_Version::get_processor_features() {
   if (!supports_avx ()) // Drop to 0 if no AVX  support
     UseAVX = 0;
 
+  // Use AES instructions if available.
+  if (supports_aes()) {
+    if (FLAG_IS_DEFAULT(UseAES)) {
+      UseAES = true;
+    }
+  } else if (UseAES) {
+    if (!FLAG_IS_DEFAULT(UseAES))
+      warning("AES instructions not available on this CPU");
+    FLAG_SET_DEFAULT(UseAES, false);
+  }
+
+  // The AES intrinsic stubs require AES instruction support (of course)
+  // but also require AVX mode for misaligned SSE access
+  if (UseAES && (UseAVX > 0)) {
+    if (FLAG_IS_DEFAULT(UseAESIntrinsics)) {
+      UseAESIntrinsics = true;
+    }
+  } else if (UseAESIntrinsics) {
+    if (!FLAG_IS_DEFAULT(UseAESIntrinsics))
+      warning("AES intrinsics not available on this CPU");
+    FLAG_SET_DEFAULT(UseAESIntrinsics, false);
+  }
+
 #ifdef COMPILER2
   if (UseFPUForSpilling) {
     if (UseSSE < 2) {
@@ -714,6 +741,9 @@ void VM_Version::get_processor_features() {
     if (UseAVX > 0) {
       tty->print("  UseAVX=%d",UseAVX);
     }
+    if (UseAES) {
+      tty->print("  UseAES=1");
+    }
     tty->cr();
     tty->print("Allocation");
     if (AllocatePrefetchStyle <= 0 || UseSSE == 0 && !supports_3dnow_prefetch()) {
diff --git a/src/cpu/x86/vm/vm_version_x86.hpp b/src/cpu/x86/vm/vm_version_x86.hpp
index 92cdbd3fd..12bd3b770 100644
--- a/src/cpu/x86/vm/vm_version_x86.hpp
+++ b/src/cpu/x86/vm/vm_version_x86.hpp
@@ -78,7 +78,9 @@ public:
                sse4_2   : 1,
                         : 2,
                popcnt   : 1,
-                        : 3,
+                        : 1,
+               aes      : 1,
+                        : 1,
                osxsave  : 1,
                avx      : 1,
                         : 3;
@@ -244,7 +246,8 @@ protected:
     CPU_TSC    = (1 << 15),
     CPU_TSCINV = (1 << 16),
     CPU_AVX    = (1 << 17),
-    CPU_AVX2   = (1 << 18)
+    CPU_AVX2   = (1 << 18),
+    CPU_AES    = (1 << 19)
   } cpuFeatureFlags;
 
   enum {
@@ -420,6 +423,8 @@ protected:
       result |= CPU_TSC;
     if (_cpuid_info.ext_cpuid7_edx.bits.tsc_invariance != 0)
       result |= CPU_TSCINV;
+    if (_cpuid_info.std_cpuid1_ecx.bits.aes != 0)
+      result |= CPU_AES;
 
     // AMD features.
     if (is_amd()) {
@@ -544,6 +549,7 @@ public:
   static bool supports_avx()      { return (_cpuFeatures & CPU_AVX) != 0; }
   static bool supports_avx2()     { return (_cpuFeatures & CPU_AVX2) != 0; }
   static bool supports_tsc()      { return (_cpuFeatures & CPU_TSC)    != 0; }
+  static bool supports_aes()      { return (_cpuFeatures & CPU_AES) != 0; }
 
   // Intel features
   static bool is_intel_family_core() { return is_intel() &&
diff --git a/src/cpu/x86/vm/x86.ad b/src/cpu/x86/vm/x86.ad
index a2cf6f794..c49d0e6c3 100644
--- a/src/cpu/x86/vm/x86.ad
+++ b/src/cpu/x86/vm/x86.ad
@@ -4102,9 +4102,158 @@ instruct vsll4L_reg_imm(vecY dst, vecY src, immI8 shift) %{
 
 // ----------------------- LogicalRightShift -----------------------------------
 
-// Shorts/Chars vector logical right shift produces incorrect Java result
+// Shorts vector logical right shift produces incorrect Java result
 // for negative data because java code convert short value into int with
-// sign extension before a shift.
+// sign extension before a shift. But char vectors are fine since chars are
+// unsigned values.
+
+instruct vsrl2S(vecS dst, vecS shift) %{
+  predicate(n->as_Vector()->length() == 2);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed2S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl2S_imm(vecS dst, immI8 shift) %{
+  predicate(n->as_Vector()->length() == 2);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed2S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl2S_reg(vecS dst, vecS src, vecS shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed2S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl2S_reg_imm(vecS dst, vecS src, immI8 shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 2);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed2S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl4S(vecD dst, vecS shift) %{
+  predicate(n->as_Vector()->length() == 4);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed4S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl4S_imm(vecD dst, immI8 shift) %{
+  predicate(n->as_Vector()->length() == 4);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed4S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl4S_reg(vecD dst, vecD src, vecS shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed4S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl4S_reg_imm(vecD dst, vecD src, immI8 shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 4);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed4S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl8S(vecX dst, vecS shift) %{
+  predicate(n->as_Vector()->length() == 8);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed8S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, $shift$$XMMRegister);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl8S_imm(vecX dst, immI8 shift) %{
+  predicate(n->as_Vector()->length() == 8);
+  match(Set dst (URShiftVS dst shift));
+  format %{ "psrlw   $dst,$shift\t! logical right shift packed8S" %}
+  ins_encode %{
+    __ psrlw($dst$$XMMRegister, (int)$shift$$constant);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl8S_reg(vecX dst, vecX src, vecS shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed8S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl8S_reg_imm(vecX dst, vecX src, immI8 shift) %{
+  predicate(UseAVX > 0 && n->as_Vector()->length() == 8);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed8S" %}
+  ins_encode %{
+    bool vector256 = false;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl16S_reg(vecY dst, vecY src, vecS shift) %{
+  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed16S" %}
+  ins_encode %{
+    bool vector256 = true;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, $shift$$XMMRegister, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
+
+instruct vsrl16S_reg_imm(vecY dst, vecY src, immI8 shift) %{
+  predicate(UseAVX > 1 && n->as_Vector()->length() == 16);
+  match(Set dst (URShiftVS src shift));
+  format %{ "vpsrlw  $dst,$src,$shift\t! logical right shift packed16S" %}
+  ins_encode %{
+    bool vector256 = true;
+    __ vpsrlw($dst$$XMMRegister, $src$$XMMRegister, (int)$shift$$constant, vector256);
+  %}
+  ins_pipe( pipe_slow );
+%}
 
 // Integers vector logical right shift
 instruct vsrl2I(vecD dst, vecS shift) %{
diff --git a/src/os/bsd/vm/perfMemory_bsd.cpp b/src/os/bsd/vm/perfMemory_bsd.cpp
index 123e6e289..cb9dada90 100644
--- a/src/os/bsd/vm/perfMemory_bsd.cpp
+++ b/src/os/bsd/vm/perfMemory_bsd.cpp
@@ -30,6 +30,7 @@
 #include "os_bsd.inline.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/perfMemory.hpp"
+#include "services/memTracker.hpp"
 #include "utilities/exceptions.hpp"
 
 // put OS-includes here
@@ -753,6 +754,10 @@ static char* mmap_create_shared(size_t size) {
   // clear the shared memory region
   (void)::memset((void*) mapAddress, 0, size);
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   return mapAddress;
 }
 
@@ -912,6 +917,10 @@ static void mmap_attach_shared(const char* user, int vmid, PerfMemory::PerfMemor
               "Could not map PerfMemory");
   }
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   *addr = mapAddress;
   *sizep = size;
 
diff --git a/src/os/linux/vm/perfMemory_linux.cpp b/src/os/linux/vm/perfMemory_linux.cpp
index 2adae8d18..b54c5db14 100644
--- a/src/os/linux/vm/perfMemory_linux.cpp
+++ b/src/os/linux/vm/perfMemory_linux.cpp
@@ -30,6 +30,7 @@
 #include "os_linux.inline.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/perfMemory.hpp"
+#include "services/memTracker.hpp"
 #include "utilities/exceptions.hpp"
 
 // put OS-includes here
@@ -753,6 +754,10 @@ static char* mmap_create_shared(size_t size) {
   // clear the shared memory region
   (void)::memset((void*) mapAddress, 0, size);
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   return mapAddress;
 }
 
@@ -912,6 +917,10 @@ static void mmap_attach_shared(const char* user, int vmid, PerfMemory::PerfMemor
               "Could not map PerfMemory");
   }
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   *addr = mapAddress;
   *sizep = size;
 
diff --git a/src/os/solaris/vm/os_solaris.cpp b/src/os/solaris/vm/os_solaris.cpp
index 0483dcbfb..e3415b8d8 100644
--- a/src/os/solaris/vm/os_solaris.cpp
+++ b/src/os/solaris/vm/os_solaris.cpp
@@ -55,6 +55,7 @@
 #include "runtime/threadCritical.hpp"
 #include "runtime/timer.hpp"
 #include "services/attachListener.hpp"
+#include "services/memTracker.hpp"
 #include "services/runtimeService.hpp"
 #include "thread_solaris.inline.hpp"
 #include "utilities/decoder.hpp"
@@ -3072,11 +3073,12 @@ char* os::pd_attempt_reserve_memory_at(size_t bytes, char* requested_addr) {
   // Since snv_84, Solaris attempts to honor the address hint - see 5003415.
   // Give it a try, if the kernel honors the hint we can return immediately.
   char* addr = Solaris::anon_mmap(requested_addr, bytes, 0, false);
+
   volatile int err = errno;
   if (addr == requested_addr) {
     return addr;
   } else if (addr != NULL) {
-    unmap_memory(addr, bytes);
+    pd_unmap_memory(addr, bytes);
   }
 
   if (PrintMiscellaneous && Verbose) {
diff --git a/src/os/solaris/vm/perfMemory_solaris.cpp b/src/os/solaris/vm/perfMemory_solaris.cpp
index c24789cbb..ebbc00b30 100644
--- a/src/os/solaris/vm/perfMemory_solaris.cpp
+++ b/src/os/solaris/vm/perfMemory_solaris.cpp
@@ -30,6 +30,7 @@
 #include "os_solaris.inline.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/perfMemory.hpp"
+#include "services/memTracker.hpp"
 #include "utilities/exceptions.hpp"
 
 // put OS-includes here
@@ -768,6 +769,10 @@ static char* mmap_create_shared(size_t size) {
   // clear the shared memory region
   (void)::memset((void*) mapAddress, 0, size);
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   return mapAddress;
 }
 
@@ -927,6 +932,10 @@ static void mmap_attach_shared(const char* user, int vmid, PerfMemory::PerfMemor
               "Could not map PerfMemory");
   }
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   *addr = mapAddress;
   *sizep = size;
 
diff --git a/src/os/windows/vm/perfMemory_windows.cpp b/src/os/windows/vm/perfMemory_windows.cpp
index 2c75539d8..061c9d84a 100644
--- a/src/os/windows/vm/perfMemory_windows.cpp
+++ b/src/os/windows/vm/perfMemory_windows.cpp
@@ -30,6 +30,7 @@
 #include "os_windows.inline.hpp"
 #include "runtime/handles.inline.hpp"
 #include "runtime/perfMemory.hpp"
+#include "services/memTracker.hpp"
 #include "utilities/exceptions.hpp"
 
 #include <windows.h>
@@ -1496,6 +1497,10 @@ static char* mapping_create_shared(size_t size) {
   // clear the shared memory region
   (void)memset(mapAddress, '\0', size);
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
   return (char*) mapAddress;
 }
 
@@ -1672,6 +1677,11 @@ static void open_file_mapping(const char* user, int vmid,
               "Could not map PerfMemory");
   }
 
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_reserve((address)mapAddress, size, CURRENT_PC);
+  MemTracker::record_virtual_memory_type((address)mapAddress, mtInternal);
+
+
   *addrp = (char*)mapAddress;
   *sizep = size;
 
@@ -1824,6 +1834,8 @@ void PerfMemory::detach(char* addr, size_t bytes, TRAPS) {
   }
 
   remove_file_mapping(addr);
+  // it does not go through os api, the operation has to record from here
+  MemTracker::record_virtual_memory_release((address)addr, bytes);
 }
 
 char* PerfMemory::backing_store_filename() {
diff --git a/src/share/vm/c1/c1_GraphBuilder.cpp b/src/share/vm/c1/c1_GraphBuilder.cpp
index 1bbdc5afb..941dd120a 100644
--- a/src/share/vm/c1/c1_GraphBuilder.cpp
+++ b/src/share/vm/c1/c1_GraphBuilder.cpp
@@ -1844,17 +1844,12 @@ void GraphBuilder::invoke(Bytecodes::Code code) {
         code == Bytecodes::_invokevirtual && target->is_final_method() ||
         code == Bytecodes::_invokedynamic) {
       ciMethod* inline_target = (cha_monomorphic_target != NULL) ? cha_monomorphic_target : target;
-      bool success = false;
-      if (target->is_method_handle_intrinsic()) {
-        // method handle invokes
-        success = try_method_handle_inline(target);
-      } else {
-        // static binding => check if callee is ok
-        success = try_inline(inline_target, (cha_monomorphic_target != NULL) || (exact_target != NULL), code, better_receiver);
-      }
-      CHECK_BAILOUT();
+      // static binding => check if callee is ok
+      bool success = try_inline(inline_target, (cha_monomorphic_target != NULL) || (exact_target != NULL), code, better_receiver);
 
+      CHECK_BAILOUT();
       clear_inline_bailout();
+
       if (success) {
         // Register dependence if JVMTI has either breakpoint
         // setting or hotswapping of methods capabilities since they may
@@ -3201,6 +3196,11 @@ bool GraphBuilder::try_inline(ciMethod* callee, bool holder_known, Bytecodes::Co
     return false;
   }
 
+  // method handle invokes
+  if (callee->is_method_handle_intrinsic()) {
+    return try_method_handle_inline(callee);
+  }
+
   // handle intrinsics
   if (callee->intrinsic_id() != vmIntrinsics::_none) {
     if (try_inline_intrinsics(callee)) {
@@ -3885,10 +3885,14 @@ bool GraphBuilder::try_method_handle_inline(ciMethod* callee) {
       ValueType* type = state()->stack_at(args_base)->type();
       if (type->is_constant()) {
         ciMethod* target = type->as_ObjectType()->constant_value()->as_method_handle()->get_vmtarget();
-        guarantee(!target->is_method_handle_intrinsic(), "should not happen");  // XXX remove
-        Bytecodes::Code bc = target->is_static() ? Bytecodes::_invokestatic : Bytecodes::_invokevirtual;
-        if (try_inline(target, /*holder_known*/ true, bc)) {
-          return true;
+        // We don't do CHA here so only inline static and statically bindable methods.
+        if (target->is_static() || target->can_be_statically_bound()) {
+          Bytecodes::Code bc = target->is_static() ? Bytecodes::_invokestatic : Bytecodes::_invokevirtual;
+          if (try_inline(target, /*holder_known*/ true, bc)) {
+            return true;
+          }
+        } else {
+          print_inlining(target, "not static or statically bindable", /*success*/ false);
         }
       } else {
         print_inlining(callee, "receiver not constant", /*success*/ false);
@@ -3941,9 +3945,14 @@ bool GraphBuilder::try_method_handle_inline(ciMethod* callee) {
             }
             j += t->size();  // long and double take two slots
           }
-          Bytecodes::Code bc = target->is_static() ? Bytecodes::_invokestatic : Bytecodes::_invokevirtual;
-          if (try_inline(target, /*holder_known*/ true, bc)) {
-            return true;
+          // We don't do CHA here so only inline static and statically bindable methods.
+          if (target->is_static() || target->can_be_statically_bound()) {
+            Bytecodes::Code bc = target->is_static() ? Bytecodes::_invokestatic : Bytecodes::_invokevirtual;
+            if (try_inline(target, /*holder_known*/ true, bc)) {
+              return true;
+            }
+          } else {
+            print_inlining(target, "not static or statically bindable", /*success*/ false);
           }
         }
       } else {
diff --git a/src/share/vm/classfile/vmSymbols.hpp b/src/share/vm/classfile/vmSymbols.hpp
index 06fdb35be..2febc7b56 100644
--- a/src/share/vm/classfile/vmSymbols.hpp
+++ b/src/share/vm/classfile/vmSymbols.hpp
@@ -110,6 +110,7 @@
   template(sun_jkernel_DownloadManager,               "sun/jkernel/DownloadManager")              \
   template(getBootClassPathEntryForClass_name,        "getBootClassPathEntryForClass")            \
   template(sun_misc_PostVMInitHook,                   "sun/misc/PostVMInitHook")                  \
+  template(sun_misc_Launcher_ExtClassLoader,          "sun/misc/Launcher$ExtClassLoader")         \
                                                                                                   \
   /* Java runtime version access */                                                               \
   template(sun_misc_Version,                          "sun/misc/Version")                         \
@@ -723,6 +724,21 @@
   /* java/lang/ref/Reference */                                                                                         \
   do_intrinsic(_Reference_get,            java_lang_ref_Reference, get_name,    void_object_signature, F_R)             \
                                                                                                                         \
+  /* support for com.sum.crypto.provider.AESCrypt and some of its callers */                                            \
+  do_class(com_sun_crypto_provider_aescrypt,      "com/sun/crypto/provider/AESCrypt")                                   \
+  do_intrinsic(_aescrypt_encryptBlock, com_sun_crypto_provider_aescrypt, encryptBlock_name, byteArray_int_byteArray_int_signature, F_R)   \
+  do_intrinsic(_aescrypt_decryptBlock, com_sun_crypto_provider_aescrypt, decryptBlock_name, byteArray_int_byteArray_int_signature, F_R)   \
+   do_name(     encryptBlock_name,                                 "encryptBlock")                                      \
+   do_name(     decryptBlock_name,                                 "decryptBlock")                                      \
+   do_signature(byteArray_int_byteArray_int_signature,             "([BI[BI)V")                                         \
+                                                                                                                        \
+  do_class(com_sun_crypto_provider_cipherBlockChaining,            "com/sun/crypto/provider/CipherBlockChaining")       \
+   do_intrinsic(_cipherBlockChaining_encryptAESCrypt, com_sun_crypto_provider_cipherBlockChaining, encrypt_name, byteArray_int_int_byteArray_int_signature, F_R)   \
+   do_intrinsic(_cipherBlockChaining_decryptAESCrypt, com_sun_crypto_provider_cipherBlockChaining, decrypt_name, byteArray_int_int_byteArray_int_signature, F_R)   \
+   do_name(     encrypt_name,                                      "encrypt")                                           \
+   do_name(     decrypt_name,                                      "decrypt")                                           \
+   do_signature(byteArray_int_int_byteArray_int_signature,         "([BII[BI)V")                                        \
+                                                                                                                        \
   /* support for sun.misc.Unsafe */                                                                                     \
   do_class(sun_misc_Unsafe,               "sun/misc/Unsafe")                                                            \
                                                                                                                         \
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp b/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp
new file mode 100644
index 000000000..01e0e8745
--- /dev/null
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp
@@ -0,0 +1,175 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+#include "precompiled.hpp"
+#include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
+#include "gc_implementation/concurrentMarkSweep/freeChunk.hpp"
+#include "memory/freeBlockDictionary.hpp"
+#include "memory/sharedHeap.hpp"
+#include "runtime/globals.hpp"
+#include "runtime/mutex.hpp"
+#include "runtime/vmThread.hpp"
+
+template <>
+void AdaptiveFreeList<FreeChunk>::print_on(outputStream* st, const char* c) const {
+  if (c != NULL) {
+    st->print("%16s", c);
+  } else {
+    st->print(SIZE_FORMAT_W(16), size());
+  }
+  st->print("\t"
+           SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t"
+           SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\n",
+           bfr_surp(),             surplus(),             desired(),             prev_sweep(),           before_sweep(),
+           count(),               coal_births(),          coal_deaths(),          split_births(),         split_deaths());
+}
+
+template <class Chunk>
+AdaptiveFreeList<Chunk>::AdaptiveFreeList() : FreeList<Chunk>(), _hint(0) {
+  init_statistics();
+}
+
+template <class Chunk>
+AdaptiveFreeList<Chunk>::AdaptiveFreeList(Chunk* fc) : FreeList<Chunk>(fc), _hint(0) {
+  init_statistics();
+#ifndef PRODUCT
+  _allocation_stats.set_returned_bytes(size() * HeapWordSize);
+#endif
+}
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::initialize() {
+  FreeList<Chunk>::initialize();
+  set_hint(0);
+  init_statistics(true /* split_birth */);
+}
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::reset(size_t hint) {
+  FreeList<Chunk>::reset();
+  set_hint(hint);
+}
+
+#ifndef PRODUCT
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::assert_proper_lock_protection_work() const {
+  assert(protecting_lock() != NULL, "Don't call this directly");
+  assert(ParallelGCThreads > 0, "Don't call this directly");
+  Thread* thr = Thread::current();
+  if (thr->is_VM_thread() || thr->is_ConcurrentGC_thread()) {
+    // assert that we are holding the freelist lock
+  } else if (thr->is_GC_task_thread()) {
+    assert(protecting_lock()->owned_by_self(), "FreeList RACE DETECTED");
+  } else if (thr->is_Java_thread()) {
+    assert(!SafepointSynchronize::is_at_safepoint(), "Should not be executing");
+  } else {
+    ShouldNotReachHere();  // unaccounted thread type?
+  }
+}
+#endif
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::init_statistics(bool split_birth) {
+  _allocation_stats.initialize(split_birth);
+}
+
+template <class Chunk>
+size_t AdaptiveFreeList<Chunk>::get_better_size() {
+
+  // A candidate chunk has been found.  If it is already under
+  // populated and there is a hinT, REturn the hint().  Else
+  // return the size of this chunk.
+  if (surplus() <= 0) {
+    if (hint() != 0) {
+      return hint();
+    } else {
+      return size();
+    }
+  } else {
+    // This list has a surplus so use it.
+    return size();
+  }
+}
+
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::return_chunk_at_head(Chunk* chunk) {
+  assert_proper_lock_protection();
+  return_chunk_at_head(chunk, true);
+}
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::return_chunk_at_head(Chunk* chunk, bool record_return) {
+  FreeList<Chunk>::return_chunk_at_head(chunk, record_return);
+#ifdef ASSERT
+  if (record_return) {
+    increment_returned_bytes_by(size()*HeapWordSize);
+  }
+#endif
+}
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::return_chunk_at_tail(Chunk* chunk) {
+  return_chunk_at_tail(chunk, true);
+}
+
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::return_chunk_at_tail(Chunk* chunk, bool record_return) {
+  FreeList<Chunk>::return_chunk_at_tail(chunk, record_return);
+#ifdef ASSERT
+  if (record_return) {
+    increment_returned_bytes_by(size()*HeapWordSize);
+  }
+#endif
+}
+
+#ifndef PRODUCT
+template <class Chunk>
+void AdaptiveFreeList<Chunk>::verify_stats() const {
+  // The +1 of the LH comparand is to allow some "looseness" in
+  // checking: we usually call this interface when adding a block
+  // and we'll subsequently update the stats; we cannot update the
+  // stats beforehand because in the case of the large-block BT
+  // dictionary for example, this might be the first block and
+  // in that case there would be no place that we could record
+  // the stats (which are kept in the block itself).
+  assert((_allocation_stats.prev_sweep() + _allocation_stats.split_births()
+          + _allocation_stats.coal_births() + 1)   // Total Production Stock + 1
+         >= (_allocation_stats.split_deaths() + _allocation_stats.coal_deaths()
+             + (ssize_t)count()),                // Total Current Stock + depletion
+         err_msg("FreeList " PTR_FORMAT " of size " SIZE_FORMAT
+                 " violates Conservation Principle: "
+                 "prev_sweep(" SIZE_FORMAT ")"
+                 " + split_births(" SIZE_FORMAT ")"
+                 " + coal_births(" SIZE_FORMAT ") + 1 >= "
+                 " split_deaths(" SIZE_FORMAT ")"
+                 " coal_deaths(" SIZE_FORMAT ")"
+                 " + count(" SSIZE_FORMAT ")",
+                 this, size(), _allocation_stats.prev_sweep(), _allocation_stats.split_births(),
+                 _allocation_stats.split_births(), _allocation_stats.split_deaths(),
+                 _allocation_stats.coal_deaths(), count()));
+}
+#endif
+
+// Needs to be after the definitions have been seen.
+template class AdaptiveFreeList<FreeChunk>;
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp b/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp
new file mode 100644
index 000000000..8b56bb11d
--- /dev/null
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp
@@ -0,0 +1,232 @@
+/*
+ * Copyright (c) 2001, 2010, 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_ADAPTIVEFREELIST_HPP
+#define SHARE_VM_MEMORY_ADAPTIVEFREELIST_HPP
+
+#include "memory/freeList.hpp"
+#include "gc_implementation/shared/allocationStats.hpp"
+
+class CompactibleFreeListSpace;
+
+// A class for maintaining a free list of Chunk's.  The FreeList
+// maintains a the structure of the list (head, tail, etc.) plus
+// statistics for allocations from the list.  The links between items
+// are not part of FreeList.  The statistics are
+// used to make decisions about coalescing Chunk's when they
+// are swept during collection.
+//
+// See the corresponding .cpp file for a description of the specifics
+// for that implementation.
+
+class Mutex;
+
+template <class Chunk>
+class AdaptiveFreeList : public FreeList<Chunk> {
+  friend class CompactibleFreeListSpace;
+  friend class VMStructs;
+  // friend class PrintTreeCensusClosure<Chunk, FreeList_t>;
+
+  size_t        _hint;          // next larger size list with a positive surplus
+
+  AllocationStats _allocation_stats; // allocation-related statistics
+
+ public:
+
+  AdaptiveFreeList();
+  AdaptiveFreeList(Chunk* fc);
+
+  using FreeList<Chunk>::assert_proper_lock_protection;
+#ifdef ASSERT
+  using FreeList<Chunk>::protecting_lock;
+#endif
+  using FreeList<Chunk>::count;
+  using FreeList<Chunk>::size;
+  using FreeList<Chunk>::verify_chunk_in_free_list;
+  using FreeList<Chunk>::getFirstNChunksFromList;
+  using FreeList<Chunk>::print_on;
+  void return_chunk_at_head(Chunk* fc, bool record_return);
+  void return_chunk_at_head(Chunk* fc);
+  void return_chunk_at_tail(Chunk* fc, bool record_return);
+  void return_chunk_at_tail(Chunk* fc);
+  using FreeList<Chunk>::return_chunk_at_tail;
+  using FreeList<Chunk>::remove_chunk;
+  using FreeList<Chunk>::prepend;
+  using FreeList<Chunk>::print_labels_on;
+  using FreeList<Chunk>::get_chunk_at_head;
+
+  // Initialize.
+  void initialize();
+
+  // Reset the head, tail, hint, and count of a free list.
+  void reset(size_t hint);
+
+  void assert_proper_lock_protection_work() const PRODUCT_RETURN;
+
+  void print_on(outputStream* st, const char* c = NULL) const;
+
+  size_t hint() const {
+    return _hint;
+  }
+  void set_hint(size_t v) {
+    assert_proper_lock_protection();
+    assert(v == 0 || size() < v, "Bad hint");
+    _hint = v;
+  }
+
+  size_t get_better_size();
+
+  // Accessors for statistics
+  void init_statistics(bool split_birth = false);
+
+  AllocationStats* allocation_stats() {
+    assert_proper_lock_protection();
+    return &_allocation_stats;
+  }
+
+  ssize_t desired() const {
+    return _allocation_stats.desired();
+  }
+  void set_desired(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_desired(v);
+  }
+  void compute_desired(float inter_sweep_current,
+                       float inter_sweep_estimate,
+                       float intra_sweep_estimate) {
+    assert_proper_lock_protection();
+    _allocation_stats.compute_desired(count(),
+                                      inter_sweep_current,
+                                      inter_sweep_estimate,
+                                      intra_sweep_estimate);
+  }
+  ssize_t coal_desired() const {
+    return _allocation_stats.coal_desired();
+  }
+  void set_coal_desired(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_coal_desired(v);
+  }
+
+  ssize_t surplus() const {
+    return _allocation_stats.surplus();
+  }
+  void set_surplus(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_surplus(v);
+  }
+  void increment_surplus() {
+    assert_proper_lock_protection();
+    _allocation_stats.increment_surplus();
+  }
+  void decrement_surplus() {
+    assert_proper_lock_protection();
+    _allocation_stats.decrement_surplus();
+  }
+
+  ssize_t bfr_surp() const {
+    return _allocation_stats.bfr_surp();
+  }
+  void set_bfr_surp(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_bfr_surp(v);
+  }
+  ssize_t prev_sweep() const {
+    return _allocation_stats.prev_sweep();
+  }
+  void set_prev_sweep(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_prev_sweep(v);
+  }
+  ssize_t before_sweep() const {
+    return _allocation_stats.before_sweep();
+  }
+  void set_before_sweep(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_before_sweep(v);
+  }
+
+  ssize_t coal_births() const {
+    return _allocation_stats.coal_births();
+  }
+  void set_coal_births(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_coal_births(v);
+  }
+  void increment_coal_births() {
+    assert_proper_lock_protection();
+    _allocation_stats.increment_coal_births();
+  }
+
+  ssize_t coal_deaths() const {
+    return _allocation_stats.coal_deaths();
+  }
+  void set_coal_deaths(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_coal_deaths(v);
+  }
+  void increment_coal_deaths() {
+    assert_proper_lock_protection();
+    _allocation_stats.increment_coal_deaths();
+  }
+
+  ssize_t split_births() const {
+    return _allocation_stats.split_births();
+  }
+  void set_split_births(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_split_births(v);
+  }
+  void increment_split_births() {
+    assert_proper_lock_protection();
+    _allocation_stats.increment_split_births();
+  }
+
+  ssize_t split_deaths() const {
+    return _allocation_stats.split_deaths();
+  }
+  void set_split_deaths(ssize_t v) {
+    assert_proper_lock_protection();
+    _allocation_stats.set_split_deaths(v);
+  }
+  void increment_split_deaths() {
+    assert_proper_lock_protection();
+    _allocation_stats.increment_split_deaths();
+  }
+
+#ifndef PRODUCT
+  // For debugging.  The "_returned_bytes" in all the lists are summed
+  // and compared with the total number of bytes swept during a
+  // collection.
+  size_t returned_bytes() const { return _allocation_stats.returned_bytes(); }
+  void set_returned_bytes(size_t v) { _allocation_stats.set_returned_bytes(v); }
+  void increment_returned_bytes_by(size_t v) {
+    _allocation_stats.set_returned_bytes(_allocation_stats.returned_bytes() + v);
+  }
+  // Stats verification
+  void verify_stats() const;
+#endif  // NOT PRODUCT
+};
+
+#endif // SHARE_VM_MEMORY_ADAPTIVEFREELIST_HPP
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp b/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp
index eac32b1ea..2f43c9876 100644
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.cpp
@@ -91,7 +91,7 @@ CompactibleFreeListSpace::CompactibleFreeListSpace(BlockOffsetSharedArray* bs,
   _collector(NULL)
 {
   assert(sizeof(FreeChunk) / BytesPerWord <= MinChunkSize,
-    "FreeChunk is larger than expected");
+         "FreeChunk is larger than expected");
   _bt.set_space(this);
   initialize(mr, SpaceDecorator::Clear, SpaceDecorator::Mangle);
   // We have all of "mr", all of which we place in the dictionary
@@ -101,14 +101,14 @@ CompactibleFreeListSpace::CompactibleFreeListSpace(BlockOffsetSharedArray* bs,
   // implementation, namely, the simple binary tree (splaying
   // temporarily disabled).
   switch (dictionaryChoice) {
+    case FreeBlockDictionary<FreeChunk>::dictionaryBinaryTree:
+      _dictionary = new BinaryTreeDictionary<FreeChunk, AdaptiveFreeList>(mr);
+      break;
     case FreeBlockDictionary<FreeChunk>::dictionarySplayTree:
     case FreeBlockDictionary<FreeChunk>::dictionarySkipList:
     default:
       warning("dictionaryChoice: selected option not understood; using"
               " default BinaryTreeDictionary implementation instead.");
-    case FreeBlockDictionary<FreeChunk>::dictionaryBinaryTree:
-      _dictionary = new BinaryTreeDictionary<FreeChunk>(mr, use_adaptive_freelists);
-      break;
   }
   assert(_dictionary != NULL, "CMS dictionary initialization");
   // The indexed free lists are initially all empty and are lazily
@@ -453,7 +453,7 @@ const {
   reportIndexedFreeListStatistics();
   gclog_or_tty->print_cr("Layout of Indexed Freelists");
   gclog_or_tty->print_cr("---------------------------");
-  FreeList<FreeChunk>::print_labels_on(st, "size");
+  AdaptiveFreeList<FreeChunk>::print_labels_on(st, "size");
   for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
     _indexedFreeList[i].print_on(gclog_or_tty);
     for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
@@ -1319,7 +1319,7 @@ FreeChunk* CompactibleFreeListSpace::getChunkFromGreater(size_t numWords) {
   size_t currSize = numWords + MinChunkSize;
   assert(currSize % MinObjAlignment == 0, "currSize should be aligned");
   for (i = currSize; i < IndexSetSize; i += IndexSetStride) {
-    FreeList<FreeChunk>* fl = &_indexedFreeList[i];
+    AdaptiveFreeList<FreeChunk>* fl = &_indexedFreeList[i];
     if (fl->head()) {
       ret = getFromListGreater(fl, numWords);
       assert(ret == NULL || ret->is_free(), "Should be returning a free chunk");
@@ -1702,7 +1702,9 @@ CompactibleFreeListSpace::returnChunkToDictionary(FreeChunk* chunk) {
   _dictionary->return_chunk(chunk);
 #ifndef PRODUCT
   if (CMSCollector::abstract_state() != CMSCollector::Sweeping) {
-    TreeChunk<FreeChunk>::as_TreeChunk(chunk)->list()->verify_stats();
+    TreeChunk<FreeChunk, AdaptiveFreeList>* tc = TreeChunk<FreeChunk, AdaptiveFreeList>::as_TreeChunk(chunk);
+    TreeList<FreeChunk, AdaptiveFreeList>* tl = tc->list();
+    tl->verify_stats();
   }
 #endif // PRODUCT
 }
@@ -1745,7 +1747,7 @@ CompactibleFreeListSpace::addChunkToFreeListsAtEndRecordingStats(
   {
     MutexLockerEx x(lock, Mutex::_no_safepoint_check_flag);
     ec = dictionary()->find_largest_dict();  // get largest block
-    if (ec != NULL && ec->end() == chunk) {
+    if (ec != NULL && ec->end() == (uintptr_t*) chunk) {
       // It's a coterminal block - we can coalesce.
       size_t old_size = ec->size();
       coalDeath(old_size);
@@ -1850,11 +1852,11 @@ FreeChunk* CompactibleFreeListSpace::bestFitSmall(size_t numWords) {
      the excess is >= MIN_CHUNK. */
   size_t start = align_object_size(numWords + MinChunkSize);
   if (start < IndexSetSize) {
-    FreeList<FreeChunk>* it   = _indexedFreeList;
+    AdaptiveFreeList<FreeChunk>* it   = _indexedFreeList;
     size_t    hint = _indexedFreeList[start].hint();
     while (hint < IndexSetSize) {
       assert(hint % MinObjAlignment == 0, "hint should be aligned");
-      FreeList<FreeChunk> *fl = &_indexedFreeList[hint];
+      AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[hint];
       if (fl->surplus() > 0 && fl->head() != NULL) {
         // Found a list with surplus, reset original hint
         // and split out a free chunk which is returned.
@@ -1873,7 +1875,7 @@ FreeChunk* CompactibleFreeListSpace::bestFitSmall(size_t numWords) {
 }
 
 /* Requires fl->size >= numWords + MinChunkSize */
-FreeChunk* CompactibleFreeListSpace::getFromListGreater(FreeList<FreeChunk>* fl,
+FreeChunk* CompactibleFreeListSpace::getFromListGreater(AdaptiveFreeList<FreeChunk>* fl,
   size_t numWords) {
   FreeChunk *curr = fl->head();
   size_t oldNumWords = curr->size();
@@ -2155,7 +2157,7 @@ void CompactibleFreeListSpace::beginSweepFLCensus(
   assert_locked();
   size_t i;
   for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
-    FreeList<FreeChunk>* fl = &_indexedFreeList[i];
+    AdaptiveFreeList<FreeChunk>* fl    = &_indexedFreeList[i];
     if (PrintFLSStatistics > 1) {
       gclog_or_tty->print("size[%d] : ", i);
     }
@@ -2174,7 +2176,7 @@ void CompactibleFreeListSpace::setFLSurplus() {
   assert_locked();
   size_t i;
   for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
-    FreeList<FreeChunk> *fl = &_indexedFreeList[i];
+    AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
     fl->set_surplus(fl->count() -
                     (ssize_t)((double)fl->desired() * CMSSmallSplitSurplusPercent));
   }
@@ -2185,7 +2187,7 @@ void CompactibleFreeListSpace::setFLHints() {
   size_t i;
   size_t h = IndexSetSize;
   for (i = IndexSetSize - 1; i != 0; i -= IndexSetStride) {
-    FreeList<FreeChunk> *fl = &_indexedFreeList[i];
+    AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
     fl->set_hint(h);
     if (fl->surplus() > 0) {
       h = i;
@@ -2197,7 +2199,7 @@ void CompactibleFreeListSpace::clearFLCensus() {
   assert_locked();
   size_t i;
   for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
-    FreeList<FreeChunk> *fl = &_indexedFreeList[i];
+    AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
     fl->set_prev_sweep(fl->count());
     fl->set_coal_births(0);
     fl->set_coal_deaths(0);
@@ -2224,7 +2226,7 @@ void CompactibleFreeListSpace::endSweepFLCensus(size_t sweep_count) {
 
 bool CompactibleFreeListSpace::coalOverPopulated(size_t size) {
   if (size < SmallForDictionary) {
-    FreeList<FreeChunk> *fl = &_indexedFreeList[size];
+    AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
     return (fl->coal_desired() < 0) ||
            ((int)fl->count() > fl->coal_desired());
   } else {
@@ -2234,14 +2236,14 @@ bool CompactibleFreeListSpace::coalOverPopulated(size_t size) {
 
 void CompactibleFreeListSpace::smallCoalBirth(size_t size) {
   assert(size < SmallForDictionary, "Size too large for indexed list");
-  FreeList<FreeChunk> *fl = &_indexedFreeList[size];
+  AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
   fl->increment_coal_births();
   fl->increment_surplus();
 }
 
 void CompactibleFreeListSpace::smallCoalDeath(size_t size) {
   assert(size < SmallForDictionary, "Size too large for indexed list");
-  FreeList<FreeChunk> *fl = &_indexedFreeList[size];
+  AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
   fl->increment_coal_deaths();
   fl->decrement_surplus();
 }
@@ -2250,7 +2252,7 @@ void CompactibleFreeListSpace::coalBirth(size_t size) {
   if (size  < SmallForDictionary) {
     smallCoalBirth(size);
   } else {
-    dictionary()->dict_census_udpate(size,
+    dictionary()->dict_census_update(size,
                                    false /* split */,
                                    true /* birth */);
   }
@@ -2260,7 +2262,7 @@ void CompactibleFreeListSpace::coalDeath(size_t size) {
   if(size  < SmallForDictionary) {
     smallCoalDeath(size);
   } else {
-    dictionary()->dict_census_udpate(size,
+    dictionary()->dict_census_update(size,
                                    false /* split */,
                                    false /* birth */);
   }
@@ -2268,14 +2270,14 @@ void CompactibleFreeListSpace::coalDeath(size_t size) {
 
 void CompactibleFreeListSpace::smallSplitBirth(size_t size) {
   assert(size < SmallForDictionary, "Size too large for indexed list");
-  FreeList<FreeChunk> *fl = &_indexedFreeList[size];
+  AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
   fl->increment_split_births();
   fl->increment_surplus();
 }
 
 void CompactibleFreeListSpace::smallSplitDeath(size_t size) {
   assert(size < SmallForDictionary, "Size too large for indexed list");
-  FreeList<FreeChunk> *fl = &_indexedFreeList[size];
+  AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
   fl->increment_split_deaths();
   fl->decrement_surplus();
 }
@@ -2284,7 +2286,7 @@ void CompactibleFreeListSpace::split_birth(size_t size) {
   if (size  < SmallForDictionary) {
     smallSplitBirth(size);
   } else {
-    dictionary()->dict_census_udpate(size,
+    dictionary()->dict_census_update(size,
                                    true /* split */,
                                    true /* birth */);
   }
@@ -2294,7 +2296,7 @@ void CompactibleFreeListSpace::splitDeath(size_t size) {
   if (size  < SmallForDictionary) {
     smallSplitDeath(size);
   } else {
-    dictionary()->dict_census_udpate(size,
+    dictionary()->dict_census_update(size,
                                    true /* split */,
                                    false /* birth */);
   }
@@ -2517,10 +2519,10 @@ void CompactibleFreeListSpace::verifyIndexedFreeList(size_t size) const {
 
 #ifndef PRODUCT
 void CompactibleFreeListSpace::check_free_list_consistency() const {
-  assert(_dictionary->min_size() <= IndexSetSize,
+  assert((TreeChunk<FreeChunk, AdaptiveFreeList>::min_size() <= IndexSetSize),
     "Some sizes can't be allocated without recourse to"
     " linear allocation buffers");
-  assert(BinaryTreeDictionary<FreeChunk>::min_tree_chunk_size*HeapWordSize == sizeof(TreeChunk<FreeChunk>),
+  assert((TreeChunk<FreeChunk, AdaptiveFreeList>::min_size()*HeapWordSize == sizeof(TreeChunk<FreeChunk, AdaptiveFreeList>)),
     "else MIN_TREE_CHUNK_SIZE is wrong");
   assert(IndexSetStart != 0, "IndexSetStart not initialized");
   assert(IndexSetStride != 0, "IndexSetStride not initialized");
@@ -2529,15 +2531,15 @@ void CompactibleFreeListSpace::check_free_list_consistency() const {
 
 void CompactibleFreeListSpace::printFLCensus(size_t sweep_count) const {
   assert_lock_strong(&_freelistLock);
-  FreeList<FreeChunk> total;
+  AdaptiveFreeList<FreeChunk> total;
   gclog_or_tty->print("end sweep# " SIZE_FORMAT "\n", sweep_count);
-  FreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
+  AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
   size_t total_free = 0;
   for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
-    const FreeList<FreeChunk> *fl = &_indexedFreeList[i];
+    const AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
     total_free += fl->count() * fl->size();
     if (i % (40*IndexSetStride) == 0) {
-      FreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
+      AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
     }
     fl->print_on(gclog_or_tty);
     total.set_bfr_surp(    total.bfr_surp()     + fl->bfr_surp()    );
@@ -2620,7 +2622,7 @@ HeapWord* CFLS_LAB::alloc(size_t word_sz) {
     res = _cfls->getChunkFromDictionaryExact(word_sz);
     if (res == NULL) return NULL;
   } else {
-    FreeList<FreeChunk>* fl = &_indexedFreeList[word_sz];
+    AdaptiveFreeList<FreeChunk>* fl = &_indexedFreeList[word_sz];
     if (fl->count() == 0) {
       // Attempt to refill this local free list.
       get_from_global_pool(word_sz, fl);
@@ -2640,7 +2642,7 @@ HeapWord* CFLS_LAB::alloc(size_t word_sz) {
 
 // Get a chunk of blocks of the right size and update related
 // book-keeping stats
-void CFLS_LAB::get_from_global_pool(size_t word_sz, FreeList<FreeChunk>* fl) {
+void CFLS_LAB::get_from_global_pool(size_t word_sz, AdaptiveFreeList<FreeChunk>* fl) {
   // Get the #blocks we want to claim
   size_t n_blks = (size_t)_blocks_to_claim[word_sz].average();
   assert(n_blks > 0, "Error");
@@ -2722,7 +2724,7 @@ void CFLS_LAB::retire(int tid) {
         if (num_retire > 0) {
           _cfls->_indexedFreeList[i].prepend(&_indexedFreeList[i]);
           // Reset this list.
-          _indexedFreeList[i] = FreeList<FreeChunk>();
+          _indexedFreeList[i] = AdaptiveFreeList<FreeChunk>();
           _indexedFreeList[i].set_size(i);
         }
       }
@@ -2736,7 +2738,7 @@ void CFLS_LAB::retire(int tid) {
   }
 }
 
-void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n, FreeList<FreeChunk>* fl) {
+void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl) {
   assert(fl->count() == 0, "Precondition.");
   assert(word_sz < CompactibleFreeListSpace::IndexSetSize,
          "Precondition");
@@ -2752,12 +2754,12 @@ void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n
          (cur_sz < CompactibleFreeListSpace::IndexSetSize) &&
          (CMSSplitIndexedFreeListBlocks || k <= 1);
          k++, cur_sz = k * word_sz) {
-      FreeList<FreeChunk> fl_for_cur_sz;  // Empty.
+      AdaptiveFreeList<FreeChunk> fl_for_cur_sz;  // Empty.
       fl_for_cur_sz.set_size(cur_sz);
       {
         MutexLockerEx x(_indexedFreeListParLocks[cur_sz],
                         Mutex::_no_safepoint_check_flag);
-        FreeList<FreeChunk>* gfl = &_indexedFreeList[cur_sz];
+        AdaptiveFreeList<FreeChunk>* gfl = &_indexedFreeList[cur_sz];
         if (gfl->count() != 0) {
           // nn is the number of chunks of size cur_sz that
           // we'd need to split k-ways each, in order to create
@@ -2832,12 +2834,11 @@ void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n
     MutexLockerEx x(parDictionaryAllocLock(),
                     Mutex::_no_safepoint_check_flag);
     while (n > 0) {
-      fc = dictionary()->get_chunk(MAX2(n * word_sz,
-                                  _dictionary->min_size()),
+      fc = dictionary()->get_chunk(MAX2(n * word_sz, _dictionary->min_size()),
                                   FreeBlockDictionary<FreeChunk>::atLeast);
       if (fc != NULL) {
         _bt.allocated((HeapWord*)fc, fc->size(), true /* reducing */);  // update _unallocated_blk
-        dictionary()->dict_census_udpate(fc->size(),
+        dictionary()->dict_census_update(fc->size(),
                                        true /*split*/,
                                        false /*birth*/);
         break;
@@ -2890,7 +2891,7 @@ void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n
       fc->set_size(prefix_size);
       if (rem >= IndexSetSize) {
         returnChunkToDictionary(rem_fc);
-        dictionary()->dict_census_udpate(rem, true /*split*/, true /*birth*/);
+        dictionary()->dict_census_update(rem, true /*split*/, true /*birth*/);
         rem_fc = NULL;
       }
       // Otherwise, return it to the small list below.
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp b/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp
index 4d247356d..1b3d93ed2 100644
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp
@@ -25,6 +25,7 @@
 #ifndef SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_COMPACTIBLEFREELISTSPACE_HPP
 #define SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_COMPACTIBLEFREELISTSPACE_HPP
 
+#include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
 #include "gc_implementation/concurrentMarkSweep/promotionInfo.hpp"
 #include "memory/binaryTreeDictionary.hpp"
 #include "memory/blockOffsetTable.inline.hpp"
@@ -38,6 +39,7 @@
 class CompactibleFreeListSpace;
 class BlkClosure;
 class BlkClosureCareful;
+class FreeChunk;
 class UpwardsObjectClosure;
 class ObjectClosureCareful;
 class Klass;
@@ -131,7 +133,7 @@ class CompactibleFreeListSpace: public CompactibleSpace {
   FreeBlockDictionary<FreeChunk>::DictionaryChoice _dictionaryChoice;
   FreeBlockDictionary<FreeChunk>* _dictionary;    // ptr to dictionary for large size blocks
 
-  FreeList<FreeChunk> _indexedFreeList[IndexSetSize];
+  AdaptiveFreeList<FreeChunk> _indexedFreeList[IndexSetSize];
                                        // indexed array for small size blocks
   // allocation stategy
   bool       _fitStrategy;      // Use best fit strategy.
@@ -168,7 +170,7 @@ class CompactibleFreeListSpace: public CompactibleSpace {
   // If the count of "fl" is negative, it's absolute value indicates a
   // number of free chunks that had been previously "borrowed" from global
   // list of size "word_sz", and must now be decremented.
-  void par_get_chunk_of_blocks(size_t word_sz, size_t n, FreeList<FreeChunk>* fl);
+  void par_get_chunk_of_blocks(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl);
 
   // Allocation helper functions
   // Allocate using a strategy that takes from the indexed free lists
@@ -214,7 +216,7 @@ class CompactibleFreeListSpace: public CompactibleSpace {
   // and return it.  The split off remainder is returned to
   // the free lists.  The old name for getFromListGreater
   // was lookInListGreater.
-  FreeChunk* getFromListGreater(FreeList<FreeChunk>* fl, size_t numWords);
+  FreeChunk* getFromListGreater(AdaptiveFreeList<FreeChunk>* fl, size_t numWords);
   // Get a chunk in the indexed free list or dictionary,
   // by considering a larger chunk and splitting it.
   FreeChunk* getChunkFromGreater(size_t numWords);
@@ -621,7 +623,7 @@ class CFLS_LAB : public CHeapObj<mtGC> {
   CompactibleFreeListSpace* _cfls;
 
   // Our local free lists.
-  FreeList<FreeChunk> _indexedFreeList[CompactibleFreeListSpace::IndexSetSize];
+  AdaptiveFreeList<FreeChunk> _indexedFreeList[CompactibleFreeListSpace::IndexSetSize];
 
   // Initialized from a command-line arg.
 
@@ -634,7 +636,7 @@ class CFLS_LAB : public CHeapObj<mtGC> {
   size_t        _num_blocks        [CompactibleFreeListSpace::IndexSetSize];
 
   // Internal work method
-  void get_from_global_pool(size_t word_sz, FreeList<FreeChunk>* fl);
+  void get_from_global_pool(size_t word_sz, AdaptiveFreeList<FreeChunk>* fl);
 
 public:
   CFLS_LAB(CompactibleFreeListSpace* cfls);
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp
index ac8b70017..475f2b8fe 100644
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.cpp
@@ -9143,7 +9143,7 @@ void ASConcurrentMarkSweepGeneration::shrink_by(size_t desired_bytes) {
     size_t shrinkable_size_in_bytes = chunk_at_end->size();
     size_t aligned_shrinkable_size_in_bytes =
       align_size_down(shrinkable_size_in_bytes, os::vm_page_size());
-    assert(unallocated_start <= chunk_at_end->end(),
+    assert(unallocated_start <= (HeapWord*) chunk_at_end->end(),
       "Inconsistent chunk at end of space");
     size_t bytes = MIN2(desired_bytes, aligned_shrinkable_size_in_bytes);
     size_t word_size_before = heap_word_size(_virtual_space.committed_size());
@@ -9210,7 +9210,7 @@ void ASConcurrentMarkSweepGeneration::shrink_by(size_t desired_bytes) {
 
     assert(_cmsSpace->unallocated_block() <= _cmsSpace->end(),
       "Inconsistency at end of space");
-    assert(chunk_at_end->end() == _cmsSpace->end(),
+    assert(chunk_at_end->end() == (uintptr_t*) _cmsSpace->end(),
       "Shrinking is inconsistent");
     return;
   }
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/freeChunk.hpp b/src/share/vm/gc_implementation/concurrentMarkSweep/freeChunk.hpp
index 21934bca8..8376e8798 100644
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/freeChunk.hpp
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/freeChunk.hpp
@@ -133,7 +133,7 @@ class FreeChunk VALUE_OBJ_CLASS_SPEC {
   }
 
   // Return the address past the end of this chunk
-  HeapWord* end() const { return ((HeapWord*) this) + size(); }
+  uintptr_t* end() const { return ((uintptr_t*) this) + size(); }
 
   // debugging
   void verify()             const PRODUCT_RETURN;
diff --git a/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp b/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp
index c393a5528..b722779b8 100644
--- a/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp
+++ b/src/share/vm/gc_implementation/concurrentMarkSweep/vmStructs_cms.hpp
@@ -25,6 +25,8 @@
 #ifndef SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_VMSTRUCTS_CMS_HPP
 #define SHARE_VM_GC_IMPLEMENTATION_CONCURRENTMARKSWEEP_VMSTRUCTS_CMS_HPP
 
+typedef BinaryTreeDictionary<FreeChunk, AdaptiveFreeList> AFLBinaryTreeDictionary;
+
 #define VM_STRUCTS_CMS(nonstatic_field, \
                    volatile_nonstatic_field, \
                    static_field) \
@@ -38,14 +40,8 @@
   nonstatic_field(CMSCollector,                _markBitMap,                                   CMSBitMap)                             \
   nonstatic_field(ConcurrentMarkSweepGeneration, _cmsSpace,                                   CompactibleFreeListSpace*)             \
      static_field(ConcurrentMarkSweepThread,   _collector,                                    CMSCollector*)                         \
-  volatile_nonstatic_field(FreeChunk,          _size,                                         size_t)                                \
-  nonstatic_field(FreeChunk,                   _next,                                         FreeChunk*)                            \
-  nonstatic_field(FreeChunk,                   _prev,                                         FreeChunk*)                            \
   nonstatic_field(LinearAllocBlock,            _word_size,                                    size_t)                                \
-  nonstatic_field(FreeList<FreeChunk>,         _size,                                         size_t)                                \
-  nonstatic_field(FreeList<FreeChunk>,         _count,                                        ssize_t)                               \
-  nonstatic_field(BinaryTreeDictionary<FreeChunk>,_total_size,                                 size_t)                                \
-  nonstatic_field(CompactibleFreeListSpace,    _dictionary,                                   FreeBlockDictionary<FreeChunk>*)       \
+  nonstatic_field(AFLBinaryTreeDictionary,     _total_size,                                   size_t)                                \
   nonstatic_field(CompactibleFreeListSpace,    _indexedFreeList[0],                           FreeList<FreeChunk>)                   \
   nonstatic_field(CompactibleFreeListSpace,    _smallLinearAllocBlock,                        LinearAllocBlock)
 
@@ -60,19 +56,17 @@
   declare_toplevel_type(CMSCollector)                                     \
   declare_toplevel_type(CMSBitMap)                                        \
   declare_toplevel_type(FreeChunk)                                        \
+  declare_toplevel_type(Metablock)                                        \
   declare_toplevel_type(ConcurrentMarkSweepThread*)                       \
   declare_toplevel_type(ConcurrentMarkSweepGeneration*)                   \
   declare_toplevel_type(SurrogateLockerThread*)                           \
   declare_toplevel_type(CompactibleFreeListSpace*)                        \
   declare_toplevel_type(CMSCollector*)                                    \
-  declare_toplevel_type(FreeChunk*)                                       \
-  declare_toplevel_type(BinaryTreeDictionary<FreeChunk>*)                 \
-  declare_toplevel_type(FreeBlockDictionary<FreeChunk>*)                  \
-  declare_toplevel_type(FreeList<FreeChunk>*)                             \
-  declare_toplevel_type(FreeList<FreeChunk>)                              \
+  declare_toplevel_type(AFLBinaryTreeDictionary*)                         \
   declare_toplevel_type(LinearAllocBlock)                                 \
   declare_toplevel_type(FreeBlockDictionary<FreeChunk>)                   \
-            declare_type(BinaryTreeDictionary<FreeChunk>, FreeBlockDictionary<FreeChunk>)
+  declare_type(AFLBinaryTreeDictionary, FreeBlockDictionary<FreeChunk>)   \
+            declare_type(AFLBinaryTreeDictionary, FreeBlockDictionary<FreeChunk>) \
 
 #define VM_INT_CONSTANTS_CMS(declare_constant)                            \
   declare_constant(Generation::ConcurrentMarkSweep)                       \
diff --git a/src/share/vm/gc_implementation/shared/vmGCOperations.hpp b/src/share/vm/gc_implementation/shared/vmGCOperations.hpp
index 3966877e4..285ef97e3 100644
--- a/src/share/vm/gc_implementation/shared/vmGCOperations.hpp
+++ b/src/share/vm/gc_implementation/shared/vmGCOperations.hpp
@@ -191,7 +191,7 @@ class VM_GenCollectFull: public VM_GC_Operation {
 class VM_CollectForMetadataAllocation: public VM_GC_Operation {
  private:
   MetaWord*                _result;
-  size_t      _size;                       // size of object to be allocated
+  size_t                   _size;     // size of object to be allocated
   Metaspace::MetadataType  _mdtype;
   ClassLoaderData*         _loader_data;
  public:
diff --git a/src/share/vm/memory/allocation.cpp b/src/share/vm/memory/allocation.cpp
index 56c7b63fb..92b335b10 100644
--- a/src/share/vm/memory/allocation.cpp
+++ b/src/share/vm/memory/allocation.cpp
@@ -433,19 +433,18 @@ Arena::Arena() {
   NOT_PRODUCT(Atomic::inc(&_instance_count);)
 }
 
-Arena::Arena(Arena *a) : _chunk(a->_chunk), _hwm(a->_hwm), _max(a->_max), _first(a->_first) {
-  set_size_in_bytes(a->size_in_bytes());
-  NOT_PRODUCT(Atomic::inc(&_instance_count);)
-}
-
-
 Arena *Arena::move_contents(Arena *copy) {
   copy->destruct_contents();
   copy->_chunk = _chunk;
   copy->_hwm   = _hwm;
   copy->_max   = _max;
   copy->_first = _first;
-  copy->set_size_in_bytes(size_in_bytes());
+
+  // workaround rare racing condition, which could double count
+  // the arena size by native memory tracking
+  size_t size = size_in_bytes();
+  set_size_in_bytes(0);
+  copy->set_size_in_bytes(size);
   // Destroy original arena
   reset();
   return copy;            // Return Arena with contents
@@ -497,6 +496,9 @@ void Arena::destruct_contents() {
     char* end = _first->next() ? _first->top() : _hwm;
     free_malloced_objects(_first, _first->bottom(), end, _hwm);
   }
+  // reset size before chop to avoid a rare racing condition
+  // that can have total arena memory exceed total chunk memory
+  set_size_in_bytes(0);
   _first->chop();
   reset();
 }
diff --git a/src/share/vm/memory/allocation.hpp b/src/share/vm/memory/allocation.hpp
index 30662b8e5..892e50ddc 100644
--- a/src/share/vm/memory/allocation.hpp
+++ b/src/share/vm/memory/allocation.hpp
@@ -144,8 +144,10 @@ enum MemoryType {
   mtNMT               = 0x0A00,  // memory used by native memory tracking
   mtChunk             = 0x0B00,  // chunk that holds content of arenas
   mtJavaHeap          = 0x0C00,  // Java heap
-  mtDontTrack         = 0x0D00,  // memory we donot or cannot track
-  mt_number_of_types  = 0x000C,  // number of memory types
+  mtClassShared       = 0x0D00,  // class data sharing
+  mt_number_of_types  = 0x000D,  // number of memory types (mtDontTrack
+                                 // is not included as validate type)
+  mtDontTrack         = 0x0E00,  // memory we do not or cannot track
   mt_masks            = 0x7F00,
 
   // object type mask
@@ -342,7 +344,6 @@ protected:
  public:
   Arena();
   Arena(size_t init_size);
-  Arena(Arena *old);
   ~Arena();
   void  destruct_contents();
   char* hwm() const             { return _hwm; }
diff --git a/src/share/vm/memory/binaryTreeDictionary.cpp b/src/share/vm/memory/binaryTreeDictionary.cpp
index 08a9b0334..f79d149f1 100644
--- a/src/share/vm/memory/binaryTreeDictionary.cpp
+++ b/src/share/vm/memory/binaryTreeDictionary.cpp
@@ -25,9 +25,15 @@
 #include "precompiled.hpp"
 #include "gc_implementation/shared/allocationStats.hpp"
 #include "memory/binaryTreeDictionary.hpp"
+#include "memory/freeList.hpp"
+#include "memory/freeBlockDictionary.hpp"
+#include "memory/metablock.hpp"
+#include "memory/metachunk.hpp"
 #include "runtime/globals.hpp"
 #include "utilities/ostream.hpp"
 #ifndef SERIALGC
+#include "gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp"
+#include "gc_implementation/concurrentMarkSweep/freeChunk.hpp"
 #include "gc_implementation/shared/spaceDecorator.hpp"
 #include "gc_implementation/concurrentMarkSweep/freeChunk.hpp"
 #endif // SERIALGC
@@ -37,15 +43,18 @@
 // This is currently used in the Concurrent Mark&Sweep implementation.
 ////////////////////////////////////////////////////////////////////////////////
 
-template <class Chunk>
-TreeChunk<Chunk>* TreeChunk<Chunk>::as_TreeChunk(Chunk* fc) {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t TreeChunk<Chunk_t, FreeList_t>::_min_tree_chunk_size = sizeof(TreeChunk<Chunk_t,  FreeList_t>)/HeapWordSize;
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>* TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(Chunk_t* fc) {
   // Do some assertion checking here.
-  return (TreeChunk<Chunk>*) fc;
+  return (TreeChunk<Chunk_t, FreeList_t>*) fc;
 }
 
-template <class Chunk>
-void TreeChunk<Chunk>::verify_tree_chunk_list() const {
-  TreeChunk<Chunk>* nextTC = (TreeChunk<Chunk>*)next();
+template <class Chunk_t, template <class> class FreeList_t>
+void TreeChunk<Chunk_t, FreeList_t>::verify_tree_chunk_list() const {
+  TreeChunk<Chunk_t, FreeList_t>* nextTC = (TreeChunk<Chunk_t, FreeList_t>*)next();
   if (prev() != NULL) { // interior list node shouldn'r have tree fields
     guarantee(embedded_list()->parent() == NULL && embedded_list()->left() == NULL &&
               embedded_list()->right()  == NULL, "should be clear");
@@ -57,53 +66,113 @@ void TreeChunk<Chunk>::verify_tree_chunk_list() const {
   }
 }
 
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>::TreeList() {}
 
-template <class Chunk>
-TreeList<Chunk>* TreeList<Chunk>::as_TreeList(TreeChunk<Chunk>* tc) {
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>*
+TreeList<Chunk_t, FreeList_t>::as_TreeList(TreeChunk<Chunk_t,FreeList_t>* tc) {
   // This first free chunk in the list will be the tree list.
-  assert(tc->size() >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "Chunk is too small for a TreeChunk");
-  TreeList<Chunk>* tl = tc->embedded_list();
+  assert((tc->size() >= (TreeChunk<Chunk_t, FreeList_t>::min_size())),
+    "Chunk is too small for a TreeChunk");
+  TreeList<Chunk_t, FreeList_t>* tl = tc->embedded_list();
+  tl->initialize();
   tc->set_list(tl);
-#ifdef ASSERT
-  tl->set_protecting_lock(NULL);
-#endif
-  tl->set_hint(0);
   tl->set_size(tc->size());
   tl->link_head(tc);
   tl->link_tail(tc);
   tl->set_count(1);
-  tl->init_statistics(true /* split_birth */);
-  tl->set_parent(NULL);
-  tl->set_left(NULL);
-  tl->set_right(NULL);
+
   return tl;
 }
 
-template <class Chunk>
-TreeList<Chunk>* TreeList<Chunk>::as_TreeList(HeapWord* addr, size_t size) {
-  TreeChunk<Chunk>* tc = (TreeChunk<Chunk>*) addr;
-  assert(size >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "Chunk is too small for a TreeChunk");
-  // The space in the heap will have been mangled initially but
-  // is not remangled when a free chunk is returned to the free list
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>*
+get_chunk(size_t size, enum FreeBlockDictionary<Chunk_t>::Dither dither) {
+  FreeBlockDictionary<Chunk_t>::verify_par_locked();
+  Chunk_t* res = get_chunk_from_tree(size, dither);
+  assert(res == NULL || res->is_free(),
+         "Should be returning a free chunk");
+  assert(dither != FreeBlockDictionary<Chunk_t>::exactly ||
+         res->size() == size, "Not correct size");
+  return res;
+}
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>*
+TreeList<Chunk_t, FreeList_t>::as_TreeList(HeapWord* addr, size_t size) {
+  TreeChunk<Chunk_t, FreeList_t>* tc = (TreeChunk<Chunk_t, FreeList_t>*) addr;
+  assert((size >= TreeChunk<Chunk_t, FreeList_t>::min_size()),
+    "Chunk is too small for a TreeChunk");
+  // The space will have been mangled initially but
+  // is not remangled when a Chunk_t is returned to the free list
   // (since it is used to maintain the chunk on the free list).
-  assert((ZapUnusedHeapArea &&
-          SpaceMangler::is_mangled((HeapWord*) tc->size_addr()) &&
-          SpaceMangler::is_mangled((HeapWord*) tc->prev_addr()) &&
-          SpaceMangler::is_mangled((HeapWord*) tc->next_addr())) ||
-          (tc->size() == 0 && tc->prev() == NULL && tc->next() == NULL),
-    "Space should be clear or mangled");
+  tc->assert_is_mangled();
   tc->set_size(size);
   tc->link_prev(NULL);
   tc->link_next(NULL);
-  TreeList<Chunk>* tl = TreeList<Chunk>::as_TreeList(tc);
+  TreeList<Chunk_t, FreeList_t>* tl = TreeList<Chunk_t, FreeList_t>::as_TreeList(tc);
   return tl;
 }
 
-template <class Chunk>
-TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk>* tc) {
 
-  TreeList<Chunk>* retTL = this;
-  Chunk* list = head();
+#ifndef SERIALGC
+// Specialize for AdaptiveFreeList which tries to avoid
+// splitting a chunk of a size that is under populated in favor of
+// an over populated size.  The general get_better_list() just returns
+// the current list.
+template <>
+TreeList<FreeChunk, AdaptiveFreeList>*
+TreeList<FreeChunk, AdaptiveFreeList>::get_better_list(
+  BinaryTreeDictionary<FreeChunk, ::AdaptiveFreeList>* dictionary) {
+  // A candidate chunk has been found.  If it is already under
+  // populated, get a chunk associated with the hint for this
+  // chunk.
+
+  TreeList<FreeChunk, ::AdaptiveFreeList>* curTL = this;
+  if (surplus() <= 0) {
+    /* Use the hint to find a size with a surplus, and reset the hint. */
+    TreeList<FreeChunk, ::AdaptiveFreeList>* hintTL = this;
+    while (hintTL->hint() != 0) {
+      assert(hintTL->hint() > hintTL->size(),
+        "hint points in the wrong direction");
+      hintTL = dictionary->find_list(hintTL->hint());
+      assert(curTL != hintTL, "Infinite loop");
+      if (hintTL == NULL ||
+          hintTL == curTL /* Should not happen but protect against it */ ) {
+        // No useful hint.  Set the hint to NULL and go on.
+        curTL->set_hint(0);
+        break;
+      }
+      assert(hintTL->size() > curTL->size(), "hint is inconsistent");
+      if (hintTL->surplus() > 0) {
+        // The hint led to a list that has a surplus.  Use it.
+        // Set the hint for the candidate to an overpopulated
+        // size.
+        curTL->set_hint(hintTL->size());
+        // Change the candidate.
+        curTL = hintTL;
+        break;
+      }
+    }
+  }
+  return curTL;
+}
+#endif // SERIALGC
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>*
+TreeList<Chunk_t, FreeList_t>::get_better_list(
+  BinaryTreeDictionary<Chunk_t, FreeList_t>* dictionary) {
+  return this;
+}
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>* TreeList<Chunk_t, FreeList_t>::remove_chunk_replace_if_needed(TreeChunk<Chunk_t, FreeList_t>* tc) {
+
+  TreeList<Chunk_t, FreeList_t>* retTL = this;
+  Chunk_t* list = head();
   assert(!list || list != list->next(), "Chunk on list twice");
   assert(tc != NULL, "Chunk being removed is NULL");
   assert(parent() == NULL || this == parent()->left() ||
@@ -112,13 +181,13 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
 
-  Chunk* prevFC = tc->prev();
-  TreeChunk<Chunk>* nextTC = TreeChunk<Chunk>::as_TreeChunk(tc->next());
+  Chunk_t* prevFC = tc->prev();
+  TreeChunk<Chunk_t, FreeList_t>* nextTC = TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(tc->next());
   assert(list != NULL, "should have at least the target chunk");
 
   // Is this the first item on the list?
   if (tc == list) {
-    // The "getChunk..." functions for a TreeList<Chunk> will not return the
+    // The "getChunk..." functions for a TreeList<Chunk_t, FreeList_t> will not return the
     // first chunk in the list unless it is the last chunk in the list
     // because the first chunk is also acting as the tree node.
     // When coalescing happens, however, the first chunk in the a tree
@@ -127,8 +196,8 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
     // allocated when the sweeper yields (giving up the free list lock)
     // to allow mutator activity.  If this chunk is the first in the
     // list and is not the last in the list, do the work to copy the
-    // TreeList<Chunk> from the first chunk to the next chunk and update all
-    // the TreeList<Chunk> pointers in the chunks in the list.
+    // TreeList<Chunk_t, FreeList_t> from the first chunk to the next chunk and update all
+    // the TreeList<Chunk_t, FreeList_t> pointers in the chunks in the list.
     if (nextTC == NULL) {
       assert(prevFC == NULL, "Not last chunk in the list");
       set_tail(NULL);
@@ -141,11 +210,11 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
       // This can be slow for a long list.  Consider having
       // an option that does not allow the first chunk on the
       // list to be coalesced.
-      for (TreeChunk<Chunk>* curTC = nextTC; curTC != NULL;
-          curTC = TreeChunk<Chunk>::as_TreeChunk(curTC->next())) {
+      for (TreeChunk<Chunk_t, FreeList_t>* curTC = nextTC; curTC != NULL;
+          curTC = TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(curTC->next())) {
         curTC->set_list(retTL);
       }
-      // Fix the parent to point to the new TreeList<Chunk>.
+      // Fix the parent to point to the new TreeList<Chunk_t, FreeList_t>.
       if (retTL->parent() != NULL) {
         if (this == retTL->parent()->left()) {
           retTL->parent()->set_left(retTL);
@@ -176,9 +245,9 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
     prevFC->link_after(nextTC);
   }
 
-  // Below this point the embeded TreeList<Chunk> being used for the
+  // Below this point the embeded TreeList<Chunk_t, FreeList_t> being used for the
   // tree node may have changed. Don't use "this"
-  // TreeList<Chunk>*.
+  // TreeList<Chunk_t, FreeList_t>*.
   // chunk should still be a free chunk (bit set in _prev)
   assert(!retTL->head() || retTL->size() == retTL->head()->size(),
     "Wrong sized chunk in list");
@@ -188,7 +257,7 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
     tc->set_list(NULL);
     bool prev_found = false;
     bool next_found = false;
-    for (Chunk* curFC = retTL->head();
+    for (Chunk_t* curFC = retTL->head();
          curFC != NULL; curFC = curFC->next()) {
       assert(curFC != tc, "Chunk is still in list");
       if (curFC == prevFC) {
@@ -215,8 +284,8 @@ TreeList<Chunk>* TreeList<Chunk>::remove_chunk_replace_if_needed(TreeChunk<Chunk
   return retTL;
 }
 
-template <class Chunk>
-void TreeList<Chunk>::return_chunk_at_tail(TreeChunk<Chunk>* chunk) {
+template <class Chunk_t, template <class> class FreeList_t>
+void TreeList<Chunk_t, FreeList_t>::return_chunk_at_tail(TreeChunk<Chunk_t, FreeList_t>* chunk) {
   assert(chunk != NULL, "returning NULL chunk");
   assert(chunk->list() == this, "list should be set for chunk");
   assert(tail() != NULL, "The tree list is embedded in the first chunk");
@@ -225,12 +294,12 @@ void TreeList<Chunk>::return_chunk_at_tail(TreeChunk<Chunk>* chunk) {
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
 
-  Chunk* fc = tail();
+  Chunk_t* fc = tail();
   fc->link_after(chunk);
   link_tail(chunk);
 
   assert(!tail() || size() == tail()->size(), "Wrong sized chunk in list");
-  increment_count();
+  FreeList_t<Chunk_t>::increment_count();
   debug_only(increment_returned_bytes_by(chunk->size()*sizeof(HeapWord));)
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
@@ -238,10 +307,10 @@ void TreeList<Chunk>::return_chunk_at_tail(TreeChunk<Chunk>* chunk) {
 
 // Add this chunk at the head of the list.  "At the head of the list"
 // is defined to be after the chunk pointer to by head().  This is
-// because the TreeList<Chunk> is embedded in the first TreeChunk<Chunk> in the
-// list.  See the definition of TreeChunk<Chunk>.
-template <class Chunk>
-void TreeList<Chunk>::return_chunk_at_head(TreeChunk<Chunk>* chunk) {
+// because the TreeList<Chunk_t, FreeList_t> is embedded in the first TreeChunk<Chunk_t, FreeList_t> in the
+// list.  See the definition of TreeChunk<Chunk_t, FreeList_t>.
+template <class Chunk_t, template <class> class FreeList_t>
+void TreeList<Chunk_t, FreeList_t>::return_chunk_at_head(TreeChunk<Chunk_t, FreeList_t>* chunk) {
   assert(chunk->list() == this, "list should be set for chunk");
   assert(head() != NULL, "The tree list is embedded in the first chunk");
   assert(chunk != NULL, "returning NULL chunk");
@@ -249,7 +318,7 @@ void TreeList<Chunk>::return_chunk_at_head(TreeChunk<Chunk>* chunk) {
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
 
-  Chunk* fc = head()->next();
+  Chunk_t* fc = head()->next();
   if (fc != NULL) {
     chunk->link_after(fc);
   } else {
@@ -258,28 +327,38 @@ void TreeList<Chunk>::return_chunk_at_head(TreeChunk<Chunk>* chunk) {
   }
   head()->link_after(chunk);
   assert(!head() || size() == head()->size(), "Wrong sized chunk in list");
-  increment_count();
+  FreeList_t<Chunk_t>::increment_count();
   debug_only(increment_returned_bytes_by(chunk->size()*sizeof(HeapWord));)
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
 }
 
-template <class Chunk>
-TreeChunk<Chunk>* TreeList<Chunk>::head_as_TreeChunk() {
-  assert(head() == NULL || TreeChunk<Chunk>::as_TreeChunk(head())->list() == this,
+template <class Chunk_t, template <class> class FreeList_t>
+void TreeChunk<Chunk_t, FreeList_t>::assert_is_mangled() const {
+  assert((ZapUnusedHeapArea &&
+          SpaceMangler::is_mangled((HeapWord*) Chunk_t::size_addr()) &&
+          SpaceMangler::is_mangled((HeapWord*) Chunk_t::prev_addr()) &&
+          SpaceMangler::is_mangled((HeapWord*) Chunk_t::next_addr())) ||
+          (size() == 0 && prev() == NULL && next() == NULL),
+    "Space should be clear or mangled");
+}
+
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>* TreeList<Chunk_t, FreeList_t>::head_as_TreeChunk() {
+  assert(head() == NULL || (TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(head())->list() == this),
     "Wrong type of chunk?");
-  return TreeChunk<Chunk>::as_TreeChunk(head());
+  return TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(head());
 }
 
-template <class Chunk>
-TreeChunk<Chunk>* TreeList<Chunk>::first_available() {
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>* TreeList<Chunk_t, FreeList_t>::first_available() {
   assert(head() != NULL, "The head of the list cannot be NULL");
-  Chunk* fc = head()->next();
-  TreeChunk<Chunk>* retTC;
+  Chunk_t* fc = head()->next();
+  TreeChunk<Chunk_t, FreeList_t>* retTC;
   if (fc == NULL) {
     retTC = head_as_TreeChunk();
   } else {
-    retTC = TreeChunk<Chunk>::as_TreeChunk(fc);
+    retTC = TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(fc);
   }
   assert(retTC->list() == this, "Wrong type of chunk.");
   return retTC;
@@ -288,41 +367,32 @@ TreeChunk<Chunk>* TreeList<Chunk>::first_available() {
 // Returns the block with the largest heap address amongst
 // those in the list for this size; potentially slow and expensive,
 // use with caution!
-template <class Chunk>
-TreeChunk<Chunk>* TreeList<Chunk>::largest_address() {
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>* TreeList<Chunk_t, FreeList_t>::largest_address() {
   assert(head() != NULL, "The head of the list cannot be NULL");
-  Chunk* fc = head()->next();
-  TreeChunk<Chunk>* retTC;
+  Chunk_t* fc = head()->next();
+  TreeChunk<Chunk_t, FreeList_t>* retTC;
   if (fc == NULL) {
     retTC = head_as_TreeChunk();
   } else {
     // walk down the list and return the one with the highest
     // heap address among chunks of this size.
-    Chunk* last = fc;
+    Chunk_t* last = fc;
     while (fc->next() != NULL) {
       if ((HeapWord*)last < (HeapWord*)fc) {
         last = fc;
       }
       fc = fc->next();
     }
-    retTC = TreeChunk<Chunk>::as_TreeChunk(last);
+    retTC = TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(last);
   }
   assert(retTC->list() == this, "Wrong type of chunk.");
   return retTC;
 }
 
-template <class Chunk>
-BinaryTreeDictionary<Chunk>::BinaryTreeDictionary(bool adaptive_freelists, bool splay) :
-  _splay(splay), _adaptive_freelists(adaptive_freelists),
-  _total_size(0), _total_free_blocks(0), _root(0) {}
-
-template <class Chunk>
-BinaryTreeDictionary<Chunk>::BinaryTreeDictionary(MemRegion mr,
-                                           bool adaptive_freelists,
-                                           bool splay):
-  _adaptive_freelists(adaptive_freelists), _splay(splay)
-{
-  assert(mr.word_size() >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "minimum chunk size");
+template <class Chunk_t, template <class> class FreeList_t>
+BinaryTreeDictionary<Chunk_t, FreeList_t>::BinaryTreeDictionary(MemRegion mr) {
+  assert((mr.byte_size() > min_size()), "minimum chunk size");
 
   reset(mr);
   assert(root()->left() == NULL, "reset check failed");
@@ -333,52 +403,48 @@ BinaryTreeDictionary<Chunk>::BinaryTreeDictionary(MemRegion mr,
   assert(total_free_blocks() == 1, "reset check failed");
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::inc_total_size(size_t inc) {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::inc_total_size(size_t inc) {
   _total_size = _total_size + inc;
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::dec_total_size(size_t dec) {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::dec_total_size(size_t dec) {
   _total_size = _total_size - dec;
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::reset(MemRegion mr) {
-  assert(mr.word_size() >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "minimum chunk size");
-  set_root(TreeList<Chunk>::as_TreeList(mr.start(), mr.word_size()));
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::reset(MemRegion mr) {
+  assert((mr.byte_size() > min_size()), "minimum chunk size");
+  set_root(TreeList<Chunk_t, FreeList_t>::as_TreeList(mr.start(), mr.word_size()));
   set_total_size(mr.word_size());
   set_total_free_blocks(1);
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::reset(HeapWord* addr, size_t byte_size) {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::reset(HeapWord* addr, size_t byte_size) {
   MemRegion mr(addr, heap_word_size(byte_size));
   reset(mr);
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::reset() {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::reset() {
   set_root(NULL);
   set_total_size(0);
   set_total_free_blocks(0);
 }
 
 // Get a free block of size at least size from tree, or NULL.
-// If a splay step is requested, the removal algorithm (only) incorporates
-// a splay step as follows:
-// . the search proceeds down the tree looking for a possible
-//   match. At the (closest) matching location, an appropriate splay step is applied
-//   (zig, zig-zig or zig-zag). A chunk of the appropriate size is then returned
-//   if available, and if it's the last chunk, the node is deleted. A deteleted
-//   node is replaced in place by its tree successor.
-template <class Chunk>
-TreeChunk<Chunk>*
-BinaryTreeDictionary<Chunk>::get_chunk_from_tree(size_t size, enum FreeBlockDictionary<Chunk>::Dither dither, bool splay)
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>*
+BinaryTreeDictionary<Chunk_t, FreeList_t>::get_chunk_from_tree(
+                              size_t size,
+                              enum FreeBlockDictionary<Chunk_t>::Dither dither)
 {
-  TreeList<Chunk> *curTL, *prevTL;
-  TreeChunk<Chunk>* retTC = NULL;
-  assert(size >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "minimum chunk size");
+  TreeList<Chunk_t, FreeList_t> *curTL, *prevTL;
+  TreeChunk<Chunk_t, FreeList_t>* retTC = NULL;
+
+  assert((size >= min_size()), "minimum chunk size");
   if (FLSVerifyDictionary) {
     verify_tree();
   }
@@ -398,7 +464,7 @@ BinaryTreeDictionary<Chunk>::get_chunk_from_tree(size_t size, enum FreeBlockDict
   }
   if (curTL == NULL) { // couldn't find exact match
 
-    if (dither == FreeBlockDictionary<Chunk>::exactly) return NULL;
+    if (dither == FreeBlockDictionary<Chunk_t>::exactly) return NULL;
 
     // try and find the next larger size by walking back up the search path
     for (curTL = prevTL; curTL != NULL;) {
@@ -410,46 +476,9 @@ BinaryTreeDictionary<Chunk>::get_chunk_from_tree(size_t size, enum FreeBlockDict
   }
   if (curTL != NULL) {
     assert(curTL->size() >= size, "size inconsistency");
-    if (adaptive_freelists()) {
-
-      // A candidate chunk has been found.  If it is already under
-      // populated, get a chunk associated with the hint for this
-      // chunk.
-      if (curTL->surplus() <= 0) {
-        /* Use the hint to find a size with a surplus, and reset the hint. */
-        TreeList<Chunk>* hintTL = curTL;
-        while (hintTL->hint() != 0) {
-          assert(hintTL->hint() == 0 || hintTL->hint() > hintTL->size(),
-            "hint points in the wrong direction");
-          hintTL = find_list(hintTL->hint());
-          assert(curTL != hintTL, "Infinite loop");
-          if (hintTL == NULL ||
-              hintTL == curTL /* Should not happen but protect against it */ ) {
-            // No useful hint.  Set the hint to NULL and go on.
-            curTL->set_hint(0);
-            break;
-          }
-          assert(hintTL->size() > size, "hint is inconsistent");
-          if (hintTL->surplus() > 0) {
-            // The hint led to a list that has a surplus.  Use it.
-            // Set the hint for the candidate to an overpopulated
-            // size.
-            curTL->set_hint(hintTL->size());
-            // Change the candidate.
-            curTL = hintTL;
-            break;
-          }
-          // The evm code reset the hint of the candidate as
-          // at an interim point.  Why?  Seems like this leaves
-          // the hint pointing to a list that didn't work.
-          // curTL->set_hint(hintTL->size());
-        }
-      }
-    }
-    // don't waste time splaying if chunk's singleton
-    if (splay && curTL->head()->next() != NULL) {
-      semi_splay_step(curTL);
-    }
+
+    curTL = curTL->get_better_list(this);
+
     retTC = curTL->first_available();
     assert((retTC != NULL) && (curTL->count() > 0),
       "A list in the binary tree should not be NULL");
@@ -465,9 +494,9 @@ BinaryTreeDictionary<Chunk>::get_chunk_from_tree(size_t size, enum FreeBlockDict
   return retTC;
 }
 
-template <class Chunk>
-TreeList<Chunk>* BinaryTreeDictionary<Chunk>::find_list(size_t size) const {
-  TreeList<Chunk>* curTL;
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>* BinaryTreeDictionary<Chunk_t, FreeList_t>::find_list(size_t size) const {
+  TreeList<Chunk_t, FreeList_t>* curTL;
   for (curTL = root(); curTL != NULL;) {
     if (curTL->size() == size) {        // exact match
       break;
@@ -484,10 +513,10 @@ TreeList<Chunk>* BinaryTreeDictionary<Chunk>::find_list(size_t size) const {
 }
 
 
-template <class Chunk>
-bool BinaryTreeDictionary<Chunk>::verify_chunk_in_free_list(Chunk* tc) const {
+template <class Chunk_t, template <class> class FreeList_t>
+bool BinaryTreeDictionary<Chunk_t, FreeList_t>::verify_chunk_in_free_list(Chunk_t* tc) const {
   size_t size = tc->size();
-  TreeList<Chunk>* tl = find_list(size);
+  TreeList<Chunk_t, FreeList_t>* tl = find_list(size);
   if (tl == NULL) {
     return false;
   } else {
@@ -495,9 +524,9 @@ bool BinaryTreeDictionary<Chunk>::verify_chunk_in_free_list(Chunk* tc) const {
   }
 }
 
-template <class Chunk>
-Chunk* BinaryTreeDictionary<Chunk>::find_largest_dict() const {
-  TreeList<Chunk> *curTL = root();
+template <class Chunk_t, template <class> class FreeList_t>
+Chunk_t* BinaryTreeDictionary<Chunk_t, FreeList_t>::find_largest_dict() const {
+  TreeList<Chunk_t, FreeList_t> *curTL = root();
   if (curTL != NULL) {
     while(curTL->right() != NULL) curTL = curTL->right();
     return curTL->largest_address();
@@ -510,15 +539,15 @@ Chunk* BinaryTreeDictionary<Chunk>::find_largest_dict() const {
 // chunk in a list on a tree node, just unlink it.
 // If it is the last chunk in the list (the next link is NULL),
 // remove the node and repair the tree.
-template <class Chunk>
-TreeChunk<Chunk>*
-BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
+template <class Chunk_t, template <class> class FreeList_t>
+TreeChunk<Chunk_t, FreeList_t>*
+BinaryTreeDictionary<Chunk_t, FreeList_t>::remove_chunk_from_tree(TreeChunk<Chunk_t, FreeList_t>* tc) {
   assert(tc != NULL, "Should not call with a NULL chunk");
   assert(tc->is_free(), "Header is not marked correctly");
 
-  TreeList<Chunk> *newTL, *parentTL;
-  TreeChunk<Chunk>* retTC;
-  TreeList<Chunk>* tl = tc->list();
+  TreeList<Chunk_t, FreeList_t> *newTL, *parentTL;
+  TreeChunk<Chunk_t, FreeList_t>* retTC;
+  TreeList<Chunk_t, FreeList_t>* tl = tc->list();
   debug_only(
     bool removing_only_chunk = false;
     if (tl == _root) {
@@ -538,8 +567,8 @@ BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
 
   retTC = tc;
   // Removing this chunk can have the side effect of changing the node
-  // (TreeList<Chunk>*) in the tree.  If the node is the root, update it.
-  TreeList<Chunk>* replacementTL = tl->remove_chunk_replace_if_needed(tc);
+  // (TreeList<Chunk_t, FreeList_t>*) in the tree.  If the node is the root, update it.
+  TreeList<Chunk_t, FreeList_t>* replacementTL = tl->remove_chunk_replace_if_needed(tc);
   assert(tc->is_free(), "Chunk should still be free");
   assert(replacementTL->parent() == NULL ||
          replacementTL == replacementTL->parent()->left() ||
@@ -549,17 +578,18 @@ BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
     assert(replacementTL->parent() == NULL, "Incorrectly replacing root");
     set_root(replacementTL);
   }
-  debug_only(
+#ifdef ASSERT
     if (tl != replacementTL) {
       assert(replacementTL->head() != NULL,
         "If the tree list was replaced, it should not be a NULL list");
-      TreeList<Chunk>* rhl = replacementTL->head_as_TreeChunk()->list();
-      TreeList<Chunk>* rtl = TreeChunk<Chunk>::as_TreeChunk(replacementTL->tail())->list();
+      TreeList<Chunk_t, FreeList_t>* rhl = replacementTL->head_as_TreeChunk()->list();
+      TreeList<Chunk_t, FreeList_t>* rtl =
+        TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(replacementTL->tail())->list();
       assert(rhl == replacementTL, "Broken head");
       assert(rtl == replacementTL, "Broken tail");
       assert(replacementTL->size() == tc->size(),  "Broken size");
     }
-  )
+#endif
 
   // Does the tree need to be repaired?
   if (replacementTL->count() == 0) {
@@ -574,7 +604,7 @@ BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
     } else if (replacementTL->right() == NULL) {
       // right is NULL
       newTL = replacementTL->left();
-      debug_only(replacementTL->clearLeft();)
+      debug_only(replacementTL->clear_left();)
     } else {  // we have both children, so, by patriarchal convention,
               // my replacement is least node in right sub-tree
       complicated_splice = true;
@@ -623,7 +653,7 @@ BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
       newTL->set_right(replacementTL->right());
       debug_only(
         replacementTL->clear_right();
-        replacementTL->clearLeft();
+        replacementTL->clear_left();
       )
     }
     assert(replacementTL->right() == NULL &&
@@ -644,21 +674,21 @@ BinaryTreeDictionary<Chunk>::remove_chunk_from_tree(TreeChunk<Chunk>* tc) {
     verify_tree();
   }
   assert(!removing_only_chunk || _root == NULL, "root should be NULL");
-  return TreeChunk<Chunk>::as_TreeChunk(retTC);
+  return TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(retTC);
 }
 
 // Remove the leftmost node (lm) in the tree and return it.
 // If lm has a right child, link it to the left node of
 // the parent of lm.
-template <class Chunk>
-TreeList<Chunk>* BinaryTreeDictionary<Chunk>::remove_tree_minimum(TreeList<Chunk>* tl) {
+template <class Chunk_t, template <class> class FreeList_t>
+TreeList<Chunk_t, FreeList_t>* BinaryTreeDictionary<Chunk_t, FreeList_t>::remove_tree_minimum(TreeList<Chunk_t, FreeList_t>* tl) {
   assert(tl != NULL && tl->parent() != NULL, "really need a proper sub-tree");
   // locate the subtree minimum by walking down left branches
-  TreeList<Chunk>* curTL = tl;
+  TreeList<Chunk_t, FreeList_t>* curTL = tl;
   for (; curTL->left() != NULL; curTL = curTL->left());
   // obviously curTL now has at most one child, a right child
   if (curTL != root()) {  // Should this test just be removed?
-    TreeList<Chunk>* parentTL = curTL->parent();
+    TreeList<Chunk_t, FreeList_t>* parentTL = curTL->parent();
     if (parentTL->left() == curTL) { // curTL is a left child
       parentTL->set_left(curTL->right());
     } else {
@@ -685,31 +715,14 @@ TreeList<Chunk>* BinaryTreeDictionary<Chunk>::remove_tree_minimum(TreeList<Chunk
   return curTL;
 }
 
-// Based on a simplification of the algorithm by Sleator and Tarjan (JACM 1985).
-// The simplifications are the following:
-// . we splay only when we delete (not when we insert)
-// . we apply a single spay step per deletion/access
-// By doing such partial splaying, we reduce the amount of restructuring,
-// while getting a reasonably efficient search tree (we think).
-// [Measurements will be needed to (in)validate this expectation.]
-
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::semi_splay_step(TreeList<Chunk>* tc) {
-  // apply a semi-splay step at the given node:
-  // . if root, norting needs to be done
-  // . if child of root, splay once
-  // . else zig-zig or sig-zag depending on path from grandparent
-  if (root() == tc) return;
-  warning("*** Splaying not yet implemented; "
-          "tree operations may be inefficient ***");
-}
-
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::insert_chunk_in_tree(Chunk* fc) {
-  TreeList<Chunk> *curTL, *prevTL;
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::insert_chunk_in_tree(Chunk_t* fc) {
+  TreeList<Chunk_t, FreeList_t> *curTL, *prevTL;
   size_t size = fc->size();
 
-  assert(size >= BinaryTreeDictionary<Chunk>::min_tree_chunk_size, "too small to be a TreeList<Chunk>");
+  assert((size >= min_size()),
+    err_msg(SIZE_FORMAT " is too small to be a TreeChunk<Chunk_t, FreeList_t> " SIZE_FORMAT,
+      size, min_size()));
   if (FLSVerifyDictionary) {
     verify_tree();
   }
@@ -729,9 +742,9 @@ void BinaryTreeDictionary<Chunk>::insert_chunk_in_tree(Chunk* fc) {
       curTL = curTL->right();
     }
   }
-  TreeChunk<Chunk>* tc = TreeChunk<Chunk>::as_TreeChunk(fc);
+  TreeChunk<Chunk_t, FreeList_t>* tc = TreeChunk<Chunk_t, FreeList_t>::as_TreeChunk(fc);
   // This chunk is being returned to the binary tree.  Its embedded
-  // TreeList<Chunk> should be unused at this point.
+  // TreeList<Chunk_t, FreeList_t> should be unused at this point.
   tc->initialize();
   if (curTL != NULL) {          // exact match
     tc->set_list(curTL);
@@ -739,8 +752,8 @@ void BinaryTreeDictionary<Chunk>::insert_chunk_in_tree(Chunk* fc) {
   } else {                     // need a new node in tree
     tc->clear_next();
     tc->link_prev(NULL);
-    TreeList<Chunk>* newTL = TreeList<Chunk>::as_TreeList(tc);
-    assert(((TreeChunk<Chunk>*)tc)->list() == newTL,
+    TreeList<Chunk_t, FreeList_t>* newTL = TreeList<Chunk_t, FreeList_t>::as_TreeList(tc);
+    assert(((TreeChunk<Chunk_t, FreeList_t>*)tc)->list() == newTL,
       "List was not initialized correctly");
     if (prevTL == NULL) {      // we are the only tree node
       assert(root() == NULL, "control point invariant");
@@ -768,30 +781,30 @@ void BinaryTreeDictionary<Chunk>::insert_chunk_in_tree(Chunk* fc) {
   }
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::max_chunk_size() const {
-  FreeBlockDictionary<Chunk>::verify_par_locked();
-  TreeList<Chunk>* tc = root();
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::max_chunk_size() const {
+  FreeBlockDictionary<Chunk_t>::verify_par_locked();
+  TreeList<Chunk_t, FreeList_t>* tc = root();
   if (tc == NULL) return 0;
   for (; tc->right() != NULL; tc = tc->right());
   return tc->size();
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_list_length(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_list_length(TreeList<Chunk_t, FreeList_t>* tl) const {
   size_t res;
   res = tl->count();
 #ifdef ASSERT
   size_t cnt;
-  Chunk* tc = tl->head();
+  Chunk_t* tc = tl->head();
   for (cnt = 0; tc != NULL; tc = tc->next(), cnt++);
   assert(res == cnt, "The count is not being maintained correctly");
 #endif
   return res;
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_size_in_tree(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_size_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const {
   if (tl == NULL)
     return 0;
   return (tl->size() * total_list_length(tl)) +
@@ -799,8 +812,8 @@ size_t BinaryTreeDictionary<Chunk>::total_size_in_tree(TreeList<Chunk>* tl) cons
          total_size_in_tree(tl->right());
 }
 
-template <class Chunk>
-double BinaryTreeDictionary<Chunk>::sum_of_squared_block_sizes(TreeList<Chunk>* const tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+double BinaryTreeDictionary<Chunk_t, FreeList_t>::sum_of_squared_block_sizes(TreeList<Chunk_t, FreeList_t>* const tl) const {
   if (tl == NULL) {
     return 0.0;
   }
@@ -811,8 +824,8 @@ double BinaryTreeDictionary<Chunk>::sum_of_squared_block_sizes(TreeList<Chunk>*
   return curr;
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_free_blocks_in_tree(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_free_blocks_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const {
   if (tl == NULL)
     return 0;
   return total_list_length(tl) +
@@ -820,28 +833,28 @@ size_t BinaryTreeDictionary<Chunk>::total_free_blocks_in_tree(TreeList<Chunk>* t
          total_free_blocks_in_tree(tl->right());
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::num_free_blocks() const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::num_free_blocks() const {
   assert(total_free_blocks_in_tree(root()) == total_free_blocks(),
          "_total_free_blocks inconsistency");
   return total_free_blocks();
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::tree_height_helper(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::tree_height_helper(TreeList<Chunk_t, FreeList_t>* tl) const {
   if (tl == NULL)
     return 0;
   return 1 + MAX2(tree_height_helper(tl->left()),
                   tree_height_helper(tl->right()));
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::treeHeight() const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::tree_height() const {
   return tree_height_helper(root());
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_nodes_helper(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_nodes_helper(TreeList<Chunk_t, FreeList_t>* tl) const {
   if (tl == NULL) {
     return 0;
   }
@@ -849,14 +862,18 @@ size_t BinaryTreeDictionary<Chunk>::total_nodes_helper(TreeList<Chunk>* tl) cons
     total_nodes_helper(tl->right());
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_nodes_in_tree(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_nodes_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const {
   return total_nodes_helper(root());
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::dict_census_udpate(size_t size, bool split, bool birth){
-  TreeList<Chunk>* nd = find_list(size);
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::dict_census_update(size_t size, bool split, bool birth){}
+
+#ifndef SERIALGC
+template <>
+void BinaryTreeDictionary<FreeChunk, AdaptiveFreeList>::dict_census_update(size_t size, bool split, bool birth){
+  TreeList<FreeChunk, AdaptiveFreeList>* nd = find_list(size);
   if (nd) {
     if (split) {
       if (birth) {
@@ -882,16 +899,26 @@ void BinaryTreeDictionary<Chunk>::dict_census_udpate(size_t size, bool split, bo
   //   This is a birth associated with a LinAB.  The chunk
   //     for the LinAB is not in the dictionary.
 }
+#endif // SERIALGC
+
+template <class Chunk_t, template <class> class FreeList_t>
+bool BinaryTreeDictionary<Chunk_t, FreeList_t>::coal_dict_over_populated(size_t size) {
+  // For the general type of freelists, encourage coalescing by
+  // returning true.
+  return true;
+}
 
-template <class Chunk>
-bool BinaryTreeDictionary<Chunk>::coal_dict_over_populated(size_t size) {
+#ifndef SERIALGC
+template <>
+bool BinaryTreeDictionary<FreeChunk, AdaptiveFreeList>::coal_dict_over_populated(size_t size) {
   if (FLSAlwaysCoalesceLarge) return true;
 
-  TreeList<Chunk>* list_of_size = find_list(size);
+  TreeList<FreeChunk, AdaptiveFreeList>* list_of_size = find_list(size);
   // None of requested size implies overpopulated.
   return list_of_size == NULL || list_of_size->coal_desired() <= 0 ||
          list_of_size->count() > list_of_size->coal_desired();
 }
+#endif  // SERIALGC
 
 // Closures for walking the binary tree.
 //   do_list() walks the free list in a node applying the closure
@@ -899,19 +926,18 @@ bool BinaryTreeDictionary<Chunk>::coal_dict_over_populated(size_t size) {
 //   do_tree() walks the nodes in the binary tree applying do_list()
 //     to each list at each node.
 
-template <class Chunk>
+template <class Chunk_t, template <class> class FreeList_t>
 class TreeCensusClosure : public StackObj {
  protected:
-  virtual void do_list(FreeList<Chunk>* fl) = 0;
+  virtual void do_list(FreeList_t<Chunk_t>* fl) = 0;
  public:
-  virtual void do_tree(TreeList<Chunk>* tl) = 0;
+  virtual void do_tree(TreeList<Chunk_t, FreeList_t>* tl) = 0;
 };
 
-template <class Chunk>
-class AscendTreeCensusClosure : public TreeCensusClosure<Chunk> {
-  using TreeCensusClosure<Chunk>::do_list;
+template <class Chunk_t, template <class> class FreeList_t>
+class AscendTreeCensusClosure : public TreeCensusClosure<Chunk_t, FreeList_t> {
  public:
-  void do_tree(TreeList<Chunk>* tl) {
+  void do_tree(TreeList<Chunk_t, FreeList_t>* tl) {
     if (tl != NULL) {
       do_tree(tl->left());
       do_list(tl);
@@ -920,11 +946,10 @@ class AscendTreeCensusClosure : public TreeCensusClosure<Chunk> {
   }
 };
 
-template <class Chunk>
-class DescendTreeCensusClosure : public TreeCensusClosure<Chunk> {
-  using TreeCensusClosure<Chunk>::do_list;
+template <class Chunk_t, template <class> class FreeList_t>
+class DescendTreeCensusClosure : public TreeCensusClosure<Chunk_t, FreeList_t> {
  public:
-  void do_tree(TreeList<Chunk>* tl) {
+  void do_tree(TreeList<Chunk_t, FreeList_t>* tl) {
     if (tl != NULL) {
       do_tree(tl->right());
       do_list(tl);
@@ -935,8 +960,8 @@ class DescendTreeCensusClosure : public TreeCensusClosure<Chunk> {
 
 // For each list in the tree, calculate the desired, desired
 // coalesce, count before sweep, and surplus before sweep.
-template <class Chunk>
-class BeginSweepClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class BeginSweepClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
   double _percentage;
   float _inter_sweep_current;
   float _inter_sweep_estimate;
@@ -951,32 +976,36 @@ class BeginSweepClosure : public AscendTreeCensusClosure<Chunk> {
    _inter_sweep_estimate(inter_sweep_estimate),
    _intra_sweep_estimate(intra_sweep_estimate) { }
 
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList<Chunk_t>* fl) {}
+
+#ifndef SERIALGC
+  void do_list(AdaptiveFreeList<Chunk_t>* fl) {
     double coalSurplusPercent = _percentage;
     fl->compute_desired(_inter_sweep_current, _inter_sweep_estimate, _intra_sweep_estimate);
     fl->set_coal_desired((ssize_t)((double)fl->desired() * coalSurplusPercent));
     fl->set_before_sweep(fl->count());
     fl->set_bfr_surp(fl->surplus());
   }
+#endif // SERIALGC
 };
 
 // Used to search the tree until a condition is met.
 // Similar to TreeCensusClosure but searches the
 // tree and returns promptly when found.
 
-template <class Chunk>
+template <class Chunk_t, template <class> class FreeList_t>
 class TreeSearchClosure : public StackObj {
  protected:
-  virtual bool do_list(FreeList<Chunk>* fl) = 0;
+  virtual bool do_list(FreeList_t<Chunk_t>* fl) = 0;
  public:
-  virtual bool do_tree(TreeList<Chunk>* tl) = 0;
+  virtual bool do_tree(TreeList<Chunk_t, FreeList_t>* tl) = 0;
 };
 
 #if 0 //  Don't need this yet but here for symmetry.
-template <class Chunk>
-class AscendTreeSearchClosure : public TreeSearchClosure {
+template <class Chunk_t, template <class> class FreeList_t>
+class AscendTreeSearchClosure : public TreeSearchClosure<Chunk_t> {
  public:
-  bool do_tree(TreeList<Chunk>* tl) {
+  bool do_tree(TreeList<Chunk_t, FreeList_t>* tl) {
     if (tl != NULL) {
       if (do_tree(tl->left())) return true;
       if (do_list(tl)) return true;
@@ -987,11 +1016,10 @@ class AscendTreeSearchClosure : public TreeSearchClosure {
 };
 #endif
 
-template <class Chunk>
-class DescendTreeSearchClosure : public TreeSearchClosure<Chunk> {
-  using TreeSearchClosure<Chunk>::do_list;
+template <class Chunk_t, template <class> class FreeList_t>
+class DescendTreeSearchClosure : public TreeSearchClosure<Chunk_t, FreeList_t> {
  public:
-  bool do_tree(TreeList<Chunk>* tl) {
+  bool do_tree(TreeList<Chunk_t, FreeList_t>* tl) {
     if (tl != NULL) {
       if (do_tree(tl->right())) return true;
       if (do_list(tl)) return true;
@@ -1003,17 +1031,17 @@ class DescendTreeSearchClosure : public TreeSearchClosure<Chunk> {
 
 // Searches the tree for a chunk that ends at the
 // specified address.
-template <class Chunk>
-class EndTreeSearchClosure : public DescendTreeSearchClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class EndTreeSearchClosure : public DescendTreeSearchClosure<Chunk_t, FreeList_t> {
   HeapWord* _target;
-  Chunk* _found;
+  Chunk_t* _found;
 
  public:
   EndTreeSearchClosure(HeapWord* target) : _target(target), _found(NULL) {}
-  bool do_list(FreeList<Chunk>* fl) {
-    Chunk* item = fl->head();
+  bool do_list(FreeList_t<Chunk_t>* fl) {
+    Chunk_t* item = fl->head();
     while (item != NULL) {
-      if (item->end() == _target) {
+      if (item->end() == (uintptr_t*) _target) {
         _found = item;
         return true;
       }
@@ -1021,22 +1049,22 @@ class EndTreeSearchClosure : public DescendTreeSearchClosure<Chunk> {
     }
     return false;
   }
-  Chunk* found() { return _found; }
+  Chunk_t* found() { return _found; }
 };
 
-template <class Chunk>
-Chunk* BinaryTreeDictionary<Chunk>::find_chunk_ends_at(HeapWord* target) const {
-  EndTreeSearchClosure<Chunk> etsc(target);
+template <class Chunk_t, template <class> class FreeList_t>
+Chunk_t* BinaryTreeDictionary<Chunk_t, FreeList_t>::find_chunk_ends_at(HeapWord* target) const {
+  EndTreeSearchClosure<Chunk_t, FreeList_t> etsc(target);
   bool found_target = etsc.do_tree(root());
   assert(found_target || etsc.found() == NULL, "Consistency check");
   assert(!found_target || etsc.found() != NULL, "Consistency check");
   return etsc.found();
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::begin_sweep_dict_census(double coalSurplusPercent,
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::begin_sweep_dict_census(double coalSurplusPercent,
   float inter_sweep_current, float inter_sweep_estimate, float intra_sweep_estimate) {
-  BeginSweepClosure<Chunk> bsc(coalSurplusPercent, inter_sweep_current,
+  BeginSweepClosure<Chunk_t, FreeList_t> bsc(coalSurplusPercent, inter_sweep_current,
                                             inter_sweep_estimate,
                                             intra_sweep_estimate);
   bsc.do_tree(root());
@@ -1045,84 +1073,91 @@ void BinaryTreeDictionary<Chunk>::begin_sweep_dict_census(double coalSurplusPerc
 // Closures and methods for calculating total bytes returned to the
 // free lists in the tree.
 #ifndef PRODUCT
-template <class Chunk>
-class InitializeDictReturnedBytesClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class InitializeDictReturnedBytesClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
    public:
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList_t<Chunk_t>* fl) {
     fl->set_returned_bytes(0);
   }
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::initialize_dict_returned_bytes() {
-  InitializeDictReturnedBytesClosure<Chunk> idrb;
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::initialize_dict_returned_bytes() {
+  InitializeDictReturnedBytesClosure<Chunk_t, FreeList_t> idrb;
   idrb.do_tree(root());
 }
 
-template <class Chunk>
-class ReturnedBytesClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class ReturnedBytesClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
   size_t _dict_returned_bytes;
  public:
   ReturnedBytesClosure() { _dict_returned_bytes = 0; }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList_t<Chunk_t>* fl) {
     _dict_returned_bytes += fl->returned_bytes();
   }
   size_t dict_returned_bytes() { return _dict_returned_bytes; }
 };
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::sum_dict_returned_bytes() {
-  ReturnedBytesClosure<Chunk> rbc;
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::sum_dict_returned_bytes() {
+  ReturnedBytesClosure<Chunk_t, FreeList_t> rbc;
   rbc.do_tree(root());
 
   return rbc.dict_returned_bytes();
 }
 
 // Count the number of entries in the tree.
-template <class Chunk>
-class treeCountClosure : public DescendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class treeCountClosure : public DescendTreeCensusClosure<Chunk_t, FreeList_t> {
  public:
   uint count;
   treeCountClosure(uint c) { count = c; }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList_t<Chunk_t>* fl) {
     count++;
   }
 };
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::total_count() {
-  treeCountClosure<Chunk> ctc(0);
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::total_count() {
+  treeCountClosure<Chunk_t, FreeList_t> ctc(0);
   ctc.do_tree(root());
   return ctc.count;
 }
 #endif // PRODUCT
 
 // Calculate surpluses for the lists in the tree.
-template <class Chunk>
-class setTreeSurplusClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class setTreeSurplusClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
   double percentage;
  public:
   setTreeSurplusClosure(double v) { percentage = v; }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList<Chunk_t>* fl) {}
+
+#ifndef SERIALGC
+  void do_list(AdaptiveFreeList<Chunk_t>* fl) {
     double splitSurplusPercent = percentage;
     fl->set_surplus(fl->count() -
                    (ssize_t)((double)fl->desired() * splitSurplusPercent));
   }
+#endif // SERIALGC
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::set_tree_surplus(double splitSurplusPercent) {
-  setTreeSurplusClosure<Chunk> sts(splitSurplusPercent);
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::set_tree_surplus(double splitSurplusPercent) {
+  setTreeSurplusClosure<Chunk_t, FreeList_t> sts(splitSurplusPercent);
   sts.do_tree(root());
 }
 
 // Set hints for the lists in the tree.
-template <class Chunk>
-class setTreeHintsClosure : public DescendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class setTreeHintsClosure : public DescendTreeCensusClosure<Chunk_t, FreeList_t> {
   size_t hint;
  public:
   setTreeHintsClosure(size_t v) { hint = v; }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList<Chunk_t>* fl) {}
+
+#ifndef SERIALGC
+  void do_list(AdaptiveFreeList<Chunk_t>* fl) {
     fl->set_hint(hint);
     assert(fl->hint() == 0 || fl->hint() > fl->size(),
       "Current hint is inconsistent");
@@ -1130,35 +1165,40 @@ class setTreeHintsClosure : public DescendTreeCensusClosure<Chunk> {
       hint = fl->size();
     }
   }
+#endif // SERIALGC
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::set_tree_hints(void) {
-  setTreeHintsClosure<Chunk> sth(0);
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::set_tree_hints(void) {
+  setTreeHintsClosure<Chunk_t, FreeList_t> sth(0);
   sth.do_tree(root());
 }
 
 // Save count before previous sweep and splits and coalesces.
-template <class Chunk>
-class clearTreeCensusClosure : public AscendTreeCensusClosure<Chunk> {
-  void do_list(FreeList<Chunk>* fl) {
+template <class Chunk_t, template <class> class FreeList_t>
+class clearTreeCensusClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
+  void do_list(FreeList<Chunk_t>* fl) {}
+
+#ifndef SERIALGC
+  void do_list(AdaptiveFreeList<Chunk_t>* fl) {
     fl->set_prev_sweep(fl->count());
     fl->set_coal_births(0);
     fl->set_coal_deaths(0);
     fl->set_split_births(0);
     fl->set_split_deaths(0);
   }
+#endif  // SERIALGC
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::clear_tree_census(void) {
-  clearTreeCensusClosure<Chunk> ctc;
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::clear_tree_census(void) {
+  clearTreeCensusClosure<Chunk_t, FreeList_t> ctc;
   ctc.do_tree(root());
 }
 
 // Do reporting and post sweep clean up.
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::end_sweep_dict_census(double splitSurplusPercent) {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::end_sweep_dict_census(double splitSurplusPercent) {
   // Does walking the tree 3 times hurt?
   set_tree_surplus(splitSurplusPercent);
   set_tree_hints();
@@ -1169,9 +1209,9 @@ void BinaryTreeDictionary<Chunk>::end_sweep_dict_census(double splitSurplusPerce
 }
 
 // Print summary statistics
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::report_statistics() const {
-  FreeBlockDictionary<Chunk>::verify_par_locked();
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::report_statistics() const {
+  FreeBlockDictionary<Chunk_t>::verify_par_locked();
   gclog_or_tty->print("Statistics for BinaryTreeDictionary:\n"
          "------------------------------------\n");
   size_t total_size = total_chunk_size(debug_only(NULL));
@@ -1182,36 +1222,47 @@ void BinaryTreeDictionary<Chunk>::report_statistics() const {
   if (free_blocks > 0) {
     gclog_or_tty->print("Av.  Block  Size: %d\n", total_size/free_blocks);
   }
-  gclog_or_tty->print("Tree      Height: %d\n", treeHeight());
+  gclog_or_tty->print("Tree      Height: %d\n", tree_height());
 }
 
 // Print census information - counts, births, deaths, etc.
 // for each list in the tree.  Also print some summary
 // information.
-template <class Chunk>
-class PrintTreeCensusClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class PrintTreeCensusClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
   int _print_line;
   size_t _total_free;
-  FreeList<Chunk> _total;
+  FreeList_t<Chunk_t> _total;
 
  public:
   PrintTreeCensusClosure() {
     _print_line = 0;
     _total_free = 0;
   }
-  FreeList<Chunk>* total() { return &_total; }
+  FreeList_t<Chunk_t>* total() { return &_total; }
   size_t total_free() { return _total_free; }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList<Chunk_t>* fl) {
     if (++_print_line >= 40) {
-      FreeList<Chunk>::print_labels_on(gclog_or_tty, "size");
+      FreeList_t<Chunk_t>::print_labels_on(gclog_or_tty, "size");
       _print_line = 0;
     }
     fl->print_on(gclog_or_tty);
     _total_free +=            fl->count()            * fl->size()        ;
     total()->set_count(      total()->count()       + fl->count()      );
-    total()->set_bfr_surp(    total()->bfr_surp()     + fl->bfr_surp()    );
+  }
+
+#ifndef SERIALGC
+  void do_list(AdaptiveFreeList<Chunk_t>* fl) {
+    if (++_print_line >= 40) {
+      FreeList_t<Chunk_t>::print_labels_on(gclog_or_tty, "size");
+      _print_line = 0;
+    }
+    fl->print_on(gclog_or_tty);
+    _total_free +=           fl->count()             * fl->size()        ;
+    total()->set_count(      total()->count()        + fl->count()      );
+    total()->set_bfr_surp(   total()->bfr_surp()     + fl->bfr_surp()    );
     total()->set_surplus(    total()->split_deaths() + fl->surplus()    );
-    total()->set_desired(    total()->desired()     + fl->desired()    );
+    total()->set_desired(    total()->desired()      + fl->desired()    );
     total()->set_prev_sweep(  total()->prev_sweep()   + fl->prev_sweep()  );
     total()->set_before_sweep(total()->before_sweep() + fl->before_sweep());
     total()->set_coal_births( total()->coal_births()  + fl->coal_births() );
@@ -1219,18 +1270,32 @@ class PrintTreeCensusClosure : public AscendTreeCensusClosure<Chunk> {
     total()->set_split_births(total()->split_births() + fl->split_births());
     total()->set_split_deaths(total()->split_deaths() + fl->split_deaths());
   }
+#endif  // SERIALGC
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::print_dict_census(void) const {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::print_dict_census(void) const {
 
   gclog_or_tty->print("\nBinaryTree\n");
-  FreeList<Chunk>::print_labels_on(gclog_or_tty, "size");
-  PrintTreeCensusClosure<Chunk> ptc;
+  FreeList_t<Chunk_t>::print_labels_on(gclog_or_tty, "size");
+  PrintTreeCensusClosure<Chunk_t, FreeList_t> ptc;
   ptc.do_tree(root());
 
-  FreeList<Chunk>* total = ptc.total();
-  FreeList<Chunk>::print_labels_on(gclog_or_tty, " ");
+  FreeList_t<Chunk_t>* total = ptc.total();
+  FreeList_t<Chunk_t>::print_labels_on(gclog_or_tty, " ");
+}
+
+#ifndef SERIALGC
+template <>
+void BinaryTreeDictionary<FreeChunk, AdaptiveFreeList>::print_dict_census(void) const {
+
+  gclog_or_tty->print("\nBinaryTree\n");
+  AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
+  PrintTreeCensusClosure<FreeChunk, AdaptiveFreeList> ptc;
+  ptc.do_tree(root());
+
+  AdaptiveFreeList<FreeChunk>* total = ptc.total();
+  AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, " ");
   total->print_on(gclog_or_tty, "TOTAL\t");
   gclog_or_tty->print(
               "total_free(words): " SIZE_FORMAT_W(16)
@@ -1242,9 +1307,10 @@ void BinaryTreeDictionary<Chunk>::print_dict_census(void) const {
              (double)(total->desired() - total->count())
              /(total->desired() != 0 ? (double)total->desired() : 1.0));
 }
+#endif  // SERIALGC
 
-template <class Chunk>
-class PrintFreeListsClosure : public AscendTreeCensusClosure<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class PrintFreeListsClosure : public AscendTreeCensusClosure<Chunk_t, FreeList_t> {
   outputStream* _st;
   int _print_line;
 
@@ -1253,14 +1319,14 @@ class PrintFreeListsClosure : public AscendTreeCensusClosure<Chunk> {
     _st = st;
     _print_line = 0;
   }
-  void do_list(FreeList<Chunk>* fl) {
+  void do_list(FreeList_t<Chunk_t>* fl) {
     if (++_print_line >= 40) {
-      FreeList<Chunk>::print_labels_on(_st, "size");
+      FreeList_t<Chunk_t>::print_labels_on(_st, "size");
       _print_line = 0;
     }
     fl->print_on(gclog_or_tty);
     size_t sz = fl->size();
-    for (Chunk* fc = fl->head(); fc != NULL;
+    for (Chunk_t* fc = fl->head(); fc != NULL;
          fc = fc->next()) {
       _st->print_cr("\t[" PTR_FORMAT "," PTR_FORMAT ")  %s",
                     fc, (HeapWord*)fc + sz,
@@ -1269,11 +1335,11 @@ class PrintFreeListsClosure : public AscendTreeCensusClosure<Chunk> {
   }
 };
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::print_free_lists(outputStream* st) const {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::print_free_lists(outputStream* st) const {
 
-  FreeList<Chunk>::print_labels_on(st, "size");
-  PrintFreeListsClosure<Chunk> pflc(st);
+  FreeList_t<Chunk_t>::print_labels_on(st, "size");
+  PrintFreeListsClosure<Chunk_t, FreeList_t> pflc(st);
   pflc.do_tree(root());
 }
 
@@ -1281,18 +1347,18 @@ void BinaryTreeDictionary<Chunk>::print_free_lists(outputStream* st) const {
 // . _root has no parent
 // . parent and child point to each other
 // . each node's key correctly related to that of its child(ren)
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::verify_tree() const {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::verify_tree() const {
   guarantee(root() == NULL || total_free_blocks() == 0 ||
     total_size() != 0, "_total_size should't be 0?");
   guarantee(root() == NULL || root()->parent() == NULL, "_root shouldn't have parent");
   verify_tree_helper(root());
 }
 
-template <class Chunk>
-size_t BinaryTreeDictionary<Chunk>::verify_prev_free_ptrs(TreeList<Chunk>* tl) {
+template <class Chunk_t, template <class> class FreeList_t>
+size_t BinaryTreeDictionary<Chunk_t, FreeList_t>::verify_prev_free_ptrs(TreeList<Chunk_t, FreeList_t>* tl) {
   size_t ct = 0;
-  for (Chunk* curFC = tl->head(); curFC != NULL; curFC = curFC->next()) {
+  for (Chunk_t* curFC = tl->head(); curFC != NULL; curFC = curFC->next()) {
     ct++;
     assert(curFC->prev() == NULL || curFC->prev()->is_free(),
       "Chunk should be free");
@@ -1303,8 +1369,8 @@ size_t BinaryTreeDictionary<Chunk>::verify_prev_free_ptrs(TreeList<Chunk>* tl) {
 // Note: this helper is recursive rather than iterative, so use with
 // caution on very deep trees; and watch out for stack overflow errors;
 // In general, to be used only for debugging.
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::verify_tree_helper(TreeList<Chunk>* tl) const {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::verify_tree_helper(TreeList<Chunk_t, FreeList_t>* tl) const {
   if (tl == NULL)
     return;
   guarantee(tl->size() != 0, "A list must has a size");
@@ -1332,15 +1398,25 @@ void BinaryTreeDictionary<Chunk>::verify_tree_helper(TreeList<Chunk>* tl) const
   verify_tree_helper(tl->right());
 }
 
-template <class Chunk>
-void BinaryTreeDictionary<Chunk>::verify() const {
+template <class Chunk_t, template <class> class FreeList_t>
+void BinaryTreeDictionary<Chunk_t, FreeList_t>::verify() const {
   verify_tree();
   guarantee(total_size() == total_size_in_tree(root()), "Total Size inconsistency");
 }
 
+template class TreeList<Metablock, FreeList>;
+template class BinaryTreeDictionary<Metablock, FreeList>;
+template class TreeChunk<Metablock, FreeList>;
+
+template class TreeList<Metachunk, FreeList>;
+template class BinaryTreeDictionary<Metachunk, FreeList>;
+template class TreeChunk<Metachunk, FreeList>;
+
+
 #ifndef SERIALGC
 // Explicitly instantiate these types for FreeChunk.
-template class BinaryTreeDictionary<FreeChunk>;
-template class TreeChunk<FreeChunk>;
-template class TreeList<FreeChunk>;
+template class TreeList<FreeChunk, AdaptiveFreeList>;
+template class BinaryTreeDictionary<FreeChunk, AdaptiveFreeList>;
+template class TreeChunk<FreeChunk, AdaptiveFreeList>;
+
 #endif // SERIALGC
diff --git a/src/share/vm/memory/binaryTreeDictionary.hpp b/src/share/vm/memory/binaryTreeDictionary.hpp
index 4ab60c2dc..757eb4fda 100644
--- a/src/share/vm/memory/binaryTreeDictionary.hpp
+++ b/src/share/vm/memory/binaryTreeDictionary.hpp
@@ -37,77 +37,78 @@
 // A TreeList is a FreeList which can be used to maintain a
 // binary tree of free lists.
 
-template <class Chunk> class TreeChunk;
-template <class Chunk> class BinaryTreeDictionary;
-template <class Chunk> class AscendTreeCensusClosure;
-template <class Chunk> class DescendTreeCensusClosure;
-template <class Chunk> class DescendTreeSearchClosure;
-
-template <class Chunk>
-class TreeList: public FreeList<Chunk> {
-  friend class TreeChunk<Chunk>;
-  friend class BinaryTreeDictionary<Chunk>;
-  friend class AscendTreeCensusClosure<Chunk>;
-  friend class DescendTreeCensusClosure<Chunk>;
-  friend class DescendTreeSearchClosure<Chunk>;
-
-  TreeList<Chunk>* _parent;
-  TreeList<Chunk>* _left;
-  TreeList<Chunk>* _right;
+template <class Chunk_t, template <class> class FreeList_t> class TreeChunk;
+template <class Chunk_t, template <class> class FreeList_t> class BinaryTreeDictionary;
+template <class Chunk_t, template <class> class FreeList_t> class AscendTreeCensusClosure;
+template <class Chunk_t, template <class> class FreeList_t> class DescendTreeCensusClosure;
+template <class Chunk_t, template <class> class FreeList_t> class DescendTreeSearchClosure;
+
+template <class Chunk_t, template <class> class FreeList_t>
+class TreeList : public FreeList_t<Chunk_t> {
+  friend class TreeChunk<Chunk_t, FreeList_t>;
+  friend class BinaryTreeDictionary<Chunk_t, FreeList_t>;
+  friend class AscendTreeCensusClosure<Chunk_t, FreeList_t>;
+  friend class DescendTreeCensusClosure<Chunk_t, FreeList_t>;
+  friend class DescendTreeSearchClosure<Chunk_t, FreeList_t>;
+
+  TreeList<Chunk_t, FreeList_t>* _parent;
+  TreeList<Chunk_t, FreeList_t>* _left;
+  TreeList<Chunk_t, FreeList_t>* _right;
 
  protected:
-  TreeList<Chunk>* parent() const { return _parent; }
-  TreeList<Chunk>* left()   const { return _left;   }
-  TreeList<Chunk>* right()  const { return _right;  }
 
-  // Explicitly import these names into our namespace to fix name lookup with templates
-  using FreeList<Chunk>::head;
-  using FreeList<Chunk>::set_head;
+  TreeList<Chunk_t, FreeList_t>* parent() const { return _parent; }
+  TreeList<Chunk_t, FreeList_t>* left()   const { return _left;   }
+  TreeList<Chunk_t, FreeList_t>* right()  const { return _right;  }
 
-  using FreeList<Chunk>::tail;
-  using FreeList<Chunk>::set_tail;
-  using FreeList<Chunk>::link_tail;
+  // Wrapper on call to base class, to get the template to compile.
+  Chunk_t* head() const { return FreeList_t<Chunk_t>::head(); }
+  Chunk_t* tail() const { return FreeList_t<Chunk_t>::tail(); }
+  void set_head(Chunk_t* head) { FreeList_t<Chunk_t>::set_head(head); }
+  void set_tail(Chunk_t* tail) { FreeList_t<Chunk_t>::set_tail(tail); }
 
-  using FreeList<Chunk>::increment_count;
-  NOT_PRODUCT(using FreeList<Chunk>::increment_returned_bytes_by;)
-  using FreeList<Chunk>::verify_chunk_in_free_list;
-  using FreeList<Chunk>::size;
+  size_t size() const { return FreeList_t<Chunk_t>::size(); }
 
   // Accessors for links in tree.
 
-  void set_left(TreeList<Chunk>* tl) {
+  void set_left(TreeList<Chunk_t, FreeList_t>* tl) {
     _left   = tl;
     if (tl != NULL)
       tl->set_parent(this);
   }
-  void set_right(TreeList<Chunk>* tl) {
+  void set_right(TreeList<Chunk_t, FreeList_t>* tl) {
     _right  = tl;
     if (tl != NULL)
       tl->set_parent(this);
   }
-  void set_parent(TreeList<Chunk>* tl)  { _parent = tl;   }
+  void set_parent(TreeList<Chunk_t, FreeList_t>* tl)  { _parent = tl;   }
 
-  void clearLeft()               { _left = NULL;   }
+  void clear_left()               { _left = NULL;   }
   void clear_right()              { _right = NULL;  }
   void clear_parent()             { _parent = NULL; }
-  void initialize()              { clearLeft(); clear_right(), clear_parent(); }
+  void initialize()               { clear_left(); clear_right(), clear_parent(); FreeList_t<Chunk_t>::initialize(); }
 
   // For constructing a TreeList from a Tree chunk or
   // address and size.
-  static TreeList<Chunk>* as_TreeList(TreeChunk<Chunk>* tc);
-  static TreeList<Chunk>* as_TreeList(HeapWord* addr, size_t size);
+  TreeList();
+  static TreeList<Chunk_t, FreeList_t>*
+          as_TreeList(TreeChunk<Chunk_t, FreeList_t>* tc);
+  static TreeList<Chunk_t, FreeList_t>* as_TreeList(HeapWord* addr, size_t size);
 
   // Returns the head of the free list as a pointer to a TreeChunk.
-  TreeChunk<Chunk>* head_as_TreeChunk();
+  TreeChunk<Chunk_t, FreeList_t>* head_as_TreeChunk();
 
   // Returns the first available chunk in the free list as a pointer
   // to a TreeChunk.
-  TreeChunk<Chunk>* first_available();
+  TreeChunk<Chunk_t, FreeList_t>* first_available();
 
   // Returns the block with the largest heap address amongst
   // those in the list for this size; potentially slow and expensive,
   // use with caution!
-  TreeChunk<Chunk>* largest_address();
+  TreeChunk<Chunk_t, FreeList_t>* largest_address();
+
+  TreeList<Chunk_t, FreeList_t>* get_better_list(
+    BinaryTreeDictionary<Chunk_t, FreeList_t>* dictionary);
 
   // remove_chunk_replace_if_needed() removes the given "tc" from the TreeList.
   // If "tc" is the first chunk in the list, it is also the
@@ -115,10 +116,10 @@ class TreeList: public FreeList<Chunk> {
   // returns the possibly replaced TreeList* for the node in
   // the tree.  It also updates the parent of the original
   // node to point to the new node.
-  TreeList<Chunk>* remove_chunk_replace_if_needed(TreeChunk<Chunk>* tc);
+  TreeList<Chunk_t, FreeList_t>* remove_chunk_replace_if_needed(TreeChunk<Chunk_t, FreeList_t>* tc);
   // See FreeList.
-  void return_chunk_at_head(TreeChunk<Chunk>* tc);
-  void return_chunk_at_tail(TreeChunk<Chunk>* tc);
+  void return_chunk_at_head(TreeChunk<Chunk_t, FreeList_t>* tc);
+  void return_chunk_at_tail(TreeChunk<Chunk_t, FreeList_t>* tc);
 };
 
 // A TreeChunk is a subclass of a Chunk that additionally
@@ -134,52 +135,54 @@ class TreeList: public FreeList<Chunk> {
 // on the free list for a node in the tree and is only removed if
 // it is the last chunk on the free list.
 
-template <class Chunk>
-class TreeChunk : public Chunk {
-  friend class TreeList<Chunk>;
-  TreeList<Chunk>* _list;
-  TreeList<Chunk> _embedded_list;  // if non-null, this chunk is on _list
+template <class Chunk_t, template <class> class FreeList_t>
+class TreeChunk : public Chunk_t {
+  friend class TreeList<Chunk_t, FreeList_t>;
+  TreeList<Chunk_t, FreeList_t>* _list;
+  TreeList<Chunk_t, FreeList_t> _embedded_list;  // if non-null, this chunk is on _list
+
+  static size_t _min_tree_chunk_size;
+
  protected:
-  TreeList<Chunk>* embedded_list() const { return (TreeList<Chunk>*) &_embedded_list; }
-  void set_embedded_list(TreeList<Chunk>* v) { _embedded_list = *v; }
+  TreeList<Chunk_t, FreeList_t>* embedded_list() const { return (TreeList<Chunk_t, FreeList_t>*) &_embedded_list; }
+  void set_embedded_list(TreeList<Chunk_t, FreeList_t>* v) { _embedded_list = *v; }
  public:
-  TreeList<Chunk>* list() { return _list; }
-  void set_list(TreeList<Chunk>* v) { _list = v; }
-  static TreeChunk<Chunk>* as_TreeChunk(Chunk* fc);
+  TreeList<Chunk_t, FreeList_t>* list() { return _list; }
+  void set_list(TreeList<Chunk_t, FreeList_t>* v) { _list = v; }
+  static TreeChunk<Chunk_t, FreeList_t>* as_TreeChunk(Chunk_t* fc);
   // Initialize fields in a TreeChunk that should be
   // initialized when the TreeChunk is being added to
   // a free list in the tree.
   void initialize() { embedded_list()->initialize(); }
 
-  Chunk* next() const { return Chunk::next(); }
-  Chunk* prev() const { return Chunk::prev(); }
-  size_t size() const volatile { return Chunk::size(); }
+  Chunk_t* next() const { return Chunk_t::next(); }
+  Chunk_t* prev() const { return Chunk_t::prev(); }
+  size_t size() const volatile { return Chunk_t::size(); }
+
+  static size_t min_size() {
+    return _min_tree_chunk_size;
+  }
 
   // debugging
   void verify_tree_chunk_list() const;
+  void assert_is_mangled() const;
 };
 
 
-template <class Chunk>
-class BinaryTreeDictionary: public FreeBlockDictionary<Chunk> {
+template <class Chunk_t, template <class> class FreeList_t>
+class BinaryTreeDictionary: public FreeBlockDictionary<Chunk_t> {
   friend class VMStructs;
-  bool       _splay;
-  bool       _adaptive_freelists;
   size_t     _total_size;
   size_t     _total_free_blocks;
-  TreeList<Chunk>* _root;
+  TreeList<Chunk_t, FreeList_t>* _root;
 
   // private accessors
-  bool splay() const { return _splay; }
-  void set_splay(bool v) { _splay = v; }
   void set_total_size(size_t v) { _total_size = v; }
   virtual void inc_total_size(size_t v);
   virtual void dec_total_size(size_t v);
-  size_t total_free_blocks() const { return _total_free_blocks; }
   void set_total_free_blocks(size_t v) { _total_free_blocks = v; }
-  TreeList<Chunk>* root() const { return _root; }
-  void set_root(TreeList<Chunk>* v) { _root = v; }
-  bool adaptive_freelists() { return _adaptive_freelists; }
+  TreeList<Chunk_t, FreeList_t>* root() const { return _root; }
+  void set_root(TreeList<Chunk_t, FreeList_t>* v) { _root = v; }
 
   // This field is added and can be set to point to the
   // the Mutex used to synchronize access to the
@@ -191,54 +194,55 @@ class BinaryTreeDictionary: public FreeBlockDictionary<Chunk> {
   // return it.  If the chunk
   // is the last chunk of that size, remove the node for that size
   // from the tree.
-  TreeChunk<Chunk>* get_chunk_from_tree(size_t size, enum FreeBlockDictionary<Chunk>::Dither dither, bool splay);
-  // Return a list of the specified size or NULL from the tree.
-  // The list is not removed from the tree.
-  TreeList<Chunk>* find_list (size_t size) const;
+  TreeChunk<Chunk_t, FreeList_t>* get_chunk_from_tree(size_t size, enum FreeBlockDictionary<Chunk_t>::Dither dither);
   // Remove this chunk from the tree.  If the removal results
   // in an empty list in the tree, remove the empty list.
-  TreeChunk<Chunk>* remove_chunk_from_tree(TreeChunk<Chunk>* tc);
+  TreeChunk<Chunk_t, FreeList_t>* remove_chunk_from_tree(TreeChunk<Chunk_t, FreeList_t>* tc);
   // Remove the node in the trees starting at tl that has the
   // minimum value and return it.  Repair the tree as needed.
-  TreeList<Chunk>* remove_tree_minimum(TreeList<Chunk>* tl);
-  void       semi_splay_step(TreeList<Chunk>* tl);
+  TreeList<Chunk_t, FreeList_t>* remove_tree_minimum(TreeList<Chunk_t, FreeList_t>* tl);
   // Add this free chunk to the tree.
-  void       insert_chunk_in_tree(Chunk* freeChunk);
+  void       insert_chunk_in_tree(Chunk_t* freeChunk);
  public:
 
-  static const size_t min_tree_chunk_size  = sizeof(TreeChunk<Chunk>)/HeapWordSize;
+  // Return a list of the specified size or NULL from the tree.
+  // The list is not removed from the tree.
+  TreeList<Chunk_t, FreeList_t>* find_list (size_t size) const;
 
   void       verify_tree() const;
   // verify that the given chunk is in the tree.
-  bool       verify_chunk_in_free_list(Chunk* tc) const;
+  bool       verify_chunk_in_free_list(Chunk_t* tc) const;
  private:
-  void          verify_tree_helper(TreeList<Chunk>* tl) const;
-  static size_t verify_prev_free_ptrs(TreeList<Chunk>* tl);
+  void          verify_tree_helper(TreeList<Chunk_t, FreeList_t>* tl) const;
+  static size_t verify_prev_free_ptrs(TreeList<Chunk_t, FreeList_t>* tl);
 
   // Returns the total number of chunks in the list.
-  size_t     total_list_length(TreeList<Chunk>* tl) const;
+  size_t     total_list_length(TreeList<Chunk_t, FreeList_t>* tl) const;
   // Returns the total number of words in the chunks in the tree
   // starting at "tl".
-  size_t     total_size_in_tree(TreeList<Chunk>* tl) const;
+  size_t     total_size_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const;
   // Returns the sum of the square of the size of each block
   // in the tree starting at "tl".
-  double     sum_of_squared_block_sizes(TreeList<Chunk>* const tl) const;
+  double     sum_of_squared_block_sizes(TreeList<Chunk_t, FreeList_t>* const tl) const;
   // Returns the total number of free blocks in the tree starting
   // at "tl".
-  size_t     total_free_blocks_in_tree(TreeList<Chunk>* tl) const;
-  size_t     num_free_blocks() const;
-  size_t     treeHeight() const;
-  size_t     tree_height_helper(TreeList<Chunk>* tl) const;
-  size_t     total_nodes_in_tree(TreeList<Chunk>* tl) const;
-  size_t     total_nodes_helper(TreeList<Chunk>* tl) const;
+  size_t     total_free_blocks_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const;
+  size_t     num_free_blocks()  const;
+  size_t     tree_height() const;
+  size_t     tree_height_helper(TreeList<Chunk_t, FreeList_t>* tl) const;
+  size_t     total_nodes_in_tree(TreeList<Chunk_t, FreeList_t>* tl) const;
+  size_t     total_nodes_helper(TreeList<Chunk_t, FreeList_t>* tl) const;
 
  public:
   // Constructor
-  BinaryTreeDictionary(bool adaptive_freelists, bool splay = false);
-  BinaryTreeDictionary(MemRegion mr, bool adaptive_freelists, bool splay = false);
+  BinaryTreeDictionary() :
+    _total_size(0), _total_free_blocks(0), _root(0) {}
+
+  BinaryTreeDictionary(MemRegion mr);
 
   // Public accessors
   size_t total_size() const { return _total_size; }
+  size_t total_free_blocks() const { return _total_free_blocks; }
 
   // Reset the dictionary to the initial conditions with
   // a single free chunk.
@@ -249,23 +253,24 @@ class BinaryTreeDictionary: public FreeBlockDictionary<Chunk> {
 
   // Return a chunk of size "size" or greater from
   // the tree.
-  // want a better dynamic splay strategy for the future.
-  Chunk* get_chunk(size_t size, enum FreeBlockDictionary<Chunk>::Dither dither) {
-    FreeBlockDictionary<Chunk>::verify_par_locked();
-    Chunk* res = get_chunk_from_tree(size, dither, splay());
+  Chunk_t* get_chunk(size_t size, enum FreeBlockDictionary<Chunk_t>::Dither dither) {
+    FreeBlockDictionary<Chunk_t>::verify_par_locked();
+    Chunk_t* res = get_chunk_from_tree(size, dither);
     assert(res == NULL || res->is_free(),
            "Should be returning a free chunk");
+    assert(dither != FreeBlockDictionary<Chunk_t>::exactly ||
+           res == NULL || res->size() == size, "Not correct size");
     return res;
   }
 
-  void return_chunk(Chunk* chunk) {
-    FreeBlockDictionary<Chunk>::verify_par_locked();
+  void return_chunk(Chunk_t* chunk) {
+    FreeBlockDictionary<Chunk_t>::verify_par_locked();
     insert_chunk_in_tree(chunk);
   }
 
-  void remove_chunk(Chunk* chunk) {
-    FreeBlockDictionary<Chunk>::verify_par_locked();
-    remove_chunk_from_tree((TreeChunk<Chunk>*)chunk);
+  void remove_chunk(Chunk_t* chunk) {
+    FreeBlockDictionary<Chunk_t>::verify_par_locked();
+    remove_chunk_from_tree((TreeChunk<Chunk_t, FreeList_t>*)chunk);
     assert(chunk->is_free(), "Should still be a free chunk");
   }
 
@@ -281,19 +286,19 @@ class BinaryTreeDictionary: public FreeBlockDictionary<Chunk> {
   }
 
   size_t     min_size() const {
-    return min_tree_chunk_size;
+    return TreeChunk<Chunk_t, FreeList_t>::min_size();
   }
 
   double     sum_of_squared_block_sizes() const {
     return sum_of_squared_block_sizes(root());
   }
 
-  Chunk* find_chunk_ends_at(HeapWord* target) const;
+  Chunk_t* find_chunk_ends_at(HeapWord* target) const;
 
   // Find the list with size "size" in the binary tree and update
   // the statistics in the list according to "split" (chunk was
   // split or coalesce) and "birth" (chunk was added or removed).
-  void       dict_census_udpate(size_t size, bool split, bool birth);
+  void       dict_census_update(size_t size, bool split, bool birth);
   // Return true if the dictionary is overpopulated (more chunks of
   // this size than desired) for size "size".
   bool       coal_dict_over_populated(size_t size);
@@ -307,7 +312,7 @@ class BinaryTreeDictionary: public FreeBlockDictionary<Chunk> {
   // statistics for the sweep.
   void       end_sweep_dict_census(double splitSurplusPercent);
   // Return the largest free chunk in the tree.
-  Chunk* find_largest_dict() const;
+  Chunk_t* find_largest_dict() const;
   // Accessors for statistics
   void       set_tree_surplus(double splitSurplusPercent);
   void       set_tree_hints(void);
diff --git a/src/share/vm/memory/filemap.cpp b/src/share/vm/memory/filemap.cpp
index d4a11f408..c3d1bf5f9 100644
--- a/src/share/vm/memory/filemap.cpp
+++ b/src/share/vm/memory/filemap.cpp
@@ -29,6 +29,7 @@
 #include "runtime/arguments.hpp"
 #include "runtime/java.hpp"
 #include "runtime/os.hpp"
+#include "services/memTracker.hpp"
 #include "utilities/defaultStream.hpp"
 
 # include <sys/stat.h>
@@ -344,25 +345,14 @@ ReservedSpace FileMapInfo::reserve_shared_memory() {
     fail_continue(err_msg("Unable to reserved shared space at required address " INTPTR_FORMAT, requested_addr));
     return rs;
   }
+  // the reserved virtual memory is for mapping class data sharing archive
+  if (MemTracker::is_on()) {
+    MemTracker::record_virtual_memory_type((address)rs.base(), mtClassShared);
+  }
   return rs;
 }
 
 // Memory map a region in the address space.
-
-char* FileMapInfo::map_region(int i, ReservedSpace rs) {
-  struct FileMapInfo::FileMapHeader::space_info* si = &_header._space[i];
-  size_t used = si->_used;
-  size_t size = align_size_up(used, os::vm_allocation_granularity());
-
-  ReservedSpace mapped_rs = rs.first_part(size, true, true);
-  ReservedSpace unmapped_rs = rs.last_part(size);
-  mapped_rs.release();
-
-  return map_region(i);
-}
-
-
-// Memory map a region in the address space.
 static const char* shared_region_name[] = { "ReadOnly", "ReadWrite", "MiscData", "MiscCode"};
 
 char* FileMapInfo::map_region(int i) {
diff --git a/src/share/vm/memory/filemap.hpp b/src/share/vm/memory/filemap.hpp
index 760ddfe17..7cdd9616c 100644
--- a/src/share/vm/memory/filemap.hpp
+++ b/src/share/vm/memory/filemap.hpp
@@ -125,7 +125,6 @@ public:
                      size_t capacity, bool read_only, bool allow_exec);
   void  write_bytes(const void* buffer, int count);
   void  write_bytes_aligned(const void* buffer, int count);
-  char* map_region(int i, ReservedSpace rs);
   char* map_region(int i);
   void  unmap_region(int i);
   void  close();
diff --git a/src/share/vm/memory/freeBlockDictionary.cpp b/src/share/vm/memory/freeBlockDictionary.cpp
index 13b4daa89..9b2221281 100644
--- a/src/share/vm/memory/freeBlockDictionary.cpp
+++ b/src/share/vm/memory/freeBlockDictionary.cpp
@@ -27,6 +27,8 @@
 #include "gc_implementation/concurrentMarkSweep/freeChunk.hpp"
 #endif // SERIALGC
 #include "memory/freeBlockDictionary.hpp"
+#include "memory/metablock.hpp"
+#include "memory/metachunk.hpp"
 #ifdef TARGET_OS_FAMILY_linux
 # include "thread_linux.inline.hpp"
 #endif
@@ -62,6 +64,9 @@ template <class Chunk> void FreeBlockDictionary<Chunk>::verify_par_locked() cons
 }
 #endif
 
+template class FreeBlockDictionary<Metablock>;
+template class FreeBlockDictionary<Metachunk>;
+
 #ifndef SERIALGC
 // Explicitly instantiate for FreeChunk
 template class FreeBlockDictionary<FreeChunk>;
diff --git a/src/share/vm/memory/freeBlockDictionary.hpp b/src/share/vm/memory/freeBlockDictionary.hpp
index 573cfed3a..2502e362d 100644
--- a/src/share/vm/memory/freeBlockDictionary.hpp
+++ b/src/share/vm/memory/freeBlockDictionary.hpp
@@ -66,7 +66,7 @@ class FreeBlockDictionary: public CHeapObj<mtGC> {
   virtual void       reset(HeapWord* addr, size_t size) = 0;
   virtual void       reset() = 0;
 
-  virtual void       dict_census_udpate(size_t size, bool split, bool birth) = 0;
+  virtual void       dict_census_update(size_t size, bool split, bool birth) = 0;
   virtual bool       coal_dict_over_populated(size_t size) = 0;
   virtual void       begin_sweep_dict_census(double coalSurplusPercent,
                        float inter_sweep_current, float inter_sweep_estimate,
diff --git a/src/share/vm/memory/freeList.cpp b/src/share/vm/memory/freeList.cpp
index a5fbc06ee..f5cd80545 100644
--- a/src/share/vm/memory/freeList.cpp
+++ b/src/share/vm/memory/freeList.cpp
@@ -25,6 +25,8 @@
 #include "precompiled.hpp"
 #include "memory/freeBlockDictionary.hpp"
 #include "memory/freeList.hpp"
+#include "memory/metablock.hpp"
+#include "memory/metachunk.hpp"
 #include "memory/sharedHeap.hpp"
 #include "runtime/globals.hpp"
 #include "runtime/mutex.hpp"
@@ -49,8 +51,6 @@ FreeList<Chunk>::FreeList() :
 {
   _size         = 0;
   _count        = 0;
-  _hint         = 0;
-  init_statistics();
 }
 
 template <class Chunk>
@@ -62,34 +62,50 @@ FreeList<Chunk>::FreeList(Chunk* fc) :
 {
   _size         = fc->size();
   _count        = 1;
-  _hint         = 0;
-  init_statistics();
-#ifndef PRODUCT
-  _allocation_stats.set_returned_bytes(size() * HeapWordSize);
-#endif
 }
 
 template <class Chunk>
-void FreeList<Chunk>::reset(size_t hint) {
+void FreeList<Chunk>::link_head(Chunk* v) {
+  assert_proper_lock_protection();
+  set_head(v);
+  // If this method is not used (just set the head instead),
+  // this check can be avoided.
+  if (v != NULL) {
+    v->link_prev(NULL);
+  }
+}
+
+
+
+template <class Chunk>
+void FreeList<Chunk>::reset() {
+  // Don't set the _size to 0 because this method is
+  // used with a existing list that has a size but which has
+  // been emptied.
+  // Don't clear the _protecting_lock of an existing list.
   set_count(0);
   set_head(NULL);
   set_tail(NULL);
-  set_hint(hint);
 }
 
 template <class Chunk>
-void FreeList<Chunk>::init_statistics(bool split_birth) {
-  _allocation_stats.initialize(split_birth);
+void FreeList<Chunk>::initialize() {
+#ifdef ASSERT
+  // Needed early because it might be checked in other initializing code.
+  set_protecting_lock(NULL);
+#endif
+  reset();
+  set_size(0);
 }
 
-template <class Chunk>
-Chunk* FreeList<Chunk>::get_chunk_at_head() {
+template <class Chunk_t>
+Chunk_t* FreeList<Chunk_t>::get_chunk_at_head() {
   assert_proper_lock_protection();
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
-  Chunk* fc = head();
+  Chunk_t* fc = head();
   if (fc != NULL) {
-    Chunk* nextFC = fc->next();
+    Chunk_t* nextFC = fc->next();
     if (nextFC != NULL) {
       // The chunk fc being removed has a "next".  Set the "next" to the
       // "prev" of fc.
@@ -197,11 +213,6 @@ void FreeList<Chunk>::return_chunk_at_head(Chunk* chunk, bool record_return) {
     link_tail(chunk);
   }
   increment_count(); // of # of chunks in list
-  DEBUG_ONLY(
-    if (record_return) {
-      increment_returned_bytes_by(size()*HeapWordSize);
-    }
-  )
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
   assert(head() == NULL || head()->size() == size(), "wrong item on list");
@@ -233,11 +244,6 @@ void FreeList<Chunk>::return_chunk_at_tail(Chunk* chunk, bool record_return) {
   }
   link_tail(chunk);
   increment_count();  // of # of chunks in list
-  DEBUG_ONLY(
-    if (record_return) {
-      increment_returned_bytes_by(size()*HeapWordSize);
-    }
-  )
   assert(head() == NULL || head()->prev() == NULL, "list invariant");
   assert(tail() == NULL || tail()->next() == NULL, "list invariant");
   assert(head() == NULL || head()->size() == size(), "wrong item on list");
@@ -273,7 +279,7 @@ void FreeList<Chunk>::prepend(FreeList<Chunk>* fl) {
   }
 }
 
-// verify_chunk_in_free_list() is used to verify that an item is in this free list.
+// verify_chunk_in_free_lists() is used to verify that an item is in this free list.
 // It is used as a debugging aid.
 template <class Chunk>
 bool FreeList<Chunk>::verify_chunk_in_free_list(Chunk* fc) const {
@@ -294,40 +300,14 @@ bool FreeList<Chunk>::verify_chunk_in_free_list(Chunk* fc) const {
 
 #ifndef PRODUCT
 template <class Chunk>
-void FreeList<Chunk>::verify_stats() const {
-  // The +1 of the LH comparand is to allow some "looseness" in
-  // checking: we usually call this interface when adding a block
-  // and we'll subsequently update the stats; we cannot update the
-  // stats beforehand because in the case of the large-block BT
-  // dictionary for example, this might be the first block and
-  // in that case there would be no place that we could record
-  // the stats (which are kept in the block itself).
-  assert((_allocation_stats.prev_sweep() + _allocation_stats.split_births()
-          + _allocation_stats.coal_births() + 1)   // Total Production Stock + 1
-         >= (_allocation_stats.split_deaths() + _allocation_stats.coal_deaths()
-             + (ssize_t)count()),                // Total Current Stock + depletion
-         err_msg("FreeList " PTR_FORMAT " of size " SIZE_FORMAT
-                 " violates Conservation Principle: "
-                 "prev_sweep(" SIZE_FORMAT ")"
-                 " + split_births(" SIZE_FORMAT ")"
-                 " + coal_births(" SIZE_FORMAT ") + 1 >= "
-                 " split_deaths(" SIZE_FORMAT ")"
-                 " coal_deaths(" SIZE_FORMAT ")"
-                 " + count(" SSIZE_FORMAT ")",
-                 this, _size, _allocation_stats.prev_sweep(), _allocation_stats.split_births(),
-                 _allocation_stats.split_births(), _allocation_stats.split_deaths(),
-                 _allocation_stats.coal_deaths(), count()));
-}
-
-template <class Chunk>
 void FreeList<Chunk>::assert_proper_lock_protection_work() const {
-  assert(_protecting_lock != NULL, "Don't call this directly");
+  assert(protecting_lock() != NULL, "Don't call this directly");
   assert(ParallelGCThreads > 0, "Don't call this directly");
   Thread* thr = Thread::current();
   if (thr->is_VM_thread() || thr->is_ConcurrentGC_thread()) {
     // assert that we are holding the freelist lock
   } else if (thr->is_GC_task_thread()) {
-    assert(_protecting_lock->owned_by_self(), "FreeList RACE DETECTED");
+    assert(protecting_lock()->owned_by_self(), "FreeList RACE DETECTED");
   } else if (thr->is_Java_thread()) {
     assert(!SafepointSynchronize::is_at_safepoint(), "Should not be executing");
   } else {
@@ -350,21 +330,17 @@ void FreeList<Chunk>::print_labels_on(outputStream* st, const char* c) {
 // to the call is a non-null string, it is printed in the first column;
 // otherwise, if the argument is null (the default), then the size of the
 // (free list) block is printed in the first column.
-template <class Chunk>
-void FreeList<Chunk>::print_on(outputStream* st, const char* c) const {
+template <class Chunk_t>
+void FreeList<Chunk_t>::print_on(outputStream* st, const char* c) const {
   if (c != NULL) {
     st->print("%16s", c);
   } else {
     st->print(SIZE_FORMAT_W(16), size());
   }
-  st->print("\t"
-           SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t"
-           SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\t" SSIZE_FORMAT_W(14) "\n",
-           bfr_surp(),             surplus(),             desired(),             prev_sweep(),           before_sweep(),
-           count(),               coal_births(),          coal_deaths(),          split_births(),         split_deaths());
 }
 
+template class FreeList<Metablock>;
+template class FreeList<Metachunk>;
 #ifndef SERIALGC
-// Needs to be after the definitions have been seen.
 template class FreeList<FreeChunk>;
 #endif // SERIALGC
diff --git a/src/share/vm/memory/freeList.hpp b/src/share/vm/memory/freeList.hpp
index a982cfbda..37438cc38 100644
--- a/src/share/vm/memory/freeList.hpp
+++ b/src/share/vm/memory/freeList.hpp
@@ -40,23 +40,19 @@ class CompactibleFreeListSpace;
 // for that implementation.
 
 class Mutex;
-template <class Chunk> class TreeList;
-template <class Chunk> class PrintTreeCensusClosure;
 
-template <class Chunk>
+template <class Chunk_t>
 class FreeList VALUE_OBJ_CLASS_SPEC {
   friend class CompactibleFreeListSpace;
   friend class VMStructs;
-  friend class PrintTreeCensusClosure<Chunk>;
 
  private:
-  Chunk*        _head;          // Head of list of free chunks
-  Chunk*        _tail;          // Tail of list of free chunks
+  Chunk_t*      _head;          // Head of list of free chunks
+  Chunk_t*      _tail;          // Tail of list of free chunks
   size_t        _size;          // Size in Heap words of each chunk
   ssize_t       _count;         // Number of entries in list
-  size_t        _hint;          // next larger size list with a positive surplus
 
-  AllocationStats _allocation_stats; // allocation-related statistics
+ protected:
 
 #ifdef ASSERT
   Mutex*        _protecting_lock;
@@ -71,10 +67,6 @@ class FreeList VALUE_OBJ_CLASS_SPEC {
 #endif
   }
 
-  // Initialize the allocation statistics.
- protected:
-  void init_statistics(bool split_birth = false);
-  void set_count(ssize_t v) { _count = v;}
   void increment_count()    {
     _count++;
   }
@@ -89,52 +81,48 @@ class FreeList VALUE_OBJ_CLASS_SPEC {
   // Construct a list without any entries.
   FreeList();
   // Construct a list with "fc" as the first (and lone) entry in the list.
-  FreeList(Chunk* fc);
+  FreeList(Chunk_t* fc);
 
-  // Reset the head, tail, hint, and count of a free list.
-  void reset(size_t hint);
+  // Do initialization
+  void initialize();
+
+  // Reset the head, tail, and count of a free list.
+  void reset();
 
   // Declare the current free list to be protected by the given lock.
 #ifdef ASSERT
-  void set_protecting_lock(Mutex* protecting_lock) {
-    _protecting_lock = protecting_lock;
+  Mutex* protecting_lock() const { return _protecting_lock; }
+  void set_protecting_lock(Mutex* v) {
+    _protecting_lock = v;
   }
 #endif
 
   // Accessors.
-  Chunk* head() const {
+  Chunk_t* head() const {
     assert_proper_lock_protection();
     return _head;
   }
-  void set_head(Chunk* v) {
+  void set_head(Chunk_t* v) {
     assert_proper_lock_protection();
     _head = v;
     assert(!_head || _head->size() == _size, "bad chunk size");
   }
   // Set the head of the list and set the prev field of non-null
   // values to NULL.
-  void link_head(Chunk* v) {
-    assert_proper_lock_protection();
-    set_head(v);
-    // If this method is not used (just set the head instead),
-    // this check can be avoided.
-    if (v != NULL) {
-      v->link_prev(NULL);
-    }
-  }
+  void link_head(Chunk_t* v);
 
-  Chunk* tail() const {
+  Chunk_t* tail() const {
     assert_proper_lock_protection();
     return _tail;
   }
-  void set_tail(Chunk* v) {
+  void set_tail(Chunk_t* v) {
     assert_proper_lock_protection();
     _tail = v;
     assert(!_tail || _tail->size() == _size, "bad chunk size");
   }
   // Set the tail of the list and set the next field of non-null
   // values to NULL.
-  void link_tail(Chunk* v) {
+  void link_tail(Chunk_t* v) {
     assert_proper_lock_protection();
     set_tail(v);
     if (v != NULL) {
@@ -152,174 +140,45 @@ class FreeList VALUE_OBJ_CLASS_SPEC {
     assert_proper_lock_protection();
     _size = v;
   }
-  ssize_t count() const {
-    return _count;
-  }
-  size_t hint() const {
-    return _hint;
-  }
-  void set_hint(size_t v) {
-    assert_proper_lock_protection();
-    assert(v == 0 || _size < v, "Bad hint"); _hint = v;
-  }
-
-  // Accessors for statistics
-  AllocationStats* allocation_stats() {
-    assert_proper_lock_protection();
-    return &_allocation_stats;
-  }
-
-  ssize_t desired() const {
-    return _allocation_stats.desired();
-  }
-  void set_desired(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_desired(v);
-  }
-  void compute_desired(float inter_sweep_current,
-                       float inter_sweep_estimate,
-                       float intra_sweep_estimate) {
-    assert_proper_lock_protection();
-    _allocation_stats.compute_desired(_count,
-                                      inter_sweep_current,
-                                      inter_sweep_estimate,
-                                      intra_sweep_estimate);
-  }
-  ssize_t coal_desired() const {
-    return _allocation_stats.coal_desired();
-  }
-  void set_coal_desired(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_coal_desired(v);
-  }
-
-  ssize_t surplus() const {
-    return _allocation_stats.surplus();
-  }
-  void set_surplus(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_surplus(v);
-  }
-  void increment_surplus() {
-    assert_proper_lock_protection();
-    _allocation_stats.increment_surplus();
-  }
-  void decrement_surplus() {
-    assert_proper_lock_protection();
-    _allocation_stats.decrement_surplus();
-  }
-
-  ssize_t bfr_surp() const {
-    return _allocation_stats.bfr_surp();
-  }
-  void set_bfr_surp(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_bfr_surp(v);
-  }
-  ssize_t prev_sweep() const {
-    return _allocation_stats.prev_sweep();
-  }
-  void set_prev_sweep(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_prev_sweep(v);
-  }
-  ssize_t before_sweep() const {
-    return _allocation_stats.before_sweep();
-  }
-  void set_before_sweep(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_before_sweep(v);
-  }
-
-  ssize_t coal_births() const {
-    return _allocation_stats.coal_births();
-  }
-  void set_coal_births(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_coal_births(v);
-  }
-  void increment_coal_births() {
-    assert_proper_lock_protection();
-    _allocation_stats.increment_coal_births();
-  }
-
-  ssize_t coal_deaths() const {
-    return _allocation_stats.coal_deaths();
-  }
-  void set_coal_deaths(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_coal_deaths(v);
-  }
-  void increment_coal_deaths() {
-    assert_proper_lock_protection();
-    _allocation_stats.increment_coal_deaths();
-  }
-
-  ssize_t split_births() const {
-    return _allocation_stats.split_births();
-  }
-  void set_split_births(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_split_births(v);
-  }
-  void increment_split_births() {
-    assert_proper_lock_protection();
-    _allocation_stats.increment_split_births();
-  }
+  ssize_t count() const { return _count; }
+  void set_count(ssize_t v) { _count = v;}
 
-  ssize_t split_deaths() const {
-    return _allocation_stats.split_deaths();
-  }
-  void set_split_deaths(ssize_t v) {
-    assert_proper_lock_protection();
-    _allocation_stats.set_split_deaths(v);
-  }
-  void increment_split_deaths() {
-    assert_proper_lock_protection();
-    _allocation_stats.increment_split_deaths();
-  }
+  size_t get_better_size() { return size(); }
 
-  NOT_PRODUCT(
-    // For debugging.  The "_returned_bytes" in all the lists are summed
-    // and compared with the total number of bytes swept during a
-    // collection.
-    size_t returned_bytes() const { return _allocation_stats.returned_bytes(); }
-    void set_returned_bytes(size_t v) { _allocation_stats.set_returned_bytes(v); }
-    void increment_returned_bytes_by(size_t v) {
-      _allocation_stats.set_returned_bytes(_allocation_stats.returned_bytes() + v);
-    }
-  )
+  size_t returned_bytes() const { ShouldNotReachHere(); return 0; }
+  void set_returned_bytes(size_t v) {}
+  void increment_returned_bytes_by(size_t v) {}
 
   // Unlink head of list and return it.  Returns NULL if
   // the list is empty.
-  Chunk* get_chunk_at_head();
+  Chunk_t* get_chunk_at_head();
 
   // Remove the first "n" or "count", whichever is smaller, chunks from the
   // list, setting "fl", which is required to be empty, to point to them.
-  void getFirstNChunksFromList(size_t n, FreeList<Chunk>* fl);
+  void getFirstNChunksFromList(size_t n, FreeList<Chunk_t>* fl);
 
   // Unlink this chunk from it's free list
-  void remove_chunk(Chunk* fc);
+  void remove_chunk(Chunk_t* fc);
 
   // Add this chunk to this free list.
-  void return_chunk_at_head(Chunk* fc);
-  void return_chunk_at_tail(Chunk* fc);
+  void return_chunk_at_head(Chunk_t* fc);
+  void return_chunk_at_tail(Chunk_t* fc);
 
   // Similar to returnChunk* but also records some diagnostic
   // information.
-  void return_chunk_at_head(Chunk* fc, bool record_return);
-  void return_chunk_at_tail(Chunk* fc, bool record_return);
+  void return_chunk_at_head(Chunk_t* fc, bool record_return);
+  void return_chunk_at_tail(Chunk_t* fc, bool record_return);
 
   // Prepend "fl" (whose size is required to be the same as that of "this")
   // to the front of "this" list.
-  void prepend(FreeList<Chunk>* fl);
+  void prepend(FreeList<Chunk_t>* fl);
 
   // Verify that the chunk is in the list.
   // found.  Return NULL if "fc" is not found.
-  bool verify_chunk_in_free_list(Chunk* fc) const;
+  bool verify_chunk_in_free_list(Chunk_t* fc) const;
 
   // Stats verification
-  void verify_stats() const PRODUCT_RETURN;
+//  void verify_stats() const { ShouldNotReachHere(); };
 
   // Printing support
   static void print_labels_on(outputStream* st, const char* c);
diff --git a/src/share/vm/memory/metablock.hpp b/src/share/vm/memory/metablock.hpp
new file mode 100644
index 000000000..220d36148
--- /dev/null
+++ b/src/share/vm/memory/metablock.hpp
@@ -0,0 +1,103 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+#ifndef SHARE_VM_MEMORY_METABLOCK_HPP
+#define SHARE_VM_MEMORY_METABLOCK_HPP
+
+// Metablock are the unit of allocation from a Chunk.  It is initialized
+// with the size of the requested allocation.  That size is overwritten
+// once the allocation returns.
+//
+// A Metablock may be reused by its SpaceManager but are never moved between
+// SpaceManagers.  There is no explicit link to the Metachunk
+// from which it was allocated.  Metablock may be deallocated and
+// put on a freelist but the space is never freed, rather
+// the Metachunk it is a part of will be deallocated when it's
+// associated class loader is collected.
+
+class Metablock VALUE_OBJ_CLASS_SPEC {
+  friend class VMStructs;
+ private:
+  // Used to align the allocation (see below).
+  union block_t {
+    void* _data[3];
+    struct header_t {
+      size_t _word_size;
+      Metablock* _next;
+      Metablock* _prev;
+    } _header;
+  } _block;
+  static size_t _min_block_byte_size;
+  static size_t _overhead;
+
+  typedef union block_t Block;
+  typedef struct header_t Header;
+  const Block* block() const { return &_block; }
+  const Block::header_t* header() const { return &(block()->_header); }
+ public:
+
+  static Metablock* initialize(MetaWord* p, size_t word_size);
+
+  // This places the body of the block at a 2 word boundary
+  // because every block starts on a 2 word boundary.  Work out
+  // how to make the body on a 2 word boundary if the block
+  // starts on a arbitrary boundary.  JJJ
+
+  size_t word_size() const  { return header()->_word_size; }
+  void set_word_size(size_t v) { _block._header._word_size = v; }
+  size_t size() const volatile { return _block._header._word_size; }
+  void set_size(size_t v) { _block._header._word_size = v; }
+  Metablock* next() const { return header()->_next; }
+  void set_next(Metablock* v) { _block._header._next = v; }
+  Metablock* prev() const { return header()->_prev; }
+  void set_prev(Metablock* v) { _block._header._prev = v; }
+
+  static size_t min_block_byte_size() { return _min_block_byte_size; }
+  static size_t overhead() { return _overhead; }
+
+  bool is_free()                 { return header()->_word_size != 0; }
+  void clear_next()              { set_next(NULL); }
+  void link_prev(Metablock* ptr) { set_prev(ptr); }
+  uintptr_t* end()              { return ((uintptr_t*) this) + size(); }
+  bool cantCoalesce() const     { return false; }
+  void link_next(Metablock* ptr) { set_next(ptr); }
+  void link_after(Metablock* ptr){
+    link_next(ptr);
+    if (ptr != NULL) ptr->link_prev(this);
+  }
+
+  // Should not be needed in a free list of Metablocks
+  void markNotFree()            { ShouldNotReachHere(); }
+
+  // Debug support
+#ifdef ASSERT
+  void* prev_addr() const { return (void*)&_block._header._prev; }
+  void* next_addr() const { return (void*)&_block._header._next; }
+  void* size_addr() const { return (void*)&_block._header._word_size; }
+#endif
+  bool verify_chunk_in_free_list(Metablock* tc) const { return true; }
+  bool verify_par_locked() { return true; }
+
+  void assert_is_mangled() const {/* Don't check "\*/}
+};
+#endif // SHARE_VM_MEMORY_METABLOCK_HPP
diff --git a/src/share/vm/memory/metachunk.hpp b/src/share/vm/memory/metachunk.hpp
new file mode 100644
index 000000000..dd461972a
--- /dev/null
+++ b/src/share/vm/memory/metachunk.hpp
@@ -0,0 +1,133 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+#ifndef SHARE_VM_MEMORY_METACHUNK_HPP
+#define SHARE_VM_MEMORY_METACHUNK_HPP
+
+//  Metachunk - Quantum of allocation from a Virtualspace
+//    Metachunks are reused (when freed are put on a global freelist) and
+//    have no permanent association to a SpaceManager.
+
+//            +--------------+ <- end
+//            |              |          --+       ---+
+//            |              |            | free     |
+//            |              |            |          |
+//            |              |            |          | capacity
+//            |              |            |          |
+//            |              | <- top   --+          |
+//            |              |           ---+        |
+//            |              |              | used   |
+//            |              |              |        |
+//            |              |              |        |
+//            +--------------+ <- bottom ---+     ---+
+
+class Metachunk VALUE_OBJ_CLASS_SPEC {
+  // link to support lists of chunks
+  Metachunk* _next;
+  Metachunk* _prev;
+
+  MetaWord* _bottom;
+  MetaWord* _end;
+  MetaWord* _top;
+  size_t _word_size;
+  // Used in a guarantee() so included in the Product builds
+  // even through it is only for debugging.
+  bool _is_free;
+
+  // Metachunks are allocated out of a MetadataVirtualSpace and
+  // and use some of its space to describe itself (plus alignment
+  // considerations).  Metadata is allocated in the rest of the chunk.
+  // This size is the overhead of maintaining the Metachunk within
+  // the space.
+  static size_t _overhead;
+
+  void set_bottom(MetaWord* v) { _bottom = v; }
+  void set_end(MetaWord* v) { _end = v; }
+  void set_top(MetaWord* v) { _top = v; }
+  void set_word_size(size_t v) { _word_size = v; }
+ public:
+#ifdef ASSERT
+  Metachunk() : _bottom(NULL), _end(NULL), _top(NULL), _is_free(false) {}
+#else
+  Metachunk() : _bottom(NULL), _end(NULL), _top(NULL) {}
+#endif
+
+  // Used to add a Metachunk to a list of Metachunks
+  void set_next(Metachunk* v) { _next = v; assert(v != this, "Boom");}
+  void set_prev(Metachunk* v) { _prev = v; assert(v != this, "Boom");}
+
+  MetaWord* allocate(size_t word_size);
+  static Metachunk* initialize(MetaWord* ptr, size_t word_size);
+
+  // Accessors
+  Metachunk* next() const { return _next; }
+  Metachunk* prev() const { return _prev; }
+  MetaWord* bottom() const { return _bottom; }
+  MetaWord* end() const { return _end; }
+  MetaWord* top() const { return _top; }
+  size_t word_size() const { return _word_size; }
+  size_t size() const volatile { return _word_size; }
+  void set_size(size_t v) { _word_size = v; }
+  bool is_free() { return _is_free; }
+  void set_is_free(bool v) { _is_free = v; }
+  static size_t overhead() { return _overhead; }
+  void clear_next()              { set_next(NULL); }
+  void link_prev(Metachunk* ptr) { set_prev(ptr); }
+  uintptr_t* end()              { return ((uintptr_t*) this) + size(); }
+  bool cantCoalesce() const     { return false; }
+  void link_next(Metachunk* ptr) { set_next(ptr); }
+  void link_after(Metachunk* ptr){
+    link_next(ptr);
+    if (ptr != NULL) ptr->link_prev(this);
+  }
+
+  // Reset top to bottom so chunk can be reused.
+  void reset_empty() { _top = (_bottom + _overhead); }
+  bool is_empty() { return _top == (_bottom + _overhead); }
+
+  // used (has been allocated)
+  // free (available for future allocations)
+  // capacity (total size of chunk)
+  size_t used_word_size();
+  size_t free_word_size();
+  size_t capacity_word_size();
+
+  // Debug support
+#ifdef ASSERT
+  void* prev_addr() const { return (void*)&_prev; }
+  void* next_addr() const { return (void*)&_next; }
+  void* size_addr() const { return (void*)&_word_size; }
+#endif
+  bool verify_chunk_in_free_list(Metachunk* tc) const { return true; }
+  bool verify_par_locked() { return true; }
+
+  void assert_is_mangled() const {/* Don't check "\*/}
+
+#ifdef ASSERT
+  void mangle();
+#endif // ASSERT
+
+  void print_on(outputStream* st) const;
+  void verify();
+};
+#endif  // SHARE_VM_MEMORY_METACHUNK_HPP
diff --git a/src/share/vm/memory/metaspace.cpp b/src/share/vm/memory/metaspace.cpp
index 0396eef2d..fc2609f2a 100644
--- a/src/share/vm/memory/metaspace.cpp
+++ b/src/share/vm/memory/metaspace.cpp
@@ -24,9 +24,12 @@
 #include "precompiled.hpp"
 #include "gc_interface/collectedHeap.hpp"
 #include "memory/binaryTreeDictionary.hpp"
+#include "memory/freeList.hpp"
 #include "memory/collectorPolicy.hpp"
 #include "memory/filemap.hpp"
 #include "memory/freeList.hpp"
+#include "memory/metablock.hpp"
+#include "memory/metachunk.hpp"
 #include "memory/metaspace.hpp"
 #include "memory/metaspaceShared.hpp"
 #include "memory/resourceArea.hpp"
@@ -37,15 +40,8 @@
 #include "utilities/copy.hpp"
 #include "utilities/debug.hpp"
 
-// Define this macro to deallocate Metablock.  If not defined,
-// blocks are not yet deallocated and are only mangled.
-#undef DEALLOCATE_BLOCKS
-
-// Easily recognizable patterns
-// These patterns can be the same in 32bit or 64bit since
-// they only have to be easily recognizable.
-const void* metaspace_allocation_leader = (void*) 0X11111111;
-const void* metaspace_allocation_trailer = (void*) 0X77777777;
+typedef BinaryTreeDictionary<Metablock, FreeList> BlockTreeDictionary;
+typedef BinaryTreeDictionary<Metachunk, FreeList> ChunkTreeDictionary;
 
 // Parameters for stress mode testing
 const uint metadata_deallocate_a_lot_block = 10;
@@ -53,7 +49,6 @@ const uint metadata_deallocate_a_lock_chunk = 3;
 size_t const allocation_from_dictionary_limit = 64 * K;
 const size_t metadata_chunk_initialize = 0xf7f7f7f7;
 const size_t metadata_deallocate = 0xf5f5f5f5;
-const size_t metadata_space_manager_allocate = 0xf3f3f3f3;
 
 MetaWord* last_allocated = 0;
 
@@ -62,11 +57,12 @@ enum ChunkIndex {
   SmallIndex = 0,
   MediumIndex = 1,
   HumongousIndex = 2,
-  NumberOfFreeLists = 3
+  NumberOfFreeLists = 2,
+  NumberOfInUseLists = 3
 };
 
 static ChunkIndex next_chunk_index(ChunkIndex i) {
-  assert(i < NumberOfFreeLists, "Out of bound");
+  assert(i < NumberOfInUseLists, "Out of bound");
   return (ChunkIndex) (i+1);
 }
 
@@ -100,164 +96,13 @@ bool MetaspaceGC::_should_concurrent_collect = false;
 // the Chunk after the header for the Chunk) where as Metachunks
 // point to space in a VirtualSpace.  To replace Metachunks with
 // Chunks, change Chunks so that they can be allocated out of a VirtualSpace.
-//
-
-// Metablock are the unit of allocation from a Chunk.  It contains
-// the size of the requested allocation in a debug build.
-// Also in a debug build it has a marker before and after the
-// body of the block. The address of the body is the address returned
-// by the allocation.
-//
-// Layout in a debug build.  In a product build only the body is present.
-//
-//     +-----------+-----------+------------+     +-----------+
-//     | word size | leader    | body       | ... | trailer   |
-//     +-----------+-----------+------------+     +-----------+
-//
-// A Metablock may be reused by its SpaceManager but are never moved between
-// SpaceManagers.  There is no explicit link to the Metachunk
-// from which it was allocated.  Metablock are not deallocated, rather
-// the Metachunk it is a part of will be deallocated when it's
-// associated class loader is collected.
-//
-// When the word size of a block is passed in to the deallocation
-// call the word size no longer needs to be part of a Metablock.
-
-class Metablock {
-  friend class VMStructs;
- private:
-  // Used to align the allocation (see below) and for debugging.
-#ifdef ASSERT
-  struct {
-    size_t _word_size;
-    void*  _leader;
-  } _header;
-  void* _data[1];
-#endif
-  static size_t _overhead;
-
-#ifdef ASSERT
-  void set_word_size(size_t v) { _header._word_size = v; }
-  void* leader() { return _header._leader; }
-  void* trailer() {
-    jlong index = (jlong) _header._word_size - sizeof(_header)/BytesPerWord - 1;
-    assert(index > 0, err_msg("Bad indexling of trailer %d", index));
-    void** ptr = &_data[index];
-    return *ptr;
-  }
-  void set_leader(void* v) { _header._leader = v; }
-  void set_trailer(void* v) {
-    void** ptr = &_data[_header._word_size - sizeof(_header)/BytesPerWord - 1];
-    *ptr = v;
-  }
- public:
-  size_t word_size() { return _header._word_size; }
-#endif
- public:
-
-  static Metablock* initialize(MetaWord* p, size_t word_size);
-
-  // This places the body of the block at a 2 word boundary
-  // because every block starts on a 2 word boundary.  Work out
-  // how to make the body on a 2 word boundary if the block
-  // starts on a arbitrary boundary.  JJJ
-
-#ifdef ASSERT
-  MetaWord* data() { return (MetaWord*) &_data[0]; }
-#else
-  MetaWord* data() { return (MetaWord*) this; }
-#endif
-  static Metablock* metablock_from_data(MetaWord* p) {
+size_t Metablock::_min_block_byte_size = sizeof(Metablock);
 #ifdef ASSERT
-    size_t word_offset = offset_of(Metablock, _data)/BytesPerWord;
-    Metablock* result = (Metablock*) (p - word_offset);
-    return result;
+  size_t Metablock::_overhead =
+    Chunk::aligned_overhead_size(sizeof(Metablock)) / BytesPerWord;
 #else
-    return (Metablock*) p;
+  size_t Metablock::_overhead = 0;
 #endif
-  }
-
-  static size_t overhead() { return _overhead; }
-  void verify();
-};
-
-//  Metachunk - Quantum of allocation from a Virtualspace
-//    Metachunks are reused (when freed are put on a global freelist) and
-//    have no permanent association to a SpaceManager.
-
-//            +--------------+ <- end
-//            |              |          --+       ---+
-//            |              |            | free     |
-//            |              |            |          |
-//            |              |            |          | capacity
-//            |              |            |          |
-//            |              | <- top   --+          |
-//            |              |           ---+        |
-//            |              |              | used   |
-//            |              |              |        |
-//            |              |              |        |
-//            +--------------+ <- bottom ---+     ---+
-
-class Metachunk VALUE_OBJ_CLASS_SPEC {
-  // link to support lists of chunks
-  Metachunk* _next;
-
-  MetaWord* _bottom;
-  MetaWord* _end;
-  MetaWord* _top;
-  size_t _word_size;
-
-  // Metachunks are allocated out of a MetadataVirtualSpace and
-  // and use some of its space to describe itself (plus alignment
-  // considerations).  Metadata is allocated in the rest of the chunk.
-  // This size is the overhead of maintaining the Metachunk within
-  // the space.
-  static size_t _overhead;
-
-  void set_bottom(MetaWord* v) { _bottom = v; }
-  void set_end(MetaWord* v) { _end = v; }
-  void set_top(MetaWord* v) { _top = v; }
-  void set_word_size(size_t v) { _word_size = v; }
- public:
-
-  // Used to add a Metachunk to a list of Metachunks
-  void set_next(Metachunk* v) { _next = v; assert(v != this, "Boom");}
-
-  Metablock* allocate(size_t word_size);
-  static Metachunk* initialize(MetaWord* ptr, size_t word_size);
-
-  // Accessors
-  Metachunk* next() const { return _next; }
-  MetaWord* bottom() const { return _bottom; }
-  MetaWord* end() const { return _end; }
-  MetaWord* top() const { return _top; }
-  size_t word_size() const { return _word_size; }
-  static size_t overhead() { return _overhead; }
-
-  // Reset top to bottom so chunk can be reused.
-  void reset_empty() { _top = (_bottom + _overhead); }
-  bool is_empty() { return _top == (_bottom + _overhead); }
-
-  // used (has been allocated)
-  // free (available for future allocations)
-  // capacity (total size of chunk)
-  size_t used_word_size();
-  size_t free_word_size();
-  size_t capacity_word_size();
-
-#ifdef ASSERT
-  void mangle() {
-    // Mangle the payload of the chunk and not the links that
-    // maintain list of chunks.
-    HeapWord* start = (HeapWord*)(bottom() + overhead());
-    size_t word_size = capacity_word_size() - overhead();
-    Copy::fill_to_words(start, word_size, metadata_chunk_initialize);
-  }
-#endif // ASSERT
-
-  void print_on(outputStream* st) const;
-  void verify();
-};
 
 
 // Pointer to list of Metachunks.
@@ -292,7 +137,10 @@ class ChunkManager VALUE_OBJ_CLASS_SPEC {
   //   SmallChunk
   //   MediumChunk
   //   HumongousChunk
-  ChunkList _free_chunks[3];
+  ChunkList _free_chunks[NumberOfFreeLists];
+
+  //   HumongousChunk
+  ChunkTreeDictionary _humongous_dictionary;
 
   // ChunkManager in all lists of this type
   size_t _free_chunks_total;
@@ -337,7 +185,9 @@ class ChunkManager VALUE_OBJ_CLASS_SPEC {
   }
   ChunkList* free_medium_chunks() { return &_free_chunks[1]; }
   ChunkList* free_small_chunks() { return &_free_chunks[0]; }
-  ChunkList* free_humongous_chunks() { return &_free_chunks[2]; }
+  ChunkTreeDictionary* humongous_dictionary() {
+    return &_humongous_dictionary;
+  }
 
   ChunkList* free_chunks(ChunkIndex index);
 
@@ -356,41 +206,35 @@ class ChunkManager VALUE_OBJ_CLASS_SPEC {
 
   void locked_print_free_chunks(outputStream* st);
   void locked_print_sum_free_chunks(outputStream* st);
+
+  void print_on(outputStream* st);
 };
 
 
 // Used to manage the free list of Metablocks (a block corresponds
 // to the allocation of a quantum of metadata).
 class BlockFreelist VALUE_OBJ_CLASS_SPEC {
-#ifdef DEALLOCATE_BLOCKS
-  BinaryTreeDictionary<Metablock>* _dictionary;
-#endif
-  static Metablock* initialize_free_chunk(Metablock* block, size_t word_size);
+  BlockTreeDictionary* _dictionary;
+  static Metablock* initialize_free_chunk(MetaWord* p, size_t word_size);
 
-#ifdef DEALLOCATE_BLOCKS
   // Accessors
-  BinaryTreeDictionary<Metablock>* dictionary() const { return _dictionary; }
-#endif
+  BlockTreeDictionary* dictionary() const { return _dictionary; }
 
  public:
   BlockFreelist();
   ~BlockFreelist();
 
   // Get and return a block to the free list
-  Metablock* get_block(size_t word_size);
-  void return_block(Metablock* block, size_t word_size);
+  MetaWord* get_block(size_t word_size);
+  void return_block(MetaWord* p, size_t word_size);
 
-  size_t totalSize() {
-#ifdef DEALLOCATE_BLOCKS
-    if (dictionary() == NULL) {
-      return 0;
-    } else {
-      return dictionary()->totalSize();
-    }
-#else
+  size_t total_size() {
+  if (dictionary() == NULL) {
     return 0;
-#endif
+  } else {
+    return dictionary()->total_size();
   }
+}
 
   void print_on(outputStream* st) const;
 };
@@ -600,7 +444,6 @@ class VirtualSpaceList : public CHeapObj<mtClass> {
   };
 };
 
-
 class Metadebug : AllStatic {
   // Debugging support for Metaspaces
   static int _deallocate_block_a_lot_count;
@@ -655,7 +498,7 @@ class SpaceManager : public CHeapObj<mtClass> {
   // List of chunks in use by this SpaceManager.  Allocations
   // are done from the current chunk.  The list is used for deallocating
   // chunks when the SpaceManager is freed.
-  Metachunk* _chunks_in_use[NumberOfFreeLists];
+  Metachunk* _chunks_in_use[NumberOfInUseLists];
   Metachunk* _current_chunk;
 
   // Virtual space where allocation comes from.
@@ -700,24 +543,6 @@ class SpaceManager : public CHeapObj<mtClass> {
   // Add chunk to the list of chunks in use
   void add_chunk(Metachunk* v, bool make_current);
 
-  // Debugging support
-  void verify_chunks_in_use_index(ChunkIndex index, Metachunk* v) {
-    switch (index) {
-    case 0:
-      assert(v->word_size() == SmallChunk, "Not a SmallChunk");
-      break;
-    case 1:
-      assert(v->word_size() == MediumChunk, "Not a MediumChunk");
-      break;
-    case 2:
-      assert(v->word_size() > MediumChunk, "Not a HumongousChunk");
-      break;
-    default:
-      assert(false, "Wrong list.");
-    }
-  }
-
- protected:
   Mutex* lock() const { return _lock; }
 
  public:
@@ -751,10 +576,10 @@ class SpaceManager : public CHeapObj<mtClass> {
   MetaWord* allocate(size_t word_size);
 
   // Helper for allocations
-  Metablock* allocate_work(size_t word_size);
+  MetaWord* allocate_work(size_t word_size);
 
   // Returns a block to the per manager freelist
-  void deallocate(MetaWord* p);
+  void deallocate(MetaWord* p, size_t word_size);
 
   // Based on the allocation size and a minimum chunk size,
   // returned chunk size (for expanding space for chunk allocation).
@@ -763,7 +588,7 @@ class SpaceManager : public CHeapObj<mtClass> {
   // Called when an allocation from the current chunk fails.
   // Gets a new chunk (may require getting a new virtual space),
   // and allocates from that chunk.
-  Metablock* grow_and_allocate(size_t word_size);
+  MetaWord* grow_and_allocate(size_t word_size);
 
   // debugging support.
 
@@ -780,6 +605,8 @@ class SpaceManager : public CHeapObj<mtClass> {
 
 uint const SpaceManager::_small_chunk_limit = 4;
 
+
+
 const char* SpaceManager::_expand_lock_name =
   "SpaceManager chunk allocation lock";
 const int SpaceManager::_expand_lock_rank = Monitor::leaf - 1;
@@ -788,39 +615,26 @@ Mutex* const SpaceManager::_expand_lock =
             SpaceManager::_expand_lock_name,
             Mutex::_allow_vm_block_flag);
 
-#ifdef ASSERT
-size_t Metablock::_overhead =
-  Chunk::aligned_overhead_size(sizeof(Metablock)) / BytesPerWord;
-#else
-size_t Metablock::_overhead = 0;
-#endif
 size_t Metachunk::_overhead =
   Chunk::aligned_overhead_size(sizeof(Metachunk)) / BytesPerWord;
 
 // New blocks returned by the Metaspace are zero initialized.
 // We should fix the constructors to not assume this instead.
 Metablock* Metablock::initialize(MetaWord* p, size_t word_size) {
+  if (p == NULL) {
+    return NULL;
+  }
+
   Metablock* result = (Metablock*) p;
 
   // Clear the memory
   Copy::fill_to_aligned_words((HeapWord*)result, word_size);
 #ifdef ASSERT
   result->set_word_size(word_size);
-  // Check after work size is set.
-  result->set_leader((void*) metaspace_allocation_leader);
-  result->set_trailer((void*) metaspace_allocation_trailer);
 #endif
   return result;
 }
 
-void Metablock::verify() {
-#ifdef ASSERT
-  assert(leader() == metaspace_allocation_leader &&
-         trailer() == metaspace_allocation_trailer,
-         "block has been corrupted");
-#endif
-}
-
 // Metachunk methods
 
 Metachunk* Metachunk::initialize(MetaWord* ptr, size_t word_size) {
@@ -843,18 +657,13 @@ Metachunk* Metachunk::initialize(MetaWord* ptr, size_t word_size) {
 }
 
 
-Metablock* Metachunk::allocate(size_t word_size) {
-  Metablock* result = NULL;
+MetaWord* Metachunk::allocate(size_t word_size) {
+  MetaWord* result = NULL;
   // If available, bump the pointer to allocate.
   if (free_word_size() >= word_size) {
-    result = Metablock::initialize(_top, word_size);
+    result = _top;
     _top = _top + word_size;
   }
-#ifdef ASSERT
-  assert(result == NULL ||
-         result->word_size() == word_size,
-         "Block size is not set correctly");
-#endif
   return result;
 }
 
@@ -878,103 +687,85 @@ void Metachunk::print_on(outputStream* st) const {
                bottom(), top(), end(), word_size());
 }
 
+#ifdef ASSERT
+void Metachunk::mangle() {
+  // Mangle the payload of the chunk and not the links that
+  // maintain list of chunks.
+  HeapWord* start = (HeapWord*)(bottom() + overhead());
+  size_t word_size = capacity_word_size() - overhead();
+  Copy::fill_to_words(start, word_size, metadata_chunk_initialize);
+}
+#endif // ASSERT
 
 void Metachunk::verify() {
 #ifdef ASSERT
   // Cannot walk through the blocks unless the blocks have
   // headers with sizes.
-  MetaWord* curr = bottom() + overhead();
-  while (curr < top()) {
-    Metablock* block = (Metablock*) curr;
-    size_t word_size = block->word_size();
-    block->verify();
-    curr = curr + word_size;
-  }
+  assert(_bottom <= _top &&
+         _top <= _end,
+         "Chunk has been smashed");
+  assert(SpaceManager::is_humongous(_word_size) ||
+         _word_size == SpaceManager::MediumChunk ||
+         _word_size == SpaceManager::SmallChunk,
+         "Chunk size is wrong");
 #endif
   return;
 }
 
 // BlockFreelist methods
 
-#ifdef DEALLOCATE_BLOCKS
 BlockFreelist::BlockFreelist() : _dictionary(NULL) {}
-#else
-BlockFreelist::BlockFreelist() {}
-#endif
 
 BlockFreelist::~BlockFreelist() {
-#ifdef DEALLOCATE_BLOCKS
   if (_dictionary != NULL) {
     if (Verbose && TraceMetadataChunkAllocation) {
       _dictionary->print_free_lists(gclog_or_tty);
     }
     delete _dictionary;
   }
-#endif
 }
 
-Metablock* BlockFreelist::initialize_free_chunk(Metablock* block, size_t word_size) {
-#ifdef DEALLOCATE_BLOCKS
-#ifdef ASSERT
-  assert(word_size = block->word_size(), "Wrong chunk size");
-#endif
-  Metablock* result = block;
-  result->setSize(word_size);
-  result->linkPrev(NULL);
-  result->linkNext(NULL);
+Metablock* BlockFreelist::initialize_free_chunk(MetaWord* p, size_t word_size) {
+  Metablock* block = (Metablock*) p;
+  block->set_word_size(word_size);
+  block->set_prev(NULL);
+  block->set_next(NULL);
 
-  return result;
-#else
-  ShouldNotReachHere();
   return block;
-#endif
 }
 
-void BlockFreelist::return_block(Metablock* block, size_t word_size) {
-#ifdef ASSERT
-  assert(word_size = block->word_size(), "Block size is wrong");;
-#endif
-  Metablock* free_chunk = initialize_free_chunk(block, word_size);
-#ifdef DEALLOCATE_BLOCKS
+void BlockFreelist::return_block(MetaWord* p, size_t word_size) {
+  Metablock* free_chunk = initialize_free_chunk(p, word_size);
   if (dictionary() == NULL) {
-   _dictionary = new BinaryTreeDictionary<Metablock>(false /* adaptive_freelists */);
+   _dictionary = new BlockTreeDictionary();
   }
-  dictionary()->returnChunk(free_chunk);
-#endif
+  dictionary()->return_chunk(free_chunk);
 }
 
-Metablock* BlockFreelist::get_block(size_t word_size) {
-#ifdef DEALLOCATE_BLOCKS
+MetaWord* BlockFreelist::get_block(size_t word_size) {
   if (dictionary() == NULL) {
     return NULL;
   }
 
-  Metablock* free_chunk =
-    dictionary()->getChunk(word_size, FreeBlockDictionary<Metablock>::exactly);
-#else
-  Metablock* free_chunk = NULL;
-#endif
-  if (free_chunk == NULL) {
+  if (word_size < TreeChunk<Metablock, FreeList>::min_size()) {
+    // Dark matter.  Too small for dictionary.
     return NULL;
   }
-  assert(free_chunk->word_size() == word_size, "Size of chunk is incorrect");
-  Metablock* block = Metablock::initialize((MetaWord*) free_chunk, word_size);
-#ifdef ASSERT
-  assert(block->word_size() == word_size, "Block size is not set correctly");
-#endif
 
-  return block;
+  Metablock* free_block =
+    dictionary()->get_chunk(word_size, FreeBlockDictionary<Metablock>::exactly);
+  if (free_block == NULL) {
+    return NULL;
+  }
+
+  return (MetaWord*) free_block;
 }
 
 void BlockFreelist::print_on(outputStream* st) const {
-#ifdef DEALLOCATE_BLOCKS
   if (dictionary() == NULL) {
     return;
   }
   dictionary()->print_free_lists(st);
-#else
-  return;
-#endif
 }
 
 // VirtualSpaceNode methods
@@ -1597,14 +1388,11 @@ void Metadebug::deallocate_block_a_lot(SpaceManager* sm,
         Metadebug::deallocate_block_a_lot_count() % MetaDataDeallocateALotInterval == 0 ) {
     Metadebug::set_deallocate_block_a_lot_count(0);
     for (uint i = 0; i < metadata_deallocate_a_lot_block; i++) {
-      Metablock* dummy_block = sm->allocate_work(raw_word_size);
+      MetaWord* dummy_block = sm->allocate_work(raw_word_size);
       if (dummy_block == 0) {
         break;
       }
-#ifdef ASSERT
-      assert(dummy_block->word_size() == raw_word_size, "Block size is not set correctly");
-#endif
-      sm->deallocate(dummy_block->data());
+      sm->deallocate(dummy_block, raw_word_size);
     }
   } else {
     Metadebug::inc_deallocate_block_a_lot_count();
@@ -1784,8 +1572,8 @@ void ChunkManager::verify() {
 }
 
 void ChunkManager::locked_verify() {
-  locked_verify_free_chunks_total();
   locked_verify_free_chunks_count();
+  locked_verify_free_chunks_total();
 }
 
 void ChunkManager::locked_print_free_chunks(outputStream* st) {
@@ -1803,7 +1591,6 @@ ChunkList* ChunkManager::free_chunks(ChunkIndex index) {
   return &_free_chunks[index];
 }
 
-
 // These methods that sum the free chunk lists are used in printing
 // methods that are used in product builds.
 size_t ChunkManager::sum_free_chunks() {
@@ -1818,6 +1605,7 @@ size_t ChunkManager::sum_free_chunks() {
 
     result = result + list->sum_list_capacity();
   }
+  result = result + humongous_dictionary()->total_size();
   return result;
 }
 
@@ -1831,6 +1619,7 @@ size_t ChunkManager::sum_free_chunks_count() {
     }
     count = count + list->sum_list_count();
   }
+  count = count + humongous_dictionary()->total_free_blocks();
   return count;
 }
 
@@ -1875,23 +1664,24 @@ Metachunk* ChunkManager::free_chunks_get(size_t word_size) {
   assert_lock_strong(SpaceManager::expand_lock());
 
   locked_verify();
-  ChunkList* free_list = find_free_chunks_list(word_size);
-  assert(free_list != NULL, "Sanity check");
 
-  Metachunk* chunk = free_list->head();
-  debug_only(Metachunk* debug_head = chunk;)
+  Metachunk* chunk = NULL;
+  if (!SpaceManager::is_humongous(word_size)) {
+    ChunkList* free_list = find_free_chunks_list(word_size);
+    assert(free_list != NULL, "Sanity check");
 
-  if (chunk == NULL) {
-    return NULL;
-  }
+    chunk = free_list->head();
+    debug_only(Metachunk* debug_head = chunk;)
+
+    if (chunk == NULL) {
+      return NULL;
+    }
 
-  Metachunk* prev_chunk = chunk;
-  if (chunk->word_size() == word_size) {
-    // Chunk is being removed from the chunks free list.
-    dec_free_chunks_total(chunk->capacity_word_size());
     // Remove the chunk as the head of the list.
     free_list->set_head(chunk->next());
     chunk->set_next(NULL);
+    // Chunk has been removed from the chunks free list.
+    dec_free_chunks_total(chunk->capacity_word_size());
 
     if (TraceMetadataChunkAllocation && Verbose) {
       tty->print_cr("ChunkManager::free_chunks_get: free_list "
@@ -1899,79 +1689,24 @@ Metachunk* ChunkManager::free_chunks_get(size_t word_size) {
                     free_list, chunk, chunk->word_size());
     }
   } else {
-    assert(SpaceManager::is_humongous(word_size),
-      "Should only need to check humongous");
-    // This code to find the best fit is just for purposes of
-    // investigating the loss due to fragmentation on a humongous
-    // chunk.  It will be replace by a binaryTreeDictionary for
-    // the humongous chunks.
-    uint count = 0;
-    Metachunk* best_fit = NULL;
-    Metachunk* best_fit_prev = NULL;
-    while (chunk != NULL) {
-      count++;
-      if (chunk->word_size() < word_size) {
-        prev_chunk = chunk;
-        chunk = chunk->next();
-      } else if (chunk->word_size() == word_size) {
-        break;
-      } else {
-        if (best_fit == NULL ||
-            best_fit->word_size() > chunk->word_size()) {
-          best_fit_prev = prev_chunk;
-          best_fit = chunk;
-        }
-        prev_chunk = chunk;
-        chunk = chunk->next();
+    chunk = humongous_dictionary()->get_chunk(
+      word_size,
+      FreeBlockDictionary<Metachunk>::atLeast);
+
+    if (chunk != NULL) {
+      if (TraceMetadataHumongousAllocation) {
+        size_t waste = chunk->word_size() - word_size;
+        tty->print_cr("Free list allocate humongous chunk size " SIZE_FORMAT
+                      " for requested size " SIZE_FORMAT
+                      " waste " SIZE_FORMAT,
+                      chunk->word_size(), word_size, waste);
       }
+      // Chunk is being removed from the chunks free list.
+      dec_free_chunks_total(chunk->capacity_word_size());
+#ifdef ASSERT
+      chunk->set_is_free(false);
+#endif
     }
-      if (chunk == NULL) {
-        prev_chunk = best_fit_prev;
-        chunk = best_fit;
-      }
-      if (chunk != NULL) {
-        if (TraceMetadataHumongousAllocation) {
-          size_t waste = chunk->word_size() - word_size;
-          tty->print_cr("Free list allocate humongous chunk size " SIZE_FORMAT
-                        " for requested size " SIZE_FORMAT
-                        " waste " SIZE_FORMAT
-                        " found at " SIZE_FORMAT " of " SIZE_FORMAT,
-                        chunk->word_size(), word_size, waste,
-                        count, free_list->sum_list_count());
-        }
-        // Chunk is being removed from the chunks free list.
-        dec_free_chunks_total(chunk->capacity_word_size());
-        // Remove the chunk if it is at the head of the list.
-        if (chunk == free_list->head()) {
-          free_list->set_head(chunk->next());
-
-          if (TraceMetadataHumongousAllocation) {
-            tty->print_cr("ChunkManager::free_chunks_get: humongous free_list "
-                          PTR_FORMAT " chunk " PTR_FORMAT " size " SIZE_FORMAT
-                          " new head " PTR_FORMAT,
-                          free_list, chunk, chunk->word_size(),
-                          free_list->head());
-          }
-        } else {
-          // Remove a chunk in the interior of the list
-          prev_chunk->set_next(chunk->next());
-
-          if (TraceMetadataHumongousAllocation) {
-            tty->print_cr("ChunkManager::free_chunks_get: humongous free_list "
-                          PTR_FORMAT " chunk " PTR_FORMAT " size " SIZE_FORMAT
-                          PTR_FORMAT "  prev " PTR_FORMAT " next " PTR_FORMAT,
-                          free_list, chunk, chunk->word_size(),
-                          prev_chunk, chunk->next());
-          }
-        }
-        chunk->set_next(NULL);
-      } else {
-        if (TraceMetadataHumongousAllocation) {
-          tty->print_cr("ChunkManager::free_chunks_get: New humongous chunk of size "
-                        SIZE_FORMAT,
-                        word_size);
-        }
-      }
   }
   locked_verify();
   return chunk;
@@ -2000,12 +1735,18 @@ Metachunk* ChunkManager::chunk_freelist_allocate(size_t word_size) {
   return chunk;
 }
 
+void ChunkManager::print_on(outputStream* out) {
+  if (PrintFLSStatistics != 0) {
+    humongous_dictionary()->report_statistics();
+  }
+}
+
 // SpaceManager methods
 
 size_t SpaceManager::sum_free_in_chunks_in_use() const {
   MutexLockerEx cl(lock(), Mutex::_no_safepoint_check_flag);
   size_t free = 0;
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
     Metachunk* chunk = chunks_in_use(i);
     while (chunk != NULL) {
       free += chunk->free_word_size();
@@ -2018,11 +1759,12 @@ size_t SpaceManager::sum_free_in_chunks_in_use() const {
 size_t SpaceManager::sum_waste_in_chunks_in_use() const {
   MutexLockerEx cl(lock(), Mutex::_no_safepoint_check_flag);
   size_t result = 0;
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
-   // Count the free space in all the chunk but not the
-   // current chunk from which allocations are still being done.
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
+
+
    result += sum_waste_in_chunks_in_use(i);
   }
+
   return result;
 }
 
@@ -2033,10 +1775,10 @@ size_t SpaceManager::sum_waste_in_chunks_in_use(ChunkIndex index) const {
   // Count the free space in all the chunk but not the
   // current chunk from which allocations are still being done.
   if (chunk != NULL) {
-    while (chunk != NULL) {
-      if (chunk != current_chunk()) {
-        result += chunk->free_word_size();
-      }
+    Metachunk* prev = chunk;
+    while (chunk != NULL && chunk != current_chunk()) {
+      result += chunk->free_word_size();
+      prev = chunk;
       chunk = chunk->next();
       count++;
     }
@@ -2047,7 +1789,7 @@ size_t SpaceManager::sum_waste_in_chunks_in_use(ChunkIndex index) const {
 size_t SpaceManager::sum_capacity_in_chunks_in_use() const {
   MutexLockerEx cl(lock(), Mutex::_no_safepoint_check_flag);
   size_t sum = 0;
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
     Metachunk* chunk = chunks_in_use(i);
     while (chunk != NULL) {
       // Just changed this sum += chunk->capacity_word_size();
@@ -2061,9 +1803,10 @@ size_t SpaceManager::sum_capacity_in_chunks_in_use() const {
 
 size_t SpaceManager::sum_count_in_chunks_in_use() {
   size_t count = 0;
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
     count = count + sum_count_in_chunks_in_use(i);
   }
+
   return count;
 }
 
@@ -2081,7 +1824,7 @@ size_t SpaceManager::sum_count_in_chunks_in_use(ChunkIndex i) {
 size_t SpaceManager::sum_used_in_chunks_in_use() const {
   MutexLockerEx cl(lock(), Mutex::_no_safepoint_check_flag);
   size_t used = 0;
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
     Metachunk* chunk = chunks_in_use(i);
     while (chunk != NULL) {
       used += chunk->used_word_size();
@@ -2139,15 +1882,13 @@ size_t SpaceManager::calc_chunk_size(size_t word_size) {
     gclog_or_tty->print_cr("  word_size " PTR_FORMAT, word_size);
     gclog_or_tty->print_cr("  chunk_word_size " PTR_FORMAT,
                            chunk_word_size);
-    gclog_or_tty->print_cr("    block overhead " PTR_FORMAT
-                           " chunk overhead " PTR_FORMAT,
-                           Metablock::overhead(),
+    gclog_or_tty->print_cr("    chunk overhead " PTR_FORMAT,
                            Metachunk::overhead());
   }
   return chunk_word_size;
 }
 
-Metablock* SpaceManager::grow_and_allocate(size_t word_size) {
+MetaWord* SpaceManager::grow_and_allocate(size_t word_size) {
   assert(vs_list()->current_virtual_space() != NULL,
          "Should have been set");
   assert(current_chunk() == NULL ||
@@ -2180,7 +1921,7 @@ Metablock* SpaceManager::grow_and_allocate(size_t word_size) {
 void SpaceManager::print_on(outputStream* st) const {
 
   for (ChunkIndex i = SmallIndex;
-       i < NumberOfFreeLists ;
+       i < NumberOfInUseLists ;
        i = next_chunk_index(i) ) {
     st->print_cr("  chunks_in_use " PTR_FORMAT " chunk size " PTR_FORMAT,
                  chunks_in_use(i),
@@ -2191,8 +1932,11 @@ void SpaceManager::print_on(outputStream* st) const {
                sum_waste_in_chunks_in_use(SmallIndex),
                sum_waste_in_chunks_in_use(MediumIndex),
                sum_waste_in_chunks_in_use(HumongousIndex));
-  // Nothing in them yet
-  // block_freelists()->print_on(st);
+  // block free lists
+  if (block_freelists() != NULL) {
+    st->print_cr("total in block free lists " SIZE_FORMAT,
+      block_freelists()->total_size());
+  }
 }
 
 SpaceManager::SpaceManager(Mutex* lock, VirtualSpaceList* vs_list) :
@@ -2200,7 +1944,7 @@ SpaceManager::SpaceManager(Mutex* lock, VirtualSpaceList* vs_list) :
   _allocation_total(0),
   _lock(lock) {
   Metadebug::init_allocation_fail_alot_count();
-  for (ChunkIndex i = SmallIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+  for (ChunkIndex i = SmallIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
     _chunks_in_use[i] = NULL;
   }
   _current_chunk = NULL;
@@ -2262,22 +2006,24 @@ SpaceManager::~SpaceManager() {
   // Humongous chunks are never the current chunk.
   Metachunk* humongous_chunks = chunks_in_use(HumongousIndex);
 
-  if (humongous_chunks != NULL) {
-    chunk_manager->free_humongous_chunks()->add_at_head(humongous_chunks);
-    set_chunks_in_use(HumongousIndex, NULL);
+  while (humongous_chunks != NULL) {
+#ifdef ASSERT
+    humongous_chunks->set_is_free(true);
+#endif
+    Metachunk* next_humongous_chunks = humongous_chunks->next();
+    chunk_manager->humongous_dictionary()->return_chunk(humongous_chunks);
+    humongous_chunks = next_humongous_chunks;
   }
+  set_chunks_in_use(HumongousIndex, NULL);
   chunk_manager->locked_verify();
 }
 
-void SpaceManager::deallocate(MetaWord* p) {
+void SpaceManager::deallocate(MetaWord* p, size_t word_size) {
   assert_lock_strong(_lock);
-  ShouldNotReachHere();  // Where is this needed.
-#ifdef DEALLOCATE_BLOCKS
-  Metablock* block = Metablock::metablock_from_data(p);
-  // This is expense but kept it until integration JJJ
-  assert(contains((address)block), "Block does not belong to this metaspace");
-  block_freelists()->return_block(block, word_size);
-#endif
+  size_t min_size = TreeChunk<Metablock, FreeList>::min_size();
+  assert(word_size >= min_size,
+    err_msg("Should not deallocate dark matter " SIZE_FORMAT, word_size));
+  block_freelists()->return_block(p, word_size);
 }
 
 // Adds a chunk to the list of chunks in use.
@@ -2366,50 +2112,40 @@ void SpaceManager::add_chunk(Metachunk* new_chunk, bool make_current) {
 MetaWord* SpaceManager::allocate(size_t word_size) {
   MutexLockerEx cl(lock(), Mutex::_no_safepoint_check_flag);
 
-  size_t block_overhead = Metablock::overhead();
   // If only the dictionary is going to be used (i.e., no
   // indexed free list), then there is a minimum size requirement.
   // MinChunkSize is a placeholder for the real minimum size JJJ
-  size_t byte_size_with_overhead = (word_size + block_overhead) * BytesPerWord;
-#ifdef DEALLOCATE_BLOCKS
-  size_t raw_bytes_size = MAX2(ARENA_ALIGN(byte_size_with_overhead),
-                               MinChunkSize * BytesPerWord);
-#else
-  size_t raw_bytes_size = ARENA_ALIGN(byte_size_with_overhead);
-#endif
+  size_t byte_size = word_size * BytesPerWord;
+
+  size_t byte_size_with_overhead = byte_size + Metablock::overhead();
+
+  size_t raw_bytes_size = MAX2(byte_size_with_overhead,
+                               Metablock::min_block_byte_size());
+  raw_bytes_size = ARENA_ALIGN(raw_bytes_size);
   size_t raw_word_size = raw_bytes_size / BytesPerWord;
   assert(raw_word_size * BytesPerWord == raw_bytes_size, "Size problem");
 
   BlockFreelist* fl =  block_freelists();
-  Metablock* block = NULL;
+  MetaWord* p = NULL;
   // Allocation from the dictionary is expensive in the sense that
   // the dictionary has to be searched for a size.  Don't allocate
   // from the dictionary until it starts to get fat.  Is this
   // a reasonable policy?  Maybe an skinny dictionary is fast enough
   // for allocations.  Do some profiling.  JJJ
-  if (fl->totalSize() > allocation_from_dictionary_limit) {
-    block = fl->get_block(raw_word_size);
+  if (fl->total_size() > allocation_from_dictionary_limit) {
+    p = fl->get_block(raw_word_size);
   }
-  if (block == NULL) {
-    block = allocate_work(raw_word_size);
-    if (block == NULL) {
-      return NULL;
-    }
+  if (p == NULL) {
+    p = allocate_work(raw_word_size);
   }
   Metadebug::deallocate_block_a_lot(this, raw_word_size);
 
-  // Push the allocation past the word containing the size and leader.
-#ifdef ASSERT
-  MetaWord* result =  block->data();
-  return result;
-#else
-  return (MetaWord*) block;
-#endif
+  return p;
 }
 
 // Returns the address of spaced allocated for "word_size".
 // This methods does not know about blocks (Metablocks)
-Metablock* SpaceManager::allocate_work(size_t word_size) {
+MetaWord* SpaceManager::allocate_work(size_t word_size) {
   assert_lock_strong(_lock);
 #ifdef ASSERT
   if (Metadebug::test_metadata_failure()) {
@@ -2417,7 +2153,7 @@ Metablock* SpaceManager::allocate_work(size_t word_size) {
   }
 #endif
   // Is there space in the current chunk?
-  Metablock* result = NULL;
+  MetaWord* result = NULL;
 
   // For DumpSharedSpaces, only allocate out of the current chunk which is
   // never null because we gave it the size we wanted.   Caller reports out
@@ -2436,8 +2172,8 @@ Metablock* SpaceManager::allocate_work(size_t word_size) {
   }
   if (result > 0) {
     inc_allocation_total(word_size);
-    assert(result != (Metablock*) chunks_in_use(MediumIndex), "Head of the list is being allocated");
-    assert(result->word_size() == word_size, "Size not set correctly");
+    assert(result != (MetaWord*) chunks_in_use(MediumIndex),
+           "Head of the list is being allocated");
   }
 
   return result;
@@ -2447,13 +2183,13 @@ void SpaceManager::verify() {
   // If there are blocks in the dictionary, then
   // verfication of chunks does not work since
   // being in the dictionary alters a chunk.
-  if (block_freelists()->totalSize() == 0) {
+  if (block_freelists()->total_size() == 0) {
     // Skip the small chunks because their next link points to
     // medium chunks.  This is because the small chunk is the
     // current chunk (for allocations) until it is full and the
     // the addition of the next chunk does not NULL the next
     // like of the small chunk.
-    for (ChunkIndex i = MediumIndex; i < NumberOfFreeLists; i = next_chunk_index(i)) {
+    for (ChunkIndex i = MediumIndex; i < NumberOfInUseLists; i = next_chunk_index(i)) {
       Metachunk* curr = chunks_in_use(i);
       while (curr != NULL) {
         curr->verify();
@@ -2492,7 +2228,7 @@ void SpaceManager::dump(outputStream* const out) const {
 
   // Add up statistics for all chunks in this SpaceManager.
   for (ChunkIndex index = SmallIndex;
-       index < NumberOfFreeLists;
+       index < NumberOfInUseLists;
        index = next_chunk_index(index)) {
     for (Metachunk* curr = chunks_in_use(index);
          curr != NULL;
@@ -2521,7 +2257,7 @@ void SpaceManager::dump(outputStream* const out) const {
 #ifdef ASSERT
 void SpaceManager::mangle_freed_chunks() {
   for (ChunkIndex index = SmallIndex;
-       index < NumberOfFreeLists;
+       index < NumberOfInUseLists;
        index = next_chunk_index(index)) {
     for (Metachunk* curr = chunks_in_use(index);
          curr != NULL;
@@ -2833,13 +2569,12 @@ void Metaspace::initialize(Mutex* lock, size_t initial_size) {
   }
 }
 
-
 MetaWord* Metaspace::allocate(size_t word_size, MetadataType mdtype) {
   // DumpSharedSpaces doesn't use class metadata area (yet)
   if (mdtype == ClassType && !DumpSharedSpaces) {
-    return class_vsm()->allocate(word_size);
+    return  class_vsm()->allocate(word_size);
   } else {
-    return vsm()->allocate(word_size);
+    return  vsm()->allocate(word_size);
   }
 }
 
@@ -2853,6 +2588,7 @@ MetaWord* Metaspace::expand_and_allocate(size_t word_size, MetadataType mdtype)
     gclog_or_tty->print_cr("Increase capacity to GC from " SIZE_FORMAT
       " to " SIZE_FORMAT, before_inc, MetaspaceGC::capacity_until_GC());
   }
+
   result = allocate(word_size, mdtype);
 
   return result;
@@ -2889,37 +2625,39 @@ size_t Metaspace::capacity_words(MetadataType mdtype) const {
 void Metaspace::deallocate(MetaWord* ptr, size_t word_size, bool is_class) {
   if (SafepointSynchronize::is_at_safepoint()) {
     assert(Thread::current()->is_VM_thread(), "should be the VM thread");
-    // Don't take lock
-#ifdef DEALLOCATE_BLOCKS
+    // Don't take Heap_lock
+    MutexLocker ml(vsm()->lock());
+    if (word_size < TreeChunk<Metablock, FreeList>::min_size()) {
+      // Dark matter.  Too small for dictionary.
+#ifdef ASSERT
+      Copy::fill_to_words((HeapWord*)ptr, word_size, 0xf5f5f5f5);
+#endif
+      return;
+    }
     if (is_class) {
-      class_vsm()->deallocate(ptr);
+       class_vsm()->deallocate(ptr, word_size);
     } else {
-      vsm()->deallocate(ptr);
+      vsm()->deallocate(ptr, word_size);
     }
-#else
-#ifdef ASSERT
-    Copy::fill_to_words((HeapWord*)ptr, word_size, metadata_deallocate);
-#endif
-#endif
-
   } else {
     MutexLocker ml(vsm()->lock());
 
-#ifdef DEALLOCATE_BLOCKS
+    if (word_size < TreeChunk<Metablock, FreeList>::min_size()) {
+      // Dark matter.  Too small for dictionary.
+#ifdef ASSERT
+      Copy::fill_to_words((HeapWord*)ptr, word_size, 0xf5f5f5f5);
+#endif
+      return;
+    }
     if (is_class) {
-      class_vsm()->deallocate(ptr);
+      class_vsm()->deallocate(ptr, word_size);
     } else {
-      vsm()->deallocate(ptr);
+      vsm()->deallocate(ptr, word_size);
     }
-#else
-#ifdef ASSERT
-    Copy::fill_to_words((HeapWord*)ptr, word_size, metadata_deallocate);
-#endif
-#endif
   }
 }
 
-MetaWord* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
+Metablock* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
                               bool read_only, MetadataType mdtype, TRAPS) {
   if (HAS_PENDING_EXCEPTION) {
     assert(false, "Should not allocate with exception pending");
@@ -2943,7 +2681,7 @@ MetaWord* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
     if (result == NULL) {
       report_out_of_shared_space(read_only ? SharedReadOnly : SharedReadWrite);
     }
-    return result;
+    return Metablock::initialize(result, word_size);
   }
 
   result = loader_data->metaspace_non_null()->allocate(word_size, mdtype);
@@ -2951,7 +2689,7 @@ MetaWord* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
   if (result == NULL) {
     // Try to clean out some memory and retry.
     result =
-    Universe::heap()->collector_policy()->satisfy_failed_metadata_allocation(
+      Universe::heap()->collector_policy()->satisfy_failed_metadata_allocation(
         loader_data, word_size, mdtype);
 
     // If result is still null, we are out of memory.
@@ -2967,7 +2705,7 @@ MetaWord* Metaspace::allocate(ClassLoaderData* loader_data, size_t word_size,
       THROW_OOP_0(Universe::out_of_memory_error_perm_gen());
     }
   }
-  return result;
+  return Metablock::initialize(result, word_size);
 }
 
 void Metaspace::print_on(outputStream* out) const {
diff --git a/src/share/vm/memory/metaspace.hpp b/src/share/vm/memory/metaspace.hpp
index 169a3b71d..c42a979cb 100644
--- a/src/share/vm/memory/metaspace.hpp
+++ b/src/share/vm/memory/metaspace.hpp
@@ -57,12 +57,10 @@
 //
 
 class ClassLoaderData;
+class Metablock;
 class MetaWord;
 class Mutex;
 class outputStream;
-class FreeChunk;
-template <class Chunk_t> class FreeList;
-template <class Chunk_t> class BinaryTreeDictionary;
 class SpaceManager;
 
 // Metaspaces each have a  SpaceManager and allocations
@@ -128,7 +126,7 @@ class Metaspace : public CHeapObj<mtClass> {
   size_t capacity_words(MetadataType mdtype) const;
   size_t waste_words(MetadataType mdtype) const;
 
-  static MetaWord* allocate(ClassLoaderData* loader_data, size_t size,
+  static Metablock* allocate(ClassLoaderData* loader_data, size_t size,
                             bool read_only, MetadataType mdtype, TRAPS);
   void deallocate(MetaWord* ptr, size_t byte_size, bool is_class);
 
diff --git a/src/share/vm/memory/metaspaceShared.cpp b/src/share/vm/memory/metaspaceShared.cpp
index f69e7586e..dde26b5f8 100644
--- a/src/share/vm/memory/metaspaceShared.cpp
+++ b/src/share/vm/memory/metaspaceShared.cpp
@@ -663,8 +663,8 @@ bool MetaspaceShared::is_in_shared_space(const void* p) {
   if (_ro_base == NULL || _rw_base == NULL) {
     return false;
   } else {
-    return ((p > _ro_base && p < (_ro_base + SharedReadOnlySize)) ||
-            (p > _rw_base && p < (_rw_base + SharedReadWriteSize)));
+    return ((p >= _ro_base && p < (_ro_base + SharedReadOnlySize)) ||
+            (p >= _rw_base && p < (_rw_base + SharedReadWriteSize)));
   }
 }
 
@@ -693,14 +693,6 @@ bool MetaspaceShared::map_shared_spaces(FileMapInfo* mapinfo) {
   ReservedSpace shared_rs = mapinfo->reserve_shared_memory();
   if (!shared_rs.is_reserved()) return false;
 
-  // Split reserved memory into pieces (windows needs this)
-  ReservedSpace ro_rs   = shared_rs.first_part(SharedReadOnlySize);
-  ReservedSpace tmp_rs1 = shared_rs.last_part(SharedReadOnlySize);
-  ReservedSpace rw_rs   = tmp_rs1.first_part(SharedReadWriteSize);
-  ReservedSpace tmp_rs2 = tmp_rs1.last_part(SharedReadWriteSize);
-  ReservedSpace md_rs   = tmp_rs2.first_part(SharedMiscDataSize);
-  ReservedSpace mc_rs   = tmp_rs2.last_part(SharedMiscDataSize);
-
   // Map each shared region
   if ((_ro_base = mapinfo->map_region(ro)) != NULL &&
       (_rw_base = mapinfo->map_region(rw)) != NULL &&
diff --git a/src/share/vm/memory/resourceArea.hpp b/src/share/vm/memory/resourceArea.hpp
index f42c0f624..f1418ce19 100644
--- a/src/share/vm/memory/resourceArea.hpp
+++ b/src/share/vm/memory/resourceArea.hpp
@@ -127,15 +127,21 @@ protected:
   void reset_to_mark() {
     if (UseMallocOnly) free_malloced_objects();
 
-    if( _chunk->next() )        // Delete later chunks
+    if( _chunk->next() ) {       // Delete later chunks
+      // reset arena size before delete chunks. Otherwise, the total
+      // arena size could exceed total chunk size
+      assert(_area->size_in_bytes() > size_in_bytes(), "Sanity check");
+      _area->set_size_in_bytes(size_in_bytes());
       _chunk->next_chop();
+    } else {
+      assert(_area->size_in_bytes() == size_in_bytes(), "Sanity check");
+    }
     _area->_chunk = _chunk;     // Roll back arena to saved chunk
     _area->_hwm = _hwm;
     _area->_max = _max;
 
     // clear out this chunk (to detect allocation bugs)
     if (ZapResourceArea) memset(_hwm, badResourceValue, _max - _hwm);
-    _area->set_size_in_bytes(size_in_bytes());
   }
 
   ~ResourceMark() {
@@ -219,15 +225,21 @@ protected:
   void reset_to_mark() {
     if (UseMallocOnly) free_malloced_objects();
 
-    if( _chunk->next() )        // Delete later chunks
+    if( _chunk->next() ) {        // Delete later chunks
+      // reset arena size before delete chunks. Otherwise, the total
+      // arena size could exceed total chunk size
+      assert(_area->size_in_bytes() > size_in_bytes(), "Sanity check");
+      _area->set_size_in_bytes(size_in_bytes());
       _chunk->next_chop();
+    } else {
+      assert(_area->size_in_bytes() == size_in_bytes(), "Sanity check");
+    }
     _area->_chunk = _chunk;     // Roll back arena to saved chunk
     _area->_hwm = _hwm;
     _area->_max = _max;
 
     // clear out this chunk (to detect allocation bugs)
     if (ZapResourceArea) memset(_hwm, badResourceValue, _max - _hwm);
-    _area->set_size_in_bytes(size_in_bytes());
   }
 
   ~DeoptResourceMark() {
diff --git a/src/share/vm/oops/method.cpp b/src/share/vm/oops/method.cpp
index 5a1032f77..9849829ea 100644
--- a/src/share/vm/oops/method.cpp
+++ b/src/share/vm/oops/method.cpp
@@ -1155,8 +1155,12 @@ methodHandle Method::clone_with_new_data(methodHandle m, u_char* new_code, int n
 vmSymbols::SID Method::klass_id_for_intrinsics(Klass* holder) {
   // if loader is not the default loader (i.e., != NULL), we can't know the intrinsics
   // because we are not loading from core libraries
-  if (InstanceKlass::cast(holder)->class_loader() != NULL)
+  // exception: the AES intrinsics come from lib/ext/sunjce_provider.jar
+  // which does not use the class default class loader so we check for its loader here
+  if ((InstanceKlass::cast(holder)->class_loader() != NULL) &&
+       InstanceKlass::cast(holder)->class_loader()->klass()->name() != vmSymbols::sun_misc_Launcher_ExtClassLoader()) {
     return vmSymbols::NO_SID;   // regardless of name, no intrinsics here
+  }
 
   // see if the klass name is well-known:
   Symbol* klass_name = InstanceKlass::cast(holder)->name();
diff --git a/src/share/vm/opto/c2_globals.hpp b/src/share/vm/opto/c2_globals.hpp
index 80996a5d7..4fdebf526 100644
--- a/src/share/vm/opto/c2_globals.hpp
+++ b/src/share/vm/opto/c2_globals.hpp
@@ -439,6 +439,9 @@
   product(bool, DoEscapeAnalysis, true,                                     \
           "Perform escape analysis")                                        \
                                                                             \
+  develop(bool, ExitEscapeAnalysisOnTimeout, true,                          \
+          "Exit or throw assert in EA when it reaches time limit")          \
+                                                                            \
   notproduct(bool, PrintEscapeAnalysis, false,                              \
           "Print the results of escape analysis")                           \
                                                                             \
diff --git a/src/share/vm/opto/callGenerator.cpp b/src/share/vm/opto/callGenerator.cpp
index 547096b3d..93f2b859b 100644
--- a/src/share/vm/opto/callGenerator.cpp
+++ b/src/share/vm/opto/callGenerator.cpp
@@ -670,6 +670,129 @@ CallGenerator* CallGenerator::for_method_handle_inline(JVMState* jvms, ciMethod*
 }
 
 
+//------------------------PredictedIntrinsicGenerator------------------------------
+// Internal class which handles all predicted Intrinsic calls.
+class PredictedIntrinsicGenerator : public CallGenerator {
+  CallGenerator* _intrinsic;
+  CallGenerator* _cg;
+
+public:
+  PredictedIntrinsicGenerator(CallGenerator* intrinsic,
+                              CallGenerator* cg)
+    : CallGenerator(cg->method())
+  {
+    _intrinsic = intrinsic;
+    _cg        = cg;
+  }
+
+  virtual bool      is_virtual()   const    { return true; }
+  virtual bool      is_inlined()   const    { return true; }
+  virtual bool      is_intrinsic() const    { return true; }
+
+  virtual JVMState* generate(JVMState* jvms);
+};
+
+
+CallGenerator* CallGenerator::for_predicted_intrinsic(CallGenerator* intrinsic,
+                                                      CallGenerator* cg) {
+  return new PredictedIntrinsicGenerator(intrinsic, cg);
+}
+
+
+JVMState* PredictedIntrinsicGenerator::generate(JVMState* jvms) {
+  GraphKit kit(jvms);
+  PhaseGVN& gvn = kit.gvn();
+
+  CompileLog* log = kit.C->log();
+  if (log != NULL) {
+    log->elem("predicted_intrinsic bci='%d' method='%d'",
+              jvms->bci(), log->identify(method()));
+  }
+
+  Node* slow_ctl = _intrinsic->generate_predicate(kit.sync_jvms());
+  if (kit.failing())
+    return NULL;  // might happen because of NodeCountInliningCutoff
+
+  SafePointNode* slow_map = NULL;
+  JVMState* slow_jvms;
+  if (slow_ctl != NULL) {
+    PreserveJVMState pjvms(&kit);
+    kit.set_control(slow_ctl);
+    if (!kit.stopped()) {
+      slow_jvms = _cg->generate(kit.sync_jvms());
+      if (kit.failing())
+        return NULL;  // might happen because of NodeCountInliningCutoff
+      assert(slow_jvms != NULL, "must be");
+      kit.add_exception_states_from(slow_jvms);
+      kit.set_map(slow_jvms->map());
+      if (!kit.stopped())
+        slow_map = kit.stop();
+    }
+  }
+
+  if (kit.stopped()) {
+    // Predicate is always false.
+    kit.set_jvms(slow_jvms);
+    return kit.transfer_exceptions_into_jvms();
+  }
+
+  // Generate intrinsic code:
+  JVMState* new_jvms = _intrinsic->generate(kit.sync_jvms());
+  if (new_jvms == NULL) {
+    // Intrinsic failed, so use slow code or make a direct call.
+    if (slow_map == NULL) {
+      CallGenerator* cg = CallGenerator::for_direct_call(method());
+      new_jvms = cg->generate(kit.sync_jvms());
+    } else {
+      kit.set_jvms(slow_jvms);
+      return kit.transfer_exceptions_into_jvms();
+    }
+  }
+  kit.add_exception_states_from(new_jvms);
+  kit.set_jvms(new_jvms);
+
+  // Need to merge slow and fast?
+  if (slow_map == NULL) {
+    // The fast path is the only path remaining.
+    return kit.transfer_exceptions_into_jvms();
+  }
+
+  if (kit.stopped()) {
+    // Intrinsic method threw an exception, so it's just the slow path after all.
+    kit.set_jvms(slow_jvms);
+    return kit.transfer_exceptions_into_jvms();
+  }
+
+  // Finish the diamond.
+  kit.C->set_has_split_ifs(true); // Has chance for split-if optimization
+  RegionNode* region = new (kit.C) RegionNode(3);
+  region->init_req(1, kit.control());
+  region->init_req(2, slow_map->control());
+  kit.set_control(gvn.transform(region));
+  Node* iophi = PhiNode::make(region, kit.i_o(), Type::ABIO);
+  iophi->set_req(2, slow_map->i_o());
+  kit.set_i_o(gvn.transform(iophi));
+  kit.merge_memory(slow_map->merged_memory(), region, 2);
+  uint tos = kit.jvms()->stkoff() + kit.sp();
+  uint limit = slow_map->req();
+  for (uint i = TypeFunc::Parms; i < limit; i++) {
+    // Skip unused stack slots; fast forward to monoff();
+    if (i == tos) {
+      i = kit.jvms()->monoff();
+      if( i >= limit ) break;
+    }
+    Node* m = kit.map()->in(i);
+    Node* n = slow_map->in(i);
+    if (m != n) {
+      const Type* t = gvn.type(m)->meet(gvn.type(n));
+      Node* phi = PhiNode::make(region, m, t);
+      phi->set_req(2, n);
+      kit.map()->set_req(i, gvn.transform(phi));
+    }
+  }
+  return kit.transfer_exceptions_into_jvms();
+}
+
 //-------------------------UncommonTrapCallGenerator-----------------------------
 // Internal class which handles all out-of-line calls checking receiver type.
 class UncommonTrapCallGenerator : public CallGenerator {
diff --git a/src/share/vm/opto/callGenerator.hpp b/src/share/vm/opto/callGenerator.hpp
index 3cfd39df6..ae59173bf 100644
--- a/src/share/vm/opto/callGenerator.hpp
+++ b/src/share/vm/opto/callGenerator.hpp
@@ -143,6 +143,9 @@ class CallGenerator : public ResourceObj {
   // Registry for intrinsics:
   static CallGenerator* for_intrinsic(ciMethod* m);
   static void register_intrinsic(ciMethod* m, CallGenerator* cg);
+  static CallGenerator* for_predicted_intrinsic(CallGenerator* intrinsic,
+                                                CallGenerator* cg);
+  virtual Node* generate_predicate(JVMState* jvms) { return NULL; };
 
   static void print_inlining(ciMethod* callee, int inline_level, int bci, const char* msg) {
     if (PrintInlining)
diff --git a/src/share/vm/opto/compile.cpp b/src/share/vm/opto/compile.cpp
index d870872a9..c4da70ee3 100644
--- a/src/share/vm/opto/compile.cpp
+++ b/src/share/vm/opto/compile.cpp
@@ -3047,9 +3047,9 @@ bool Compile::Constant::operator==(const Constant& other) {
   case T_LONG:
   case T_DOUBLE:  return (_v._value.j == other._v._value.j);
   case T_OBJECT:
-  case T_METADATA: return (_v._metadata == other._v._metadata);
   case T_ADDRESS: return (_v._value.l == other._v._value.l);
   case T_VOID:    return (_v._value.l == other._v._value.l);  // jump-table entries
+  case T_METADATA: return (_v._metadata == other._v._metadata);
   default: ShouldNotReachHere();
   }
   return false;
diff --git a/src/share/vm/opto/compile.hpp b/src/share/vm/opto/compile.hpp
index 6bcf3d3da..44bf27768 100644
--- a/src/share/vm/opto/compile.hpp
+++ b/src/share/vm/opto/compile.hpp
@@ -149,7 +149,7 @@ class Compile : public Phase {
   private:
     BasicType _type;
     union {
-    jvalue    _value;
+      jvalue    _value;
       Metadata* _metadata;
     } _v;
     int       _offset;         // offset of this constant (in bytes) relative to the constant table base.
diff --git a/src/share/vm/opto/doCall.cpp b/src/share/vm/opto/doCall.cpp
index 30a01f34b..95d148841 100644
--- a/src/share/vm/opto/doCall.cpp
+++ b/src/share/vm/opto/doCall.cpp
@@ -107,7 +107,17 @@ CallGenerator* Compile::call_generator(ciMethod* callee, int vtable_index, bool
   // intrinsics handle strict f.p. correctly.
   if (allow_inline && allow_intrinsics) {
     CallGenerator* cg = find_intrinsic(callee, call_is_virtual);
-    if (cg != NULL)  return cg;
+    if (cg != NULL) {
+      if (cg->is_predicted()) {
+        // Code without intrinsic but, hopefully, inlined.
+        CallGenerator* inline_cg = this->call_generator(callee,
+              vtable_index, call_is_virtual, jvms, allow_inline, prof_factor, false);
+        if (inline_cg != NULL) {
+          cg = CallGenerator::for_predicted_intrinsic(cg, inline_cg);
+        }
+      }
+      return cg;
+    }
   }
 
   // Do method handle calls.
diff --git a/src/share/vm/opto/escape.cpp b/src/share/vm/opto/escape.cpp
index 9fd318050..a5aa47119 100644
--- a/src/share/vm/opto/escape.cpp
+++ b/src/share/vm/opto/escape.cpp
@@ -893,12 +893,16 @@ void ConnectionGraph::process_call_arguments(CallNode *call) {
                                        arg_has_oops && (i > TypeFunc::Parms);
 #ifdef ASSERT
           if (!(is_arraycopy ||
-                call->as_CallLeaf()->_name != NULL &&
-                (strcmp(call->as_CallLeaf()->_name, "g1_wb_pre")  == 0 ||
-                 strcmp(call->as_CallLeaf()->_name, "g1_wb_post") == 0 ))
-          ) {
+                (call->as_CallLeaf()->_name != NULL &&
+                 (strcmp(call->as_CallLeaf()->_name, "g1_wb_pre")  == 0 ||
+                  strcmp(call->as_CallLeaf()->_name, "g1_wb_post") == 0 ||
+                  strcmp(call->as_CallLeaf()->_name, "aescrypt_encryptBlock") == 0 ||
+                  strcmp(call->as_CallLeaf()->_name, "aescrypt_decryptBlock") == 0 ||
+                  strcmp(call->as_CallLeaf()->_name, "cipherBlockChaining_encryptAESCrypt") == 0 ||
+                  strcmp(call->as_CallLeaf()->_name, "cipherBlockChaining_decryptAESCrypt") == 0)
+                  ))) {
             call->dump();
-            assert(false, "EA: unexpected CallLeaf");
+            fatal(err_msg_res("EA unexpected CallLeaf %s", call->as_CallLeaf()->_name));
           }
 #endif
           // Always process arraycopy's destination object since
@@ -1080,7 +1084,7 @@ bool ConnectionGraph::complete_connection_graph(
       C->log()->text("%s", (iterations >= CG_BUILD_ITER_LIMIT) ? "iterations" : "time");
       C->log()->end_elem(" limit'");
     }
-    assert(false, err_msg_res("infinite EA connection graph build (%f sec, %d iterations) with %d nodes and worklist size %d",
+    assert(ExitEscapeAnalysisOnTimeout, err_msg_res("infinite EA connection graph build (%f sec, %d iterations) with %d nodes and worklist size %d",
            time.seconds(), iterations, nodes_size(), ptnodes_worklist.length()));
     // Possible infinite build_connection_graph loop,
     // bailout (no changes to ideal graph were made).
diff --git a/src/share/vm/opto/library_call.cpp b/src/share/vm/opto/library_call.cpp
index f34df79b4..6b90061ff 100644
--- a/src/share/vm/opto/library_call.cpp
+++ b/src/share/vm/opto/library_call.cpp
@@ -44,18 +44,22 @@ class LibraryIntrinsic : public InlineCallGenerator {
  public:
  private:
   bool             _is_virtual;
+  bool             _is_predicted;
   vmIntrinsics::ID _intrinsic_id;
 
  public:
-  LibraryIntrinsic(ciMethod* m, bool is_virtual, vmIntrinsics::ID id)
+  LibraryIntrinsic(ciMethod* m, bool is_virtual, bool is_predicted, vmIntrinsics::ID id)
     : InlineCallGenerator(m),
       _is_virtual(is_virtual),
+      _is_predicted(is_predicted),
       _intrinsic_id(id)
   {
   }
   virtual bool is_intrinsic() const { return true; }
   virtual bool is_virtual()   const { return _is_virtual; }
+  virtual bool is_predicted()   const { return _is_predicted; }
   virtual JVMState* generate(JVMState* jvms);
+  virtual Node* generate_predicate(JVMState* jvms);
   vmIntrinsics::ID intrinsic_id() const { return _intrinsic_id; }
 };
 
@@ -83,6 +87,7 @@ class LibraryCallKit : public GraphKit {
   int               arg_size()  const    { return callee()->arg_size(); }
 
   bool try_to_inline();
+  Node* try_to_predicate();
 
   // Helper functions to inline natives
   void push_result(RegionNode* region, PhiNode* value);
@@ -148,6 +153,7 @@ class LibraryCallKit : public GraphKit {
   CallJavaNode* generate_method_call_virtual(vmIntrinsics::ID method_id) {
     return generate_method_call(method_id, true, false);
   }
+  Node * load_field_from_object(Node * fromObj, const char * fieldName, const char * fieldTypeString, bool is_exact, bool is_static);
 
   Node* make_string_method_node(int opcode, Node* str1_start, Node* cnt1, Node* str2_start, Node* cnt2);
   Node* make_string_method_node(int opcode, Node* str1, Node* str2);
@@ -253,6 +259,10 @@ class LibraryCallKit : public GraphKit {
   bool inline_reverseBytes(vmIntrinsics::ID id);
 
   bool inline_reference_get();
+  bool inline_aescrypt_Block(vmIntrinsics::ID id);
+  bool inline_cipherBlockChaining_AESCrypt(vmIntrinsics::ID id);
+  Node* inline_cipherBlockChaining_AESCrypt_predicate(bool decrypting);
+  Node* get_key_start_from_aescrypt_object(Node* aescrypt_object);
 };
 
 
@@ -306,6 +316,8 @@ CallGenerator* Compile::make_vm_intrinsic(ciMethod* m, bool is_virtual) {
     }
   }
 
+  bool is_predicted = false;
+
   switch (id) {
   case vmIntrinsics::_compareTo:
     if (!SpecialStringCompareTo)  return NULL;
@@ -413,6 +425,18 @@ CallGenerator* Compile::make_vm_intrinsic(ciMethod* m, bool is_virtual) {
     break;
 #endif
 
+  case vmIntrinsics::_aescrypt_encryptBlock:
+  case vmIntrinsics::_aescrypt_decryptBlock:
+    if (!UseAESIntrinsics) return NULL;
+    break;
+
+  case vmIntrinsics::_cipherBlockChaining_encryptAESCrypt:
+  case vmIntrinsics::_cipherBlockChaining_decryptAESCrypt:
+    if (!UseAESIntrinsics) return NULL;
+    // these two require the predicated logic
+    is_predicted = true;
+    break;
+
  default:
     assert(id <= vmIntrinsics::LAST_COMPILER_INLINE, "caller responsibility");
     assert(id != vmIntrinsics::_Object_init && id != vmIntrinsics::_invoke, "enum out of order?");
@@ -444,7 +468,7 @@ CallGenerator* Compile::make_vm_intrinsic(ciMethod* m, bool is_virtual) {
     if (!InlineUnsafeOps)  return NULL;
   }
 
-  return new LibraryIntrinsic(m, is_virtual, (vmIntrinsics::ID) id);
+  return new LibraryIntrinsic(m, is_virtual, is_predicted, (vmIntrinsics::ID) id);
 }
 
 //----------------------register_library_intrinsics-----------------------
@@ -496,6 +520,47 @@ JVMState* LibraryIntrinsic::generate(JVMState* jvms) {
   return NULL;
 }
 
+Node* LibraryIntrinsic::generate_predicate(JVMState* jvms) {
+  LibraryCallKit kit(jvms, this);
+  Compile* C = kit.C;
+  int nodes = C->unique();
+#ifndef PRODUCT
+  assert(is_predicted(), "sanity");
+  if ((PrintIntrinsics || PrintInlining NOT_PRODUCT( || PrintOptoInlining) ) && Verbose) {
+    char buf[1000];
+    const char* str = vmIntrinsics::short_name_as_C_string(intrinsic_id(), buf, sizeof(buf));
+    tty->print_cr("Predicate for intrinsic %s", str);
+  }
+#endif
+
+  Node* slow_ctl = kit.try_to_predicate();
+  if (!kit.failing()) {
+    if (C->log()) {
+      C->log()->elem("predicate_intrinsic id='%s'%s nodes='%d'",
+                     vmIntrinsics::name_at(intrinsic_id()),
+                     (is_virtual() ? " virtual='1'" : ""),
+                     C->unique() - nodes);
+    }
+    return slow_ctl; // Could be NULL if the check folds.
+  }
+
+  // The intrinsic bailed out
+  if (PrintIntrinsics || PrintInlining NOT_PRODUCT( || PrintOptoInlining) ) {
+    if (jvms->has_method()) {
+      // Not a root compile.
+      const char* msg = "failed to generate predicate for intrinsic";
+      CompileTask::print_inlining(kit.callee(), jvms->depth() - 1, kit.bci(), msg);
+    } else {
+      // Root compile
+      tty->print("Did not generate predicate for intrinsic %s%s at bci:%d in",
+               vmIntrinsics::name_at(intrinsic_id()),
+               (is_virtual() ? " (virtual)" : ""), kit.bci());
+    }
+  }
+  C->gather_intrinsic_statistics(intrinsic_id(), is_virtual(), Compile::_intrinsic_failed);
+  return NULL;
+}
+
 bool LibraryCallKit::try_to_inline() {
   // Handle symbolic names for otherwise undistinguished boolean switches:
   const bool is_store       = true;
@@ -767,6 +832,14 @@ bool LibraryCallKit::try_to_inline() {
   case vmIntrinsics::_Reference_get:
     return inline_reference_get();
 
+  case vmIntrinsics::_aescrypt_encryptBlock:
+  case vmIntrinsics::_aescrypt_decryptBlock:
+    return inline_aescrypt_Block(intrinsic_id());
+
+  case vmIntrinsics::_cipherBlockChaining_encryptAESCrypt:
+  case vmIntrinsics::_cipherBlockChaining_decryptAESCrypt:
+    return inline_cipherBlockChaining_AESCrypt(intrinsic_id());
+
   default:
     // If you get here, it may be that someone has added a new intrinsic
     // to the list in vmSymbols.hpp without implementing it here.
@@ -780,6 +853,36 @@ bool LibraryCallKit::try_to_inline() {
   }
 }
 
+Node* LibraryCallKit::try_to_predicate() {
+  if (!jvms()->has_method()) {
+    // Root JVMState has a null method.
+    assert(map()->memory()->Opcode() == Op_Parm, "");
+    // Insert the memory aliasing node
+    set_all_memory(reset_memory());
+  }
+  assert(merged_memory(), "");
+
+  switch (intrinsic_id()) {
+  case vmIntrinsics::_cipherBlockChaining_encryptAESCrypt:
+    return inline_cipherBlockChaining_AESCrypt_predicate(false);
+  case vmIntrinsics::_cipherBlockChaining_decryptAESCrypt:
+    return inline_cipherBlockChaining_AESCrypt_predicate(true);
+
+  default:
+    // If you get here, it may be that someone has added a new intrinsic
+    // to the list in vmSymbols.hpp without implementing it here.
+#ifndef PRODUCT
+    if ((PrintMiscellaneous && (Verbose || WizardMode)) || PrintOpto) {
+      tty->print_cr("*** Warning: Unimplemented predicate for intrinsic %s(%d)",
+                    vmIntrinsics::name_at(intrinsic_id()), intrinsic_id());
+    }
+#endif
+    Node* slow_ctl = control();
+    set_control(top()); // No fast path instrinsic
+    return slow_ctl;
+  }
+}
+
 //------------------------------push_result------------------------------
 // Helper function for finishing intrinsics.
 void LibraryCallKit::push_result(RegionNode* region, PhiNode* value) {
@@ -3830,7 +3933,7 @@ Node* LibraryCallKit::generate_virtual_guard(Node* obj_klass,
                      vtable_index*vtableEntry::size()) * wordSize +
                      vtableEntry::method_offset_in_bytes();
   Node* entry_addr  = basic_plus_adr(obj_klass, entry_offset);
-  Node* target_call = make_load(NULL, entry_addr, TypeInstPtr::NOTNULL, T_OBJECT);
+  Node* target_call = make_load(NULL, entry_addr, TypePtr::NOTNULL, T_ADDRESS);
 
   // Compare the target method with the expected method (e.g., Object.hashCode).
   const TypePtr* native_call_addr = TypeMetadataPtr::make(method);
@@ -5613,3 +5716,265 @@ bool LibraryCallKit::inline_reference_get() {
   push(result);
   return true;
 }
+
+
+Node * LibraryCallKit::load_field_from_object(Node * fromObj, const char * fieldName, const char * fieldTypeString,
+                                              bool is_exact=true, bool is_static=false) {
+
+  const TypeInstPtr* tinst = _gvn.type(fromObj)->isa_instptr();
+  assert(tinst != NULL, "obj is null");
+  assert(tinst->klass()->is_loaded(), "obj is not loaded");
+  assert(!is_exact || tinst->klass_is_exact(), "klass not exact");
+
+  ciField* field = tinst->klass()->as_instance_klass()->get_field_by_name(ciSymbol::make(fieldName),
+                                                                          ciSymbol::make(fieldTypeString),
+                                                                          is_static);
+  if (field == NULL) return (Node *) NULL;
+  assert (field != NULL, "undefined field");
+
+  // Next code  copied from Parse::do_get_xxx():
+
+  // Compute address and memory type.
+  int offset  = field->offset_in_bytes();
+  bool is_vol = field->is_volatile();
+  ciType* field_klass = field->type();
+  assert(field_klass->is_loaded(), "should be loaded");
+  const TypePtr* adr_type = C->alias_type(field)->adr_type();
+  Node *adr = basic_plus_adr(fromObj, fromObj, offset);
+  BasicType bt = field->layout_type();
+
+  // Build the resultant type of the load
+  const Type *type = TypeOopPtr::make_from_klass(field_klass->as_klass());
+
+  // Build the load.
+  Node* loadedField = make_load(NULL, adr, type, bt, adr_type, is_vol);
+  return loadedField;
+}
+
+
+//------------------------------inline_aescrypt_Block-----------------------
+bool LibraryCallKit::inline_aescrypt_Block(vmIntrinsics::ID id) {
+  address stubAddr;
+  const char *stubName;
+  assert(UseAES, "need AES instruction support");
+
+  switch(id) {
+  case vmIntrinsics::_aescrypt_encryptBlock:
+    stubAddr = StubRoutines::aescrypt_encryptBlock();
+    stubName = "aescrypt_encryptBlock";
+    break;
+  case vmIntrinsics::_aescrypt_decryptBlock:
+    stubAddr = StubRoutines::aescrypt_decryptBlock();
+    stubName = "aescrypt_decryptBlock";
+    break;
+  }
+  if (stubAddr == NULL) return false;
+
+  // Restore the stack and pop off the arguments.
+  int nargs = 5;  // this + 2 oop/offset combos
+  assert(callee()->signature()->size() == nargs-1, "encryptBlock has 4 arguments");
+
+  Node *aescrypt_object  = argument(0);
+  Node *src         = argument(1);
+  Node *src_offset  = argument(2);
+  Node *dest        = argument(3);
+  Node *dest_offset = argument(4);
+
+  // (1) src and dest are arrays.
+  const Type* src_type = src->Value(&_gvn);
+  const Type* dest_type = dest->Value(&_gvn);
+  const TypeAryPtr* top_src = src_type->isa_aryptr();
+  const TypeAryPtr* top_dest = dest_type->isa_aryptr();
+  assert (top_src  != NULL && top_src->klass()  != NULL &&  top_dest != NULL && top_dest->klass() != NULL, "args are strange");
+
+  // for the quick and dirty code we will skip all the checks.
+  // we are just trying to get the call to be generated.
+  Node* src_start  = src;
+  Node* dest_start = dest;
+  if (src_offset != NULL || dest_offset != NULL) {
+    assert(src_offset != NULL && dest_offset != NULL, "");
+    src_start  = array_element_address(src,  src_offset,  T_BYTE);
+    dest_start = array_element_address(dest, dest_offset, T_BYTE);
+  }
+
+  // now need to get the start of its expanded key array
+  // this requires a newer class file that has this array as littleEndian ints, otherwise we revert to java
+  Node* k_start = get_key_start_from_aescrypt_object(aescrypt_object);
+  if (k_start == NULL) return false;
+
+  // Call the stub.
+  make_runtime_call(RC_LEAF|RC_NO_FP, OptoRuntime::aescrypt_block_Type(),
+                    stubAddr, stubName, TypePtr::BOTTOM,
+                    src_start, dest_start, k_start);
+
+  return true;
+}
+
+//------------------------------inline_cipherBlockChaining_AESCrypt-----------------------
+bool LibraryCallKit::inline_cipherBlockChaining_AESCrypt(vmIntrinsics::ID id) {
+  address stubAddr;
+  const char *stubName;
+
+  assert(UseAES, "need AES instruction support");
+
+  switch(id) {
+  case vmIntrinsics::_cipherBlockChaining_encryptAESCrypt:
+    stubAddr = StubRoutines::cipherBlockChaining_encryptAESCrypt();
+    stubName = "cipherBlockChaining_encryptAESCrypt";
+    break;
+  case vmIntrinsics::_cipherBlockChaining_decryptAESCrypt:
+    stubAddr = StubRoutines::cipherBlockChaining_decryptAESCrypt();
+    stubName = "cipherBlockChaining_decryptAESCrypt";
+    break;
+  }
+  if (stubAddr == NULL) return false;
+
+
+  // Restore the stack and pop off the arguments.
+  int nargs = 6;  // this + oop/offset + len + oop/offset
+  assert(callee()->signature()->size() == nargs-1, "wrong number of arguments");
+  Node *cipherBlockChaining_object  = argument(0);
+  Node *src         = argument(1);
+  Node *src_offset  = argument(2);
+  Node *len         = argument(3);
+  Node *dest        = argument(4);
+  Node *dest_offset = argument(5);
+
+  // (1) src and dest are arrays.
+  const Type* src_type = src->Value(&_gvn);
+  const Type* dest_type = dest->Value(&_gvn);
+  const TypeAryPtr* top_src = src_type->isa_aryptr();
+  const TypeAryPtr* top_dest = dest_type->isa_aryptr();
+  assert (top_src  != NULL && top_src->klass()  != NULL
+          &&  top_dest != NULL && top_dest->klass() != NULL, "args are strange");
+
+  // checks are the responsibility of the caller
+  Node* src_start  = src;
+  Node* dest_start = dest;
+  if (src_offset != NULL || dest_offset != NULL) {
+    assert(src_offset != NULL && dest_offset != NULL, "");
+    src_start  = array_element_address(src,  src_offset,  T_BYTE);
+    dest_start = array_element_address(dest, dest_offset, T_BYTE);
+  }
+
+  // if we are in this set of code, we "know" the embeddedCipher is an AESCrypt object
+  // (because of the predicated logic executed earlier).
+  // so we cast it here safely.
+  // this requires a newer class file that has this array as littleEndian ints, otherwise we revert to java
+
+  Node* embeddedCipherObj = load_field_from_object(cipherBlockChaining_object, "embeddedCipher", "Lcom/sun/crypto/provider/SymmetricCipher;", /*is_exact*/ false);
+  if (embeddedCipherObj == NULL) return false;
+
+  // cast it to what we know it will be at runtime
+  const TypeInstPtr* tinst = _gvn.type(cipherBlockChaining_object)->isa_instptr();
+  assert(tinst != NULL, "CBC obj is null");
+  assert(tinst->klass()->is_loaded(), "CBC obj is not loaded");
+  ciKlass* klass_AESCrypt = tinst->klass()->as_instance_klass()->find_klass(ciSymbol::make("com/sun/crypto/provider/AESCrypt"));
+  if (!klass_AESCrypt->is_loaded()) return false;
+
+  ciInstanceKlass* instklass_AESCrypt = klass_AESCrypt->as_instance_klass();
+  const TypeKlassPtr* aklass = TypeKlassPtr::make(instklass_AESCrypt);
+  const TypeOopPtr* xtype = aklass->as_instance_type();
+  Node* aescrypt_object = new(C) CheckCastPPNode(control(), embeddedCipherObj, xtype);
+  aescrypt_object = _gvn.transform(aescrypt_object);
+
+  // we need to get the start of the aescrypt_object's expanded key array
+  Node* k_start = get_key_start_from_aescrypt_object(aescrypt_object);
+  if (k_start == NULL) return false;
+
+  // similarly, get the start address of the r vector
+  Node* objRvec = load_field_from_object(cipherBlockChaining_object, "r", "[B", /*is_exact*/ false);
+  if (objRvec == NULL) return false;
+  Node* r_start = array_element_address(objRvec, intcon(0), T_BYTE);
+
+  // Call the stub, passing src_start, dest_start, k_start, r_start and src_len
+  make_runtime_call(RC_LEAF|RC_NO_FP,
+                    OptoRuntime::cipherBlockChaining_aescrypt_Type(),
+                    stubAddr, stubName, TypePtr::BOTTOM,
+                    src_start, dest_start, k_start, r_start, len);
+
+  // return is void so no result needs to be pushed
+
+  return true;
+}
+
+//------------------------------get_key_start_from_aescrypt_object-----------------------
+Node * LibraryCallKit::get_key_start_from_aescrypt_object(Node *aescrypt_object) {
+  Node* objAESCryptKey = load_field_from_object(aescrypt_object, "K", "[I", /*is_exact*/ false);
+  assert (objAESCryptKey != NULL, "wrong version of com.sun.crypto.provider.AESCrypt");
+  if (objAESCryptKey == NULL) return (Node *) NULL;
+
+  // now have the array, need to get the start address of the K array
+  Node* k_start = array_element_address(objAESCryptKey, intcon(0), T_INT);
+  return k_start;
+}
+
+//----------------------------inline_cipherBlockChaining_AESCrypt_predicate----------------------------
+// Return node representing slow path of predicate check.
+// the pseudo code we want to emulate with this predicate is:
+// for encryption:
+//    if (embeddedCipherObj instanceof AESCrypt) do_intrinsic, else do_javapath
+// for decryption:
+//    if ((embeddedCipherObj instanceof AESCrypt) && (cipher!=plain)) do_intrinsic, else do_javapath
+//    note cipher==plain is more conservative than the original java code but that's OK
+//
+Node* LibraryCallKit::inline_cipherBlockChaining_AESCrypt_predicate(bool decrypting) {
+  // First, check receiver for NULL since it is virtual method.
+  int nargs = arg_size();
+  Node* objCBC = argument(0);
+  _sp += nargs;
+  objCBC = do_null_check(objCBC, T_OBJECT);
+  _sp -= nargs;
+
+  if (stopped()) return NULL; // Always NULL
+
+  // Load embeddedCipher field of CipherBlockChaining object.
+  Node* embeddedCipherObj = load_field_from_object(objCBC, "embeddedCipher", "Lcom/sun/crypto/provider/SymmetricCipher;", /*is_exact*/ false);
+
+  // get AESCrypt klass for instanceOf check
+  // AESCrypt might not be loaded yet if some other SymmetricCipher got us to this compile point
+  // will have same classloader as CipherBlockChaining object
+  const TypeInstPtr* tinst = _gvn.type(objCBC)->isa_instptr();
+  assert(tinst != NULL, "CBCobj is null");
+  assert(tinst->klass()->is_loaded(), "CBCobj is not loaded");
+
+  // we want to do an instanceof comparison against the AESCrypt class
+  ciKlass* klass_AESCrypt = tinst->klass()->as_instance_klass()->find_klass(ciSymbol::make("com/sun/crypto/provider/AESCrypt"));
+  if (!klass_AESCrypt->is_loaded()) {
+    // if AESCrypt is not even loaded, we never take the intrinsic fast path
+    Node* ctrl = control();
+    set_control(top()); // no regular fast path
+    return ctrl;
+  }
+  ciInstanceKlass* instklass_AESCrypt = klass_AESCrypt->as_instance_klass();
+
+  _sp += nargs;          // gen_instanceof might do an uncommon trap
+  Node* instof = gen_instanceof(embeddedCipherObj, makecon(TypeKlassPtr::make(instklass_AESCrypt)));
+  _sp -= nargs;
+  Node* cmp_instof  = _gvn.transform(new (C) CmpINode(instof, intcon(1)));
+  Node* bool_instof  = _gvn.transform(new (C) BoolNode(cmp_instof, BoolTest::ne));
+
+  Node* instof_false = generate_guard(bool_instof, NULL, PROB_MIN);
+
+  // for encryption, we are done
+  if (!decrypting)
+    return instof_false;  // even if it is NULL
+
+  // for decryption, we need to add a further check to avoid
+  // taking the intrinsic path when cipher and plain are the same
+  // see the original java code for why.
+  RegionNode* region = new(C) RegionNode(3);
+  region->init_req(1, instof_false);
+  Node* src = argument(1);
+  Node *dest = argument(4);
+  Node* cmp_src_dest = _gvn.transform(new (C) CmpPNode(src, dest));
+  Node* bool_src_dest = _gvn.transform(new (C) BoolNode(cmp_src_dest, BoolTest::eq));
+  Node* src_dest_conjoint = generate_guard(bool_src_dest, NULL, PROB_MIN);
+  region->init_req(2, src_dest_conjoint);
+
+  record_for_igvn(region);
+  return _gvn.transform(region);
+
+}
+
+
diff --git a/src/share/vm/opto/mulnode.cpp b/src/share/vm/opto/mulnode.cpp
index 4572a265e..4047b933f 100644
--- a/src/share/vm/opto/mulnode.cpp
+++ b/src/share/vm/opto/mulnode.cpp
@@ -479,24 +479,27 @@ Node *AndINode::Ideal(PhaseGVN *phase, bool can_reshape) {
     return new (phase->C) AndINode(load,phase->intcon(mask&0xFFFF));
 
   // Masking bits off of a Short?  Loading a Character does some masking
-  if (lop == Op_LoadS && (mask & 0xFFFF0000) == 0 ) {
-    Node *ldus = new (phase->C) LoadUSNode(load->in(MemNode::Control),
-                                              load->in(MemNode::Memory),
-                                              load->in(MemNode::Address),
-                                              load->adr_type());
-    ldus = phase->transform(ldus);
-    return new (phase->C) AndINode(ldus, phase->intcon(mask & 0xFFFF));
-  }
+  if (can_reshape &&
+      load->outcnt() == 1 && load->unique_out() == this) {
+    if (lop == Op_LoadS && (mask & 0xFFFF0000) == 0 ) {
+      Node *ldus = new (phase->C) LoadUSNode(load->in(MemNode::Control),
+                                             load->in(MemNode::Memory),
+                                             load->in(MemNode::Address),
+                                             load->adr_type());
+      ldus = phase->transform(ldus);
+      return new (phase->C) AndINode(ldus, phase->intcon(mask & 0xFFFF));
+    }
 
-  // Masking sign bits off of a Byte?  Do an unsigned byte load plus
-  // an and.
-  if (lop == Op_LoadB && (mask & 0xFFFFFF00) == 0) {
-    Node* ldub = new (phase->C) LoadUBNode(load->in(MemNode::Control),
-                                              load->in(MemNode::Memory),
-                                              load->in(MemNode::Address),
-                                              load->adr_type());
-    ldub = phase->transform(ldub);
-    return new (phase->C) AndINode(ldub, phase->intcon(mask));
+    // Masking sign bits off of a Byte?  Do an unsigned byte load plus
+    // an and.
+    if (lop == Op_LoadB && (mask & 0xFFFFFF00) == 0) {
+      Node* ldub = new (phase->C) LoadUBNode(load->in(MemNode::Control),
+                                             load->in(MemNode::Memory),
+                                             load->in(MemNode::Address),
+                                             load->adr_type());
+      ldub = phase->transform(ldub);
+      return new (phase->C) AndINode(ldub, phase->intcon(mask));
+    }
   }
 
   // Masking off sign bits?  Dont make them!
@@ -923,7 +926,9 @@ Node *RShiftINode::Ideal(PhaseGVN *phase, bool can_reshape) {
       set_req(2, phase->intcon(0));
       return this;
     }
-    else if( ld->Opcode() == Op_LoadUS )
+    else if( can_reshape &&
+             ld->Opcode() == Op_LoadUS &&
+             ld->outcnt() == 1 && ld->unique_out() == shl)
       // Replace zero-extension-load with sign-extension-load
       return new (phase->C) LoadSNode( ld->in(MemNode::Control),
                                 ld->in(MemNode::Memory),
diff --git a/src/share/vm/opto/runtime.cpp b/src/share/vm/opto/runtime.cpp
index bb050533d..51987e25e 100644
--- a/src/share/vm/opto/runtime.cpp
+++ b/src/share/vm/opto/runtime.cpp
@@ -811,6 +811,48 @@ const TypeFunc* OptoRuntime::array_fill_Type() {
   return TypeFunc::make(domain, range);
 }
 
+// for aescrypt encrypt/decrypt operations, just three pointers returning void (length is constant)
+const TypeFunc* OptoRuntime::aescrypt_block_Type() {
+  // create input type (domain)
+  int num_args      = 3;
+  int argcnt = num_args;
+  const Type** fields = TypeTuple::fields(argcnt);
+  int argp = TypeFunc::Parms;
+  fields[argp++] = TypePtr::NOTNULL;    // src
+  fields[argp++] = TypePtr::NOTNULL;    // dest
+  fields[argp++] = TypePtr::NOTNULL;    // k array
+  assert(argp == TypeFunc::Parms+argcnt, "correct decoding");
+  const TypeTuple* domain = TypeTuple::make(TypeFunc::Parms+argcnt, fields);
+
+  // no result type needed
+  fields = TypeTuple::fields(1);
+  fields[TypeFunc::Parms+0] = NULL; // void
+  const TypeTuple* range = TypeTuple::make(TypeFunc::Parms, fields);
+  return TypeFunc::make(domain, range);
+}
+
+// for cipherBlockChaining calls of aescrypt encrypt/decrypt, four pointers and a length, returning void
+const TypeFunc* OptoRuntime::cipherBlockChaining_aescrypt_Type() {
+  // create input type (domain)
+  int num_args      = 5;
+  int argcnt = num_args;
+  const Type** fields = TypeTuple::fields(argcnt);
+  int argp = TypeFunc::Parms;
+  fields[argp++] = TypePtr::NOTNULL;    // src
+  fields[argp++] = TypePtr::NOTNULL;    // dest
+  fields[argp++] = TypePtr::NOTNULL;    // k array
+  fields[argp++] = TypePtr::NOTNULL;    // r array
+  fields[argp++] = TypeInt::INT;        // src len
+  assert(argp == TypeFunc::Parms+argcnt, "correct decoding");
+  const TypeTuple* domain = TypeTuple::make(TypeFunc::Parms+argcnt, fields);
+
+  // no result type needed
+  fields = TypeTuple::fields(1);
+  fields[TypeFunc::Parms+0] = NULL; // void
+  const TypeTuple* range = TypeTuple::make(TypeFunc::Parms, fields);
+  return TypeFunc::make(domain, range);
+}
+
 //------------- Interpreter state access for on stack replacement
 const TypeFunc* OptoRuntime::osr_end_Type() {
   // create input type (domain)
diff --git a/src/share/vm/opto/runtime.hpp b/src/share/vm/opto/runtime.hpp
index c70777267..13da255b7 100644
--- a/src/share/vm/opto/runtime.hpp
+++ b/src/share/vm/opto/runtime.hpp
@@ -280,6 +280,9 @@ private:
 
   static const TypeFunc* array_fill_Type();
 
+  static const TypeFunc* aescrypt_block_Type();
+  static const TypeFunc* cipherBlockChaining_aescrypt_Type();
+
   // leaf on stack replacement interpreter accessor types
   static const TypeFunc* osr_end_Type();
 
diff --git a/src/share/vm/opto/superword.cpp b/src/share/vm/opto/superword.cpp
index ffc5394bb..f53c64837 100644
--- a/src/share/vm/opto/superword.cpp
+++ b/src/share/vm/opto/superword.cpp
@@ -1776,16 +1776,15 @@ void SuperWord::compute_vector_element_type() {
     set_velt_type(n, container_type(n));
   }
 
-  // Propagate narrowed type backwards through operations
+  // Propagate integer narrowed type backwards through operations
   // that don't depend on higher order bits
   for (int i = _block.length() - 1; i >= 0; i--) {
     Node* n = _block.at(i);
     // Only integer types need be examined
-    const Type* vt = velt_type(n);
-    if (vt->basic_type() == T_INT) {
+    const Type* vtn = velt_type(n);
+    if (vtn->basic_type() == T_INT) {
       uint start, end;
       VectorNode::vector_operands(n, &start, &end);
-      const Type* vt = velt_type(n);
 
       for (uint j = start; j < end; j++) {
         Node* in  = n->in(j);
@@ -1801,6 +1800,24 @@ void SuperWord::compute_vector_element_type() {
             }
           }
           if (same_type) {
+            // For right shifts of small integer types (bool, byte, char, short)
+            // we need precise information about sign-ness. Only Load nodes have
+            // this information because Store nodes are the same for signed and
+            // unsigned values. And any arithmetic operation after a load may
+            // expand a value to signed Int so such right shifts can't be used
+            // because vector elements do not have upper bits of Int.
+            const Type* vt = vtn;
+            if (VectorNode::is_shift(in)) {
+              Node* load = in->in(1);
+              if (load->is_Load() && in_bb(load) && (velt_type(load)->basic_type() == T_INT)) {
+                vt = velt_type(load);
+              } else if (in->Opcode() != Op_LShiftI) {
+                // Widen type to Int to avoid creation of right shift vector
+                // (align + data_size(s1) check in stmts_can_pack() will fail).
+                // Note, left shifts work regardless type.
+                vt = TypeInt::INT;
+              }
+            }
             set_velt_type(in, vt);
           }
         }
@@ -1841,7 +1858,20 @@ int SuperWord::memory_alignment(MemNode* s, int iv_adjust) {
 // Smallest type containing range of values
 const Type* SuperWord::container_type(Node* n) {
   if (n->is_Mem()) {
-    return Type::get_const_basic_type(n->as_Mem()->memory_type());
+    BasicType bt = n->as_Mem()->memory_type();
+    if (n->is_Store() && (bt == T_CHAR)) {
+      // Use T_SHORT type instead of T_CHAR for stored values because any
+      // preceding arithmetic operation extends values to signed Int.
+      bt = T_SHORT;
+    }
+    if (n->Opcode() == Op_LoadUB) {
+      // Adjust type for unsigned byte loads, it is important for right shifts.
+      // T_BOOLEAN is used because there is no basic type representing type
+      // TypeInt::UBYTE. Use of T_BOOLEAN for vectors is fine because only
+      // size (one byte) and sign is important.
+      bt = T_BOOLEAN;
+    }
+    return Type::get_const_basic_type(bt);
   }
   const Type* t = _igvn.type(n);
   if (t->basic_type() == T_INT) {
diff --git a/src/share/vm/opto/type.cpp b/src/share/vm/opto/type.cpp
index f982799f6..1a8ee2597 100644
--- a/src/share/vm/opto/type.cpp
+++ b/src/share/vm/opto/type.cpp
@@ -61,7 +61,7 @@ Type::TypeInfo Type::_type_info[Type::lastype] = {
   { Bad,             T_ILLEGAL,    "tuple:",        false, Node::NotAMachineReg, relocInfo::none          },  // Tuple
   { Bad,             T_ARRAY,      "array:",        false, Node::NotAMachineReg, relocInfo::none          },  // Array
 
-#if defined(IA32) || defined(AMD64)
+#ifndef SPARC
   { Bad,             T_ILLEGAL,    "vectors:",      false, Op_VecS,              relocInfo::none          },  // VectorS
   { Bad,             T_ILLEGAL,    "vectord:",      false, Op_VecD,              relocInfo::none          },  // VectorD
   { Bad,             T_ILLEGAL,    "vectorx:",      false, Op_VecX,              relocInfo::none          },  // VectorX
diff --git a/src/share/vm/opto/vectornode.cpp b/src/share/vm/opto/vectornode.cpp
index d0955a819..9660d4ed8 100644
--- a/src/share/vm/opto/vectornode.cpp
+++ b/src/share/vm/opto/vectornode.cpp
@@ -29,8 +29,7 @@
 //------------------------------VectorNode--------------------------------------
 
 // Return the vector operator for the specified scalar operation
-// and vector length.  Also used to check if the code generator
-// supports the vector operation.
+// and vector length.
 int VectorNode::opcode(int sopc, BasicType bt) {
   switch (sopc) {
   case Op_AddI:
@@ -75,7 +74,7 @@ int VectorNode::opcode(int sopc, BasicType bt) {
     case T_BYTE:   return 0;   // Unimplemented
     case T_CHAR:
     case T_SHORT:  return Op_MulVS;
-    case T_INT:    return Matcher::match_rule_supported(Op_MulVI) ? Op_MulVI : 0; // SSE4_1
+    case T_INT:    return Op_MulVI;
     }
     ShouldNotReachHere();
   case Op_MulF:
@@ -104,9 +103,9 @@ int VectorNode::opcode(int sopc, BasicType bt) {
     return Op_LShiftVL;
   case Op_RShiftI:
     switch (bt) {
-    case T_BOOLEAN:
+    case T_BOOLEAN:return Op_URShiftVB; // boolean is unsigned value
+    case T_CHAR:   return Op_URShiftVS; // char is unsigned value
     case T_BYTE:   return Op_RShiftVB;
-    case T_CHAR:
     case T_SHORT:  return Op_RShiftVS;
     case T_INT:    return Op_RShiftVI;
     }
@@ -116,10 +115,14 @@ int VectorNode::opcode(int sopc, BasicType bt) {
     return Op_RShiftVL;
   case Op_URShiftI:
     switch (bt) {
-    case T_BOOLEAN:
-    case T_BYTE:   return Op_URShiftVB;
-    case T_CHAR:
-    case T_SHORT:  return Op_URShiftVS;
+    case T_BOOLEAN:return Op_URShiftVB;
+    case T_CHAR:   return Op_URShiftVS;
+    case T_BYTE:
+    case T_SHORT:  return 0; // Vector logical right shift for signed short
+                             // values produces incorrect Java result for
+                             // negative data because java code should convert
+                             // a short value into int value with sign
+                             // extension before a shift.
     case T_INT:    return Op_URShiftVI;
     }
     ShouldNotReachHere();
@@ -157,12 +160,14 @@ int VectorNode::opcode(int sopc, BasicType bt) {
   return 0; // Unimplemented
 }
 
+// Also used to check if the code generator
+// supports the vector operation.
 bool VectorNode::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 = VectorNode::opcode(opc, bt);
-    return vopc > 0 && Matcher::has_match_rule(vopc);
+    return vopc > 0 && Matcher::match_rule_supported(vopc);
   }
   return false;
 }
diff --git a/src/share/vm/prims/unsafe.cpp b/src/share/vm/prims/unsafe.cpp
index 051c85975..e3750e66f 100644
--- a/src/share/vm/prims/unsafe.cpp
+++ b/src/share/vm/prims/unsafe.cpp
@@ -124,6 +124,8 @@ inline void* index_oop_from_field_offset_long(oop p, jlong field_offset) {
       assert((void*)p->obj_field_addr<oop>((jint)byte_offset) == ptr_plus_disp,
              "raw [ptr+disp] must be consistent with oop::field_base");
     }
+    jlong p_size = HeapWordSize * (jlong)(p->size());
+    assert(byte_offset < p_size, err_msg("Unsafe access: offset " INT64_FORMAT " > object's size " INT64_FORMAT, byte_offset, p_size));
   }
 #endif
   if (sizeof(char*) == sizeof(jint))    // (this constant folds!)
diff --git a/src/share/vm/runtime/arguments.cpp b/src/share/vm/runtime/arguments.cpp
index 0b55da2c0..12fbddc29 100644
--- a/src/share/vm/runtime/arguments.cpp
+++ b/src/share/vm/runtime/arguments.cpp
@@ -791,6 +791,10 @@ void Arguments::print_on(outputStream* st) {
     st->print("jvm_args: "); print_jvm_args_on(st);
   }
   st->print_cr("java_command: %s", java_command() ? java_command() : "<unknown>");
+  if (_java_class_path != NULL) {
+    char* path = _java_class_path->value();
+    st->print_cr("java_class_path (initial): %s", strlen(path) == 0 ? "<not set>" : path );
+  }
   st->print_cr("Launcher Type: %s", _sun_java_launcher);
 }
 
@@ -2771,6 +2775,11 @@ SOLARIS_ONLY(
         return JNI_EINVAL;
       }
       FLAG_SET_CMDLINE(uintx, MaxDirectMemorySize, max_direct_memory_size);
+    } else if (match_option(option, "-XX:+UseVMInterruptibleIO", &tail)) {
+      // NOTE! In JDK 9, the UseVMInterruptibleIO flag will completely go
+      //       away and will cause VM initialization failures!
+      warning("-XX:+UseVMInterruptibleIO is obsolete and will be removed in a future release.");
+      FLAG_SET_CMDLINE(bool, UseVMInterruptibleIO, true);
     } else if (match_option(option, "-XX:", &tail)) { // -XX:xxxx
       // Skip -XX:Flags= since that case has already been handled
       if (strncmp(tail, "Flags=", strlen("Flags=")) != 0) {
@@ -2786,10 +2795,6 @@ SOLARIS_ONLY(
 
   // Change the default value for flags  which have different default values
   // when working with older JDKs.
-  if (JDK_Version::current().compare_major(6) <= 0 &&
-      FLAG_IS_DEFAULT(UseVMInterruptibleIO)) {
-    FLAG_SET_DEFAULT(UseVMInterruptibleIO, true);
-  }
 #ifdef LINUX
  if (JDK_Version::current().compare_major(6) <= 0 &&
       FLAG_IS_DEFAULT(UseLinuxPosixThreadCPUClocks)) {
diff --git a/src/share/vm/runtime/globals.hpp b/src/share/vm/runtime/globals.hpp
index e783883eb..39e6a98ab 100644
--- a/src/share/vm/runtime/globals.hpp
+++ b/src/share/vm/runtime/globals.hpp
@@ -533,6 +533,9 @@ class CommandLineFlags {
   product(intx, UseSSE, 99,                                                 \
           "Highest supported SSE instructions set on x86/x64")              \
                                                                             \
+  product(bool, UseAES, false,                                               \
+          "Control whether AES instructions can be used on x86/x64")        \
+                                                                            \
   product(uintx, LargePageSizeInBytes, 0,                                   \
           "Large page size (0 to let VM choose the page size")              \
                                                                             \
@@ -635,6 +638,9 @@ class CommandLineFlags {
   product(bool, UseSSE42Intrinsics, false,                                  \
           "SSE4.2 versions of intrinsics")                                  \
                                                                             \
+  product(bool, UseAESIntrinsics, false,                                    \
+          "use intrinsics for AES versions of crypto")                      \
+                                                                            \
   develop(bool, TraceCallFixup, false,                                      \
           "traces all call fixups")                                         \
                                                                             \
diff --git a/src/share/vm/runtime/handles.cpp b/src/share/vm/runtime/handles.cpp
index cb53088ee..a37e46da1 100644
--- a/src/share/vm/runtime/handles.cpp
+++ b/src/share/vm/runtime/handles.cpp
@@ -158,13 +158,18 @@ HandleMark::~HandleMark() {
 
   // Delete later chunks
   if( _chunk->next() ) {
+    // reset arena size before delete chunks. Otherwise, the total
+    // arena size could exceed total chunk size
+    assert(area->size_in_bytes() > size_in_bytes(), "Sanity check");
+    area->set_size_in_bytes(size_in_bytes());
     _chunk->next_chop();
+  } else {
+    assert(area->size_in_bytes() == size_in_bytes(), "Sanity check");
   }
   // Roll back arena to saved top markers
   area->_chunk = _chunk;
   area->_hwm = _hwm;
   area->_max = _max;
-  area->set_size_in_bytes(_size_in_bytes);
 #ifdef ASSERT
   // clear out first chunk (to detect allocation bugs)
   if (ZapVMHandleArea) {
diff --git a/src/share/vm/runtime/handles.hpp b/src/share/vm/runtime/handles.hpp
index cab3dc581..8c643d7c2 100644
--- a/src/share/vm/runtime/handles.hpp
+++ b/src/share/vm/runtime/handles.hpp
@@ -297,6 +297,7 @@ class HandleMark {
   void set_previous_handle_mark(HandleMark* mark) { _previous_handle_mark = mark; }
   HandleMark* previous_handle_mark() const        { return _previous_handle_mark; }
 
+  size_t size_in_bytes() const { return _size_in_bytes; }
  public:
   HandleMark();                            // see handles_inline.hpp
   HandleMark(Thread* thread)                      { initialize(thread); }
diff --git a/src/share/vm/runtime/handles.inline.hpp b/src/share/vm/runtime/handles.inline.hpp
index 96dd94cc5..51c31ff4b 100644
--- a/src/share/vm/runtime/handles.inline.hpp
+++ b/src/share/vm/runtime/handles.inline.hpp
@@ -136,13 +136,18 @@ inline void HandleMark::pop_and_restore() {
   HandleArea* area = _area;   // help compilers with poor alias analysis
   // Delete later chunks
   if( _chunk->next() ) {
+    // reset arena size before delete chunks. Otherwise, the total
+    // arena size could exceed total chunk size
+    assert(area->size_in_bytes() > size_in_bytes(), "Sanity check");
+    area->set_size_in_bytes(size_in_bytes());
     _chunk->next_chop();
+  } else {
+    assert(area->size_in_bytes() == size_in_bytes(), "Sanity check");
   }
   // Roll back arena to saved top markers
   area->_chunk = _chunk;
   area->_hwm = _hwm;
   area->_max = _max;
-  area->set_size_in_bytes(_size_in_bytes);
   debug_only(area->_handle_mark_nesting--);
 }
 
diff --git a/src/share/vm/runtime/os.cpp b/src/share/vm/runtime/os.cpp
index 550e64a27..56bd82613 100644
--- a/src/share/vm/runtime/os.cpp
+++ b/src/share/vm/runtime/os.cpp
@@ -600,9 +600,7 @@ void* os::malloc(size_t size, MEMFLAGS memflags, address caller) {
   if (PrintMalloc && tty != NULL) tty->print_cr("os::malloc " SIZE_FORMAT " bytes --> " PTR_FORMAT, size, memblock);
 
   // we do not track MallocCushion memory
-  if (MemTracker::is_on()) {
     MemTracker::record_malloc((address)memblock, size, memflags, caller == 0 ? CALLER_PC : caller);
-  }
 
   return memblock;
 }
@@ -613,7 +611,7 @@ void* os::realloc(void *memblock, size_t size, MEMFLAGS memflags, address caller
   NOT_PRODUCT(inc_stat_counter(&num_mallocs, 1));
   NOT_PRODUCT(inc_stat_counter(&alloc_bytes, size));
   void* ptr = ::realloc(memblock, size);
-  if (ptr != NULL && MemTracker::is_on()) {
+  if (ptr != NULL) {
     MemTracker::record_realloc((address)memblock, (address)ptr, size, memflags,
      caller == 0 ? CALLER_PC : caller);
   }
@@ -1401,7 +1399,7 @@ bool os::create_stack_guard_pages(char* addr, size_t bytes) {
 
 char* os::reserve_memory(size_t bytes, char* addr, size_t alignment_hint) {
   char* result = pd_reserve_memory(bytes, addr, alignment_hint);
-  if (result != NULL && MemTracker::is_on()) {
+  if (result != NULL) {
     MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
   }
 
@@ -1409,7 +1407,7 @@ char* os::reserve_memory(size_t bytes, char* addr, size_t alignment_hint) {
 }
 char* os::attempt_reserve_memory_at(size_t bytes, char* addr) {
   char* result = pd_attempt_reserve_memory_at(bytes, addr);
-  if (result != NULL && MemTracker::is_on()) {
+  if (result != NULL) {
     MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
   }
   return result;
@@ -1422,7 +1420,7 @@ void os::split_reserved_memory(char *base, size_t size,
 
 bool os::commit_memory(char* addr, size_t bytes, bool executable) {
   bool res = pd_commit_memory(addr, bytes, executable);
-  if (res && MemTracker::is_on()) {
+  if (res) {
     MemTracker::record_virtual_memory_commit((address)addr, bytes, CALLER_PC);
   }
   return res;
@@ -1431,7 +1429,7 @@ bool os::commit_memory(char* addr, size_t bytes, bool executable) {
 bool os::commit_memory(char* addr, size_t size, size_t alignment_hint,
                               bool executable) {
   bool res = os::pd_commit_memory(addr, size, alignment_hint, executable);
-  if (res && MemTracker::is_on()) {
+  if (res) {
     MemTracker::record_virtual_memory_commit((address)addr, size, CALLER_PC);
   }
   return res;
@@ -1458,8 +1456,9 @@ char* os::map_memory(int fd, const char* file_name, size_t file_offset,
                            char *addr, size_t bytes, bool read_only,
                            bool allow_exec) {
   char* result = pd_map_memory(fd, file_name, file_offset, addr, bytes, read_only, allow_exec);
-  if (result != NULL && MemTracker::is_on()) {
+  if (result != NULL) {
     MemTracker::record_virtual_memory_reserve((address)result, bytes, CALLER_PC);
+    MemTracker::record_virtual_memory_commit((address)result, bytes, CALLER_PC);
   }
   return result;
 }
@@ -1474,6 +1473,7 @@ char* os::remap_memory(int fd, const char* file_name, size_t file_offset,
 bool os::unmap_memory(char *addr, size_t bytes) {
   bool result = pd_unmap_memory(addr, bytes);
   if (result) {
+    MemTracker::record_virtual_memory_uncommit((address)addr, bytes);
     MemTracker::record_virtual_memory_release((address)addr, bytes);
   }
   return result;
diff --git a/src/share/vm/runtime/stubRoutines.cpp b/src/share/vm/runtime/stubRoutines.cpp
index 5ca4ba599..98d428abd 100644
--- a/src/share/vm/runtime/stubRoutines.cpp
+++ b/src/share/vm/runtime/stubRoutines.cpp
@@ -120,6 +120,10 @@ address StubRoutines::_arrayof_jbyte_fill;
 address StubRoutines::_arrayof_jshort_fill;
 address StubRoutines::_arrayof_jint_fill;
 
+address StubRoutines::_aescrypt_encryptBlock               = NULL;
+address StubRoutines::_aescrypt_decryptBlock               = NULL;
+address StubRoutines::_cipherBlockChaining_encryptAESCrypt = NULL;
+address StubRoutines::_cipherBlockChaining_decryptAESCrypt = NULL;
 
 double (* StubRoutines::_intrinsic_log   )(double) = NULL;
 double (* StubRoutines::_intrinsic_log10 )(double) = NULL;
diff --git a/src/share/vm/runtime/stubRoutines.hpp b/src/share/vm/runtime/stubRoutines.hpp
index 0e583aea0..91f273e65 100644
--- a/src/share/vm/runtime/stubRoutines.hpp
+++ b/src/share/vm/runtime/stubRoutines.hpp
@@ -199,6 +199,11 @@ class StubRoutines: AllStatic {
   // zero heap space aligned to jlong (8 bytes)
   static address _zero_aligned_words;
 
+  static address _aescrypt_encryptBlock;
+  static address _aescrypt_decryptBlock;
+  static address _cipherBlockChaining_encryptAESCrypt;
+  static address _cipherBlockChaining_decryptAESCrypt;
+
   // These are versions of the java.lang.Math methods which perform
   // the same operations as the intrinsic version.  They are used for
   // constant folding in the compiler to ensure equivalence.  If the
@@ -330,6 +335,11 @@ class StubRoutines: AllStatic {
   static address arrayof_jshort_fill() { return _arrayof_jshort_fill; }
   static address arrayof_jint_fill()   { return _arrayof_jint_fill; }
 
+  static address aescrypt_encryptBlock()                { return _aescrypt_encryptBlock; }
+  static address aescrypt_decryptBlock()                { return _aescrypt_decryptBlock; }
+  static address cipherBlockChaining_encryptAESCrypt()  { return _cipherBlockChaining_encryptAESCrypt; }
+  static address cipherBlockChaining_decryptAESCrypt()  { return _cipherBlockChaining_decryptAESCrypt; }
+
   static address select_fill_function(BasicType t, bool aligned, const char* &name);
 
   static address zero_aligned_words()   { return _zero_aligned_words; }
diff --git a/src/share/vm/runtime/thread.cpp b/src/share/vm/runtime/thread.cpp
index 0ed9a4b1f..5c96224a4 100644
--- a/src/share/vm/runtime/thread.cpp
+++ b/src/share/vm/runtime/thread.cpp
@@ -323,12 +323,10 @@ void Thread::record_stack_base_and_size() {
   os::initialize_thread(this);
 
 #if INCLUDE_NMT
-   // record thread's native stack, stack grows downward
-  if (MemTracker::is_on()) {
-    address stack_low_addr = stack_base() - stack_size();
-    MemTracker::record_thread_stack(stack_low_addr, stack_size(), this,
+  // record thread's native stack, stack grows downward
+  address stack_low_addr = stack_base() - stack_size();
+  MemTracker::record_thread_stack(stack_low_addr, stack_size(), this,
       CURRENT_PC);
-  }
 #endif // INCLUDE_NMT
 }
 
@@ -345,6 +343,9 @@ Thread::~Thread() {
   if (_stack_base != NULL) {
     address low_stack_addr = stack_base() - stack_size();
     MemTracker::release_thread_stack(low_stack_addr, stack_size(), this);
+#ifdef ASSERT
+    set_stack_base(NULL);
+#endif
   }
 #endif // INCLUDE_NMT
 
@@ -1521,10 +1522,12 @@ JavaThread::~JavaThread() {
       tty->print_cr("terminate thread %p", this);
   }
 
-  // Info NMT that this JavaThread is exiting, its memory
-  // recorder should be collected
+  // By now, this thread should already be invisible to safepoint,
+  // and its per-thread recorder also collected.
   assert(!is_safepoint_visible(), "wrong state");
-  MemTracker::thread_exiting(this);
+#if INCLUDE_NMT
+  assert(get_recorder() == NULL, "Already collected");
+#endif // INCLUDE_NMT
 
   // JSR166 -- return the parker to the free list
   Parker::Release(_parker);
@@ -2425,6 +2428,7 @@ void JavaThread::create_stack_guard_pages() {
 }
 
 void JavaThread::remove_stack_guard_pages() {
+  assert(Thread::current() == this, "from different thread");
   if (_stack_guard_state == stack_guard_unused) return;
   address low_addr = stack_base() - stack_size();
   size_t len = (StackYellowPages + StackRedPages) * os::vm_page_size();
@@ -4093,7 +4097,10 @@ void Threads::remove(JavaThread* p) {
 
     // Now, this thread is not visible to safepoint
     p->set_safepoint_visible(false);
-
+    // once the thread becomes safepoint invisible, we can not use its per-thread
+    // recorder. And Threads::do_threads() no longer walks this thread, so we have
+    // to release its per-thread recorder here.
+    MemTracker::thread_exiting(p);
   } // unlock Threads_lock
 
   // Since Events::log uses a lock, we grab it outside the Threads_lock
diff --git a/src/share/vm/runtime/vmStructs.cpp b/src/share/vm/runtime/vmStructs.cpp
index 51e3ab5f7..9159ad945 100644
--- a/src/share/vm/runtime/vmStructs.cpp
+++ b/src/share/vm/runtime/vmStructs.cpp
@@ -59,6 +59,7 @@
 #include "memory/generation.hpp"
 #include "memory/generationSpec.hpp"
 #include "memory/heap.hpp"
+#include "memory/metablock.hpp"
 #include "memory/space.hpp"
 #include "memory/tenuredGeneration.hpp"
 #include "memory/universe.hpp"
@@ -249,6 +250,7 @@ typedef TwoOopHashtable<Klass*, mtClass>      KlassTwoOopHashtable;
 typedef Hashtable<Klass*, mtClass>            KlassHashtable;
 typedef HashtableEntry<Klass*, mtClass>       KlassHashtableEntry;
 typedef TwoOopHashtable<Symbol*, mtClass>     SymbolTwoOopHashtable;
+typedef BinaryTreeDictionary<Metablock, FreeList> MetablockTreeDictionary;
 
 //--------------------------------------------------------------------------------
 // VM_STRUCTS
@@ -1237,7 +1239,15 @@ typedef TwoOopHashtable<Symbol*, mtClass>     SymbolTwoOopHashtable;
   nonstatic_field(AccessFlags,                 _flags,                                       jint)                                   \
   nonstatic_field(elapsedTimer,                _counter,                                     jlong)                                  \
   nonstatic_field(elapsedTimer,                _active,                                      bool)                                   \
-  nonstatic_field(InvocationCounter,           _counter,                                     unsigned int)
+  nonstatic_field(InvocationCounter,           _counter,                                     unsigned int)                           \
+  volatile_nonstatic_field(FreeChunk,          _size,                                        size_t)                                 \
+  nonstatic_field(FreeChunk,                   _next,                                        FreeChunk*)                             \
+  nonstatic_field(FreeChunk,                   _prev,                                        FreeChunk*)                             \
+  nonstatic_field(FreeList<FreeChunk>,         _size,                                        size_t)                                 \
+  nonstatic_field(FreeList<Metablock>,         _size,                                        size_t)                                 \
+  nonstatic_field(FreeList<FreeChunk>,         _count,                                       ssize_t)                                \
+  nonstatic_field(FreeList<Metablock>,         _count,                                       ssize_t)                                \
+  nonstatic_field(MetablockTreeDictionary,     _total_size,                                  size_t)
 
   /* NOTE that we do not use the last_entry() macro here; it is used  */
   /* in vmStructs_<os>_<cpu>.hpp's VM_STRUCTS_OS_CPU macro (and must  */
@@ -2080,7 +2090,24 @@ typedef TwoOopHashtable<Symbol*, mtClass>     SymbolTwoOopHashtable;
   declare_toplevel_type(Universe)                                         \
   declare_toplevel_type(vframeArray)                                      \
   declare_toplevel_type(vframeArrayElement)                               \
-  declare_toplevel_type(Annotations*)
+  declare_toplevel_type(Annotations*)                                     \
+                                                                          \
+  /***************/                                                       \
+  /* Miscellaneous types */                                               \
+  /***************/                                                       \
+                                                                          \
+  /* freelist */                                                          \
+  declare_toplevel_type(FreeChunk*)                                       \
+  declare_toplevel_type(Metablock*)                                       \
+  declare_toplevel_type(FreeBlockDictionary<FreeChunk>*)                  \
+  declare_toplevel_type(FreeList<FreeChunk>*)                             \
+  declare_toplevel_type(FreeList<FreeChunk>)                              \
+  declare_toplevel_type(FreeBlockDictionary<Metablock>*)                  \
+  declare_toplevel_type(FreeList<Metablock>*)                             \
+  declare_toplevel_type(FreeList<Metablock>)                              \
+  declare_toplevel_type(MetablockTreeDictionary*)                         \
+  declare_type(MetablockTreeDictionary, FreeBlockDictionary<Metablock>)   \
+              declare_type(MetablockTreeDictionary, FreeBlockDictionary<Metablock>)
 
 
   /* NOTE that we do not use the last_entry() macro here; it is used  */
diff --git a/src/share/vm/services/attachListener.cpp b/src/share/vm/services/attachListener.cpp
index 983f3f95f..80dfd7eff 100644
--- a/src/share/vm/services/attachListener.cpp
+++ b/src/share/vm/services/attachListener.cpp
@@ -404,6 +404,8 @@ static AttachOperationFunctionInfo funcs[] = {
 static void attach_listener_thread_entry(JavaThread* thread, TRAPS) {
   os::set_priority(thread, NearMaxPriority);
 
+  thread->record_stack_base_and_size();
+
   if (AttachListener::pd_init() != 0) {
     return;
   }
diff --git a/src/share/vm/services/memBaseline.cpp b/src/share/vm/services/memBaseline.cpp
index 5b8297516..57bdb0138 100644
--- a/src/share/vm/services/memBaseline.cpp
+++ b/src/share/vm/services/memBaseline.cpp
@@ -40,6 +40,7 @@ MemType2Name MemBaseline::MemType2NameMap[NUMBER_OF_MEMORY_TYPE] = {
   {mtSymbol,     "Symbol"},
   {mtNMT,        "Memory Tracking"},
   {mtChunk,      "Pooled Free Chunks"},
+  {mtClassShared,"Shared spaces for classes"},
   {mtNone,       "Unknown"}  // It can happen when type tagging records are lagging
                              // behind
 };
@@ -55,6 +56,7 @@ MemBaseline::MemBaseline() {
 
   _malloc_cs = NULL;
   _vm_cs = NULL;
+  _vm_map = NULL;
 
   _number_of_classes = 0;
   _number_of_threads = 0;
@@ -72,6 +74,11 @@ void MemBaseline::clear() {
     _vm_cs = NULL;
   }
 
+  if (_vm_map != NULL) {
+    delete _vm_map;
+    _vm_map = NULL;
+  }
+
   reset();
 }
 
@@ -85,6 +92,7 @@ void MemBaseline::reset() {
 
   if (_malloc_cs != NULL) _malloc_cs->clear();
   if (_vm_cs != NULL) _vm_cs->clear();
+  if (_vm_map != NULL) _vm_map->clear();
 
   for (int index = 0; index < NUMBER_OF_MEMORY_TYPE; index ++) {
     _malloc_data[index].clear();
@@ -94,39 +102,33 @@ void MemBaseline::reset() {
 }
 
 MemBaseline::~MemBaseline() {
-  if (_malloc_cs != NULL) {
-    delete _malloc_cs;
-  }
-
-  if (_vm_cs != NULL) {
-    delete _vm_cs;
-  }
+  clear();
 }
 
 // baseline malloc'd memory records, generate overall summary and summaries by
 // memory types
 bool MemBaseline::baseline_malloc_summary(const MemPointerArray* malloc_records) {
-  MemPointerArrayIteratorImpl mItr((MemPointerArray*)malloc_records);
-  MemPointerRecord* mptr = (MemPointerRecord*)mItr.current();
+  MemPointerArrayIteratorImpl malloc_itr((MemPointerArray*)malloc_records);
+  MemPointerRecord* malloc_ptr = (MemPointerRecord*)malloc_itr.current();
   size_t used_arena_size = 0;
   int index;
-  while (mptr != NULL) {
-    index = flag2index(FLAGS_TO_MEMORY_TYPE(mptr->flags()));
-    size_t size = mptr->size();
+  while (malloc_ptr != NULL) {
+    index = flag2index(FLAGS_TO_MEMORY_TYPE(malloc_ptr->flags()));
+    size_t size = malloc_ptr->size();
     _total_malloced += size;
     _malloc_data[index].inc(size);
-    if (MemPointerRecord::is_arena_record(mptr->flags())) {
+    if (MemPointerRecord::is_arena_record(malloc_ptr->flags())) {
       // see if arena size record present
-      MemPointerRecord* next_p = (MemPointerRecordEx*)mItr.peek_next();
-      if (MemPointerRecord::is_arena_size_record(next_p->flags())) {
-        assert(next_p->is_size_record_of_arena(mptr), "arena records do not match");
-        size = next_p->size();
+      MemPointerRecord* next_malloc_ptr = (MemPointerRecordEx*)malloc_itr.peek_next();
+      if (MemPointerRecord::is_arena_size_record(next_malloc_ptr->flags())) {
+        assert(next_malloc_ptr->is_size_record_of_arena(malloc_ptr), "arena records do not match");
+        size = next_malloc_ptr->size();
         _arena_data[index].inc(size);
         used_arena_size += size;
-        mItr.next();
+        malloc_itr.next();
       }
     }
-    mptr = (MemPointerRecordEx*)mItr.next();
+    malloc_ptr = (MemPointerRecordEx*)malloc_itr.next();
   }
 
   // substract used arena size to get size of arena chunk in free list
@@ -142,20 +144,23 @@ bool MemBaseline::baseline_malloc_summary(const MemPointerArray* malloc_records)
 // baseline mmap'd memory records, generate overall summary and summaries by
 // memory types
 bool MemBaseline::baseline_vm_summary(const MemPointerArray* vm_records) {
-  MemPointerArrayIteratorImpl vItr((MemPointerArray*)vm_records);
-  VMMemRegion* vptr = (VMMemRegion*)vItr.current();
+  MemPointerArrayIteratorImpl vm_itr((MemPointerArray*)vm_records);
+  VMMemRegion* vm_ptr = (VMMemRegion*)vm_itr.current();
   int index;
-  while (vptr != NULL) {
-    index = flag2index(FLAGS_TO_MEMORY_TYPE(vptr->flags()));
-
+  while (vm_ptr != NULL) {
+    if (vm_ptr->is_reserved_region()) {
+      index = flag2index(FLAGS_TO_MEMORY_TYPE(vm_ptr->flags()));
     // we use the number of thread stack to count threads
-    if (IS_MEMORY_TYPE(vptr->flags(), mtThreadStack)) {
+      if (IS_MEMORY_TYPE(vm_ptr->flags(), mtThreadStack)) {
       _number_of_threads ++;
     }
-    _total_vm_reserved += vptr->reserved_size();
-    _total_vm_committed += vptr->committed_size();
-    _vm_data[index].inc(vptr->reserved_size(), vptr->committed_size());
-    vptr = (VMMemRegion*)vItr.next();
+      _total_vm_reserved += vm_ptr->size();
+      _vm_data[index].inc(vm_ptr->size(), 0);
+    } else {
+      _total_vm_committed += vm_ptr->size();
+      _vm_data[index].inc(0, vm_ptr->size());
+    }
+    vm_ptr = (VMMemRegion*)vm_itr.next();
   }
   return true;
 }
@@ -165,41 +170,57 @@ bool MemBaseline::baseline_vm_summary(const MemPointerArray* vm_records) {
 bool MemBaseline::baseline_malloc_details(const MemPointerArray* malloc_records) {
   assert(MemTracker::track_callsite(), "detail tracking is off");
 
-  MemPointerArrayIteratorImpl mItr((MemPointerArray*)malloc_records);
-  MemPointerRecordEx* mptr = (MemPointerRecordEx*)mItr.current();
-  MallocCallsitePointer mp;
+  MemPointerArrayIteratorImpl malloc_itr(const_cast<MemPointerArray*>(malloc_records));
+  MemPointerRecordEx* malloc_ptr = (MemPointerRecordEx*)malloc_itr.current();
+  MallocCallsitePointer malloc_callsite;
 
+  // initailize malloc callsite array
   if (_malloc_cs == NULL) {
     _malloc_cs = new (std::nothrow) MemPointerArrayImpl<MallocCallsitePointer>(64);
     // out of native memory
-    if (_malloc_cs == NULL) {
+    if (_malloc_cs == NULL || _malloc_cs->out_of_memory()) {
       return false;
     }
   } else {
     _malloc_cs->clear();
   }
 
+  MemPointerArray* malloc_data = const_cast<MemPointerArray*>(malloc_records);
+
+  // sort into callsite pc order. Details are aggregated by callsites
+  malloc_data->sort((FN_SORT)malloc_sort_by_pc);
+  bool ret = true;
+
   // baseline memory that is totaled over 1 KB
-  while (mptr != NULL) {
-    if (!MemPointerRecord::is_arena_size_record(mptr->flags())) {
+  while (malloc_ptr != NULL) {
+    if (!MemPointerRecord::is_arena_size_record(malloc_ptr->flags())) {
       // skip thread stacks
-      if (!IS_MEMORY_TYPE(mptr->flags(), mtThreadStack)) {
-        if (mp.addr() != mptr->pc()) {
-          if ((mp.amount()/K) > 0) {
-            if (!_malloc_cs->append(&mp)) {
-              return false;
+      if (!IS_MEMORY_TYPE(malloc_ptr->flags(), mtThreadStack)) {
+        if (malloc_callsite.addr() != malloc_ptr->pc()) {
+          if ((malloc_callsite.amount()/K) > 0) {
+            if (!_malloc_cs->append(&malloc_callsite)) {
+              ret = false;
+              break;
             }
           }
-          mp = MallocCallsitePointer(mptr->pc());
+          malloc_callsite = MallocCallsitePointer(malloc_ptr->pc());
         }
-        mp.inc(mptr->size());
+        malloc_callsite.inc(malloc_ptr->size());
       }
     }
-    mptr = (MemPointerRecordEx*)mItr.next();
+    malloc_ptr = (MemPointerRecordEx*)malloc_itr.next();
   }
 
-  if (mp.addr() != 0 && (mp.amount()/K) > 0) {
-    if (!_malloc_cs->append(&mp)) {
+  // restore to address order. Snapshot malloc data is maintained in memory
+  // address order.
+  malloc_data->sort((FN_SORT)malloc_sort_by_addr);
+
+  if (!ret) {
+              return false;
+            }
+  // deal with last record
+  if (malloc_callsite.addr() != 0 && (malloc_callsite.amount()/K) > 0) {
+    if (!_malloc_cs->append(&malloc_callsite)) {
       return false;
     }
   }
@@ -210,34 +231,106 @@ bool MemBaseline::baseline_malloc_details(const MemPointerArray* malloc_records)
 bool MemBaseline::baseline_vm_details(const MemPointerArray* vm_records) {
   assert(MemTracker::track_callsite(), "detail tracking is off");
 
-  VMCallsitePointer vp;
-  MemPointerArrayIteratorImpl vItr((MemPointerArray*)vm_records);
-  VMMemRegionEx* vptr = (VMMemRegionEx*)vItr.current();
+  VMCallsitePointer  vm_callsite;
+  VMCallsitePointer* cur_callsite = NULL;
+  MemPointerArrayIteratorImpl vm_itr((MemPointerArray*)vm_records);
+  VMMemRegionEx* vm_ptr = (VMMemRegionEx*)vm_itr.current();
+
+  // initialize virtual memory map array
+  if (_vm_map == NULL) {
+    _vm_map = new (std::nothrow) MemPointerArrayImpl<VMMemRegionEx>(vm_records->length());
+   if (_vm_map == NULL || _vm_map->out_of_memory()) {
+     return false;
+   }
+  } else {
+    _vm_map->clear();
+  }
 
+  // initialize virtual memory callsite array
   if (_vm_cs == NULL) {
     _vm_cs = new (std::nothrow) MemPointerArrayImpl<VMCallsitePointer>(64);
-    if (_vm_cs == NULL) {
+    if (_vm_cs == NULL || _vm_cs->out_of_memory()) {
       return false;
     }
   } else {
     _vm_cs->clear();
   }
 
-  while (vptr != NULL) {
-    if (vp.addr() != vptr->pc()) {
-      if (!_vm_cs->append(&vp)) {
+  // consolidate virtual memory data
+  VMMemRegionEx*     reserved_rec = NULL;
+  VMMemRegionEx*     committed_rec = NULL;
+
+  // vm_ptr is coming in increasing base address order
+  while (vm_ptr != NULL) {
+    if (vm_ptr->is_reserved_region()) {
+      // consolidate reserved memory regions for virtual memory map.
+      // The criteria for consolidation is:
+      // 1. two adjacent reserved memory regions
+      // 2. belong to the same memory type
+      // 3. reserved from the same callsite
+      if (reserved_rec == NULL ||
+        reserved_rec->base() + reserved_rec->size() != vm_ptr->addr() ||
+        FLAGS_TO_MEMORY_TYPE(reserved_rec->flags()) != FLAGS_TO_MEMORY_TYPE(vm_ptr->flags()) ||
+        reserved_rec->pc() != vm_ptr->pc()) {
+        if (!_vm_map->append(vm_ptr)) {
         return false;
       }
-      vp = VMCallsitePointer(vptr->pc());
+        // inserted reserved region, we need the pointer to the element in virtual
+        // memory map array.
+        reserved_rec = (VMMemRegionEx*)_vm_map->at(_vm_map->length() - 1);
+      } else {
+        reserved_rec->expand_region(vm_ptr->addr(), vm_ptr->size());
     }
-    vp.inc(vptr->size(), vptr->committed_size());
-    vptr = (VMMemRegionEx*)vItr.next();
-  }
-  if (vp.addr() != 0) {
-    if (!_vm_cs->append(&vp)) {
+
+      if (cur_callsite != NULL && !_vm_cs->append(cur_callsite)) {
       return false;
     }
+      vm_callsite = VMCallsitePointer(vm_ptr->pc());
+      cur_callsite = &vm_callsite;
+      vm_callsite.inc(vm_ptr->size(), 0);
+    } else {
+      // consolidate committed memory regions for virtual memory map
+      // The criterial is:
+      // 1. two adjacent committed memory regions
+      // 2. committed from the same callsite
+      if (committed_rec == NULL ||
+        committed_rec->base() + committed_rec->size() != vm_ptr->addr() ||
+        committed_rec->pc() != vm_ptr->pc()) {
+        if (!_vm_map->append(vm_ptr)) {
+          return false;
+  }
+        committed_rec = (VMMemRegionEx*)_vm_map->at(_vm_map->length() - 1);
+    } else {
+        committed_rec->expand_region(vm_ptr->addr(), vm_ptr->size());
+      }
+      vm_callsite.inc(0, vm_ptr->size());
+    }
+    vm_ptr = (VMMemRegionEx*)vm_itr.next();
   }
+  // deal with last record
+  if (cur_callsite != NULL && !_vm_cs->append(cur_callsite)) {
+    return false;
+  }
+
+  // sort it into callsite pc order. Details are aggregated by callsites
+  _vm_cs->sort((FN_SORT)bl_vm_sort_by_pc);
+
+  // walk the array to consolidate record by pc
+  MemPointerArrayIteratorImpl itr(_vm_cs);
+  VMCallsitePointer* callsite_rec = (VMCallsitePointer*)itr.current();
+  VMCallsitePointer* next_rec = (VMCallsitePointer*)itr.next();
+  while (next_rec != NULL) {
+    assert(callsite_rec != NULL, "Sanity check");
+    if (next_rec->addr() == callsite_rec->addr()) {
+      callsite_rec->inc(next_rec->reserved_amount(), next_rec->committed_amount());
+      itr.remove();
+      next_rec = (VMCallsitePointer*)itr.current();
+    } else {
+      callsite_rec = next_rec;
+      next_rec = (VMCallsitePointer*)itr.next();
+    }
+  }
+
   return true;
 }
 
@@ -251,12 +344,8 @@ bool MemBaseline::baseline(MemSnapshot& snapshot, bool summary_only) {
   _number_of_classes = SystemDictionary::number_of_classes();
 
   if (!summary_only && MemTracker::track_callsite() && _baselined) {
-    ((MemPointerArray*)snapshot._alloc_ptrs)->sort((FN_SORT)malloc_sort_by_pc);
-    ((MemPointerArray*)snapshot._vm_ptrs)->sort((FN_SORT)vm_sort_by_pc);
     _baselined =  baseline_malloc_details(snapshot._alloc_ptrs) &&
       baseline_vm_details(snapshot._vm_ptrs);
-    ((MemPointerArray*)snapshot._alloc_ptrs)->sort((FN_SORT)malloc_sort_by_addr);
-    ((MemPointerArray*)snapshot._vm_ptrs)->sort((FN_SORT)vm_sort_by_addr);
   }
   return _baselined;
 }
@@ -278,7 +367,7 @@ const char* MemBaseline::type2name(MEMFLAGS type) {
       return MemType2NameMap[index]._name;
     }
   }
-  assert(false, "no type");
+  assert(false, err_msg("bad type %x", type));
   return NULL;
 }
 
@@ -341,13 +430,6 @@ int MemBaseline::bl_malloc_sort_by_pc(const void* p1, const void* p2) {
   return UNSIGNED_COMPARE(mp1->addr(), mp2->addr());
 }
 
-// sort snapshot mmap'd records in callsite pc order
-int MemBaseline::vm_sort_by_pc(const void* p1, const void* p2) {
-  assert(MemTracker::track_callsite(),"Just check");
-  const VMMemRegionEx* mp1 = (const VMMemRegionEx*)p1;
-  const VMMemRegionEx* mp2 = (const VMMemRegionEx*)p2;
-  return UNSIGNED_COMPARE(mp1->pc(), mp2->pc());
-}
 
 // sort baselined mmap'd records in size (reserved size) order
 int MemBaseline::bl_vm_sort_by_size(const void* p1, const void* p2) {
@@ -376,12 +458,3 @@ int MemBaseline::malloc_sort_by_addr(const void* p1, const void* p2) {
   return delta;
 }
 
-// sort snapshot mmap'd records in memory block address order
-int MemBaseline::vm_sort_by_addr(const void* p1, const void* p2) {
-  assert(MemTracker::is_on(), "Just check");
-  const VMMemRegion* mp1 = (const VMMemRegion*)p1;
-  const VMMemRegion* mp2 = (const VMMemRegion*)p2;
-  int delta = UNSIGNED_COMPARE(mp1->addr(), mp2->addr());
-  assert(delta != 0, "dup pointer");
-  return delta;
-}
diff --git a/src/share/vm/services/memBaseline.hpp b/src/share/vm/services/memBaseline.hpp
index 2affa2b96..5f98e30d4 100644
--- a/src/share/vm/services/memBaseline.hpp
+++ b/src/share/vm/services/memBaseline.hpp
@@ -320,6 +320,8 @@ class MemBaseline : public _ValueObj {
   // only available when detail tracking is on.
   MemPointerArray*  _malloc_cs;
   MemPointerArray*  _vm_cs;
+  // virtual memory map
+  MemPointerArray*  _vm_map;
 
  private:
   static MemType2Name  MemType2NameMap[NUMBER_OF_MEMORY_TYPE];
@@ -432,9 +434,6 @@ class MemBaseline : public _ValueObj {
   static int malloc_sort_by_pc(const void* p1, const void* p2);
   static int malloc_sort_by_addr(const void* p1, const void* p2);
 
-  static int vm_sort_by_pc(const void* p1, const void* p2);
-  static int vm_sort_by_addr(const void* p1, const void* p2);
-
  private:
   // sorting functions for baselined records
   static int bl_malloc_sort_by_size(const void* p1, const void* p2);
diff --git a/src/share/vm/services/memPtr.cpp b/src/share/vm/services/memPtr.cpp
index d2a18765a..5d0fbf5bf 100644
--- a/src/share/vm/services/memPtr.cpp
+++ b/src/share/vm/services/memPtr.cpp
@@ -40,35 +40,3 @@ jint SequenceGenerator::next() {
   return seq;
 }
 
-
-
-bool VMMemRegion::contains(const VMMemRegion* mr) const {
-  assert(base() != 0, "Sanity check");
-  assert(size() != 0 || committed_size() != 0,
-    "Sanity check");
-  address base_addr = base();
-  address end_addr = base_addr +
-    (is_reserve_record()? reserved_size(): committed_size());
-  if (mr->is_reserve_record()) {
-    if (mr->base() == base_addr && mr->size() == size()) {
-      // the same range
-      return true;
-    }
-    return false;
-  } else if (mr->is_commit_record() || mr->is_uncommit_record()) {
-    assert(mr->base() != 0 && mr->committed_size() > 0,
-      "bad record");
-    return (mr->base() >= base_addr &&
-      (mr->base() + mr->committed_size()) <= end_addr);
-  } else if (mr->is_type_tagging_record()) {
-    assert(mr->base() != NULL, "Sanity check");
-    return (mr->base() >= base_addr && mr->base() < end_addr);
-  } else if (mr->is_release_record()) {
-    assert(mr->base() != 0 && mr->size() > 0,
-      "bad record");
-    return (mr->base() == base_addr && mr->size() == size());
-  } else {
-    ShouldNotReachHere();
-    return false;
-  }
-}
diff --git a/src/share/vm/services/memPtr.hpp b/src/share/vm/services/memPtr.hpp
index bef49b2b4..0618a3c80 100644
--- a/src/share/vm/services/memPtr.hpp
+++ b/src/share/vm/services/memPtr.hpp
@@ -291,6 +291,26 @@ public:
   inline bool is_type_tagging_record() const {
     return is_virtual_memory_type_record(_flags);
   }
+
+  // if the two memory pointer records actually represent the same
+  // memory block
+  inline bool is_same_region(const MemPointerRecord* other) const {
+    return (addr() == other->addr() && size() == other->size());
+  }
+
+  // if this memory region fully contains another one
+  inline bool contains_region(const MemPointerRecord* other) const {
+    return contains_region(other->addr(), other->size());
+  }
+
+  // if this memory region fully contains specified memory range
+  inline bool contains_region(address add, size_t sz) const {
+    return (addr() <= add && addr() + size() >= add + sz);
+  }
+
+  inline bool contains_address(address add) const {
+    return (addr() <= add && addr() + size() > add);
+  }
 };
 
 // MemPointerRecordEx also records callsite pc, from where
@@ -321,66 +341,32 @@ class MemPointerRecordEx : public MemPointerRecord {
   }
 };
 
-// a virtual memory region
+// a virtual memory region. The region can represent a reserved
+// virtual memory region or a committed memory region
 class VMMemRegion : public MemPointerRecord {
- private:
-  // committed size
-  size_t       _committed_size;
-
 public:
-  VMMemRegion(): _committed_size(0) { }
+  VMMemRegion() { }
 
   void init(const MemPointerRecord* mp) {
-    assert(mp->is_vm_pointer(), "not virtual memory pointer");
+    assert(mp->is_vm_pointer(), "Sanity check");
     _addr = mp->addr();
-    if (mp->is_commit_record() || mp->is_uncommit_record()) {
-      _committed_size = mp->size();
-      set_size(_committed_size);
-    } else {
       set_size(mp->size());
-      _committed_size = 0;
-    }
     set_flags(mp->flags());
   }
 
   VMMemRegion& operator=(const VMMemRegion& other) {
     MemPointerRecord::operator=(other);
-    _committed_size = other.committed_size();
     return *this;
   }
 
-  inline bool is_reserve_record() const {
-    return is_virtual_memory_reserve_record(flags());
-  }
-
-  inline bool is_release_record() const {
-    return is_virtual_memory_release_record(flags());
-  }
-
-  // resize reserved VM range
-  inline void set_reserved_size(size_t new_size) {
-    assert(new_size >= committed_size(), "resize");
-    set_size(new_size);
-  }
-
-  inline void commit(size_t size) {
-    _committed_size += size;
+  inline bool is_reserved_region() const {
+    return is_allocation_record();
   }
 
-  inline void uncommit(size_t size) {
-    if (_committed_size >= size) {
-      _committed_size -= size;
-    } else {
-      _committed_size = 0;
-    }
+  inline bool is_committed_region() const {
+    return is_commit_record();
   }
 
-  /*
-   * if this virtual memory range covers whole range of
-   * the other VMMemRegion
-   */
-  bool contains(const VMMemRegion* mr) const;
-
   /* base address of this virtual memory range */
   inline address base() const {
     return addr();
@@ -391,13 +377,28 @@ public:
     set_flags(flags() | (f & mt_masks));
   }
 
-  // release part of memory range
-  inline void partial_release(address add, size_t sz) {
-    assert(add >= addr() && add < addr() + size(), "not valid address");
-    // for now, it can partially release from the both ends,
-    // but not in the middle
+  // expand this region to also cover specified range.
+  // The range has to be on either end of the memory region.
+  void expand_region(address addr, size_t sz) {
+    if (addr < base()) {
+      assert(addr + sz == base(), "Sanity check");
+      _addr = addr;
+      set_size(size() + sz);
+    } else {
+      assert(base() + size() == addr, "Sanity check");
+      set_size(size() + sz);
+    }
+  }
+
+  // exclude the specified address range from this region.
+  // The excluded memory range has to be on either end of this memory
+  // region.
+  inline void exclude_region(address add, size_t sz) {
+    assert(is_reserved_region() || is_committed_region(), "Sanity check");
+    assert(addr() != NULL && size() != 0, "Sanity check");
+    assert(add >= addr() && add < addr() + size(), "Sanity check");
     assert(add == addr() || (add + sz) == (addr() + size()),
-      "release in the middle");
+      "exclude in the middle");
     if (add == addr()) {
       set_addr(add + sz);
       set_size(size() - sz);
@@ -405,16 +406,6 @@ public:
       set_size(size() - sz);
     }
   }
-
-  // the committed size of the virtual memory block
-  inline size_t committed_size() const {
-    return _committed_size;
-  }
-
-  // the reserved size of the virtual memory block
-  inline size_t reserved_size() const {
-    return size();
-  }
 };
 
 class VMMemRegionEx : public VMMemRegion {
diff --git a/src/share/vm/services/memRecorder.cpp b/src/share/vm/services/memRecorder.cpp
index 33db875af..5ec865af3 100644
--- a/src/share/vm/services/memRecorder.cpp
+++ b/src/share/vm/services/memRecorder.cpp
@@ -31,14 +31,19 @@
 #include "services/memTracker.hpp"
 
 MemPointer* SequencedRecordIterator::next_record() {
-  MemPointer* itr_cur = _itr.current();
-  if (itr_cur == NULL) return NULL;
-  MemPointer* itr_next = _itr.next();
+  MemPointerRecord* itr_cur = (MemPointerRecord*)_itr.current();
+  if (itr_cur == NULL)  {
+    return itr_cur;
+  }
+
+  MemPointerRecord* itr_next = (MemPointerRecord*)_itr.next();
 
-  while (itr_next != NULL &&
-    same_kind((MemPointerRecord*)itr_cur, (MemPointerRecord*)itr_next)) {
+  // don't collapse virtual memory records
+  while (itr_next != NULL && !itr_cur->is_vm_pointer() &&
+    !itr_next->is_vm_pointer() &&
+    same_kind(itr_cur, itr_next)) {
     itr_cur = itr_next;
-    itr_next = _itr.next();
+    itr_next = (MemPointerRecord*)_itr.next();
   }
 
   return itr_cur;
diff --git a/src/share/vm/services/memRecorder.hpp b/src/share/vm/services/memRecorder.hpp
index 754cbd6c3..2afeeb09b 100644
--- a/src/share/vm/services/memRecorder.hpp
+++ b/src/share/vm/services/memRecorder.hpp
@@ -188,6 +188,7 @@ class SequencedRecordIterator : public MemPointerArrayIterator {
   // Test if the two records are the same kind: the same memory block and allocation
   // type.
   inline bool same_kind(const MemPointerRecord* p1, const MemPointerRecord* p2) const {
+    assert(!p1->is_vm_pointer() && !p2->is_vm_pointer(), "malloc pointer only");
     return (p1->addr() == p2->addr() &&
       (p1->flags() &MemPointerRecord::tag_masks) ==
       (p2->flags() & MemPointerRecord::tag_masks));
diff --git a/src/share/vm/services/memReporter.cpp b/src/share/vm/services/memReporter.cpp
index 783d951de..602ac1c08 100644
--- a/src/share/vm/services/memReporter.cpp
+++ b/src/share/vm/services/memReporter.cpp
@@ -51,6 +51,7 @@ void BaselineReporter::report_baseline(const MemBaseline& baseline, bool summary
 
   report_summaries(baseline);
   if (!summary_only && MemTracker::track_callsite()) {
+    report_virtual_memory_map(baseline);
     report_callsites(baseline);
   }
   _outputer.done();
@@ -74,6 +75,25 @@ void BaselineReporter::report_summaries(const MemBaseline& baseline) {
   _outputer.done_category_summary();
 }
 
+void BaselineReporter::report_virtual_memory_map(const MemBaseline& baseline) {
+  _outputer.start_virtual_memory_map();
+  MemBaseline* pBL = const_cast<MemBaseline*>(&baseline);
+  MemPointerArrayIteratorImpl itr = MemPointerArrayIteratorImpl(pBL->_vm_map);
+  VMMemRegionEx* rgn = (VMMemRegionEx*)itr.current();
+  while (rgn != NULL) {
+    if (rgn->is_reserved_region()) {
+      _outputer.reserved_memory_region(FLAGS_TO_MEMORY_TYPE(rgn->flags()),
+        rgn->base(), rgn->base() + rgn->size(), amount_in_current_scale(rgn->size()), rgn->pc());
+    } else {
+      _outputer.committed_memory_region(rgn->base(), rgn->base() + rgn->size(),
+        amount_in_current_scale(rgn->size()), rgn->pc());
+    }
+    rgn = (VMMemRegionEx*)itr.next();
+  }
+
+  _outputer.done_virtual_memory_map();
+}
+
 void BaselineReporter::report_callsites(const MemBaseline& baseline) {
   _outputer.start_callsite();
   MemBaseline* pBL = const_cast<MemBaseline*>(&baseline);
@@ -324,6 +344,40 @@ void BaselineTTYOutputer::done_category_summary() {
   _output->print_cr(" ");
 }
 
+
+void BaselineTTYOutputer::start_virtual_memory_map() {
+  _output->print_cr("Virtual memory map:");
+}
+
+void BaselineTTYOutputer::reserved_memory_region(MEMFLAGS type, address base, address end,
+                                                 size_t size, address pc) {
+  const char* unit = memory_unit(_scale);
+  char buf[128];
+  int  offset;
+  _output->print_cr(" ");
+  _output->print_cr("[" PTR_FORMAT " - " PTR_FORMAT "] reserved %d%s for %s", base, end, size, unit,
+            MemBaseline::type2name(type));
+  if (os::dll_address_to_function_name(pc, buf, sizeof(buf), &offset)) {
+      _output->print_cr("\t\tfrom [%s+0x%x]", buf, offset);
+  }
+}
+
+void BaselineTTYOutputer::committed_memory_region(address base, address end, size_t size, address pc) {
+  const char* unit = memory_unit(_scale);
+  char buf[128];
+  int  offset;
+  _output->print("\t[" PTR_FORMAT " - " PTR_FORMAT "] committed %d%s", base, end, size, unit);
+  if (os::dll_address_to_function_name(pc, buf, sizeof(buf), &offset)) {
+      _output->print_cr(" from [%s+0x%x]", buf, offset);
+  }
+}
+
+void BaselineTTYOutputer::done_virtual_memory_map() {
+  _output->print_cr(" ");
+}
+
+
+
 void BaselineTTYOutputer::start_callsite() {
   _output->print_cr("Details:");
   _output->print_cr(" ");
@@ -337,7 +391,7 @@ void BaselineTTYOutputer::malloc_callsite(address pc, size_t malloc_amt,
   size_t malloc_count) {
   if (malloc_amt > 0) {
     const char* unit = memory_unit(_scale);
-    char buf[64];
+    char buf[128];
     int  offset;
     if (pc == 0) {
       _output->print("[BOOTSTRAP]%18s", " ");
@@ -357,7 +411,7 @@ void BaselineTTYOutputer::virtual_memory_callsite(address pc, size_t reserved_am
   size_t committed_amt) {
   if (reserved_amt > 0) {
     const char* unit = memory_unit(_scale);
-    char buf[64];
+    char buf[128];
     int  offset;
     if (pc == 0) {
       _output->print("[BOOTSTRAP]%18s", " ");
@@ -502,7 +556,7 @@ void BaselineTTYOutputer::diff_malloc_callsite(address pc,
     int malloc_diff, int malloc_count_diff) {
   if (malloc_diff != 0) {
     const char* unit = memory_unit(_scale);
-    char buf[64];
+    char buf[128];
     int  offset;
     if (pc == 0) {
       _output->print_cr("[BOOTSTRAP]%18s", " ");
diff --git a/src/share/vm/services/memReporter.hpp b/src/share/vm/services/memReporter.hpp
index 4595b2797..7a5372d49 100644
--- a/src/share/vm/services/memReporter.hpp
+++ b/src/share/vm/services/memReporter.hpp
@@ -93,6 +93,11 @@ class BaselineOutputer : public StackObj {
 
   virtual void done_category_summary() = 0;
 
+  virtual void start_virtual_memory_map() = 0;
+  virtual void reserved_memory_region(MEMFLAGS type, address base, address end, size_t size, address pc) = 0;
+  virtual void committed_memory_region(address base, address end, size_t size, address pc) = 0;
+  virtual void done_virtual_memory_map() = 0;
+
   /*
    *  Report callsite information
    */
@@ -136,6 +141,7 @@ class BaselineReporter : public StackObj {
 
  private:
   void report_summaries(const MemBaseline& baseline);
+  void report_virtual_memory_map(const MemBaseline& baseline);
   void report_callsites(const MemBaseline& baseline);
 
   void diff_summaries(const MemBaseline& cur, const MemBaseline& prev);
@@ -251,6 +257,13 @@ class BaselineTTYOutputer : public BaselineOutputer {
 
   void done_category_summary();
 
+  // virtual memory map
+  void start_virtual_memory_map();
+  void reserved_memory_region(MEMFLAGS type, address base, address end, size_t size, address pc);
+  void committed_memory_region(address base, address end, size_t size, address pc);
+  void done_virtual_memory_map();
+
+
   /*
    *  Report callsite information
    */
diff --git a/src/share/vm/services/memSnapshot.cpp b/src/share/vm/services/memSnapshot.cpp
index f86057694..20e1ce61f 100644
--- a/src/share/vm/services/memSnapshot.cpp
+++ b/src/share/vm/services/memSnapshot.cpp
@@ -31,6 +31,220 @@
 #include "services/memSnapshot.hpp"
 #include "services/memTracker.hpp"
 
+
+bool VMMemPointerIterator::insert_record(MemPointerRecord* rec) {
+  VMMemRegionEx new_rec;
+  assert(rec->is_allocation_record() || rec->is_commit_record(),
+    "Sanity check");
+  if (MemTracker::track_callsite()) {
+    new_rec.init((MemPointerRecordEx*)rec);
+  } else {
+    new_rec.init(rec);
+  }
+  return insert(&new_rec);
+}
+
+bool VMMemPointerIterator::insert_record_after(MemPointerRecord* rec) {
+  VMMemRegionEx new_rec;
+  assert(rec->is_allocation_record() || rec->is_commit_record(),
+    "Sanity check");
+  if (MemTracker::track_callsite()) {
+    new_rec.init((MemPointerRecordEx*)rec);
+  } else {
+    new_rec.init(rec);
+  }
+  return insert_after(&new_rec);
+}
+
+// we don't consolidate reserved regions, since they may be categorized
+// in different types.
+bool VMMemPointerIterator::add_reserved_region(MemPointerRecord* rec) {
+  assert(rec->is_allocation_record(), "Sanity check");
+  VMMemRegion* cur = (VMMemRegion*)current();
+
+  // we don't have anything yet
+  if (cur == NULL) {
+    return insert_record(rec);
+  }
+
+  assert(cur->is_reserved_region(), "Sanity check");
+  // duplicated records
+  if (cur->is_same_region(rec)) {
+    return true;
+  }
+  assert(cur->base() > rec->addr(), "Just check: locate()");
+  assert(rec->addr() + rec->size() <= cur->base(), "Can not overlap");
+  return insert_record(rec);
+}
+
+// we do consolidate committed regions
+bool VMMemPointerIterator::add_committed_region(MemPointerRecord* rec) {
+  assert(rec->is_commit_record(), "Sanity check");
+  VMMemRegion* cur;
+  cur = (VMMemRegion*)current();
+  assert(cur->is_reserved_region() && cur->contains_region(rec),
+    "Sanity check");
+
+  // thread's native stack is always marked as "committed", ignore
+  // the "commit" operation for creating stack guard pages
+  if (FLAGS_TO_MEMORY_TYPE(cur->flags()) == mtThreadStack &&
+      FLAGS_TO_MEMORY_TYPE(rec->flags()) != mtThreadStack) {
+    return true;
+  }
+
+  cur = (VMMemRegion*)next();
+  while (cur != NULL && cur->is_committed_region()) {
+    // duplicated commit records
+    if(cur->contains_region(rec)) {
+      return true;
+    }
+    if (cur->base() > rec->addr()) {
+      // committed regions can not overlap
+      assert(rec->addr() + rec->size() <= cur->base(), "Can not overlap");
+      if (rec->addr() + rec->size() == cur->base()) {
+        cur->expand_region(rec->addr(), rec->size());
+        return true;
+      } else {
+        return insert_record(rec);
+      }
+    } else if (cur->base() + cur->size() == rec->addr()) {
+      cur->expand_region(rec->addr(), rec->size());
+      VMMemRegion* next_reg = (VMMemRegion*)next();
+      // see if we can consolidate next committed region
+      if (next_reg != NULL && next_reg->is_committed_region() &&
+        next_reg->base() == cur->base() + cur->size()) {
+          cur->expand_region(next_reg->base(), next_reg->size());
+          remove();
+      }
+      return true;
+    }
+    cur = (VMMemRegion*)next();
+  }
+  return insert_record(rec);
+}
+
+bool VMMemPointerIterator::remove_uncommitted_region(MemPointerRecord* rec) {
+  assert(rec->is_uncommit_record(), "sanity check");
+  VMMemRegion* cur;
+  cur = (VMMemRegion*)current();
+  assert(cur->is_reserved_region() && cur->contains_region(rec),
+    "Sanity check");
+  // thread's native stack is always marked as "committed", ignore
+  // the "commit" operation for creating stack guard pages
+  if (FLAGS_TO_MEMORY_TYPE(cur->flags()) == mtThreadStack &&
+      FLAGS_TO_MEMORY_TYPE(rec->flags()) != mtThreadStack) {
+    return true;
+  }
+
+  cur = (VMMemRegion*)next();
+  while (cur != NULL && cur->is_committed_region()) {
+    // region already uncommitted, must be due to duplicated record
+    if (cur->addr() >= rec->addr() + rec->size()) {
+      break;
+    } else if (cur->contains_region(rec)) {
+      // uncommit whole region
+      if (cur->is_same_region(rec)) {
+        remove();
+        break;
+      } else if (rec->addr() == cur->addr() ||
+        rec->addr() + rec->size() == cur->addr() + cur->size()) {
+        // uncommitted from either end of current memory region.
+        cur->exclude_region(rec->addr(), rec->size());
+        break;
+      } else { // split the committed region and release the middle
+        address high_addr = cur->addr() + cur->size();
+        size_t sz = high_addr - rec->addr();
+        cur->exclude_region(rec->addr(), sz);
+        sz = high_addr - (rec->addr() + rec->size());
+        if (MemTracker::track_callsite()) {
+          MemPointerRecordEx tmp(rec->addr() + rec->size(), cur->flags(), sz,
+             ((VMMemRegionEx*)cur)->pc());
+          return insert_record_after(&tmp);
+        } else {
+          MemPointerRecord tmp(rec->addr() + rec->size(), cur->flags(), sz);
+          return insert_record_after(&tmp);
+        }
+      }
+    }
+    cur = (VMMemRegion*)next();
+  }
+
+  // we may not find committed record due to duplicated records
+  return true;
+}
+
+bool VMMemPointerIterator::remove_released_region(MemPointerRecord* rec) {
+  assert(rec->is_deallocation_record(), "Sanity check");
+  VMMemRegion* cur = (VMMemRegion*)current();
+  assert(cur->is_reserved_region() && cur->contains_region(rec),
+    "Sanity check");
+#ifdef ASSERT
+  VMMemRegion* next_reg = (VMMemRegion*)peek_next();
+  // should not have any committed memory in this reserved region
+  assert(next_reg == NULL || !next_reg->is_committed_region(), "Sanity check");
+#endif
+  if (rec->is_same_region(cur)) {
+    remove();
+  } else if (rec->addr() == cur->addr() ||
+    rec->addr() + rec->size() == cur->addr() + cur->size()) {
+    // released region is at either end of this region
+    cur->exclude_region(rec->addr(), rec->size());
+  } else { // split the reserved region and release the middle
+    address high_addr = cur->addr() + cur->size();
+    size_t sz = high_addr - rec->addr();
+    cur->exclude_region(rec->addr(), sz);
+    sz = high_addr - rec->addr() - rec->size();
+    if (MemTracker::track_callsite()) {
+      MemPointerRecordEx tmp(rec->addr() + rec->size(), cur->flags(), sz,
+        ((VMMemRegionEx*)cur)->pc());
+      return insert_reserved_region(&tmp);
+    } else {
+      MemPointerRecord tmp(rec->addr() + rec->size(), cur->flags(), sz);
+      return insert_reserved_region(&tmp);
+    }
+  }
+  return true;
+}
+
+bool VMMemPointerIterator::insert_reserved_region(MemPointerRecord* rec) {
+  // skip all 'commit' records associated with previous reserved region
+  VMMemRegion* p = (VMMemRegion*)next();
+  while (p != NULL && p->is_committed_region() &&
+         p->base() + p->size() < rec->addr()) {
+    p = (VMMemRegion*)next();
+  }
+  return insert_record(rec);
+}
+
+bool VMMemPointerIterator::split_reserved_region(VMMemRegion* rgn, address new_rgn_addr, size_t new_rgn_size) {
+  assert(rgn->contains_region(new_rgn_addr, new_rgn_size), "Not fully contained");
+  address pc = (MemTracker::track_callsite() ? ((VMMemRegionEx*)rgn)->pc() : NULL);
+  if (rgn->base() == new_rgn_addr) { // new region is at the beginning of the region
+    size_t sz = rgn->size() - new_rgn_size;
+    // the original region becomes 'new' region
+    rgn->exclude_region(new_rgn_addr + new_rgn_size, sz);
+     // remaining becomes next region
+    MemPointerRecordEx next_rgn(new_rgn_addr + new_rgn_size, rgn->flags(), sz, pc);
+    return insert_reserved_region(&next_rgn);
+  } else if (rgn->base() + rgn->size() == new_rgn_addr + new_rgn_size) {
+    rgn->exclude_region(new_rgn_addr, new_rgn_size);
+    MemPointerRecordEx next_rgn(new_rgn_addr, rgn->flags(), new_rgn_size, pc);
+    return insert_reserved_region(&next_rgn);
+  } else {
+    // the orginal region will be split into three
+    address rgn_high_addr = rgn->base() + rgn->size();
+    // first region
+    rgn->exclude_region(new_rgn_addr, (rgn_high_addr - new_rgn_addr));
+    // the second region is the new region
+    MemPointerRecordEx new_rgn(new_rgn_addr, rgn->flags(), new_rgn_size, pc);
+    if (!insert_reserved_region(&new_rgn)) return false;
+    // the remaining region
+    MemPointerRecordEx rem_rgn(new_rgn_addr + new_rgn_size, rgn->flags(),
+      rgn_high_addr - (new_rgn_addr + new_rgn_size), pc);
+    return insert_reserved_region(&rem_rgn);
+  }
+}
+
 static int sort_in_seq_order(const void* p1, const void* p2) {
   assert(p1 != NULL && p2 != NULL, "Sanity check");
   const MemPointerRecord* mp1 = (MemPointerRecord*)p1;
@@ -61,11 +275,11 @@ bool StagingArea::init() {
 }
 
 
-MemPointerArrayIteratorImpl StagingArea::virtual_memory_record_walker() {
+VMRecordIterator StagingArea::virtual_memory_record_walker() {
   MemPointerArray* arr = vm_data();
   // sort into seq number order
   arr->sort((FN_SORT)sort_in_seq_order);
-  return MemPointerArrayIteratorImpl(arr);
+  return VMRecordIterator(arr);
 }
 
 
@@ -135,6 +349,8 @@ bool MemSnapshot::merge(MemRecorder* rec) {
         return false;
       }
     } else {
+      // locate matched record and/or also position the iterator to proper
+      // location for this incoming record.
       p2 = (MemPointerRecord*)malloc_staging_itr.locate(p1->addr());
       // we have not seen this memory block, so just add to staging area
       if (p2 == NULL) {
@@ -199,7 +415,7 @@ bool MemSnapshot::promote() {
   MallocRecordIterator  malloc_itr = _staging_area.malloc_record_walker();
   bool promoted = false;
   if (promote_malloc_records(&malloc_itr)) {
-    MemPointerArrayIteratorImpl vm_itr = _staging_area.virtual_memory_record_walker();
+    VMRecordIterator vm_itr = _staging_area.virtual_memory_record_walker();
     if (promote_virtual_memory_records(&vm_itr)) {
       promoted = true;
     }
@@ -218,7 +434,7 @@ bool MemSnapshot::promote_malloc_records(MemPointerArrayIterator* itr) {
     matched_rec = (MemPointerRecord*)malloc_snapshot_itr.locate(new_rec->addr());
     // found matched memory block
     if (matched_rec != NULL && new_rec->addr() == matched_rec->addr()) {
-      // snapshot already contains 'lived' records
+      // snapshot already contains 'live' records
       assert(matched_rec->is_allocation_record() || matched_rec->is_arena_size_record(),
              "Sanity check");
       // update block states
@@ -277,87 +493,60 @@ bool MemSnapshot::promote_malloc_records(MemPointerArrayIterator* itr) {
 bool MemSnapshot::promote_virtual_memory_records(MemPointerArrayIterator* itr) {
   VMMemPointerIterator vm_snapshot_itr(_vm_ptrs);
   MemPointerRecord* new_rec = (MemPointerRecord*)itr->current();
-  VMMemRegionEx new_vm_rec;
-  VMMemRegion*  matched_rec;
+  VMMemRegion*  reserved_rec;
   while (new_rec != NULL) {
     assert(new_rec->is_vm_pointer(), "Sanity check");
-    if (MemTracker::track_callsite()) {
-      new_vm_rec.init((MemPointerRecordEx*)new_rec);
-    } else {
-      new_vm_rec.init(new_rec);
-    }
-    matched_rec = (VMMemRegion*)vm_snapshot_itr.locate(new_rec->addr());
-    if (matched_rec != NULL &&
-        (matched_rec->contains(&new_vm_rec) || matched_rec->base() == new_vm_rec.base())) {
+
+    // locate a reserved region that contains the specified address, or
+    // the nearest reserved region has base address just above the specified
+    // address
+    reserved_rec = (VMMemRegion*)vm_snapshot_itr.locate(new_rec->addr());
+    if (reserved_rec != NULL && reserved_rec->contains_region(new_rec)) {
       // snapshot can only have 'live' records
-      assert(matched_rec->is_reserve_record(), "Sanity check");
-      if (new_vm_rec.is_reserve_record() && matched_rec->base() == new_vm_rec.base()) {
-        // resize reserved virtual memory range
-        // resize has to cover committed area
-        assert(new_vm_rec.size() >= matched_rec->committed_size(), "Sanity check");
-        matched_rec->set_reserved_size(new_vm_rec.size());
-      } else if (new_vm_rec.is_commit_record()) {
-        // commit memory inside reserved memory range
-        assert(new_vm_rec.committed_size() <= matched_rec->reserved_size(), "Sanity check");
-        // thread stacks are marked committed, so we ignore 'commit' record for creating
-        // stack guard pages
-        if (FLAGS_TO_MEMORY_TYPE(matched_rec->flags()) != mtThreadStack) {
-          matched_rec->commit(new_vm_rec.committed_size());
-        }
-      } else if (new_vm_rec.is_uncommit_record()) {
-        if (FLAGS_TO_MEMORY_TYPE(matched_rec->flags()) == mtThreadStack) {
-          // ignore 'uncommit' record from removing stack guard pages, uncommit
-          // thread stack as whole
-          if (matched_rec->committed_size() == new_vm_rec.committed_size()) {
-            matched_rec->uncommit(new_vm_rec.committed_size());
+      assert(reserved_rec->is_reserved_region(), "Sanity check");
+      if (new_rec->is_allocation_record()) {
+        if (!reserved_rec->is_same_region(new_rec)) {
+          // only deal with split a bigger reserved region into smaller regions.
+          // So far, CDS is the only use case.
+          if (!vm_snapshot_itr.split_reserved_region(reserved_rec, new_rec->addr(), new_rec->size())) {
+            return false;
           }
-        } else {
-          // uncommit memory inside reserved memory range
-          assert(new_vm_rec.committed_size() <= matched_rec->committed_size(),
-                "Sanity check");
-          matched_rec->uncommit(new_vm_rec.committed_size());
         }
-      } else if (new_vm_rec.is_type_tagging_record()) {
-        // tag this virtual memory range to a memory type
-        // can not re-tag a memory range to different type
-        assert(FLAGS_TO_MEMORY_TYPE(matched_rec->flags()) == mtNone ||
-               FLAGS_TO_MEMORY_TYPE(matched_rec->flags()) == FLAGS_TO_MEMORY_TYPE(new_vm_rec.flags()),
-               "Sanity check");
-        matched_rec->tag(new_vm_rec.flags());
-      } else if (new_vm_rec.is_release_record()) {
-        // release part or whole memory range
-        if (new_vm_rec.base() == matched_rec->base() &&
-            new_vm_rec.size() == matched_rec->size()) {
-          // release whole virtual memory range
-          assert(matched_rec->committed_size() == 0, "Sanity check");
-          vm_snapshot_itr.remove();
-        } else {
-          // partial release
-          matched_rec->partial_release(new_vm_rec.base(), new_vm_rec.size());
+      } else if (new_rec->is_uncommit_record()) {
+        if (!vm_snapshot_itr.remove_uncommitted_region(new_rec)) {
+          return false;
         }
-      } else {
-        // multiple reserve/commit on the same virtual memory range
-        assert((new_vm_rec.is_reserve_record() || new_vm_rec.is_commit_record()) &&
-          (new_vm_rec.base() == matched_rec->base() && new_vm_rec.size() == matched_rec->size()),
-          "Sanity check");
-        matched_rec->tag(new_vm_rec.flags());
-      }
+      } else if (new_rec->is_commit_record()) {
+        // insert or expand existing committed region to cover this
+        // newly committed region
+        if (!vm_snapshot_itr.add_committed_region(new_rec)) {
+          return false;
+        }
+      } else if (new_rec->is_deallocation_record()) {
+        // release part or all memory region
+        if (!vm_snapshot_itr.remove_released_region(new_rec)) {
+          return false;
+        }
+      } else if (new_rec->is_type_tagging_record()) {
+        // tag this reserved virtual memory range to a memory type. Can not re-tag a memory range
+        // to different type.
+        assert(FLAGS_TO_MEMORY_TYPE(reserved_rec->flags()) == mtNone ||
+               FLAGS_TO_MEMORY_TYPE(reserved_rec->flags()) == FLAGS_TO_MEMORY_TYPE(new_rec->flags()),
+               "Sanity check");
+        reserved_rec->tag(new_rec->flags());
     } else {
-      // no matched record
-      if (new_vm_rec.is_reserve_record()) {
-        if (matched_rec == NULL || matched_rec->base() > new_vm_rec.base()) {
-          if (!vm_snapshot_itr.insert(&new_vm_rec)) {
-            return false;
+        ShouldNotReachHere();
           }
         } else {
-          if (!vm_snapshot_itr.insert_after(&new_vm_rec)) {
+      /*
+       * The assertion failure indicates mis-matched virtual memory records. The likely
+       * scenario is, that some virtual memory operations are not going through os::xxxx_memory()
+       * api, which have to be tracked manually. (perfMemory is an example).
+      */
+      assert(new_rec->is_allocation_record(), "Sanity check");
+      if (!vm_snapshot_itr.add_reserved_region(new_rec)) {
             return false;
           }
-        }
-      } else {
-        // throw out obsolete records, which are the commit/uncommit/release/tag records
-        // on memory regions that are already released.
-      }
   }
     new_rec = (MemPointerRecord*)itr->next();
   }
@@ -433,5 +622,33 @@ void MemSnapshot::check_staging_data() {
     cur = (MemPointerRecord*)vm_itr.next();
   }
 }
+
+void MemSnapshot::dump_all_vm_pointers() {
+  MemPointerArrayIteratorImpl itr(_vm_ptrs);
+  VMMemRegion* ptr = (VMMemRegion*)itr.current();
+  tty->print_cr("dump virtual memory pointers:");
+  while (ptr != NULL) {
+    if (ptr->is_committed_region()) {
+      tty->print("\t");
+    }
+    tty->print("[" PTR_FORMAT " - " PTR_FORMAT "] [%x]", ptr->addr(),
+      (ptr->addr() + ptr->size()), ptr->flags());
+
+    if (MemTracker::track_callsite()) {
+      VMMemRegionEx* ex = (VMMemRegionEx*)ptr;
+      if (ex->pc() != NULL) {
+        char buf[1024];
+        if (os::dll_address_to_function_name(ex->pc(), buf, sizeof(buf), NULL)) {
+          tty->print_cr("\t%s", buf);
+        } else {
+          tty->print_cr("");
+        }
+      }
+    }
+
+    ptr = (VMMemRegion*)itr.next();
+  }
+  tty->flush();
+}
 #endif // ASSERT
 
diff --git a/src/share/vm/services/memSnapshot.hpp b/src/share/vm/services/memSnapshot.hpp
index 9ac6e4cf3..dd52f4cd7 100644
--- a/src/share/vm/services/memSnapshot.hpp
+++ b/src/share/vm/services/memSnapshot.hpp
@@ -111,33 +111,41 @@ class VMMemPointerIterator : public MemPointerIterator {
       MemPointerIterator(arr) {
   }
 
-  // locate an existing record that contains specified address, or
-  // the record, where the record with specified address, should
-  // be inserted.
-  // virtual memory record array is sorted in address order, so
-  // binary search is performed
+  // locate an existing reserved memory region that contains specified address,
+  // or the reserved region just above this address, where the incoming
+  // reserved region should be inserted.
   virtual MemPointer* locate(address addr) {
-    int index_low = 0;
-    int index_high = _array->length();
-    int index_mid = (index_high + index_low) / 2;
-    int r = 1;
-    while (index_low < index_high && (r = compare(index_mid, addr)) != 0) {
-      if (r > 0) {
-        index_high = index_mid;
-      } else {
-        index_low = index_mid;
+    reset();
+    VMMemRegion* reg = (VMMemRegion*)current();
+    while (reg != NULL) {
+      if (reg->is_reserved_region()) {
+        if (reg->contains_address(addr) || addr < reg->base()) {
+          return reg;
       }
-      index_mid = (index_high + index_low) / 2;
     }
-    if (r == 0) {
-      // update current location
-      _pos = index_mid;
-      return _array->at(index_mid);
-    } else {
+      reg = (VMMemRegion*)next();
+    }
       return NULL;
     }
-  }
 
+  // following methods update virtual memory in the context
+  // of 'current' position, which is properly positioned by
+  // callers via locate method.
+  bool add_reserved_region(MemPointerRecord* rec);
+  bool add_committed_region(MemPointerRecord* rec);
+  bool remove_uncommitted_region(MemPointerRecord* rec);
+  bool remove_released_region(MemPointerRecord* rec);
+
+  // split a reserved region to create a new memory region with specified base and size
+  bool split_reserved_region(VMMemRegion* rgn, address new_rgn_addr, size_t new_rgn_size);
+ private:
+  bool insert_record(MemPointerRecord* rec);
+  bool insert_record_after(MemPointerRecord* rec);
+
+  bool insert_reserved_region(MemPointerRecord* rec);
+
+  // reset current position
+  inline void reset() { _pos = 0; }
 #ifdef ASSERT
   virtual bool is_dup_pointer(const MemPointer* ptr1,
     const MemPointer* ptr2) const {
@@ -154,32 +162,17 @@ class VMMemPointerIterator : public MemPointerIterator {
            (p1->flags() & MemPointerRecord::tag_masks) == MemPointerRecord::tag_release;
   }
 #endif
-  // compare if an address falls into a memory region,
-  // return 0, if the address falls into a memory region at specified index
-  // return 1, if memory region pointed by specified index is higher than the address
-  // return -1, if memory region pointed by specified index is lower than the address
-  int compare(int index, address addr) const {
-    VMMemRegion* r = (VMMemRegion*)_array->at(index);
-    assert(r->is_reserve_record(), "Sanity check");
-    if (r->addr() > addr) {
-      return 1;
-    } else if (r->addr() + r->reserved_size() <= addr) {
-      return -1;
-    } else {
-      return 0;
-    }
-  }
 };
 
 class MallocRecordIterator : public MemPointerArrayIterator {
- private:
+ protected:
   MemPointerArrayIteratorImpl  _itr;
 
  public:
   MallocRecordIterator(MemPointerArray* arr) : _itr(arr) {
   }
 
-  MemPointer* current() const {
+  virtual MemPointer* current() const {
     MemPointerRecord* cur = (MemPointerRecord*)_itr.current();
     assert(cur == NULL || !cur->is_vm_pointer(), "seek error");
     MemPointerRecord* next = (MemPointerRecord*)_itr.peek_next();
@@ -194,7 +187,7 @@ class MallocRecordIterator : public MemPointerArrayIterator {
     }
   }
 
-  MemPointer* next() {
+  virtual MemPointer* next() {
     MemPointerRecord* cur = (MemPointerRecord*)_itr.current();
     assert(cur == NULL || !cur->is_vm_pointer(), "Sanity check");
     MemPointerRecord* next = (MemPointerRecord*)_itr.next();
@@ -214,6 +207,63 @@ class MallocRecordIterator : public MemPointerArrayIterator {
   bool insert_after(MemPointer* ptr) { ShouldNotReachHere(); return false; }
 };
 
+// collapse duplicated records. Eliminating duplicated records here, is much
+// cheaper than during promotion phase. However, it does have limitation - it
+// can only eliminate duplicated records within the generation, there are
+// still chances seeing duplicated records during promotion.
+// We want to use the record with higher sequence number, because it has
+// more accurate callsite pc.
+class VMRecordIterator : public MallocRecordIterator {
+ public:
+  VMRecordIterator(MemPointerArray* arr) : MallocRecordIterator(arr) {
+    MemPointerRecord* cur = (MemPointerRecord*)_itr.current();
+    MemPointerRecord* next = (MemPointerRecord*)_itr.peek_next();
+    while (next != NULL) {
+      assert(cur != NULL, "Sanity check");
+      assert(((SeqMemPointerRecord*)next)->seq() > ((SeqMemPointerRecord*)cur)->seq(),
+        "pre-sort order");
+
+      if (is_duplicated_record(cur, next)) {
+        _itr.next();
+        next = (MemPointerRecord*)_itr.peek_next();
+      } else {
+        break;
+      }
+    }
+  }
+
+  virtual MemPointer* current() const {
+    return _itr.current();
+  }
+
+  // get next record, but skip the duplicated records
+  virtual MemPointer* next() {
+    MemPointerRecord* cur = (MemPointerRecord*)_itr.next();
+    MemPointerRecord* next = (MemPointerRecord*)_itr.peek_next();
+    while (next != NULL) {
+      assert(cur != NULL, "Sanity check");
+      assert(((SeqMemPointerRecord*)next)->seq() > ((SeqMemPointerRecord*)cur)->seq(),
+        "pre-sort order");
+
+      if (is_duplicated_record(cur, next)) {
+        _itr.next();
+        cur = next;
+        next = (MemPointerRecord*)_itr.peek_next();
+      } else {
+        break;
+      }
+    }
+    return cur;
+  }
+
+ private:
+  bool is_duplicated_record(MemPointerRecord* p1, MemPointerRecord* p2) const {
+    bool ret = (p1->addr() == p2->addr() && p1->size() == p2->size() && p1->flags() == p2->flags());
+    assert(!(ret && FLAGS_TO_MEMORY_TYPE(p1->flags()) == mtThreadStack), "dup on stack record");
+    return ret;
+  }
+};
+
 class StagingArea : public _ValueObj {
  private:
   MemPointerArray*   _malloc_data;
@@ -233,7 +283,8 @@ class StagingArea : public _ValueObj {
     return MallocRecordIterator(malloc_data());
   }
 
-  MemPointerArrayIteratorImpl virtual_memory_record_walker();
+  VMRecordIterator virtual_memory_record_walker();
+
   bool init();
   void clear() {
     assert(_malloc_data != NULL && _vm_data != NULL, "Just check");
@@ -293,6 +344,8 @@ class MemSnapshot : public CHeapObj<mtNMT> {
   NOT_PRODUCT(void check_staging_data();)
   NOT_PRODUCT(void check_malloc_pointers();)
   NOT_PRODUCT(bool has_allocation_record(address addr);)
+  // dump all virtual memory pointers in snapshot
+  DEBUG_ONLY( void dump_all_vm_pointers();)
 
  private:
    // copy pointer data from src to dest
@@ -302,5 +355,4 @@ class MemSnapshot : public CHeapObj<mtNMT> {
    bool promote_virtual_memory_records(MemPointerArrayIterator* itr);
 };
 
-
 #endif // SHARE_VM_SERVICES_MEM_SNAPSHOT_HPP
diff --git a/src/share/vm/services/memTracker.cpp b/src/share/vm/services/memTracker.cpp
index 2af6e10d6..3cec67868 100644
--- a/src/share/vm/services/memTracker.cpp
+++ b/src/share/vm/services/memTracker.cpp
@@ -364,7 +364,7 @@ void MemTracker::create_memory_record(address addr, MEMFLAGS flags,
 
     if (thread != NULL) {
       if (thread->is_Java_thread() && ((JavaThread*)thread)->is_safepoint_visible()) {
-        JavaThread*      java_thread = static_cast<JavaThread*>(thread);
+        JavaThread*      java_thread = (JavaThread*)thread;
         JavaThreadState  state = java_thread->thread_state();
         if (SafepointSynchronize::safepoint_safe(java_thread, state)) {
           // JavaThreads that are safepoint safe, can run through safepoint,
@@ -472,6 +472,8 @@ void MemTracker::sync() {
       // it should guarantee that NMT is fully sync-ed.
       ThreadCritical tc;
 
+      SequenceGenerator::reset();
+
       // walk all JavaThreads to collect recorders
       SyncThreadRecorderClosure stc;
       Threads::threads_do(&stc);
@@ -484,11 +486,12 @@ void MemTracker::sync() {
         pending_recorders = _global_recorder;
         _global_recorder = NULL;
       }
-      SequenceGenerator::reset();
       // check _worker_thread with lock to avoid racing condition
       if (_worker_thread != NULL) {
         _worker_thread->at_sync_point(pending_recorders);
       }
+
+      assert(SequenceGenerator::peek() == 1, "Should not have memory activities during sync-point");
     }
   }
 
diff --git a/src/share/vm/services/memTracker.hpp b/src/share/vm/services/memTracker.hpp
index 62637d8a8..c42f01bf6 100644
--- a/src/share/vm/services/memTracker.hpp
+++ b/src/share/vm/services/memTracker.hpp
@@ -113,8 +113,10 @@ class MemTracker : AllStatic {
 #include "thread_solaris.inline.hpp"
 #endif
 
-#ifdef _DEBUG
-  #define DEBUG_CALLER_PC  os::get_caller_pc(3)
+extern bool NMT_track_callsite;
+
+#ifdef ASSERT
+  #define DEBUG_CALLER_PC  (NMT_track_callsite ? os::get_caller_pc(2) : 0)
 #else
   #define DEBUG_CALLER_PC  0
 #endif
@@ -261,7 +263,7 @@ class MemTracker : AllStatic {
   // record a 'malloc' call
   static inline void record_malloc(address addr, size_t size, MEMFLAGS flags,
                             address pc = 0, Thread* thread = NULL) {
-    if (NMT_CAN_TRACK(flags)) {
+    if (is_on() && NMT_CAN_TRACK(flags)) {
       assert(size > 0, "Sanity check");
       create_memory_record(addr, (flags|MemPointerRecord::malloc_tag()), size, pc, thread);
     }
@@ -275,7 +277,7 @@ class MemTracker : AllStatic {
   // record a 'realloc' call
   static inline void record_realloc(address old_addr, address new_addr, size_t size,
        MEMFLAGS flags, address pc = 0, Thread* thread = NULL) {
-    if (is_on()) {
+    if (is_on() && NMT_CAN_TRACK(flags)) {
       assert(size > 0, "Sanity check");
       record_free(old_addr, flags, thread);
       record_malloc(new_addr, size, flags, pc, thread);
@@ -317,6 +319,7 @@ class MemTracker : AllStatic {
   static inline void release_thread_stack(address addr, size_t size, Thread* thr) {
     if (is_on()) {
       assert(size > 0 && thr != NULL, "Sanity check");
+      assert(!thr->is_Java_thread(), "too early");
       create_memory_record(addr, MemPointerRecord::virtual_memory_uncommit_tag() | mtThreadStack,
                           size, DEBUG_CALLER_PC, thr);
       create_memory_record(addr, MemPointerRecord::virtual_memory_release_tag() | mtThreadStack,
@@ -326,11 +329,11 @@ class MemTracker : AllStatic {
 
   // record a virtual memory 'commit' call
   static inline void record_virtual_memory_commit(address addr, size_t size,
-                            address pc = 0, Thread* thread = NULL) {
+                            address pc, Thread* thread = NULL) {
     if (is_on()) {
       assert(size > 0, "Sanity check");
       create_memory_record(addr, MemPointerRecord::virtual_memory_commit_tag(),
-                           size, DEBUG_CALLER_PC, thread);
+                           size, pc, thread);
     }
   }
 
diff --git a/test/compiler/6340864/TestByteVect.java b/test/compiler/6340864/TestByteVect.java
index ec4ba9fcd..5db3687e6 100644
--- a/test/compiler/6340864/TestByteVect.java
+++ b/test/compiler/6340864/TestByteVect.java
@@ -33,7 +33,7 @@
 public class TestByteVect {
   private static final int ARRLEN = 997;
   private static final int ITERS  = 11000;
-  private static final int ADD_INIT = 0;
+  private static final int ADD_INIT = 63;
   private static final int BIT_MASK = 0xB7;
   private static final int VALUE = 3;
   private static final int SHIFT = 8;
@@ -76,6 +76,7 @@ public class TestByteVect {
       test_subc(a0, a1);
       test_subv(a0, a1, (byte)VALUE);
       test_suba(a0, a1, a2);
+
       test_mulc(a0, a1);
       test_mulv(a0, a1, (byte)VALUE);
       test_mula(a0, a1, a2);
@@ -88,6 +89,7 @@ public class TestByteVect {
       test_divc_n(a0, a1);
       test_divv(a0, a1, (byte)-VALUE);
       test_diva(a0, a1, a3);
+
       test_andc(a0, a1);
       test_andv(a0, a1, (byte)BIT_MASK);
       test_anda(a0, a1, a4);
@@ -97,30 +99,49 @@ public class TestByteVect {
       test_xorc(a0, a1);
       test_xorv(a0, a1, (byte)BIT_MASK);
       test_xora(a0, a1, a4);
+
       test_sllc(a0, a1);
       test_sllv(a0, a1, VALUE);
       test_srlc(a0, a1);
       test_srlv(a0, a1, VALUE);
       test_srac(a0, a1);
       test_srav(a0, a1, VALUE);
+
       test_sllc_n(a0, a1);
       test_sllv(a0, a1, -VALUE);
       test_srlc_n(a0, a1);
       test_srlv(a0, a1, -VALUE);
       test_srac_n(a0, a1);
       test_srav(a0, a1, -VALUE);
+
       test_sllc_o(a0, a1);
       test_sllv(a0, a1, SHIFT);
       test_srlc_o(a0, a1);
       test_srlv(a0, a1, SHIFT);
       test_srac_o(a0, a1);
       test_srav(a0, a1, SHIFT);
+
       test_sllc_on(a0, a1);
       test_sllv(a0, a1, -SHIFT);
       test_srlc_on(a0, a1);
       test_srlv(a0, a1, -SHIFT);
       test_srac_on(a0, a1);
       test_srav(a0, a1, -SHIFT);
+
+      test_sllc_add(a0, a1);
+      test_sllv_add(a0, a1, ADD_INIT);
+      test_srlc_add(a0, a1);
+      test_srlv_add(a0, a1, ADD_INIT);
+      test_srac_add(a0, a1);
+      test_srav_add(a0, a1, ADD_INIT);
+
+      test_sllc_and(a0, a1);
+      test_sllv_and(a0, a1, BIT_MASK);
+      test_srlc_and(a0, a1);
+      test_srlv_and(a0, a1, BIT_MASK);
+      test_srac_and(a0, a1);
+      test_srav_and(a0, a1, BIT_MASK);
+
       test_pack2(p2, a1);
       test_unpack2(a0, p2);
       test_pack2_swap(p2, a1);
@@ -369,6 +390,60 @@ public class TestByteVect {
         errn += verify("test_srav_on: ", i, a0[i], (byte)((byte)(ADD_INIT+i)>>(-SHIFT)));
       }
 
+      test_sllc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+      test_sllv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+
+      test_srlc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+      test_srlv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+
+      test_srac_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+      test_srav_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_add: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+
+      test_sllc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+      test_sllv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+
+      test_srlc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+      test_srlv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+
+      test_srac_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+      test_srav_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_and: ", i, a0[i], (byte)(((byte)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+
       test_pack2(p2, a1);
       for (int i=0; i<ARRLEN/2; i++) {
         errn += verify("test_pack2: ", i, p2[i], (short)(((short)(ADD_INIT+2*i) & 0xFF) | ((short)(ADD_INIT+2*i+1) << 8)));
@@ -805,6 +880,84 @@ public class TestByteVect {
 
     start = System.currentTimeMillis();
     for (int i=0; i<ITERS; i++) {
+      test_sllc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
       test_pack2(p2, a1);
     }
     end = System.currentTimeMillis();
@@ -1036,6 +1189,26 @@ public class TestByteVect {
       a0[i] = (byte)(a1[i]<<b);
     }
   }
+  static void test_sllc_add(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + ADD_INIT)<<VALUE);
+    }
+  }
+  static void test_sllv_add(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + b)<<VALUE);
+    }
+  }
+  static void test_sllc_and(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & BIT_MASK)<<VALUE);
+    }
+  }
+  static void test_sllv_and(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & b)<<VALUE);
+    }
+  }
 
   static void test_srlc(byte[] a0, byte[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -1062,6 +1235,26 @@ public class TestByteVect {
       a0[i] = (byte)(a1[i]>>>b);
     }
   }
+  static void test_srlc_add(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + ADD_INIT)>>>VALUE);
+    }
+  }
+  static void test_srlv_add(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + b)>>>VALUE);
+    }
+  }
+  static void test_srlc_and(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & BIT_MASK)>>>VALUE);
+    }
+  }
+  static void test_srlv_and(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & b)>>>VALUE);
+    }
+  }
 
   static void test_srac(byte[] a0, byte[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -1088,6 +1281,26 @@ public class TestByteVect {
       a0[i] = (byte)(a1[i]>>b);
     }
   }
+  static void test_srac_add(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + ADD_INIT)>>VALUE);
+    }
+  }
+  static void test_srav_add(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] + b)>>VALUE);
+    }
+  }
+  static void test_srac_and(byte[] a0, byte[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & BIT_MASK)>>VALUE);
+    }
+  }
+  static void test_srav_and(byte[] a0, byte[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (byte)((a1[i] & b)>>VALUE);
+    }
+  }
 
   static void test_pack2(short[] p2, byte[] a1) {
     if (p2.length*2 > a1.length) return;
diff --git a/test/compiler/6340864/TestIntVect.java b/test/compiler/6340864/TestIntVect.java
index 36e277f73..5866b34a5 100644
--- a/test/compiler/6340864/TestIntVect.java
+++ b/test/compiler/6340864/TestIntVect.java
@@ -74,6 +74,7 @@ public class TestIntVect {
       test_subc(a0, a1);
       test_subv(a0, a1, (int)VALUE);
       test_suba(a0, a1, a2);
+
       test_mulc(a0, a1);
       test_mulv(a0, a1, (int)VALUE);
       test_mula(a0, a1, a2);
@@ -86,6 +87,7 @@ public class TestIntVect {
       test_divc_n(a0, a1);
       test_divv(a0, a1, (int)-VALUE);
       test_diva(a0, a1, a3);
+
       test_andc(a0, a1);
       test_andv(a0, a1, (int)BIT_MASK);
       test_anda(a0, a1, a4);
@@ -95,30 +97,49 @@ public class TestIntVect {
       test_xorc(a0, a1);
       test_xorv(a0, a1, (int)BIT_MASK);
       test_xora(a0, a1, a4);
+
       test_sllc(a0, a1);
       test_sllv(a0, a1, VALUE);
       test_srlc(a0, a1);
       test_srlv(a0, a1, VALUE);
       test_srac(a0, a1);
       test_srav(a0, a1, VALUE);
+
       test_sllc_n(a0, a1);
       test_sllv(a0, a1, -VALUE);
       test_srlc_n(a0, a1);
       test_srlv(a0, a1, -VALUE);
       test_srac_n(a0, a1);
       test_srav(a0, a1, -VALUE);
+
       test_sllc_o(a0, a1);
       test_sllv(a0, a1, SHIFT);
       test_srlc_o(a0, a1);
       test_srlv(a0, a1, SHIFT);
       test_srac_o(a0, a1);
       test_srav(a0, a1, SHIFT);
+
       test_sllc_on(a0, a1);
       test_sllv(a0, a1, -SHIFT);
       test_srlc_on(a0, a1);
       test_srlv(a0, a1, -SHIFT);
       test_srac_on(a0, a1);
       test_srav(a0, a1, -SHIFT);
+
+      test_sllc_add(a0, a1);
+      test_sllv_add(a0, a1, ADD_INIT);
+      test_srlc_add(a0, a1);
+      test_srlv_add(a0, a1, ADD_INIT);
+      test_srac_add(a0, a1);
+      test_srav_add(a0, a1, ADD_INIT);
+
+      test_sllc_and(a0, a1);
+      test_sllv_and(a0, a1, BIT_MASK);
+      test_srlc_and(a0, a1);
+      test_srlv_and(a0, a1, BIT_MASK);
+      test_srac_and(a0, a1);
+      test_srav_and(a0, a1, BIT_MASK);
+
       test_pack2(p2, a1);
       test_unpack2(a0, p2);
       test_pack2_swap(p2, a1);
@@ -359,6 +380,60 @@ public class TestIntVect {
         errn += verify("test_srav_on: ", i, a0[i], (int)((int)(ADD_INIT+i)>>(-SHIFT)));
       }
 
+      test_sllc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+      test_sllv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+
+      test_srlc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+      test_srlv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+
+      test_srac_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+      test_srav_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_add: ", i, a0[i], (int)(((int)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+
+      test_sllc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+      test_sllv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+
+      test_srlc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+      test_srlv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+
+      test_srac_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+      test_srav_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_and: ", i, a0[i], (int)(((int)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+
       test_pack2(p2, a1);
       for (int i=0; i<ARRLEN/2; i++) {
         errn += verify("test_pack2: ", i, p2[i], ((long)(ADD_INIT+2*i) & 0xFFFFFFFFl) | ((long)(ADD_INIT+2*i+1) << 32));
@@ -727,6 +802,84 @@ public class TestIntVect {
 
     start = System.currentTimeMillis();
     for (int i=0; i<ITERS; i++) {
+      test_sllc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
       test_pack2(p2, a1);
     }
     end = System.currentTimeMillis();
@@ -908,6 +1061,26 @@ public class TestIntVect {
       a0[i] = (int)(a1[i]<<b);
     }
   }
+  static void test_sllc_add(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + ADD_INIT)<<VALUE);
+    }
+  }
+  static void test_sllv_add(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + b)<<VALUE);
+    }
+  }
+  static void test_sllc_and(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & BIT_MASK)<<VALUE);
+    }
+  }
+  static void test_sllv_and(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & b)<<VALUE);
+    }
+  }
 
   static void test_srlc(int[] a0, int[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -934,6 +1107,26 @@ public class TestIntVect {
       a0[i] = (int)(a1[i]>>>b);
     }
   }
+  static void test_srlc_add(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + ADD_INIT)>>>VALUE);
+    }
+  }
+  static void test_srlv_add(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + b)>>>VALUE);
+    }
+  }
+  static void test_srlc_and(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & BIT_MASK)>>>VALUE);
+    }
+  }
+  static void test_srlv_and(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & b)>>>VALUE);
+    }
+  }
 
   static void test_srac(int[] a0, int[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -960,6 +1153,26 @@ public class TestIntVect {
       a0[i] = (int)(a1[i]>>b);
     }
   }
+  static void test_srac_add(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + ADD_INIT)>>VALUE);
+    }
+  }
+  static void test_srav_add(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] + b)>>VALUE);
+    }
+  }
+  static void test_srac_and(int[] a0, int[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & BIT_MASK)>>VALUE);
+    }
+  }
+  static void test_srav_and(int[] a0, int[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (int)((a1[i] & b)>>VALUE);
+    }
+  }
 
   static void test_pack2(long[] p2, int[] a1) {
     if (p2.length*2 > a1.length) return;
diff --git a/test/compiler/6340864/TestLongVect.java b/test/compiler/6340864/TestLongVect.java
index 70b41f4b4..436a8472d 100644
--- a/test/compiler/6340864/TestLongVect.java
+++ b/test/compiler/6340864/TestLongVect.java
@@ -73,6 +73,7 @@ public class TestLongVect {
       test_subc(a0, a1);
       test_subv(a0, a1, (long)VALUE);
       test_suba(a0, a1, a2);
+
       test_mulc(a0, a1);
       test_mulv(a0, a1, (long)VALUE);
       test_mula(a0, a1, a2);
@@ -85,6 +86,7 @@ public class TestLongVect {
       test_divc_n(a0, a1);
       test_divv(a0, a1, (long)-VALUE);
       test_diva(a0, a1, a3);
+
       test_andc(a0, a1);
       test_andv(a0, a1, (long)BIT_MASK);
       test_anda(a0, a1, a4);
@@ -94,30 +96,48 @@ public class TestLongVect {
       test_xorc(a0, a1);
       test_xorv(a0, a1, (long)BIT_MASK);
       test_xora(a0, a1, a4);
+
       test_sllc(a0, a1);
       test_sllv(a0, a1, VALUE);
       test_srlc(a0, a1);
       test_srlv(a0, a1, VALUE);
       test_srac(a0, a1);
       test_srav(a0, a1, VALUE);
+
       test_sllc_n(a0, a1);
       test_sllv(a0, a1, -VALUE);
       test_srlc_n(a0, a1);
       test_srlv(a0, a1, -VALUE);
       test_srac_n(a0, a1);
       test_srav(a0, a1, -VALUE);
+
       test_sllc_o(a0, a1);
       test_sllv(a0, a1, SHIFT);
       test_srlc_o(a0, a1);
       test_srlv(a0, a1, SHIFT);
       test_srac_o(a0, a1);
       test_srav(a0, a1, SHIFT);
+
       test_sllc_on(a0, a1);
       test_sllv(a0, a1, -SHIFT);
       test_srlc_on(a0, a1);
       test_srlv(a0, a1, -SHIFT);
       test_srac_on(a0, a1);
       test_srav(a0, a1, -SHIFT);
+
+      test_sllc_add(a0, a1);
+      test_sllv_add(a0, a1, ADD_INIT);
+      test_srlc_add(a0, a1);
+      test_srlv_add(a0, a1, ADD_INIT);
+      test_srac_add(a0, a1);
+      test_srav_add(a0, a1, ADD_INIT);
+
+      test_sllc_and(a0, a1);
+      test_sllv_and(a0, a1, BIT_MASK);
+      test_srlc_and(a0, a1);
+      test_srlv_and(a0, a1, BIT_MASK);
+      test_srac_and(a0, a1);
+      test_srav_and(a0, a1, BIT_MASK);
     }
     // Test and verify results
     System.out.println("Verification");
@@ -354,6 +374,60 @@ public class TestLongVect {
         errn += verify("test_srav_on: ", i, a0[i], (long)((long)(ADD_INIT+i)>>(-SHIFT)));
       }
 
+      test_sllc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+      test_sllv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+
+      test_srlc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+      test_srlv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+
+      test_srac_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+      test_srav_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_add: ", i, a0[i], (long)(((long)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+
+      test_sllc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+      test_sllv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+
+      test_srlc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+      test_srlv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+
+      test_srac_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+      test_srav_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_and: ", i, a0[i], (long)(((long)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+
     }
 
     if (errn > 0)
@@ -696,6 +770,84 @@ public class TestLongVect {
     end = System.currentTimeMillis();
     System.out.println("test_srav_on: " + (end - start));
 
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_and: " + (end - start));
+
     return errn;
   }
 
@@ -854,6 +1006,26 @@ public class TestLongVect {
       a0[i] = (long)(a1[i]<<b);
     }
   }
+  static void test_sllc_add(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + ADD_INIT)<<VALUE);
+    }
+  }
+  static void test_sllv_add(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + b)<<VALUE);
+    }
+  }
+  static void test_sllc_and(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & BIT_MASK)<<VALUE);
+    }
+  }
+  static void test_sllv_and(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & b)<<VALUE);
+    }
+  }
 
   static void test_srlc(long[] a0, long[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -880,6 +1052,26 @@ public class TestLongVect {
       a0[i] = (long)(a1[i]>>>b);
     }
   }
+  static void test_srlc_add(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + ADD_INIT)>>>VALUE);
+    }
+  }
+  static void test_srlv_add(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + b)>>>VALUE);
+    }
+  }
+  static void test_srlc_and(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & BIT_MASK)>>>VALUE);
+    }
+  }
+  static void test_srlv_and(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & b)>>>VALUE);
+    }
+  }
 
   static void test_srac(long[] a0, long[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -906,6 +1098,26 @@ public class TestLongVect {
       a0[i] = (long)(a1[i]>>b);
     }
   }
+  static void test_srac_add(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + ADD_INIT)>>VALUE);
+    }
+  }
+  static void test_srav_add(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] + b)>>VALUE);
+    }
+  }
+  static void test_srac_and(long[] a0, long[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & BIT_MASK)>>VALUE);
+    }
+  }
+  static void test_srav_and(long[] a0, long[] a1, long b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (long)((a1[i] & b)>>VALUE);
+    }
+  }
 
   static int verify(String text, int i, long elem, long val) {
     if (elem != val) {
diff --git a/test/compiler/6340864/TestShortVect.java b/test/compiler/6340864/TestShortVect.java
index a688e0de0..9f59c8f22 100644
--- a/test/compiler/6340864/TestShortVect.java
+++ b/test/compiler/6340864/TestShortVect.java
@@ -75,6 +75,7 @@ public class TestShortVect {
       test_subc(a0, a1);
       test_subv(a0, a1, (short)VALUE);
       test_suba(a0, a1, a2);
+
       test_mulc(a0, a1);
       test_mulv(a0, a1, (short)VALUE);
       test_mula(a0, a1, a2);
@@ -87,6 +88,7 @@ public class TestShortVect {
       test_divc_n(a0, a1);
       test_divv(a0, a1, (short)-VALUE);
       test_diva(a0, a1, a3);
+
       test_andc(a0, a1);
       test_andv(a0, a1, (short)BIT_MASK);
       test_anda(a0, a1, a4);
@@ -96,30 +98,49 @@ public class TestShortVect {
       test_xorc(a0, a1);
       test_xorv(a0, a1, (short)BIT_MASK);
       test_xora(a0, a1, a4);
+
       test_sllc(a0, a1);
       test_sllv(a0, a1, VALUE);
       test_srlc(a0, a1);
       test_srlv(a0, a1, VALUE);
       test_srac(a0, a1);
       test_srav(a0, a1, VALUE);
+
       test_sllc_n(a0, a1);
       test_sllv(a0, a1, -VALUE);
       test_srlc_n(a0, a1);
       test_srlv(a0, a1, -VALUE);
       test_srac_n(a0, a1);
       test_srav(a0, a1, -VALUE);
+
       test_sllc_o(a0, a1);
       test_sllv(a0, a1, SHIFT);
       test_srlc_o(a0, a1);
       test_srlv(a0, a1, SHIFT);
       test_srac_o(a0, a1);
       test_srav(a0, a1, SHIFT);
+
       test_sllc_on(a0, a1);
       test_sllv(a0, a1, -SHIFT);
       test_srlc_on(a0, a1);
       test_srlv(a0, a1, -SHIFT);
       test_srac_on(a0, a1);
       test_srav(a0, a1, -SHIFT);
+
+      test_sllc_add(a0, a1);
+      test_sllv_add(a0, a1, ADD_INIT);
+      test_srlc_add(a0, a1);
+      test_srlv_add(a0, a1, ADD_INIT);
+      test_srac_add(a0, a1);
+      test_srav_add(a0, a1, ADD_INIT);
+
+      test_sllc_and(a0, a1);
+      test_sllv_and(a0, a1, BIT_MASK);
+      test_srlc_and(a0, a1);
+      test_srlv_and(a0, a1, BIT_MASK);
+      test_srac_and(a0, a1);
+      test_srav_and(a0, a1, BIT_MASK);
+
       test_pack2(p2, a1);
       test_unpack2(a0, p2);
       test_pack2_swap(p2, a1);
@@ -364,6 +385,60 @@ public class TestShortVect {
         errn += verify("test_srav_on: ", i, a0[i], (short)((short)(ADD_INIT+i)>>(-SHIFT)));
       }
 
+      test_sllc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+      test_sllv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+
+      test_srlc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+      test_srlv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+
+      test_srac_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+      test_srav_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_add: ", i, a0[i], (short)(((short)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+
+      test_sllc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+      test_sllv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+
+      test_srlc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+      test_srlv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+
+      test_srac_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+      test_srav_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_and: ", i, a0[i], (short)(((short)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+
       test_pack2(p2, a1);
       for (int i=0; i<ARRLEN/2; i++) {
         errn += verify("test_pack2: ", i, p2[i], ((int)(ADD_INIT+2*i) & 0xFFFF) | ((int)(ADD_INIT+2*i+1) << 16));
@@ -762,6 +837,84 @@ public class TestShortVect {
 
     start = System.currentTimeMillis();
     for (int i=0; i<ITERS; i++) {
+      test_sllc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
       test_pack2(p2, a1);
     }
     end = System.currentTimeMillis();
@@ -968,6 +1121,26 @@ public class TestShortVect {
       a0[i] = (short)(a1[i]<<b);
     }
   }
+  static void test_sllc_add(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + ADD_INIT)<<VALUE);
+    }
+  }
+  static void test_sllv_add(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + b)<<VALUE);
+    }
+  }
+  static void test_sllc_and(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & BIT_MASK)<<VALUE);
+    }
+  }
+  static void test_sllv_and(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & b)<<VALUE);
+    }
+  }
 
   static void test_srlc(short[] a0, short[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -994,6 +1167,26 @@ public class TestShortVect {
       a0[i] = (short)(a1[i]>>>b);
     }
   }
+  static void test_srlc_add(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + ADD_INIT)>>>VALUE);
+    }
+  }
+  static void test_srlv_add(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + b)>>>VALUE);
+    }
+  }
+  static void test_srlc_and(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & BIT_MASK)>>>VALUE);
+    }
+  }
+  static void test_srlv_and(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & b)>>>VALUE);
+    }
+  }
 
   static void test_srac(short[] a0, short[] a1) {
     for (int i = 0; i < a0.length; i+=1) {
@@ -1020,6 +1213,26 @@ public class TestShortVect {
       a0[i] = (short)(a1[i]>>b);
     }
   }
+  static void test_srac_add(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + ADD_INIT)>>VALUE);
+    }
+  }
+  static void test_srav_add(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] + b)>>VALUE);
+    }
+  }
+  static void test_srac_and(short[] a0, short[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & BIT_MASK)>>VALUE);
+    }
+  }
+  static void test_srav_and(short[] a0, short[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (short)((a1[i] & b)>>VALUE);
+    }
+  }
 
   static void test_pack2(int[] p2, short[] a1) {
     if (p2.length*2 > a1.length) return;
diff --git a/test/compiler/7184394/TestAESBase.java b/test/compiler/7184394/TestAESBase.java
new file mode 100644
index 000000000..ad6c835cc
--- /dev/null
+++ b/test/compiler/7184394/TestAESBase.java
@@ -0,0 +1,154 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+/**
+ * @author Tom Deneau
+ */
+
+import javax.crypto.Cipher;
+import javax.crypto.KeyGenerator;
+import javax.crypto.SecretKey;
+import javax.crypto.spec.IvParameterSpec;
+import javax.crypto.spec.SecretKeySpec;
+import java.security.AlgorithmParameters;
+
+import java.util.Random;
+import java.util.Arrays;
+
+abstract public class TestAESBase {
+  int msgSize = Integer.getInteger("msgSize", 646);
+  boolean checkOutput = Boolean.getBoolean("checkOutput");
+  boolean noReinit = Boolean.getBoolean("noReinit");
+  int keySize = Integer.getInteger("keySize", 128);
+  String algorithm = System.getProperty("algorithm", "AES");
+  String mode = System.getProperty("mode", "CBC");
+  byte[] input;
+  byte[] encode;
+  byte[] expectedEncode;
+  byte[] decode;
+  byte[] expectedDecode;
+  Random random = new Random(0);
+  Cipher cipher;
+  Cipher dCipher;
+  String paddingStr = "PKCS5Padding";
+  AlgorithmParameters algParams;
+  SecretKey key;
+  int ivLen;
+
+  static int numThreads = 0;
+  int  threadId;
+  static synchronized int getThreadId() {
+    int id = numThreads;
+    numThreads++;
+    return id;
+  }
+
+  abstract public void run();
+
+  public void prepare() {
+    try {
+    System.out.println("\nmsgSize=" + msgSize + ", key size=" + keySize + ", reInit=" + !noReinit + ", checkOutput=" + checkOutput);
+
+      int keyLenBytes = (keySize == 0 ? 16 : keySize/8);
+      byte keyBytes[] = new byte[keyLenBytes];
+      if (keySize == 128)
+        keyBytes = new byte[] {-8, -7, -6, -5, -4, -3, -2, -1, 0, 1, 2, 3, 4, 5, 6, 7};
+      else
+        random.nextBytes(keyBytes);
+
+      key = new SecretKeySpec(keyBytes, algorithm);
+      if (threadId == 0) {
+        System.out.println("Algorithm: " + key.getAlgorithm() + "("
+                           + key.getEncoded().length * 8 + "bit)");
+      }
+      input = new byte[msgSize];
+      for (int i=0; i<input.length; i++) {
+        input[i] = (byte) (i & 0xff);
+      }
+
+      cipher = Cipher.getInstance(algorithm + "/" + mode + "/" + paddingStr, "SunJCE");
+      dCipher = Cipher.getInstance(algorithm + "/" + mode + "/" + paddingStr, "SunJCE");
+
+      ivLen = (algorithm.equals("AES") ? 16 : algorithm.equals("DES") ? 8 : 0);
+      IvParameterSpec initVector = new IvParameterSpec(new byte[ivLen]);
+
+      cipher.init(Cipher.ENCRYPT_MODE, key, initVector);
+      algParams = cipher.getParameters();
+      dCipher.init(Cipher.DECRYPT_MODE, key, algParams);
+      if (threadId == 0) {
+        childShowCipher();
+      }
+
+      // do one encode and decode in preparation
+      // this will also create the encode buffer and decode buffer
+      encode = cipher.doFinal(input);
+      decode = dCipher.doFinal(encode);
+      if (checkOutput) {
+        expectedEncode = (byte[]) encode.clone();
+        expectedDecode = (byte[]) decode.clone();
+        showArray(key.getEncoded()  ,  "key:    ");
+        showArray(input,  "input:  ");
+        showArray(encode, "encode: ");
+        showArray(decode, "decode: ");
+      }
+    }
+    catch (Exception e) {
+      e.printStackTrace();
+      System.exit(1);
+    }
+  }
+
+  void showArray(byte b[], String name) {
+    System.out.format("%s [%d]: ", name, b.length);
+    for (int i=0; i<Math.min(b.length, 32); i++) {
+      System.out.format("%02x ", b[i] & 0xff);
+    }
+    System.out.println();
+  }
+
+  void compareArrays(byte b[], byte exp[]) {
+    if (b.length != exp.length) {
+      System.out.format("different lengths for actual and expected output arrays\n");
+      showArray(b, "test: ");
+      showArray(exp, "exp : ");
+      System.exit(1);
+    }
+    for (int i=0; i< exp.length; i++) {
+      if (b[i] != exp[i]) {
+        System.out.format("output error at index %d: got %02x, expected %02x\n", i, b[i] & 0xff, exp[i] & 0xff);
+        showArray(b, "test: ");
+        showArray(exp, "exp : ");
+        System.exit(1);
+      }
+    }
+  }
+
+
+  void showCipher(Cipher c, String kind) {
+    System.out.println(kind + " cipher provider: " + cipher.getProvider());
+    System.out.println(kind + " cipher algorithm: " + cipher.getAlgorithm());
+  }
+
+  abstract void childShowCipher();
+}
diff --git a/test/compiler/7184394/TestAESDecode.java b/test/compiler/7184394/TestAESDecode.java
new file mode 100644
index 000000000..f9ec02d15
--- /dev/null
+++ b/test/compiler/7184394/TestAESDecode.java
@@ -0,0 +1,57 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+/**
+ * @author Tom Deneau
+ */
+
+import javax.crypto.Cipher;
+
+public class TestAESDecode extends TestAESBase {
+  @Override
+  public void run() {
+    try {
+      if (!noReinit) dCipher.init(Cipher.DECRYPT_MODE, key, algParams);
+      if (checkOutput) {
+        // checked version creates new output buffer each time
+        decode = dCipher.doFinal(encode, 0, encode.length);
+        compareArrays(decode, expectedDecode);
+      } else {
+        // non-checked version outputs to existing encode buffer for maximum speed
+        decode = new byte[dCipher.getOutputSize(encode.length)];
+        dCipher.doFinal(encode, 0, encode.length, decode);
+      }
+    }
+    catch (Exception e) {
+      e.printStackTrace();
+      System.exit(1);
+    }
+  }
+
+  @Override
+  void childShowCipher() {
+    showCipher(dCipher, "Decryption");
+  }
+
+}
diff --git a/test/compiler/7184394/TestAESEncode.java b/test/compiler/7184394/TestAESEncode.java
new file mode 100644
index 000000000..1d6bf7fbd
--- /dev/null
+++ b/test/compiler/7184394/TestAESEncode.java
@@ -0,0 +1,57 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+/**
+ * @author Tom Deneau
+ */
+
+import javax.crypto.Cipher;
+
+public class TestAESEncode extends TestAESBase {
+  @Override
+  public void run() {
+    try {
+      if (!noReinit) cipher.init(Cipher.ENCRYPT_MODE, key, algParams);
+      if (checkOutput) {
+        // checked version creates new output buffer each time
+        encode = cipher.doFinal(input, 0, msgSize);
+        compareArrays(encode, expectedEncode);
+      } else {
+        // non-checked version outputs to existing encode buffer for maximum speed
+        encode = new byte[cipher.getOutputSize(msgSize)];
+        cipher.doFinal(input, 0, msgSize, encode);
+      }
+    }
+    catch (Exception e) {
+      e.printStackTrace();
+      System.exit(1);
+    }
+  }
+
+  @Override
+  void childShowCipher() {
+    showCipher(cipher, "Encryption");
+  }
+
+}
diff --git a/test/compiler/7184394/TestAESMain.java b/test/compiler/7184394/TestAESMain.java
new file mode 100644
index 000000000..ca2cb43dc
--- /dev/null
+++ b/test/compiler/7184394/TestAESMain.java
@@ -0,0 +1,57 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+/**
+ * @test
+ * @bug 7184394
+ * @summary add intrinsics to use AES instructions
+ *
+ * @run main/othervm/timeout=600 -Xbatch -DcheckOutput=true TestAESMain
+ *
+ * @author Tom Deneau
+ */
+
+public class TestAESMain {
+  public static void main(String[] args) {
+    int iters = (args.length > 0 ? Integer.valueOf(args[0]) : 1000000);
+    System.out.println(iters + " iterations");
+    TestAESEncode etest = new TestAESEncode();
+    etest.prepare();
+    long start = System.nanoTime();
+    for (int i=0; i<iters; i++) {
+      etest.run();
+    }
+    long end = System.nanoTime();
+    System.out.println("TestAESEncode runtime was " + (double)((end - start)/1000000000.0) + " ms");
+
+    TestAESDecode dtest = new TestAESDecode();
+    dtest.prepare();
+    start = System.nanoTime();
+    for (int i=0; i<iters; i++) {
+      dtest.run();
+    }
+    end = System.nanoTime();
+    System.out.println("TestAESDecode runtime was " + (double)((end - start)/1000000000.0) + " ms");
+  }
+}
diff --git a/test/compiler/8000805/Test8000805.java b/test/compiler/8000805/Test8000805.java
new file mode 100644
index 000000000..bd0b7b41c
--- /dev/null
+++ b/test/compiler/8000805/Test8000805.java
@@ -0,0 +1,85 @@
+/*
+ * Copyright (c) 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.
+ */
+
+/**
+ * @test
+ * @bug 8000805
+ * @summary JMM issue: short loads are non-atomic
+ *
+ * @run main/othervm -server -XX:-TieredCompilation -Xcomp -XX:+PrintCompilation -XX:CompileOnly=Test8000805.loadS2LmaskFF,Test8000805.loadS2Lmask16,Test8000805.loadS2Lmask13,Test8000805.loadUS_signExt,Test8000805.loadB2L_mask8 Test8000805
+ */
+
+public class Test8000805 {
+    static long loadS2LmaskFF   (short[] sa) { return sa[0] & 0xFF; }
+    static long loadS2LmaskFF_1 (short[] sa) { return sa[0] & 0xFF; }
+
+    static long loadS2Lmask16   (short[] sa) { return sa[0] & 0xFFFE; }
+    static long loadS2Lmask16_1 (short[] sa) { return sa[0] & 0xFFFE; }
+
+    static long loadS2Lmask13   (short[] sa) { return sa[0] & 0x0FFF; }
+    static long loadS2Lmask13_1 (short[] sa) { return sa[0] & 0x0FFF; }
+
+    static int loadUS_signExt   (char[] ca) { return (ca[0] << 16) >> 16; }
+    static int loadUS_signExt_1 (char[] ca) { return (ca[0] << 16) >> 16; }
+
+    static long loadB2L_mask8   (byte[] ba) { return ba[0] & 0x55; }
+    static long loadB2L_mask8_1 (byte[] ba) { return ba[0] & 0x55; }
+
+    public static void main(String[] args) {
+        for (int i = Byte.MIN_VALUE; i < Byte.MAX_VALUE; i++) {
+            byte[] ba = new byte[]  { (byte) i};
+
+            { long v1 = loadB2L_mask8(ba);
+              long v2 = loadB2L_mask8_1(ba);
+              if (v1 != v2)
+              throw new InternalError(String.format("loadB2L_mask8 failed: %x != %x", v1, v2)); }
+        }
+
+        for (int i = Short.MIN_VALUE; i < Short.MAX_VALUE; i++) {
+            short[] sa = new short[] { (short)i };
+            char[] ca = new char[] { (char)i };
+
+            { long v1 = loadS2LmaskFF(sa);
+              long v2 = loadS2LmaskFF_1(sa);
+              if (v1 != v2)
+              throw new InternalError(String.format("loadS2LmaskFF failed: %x != %x", v1, v2)); }
+
+            { long v1 = loadS2Lmask16(sa);
+              long v2 = loadS2Lmask16_1(sa);
+              if (v1 != v2)
+              throw new InternalError(String.format("loadS2Lmask16 failed: %x != %x", v1, v2)); }
+
+            { long v1 = loadS2Lmask13(sa);
+              long v2 = loadS2Lmask13_1(sa);
+              if (v1 != v2)
+              throw new InternalError(String.format("loadS2Lmask13 failed: %x != %x", v1, v2)); }
+
+            { int v1 = loadUS_signExt(ca);
+              int v2 = loadUS_signExt_1(ca);
+              if (v1 != v2)
+                throw new InternalError(String.format("loadUS_signExt failed: %x != %x", v1, v2)); }
+        }
+
+        System.out.println("TEST PASSED.");
+    }
+}
diff --git a/test/compiler/8001183/TestCharVect.java b/test/compiler/8001183/TestCharVect.java
new file mode 100644
index 000000000..a6ff1e2b9
--- /dev/null
+++ b/test/compiler/8001183/TestCharVect.java
@@ -0,0 +1,1332 @@
+/*
+ * Copyright (c) 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.
+ *
+ */
+
+/**
+ * @test
+ * @bug 8001183
+ * @summary incorrect results of char vectors right shift operaiton
+ *
+ * @run main/othervm/timeout=400 -Xbatch -Xmx64m TestCharVect
+ */
+
+public class TestCharVect {
+  private static final int ARRLEN = 997;
+  private static final int ITERS  = 11000;
+  private static final int ADD_INIT = Character.MAX_VALUE-500;
+  private static final int BIT_MASK = 0xB731;
+  private static final int VALUE = 7;
+  private static final int SHIFT = 16;
+
+  public static void main(String args[]) {
+    System.out.println("Testing Char vectors");
+    int errn = test();
+    if (errn > 0) {
+      System.err.println("FAILED: " + errn + " errors");
+      System.exit(97);
+    }
+    System.out.println("PASSED");
+  }
+
+  static int test() {
+    char[] a0 = new char[ARRLEN];
+    char[] a1 = new char[ARRLEN];
+    short[] a2 = new short[ARRLEN];
+    short[] a3 = new short[ARRLEN];
+    short[] a4 = new short[ARRLEN];
+     int[] p2 = new  int[ARRLEN/2];
+    long[] p4 = new long[ARRLEN/4];
+    // Initialize
+    int gold_sum = 0;
+    for (int i=0; i<ARRLEN; i++) {
+      char val = (char)(ADD_INIT+i);
+      gold_sum += val;
+      a1[i] = val;
+      a2[i] = VALUE;
+      a3[i] = -VALUE;
+      a4[i] = (short)BIT_MASK;
+    }
+    System.out.println("Warmup");
+    for (int i=0; i<ITERS; i++) {
+      test_sum(a1);
+      test_addc(a0, a1);
+      test_addv(a0, a1, (char)VALUE);
+      test_adda(a0, a1, a2);
+      test_subc(a0, a1);
+      test_subv(a0, a1, (char)VALUE);
+      test_suba(a0, a1, a2);
+
+      test_mulc(a0, a1);
+      test_mulv(a0, a1, (char)VALUE);
+      test_mula(a0, a1, a2);
+      test_divc(a0, a1);
+      test_divv(a0, a1, VALUE);
+      test_diva(a0, a1, a2);
+      test_mulc_n(a0, a1);
+      test_mulv(a0, a1, (char)-VALUE);
+      test_mula(a0, a1, a3);
+      test_divc_n(a0, a1);
+      test_divv(a0, a1, -VALUE);
+      test_diva(a0, a1, a3);
+
+      test_andc(a0, a1);
+      test_andv(a0, a1, (short)BIT_MASK);
+      test_anda(a0, a1, a4);
+      test_orc(a0, a1);
+      test_orv(a0, a1, (short)BIT_MASK);
+      test_ora(a0, a1, a4);
+      test_xorc(a0, a1);
+      test_xorv(a0, a1, (short)BIT_MASK);
+      test_xora(a0, a1, a4);
+
+      test_sllc(a0, a1);
+      test_sllv(a0, a1, VALUE);
+      test_srlc(a0, a1);
+      test_srlv(a0, a1, VALUE);
+      test_srac(a0, a1);
+      test_srav(a0, a1, VALUE);
+
+      test_sllc_n(a0, a1);
+      test_sllv(a0, a1, -VALUE);
+      test_srlc_n(a0, a1);
+      test_srlv(a0, a1, -VALUE);
+      test_srac_n(a0, a1);
+      test_srav(a0, a1, -VALUE);
+
+      test_sllc_o(a0, a1);
+      test_sllv(a0, a1, SHIFT);
+      test_srlc_o(a0, a1);
+      test_srlv(a0, a1, SHIFT);
+      test_srac_o(a0, a1);
+      test_srav(a0, a1, SHIFT);
+
+      test_sllc_on(a0, a1);
+      test_sllv(a0, a1, -SHIFT);
+      test_srlc_on(a0, a1);
+      test_srlv(a0, a1, -SHIFT);
+      test_srac_on(a0, a1);
+      test_srav(a0, a1, -SHIFT);
+
+      test_sllc_add(a0, a1);
+      test_sllv_add(a0, a1, ADD_INIT);
+      test_srlc_add(a0, a1);
+      test_srlv_add(a0, a1, ADD_INIT);
+      test_srac_add(a0, a1);
+      test_srav_add(a0, a1, ADD_INIT);
+
+      test_sllc_and(a0, a1);
+      test_sllv_and(a0, a1, BIT_MASK);
+      test_srlc_and(a0, a1);
+      test_srlv_and(a0, a1, BIT_MASK);
+      test_srac_and(a0, a1);
+      test_srav_and(a0, a1, BIT_MASK);
+
+      test_pack2(p2, a1);
+      test_unpack2(a0, p2);
+      test_pack2_swap(p2, a1);
+      test_unpack2_swap(a0, p2);
+      test_pack4(p4, a1);
+      test_unpack4(a0, p4);
+      test_pack4_swap(p4, a1);
+      test_unpack4_swap(a0, p4);
+    }
+    // Test and verify results
+    System.out.println("Verification");
+    int errn = 0;
+    {
+      int sum = test_sum(a1);
+      if (sum != gold_sum) {
+        System.err.println("test_sum:  " + sum + " != " + gold_sum);
+        errn++;
+      }
+
+      test_addc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_addc: ", i, a0[i], (char)((char)(ADD_INIT+i)+VALUE));
+      }
+      test_addv(a0, a1, (char)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_addv: ", i, a0[i], (char)((char)(ADD_INIT+i)+VALUE));
+      }
+      test_adda(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_adda: ", i, a0[i], (char)((char)(ADD_INIT+i)+VALUE));
+      }
+
+      test_subc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_subc: ", i, a0[i], (char)((char)(ADD_INIT+i)-VALUE));
+      }
+      test_subv(a0, a1, (char)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_subv: ", i, a0[i], (char)((char)(ADD_INIT+i)-VALUE));
+      }
+      test_suba(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_suba: ", i, a0[i], (char)((char)(ADD_INIT+i)-VALUE));
+      }
+
+      test_mulc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulc: ", i, a0[i], (char)((char)(ADD_INIT+i)*VALUE));
+      }
+      test_mulv(a0, a1, (char)VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulv: ", i, a0[i], (char)((char)(ADD_INIT+i)*VALUE));
+      }
+      test_mula(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mula: ", i, a0[i], (char)((char)(ADD_INIT+i)*VALUE));
+      }
+
+      test_divc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divc: ", i, a0[i], (char)((char)(ADD_INIT+i)/VALUE));
+      }
+      test_divv(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divv: ", i, a0[i], (char)((char)(ADD_INIT+i)/VALUE));
+      }
+      test_diva(a0, a1, a2);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_diva: ", i, a0[i], (char)((char)(ADD_INIT+i)/VALUE));
+      }
+
+      test_mulc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulc_n: ", i, a0[i], (char)((char)(ADD_INIT+i)*(-VALUE)));
+      }
+      test_mulv(a0, a1, (char)-VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mulv_n: ", i, a0[i], (char)((char)(ADD_INIT+i)*(-VALUE)));
+      }
+      test_mula(a0, a1, a3);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_mula_n: ", i, a0[i], (char)((char)(ADD_INIT+i)*(-VALUE)));
+      }
+
+      test_divc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divc_n: ", i, a0[i], (char)((char)(ADD_INIT+i)/(-VALUE)));
+      }
+      test_divv(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_divv_n: ", i, a0[i], (char)((char)(ADD_INIT+i)/(-VALUE)));
+      }
+      test_diva(a0, a1, a3);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_diva_n: ", i, a0[i], (char)((char)(ADD_INIT+i)/(-VALUE)));
+      }
+
+      test_andc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_andc: ", i, a0[i], (char)((char)(ADD_INIT+i)&BIT_MASK));
+      }
+      test_andv(a0, a1, (short)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_andv: ", i, a0[i], (char)((char)(ADD_INIT+i)&BIT_MASK));
+      }
+      test_anda(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_anda: ", i, a0[i], (char)((char)(ADD_INIT+i)&BIT_MASK));
+      }
+
+      test_orc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_orc: ", i, a0[i], (char)((char)(ADD_INIT+i)|BIT_MASK));
+      }
+      test_orv(a0, a1, (short)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_orv: ", i, a0[i], (char)((char)(ADD_INIT+i)|BIT_MASK));
+      }
+      test_ora(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_ora: ", i, a0[i], (char)((char)(ADD_INIT+i)|BIT_MASK));
+      }
+
+      test_xorc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xorc: ", i, a0[i], (char)((char)(ADD_INIT+i)^BIT_MASK));
+      }
+      test_xorv(a0, a1, (short)BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xorv: ", i, a0[i], (char)((char)(ADD_INIT+i)^BIT_MASK));
+      }
+      test_xora(a0, a1, a4);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_xora: ", i, a0[i], (char)((char)(ADD_INIT+i)^BIT_MASK));
+      }
+
+      test_sllc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc: ", i, a0[i], (char)((char)(ADD_INIT+i)<<VALUE));
+      }
+      test_sllv(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv: ", i, a0[i], (char)((char)(ADD_INIT+i)<<VALUE));
+      }
+
+      test_srlc(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>VALUE));
+      }
+      test_srlv(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>VALUE));
+      }
+
+      test_srac(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac: ", i, a0[i], (char)((char)(ADD_INIT+i)>>VALUE));
+      }
+      test_srav(a0, a1, VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav: ", i, a0[i], (char)((char)(ADD_INIT+i)>>VALUE));
+      }
+
+      test_sllc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_n: ", i, a0[i], (char)((char)(ADD_INIT+i)<<(-VALUE)));
+      }
+      test_sllv(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_n: ", i, a0[i], (char)((char)(ADD_INIT+i)<<(-VALUE)));
+      }
+
+      test_srlc_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_n: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>(-VALUE)));
+      }
+      test_srlv(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_n: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>(-VALUE)));
+      }
+
+      test_srac_n(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_n: ", i, a0[i], (char)((char)(ADD_INIT+i)>>(-VALUE)));
+      }
+      test_srav(a0, a1, -VALUE);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_n: ", i, a0[i], (char)((char)(ADD_INIT+i)>>(-VALUE)));
+      }
+
+      test_sllc_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_o: ", i, a0[i], (char)((char)(ADD_INIT+i)<<SHIFT));
+      }
+      test_sllv(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_o: ", i, a0[i], (char)((char)(ADD_INIT+i)<<SHIFT));
+      }
+
+      test_srlc_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_o: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>SHIFT));
+      }
+      test_srlv(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_o: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>SHIFT));
+      }
+
+      test_srac_o(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_o: ", i, a0[i], (char)((char)(ADD_INIT+i)>>SHIFT));
+      }
+      test_srav(a0, a1, SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_o: ", i, a0[i], (char)((char)(ADD_INIT+i)>>SHIFT));
+      }
+
+      test_sllc_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_on: ", i, a0[i], (char)((char)(ADD_INIT+i)<<(-SHIFT)));
+      }
+      test_sllv(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_on: ", i, a0[i], (char)((char)(ADD_INIT+i)<<(-SHIFT)));
+      }
+
+      test_srlc_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_on: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>(-SHIFT)));
+      }
+      test_srlv(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_on: ", i, a0[i], (char)((char)(ADD_INIT+i)>>>(-SHIFT)));
+      }
+
+      test_srac_on(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_on: ", i, a0[i], (char)((char)(ADD_INIT+i)>>(-SHIFT)));
+      }
+      test_srav(a0, a1, -SHIFT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_on: ", i, a0[i], (char)((char)(ADD_INIT+i)>>(-SHIFT)));
+      }
+
+      test_sllc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+      test_sllv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)<<VALUE));
+      }
+
+      test_srlc_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+      test_srlv_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)>>>VALUE));
+      }
+
+      test_srac_add(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+      test_srav_add(a0, a1, ADD_INIT);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_add: ", i, a0[i], (char)(((char)(ADD_INIT+i) + ADD_INIT)>>VALUE));
+      }
+
+      test_sllc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllc_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+      test_sllv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_sllv_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)<<VALUE));
+      }
+
+      test_srlc_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlc_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+      test_srlv_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srlv_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)>>>VALUE));
+      }
+
+      test_srac_and(a0, a1);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srac_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+      test_srav_and(a0, a1, BIT_MASK);
+      for (int i=0; i<ARRLEN; i++) {
+        errn += verify("test_srav_and: ", i, a0[i], (char)(((char)(ADD_INIT+i) & BIT_MASK)>>VALUE));
+      }
+
+      test_pack2(p2, a1);
+      for (int i=0; i<ARRLEN/2; i++) {
+        errn += verify("test_pack2: ", i, p2[i], ((int)(ADD_INIT+2*i) & 0xFFFF) | ((int)(ADD_INIT+2*i+1) << 16));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = (char)-1;
+      }
+      test_unpack2(a0, p2);
+      for (int i=0; i<(ARRLEN&(-2)); i++) {
+        errn += verify("test_unpack2: ", i, a0[i], (char)(ADD_INIT+i));
+      }
+
+      test_pack2_swap(p2, a1);
+      for (int i=0; i<ARRLEN/2; i++) {
+        errn += verify("test_pack2_swap: ", i, p2[i], ((int)(ADD_INIT+2*i+1) & 0xFFFF) | ((int)(ADD_INIT+2*i) << 16));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = (char)-1;
+      }
+      test_unpack2_swap(a0, p2);
+      for (int i=0; i<(ARRLEN&(-2)); i++) {
+        errn += verify("test_unpack2_swap: ", i, a0[i], (char)(ADD_INIT+i));
+      }
+
+      test_pack4(p4, a1);
+      for (int i=0; i<ARRLEN/4; i++) {
+        errn += verify("test_pack4: ", i, p4[i],  ((long)(ADD_INIT+4*i+0) & 0xFFFFl) |
+                                                 (((long)(ADD_INIT+4*i+1) & 0xFFFFl) << 16)  |
+                                                 (((long)(ADD_INIT+4*i+2) & 0xFFFFl) << 32)  |
+                                                 (((long)(ADD_INIT+4*i+3) & 0xFFFFl) << 48));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = (char)-1;
+      }
+      test_unpack4(a0, p4);
+      for (int i=0; i<(ARRLEN&(-4)); i++) {
+        errn += verify("test_unpack4: ", i, a0[i], (char)(ADD_INIT+i));
+      }
+
+      test_pack4_swap(p4, a1);
+      for (int i=0; i<ARRLEN/4; i++) {
+        errn += verify("test_pack4_swap: ", i, p4[i],  ((long)(ADD_INIT+4*i+3) & 0xFFFFl) |
+                                                      (((long)(ADD_INIT+4*i+2) & 0xFFFFl) << 16)  |
+                                                      (((long)(ADD_INIT+4*i+1) & 0xFFFFl) << 32)  |
+                                                      (((long)(ADD_INIT+4*i+0) & 0xFFFFl) << 48));
+      }
+      for (int i=0; i<ARRLEN; i++) {
+        a0[i] = (char)-1;
+      }
+      test_unpack4_swap(a0, p4);
+      for (int i=0; i<(ARRLEN&(-4)); i++) {
+        errn += verify("test_unpack4_swap: ", i, a0[i], (char)(ADD_INIT+i));
+      }
+
+    }
+
+    if (errn > 0)
+      return errn;
+
+    System.out.println("Time");
+    long start, end;
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sum(a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sum: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_addc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_addc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_addv(a0, a1, (char)VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_addv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_adda(a0, a1, a2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_adda: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_subc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_subc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_subv(a0, a1, (char)VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_subv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_suba(a0, a1, a2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_suba: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mulc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mulc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mulv(a0, a1, (char)VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mulv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mula(a0, a1, a2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mula: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_divc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_divc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_divv(a0, a1, VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_divv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_diva(a0, a1, a2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_diva: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mulc_n(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mulc_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mulv(a0, a1, (char)-VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mulv_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_mula(a0, a1, a3);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_mula_n: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_divc_n(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_divc_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_divv(a0, a1, -VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_divv_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_diva(a0, a1, a3);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_diva_n: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_andc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_andc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_andv(a0, a1, (short)BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_andv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_anda(a0, a1, a4);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_anda: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_orc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_orc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_orv(a0, a1, (short)BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_orv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_ora(a0, a1, a4);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_ora: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_xorc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_xorc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_xorv(a0, a1, (short)BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_xorv: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_xora(a0, a1, a4);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_xora: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv(a0, a1, VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv(a0, a1, VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav(a0, a1, VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_n(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv(a0, a1, -VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_n: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_n(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv(a0, a1, -VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_n: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_n(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_n: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav(a0, a1, -VALUE);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_n: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_o(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_o: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv(a0, a1, SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_o: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_o(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_o: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv(a0, a1, SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_o: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_o(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_o: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav(a0, a1, SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_o: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_on(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_on: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv(a0, a1, -SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_on: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_on(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_on: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv(a0, a1, -SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_on: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_on(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_on: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav(a0, a1, -SHIFT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_on: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_add(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_add: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_add(a0, a1, ADD_INIT);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_add: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_sllv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_sllv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlc_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlc_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srlv_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srlv_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srac_and(a0, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srac_and: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_srav_and(a0, a1, BIT_MASK);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_srav_and: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_pack2(p2, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_pack2: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_unpack2(a0, p2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_unpack2: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_pack2_swap(p2, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_pack2_swap: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_unpack2_swap(a0, p2);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_unpack2_swap: " + (end - start));
+
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_pack4(p4, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_pack4: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_unpack4(a0, p4);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_unpack4: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_pack4_swap(p4, a1);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_pack4_swap: " + (end - start));
+    start = System.currentTimeMillis();
+    for (int i=0; i<ITERS; i++) {
+      test_unpack4_swap(a0, p4);
+    }
+    end = System.currentTimeMillis();
+    System.out.println("test_unpack4_swap: " + (end - start));
+
+    return errn;
+  }
+
+  static int test_sum(char[] a1) {
+    int sum = 0;
+    for (int i = 0; i < a1.length; i+=1) {
+      sum += a1[i];
+    }
+    return sum;
+  }
+
+  static void test_addc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]+VALUE);
+    }
+  }
+  static void test_addv(char[] a0, char[] a1, char b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]+b);
+    }
+  }
+  static void test_adda(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]+a2[i]);
+    }
+  }
+
+  static void test_subc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]-VALUE);
+    }
+  }
+  static void test_subv(char[] a0, char[] a1, char b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]-b);
+    }
+  }
+  static void test_suba(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]-a2[i]);
+    }
+  }
+
+  static void test_mulc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]*VALUE);
+    }
+  }
+  static void test_mulc_n(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]*(-VALUE));
+    }
+  }
+  static void test_mulv(char[] a0, char[] a1, char b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]*b);
+    }
+  }
+  static void test_mula(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]*a2[i]);
+    }
+  }
+
+  static void test_divc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]/VALUE);
+    }
+  }
+  static void test_divc_n(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]/(-VALUE));
+    }
+  }
+  static void test_divv(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]/b);
+    }
+  }
+  static void test_diva(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]/a2[i]);
+    }
+  }
+
+  static void test_andc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]&BIT_MASK);
+    }
+  }
+  static void test_andv(char[] a0, char[] a1, short b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]&b);
+    }
+  }
+  static void test_anda(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]&a2[i]);
+    }
+  }
+
+  static void test_orc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]|BIT_MASK);
+    }
+  }
+  static void test_orv(char[] a0, char[] a1, short b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]|b);
+    }
+  }
+  static void test_ora(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]|a2[i]);
+    }
+  }
+
+  static void test_xorc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]^BIT_MASK);
+    }
+  }
+  static void test_xorv(char[] a0, char[] a1, short b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]^b);
+    }
+  }
+  static void test_xora(char[] a0, char[] a1, short[] a2) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]^a2[i]);
+    }
+  }
+
+  static void test_sllc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]<<VALUE);
+    }
+  }
+  static void test_sllc_n(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]<<(-VALUE));
+    }
+  }
+  static void test_sllc_o(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]<<SHIFT);
+    }
+  }
+  static void test_sllc_on(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]<<(-SHIFT));
+    }
+  }
+  static void test_sllv(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]<<b);
+    }
+  }
+  static void test_sllc_add(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + ADD_INIT)<<VALUE);
+    }
+  }
+  static void test_sllv_add(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + b)<<VALUE);
+    }
+  }
+  static void test_sllc_and(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & BIT_MASK)<<VALUE);
+    }
+  }
+  static void test_sllv_and(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & b)<<VALUE);
+    }
+  }
+
+  static void test_srlc(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>>VALUE);
+    }
+  }
+  static void test_srlc_n(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>>(-VALUE));
+    }
+  }
+  static void test_srlc_o(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>>SHIFT);
+    }
+  }
+  static void test_srlc_on(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>>(-SHIFT));
+    }
+  }
+  static void test_srlv(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>>b);
+    }
+  }
+  static void test_srlc_add(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + ADD_INIT)>>>VALUE);
+    }
+  }
+  static void test_srlv_add(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + b)>>>VALUE);
+    }
+  }
+  static void test_srlc_and(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & BIT_MASK)>>>VALUE);
+    }
+  }
+  static void test_srlv_and(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & b)>>>VALUE);
+    }
+  }
+
+  static void test_srac(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>VALUE);
+    }
+  }
+  static void test_srac_n(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>(-VALUE));
+    }
+  }
+  static void test_srac_o(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>SHIFT);
+    }
+  }
+  static void test_srac_on(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>(-SHIFT));
+    }
+  }
+  static void test_srav(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)(a1[i]>>b);
+    }
+  }
+  static void test_srac_add(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + ADD_INIT)>>VALUE);
+    }
+  }
+  static void test_srav_add(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] + b)>>VALUE);
+    }
+  }
+  static void test_srac_and(char[] a0, char[] a1) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & BIT_MASK)>>VALUE);
+    }
+  }
+  static void test_srav_and(char[] a0, char[] a1, int b) {
+    for (int i = 0; i < a0.length; i+=1) {
+      a0[i] = (char)((a1[i] & b)>>VALUE);
+    }
+  }
+
+  static void test_pack2(int[] p2, char[] a1) {
+    if (p2.length*2 > a1.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      int l0 = (int)a1[i*2+0];
+      int l1 = (int)a1[i*2+1];
+      p2[i] = (l1 << 16) | (l0 & 0xFFFF);
+    }
+  }
+  static void test_unpack2(char[] a0, int[] p2) {
+    if (p2.length*2 > a0.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      int l = p2[i];
+      a0[i*2+0] = (char)(l & 0xFFFF);
+      a0[i*2+1] = (char)(l >> 16);
+    }
+  }
+  static void test_pack2_swap(int[] p2, char[] a1) {
+    if (p2.length*2 > a1.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      int l0 = (int)a1[i*2+0];
+      int l1 = (int)a1[i*2+1];
+      p2[i] = (l0 << 16) | (l1 & 0xFFFF);
+    }
+  }
+  static void test_unpack2_swap(char[] a0, int[] p2) {
+    if (p2.length*2 > a0.length) return;
+    for (int i = 0; i < p2.length; i+=1) {
+      int l = p2[i];
+      a0[i*2+0] = (char)(l >> 16);
+      a0[i*2+1] = (char)(l & 0xFFFF);
+    }
+  }
+
+  static void test_pack4(long[] p4, char[] a1) {
+    if (p4.length*4 > a1.length) return;
+    for (int i = 0; i < p4.length; i+=1) {
+      long l0 = (long)a1[i*4+0];
+      long l1 = (long)a1[i*4+1];
+      long l2 = (long)a1[i*4+2];
+      long l3 = (long)a1[i*4+3];
+      p4[i] = (l0 & 0xFFFFl) |
+             ((l1 & 0xFFFFl) << 16) |
+             ((l2 & 0xFFFFl) << 32) |
+             ((l3 & 0xFFFFl) << 48);
+    }
+  }
+  static void test_unpack4(char[] a0, long[] p4) {
+    if (p4.length*4 > a0.length) return;
+    for (int i = 0; i < p4.length; i+=1) {
+      long l = p4[i];
+      a0[i*4+0] = (char)(l & 0xFFFFl);
+      a0[i*4+1] = (char)(l >> 16);
+      a0[i*4+2] = (char)(l >> 32);
+      a0[i*4+3] = (char)(l >> 48);
+    }
+  }
+  static void test_pack4_swap(long[] p4, char[] a1) {
+    if (p4.length*4 > a1.length) return;
+    for (int i = 0; i < p4.length; i+=1) {
+      long l0 = (long)a1[i*4+0];
+      long l1 = (long)a1[i*4+1];
+      long l2 = (long)a1[i*4+2];
+      long l3 = (long)a1[i*4+3];
+      p4[i] = (l3 & 0xFFFFl) |
+             ((l2 & 0xFFFFl) << 16) |
+             ((l1 & 0xFFFFl) << 32) |
+             ((l0 & 0xFFFFl) << 48);
+    }
+  }
+  static void test_unpack4_swap(char[] a0, long[] p4) {
+    if (p4.length*4 > a0.length) return;
+    for (int i = 0; i < p4.length; i+=1) {
+      long l = p4[i];
+      a0[i*4+0] = (char)(l >> 48);
+      a0[i*4+1] = (char)(l >> 32);
+      a0[i*4+2] = (char)(l >> 16);
+      a0[i*4+3] = (char)(l & 0xFFFFl);
+    }
+  }
+
+  static int verify(String text, int i, int elem, int val) {
+    if (elem != val) {
+      System.err.println(text + "[" + i + "] = " + elem + " != " + val);
+      return 1;
+    }
+    return 0;
+  }
+
+  static int verify(String text, int i, long elem, long val) {
+    if (elem != val) {
+      System.err.println(text + "[" + i + "] = " + Long.toHexString(elem) + " != " + Long.toHexString(val));
+      return 1;
+    }
+    return 0;
+  }
+}