Merge

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);
 

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

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;

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 \

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

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) {

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); }

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;

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; }

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;

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

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()) {

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() &&

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) %{

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;
 

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;
 

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) {

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;
 

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() {

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 {

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")                                                            \
                                                                                                                         \

src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.cpp

+/*
+ * 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>;

src/share/vm/gc_implementation/concurrentMarkSweep/adaptiveFreeList.hpp

+/*
+ * 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

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.

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);

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;
   }

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;

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)                       \

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:

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();
 }

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; }

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);

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>
@@ -344,24 +345,13 @@ 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"};
 

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();

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>;

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,

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 {
@@ -293,41 +299,15 @@ 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

src/share/vm/memory/metablock.hpp

+/*
+ * 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

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);
 

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 &&

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();