# HG changeset patch
# User kvn
# Date 1238518921 25200
# Node ID d3676b4cb78c62057456bc65bf4dc445d1c5825d
# Parent  9ab385cb0c42997e16a7761ebcd25c90560a2714# Parent  f6da6f0174ac01a4827628ef2bf5d93efecd2ad6
Merge

diff -r 9ab385cb0c42 -r d3676b4cb78c agent/src/share/classes/sun/jvm/hotspot/runtime/StubRoutines.java
--- a/agent/src/share/classes/sun/jvm/hotspot/runtime/StubRoutines.java	Fri Mar 27 16:58:30 2009 -0700
+++ b/agent/src/share/classes/sun/jvm/hotspot/runtime/StubRoutines.java	Tue Mar 31 10:02:01 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() {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/assembler_sparc.cpp
--- a/src/cpu/sparc/vm/assembler_sparc.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/assembler_sparc.cpp	Tue Mar 31 10:02:01 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);
@@ -4472,7 +4472,7 @@
 }
 
 // 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:
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/assembler_sparc.hpp
--- a/src/cpu/sparc/vm/assembler_sparc.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/assembler_sparc.hpp	Tue Mar 31 10:02:01 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();
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/assembler_sparc.inline.hpp
--- a/src/cpu/sparc/vm/assembler_sparc.inline.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/assembler_sparc.inline.hpp	Tue Mar 31 10:02:01 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);
 }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp
--- a/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/c1_LIRAssembler_sparc.cpp	Tue Mar 31 10:02:01 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) {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/globals_sparc.hpp
--- a/src/cpu/sparc/vm/globals_sparc.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/globals_sparc.hpp	Tue Mar 31 10:02:01 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);
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/stubGenerator_sparc.cpp
--- a/src/cpu/sparc/vm/stubGenerator_sparc.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/stubGenerator_sparc.cpp	Tue Mar 31 10:02:01 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
   }
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/sparc/vm/vm_version_sparc.cpp
--- a/src/cpu/sparc/vm/vm_version_sparc.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/sparc/vm/vm_version_sparc.cpp	Tue Mar 31 10:02:01 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);
     }
   }
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/assembler_x86.cpp
--- a/src/cpu/x86/vm/assembler_x86.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/assembler_x86.cpp	Tue Mar 31 10:02:01 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) {
@@ -7218,7 +7204,7 @@
 // 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 +7289,7 @@
                                                    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 +7338,7 @@
   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 +7536,12 @@
 }
 
 
-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 +7557,7 @@
   if (offset != 0)
     addptr(tmp, offset);
 
-  return RegisterConstant(tmp);
+  return RegisterOrConstant(tmp);
 }
 
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/assembler_x86.hpp
--- a/src/cpu/x86/vm/assembler_x86.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/assembler_x86.hpp	Tue Mar 31 10:02:01 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();
@@ -1802,7 +1810,7 @@
   // 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 +1827,7 @@
                                      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 +1891,9 @@
   // 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);
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/globals_x86.hpp
--- a/src/cpu/x86/vm/globals_x86.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/globals_x86.hpp	Tue Mar 31 10:02:01 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);
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/stubGenerator_x86_64.cpp
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Tue Mar 31 10:02:01 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;
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/x86_32.ad
--- a/src/cpu/x86/vm/x86_32.ad	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/x86_32.ad	Tue Mar 31 10:02:01 2009 -0700
@@ -4288,24 +4288,6 @@
     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 +7480,9 @@
   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 +7501,9 @@
   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 +7517,22 @@
   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);
 %}
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/cpu/x86/vm/x86_64.ad
--- a/src/cpu/x86/vm/x86_64.ad	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/cpu/x86/vm/x86_64.ad	Tue Mar 31 10:02:01 2009 -0700
@@ -4162,33 +4162,6 @@
     // 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.
@@ -7458,7 +7431,7 @@
   ins_cost(0);
 
   size(0);
-  format %{ "MEMBAR-acquire" %}
+  format %{ "MEMBAR-acquire ! (empty encoding)" %}
   ins_encode();
   ins_pipe(empty);
 %}
@@ -7481,7 +7454,7 @@
   ins_cost(0);
 
   size(0);
-  format %{ "MEMBAR-release" %}
+  format %{ "MEMBAR-release ! (empty encoding)" %}
   ins_encode();
   ins_pipe(empty);
 %}
@@ -7498,13 +7471,22 @@
   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);
 %}
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/linux_sparc/vm/os_linux_sparc.hpp
--- a/src/os_cpu/linux_sparc/vm/os_linux_sparc.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/linux_sparc/vm/os_linux_sparc.hpp	Tue Mar 31 10:02:01 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() {}
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/linux_x86/vm/orderAccess_linux_x86.inline.hpp
--- a/src/os_cpu/linux_x86/vm/orderAccess_linux_x86.inline.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/linux_x86/vm/orderAccess_linux_x86.inline.hpp	Tue Mar 31 10:02:01 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
   }
 }
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_sparc/vm/orderAccess_solaris_sparc.inline.hpp
--- a/src/os_cpu/solaris_sparc/vm/orderAccess_solaris_sparc.inline.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_sparc/vm/orderAccess_solaris_sparc.inline.hpp	Tue Mar 31 10:02:01 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; }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.cpp	Tue Mar 31 10:02:01 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
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_sparc/vm/os_solaris_sparc.hpp
--- a/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_sparc/vm/os_solaris_sparc.hpp	Tue Mar 31 10:02:01 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() {}
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_x86/vm/orderAccess_solaris_x86.inline.hpp
--- a/src/os_cpu/solaris_x86/vm/orderAccess_solaris_x86.inline.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_x86/vm/orderAccess_solaris_x86.inline.hpp	Tue Mar 31 10:02:01 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
   }
 
 }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp
--- a/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_x86/vm/os_solaris_x86.cpp	Tue Mar 31 10:02:01 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() {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/solaris_x86/vm/os_solaris_x86.hpp
--- a/src/os_cpu/solaris_x86/vm/os_solaris_x86.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/solaris_x86/vm/os_solaris_x86.hpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/windows_x86/vm/orderAccess_windows_x86.inline.hpp
--- a/src/os_cpu/windows_x86/vm/orderAccess_windows_x86.inline.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/windows_x86/vm/orderAccess_windows_x86.inline.hpp	Tue Mar 31 10:02:01 2009 -0700
@@ -46,7 +46,7 @@
 
 inline void OrderAccess::fence() {
 #ifdef AMD64
-  (*os::fence_func)();
+  StubRoutines_fence();
 #else
   if (os::is_MP()) {
     __asm {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/windows_x86/vm/os_windows_x86.cpp
--- a/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/windows_x86/vm/os_windows_x86.cpp	Tue Mar 31 10:02:01 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
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/os_cpu/windows_x86/vm/os_windows_x86.hpp
--- a/src/os_cpu/windows_x86/vm/os_windows_x86.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/os_cpu/windows_x86/vm/os_windows_x86.hpp	Tue Mar 31 10:02:01 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();
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/asm/assembler.hpp
--- a/src/share/vm/asm/assembler.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/asm/assembler.hpp	Tue Mar 31 10:02:01 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)());
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/ci/ciTypeFlow.cpp
--- a/src/share/vm/ci/ciTypeFlow.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/ci/ciTypeFlow.cpp	Tue Mar 31 10:02:01 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);
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/javaClasses.cpp
--- a/src/share/vm/classfile/javaClasses.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/javaClasses.cpp	Tue Mar 31 10:02:01 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());
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/javaClasses.hpp
--- a/src/share/vm/classfile/javaClasses.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/javaClasses.hpp	Tue Mar 31 10:02:01 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 :
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/loaderConstraints.hpp
--- a/src/share/vm/classfile/loaderConstraints.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/loaderConstraints.hpp	Tue Mar 31 10:02:01 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,
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/symbolTable.cpp
--- a/src/share/vm/classfile/symbolTable.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/symbolTable.cpp	Tue Mar 31 10:02:01 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(),
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/symbolTable.hpp
--- a/src/share/vm/classfile/symbolTable.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/symbolTable.hpp	Tue Mar 31 10:02:01 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;
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/systemDictionary.cpp
--- a/src/share/vm/classfile/systemDictionary.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/systemDictionary.cpp	Tue Mar 31 10:02:01 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)  {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/systemDictionary.hpp
--- a/src/share/vm/classfile/systemDictionary.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/systemDictionary.hpp	Tue Mar 31 10:02:01 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);
+};
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/classfile/vmSymbols.hpp
--- a/src/share/vm/classfile/vmSymbols.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/classfile/vmSymbols.hpp	Tue Mar 31 10:02:01 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 +287,7 @@
   template(frontCacheEnabled_name,                    "frontCacheEnabled")                        \
   template(stringCacheEnabled_name,                   "stringCacheEnabled")                       \
   template(bitCount_name,                             "bitCount")                                 \
+  template(profile_name,                              "profile")                                  \
                                                                                                   \
   /* non-intrinsic name/signature pairs: */                                                       \
   template(register_method_name,                      "register")                                 \
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp
--- a/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/gc_implementation/parallelScavenge/parallelScavengeHeap.cpp	Tue Mar 31 10:02:01 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),
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/includeDB_core
--- a/src/share/vm/includeDB_core	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/includeDB_core	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/instanceKlass.cpp
--- a/src/share/vm/oops/instanceKlass.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/instanceKlass.cpp	Tue Mar 31 10:02:01 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;
 
@@ -1879,6 +1881,56 @@
    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();
   jint access = access_flags().as_int();
@@ -2015,9 +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 {
@@ -2038,7 +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();
@@ -2046,23 +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();
   }
 }
 
@@ -2070,6 +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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/instanceKlass.hpp
--- a/src/share/vm/oops/instanceKlass.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/instanceKlass.hpp	Tue Mar 31 10:02:01 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
@@ -311,6 +311,22 @@
   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; }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/instanceKlassKlass.cpp
--- a/src/share/vm/oops/instanceKlassKlass.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/instanceKlassKlass.cpp	Tue Mar 31 10:02:01 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) {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/klassVtable.cpp
--- a/src/share/vm/oops/klassVtable.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/klassVtable.cpp	Tue Mar 31 10:02:01 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
@@ -1148,6 +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;
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/klassVtable.hpp
--- a/src/share/vm/oops/klassVtable.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/klassVtable.hpp	Tue Mar 31 10:02:01 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
@@ -301,6 +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;
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/methodKlass.cpp
--- a/src/share/vm/oops/methodKlass.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/methodKlass.cpp	Tue Mar 31 10:02:01 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());
+  }
 }
 
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/objArrayKlass.cpp
--- a/src/share/vm/oops/objArrayKlass.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/objArrayKlass.cpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/oops/oop.cpp
--- a/src/share/vm/oops/oop.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/oops/oop.cpp	Tue Mar 31 10:02:01 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);
   }
 }
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/block.hpp
--- a/src/share/vm/opto/block.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/block.hpp	Tue Mar 31 10:02:01 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; }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/bytecodeInfo.cpp
--- a/src/share/vm/opto/bytecodeInfo.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/bytecodeInfo.cpp	Tue Mar 31 10:02:01 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;
 }
 
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/c2_globals.hpp
--- a/src/share/vm/opto/c2_globals.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/c2_globals.hpp	Tue Mar 31 10:02:01 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)
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/chaitin.cpp
--- a/src/share/vm/opto/chaitin.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/chaitin.cpp	Tue Mar 31 10:02:01 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 );
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/chaitin.hpp
--- a/src/share/vm/opto/chaitin.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/chaitin.hpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/coalesce.cpp
--- a/src/share/vm/opto/coalesce.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/coalesce.cpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/gcm.cpp
--- a/src/share/vm/opto/gcm.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/gcm.cpp	Tue Mar 31 10:02:01 2009 -0700
@@ -1374,6 +1374,9 @@
   _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 +1901,7 @@
 // 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 +1916,14 @@
   }
 }
 
+// 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 {
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/machnode.cpp
--- a/src/share/vm/opto/machnode.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/machnode.cpp	Tue Mar 31 10:02:01 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");
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/subnode.cpp
--- a/src/share/vm/opto/subnode.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/subnode.cpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/superword.cpp
--- a/src/share/vm/opto/superword.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/superword.cpp	Tue Mar 31 10:02:01 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);
     }
   }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/opto/superword.hpp
--- a/src/share/vm/opto/superword.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/opto/superword.hpp	Tue Mar 31 10:02:01 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();
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/prims/jvm.cpp
--- a/src/share/vm/prims/jvm.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/prims/jvm.cpp	Tue Mar 31 10:02:01 2009 -0700
@@ -1252,7 +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++;
@@ -1275,16 +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
@@ -1298,35 +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");
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/fieldDescriptor.cpp
--- a/src/share/vm/runtime/fieldDescriptor.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/fieldDescriptor.cpp	Tue Mar 31 10:02:01 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);
   }
 }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/globals.hpp
--- a/src/share/vm/runtime/globals.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/globals.hpp	Tue Mar 31 10:02:01 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);
@@ -2616,7 +2617,7 @@
   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")                                                 \
                                                                             \
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/handles.hpp
--- a/src/share/vm/runtime/handles.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/handles.hpp	Tue Mar 31 10:02:01 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(); }
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/orderAccess.cpp
--- a/src/share/vm/runtime/orderAccess.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/orderAccess.cpp	Tue Mar 31 10:02:01 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");
+}
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/orderAccess.hpp
--- a/src/share/vm/runtime/orderAccess.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/orderAccess.hpp	Tue Mar 31 10:02:01 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();
 };
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/reflection.cpp
--- a/src/share/vm/runtime/reflection.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/reflection.cpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/reflection.hpp
--- a/src/share/vm/runtime/reflection.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/reflection.hpp	Tue Mar 31 10:02:01 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)
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/sharedRuntime.cpp
--- a/src/share/vm/runtime/sharedRuntime.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/sharedRuntime.cpp	Tue Mar 31 10:02:01 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");
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/sharedRuntime.hpp
--- a/src/share/vm/runtime/sharedRuntime.hpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/sharedRuntime.hpp	Tue Mar 31 10:02:01 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
diff -r 9ab385cb0c42 -r d3676b4cb78c src/share/vm/runtime/thread.cpp
--- a/src/share/vm/runtime/thread.cpp	Fri Mar 27 16:58:30 2009 -0700
+++ b/src/share/vm/runtime/thread.cpp	Tue Mar 31 10:02:01 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);
+          }
         }
       }
     }
diff -r 9ab385cb0c42 -r d3676b4cb78c test/compiler/6636138/Test1.java
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/6636138/Test1.java	Tue Mar 31 10:02:01 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();
+        }
+    }
+}
diff -r 9ab385cb0c42 -r d3676b4cb78c test/compiler/6636138/Test2.java
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/compiler/6636138/Test2.java	Tue Mar 31 10:02:01 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();
+    }
+}