Mercurial > hg > openjdk > hsx19

changeset 690:922aedc96ef5

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Merge
author ysr
date Fri, 03 Apr 2009 15:59:19 -0700
parents f18338cf04b0 (current diff) 2c1dbb844832 (diff)
children fafab5d5349c b9fba36710f2
files src/share/vm/runtime/arguments.cpp src/share/vm/runtime/globals.hpp
diffstat 110 files changed, 3260 insertions(+), 1005 deletions(-) [+]
line wrap: on
line diff
--- a/.hgtags	Tue Mar 03 14:23:18 2009 -0800
+++ b/.hgtags	Fri Apr 03 15:59:19 2009 -0700
@@ -25,3 +25,4 @@
 bcb33806d186561c781992e5f4d8a90bb033f9f0 jdk7-b48
 8b22ccb5aba2c6c11bddf6488a7bb7ef5b4bf2be jdk7-b49
 dae503d9f04c1a11e182dbf7f770509c28dc0609 jdk7-b50
+2581d90c6c9b2012da930eb4742add94a03069a0 jdk7-b51
--- a/agent/src/share/classes/sun/jvm/hotspot/runtime/StubRoutines.java	Tue Mar 03 14:23:18 2009 -0800
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/StubRoutines.java	Fri Apr 03 15:59:19 2009 -0700
@@ -46,12 +46,18 @@
     Type type = db.lookupType("StubRoutines");
 
     callStubReturnAddressField = type.getAddressField("_call_stub_return_address");
-    // Only some platforms have specif return from compiled to call_stub
+    // Only some platforms have specific return from compiled to call_stub
     try {
-      callStubCompiledReturnAddressField = type.getAddressField("_call_stub_compiled_return");
+      type = db.lookupType("StubRoutines::x86");
+      if (type != null) {
+        callStubCompiledReturnAddressField = type.getAddressField("_call_stub_compiled_return");
+      }
     } catch (RuntimeException re) {
       callStubCompiledReturnAddressField = null;
     }
+    if (callStubCompiledReturnAddressField == null && VM.getVM().getCPU().equals("x86")) {
+      throw new InternalError("Missing definition for _call_stub_compiled_return");
+    }
   }
 
   public StubRoutines() {
--- a/agent/src/share/classes/sun/jvm/hotspot/runtime/VM.java	Tue Mar 03 14:23:18 2009 -0800
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/VM.java	Fri Apr 03 15:59:19 2009 -0700
@@ -342,11 +342,14 @@
       throw new RuntimeException("Attempt to initialize VM twice");
     }
     soleInstance = new VM(db, debugger, debugger.getMachineDescription().isBigEndian());
+
     debugger.putHeapConst(soleInstance.getHeapOopSize(), Universe.getNarrowOopBase(),
                           Universe.getNarrowOopShift());
+
     for (Iterator iter = vmInitializedObservers.iterator(); iter.hasNext(); ) {
       ((Observer) iter.next()).update(null, null);
     }
+
   }
 
   /** This is used by the debugging system */
--- a/make/hotspot_version	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/hotspot_version	Fri Apr 03 15:59:19 2009 -0700
@@ -35,7 +35,7 @@
 
 HS_MAJOR_VER=15
 HS_MINOR_VER=0
-HS_BUILD_NUMBER=03
+HS_BUILD_NUMBER=04
 
 JDK_MAJOR_VER=1
 JDK_MINOR_VER=7
--- a/make/windows/build_vm_def.sh	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/build_vm_def.sh	Fri Apr 03 15:59:19 2009 -0700
@@ -52,6 +52,19 @@
 RM="$MKS_HOME/rm.exe"
 DUMPBIN="link.exe /dump"
 
+# When called from IDE the first param should contain the link version, otherwise may be nill
+if [ "x$1" != "x" ]; then
+LINK_VER="$1"
+fi
+
+if [ "x$LINK_VER" != "x800" -a  "x$LINK_VER" != "x900" ]; then
 $DUMPBIN /symbols *.obj | "$GREP" "??_7.*@@6B@" | "$AWK" '{print $7}' | "$SORT" | "$UNIQ" > vm2.def
+else
+# Can't use pipes when calling cl.exe or link.exe from IDE. Using transit file vm3.def
+$DUMPBIN /OUT:vm3.def /symbols *.obj 
+"$CAT" vm3.def | "$GREP" "??_7.*@@6B@" | "$AWK" '{print $7}' | "$SORT" | "$UNIQ" > vm2.def
+"$RM" -f vm3.def
+fi
+
 "$CAT" vm1.def vm2.def > vm.def
 "$RM" -f vm1.def vm2.def
--- a/make/windows/create.bat	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/create.bat	Fri Apr 03 15:59:19 2009 -0700
@@ -72,12 +72,20 @@
 for /F %%i in ('sh %HotSpotWorkSpace%/make/windows/get_msc_ver.sh') do set %%i
 
 echo **************************************************************
+set ProjectFile=vm.vcproj
 if "%MSC_VER%" == "1200" (
 set ProjectFile=vm.dsp
 echo Will generate VC6 project {unsupported}
 ) else (
-set ProjectFile=vm.vcproj
-echo Will generate VC7 project
+if "%MSC_VER%" == "1400" (
+echo Will generate VC8 {Visual Studio 2005}
+) else (
+if "%MSC_VER%" == "1500" (
+echo Will generate VC9 {Visual Studio 2008}
+) else (
+echo Will generate VC7 project {Visual Studio 2003 .NET}
+)
+)
 )
 echo                            %ProjectFile%
 echo **************************************************************
--- a/make/windows/makefiles/adlc.make	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/makefiles/adlc.make	Fri Apr 03 15:59:19 2009 -0700
@@ -46,6 +46,7 @@
 ADLCFLAGS=-q -T -U_LP64
 !endif
 
+CPP_FLAGS=$(CPP_FLAGS) /D _CRT_SECURE_NO_WARNINGS /D _CRT_SECURE_NO_DEPRECATE  
 
 CPP_INCLUDE_DIRS=\
   /I "..\generated"                          \
--- a/make/windows/makefiles/compile.make	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/makefiles/compile.make	Fri Apr 03 15:59:19 2009 -0700
@@ -170,8 +170,6 @@
 # Manifest Tool - used in VS2005 and later to adjust manifests stored
 # as resources inside build artifacts.
 MT=mt.exe
-# VS2005 and later restricts the use of certain libc functions without this
-CPP_FLAGS=$(CPP_FLAGS) /D _CRT_SECURE_NO_DEPRECATE
 !endif
 
 !if "$(COMPILER_NAME)" == "VS2008"
@@ -183,8 +181,6 @@
 # Manifest Tool - used in VS2005 and later to adjust manifests stored
 # as resources inside build artifacts.
 MT=mt.exe
-# VS2005 and later restricts the use of certain libc functions without this
-CPP_FLAGS=$(CPP_FLAGS) /D _CRT_SECURE_NO_DEPRECATE
 !endif
 
 # Compile for space above time.
--- a/make/windows/makefiles/makedeps.make	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/makefiles/makedeps.make	Fri Apr 03 15:59:19 2009 -0700
@@ -48,6 +48,8 @@
         $(WorkSpace)\src\share\tools\MakeDeps\WinGammaPlatform.java \
         $(WorkSpace)\src\share\tools\MakeDeps\WinGammaPlatformVC6.java \
         $(WorkSpace)\src\share\tools\MakeDeps\WinGammaPlatformVC7.java \
+        $(WorkSpace)\src\share\tools\MakeDeps\WinGammaPlatformVC8.java \
+        $(WorkSpace)\src\share\tools\MakeDeps\WinGammaPlatformVC9.java \
         $(WorkSpace)\src\share\tools\MakeDeps\Util.java \
         $(WorkSpace)\src\share\tools\MakeDeps\BuildConfig.java \
         $(WorkSpace)\src\share\tools\MakeDeps\ArgsParser.java
@@ -121,7 +123,7 @@
         -additionalFile includeDB_gc_shared \
         -additionalFile includeDB_gc_serial \
         -additionalGeneratedFile $(HOTSPOTBUILDSPACE)\%f\%b vm.def \
-        -prelink  "" "Generating vm.def..." "cd $(HOTSPOTBUILDSPACE)\%f\%b	$(HOTSPOTMKSHOME)\sh $(HOTSPOTWORKSPACE)\make\windows\build_vm_def.sh" \
+        -prelink  "" "Generating vm.def..." "cd $(HOTSPOTBUILDSPACE)\%f\%b	set HOTSPOTMKSHOME=$(HOTSPOTMKSHOME)	$(HOTSPOTMKSHOME)\sh $(HOTSPOTWORKSPACE)\make\windows\build_vm_def.sh $(LINK_VER)" \
        $(MakeDepsIncludesPRIVATE)
 
 # Add in build-specific options
--- a/make/windows/makefiles/rules.make	Tue Mar 03 14:23:18 2009 -0800
+++ b/make/windows/makefiles/rules.make	Fri Apr 03 15:59:19 2009 -0700
@@ -42,10 +42,23 @@
 BOOT_JAVA_HOME=
 !endif
 
+ProjectFile=vm.vcproj
+
 !if "$(MSC_VER)" == "1200"
+
 VcVersion=VC6
 ProjectFile=vm.dsp
+
+!elseif "$(MSC_VER)" == "1400"
+
+VcVersion=VC8
+
+!elseif "$(MSC_VER)" == "1500"
+
+VcVersion=VC9
+
 !else
+
 VcVersion=VC7
-ProjectFile=vm.vcproj
+
 !endif
--- a/src/cpu/sparc/vm/assembler_sparc.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/assembler_sparc.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2615,12 +2615,12 @@
   }
 }
 
-RegisterConstant MacroAssembler::delayed_value(intptr_t* delayed_value_addr,
-                                               Register tmp,
-                                               int offset) {
+RegisterOrConstant MacroAssembler::delayed_value_impl(intptr_t* delayed_value_addr,
+                                                      Register tmp,
+                                                      int offset) {
   intptr_t value = *delayed_value_addr;
   if (value != 0)
-    return RegisterConstant(value + offset);
+    return RegisterOrConstant(value + offset);
 
   // load indirectly to solve generation ordering problem
   Address a(tmp, (address) delayed_value_addr);
@@ -2634,11 +2634,11 @@
   if (offset != 0)
     add(tmp, offset, tmp);
 
-  return RegisterConstant(tmp);
+  return RegisterOrConstant(tmp);
 }
 
 
-void MacroAssembler::regcon_inc_ptr( RegisterConstant& dest, RegisterConstant src, Register temp ) {
+void MacroAssembler::regcon_inc_ptr( RegisterOrConstant& dest, RegisterOrConstant src, Register temp ) {
   assert(dest.register_or_noreg() != G0, "lost side effect");
   if ((src.is_constant() && src.as_constant() == 0) ||
       (src.is_register() && src.as_register() == G0)) {
@@ -2647,15 +2647,15 @@
     add(dest.as_register(), ensure_rs2(src, temp), dest.as_register());
   } else if (src.is_constant()) {
     intptr_t res = dest.as_constant() + src.as_constant();
-    dest = RegisterConstant(res); // side effect seen by caller
+    dest = RegisterOrConstant(res); // side effect seen by caller
   } else {
     assert(temp != noreg, "cannot handle constant += register");
     add(src.as_register(), ensure_rs2(dest, temp), temp);
-    dest = RegisterConstant(temp); // side effect seen by caller
+    dest = RegisterOrConstant(temp); // side effect seen by caller
   }
 }
 
-void MacroAssembler::regcon_sll_ptr( RegisterConstant& dest, RegisterConstant src, Register temp ) {
+void MacroAssembler::regcon_sll_ptr( RegisterOrConstant& dest, RegisterOrConstant src, Register temp ) {
   assert(dest.register_or_noreg() != G0, "lost side effect");
   if (!is_simm13(src.constant_or_zero()))
     src = (src.as_constant() & 0xFF);
@@ -2666,12 +2666,12 @@
     sll_ptr(dest.as_register(), src, dest.as_register());
   } else if (src.is_constant()) {
     intptr_t res = dest.as_constant() << src.as_constant();
-    dest = RegisterConstant(res); // side effect seen by caller
+    dest = RegisterOrConstant(res); // side effect seen by caller
   } else {
     assert(temp != noreg, "cannot handle constant <<= register");
     set(dest.as_constant(), temp);
     sll_ptr(temp, src, temp);
-    dest = RegisterConstant(temp); // side effect seen by caller
+    dest = RegisterOrConstant(temp); // side effect seen by caller
   }
 }
 
@@ -2683,7 +2683,7 @@
 // On failure, execution transfers to the given label.
 void MacroAssembler::lookup_interface_method(Register recv_klass,
                                              Register intf_klass,
-                                             RegisterConstant itable_index,
+                                             RegisterOrConstant itable_index,
                                              Register method_result,
                                              Register scan_temp,
                                              Register sethi_temp,
@@ -2720,7 +2720,7 @@
   add(recv_klass, scan_temp, scan_temp);
 
   // Adjust recv_klass by scaled itable_index, so we can free itable_index.
-  RegisterConstant itable_offset = itable_index;
+  RegisterOrConstant itable_offset = itable_index;
   regcon_sll_ptr(itable_offset, exact_log2(itableMethodEntry::size() * wordSize));
   regcon_inc_ptr(itable_offset, itableMethodEntry::method_offset_in_bytes());
   add(recv_klass, ensure_rs2(itable_offset, sethi_temp), recv_klass);
@@ -2805,7 +2805,7 @@
                                                    Label* L_success,
                                                    Label* L_failure,
                                                    Label* L_slow_path,
-                                        RegisterConstant super_check_offset,
+                                        RegisterOrConstant super_check_offset,
                                         Register instanceof_hack) {
   int sc_offset = (klassOopDesc::header_size() * HeapWordSize +
                    Klass::secondary_super_cache_offset_in_bytes());
@@ -2867,7 +2867,7 @@
   if (must_load_sco) {
     // The super check offset is always positive...
     lduw(super_klass, sco_offset, temp2_reg);
-    super_check_offset = RegisterConstant(temp2_reg);
+    super_check_offset = RegisterOrConstant(temp2_reg);
   }
   ld_ptr(sub_klass, super_check_offset, temp_reg);
   cmp(super_klass, temp_reg);
@@ -4465,7 +4465,7 @@
 }
 
 // Loading values by size and signed-ness
-void MacroAssembler::load_sized_value(Register s1, RegisterConstant s2, Register d,
+void MacroAssembler::load_sized_value(Register s1, RegisterOrConstant s2, Register d,
                                       int size_in_bytes, bool is_signed) {
   switch (size_in_bytes ^ (is_signed ? -1 : 0)) {
   case ~8:  // fall through:
--- a/src/cpu/sparc/vm/assembler_sparc.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/assembler_sparc.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1088,8 +1088,8 @@
   inline void add(    Register s1, Register s2, Register d );
   inline void add(    Register s1, int simm13a, Register d, relocInfo::relocType rtype = relocInfo::none);
   inline void add(    Register s1, int simm13a, Register d, RelocationHolder const& rspec);
-  inline void add(    Register s1, RegisterConstant s2, Register d, int offset = 0);
-  inline void add(    const Address&  a,              Register d, int offset = 0);
+  inline void add(    Register s1, RegisterOrConstant s2, Register d, int offset = 0);
+  inline void add(    const Address&  a,                  Register d, int offset = 0);
 
   void addcc(  Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3  | cc_bit_op3) | rs1(s1) | rs2(s2) ); }
   void addcc(  Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3  | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); }
@@ -1305,15 +1305,15 @@
   inline void ld(   const Address& a, Register d, int offset = 0 );
   inline void ldd(  const Address& a, Register d, int offset = 0 );
 
-  inline void ldub(  Register s1, RegisterConstant s2, Register d );
-  inline void ldsb(  Register s1, RegisterConstant s2, Register d );
-  inline void lduh(  Register s1, RegisterConstant s2, Register d );
-  inline void ldsh(  Register s1, RegisterConstant s2, Register d );
-  inline void lduw(  Register s1, RegisterConstant s2, Register d );
-  inline void ldsw(  Register s1, RegisterConstant s2, Register d );
-  inline void ldx(   Register s1, RegisterConstant s2, Register d );
-  inline void ld(    Register s1, RegisterConstant s2, Register d );
-  inline void ldd(   Register s1, RegisterConstant s2, Register d );
+  inline void ldub(  Register s1, RegisterOrConstant s2, Register d );
+  inline void ldsb(  Register s1, RegisterOrConstant s2, Register d );
+  inline void lduh(  Register s1, RegisterOrConstant s2, Register d );
+  inline void ldsh(  Register s1, RegisterOrConstant s2, Register d );
+  inline void lduw(  Register s1, RegisterOrConstant s2, Register d );
+  inline void ldsw(  Register s1, RegisterOrConstant s2, Register d );
+  inline void ldx(   Register s1, RegisterOrConstant s2, Register d );
+  inline void ld(    Register s1, RegisterOrConstant s2, Register d );
+  inline void ldd(   Register s1, RegisterOrConstant s2, Register d );
 
   // pp 177
 
@@ -1535,12 +1535,12 @@
   inline void st(   Register d, const Address& a, int offset = 0 );
   inline void std(  Register d, const Address& a, int offset = 0 );
 
-  inline void stb(  Register d, Register s1, RegisterConstant s2 );
-  inline void sth(  Register d, Register s1, RegisterConstant s2 );
-  inline void stw(  Register d, Register s1, RegisterConstant s2 );
-  inline void stx(  Register d, Register s1, RegisterConstant s2 );
-  inline void std(  Register d, Register s1, RegisterConstant s2 );
-  inline void st(   Register d, Register s1, RegisterConstant s2 );
+  inline void stb(  Register d, Register s1, RegisterOrConstant s2 );
+  inline void sth(  Register d, Register s1, RegisterOrConstant s2 );
+  inline void stw(  Register d, Register s1, RegisterOrConstant s2 );
+  inline void stx(  Register d, Register s1, RegisterOrConstant s2 );
+  inline void std(  Register d, Register s1, RegisterOrConstant s2 );
+  inline void st(   Register d, Register s1, RegisterOrConstant s2 );
 
   // pp 177
 
@@ -1859,7 +1859,7 @@
   // Functions for isolating 64 bit shifts for LP64
   inline void sll_ptr( Register s1, Register s2, Register d );
   inline void sll_ptr( Register s1, int imm6a,   Register d );
-  inline void sll_ptr( Register s1, RegisterConstant s2, Register d );
+  inline void sll_ptr( Register s1, RegisterOrConstant s2, Register d );
   inline void srl_ptr( Register s1, Register s2, Register d );
   inline void srl_ptr( Register s1, int imm6a,   Register d );
 
@@ -1965,26 +1965,26 @@
   // st_ptr will perform st for 32 bit VM's and stx for 64 bit VM's
   inline void ld_ptr(   Register s1, Register s2, Register d );
   inline void ld_ptr(   Register s1, int simm13a, Register d);
-  inline void ld_ptr(   Register s1, RegisterConstant s2, Register d );
+  inline void ld_ptr(   Register s1, RegisterOrConstant s2, Register d );
   inline void ld_ptr(  const Address& a, Register d, int offset = 0 );
   inline void st_ptr(  Register d, Register s1, Register s2 );
   inline void st_ptr(  Register d, Register s1, int simm13a);
-  inline void st_ptr(  Register d, Register s1, RegisterConstant s2 );
+  inline void st_ptr(  Register d, Register s1, RegisterOrConstant s2 );
   inline void st_ptr(  Register d, const Address& a, int offset = 0 );
 
   // ld_long will perform ld for 32 bit VM's and ldx for 64 bit VM's
   // st_long will perform st for 32 bit VM's and stx for 64 bit VM's
   inline void ld_long( Register s1, Register s2, Register d );
   inline void ld_long( Register s1, int simm13a, Register d );
-  inline void ld_long( Register s1, RegisterConstant s2, Register d );
+  inline void ld_long( Register s1, RegisterOrConstant s2, Register d );
   inline void ld_long( const Address& a, Register d, int offset = 0 );
   inline void st_long( Register d, Register s1, Register s2 );
   inline void st_long( Register d, Register s1, int simm13a );
-  inline void st_long( Register d, Register s1, RegisterConstant s2 );
+  inline void st_long( Register d, Register s1, RegisterOrConstant s2 );
   inline void st_long( Register d, const Address& a, int offset = 0 );
 
   // Loading values by size and signed-ness
-  void load_sized_value(Register s1, RegisterConstant s2, Register d,
+  void load_sized_value(Register s1, RegisterOrConstant s2, Register d,
                         int size_in_bytes, bool is_signed);
 
   // Helpers for address formation.
@@ -1994,11 +1994,11 @@
   // is required, and becomes the result.
   // If dest is a register and src is a non-simm13 constant,
   // the temp argument is required, and is used to materialize the constant.
-  void regcon_inc_ptr( RegisterConstant& dest, RegisterConstant src,
+  void regcon_inc_ptr( RegisterOrConstant& dest, RegisterOrConstant src,
                        Register temp = noreg );
-  void regcon_sll_ptr( RegisterConstant& dest, RegisterConstant src,
+  void regcon_sll_ptr( RegisterOrConstant& dest, RegisterOrConstant src,
                        Register temp = noreg );
-  RegisterConstant ensure_rs2(RegisterConstant rs2, Register sethi_temp) {
+  RegisterOrConstant ensure_rs2(RegisterOrConstant rs2, Register sethi_temp) {
     guarantee(sethi_temp != noreg, "constant offset overflow");
     if (is_simm13(rs2.constant_or_zero()))
       return rs2;               // register or short constant
@@ -2322,7 +2322,7 @@
   // interface method calling
   void lookup_interface_method(Register recv_klass,
                                Register intf_klass,
-                               RegisterConstant itable_index,
+                               RegisterOrConstant itable_index,
                                Register method_result,
                                Register temp_reg, Register temp2_reg,
                                Label& no_such_interface);
@@ -2341,7 +2341,7 @@
                                      Label* L_success,
                                      Label* L_failure,
                                      Label* L_slow_path,
-                RegisterConstant super_check_offset = RegisterConstant(-1),
+                RegisterOrConstant super_check_offset = RegisterOrConstant(-1),
                 Register instanceof_hack = noreg);
 
   // The rest of the type check; must be wired to a corresponding fast path.
@@ -2381,7 +2381,7 @@
   // stack overflow + shadow pages.  Clobbers tsp and scratch registers.
   void bang_stack_size(Register Rsize, Register Rtsp, Register Rscratch);
 
-  virtual RegisterConstant delayed_value(intptr_t* delayed_value_addr, Register tmp, int offset);
+  virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr, Register tmp, int offset);
 
   void verify_tlab();
 
--- a/src/cpu/sparc/vm/assembler_sparc.inline.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/assembler_sparc.inline.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -143,45 +143,45 @@
 inline void Assembler::ld(  Register s1, int simm13a, Register d) { lduw( s1, simm13a, d); }
 #endif
 
-inline void Assembler::ldub(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldub(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsb(s1, s2.as_register(), d);
   else                   ldsb(s1, s2.as_constant(), d);
 }
-inline void Assembler::ldsb(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldsb(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsb(s1, s2.as_register(), d);
   else                   ldsb(s1, s2.as_constant(), d);
 }
-inline void Assembler::lduh(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::lduh(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsh(s1, s2.as_register(), d);
   else                   ldsh(s1, s2.as_constant(), d);
 }
-inline void Assembler::ldsh(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldsh(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsh(s1, s2.as_register(), d);
   else                   ldsh(s1, s2.as_constant(), d);
 }
-inline void Assembler::lduw(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::lduw(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsw(s1, s2.as_register(), d);
   else                   ldsw(s1, s2.as_constant(), d);
 }
-inline void Assembler::ldsw(  Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldsw(  Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldsw(s1, s2.as_register(), d);
   else                   ldsw(s1, s2.as_constant(), d);
 }
-inline void Assembler::ldx(   Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldx(   Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldx(s1, s2.as_register(), d);
   else                   ldx(s1, s2.as_constant(), d);
 }
-inline void Assembler::ld(    Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ld(    Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ld(s1, s2.as_register(), d);
   else                   ld(s1, s2.as_constant(), d);
 }
-inline void Assembler::ldd(   Register s1, RegisterConstant s2, Register d) {
+inline void Assembler::ldd(   Register s1, RegisterOrConstant s2, Register d) {
   if (s2.is_register())  ldd(s1, s2.as_register(), d);
   else                   ldd(s1, s2.as_constant(), d);
 }
 
 // form effective addresses this way:
-inline void Assembler::add(   Register s1, RegisterConstant s2, Register d, int offset) {
+inline void Assembler::add(   Register s1, RegisterOrConstant s2, Register d, int offset) {
   if (s2.is_register())  add(s1, s2.as_register(), d);
   else                 { add(s1, s2.as_constant() + offset, d); offset = 0; }
   if (offset != 0)       add(d,  offset,                    d);
@@ -243,23 +243,23 @@
 inline void Assembler::st(  Register d, Register s1, Register s2) { stw(d, s1, s2); }
 inline void Assembler::st(  Register d, Register s1, int simm13a) { stw(d, s1, simm13a); }
 
-inline void Assembler::stb(  Register d, Register s1, RegisterConstant s2) {
+inline void Assembler::stb(  Register d, Register s1, RegisterOrConstant s2) {
   if (s2.is_register())  stb(d, s1, s2.as_register());
   else                   stb(d, s1, s2.as_constant());
 }
-inline void Assembler::sth(  Register d, Register s1, RegisterConstant s2) {
+inline void Assembler::sth(  Register d, Register s1, RegisterOrConstant s2) {
   if (s2.is_register())  sth(d, s1, s2.as_register());
   else                   sth(d, s1, s2.as_constant());
 }
-inline void Assembler::stx(  Register d, Register s1, RegisterConstant s2) {
+inline void Assembler::stx(  Register d, Register s1, RegisterOrConstant s2) {
   if (s2.is_register())  stx(d, s1, s2.as_register());
   else                   stx(d, s1, s2.as_constant());
 }
-inline void Assembler::std( Register d, Register s1, RegisterConstant s2) {
+inline void Assembler::std( Register d, Register s1, RegisterOrConstant s2) {
   if (s2.is_register())  std(d, s1, s2.as_register());
   else                   std(d, s1, s2.as_constant());
 }
-inline void Assembler::st(  Register d, Register s1, RegisterConstant s2) {
+inline void Assembler::st(  Register d, Register s1, RegisterOrConstant s2) {
   if (s2.is_register())  st(d, s1, s2.as_register());
   else                   st(d, s1, s2.as_constant());
 }
@@ -308,7 +308,7 @@
 #endif
 }
 
-inline void MacroAssembler::ld_ptr( Register s1, RegisterConstant s2, Register d ) {
+inline void MacroAssembler::ld_ptr( Register s1, RegisterOrConstant s2, Register d ) {
 #ifdef _LP64
   Assembler::ldx( s1, s2, d);
 #else
@@ -340,7 +340,7 @@
 #endif
 }
 
-inline void MacroAssembler::st_ptr( Register d, Register s1, RegisterConstant s2 ) {
+inline void MacroAssembler::st_ptr( Register d, Register s1, RegisterOrConstant s2 ) {
 #ifdef _LP64
   Assembler::stx( d, s1, s2);
 #else
@@ -373,7 +373,7 @@
 #endif
 }
 
-inline void MacroAssembler::ld_long( Register s1, RegisterConstant s2, Register d ) {
+inline void MacroAssembler::ld_long( Register s1, RegisterOrConstant s2, Register d ) {
 #ifdef _LP64
   Assembler::ldx(s1, s2, d);
 #else
@@ -405,7 +405,7 @@
 #endif
 }
 
-inline void MacroAssembler::st_long( Register d, Register s1, RegisterConstant s2 ) {
+inline void MacroAssembler::st_long( Register d, Register s1, RegisterOrConstant s2 ) {
 #ifdef _LP64
   Assembler::stx(d, s1, s2);
 #else
@@ -455,7 +455,7 @@
 #endif
 }
 
-inline void MacroAssembler::sll_ptr( Register s1, RegisterConstant s2, Register d ) {
+inline void MacroAssembler::sll_ptr( Register s1, RegisterOrConstant s2, Register d ) {
   if (s2.is_register())  sll_ptr(s1, s2.as_register(), d);
   else                   sll_ptr(s1, s2.as_constant(), d);
 }
--- a/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 2000-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 2000-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -2489,7 +2489,7 @@
         __ check_klass_subtype_fast_path(klass_RInfo, k_RInfo, Rtmp1, noreg,
                                          (need_slow_path ? &done : NULL),
                                          stub->entry(), NULL,
-                                         RegisterConstant(k->super_check_offset()));
+                                         RegisterOrConstant(k->super_check_offset()));
       } else {
         // perform the fast part of the checking logic
         __ check_klass_subtype_fast_path(klass_RInfo, k_RInfo, Rtmp1, O7,
@@ -2550,14 +2550,14 @@
         __ check_klass_subtype_fast_path(klass_RInfo, k_RInfo, O7, noreg,
                                          (need_slow_path ? &done : NULL),
                                          (need_slow_path ? &done : NULL), NULL,
-                                         RegisterConstant(k->super_check_offset()),
+                                         RegisterOrConstant(k->super_check_offset()),
                                          dst);
       } else {
         assert(dst != klass_RInfo && dst != k_RInfo, "need 3 registers");
         // perform the fast part of the checking logic
         __ check_klass_subtype_fast_path(klass_RInfo, k_RInfo, O7, dst,
                                          &done, &done, NULL,
-                                         RegisterConstant(-1),
+                                         RegisterOrConstant(-1),
                                          dst);
       }
       if (need_slow_path) {
--- a/src/cpu/sparc/vm/globals_sparc.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/globals_sparc.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -46,6 +46,7 @@
 define_pd_global(uintx, NewSize, ScaleForWordSize((2048 * K) + (2 * (64 * K))));
 define_pd_global(intx,  SurvivorRatio,         8);
 define_pd_global(intx,  InlineFrequencyCount,  50);  // we can use more inlining on the SPARC
+define_pd_global(intx,  InlineSmallCode,       1500);
 #ifdef _LP64
 // Stack slots are 2X larger in LP64 than in the 32 bit VM.
 define_pd_global(intx,  ThreadStackSize,       1024);
--- a/src/cpu/sparc/vm/sparc.ad	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/sparc.ad	Fri Apr 03 15:59:19 2009 -0700
@@ -3003,6 +3003,202 @@
     __ bind(Ldone);
   %}
 
+enc_class enc_String_Equals(o0RegP str1, o1RegP str2, g3RegP tmp1, g4RegP tmp2, notemp_iRegI result) %{
+    Label Lword, Lword_loop, Lpost_word, Lchar, Lchar_loop, Ldone;
+    MacroAssembler _masm(&cbuf);
+
+    Register   str1_reg = reg_to_register_object($str1$$reg);
+    Register   str2_reg = reg_to_register_object($str2$$reg);
+    Register   tmp1_reg = reg_to_register_object($tmp1$$reg);
+    Register   tmp2_reg = reg_to_register_object($tmp2$$reg);
+    Register result_reg = reg_to_register_object($result$$reg);
+
+    // Get the first character position in both strings
+    //         [8] char array, [12] offset, [16] count
+    int  value_offset = java_lang_String:: value_offset_in_bytes();
+    int offset_offset = java_lang_String::offset_offset_in_bytes();
+    int  count_offset = java_lang_String:: count_offset_in_bytes();
+
+    // load str1 (jchar*) base address into tmp1_reg
+    __ load_heap_oop(Address(str1_reg, 0,  value_offset), tmp1_reg);
+    __ ld(Address(str1_reg, 0, offset_offset), result_reg);
+    __ add(tmp1_reg, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp1_reg);
+    __    ld(Address(str1_reg, 0, count_offset), str1_reg); // hoisted
+    __ sll(result_reg, exact_log2(sizeof(jchar)), result_reg);
+    __    load_heap_oop(Address(str2_reg, 0,  value_offset), tmp2_reg); // hoisted
+    __ add(result_reg, tmp1_reg, tmp1_reg);
+
+    // load str2 (jchar*) base address into tmp2_reg
+    // __ ld_ptr(Address(str2_reg, 0,  value_offset), tmp2_reg); // hoisted
+    __ ld(Address(str2_reg, 0, offset_offset), result_reg);
+    __ add(tmp2_reg, arrayOopDesc::base_offset_in_bytes(T_CHAR), tmp2_reg);
+    __    ld(Address(str2_reg, 0, count_offset), str2_reg); // hoisted
+    __ sll(result_reg, exact_log2(sizeof(jchar)), result_reg);
+    __   cmp(str1_reg, str2_reg); // hoisted
+    __ add(result_reg, tmp2_reg, tmp2_reg);
+
+    __ sll(str1_reg, exact_log2(sizeof(jchar)), str1_reg);
+    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg);    // not equal
+
+    __ br_zero(Assembler::equal, true, Assembler::pn, str1_reg, Ldone);
+    __ delayed()->add(G0, 1, result_reg); //equals
+
+    __ cmp(tmp1_reg, tmp2_reg); //same string ?
+    __ brx(Assembler::equal, true, Assembler::pn, Ldone);
+    __ delayed()->add(G0, 1, result_reg);
+
+    //rename registers
+    Register limit_reg =   str1_reg;
+    Register  chr2_reg =   str2_reg;
+    Register  chr1_reg = result_reg;
+    // tmp{12} are the base pointers
+
+    //check for alignment and position the pointers to the ends
+    __ or3(tmp1_reg, tmp2_reg, chr1_reg);
+    __ andcc(chr1_reg, 0x3, chr1_reg); // notZero means at least one not 4-byte aligned
+    __ br(Assembler::notZero, false, Assembler::pn, Lchar);
+    __ delayed()->nop();
+
+    __ bind(Lword);
+    __ and3(limit_reg, 0x2, O7); //remember the remainder (either 0 or 2)
+    __ andn(limit_reg, 0x3, limit_reg);
+    __ br_zero(Assembler::zero, false, Assembler::pn, limit_reg, Lpost_word);
+    __ delayed()->nop();
+
+    __ add(tmp1_reg, limit_reg, tmp1_reg);
+    __ add(tmp2_reg, limit_reg, tmp2_reg);
+    __ neg(limit_reg);
+
+    __ lduw(tmp1_reg, limit_reg, chr1_reg);
+    __ bind(Lword_loop);
+    __ lduw(tmp2_reg, limit_reg, chr2_reg);
+    __ cmp(chr1_reg, chr2_reg);
+    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg);
+    __ inccc(limit_reg, 2*sizeof(jchar));
+    // annul LDUW if branch i  s not taken to prevent access past end of string
+    __ br(Assembler::notZero, true, Assembler::pt, Lword_loop); //annul on taken
+    __ delayed()->lduw(tmp1_reg, limit_reg, chr1_reg); // hoisted
+
+    __ bind(Lpost_word);
+    __ br_zero(Assembler::zero, true, Assembler::pt, O7, Ldone);
+    __ delayed()->add(G0, 1, result_reg);
+
+    __ lduh(tmp1_reg, 0, chr1_reg);
+    __ lduh(tmp2_reg, 0, chr2_reg);
+    __ cmp (chr1_reg, chr2_reg);
+    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg);
+    __ ba(false,Ldone);
+    __ delayed()->add(G0, 1, result_reg);
+
+    __ bind(Lchar);
+    __ add(tmp1_reg, limit_reg, tmp1_reg);
+    __ add(tmp2_reg, limit_reg, tmp2_reg);
+    __ neg(limit_reg); //negate count
+
+    __ lduh(tmp1_reg, limit_reg, chr1_reg);
+    __ bind(Lchar_loop);
+    __ lduh(tmp2_reg, limit_reg, chr2_reg);
+    __ cmp(chr1_reg, chr2_reg);
+    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg); //not equal
+    __ inccc(limit_reg, sizeof(jchar));
+    // annul LDUH if branch is not taken to prevent access past end of string
+    __ br(Assembler::notZero, true, Assembler::pt, Lchar_loop); //annul on taken
+    __ delayed()->lduh(tmp1_reg, limit_reg, chr1_reg); // hoisted
+
+    __ add(G0, 1, result_reg);  //equal
+
+    __ bind(Ldone);
+  %}
+
+enc_class enc_Array_Equals(o0RegP ary1, o1RegP ary2, g3RegP tmp1, g4RegP tmp2, notemp_iRegI result) %{
+    Label Lvector, Ldone, Lloop;
+    MacroAssembler _masm(&cbuf);
+
+    Register   ary1_reg = reg_to_register_object($ary1$$reg);
+    Register   ary2_reg = reg_to_register_object($ary2$$reg);
+    Register   tmp1_reg = reg_to_register_object($tmp1$$reg);
+    Register   tmp2_reg = reg_to_register_object($tmp2$$reg);
+    Register result_reg = reg_to_register_object($result$$reg);
+
+    int length_offset  = arrayOopDesc::length_offset_in_bytes();
+    int base_offset    = arrayOopDesc::base_offset_in_bytes(T_CHAR);
+
+    // return true if the same array
+    __ cmp(ary1_reg, ary2_reg);
+    __ br(Assembler::equal, true, Assembler::pn, Ldone);
+    __ delayed()->add(G0, 1, result_reg); // equal
+
+    __ br_null(ary1_reg, true, Assembler::pn, Ldone);
+    __ delayed()->mov(G0, result_reg);    // not equal
+
+    __ br_null(ary2_reg, true, Assembler::pn, Ldone);
+    __ delayed()->mov(G0, result_reg);    // not equal
+
+    //load the lengths of arrays
+    __ ld(Address(ary1_reg, 0, length_offset), tmp1_reg);
+    __ ld(Address(ary2_reg, 0, length_offset), tmp2_reg);
+
+    // return false if the two arrays are not equal length
+    __ cmp(tmp1_reg, tmp2_reg);
+    __ br(Assembler::notEqual, true, Assembler::pn, Ldone);
+    __ delayed()->mov(G0, result_reg);     // not equal
+
+    __ br_zero(Assembler::zero, true, Assembler::pn, tmp1_reg, Ldone);
+    __ delayed()->add(G0, 1, result_reg); // zero-length arrays are equal
+
+    // load array addresses
+    __ add(ary1_reg, base_offset, ary1_reg);
+    __ add(ary2_reg, base_offset, ary2_reg);
+
+    // renaming registers
+    Register chr1_reg  =  tmp2_reg;   // for characters in ary1
+    Register chr2_reg  =  result_reg; // for characters in ary2
+    Register limit_reg =  tmp1_reg;   // length
+
+    // set byte count
+    __ sll(limit_reg, exact_log2(sizeof(jchar)), limit_reg);
+    __ andcc(limit_reg, 0x2, chr1_reg); //trailing character ?
+    __ br(Assembler::zero, false, Assembler::pt, Lvector);
+    __ delayed()->nop();
+
+    //compare the trailing char
+    __ sub(limit_reg, sizeof(jchar), limit_reg);
+    __ lduh(ary1_reg, limit_reg, chr1_reg);
+    __ lduh(ary2_reg, limit_reg, chr2_reg);
+    __ cmp(chr1_reg, chr2_reg);
+    __ br(Assembler::notEqual, true, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg);     // not equal
+
+    // only one char ?
+    __ br_zero(Assembler::zero, true, Assembler::pn, limit_reg, Ldone);
+    __ delayed()->add(G0, 1, result_reg); // zero-length arrays are equal
+
+    __ bind(Lvector);
+    // Shift ary1_reg and ary2_reg to the end of the arrays, negate limit
+    __ add(ary1_reg, limit_reg, ary1_reg);
+    __ add(ary2_reg, limit_reg, ary2_reg);
+    __ neg(limit_reg, limit_reg);
+
+    __ lduw(ary1_reg, limit_reg, chr1_reg);
+    __ bind(Lloop);
+    __ lduw(ary2_reg, limit_reg, chr2_reg);
+    __ cmp(chr1_reg, chr2_reg);
+    __ br(Assembler::notEqual, false, Assembler::pt, Ldone);
+    __ delayed()->mov(G0, result_reg);     // not equal
+    __ inccc(limit_reg, 2*sizeof(jchar));
+    // annul LDUW if branch is not taken to prevent access past end of string
+    __ br(Assembler::notZero, true, Assembler::pt, Lloop); //annul on taken
+    __ delayed()->lduw(ary1_reg, limit_reg, chr1_reg); // hoisted
+
+    __ add(G0, 1, result_reg); // equals
+
+    __ bind(Ldone);
+  %}
+
   enc_class enc_rethrow() %{
     cbuf.set_inst_mark();
     Register temp_reg = G3;
@@ -9015,6 +9211,25 @@
   ins_pipe(long_memory_op);
 %}
 
+instruct string_equals(o0RegP str1, o1RegP str2, g3RegP tmp1, g4RegP tmp2, notemp_iRegI result,
+                       o7RegI tmp3, flagsReg ccr) %{
+  match(Set result (StrEquals str1 str2));
+  effect(USE_KILL str1, USE_KILL str2, KILL tmp1, KILL tmp2, KILL ccr, KILL tmp3);
+  ins_cost(300);
+  format %{ "String Equals $str1,$str2 -> $result" %}
+  ins_encode( enc_String_Equals(str1, str2, tmp1, tmp2, result) );
+  ins_pipe(long_memory_op);
+%}
+
+instruct array_equals(o0RegP ary1, o1RegP ary2, g3RegP tmp1, g4RegP tmp2, notemp_iRegI result,
+                        flagsReg ccr) %{
+  match(Set result (AryEq ary1 ary2));
+  effect(USE_KILL ary1, USE_KILL ary2, KILL tmp1, KILL tmp2, KILL ccr);
+  ins_cost(300);
+  format %{ "Array Equals $ary1,$ary2 -> $result" %}
+  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, tmp2, result));
+  ins_pipe(long_memory_op);
+%}
 
 //---------- Population Count Instructions -------------------------------------
 
--- a/src/cpu/sparc/vm/stubGenerator_sparc.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/stubGenerator_sparc.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -817,21 +817,6 @@
   Label _atomic_add_stub;  // called from other stubs
 
 
-  // Support for void OrderAccess::fence().
-  //
-  address generate_fence() {
-    StubCodeMark mark(this, "StubRoutines", "fence");
-    address start = __ pc();
-
-    __ membar(Assembler::Membar_mask_bits(Assembler::LoadLoad  | Assembler::LoadStore |
-                                          Assembler::StoreLoad | Assembler::StoreStore));
-    __ retl(false);
-    __ delayed()->nop();
-
-    return start;
-  }
-
-
   //------------------------------------------------------------------------------------------------------------------------
   // The following routine generates a subroutine to throw an asynchronous
   // UnknownError when an unsafe access gets a fault that could not be
@@ -2861,7 +2846,6 @@
     StubRoutines::_atomic_cmpxchg_ptr_entry  = StubRoutines::_atomic_cmpxchg_entry;
     StubRoutines::_atomic_cmpxchg_long_entry = generate_atomic_cmpxchg_long();
     StubRoutines::_atomic_add_ptr_entry      = StubRoutines::_atomic_add_entry;
-    StubRoutines::_fence_entry               = generate_fence();
 #endif  // COMPILER2 !=> _LP64
   }
 
--- a/src/cpu/sparc/vm/vm_version_sparc.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/sparc/vm/vm_version_sparc.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -62,7 +62,7 @@
   if (is_niagara1()) {
     // Indirect branch is the same cost as direct
     if (FLAG_IS_DEFAULT(UseInlineCaches)) {
-      UseInlineCaches         = false;
+      FLAG_SET_DEFAULT(UseInlineCaches, false);
     }
 #ifdef _LP64
     // Single issue niagara1 is slower for CompressedOops
@@ -79,15 +79,19 @@
 #ifdef COMPILER2
     // Indirect branch is the same cost as direct
     if (FLAG_IS_DEFAULT(UseJumpTables)) {
-      UseJumpTables           = true;
+      FLAG_SET_DEFAULT(UseJumpTables, true);
     }
     // Single-issue, so entry and loop tops are
     // aligned on a single instruction boundary
     if (FLAG_IS_DEFAULT(InteriorEntryAlignment)) {
-      InteriorEntryAlignment  = 4;
+      FLAG_SET_DEFAULT(InteriorEntryAlignment, 4);
     }
     if (FLAG_IS_DEFAULT(OptoLoopAlignment)) {
-      OptoLoopAlignment       = 4;
+      FLAG_SET_DEFAULT(OptoLoopAlignment, 4);
+    }
+    if (is_niagara1_plus() && FLAG_IS_DEFAULT(AllocatePrefetchDistance)) {
+      // Use smaller prefetch distance on N2
+      FLAG_SET_DEFAULT(AllocatePrefetchDistance, 256);
     }
 #endif
   }
@@ -95,7 +99,7 @@
   // Use hardware population count instruction if available.
   if (has_hardware_popc()) {
     if (FLAG_IS_DEFAULT(UsePopCountInstruction)) {
-      UsePopCountInstruction = true;
+      FLAG_SET_DEFAULT(UsePopCountInstruction, true);
     }
   }
 
--- a/src/cpu/x86/vm/assembler_x86.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/assembler_x86.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1438,26 +1438,12 @@
   }
 }
 
-// Serializes memory.
+// Emit mfence instruction
 void Assembler::mfence() {
-    // Memory barriers are only needed on multiprocessors
-  if (os::is_MP()) {
-    if( LP64_ONLY(true ||) VM_Version::supports_sse2() ) {
-      emit_byte( 0x0F );                // MFENCE; faster blows no regs
-      emit_byte( 0xAE );
-      emit_byte( 0xF0 );
-    } else {
-      // All usable chips support "locked" instructions which suffice
-      // as barriers, and are much faster than the alternative of
-      // using cpuid instruction. We use here a locked add [esp],0.
-      // This is conveniently otherwise a no-op except for blowing
-      // flags (which we save and restore.)
-      pushf();                // Save eflags register
-      lock();
-      addl(Address(rsp, 0), 0);// Assert the lock# signal here
-      popf();                 // Restore eflags register
-    }
-  }
+  NOT_LP64(assert(VM_Version::supports_sse2(), "unsupported");)
+  emit_byte( 0x0F );
+  emit_byte( 0xAE );
+  emit_byte( 0xF0 );
 }
 
 void Assembler::mov(Register dst, Register src) {
@@ -2187,6 +2173,31 @@
   emit_arith(0x0B, 0xC0, dst, src);
 }
 
+void Assembler::pcmpestri(XMMRegister dst, Address src, int imm8) {
+  assert(VM_Version::supports_sse4_2(), "");
+
+  InstructionMark im(this);
+  emit_byte(0x66);
+  prefix(src, dst);
+  emit_byte(0x0F);
+  emit_byte(0x3A);
+  emit_byte(0x61);
+  emit_operand(dst, src);
+  emit_byte(imm8);
+}
+
+void Assembler::pcmpestri(XMMRegister dst, XMMRegister src, int imm8) {
+  assert(VM_Version::supports_sse4_2(), "");
+
+  emit_byte(0x66);
+  int encode = prefixq_and_encode(dst->encoding(), src->encoding());
+  emit_byte(0x0F);
+  emit_byte(0x3A);
+  emit_byte(0x61);
+  emit_byte(0xC0 | encode);
+  emit_byte(imm8);
+}
+
 // generic
 void Assembler::pop(Register dst) {
   int encode = prefix_and_encode(dst->encoding());
@@ -2344,6 +2355,29 @@
   emit_byte(shift);
 }
 
+void Assembler::ptest(XMMRegister dst, Address src) {
+  assert(VM_Version::supports_sse4_1(), "");
+
+  InstructionMark im(this);
+  emit_byte(0x66);
+  prefix(src, dst);
+  emit_byte(0x0F);
+  emit_byte(0x38);
+  emit_byte(0x17);
+  emit_operand(dst, src);
+}
+
+void Assembler::ptest(XMMRegister dst, XMMRegister src) {
+  assert(VM_Version::supports_sse4_1(), "");
+
+  emit_byte(0x66);
+  int encode = prefixq_and_encode(dst->encoding(), src->encoding());
+  emit_byte(0x0F);
+  emit_byte(0x38);
+  emit_byte(0x17);
+  emit_byte(0xC0 | encode);
+}
+
 void Assembler::punpcklbw(XMMRegister dst, XMMRegister src) {
   NOT_LP64(assert(VM_Version::supports_sse2(), ""));
   emit_byte(0x66);
@@ -7218,7 +7252,7 @@
 // On failure, execution transfers to the given label.
 void MacroAssembler::lookup_interface_method(Register recv_klass,
                                              Register intf_klass,
-                                             RegisterConstant itable_index,
+                                             RegisterOrConstant itable_index,
                                              Register method_result,
                                              Register scan_temp,
                                              Label& L_no_such_interface) {
@@ -7303,7 +7337,7 @@
                                                    Label* L_success,
                                                    Label* L_failure,
                                                    Label* L_slow_path,
-                                        RegisterConstant super_check_offset) {
+                                        RegisterOrConstant super_check_offset) {
   assert_different_registers(sub_klass, super_klass, temp_reg);
   bool must_load_sco = (super_check_offset.constant_or_zero() == -1);
   if (super_check_offset.is_register()) {
@@ -7352,7 +7386,7 @@
   if (must_load_sco) {
     // Positive movl does right thing on LP64.
     movl(temp_reg, super_check_offset_addr);
-    super_check_offset = RegisterConstant(temp_reg);
+    super_check_offset = RegisterOrConstant(temp_reg);
   }
   Address super_check_addr(sub_klass, super_check_offset, Address::times_1, 0);
   cmpptr(super_klass, super_check_addr); // load displayed supertype
@@ -7550,12 +7584,12 @@
 }
 
 
-RegisterConstant MacroAssembler::delayed_value(intptr_t* delayed_value_addr,
-                                               Register tmp,
-                                               int offset) {
+RegisterOrConstant MacroAssembler::delayed_value_impl(intptr_t* delayed_value_addr,
+                                                      Register tmp,
+                                                      int offset) {
   intptr_t value = *delayed_value_addr;
   if (value != 0)
-    return RegisterConstant(value + offset);
+    return RegisterOrConstant(value + offset);
 
   // load indirectly to solve generation ordering problem
   movptr(tmp, ExternalAddress((address) delayed_value_addr));
@@ -7571,7 +7605,7 @@
   if (offset != 0)
     addptr(tmp, offset);
 
-  return RegisterConstant(tmp);
+  return RegisterOrConstant(tmp);
 }
 
 
--- a/src/cpu/x86/vm/assembler_x86.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/assembler_x86.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -212,7 +212,7 @@
            "inconsistent address");
   }
 
-  Address(Register base, RegisterConstant index, ScaleFactor scale = times_1, int disp = 0)
+  Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0)
     : _base (base),
       _index(index.register_or_noreg()),
       _scale(scale),
@@ -256,7 +256,7 @@
            "inconsistent address");
   }
 
-  Address(Register base, RegisterConstant index, ScaleFactor scale, ByteSize disp)
+  Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp)
     : _base (base),
       _index(index.register_or_noreg()),
       _scale(scale),
@@ -1068,15 +1068,23 @@
     LoadLoad   = 1 << 0
   };
 
-  // Serializes memory.
+  // Serializes memory and blows flags
   void membar(Membar_mask_bits order_constraint) {
-    // We only have to handle StoreLoad and LoadLoad
-    if (order_constraint & StoreLoad) {
-      // MFENCE subsumes LFENCE
-      mfence();
-    } /* [jk] not needed currently: else if (order_constraint & LoadLoad) {
-         lfence();
-    } */
+    if (os::is_MP()) {
+      // We only have to handle StoreLoad
+      if (order_constraint & StoreLoad) {
+        // All usable chips support "locked" instructions which suffice
+        // as barriers, and are much faster than the alternative of
+        // using cpuid instruction. We use here a locked add [esp],0.
+        // This is conveniently otherwise a no-op except for blowing
+        // flags.
+        // Any change to this code may need to revisit other places in
+        // the code where this idiom is used, in particular the
+        // orderAccess code.
+        lock();
+        addl(Address(rsp, 0), 0);// Assert the lock# signal here
+      }
+    }
   }
 
   void mfence();
@@ -1218,6 +1226,10 @@
   void orq(Register dst, Address src);
   void orq(Register dst, Register src);
 
+  // SSE4.2 string instructions
+  void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
+  void pcmpestri(XMMRegister xmm1, Address src, int imm8);
+
   void popl(Address dst);
 
 #ifdef _LP64
@@ -1252,6 +1264,10 @@
   // Shift Right Logical Quadword Immediate
   void psrlq(XMMRegister dst, int shift);
 
+  // Logical Compare Double Quadword
+  void ptest(XMMRegister dst, XMMRegister src);
+  void ptest(XMMRegister dst, Address src);
+
   // Interleave Low Bytes
   void punpcklbw(XMMRegister dst, XMMRegister src);
 
@@ -1802,7 +1818,7 @@
   // interface method calling
   void lookup_interface_method(Register recv_klass,
                                Register intf_klass,
-                               RegisterConstant itable_index,
+                               RegisterOrConstant itable_index,
                                Register method_result,
                                Register scan_temp,
                                Label& no_such_interface);
@@ -1819,7 +1835,7 @@
                                      Label* L_success,
                                      Label* L_failure,
                                      Label* L_slow_path,
-                RegisterConstant super_check_offset = RegisterConstant(-1));
+                RegisterOrConstant super_check_offset = RegisterOrConstant(-1));
 
   // The rest of the type check; must be wired to a corresponding fast path.
   // It does not repeat the fast path logic, so don't use it standalone.
@@ -1883,9 +1899,9 @@
   // stack overflow + shadow pages.  Also, clobbers tmp
   void bang_stack_size(Register size, Register tmp);
 
-  virtual RegisterConstant delayed_value(intptr_t* delayed_value_addr,
-                                         Register tmp,
-                                         int offset);
+  virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr,
+                                                Register tmp,
+                                                int offset);
 
   // Support for serializing memory accesses between threads
   void serialize_memory(Register thread, Register tmp);
--- a/src/cpu/x86/vm/globals_x86.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/globals_x86.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -60,6 +60,7 @@
 define_pd_global(intx, StackShadowPages, 3 DEBUG_ONLY(+1));
 #endif // AMD64
 define_pd_global(intx,  InlineFrequencyCount,     100);
+define_pd_global(intx,  InlineSmallCode,          1000);
 define_pd_global(intx,  PreInflateSpin,           10);
 
 define_pd_global(intx, StackYellowPages, 2);
--- a/src/cpu/x86/vm/sharedRuntime_x86_64.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/sharedRuntime_x86_64.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -2691,7 +2691,7 @@
   __ mov(rdi, rax);
 
    Label noException;
-  __ cmpl(r12, Deoptimization::Unpack_exception);   // Was exception pending?
+  __ cmpl(r14, Deoptimization::Unpack_exception);   // Was exception pending?
   __ jcc(Assembler::notEqual, noException);
   __ movptr(rax, Address(r15_thread, JavaThread::exception_oop_offset()));
   // QQQ this is useless it was NULL above
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -637,7 +637,7 @@
   address generate_orderaccess_fence() {
     StubCodeMark mark(this, "StubRoutines", "orderaccess_fence");
     address start = __ pc();
-    __ mfence();
+    __ membar(Assembler::StoreLoad);
     __ ret(0);
 
     return start;
--- a/src/cpu/x86/vm/vm_version_x86.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/vm_version_x86.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -408,6 +408,11 @@
           UseUnalignedLoadStores = true; // use movdqu on newest Intel cpus
         }
       }
+      if( supports_sse4_2() && UseSSE >= 4 ) {
+        if( FLAG_IS_DEFAULT(UseSSE42Intrinsics)) {
+          UseSSE42Intrinsics = true;
+        }
+      }
     }
   }
 
--- a/src/cpu/x86/vm/x86_32.ad	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/x86_32.ad	Fri Apr 03 15:59:19 2009 -0700
@@ -3694,12 +3694,16 @@
     }
   %}
 
-  enc_class enc_String_Compare() %{
+  enc_class enc_String_Compare(eDIRegP str1, eSIRegP str2, regXD tmp1, regXD tmp2,
+                        eAXRegI tmp3, eBXRegI tmp4, eCXRegI result) %{
     Label ECX_GOOD_LABEL, LENGTH_DIFF_LABEL,
           POP_LABEL, DONE_LABEL, CONT_LABEL,
           WHILE_HEAD_LABEL;
     MacroAssembler masm(&cbuf);
 
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+
     // Get the first character position in both strings
     //         [8] char array, [12] offset, [16] count
     int value_offset  = java_lang_String::value_offset_in_bytes();
@@ -3717,7 +3721,6 @@
     // Compute the minimum of the string lengths(rsi) and the
     // difference of the string lengths (stack)
 
-
     if (VM_Version::supports_cmov()) {
       masm.movl(rdi, Address(rdi, count_offset));
       masm.movl(rsi, Address(rsi, count_offset));
@@ -3731,7 +3734,7 @@
       masm.movl(rsi, rdi);
       masm.subl(rdi, rcx);
       masm.push(rdi);
-      masm.jcc(Assembler::lessEqual, ECX_GOOD_LABEL);
+      masm.jccb(Assembler::lessEqual, ECX_GOOD_LABEL);
       masm.movl(rsi, rcx);
       // rsi holds min, rcx is unused
     }
@@ -3756,7 +3759,7 @@
       Label LSkip2;
       // Check if the strings start at same location
       masm.cmpptr(rbx,rax);
-      masm.jcc(Assembler::notEqual, LSkip2);
+      masm.jccb(Assembler::notEqual, LSkip2);
 
       // Check if the length difference is zero (from stack)
       masm.cmpl(Address(rsp, 0), 0x0);
@@ -3766,9 +3769,52 @@
       masm.bind(LSkip2);
     }
 
-    // Shift rax, and rbx, to the end of the arrays, negate min
-    masm.lea(rax, Address(rax, rsi, Address::times_2, 2));
-    masm.lea(rbx, Address(rbx, rsi, Address::times_2, 2));
+   // Advance to next character
+    masm.addptr(rax, 2);
+    masm.addptr(rbx, 2);
+
+    if (UseSSE42Intrinsics) {
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_VECTORS, VECTOR_NOT_EQUAL, COMPARE_TAIL;
+      // Setup to compare 16-byte vectors
+      masm.movl(rdi, rsi);
+      masm.andl(rsi, 0xfffffff8); // rsi holds the vector count
+      masm.andl(rdi, 0x00000007); // rdi holds the tail count
+      masm.testl(rsi, rsi);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+
+      masm.lea(rax, Address(rax, rsi, Address::times_2));
+      masm.lea(rbx, Address(rbx, rsi, Address::times_2));
+      masm.negl(rsi);
+
+      masm.bind(COMPARE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(rax, rsi, Address::times_2));
+      masm.movdqu(tmp2Reg, Address(rbx, rsi, Address::times_2));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+      masm.jccb(Assembler::notZero, VECTOR_NOT_EQUAL);
+      masm.addl(rsi, 8);
+      masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+      masm.jmpb(COMPARE_TAIL);
+
+      // Mismatched characters in the vectors
+      masm.bind(VECTOR_NOT_EQUAL);
+      masm.lea(rax, Address(rax, rsi, Address::times_2));
+      masm.lea(rbx, Address(rbx, rsi, Address::times_2));
+      masm.movl(rdi, 8);
+
+      // Compare tail (< 8 chars), or rescan last vectors to
+      // find 1st mismatched characters
+      masm.bind(COMPARE_TAIL);
+      masm.testl(rdi, rdi);
+      masm.jccb(Assembler::zero, LENGTH_DIFF_LABEL);
+      masm.movl(rsi, rdi);
+      // Fallthru to tail compare
+    }
+
+    //Shift rax, and rbx, to the end of the arrays, negate min
+    masm.lea(rax, Address(rax, rsi, Address::times_2, 0));
+    masm.lea(rbx, Address(rbx, rsi, Address::times_2, 0));
     masm.negl(rsi);
 
     // Compare the rest of the characters
@@ -3776,93 +3822,329 @@
     masm.load_unsigned_short(rcx, Address(rbx, rsi, Address::times_2, 0));
     masm.load_unsigned_short(rdi, Address(rax, rsi, Address::times_2, 0));
     masm.subl(rcx, rdi);
-    masm.jcc(Assembler::notZero, POP_LABEL);
+    masm.jccb(Assembler::notZero, POP_LABEL);
     masm.incrementl(rsi);
     masm.jcc(Assembler::notZero, WHILE_HEAD_LABEL);
 
     // Strings are equal up to min length.  Return the length difference.
     masm.bind(LENGTH_DIFF_LABEL);
     masm.pop(rcx);
-    masm.jmp(DONE_LABEL);
+    masm.jmpb(DONE_LABEL);
 
     // Discard the stored length difference
     masm.bind(POP_LABEL);
     masm.addptr(rsp, 4);
-       
+
     // That's it
     masm.bind(DONE_LABEL);
   %}
 
-  enc_class enc_Array_Equals(eDIRegP ary1, eSIRegP ary2, eAXRegI tmp1, eBXRegI tmp2, eCXRegI result) %{
-    Label TRUE_LABEL, FALSE_LABEL, DONE_LABEL, COMPARE_LOOP_HDR, COMPARE_LOOP;
+ enc_class enc_String_Equals(eDIRegP str1, eSIRegP str2, regXD tmp1, regXD tmp2,
+                       eBXRegI tmp3, eCXRegI tmp4, eAXRegI result) %{
+    Label RET_TRUE, RET_FALSE, DONE, COMPARE_VECTORS, COMPARE_CHAR;
+    MacroAssembler masm(&cbuf);
+
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+
+    int value_offset  = java_lang_String::value_offset_in_bytes();
+    int offset_offset = java_lang_String::offset_offset_in_bytes();
+    int count_offset  = java_lang_String::count_offset_in_bytes();
+    int base_offset   = arrayOopDesc::base_offset_in_bytes(T_CHAR);
+
+    // does source == target string?
+    masm.cmpptr(rdi, rsi);
+    masm.jcc(Assembler::equal, RET_TRUE);
+
+    // get and compare counts
+    masm.movl(rcx, Address(rdi, count_offset));
+    masm.movl(rax, Address(rsi, count_offset));
+    masm.cmpl(rcx, rax);
+    masm.jcc(Assembler::notEqual, RET_FALSE);
+    masm.testl(rax, rax);
+    masm.jcc(Assembler::zero, RET_TRUE);
+
+    // get source string offset and value
+    masm.movptr(rbx, Address(rsi, value_offset));
+    masm.movl(rax, Address(rsi, offset_offset));
+    masm.leal(rsi, Address(rbx, rax, Address::times_2, base_offset));
+
+    // get compare string offset and value
+    masm.movptr(rbx, Address(rdi, value_offset));
+    masm.movl(rax, Address(rdi, offset_offset));
+    masm.leal(rdi, Address(rbx, rax, Address::times_2, base_offset));
+
+    // Set byte count
+    masm.shll(rcx, 1);
+    masm.movl(rax, rcx);
+
+    if (UseSSE42Intrinsics) {
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+      // Compare 16-byte vectors
+      masm.andl(rcx, 0xfffffff0);  // vector count (in bytes)
+      masm.andl(rax, 0x0000000e);  // tail count (in bytes)
+      masm.testl(rcx, rcx);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+      masm.lea(rdi, Address(rdi, rcx, Address::times_1));
+      masm.lea(rsi, Address(rsi, rcx, Address::times_1));
+      masm.negl(rcx);
+
+      masm.bind(COMPARE_WIDE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(rdi, rcx, Address::times_1));
+      masm.movdqu(tmp2Reg, Address(rsi, rcx, Address::times_1));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+      masm.jccb(Assembler::notZero, RET_FALSE);
+      masm.addl(rcx, 16);
+      masm.jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+      masm.bind(COMPARE_TAIL);
+      masm.movl(rcx, rax);
+      // Fallthru to tail compare
+    }
+
+    // Compare 4-byte vectors
+    masm.andl(rcx, 0xfffffffc);  // vector count (in bytes)
+    masm.andl(rax, 0x00000002);  // tail char (in bytes)
+    masm.testl(rcx, rcx);
+    masm.jccb(Assembler::zero, COMPARE_CHAR);
+    masm.lea(rdi, Address(rdi, rcx, Address::times_1));
+    masm.lea(rsi, Address(rsi, rcx, Address::times_1));
+    masm.negl(rcx);
+
+    masm.bind(COMPARE_VECTORS);
+    masm.movl(rbx, Address(rdi, rcx, Address::times_1));
+    masm.cmpl(rbx, Address(rsi, rcx, Address::times_1));
+    masm.jccb(Assembler::notEqual, RET_FALSE);
+    masm.addl(rcx, 4);
+    masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+
+    // Compare trailing char (final 2 bytes), if any
+    masm.bind(COMPARE_CHAR);
+    masm.testl(rax, rax);
+    masm.jccb(Assembler::zero, RET_TRUE);
+    masm.load_unsigned_short(rbx, Address(rdi, 0));
+    masm.load_unsigned_short(rcx, Address(rsi, 0));
+    masm.cmpl(rbx, rcx);
+    masm.jccb(Assembler::notEqual, RET_FALSE);
+
+    masm.bind(RET_TRUE);
+    masm.movl(rax, 1);   // return true
+    masm.jmpb(DONE);
+
+    masm.bind(RET_FALSE);
+    masm.xorl(rax, rax); // return false
+
+    masm.bind(DONE);
+    %}
+
+ enc_class enc_String_IndexOf(eSIRegP str1, eDIRegP str2, regXD tmp1, eAXRegI tmp2,
+                        eCXRegI tmp3, eDXRegI tmp4, eBXRegI result) %{
+    // SSE4.2 version
+    Label LOAD_SUBSTR, PREP_FOR_SCAN, SCAN_TO_SUBSTR,
+          SCAN_SUBSTR, RET_NEG_ONE, RET_NOT_FOUND, CLEANUP, DONE;
     MacroAssembler masm(&cbuf);
 
-    Register ary1Reg   = as_Register($ary1$$reg);
-    Register ary2Reg   = as_Register($ary2$$reg);
-    Register tmp1Reg   = as_Register($tmp1$$reg);
-    Register tmp2Reg   = as_Register($tmp2$$reg);
-    Register resultReg = as_Register($result$$reg);
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+
+    // Get the first character position in both strings
+    //         [8] char array, [12] offset, [16] count
+    int value_offset  = java_lang_String::value_offset_in_bytes();
+    int offset_offset = java_lang_String::offset_offset_in_bytes();
+    int count_offset  = java_lang_String::count_offset_in_bytes();
+    int base_offset   = arrayOopDesc::base_offset_in_bytes(T_CHAR);
+
+    // Get counts for string and substr
+    masm.movl(rdx, Address(rsi, count_offset));
+    masm.movl(rax, Address(rdi, count_offset));
+    // Check for substr count > string count
+    masm.cmpl(rax, rdx);
+    masm.jcc(Assembler::greater, RET_NEG_ONE);
+
+    // Start the indexOf operation
+    // Get start addr of string
+    masm.movptr(rbx, Address(rsi, value_offset));
+    masm.movl(rcx, Address(rsi, offset_offset));
+    masm.lea(rsi, Address(rbx, rcx, Address::times_2, base_offset));
+    masm.push(rsi);
+
+    // Get start addr of substr
+    masm.movptr(rbx, Address(rdi, value_offset));
+    masm.movl(rcx, Address(rdi, offset_offset));
+    masm.lea(rdi, Address(rbx, rcx, Address::times_2, base_offset));
+    masm.push(rdi);
+    masm.push(rax);
+    masm.jmpb(PREP_FOR_SCAN);
+
+    // Substr count saved at sp
+    // Substr saved at sp+4
+    // String saved at sp+8
+
+    // Prep to load substr for scan
+    masm.bind(LOAD_SUBSTR);
+    masm.movptr(rdi, Address(rsp, 4));
+    masm.movl(rax, Address(rsp, 0));
+
+    // Load substr
+    masm.bind(PREP_FOR_SCAN);
+    masm.movdqu(tmp1Reg, Address(rdi, 0));
+    masm.addl(rdx, 8);        // prime the loop
+    masm.subptr(rsi, 16);
+
+    // Scan string for substr in 16-byte vectors
+    masm.bind(SCAN_TO_SUBSTR);
+    masm.subl(rdx, 8);
+    masm.addptr(rsi, 16);
+    masm.pcmpestri(tmp1Reg, Address(rsi, 0), 0x0d);
+    masm.jcc(Assembler::above, SCAN_TO_SUBSTR);     // CF == 0 && ZF == 0
+    masm.jccb(Assembler::aboveEqual, RET_NOT_FOUND); // CF == 0
+
+    // Fallthru: found a potential substr
+
+    // Make sure string is still long enough
+    masm.subl(rdx, rcx);
+    masm.cmpl(rdx, rax);
+    masm.jccb(Assembler::negative, RET_NOT_FOUND);
+    // Compute start addr of substr
+    masm.lea(rsi, Address(rsi, rcx, Address::times_2));
+    masm.movptr(rbx, rsi);
+
+    // Compare potential substr
+    masm.addl(rdx, 8);        // prime the loop
+    masm.addl(rax, 8);
+    masm.subptr(rsi, 16);
+    masm.subptr(rdi, 16);
+
+    // Scan 16-byte vectors of string and substr
+    masm.bind(SCAN_SUBSTR);
+    masm.subl(rax, 8);
+    masm.subl(rdx, 8);
+    masm.addptr(rsi, 16);
+    masm.addptr(rdi, 16);
+    masm.movdqu(tmp1Reg, Address(rdi, 0));
+    masm.pcmpestri(tmp1Reg, Address(rsi, 0), 0x0d);
+    masm.jcc(Assembler::noOverflow, LOAD_SUBSTR);   // OF == 0
+    masm.jcc(Assembler::positive, SCAN_SUBSTR);     // SF == 0
+
+    // Compute substr offset
+    masm.movptr(rsi, Address(rsp, 8));
+    masm.subptr(rbx, rsi);
+    masm.shrl(rbx, 1);
+    masm.jmpb(CLEANUP);
+
+    masm.bind(RET_NEG_ONE);
+    masm.movl(rbx, -1);
+    masm.jmpb(DONE);
+
+    masm.bind(RET_NOT_FOUND);
+    masm.movl(rbx, -1);
+
+    masm.bind(CLEANUP);
+    masm.addptr(rsp, 12);
+
+    masm.bind(DONE);
+  %}
+
+  enc_class enc_Array_Equals(eDIRegP ary1, eSIRegP ary2, regXD tmp1, regXD tmp2,
+                             eBXRegI tmp3, eDXRegI tmp4, eAXRegI result) %{
+    Label TRUE_LABEL, FALSE_LABEL, DONE, COMPARE_VECTORS, COMPARE_CHAR;
+    MacroAssembler masm(&cbuf);
+
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+    Register ary1Reg      = as_Register($ary1$$reg);
+    Register ary2Reg      = as_Register($ary2$$reg);
+    Register tmp3Reg      = as_Register($tmp3$$reg);
+    Register tmp4Reg      = as_Register($tmp4$$reg);
+    Register resultReg    = as_Register($result$$reg);
 
     int length_offset  = arrayOopDesc::length_offset_in_bytes();
     int base_offset    = arrayOopDesc::base_offset_in_bytes(T_CHAR);
 
     // Check the input args
-    masm.cmpl(ary1Reg, ary2Reg);
+    masm.cmpptr(ary1Reg, ary2Reg);
     masm.jcc(Assembler::equal, TRUE_LABEL);
-    masm.testl(ary1Reg, ary1Reg);
+    masm.testptr(ary1Reg, ary1Reg);
     masm.jcc(Assembler::zero, FALSE_LABEL);
-    masm.testl(ary2Reg, ary2Reg);
+    masm.testptr(ary2Reg, ary2Reg);
     masm.jcc(Assembler::zero, FALSE_LABEL);
 
     // Check the lengths
-    masm.movl(tmp2Reg, Address(ary1Reg, length_offset));
+    masm.movl(tmp4Reg, Address(ary1Reg, length_offset));
     masm.movl(resultReg, Address(ary2Reg, length_offset));
-    masm.cmpl(tmp2Reg, resultReg);
+    masm.cmpl(tmp4Reg, resultReg);
     masm.jcc(Assembler::notEqual, FALSE_LABEL);
     masm.testl(resultReg, resultReg);
     masm.jcc(Assembler::zero, TRUE_LABEL);
 
-    // Get the number of 4 byte vectors to compare
-    masm.shrl(resultReg, 1);
-
-    // Check for odd-length arrays
-    masm.andl(tmp2Reg, 1);
-    masm.testl(tmp2Reg, tmp2Reg);
-    masm.jcc(Assembler::zero, COMPARE_LOOP_HDR);
-
-    // Compare 2-byte "tail" at end of arrays
-    masm.load_unsigned_short(tmp1Reg, Address(ary1Reg, resultReg, Address::times_4, base_offset));
-    masm.load_unsigned_short(tmp2Reg, Address(ary2Reg, resultReg, Address::times_4, base_offset));
-    masm.cmpl(tmp1Reg, tmp2Reg);
-    masm.jcc(Assembler::notEqual, FALSE_LABEL);
+    // Load array addrs
+    masm.lea(ary1Reg, Address(ary1Reg, base_offset));
+    masm.lea(ary2Reg, Address(ary2Reg, base_offset));
+
+    // Set byte count
+    masm.shll(tmp4Reg, 1);
+    masm.movl(resultReg, tmp4Reg);
+
+    if (UseSSE42Intrinsics) {
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+      // Compare 16-byte vectors
+      masm.andl(tmp4Reg, 0xfffffff0);    // vector count (in bytes)
+      masm.andl(resultReg, 0x0000000e);  // tail count (in bytes)
+      masm.testl(tmp4Reg, tmp4Reg);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+      masm.lea(ary1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+      masm.lea(ary2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+      masm.negl(tmp4Reg);
+
+      masm.bind(COMPARE_WIDE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+      masm.movdqu(tmp2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+
+      masm.jccb(Assembler::notZero, FALSE_LABEL);
+      masm.addl(tmp4Reg, 16);
+      masm.jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+      masm.bind(COMPARE_TAIL);
+      masm.movl(tmp4Reg, resultReg);
+      // Fallthru to tail compare
+    }
+
+    // Compare 4-byte vectors
+    masm.andl(tmp4Reg, 0xfffffffc);    // vector count (in bytes)
+    masm.andl(resultReg, 0x00000002);  // tail char (in bytes)
+    masm.testl(tmp4Reg, tmp4Reg);
+    masm.jccb(Assembler::zero, COMPARE_CHAR);
+    masm.lea(ary1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+    masm.lea(ary2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+    masm.negl(tmp4Reg);
+
+    masm.bind(COMPARE_VECTORS);
+    masm.movl(tmp3Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+    masm.cmpl(tmp3Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+    masm.jccb(Assembler::notEqual, FALSE_LABEL);
+    masm.addl(tmp4Reg, 4);
+    masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+
+    // Compare trailing char (final 2 bytes), if any
+    masm.bind(COMPARE_CHAR);
     masm.testl(resultReg, resultReg);
-    masm.jcc(Assembler::zero, TRUE_LABEL);
-
-    // Setup compare loop
-    masm.bind(COMPARE_LOOP_HDR);
-    // Shift tmp1Reg and tmp2Reg to the last 4-byte boundary of the arrays
-    masm.leal(tmp1Reg, Address(ary1Reg, resultReg, Address::times_4, base_offset));
-    masm.leal(tmp2Reg, Address(ary2Reg, resultReg, Address::times_4, base_offset));
-    masm.negl(resultReg);
-
-    // 4-byte-wide compare loop
-    masm.bind(COMPARE_LOOP);
-    masm.movl(ary1Reg, Address(tmp1Reg, resultReg, Address::times_4, 0));
-    masm.movl(ary2Reg, Address(tmp2Reg, resultReg, Address::times_4, 0));
-    masm.cmpl(ary1Reg, ary2Reg);
-    masm.jcc(Assembler::notEqual, FALSE_LABEL);
-    masm.increment(resultReg);
-    masm.jcc(Assembler::notZero, COMPARE_LOOP);
+    masm.jccb(Assembler::zero, TRUE_LABEL);
+    masm.load_unsigned_short(tmp3Reg, Address(ary1Reg, 0));
+    masm.load_unsigned_short(tmp4Reg, Address(ary2Reg, 0));
+    masm.cmpl(tmp3Reg, tmp4Reg);
+    masm.jccb(Assembler::notEqual, FALSE_LABEL);
 
     masm.bind(TRUE_LABEL);
     masm.movl(resultReg, 1);   // return true
-    masm.jmp(DONE_LABEL);
+    masm.jmpb(DONE);
 
     masm.bind(FALSE_LABEL);
     masm.xorl(resultReg, resultReg); // return false
 
     // That's it
-    masm.bind(DONE_LABEL);
+    masm.bind(DONE);
   %}
 
   enc_class enc_pop_rdx() %{
@@ -4288,24 +4570,6 @@
     emit_opcode(cbuf, 0xC8 + $src2$$reg);
   %}
 
-  enc_class enc_membar_acquire %{
-    // Doug Lea believes this is not needed with current Sparcs and TSO.
-    // MacroAssembler masm(&cbuf);
-    // masm.membar();
-  %}
-
-  enc_class enc_membar_release %{
-    // Doug Lea believes this is not needed with current Sparcs and TSO.
-    // MacroAssembler masm(&cbuf);
-    // masm.membar();
-  %}
-
-  enc_class enc_membar_volatile %{
-    MacroAssembler masm(&cbuf);
-    masm.membar(Assembler::Membar_mask_bits(Assembler::StoreLoad |
-                                            Assembler::StoreStore));
-  %}
-
   // Atomically load the volatile long
   enc_class enc_loadL_volatile( memory mem, stackSlotL dst ) %{
     emit_opcode(cbuf,0xDF);
@@ -7498,9 +7762,9 @@
   ins_cost(400);
 
   size(0);
-  format %{ "MEMBAR-acquire" %}
-  ins_encode( enc_membar_acquire );
-  ins_pipe(pipe_slow);
+  format %{ "MEMBAR-acquire ! (empty encoding)" %}
+  ins_encode();
+  ins_pipe(empty);
 %}
 
 instruct membar_acquire_lock() %{
@@ -7519,9 +7783,9 @@
   ins_cost(400);
 
   size(0);
-  format %{ "MEMBAR-release" %}
-  ins_encode( enc_membar_release );
-  ins_pipe(pipe_slow);
+  format %{ "MEMBAR-release ! (empty encoding)" %}
+  ins_encode( );
+  ins_pipe(empty);
 %}
 
 instruct membar_release_lock() %{
@@ -7535,12 +7799,22 @@
   ins_pipe(empty);
 %}
 
-instruct membar_volatile() %{
+instruct membar_volatile(eFlagsReg cr) %{
   match(MemBarVolatile);
+  effect(KILL cr);
   ins_cost(400);
 
-  format %{ "MEMBAR-volatile" %}
-  ins_encode( enc_membar_volatile );
+  format %{ 
+    $$template
+    if (os::is_MP()) {
+      $$emit$$"LOCK ADDL [ESP + #0], 0\t! membar_volatile"
+    } else {
+      $$emit$$"MEMBAR-volatile ! (empty encoding)"
+    }
+  %}
+  ins_encode %{
+    __ membar(Assembler::StoreLoad);
+  %}
   ins_pipe(pipe_slow);
 %}
 
@@ -12082,11 +12356,8 @@
   ins_pipe( fpu_reg_reg );
 %}
 
-
-
 // =======================================================================
 // fast clearing of an array
-
 instruct rep_stos(eCXRegI cnt, eDIRegP base, eAXRegI zero, Universe dummy, eFlagsReg cr) %{
   match(Set dummy (ClearArray cnt base));
   effect(USE_KILL cnt, USE_KILL base, KILL zero, KILL cr);
@@ -12100,24 +12371,48 @@
   ins_pipe( pipe_slow );
 %}
 
-instruct string_compare(eDIRegP str1, eSIRegP str2, eAXRegI tmp1, eBXRegI tmp2, eCXRegI result, eFlagsReg cr) %{
+instruct string_compare(eDIRegP str1, eSIRegP str2, regXD tmp1, regXD tmp2,
+                        eAXRegI tmp3, eBXRegI tmp4, eCXRegI result, eFlagsReg cr) %{
   match(Set result (StrComp str1 str2));
-  effect(USE_KILL str1, USE_KILL str2, KILL tmp1, KILL tmp2, KILL cr);
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL str1, USE_KILL str2, KILL tmp3, KILL tmp4, KILL cr);
   //ins_cost(300);
 
   format %{ "String Compare $str1,$str2 -> $result    // KILL EAX, EBX" %}
-  ins_encode( enc_String_Compare() );
+  ins_encode( enc_String_Compare(str1, str2, tmp1, tmp2, tmp3, tmp4, result) );
+  ins_pipe( pipe_slow );
+%}
+
+// fast string equals
+instruct string_equals(eDIRegP str1, eSIRegP str2, regXD tmp1, regXD tmp2,
+                       eBXRegI tmp3, eCXRegI tmp4, eAXRegI result, eFlagsReg cr) %{
+  match(Set result (StrEquals str1 str2));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL str1, USE_KILL str2, KILL tmp3, KILL tmp4, KILL cr);
+
+  format %{ "String Equals $str1,$str2 -> $result    // KILL EBX, ECX" %}
+  ins_encode( enc_String_Equals(tmp1, tmp2, str1, str2, tmp3, tmp4, result) );
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexof(eSIRegP str1, eDIRegP str2, regXD tmp1, eAXRegI tmp2,
+                        eCXRegI tmp3, eDXRegI tmp4, eBXRegI result, eFlagsReg cr) %{
+  predicate(UseSSE42Intrinsics);
+  match(Set result (StrIndexOf str1 str2));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, KILL tmp2, KILL tmp3, KILL tmp4, KILL cr);
+
+  format %{ "String IndexOf $str1,$str2 -> $result    // KILL EAX, ECX, EDX" %}
+  ins_encode( enc_String_IndexOf(str1, str2, tmp1, tmp2, tmp3, tmp4, result) );
   ins_pipe( pipe_slow );
 %}
 
 // fast array equals
-instruct array_equals(eDIRegP ary1, eSIRegP ary2, eAXRegI tmp1, eBXRegI tmp2, eCXRegI result, eFlagsReg cr) %{
+instruct array_equals(eDIRegP ary1, eSIRegP ary2, regXD tmp1, regXD tmp2, eBXRegI tmp3,
+                      eDXRegI tmp4, eAXRegI result, eFlagsReg cr) %{
   match(Set result (AryEq ary1 ary2));
-  effect(USE_KILL ary1, USE_KILL ary2, KILL tmp1, KILL tmp2, KILL cr);
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
   //ins_cost(300);
 
-  format %{ "Array Equals $ary1,$ary2 -> $result    // KILL EAX, EBX" %}
-  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, tmp2, result) );
+  format %{ "Array Equals $ary1,$ary2 -> $result    // KILL EBX, EDX" %}
+  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, tmp2, tmp3, tmp4, result) );
   ins_pipe( pipe_slow );
 %}
 
--- a/src/cpu/x86/vm/x86_64.ad	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/cpu/x86/vm/x86_64.ad	Fri Apr 03 15:59:19 2009 -0700
@@ -3694,13 +3694,16 @@
     }
   %}
 
-  enc_class enc_String_Compare()
-  %{
+  enc_class enc_String_Compare(rdi_RegP str1, rsi_RegP str2, regD tmp1, regD tmp2,
+                        rax_RegI tmp3, rbx_RegI tmp4, rcx_RegI result) %{
     Label RCX_GOOD_LABEL, LENGTH_DIFF_LABEL,
           POP_LABEL, DONE_LABEL, CONT_LABEL,
           WHILE_HEAD_LABEL;
     MacroAssembler masm(&cbuf);
 
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+
     // Get the first character position in both strings
     //         [8] char array, [12] offset, [16] count
     int value_offset  = java_lang_String::value_offset_in_bytes();
@@ -3718,6 +3721,7 @@
     // Compute the minimum of the string lengths(rsi) and the
     // difference of the string lengths (stack)
 
+    // do the conditional move stuff
     masm.movl(rdi, Address(rdi, count_offset));
     masm.movl(rsi, Address(rsi, count_offset));
     masm.movl(rcx, rdi);
@@ -3745,7 +3749,7 @@
       Label LSkip2;
       // Check if the strings start at same location
       masm.cmpptr(rbx, rax);
-      masm.jcc(Assembler::notEqual, LSkip2);
+      masm.jccb(Assembler::notEqual, LSkip2);
 
       // Check if the length difference is zero (from stack)
       masm.cmpl(Address(rsp, 0), 0x0);
@@ -3755,9 +3759,52 @@
       masm.bind(LSkip2);
     }
 
+    // Advance to next character
+    masm.addptr(rax, 2);
+    masm.addptr(rbx, 2);
+
+    if (UseSSE42Intrinsics) {
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_VECTORS, VECTOR_NOT_EQUAL, COMPARE_TAIL;
+      // Setup to compare 16-byte vectors
+      masm.movl(rdi, rsi);
+      masm.andl(rsi, 0xfffffff8); // rsi holds the vector count
+      masm.andl(rdi, 0x00000007); // rdi holds the tail count
+      masm.testl(rsi, rsi);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+
+      masm.lea(rax, Address(rax, rsi, Address::times_2));
+      masm.lea(rbx, Address(rbx, rsi, Address::times_2));
+      masm.negptr(rsi);
+
+      masm.bind(COMPARE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(rax, rsi, Address::times_2));
+      masm.movdqu(tmp2Reg, Address(rbx, rsi, Address::times_2));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+      masm.jccb(Assembler::notZero, VECTOR_NOT_EQUAL);
+      masm.addptr(rsi, 8);
+      masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+      masm.jmpb(COMPARE_TAIL);
+
+      // Mismatched characters in the vectors
+      masm.bind(VECTOR_NOT_EQUAL);
+      masm.lea(rax, Address(rax, rsi, Address::times_2));
+      masm.lea(rbx, Address(rbx, rsi, Address::times_2));
+      masm.movl(rdi, 8);
+
+      // Compare tail (< 8 chars), or rescan last vectors to
+      // find 1st mismatched characters
+      masm.bind(COMPARE_TAIL);
+      masm.testl(rdi, rdi);
+      masm.jccb(Assembler::zero, LENGTH_DIFF_LABEL);
+      masm.movl(rsi, rdi);
+      // Fallthru to tail compare
+    }
+
     // Shift RAX and RBX to the end of the arrays, negate min
-    masm.lea(rax, Address(rax, rsi, Address::times_2, 2));
-    masm.lea(rbx, Address(rbx, rsi, Address::times_2, 2));
+    masm.lea(rax, Address(rax, rsi, Address::times_2, 0));
+    masm.lea(rbx, Address(rbx, rsi, Address::times_2, 0));
     masm.negptr(rsi);
 
     // Compare the rest of the characters
@@ -3765,93 +3812,329 @@
     masm.load_unsigned_short(rcx, Address(rbx, rsi, Address::times_2, 0));
     masm.load_unsigned_short(rdi, Address(rax, rsi, Address::times_2, 0));
     masm.subl(rcx, rdi);
-    masm.jcc(Assembler::notZero, POP_LABEL);
+    masm.jccb(Assembler::notZero, POP_LABEL);
     masm.increment(rsi);
     masm.jcc(Assembler::notZero, WHILE_HEAD_LABEL);
 
     // Strings are equal up to min length.  Return the length difference.
     masm.bind(LENGTH_DIFF_LABEL);
     masm.pop(rcx);
-    masm.jmp(DONE_LABEL);
+    masm.jmpb(DONE_LABEL);
 
     // Discard the stored length difference
     masm.bind(POP_LABEL);
     masm.addptr(rsp, 8);
-       
+
     // That's it
     masm.bind(DONE_LABEL);
   %}
 
-  enc_class enc_Array_Equals(rdi_RegP ary1, rsi_RegP ary2, rax_RegI tmp1, rbx_RegI tmp2, rcx_RegI result) %{
-    Label TRUE_LABEL, FALSE_LABEL, DONE_LABEL, COMPARE_LOOP_HDR, COMPARE_LOOP;
+ enc_class enc_String_IndexOf(rsi_RegP str1, rdi_RegP str2, regD tmp1, rax_RegI tmp2,
+                        rcx_RegI tmp3, rdx_RegI tmp4, rbx_RegI result) %{
+    // SSE4.2 version
+    Label LOAD_SUBSTR, PREP_FOR_SCAN, SCAN_TO_SUBSTR,
+          SCAN_SUBSTR, RET_NEG_ONE, RET_NOT_FOUND, CLEANUP, DONE;
     MacroAssembler masm(&cbuf);
 
-    Register ary1Reg   = as_Register($ary1$$reg);
-    Register ary2Reg   = as_Register($ary2$$reg);
-    Register tmp1Reg   = as_Register($tmp1$$reg);
-    Register tmp2Reg   = as_Register($tmp2$$reg);
-    Register resultReg = as_Register($result$$reg);
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+
+    // Get the first character position in both strings
+    //         [8] char array, [12] offset, [16] count
+    int value_offset  = java_lang_String::value_offset_in_bytes();
+    int offset_offset = java_lang_String::offset_offset_in_bytes();
+    int count_offset  = java_lang_String::count_offset_in_bytes();
+    int base_offset   = arrayOopDesc::base_offset_in_bytes(T_CHAR);
+
+    // Get counts for string and substr
+    masm.movl(rdx, Address(rsi, count_offset));
+    masm.movl(rax, Address(rdi, count_offset));
+    // Check for substr count > string count
+    masm.cmpl(rax, rdx);
+    masm.jcc(Assembler::greater, RET_NEG_ONE);
+
+    // Start the indexOf operation
+    // Get start addr of string
+    masm.load_heap_oop(rbx, Address(rsi, value_offset));
+    masm.movl(rcx, Address(rsi, offset_offset));
+    masm.lea(rsi, Address(rbx, rcx, Address::times_2, base_offset));
+    masm.push(rsi);
+
+    // Get start addr of substr
+    masm.load_heap_oop(rbx, Address(rdi, value_offset));
+    masm.movl(rcx, Address(rdi, offset_offset));
+    masm.lea(rdi, Address(rbx, rcx, Address::times_2, base_offset));
+    masm.push(rdi);
+    masm.push(rax);
+    masm.jmpb(PREP_FOR_SCAN);
+
+    // Substr count saved at sp
+    // Substr saved at sp+8
+    // String saved at sp+16
+
+    // Prep to load substr for scan
+    masm.bind(LOAD_SUBSTR);
+    masm.movptr(rdi, Address(rsp, 8));
+    masm.movl(rax, Address(rsp, 0));
+
+    // Load substr
+    masm.bind(PREP_FOR_SCAN);
+    masm.movdqu(tmp1Reg, Address(rdi, 0));
+    masm.addq(rdx, 8);    // prime the loop
+    masm.subptr(rsi, 16);
+
+    // Scan string for substr in 16-byte vectors
+    masm.bind(SCAN_TO_SUBSTR);
+    masm.subq(rdx, 8);
+    masm.addptr(rsi, 16);
+    masm.pcmpestri(tmp1Reg, Address(rsi, 0), 0x0d);
+    masm.jcc(Assembler::above, SCAN_TO_SUBSTR);
+    masm.jccb(Assembler::aboveEqual, RET_NOT_FOUND);
+
+    // Fallthru: found a potential substr
+
+    //Make sure string is still long enough
+    masm.subl(rdx, rcx);
+    masm.cmpl(rdx, rax);
+    masm.jccb(Assembler::negative, RET_NOT_FOUND);
+    // Compute start addr of substr
+    masm.lea(rsi, Address(rsi, rcx, Address::times_2));
+    masm.movptr(rbx, rsi);
+
+    // Compare potential substr
+    masm.addq(rdx, 8);        // prime the loop
+    masm.addq(rax, 8);
+    masm.subptr(rsi, 16);
+    masm.subptr(rdi, 16);
+
+    // Scan 16-byte vectors of string and substr
+    masm.bind(SCAN_SUBSTR);
+    masm.subq(rax, 8);
+    masm.subq(rdx, 8);
+    masm.addptr(rsi, 16);
+    masm.addptr(rdi, 16);
+    masm.movdqu(tmp1Reg, Address(rdi, 0));
+    masm.pcmpestri(tmp1Reg, Address(rsi, 0), 0x0d);
+    masm.jcc(Assembler::noOverflow, LOAD_SUBSTR);   // OF == 0
+    masm.jcc(Assembler::positive, SCAN_SUBSTR);     // SF == 0
+
+    // Compute substr offset
+    masm.movptr(rsi, Address(rsp, 16));
+    masm.subptr(rbx, rsi);
+    masm.shrl(rbx, 1);
+    masm.jmpb(CLEANUP);
+
+    masm.bind(RET_NEG_ONE);
+    masm.movl(rbx, -1);
+    masm.jmpb(DONE);
+
+    masm.bind(RET_NOT_FOUND);
+    masm.movl(rbx, -1);
+
+    masm.bind(CLEANUP);
+    masm.addptr(rsp, 24);
+
+    masm.bind(DONE);
+  %}
+
+  enc_class enc_String_Equals(rdi_RegP str1, rsi_RegP str2, regD tmp1, regD tmp2,
+                              rbx_RegI tmp3, rcx_RegI tmp2, rax_RegI result) %{
+    Label RET_TRUE, RET_FALSE, DONE, COMPARE_VECTORS, COMPARE_CHAR;
+    MacroAssembler masm(&cbuf);
+
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+
+    int value_offset  = java_lang_String::value_offset_in_bytes();
+    int offset_offset = java_lang_String::offset_offset_in_bytes();
+    int count_offset  = java_lang_String::count_offset_in_bytes();
+    int base_offset   = arrayOopDesc::base_offset_in_bytes(T_CHAR);
+
+    // does source == target string?
+    masm.cmpptr(rdi, rsi);
+    masm.jcc(Assembler::equal, RET_TRUE);
+
+    // get and compare counts
+    masm.movl(rcx, Address(rdi, count_offset));
+    masm.movl(rax, Address(rsi, count_offset));
+    masm.cmpl(rcx, rax);
+    masm.jcc(Assembler::notEqual, RET_FALSE);
+    masm.testl(rax, rax);
+    masm.jcc(Assembler::zero, RET_TRUE);
+
+    // get source string offset and value
+    masm.load_heap_oop(rbx, Address(rsi, value_offset));
+    masm.movl(rax, Address(rsi, offset_offset));
+    masm.lea(rsi, Address(rbx, rax, Address::times_2, base_offset));
+
+    // get compare string offset and value
+    masm.load_heap_oop(rbx, Address(rdi, value_offset));
+    masm.movl(rax, Address(rdi, offset_offset));
+    masm.lea(rdi, Address(rbx, rax, Address::times_2, base_offset));
+
+    // Set byte count
+    masm.shll(rcx, 1);
+    masm.movl(rax, rcx);
+
+    if (UseSSE42Intrinsics) {
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+      // Compare 16-byte vectors
+      masm.andl(rcx, 0xfffffff0);  // vector count (in bytes)
+      masm.andl(rax, 0x0000000e);  // tail count (in bytes)
+      masm.testl(rcx, rcx);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+      masm.lea(rdi, Address(rdi, rcx, Address::times_1));
+      masm.lea(rsi, Address(rsi, rcx, Address::times_1));
+      masm.negptr(rcx);
+
+      masm.bind(COMPARE_WIDE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(rdi, rcx, Address::times_1));
+      masm.movdqu(tmp2Reg, Address(rsi, rcx, Address::times_1));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+      masm.jccb(Assembler::notZero, RET_FALSE);
+      masm.addptr(rcx, 16);
+      masm.jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+      masm.bind(COMPARE_TAIL);
+      masm.movl(rcx, rax);
+      // Fallthru to tail compare
+    }
+
+    // Compare 4-byte vectors
+    masm.andl(rcx, 0xfffffffc);  // vector count (in bytes)
+    masm.andl(rax, 0x00000002);  // tail char (in bytes)
+    masm.testl(rcx, rcx);
+    masm.jccb(Assembler::zero, COMPARE_CHAR);
+    masm.lea(rdi, Address(rdi, rcx, Address::times_1));
+    masm.lea(rsi, Address(rsi, rcx, Address::times_1));
+    masm.negptr(rcx);
+
+    masm.bind(COMPARE_VECTORS);
+    masm.movl(rbx, Address(rdi, rcx, Address::times_1));
+    masm.cmpl(rbx, Address(rsi, rcx, Address::times_1));
+    masm.jccb(Assembler::notEqual, RET_FALSE);
+    masm.addptr(rcx, 4);
+    masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+
+    // Compare trailing char (final 2 bytes), if any
+    masm.bind(COMPARE_CHAR);
+    masm.testl(rax, rax);
+    masm.jccb(Assembler::zero, RET_TRUE);
+    masm.load_unsigned_short(rbx, Address(rdi, 0));
+    masm.load_unsigned_short(rcx, Address(rsi, 0));
+    masm.cmpl(rbx, rcx);
+    masm.jccb(Assembler::notEqual, RET_FALSE);
+
+    masm.bind(RET_TRUE);
+    masm.movl(rax, 1);   // return true
+    masm.jmpb(DONE);
+
+    masm.bind(RET_FALSE);
+    masm.xorl(rax, rax); // return false
+
+    masm.bind(DONE);
+  %}
+
+  enc_class enc_Array_Equals(rdi_RegP ary1, rsi_RegP ary2, regD tmp1, regD tmp2,
+                             rax_RegI tmp3, rbx_RegI tmp4, rcx_RegI result) %{
+    Label TRUE_LABEL, FALSE_LABEL, DONE, COMPARE_VECTORS, COMPARE_CHAR;
+    MacroAssembler masm(&cbuf);
+
+    XMMRegister tmp1Reg   = as_XMMRegister($tmp1$$reg);
+    XMMRegister tmp2Reg   = as_XMMRegister($tmp2$$reg);
+    Register ary1Reg      = as_Register($ary1$$reg);
+    Register ary2Reg      = as_Register($ary2$$reg);
+    Register tmp3Reg      = as_Register($tmp3$$reg);
+    Register tmp4Reg      = as_Register($tmp4$$reg);
+    Register resultReg    = as_Register($result$$reg);
 
     int length_offset  = arrayOopDesc::length_offset_in_bytes();
     int base_offset    = arrayOopDesc::base_offset_in_bytes(T_CHAR);
 
     // Check the input args
-    masm.cmpq(ary1Reg, ary2Reg);                        
+    masm.cmpq(ary1Reg, ary2Reg);
     masm.jcc(Assembler::equal, TRUE_LABEL);
-    masm.testq(ary1Reg, ary1Reg);                       
+    masm.testq(ary1Reg, ary1Reg);
     masm.jcc(Assembler::zero, FALSE_LABEL);
-    masm.testq(ary2Reg, ary2Reg);                       
+    masm.testq(ary2Reg, ary2Reg);
     masm.jcc(Assembler::zero, FALSE_LABEL);
 
     // Check the lengths
-    masm.movl(tmp2Reg, Address(ary1Reg, length_offset));
+    masm.movl(tmp4Reg, Address(ary1Reg, length_offset));
     masm.movl(resultReg, Address(ary2Reg, length_offset));
-    masm.cmpl(tmp2Reg, resultReg);
+    masm.cmpl(tmp4Reg, resultReg);
     masm.jcc(Assembler::notEqual, FALSE_LABEL);
     masm.testl(resultReg, resultReg);
     masm.jcc(Assembler::zero, TRUE_LABEL);
 
-    // Get the number of 4 byte vectors to compare
-    masm.shrl(resultReg, 1);
-
-    // Check for odd-length arrays
-    masm.andl(tmp2Reg, 1);
-    masm.testl(tmp2Reg, tmp2Reg);
-    masm.jcc(Assembler::zero, COMPARE_LOOP_HDR);
-
-    // Compare 2-byte "tail" at end of arrays
-    masm.load_unsigned_short(tmp1Reg, Address(ary1Reg, resultReg, Address::times_4, base_offset));
-    masm.load_unsigned_short(tmp2Reg, Address(ary2Reg, resultReg, Address::times_4, base_offset));
-    masm.cmpl(tmp1Reg, tmp2Reg);
-    masm.jcc(Assembler::notEqual, FALSE_LABEL);
+    //load array address
+    masm.lea(ary1Reg, Address(ary1Reg, base_offset));
+    masm.lea(ary2Reg, Address(ary2Reg, base_offset));
+
+    //set byte count
+    masm.shll(tmp4Reg, 1);
+    masm.movl(resultReg,tmp4Reg);
+
+    if (UseSSE42Intrinsics){
+      // With SSE4.2, use double quad vector compare
+      Label COMPARE_WIDE_VECTORS, COMPARE_TAIL;
+      // Compare 16-byte vectors
+      masm.andl(tmp4Reg, 0xfffffff0);    // vector count (in bytes)
+      masm.andl(resultReg, 0x0000000e);  // tail count (in bytes)
+      masm.testl(tmp4Reg, tmp4Reg);
+      masm.jccb(Assembler::zero, COMPARE_TAIL);
+      masm.lea(ary1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+      masm.lea(ary2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+      masm.negptr(tmp4Reg);
+
+      masm.bind(COMPARE_WIDE_VECTORS);
+      masm.movdqu(tmp1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+      masm.movdqu(tmp2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+      masm.pxor(tmp1Reg, tmp2Reg);
+      masm.ptest(tmp1Reg, tmp1Reg);
+
+      masm.jccb(Assembler::notZero, FALSE_LABEL);
+      masm.addptr(tmp4Reg, 16);
+      masm.jcc(Assembler::notZero, COMPARE_WIDE_VECTORS);
+      masm.bind(COMPARE_TAIL);
+      masm.movl(tmp4Reg, resultReg);
+      // Fallthru to tail compare
+    }
+
+   // Compare 4-byte vectors
+    masm.andl(tmp4Reg, 0xfffffffc);    // vector count (in bytes)
+    masm.andl(resultReg, 0x00000002);  // tail char (in bytes)
+    masm.testl(tmp4Reg, tmp4Reg); //if tmp2 == 0, only compare char
+    masm.jccb(Assembler::zero, COMPARE_CHAR);
+    masm.lea(ary1Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+    masm.lea(ary2Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+    masm.negptr(tmp4Reg);
+
+    masm.bind(COMPARE_VECTORS);
+    masm.movl(tmp3Reg, Address(ary1Reg, tmp4Reg, Address::times_1));
+    masm.cmpl(tmp3Reg, Address(ary2Reg, tmp4Reg, Address::times_1));
+    masm.jccb(Assembler::notEqual, FALSE_LABEL);
+    masm.addptr(tmp4Reg, 4);
+    masm.jcc(Assembler::notZero, COMPARE_VECTORS);
+
+    // Compare trailing char (final 2 bytes), if any
+    masm.bind(COMPARE_CHAR);
     masm.testl(resultReg, resultReg);
-    masm.jcc(Assembler::zero, TRUE_LABEL);
-
-    // Setup compare loop
-    masm.bind(COMPARE_LOOP_HDR);
-    // Shift tmp1Reg and tmp2Reg to the last 4-byte boundary of the arrays
-    masm.leaq(tmp1Reg, Address(ary1Reg, resultReg, Address::times_4, base_offset));
-    masm.leaq(tmp2Reg, Address(ary2Reg, resultReg, Address::times_4, base_offset));
-    masm.negq(resultReg);
-
-    // 4-byte-wide compare loop
-    masm.bind(COMPARE_LOOP);
-    masm.movl(ary1Reg, Address(tmp1Reg, resultReg, Address::times_4, 0));
-    masm.movl(ary2Reg, Address(tmp2Reg, resultReg, Address::times_4, 0));
-    masm.cmpl(ary1Reg, ary2Reg);
-    masm.jcc(Assembler::notEqual, FALSE_LABEL);
-    masm.incrementq(resultReg);
-    masm.jcc(Assembler::notZero, COMPARE_LOOP);
+    masm.jccb(Assembler::zero, TRUE_LABEL);
+    masm.load_unsigned_short(tmp3Reg, Address(ary1Reg, 0));
+    masm.load_unsigned_short(tmp4Reg, Address(ary2Reg, 0));
+    masm.cmpl(tmp3Reg, tmp4Reg);
+    masm.jccb(Assembler::notEqual, FALSE_LABEL);
 
     masm.bind(TRUE_LABEL);
     masm.movl(resultReg, 1);   // return true
-    masm.jmp(DONE_LABEL);
+    masm.jmpb(DONE);
 
     masm.bind(FALSE_LABEL);
     masm.xorl(resultReg, resultReg); // return false
 
     // That's it
-    masm.bind(DONE_LABEL);
+    masm.bind(DONE);
   %}
 
   enc_class enc_rethrow()
@@ -4162,33 +4445,6 @@
     // done:
   %}
 
-  enc_class enc_membar_acquire
-  %{
-    // [jk] not needed currently, if you enable this and it really
-    // emits code don't forget to the remove the "size(0)" line in
-    // membar_acquire()
-    // MacroAssembler masm(&cbuf);
-    // masm.membar(Assembler::Membar_mask_bits(Assembler::LoadStore |
-    //                                         Assembler::LoadLoad));
-  %}
-
-  enc_class enc_membar_release
-  %{
-    // [jk] not needed currently, if you enable this and it really
-    // emits code don't forget to the remove the "size(0)" line in
-    // membar_release()
-    // MacroAssembler masm(&cbuf);
-    // masm.membar(Assembler::Membar_mask_bits(Assembler::LoadStore |
-    //                                         Assembler::StoreStore));
-  %}
-
-  enc_class enc_membar_volatile
-  %{
-    MacroAssembler masm(&cbuf);
-    masm.membar(Assembler::Membar_mask_bits(Assembler::StoreLoad |
-                                            Assembler::StoreStore));
-  %}
-
   // Safepoint Poll.  This polls the safepoint page, and causes an
   // exception if it is not readable. Unfortunately, it kills
   // RFLAGS in the process.
@@ -5114,7 +5370,7 @@
 %}
 
 // Double register operands
-operand regD()
+operand regD() 
 %{
   constraint(ALLOC_IN_RC(double_reg));
   match(RegD);
@@ -7458,7 +7714,7 @@
   ins_cost(0);
 
   size(0);
-  format %{ "MEMBAR-acquire" %}
+  format %{ "MEMBAR-acquire ! (empty encoding)" %}
   ins_encode();
   ins_pipe(empty);
 %}
@@ -7481,7 +7737,7 @@
   ins_cost(0);
 
   size(0);
-  format %{ "MEMBAR-release" %}
+  format %{ "MEMBAR-release ! (empty encoding)" %}
   ins_encode();
   ins_pipe(empty);
 %}
@@ -7498,13 +7754,22 @@
   ins_pipe(empty);
 %}
 
-instruct membar_volatile()
-%{
+instruct membar_volatile(rFlagsReg cr) %{
   match(MemBarVolatile);
+  effect(KILL cr);
   ins_cost(400);
 
-  format %{ "MEMBAR-volatile" %}
-  ins_encode(enc_membar_volatile);
+  format %{ 
+    $$template
+    if (os::is_MP()) {
+      $$emit$$"lock addl [rsp + #0], 0\t! membar_volatile"
+    } else {
+      $$emit$$"MEMBAR-volatile ! (empty encoding)"
+    }
+  %}
+  ins_encode %{
+    __ membar(Assembler::StoreLoad);
+  %}
   ins_pipe(pipe_slow);
 %}
 
@@ -11558,27 +11823,52 @@
   ins_pipe(pipe_slow);
 %}
 
-instruct string_compare(rdi_RegP str1, rsi_RegP str2, rax_RegI tmp1,
-                        rbx_RegI tmp2, rcx_RegI result, rFlagsReg cr)
+instruct string_compare(rdi_RegP str1, rsi_RegP str2, regD tmp1, regD tmp2,
+                        rax_RegI tmp3, rbx_RegI tmp4, rcx_RegI result, rFlagsReg cr)
 %{
   match(Set result (StrComp str1 str2));
-  effect(USE_KILL str1, USE_KILL str2, KILL tmp1, KILL tmp2, KILL cr);
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL str1, USE_KILL str2, KILL tmp3, KILL tmp4, KILL cr);
   //ins_cost(300);
 
   format %{ "String Compare $str1, $str2 -> $result    // XXX KILL RAX, RBX" %}
-  ins_encode( enc_String_Compare() );
+  ins_encode( enc_String_Compare(str1, str2, tmp1, tmp2, tmp3, tmp4, result) );
+  ins_pipe( pipe_slow );
+%}
+
+instruct string_indexof(rsi_RegP str1, rdi_RegP str2, regD tmp1, rax_RegI tmp2,
+                        rcx_RegI tmp3, rdx_RegI tmp4, rbx_RegI result, rFlagsReg cr)
+%{
+  predicate(UseSSE42Intrinsics);
+  match(Set result (StrIndexOf str1 str2));
+  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, KILL tmp2, KILL tmp3, KILL tmp4, KILL cr);
+
+  format %{ "String IndexOf $str1,$str2 -> $result   // KILL RAX, RCX, RDX" %}
+  ins_encode( enc_String_IndexOf(str1, str2, tmp1, tmp2, tmp3, tmp4, result) );
+  ins_pipe( pipe_slow );
+%}
+
+// fast string equals
+instruct string_equals(rdi_RegP str1, rsi_RegP str2, regD tmp1, regD tmp2, rbx_RegI tmp3,
+                       rcx_RegI tmp4, rax_RegI result, rFlagsReg cr)
+%{
+  match(Set result (StrEquals str1 str2));
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL str1, USE_KILL str2, KILL tmp3, KILL tmp4, KILL cr);
+
+  format %{ "String Equals $str1,$str2 -> $result    // KILL RBX, RCX" %}
+  ins_encode( enc_String_Equals(str1, str2, tmp1, tmp2, tmp3, tmp4, result) );
   ins_pipe( pipe_slow );
 %}
 
 // fast array equals
-instruct array_equals(rdi_RegP ary1, rsi_RegP ary2, rax_RegI tmp1, 
-                      rbx_RegI tmp2, rcx_RegI result, rFlagsReg cr) %{
+instruct array_equals(rdi_RegP ary1, rsi_RegP ary2, regD tmp1, regD tmp2, rax_RegI tmp3,
+                      rbx_RegI tmp4, rcx_RegI result, rFlagsReg cr)
+%{
   match(Set result (AryEq ary1 ary2));
-  effect(USE_KILL ary1, USE_KILL ary2, KILL tmp1, KILL tmp2, KILL cr);
+  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);
   //ins_cost(300);
 
-  format %{ "Array Equals $ary1,$ary2 -> $result    // KILL RAX, RBX" %}
-  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, tmp2, result) );
+  format %{ "Array Equals $ary1,$ary2 -> $result   // KILL RAX, RBX" %}
+  ins_encode( enc_Array_Equals(ary1, ary2, tmp1, tmp2, tmp3, tmp4, result) );
   ins_pipe( pipe_slow );
 %}
 
--- a/src/os/linux/vm/os_linux.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os/linux/vm/os_linux.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1518,21 +1518,51 @@
 
 const char* os::get_temp_directory() { return "/tmp/"; }
 
-void os::dll_build_name(
-    char* buffer, size_t buflen, const char* pname, const char* fname) {
-  // copied from libhpi
+static bool file_exists(const char* filename) {
+  struct stat statbuf;
+  if (filename == NULL || strlen(filename) == 0) {
+    return false;
+  }
+  return os::stat(filename, &statbuf) == 0;
+}
+
+void os::dll_build_name(char* buffer, size_t buflen,
+                        const char* pname, const char* fname) {
+  // Copied from libhpi
   const size_t pnamelen = pname ? strlen(pname) : 0;
 
-  /* Quietly truncate on buffer overflow.  Should be an error. */
+  // Quietly truncate on buffer overflow.  Should be an error.
   if (pnamelen + strlen(fname) + 10 > (size_t) buflen) {
       *buffer = '\0';
       return;
   }
 
   if (pnamelen == 0) {
-      sprintf(buffer, "lib%s.so", fname);
+    snprintf(buffer, buflen, "lib%s.so", fname);
+  } else if (strchr(pname, *os::path_separator()) != NULL) {
+    int n;
+    char** pelements = split_path(pname, &n);
+    for (int i = 0 ; i < n ; i++) {
+      // Really shouldn't be NULL, but check can't hurt
+      if (pelements[i] == NULL || strlen(pelements[i]) == 0) {
+        continue; // skip the empty path values
+      }
+      snprintf(buffer, buflen, "%s/lib%s.so", pelements[i], fname);
+      if (file_exists(buffer)) {
+        break;
+      }
+    }
+    // release the storage
+    for (int i = 0 ; i < n ; i++) {
+      if (pelements[i] != NULL) {
+        FREE_C_HEAP_ARRAY(char, pelements[i]);
+      }
+    }
+    if (pelements != NULL) {
+      FREE_C_HEAP_ARRAY(char*, pelements);
+    }
   } else {
-      sprintf(buffer, "%s/lib%s.so", pname, fname);
+    snprintf(buffer, buflen, "%s/lib%s.so", pname, fname);
   }
 }
 
@@ -2269,15 +2299,16 @@
 //       All it does is to check if there are enough free pages
 //       left at the time of mmap(). This could be a potential
 //       problem.
-bool os::commit_memory(char* addr, size_t size) {
-  uintptr_t res = (uintptr_t) ::mmap(addr, size,
-                                   PROT_READ|PROT_WRITE|PROT_EXEC,
+bool os::commit_memory(char* addr, size_t size, bool exec) {
+  int prot = exec ? PROT_READ|PROT_WRITE|PROT_EXEC : PROT_READ|PROT_WRITE;
+  uintptr_t res = (uintptr_t) ::mmap(addr, size, prot,
                                    MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0);
   return res != (uintptr_t) MAP_FAILED;
 }
 
-bool os::commit_memory(char* addr, size_t size, size_t alignment_hint) {
-  return commit_memory(addr, size);
+bool os::commit_memory(char* addr, size_t size, size_t alignment_hint,
+                       bool exec) {
+  return commit_memory(addr, size, exec);
 }
 
 void os::realign_memory(char *addr, size_t bytes, size_t alignment_hint) { }
@@ -2417,8 +2448,7 @@
 unsigned long* os::Linux::_numa_all_nodes;
 
 bool os::uncommit_memory(char* addr, size_t size) {
-  return ::mmap(addr, size,
-                PROT_READ|PROT_WRITE|PROT_EXEC,
+  return ::mmap(addr, size, PROT_NONE,
                 MAP_PRIVATE|MAP_FIXED|MAP_NORESERVE|MAP_ANONYMOUS, -1, 0)
     != MAP_FAILED;
 }
@@ -2441,7 +2471,9 @@
     flags |= MAP_FIXED;
   }
 
-  addr = (char*)::mmap(requested_addr, bytes, PROT_READ|PROT_WRITE|PROT_EXEC,
+  // Map uncommitted pages PROT_READ and PROT_WRITE, change access
+  // to PROT_EXEC if executable when we commit the page.
+  addr = (char*)::mmap(requested_addr, bytes, PROT_READ|PROT_WRITE,
                        flags, -1, 0);
 
   if (addr != MAP_FAILED) {
@@ -2582,7 +2614,9 @@
 #define SHM_HUGETLB 04000
 #endif
 
-char* os::reserve_memory_special(size_t bytes, char* req_addr) {
+char* os::reserve_memory_special(size_t bytes, char* req_addr, bool exec) {
+  // "exec" is passed in but not used.  Creating the shared image for
+  // the code cache doesn't have an SHM_X executable permission to check.
   assert(UseLargePages, "only for large pages");
 
   key_t key = IPC_PRIVATE;
--- a/src/os/solaris/vm/os_solaris.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os/solaris/vm/os_solaris.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1827,21 +1827,51 @@
 
 const char* os::get_temp_directory() { return "/tmp/"; }
 
-void os::dll_build_name(
-    char* buffer, size_t buflen, const char* pname, const char* fname) {
-  // copied from libhpi
+static bool file_exists(const char* filename) {
+  struct stat statbuf;
+  if (filename == NULL || strlen(filename) == 0) {
+    return false;
+  }
+  return os::stat(filename, &statbuf) == 0;
+}
+
+void os::dll_build_name(char* buffer, size_t buflen,
+                        const char* pname, const char* fname) {
+  // Copied from libhpi
   const size_t pnamelen = pname ? strlen(pname) : 0;
 
-  /* Quietly truncate on buffer overflow.  Should be an error. */
+  // Quietly truncate on buffer overflow.  Should be an error.
   if (pnamelen + strlen(fname) + 10 > (size_t) buflen) {
       *buffer = '\0';
       return;
   }
 
   if (pnamelen == 0) {
-      sprintf(buffer, "lib%s.so", fname);
+    snprintf(buffer, buflen, "lib%s.so", fname);
+  } else if (strchr(pname, *os::path_separator()) != NULL) {
+    int n;
+    char** pelements = split_path(pname, &n);
+    for (int i = 0 ; i < n ; i++) {
+      // really shouldn't be NULL but what the heck, check can't hurt
+      if (pelements[i] == NULL || strlen(pelements[i]) == 0) {
+        continue; // skip the empty path values
+      }
+      snprintf(buffer, buflen, "%s/lib%s.so", pelements[i], fname);
+      if (file_exists(buffer)) {
+        break;
+      }
+    }
+    // release the storage
+    for (int i = 0 ; i < n ; i++) {
+      if (pelements[i] != NULL) {
+        FREE_C_HEAP_ARRAY(char, pelements[i]);
+      }
+    }
+    if (pelements != NULL) {
+      FREE_C_HEAP_ARRAY(char*, pelements);
+    }
   } else {
-      sprintf(buffer, "%s/lib%s.so", pname, fname);
+    snprintf(buffer, buflen, "%s/lib%s.so", pname, fname);
   }
 }
 
@@ -2623,15 +2653,16 @@
   return page_size;
 }
 
-bool os::commit_memory(char* addr, size_t bytes) {
+bool os::commit_memory(char* addr, size_t bytes, bool exec) {
+  int prot = exec ? PROT_READ|PROT_WRITE|PROT_EXEC : PROT_READ|PROT_WRITE;
   size_t size = bytes;
   return
-     NULL != Solaris::mmap_chunk(addr, size, MAP_PRIVATE|MAP_FIXED,
-                                 PROT_READ | PROT_WRITE | PROT_EXEC);
-}
-
-bool os::commit_memory(char* addr, size_t bytes, size_t alignment_hint) {
-  if (commit_memory(addr, bytes)) {
+     NULL != Solaris::mmap_chunk(addr, size, MAP_PRIVATE|MAP_FIXED, prot);
+}
+
+bool os::commit_memory(char* addr, size_t bytes, size_t alignment_hint,
+                       bool exec) {
+  if (commit_memory(addr, bytes, exec)) {
     if (UseMPSS && alignment_hint > (size_t)vm_page_size()) {
       // If the large page size has been set and the VM
       // is using large pages, use the large page size
@@ -3220,7 +3251,9 @@
   return true;
 }
 
-char* os::reserve_memory_special(size_t bytes, char* addr) {
+char* os::reserve_memory_special(size_t bytes, char* addr, bool exec) {
+  // "exec" is passed in but not used.  Creating the shared image for
+  // the code cache doesn't have an SHM_X executable permission to check.
   assert(UseLargePages && UseISM, "only for ISM large pages");
 
   size_t size = bytes;
--- a/src/os/windows/vm/os_windows.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os/windows/vm/os_windows.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1004,26 +1004,61 @@
     }
 }
 
-void os::dll_build_name(char *holder, size_t holderlen,
-                        const char* pname, const char* fname)
-{
-    // copied from libhpi
-    const size_t pnamelen = pname ? strlen(pname) : 0;
-    const char c = (pnamelen > 0) ? pname[pnamelen-1] : 0;
-
-    /* Quietly truncates on buffer overflow. Should be an error. */
-    if (pnamelen + strlen(fname) + 10 > holderlen) {
-        *holder = '\0';
-        return;
+static bool file_exists(const char* filename) {
+  if (filename == NULL || strlen(filename) == 0) {
+    return false;
+  }
+  return GetFileAttributes(filename) != INVALID_FILE_ATTRIBUTES;
+}
+
+void os::dll_build_name(char *buffer, size_t buflen,
+                        const char* pname, const char* fname) {
+  // Copied from libhpi
+  const size_t pnamelen = pname ? strlen(pname) : 0;
+  const char c = (pnamelen > 0) ? pname[pnamelen-1] : 0;
+
+  // Quietly truncates on buffer overflow. Should be an error.
+  if (pnamelen + strlen(fname) + 10 > buflen) {
+    *buffer = '\0';
+    return;
+  }
+
+  if (pnamelen == 0) {
+    jio_snprintf(buffer, buflen, "%s.dll", fname);
+  } else if (c == ':' || c == '\\') {
+    jio_snprintf(buffer, buflen, "%s%s.dll", pname, fname);
+  } else if (strchr(pname, *os::path_separator()) != NULL) {
+    int n;
+    char** pelements = split_path(pname, &n);
+    for (int i = 0 ; i < n ; i++) {
+      char* path = pelements[i];
+      // Really shouldn't be NULL, but check can't hurt
+      size_t plen = (path == NULL) ? 0 : strlen(path);
+      if (plen == 0) {
+        continue; // skip the empty path values
+      }
+      const char lastchar = path[plen - 1];
+      if (lastchar == ':' || lastchar == '\\') {
+        jio_snprintf(buffer, buflen, "%s%s.dll", path, fname);
+      } else {
+        jio_snprintf(buffer, buflen, "%s\\%s.dll", path, fname);
+      }
+      if (file_exists(buffer)) {
+        break;
+      }
     }
-
-    if (pnamelen == 0) {
-        sprintf(holder, "%s.dll", fname);
-    } else if (c == ':' || c == '\\') {
-        sprintf(holder, "%s%s.dll", pname, fname);
-    } else {
-        sprintf(holder, "%s\\%s.dll", pname, fname);
+    // release the storage
+    for (int i = 0 ; i < n ; i++) {
+      if (pelements[i] != NULL) {
+        FREE_C_HEAP_ARRAY(char, pelements[i]);
+      }
     }
+    if (pelements != NULL) {
+      FREE_C_HEAP_ARRAY(char*, pelements);
+    }
+  } else {
+    jio_snprintf(buffer, buflen, "%s\\%s.dll", pname, fname);
+  }
 }
 
 // Needs to be in os specific directory because windows requires another
@@ -2189,7 +2224,8 @@
           if (addr > thread->stack_yellow_zone_base() && addr < thread->stack_base() ) {
                   addr = (address)((uintptr_t)addr &
                          (~((uintptr_t)os::vm_page_size() - (uintptr_t)1)));
-                  os::commit_memory( (char *)addr, thread->stack_base() - addr );
+                  os::commit_memory((char *)addr, thread->stack_base() - addr,
+                                    false );
                   return EXCEPTION_CONTINUE_EXECUTION;
           }
           else
@@ -2565,8 +2601,7 @@
   assert((size_t)addr % os::vm_allocation_granularity() == 0,
          "reserve alignment");
   assert(bytes % os::vm_allocation_granularity() == 0, "reserve block size");
-  char* res = (char*)VirtualAlloc(addr, bytes, MEM_RESERVE,
-                                  PAGE_EXECUTE_READWRITE);
+  char* res = (char*)VirtualAlloc(addr, bytes, MEM_RESERVE, PAGE_READWRITE);
   assert(res == NULL || addr == NULL || addr == res,
          "Unexpected address from reserve.");
   return res;
@@ -2595,7 +2630,7 @@
   return true;
 }
 
-char* os::reserve_memory_special(size_t bytes, char* addr) {
+char* os::reserve_memory_special(size_t bytes, char* addr, bool exec) {
 
   if (UseLargePagesIndividualAllocation) {
     if (TracePageSizes && Verbose) {
@@ -2618,7 +2653,7 @@
     p_buf = (char *) VirtualAlloc(addr,
                                  size_of_reserve,  // size of Reserve
                                  MEM_RESERVE,
-                                 PAGE_EXECUTE_READWRITE);
+                                 PAGE_READWRITE);
     // If reservation failed, return NULL
     if (p_buf == NULL) return NULL;
 
@@ -2659,7 +2694,13 @@
         p_new = (char *) VirtualAlloc(next_alloc_addr,
                                     bytes_to_rq,
                                     MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES,
-                                    PAGE_EXECUTE_READWRITE);
+                                    PAGE_READWRITE);
+        if (p_new != NULL && exec) {
+          DWORD oldprot;
+          // Windows doc says to use VirtualProtect to get execute permissions
+          VirtualProtect(next_alloc_addr, bytes_to_rq,
+                         PAGE_EXECUTE_READWRITE, &oldprot);
+        }
       }
 
       if (p_new == NULL) {
@@ -2688,10 +2729,12 @@
   } else {
     // normal policy just allocate it all at once
     DWORD flag = MEM_RESERVE | MEM_COMMIT | MEM_LARGE_PAGES;
-    char * res = (char *)VirtualAlloc(NULL,
-                                      bytes,
-                                      flag,
-                                      PAGE_EXECUTE_READWRITE);
+    char * res = (char *)VirtualAlloc(NULL, bytes, flag, PAGE_READWRITE);
+    if (res != NULL && exec) {
+      DWORD oldprot;
+      // Windows doc says to use VirtualProtect to get execute permissions
+      VirtualProtect(res, bytes, PAGE_EXECUTE_READWRITE, &oldprot);
+    }
     return res;
   }
 }
@@ -2703,7 +2746,7 @@
 void os::print_statistics() {
 }
 
-bool os::commit_memory(char* addr, size_t bytes) {
+bool os::commit_memory(char* addr, size_t bytes, bool exec) {
   if (bytes == 0) {
     // Don't bother the OS with noops.
     return true;
@@ -2712,11 +2755,19 @@
   assert(bytes % os::vm_page_size() == 0, "commit in page-sized chunks");
   // Don't attempt to print anything if the OS call fails. We're
   // probably low on resources, so the print itself may cause crashes.
-  return VirtualAlloc(addr, bytes, MEM_COMMIT, PAGE_EXECUTE_READWRITE) != NULL;
+  bool result = VirtualAlloc(addr, bytes, MEM_COMMIT, PAGE_READWRITE) != 0;
+  if (result != NULL && exec) {
+    DWORD oldprot;
+    // Windows doc says to use VirtualProtect to get execute permissions
+    return VirtualProtect(addr, bytes, PAGE_EXECUTE_READWRITE, &oldprot) != 0;
+  } else {
+    return result;
+  }
 }
 
-bool os::commit_memory(char* addr, size_t size, size_t alignment_hint) {
-  return commit_memory(addr, size);
+bool os::commit_memory(char* addr, size_t size, size_t alignment_hint,
+                       bool exec) {
+  return commit_memory(addr, size, exec);
 }
 
 bool os::uncommit_memory(char* addr, size_t bytes) {
@@ -2750,7 +2801,7 @@
 
   // Strange enough, but on Win32 one can change protection only for committed
   // memory, not a big deal anyway, as bytes less or equal than 64K
-  if (!is_committed && !commit_memory(addr, bytes)) {
+  if (!is_committed && !commit_memory(addr, bytes, prot == MEM_PROT_RWX)) {
     fatal("cannot commit protection page");
   }
   // One cannot use os::guard_memory() here, as on Win32 guard page
@@ -3248,10 +3299,10 @@
 #endif
 
   if (!UseMembar) {
-    address mem_serialize_page = (address)VirtualAlloc(NULL, os::vm_page_size(), MEM_RESERVE, PAGE_EXECUTE_READWRITE);
+    address mem_serialize_page = (address)VirtualAlloc(NULL, os::vm_page_size(), MEM_RESERVE, PAGE_READWRITE);
     guarantee( mem_serialize_page != NULL, "Reserve Failed for memory serialize page");
 
-    return_page  = (address)VirtualAlloc(mem_serialize_page, os::vm_page_size(), MEM_COMMIT, PAGE_EXECUTE_READWRITE);
+    return_page  = (address)VirtualAlloc(mem_serialize_page, os::vm_page_size(), MEM_COMMIT, PAGE_READWRITE);
     guarantee( return_page != NULL, "Commit Failed for memory serialize page");
 
     os::set_memory_serialize_page( mem_serialize_page );
--- a/src/os_cpu/linux_sparc/vm/os_linux_sparc.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/linux_sparc/vm/os_linux_sparc.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -29,13 +29,11 @@
   static jint  (*atomic_cmpxchg_func)     (jint,  volatile jint*,  jint);
   static jlong (*atomic_cmpxchg_long_func)(jlong, volatile jlong*, jlong);
   static jint  (*atomic_add_func)         (jint,  volatile jint*);
-  static void  (*fence_func)              ();
 
   static jint  atomic_xchg_bootstrap        (jint,  volatile jint*);
   static jint  atomic_cmpxchg_bootstrap     (jint,  volatile jint*,  jint);
   static jlong atomic_cmpxchg_long_bootstrap(jlong, volatile jlong*, jlong);
   static jint  atomic_add_bootstrap         (jint,  volatile jint*);
-  static void  fence_bootstrap              ();
 
   static void setup_fpu() {}
 
--- a/src/os_cpu/linux_x86/vm/orderAccess_linux_x86.inline.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/linux_x86/vm/orderAccess_linux_x86.inline.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -44,11 +44,12 @@
 
 inline void OrderAccess::fence() {
   if (os::is_MP()) {
+    // always use locked addl since mfence is sometimes expensive
 #ifdef AMD64
-    __asm__ __volatile__ ("mfence":::"memory");
+    __asm__ volatile ("lock; addl $0,0(%%rsp)" : : : "cc", "memory");
 #else
     __asm__ volatile ("lock; addl $0,0(%%esp)" : : : "cc", "memory");
-#endif // AMD64
+#endif
   }
 }
 
--- a/src/os_cpu/solaris_sparc/vm/orderAccess_solaris_sparc.inline.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_sparc/vm/orderAccess_solaris_sparc.inline.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -60,22 +60,10 @@
   dummy = 0;
 }
 
-#if defined(COMPILER2) || defined(_LP64)
-
 inline void OrderAccess::fence() {
   _OrderAccess_fence();
 }
 
-#else  // defined(COMPILER2) || defined(_LP64)
-
-inline void OrderAccess::fence() {
-  if (os::is_MP()) {
-    (*os::fence_func)();
-  }
-}
-
-#endif // defined(COMPILER2) || defined(_LP64)
-
 #endif // _GNU_SOURCE
 
 inline jbyte    OrderAccess::load_acquire(volatile jbyte*   p) { return *p; }
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -619,7 +619,6 @@
 typedef jint  cmpxchg_func_t     (jint,  volatile jint*,  jint);
 typedef jlong cmpxchg_long_func_t(jlong, volatile jlong*, jlong);
 typedef jint  add_func_t         (jint,  volatile jint*);
-typedef void  fence_func_t       ();
 
 jint os::atomic_xchg_bootstrap(jint exchange_value, volatile jint* dest) {
   // try to use the stub:
@@ -681,25 +680,10 @@
   return (*dest) += add_value;
 }
 
-void os::fence_bootstrap() {
-  // try to use the stub:
-  fence_func_t* func = CAST_TO_FN_PTR(fence_func_t*, StubRoutines::fence_entry());
-
-  if (func != NULL) {
-    os::fence_func = func;
-    (*func)();
-    return;
-  }
-  assert(Threads::number_of_threads() == 0, "for bootstrap only");
-
-  // don't have to do anything for a single thread
-}
-
 xchg_func_t*         os::atomic_xchg_func         = os::atomic_xchg_bootstrap;
 cmpxchg_func_t*      os::atomic_cmpxchg_func      = os::atomic_cmpxchg_bootstrap;
 cmpxchg_long_func_t* os::atomic_cmpxchg_long_func = os::atomic_cmpxchg_long_bootstrap;
 add_func_t*          os::atomic_add_func          = os::atomic_add_bootstrap;
-fence_func_t*        os::fence_func               = os::fence_bootstrap;
 
 #endif // !_LP64 && !COMPILER2
 
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -29,13 +29,11 @@
   static jint  (*atomic_cmpxchg_func)     (jint,  volatile jint*,  jint);
   static jlong (*atomic_cmpxchg_long_func)(jlong, volatile jlong*, jlong);
   static jint  (*atomic_add_func)         (jint,  volatile jint*);
-  static void  (*fence_func)              ();
 
   static jint  atomic_xchg_bootstrap        (jint,  volatile jint*);
   static jint  atomic_cmpxchg_bootstrap     (jint,  volatile jint*,  jint);
   static jlong atomic_cmpxchg_long_bootstrap(jlong, volatile jlong*, jlong);
   static jint  atomic_add_bootstrap         (jint,  volatile jint*);
-  static void  fence_bootstrap              ();
 
   static void setup_fpu() {}
 
--- a/src/os_cpu/solaris_x86/vm/orderAccess_solaris_x86.inline.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_x86/vm/orderAccess_solaris_x86.inline.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -61,11 +61,8 @@
 #endif // AMD64
   }
   inline void _OrderAccess_fence() {
-#ifdef AMD64
-    __asm__ __volatile__ ("mfence":::"memory");
-#else
+    // Always use locked addl since mfence is sometimes expensive
     __asm__ volatile ("lock; addl $0,0(%%esp)" : : : "cc", "memory");
-#endif // AMD64
   }
 
 }
--- a/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -794,7 +794,6 @@
 typedef jint  cmpxchg_func_t     (jint,  volatile jint*,  jint);
 typedef jlong cmpxchg_long_func_t(jlong, volatile jlong*, jlong);
 typedef jint  add_func_t         (jint,  volatile jint*);
-typedef void  fence_func_t       ();
 
 jint os::atomic_xchg_bootstrap(jint exchange_value, volatile jint* dest) {
   // try to use the stub:
@@ -856,25 +855,10 @@
   return (*dest) += add_value;
 }
 
-void os::fence_bootstrap() {
-  // try to use the stub:
-  fence_func_t* func = CAST_TO_FN_PTR(fence_func_t*, StubRoutines::fence_entry());
-
-  if (func != NULL) {
-    os::fence_func = func;
-    (*func)();
-    return;
-  }
-  assert(Threads::number_of_threads() == 0, "for bootstrap only");
-
-  // don't have to do anything for a single thread
-}
-
 xchg_func_t*         os::atomic_xchg_func         = os::atomic_xchg_bootstrap;
 cmpxchg_func_t*      os::atomic_cmpxchg_func      = os::atomic_cmpxchg_bootstrap;
 cmpxchg_long_func_t* os::atomic_cmpxchg_long_func = os::atomic_cmpxchg_long_bootstrap;
 add_func_t*          os::atomic_add_func          = os::atomic_add_bootstrap;
-fence_func_t*        os::fence_func               = os::fence_bootstrap;
 
 extern "C" _solaris_raw_setup_fpu(address ptr);
 void os::setup_fpu() {
--- a/src/os_cpu/solaris_x86/vm/os_solaris_x86.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/solaris_x86/vm/os_solaris_x86.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -32,13 +32,11 @@
   static jint  (*atomic_cmpxchg_func)     (jint,  volatile jint*,  jint);
   static jlong (*atomic_cmpxchg_long_func)(jlong, volatile jlong*, jlong);
   static jint  (*atomic_add_func)         (jint,  volatile jint*);
-  static void  (*fence_func)              ();
 
   static jint  atomic_xchg_bootstrap        (jint,  volatile jint*);
   static jint  atomic_cmpxchg_bootstrap     (jint,  volatile jint*,  jint);
   static jlong atomic_cmpxchg_long_bootstrap(jlong, volatile jlong*, jlong);
   static jint  atomic_add_bootstrap         (jint,  volatile jint*);
-  static void  fence_bootstrap              ();
 
   static void setup_fpu();
 #endif // AMD64
--- a/src/os_cpu/windows_x86/vm/orderAccess_windows_x86.inline.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/windows_x86/vm/orderAccess_windows_x86.inline.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -46,7 +46,7 @@
 
 inline void OrderAccess::fence() {
 #ifdef AMD64
-  (*os::fence_func)();
+  StubRoutines_fence();
 #else
   if (os::is_MP()) {
     __asm {
--- a/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -196,7 +196,6 @@
 typedef jlong     cmpxchg_long_func_t    (jlong,    volatile jlong*, jlong);
 typedef jint      add_func_t             (jint,     volatile jint*);
 typedef intptr_t  add_ptr_func_t         (intptr_t, volatile intptr_t*);
-typedef void      fence_func_t           ();
 
 #ifdef AMD64
 
@@ -292,27 +291,11 @@
   return (*dest) += add_value;
 }
 
-void os::fence_bootstrap() {
-  // try to use the stub:
-  fence_func_t* func = CAST_TO_FN_PTR(fence_func_t*, StubRoutines::fence_entry());
-
-  if (func != NULL) {
-    os::fence_func = func;
-    (*func)();
-    return;
-  }
-  assert(Threads::number_of_threads() == 0, "for bootstrap only");
-
-  // don't have to do anything for a single thread
-}
-
-
 xchg_func_t*         os::atomic_xchg_func         = os::atomic_xchg_bootstrap;
 xchg_ptr_func_t*     os::atomic_xchg_ptr_func     = os::atomic_xchg_ptr_bootstrap;
 cmpxchg_func_t*      os::atomic_cmpxchg_func      = os::atomic_cmpxchg_bootstrap;
 add_func_t*          os::atomic_add_func          = os::atomic_add_bootstrap;
 add_ptr_func_t*      os::atomic_add_ptr_func      = os::atomic_add_ptr_bootstrap;
-fence_func_t*        os::fence_func               = os::fence_bootstrap;
 
 #endif // AMD64
 
--- a/src/os_cpu/windows_x86/vm/os_windows_x86.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/os_cpu/windows_x86/vm/os_windows_x86.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -35,9 +35,6 @@
   static jint      (*atomic_add_func)           (jint,      volatile jint*);
   static intptr_t  (*atomic_add_ptr_func)       (intptr_t,  volatile intptr_t*);
 
-  static void      (*fence_func)                ();
-
-
   static jint      atomic_xchg_bootstrap        (jint,      volatile jint*);
   static intptr_t  atomic_xchg_ptr_bootstrap    (intptr_t,  volatile intptr_t*);
 
@@ -53,8 +50,6 @@
 #ifdef AMD64
   static jint      atomic_add_bootstrap         (jint,      volatile jint*);
   static intptr_t  atomic_add_ptr_bootstrap     (intptr_t,  volatile intptr_t*);
-
-  static void      fence_bootstrap              ();
 #endif // AMD64
 
   static void setup_fpu();
--- a/src/share/tools/MakeDeps/WinGammaPlatformVC7.java	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/tools/MakeDeps/WinGammaPlatformVC7.java	Fri Apr 03 15:59:19 2009 -0700
@@ -27,6 +27,8 @@
 
 public class WinGammaPlatformVC7 extends WinGammaPlatform {
 
+    String projectVersion() {return "7.10";};
+
     public void writeProjectFile(String projectFileName, String projectName,
                                  Vector allConfigs) throws IOException {
         System.out.println();
@@ -40,7 +42,7 @@
             "VisualStudioProject",
             new String[] {
                 "ProjectType", "Visual C++",
-                "Version", "7.10",
+                "Version", projectVersion(),
                 "Name", projectName,
                 "ProjectGUID", "{8822CB5C-1C41-41C2-8493-9F6E1994338B}",
                 "SccProjectName", "",
@@ -417,7 +419,9 @@
             new String[] {
                 "Name", "VCPreLinkEventTool",
                 "Description", BuildConfig.getFieldString(null, "PrelinkDescription"),
-                "CommandLine", cfg.expandFormat(BuildConfig.getFieldString(null, "PrelinkCommand").replace('\t', '\n'))
+                //Caution: String.replace(String,String) is available from JDK5 onwards only
+                "CommandLine", cfg.expandFormat(BuildConfig.getFieldString(null, "PrelinkCommand").replace
+                   ("\t", "&#x0D;&#x0A;"))
             }
             );
 
@@ -542,25 +546,41 @@
 }
 
 class CompilerInterfaceVC7 extends CompilerInterface {
-    Vector getBaseCompilerFlags(Vector defines, Vector includes, String outDir) {
-        Vector rv = new Vector();
+    void getBaseCompilerFlags_common(Vector defines, Vector includes, String outDir,Vector rv) {
 
         // advanced M$ IDE (2003) can only recognize name if it's first or
         // second attribute in the tag - go guess
         addAttr(rv, "Name", "VCCLCompilerTool");
         addAttr(rv, "AdditionalIncludeDirectories", Util.join(",", includes));
-        addAttr(rv, "PreprocessorDefinitions", Util.join(";", defines).replace("\"","&quot;"));
-        addAttr(rv, "UsePrecompiledHeader", "3");
-        addAttr(rv, "PrecompiledHeaderThrough", "incls"+Util.sep+"_precompiled.incl");
+        addAttr(rv, "PreprocessorDefinitions",
+                                Util.join(";", defines).replace("\"","&quot;"));
+        addAttr(rv, "PrecompiledHeaderThrough",
+                                "incls"+Util.sep+"_precompiled.incl");
         addAttr(rv, "PrecompiledHeaderFile", outDir+Util.sep+"vm.pch");
         addAttr(rv, "AssemblerListingLocation", outDir);
         addAttr(rv, "ObjectFile", outDir+Util.sep);
         addAttr(rv, "ProgramDataBaseFileName", outDir+Util.sep+"vm.pdb");
+        // Set /nologo optin
         addAttr(rv, "SuppressStartupBanner", "TRUE");
+        // Surpass the default /Tc or /Tp. 0 is compileAsDefault
         addAttr(rv, "CompileAs", "0");
+        // Set /W3 option. 3 is warningLevel_3
         addAttr(rv, "WarningLevel", "3");
+        // Set /WX option,
         addAttr(rv, "WarnAsError", "TRUE");
+        // Set /GS option
         addAttr(rv, "BufferSecurityCheck", "FALSE");
+        // Set /Zi option. 3 is debugEnabled
+        addAttr(rv, "DebugInformationFormat", "3");
+    }
+    Vector getBaseCompilerFlags(Vector defines, Vector includes, String outDir) {
+        Vector rv = new Vector();
+
+        getBaseCompilerFlags_common(defines,includes, outDir, rv);
+        // Set /Yu option. 3 is pchUseUsingSpecific
+        // Note: Starting VC8 pchUseUsingSpecific is 2 !!!
+        addAttr(rv, "UsePrecompiledHeader", "3");
+        // Set /EHsc- option
         addAttr(rv, "ExceptionHandling", "FALSE");
 
         return rv;
@@ -579,27 +599,39 @@
                 "/export:jio_vsnprintf ");
         addAttr(rv, "AdditionalDependencies", "Wsock32.lib winmm.lib");
         addAttr(rv, "OutputFile", outDll);
+        // Set /INCREMENTAL option. 1 is linkIncrementalNo
         addAttr(rv, "LinkIncremental", "1");
         addAttr(rv, "SuppressStartupBanner", "TRUE");
         addAttr(rv, "ModuleDefinitionFile", outDir+Util.sep+"vm.def");
         addAttr(rv, "ProgramDatabaseFile", outDir+Util.sep+"vm.pdb");
+        // Set /SUBSYSTEM option. 2 is subSystemWindows
         addAttr(rv, "SubSystem", "2");
         addAttr(rv, "BaseAddress", "0x8000000");
         addAttr(rv, "ImportLibrary", outDir+Util.sep+"jvm.lib");
+        // Set /MACHINE option. 1 is machineX86
         addAttr(rv, "TargetMachine", "1");
 
         return rv;
     }
 
+    void  getDebugCompilerFlags_common(String opt,Vector rv) {
+
+        // Set /On option
+        addAttr(rv, "Optimization", opt);
+        // Set /FR option. 1 is brAllInfo
+        addAttr(rv, "BrowseInformation", "1");
+        addAttr(rv, "BrowseInformationFile", "$(IntDir)" + Util.sep);
+        // Set /MD option. 2 is rtMultiThreadedDLL
+        addAttr(rv, "RuntimeLibrary", "2");
+        // Set /Oy- option
+        addAttr(rv, "OmitFramePointers", "FALSE");
+
+    }
+
     Vector getDebugCompilerFlags(String opt) {
         Vector rv = new Vector();
 
-        addAttr(rv, "Optimization", opt);
-        addAttr(rv, "OptimizeForProcessor", "1");
-        addAttr(rv, "DebugInformationFormat", "3");
-        addAttr(rv, "RuntimeLibrary", "2");
-        addAttr(rv, "BrowseInformation", "1");
-        addAttr(rv, "BrowseInformationFile", "$(IntDir)" + Util.sep);
+        getDebugCompilerFlags_common(opt,rv);
 
         return rv;
     }
@@ -607,18 +639,29 @@
     Vector getDebugLinkerFlags() {
         Vector rv = new Vector();
 
-        addAttr(rv, "GenerateDebugInformation", "TRUE");
+        addAttr(rv, "GenerateDebugInformation", "TRUE"); // == /DEBUG option
 
         return rv;
     }
 
+    void getProductCompilerFlags_common(Vector rv) {
+        // Set /O2 option. 2 is optimizeMaxSpeed
+        addAttr(rv, "Optimization", "2");
+        // Set /Oy- option
+        addAttr(rv, "OmitFramePointers", "FALSE");
+    }
+
     Vector getProductCompilerFlags() {
         Vector rv = new Vector();
 
-        addAttr(rv, "Optimization", "2");
+        getProductCompilerFlags_common(rv);
+        // Set /Ob option.  1 is expandOnlyInline
         addAttr(rv, "InlineFunctionExpansion", "1");
+        // Set /GF option.
         addAttr(rv, "StringPooling", "TRUE");
+        // Set /MD option. 2 is rtMultiThreadedDLL
         addAttr(rv, "RuntimeLibrary", "2");
+        // Set /Gy option
         addAttr(rv, "EnableFunctionLevelLinking", "TRUE");
 
         return rv;
@@ -627,7 +670,9 @@
     Vector getProductLinkerFlags() {
         Vector rv = new Vector();
 
+        // Set /OPT:REF option. 2 is optReferences
         addAttr(rv, "OptimizeReferences", "2");
+        // Set /OPT:optFolding option. 2 is optFolding
         addAttr(rv, "EnableCOMDATFolding", "2");
 
         return rv;
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/share/tools/MakeDeps/WinGammaPlatformVC8.java	Fri Apr 03 15:59:19 2009 -0700
@@ -0,0 +1,66 @@
+/*
+ * Copyright 2005-2007 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ */
+
+import java.io.*;
+import java.util.*;
+
+public class WinGammaPlatformVC8 extends WinGammaPlatformVC7 {
+
+    String projectVersion() {return "8.00";};
+
+}
+
+class CompilerInterfaceVC8 extends CompilerInterfaceVC7 {
+
+    Vector getBaseCompilerFlags(Vector defines, Vector includes, String outDir) {
+        Vector rv = new Vector();
+
+        getBaseCompilerFlags_common(defines,includes, outDir, rv);
+        // Set /Yu option. 2 is pchUseUsingSpecific
+        addAttr(rv, "UsePrecompiledHeader", "2");
+        // Set /EHsc- option. 0 is cppExceptionHandlingNo
+        addAttr(rv, "ExceptionHandling", "0");
+
+        return rv;
+    }
+
+
+    Vector getDebugCompilerFlags(String opt) {
+        Vector rv = new Vector();
+
+        getDebugCompilerFlags_common(opt,rv);
+
+        return rv;
+    }
+
+    Vector getProductCompilerFlags() {
+        Vector rv = new Vector();
+
+        getProductCompilerFlags_common(rv);
+
+        return rv;
+    }
+
+
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/src/share/tools/MakeDeps/WinGammaPlatformVC9.java	Fri Apr 03 15:59:19 2009 -0700
@@ -0,0 +1,35 @@
+/*
+ * Copyright 2005-2007 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ *
+ */
+
+import java.io.*;
+import java.util.*;
+
+public class WinGammaPlatformVC9 extends WinGammaPlatformVC8 {
+
+    String projectVersion() {return "9.00";};
+
+}
+
+class CompilerInterfaceVC9 extends CompilerInterfaceVC8 {
+}
--- a/src/share/vm/adlc/formssel.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/adlc/formssel.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -574,9 +574,13 @@
   // TEMPORARY
   // if( is_simple_chain_rule(globals) )  return false;
 
-  // String-compare uses many memorys edges, but writes none
+  // String.(compareTo/equals/indexOf) and Arrays.equals use many memorys edges,
+  // but writes none
   if( _matrule && _matrule->_rChild &&
-      strcmp(_matrule->_rChild->_opType,"StrComp")==0 )
+      ( strcmp(_matrule->_rChild->_opType,"StrComp"    )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrEquals"  )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrIndexOf" )==0 ||
+        strcmp(_matrule->_rChild->_opType,"AryEq"      )==0 ))
     return true;
 
   // Check if instruction has a USE of a memory operand class, but no defs
@@ -815,8 +819,10 @@
     return AdlcVMDeps::Parms;   // Skip the machine-state edges
 
   if( _matrule->_rChild &&
-          strcmp(_matrule->_rChild->_opType,"StrComp")==0 ) {
-        // String compare takes 1 control and 4 memory edges.
+      ( strcmp(_matrule->_rChild->_opType,"StrComp"   )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrEquals" )==0 ||
+        strcmp(_matrule->_rChild->_opType,"StrIndexOf")==0 )) {
+        // String.(compareTo/equals/indexOf) take 1 control and 4 memory edges.
     return 5;
   }
 
--- a/src/share/vm/asm/assembler.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/asm/assembler.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -143,15 +143,15 @@
 // A union type for code which has to assemble both constant and
 // non-constant operands, when the distinction cannot be made
 // statically.
-class RegisterConstant VALUE_OBJ_CLASS_SPEC {
+class RegisterOrConstant VALUE_OBJ_CLASS_SPEC {
  private:
   Register _r;
   intptr_t _c;
 
  public:
-  RegisterConstant(): _r(noreg), _c(0) {}
-  RegisterConstant(Register r): _r(r), _c(0) {}
-  RegisterConstant(intptr_t c): _r(noreg), _c(c) {}
+  RegisterOrConstant(): _r(noreg), _c(0) {}
+  RegisterOrConstant(Register r): _r(r), _c(0) {}
+  RegisterOrConstant(intptr_t c): _r(noreg), _c(c) {}
 
   Register as_register() const { assert(is_register(),""); return _r; }
   intptr_t as_constant() const { assert(is_constant(),""); return _c; }
@@ -310,13 +310,13 @@
   // offsets in code which must be generated before the object class is loaded.
   // Field offsets are never zero, since an object's header (mark word)
   // is located at offset zero.
-  RegisterConstant delayed_value(int(*value_fn)(), Register tmp, int offset = 0) {
-    return delayed_value(delayed_value_addr(value_fn), tmp, offset);
+  RegisterOrConstant delayed_value(int(*value_fn)(), Register tmp, int offset = 0) {
+    return delayed_value_impl(delayed_value_addr(value_fn), tmp, offset);
   }
-  RegisterConstant delayed_value(address(*value_fn)(), Register tmp, int offset = 0) {
-    return delayed_value(delayed_value_addr(value_fn), tmp, offset);
+  RegisterOrConstant delayed_value(address(*value_fn)(), Register tmp, int offset = 0) {
+    return delayed_value_impl(delayed_value_addr(value_fn), tmp, offset);
   }
-  virtual RegisterConstant delayed_value(intptr_t* delayed_value_addr, Register tmp, int offset) = 0;
+  virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr, Register tmp, int offset) = 0;
   // Last overloading is platform-dependent; look in assembler_<arch>.cpp.
   static intptr_t* delayed_value_addr(int(*constant_fn)());
   static intptr_t* delayed_value_addr(address(*constant_fn)());
--- a/src/share/vm/ci/ciTypeFlow.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/ci/ciTypeFlow.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -2237,7 +2237,6 @@
   for (SuccIter iter(tail); !iter.done(); iter.next()) {
     if (iter.succ() == head) {
       iter.set_succ(clone);
-      break;
     }
   }
   flow_block(tail, temp_vector, temp_set);
--- a/src/share/vm/classfile/classFileParser.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/classFileParser.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -2747,9 +2747,10 @@
                                                       super_klass(),
                                                       methods(),
                                                       access_flags,
-                                                      class_loader(),
-                                                      class_name(),
-                                                      local_interfaces());
+                                                      class_loader,
+                                                      class_name,
+                                                      local_interfaces(),
+                                                      CHECK_(nullHandle));
 
     // Size of Java itable (in words)
     itable_size = access_flags.is_interface() ? 0 : klassItable::compute_itable_size(transitive_interfaces);
@@ -3229,7 +3230,7 @@
       // print out the superclass.
       const char * from = Klass::cast(this_klass())->external_name();
       if (this_klass->java_super() != NULL) {
-        tty->print("RESOLVE %s %s\n", from, instanceKlass::cast(this_klass->java_super())->external_name());
+        tty->print("RESOLVE %s %s (super)\n", from, instanceKlass::cast(this_klass->java_super())->external_name());
       }
       // print out each of the interface classes referred to by this class.
       objArrayHandle local_interfaces(THREAD, this_klass->local_interfaces());
@@ -3239,7 +3240,7 @@
           klassOop k = klassOop(local_interfaces->obj_at(i));
           instanceKlass* to_class = instanceKlass::cast(k);
           const char * to = to_class->external_name();
-          tty->print("RESOLVE %s %s\n", from, to);
+          tty->print("RESOLVE %s %s (interface)\n", from, to);
         }
       }
     }
--- a/src/share/vm/classfile/javaClasses.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/javaClasses.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -239,22 +239,20 @@
   typeArrayOop value  = java_lang_String::value(obj);
   int          offset = java_lang_String::offset(obj);
   int          length = java_lang_String::length(obj);
-
-  ResourceMark rm(THREAD);
-  symbolHandle result;
-
-  if (length > 0) {
-    int utf8_length = UNICODE::utf8_length(value->char_at_addr(offset), length);
-    char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);
-    UNICODE::convert_to_utf8(value->char_at_addr(offset), length, chars);
-    // Allocate the symbol
-    result = oopFactory::new_symbol_handle(chars, utf8_length, CHECK_(symbolHandle()));
-  } else {
-    result = oopFactory::new_symbol_handle("", 0, CHECK_(symbolHandle()));
-  }
-  return result;
+  jchar* base = value->char_at_addr(offset);
+  symbolOop sym = SymbolTable::lookup_unicode(base, length, THREAD);
+  return symbolHandle(THREAD, sym);
 }
 
+symbolOop java_lang_String::as_symbol_or_null(oop java_string) {
+  typeArrayOop value  = java_lang_String::value(java_string);
+  int          offset = java_lang_String::offset(java_string);
+  int          length = java_lang_String::length(java_string);
+  jchar* base = value->char_at_addr(offset);
+  return SymbolTable::probe_unicode(base, length);
+}
+
+
 int java_lang_String::utf8_length(oop java_string) {
   typeArrayOop value  = java_lang_String::value(java_string);
   int          offset = java_lang_String::offset(java_string);
@@ -385,6 +383,48 @@
 }
 
 
+void java_lang_Class::print_signature(oop java_class, outputStream* st) {
+  assert(java_lang_Class::is_instance(java_class), "must be a Class object");
+  symbolOop name = NULL;
+  bool is_instance = false;
+  if (is_primitive(java_class)) {
+    name = vmSymbols::type_signature(primitive_type(java_class));
+  } else {
+    klassOop k = as_klassOop(java_class);
+    is_instance = Klass::cast(k)->oop_is_instance();
+    name = Klass::cast(k)->name();
+  }
+  if (name == NULL) {
+    st->print("<null>");
+    return;
+  }
+  if (is_instance)  st->print("L");
+  st->write((char*) name->base(), (int) name->utf8_length());
+  if (is_instance)  st->print(";");
+}
+
+symbolOop java_lang_Class::as_signature(oop java_class, bool intern_if_not_found, TRAPS) {
+  assert(java_lang_Class::is_instance(java_class), "must be a Class object");
+  symbolOop name = NULL;
+  if (is_primitive(java_class)) {
+    return vmSymbols::type_signature(primitive_type(java_class));
+  } else {
+    klassOop k = as_klassOop(java_class);
+    if (!Klass::cast(k)->oop_is_instance()) {
+      return Klass::cast(k)->name();
+    } else {
+      ResourceMark rm;
+      const char* sigstr = Klass::cast(k)->signature_name();
+      int         siglen = (int) strlen(sigstr);
+      if (!intern_if_not_found)
+        return SymbolTable::probe(sigstr, siglen);
+      else
+        return oopFactory::new_symbol(sigstr, siglen, THREAD);
+    }
+  }
+}
+
+
 klassOop java_lang_Class::array_klass(oop java_class) {
   klassOop k = klassOop(java_class->obj_field(array_klass_offset));
   assert(k == NULL || k->is_klass() && Klass::cast(k)->oop_is_javaArray(), "should be array klass");
@@ -412,6 +452,8 @@
 
 
 bool java_lang_Class::is_primitive(oop java_class) {
+  // should assert:
+  //assert(java_lang_Class::is_instance(java_class), "must be a Class object");
   klassOop k = klassOop(java_class->obj_field(klass_offset));
   return k == NULL;
 }
@@ -431,6 +473,19 @@
   return type;
 }
 
+BasicType java_lang_Class::as_BasicType(oop java_class, klassOop* reference_klass) {
+  assert(java_lang_Class::is_instance(java_class), "must be a Class object");
+  if (is_primitive(java_class)) {
+    if (reference_klass != NULL)
+      (*reference_klass) = NULL;
+    return primitive_type(java_class);
+  } else {
+    if (reference_klass != NULL)
+      (*reference_klass) = as_klassOop(java_class);
+    return T_OBJECT;
+  }
+}
+
 
 oop java_lang_Class::primitive_mirror(BasicType t) {
   oop mirror = Universe::java_mirror(t);
@@ -1988,6 +2043,21 @@
 }
 
 
+void java_lang_boxing_object::print(BasicType type, jvalue* value, outputStream* st) {
+  switch (type) {
+  case T_BOOLEAN:   st->print("%s", value->z ? "true" : "false");   break;
+  case T_CHAR:      st->print("%d", value->c);                      break;
+  case T_BYTE:      st->print("%d", value->b);                      break;
+  case T_SHORT:     st->print("%d", value->s);                      break;
+  case T_INT:       st->print("%d", value->i);                      break;
+  case T_LONG:      st->print(INT64_FORMAT, value->j);              break;
+  case T_FLOAT:     st->print("%f", value->f);                      break;
+  case T_DOUBLE:    st->print("%lf", value->d);                     break;
+  default:          st->print("type %d?", type);                    break;
+  }
+}
+
+
 // Support for java_lang_ref_Reference
 oop java_lang_ref_Reference::pending_list_lock() {
   instanceKlass* ik = instanceKlass::cast(SystemDictionary::reference_klass());
--- a/src/share/vm/classfile/javaClasses.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/javaClasses.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -107,6 +107,7 @@
 
   // Conversion
   static symbolHandle as_symbol(Handle java_string, TRAPS);
+  static symbolOop as_symbol_or_null(oop java_string);
 
   // Testers
   static bool is_instance(oop obj) {
@@ -149,6 +150,9 @@
   static oop  create_basic_type_mirror(const char* basic_type_name, BasicType type, TRAPS);
   // Conversion
   static klassOop as_klassOop(oop java_class);
+  static BasicType as_BasicType(oop java_class, klassOop* reference_klass = NULL);
+  static symbolOop as_signature(oop java_class, bool intern_if_not_found, TRAPS);
+  static void print_signature(oop java_class, outputStream *st);
   // Testing
   static bool is_instance(oop obj) {
     return obj != NULL && obj->klass() == SystemDictionary::class_klass();
@@ -668,6 +672,8 @@
   static BasicType basic_type(oop box);
   static bool is_instance(oop box)                 { return basic_type(box) != T_ILLEGAL; }
   static bool is_instance(oop box, BasicType type) { return basic_type(box) == type; }
+  static void print(oop box, outputStream* st)     { jvalue value;  print(get_value(box, &value), &value, st); }
+  static void print(BasicType type, jvalue* value, outputStream* st);
 
   static int value_offset_in_bytes(BasicType type) {
     return ( type == T_LONG || type == T_DOUBLE ) ? long_value_offset :
--- a/src/share/vm/classfile/loaderConstraints.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/loaderConstraints.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 2003-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 2003-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -60,8 +60,10 @@
   bool add_entry(symbolHandle name, klassOop klass1, Handle loader1,
                                     klassOop klass2, Handle loader2);
 
-  void check_signature_loaders(symbolHandle signature, Handle loader1,
-                               Handle loader2, bool is_method, TRAPS);
+  // Note:  The main entry point for this module is via SystemDictionary.
+  // SystemDictionary::check_signature_loaders(symbolHandle signature,
+  //                                           Handle loader1, Handle loader2,
+  //                                           bool is_method, TRAPS)
 
   klassOop find_constrained_klass(symbolHandle name, Handle loader);
   klassOop find_constrained_elem_klass(symbolHandle name, symbolHandle elem_name,
--- a/src/share/vm/classfile/symbolTable.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/symbolTable.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -109,6 +109,40 @@
   return the_table()->lookup(index, name, len, hash);
 }
 
+// Suggestion: Push unicode-based lookup all the way into the hashing
+// and probing logic, so there is no need for convert_to_utf8 until
+// an actual new symbolOop is created.
+symbolOop SymbolTable::lookup_unicode(const jchar* name, int utf16_length, TRAPS) {
+  int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
+  char stack_buf[128];
+  if (utf8_length < (int) sizeof(stack_buf)) {
+    char* chars = stack_buf;
+    UNICODE::convert_to_utf8(name, utf16_length, chars);
+    return lookup(chars, utf8_length, THREAD);
+  } else {
+    ResourceMark rm(THREAD);
+    char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
+    UNICODE::convert_to_utf8(name, utf16_length, chars);
+    return lookup(chars, utf8_length, THREAD);
+  }
+}
+
+symbolOop SymbolTable::lookup_only_unicode(const jchar* name, int utf16_length,
+                                           unsigned int& hash) {
+  int utf8_length = UNICODE::utf8_length((jchar*) name, utf16_length);
+  char stack_buf[128];
+  if (utf8_length < (int) sizeof(stack_buf)) {
+    char* chars = stack_buf;
+    UNICODE::convert_to_utf8(name, utf16_length, chars);
+    return lookup_only(chars, utf8_length, hash);
+  } else {
+    ResourceMark rm;
+    char* chars = NEW_RESOURCE_ARRAY(char, utf8_length + 1);;
+    UNICODE::convert_to_utf8(name, utf16_length, chars);
+    return lookup_only(chars, utf8_length, hash);
+  }
+}
+
 void SymbolTable::add(constantPoolHandle cp, int names_count,
                       const char** names, int* lengths, int* cp_indices,
                       unsigned int* hashValues, TRAPS) {
@@ -126,15 +160,6 @@
   }
 }
 
-// Needed for preloading classes in signatures when compiling.
-
-symbolOop SymbolTable::probe(const char* name, int len) {
-  unsigned int hashValue = hash_symbol(name, len);
-  int index = the_table()->hash_to_index(hashValue);
-  return the_table()->lookup(index, name, len, hashValue);
-}
-
-
 symbolOop SymbolTable::basic_add(int index, u1 *name, int len,
                                  unsigned int hashValue, TRAPS) {
   assert(!Universe::heap()->is_in_reserved(name) || GC_locker::is_active(),
--- a/src/share/vm/classfile/symbolTable.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/symbolTable.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -91,6 +91,10 @@
   // Only copy to C string to be added if lookup failed.
   static symbolOop lookup(symbolHandle sym, int begin, int end, TRAPS);
 
+  // jchar (utf16) version of lookups
+  static symbolOop lookup_unicode(const jchar* name, int len, TRAPS);
+  static symbolOop lookup_only_unicode(const jchar* name, int len, unsigned int& hash);
+
   static void add(constantPoolHandle cp, int names_count,
                   const char** names, int* lengths, int* cp_indices,
                   unsigned int* hashValues, TRAPS);
@@ -112,7 +116,14 @@
   // Needed for preloading classes in signatures when compiling.
   // Returns the symbol is already present in symbol table, otherwise
   // NULL.  NO ALLOCATION IS GUARANTEED!
-  static symbolOop probe(const char* name, int len);
+  static symbolOop probe(const char* name, int len) {
+    unsigned int ignore_hash;
+    return lookup_only(name, len, ignore_hash);
+  }
+  static symbolOop probe_unicode(const jchar* name, int len) {
+    unsigned int ignore_hash;
+    return lookup_only_unicode(name, len, ignore_hash);
+  }
 
   // Histogram
   static void print_histogram()     PRODUCT_RETURN;
--- a/src/share/vm/classfile/systemDictionary.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/systemDictionary.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1964,6 +1964,13 @@
   return T_OBJECT;
 }
 
+KlassHandle SystemDictionaryHandles::box_klass(BasicType t) {
+  if (t >= T_BOOLEAN && t <= T_VOID)
+    return KlassHandle(&SystemDictionary::_box_klasses[t], true);
+  else
+    return KlassHandle();
+}
+
 // Constraints on class loaders. The details of the algorithm can be
 // found in the OOPSLA'98 paper "Dynamic Class Loading in the Java
 // Virtual Machine" by Sheng Liang and Gilad Bracha.  The basic idea is
@@ -2174,11 +2181,56 @@
 }
 
 
+// Signature constraints ensure that callers and callees agree about
+// the meaning of type names in their signatures.  This routine is the
+// intake for constraints.  It collects them from several places:
+//
+//  * LinkResolver::resolve_method (if check_access is true) requires
+//    that the resolving class (the caller) and the defining class of
+//    the resolved method (the callee) agree on each type in the
+//    method's signature.
+//
+//  * LinkResolver::resolve_interface_method performs exactly the same
+//    checks.
+//
+//  * LinkResolver::resolve_field requires that the constant pool
+//    attempting to link to a field agree with the field's defining
+//    class about the type of the field signature.
+//
+//  * klassVtable::initialize_vtable requires that, when a class
+//    overrides a vtable entry allocated by a superclass, that the
+//    overriding method (i.e., the callee) agree with the superclass
+//    on each type in the method's signature.
+//
+//  * klassItable::initialize_itable requires that, when a class fills
+//    in its itables, for each non-abstract method installed in an
+//    itable, the method (i.e., the callee) agree with the interface
+//    on each type in the method's signature.
+//
+// All those methods have a boolean (check_access, checkconstraints)
+// which turns off the checks.  This is used from specialized contexts
+// such as bootstrapping, dumping, and debugging.
+//
+// No direct constraint is placed between the class and its
+// supertypes.  Constraints are only placed along linked relations
+// between callers and callees.  When a method overrides or implements
+// an abstract method in a supertype (superclass or interface), the
+// constraints are placed as if the supertype were the caller to the
+// overriding method.  (This works well, since callers to the
+// supertype have already established agreement between themselves and
+// the supertype.)  As a result of all this, a class can disagree with
+// its supertype about the meaning of a type name, as long as that
+// class neither calls a relevant method of the supertype, nor is
+// called (perhaps via an override) from the supertype.
+//
+//
+// SystemDictionary::check_signature_loaders(sig, l1, l2)
+//
 // Make sure all class components (including arrays) in the given
 // signature will be resolved to the same class in both loaders.
 // Returns the name of the type that failed a loader constraint check, or
 // NULL if no constraint failed. The returned C string needs cleaning up
-// with a ResourceMark in the caller
+// with a ResourceMark in the caller.  No exception except OOME is thrown.
 char* SystemDictionary::check_signature_loaders(symbolHandle signature,
                                                Handle loader1, Handle loader2,
                                                bool is_method, TRAPS)  {
--- a/src/share/vm/classfile/systemDictionary.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/systemDictionary.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -161,6 +161,7 @@
 class SystemDictionary : AllStatic {
   friend class VMStructs;
   friend class CompactingPermGenGen;
+  friend class SystemDictionaryHandles;
   NOT_PRODUCT(friend class instanceKlassKlass;)
 
  public:
@@ -595,3 +596,18 @@
   static bool _has_loadClassInternal;
   static bool _has_checkPackageAccess;
 };
+
+// Cf. vmSymbols vs. vmSymbolHandles
+class SystemDictionaryHandles : AllStatic {
+public:
+  #define WK_KLASS_HANDLE_DECLARE(name, ignore_symbol, option) \
+    static KlassHandle name() { \
+      SystemDictionary::name(); \
+      klassOop* loc = &SystemDictionary::_well_known_klasses[SystemDictionary::WK_KLASS_ENUM_NAME(name)]; \
+      return KlassHandle(loc, true); \
+    }
+  WK_KLASSES_DO(WK_KLASS_HANDLE_DECLARE);
+  #undef WK_KLASS_HANDLE_DECLARE
+
+  static KlassHandle box_klass(BasicType t);
+};
--- a/src/share/vm/classfile/vmSymbols.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/classfile/vmSymbols.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -49,6 +49,8 @@
   template(java_lang_Object,                          "java/lang/Object")                         \
   template(java_lang_Class,                           "java/lang/Class")                          \
   template(java_lang_String,                          "java/lang/String")                         \
+  template(java_lang_StringValue,                     "java/lang/StringValue")                    \
+  template(java_lang_StringCache,                     "java/lang/StringValue$StringCache")        \
   template(java_lang_Thread,                          "java/lang/Thread")                         \
   template(java_lang_ThreadGroup,                     "java/lang/ThreadGroup")                    \
   template(java_lang_Cloneable,                       "java/lang/Cloneable")                      \
@@ -285,6 +287,8 @@
   template(frontCacheEnabled_name,                    "frontCacheEnabled")                        \
   template(stringCacheEnabled_name,                   "stringCacheEnabled")                       \
   template(bitCount_name,                             "bitCount")                                 \
+  template(profile_name,                              "profile")                                  \
+  template(equals_name,                               "equals")                                   \
                                                                                                   \
   /* non-intrinsic name/signature pairs: */                                                       \
   template(register_method_name,                      "register")                                 \
@@ -378,7 +382,7 @@
   template(unknown_class_name,                        "<Unknown>")                                                \
                                                                                                                   \
   /* used to identify class loaders handling parallel class loading */                                            \
-  template(parallelCapable_name,                      "parallelLockMap;")                                         \
+  template(parallelCapable_name,                      "parallelLockMap")                                          \
                                                                                                                   \
   /* JVM monitoring and management support */                                                                     \
   template(java_lang_StackTraceElement_array,          "[Ljava/lang/StackTraceElement;")                          \
@@ -576,7 +580,6 @@
    do_signature(copyOfRange_signature,        "([Ljava/lang/Object;IILjava/lang/Class;)[Ljava/lang/Object;")            \
                                                                                                                         \
   do_intrinsic(_equalsC,                  java_util_Arrays,       equals_name,    equalsC_signature,             F_S)   \
-   do_name(     equals_name,                                     "equals")                                              \
    do_signature(equalsC_signature,                               "([C[C)Z")                                             \
                                                                                                                         \
   do_intrinsic(_invoke,                   java_lang_reflect_Method, invoke_name, object_array_object_object_signature, F_R) \
@@ -586,6 +589,7 @@
    do_name(     compareTo_name,                                  "compareTo")                                           \
   do_intrinsic(_indexOf,                  java_lang_String,       indexOf_name, string_int_signature,            F_R)   \
    do_name(     indexOf_name,                                    "indexOf")                                             \
+  do_intrinsic(_equals,                   java_lang_String,       equals_name, object_boolean_signature,         F_R)   \
                                                                                                                         \
   do_class(java_nio_Buffer,               "java/nio/Buffer")                                                            \
   do_intrinsic(_checkIndex,               java_nio_Buffer,        checkIndex_name, int_int_signature,            F_R)   \
--- a/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 2001-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 2001-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -825,6 +825,7 @@
   if (young_gen()->is_in_reserved(addr)) {
     assert(young_gen()->is_in(addr),
            "addr should be in allocated part of young gen");
+    if (Debugging)  return NULL;  // called from find() in debug.cpp
     Unimplemented();
   } else if (old_gen()->is_in_reserved(addr)) {
     assert(old_gen()->is_in(addr),
--- a/src/share/vm/includeDB_core	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/includeDB_core	Fri Apr 03 15:59:19 2009 -0700
@@ -3154,6 +3154,8 @@
 oopsHierarchy.cpp                       thread_<os_family>.inline.hpp
 
 orderAccess.cpp                         orderAccess.hpp
+orderAccess.cpp                         stubRoutines.hpp
+orderAccess.cpp                         thread.hpp
 
 orderAccess.hpp                         allocation.hpp
 orderAccess.hpp                         os.hpp
--- a/src/share/vm/interpreter/invocationCounter.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/interpreter/invocationCounter.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -47,6 +47,8 @@
   // executed many more times before re-entering the VM.
   int old_count = count();
   int new_count = MIN2(old_count, (int) (CompileThreshold / 2));
+  // prevent from going to zero, to distinguish from never-executed methods
+  if (new_count == 0)  new_count = 1;
   if (old_count != new_count)  set(state(), new_count);
 }
 
--- a/src/share/vm/memory/heap.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/memory/heap.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -112,7 +112,7 @@
 
   const size_t rs_align = page_size == (size_t) os::vm_page_size() ? 0 :
     MAX2(page_size, granularity);
-  ReservedSpace rs(r_size, rs_align, rs_align > 0);
+  ReservedCodeSpace rs(r_size, rs_align, rs_align > 0);
   os::trace_page_sizes("code heap", committed_size, reserved_size, page_size,
                        rs.base(), rs.size());
   if (!_memory.initialize(rs, c_size)) {
--- a/src/share/vm/oops/instanceKlass.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/instanceKlass.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1813,6 +1813,8 @@
                                           oop class_loader2, symbolOop class_name2) {
   if (class_loader1 != class_loader2) {
     return false;
+  } else if (class_name1 == class_name2) {
+    return true;                // skip painful bytewise comparison
   } else {
     ResourceMark rm;
 
@@ -1859,6 +1861,75 @@
   }
 }
 
+// Returns true iff super_method can be overridden by a method in targetclassname
+// See JSL 3rd edition 8.4.6.1
+// Assumes name-signature match
+// "this" is instanceKlass of super_method which must exist
+// note that the instanceKlass of the method in the targetclassname has not always been created yet
+bool instanceKlass::is_override(methodHandle super_method, Handle targetclassloader, symbolHandle targetclassname, TRAPS) {
+   // Private methods can not be overridden
+   if (super_method->is_private()) {
+     return false;
+   }
+   // If super method is accessible, then override
+   if ((super_method->is_protected()) ||
+       (super_method->is_public())) {
+     return true;
+   }
+   // Package-private methods are not inherited outside of package
+   assert(super_method->is_package_private(), "must be package private");
+   return(is_same_class_package(targetclassloader(), targetclassname()));
+}
+
+/* defined for now in jvm.cpp, for historical reasons *--
+klassOop instanceKlass::compute_enclosing_class_impl(instanceKlassHandle self,
+                                                     symbolOop& simple_name_result, TRAPS) {
+  ...
+}
+*/
+
+// tell if two classes have the same enclosing class (at package level)
+bool instanceKlass::is_same_package_member_impl(instanceKlassHandle class1,
+                                                klassOop class2_oop, TRAPS) {
+  if (class2_oop == class1->as_klassOop())          return true;
+  if (!Klass::cast(class2_oop)->oop_is_instance())  return false;
+  instanceKlassHandle class2(THREAD, class2_oop);
+
+  // must be in same package before we try anything else
+  if (!class1->is_same_class_package(class2->class_loader(), class2->name()))
+    return false;
+
+  // As long as there is an outer1.getEnclosingClass,
+  // shift the search outward.
+  instanceKlassHandle outer1 = class1;
+  for (;;) {
+    // As we walk along, look for equalities between outer1 and class2.
+    // Eventually, the walks will terminate as outer1 stops
+    // at the top-level class around the original class.
+    symbolOop ignore_name;
+    klassOop next = outer1->compute_enclosing_class(ignore_name, CHECK_false);
+    if (next == NULL)  break;
+    if (next == class2())  return true;
+    outer1 = instanceKlassHandle(THREAD, next);
+  }
+
+  // Now do the same for class2.
+  instanceKlassHandle outer2 = class2;
+  for (;;) {
+    symbolOop ignore_name;
+    klassOop next = outer2->compute_enclosing_class(ignore_name, CHECK_false);
+    if (next == NULL)  break;
+    // Might as well check the new outer against all available values.
+    if (next == class1())  return true;
+    if (next == outer1())  return true;
+    outer2 = instanceKlassHandle(THREAD, next);
+  }
+
+  // If by this point we have not found an equality between the
+  // two classes, we know they are in separate package members.
+  return false;
+}
+
 
 jint instanceKlass::compute_modifier_flags(TRAPS) const {
   klassOop k = as_klassOop();
@@ -1996,9 +2067,11 @@
 
 // Printing
 
+#define BULLET  " - "
+
 void FieldPrinter::do_field(fieldDescriptor* fd) {
-   if (fd->is_static() == (_obj == NULL)) {
-     _st->print("   - ");
+  _st->print(BULLET);
+   if (fd->is_static() || (_obj == NULL)) {
      fd->print_on(_st);
      _st->cr();
    } else {
@@ -2019,7 +2092,7 @@
         value->is_typeArray() &&
         offset          <= (juint) value->length() &&
         offset + length <= (juint) value->length()) {
-      st->print("string: ");
+      st->print(BULLET"string: ");
       Handle h_obj(obj);
       java_lang_String::print(h_obj, st);
       st->cr();
@@ -2027,23 +2100,26 @@
     }
   }
 
-  st->print_cr("fields:");
+  st->print_cr(BULLET"---- fields (total size %d words):", oop_size(obj));
   FieldPrinter print_nonstatic_field(st, obj);
   do_nonstatic_fields(&print_nonstatic_field);
 
   if (as_klassOop() == SystemDictionary::class_klass()) {
+    st->print(BULLET"signature: ");
+    java_lang_Class::print_signature(obj, st);
+    st->cr();
     klassOop mirrored_klass = java_lang_Class::as_klassOop(obj);
-    st->print("   - fake entry for mirror: ");
+    st->print(BULLET"fake entry for mirror: ");
     mirrored_klass->print_value_on(st);
     st->cr();
-    st->print("   - fake entry resolved_constructor: ");
+    st->print(BULLET"fake entry resolved_constructor: ");
     methodOop ctor = java_lang_Class::resolved_constructor(obj);
     ctor->print_value_on(st);
     klassOop array_klass = java_lang_Class::array_klass(obj);
-    st->print("   - fake entry for array: ");
+    st->cr();
+    st->print(BULLET"fake entry for array: ");
     array_klass->print_value_on(st);
     st->cr();
-    st->cr();
   }
 }
 
@@ -2051,6 +2127,28 @@
   st->print("a ");
   name()->print_value_on(st);
   obj->print_address_on(st);
+  if (as_klassOop() == SystemDictionary::string_klass()
+      && java_lang_String::value(obj) != NULL) {
+    ResourceMark rm;
+    int len = java_lang_String::length(obj);
+    int plen = (len < 24 ? len : 12);
+    char* str = java_lang_String::as_utf8_string(obj, 0, plen);
+    st->print(" = \"%s\"", str);
+    if (len > plen)
+      st->print("...[%d]", len);
+  } else if (as_klassOop() == SystemDictionary::class_klass()) {
+    klassOop k = java_lang_Class::as_klassOop(obj);
+    st->print(" = ");
+    if (k != NULL) {
+      k->print_value_on(st);
+    } else {
+      const char* tname = type2name(java_lang_Class::primitive_type(obj));
+      st->print("%s", tname ? tname : "type?");
+    }
+  } else if (java_lang_boxing_object::is_instance(obj)) {
+    st->print(" = ");
+    java_lang_boxing_object::print(obj, st);
+  }
 }
 
 #endif // ndef PRODUCT
--- a/src/share/vm/oops/instanceKlass.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/instanceKlass.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -303,11 +303,30 @@
     inner_class_next_offset = 4
   };
 
+  // method override check
+  bool is_override(methodHandle super_method, Handle targetclassloader, symbolHandle targetclassname, TRAPS);
+
   // package
   bool is_same_class_package(klassOop class2);
   bool is_same_class_package(oop classloader2, symbolOop classname2);
   static bool is_same_class_package(oop class_loader1, symbolOop class_name1, oop class_loader2, symbolOop class_name2);
 
+  // find an enclosing class (defined where original code was, in jvm.cpp!)
+  klassOop compute_enclosing_class(symbolOop& simple_name_result, TRAPS) {
+    instanceKlassHandle self(THREAD, this->as_klassOop());
+    return compute_enclosing_class_impl(self, simple_name_result, THREAD);
+  }
+  static klassOop compute_enclosing_class_impl(instanceKlassHandle self,
+                                               symbolOop& simple_name_result, TRAPS);
+
+  // tell if two classes have the same enclosing class (at package level)
+  bool is_same_package_member(klassOop class2, TRAPS) {
+    instanceKlassHandle self(THREAD, this->as_klassOop());
+    return is_same_package_member_impl(self, class2, THREAD);
+  }
+  static bool is_same_package_member_impl(instanceKlassHandle self,
+                                          klassOop class2, TRAPS);
+
   // initialization state
   bool is_loaded() const                   { return _init_state >= loaded; }
   bool is_linked() const                   { return _init_state >= linked; }
--- a/src/share/vm/oops/instanceKlassKlass.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/instanceKlassKlass.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -487,6 +487,8 @@
 
 // Printing
 
+#define BULLET  " - "
+
 static const char* state_names[] = {
   "unparseable_by_gc", "allocated", "loaded", "linked", "being_initialized", "fully_initialized", "initialization_error"
 };
@@ -497,13 +499,13 @@
   instanceKlass* ik = instanceKlass::cast(klassOop(obj));
   klassKlass::oop_print_on(obj, st);
 
-  st->print(" - instance size:     %d", ik->size_helper());                        st->cr();
-  st->print(" - klass size:        %d", ik->object_size());                        st->cr();
-  st->print(" - access:            "); ik->access_flags().print_on(st);            st->cr();
-  st->print(" - state:             "); st->print_cr(state_names[ik->_init_state]);
-  st->print(" - name:              "); ik->name()->print_value_on(st);             st->cr();
-  st->print(" - super:             "); ik->super()->print_value_on(st);            st->cr();
-  st->print(" - sub:               ");
+  st->print(BULLET"instance size:     %d", ik->size_helper());                        st->cr();
+  st->print(BULLET"klass size:        %d", ik->object_size());                        st->cr();
+  st->print(BULLET"access:            "); ik->access_flags().print_on(st);            st->cr();
+  st->print(BULLET"state:             "); st->print_cr(state_names[ik->_init_state]);
+  st->print(BULLET"name:              "); ik->name()->print_value_on(st);             st->cr();
+  st->print(BULLET"super:             "); ik->super()->print_value_on(st);            st->cr();
+  st->print(BULLET"sub:               ");
   Klass* sub = ik->subklass();
   int n;
   for (n = 0; sub != NULL; n++, sub = sub->next_sibling()) {
@@ -516,12 +518,12 @@
   st->cr();
 
   if (ik->is_interface()) {
-    st->print_cr(" - nof implementors:  %d", ik->nof_implementors());
+    st->print_cr(BULLET"nof implementors:  %d", ik->nof_implementors());
     int print_impl = 0;
     for (int i = 0; i < instanceKlass::implementors_limit; i++) {
       if (ik->implementor(i) != NULL) {
         if (++print_impl == 1)
-          st->print_cr(" - implementor:    ");
+          st->print_cr(BULLET"implementor:    ");
         st->print("   ");
         ik->implementor(i)->print_value_on(st);
       }
@@ -529,34 +531,33 @@
     if (print_impl > 0)  st->cr();
   }
 
-  st->print(" - arrays:            "); ik->array_klasses()->print_value_on(st);     st->cr();
-  st->print(" - methods:           "); ik->methods()->print_value_on(st);           st->cr();
+  st->print(BULLET"arrays:            "); ik->array_klasses()->print_value_on(st);     st->cr();
+  st->print(BULLET"methods:           "); ik->methods()->print_value_on(st);           st->cr();
   if (Verbose) {
     objArrayOop methods = ik->methods();
     for(int i = 0; i < methods->length(); i++) {
       tty->print("%d : ", i); methods->obj_at(i)->print_value(); tty->cr();
     }
   }
-  st->print(" - method ordering:   "); ik->method_ordering()->print_value_on(st);       st->cr();
-  st->print(" - local interfaces:  "); ik->local_interfaces()->print_value_on(st);      st->cr();
-  st->print(" - trans. interfaces: "); ik->transitive_interfaces()->print_value_on(st); st->cr();
-  st->print(" - constants:         "); ik->constants()->print_value_on(st);         st->cr();
-  st->print(" - class loader:      "); ik->class_loader()->print_value_on(st);      st->cr();
-  st->print(" - protection domain: "); ik->protection_domain()->print_value_on(st); st->cr();
-  st->print(" - host class: ");        ik->host_klass()->print_value_on(st);        st->cr();
-  st->print(" - signers:           "); ik->signers()->print_value_on(st);           st->cr();
+  st->print(BULLET"method ordering:   "); ik->method_ordering()->print_value_on(st);       st->cr();
+  st->print(BULLET"local interfaces:  "); ik->local_interfaces()->print_value_on(st);      st->cr();
+  st->print(BULLET"trans. interfaces: "); ik->transitive_interfaces()->print_value_on(st); st->cr();
+  st->print(BULLET"constants:         "); ik->constants()->print_value_on(st);         st->cr();
+  st->print(BULLET"class loader:      "); ik->class_loader()->print_value_on(st);      st->cr();
+  st->print(BULLET"protection domain: "); ik->protection_domain()->print_value_on(st); st->cr();
+  st->print(BULLET"host class:        "); ik->host_klass()->print_value_on(st);        st->cr();
+  st->print(BULLET"signers:           "); ik->signers()->print_value_on(st);           st->cr();
   if (ik->source_file_name() != NULL) {
-    st->print(" - source file:       ");
+    st->print(BULLET"source file:       ");
     ik->source_file_name()->print_value_on(st);
     st->cr();
   }
   if (ik->source_debug_extension() != NULL) {
-    st->print(" - source debug extension:       ");
+    st->print(BULLET"source debug extension:       ");
     ik->source_debug_extension()->print_value_on(st);
     st->cr();
   }
 
-  st->print_cr(" - previous version:       ");
   {
     ResourceMark rm;
     // PreviousVersionInfo objects returned via PreviousVersionWalker
@@ -564,38 +565,43 @@
     // GrowableArray _after_ the PreviousVersionWalker destructor
     // has destroyed the handles.
     {
+      bool have_pv = false;
       PreviousVersionWalker pvw(ik);
       for (PreviousVersionInfo * pv_info = pvw.next_previous_version();
            pv_info != NULL; pv_info = pvw.next_previous_version()) {
+        if (!have_pv)
+          st->print(BULLET"previous version:  ");
+        have_pv = true;
         pv_info->prev_constant_pool_handle()()->print_value_on(st);
       }
-      st->cr();
+      if (have_pv)  st->cr();
     } // pvw is cleaned up
   } // rm is cleaned up
 
   if (ik->generic_signature() != NULL) {
-    st->print(" - generic signature:            ");
+    st->print(BULLET"generic signature: ");
     ik->generic_signature()->print_value_on(st);
+    st->cr();
   }
-  st->print(" - inner classes:     "); ik->inner_classes()->print_value_on(st);     st->cr();
-  st->print(" - java mirror:       "); ik->java_mirror()->print_value_on(st);       st->cr();
-  st->print(" - vtable length      %d  (start addr: " INTPTR_FORMAT ")", ik->vtable_length(), ik->start_of_vtable());  st->cr();
-  st->print(" - itable length      %d (start addr: " INTPTR_FORMAT ")", ik->itable_length(), ik->start_of_itable()); st->cr();
-  st->print_cr(" - static fields:");
+  st->print(BULLET"inner classes:     "); ik->inner_classes()->print_value_on(st);     st->cr();
+  st->print(BULLET"java mirror:       "); ik->java_mirror()->print_value_on(st);       st->cr();
+  st->print(BULLET"vtable length      %d  (start addr: " INTPTR_FORMAT ")", ik->vtable_length(), ik->start_of_vtable());  st->cr();
+  st->print(BULLET"itable length      %d (start addr: " INTPTR_FORMAT ")", ik->itable_length(), ik->start_of_itable()); st->cr();
+  st->print_cr(BULLET"---- static fields (%d words):", ik->static_field_size());
   FieldPrinter print_static_field(st);
   ik->do_local_static_fields(&print_static_field);
-  st->print_cr(" - non-static fields:");
-  FieldPrinter print_nonstatic_field(st, obj);
+  st->print_cr(BULLET"---- non-static fields (%d words):", ik->nonstatic_field_size());
+  FieldPrinter print_nonstatic_field(st);
   ik->do_nonstatic_fields(&print_nonstatic_field);
 
-  st->print(" - static oop maps:     ");
+  st->print(BULLET"static oop maps:     ");
   if (ik->static_oop_field_size() > 0) {
     int first_offset = ik->offset_of_static_fields();
     st->print("%d-%d", first_offset, first_offset + ik->static_oop_field_size() - 1);
   }
   st->cr();
 
-  st->print(" - non-static oop maps: ");
+  st->print(BULLET"non-static oop maps: ");
   OopMapBlock* map     = ik->start_of_nonstatic_oop_maps();
   OopMapBlock* end_map = map + ik->nonstatic_oop_map_size();
   while (map < end_map) {
--- a/src/share/vm/oops/klassVtable.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/klassVtable.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -45,9 +45,10 @@
                                                        klassOop super,
                                                        objArrayOop methods,
                                                        AccessFlags class_flags,
-                                                       oop classloader,
-                                                       symbolOop classname,
-                                                       objArrayOop local_interfaces
+                                                       Handle classloader,
+                                                       symbolHandle classname,
+                                                       objArrayOop local_interfaces,
+                                                       TRAPS
                                                        ) {
 
   No_Safepoint_Verifier nsv;
@@ -64,9 +65,9 @@
   int len = methods->length();
   for (int i = 0; i < len; i++) {
     assert(methods->obj_at(i)->is_method(), "must be a methodOop");
-    methodOop m = methodOop(methods->obj_at(i));
+    methodHandle mh(THREAD, methodOop(methods->obj_at(i)));
 
-    if (needs_new_vtable_entry(m, super, classloader, classname, class_flags)) {
+    if (needs_new_vtable_entry(mh, super, classloader, classname, class_flags, THREAD)) {
       vtable_length += vtableEntry::size(); // we need a new entry
     }
   }
@@ -117,6 +118,7 @@
     superVtable->copy_vtable_to(table());
 #ifndef PRODUCT
     if (PrintVtables && Verbose) {
+      ResourceMark rm;
       tty->print_cr("copy vtable from %s to %s size %d", sk->internal_name(), klass()->internal_name(), _length);
     }
 #endif
@@ -159,13 +161,13 @@
     int len = methods()->length();
     int initialized = super_vtable_len;
 
-    // update_super_vtable can stop for gc - ensure using handles
+    // update_inherited_vtable can stop for gc - ensure using handles
     for (int i = 0; i < len; i++) {
       HandleMark hm(THREAD);
       assert(methods()->obj_at(i)->is_method(), "must be a methodOop");
       methodHandle mh(THREAD, (methodOop)methods()->obj_at(i));
 
-      bool needs_new_entry = update_super_vtable(ik(), mh, super_vtable_len, checkconstraints, CHECK);
+      bool needs_new_entry = update_inherited_vtable(ik(), mh, super_vtable_len, checkconstraints, CHECK);
 
       if (needs_new_entry) {
         put_method_at(mh(), initialized);
@@ -177,7 +179,7 @@
     // add miranda methods; it will also update the value of initialized
     fill_in_mirandas(initialized);
 
-    // In class hierachieswhere the accesibility is not increasing (i.e., going from private ->
+    // In class hierarchies where the accessibility is not increasing (i.e., going from private ->
     // package_private -> publicprotected), the vtable might actually be smaller than our initial
     // calculation.
     assert(initialized <= _length, "vtable initialization failed");
@@ -188,26 +190,49 @@
   }
 }
 
-// Interates through the vtables to find the broadest access level. This
-// will always be monotomic for valid Java programs - but not neccesarily
-// for incompatible class files.
-klassVtable::AccessType klassVtable::vtable_accessibility_at(int i) {
-  // This vtable is not implementing the specific method
-  if (i >= length()) return acc_private;
+// Called for cases where a method does not override its superclass' vtable entry
+// For bytecodes not produced by javac together it is possible that a method does not override
+// the superclass's method, but might indirectly override a super-super class's vtable entry
+// If none found, return a null superk, else return the superk of the method this does override
+instanceKlass* klassVtable::find_transitive_override(instanceKlass* initialsuper, methodHandle target_method,
+                            int vtable_index, Handle target_loader, symbolHandle target_classname, Thread * THREAD) {
+  instanceKlass* superk = initialsuper;
+  while (superk != NULL && superk->super() != NULL) {
+    instanceKlass* supersuperklass = instanceKlass::cast(superk->super());
+    klassVtable* ssVtable = supersuperklass->vtable();
+    if (vtable_index < ssVtable->length()) {
+      methodOop super_method = ssVtable->method_at(vtable_index);
+#ifndef PRODUCT
+      symbolHandle name(THREAD,target_method()->name());
+      symbolHandle signature(THREAD,target_method()->signature());
+      assert(super_method->name() == name() && super_method->signature() == signature(), "vtable entry name/sig mismatch");
+#endif
+      if (supersuperklass->is_override(super_method, target_loader, target_classname, THREAD)) {
+#ifndef PRODUCT
+        if (PrintVtables && Verbose) {
+          ResourceMark rm(THREAD);
+          tty->print("transitive overriding superclass %s with %s::%s index %d, original flags: ",
+           supersuperklass->internal_name(),
+           _klass->internal_name(), (target_method() != NULL) ?
+           target_method()->name()->as_C_string() : "<NULL>", vtable_index);
+           super_method->access_flags().print_on(tty);
+           tty->print("overriders flags: ");
+           target_method->access_flags().print_on(tty);
+           tty->cr();
+        }
+#endif /*PRODUCT*/
+        break; // return found superk
+      }
+    } else  {
+      // super class has no vtable entry here, stop transitive search
+      superk = (instanceKlass*)NULL;
+      break;
+    }
+    // if no override found yet, continue to search up
+    superk = instanceKlass::cast(superk->super());
+  }
 
-  // Compute AccessType for current method. public or protected we are done.
-  methodOop m = method_at(i);
-  if (m->is_protected() || m->is_public()) return acc_publicprotected;
-
-  AccessType acc = m->is_package_private() ? acc_package_private : acc_private;
-
-  // Compute AccessType for method in super classes
-  klassOop super = klass()->super();
-  AccessType super_acc = (super != NULL) ? instanceKlass::cast(klass()->super())->vtable()->vtable_accessibility_at(i)
-                                         : acc_private;
-
-  // Merge
-  return (AccessType)MAX2((int)acc, (int)super_acc);
+  return superk;
 }
 
 
@@ -215,7 +240,8 @@
 // OR return true if a new vtable entry is required
 // Only called for instanceKlass's, i.e. not for arrays
 // If that changed, could not use _klass as handle for klass
-bool klassVtable::update_super_vtable(instanceKlass* klass, methodHandle target_method, int super_vtable_len, bool checkconstraints, TRAPS) {
+bool klassVtable::update_inherited_vtable(instanceKlass* klass, methodHandle target_method, int super_vtable_len,
+                  bool checkconstraints, TRAPS) {
   ResourceMark rm;
   bool allocate_new = true;
   assert(klass->oop_is_instance(), "must be instanceKlass");
@@ -242,58 +268,35 @@
   }
 
   // private methods always have a new entry in the vtable
+  // specification interpretation since classic has
+  // private methods not overriding
   if (target_method()->is_private()) {
     return allocate_new;
   }
 
   // search through the vtable and update overridden entries
   // Since check_signature_loaders acquires SystemDictionary_lock
-  // which can block for gc, once we are in this loop, use handles, not
-  // unhandled oops unless they are reinitialized for each loop
-  // handles for name, signature, klass, target_method
-  // not for match_method, holder
+  // which can block for gc, once we are in this loop, use handles
+  // For classfiles built with >= jdk7, we now look for transitive overrides
 
   symbolHandle name(THREAD,target_method()->name());
   symbolHandle signature(THREAD,target_method()->signature());
+  Handle target_loader(THREAD, _klass->class_loader());
+  symbolHandle target_classname(THREAD, _klass->name());
   for(int i = 0; i < super_vtable_len; i++) {
-    methodOop match_method = method_at(i);
+    methodOop super_method = method_at(i);
     // Check if method name matches
-    if (match_method->name() == name() && match_method->signature() == signature()) {
-
-      instanceKlass* holder = (THREAD, instanceKlass::cast(match_method->method_holder()));
-
-      // Check if the match_method is accessable from current class
-
-      bool same_package_init = false;
-      bool same_package_flag = false;
-      bool simple_match = match_method->is_public()  || match_method->is_protected();
-      if (!simple_match) {
-        same_package_init = true;
-        same_package_flag = holder->is_same_class_package(_klass->class_loader(), _klass->name());
+    if (super_method->name() == name() && super_method->signature() == signature()) {
 
-        simple_match = match_method->is_package_private() && same_package_flag;
-      }
-      // match_method is the superclass' method. Note we can't override
-      // and shouldn't access superclass' ACC_PRIVATE methods
-      // (although they have been copied into our vtable)
-      // A simple form of this statement is:
-      // if ( (match_method->is_public()  || match_method->is_protected()) ||
-      //    (match_method->is_package_private() && holder->is_same_class_package(klass->class_loader(), klass->name()))) {
-      //
-      // The complexity is introduced it avoid recomputing 'is_same_class_package' which is expensive.
-      if (simple_match) {
-        // Check if target_method and match_method has same level of accessibility. The accesibility of the
-        // match method is the "most-general" visibility of all entries at it's particular vtable index for
-        // all superclasses. This check must be done before we override the current entry in the vtable.
-        AccessType at = vtable_accessibility_at(i);
-        bool same_access = false;
+      // get super_klass for method_holder for the found method
+      instanceKlass* super_klass =  instanceKlass::cast(super_method->method_holder());
 
-        if (  (at == acc_publicprotected && (target_method()->is_public() || target_method()->is_protected())
-           || (at == acc_package_private && (target_method()->is_package_private() &&
-                                            (( same_package_init && same_package_flag) ||
-                                             (!same_package_init && holder->is_same_class_package(_klass->class_loader(), _klass->name()))))))) {
-           same_access = true;
-        }
+      if ((super_klass->is_override(super_method, target_loader, target_classname, THREAD)) ||
+      ((klass->major_version() >= VTABLE_TRANSITIVE_OVERRIDE_VERSION)
+        && ((super_klass = find_transitive_override(super_klass, target_method, i, target_loader,
+             target_classname, THREAD)) != (instanceKlass*)NULL))) {
+        // overriding, so no new entry
+        allocate_new = false;
 
         if (checkconstraints) {
         // Override vtable entry if passes loader constraint check
@@ -302,15 +305,12 @@
         // have already made any needed loader constraints.
         // Since loader constraints are transitive, it is enough
         // to link to the first super, and we get all the others.
-          symbolHandle signature(THREAD, target_method()->signature());
-          Handle this_loader(THREAD, _klass->class_loader());
-          instanceKlassHandle super_klass(THREAD, _klass->super());
           Handle super_loader(THREAD, super_klass->class_loader());
 
-          if (this_loader() != super_loader()) {
+          if (target_loader() != super_loader()) {
             ResourceMark rm(THREAD);
             char* failed_type_name =
-              SystemDictionary::check_signature_loaders(signature, this_loader,
+              SystemDictionary::check_signature_loaders(signature, target_loader,
                                                         super_loader, true,
                                                         CHECK_(false));
             if (failed_type_name != NULL) {
@@ -320,7 +320,7 @@
                 "(instance of %s), have different Class objects for the type "
                 "%s used in the signature";
               char* sig = target_method()->name_and_sig_as_C_string();
-              const char* loader1 = SystemDictionary::loader_name(this_loader());
+              const char* loader1 = SystemDictionary::loader_name(target_loader());
               char* current = _klass->name()->as_C_string();
               const char* loader2 = SystemDictionary::loader_name(super_loader());
               size_t buflen = strlen(msg) + strlen(sig) + strlen(loader1) +
@@ -331,59 +331,46 @@
               THROW_MSG_(vmSymbols::java_lang_LinkageError(), buf, false);
             }
           }
-        }
-        put_method_at(target_method(), i);
-
+       }
 
-        if (same_access) {
-          // target and match has same accessiblity - share entry
-          allocate_new = false;
-          target_method()->set_vtable_index(i);
+        put_method_at(target_method(), i);
+        target_method()->set_vtable_index(i);
 #ifndef PRODUCT
-          if (PrintVtables && Verbose) {
-            AccessType targetacc;
-            if (target_method()->is_protected() ||
-                 target_method()->is_public()) {
-               targetacc =  acc_publicprotected;
-            } else {
-              targetacc = target_method()->is_package_private() ? acc_package_private : acc_private;
-            }
-            tty->print_cr("overriding with %s::%s index %d, original flags: %x overriders flags: %x",
-             _klass->internal_name(), (target_method() != NULL) ?
-             target_method()->name()->as_C_string() : "<NULL>", i,
-             at, targetacc);
-          }
+        if (PrintVtables && Verbose) {
+          tty->print("overriding with %s::%s index %d, original flags: ",
+           _klass->internal_name(), (target_method() != NULL) ?
+           target_method()->name()->as_C_string() : "<NULL>", i);
+           super_method->access_flags().print_on(tty);
+           tty->print("overriders flags: ");
+           target_method->access_flags().print_on(tty);
+           tty->cr();
+        }
 #endif /*PRODUCT*/
-        } else {
+      } else {
+        // allocate_new = true; default. We might override one entry,
+        // but not override another. Once we override one, not need new
 #ifndef PRODUCT
-          if (PrintVtables && Verbose) {
-            AccessType targetacc;
-            if (target_method()->is_protected() ||
-                 target_method()->is_public()) {
-               targetacc =  acc_publicprotected;
-            } else {
-              targetacc = target_method()->is_package_private() ? acc_package_private : acc_private;
-            }
-            tty->print_cr("override %s %s::%s at index %d, original flags: %x overriders flags: %x",
-            allocate_new ? "+ new" : "only",
-            _klass->internal_name(), (target_method() != NULL) ?
-            target_method()->name()->as_C_string() : "<NULL>", i,
-            at, targetacc);
-           }
+        if (PrintVtables && Verbose) {
+          tty->print("NOT overriding with %s::%s index %d, original flags: ",
+           _klass->internal_name(), (target_method() != NULL) ?
+           target_method()->name()->as_C_string() : "<NULL>", i);
+           super_method->access_flags().print_on(tty);
+           tty->print("overriders flags: ");
+           target_method->access_flags().print_on(tty);
+           tty->cr();
+        }
 #endif /*PRODUCT*/
-        }
       }
     }
   }
   return allocate_new;
 }
 
-
-
 void klassVtable::put_method_at(methodOop m, int index) {
   assert(m->is_oop_or_null(), "Not an oop or null");
 #ifndef PRODUCT
   if (PrintVtables && Verbose) {
+    ResourceMark rm;
     tty->print_cr("adding %s::%s at index %d", _klass->internal_name(),
       (m != NULL) ? m->name()->as_C_string() : "<NULL>", index);
   }
@@ -397,19 +384,23 @@
 // by "classloader" and "classname".
 // NOTE: The logic used here is very similar to the one used for computing
 // the vtables indices for a method. We cannot directly use that function because,
-// when the Universe is boostrapping, a super's vtable might not be initialized.
-bool klassVtable::needs_new_vtable_entry(methodOop target_method,
+// we allocate the instanceKlass at load time, and that requires that the
+// superclass has been loaded.
+// However, the vtable entries are filled in at link time, and therefore
+// the superclass' vtable may not yet have been filled in.
+bool klassVtable::needs_new_vtable_entry(methodHandle target_method,
                                          klassOop super,
-                                         oop classloader,
-                                         symbolOop classname,
-                                         AccessFlags class_flags) {
-  if ((class_flags.is_final() || target_method->is_final()) ||
+                                         Handle classloader,
+                                         symbolHandle classname,
+                                         AccessFlags class_flags,
+                                         TRAPS) {
+  if ((class_flags.is_final() || target_method()->is_final()) ||
       // a final method never needs a new entry; final methods can be statically
       // resolved and they have to be present in the vtable only if they override
       // a super's method, in which case they re-use its entry
-      (target_method->is_static()) ||
+      (target_method()->is_static()) ||
       // static methods don't need to be in vtable
-      (target_method->name() ==  vmSymbols::object_initializer_name())
+      (target_method()->name() ==  vmSymbols::object_initializer_name())
       // <init> is never called dynamically-bound
       ) {
     return false;
@@ -421,55 +412,58 @@
   }
 
   // private methods always have a new entry in the vtable
-  if (target_method->is_private()) {
+  // specification interpretation since classic has
+  // private methods not overriding
+  if (target_method()->is_private()) {
     return true;
   }
 
   // search through the super class hierarchy to see if we need
   // a new entry
-  symbolOop name = target_method->name();
-  symbolOop signature = target_method->signature();
+  ResourceMark rm;
+  symbolOop name = target_method()->name();
+  symbolOop signature = target_method()->signature();
   klassOop k = super;
-  methodOop match_method = NULL;
+  methodOop super_method = NULL;
   instanceKlass *holder = NULL;
+  methodOop recheck_method =  NULL;
   while (k != NULL) {
     // lookup through the hierarchy for a method with matching name and sign.
-    match_method = instanceKlass::cast(k)->lookup_method(name, signature);
-    if (match_method == NULL) {
+    super_method = instanceKlass::cast(k)->lookup_method(name, signature);
+    if (super_method == NULL) {
       break; // we still have to search for a matching miranda method
     }
     // get the class holding the matching method
-    holder = instanceKlass::cast(match_method->method_holder());
-
-    if (!match_method->is_static()) { // we want only instance method matches
-      if ((target_method->is_public() || target_method->is_protected()) &&
-          (match_method->is_public()  || match_method->is_protected())) {
-        // target and match are public/protected; we do not need a new entry
+    // make sure you use that class for is_override
+    instanceKlass* superk = instanceKlass::cast(super_method->method_holder());
+    // we want only instance method matches
+    // pretend private methods are not in the super vtable
+    // since we do override around them: e.g. a.m pub/b.m private/c.m pub,
+    // ignore private, c.m pub does override a.m pub
+    // For classes that were not javac'd together, we also do transitive overriding around
+    // methods that have less accessibility
+    if ((!super_method->is_static()) &&
+       (!super_method->is_private())) {
+      if (superk->is_override(super_method, classloader, classname, THREAD)) {
         return false;
-      }
-
-      if (target_method->is_package_private() &&
-          match_method->is_package_private() &&
-          holder->is_same_class_package(classloader, classname)) {
-        // target and match are P private; we do not need a new entry
-        return false;
+      // else keep looking for transitive overrides
       }
     }
 
-    k = holder->super(); // haven't found a match yet; continue to look
+    // Start with lookup result and continue to search up
+    k = superk->super(); // haven't found an override match yet; continue to look
   }
 
   // if the target method is public or protected it may have a matching
   // miranda method in the super, whose entry it should re-use.
-  if (target_method->is_public() || target_method->is_protected()) {
-    instanceKlass *sk = instanceKlass::cast(super);
-    if (sk->has_miranda_methods()) {
-      if (sk->lookup_method_in_all_interfaces(name, signature) != NULL) {
-        return false;  // found a matching miranda; we do not need a new entry
-      }
+  // Actually, to handle cases that javac would not generate, we need
+  // this check for all access permissions.
+  instanceKlass *sk = instanceKlass::cast(super);
+  if (sk->has_miranda_methods()) {
+    if (sk->lookup_method_in_all_interfaces(name, signature) != NULL) {
+      return false;  // found a matching miranda; we do not need a new entry
     }
   }
-
   return true; // found no match; we need a new entry
 }
 
@@ -884,7 +878,7 @@
                                     _klass->name()->as_C_string());
 
 
-    // Interate through all interfaces
+    // Iterate through all interfaces
     int i;
     for(i = 0; i < num_interfaces; i++) {
       itableOffsetEntry* ioe = offset_entry(i);
@@ -1012,6 +1006,7 @@
             new_method->name()->as_C_string(),
             new_method->signature()->as_C_string()));
         }
+        break;
       }
       ime++;
     }
@@ -1153,6 +1148,27 @@
   return index;
 }
 
+
+// inverse to compute_itable_index
+methodOop klassItable::method_for_itable_index(klassOop intf, int itable_index) {
+  assert(instanceKlass::cast(intf)->is_interface(), "sanity check");
+  objArrayOop methods = instanceKlass::cast(intf)->methods();
+
+  int index = itable_index;
+  // Adjust for <clinit>, which is left out of table if first method
+  if (methods->length() > 0 && ((methodOop)methods->obj_at(0))->name() == vmSymbols::class_initializer_name()) {
+    index++;
+  }
+
+  if (itable_index < 0 || index >= methods->length())
+    return NULL;                // help caller defend against bad indexes
+
+  methodOop m = (methodOop)methods->obj_at(index);
+  assert(compute_itable_index(m) == itable_index, "correct inverse");
+
+  return m;
+}
+
 void klassVtable::verify(outputStream* st, bool forced) {
   // make sure table is initialized
   if (!Universe::is_fully_initialized()) return;
--- a/src/share/vm/oops/klassVtable.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/klassVtable.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -70,8 +70,9 @@
   // conputes vtable length (in words) and the number of miranda methods
   static void compute_vtable_size_and_num_mirandas(int &vtable_length, int &num_miranda_methods,
                                                    klassOop super, objArrayOop methods,
-                                                   AccessFlags class_flags, oop classloader,
-                                                   symbolOop classname, objArrayOop local_interfaces);
+                                                   AccessFlags class_flags, Handle classloader,
+                                                   symbolHandle classname, objArrayOop local_interfaces,
+                                                   TRAPS);
 
   // RedefineClasses() API support:
   // If any entry of this vtable points to any of old_methods,
@@ -111,14 +112,16 @@
  protected:
   friend class vtableEntry;
  private:
+  enum { VTABLE_TRANSITIVE_OVERRIDE_VERSION = 51 } ;
   void copy_vtable_to(vtableEntry* start);
   int  initialize_from_super(KlassHandle super);
   int  index_of(methodOop m, int len) const; // same as index_of, but search only up to len
   void put_method_at(methodOop m, int index);
-  static bool needs_new_vtable_entry(methodOop m, klassOop super, oop classloader, symbolOop classname, AccessFlags access_flags);
-  AccessType vtable_accessibility_at(int i);
+  static bool needs_new_vtable_entry(methodHandle m, klassOop super, Handle classloader, symbolHandle classname, AccessFlags access_flags, TRAPS);
 
-  bool update_super_vtable(instanceKlass* klass, methodHandle target_method, int super_vtable_len, bool checkconstraints, TRAPS);
+  bool update_inherited_vtable(instanceKlass* klass, methodHandle target_method, int super_vtable_len, bool checkconstraints, TRAPS);
+ instanceKlass* find_transitive_override(instanceKlass* initialsuper, methodHandle target_method, int vtable_index,
+                                         Handle target_loader, symbolHandle target_classname, Thread* THREAD);
 
   // support for miranda methods
   bool is_miranda_entry_at(int i);
@@ -298,6 +301,8 @@
 
   // Resolving of method to index
   static int compute_itable_index(methodOop m);
+  // ...and back again:
+  static methodOop method_for_itable_index(klassOop klass, int itable_index);
 
   // Debugging/Statistics
   static void print_statistics() PRODUCT_RETURN;
--- a/src/share/vm/oops/methodKlass.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/methodKlass.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2007 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -247,9 +247,14 @@
   st->print_cr(" - size of params:    %d",   m->size_of_parameters());
   st->print_cr(" - method size:       %d",   m->method_size());
   st->print_cr(" - vtable index:      %d",   m->_vtable_index);
+  st->print_cr(" - i2i entry:         " INTPTR_FORMAT, m->interpreter_entry());
+  st->print_cr(" - adapter:           " INTPTR_FORMAT, m->adapter());
+  st->print_cr(" - compiled entry     " INTPTR_FORMAT, m->from_compiled_entry());
   st->print_cr(" - code size:         %d",   m->code_size());
-  st->print_cr(" - code start:        " INTPTR_FORMAT, m->code_base());
-  st->print_cr(" - code end (excl):   " INTPTR_FORMAT, m->code_base() + m->code_size());
+  if (m->code_size() != 0) {
+    st->print_cr(" - code start:        " INTPTR_FORMAT, m->code_base());
+    st->print_cr(" - code end (excl):   " INTPTR_FORMAT, m->code_base() + m->code_size());
+  }
   if (m->method_data() != NULL) {
     st->print_cr(" - method data:       " INTPTR_FORMAT, (address)m->method_data());
   }
@@ -293,6 +298,10 @@
     m->code()->print_value_on(st);
     st->cr();
   }
+  if (m->is_native()) {
+    st->print_cr(" - native function:   " INTPTR_FORMAT, m->native_function());
+    st->print_cr(" - signature handler: " INTPTR_FORMAT, m->signature_handler());
+  }
 }
 
 
--- a/src/share/vm/oops/objArrayKlass.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/objArrayKlass.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -502,12 +502,25 @@
   }
 }
 
+static int max_objArray_print_length = 4;
 
 void objArrayKlass::oop_print_value_on(oop obj, outputStream* st) {
   assert(obj->is_objArray(), "must be objArray");
+  st->print("a ");
   element_klass()->print_value_on(st);
-  st->print("a [%d] ", objArrayOop(obj)->length());
-  as_klassOop()->klass()->print_value_on(st);
+  int len = objArrayOop(obj)->length();
+  st->print("[%d] ", len);
+  obj->print_address_on(st);
+  if (PrintOopAddress || PrintMiscellaneous && (WizardMode || Verbose)) {
+    st->print("{");
+    for (int i = 0; i < len; i++) {
+      if (i > max_objArray_print_length) {
+        st->print("..."); break;
+      }
+      st->print(" "INTPTR_FORMAT, (intptr_t)(void*)objArrayOop(obj)->obj_at(i));
+    }
+    st->print(" }");
+  }
 }
 
 #endif // PRODUCT
--- a/src/share/vm/oops/oop.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/oops/oop.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -65,11 +65,7 @@
 
 void oopDesc::print_address_on(outputStream* st) const {
   if (PrintOopAddress) {
-    st->print("{");
-    if (PrintOopAddress) {
-      st->print(INTPTR_FORMAT, this);
-    }
-    st->print("}");
+    st->print("{"INTPTR_FORMAT"}", this);
   }
 }
 
--- a/src/share/vm/opto/block.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/block.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -371,6 +371,7 @@
   Block *_broot;                // Basic block of root
   uint _rpo_ctr;
   CFGLoop* _root_loop;
+  float _outer_loop_freq;       // Outmost loop frequency
 
   // Per node latency estimation, valid only during GCM
   GrowableArray<uint> _node_latency;
@@ -537,6 +538,7 @@
   void compute_loop_depth(int depth);
   void compute_freq(); // compute frequency with loop assuming head freq 1.0f
   void scale_freq();   // scale frequency by loop trip count (including outer loops)
+  float outer_loop_freq() const; // frequency of outer loop
   bool in_loop_nest(Block* b);
   float trip_count() const { return 1.0f / _exit_prob; }
   virtual bool is_loop()  { return true; }
--- a/src/share/vm/opto/bytecodeInfo.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/bytecodeInfo.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -232,6 +232,14 @@
     return "disallowed by CompilerOracle";
   }
 
+  if (UseStringCache) {
+    // Do not inline StringCache::profile() method used only at the beginning.
+    if (callee_method->name() == ciSymbol::profile_name() &&
+        callee_method->holder()->name() == ciSymbol::java_lang_StringCache()) {
+      return "profiling method";
+    }
+  }
+
   return NULL;
 }
 
--- a/src/share/vm/opto/c2_globals.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/c2_globals.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -391,7 +391,7 @@
   product(intx, EliminateAllocationArraySizeLimit, 64,                      \
           "Array size (number of elements) limit for scalar replacement")   \
                                                                             \
-  product(bool, UseOptoBiasInlining, true,                                 \
+  product(bool, UseOptoBiasInlining, true,                                  \
           "Generate biased locking code in C2 ideal graph")                 \
                                                                             \
   product(intx, ValueSearchLimit, 1000,                                     \
@@ -410,7 +410,7 @@
           "Miniumum %% of a successor (predecessor) for which block layout "\
           "a will allow a fork (join) in a single chain")                   \
                                                                             \
-  product(bool, BlockLayoutRotateLoops, false,                              \
+  product(bool, BlockLayoutRotateLoops, true,                               \
           "Allow back branches to be fall throughs in the block layour")    \
 
 C2_FLAGS(DECLARE_DEVELOPER_FLAG, DECLARE_PD_DEVELOPER_FLAG, DECLARE_PRODUCT_FLAG, DECLARE_PD_PRODUCT_FLAG, DECLARE_DIAGNOSTIC_FLAG, DECLARE_NOTPRODUCT_FLAG)
--- a/src/share/vm/opto/chaitin.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/chaitin.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -149,6 +149,9 @@
 #endif
 {
   NOT_PRODUCT( Compile::TracePhase t3("ctorChaitin", &_t_ctorChaitin, TimeCompiler); )
+
+  _high_frequency_lrg = MIN2(float(OPTO_LRG_HIGH_FREQ), _cfg._outer_loop_freq);
+
   uint i,j;
   // Build a list of basic blocks, sorted by frequency
   _blks = NEW_RESOURCE_ARRAY( Block *, _cfg._num_blocks );
--- a/src/share/vm/opto/chaitin.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/chaitin.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -338,6 +338,8 @@
 
   Block **_blks;                // Array of blocks sorted by frequency for coalescing
 
+  float _high_frequency_lrg;    // Frequency at which LRG will be spilled for debug info
+
 #ifndef PRODUCT
   bool _trace_spilling;
 #endif
@@ -360,6 +362,8 @@
 
   uint n2lidx( const Node *n ) const { return _names[n->_idx]; }
 
+  float high_frequency_lrg() const { return _high_frequency_lrg; }
+
 #ifndef PRODUCT
   bool trace_spilling() const { return _trace_spilling; }
 #endif
--- a/src/share/vm/opto/classes.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/classes.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -218,6 +218,8 @@
 macro(StoreP)
 macro(StoreN)
 macro(StrComp)
+macro(StrEquals)
+macro(StrIndexOf)
 macro(SubD)
 macro(SubF)
 macro(SubI)
--- a/src/share/vm/opto/coalesce.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/coalesce.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -473,7 +473,7 @@
         } // End of is two-adr
 
         // Insert a copy at a debug use for a lrg which has high frequency
-        if( (b->_freq < OPTO_DEBUG_SPLIT_FREQ) && n->is_MachSafePoint() ) {
+        if( b->_freq < OPTO_DEBUG_SPLIT_FREQ || b->is_uncommon(_phc._cfg._bbs) ) {
           // Walk the debug inputs to the node and check for lrg freq
           JVMState* jvms = n->jvms();
           uint debug_start = jvms ? jvms->debug_start() : 999999;
@@ -487,7 +487,7 @@
             LRG &lrg = lrgs(nidx);
 
             // If this lrg has a high frequency use/def
-            if( lrg._maxfreq >= OPTO_LRG_HIGH_FREQ ) {
+            if( lrg._maxfreq >= _phc.high_frequency_lrg() ) {
               // If the live range is also live out of this block (like it
               // would be for a fast/slow idiom), the normal spill mechanism
               // does an excellent job.  If it is not live out of this block
--- a/src/share/vm/opto/gcm.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/gcm.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -438,6 +438,12 @@
 #endif
   assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrComp),
          "String compare is only known 'load' that does not conflict with any stores");
+  assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrEquals),
+         "String equals is a 'load' that does not conflict with any stores");
+  assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_StrIndexOf),
+         "String indexOf is a 'load' that does not conflict with any stores");
+  assert(load_alias_idx || (load->is_Mach() && load->as_Mach()->ideal_Opcode() == Op_AryEq),
+         "Arrays equals is a 'load' that do not conflict with any stores");
 
   if (!C->alias_type(load_alias_idx)->is_rewritable()) {
     // It is impossible to spoil this load by putting stores before it,
@@ -1374,6 +1380,9 @@
   _root_loop->_freq = 1.0;
   _root_loop->scale_freq();
 
+  // Save outmost loop frequency for LRG frequency threshold
+  _outer_loop_freq = _root_loop->outer_loop_freq();
+
   // force paths ending at uncommon traps to be infrequent
   if (!C->do_freq_based_layout()) {
     Block_List worklist;
@@ -1898,6 +1907,7 @@
 // Do a top down traversal of loop tree (visit outer loops first.)
 void CFGLoop::scale_freq() {
   float loop_freq = _freq * trip_count();
+  _freq = loop_freq;
   for (int i = 0; i < _members.length(); i++) {
     CFGElement* s = _members.at(i);
     float block_freq = s->_freq * loop_freq;
@@ -1912,6 +1922,14 @@
   }
 }
 
+// Frequency of outer loop
+float CFGLoop::outer_loop_freq() const {
+  if (_child != NULL) {
+    return _child->_freq;
+  }
+  return _freq;
+}
+
 #ifndef PRODUCT
 //------------------------------dump_tree--------------------------------------
 void CFGLoop::dump_tree() const {
--- a/src/share/vm/opto/lcm.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/lcm.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -137,6 +137,8 @@
       if( mach->in(2) != val ) continue;
       break;                    // Found a memory op?
     case Op_StrComp:
+    case Op_StrEquals:
+    case Op_StrIndexOf:
     case Op_AryEq:
       // Not a legit memory op for implicit null check regardless of
       // embedded loads
--- a/src/share/vm/opto/library_call.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/library_call.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -136,6 +136,7 @@
   bool inline_string_compareTo();
   bool inline_string_indexOf();
   Node* string_indexOf(Node* string_object, ciTypeArray* target_array, jint offset, jint cache_i, jint md2_i);
+  bool inline_string_equals();
   Node* pop_math_arg();
   bool runtime_math(const TypeFunc* call_type, address funcAddr, const char* funcName);
   bool inline_math_native(vmIntrinsics::ID id);
@@ -261,6 +262,7 @@
     switch (id) {
     case vmIntrinsics::_indexOf:
     case vmIntrinsics::_compareTo:
+    case vmIntrinsics::_equals:
     case vmIntrinsics::_equalsC:
       break;  // InlineNatives does not control String.compareTo
     default:
@@ -275,6 +277,9 @@
   case vmIntrinsics::_indexOf:
     if (!SpecialStringIndexOf)  return NULL;
     break;
+  case vmIntrinsics::_equals:
+    if (!SpecialStringEquals)  return NULL;
+    break;
   case vmIntrinsics::_equalsC:
     if (!SpecialArraysEquals)  return NULL;
     break;
@@ -442,6 +447,8 @@
     return inline_string_compareTo();
   case vmIntrinsics::_indexOf:
     return inline_string_indexOf();
+  case vmIntrinsics::_equals:
+    return inline_string_equals();
 
   case vmIntrinsics::_getObject:
     return inline_unsafe_access(!is_native_ptr, !is_store, T_OBJECT, false);
@@ -793,6 +800,8 @@
 //------------------------------inline_string_compareTo------------------------
 bool LibraryCallKit::inline_string_compareTo() {
 
+  if (!Matcher::has_match_rule(Op_StrComp)) return false;
+
   const int value_offset = java_lang_String::value_offset_in_bytes();
   const int count_offset = java_lang_String::count_offset_in_bytes();
   const int offset_offset = java_lang_String::offset_offset_in_bytes();
@@ -830,6 +839,82 @@
   return true;
 }
 
+//------------------------------inline_string_equals------------------------
+bool LibraryCallKit::inline_string_equals() {
+
+  if (!Matcher::has_match_rule(Op_StrEquals)) return false;
+
+  const int value_offset = java_lang_String::value_offset_in_bytes();
+  const int count_offset = java_lang_String::count_offset_in_bytes();
+  const int offset_offset = java_lang_String::offset_offset_in_bytes();
+
+  _sp += 2;
+  Node* argument = pop();  // pop non-receiver first:  it was pushed second
+  Node* receiver = pop();
+
+  // Null check on self without removing any arguments.  The argument
+  // null check technically happens in the wrong place, which can lead to
+  // invalid stack traces when string compare is inlined into a method
+  // which handles NullPointerExceptions.
+  _sp += 2;
+  receiver = do_null_check(receiver, T_OBJECT);
+  //should not do null check for argument for String.equals(), because spec
+  //allows to specify NULL as argument.
+  _sp -= 2;
+
+  if (stopped()) {
+    return true;
+  }
+
+  // get String klass for instanceOf
+  ciInstanceKlass* klass = env()->String_klass();
+
+  // two paths (plus control) merge
+  RegionNode* region = new (C, 3) RegionNode(3);
+  Node* phi = new (C, 3) PhiNode(region, TypeInt::BOOL);
+
+  Node* inst = gen_instanceof(argument, makecon(TypeKlassPtr::make(klass)));
+  Node* cmp  = _gvn.transform(new (C, 3) CmpINode(inst, intcon(1)));
+  Node* bol  = _gvn.transform(new (C, 2) BoolNode(cmp, BoolTest::eq));
+
+  IfNode* iff = create_and_map_if(control(), bol, PROB_MAX, COUNT_UNKNOWN);
+
+  Node* if_true  = _gvn.transform(new (C, 1) IfTrueNode(iff));
+  set_control(if_true);
+
+  const TypeInstPtr* string_type =
+    TypeInstPtr::make(TypePtr::BotPTR, klass, false, NULL, 0);
+
+  // instanceOf == true
+  Node* equals =
+    _gvn.transform(new (C, 7) StrEqualsNode(
+                        control(),
+                        memory(TypeAryPtr::CHARS),
+                        memory(string_type->add_offset(value_offset)),
+                        memory(string_type->add_offset(count_offset)),
+                        memory(string_type->add_offset(offset_offset)),
+                        receiver,
+                        argument));
+
+  phi->init_req(1, _gvn.transform(equals));
+  region->init_req(1, if_true);
+
+  //instanceOf == false, fallthrough
+  Node* if_false = _gvn.transform(new (C, 1) IfFalseNode(iff));
+  set_control(if_false);
+
+  phi->init_req(2, _gvn.transform(intcon(0)));
+  region->init_req(2, if_false);
+
+  // post merge
+  set_control(_gvn.transform(region));
+  record_for_igvn(region);
+
+  push(_gvn.transform(phi));
+
+  return true;
+}
+
 //------------------------------inline_array_equals----------------------------
 bool LibraryCallKit::inline_array_equals() {
 
@@ -994,80 +1079,115 @@
   return result;
 }
 
-
 //------------------------------inline_string_indexOf------------------------
 bool LibraryCallKit::inline_string_indexOf() {
 
-  _sp += 2;
-  Node *argument = pop();  // pop non-receiver first:  it was pushed second
-  Node *receiver = pop();
-
-  // don't intrinsify if argument isn't a constant string.
-  if (!argument->is_Con()) {
-    return false;
-  }
-  const TypeOopPtr* str_type = _gvn.type(argument)->isa_oopptr();
-  if (str_type == NULL) {
-    return false;
-  }
-  ciInstanceKlass* klass = env()->String_klass();
-  ciObject* str_const = str_type->const_oop();
-  if (str_const == NULL || str_const->klass() != klass) {
-    return false;
-  }
-  ciInstance* str = str_const->as_instance();
-  assert(str != NULL, "must be instance");
-
   const int value_offset  = java_lang_String::value_offset_in_bytes();
   const int count_offset  = java_lang_String::count_offset_in_bytes();
   const int offset_offset = java_lang_String::offset_offset_in_bytes();
 
-  ciObject* v = str->field_value_by_offset(value_offset).as_object();
-  int       o = str->field_value_by_offset(offset_offset).as_int();
-  int       c = str->field_value_by_offset(count_offset).as_int();
-  ciTypeArray* pat = v->as_type_array(); // pattern (argument) character array
-
-  // constant strings have no offset and count == length which
-  // simplifies the resulting code somewhat so lets optimize for that.
-  if (o != 0 || c != pat->length()) {
-    return false;
-  }
-
-  // Null check on self without removing any arguments.  The argument
-  // null check technically happens in the wrong place, which can lead to
-  // invalid stack traces when string compare is inlined into a method
-  // which handles NullPointerExceptions.
   _sp += 2;
-  receiver = do_null_check(receiver, T_OBJECT);
-  // No null check on the argument is needed since it's a constant String oop.
-  _sp -= 2;
-  if (stopped()) {
-    return true;
+  Node *argument = pop();  // pop non-receiver first:  it was pushed second
+  Node *receiver = pop();
+
+  Node* result;
+  if (Matcher::has_match_rule(Op_StrIndexOf) &&
+      UseSSE42Intrinsics) {
+    // Generate SSE4.2 version of indexOf
+    // We currently only have match rules that use SSE4.2
+
+    // Null check on self without removing any arguments.  The argument
+    // null check technically happens in the wrong place, which can lead to
+    // invalid stack traces when string compare is inlined into a method
+    // which handles NullPointerExceptions.
+    _sp += 2;
+    receiver = do_null_check(receiver, T_OBJECT);
+    argument = do_null_check(argument, T_OBJECT);
+    _sp -= 2;
+
+    if (stopped()) {
+      return true;
+    }
+
+    ciInstanceKlass* klass = env()->String_klass();
+    const TypeInstPtr* string_type =
+      TypeInstPtr::make(TypePtr::BotPTR, klass, false, NULL, 0);
+
+    result =
+      _gvn.transform(new (C, 7)
+                     StrIndexOfNode(control(),
+                                    memory(TypeAryPtr::CHARS),
+                                    memory(string_type->add_offset(value_offset)),
+                                    memory(string_type->add_offset(count_offset)),
+                                    memory(string_type->add_offset(offset_offset)),
+                                    receiver,
+                                    argument));
+  } else { //Use LibraryCallKit::string_indexOf
+    // don't intrinsify is argument isn't a constant string.
+    if (!argument->is_Con()) {
+     return false;
+    }
+    const TypeOopPtr* str_type = _gvn.type(argument)->isa_oopptr();
+    if (str_type == NULL) {
+      return false;
+    }
+    ciInstanceKlass* klass = env()->String_klass();
+    ciObject* str_const = str_type->const_oop();
+    if (str_const == NULL || str_const->klass() != klass) {
+      return false;
+    }
+    ciInstance* str = str_const->as_instance();
+    assert(str != NULL, "must be instance");
+
+    ciObject* v = str->field_value_by_offset(value_offset).as_object();
+    int       o = str->field_value_by_offset(offset_offset).as_int();
+    int       c = str->field_value_by_offset(count_offset).as_int();
+    ciTypeArray* pat = v->as_type_array(); // pattern (argument) character array
+
+    // constant strings have no offset and count == length which
+    // simplifies the resulting code somewhat so lets optimize for that.
+    if (o != 0 || c != pat->length()) {
+     return false;
+    }
+
+    // Null check on self without removing any arguments.  The argument
+    // null check technically happens in the wrong place, which can lead to
+    // invalid stack traces when string compare is inlined into a method
+    // which handles NullPointerExceptions.
+    _sp += 2;
+    receiver = do_null_check(receiver, T_OBJECT);
+    // No null check on the argument is needed since it's a constant String oop.
+    _sp -= 2;
+    if (stopped()) {
+     return true;
+    }
+
+    // The null string as a pattern always returns 0 (match at beginning of string)
+    if (c == 0) {
+      push(intcon(0));
+      return true;
+    }
+
+    // Generate default indexOf
+    jchar lastChar = pat->char_at(o + (c - 1));
+    int cache = 0;
+    int i;
+    for (i = 0; i < c - 1; i++) {
+      assert(i < pat->length(), "out of range");
+      cache |= (1 << (pat->char_at(o + i) & (sizeof(cache) * BitsPerByte - 1)));
+    }
+
+    int md2 = c;
+    for (i = 0; i < c - 1; i++) {
+      assert(i < pat->length(), "out of range");
+      if (pat->char_at(o + i) == lastChar) {
+        md2 = (c - 1) - i;
+      }
+    }
+
+    result = string_indexOf(receiver, pat, o, cache, md2);
   }
 
-  // The null string as a pattern always returns 0 (match at beginning of string)
-  if (c == 0) {
-    push(intcon(0));
-    return true;
-  }
-
-  jchar lastChar = pat->char_at(o + (c - 1));
-  int cache = 0;
-  int i;
-  for (i = 0; i < c - 1; i++) {
-    assert(i < pat->length(), "out of range");
-    cache |= (1 << (pat->char_at(o + i) & (sizeof(cache) * BitsPerByte - 1)));
-  }
-
-  int md2 = c;
-  for (i = 0; i < c - 1; i++) {
-    assert(i < pat->length(), "out of range");
-    if (pat->char_at(o + i) == lastChar) {
-      md2 = (c - 1) - i;
-    }
-  }
-
-  Node* result = string_indexOf(receiver, pat, o, cache, md2);
   push(result);
   return true;
 }
--- a/src/share/vm/opto/loopnode.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/loopnode.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -2668,6 +2668,8 @@
     case Op_LoadD_unaligned:
     case Op_LoadL_unaligned:
     case Op_StrComp:            // Does a bunch of load-like effects
+    case Op_StrEquals:
+    case Op_StrIndexOf:
     case Op_AryEq:
       pinned = false;
     }
--- a/src/share/vm/opto/machnode.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/machnode.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -340,6 +340,10 @@
   if (base == NodeSentinel)  return TypePtr::BOTTOM;
 
   const Type* t = base->bottom_type();
+  if (UseCompressedOops && Universe::narrow_oop_shift() == 0) {
+    // 32-bit unscaled narrow oop can be the base of any address expression
+    t = t->make_ptr();
+  }
   if (t->isa_intptr_t() && offset != 0 && offset != Type::OffsetBot) {
     // We cannot assert that the offset does not look oop-ish here.
     // Depending on the heap layout the cardmark base could land
@@ -353,6 +357,7 @@
 
   // be conservative if we do not recognize the type
   if (tp == NULL) {
+    assert(false, "this path may produce not optimal code");
     return TypePtr::BOTTOM;
   }
   assert(tp->base() != Type::AnyPtr, "not a bare pointer");
--- a/src/share/vm/opto/matcher.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/matcher.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -746,6 +746,8 @@
   if (nidx == Compile::AliasIdxBot && midx == Compile::AliasIdxTop) {
     switch (n->Opcode()) {
     case Op_StrComp:
+    case Op_StrEquals:
+    case Op_StrIndexOf:
     case Op_AryEq:
     case Op_MemBarVolatile:
     case Op_MemBarCPUOrder: // %%% these ideals should have narrower adr_type?
@@ -1788,6 +1790,8 @@
         mstack.push(n->in(0), Pre_Visit);     // Visit Control input
         continue;                             // while (mstack.is_nonempty())
       case Op_StrComp:
+      case Op_StrEquals:
+      case Op_StrIndexOf:
       case Op_AryEq:
         set_shared(n); // Force result into register (it will be anyways)
         break;
--- a/src/share/vm/opto/memnode.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/memnode.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -2481,6 +2481,31 @@
   return remove_dead_region(phase, can_reshape) ? this : NULL;
 }
 
+// Do we match on this edge? No memory edges
+uint StrEqualsNode::match_edge(uint idx) const {
+  return idx == 5 || idx == 6;
+}
+
+//------------------------------Ideal------------------------------------------
+// Return a node which is more "ideal" than the current node.  Strip out
+// control copies
+Node *StrEqualsNode::Ideal(PhaseGVN *phase, bool can_reshape){
+  return remove_dead_region(phase, can_reshape) ? this : NULL;
+}
+
+//=============================================================================
+// Do we match on this edge? No memory edges
+uint StrIndexOfNode::match_edge(uint idx) const {
+  return idx == 5 || idx == 6;
+}
+
+//------------------------------Ideal------------------------------------------
+// Return a node which is more "ideal" than the current node.  Strip out
+// control copies
+Node *StrIndexOfNode::Ideal(PhaseGVN *phase, bool can_reshape){
+  return remove_dead_region(phase, can_reshape) ? this : NULL;
+}
+
 //------------------------------Ideal------------------------------------------
 // Return a node which is more "ideal" than the current node.  Strip out
 // control copies
@@ -2488,7 +2513,6 @@
   return remove_dead_region(phase, can_reshape) ? this : NULL;
 }
 
-
 //=============================================================================
 MemBarNode::MemBarNode(Compile* C, int alias_idx, Node* precedent)
   : MultiNode(TypeFunc::Parms + (precedent == NULL? 0: 1)),
--- a/src/share/vm/opto/memnode.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/memnode.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -765,6 +765,54 @@
   virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
 };
 
+//------------------------------StrEquals-------------------------------------
+class StrEqualsNode: public Node {
+public:
+  StrEqualsNode(Node *control,
+                Node* char_array_mem,
+                Node* value_mem,
+                Node* count_mem,
+                Node* offset_mem,
+                Node* s1, Node* s2): Node(control,
+                                          char_array_mem,
+                                          value_mem,
+                                          count_mem,
+                                          offset_mem,
+                                          s1, s2) {};
+  virtual int Opcode() const;
+  virtual bool depends_only_on_test() const { return false; }
+  virtual const Type* bottom_type() const { return TypeInt::BOOL; }
+  // a StrEqualsNode (conservatively) aliases with everything:
+  virtual const TypePtr* adr_type() const { return TypePtr::BOTTOM; }
+  virtual uint match_edge(uint idx) const;
+  virtual uint ideal_reg() const { return Op_RegI; }
+  virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
+};
+
+//------------------------------StrIndexOf-------------------------------------
+class StrIndexOfNode: public Node {
+public:
+  StrIndexOfNode(Node *control,
+                 Node* char_array_mem,
+                 Node* value_mem,
+                 Node* count_mem,
+                 Node* offset_mem,
+                 Node* s1, Node* s2): Node(control,
+                                           char_array_mem,
+                                           value_mem,
+                                           count_mem,
+                                           offset_mem,
+                                           s1, s2) {};
+  virtual int Opcode() const;
+  virtual bool depends_only_on_test() const { return false; }
+  virtual const Type* bottom_type() const { return TypeInt::INT; }
+  // a StrIndexOfNode (conservatively) aliases with everything:
+  virtual const TypePtr* adr_type() const { return TypePtr::BOTTOM; }
+  virtual uint match_edge(uint idx) const;
+  virtual uint ideal_reg() const { return Op_RegI; }
+  virtual Node *Ideal(PhaseGVN *phase, bool can_reshape);
+};
+
 //------------------------------AryEq---------------------------------------
 class AryEqNode: public Node {
 public:
--- a/src/share/vm/opto/parse1.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/parse1.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -95,7 +95,7 @@
   switch( bt ) {                // Signature is flattened
   case T_INT:     l = new (C, 3) LoadINode( 0, mem, adr, TypeRawPtr::BOTTOM ); break;
   case T_FLOAT:   l = new (C, 3) LoadFNode( 0, mem, adr, TypeRawPtr::BOTTOM ); break;
-  case T_ADDRESS:
+  case T_ADDRESS: l = new (C, 3) LoadPNode( 0, mem, adr, TypeRawPtr::BOTTOM, TypeRawPtr::BOTTOM  ); break;
   case T_OBJECT:  l = new (C, 3) LoadPNode( 0, mem, adr, TypeRawPtr::BOTTOM, TypeInstPtr::BOTTOM ); break;
   case T_LONG:
   case T_DOUBLE: {
--- a/src/share/vm/opto/subnode.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/subnode.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -639,8 +639,8 @@
     int kps = (p0->isa_klassptr()?1:0) + (p1->isa_klassptr()?1:0);
     if (klass0 && klass1 &&
         kps != 1 &&             // both or neither are klass pointers
-        !klass0->is_interface() && // do not trust interfaces
-        !klass1->is_interface()) {
+        klass0->is_loaded() && !klass0->is_interface() && // do not trust interfaces
+        klass1->is_loaded() && !klass1->is_interface()) {
       bool unrelated_classes = false;
       // See if neither subclasses the other, or if the class on top
       // is precise.  In either of these cases, the compare is known
--- a/src/share/vm/opto/superword.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/superword.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -454,9 +454,13 @@
           // or need to run igvn.optimize() again before SLP
         } else if (out->is_Phi() && out->bottom_type() == Type::MEMORY && !in_bb(out)) {
           // Ditto.  Not sure what else to check further.
-        } else if (out->Opcode() == Op_StoreCM && out->in(4) == n) {
+        } else if (out->Opcode() == Op_StoreCM && out->in(MemNode::OopStore) == n) {
           // StoreCM has an input edge used as a precedence edge.
           // Maybe an issue when oop stores are vectorized.
+        } else if( out->is_MergeMem() && prev &&
+                   prev->Opcode() == Op_StoreCM && out == prev->in(MemNode::OopStore)) {
+          // Oop store is a MergeMem! This should not happen. Temporarily remove the assertion
+          // for this case because it could not be superwordized anyway.
         } else {
           assert(out == prev || prev == NULL, "no branches off of store slice");
         }
@@ -912,54 +916,175 @@
   }
 }
 
-//------------------------------co_locate_pack---------------------------
-// Within a pack, move stores down to the last executed store,
-// and move loads up to the first executed load.
+//-------------------------------remove_and_insert-------------------
+//remove "current" from its current position in the memory graph and insert
+//it after the appropriate insertion point (lip or uip)
+void SuperWord::remove_and_insert(MemNode *current, MemNode *prev, MemNode *lip,
+                                  Node *uip, Unique_Node_List &sched_before) {
+  Node* my_mem = current->in(MemNode::Memory);
+  _igvn.hash_delete(current);
+  _igvn.hash_delete(my_mem);
+
+  //remove current_store from its current position in the memmory graph
+  for (DUIterator i = current->outs(); current->has_out(i); i++) {
+    Node* use = current->out(i);
+    if (use->is_Mem()) {
+      assert(use->in(MemNode::Memory) == current, "must be");
+      _igvn.hash_delete(use);
+      if (use == prev) { // connect prev to my_mem
+        use->set_req(MemNode::Memory, my_mem);
+      } else if (sched_before.member(use)) {
+        _igvn.hash_delete(uip);
+        use->set_req(MemNode::Memory, uip);
+      } else {
+        _igvn.hash_delete(lip);
+        use->set_req(MemNode::Memory, lip);
+      }
+      _igvn._worklist.push(use);
+      --i; //deleted this edge; rescan position
+    }
+  }
+
+  bool sched_up = sched_before.member(current);
+  Node *insert_pt =  sched_up ?  uip : lip;
+  _igvn.hash_delete(insert_pt);
+
+  // all uses of insert_pt's memory state should use current's instead
+  for (DUIterator i = insert_pt->outs(); insert_pt->has_out(i); i++) {
+    Node* use = insert_pt->out(i);
+    if (use->is_Mem()) {
+      assert(use->in(MemNode::Memory) == insert_pt, "must be");
+      _igvn.hash_delete(use);
+      use->set_req(MemNode::Memory, current);
+      _igvn._worklist.push(use);
+      --i; //deleted this edge; rescan position
+    } else if (!sched_up && use->is_Phi() && use->bottom_type() == Type::MEMORY) {
+      uint pos; //lip (lower insert point) must be the last one in the memory slice
+      _igvn.hash_delete(use);
+      for (pos=1; pos < use->req(); pos++) {
+        if (use->in(pos) == insert_pt) break;
+      }
+      use->set_req(pos, current);
+      _igvn._worklist.push(use);
+      --i;
+    }
+  }
+
+  //connect current to insert_pt
+  current->set_req(MemNode::Memory, insert_pt);
+  _igvn._worklist.push(current);
+}
+
+//------------------------------co_locate_pack----------------------------------
+// To schedule a store pack, we need to move any sandwiched memory ops either before
+// or after the pack, based upon dependence information:
+// (1) If any store in the pack depends on the sandwiched memory op, the
+//     sandwiched memory op must be scheduled BEFORE the pack;
+// (2) If a sandwiched memory op depends on any store in the pack, the
+//     sandwiched memory op must be scheduled AFTER the pack;
+// (3) If a sandwiched memory op (say, memA) depends on another sandwiched
+//     memory op (say memB), memB must be scheduled before memA. So, if memA is
+//     scheduled before the pack, memB must also be scheduled before the pack;
+// (4) If there is no dependence restriction for a sandwiched memory op, we simply
+//     schedule this store AFTER the pack
+// (5) We know there is no dependence cycle, so there in no other case;
+// (6) Finally, all memory ops in another single pack should be moved in the same direction.
+//
+// To schedule a load pack: the memory edge of every loads in the pack must be
+// the same as the memory edge of the last executed load in the pack
 void SuperWord::co_locate_pack(Node_List* pk) {
   if (pk->at(0)->is_Store()) {
-    // Push Stores down towards last executed pack member
     MemNode* first     = executed_first(pk)->as_Mem();
     MemNode* last      = executed_last(pk)->as_Mem();
-    MemNode* insert_pt = last;
+    Unique_Node_List schedule_before_pack;
+    Unique_Node_List memops;
+
     MemNode* current   = last->in(MemNode::Memory)->as_Mem();
+    MemNode* previous  = last;
     while (true) {
       assert(in_bb(current), "stay in block");
+      memops.push(previous);
+      for (DUIterator i = current->outs(); current->has_out(i); i++) {
+        Node* use = current->out(i);
+        if (use->is_Mem() && use != previous)
+          memops.push(use);
+      }
+      if(current == first) break;
+      previous = current;
+      current  = current->in(MemNode::Memory)->as_Mem();
+    }
+
+    // determine which memory operations should be scheduled before the pack
+    for (uint i = 1; i < memops.size(); i++) {
+      Node *s1 = memops.at(i);
+      if (!in_pack(s1, pk) && !schedule_before_pack.member(s1)) {
+        for (uint j = 0; j< i; j++) {
+          Node *s2 = memops.at(j);
+          if (!independent(s1, s2)) {
+            if (in_pack(s2, pk) || schedule_before_pack.member(s2)) {
+              schedule_before_pack.push(s1); //s1 must be scheduled before
+              Node_List* mem_pk = my_pack(s1);
+              if (mem_pk != NULL) {
+                for (uint ii = 0; ii < mem_pk->size(); ii++) {
+                  Node* s = mem_pk->at(ii); // follow partner
+                  if (memops.member(s) && !schedule_before_pack.member(s))
+                    schedule_before_pack.push(s);
+                }
+              }
+            }
+          }
+        }
+      }
+    }
+
+    MemNode* lower_insert_pt = last;
+    Node*    upper_insert_pt = first->in(MemNode::Memory);
+    previous                 = last; //previous store in pk
+    current                  = last->in(MemNode::Memory)->as_Mem();
+
+    //start scheduling from "last" to "first"
+    while (true) {
+      assert(in_bb(current), "stay in block");
+      assert(in_pack(previous, pk), "previous stays in pack");
       Node* my_mem = current->in(MemNode::Memory);
+
       if (in_pack(current, pk)) {
-        // Forward users of my memory state to my input memory state
+        // Forward users of my memory state (except "previous) to my input memory state
         _igvn.hash_delete(current);
-        _igvn.hash_delete(my_mem);
         for (DUIterator i = current->outs(); current->has_out(i); i++) {
           Node* use = current->out(i);
-          if (use->is_Mem()) {
+          if (use->is_Mem() && use != previous) {
             assert(use->in(MemNode::Memory) == current, "must be");
             _igvn.hash_delete(use);
-            use->set_req(MemNode::Memory, my_mem);
+            if (schedule_before_pack.member(use)) {
+              _igvn.hash_delete(upper_insert_pt);
+              use->set_req(MemNode::Memory, upper_insert_pt);
+            } else {
+              _igvn.hash_delete(lower_insert_pt);
+              use->set_req(MemNode::Memory, lower_insert_pt);
+            }
             _igvn._worklist.push(use);
             --i; // deleted this edge; rescan position
           }
         }
-        // put current immediately before insert_pt
-        current->set_req(MemNode::Memory, insert_pt->in(MemNode::Memory));
-        _igvn.hash_delete(insert_pt);
-        insert_pt->set_req(MemNode::Memory, current);
-        _igvn._worklist.push(insert_pt);
-        _igvn._worklist.push(current);
-        insert_pt = current;
+        previous = current;
+      } else { // !in_pack(current, pk) ==> a sandwiched store
+        remove_and_insert(current, previous, lower_insert_pt, upper_insert_pt, schedule_before_pack);
       }
+
       if (current == first) break;
       current = my_mem->as_Mem();
-    }
-  } else if (pk->at(0)->is_Load()) {
-    // Pull Loads up towards first executed pack member
-    LoadNode* first = executed_first(pk)->as_Load();
-    Node* first_mem = first->in(MemNode::Memory);
-    _igvn.hash_delete(first_mem);
-    // Give each load same memory state as first
+    } // end while
+  } else if (pk->at(0)->is_Load()) { //load
+    // all use the memory state that the last executed load uses
+    LoadNode* last_load  = executed_last(pk)->as_Load();
+    Node* last_mem       = last_load->in(MemNode::Memory);
+    _igvn.hash_delete(last_mem);
+    // Give each load same memory state as last
     for (uint i = 0; i < pk->size(); i++) {
       LoadNode* ld = pk->at(i)->as_Load();
       _igvn.hash_delete(ld);
-      ld->set_req(MemNode::Memory, first_mem);
+      ld->set_req(MemNode::Memory, last_mem);
       _igvn._worklist.push(ld);
     }
   }
--- a/src/share/vm/opto/superword.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/opto/superword.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -341,8 +341,11 @@
   void filter_packs();
   // Adjust the memory graph for the packed operations
   void schedule();
-  // Within a pack, move stores down to the last executed store,
-  // and move loads up to the first executed load.
+  // Remove "current" from its current position in the memory graph and insert
+  // it after the appropriate insert points (lip or uip);
+  void remove_and_insert(MemNode *current, MemNode *prev, MemNode *lip, Node *uip, Unique_Node_List &schd_before);
+  // Within a store pack, schedule stores together by moving out the sandwiched memory ops according
+  // to dependence info; and within a load pack, move loads down to the last executed load.
   void co_locate_pack(Node_List* p);
   // Convert packs into vector node operations
   void output();
--- a/src/share/vm/prims/jni.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/prims/jni.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -301,6 +301,10 @@
   klassOop k = SystemDictionary::resolve_from_stream(class_name, class_loader,
                                                      Handle(), &st, CHECK_NULL);
 
+  if (TraceClassResolution && k != NULL) {
+    trace_class_resolution(k);
+  }
+
   cls = (jclass)JNIHandles::make_local(
     env, Klass::cast(k)->java_mirror());
   return cls;
@@ -365,6 +369,10 @@
   result = find_class_from_class_loader(env, sym, true, loader,
                                         protection_domain, true, thread);
 
+  if (TraceClassResolution && result != NULL) {
+    trace_class_resolution(java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(result)));
+  }
+
   // If we were the first invocation of jni_FindClass, we enable compilation again
   // rather than just allowing invocation counter to overflow and decay.
   // Controlled by flag DelayCompilationDuringStartup.
@@ -2646,7 +2654,12 @@
   Handle protection_domain; // null protection domain
 
   symbolHandle sym = oopFactory::new_symbol_handle(name, CHECK_NULL);
-  return find_class_from_class_loader(env, sym, true, loader, protection_domain, true, CHECK_NULL);
+  jclass result =  find_class_from_class_loader(env, sym, true, loader, protection_domain, true, CHECK_NULL);
+
+  if (TraceClassResolution && result != NULL) {
+    trace_class_resolution(java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(result)));
+  }
+  return result;
 }
 
 // These lookups are done with the NULL (bootstrap) ClassLoader to
--- a/src/share/vm/prims/jvm.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/prims/jvm.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -64,6 +64,7 @@
   ResourceMark rm;
   int line_number = -1;
   const char * source_file = NULL;
+  const char * trace = "explicit";
   klassOop caller = NULL;
   JavaThread* jthread = JavaThread::current();
   if (jthread->has_last_Java_frame()) {
@@ -107,12 +108,21 @@
                (last_caller->name() == vmSymbols::loadClassInternal_name() ||
                 last_caller->name() == vmSymbols::loadClass_name())) {
       found_it = true;
+    } else if (!vfst.at_end()) {
+      if (vfst.method()->is_native()) {
+        // JNI call
+        found_it = true;
+      }
     }
     if (found_it && !vfst.at_end()) {
       // found the caller
       caller = vfst.method()->method_holder();
       line_number = vfst.method()->line_number_from_bci(vfst.bci());
-      symbolOop s = instanceKlass::cast(vfst.method()->method_holder())->source_file_name();
+      if (line_number == -1) {
+        // show method name if it's a native method
+        trace = vfst.method()->name_and_sig_as_C_string();
+      }
+      symbolOop s = instanceKlass::cast(caller)->source_file_name();
       if (s != NULL) {
         source_file = s->as_C_string();
       }
@@ -124,15 +134,15 @@
       const char * to = Klass::cast(to_class)->external_name();
       // print in a single call to reduce interleaving between threads
       if (source_file != NULL) {
-        tty->print("RESOLVE %s %s %s:%d (explicit)\n", from, to, source_file, line_number);
+        tty->print("RESOLVE %s %s %s:%d (%s)\n", from, to, source_file, line_number, trace);
       } else {
-        tty->print("RESOLVE %s %s (explicit)\n", from, to);
+        tty->print("RESOLVE %s %s (%s)\n", from, to, trace);
       }
     }
   }
 }
 
-static void trace_class_resolution(klassOop to_class) {
+void trace_class_resolution(klassOop to_class) {
   EXCEPTION_MARK;
   trace_class_resolution_impl(to_class, THREAD);
   if (HAS_PENDING_EXCEPTION) {
@@ -1242,7 +1252,7 @@
 
            // Throws an exception if outer klass has not declared k as
            // an inner klass
-           Reflection::check_for_inner_class(k, inner_klass, CHECK_NULL);
+           Reflection::check_for_inner_class(k, inner_klass, true, CHECK_NULL);
 
            result->obj_at_put(members, inner_klass->java_mirror());
            members++;
@@ -1265,16 +1275,29 @@
 
 
 JVM_ENTRY(jclass, JVM_GetDeclaringClass(JNIEnv *env, jclass ofClass))
-  const int inner_class_info_index = 0;
-  const int outer_class_info_index = 1;
-
+{
   // ofClass is a reference to a java_lang_Class object.
   if (java_lang_Class::is_primitive(JNIHandles::resolve_non_null(ofClass)) ||
       ! Klass::cast(java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(ofClass)))->oop_is_instance()) {
     return NULL;
   }
 
-  instanceKlassHandle k(thread, java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(ofClass)));
+  symbolOop simple_name = NULL;
+  klassOop outer_klass
+    = instanceKlass::cast(java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(ofClass))
+                          )->compute_enclosing_class(simple_name, CHECK_NULL);
+  if (outer_klass == NULL)  return NULL;  // already a top-level class
+  if (simple_name == NULL)  return NULL;  // an anonymous class (inside a method)
+  return (jclass) JNIHandles::make_local(env, Klass::cast(outer_klass)->java_mirror());
+}
+JVM_END
+
+// should be in instanceKlass.cpp, but is here for historical reasons
+klassOop instanceKlass::compute_enclosing_class_impl(instanceKlassHandle k,
+                                                     symbolOop& simple_name_result, TRAPS) {
+  Thread* thread = THREAD;
+  const int inner_class_info_index = inner_class_inner_class_info_offset;
+  const int outer_class_info_index = inner_class_outer_class_info_offset;
 
   if (k->inner_classes()->length() == 0) {
     // No inner class info => no declaring class
@@ -1288,35 +1311,51 @@
   bool found = false;
   klassOop ok;
   instanceKlassHandle outer_klass;
+  bool inner_is_member = false;
+  int simple_name_index = 0;
 
   // Find inner_klass attribute
-  for(int i = 0; i < i_length && !found; i+= 4) {
+  for (int i = 0; i < i_length && !found; i += inner_class_next_offset) {
     int ioff = i_icls->ushort_at(i + inner_class_info_index);
     int ooff = i_icls->ushort_at(i + outer_class_info_index);
-
-    if (ioff != 0 && ooff != 0) {
+    int noff = i_icls->ushort_at(i + inner_class_inner_name_offset);
+    if (ioff != 0) {
       // Check to see if the name matches the class we're looking for
       // before attempting to find the class.
       if (i_cp->klass_name_at_matches(k, ioff)) {
         klassOop inner_klass = i_cp->klass_at(ioff, CHECK_NULL);
-        if (k() == inner_klass) {
-          found = true;
+        found = (k() == inner_klass);
+        if (found && ooff != 0) {
           ok = i_cp->klass_at(ooff, CHECK_NULL);
           outer_klass = instanceKlassHandle(thread, ok);
+          simple_name_index = noff;
+          inner_is_member = true;
         }
       }
     }
   }
 
+  if (found && outer_klass.is_null()) {
+    // It may be anonymous; try for that.
+    int encl_method_class_idx = k->enclosing_method_class_index();
+    if (encl_method_class_idx != 0) {
+      ok = i_cp->klass_at(encl_method_class_idx, CHECK_NULL);
+      outer_klass = instanceKlassHandle(thread, ok);
+      inner_is_member = false;
+    }
+  }
+
   // If no inner class attribute found for this class.
-  if (!found) return NULL;
+  if (outer_klass.is_null())  return NULL;
 
   // Throws an exception if outer klass has not declared k as an inner klass
-  Reflection::check_for_inner_class(outer_klass, k, CHECK_NULL);
-
-  return (jclass)JNIHandles::make_local(env, outer_klass->java_mirror());
-JVM_END
-
+  // We need evidence that each klass knows about the other, or else
+  // the system could allow a spoof of an inner class to gain access rights.
+  Reflection::check_for_inner_class(outer_klass, k, inner_is_member, CHECK_NULL);
+
+  simple_name_result = (inner_is_member ? i_cp->symbol_at(simple_name_index) : symbolOop(NULL));
+  return outer_klass();
+}
 
 JVM_ENTRY(jstring, JVM_GetClassSignature(JNIEnv *env, jclass cls))
   assert (cls != NULL, "illegal class");
@@ -3213,8 +3252,12 @@
   }
   Handle h_loader(THREAD, loader);
   Handle h_prot  (THREAD, protection_domain);
-  return find_class_from_class_loader(env, name, true, h_loader, h_prot,
-                                      false, thread);
+  jclass result =  find_class_from_class_loader(env, name, true, h_loader, h_prot,
+                                                false, thread);
+  if (TraceClassResolution && result != NULL) {
+    trace_class_resolution(java_lang_Class::as_klassOop(JNIHandles::resolve_non_null(result)));
+  }
+  return result;
 JVM_END
 
 
--- a/src/share/vm/prims/jvm_misc.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/prims/jvm_misc.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -27,6 +27,7 @@
 
 jclass find_class_from_class_loader(JNIEnv* env, symbolHandle name, jboolean init, Handle loader, Handle protection_domain, jboolean throwError, TRAPS);
 
+void trace_class_resolution(klassOop to_class);
 
 /*
  * Support for Serialization and RMI. Currently used by HotSpot only.
--- a/src/share/vm/runtime/arguments.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/arguments.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -852,16 +852,13 @@
       FreeHeap(value);
     }
     return true;
-  }
-  else if (strcmp(key, "sun.java.command") == 0) {
-
+  } else if (strcmp(key, "sun.java.command") == 0) {
     _java_command = value;
 
     // don't add this property to the properties exposed to the java application
     FreeHeap(key);
     return true;
-  }
-  else if (strcmp(key, "sun.java.launcher.pid") == 0) {
+  } else if (strcmp(key, "sun.java.launcher.pid") == 0) {
     // launcher.pid property is private and is processed
     // in process_sun_java_launcher_properties();
     // the sun.java.launcher property is passed on to the java application
@@ -870,13 +867,14 @@
       FreeHeap(value);
     }
     return true;
-  }
-  else if (strcmp(key, "java.vendor.url.bug") == 0) {
+  } else if (strcmp(key, "java.vendor.url.bug") == 0) {
     // save it in _java_vendor_url_bug, so JVM fatal error handler can access
     // its value without going through the property list or making a Java call.
     _java_vendor_url_bug = value;
+  } else if (strcmp(key, "sun.boot.library.path") == 0) {
+    PropertyList_unique_add(&_system_properties, key, value, true);
+    return true;
   }
-
   // Create new property and add at the end of the list
   PropertyList_unique_add(&_system_properties, key, value);
   return true;
@@ -895,7 +893,7 @@
   // Ensure Agent_OnLoad has the correct initial values.
   // This may not be the final mode; mode may change later in onload phase.
   PropertyList_unique_add(&_system_properties, "java.vm.info",
-     (char*)Abstract_VM_Version::vm_info_string());
+                          (char*)Abstract_VM_Version::vm_info_string(), false);
 
   UseInterpreter             = true;
   UseCompiler                = true;
@@ -1376,9 +1374,6 @@
   if (AggressiveOpts && FLAG_IS_DEFAULT(DoEscapeAnalysis)) {
     FLAG_SET_DEFAULT(DoEscapeAnalysis, true);
   }
-  if (AggressiveOpts && FLAG_IS_DEFAULT(SpecialArraysEquals)) {
-    FLAG_SET_DEFAULT(SpecialArraysEquals, true);
-  }
   if (AggressiveOpts && FLAG_IS_DEFAULT(BiasedLockingStartupDelay)) {
     FLAG_SET_DEFAULT(BiasedLockingStartupDelay, 500);
   }
@@ -2777,7 +2772,7 @@
 }
 
 // This add maintains unique property key in the list.
-void Arguments::PropertyList_unique_add(SystemProperty** plist, const char* k, char* v) {
+void Arguments::PropertyList_unique_add(SystemProperty** plist, const char* k, char* v, jboolean append) {
   if (plist == NULL)
     return;
 
@@ -2785,7 +2780,11 @@
   SystemProperty* prop;
   for (prop = *plist; prop != NULL; prop = prop->next()) {
     if (strcmp(k, prop->key()) == 0) {
-      prop->set_value(v);
+      if (append) {
+        prop->append_value(v);
+      } else {
+        prop->set_value(v);
+      }
       return;
     }
   }
--- a/src/share/vm/runtime/arguments.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/arguments.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -475,10 +475,13 @@
   // System properties
   static void init_system_properties();
 
-  // Proptery List manipulation
+  // Property List manipulation
   static void PropertyList_add(SystemProperty** plist, SystemProperty *element);
   static void PropertyList_add(SystemProperty** plist, const char* k, char* v);
-  static void PropertyList_unique_add(SystemProperty** plist, const char* k, char* v);
+  static void PropertyList_unique_add(SystemProperty** plist, const char* k, char* v) {
+    PropertyList_unique_add(plist, k, v, false);
+  }
+  static void PropertyList_unique_add(SystemProperty** plist, const char* k, char* v, jboolean append);
   static const char* PropertyList_get_value(SystemProperty* plist, const char* key);
   static int  PropertyList_count(SystemProperty* pl);
   static const char* PropertyList_get_key_at(SystemProperty* pl,int index);
--- a/src/share/vm/runtime/fieldDescriptor.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/fieldDescriptor.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2005 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -107,13 +107,14 @@
 void fieldDescriptor::print_on_for(outputStream* st, oop obj) {
   print_on(st);
   BasicType ft = field_type();
-  jint as_int;
+  jint as_int = 0;
   switch (ft) {
     case T_BYTE:
       as_int = (jint)obj->byte_field(offset());
       st->print(" %d", obj->byte_field(offset()));
       break;
     case T_CHAR:
+      as_int = (jint)obj->char_field(offset());
       {
         jchar c = obj->char_field(offset());
         as_int = c;
@@ -128,6 +129,7 @@
       st->print(" %f", obj->float_field(offset()));
       break;
     case T_INT:
+      as_int = obj->int_field(offset());
       st->print(" %d", obj->int_field(offset()));
       break;
     case T_LONG:
@@ -144,12 +146,12 @@
       break;
     case T_ARRAY:
       st->print(" ");
-      as_int = obj->int_field(offset());
+      NOT_LP64(as_int = obj->int_field(offset()));
       obj->obj_field(offset())->print_value_on(st);
       break;
     case T_OBJECT:
       st->print(" ");
-      as_int = obj->int_field(offset());
+      NOT_LP64(as_int = obj->int_field(offset()));
       obj->obj_field(offset())->print_value_on(st);
       break;
     default:
@@ -158,9 +160,9 @@
   }
   // Print a hint as to the underlying integer representation. This can be wrong for
   // pointers on an LP64 machine
-  if (ft == T_LONG || ft == T_DOUBLE) {
+  if (ft == T_LONG || ft == T_DOUBLE LP64_ONLY(|| !is_java_primitive(ft)) ) {
     st->print(" (%x %x)", obj->int_field(offset()), obj->int_field(offset()+sizeof(jint)));
-  } else {
+  } else if (as_int < 0 || as_int > 9) {
     st->print(" (%x)", as_int);
   }
 }
--- a/src/share/vm/runtime/globals.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/globals.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -47,6 +47,7 @@
 define_pd_global(intx, OnStackReplacePercentage,     0);
 define_pd_global(bool, ResizeTLAB,                   false);
 define_pd_global(intx, FreqInlineSize,               0);
+define_pd_global(intx, InlineSmallCode,              0);
 define_pd_global(intx, NewSizeThreadIncrease,        4*K);
 define_pd_global(intx, NewRatio,                     4);
 define_pd_global(intx, InlineClassNatives,           true);
@@ -490,9 +491,15 @@
   develop(bool, SpecialStringIndexOf, true,                                 \
           "special version of string indexOf")                              \
                                                                             \
-  product(bool, SpecialArraysEquals, false,                                 \
+  develop(bool, SpecialStringEquals, true,                                  \
+          "special version of string equals")                               \
+                                                                            \
+  develop(bool, SpecialArraysEquals, true,                                  \
           "special version of Arrays.equals(char[],char[])")                \
                                                                             \
+  product(bool, UseSSE42Intrinsics, false,                                  \
+          "SSE4.2 versions of intrinsics")                                  \
+                                                                            \
   develop(bool, TraceCallFixup, false,                                      \
           "traces all call fixups")                                         \
                                                                             \
@@ -2622,7 +2629,7 @@
   develop(intx, MaxRecursiveInlineLevel, 1,                                 \
           "maximum number of nested recursive calls that are inlined")      \
                                                                             \
-  product(intx, InlineSmallCode, 1000,                                      \
+  product_pd(intx, InlineSmallCode,                                         \
           "Only inline already compiled methods if their code size is "     \
           "less than this")                                                 \
                                                                             \
--- a/src/share/vm/runtime/handles.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/handles.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2007 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -137,6 +137,14 @@
     assert(is_null() || obj()->is_klass(), "not a klassOop");
   }
 
+  // Direct interface, use very sparingly.
+  // Used by SystemDictionaryHandles to create handles on existing WKKs.
+  // The obj of such a klass handle may be null, because the handle is formed
+  // during system bootstrapping.
+  KlassHandle(klassOop *handle, bool dummy) : Handle((oop*)handle, dummy) {
+    assert(SharedSkipVerify || is_null() || obj() == NULL || obj()->is_klass(), "not a klassOop");
+  }
+
   // General access
   klassOop    operator () () const               { return obj(); }
   Klass*      operator -> () const               { return as_klass(); }
--- a/src/share/vm/runtime/hpi.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/hpi.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -90,7 +90,7 @@
   static inline struct protoent* get_proto_by_name(char* name);
 
   // HPI_LibraryInterface
-  static inline void   dll_build_name(char *buf, int buf_len, char* path,
+  static inline void   dll_build_name(char *buf, int buf_len, const char* path,
                                       const char *name);
   static inline void*  dll_load(const char *name, char *ebuf, int ebuflen);
   static inline void   dll_unload(void *lib);
@@ -137,7 +137,15 @@
     return result;                            \
   }
 
-
+#define VM_HPIDECL_VOID(name, names, func, arg_type, arg_print, arg)   \
+  inline void  hpi::name arg_type {           \
+    if (TraceHPI) {                           \
+      tty->print("hpi::" names "(");          \
+      tty->print arg_print;                   \
+      tty->print(") = ");                     \
+    }                                         \
+    func arg;                                 \
+  }
 
 #define HPIDECL_VOID(name, names, intf, func, arg_type, arg_print, arg) \
   inline void hpi::name arg_type {            \
@@ -197,11 +205,11 @@
         (fd, size));
 
 // HPI_LibraryInterface
-HPIDECL_VOID(dll_build_name, "dll_build_name", _library, BuildLibName,
-             (char *buf, int buf_len, char *path, const char *name),
-             ("buf = %p, buflen = %d, path = %s, name = %s",
-              buf, buf_len, path, name),
-             (buf, buf_len, path, name));
+VM_HPIDECL_VOID(dll_build_name, "dll_build_name", os::dll_build_name,
+               (char *buf, int buf_len, const char *path, const char *name),
+               ("buf = %p, buflen = %d, path = %s, name = %s",
+                buf, buf_len, path, name),
+               (buf, buf_len, path, name));
 
 VM_HPIDECL(dll_load, "dll_load", os::dll_load,
         void *, "(void *)%p",
--- a/src/share/vm/runtime/orderAccess.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/orderAccess.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -26,3 +26,15 @@
 # include "incls/_orderAccess.cpp.incl"
 
 volatile intptr_t OrderAccess::dummy = 0;
+
+void OrderAccess::StubRoutines_fence() {
+  // Use a stub if it exists.  It may not exist during bootstrap so do
+  // nothing in that case but assert if no fence code exists after threads have been created
+  void (*func)() = CAST_TO_FN_PTR(void (*)(), StubRoutines::fence_entry());
+
+  if (func != NULL) {
+    (*func)();
+    return;
+  }
+  assert(Threads::number_of_threads() == 0, "for bootstrap only");
+}
--- a/src/share/vm/runtime/orderAccess.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/orderAccess.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -300,4 +300,10 @@
   // In order to force a memory access, implementations may
   // need a volatile externally visible dummy variable.
   static volatile intptr_t dummy;
+
+ private:
+  // This is a helper that invokes the StubRoutines::fence_entry()
+  // routine if it exists, It should only be used by platforms that
+  // don't another way to do the inline eassembly.
+  static void StubRoutines_fence();
 };
--- a/src/share/vm/runtime/os.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/os.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -863,7 +863,6 @@
 
 
 bool os::set_boot_path(char fileSep, char pathSep) {
-
     const char* home = Arguments::get_java_home();
     int home_len = (int)strlen(home);
 
@@ -893,6 +892,60 @@
     return true;
 }
 
+/*
+ * Splits a path, based on its separator, the number of
+ * elements is returned back in n.
+ * It is the callers responsibility to:
+ *   a> check the value of n, and n may be 0.
+ *   b> ignore any empty path elements
+ *   c> free up the data.
+ */
+char** os::split_path(const char* path, int* n) {
+  *n = 0;
+  if (path == NULL || strlen(path) == 0) {
+    return NULL;
+  }
+  const char psepchar = *os::path_separator();
+  char* inpath = (char*)NEW_C_HEAP_ARRAY(char, strlen(path) + 1);
+  if (inpath == NULL) {
+    return NULL;
+  }
+  strncpy(inpath, path, strlen(path));
+  int count = 1;
+  char* p = strchr(inpath, psepchar);
+  // Get a count of elements to allocate memory
+  while (p != NULL) {
+    count++;
+    p++;
+    p = strchr(p, psepchar);
+  }
+  char** opath = (char**) NEW_C_HEAP_ARRAY(char*, count);
+  if (opath == NULL) {
+    return NULL;
+  }
+
+  // do the actual splitting
+  p = inpath;
+  for (int i = 0 ; i < count ; i++) {
+    size_t len = strcspn(p, os::path_separator());
+    if (len > JVM_MAXPATHLEN) {
+      return NULL;
+    }
+    // allocate the string and add terminator storage
+    char* s  = (char*)NEW_C_HEAP_ARRAY(char, len + 1);
+    if (s == NULL) {
+      return NULL;
+    }
+    strncpy(s, p, len);
+    s[len] = '\0';
+    opath[i] = s;
+    p += len + 1;
+  }
+  FREE_C_HEAP_ARRAY(char, inpath);
+  *n = count;
+  return opath;
+}
+
 void os::set_memory_serialize_page(address page) {
   int count = log2_intptr(sizeof(class JavaThread)) - log2_intptr(64);
   _mem_serialize_page = (volatile int32_t *)page;
--- a/src/share/vm/runtime/os.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/os.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -202,8 +202,10 @@
   static char*  attempt_reserve_memory_at(size_t bytes, char* addr);
   static void   split_reserved_memory(char *base, size_t size,
                                       size_t split, bool realloc);
-  static bool   commit_memory(char* addr, size_t bytes);
-  static bool   commit_memory(char* addr, size_t size, size_t alignment_hint);
+  static bool   commit_memory(char* addr, size_t bytes,
+                              bool executable = false);
+  static bool   commit_memory(char* addr, size_t size, size_t alignment_hint,
+                              bool executable = false);
   static bool   uncommit_memory(char* addr, size_t bytes);
   static bool   release_memory(char* addr, size_t bytes);
 
@@ -243,7 +245,8 @@
 
   static char*  non_memory_address_word();
   // reserve, commit and pin the entire memory region
-  static char*  reserve_memory_special(size_t size, char* addr = NULL);
+  static char*  reserve_memory_special(size_t size, char* addr = NULL,
+                bool executable = false);
   static bool   release_memory_special(char* addr, size_t bytes);
   static bool   large_page_init();
   static size_t large_page_size();
@@ -604,6 +607,7 @@
                                 char fileSep,
                                 char pathSep);
   static bool set_boot_path(char fileSep, char pathSep);
+  static char** split_path(const char* path, int* n);
 };
 
 // Note that "PAUSE" is almost always used with synchronization
--- a/src/share/vm/runtime/reflection.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/reflection.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -554,10 +554,18 @@
   return instanceKlass::cast(class1)->is_same_class_package(class2);
 }
 
+bool Reflection::is_same_package_member(klassOop class1, klassOop class2, TRAPS) {
+  return instanceKlass::cast(class1)->is_same_package_member(class2, THREAD);
+}
+
 
 // Checks that the 'outer' klass has declared 'inner' as being an inner klass. If not,
 // throw an incompatible class change exception
-void Reflection::check_for_inner_class(instanceKlassHandle outer, instanceKlassHandle inner, TRAPS) {
+// If inner_is_member, require the inner to be a member of the outer.
+// If !inner_is_member, require the inner to be anonymous (a non-member).
+// Caller is responsible for figuring out in advance which case must be true.
+void Reflection::check_for_inner_class(instanceKlassHandle outer, instanceKlassHandle inner,
+                                       bool inner_is_member, TRAPS) {
   const int inner_class_info_index = 0;
   const int outer_class_info_index = 1;
 
@@ -567,7 +575,7 @@
      int ioff = icls->ushort_at(i + inner_class_info_index);
      int ooff = icls->ushort_at(i + outer_class_info_index);
 
-     if (ioff != 0 && ooff != 0) {
+     if (inner_is_member && ioff != 0 && ooff != 0) {
         klassOop o = cp->klass_at(ooff, CHECK);
         if (o == outer()) {
           klassOop i = cp->klass_at(ioff, CHECK);
@@ -576,6 +584,13 @@
           }
         }
      }
+     if (!inner_is_member && ioff != 0 && ooff == 0 &&
+         cp->klass_name_at_matches(inner, ioff)) {
+        klassOop i = cp->klass_at(ioff, CHECK);
+        if (i == inner()) {
+          return;
+        }
+     }
   }
 
   // 'inner' not declared as an inner klass in outer
--- a/src/share/vm/runtime/reflection.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/reflection.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2006 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -87,12 +87,18 @@
                                       bool classloader_only,
                                       bool protected_restriction = false);
   static bool     is_same_class_package(klassOop class1, klassOop class2);
+  static bool     is_same_package_member(klassOop class1, klassOop class2, TRAPS);
 
   static bool can_relax_access_check_for(
     klassOop accessor, klassOop accesee, bool classloader_only);
 
   // inner class reflection
-  static void check_for_inner_class(instanceKlassHandle outer, instanceKlassHandle inner, TRAPS);
+  // raise an ICCE unless the required relationship can be proven to hold
+  // If inner_is_member, require the inner to be a member of the outer.
+  // If !inner_is_member, require the inner to be anonymous (a non-member).
+  // Caller is responsible for figuring out in advance which case must be true.
+  static void check_for_inner_class(instanceKlassHandle outer, instanceKlassHandle inner,
+                                    bool inner_is_member, TRAPS);
 
   //
   // Support for reflection based on dynamic bytecode generation (JDK 1.4)
--- a/src/share/vm/runtime/sharedRuntime.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/sharedRuntime.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -675,48 +675,6 @@
 JRT_END
 
 
-// ---------------------------------------------------------------------------------------------------------
-// Non-product code
-#ifndef PRODUCT
-
-void SharedRuntime::verify_caller_frame(frame caller_frame, methodHandle callee_method) {
-  ResourceMark rm;
-  assert (caller_frame.is_interpreted_frame(), "sanity check");
-  assert (callee_method->has_compiled_code(), "callee must be compiled");
-  methodHandle caller_method (Thread::current(), caller_frame.interpreter_frame_method());
-  jint bci = caller_frame.interpreter_frame_bci();
-  methodHandle method = find_callee_method_inside_interpreter(caller_frame, caller_method, bci);
-  assert (callee_method == method, "incorrect method");
-}
-
-methodHandle SharedRuntime::find_callee_method_inside_interpreter(frame caller_frame, methodHandle caller_method, int bci) {
-  EXCEPTION_MARK;
-  Bytecode_invoke* bytecode = Bytecode_invoke_at(caller_method, bci);
-  methodHandle staticCallee = bytecode->static_target(CATCH); // Non-product code
-
-  bytecode = Bytecode_invoke_at(caller_method, bci);
-  int bytecode_index = bytecode->index();
-  Bytecodes::Code bc = bytecode->adjusted_invoke_code();
-
-  Handle receiver;
-  if (bc == Bytecodes::_invokeinterface ||
-      bc == Bytecodes::_invokevirtual ||
-      bc == Bytecodes::_invokespecial) {
-    symbolHandle signature (THREAD, staticCallee->signature());
-    receiver = Handle(THREAD, retrieve_receiver(signature, caller_frame));
-  } else {
-    receiver = Handle();
-  }
-  CallInfo result;
-  constantPoolHandle constants (THREAD, caller_method->constants());
-  LinkResolver::resolve_invoke(result, receiver, constants, bytecode_index, bc, CATCH); // Non-product code
-  methodHandle calleeMethod = result.selected_method();
-  return calleeMethod;
-}
-
-#endif  // PRODUCT
-
-
 JRT_ENTRY_NO_ASYNC(void, SharedRuntime::register_finalizer(JavaThread* thread, oopDesc* obj))
   assert(obj->is_oop(), "must be a valid oop");
   assert(obj->klass()->klass_part()->has_finalizer(), "shouldn't be here otherwise");
--- a/src/share/vm/runtime/sharedRuntime.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/sharedRuntime.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -1,5 +1,5 @@
 /*
- * Copyright 1997-2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * Copyright 1997-2009 Sun Microsystems, Inc.  All Rights Reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -180,9 +180,6 @@
 
   static oop retrieve_receiver( symbolHandle sig, frame caller );
 
-  static void verify_caller_frame(frame caller_frame, methodHandle callee_method) PRODUCT_RETURN;
-  static methodHandle find_callee_method_inside_interpreter(frame caller_frame, methodHandle caller_method, int bci) PRODUCT_RETURN_(return methodHandle(););
-
   static void register_finalizer(JavaThread* thread, oopDesc* obj);
 
   // dtrace notifications
--- a/src/share/vm/runtime/thread.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/thread.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -3007,17 +3007,19 @@
       }
 
       if (UseStringCache) {
-        // Forcibly initialize java/lang/String and mutate the private
+        // Forcibly initialize java/lang/StringValue and mutate the private
         // static final "stringCacheEnabled" field before we start creating instances
-        klassOop k_o = SystemDictionary::resolve_or_null(vmSymbolHandles::java_lang_String(), Handle(), Handle(), CHECK_0);
-        KlassHandle k = KlassHandle(THREAD, k_o);
-        guarantee(k.not_null(), "Must find java/lang/String");
-        instanceKlassHandle ik = instanceKlassHandle(THREAD, k());
-        ik->initialize(CHECK_0);
-        fieldDescriptor fd;
-        // Possible we might not find this field; if so, don't break
-        if (ik->find_local_field(vmSymbols::stringCacheEnabled_name(), vmSymbols::bool_signature(), &fd)) {
-          k()->bool_field_put(fd.offset(), true);
+        klassOop k_o = SystemDictionary::resolve_or_null(vmSymbolHandles::java_lang_StringValue(), Handle(), Handle(), CHECK_0);
+        // Possible that StringValue isn't present: if so, silently don't break
+        if (k_o != NULL) {
+          KlassHandle k = KlassHandle(THREAD, k_o);
+          instanceKlassHandle ik = instanceKlassHandle(THREAD, k());
+          ik->initialize(CHECK_0);
+          fieldDescriptor fd;
+          // Possible we might not find this field: if so, silently don't break
+          if (ik->find_local_field(vmSymbols::stringCacheEnabled_name(), vmSymbols::bool_signature(), &fd)) {
+            k()->bool_field_put(fd.offset(), true);
+          }
         }
       }
     }
--- a/src/share/vm/runtime/virtualspace.cpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/virtualspace.cpp	Fri Apr 03 15:59:19 2009 -0700
@@ -28,7 +28,7 @@
 
 // ReservedSpace
 ReservedSpace::ReservedSpace(size_t size) {
-  initialize(size, 0, false, NULL, 0);
+  initialize(size, 0, false, NULL, 0, false);
 }
 
 ReservedSpace::ReservedSpace(size_t size, size_t alignment,
@@ -36,7 +36,13 @@
                              char* requested_address,
                              const size_t noaccess_prefix) {
   initialize(size+noaccess_prefix, alignment, large, requested_address,
-             noaccess_prefix);
+             noaccess_prefix, false);
+}
+
+ReservedSpace::ReservedSpace(size_t size, size_t alignment,
+                             bool large,
+                             bool executable) {
+  initialize(size, alignment, large, NULL, 0, executable);
 }
 
 char *
@@ -132,7 +138,8 @@
   const bool try_reserve_special = UseLargePages &&
     prefix_align == os::large_page_size();
   if (!os::can_commit_large_page_memory() && try_reserve_special) {
-    initialize(size, prefix_align, true, requested_address, noaccess_prefix);
+    initialize(size, prefix_align, true, requested_address, noaccess_prefix,
+               false);
     return;
   }
 
@@ -141,6 +148,7 @@
   _alignment = 0;
   _special = false;
   _noaccess_prefix = 0;
+  _executable = false;
 
   // Assert that if noaccess_prefix is used, it is the same as prefix_align.
   assert(noaccess_prefix == 0 ||
@@ -189,7 +197,8 @@
 
 void ReservedSpace::initialize(size_t size, size_t alignment, bool large,
                                char* requested_address,
-                               const size_t noaccess_prefix) {
+                               const size_t noaccess_prefix,
+                               bool executable) {
   const size_t granularity = os::vm_allocation_granularity();
   assert((size & granularity - 1) == 0,
          "size not aligned to os::vm_allocation_granularity()");
@@ -201,6 +210,7 @@
   _base = NULL;
   _size = 0;
   _special = false;
+  _executable = executable;
   _alignment = 0;
   _noaccess_prefix = 0;
   if (size == 0) {
@@ -214,7 +224,7 @@
 
   if (special) {
 
-    base = os::reserve_memory_special(size, requested_address);
+    base = os::reserve_memory_special(size, requested_address, executable);
 
     if (base != NULL) {
       // Check alignment constraints
@@ -284,7 +294,7 @@
 
 
 ReservedSpace::ReservedSpace(char* base, size_t size, size_t alignment,
-                             bool special) {
+                             bool special, bool executable) {
   assert((size % os::vm_allocation_granularity()) == 0,
          "size not allocation aligned");
   _base = base;
@@ -292,6 +302,7 @@
   _alignment = alignment;
   _noaccess_prefix = 0;
   _special = special;
+  _executable = executable;
 }
 
 
@@ -299,9 +310,10 @@
                                         bool split, bool realloc) {
   assert(partition_size <= size(), "partition failed");
   if (split) {
-    os::split_reserved_memory(_base, _size, partition_size, realloc);
+    os::split_reserved_memory(base(), size(), partition_size, realloc);
   }
-  ReservedSpace result(base(), partition_size, alignment, special());
+  ReservedSpace result(base(), partition_size, alignment, special(),
+                       executable());
   return result;
 }
 
@@ -310,7 +322,7 @@
 ReservedSpace::last_part(size_t partition_size, size_t alignment) {
   assert(partition_size <= size(), "partition failed");
   ReservedSpace result(base() + partition_size, size() - partition_size,
-                       alignment, special());
+                       alignment, special(), executable());
   return result;
 }
 
@@ -348,6 +360,7 @@
     _size = 0;
     _noaccess_prefix = 0;
     _special = false;
+    _executable = false;
   }
 }
 
@@ -396,6 +409,14 @@
   protect_noaccess_prefix(prefix_size+suffix_size);
 }
 
+// Reserve space for code segment.  Same as Java heap only we mark this as
+// executable.
+ReservedCodeSpace::ReservedCodeSpace(size_t r_size,
+                                     size_t rs_align,
+                                     bool large) :
+  ReservedSpace(r_size, rs_align, large, /*executable*/ true) {
+}
+
 // VirtualSpace
 
 VirtualSpace::VirtualSpace() {
@@ -413,6 +434,7 @@
   _middle_alignment       = 0;
   _upper_alignment        = 0;
   _special                = false;
+  _executable             = false;
 }
 
 
@@ -426,6 +448,7 @@
   _high = low();
 
   _special = rs.special();
+  _executable = rs.executable();
 
   // When a VirtualSpace begins life at a large size, make all future expansion
   // and shrinking occur aligned to a granularity of large pages.  This avoids
@@ -483,6 +506,7 @@
   _middle_alignment       = 0;
   _upper_alignment        = 0;
   _special                = false;
+  _executable             = false;
 }
 
 
@@ -592,7 +616,7 @@
     assert(low_boundary() <= lower_high() &&
            lower_high() + lower_needs <= lower_high_boundary(),
            "must not expand beyond region");
-    if (!os::commit_memory(lower_high(), lower_needs)) {
+    if (!os::commit_memory(lower_high(), lower_needs, _executable)) {
       debug_only(warning("os::commit_memory failed"));
       return false;
     } else {
@@ -603,7 +627,8 @@
     assert(lower_high_boundary() <= middle_high() &&
            middle_high() + middle_needs <= middle_high_boundary(),
            "must not expand beyond region");
-    if (!os::commit_memory(middle_high(), middle_needs, middle_alignment())) {
+    if (!os::commit_memory(middle_high(), middle_needs, middle_alignment(),
+                           _executable)) {
       debug_only(warning("os::commit_memory failed"));
       return false;
     }
@@ -613,7 +638,7 @@
     assert(middle_high_boundary() <= upper_high() &&
            upper_high() + upper_needs <= upper_high_boundary(),
            "must not expand beyond region");
-    if (!os::commit_memory(upper_high(), upper_needs)) {
+    if (!os::commit_memory(upper_high(), upper_needs, _executable)) {
       debug_only(warning("os::commit_memory failed"));
       return false;
     } else {
--- a/src/share/vm/runtime/virtualspace.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/runtime/virtualspace.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -32,12 +32,15 @@
   size_t _noaccess_prefix;
   size_t _alignment;
   bool   _special;
+  bool   _executable;
 
   // ReservedSpace
-  ReservedSpace(char* base, size_t size, size_t alignment, bool special);
+  ReservedSpace(char* base, size_t size, size_t alignment, bool special,
+                bool executable);
   void initialize(size_t size, size_t alignment, bool large,
                   char* requested_address,
-                  const size_t noaccess_prefix);
+                  const size_t noaccess_prefix,
+                  bool executable);
 
   // Release parts of an already-reserved memory region [addr, addr + len) to
   // get a new region that has "compound alignment."  Return the start of the
@@ -75,16 +78,16 @@
                 const size_t suffix_size, const size_t suffix_align,
                 char* requested_address,
                 const size_t noaccess_prefix = 0);
+  ReservedSpace(size_t size, size_t alignment, bool large, bool executable);
 
   // Accessors
-  char*  base()      const { return _base;      }
-  size_t size()      const { return _size;      }
-  size_t alignment() const { return _alignment; }
-  bool   special()   const { return _special;   }
-
-  size_t noaccess_prefix()   const { return _noaccess_prefix;   }
-
-  bool is_reserved() const { return _base != NULL; }
+  char*  base()            const { return _base;      }
+  size_t size()            const { return _size;      }
+  size_t alignment()       const { return _alignment; }
+  bool   special()         const { return _special;   }
+  bool   executable()      const { return _executable;   }
+  size_t noaccess_prefix() const { return _noaccess_prefix;   }
+  bool is_reserved()       const { return _base != NULL; }
   void release();
 
   // Splitting
@@ -126,6 +129,13 @@
                     char* requested_address);
 };
 
+// Class encapsulating behavior specific memory space for Code
+class ReservedCodeSpace : public ReservedSpace {
+ public:
+  // Constructor
+  ReservedCodeSpace(size_t r_size, size_t rs_align, bool large);
+};
+
 // VirtualSpace is data structure for committing a previously reserved address range in smaller chunks.
 
 class VirtualSpace VALUE_OBJ_CLASS_SPEC {
@@ -143,6 +153,9 @@
   // os::commit_memory() or os::uncommit_memory().
   bool _special;
 
+  // Need to know if commit should be executable.
+  bool   _executable;
+
   // MPSS Support
   // Each virtualspace region has a lower, middle, and upper region.
   // Each region has an end boundary and a high pointer which is the
--- a/src/share/vm/utilities/globalDefinitions_visCPP.hpp	Tue Mar 03 14:23:18 2009 -0800
+++ b/src/share/vm/utilities/globalDefinitions_visCPP.hpp	Fri Apr 03 15:59:19 2009 -0700
@@ -153,14 +153,6 @@
 //----------------------------------------------------------------------------------------------------
 // Miscellaneous
 
-inline int vsnprintf(char* buf, size_t count, const char* fmt, va_list argptr) {
-  // If number of characters written == count, Windows doesn't write a
-  // terminating NULL, so we do it ourselves.
-  int ret = _vsnprintf(buf, count, fmt, argptr);
-  if (count > 0) buf[count-1] = '\0';
-  return ret;
-}
-
 // Visual Studio 2005 deprecates POSIX names - use ISO C++ names instead
 #if _MSC_VER >= 1400
 #define open _open
@@ -180,6 +172,17 @@
 #pragma warning( disable : 4201 ) // nonstandard extension used : nameless struct/union (needed in windows.h)
 #pragma warning( disable : 4511 ) // copy constructor could not be generated
 #pragma warning( disable : 4291 ) // no matching operator delete found; memory will not be freed if initialization thows an exception
+#if _MSC_VER >= 1400
+#pragma warning( disable : 4996 ) // unsafe string functions. Same as define _CRT_SECURE_NO_WARNINGS/_CRT_SECURE_NO_DEPRICATE
+#endif
+
+inline int vsnprintf(char* buf, size_t count, const char* fmt, va_list argptr) {
+  // If number of characters written == count, Windows doesn't write a
+  // terminating NULL, so we do it ourselves.
+  int ret = _vsnprintf(buf, count, fmt, argptr);
+  if (count > 0) buf[count-1] = '\0';
+  return ret;
+}
 
 // Portability macros
 #define PRAGMA_INTERFACE
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/6636138/Test1.java	Fri Apr 03 15:59:19 2009 -0700
@@ -0,0 +1,67 @@
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ */
+
+/**
+ * @test
+ * @bug 6636138
+ * @summary SuperWord::co_locate_pack(Node_List* p) generates memory graph that leads to memory order violation.
+ *
+ * @run main/othervm -server -Xbatch -XX:CompileOnly=Test1.init -XX:+UseSuperword Test1
+ */
+
+class Test1 {
+
+    public static void init(int src[], int [] dst, int[] ref) {
+        // initialize the arrays
+        for (int i =0; i<src.length; i++) {
+            src[i] =  i;
+            dst[i] = 2;      // yes, dst[i] needed(otherwise src[i] will be replaced with i)
+            ref[i] = src[i]; // src[i] depends on the store src[i]
+        }
+    }
+
+    public static void verify(int src[], int[] ref) {
+        // check whether src and ref are equal
+        for (int i = 0; i < src.length; i++) {
+            if (src[i] != ref[i]) {
+                System.out.println("Error: src and ref don't match at " + i);
+                System.exit(-1);
+            }
+        }
+    }
+
+    public static void test() {
+        int[] src = new int[34];
+        int[] dst = new int[34];
+        int[] ref = new int[34];
+
+        init(src, dst, ref);
+        verify(src, ref);
+    }
+
+    public static void main(String[] args) {
+        for (int i=0; i< 2000; i++) {
+            test();
+        }
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/6636138/Test2.java	Fri Apr 03 15:59:19 2009 -0700
@@ -0,0 +1,70 @@
+/*
+ * Copyright 2009 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ */
+
+/**
+ * @test
+ * @bug 6636138
+ * @summary SuperWord::co_locate_pack(Node_List* p) generates memory graph that leads to memory order violation.
+ *
+ * @run main/othervm -server -Xbatch -XX:CompileOnly=Test2.shift -XX:+UseSuperword Test2
+ */
+
+class Test2 {
+
+    public static void init(int src[]) {
+        // Initialize the array
+        for (int i = 0; i < src.length; i++)
+            src[i] = i;
+    }
+
+   public static void shift(int src[]) {
+       //left-shift the array
+       for (int i = src.length-1; i > 0; i--){
+           int tmp  = src[i];
+           src[i]   = src[i-1];
+           src[i-1] = tmp;
+       }
+    }
+
+    public static void verify(int src[]) {
+        for (int i = 0; i < src.length; i++){
+            int value = (i-1 + src.length)%src.length; // correct value after shifting
+                if (src[i] != value) {
+                    System.out.println("Error: src["+i+"] should be "+ value + " instead of " + src[i]);
+                    System.exit(-1);
+                }
+        }
+    }
+
+    public static void test() {
+        int[] src = new int[10];
+        init(src);
+        shift(src);
+        verify(src);
+    }
+
+    public static void main(String[] args) {
+        for (int i=0; i< 2000; i++)
+            test();
+    }
+}
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/runtime/6819213/TestBootNativeLibraryPath.java	Fri Apr 03 15:59:19 2009 -0700
@@ -0,0 +1,133 @@
+/*
+ * Copyright 2008 Sun Microsystems, Inc.  All Rights Reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
+ * CA 95054 USA or visit www.sun.com if you need additional information or
+ * have any questions.
+ */
+
+/*
+ * @test TestBootNativeLibraryPath.java
+ * @bug 6819213
+ * @compile -XDignore.symbol.file TestBootNativeLibraryPath.java
+ * @summary verify sun.boot.native.library.path is expandable on 32 bit systems
+ * @run main TestBootNativeLibraryPath
+ * @author ksrini
+*/
+
+import java.io.BufferedReader;
+import java.io.File;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.io.InputStreamReader;
+import java.io.PrintStream;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+import java.util.logging.Level;
+import java.util.logging.Logger;
+import javax.tools.JavaCompiler;
+import javax.tools.ToolProvider;
+
+public class TestBootNativeLibraryPath {
+
+    private static final String TESTFILE = "Test6";
+
+    static void createTestClass() throws IOException {
+        FileOutputStream fos = new FileOutputStream(TESTFILE + ".java");
+        PrintStream ps = new PrintStream(fos);
+        ps.println("public class " + TESTFILE + "{");
+        ps.println("public static void main(String[] args) {\n");
+        ps.println("System.out.println(System.getProperty(\"sun.boot.library.path\"));\n");
+        ps.println("}}\n");
+        ps.close();
+        fos.close();
+
+        JavaCompiler javac = ToolProvider.getSystemJavaCompiler();
+        String javacOpts[] = {TESTFILE + ".java"};
+        if (javac.run(null, null, null,  javacOpts) != 0) {
+            throw new RuntimeException("compilation of " + TESTFILE + ".java Failed");
+        }
+    }
+
+    static List<String> doExec(String... args) {
+        String javaCmd = System.getProperty("java.home") + "/bin/java";
+        if (!new File(javaCmd).exists()) {
+            javaCmd = System.getProperty("java.home") + "/bin/java.exe";
+        }
+
+        ArrayList<String> cmds = new ArrayList<String>();
+        cmds.add(javaCmd);
+        for (String x : args) {
+            cmds.add(x);
+        }
+        System.out.println("cmds=" + cmds);
+        ProcessBuilder pb = new ProcessBuilder(cmds);
+
+        Map<String, String> env = pb.environment();
+        pb.directory(new File("."));
+
+        List<String> out = new ArrayList<String>();
+        try {
+            pb.redirectErrorStream(true);
+            Process p = pb.start();
+            BufferedReader rd = new BufferedReader(new InputStreamReader(p.getInputStream()),8192);
+            String in = rd.readLine();
+            while (in != null) {
+                out.add(in);
+                System.out.println(in);
+                in = rd.readLine();
+            }
+            int retval = p.waitFor();
+            p.destroy();
+            if (retval != 0) {
+                throw new RuntimeException("Error: test returned non-zero value");
+            }
+            return out;
+        } catch (Exception ex) {
+            ex.printStackTrace();
+            throw new RuntimeException(ex.getMessage());
+        }
+    }
+
+    public static void main(String[] args) {
+        try {
+            if (!System.getProperty("sun.arch.data.model").equals("32")) {
+                System.out.println("Warning: test skipped for 64-bit systems\n");
+                return;
+            }
+            String osname = System.getProperty("os.name");
+            if (osname.startsWith("Windows")) {
+                osname = "Windows";
+            }
+
+            createTestClass();
+
+            // Test a simple path
+            String libpath = File.pathSeparator + "tmp" + File.pathSeparator + "foobar";
+            List<String> processOut = null;
+            String sunbootlibrarypath = "-Dsun.boot.library.path=" + libpath;
+            processOut = doExec(sunbootlibrarypath, "-cp", ".", TESTFILE);
+            if (processOut == null || !processOut.get(0).endsWith(libpath)) {
+                throw new RuntimeException("Error: did not get expected error string");
+            }
+        } catch (IOException ex) {
+            throw new RuntimeException("Unexpected error " + ex);
+        }
+    }
+}