# HG changeset patch
# User never
# Date 1329509932 28800
# Node ID 15085a6eb50c03a0eeb68d910e0f16b2c3fb6649
# Parent  72c425c46102a0204285c1d84657aa2e80776f5c# Parent  d3384450b64967964468290c7c2534de5c8880d3
Merge

diff -r d3384450b649 -r 15085a6eb50c make/Makefile
--- a/make/Makefile	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/Makefile	Fri Feb 17 12:18:52 2012 -0800
@@ -402,7 +402,6 @@
 	$(install-file)
 else
 $(EXPORT_INCLUDE_DIR)/jfr.h:
-	
 endif
 
 # Doc files (jvmti.html)
@@ -448,12 +447,18 @@
 	 ($(CD) $(JDK_IMAGE_DIR) && $(TAR) -xf -)
 
 test_jdk:
-  ifneq ($(ZERO_BUILD), true)
     ifeq ($(ARCH_DATA_MODEL), 32)
-	$(JDK_IMAGE_DIR)/bin/java -client -version
+      ifneq ($(ZERO_BUILD), true)
+	$(JDK_IMAGE_DIR)/bin/java -d32 -client -Xinternalversion
+	$(JDK_IMAGE_DIR)/bin/java -d32 -client -version
+      endif
+	$(JDK_IMAGE_DIR)/bin/java -d32 -server -Xinternalversion
+	$(JDK_IMAGE_DIR)/bin/java -d32 -server -version
     endif
-  endif
-	$(JDK_IMAGE_DIR)/bin/java -server -version
+    ifeq ($(ARCH_DATA_MODEL), 64)
+	$(JDK_IMAGE_DIR)/bin/java -d64 -server -Xinternalversion
+	$(JDK_IMAGE_DIR)/bin/java -d64 -server -version
+    endif
 
 copy_product_jdk::
 	$(RM) -r $(JDK_IMAGE_DIR)
@@ -545,6 +550,7 @@
 OUTPUTDIR.desc             = Output directory, default is build/<osname>
 BOOTDIR.desc               = JDK used to compile agent java source and test with
 JDK_IMPORT_PATH.desc       = Promoted JDK to copy for 'create_jdk'
+JDK_IMAGE_DIR.desc         = Directory to place JDK to copy
 EXPORT_PATH.desc           = Directory to place files to export for JDK build
 
 # Make variables to print out (description and value)
@@ -553,6 +559,7 @@
     OUTPUTDIR                   \
     BOOTDIR                     \
     JDK_IMPORT_PATH             \
+    JDK_IMAGE_DIR               \
     EXPORT_PATH
 
 # Make variables that should refer to directories that exist
diff -r d3384450b649 -r 15085a6eb50c make/bsd/makefiles/defs.make
--- a/make/bsd/makefiles/defs.make	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/bsd/makefiles/defs.make	Fri Feb 17 12:18:52 2012 -0800
@@ -191,6 +191,9 @@
 
     # Set universal image dir
     JDK_IMAGE_DIR=$(OUTPUTDIR)/jdk-universal$(EXPORT_SUBDIR)
+    ifneq ($(ALT_JDK_IMAGE_DIR),)
+      JDK_IMAGE_DIR=$(ALT_JDK_IMAGE_DIR)
+    endif
 
     # Binaries to 'universalize' if built
     UNIVERSAL_LIPO_LIST += $(EXPORT_JRE_LIB_DIR)/libjsig.$(LIBRARY_SUFFIX)
diff -r d3384450b649 -r 15085a6eb50c make/bsd/makefiles/top.make
--- a/make/bsd/makefiles/top.make	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/bsd/makefiles/top.make	Fri Feb 17 12:18:52 2012 -0800
@@ -124,8 +124,8 @@
 	@$(UpdatePCH)
 	@$(MAKE) -f vm.make $(MFLAGS-adjusted)
 
-install: the_vm
-	@$(MAKE) -f vm.make install
+install gamma: the_vm
+	@$(MAKE) -f vm.make $@
 
 # next rules support "make foo.[ois]"
 
diff -r d3384450b649 -r 15085a6eb50c make/defs.make
--- a/make/defs.make	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/defs.make	Fri Feb 17 12:18:52 2012 -0800
@@ -193,6 +193,9 @@
 
 # Default jdk image if one is created for you with create_jdk
 JDK_IMAGE_DIR=$(OUTPUTDIR)/jdk-$(PLATFORM)
+ifneq ($(ALT_JDK_IMAGE_DIR),)
+  JDK_IMAGE_DIR=$(ALT_JDK_IMAGE_DIR)
+endif
 
 # The platform dependent defs.make defines platform specific variable such 
 # as ARCH, EXPORT_LIST etc. We must place the include here after BOOTDIR is defined.
diff -r d3384450b649 -r 15085a6eb50c make/linux/makefiles/top.make
--- a/make/linux/makefiles/top.make	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/linux/makefiles/top.make	Fri Feb 17 12:18:52 2012 -0800
@@ -115,8 +115,8 @@
 	@$(UpdatePCH)
 	@$(MAKE) -f vm.make $(MFLAGS-adjusted)
 
-install: the_vm
-	@$(MAKE) -f vm.make install
+install gamma: the_vm
+	@$(MAKE) -f vm.make $@
 
 # next rules support "make foo.[ois]"
 
diff -r d3384450b649 -r 15085a6eb50c make/solaris/makefiles/top.make
--- a/make/solaris/makefiles/top.make	Fri Feb 17 06:34:49 2012 -0800
+++ b/make/solaris/makefiles/top.make	Fri Feb 17 12:18:52 2012 -0800
@@ -107,8 +107,8 @@
 the_vm: vm_build_preliminaries $(adjust-mflags)
 	@$(MAKE) -f vm.make $(MFLAGS-adjusted)
 
-install: the_vm
-	@$(MAKE) -f vm.make install
+install gamma: the_vm
+	@$(MAKE) -f vm.make $@
 
 # next rules support "make foo.[oi]"
 
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/assembler_x86.cpp
--- a/src/cpu/x86/vm/assembler_x86.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/assembler_x86.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -236,6 +236,16 @@
   }
 }
 
+// Force generation of a 4 byte immediate value even if it fits into 8bit
+void Assembler::emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32) {
+  assert(isByte(op1) && isByte(op2), "wrong opcode");
+  assert((op1 & 0x01) == 1, "should be 32bit operation");
+  assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
+  emit_byte(op1);
+  emit_byte(op2 | encode(dst));
+  emit_long(imm32);
+}
+
 // immediate-to-memory forms
 void Assembler::emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32) {
   assert((op1 & 0x01) == 1, "should be 32bit operation");
@@ -939,6 +949,7 @@
 }
 
 void Assembler::addr_nop_4() {
+  assert(UseAddressNop, "no CPU support");
   // 4 bytes: NOP DWORD PTR [EAX+0]
   emit_byte(0x0F);
   emit_byte(0x1F);
@@ -947,6 +958,7 @@
 }
 
 void Assembler::addr_nop_5() {
+  assert(UseAddressNop, "no CPU support");
   // 5 bytes: NOP DWORD PTR [EAX+EAX*0+0] 8-bits offset
   emit_byte(0x0F);
   emit_byte(0x1F);
@@ -956,6 +968,7 @@
 }
 
 void Assembler::addr_nop_7() {
+  assert(UseAddressNop, "no CPU support");
   // 7 bytes: NOP DWORD PTR [EAX+0] 32-bits offset
   emit_byte(0x0F);
   emit_byte(0x1F);
@@ -964,6 +977,7 @@
 }
 
 void Assembler::addr_nop_8() {
+  assert(UseAddressNop, "no CPU support");
   // 8 bytes: NOP DWORD PTR [EAX+EAX*0+0] 32-bits offset
   emit_byte(0x0F);
   emit_byte(0x1F);
@@ -2769,6 +2783,12 @@
   emit_arith(0x81, 0xE8, dst, imm32);
 }
 
+// Force generation of a 4 byte immediate value even if it fits into 8bit
+void Assembler::subl_imm32(Register dst, int32_t imm32) {
+  prefix(dst);
+  emit_arith_imm32(0x81, 0xE8, dst, imm32);
+}
+
 void Assembler::subl(Register dst, Address src) {
   InstructionMark im(this);
   prefix(src, dst);
@@ -4760,6 +4780,12 @@
   emit_arith(0x81, 0xE8, dst, imm32);
 }
 
+// Force generation of a 4 byte immediate value even if it fits into 8bit
+void Assembler::subq_imm32(Register dst, int32_t imm32) {
+  (void) prefixq_and_encode(dst->encoding());
+  emit_arith_imm32(0x81, 0xE8, dst, imm32);
+}
+
 void Assembler::subq(Register dst, Address src) {
   InstructionMark im(this);
   prefixq(src, dst);
@@ -5101,15 +5127,6 @@
   }
 }
 
-void MacroAssembler::fat_nop() {
-  // A 5 byte nop that is safe for patching (see patch_verified_entry)
-  emit_byte(0x26); // es:
-  emit_byte(0x2e); // cs:
-  emit_byte(0x64); // fs:
-  emit_byte(0x65); // gs:
-  emit_byte(0x90);
-}
-
 void MacroAssembler::jC2(Register tmp, Label& L) {
   // set parity bit if FPU flag C2 is set (via rax)
   save_rax(tmp);
@@ -5704,17 +5721,6 @@
   /* else */      { subq(dst, value)       ; return; }
 }
 
-void MacroAssembler::fat_nop() {
-  // A 5 byte nop that is safe for patching (see patch_verified_entry)
-  // Recommened sequence from 'Software Optimization Guide for the AMD
-  // Hammer Processor'
-  emit_byte(0x66);
-  emit_byte(0x66);
-  emit_byte(0x90);
-  emit_byte(0x66);
-  emit_byte(0x90);
-}
-
 void MacroAssembler::incrementq(Register reg, int value) {
   if (value == min_jint) { addq(reg, value); return; }
   if (value <  0) { decrementq(reg, -value); return; }
@@ -6766,6 +6772,19 @@
   mov(rbp, rsp);
 }
 
+// A 5 byte nop that is safe for patching (see patch_verified_entry)
+void MacroAssembler::fat_nop() {
+  if (UseAddressNop) {
+    addr_nop_5();
+  } else {
+    emit_byte(0x26); // es:
+    emit_byte(0x2e); // cs:
+    emit_byte(0x64); // fs:
+    emit_byte(0x65); // gs:
+    emit_byte(0x90);
+  }
+}
+
 void MacroAssembler::fcmp(Register tmp) {
   fcmp(tmp, 1, true, true);
 }
@@ -7825,6 +7844,11 @@
   LP64_ONLY(subq(dst, imm32)) NOT_LP64(subl(dst, imm32));
 }
 
+// Force generation of a 4 byte immediate value even if it fits into 8bit
+void MacroAssembler::subptr_imm32(Register dst, int32_t imm32) {
+  LP64_ONLY(subq_imm32(dst, imm32)) NOT_LP64(subl_imm32(dst, imm32));
+}
+
 void MacroAssembler::subptr(Register dst, Register src) {
   LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src));
 }
@@ -9292,6 +9316,80 @@
 }
 #endif // _LP64
 
+
+// C2 compiled method's prolog code.
+void MacroAssembler::verified_entry(int framesize, bool stack_bang, bool fp_mode_24b) {
+
+  // WARNING: Initial instruction MUST be 5 bytes or longer so that
+  // NativeJump::patch_verified_entry will be able to patch out the entry
+  // code safely. The push to verify stack depth is ok at 5 bytes,
+  // the frame allocation can be either 3 or 6 bytes. So if we don't do
+  // stack bang then we must use the 6 byte frame allocation even if
+  // we have no frame. :-(
+
+  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
+  // Remove word for return addr
+  framesize -= wordSize;
+
+  // Calls to C2R adapters often do not accept exceptional returns.
+  // We require that their callers must bang for them.  But be careful, because
+  // some VM calls (such as call site linkage) can use several kilobytes of
+  // stack.  But the stack safety zone should account for that.
+  // See bugs 4446381, 4468289, 4497237.
+  if (stack_bang) {
+    generate_stack_overflow_check(framesize);
+
+    // We always push rbp, so that on return to interpreter rbp, will be
+    // restored correctly and we can correct the stack.
+    push(rbp);
+    // Remove word for ebp
+    framesize -= wordSize;
+
+    // Create frame
+    if (framesize) {
+      subptr(rsp, framesize);
+    }
+  } else {
+    // Create frame (force generation of a 4 byte immediate value)
+    subptr_imm32(rsp, framesize);
+
+    // Save RBP register now.
+    framesize -= wordSize;
+    movptr(Address(rsp, framesize), rbp);
+  }
+
+  if (VerifyStackAtCalls) { // Majik cookie to verify stack depth
+    framesize -= wordSize;
+    movptr(Address(rsp, framesize), (int32_t)0xbadb100d);
+  }
+
+#ifndef _LP64
+  // If method sets FPU control word do it now
+  if (fp_mode_24b) {
+    fldcw(ExternalAddress(StubRoutines::addr_fpu_cntrl_wrd_24()));
+  }
+  if (UseSSE >= 2 && VerifyFPU) {
+    verify_FPU(0, "FPU stack must be clean on entry");
+  }
+#endif
+
+#ifdef ASSERT
+  if (VerifyStackAtCalls) {
+    Label L;
+    push(rax);
+    mov(rax, rsp);
+    andptr(rax, StackAlignmentInBytes-1);
+    cmpptr(rax, StackAlignmentInBytes-wordSize);
+    pop(rax);
+    jcc(Assembler::equal, L);
+    stop("Stack is not properly aligned!");
+    bind(L);
+  }
+#endif
+
+}
+
+
 // IndexOf for constant substrings with size >= 8 chars
 // which don't need to be loaded through stack.
 void MacroAssembler::string_indexofC8(Register str1, Register str2,
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/assembler_x86.hpp
--- a/src/cpu/x86/vm/assembler_x86.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/assembler_x86.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -667,6 +667,8 @@
   void emit_arith_b(int op1, int op2, Register dst, int imm8);
 
   void emit_arith(int op1, int op2, Register dst, int32_t imm32);
+  // Force generation of a 4 byte immediate value even if it fits into 8bit
+  void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32);
   // only 32bit??
   void emit_arith(int op1, int op2, Register dst, jobject obj);
   void emit_arith(int op1, int op2, Register dst, Register src);
@@ -1526,6 +1528,9 @@
   void subq(Register dst, Address src);
   void subq(Register dst, Register src);
 
+  // Force generation of a 4 byte immediate value even if it fits into 8bit
+  void subl_imm32(Register dst, int32_t imm32);
+  void subq_imm32(Register dst, int32_t imm32);
 
   // Subtract Scalar Double-Precision Floating-Point Values
   void subsd(XMMRegister dst, Address src);
@@ -1763,8 +1768,8 @@
   // Alignment
   void align(int modulus);
 
-  // Misc
-  void fat_nop(); // 5 byte nop
+  // A 5 byte nop that is safe for patching (see patch_verified_entry)
+  void fat_nop();
 
   // Stack frame creation/removal
   void enter();
@@ -2275,6 +2280,8 @@
 
   void subptr(Register dst, Address src) { LP64_ONLY(subq(dst, src)) NOT_LP64(subl(dst, src)); }
   void subptr(Register dst, int32_t src);
+  // Force generation of a 4 byte immediate value even if it fits into 8bit
+  void subptr_imm32(Register dst, int32_t src);
   void subptr(Register dst, Register src);
   void subptr(Register dst, RegisterOrConstant src) {
     if (src.is_constant()) subptr(dst, (int) src.as_constant());
@@ -2566,6 +2573,9 @@
   void movl2ptr(Register dst, Address src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(movl(dst, src)); }
   void movl2ptr(Register dst, Register src) { LP64_ONLY(movslq(dst, src)) NOT_LP64(if (dst != src) movl(dst, src)); }
 
+  // C2 compiled method's prolog code.
+  void verified_entry(int framesize, bool stack_bang, bool fp_mode_24b);
+
   // IndexOf strings.
   // Small strings are loaded through stack if they cross page boundary.
   void string_indexof(Register str1, Register str2,
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/c1_MacroAssembler_x86.cpp
--- a/src/cpu/x86/vm/c1_MacroAssembler_x86.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/c1_MacroAssembler_x86.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -381,6 +381,16 @@
 
 
 void C1_MacroAssembler::verified_entry() {
+  if (C1Breakpoint || VerifyFPU || !UseStackBanging) {
+    // Verified Entry first instruction should be 5 bytes long for correct
+    // patching by patch_verified_entry().
+    //
+    // C1Breakpoint and VerifyFPU have one byte first instruction.
+    // Also first instruction will be one byte "push(rbp)" if stack banging
+    // code is not generated (see build_frame() above).
+    // For all these cases generate long instruction first.
+    fat_nop();
+  }
   if (C1Breakpoint)int3();
   // build frame
   verify_FPU(0, "method_entry");
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/globals_x86.hpp
--- a/src/cpu/x86/vm/globals_x86.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/globals_x86.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -62,7 +62,7 @@
 // due to lack of optimization caused by C++ compiler bugs
 define_pd_global(intx, StackShadowPages, NOT_WIN64(20) WIN64_ONLY(6) DEBUG_ONLY(+2));
 #else
-define_pd_global(intx, StackShadowPages, 3 DEBUG_ONLY(+5));
+define_pd_global(intx, StackShadowPages, 4 DEBUG_ONLY(+5));
 #endif // AMD64
 
 define_pd_global(intx, PreInflateSpin,           10);
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/methodHandles_x86.cpp
--- a/src/cpu/x86/vm/methodHandles_x86.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/methodHandles_x86.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -1018,41 +1018,26 @@
 void trace_method_handle_stub(const char* adaptername,
                               oop mh,
                               intptr_t* saved_regs,
-                              intptr_t* entry_sp,
-                              intptr_t* saved_sp,
-                              intptr_t* saved_bp) {
+                              intptr_t* entry_sp) {
   // called as a leaf from native code: do not block the JVM!
   bool has_mh = (strstr(adaptername, "return/") == NULL);  // return adapters don't have rcx_mh
+  const char* mh_reg_name = has_mh ? "rcx_mh" : "rcx";
+  tty->print_cr("MH %s %s="PTR_FORMAT" sp="PTR_FORMAT, adaptername, mh_reg_name, mh, entry_sp);
 
-  intptr_t* last_sp = (intptr_t*) saved_bp[frame::interpreter_frame_last_sp_offset];
-  intptr_t* base_sp = last_sp;
-  typedef MethodHandles::RicochetFrame RicochetFrame;
-  RicochetFrame* rfp = (RicochetFrame*)((address)saved_bp - RicochetFrame::sender_link_offset_in_bytes());
-  if (Universe::heap()->is_in((address) rfp->saved_args_base())) {
-    // Probably an interpreter frame.
-    base_sp = (intptr_t*) saved_bp[frame::interpreter_frame_monitor_block_top_offset];
-  }
-  intptr_t    mh_reg = (intptr_t)mh;
-  const char* mh_reg_name = "rcx_mh";
-  if (!has_mh)  mh_reg_name = "rcx";
-  tty->print_cr("MH %s %s="PTR_FORMAT" sp=("PTR_FORMAT"+"INTX_FORMAT") stack_size="INTX_FORMAT" bp="PTR_FORMAT,
-                adaptername, mh_reg_name, mh_reg,
-                (intptr_t)entry_sp, (intptr_t)(saved_sp - entry_sp), (intptr_t)(base_sp - last_sp), (intptr_t)saved_bp);
   if (Verbose) {
-    tty->print(" reg dump: ");
-    int saved_regs_count = (entry_sp-1) - saved_regs;
-    // 32 bit: rdi rsi rbp rsp; rbx rdx rcx (*) rax
-    int i;
-    for (i = 0; i <= saved_regs_count; i++) {
-      if (i > 0 && i % 4 == 0 && i != saved_regs_count) {
+    tty->print_cr("Registers:");
+    const int saved_regs_count = RegisterImpl::number_of_registers;
+    for (int i = 0; i < saved_regs_count; i++) {
+      Register r = as_Register(i);
+      // The registers are stored in reverse order on the stack (by pusha).
+      tty->print("%3s=" PTR_FORMAT, r->name(), saved_regs[((saved_regs_count - 1) - i)]);
+      if ((i + 1) % 4 == 0) {
         tty->cr();
-        tty->print("   + dump: ");
+      } else {
+        tty->print(", ");
       }
-      tty->print(" %d: "PTR_FORMAT, i, saved_regs[i]);
     }
     tty->cr();
-    if (last_sp != saved_sp && last_sp != NULL)
-      tty->print_cr("*** last_sp="PTR_FORMAT, (intptr_t)last_sp);
 
     {
      // dumping last frame with frame::describe
@@ -1102,14 +1087,7 @@
         values.describe(-1, dump_sp, "sp for #1");
       }
 
-      // mark saved_sp if seems valid
-      if (has_mh) {
-        if ((saved_sp >= dump_sp - UNREASONABLE_STACK_MOVE) && (saved_sp < dump_fp)) {
-          values.describe(-1, saved_sp, "*saved_sp");
-        }
-      }
-
-      tty->print_cr("  stack layout:");
+      tty->print_cr("Stack layout:");
       values.print(p);
     }
     if (has_mh)
@@ -1125,16 +1103,12 @@
   oopDesc* mh;
   intptr_t* saved_regs;
   intptr_t* entry_sp;
-  intptr_t* saved_sp;
-  intptr_t* saved_bp;
 };
 void trace_method_handle_stub_wrapper(MethodHandleStubArguments* args) {
   trace_method_handle_stub(args->adaptername,
                            args->mh,
                            args->saved_regs,
-                           args->entry_sp,
-                           args->saved_sp,
-                           args->saved_bp);
+                           args->entry_sp);
 }
 
 void MethodHandles::trace_method_handle(MacroAssembler* _masm, const char* adaptername) {
@@ -1157,20 +1131,18 @@
     __ fst_d(Address(rsp, 0));
   }
 
-  // incoming state:
+  // Incoming state:
   // rcx: method handle
-  // r13 or rsi: saved sp
-  // To avoid calling convention issues, build a record on the stack and pass the pointer to that instead.
-  // Note: fix the increment below if pushing more arguments
-  __ push(rbp);               // saved_bp
-  __ push(saved_last_sp_register()); // saved_sp
+  //
+  // To avoid calling convention issues, build a record on the stack
+  // and pass the pointer to that instead.
   __ push(rbp);               // entry_sp (with extra align space)
   __ push(rbx);               // pusha saved_regs
   __ push(rcx);               // mh
   __ push(rcx);               // slot for adaptername
   __ movptr(Address(rsp, 0), (intptr_t) adaptername);
   __ super_call_VM_leaf(CAST_FROM_FN_PTR(address, trace_method_handle_stub_wrapper), rsp);
-  __ increment(rsp, 6 * wordSize); // MethodHandleStubArguments
+  __ increment(rsp, sizeof(MethodHandleStubArguments));
 
   if  (UseSSE >= 2) {
     __ movdbl(xmm0, Address(rsp, 0));
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/stubGenerator_x86_64.cpp
--- a/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/stubGenerator_x86_64.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -2997,7 +2997,7 @@
     // Generate oop map
     OopMap* map = new OopMap(framesize, 0);
 
-    oop_maps->add_gc_map(__ pc() - start, map);
+    oop_maps->add_gc_map(the_pc - start, map);
 
     __ reset_last_Java_frame(true, true);
 
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/x86.ad
--- a/src/cpu/x86/vm/x86.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/x86.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2011, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2011, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -37,10 +37,87 @@
   static address double_signmask() { return (address)double_signmask_pool; }
   static address double_signflip() { return (address)double_signflip_pool; }
 #endif
+
+#ifndef PRODUCT
+  void MachNopNode::format(PhaseRegAlloc*, outputStream* st) const {
+    st->print("nop \t# %d bytes pad for loops and calls", _count);
+  }
+#endif
+
+  void MachNopNode::emit(CodeBuffer &cbuf, PhaseRegAlloc*) const {
+    MacroAssembler _masm(&cbuf);
+    __ nop(_count);
+  }
+
+  uint MachNopNode::size(PhaseRegAlloc*) const {
+    return _count;
+  }
+
+#ifndef PRODUCT
+  void MachBreakpointNode::format(PhaseRegAlloc*, outputStream* st) const {
+    st->print("# breakpoint");
+  }
+#endif
+
+  void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc* ra_) const {
+    MacroAssembler _masm(&cbuf);
+    __ int3();
+  }
+
+  uint MachBreakpointNode::size(PhaseRegAlloc* ra_) const {
+    return MachNode::size(ra_);
+  }
+
+%}
+
+encode %{
+
+  enc_class preserve_SP %{
+    debug_only(int off0 = cbuf.insts_size());
+    MacroAssembler _masm(&cbuf);
+    // RBP is preserved across all calls, even compiled calls.
+    // Use it to preserve RSP in places where the callee might change the SP.
+    __ movptr(rbp_mh_SP_save, rsp);
+    debug_only(int off1 = cbuf.insts_size());
+    assert(off1 - off0 == preserve_SP_size(), "correct size prediction");
+  %}
+
+  enc_class restore_SP %{
+    MacroAssembler _masm(&cbuf);
+    __ movptr(rsp, rbp_mh_SP_save);
+  %}
+
+  enc_class call_epilog %{
+    if (VerifyStackAtCalls) {
+      // Check that stack depth is unchanged: find majik cookie on stack
+      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
+      MacroAssembler _masm(&cbuf);
+      Label L;
+      __ cmpptr(Address(rsp, framesize), (int32_t)0xbadb100d);
+      __ jccb(Assembler::equal, L);
+      // Die if stack mismatch
+      __ int3();
+      __ bind(L);
+    }
+  %}
+
 %}
 
 // INSTRUCTIONS -- Platform independent definitions (same for 32- and 64-bit)
 
+// ============================================================================
+
+instruct ShouldNotReachHere() %{
+  match(Halt);
+  format %{ "int3\t# ShouldNotReachHere" %}
+  ins_encode %{
+    __ int3();
+  %}
+  ins_pipe(pipe_slow);
+%}
+
+// ============================================================================
+
 instruct addF_reg(regF dst, regF src) %{
   predicate((UseSSE>=1) && (UseAVX == 0));
   match(Set dst (AddF dst src));
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/x86_32.ad
--- a/src/cpu/x86/vm/x86_32.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/x86_32.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 1997, 2011, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 1997, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -341,12 +341,6 @@
   return round_to(current_offset, alignment_required()) - current_offset;
 }
 
-#ifndef PRODUCT
-void MachBreakpointNode::format( PhaseRegAlloc *, outputStream* st ) const {
-  st->print("INT3");
-}
-#endif
-
 // EMIT_RM()
 void emit_rm(CodeBuffer &cbuf, int f1, int f2, int f3) {
   unsigned char c = (unsigned char)((f1 << 6) | (f2 << 3) | f3);
@@ -550,118 +544,66 @@
 
 //=============================================================================
 #ifndef PRODUCT
-void MachPrologNode::format( PhaseRegAlloc *ra_, outputStream* st ) const {
+void MachPrologNode::format(PhaseRegAlloc* ra_, outputStream* st) const {
   Compile* C = ra_->C;
-  if( C->in_24_bit_fp_mode() ) {
-    st->print("FLDCW  24 bit fpu control word");
-    st->print_cr(""); st->print("\t");
-  }
 
   int framesize = C->frame_slots() << LogBytesPerInt;
   assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
-  // Remove two words for return addr and rbp,
-  framesize -= 2*wordSize;
-
-  // Calls to C2R adapters often do not accept exceptional returns.
-  // We require that their callers must bang for them.  But be careful, because
-  // some VM calls (such as call site linkage) can use several kilobytes of
-  // stack.  But the stack safety zone should account for that.
-  // See bugs 4446381, 4468289, 4497237.
+  // Remove wordSize for return addr which is already pushed.
+  framesize -= wordSize;
+
   if (C->need_stack_bang(framesize)) {
-    st->print_cr("# stack bang"); st->print("\t");
-  }
-  st->print_cr("PUSHL  EBP"); st->print("\t");
-
-  if( VerifyStackAtCalls ) { // Majik cookie to verify stack depth
-    st->print("PUSH   0xBADB100D\t# Majik cookie for stack depth check");
-    st->print_cr(""); st->print("\t");
     framesize -= wordSize;
-  }
-
-  if ((C->in_24_bit_fp_mode() || VerifyStackAtCalls ) && framesize < 128 ) {
+    st->print("# stack bang");
+    st->print("\n\t");
+    st->print("PUSH   EBP\t# Save EBP");
     if (framesize) {
-      st->print("SUB    ESP,%d\t# Create frame",framesize);
+      st->print("\n\t");
+      st->print("SUB    ESP, #%d\t# Create frame",framesize);
     }
   } else {
-    st->print("SUB    ESP,%d\t# Create frame",framesize);
+    st->print("SUB    ESP, #%d\t# Create frame",framesize);
+    st->print("\n\t");
+    framesize -= wordSize;
+    st->print("MOV    [ESP + #%d], EBP\t# Save EBP",framesize);
+  }
+
+  if (VerifyStackAtCalls) {
+    st->print("\n\t");
+    framesize -= wordSize;
+    st->print("MOV    [ESP + #%d], 0xBADB100D\t# Majik cookie for stack depth check",framesize);
   }
+
+  if( C->in_24_bit_fp_mode() ) {
+    st->print("\n\t");
+    st->print("FLDCW  \t# load 24 bit fpu control word");
+  }
+  if (UseSSE >= 2 && VerifyFPU) {
+    st->print("\n\t");
+    st->print("# verify FPU stack (must be clean on entry)");
+  }
+
+#ifdef ASSERT
+  if (VerifyStackAtCalls) {
+    st->print("\n\t");
+    st->print("# stack alignment check");
+  }
+#endif
+  st->cr();
 }
 #endif
 
 
 void MachPrologNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
   Compile* C = ra_->C;
-
-  if (UseSSE >= 2 && VerifyFPU) {
-    MacroAssembler masm(&cbuf);
-    masm.verify_FPU(0, "FPU stack must be clean on entry");
-  }
-
-  // WARNING: Initial instruction MUST be 5 bytes or longer so that
-  // NativeJump::patch_verified_entry will be able to patch out the entry
-  // code safely. The fldcw is ok at 6 bytes, the push to verify stack
-  // depth is ok at 5 bytes, the frame allocation can be either 3 or
-  // 6 bytes. So if we don't do the fldcw or the push then we must
-  // use the 6 byte frame allocation even if we have no frame. :-(
-  // If method sets FPU control word do it now
-  if( C->in_24_bit_fp_mode() ) {
-    MacroAssembler masm(&cbuf);
-    masm.fldcw(ExternalAddress(StubRoutines::addr_fpu_cntrl_wrd_24()));
-  }
+  MacroAssembler _masm(&cbuf);
 
   int framesize = C->frame_slots() << LogBytesPerInt;
-  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
-  // Remove two words for return addr and rbp,
-  framesize -= 2*wordSize;
-
-  // Calls to C2R adapters often do not accept exceptional returns.
-  // We require that their callers must bang for them.  But be careful, because
-  // some VM calls (such as call site linkage) can use several kilobytes of
-  // stack.  But the stack safety zone should account for that.
-  // See bugs 4446381, 4468289, 4497237.
-  if (C->need_stack_bang(framesize)) {
-    MacroAssembler masm(&cbuf);
-    masm.generate_stack_overflow_check(framesize);
-  }
-
-  // We always push rbp, so that on return to interpreter rbp, will be
-  // restored correctly and we can correct the stack.
-  emit_opcode(cbuf, 0x50 | EBP_enc);
-
-  if( VerifyStackAtCalls ) { // Majik cookie to verify stack depth
-    emit_opcode(cbuf, 0x68); // push 0xbadb100d
-    emit_d32(cbuf, 0xbadb100d);
-    framesize -= wordSize;
-  }
-
-  if ((C->in_24_bit_fp_mode() || VerifyStackAtCalls ) && framesize < 128 ) {
-    if (framesize) {
-      emit_opcode(cbuf, 0x83);   // sub  SP,#framesize
-      emit_rm(cbuf, 0x3, 0x05, ESP_enc);
-      emit_d8(cbuf, framesize);
-    }
-  } else {
-    emit_opcode(cbuf, 0x81);   // sub  SP,#framesize
-    emit_rm(cbuf, 0x3, 0x05, ESP_enc);
-    emit_d32(cbuf, framesize);
-  }
+
+  __ verified_entry(framesize, C->need_stack_bang(framesize), C->in_24_bit_fp_mode());
+
   C->set_frame_complete(cbuf.insts_size());
 
-#ifdef ASSERT
-  if (VerifyStackAtCalls) {
-    Label L;
-    MacroAssembler masm(&cbuf);
-    masm.push(rax);
-    masm.mov(rax, rsp);
-    masm.andptr(rax, StackAlignmentInBytes-1);
-    masm.cmpptr(rax, StackAlignmentInBytes-wordSize);
-    masm.pop(rax);
-    masm.jcc(Assembler::equal, L);
-    masm.stop("Stack is not properly aligned!");
-    masm.bind(L);
-  }
-#endif
-
   if (C->has_mach_constant_base_node()) {
     // NOTE: We set the table base offset here because users might be
     // emitted before MachConstantBaseNode.
@@ -1169,7 +1111,7 @@
 }
 
 #ifndef PRODUCT
-void MachSpillCopyNode::format( PhaseRegAlloc *ra_, outputStream* st ) const {
+void MachSpillCopyNode::format(PhaseRegAlloc *ra_, outputStream* st) const {
   implementation( NULL, ra_, false, st );
 }
 #endif
@@ -1182,22 +1124,6 @@
   return implementation( NULL, ra_, true, NULL );
 }
 
-//=============================================================================
-#ifndef PRODUCT
-void MachNopNode::format( PhaseRegAlloc *, outputStream* st ) const {
-  st->print("NOP \t# %d bytes pad for loops and calls", _count);
-}
-#endif
-
-void MachNopNode::emit(CodeBuffer &cbuf, PhaseRegAlloc * ) const {
-  MacroAssembler _masm(&cbuf);
-  __ nop(_count);
-}
-
-uint MachNopNode::size(PhaseRegAlloc *) const {
-  return _count;
-}
-
 
 //=============================================================================
 #ifndef PRODUCT
@@ -1883,21 +1809,6 @@
     }
   %}
 
-  enc_class preserve_SP %{
-    debug_only(int off0 = cbuf.insts_size());
-    MacroAssembler _masm(&cbuf);
-    // RBP is preserved across all calls, even compiled calls.
-    // Use it to preserve RSP in places where the callee might change the SP.
-    __ movptr(rbp_mh_SP_save, rsp);
-    debug_only(int off1 = cbuf.insts_size());
-    assert(off1 - off0 == preserve_SP_size(), "correct size prediction");
-  %}
-
-  enc_class restore_SP %{
-    MacroAssembler _masm(&cbuf);
-    __ movptr(rsp, rbp_mh_SP_save);
-  %}
-
   enc_class Java_Static_Call (method meth) %{    // JAVA STATIC CALL
     // CALL to fixup routine.  Fixup routine uses ScopeDesc info to determine
     // who we intended to call.
@@ -3846,9 +3757,9 @@
   // Ret Addr is on stack in slot 0 if no locks or verification or alignment.
   // Otherwise, it is above the locks and verification slot and alignment word
   return_addr(STACK - 1 +
-              round_to(1+VerifyStackAtCalls+
-              Compile::current()->fixed_slots(),
-              (StackAlignmentInBytes/wordSize)));
+              round_to((Compile::current()->in_preserve_stack_slots() +
+                        Compile::current()->fixed_slots()),
+                       stack_alignment_in_slots()));
 
   // Body of function which returns an integer array locating
   // arguments either in registers or in stack slots.  Passed an array
@@ -13476,6 +13387,25 @@
   ins_pipe( ialu_reg_mem );
 %}
 
+
+// ============================================================================
+// This name is KNOWN by the ADLC and cannot be changed.
+// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
+// for this guy.
+instruct tlsLoadP(eRegP dst, eFlagsReg cr) %{
+  match(Set dst (ThreadLocal));
+  effect(DEF dst, KILL cr);
+
+  format %{ "MOV    $dst, Thread::current()" %}
+  ins_encode %{
+    Register dstReg = as_Register($dst$$reg);
+    __ get_thread(dstReg);
+  %}
+  ins_pipe( ialu_reg_fat );
+%}
+
+
+
 //----------PEEPHOLE RULES-----------------------------------------------------
 // These must follow all instruction definitions as they use the names
 // defined in the instructions definitions.
diff -r d3384450b649 -r 15085a6eb50c src/cpu/x86/vm/x86_64.ad
--- a/src/cpu/x86/vm/x86_64.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/cpu/x86/vm/x86_64.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2003, 2011, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -610,13 +610,6 @@
   return round_to(current_offset, alignment_required()) - current_offset;
 }
 
-#ifndef PRODUCT
-void MachBreakpointNode::format(PhaseRegAlloc*, outputStream* st) const
-{
-  st->print("INT3");
-}
-#endif
-
 // EMIT_RM()
 void emit_rm(CodeBuffer &cbuf, int f1, int f2, int f3) {
   unsigned char c = (unsigned char) ((f1 << 6) | (f2 << 3) | f3);
@@ -853,121 +846,53 @@
 
 //=============================================================================
 #ifndef PRODUCT
-void MachPrologNode::format(PhaseRegAlloc* ra_, outputStream* st) const
-{
+void MachPrologNode::format(PhaseRegAlloc* ra_, outputStream* st) const {
   Compile* C = ra_->C;
 
   int framesize = C->frame_slots() << LogBytesPerInt;
   assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
-  // Remove wordSize for return adr already pushed
-  // and another for the RBP we are going to save
-  framesize -= 2*wordSize;
-  bool need_nop = true;
-
-  // Calls to C2R adapters often do not accept exceptional returns.
-  // We require that their callers must bang for them.  But be
-  // careful, because some VM calls (such as call site linkage) can
-  // use several kilobytes of stack.  But the stack safety zone should
-  // account for that.  See bugs 4446381, 4468289, 4497237.
+  // Remove wordSize for return addr which is already pushed.
+  framesize -= wordSize;
+
   if (C->need_stack_bang(framesize)) {
-    st->print_cr("# stack bang"); st->print("\t");
-    need_nop = false;
+    framesize -= wordSize;
+    st->print("# stack bang");
+    st->print("\n\t");
+    st->print("pushq   rbp\t# Save rbp");
+    if (framesize) {
+      st->print("\n\t");
+      st->print("subq    rsp, #%d\t# Create frame",framesize);
+    }
+  } else {
+    st->print("subq    rsp, #%d\t# Create frame",framesize);
+    st->print("\n\t");
+    framesize -= wordSize;
+    st->print("movq    [rsp + #%d], rbp\t# Save rbp",framesize);
   }
-  st->print_cr("pushq   rbp"); st->print("\t");
 
   if (VerifyStackAtCalls) {
-    // Majik cookie to verify stack depth
-    st->print_cr("pushq   0xffffffffbadb100d"
-                  "\t# Majik cookie for stack depth check");
-    st->print("\t");
-    framesize -= wordSize; // Remove 2 for cookie
-    need_nop = false;
+    st->print("\n\t");
+    framesize -= wordSize;
+    st->print("movq    [rsp + #%d], 0xbadb100d\t# Majik cookie for stack depth check",framesize);
+#ifdef ASSERT
+    st->print("\n\t");
+    st->print("# stack alignment check");
+#endif
   }
-
-  if (framesize) {
-    st->print("subq    rsp, #%d\t# Create frame", framesize);
-    if (framesize < 0x80 && need_nop) {
-      st->print("\n\tnop\t# nop for patch_verified_entry");
-    }
-  }
+  st->cr();
 }
 #endif
 
-void MachPrologNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const
-{
+void MachPrologNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
   Compile* C = ra_->C;
-
-  // WARNING: Initial instruction MUST be 5 bytes or longer so that
-  // NativeJump::patch_verified_entry will be able to patch out the entry
-  // code safely. The fldcw is ok at 6 bytes, the push to verify stack
-  // depth is ok at 5 bytes, the frame allocation can be either 3 or
-  // 6 bytes. So if we don't do the fldcw or the push then we must
-  // use the 6 byte frame allocation even if we have no frame. :-(
-  // If method sets FPU control word do it now
+  MacroAssembler _masm(&cbuf);
 
   int framesize = C->frame_slots() << LogBytesPerInt;
-  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
-  // Remove wordSize for return adr already pushed
-  // and another for the RBP we are going to save
-  framesize -= 2*wordSize;
-  bool need_nop = true;
-
-  // Calls to C2R adapters often do not accept exceptional returns.
-  // We require that their callers must bang for them.  But be
-  // careful, because some VM calls (such as call site linkage) can
-  // use several kilobytes of stack.  But the stack safety zone should
-  // account for that.  See bugs 4446381, 4468289, 4497237.
-  if (C->need_stack_bang(framesize)) {
-    MacroAssembler masm(&cbuf);
-    masm.generate_stack_overflow_check(framesize);
-    need_nop = false;
-  }
-
-  // We always push rbp so that on return to interpreter rbp will be
-  // restored correctly and we can correct the stack.
-  emit_opcode(cbuf, 0x50 | RBP_enc);
-
-  if (VerifyStackAtCalls) {
-    // Majik cookie to verify stack depth
-    emit_opcode(cbuf, 0x68); // pushq (sign-extended) 0xbadb100d
-    emit_d32(cbuf, 0xbadb100d);
-    framesize -= wordSize; // Remove 2 for cookie
-    need_nop = false;
-  }
-
-  if (framesize) {
-    emit_opcode(cbuf, Assembler::REX_W);
-    if (framesize < 0x80) {
-      emit_opcode(cbuf, 0x83);   // sub  SP,#framesize
-      emit_rm(cbuf, 0x3, 0x05, RSP_enc);
-      emit_d8(cbuf, framesize);
-      if (need_nop) {
-        emit_opcode(cbuf, 0x90); // nop
-      }
-    } else {
-      emit_opcode(cbuf, 0x81);   // sub  SP,#framesize
-      emit_rm(cbuf, 0x3, 0x05, RSP_enc);
-      emit_d32(cbuf, framesize);
-    }
-  }
+
+  __ verified_entry(framesize, C->need_stack_bang(framesize), false);
 
   C->set_frame_complete(cbuf.insts_size());
 
-#ifdef ASSERT
-  if (VerifyStackAtCalls) {
-    Label L;
-    MacroAssembler masm(&cbuf);
-    masm.push(rax);
-    masm.mov(rax, rsp);
-    masm.andptr(rax, StackAlignmentInBytes-1);
-    masm.cmpptr(rax, StackAlignmentInBytes-wordSize);
-    masm.pop(rax);
-    masm.jcc(Assembler::equal, L);
-    masm.stop("Stack is not properly aligned!");
-    masm.bind(L);
-  }
-#endif
-
   if (C->has_mach_constant_base_node()) {
     // NOTE: We set the table base offset here because users might be
     // emitted before MachConstantBaseNode.
@@ -1598,26 +1523,6 @@
 
 //=============================================================================
 #ifndef PRODUCT
-void MachNopNode::format(PhaseRegAlloc*, outputStream* st) const
-{
-  st->print("nop \t# %d bytes pad for loops and calls", _count);
-}
-#endif
-
-void MachNopNode::emit(CodeBuffer &cbuf, PhaseRegAlloc*) const
-{
-  MacroAssembler _masm(&cbuf);
-  __ nop(_count);
-}
-
-uint MachNopNode::size(PhaseRegAlloc*) const
-{
-  return _count;
-}
-
-
-//=============================================================================
-#ifndef PRODUCT
 void BoxLockNode::format(PhaseRegAlloc* ra_, outputStream* st) const
 {
   int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
@@ -2323,21 +2228,6 @@
                    RELOC_DISP32);
   %}
 
-  enc_class preserve_SP %{
-    debug_only(int off0 = cbuf.insts_size());
-    MacroAssembler _masm(&cbuf);
-    // RBP is preserved across all calls, even compiled calls.
-    // Use it to preserve RSP in places where the callee might change the SP.
-    __ movptr(rbp_mh_SP_save, rsp);
-    debug_only(int off1 = cbuf.insts_size());
-    assert(off1 - off0 == preserve_SP_size(), "correct size prediction");
-  %}
-
-  enc_class restore_SP %{
-    MacroAssembler _masm(&cbuf);
-    __ movptr(rsp, rbp_mh_SP_save);
-  %}
-
   enc_class Java_Static_Call(method meth)
   %{
     // JAVA STATIC CALL
@@ -3276,9 +3166,9 @@
   // Ret Addr is on stack in slot 0 if no locks or verification or alignment.
   // Otherwise, it is above the locks and verification slot and alignment word
   return_addr(STACK - 2 +
-              round_to(2 + 2 * VerifyStackAtCalls +
-                       Compile::current()->fixed_slots(),
-                       WordsPerLong * 2));
+              round_to((Compile::current()->in_preserve_stack_slots() +
+                        Compile::current()->fixed_slots()),
+                       stack_alignment_in_slots()));
 
   // Body of function which returns an integer array locating
   // arguments either in registers or in stack slots.  Passed an array
@@ -11736,6 +11626,21 @@
 %}
 
 
+// ============================================================================
+// This name is KNOWN by the ADLC and cannot be changed.
+// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
+// for this guy.
+instruct tlsLoadP(r15_RegP dst) %{
+  match(Set dst (ThreadLocal));
+  effect(DEF dst);
+
+  size(0);
+  format %{ "# TLS is in R15" %}
+  ins_encode( /*empty encoding*/ );
+  ins_pipe(ialu_reg_reg);
+%}
+
+
 //----------PEEPHOLE RULES-----------------------------------------------------
 // These must follow all instruction definitions as they use the names
 // defined in the instructions definitions.
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/bsd_x86/vm/bsd_x86_32.ad
--- a/src/os_cpu/bsd_x86/vm/bsd_x86_32.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/bsd_x86/vm/bsd_x86_32.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 1999, 2008, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -24,137 +24,3 @@
 
 // X86 Bsd Architecture Description File
 
-//----------OS-DEPENDENT ENCODING BLOCK-----------------------------------------------------
-// This block specifies the encoding classes used by the compiler to output
-// byte streams.  Encoding classes generate functions which are called by
-// Machine Instruction Nodes in order to generate the bit encoding of the
-// instruction.  Operands specify their base encoding interface with the
-// interface keyword.  There are currently supported four interfaces,
-// REG_INTER, CONST_INTER, MEMORY_INTER, & COND_INTER.  REG_INTER causes an
-// operand to generate a function which returns its register number when
-// queried.   CONST_INTER causes an operand to generate a function which
-// returns the value of the constant when queried.  MEMORY_INTER causes an
-// operand to generate four functions which return the Base Register, the
-// Index Register, the Scale Value, and the Offset Value of the operand when
-// queried.  COND_INTER causes an operand to generate six functions which
-// return the encoding code (ie - encoding bits for the instruction)
-// associated with each basic boolean condition for a conditional instruction.
-// Instructions specify two basic values for encoding.  They use the
-// ins_encode keyword to specify their encoding class (which must be one of
-// the class names specified in the encoding block), and they use the
-// opcode keyword to specify, in order, their primary, secondary, and
-// tertiary opcode.  Only the opcode sections which a particular instruction
-// needs for encoding need to be specified.
-encode %{
-  // Build emit functions for each basic byte or larger field in the intel
-  // encoding scheme (opcode, rm, sib, immediate), and call them from C++
-  // code in the enc_class source block.  Emit functions will live in the
-  // main source block for now.  In future, we can generalize this by
-  // adding a syntax that specifies the sizes of fields in an order,
-  // so that the adlc can build the emit functions automagically
-
-  enc_class bsd_tlsencode (eRegP dst) %{
-    Register dstReg = as_Register($dst$$reg);
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-      masm->get_thread(dstReg);
-  %}
-
-  enc_class bsd_breakpoint  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-  %}
-
-  enc_class call_epilog %{
-    if( VerifyStackAtCalls ) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP,-3*VMRegImpl::slots_per_word));
-      if(framesize >= 128) {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0xBC);
-        emit_d8(cbuf,0x24);
-        emit_d32(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      else {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0x7C);
-        emit_d8(cbuf,0x24);
-        emit_d8(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf,0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(eRegP dst, eFlagsReg cr) %{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst, KILL cr);
-
-  format %{ "MOV    $dst, Thread::current()" %}
-  ins_encode( bsd_tlsencode(dst) );
-  ins_pipe( ialu_reg_fat );
-%}
-
-instruct TLS(eRegP dst) %{
-  match(Set dst (ThreadLocal));
-
-  expand %{
-    tlsLoadP(dst);
-  %}
-%}
-
-// Die now
-instruct ShouldNotReachHere( )
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "INT3   ; ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(bsd_breakpoint);
-  ins_pipe( pipe_slow );
-%}
-
-
-
-// Platform dependent source
-
-source %{
-
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer &cbuf) {
-
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
-  emit_break(cbuf);
-}
-
-
-uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
-  return 5;
-}
-
-%}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/bsd_x86/vm/bsd_x86_64.ad
--- a/src/os_cpu/bsd_x86/vm/bsd_x86_64.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/bsd_x86/vm/bsd_x86_64.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2003, 2006, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -55,8 +55,7 @@
   // adding a syntax that specifies the sizes of fields in an order,
   // so that the adlc can build the emit functions automagically
 
-  enc_class Java_To_Runtime(method meth)
-  %{
+  enc_class Java_To_Runtime(method meth) %{
     // No relocation needed
 
     // movq r10, <meth>
@@ -70,104 +69,15 @@
     emit_opcode(cbuf, 0xD0 | (R10_enc - 8));
   %}
 
-  enc_class bsd_breakpoint
-  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-  %}
-
-  enc_class call_epilog
-  %{
-    if (VerifyStackAtCalls) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize =
-        ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
-      if (framesize) {
-        if (framesize < 0x80) {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0x7C);
-          emit_d8(cbuf, 0x24);
-          emit_d8(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        } else {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0xBC);
-          emit_d8(cbuf, 0x24);
-          emit_d32(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        }
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf, 0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(r15_RegP dst)
-%{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst);
-
-  size(0);
-  format %{ "# TLS is in R15" %}
-  ins_encode( /*empty encoding*/ );
-  ins_pipe(ialu_reg_reg);
-%}
-
-// Die now
-instruct ShouldNotReachHere()
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "int3\t# ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(bsd_breakpoint);
-  ins_pipe(pipe_slow);
 %}
 
 
 // Platform dependent source
 
-source
-%{
+source %{
 
 int MachCallRuntimeNode::ret_addr_offset() {
   return 13; // movq r10,#addr; callq (r10)
 }
 
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer& cbuf) {
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const {
-  emit_break(cbuf);
-}
-
-uint MachBreakpointNode::size(PhaseRegAlloc* ra_) const {
-  return 5;
-}
-
 %}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/linux_x86/vm/linux_x86_32.ad
--- a/src/os_cpu/linux_x86/vm/linux_x86_32.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/linux_x86/vm/linux_x86_32.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 1999, 2008, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -24,137 +24,3 @@
 
 // X86 Linux Architecture Description File
 
-//----------OS-DEPENDENT ENCODING BLOCK-----------------------------------------------------
-// This block specifies the encoding classes used by the compiler to output
-// byte streams.  Encoding classes generate functions which are called by
-// Machine Instruction Nodes in order to generate the bit encoding of the
-// instruction.  Operands specify their base encoding interface with the
-// interface keyword.  There are currently supported four interfaces,
-// REG_INTER, CONST_INTER, MEMORY_INTER, & COND_INTER.  REG_INTER causes an
-// operand to generate a function which returns its register number when
-// queried.   CONST_INTER causes an operand to generate a function which
-// returns the value of the constant when queried.  MEMORY_INTER causes an
-// operand to generate four functions which return the Base Register, the
-// Index Register, the Scale Value, and the Offset Value of the operand when
-// queried.  COND_INTER causes an operand to generate six functions which
-// return the encoding code (ie - encoding bits for the instruction)
-// associated with each basic boolean condition for a conditional instruction.
-// Instructions specify two basic values for encoding.  They use the
-// ins_encode keyword to specify their encoding class (which must be one of
-// the class names specified in the encoding block), and they use the
-// opcode keyword to specify, in order, their primary, secondary, and
-// tertiary opcode.  Only the opcode sections which a particular instruction
-// needs for encoding need to be specified.
-encode %{
-  // Build emit functions for each basic byte or larger field in the intel
-  // encoding scheme (opcode, rm, sib, immediate), and call them from C++
-  // code in the enc_class source block.  Emit functions will live in the
-  // main source block for now.  In future, we can generalize this by
-  // adding a syntax that specifies the sizes of fields in an order,
-  // so that the adlc can build the emit functions automagically
-
-  enc_class linux_tlsencode (eRegP dst) %{
-    Register dstReg = as_Register($dst$$reg);
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-      masm->get_thread(dstReg);
-  %}
-
-  enc_class linux_breakpoint  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-  %}
-
-  enc_class call_epilog %{
-    if( VerifyStackAtCalls ) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP,-3*VMRegImpl::slots_per_word));
-      if(framesize >= 128) {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0xBC);
-        emit_d8(cbuf,0x24);
-        emit_d32(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      else {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0x7C);
-        emit_d8(cbuf,0x24);
-        emit_d8(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf,0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(eRegP dst, eFlagsReg cr) %{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst, KILL cr);
-
-  format %{ "MOV    $dst, Thread::current()" %}
-  ins_encode( linux_tlsencode(dst) );
-  ins_pipe( ialu_reg_fat );
-%}
-
-instruct TLS(eRegP dst) %{
-  match(Set dst (ThreadLocal));
-
-  expand %{
-    tlsLoadP(dst);
-  %}
-%}
-
-// Die now
-instruct ShouldNotReachHere( )
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "INT3   ; ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(linux_breakpoint);
-  ins_pipe( pipe_slow );
-%}
-
-
-
-// Platform dependent source
-
-source %{
-
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer &cbuf) {
-
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
-  emit_break(cbuf);
-}
-
-
-uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
-  return MachNode::size(ra_);
-}
-
-%}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/linux_x86/vm/linux_x86_64.ad
--- a/src/os_cpu/linux_x86/vm/linux_x86_64.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/linux_x86/vm/linux_x86_64.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2003, 2006, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -55,8 +55,7 @@
   // adding a syntax that specifies the sizes of fields in an order,
   // so that the adlc can build the emit functions automagically
 
-  enc_class Java_To_Runtime(method meth)
-  %{
+  enc_class Java_To_Runtime(method meth) %{
     // No relocation needed
 
     // movq r10, <meth>
@@ -70,105 +69,15 @@
     emit_opcode(cbuf, 0xD0 | (R10_enc - 8));
   %}
 
-  enc_class linux_breakpoint
-  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-  %}
-
-  enc_class call_epilog
-  %{
-    if (VerifyStackAtCalls) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize =
-        ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
-      if (framesize) {
-        if (framesize < 0x80) {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0x7C);
-          emit_d8(cbuf, 0x24);
-          emit_d8(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        } else {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0xBC);
-          emit_d8(cbuf, 0x24);
-          emit_d32(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        }
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf, 0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(r15_RegP dst)
-%{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst);
-
-  size(0);
-  format %{ "# TLS is in R15" %}
-  ins_encode( /*empty encoding*/ );
-  ins_pipe(ialu_reg_reg);
-%}
-
-// Die now
-instruct ShouldNotReachHere()
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "int3\t# ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(linux_breakpoint);
-  ins_pipe(pipe_slow);
 %}
 
 
 // Platform dependent source
 
-source
-%{
+source %{
 
 int MachCallRuntimeNode::ret_addr_offset() {
   return 13; // movq r10,#addr; callq (r10)
 }
 
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer& cbuf) {
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const {
-  emit_break(cbuf);
-}
-
-uint MachBreakpointNode::size(PhaseRegAlloc* ra_) const {
-  // distance could be far and requires load and call through register
-  return MachNode::size(ra_);
-}
-
 %}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/solaris_x86/vm/solaris_x86_32.ad
--- a/src/os_cpu/solaris_x86/vm/solaris_x86_32.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/solaris_x86/vm/solaris_x86_32.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 1999, 2008, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -24,144 +24,3 @@
 
 // X86 Solaris Architecture Description File
 
-//----------OS-DEPENDENT ENCODING BLOCK-----------------------------------------------------
-// This block specifies the encoding classes used by the compiler to output
-// byte streams.  Encoding classes generate functions which are called by
-// Machine Instruction Nodes in order to generate the bit encoding of the
-// instruction.  Operands specify their base encoding interface with the
-// interface keyword.  There are currently supported four interfaces,
-// REG_INTER, CONST_INTER, MEMORY_INTER, & COND_INTER.  REG_INTER causes an
-// operand to generate a function which returns its register number when
-// queried.   CONST_INTER causes an operand to generate a function which
-// returns the value of the constant when queried.  MEMORY_INTER causes an
-// operand to generate four functions which return the Base Register, the
-// Index Register, the Scale Value, and the Offset Value of the operand when
-// queried.  COND_INTER causes an operand to generate six functions which
-// return the encoding code (ie - encoding bits for the instruction)
-// associated with each basic boolean condition for a conditional instruction.
-// Instructions specify two basic values for encoding.  They use the
-// ins_encode keyword to specify their encoding class (which must be one of
-// the class names specified in the encoding block), and they use the
-// opcode keyword to specify, in order, their primary, secondary, and
-// tertiary opcode.  Only the opcode sections which a particular instruction
-// needs for encoding need to be specified.
-encode %{
-  // Build emit functions for each basic byte or larger field in the intel
-  // encoding scheme (opcode, rm, sib, immediate), and call them from C++
-  // code in the enc_class source block.  Emit functions will live in the
-  // main source block for now.  In future, we can generalize this by
-  // adding a syntax that specifies the sizes of fields in an order,
-  // so that the adlc can build the emit functions automagically
-
-  enc_class solaris_tlsencode (eRegP dst) %{
-    Register dstReg = as_Register($dst$$reg);
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->get_thread(dstReg);
-  %}
-
-  enc_class solaris_breakpoint  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    // Really need to fix this
-    masm->push(rax);
-    masm->push(rcx);
-    masm->push(rdx);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-    masm->pop(rdx);
-    masm->pop(rcx);
-    masm->pop(rax);
-  %}
-
-  enc_class call_epilog %{
-    if( VerifyStackAtCalls ) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP,-3*VMRegImpl::slots_per_word));
-      if(framesize >= 128) {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0xBC);
-        emit_d8(cbuf,0x24);
-        emit_d32(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      else {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0x7C);
-        emit_d8(cbuf,0x24);
-        emit_d8(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 11; // size of call to breakpoint (and register preserve), 1 for CC
-      emit_opcode(cbuf,0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(eRegP dst, eFlagsReg cr) %{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst, KILL cr);
-
-  format %{ "MOV    $dst, Thread::current()" %}
-  ins_encode( solaris_tlsencode(dst) );
-  ins_pipe( ialu_reg_fat );
-%}
-
-instruct TLS(eRegP dst) %{
-  match(Set dst (ThreadLocal));
-
-  expand %{
-    tlsLoadP(dst);
-  %}
-%}
-
-// Die now
-instruct ShouldNotReachHere( )
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "INT3   ; ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(solaris_breakpoint);
-  ins_pipe( pipe_slow );
-%}
-
-
-
-// Platform dependent source
-
-source %{
-
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer &cbuf) {
-
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
-  emit_break(cbuf);
-}
-
-
-uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
-  return MachNode::size(ra_);
-}
-
-%}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/solaris_x86/vm/solaris_x86_64.ad
--- a/src/os_cpu/solaris_x86/vm/solaris_x86_64.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/solaris_x86/vm/solaris_x86_64.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2004, 2006, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2004, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -55,8 +55,7 @@
   // adding a syntax that specifies the sizes of fields in an order,
   // so that the adlc can build the emit functions automagically
 
-  enc_class Java_To_Runtime(method meth)
-  %{
+  enc_class Java_To_Runtime(method meth) %{
     // No relocation needed
 
     // movq r10, <meth>
@@ -70,118 +69,24 @@
     emit_opcode(cbuf, 0xD0 | (R10_enc - 8));
   %}
 
-  enc_class solaris_breakpoint
-  %{
-    MacroAssembler* masm = new MacroAssembler(&cbuf);
-    masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-  %}
-
-  enc_class call_epilog
-  %{
-    if (VerifyStackAtCalls) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize =
-        ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
-      if (framesize) {
-        if (framesize < 0x80) {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0x7C);
-          emit_d8(cbuf, 0x24);
-          emit_d8(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        } else {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0xBC);
-          emit_d8(cbuf, 0x24);
-          emit_d32(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        }
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf, 0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
+  enc_class post_call_verify_mxcsr %{
+    MacroAssembler _masm(&cbuf);
+    if (RestoreMXCSROnJNICalls) {
+      __ ldmxcsr(ExternalAddress(StubRoutines::amd64::mxcsr_std()));
+    }
+    else if (CheckJNICalls) {
+      __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::amd64::verify_mxcsr_entry())));
     }
   %}
-
-  enc_class post_call_verify_mxcsr %{
-    MacroAssembler masm(&cbuf);
-    if (RestoreMXCSROnJNICalls) {
-      masm.ldmxcsr(ExternalAddress(StubRoutines::amd64::mxcsr_std()));
-    }
-    else if (CheckJNICalls) {
-      masm.call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::amd64::verify_mxcsr_entry())));
-    }
-  %}
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(r15_RegP dst)
-%{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst);
-
-  size(0);
-  format %{ "# TLS is in R15" %}
-  ins_encode( /*empty encoding*/ );
-  ins_pipe(ialu_reg_reg);
-%}
-
-// Die now
-instruct ShouldNotReachHere()
-%{
-  match(Halt);
-
-  // Use the following format syntax
-  format %{ "int3\t# ShouldNotReachHere" %}
-  // QQQ TODO for now call breakpoint
-  // opcode(0xCC);
-  // ins_encode(Opc);
-  ins_encode(solaris_breakpoint);
-  ins_pipe(pipe_slow);
 %}
 
 
 // Platform dependent source
 
-source
-%{
+source %{
 
-int MachCallRuntimeNode::ret_addr_offset()
-{
+int MachCallRuntimeNode::ret_addr_offset() {
   return 13; // movq r10,#addr; callq (r10)
 }
 
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer& cbuf)
-{
-  // Debugger doesn't really catch this but best we can do so far QQQ
-  MacroAssembler* masm = new MacroAssembler(&cbuf);
-  masm->call(RuntimeAddress(CAST_FROM_FN_PTR(address, os::breakpoint)));
-}
-
-void MachBreakpointNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const
-{
-  emit_break(cbuf);
-}
-
-uint MachBreakpointNode::size(PhaseRegAlloc* ra_) const
-{
-  // distance could be far and requires load and call through register
-  return MachNode::size(ra_);
-}
-
 %}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/windows_x86/vm/windows_x86_32.ad
--- a/src/os_cpu/windows_x86/vm/windows_x86_32.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/windows_x86/vm/windows_x86_32.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 1999, 2010, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 1999, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -24,134 +24,3 @@
 
 // X86 Win32 Architecture Description File
 
-//----------OS-DEPENDENT ENCODING BLOCK-----------------------------------------------------
-// This block specifies the encoding classes used by the compiler to output
-// byte streams.  Encoding classes generate functions which are called by
-// Machine Instruction Nodes in order to generate the bit encoding of the
-// instruction.  Operands specify their base encoding interface with the
-// interface keyword.  There are currently supported four interfaces,
-// REG_INTER, CONST_INTER, MEMORY_INTER, & COND_INTER.  REG_INTER causes an
-// operand to generate a function which returns its register number when
-// queried.   CONST_INTER causes an operand to generate a function which
-// returns the value of the constant when queried.  MEMORY_INTER causes an
-// operand to generate four functions which return the Base Register, the
-// Index Register, the Scale Value, and the Offset Value of the operand when
-// queried.  COND_INTER causes an operand to generate six functions which
-// return the encoding code (ie - encoding bits for the instruction)
-// associated with each basic boolean condition for a conditional instruction.
-// Instructions specify two basic values for encoding.  They use the
-// ins_encode keyword to specify their encoding class (which must be one of
-// the class names specified in the encoding block), and they use the
-// opcode keyword to specify, in order, their primary, secondary, and
-// tertiary opcode.  Only the opcode sections which a particular instruction
-// needs for encoding need to be specified.
-encode %{
-  // Build emit functions for each basic byte or larger field in the intel
-  // encoding scheme (opcode, rm, sib, immediate), and call them from C++
-  // code in the enc_class source block.  Emit functions will live in the
-  // main source block for now.  In future, we can generalize this by
-  // adding a syntax that specifies the sizes of fields in an order,
-  // so that the adlc can build the emit functions automagically
-
-  enc_class tlsencode (eRegP dst, eRegP src) %{
-    emit_rm(cbuf, 0x2, $dst$$reg, $src$$reg);
-    emit_d32(cbuf, ThreadLocalStorage::get_thread_ptr_offset() );
-  %}
-
-  enc_class call_epilog %{
-    if( VerifyStackAtCalls ) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize = ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP,-3*VMRegImpl::slots_per_word));
-      if(framesize >= 128) {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0xBC);
-        emit_d8(cbuf,0x24);
-        emit_d32(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      else {
-        emit_opcode(cbuf, 0x81); // cmp [esp+0],0xbadb1ood
-        emit_d8(cbuf,0x7C);
-        emit_d8(cbuf,0x24);
-        emit_d8(cbuf,framesize); // Find majik cookie from ESP
-        emit_d32(cbuf, 0xbadb100d);
-      }
-      // jmp EQ around INT3
-      emit_opcode(cbuf,0x74);
-      emit_d8(cbuf,1);
-      // Die if stack mismatch
-      emit_opcode(cbuf,0xCC);
-    }
-  %}
-
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-
-//----------OS and Locking Instructions----------------------------------------
-
-// The prefix of this name is KNOWN by the ADLC and cannot be changed.
-instruct tlsLoadP_prefixLoadP(eRegP t1) %{
-  effect(DEF t1);
-
-  format %{ "MOV    $t1,FS:[0x00] "%}
-  opcode(0x8B, 0x64);
-  ins_encode(OpcS, OpcP, conmemref(t1));
-  ins_pipe( ialu_reg_fat );
-%}
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-// %%% Should do this with a clause like:  bottom_type(TypeRawPtr::BOTTOM);
-instruct tlsLoadP(eRegP dst, eRegP t1) %{
-  effect(DEF dst, USE t1);
-
-  format %{ "MOV    $dst,[$t1 + TLS::thread_ptr_offset()]" %}
-  opcode(0x8B);
-  ins_encode(OpcP, tlsencode(dst, t1));
-  ins_pipe( ialu_reg_reg_fat );
-%}
-
-instruct TLS(eRegP dst) %{
-  match(Set dst (ThreadLocal));
-  expand %{
-    eRegP t1;
-    tlsLoadP_prefixLoadP(t1);
-    tlsLoadP(dst, t1);
-  %}
-%}
-
-// Die now
-instruct ShouldNotReachHere( )
-%{
-  match(Halt);
-  // Use the following format syntax
-  format %{ "INT3   ; ShouldNotReachHere" %}
-  opcode(0xCC);
-  ins_encode(OpcP);
-  ins_pipe( pipe_slow );
-%}
-
-//
-// Platform dependent source
-//
-source %{
-
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer &cbuf) {
-  cbuf.insts()->emit_int8((unsigned char) 0xcc);
-}
-
-void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
-  emit_break(cbuf);
-}
-
-
-uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
-  return 1;
-}
-
-
-%}
diff -r d3384450b649 -r 15085a6eb50c src/os_cpu/windows_x86/vm/windows_x86_64.ad
--- a/src/os_cpu/windows_x86/vm/windows_x86_64.ad	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/os_cpu/windows_x86/vm/windows_x86_64.ad	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
+// Copyright (c) 2003, 2012, Oracle and/or its affiliates. All rights reserved.
 // DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 //
 // This code is free software; you can redistribute it and/or modify it
@@ -67,69 +67,6 @@
     emit_opcode(cbuf, 0xD0 | (R10_enc - 8));
   %}
 
-  enc_class call_epilog %{
-    if (VerifyStackAtCalls) {
-      // Check that stack depth is unchanged: find majik cookie on stack
-      int framesize =
-        ra_->reg2offset_unchecked(OptoReg::add(ra_->_matcher._old_SP, -3*VMRegImpl::slots_per_word));
-      if (framesize) {
-        if (framesize < 0x80) {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0x7C);
-          emit_d8(cbuf, 0x24);
-          emit_d8(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        } else {
-          emit_opcode(cbuf, Assembler::REX_W);
-          emit_opcode(cbuf, 0x81); // cmpq [rsp+0],0xbadb1ood
-          emit_d8(cbuf, 0xBC);
-          emit_d8(cbuf, 0x24);
-          emit_d32(cbuf, framesize); // Find majik cookie from ESP
-          emit_d32(cbuf, 0xbadb100d);
-        }
-      }
-      // jmp EQ around INT3
-      // QQQ TODO
-      const int jump_around = 5; // size of call to breakpoint, 1 for CC
-      emit_opcode(cbuf, 0x74);
-      emit_d8(cbuf, jump_around);
-      // QQQ temporary
-      emit_break(cbuf);
-      // Die if stack mismatch
-      // emit_opcode(cbuf,0xCC);
-    }
-  %}
-%}
-
-// INSTRUCTIONS -- Platform dependent
-
-
-//----------OS and Locking Instructions----------------------------------------
-
-// This name is KNOWN by the ADLC and cannot be changed.
-// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
-// for this guy.
-instruct tlsLoadP(r15_RegP dst)
-%{
-  match(Set dst (ThreadLocal));
-  effect(DEF dst);
-
-  size(0);
-  format %{ "# TLS is in R15" %}
-  ins_encode( /*empty encoding*/ );
-  ins_pipe(ialu_reg_reg);
-%}
-
-// Die now
-instruct ShouldNotReachHere( )
-%{
-  match(Halt);
-  // Use the following format syntax
-  format %{ "INT3   ; ShouldNotReachHere" %}
-  opcode(0xCC);
-  ins_encode(OpcP);
-  ins_pipe( pipe_slow );
 %}
 
 //
@@ -142,17 +79,4 @@
   return 13; // movq r10,#addr; callq (r10)
 }
 
-// emit an interrupt that is caught by the debugger
-void emit_break(CodeBuffer &cbuf) {
-  cbuf.insts()->emit_int8((unsigned char) 0xcc);
-}
-
-void MachBreakpointNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
-  emit_break(cbuf);
-}
-
-uint MachBreakpointNode::size(PhaseRegAlloc *ra_) const {
-  return 1;
-}
-
 %}
diff -r d3384450b649 -r 15085a6eb50c src/share/tools/hsdis/hsdis.c
--- a/src/share/tools/hsdis/hsdis.c	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/tools/hsdis/hsdis.c	Fri Feb 17 12:18:52 2012 -0800
@@ -356,7 +356,7 @@
       if (plen > mach_size)  plen = mach_size;
       strncpy(mach_option, p, plen);
       mach_option[plen] = '\0';
-    } else if (plen > 6 && strncmp(p, "hsdis-", 6)) {
+    } else if (plen > 6 && strncmp(p, "hsdis-", 6) == 0) {
       // do not pass these to the next level
     } else {
       /* just copy it; {i386,sparc}-dis.c might like to see it  */
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/c1/c1_Canonicalizer.cpp
--- a/src/share/vm/c1/c1_Canonicalizer.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/c1/c1_Canonicalizer.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -594,6 +594,13 @@
   return false;
 }
 
+static bool is_safepoint(BlockEnd* x, BlockBegin* sux) {
+  // An Instruction with multiple successors, x, is replaced by a Goto
+  // to a single successor, sux. Is a safepoint check needed = was the
+  // instruction being replaced a safepoint and the single remaining
+  // successor a back branch?
+  return x->is_safepoint() && (sux->bci() < x->state_before()->bci());
+}
 
 void Canonicalizer::do_If(If* x) {
   // move const to right
@@ -614,7 +621,7 @@
     case If::geq: sux = x->sux_for(true);  break;
     }
     // If is a safepoint then the debug information should come from the state_before of the If.
-    set_canonical(new Goto(sux, x->state_before(), x->is_safepoint()));
+    set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
     return;
   }
 
@@ -626,7 +633,7 @@
                                                        x->sux_for(false));
       if (sux != NULL) {
         // If is a safepoint then the debug information should come from the state_before of the If.
-        set_canonical(new Goto(sux, x->state_before(), x->is_safepoint()));
+        set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
       }
     }
   } else if (rt->as_IntConstant() != NULL) {
@@ -694,10 +701,12 @@
     }
   } else if (rt == objectNull && (l->as_NewInstance() || l->as_NewArray())) {
     if (x->cond() == Instruction::eql) {
-      set_canonical(new Goto(x->fsux(), x->state_before(), x->is_safepoint()));
+      BlockBegin* sux = x->fsux();
+      set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
     } else {
       assert(x->cond() == Instruction::neq, "only other valid case");
-      set_canonical(new Goto(x->tsux(), x->state_before(), x->is_safepoint()));
+      BlockBegin* sux = x->tsux();
+      set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
     }
   }
 }
@@ -710,7 +719,7 @@
     if (v >= x->lo_key() && v <= x->hi_key()) {
       sux = x->sux_at(v - x->lo_key());
     }
-    set_canonical(new Goto(sux, x->state_before(), x->is_safepoint()));
+    set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
   } else if (x->number_of_sux() == 1) {
     // NOTE: Code permanently disabled for now since the switch statement's
     //       tag expression may produce side-effects in which case it must
@@ -741,7 +750,7 @@
         sux = x->sux_at(i);
       }
     }
-    set_canonical(new Goto(sux, x->state_before(), x->is_safepoint()));
+    set_canonical(new Goto(sux, x->state_before(), is_safepoint(x, sux)));
   } else if (x->number_of_sux() == 1) {
     // NOTE: Code permanently disabled for now since the switch statement's
     //       tag expression may produce side-effects in which case it must
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/c1/c1_GraphBuilder.cpp
--- a/src/share/vm/c1/c1_GraphBuilder.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/c1/c1_GraphBuilder.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -1181,6 +1181,11 @@
   bool is_bb = tsux->bci() < stream()->cur_bci() || fsux->bci() < stream()->cur_bci();
   Instruction *i = append(new If(x, cond, false, y, tsux, fsux, is_bb ? state_before : NULL, is_bb));
 
+  assert(i->as_Goto() == NULL ||
+         (i->as_Goto()->sux_at(0) == tsux  && i->as_Goto()->is_safepoint() == tsux->bci() < stream()->cur_bci()) ||
+         (i->as_Goto()->sux_at(0) == fsux  && i->as_Goto()->is_safepoint() == fsux->bci() < stream()->cur_bci()),
+         "safepoint state of Goto returned by canonicalizer incorrect");
+
   if (is_profiling()) {
     If* if_node = i->as_If();
     if (if_node != NULL) {
@@ -1303,7 +1308,16 @@
     // add default successor
     sux->at_put(i, block_at(bci() + sw.default_offset()));
     ValueStack* state_before = has_bb ? copy_state_before() : NULL;
-    append(new TableSwitch(ipop(), sux, sw.low_key(), state_before, has_bb));
+    Instruction* res = append(new TableSwitch(ipop(), sux, sw.low_key(), state_before, has_bb));
+#ifdef ASSERT
+    if (res->as_Goto()) {
+      for (i = 0; i < l; i++) {
+        if (sux->at(i) == res->as_Goto()->sux_at(0)) {
+          assert(res->as_Goto()->is_safepoint() == sw.dest_offset_at(i) < 0, "safepoint state of Goto returned by canonicalizer incorrect");
+        }
+      }
+    }
+#endif
   }
 }
 
@@ -1338,7 +1352,16 @@
     // add default successor
     sux->at_put(i, block_at(bci() + sw.default_offset()));
     ValueStack* state_before = has_bb ? copy_state_before() : NULL;
-    append(new LookupSwitch(ipop(), sux, keys, state_before, has_bb));
+    Instruction* res = append(new LookupSwitch(ipop(), sux, keys, state_before, has_bb));
+#ifdef ASSERT
+    if (res->as_Goto()) {
+      for (i = 0; i < l; i++) {
+        if (sux->at(i) == res->as_Goto()->sux_at(0)) {
+          assert(res->as_Goto()->is_safepoint() == sw.pair_at(i).offset() < 0, "safepoint state of Goto returned by canonicalizer incorrect");
+        }
+      }
+    }
+#endif
   }
 }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/c1/c1_LinearScan.cpp
--- a/src/share/vm/c1/c1_LinearScan.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/c1/c1_LinearScan.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -2464,12 +2464,15 @@
 
 
 // frequently used constants
-ConstantOopWriteValue LinearScan::_oop_null_scope_value = ConstantOopWriteValue(NULL);
-ConstantIntValue      LinearScan::_int_m1_scope_value = ConstantIntValue(-1);
-ConstantIntValue      LinearScan::_int_0_scope_value =  ConstantIntValue(0);
-ConstantIntValue      LinearScan::_int_1_scope_value =  ConstantIntValue(1);
-ConstantIntValue      LinearScan::_int_2_scope_value =  ConstantIntValue(2);
-LocationValue         _illegal_value = LocationValue(Location());
+// Allocate them with new so they are never destroyed (otherwise, a
+// forced exit could destroy these objects while they are still in
+// use).
+ConstantOopWriteValue* LinearScan::_oop_null_scope_value = new (ResourceObj::C_HEAP) ConstantOopWriteValue(NULL);
+ConstantIntValue*      LinearScan::_int_m1_scope_value = new (ResourceObj::C_HEAP) ConstantIntValue(-1);
+ConstantIntValue*      LinearScan::_int_0_scope_value =  new (ResourceObj::C_HEAP) ConstantIntValue(0);
+ConstantIntValue*      LinearScan::_int_1_scope_value =  new (ResourceObj::C_HEAP) ConstantIntValue(1);
+ConstantIntValue*      LinearScan::_int_2_scope_value =  new (ResourceObj::C_HEAP) ConstantIntValue(2);
+LocationValue*         _illegal_value = new (ResourceObj::C_HEAP) LocationValue(Location());
 
 void LinearScan::init_compute_debug_info() {
   // cache for frequently used scope values
@@ -2508,7 +2511,7 @@
     case T_OBJECT: {
       jobject value = c->as_jobject();
       if (value == NULL) {
-        scope_values->append(&_oop_null_scope_value);
+        scope_values->append(_oop_null_scope_value);
       } else {
         scope_values->append(new ConstantOopWriteValue(c->as_jobject()));
       }
@@ -2519,10 +2522,10 @@
     case T_FLOAT: {
       int value = c->as_jint_bits();
       switch (value) {
-        case -1: scope_values->append(&_int_m1_scope_value); break;
-        case 0:  scope_values->append(&_int_0_scope_value); break;
-        case 1:  scope_values->append(&_int_1_scope_value); break;
-        case 2:  scope_values->append(&_int_2_scope_value); break;
+        case -1: scope_values->append(_int_m1_scope_value); break;
+        case 0:  scope_values->append(_int_0_scope_value); break;
+        case 1:  scope_values->append(_int_1_scope_value); break;
+        case 2:  scope_values->append(_int_2_scope_value); break;
         default: scope_values->append(new ConstantIntValue(c->as_jint_bits())); break;
       }
       return 1;
@@ -2531,7 +2534,7 @@
     case T_LONG: // fall through
     case T_DOUBLE: {
 #ifdef _LP64
-      scope_values->append(&_int_0_scope_value);
+      scope_values->append(_int_0_scope_value);
       scope_values->append(new ConstantLongValue(c->as_jlong_bits()));
 #else
       if (hi_word_offset_in_bytes > lo_word_offset_in_bytes) {
@@ -2657,7 +2660,7 @@
       }
       // Does this reverse on x86 vs. sparc?
       first =  new LocationValue(loc1);
-      second = &_int_0_scope_value;
+      second = _int_0_scope_value;
 #else
       Location loc1, loc2;
       if (!frame_map()->locations_for_slot(opr->double_stack_ix(), Location::normal, &loc1, &loc2)) {
@@ -2671,7 +2674,7 @@
 #ifdef _LP64
       VMReg rname_first = opr->as_register_lo()->as_VMReg();
       first = new LocationValue(Location::new_reg_loc(Location::lng, rname_first));
-      second = &_int_0_scope_value;
+      second = _int_0_scope_value;
 #else
       VMReg rname_first = opr->as_register_lo()->as_VMReg();
       VMReg rname_second = opr->as_register_hi()->as_VMReg();
@@ -2694,7 +2697,7 @@
       VMReg rname_first  = opr->as_xmm_double_reg()->as_VMReg();
 #  ifdef _LP64
       first = new LocationValue(Location::new_reg_loc(Location::dbl, rname_first));
-      second = &_int_0_scope_value;
+      second = _int_0_scope_value;
 #  else
       first = new LocationValue(Location::new_reg_loc(Location::normal, rname_first));
       // %%% This is probably a waste but we'll keep things as they were for now
@@ -2741,7 +2744,7 @@
 
 #ifdef _LP64
       first = new LocationValue(Location::new_reg_loc(Location::dbl, rname_first));
-      second = &_int_0_scope_value;
+      second = _int_0_scope_value;
 #else
       first = new LocationValue(Location::new_reg_loc(Location::normal, rname_first));
       // %%% This is probably a waste but we'll keep things as they were for now
@@ -2822,7 +2825,7 @@
     }
   } else {
     // append a dummy value because real value not needed
-    scope_values->append(&_illegal_value);
+    scope_values->append(_illegal_value);
     return 1;
   }
 }
@@ -2865,7 +2868,7 @@
     nof_locals = cur_scope->method()->max_locals();
     locals = new GrowableArray<ScopeValue*>(nof_locals);
     for(int i = 0; i < nof_locals; i++) {
-      locals->append(&_illegal_value);
+      locals->append(_illegal_value);
     }
   }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/c1/c1_LinearScan.hpp
--- a/src/share/vm/c1/c1_LinearScan.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/c1/c1_LinearScan.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -160,11 +160,11 @@
   // TODO: cached scope values for registers could be static
   ScopeValueArray           _scope_value_cache;
 
-  static ConstantOopWriteValue _oop_null_scope_value;
-  static ConstantIntValue    _int_m1_scope_value;
-  static ConstantIntValue    _int_0_scope_value;
-  static ConstantIntValue    _int_1_scope_value;
-  static ConstantIntValue    _int_2_scope_value;
+  static ConstantOopWriteValue* _oop_null_scope_value;
+  static ConstantIntValue*    _int_m1_scope_value;
+  static ConstantIntValue*    _int_0_scope_value;
+  static ConstantIntValue*    _int_1_scope_value;
+  static ConstantIntValue*    _int_2_scope_value;
 
   // accessors
   IR*           ir() const                       { return _ir; }
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/compiler/compileBroker.cpp
--- a/src/share/vm/compiler/compileBroker.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/compiler/compileBroker.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -204,7 +204,8 @@
   }
 
   void log_nmethod(JavaThread* thread, nmethod* nm) {
-    log(thread, "nmethod " INTPTR_FORMAT " code ["INTPTR_FORMAT ", " INTPTR_FORMAT "]",
+    log(thread, "nmethod %d%s " INTPTR_FORMAT " code ["INTPTR_FORMAT ", " INTPTR_FORMAT "]",
+        nm->compile_id(), nm->is_osr_method() ? "%" : "",
         nm, nm->code_begin(), nm->code_end());
   }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/gc_interface/collectedHeap.cpp
--- a/src/share/vm/gc_interface/collectedHeap.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/gc_interface/collectedHeap.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -62,7 +62,7 @@
     return;
   }
 
-  jlong timestamp = os::javaTimeNanos() / NANOSECS_PER_MILLISEC;
+  double timestamp = fetch_timestamp();
   MutexLockerEx ml(&_mutex, Mutex::_no_safepoint_check_flag);
   int index = compute_log_index();
   _records[index].thread = NULL; // Its the GC thread so it's not that interesting.
@@ -70,9 +70,9 @@
   _records[index].data.is_before = before;
   stringStream st(_records[index].data.buffer(), _records[index].data.size());
   if (before) {
-    Universe::print_heap_before_gc(&st);
+    Universe::print_heap_before_gc(&st, true);
   } else {
-    Universe::print_heap_after_gc(&st);
+    Universe::print_heap_after_gc(&st, true);
   }
 }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/memory/gcLocker.cpp
--- a/src/share/vm/memory/gcLocker.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/memory/gcLocker.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -31,7 +31,6 @@
 volatile jint GC_locker::_lock_count     = 0;
 volatile bool GC_locker::_needs_gc       = false;
 volatile bool GC_locker::_doing_gc       = false;
-jlong GC_locker::_wait_begin = 0;
 
 #ifdef ASSERT
 volatile jint GC_locker::_debug_jni_lock_count = 0;
@@ -69,9 +68,8 @@
     _needs_gc = true;
     if (PrintJNIGCStalls && PrintGCDetails) {
       ResourceMark rm; // JavaThread::name() allocates to convert to UTF8
-      _wait_begin = os::javaTimeNanos() / NANOSECS_PER_MILLISEC;
-      gclog_or_tty->print_cr(INT64_FORMAT ": Setting _needs_gc. Thread \"%s\" %d locked.",
-                             _wait_begin, Thread::current()->name(), _jni_lock_count);
+      gclog_or_tty->print_cr("%.3f: Setting _needs_gc. Thread \"%s\" %d locked.",
+                             gclog_or_tty->time_stamp().seconds(), Thread::current()->name(), _jni_lock_count);
     }
 
   }
@@ -85,8 +83,8 @@
   if (needs_gc()) {
     if (PrintJNIGCStalls && PrintGCDetails) {
       ResourceMark rm; // JavaThread::name() allocates to convert to UTF8
-      gclog_or_tty->print_cr(INT64_FORMAT ": Allocation failed. Thread \"%s\" is stalled by JNI critical section, %d locked.",
-                             (os::javaTimeNanos() / NANOSECS_PER_MILLISEC) - _wait_begin, Thread::current()->name(), _jni_lock_count);
+      gclog_or_tty->print_cr("%.3f: Allocation failed. Thread \"%s\" is stalled by JNI critical section, %d locked.",
+                             gclog_or_tty->time_stamp().seconds(), Thread::current()->name(), _jni_lock_count);
     }
   }
 
@@ -131,8 +129,8 @@
         MutexUnlocker munlock(JNICritical_lock);
         if (PrintJNIGCStalls && PrintGCDetails) {
           ResourceMark rm; // JavaThread::name() allocates to convert to UTF8
-          gclog_or_tty->print_cr(INT64_FORMAT ": Thread \"%s\" is performing GC after exiting critical section, %d locked",
-                                 (os::javaTimeNanos() / NANOSECS_PER_MILLISEC) - _wait_begin, Thread::current()->name(), _jni_lock_count);
+          gclog_or_tty->print_cr("%.3f: Thread \"%s\" is performing GC after exiting critical section, %d locked",
+                                 gclog_or_tty->time_stamp().seconds(), Thread::current()->name(), _jni_lock_count);
         }
         Universe::heap()->collect(GCCause::_gc_locker);
       }
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/memory/gcLocker.hpp
--- a/src/share/vm/memory/gcLocker.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/memory/gcLocker.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -63,9 +63,6 @@
                                          // note: bool is typedef'd as jint
   static volatile bool _doing_gc;        // unlock_critical() is doing a GC
 
-  static jlong         _wait_begin;      // Timestamp for the setting of _needs_gc.
-                                         // Used only by printing code.
-
 #ifdef ASSERT
   // This lock count is updated for all operations and is used to
   // validate the jni_lock_count that is computed during safepoints.
@@ -86,13 +83,26 @@
   static void jni_lock(JavaThread* thread);
   static void jni_unlock(JavaThread* thread);
 
+  static bool is_active_internal() {
+    verify_critical_count();
+    return _lock_count > 0 || _jni_lock_count > 0;
+  }
+
  public:
   // Accessors
-  static bool is_active();
+  static bool is_active() {
+    assert(_needs_gc || SafepointSynchronize::is_at_safepoint(), "only read at safepoint");
+    return is_active_internal();
+  }
   static bool needs_gc()       { return _needs_gc;                        }
 
   // Shorthand
-  static bool is_active_and_needs_gc() { return needs_gc() && is_active(); }
+  static bool is_active_and_needs_gc() {
+    // Use is_active_internal since _needs_gc can change from true to
+    // false outside of a safepoint, triggering the assert in
+    // is_active.
+    return needs_gc() && is_active_internal();
+  }
 
   // In debug mode track the locking state at all times
   static void increment_debug_jni_lock_count() {
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/memory/gcLocker.inline.hpp
--- a/src/share/vm/memory/gcLocker.inline.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/memory/gcLocker.inline.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -27,12 +27,6 @@
 
 #include "memory/gcLocker.hpp"
 
-inline bool GC_locker::is_active() {
-  assert(_needs_gc || SafepointSynchronize::is_at_safepoint(), "only read at safepoint");
-  verify_critical_count();
-  return _lock_count > 0 || _jni_lock_count > 0;
-}
-
 inline void GC_locker::lock() {
   // cast away volatile
   Atomic::inc(&_lock_count);
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/memory/universe.cpp
--- a/src/share/vm/memory/universe.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/memory/universe.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -1303,22 +1303,22 @@
   }
 }
 
-void Universe::print_heap_before_gc(outputStream* st) {
+void Universe::print_heap_before_gc(outputStream* st, bool ignore_extended) {
   st->print_cr("{Heap before GC invocations=%u (full %u):",
                heap()->total_collections(),
                heap()->total_full_collections());
-  if (!PrintHeapAtGCExtended) {
+  if (!PrintHeapAtGCExtended || ignore_extended) {
     heap()->print_on(st);
   } else {
     heap()->print_extended_on(st);
   }
 }
 
-void Universe::print_heap_after_gc(outputStream* st) {
+void Universe::print_heap_after_gc(outputStream* st, bool ignore_extended) {
   st->print_cr("Heap after GC invocations=%u (full %u):",
                heap()->total_collections(),
                heap()->total_full_collections());
-  if (!PrintHeapAtGCExtended) {
+  if (!PrintHeapAtGCExtended || ignore_extended) {
     heap()->print_on(st);
   } else {
     heap()->print_extended_on(st);
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/memory/universe.hpp
--- a/src/share/vm/memory/universe.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/memory/universe.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -424,8 +424,8 @@
   static void print_heap_at_SIGBREAK();
   static void print_heap_before_gc() { print_heap_before_gc(gclog_or_tty); }
   static void print_heap_after_gc()  { print_heap_after_gc(gclog_or_tty); }
-  static void print_heap_before_gc(outputStream* st);
-  static void print_heap_after_gc(outputStream* st);
+  static void print_heap_before_gc(outputStream* st, bool ignore_extended = false);
+  static void print_heap_after_gc(outputStream* st, bool ignore_extended = false);
 
   // Change the number of dummy objects kept reachable by the full gc dummy
   // array; this should trigger relocation in a sliding compaction collector.
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/opto/chaitin.cpp
--- a/src/share/vm/opto/chaitin.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/opto/chaitin.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2000, 2012, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1946,18 +1946,29 @@
     reg2offset_unchecked(OptoReg::add(_matcher._old_SP,-1)) - reg2offset_unchecked(_matcher._new_SP)+jintSize);
 
   // Preserve area dump
+  int fixed_slots = C->fixed_slots();
+  OptoReg::Name begin_in_preserve = OptoReg::add(_matcher._old_SP, -(int)C->in_preserve_stack_slots());
+  OptoReg::Name return_addr = _matcher.return_addr();
+
   reg = OptoReg::add(reg, -1);
-  while( OptoReg::is_stack(reg)) {
+  while (OptoReg::is_stack(reg)) {
     tty->print("#r%3.3d %s+%2d: ",reg,fp,reg2offset_unchecked(reg));
-    if( _matcher.return_addr() == reg )
+    if (return_addr == reg) {
       tty->print_cr("return address");
-    else if( _matcher.return_addr() == OptoReg::add(reg,1) &&
-             VerifyStackAtCalls )
-      tty->print_cr("0xBADB100D   +VerifyStackAtCalls");
-    else if ((int)OptoReg::reg2stack(reg) < C->fixed_slots())
+    } else if (reg >= begin_in_preserve) {
+      // Preserved slots are present on x86
+      if (return_addr == OptoReg::add(reg, VMRegImpl::slots_per_word))
+        tty->print_cr("saved fp register");
+      else if (return_addr == OptoReg::add(reg, 2*VMRegImpl::slots_per_word) &&
+               VerifyStackAtCalls)
+        tty->print_cr("0xBADB100D   +VerifyStackAtCalls");
+      else
+        tty->print_cr("in_preserve");
+    } else if ((int)OptoReg::reg2stack(reg) < fixed_slots) {
       tty->print_cr("Fixed slot %d", OptoReg::reg2stack(reg));
-    else
-      tty->print_cr("pad2, in_preserve");
+    } else {
+      tty->print_cr("pad2, stack alignment");
+    }
     reg = OptoReg::add(reg, -1);
   }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/opto/escape.cpp
--- a/src/share/vm/opto/escape.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/opto/escape.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2005, 2011, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2005, 2012, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -1687,12 +1687,23 @@
   // Observed 8 passes in jvm2008 compiler.compiler.
   // Set limit to 20 to catch situation when something
   // did go wrong and recompile the method without EA.
+  // Also limit build time to 30 sec (60 in debug VM).
 
 #define CG_BUILD_ITER_LIMIT 20
 
+#ifdef ASSERT
+#define CG_BUILD_TIME_LIMIT 60.0
+#else
+#define CG_BUILD_TIME_LIMIT 30.0
+#endif
+
   uint length = worklist.length();
   int iterations = 0;
-  while(_progress && (iterations++ < CG_BUILD_ITER_LIMIT)) {
+  elapsedTimer time;
+  while(_progress &&
+        (iterations++   < CG_BUILD_ITER_LIMIT) &&
+        (time.seconds() < CG_BUILD_TIME_LIMIT)) {
+    time.start();
     _progress = false;
     for( uint next = 0; next < length; ++next ) {
       int ni = worklist.at(next);
@@ -1701,18 +1712,19 @@
       assert(n != NULL, "should be known node");
       build_connection_graph(n, igvn);
     }
+    time.stop();
   }
-  if (iterations >= CG_BUILD_ITER_LIMIT) {
-    assert(iterations < CG_BUILD_ITER_LIMIT,
-           err_msg("infinite EA connection graph build with %d nodes and worklist size %d",
-           nodes_size(), length));
+  if ((iterations     >= CG_BUILD_ITER_LIMIT) ||
+      (time.seconds() >= CG_BUILD_TIME_LIMIT)) {
+    assert(false, err_msg("infinite EA connection graph build (%f sec, %d iterations) with %d nodes and worklist size %d",
+           time.seconds(), iterations, nodes_size(), length));
     // Possible infinite build_connection_graph loop,
-    // retry compilation without escape analysis.
-    C->record_failure(C2Compiler::retry_no_escape_analysis());
+    // bailout (no changes to ideal graph were made).
     _collecting = false;
     return false;
   }
 #undef CG_BUILD_ITER_LIMIT
+#undef CG_BUILD_TIME_LIMIT
 
   // 5. Propagate escaped states.
   worklist.clear();
@@ -2292,9 +2304,35 @@
         PointsToNode::EscapeState arg_esc = ptnode_adr(arg->_idx)->escape_state();
         if (!arg->is_top() && at->isa_ptr() && aat->isa_ptr() &&
             (is_arraycopy || arg_esc < PointsToNode::ArgEscape)) {
-
+#ifdef ASSERT
           assert(aat == Type::TOP || aat == TypePtr::NULL_PTR ||
                  aat->isa_ptr() != NULL, "expecting an Ptr");
+          if (!(is_arraycopy ||
+                call->as_CallLeaf()->_name != NULL &&
+                (strcmp(call->as_CallLeaf()->_name, "g1_wb_pre")  == 0 ||
+                 strcmp(call->as_CallLeaf()->_name, "g1_wb_post") == 0 ))
+          ) {
+            call->dump();
+            assert(false, "EA: unexpected CallLeaf");
+          }
+#endif
+          if (arg_esc < PointsToNode::ArgEscape) {
+            set_escape_state(arg->_idx, PointsToNode::ArgEscape);
+            Node* arg_base = arg;
+            if (arg->is_AddP()) {
+              //
+              // The inline_native_clone() case when the arraycopy stub is called
+              // after the allocation before Initialize and CheckCastPP nodes.
+              // Or normal arraycopy for object arrays case.
+              //
+              // Set AddP's base (Allocate) as not scalar replaceable since
+              // pointer to the base (with offset) is passed as argument.
+              //
+              arg_base = get_addp_base(arg);
+              set_escape_state(arg_base->_idx, PointsToNode::ArgEscape);
+            }
+          }
+
           bool arg_has_oops = aat->isa_oopptr() &&
                               (aat->isa_oopptr()->klass() == NULL || aat->isa_instptr() ||
                                (aat->isa_aryptr() && aat->isa_aryptr()->klass()->is_obj_array_klass()));
@@ -2307,85 +2345,33 @@
           //   arraycopy(char[],0,Object*,0,size);
           //   arraycopy(Object*,0,char[],0,size);
           //
-          // Don't add edges from dst's fields in such cases.
+          // Do nothing special in such cases.
           //
-          bool arg_is_arraycopy_dest = src_has_oops && is_arraycopy &&
-                                       arg_has_oops && (i > TypeFunc::Parms);
-#ifdef ASSERT
-          if (!(is_arraycopy ||
-                call->as_CallLeaf()->_name != NULL &&
-                (strcmp(call->as_CallLeaf()->_name, "g1_wb_pre")  == 0 ||
-                 strcmp(call->as_CallLeaf()->_name, "g1_wb_post") == 0 ))
-          ) {
-            call->dump();
-            assert(false, "EA: unexpected CallLeaf");
-          }
-#endif
-          // Always process arraycopy's destination object since
-          // we need to add all possible edges to references in
-          // source object.
-          if (arg_esc >= PointsToNode::ArgEscape &&
-              !arg_is_arraycopy_dest) {
-            continue;
-          }
-          set_escape_state(arg->_idx, PointsToNode::ArgEscape);
-          Node* arg_base = arg;
-          if (arg->is_AddP()) {
-            //
-            // The inline_native_clone() case when the arraycopy stub is called
-            // after the allocation before Initialize and CheckCastPP nodes.
-            // Or normal arraycopy for object arrays case.
-            //
-            // Set AddP's base (Allocate) as not scalar replaceable since
-            // pointer to the base (with offset) is passed as argument.
-            //
-            arg_base = get_addp_base(arg);
-          }
-          VectorSet argset = *PointsTo(arg_base); // Clone set
-          for( VectorSetI j(&argset); j.test(); ++j ) {
-            uint pd = j.elem; // Destination object
-            set_escape_state(pd, PointsToNode::ArgEscape);
-
-            if (arg_is_arraycopy_dest) {
-              PointsToNode* ptd = ptnode_adr(pd);
-              // Conservatively reference an unknown object since
-              // not all source's fields/elements may be known.
-              add_edge_from_fields(pd, _phantom_object, Type::OffsetBot);
-
-              Node *src = call->in(TypeFunc::Parms)->uncast();
-              Node* src_base = src;
-              if (src->is_AddP()) {
-                src_base  = get_addp_base(src);
-              }
-              // Create edges from destination's fields to
-              // everything known source's fields could point to.
-              for( VectorSetI s(PointsTo(src_base)); s.test(); ++s ) {
-                uint ps = s.elem;
-                bool has_bottom_offset = false;
-                for (uint fd = 0; fd < ptd->edge_count(); fd++) {
-                  assert(ptd->edge_type(fd) == PointsToNode::FieldEdge, "expecting a field edge");
-                  int fdi = ptd->edge_target(fd);
-                  PointsToNode* pfd = ptnode_adr(fdi);
-                  int offset = pfd->offset();
-                  if (offset == Type::OffsetBot)
-                    has_bottom_offset = true;
-                  assert(offset != -1, "offset should be set");
-                  add_deferred_edge_to_fields(fdi, ps, offset);
-                }
-                // Destination object may not have access (no field edge)
-                // to fields which are accessed in source object.
-                // As result no edges will be created to those source's
-                // fields and escape state of destination object will
-                // not be propagated to those fields.
-                //
-                // Mark source object as global escape except in
-                // the case with Type::OffsetBot field (which is
-                // common case for array elements access) when
-                // edges are created to all source's fields.
-                if (!has_bottom_offset) {
-                  set_escape_state(ps, PointsToNode::GlobalEscape);
-                }
-              }
+          if (is_arraycopy && (i > TypeFunc::Parms) &&
+              src_has_oops && arg_has_oops) {
+            // Destination object's fields reference an unknown object.
+            Node* arg_base = arg;
+            if (arg->is_AddP()) {
+              arg_base = get_addp_base(arg);
+            }
+            for (VectorSetI s(PointsTo(arg_base)); s.test(); ++s) {
+              uint ps = s.elem;
+              set_escape_state(ps, PointsToNode::ArgEscape);
+              add_edge_from_fields(ps, _phantom_object, Type::OffsetBot);
+            }
+            // Conservatively all values in source object fields globally escape
+            // since we don't know if values in destination object fields
+            // escape (it could be traced but it is too expensive).
+            Node* src = call->in(TypeFunc::Parms)->uncast();
+            Node* src_base = src;
+            if (src->is_AddP()) {
+              src_base  = get_addp_base(src);
+            }
+            for (VectorSetI s(PointsTo(src_base)); s.test(); ++s) {
+              uint ps = s.elem;
+              set_escape_state(ps, PointsToNode::ArgEscape);
+              // Use OffsetTop to indicate fields global escape.
+              add_edge_from_fields(ps, _phantom_object, Type::OffsetTop);
             }
           }
         }
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/opto/output.cpp
--- a/src/share/vm/opto/output.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/opto/output.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -167,7 +167,7 @@
   // Determine if we need to generate a stack overflow check.
   // Do it if the method is not a stub function and
   // has java calls or has frame size > vm_page_size/8.
-  return (stub_function() == NULL &&
+  return (UseStackBanging && stub_function() == NULL &&
           (has_java_calls() || frame_size_in_bytes > os::vm_page_size()>>3));
 }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/runtime/arguments.cpp
--- a/src/share/vm/runtime/arguments.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/runtime/arguments.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -102,8 +102,6 @@
 char* Arguments::_meta_index_path = NULL;
 char* Arguments::_meta_index_dir = NULL;
 
-static bool force_client_mode = false;
-
 // Check if head of 'option' matches 'name', and sets 'tail' remaining part of option string
 
 static bool match_option(const JavaVMOption *option, const char* name,
@@ -1345,7 +1343,7 @@
     return;
   }
 
-  if (os::is_server_class_machine() && !force_client_mode ) {
+  if (os::is_server_class_machine()) {
     // If no other collector is requested explicitly,
     // let the VM select the collector based on
     // machine class and automatic selection policy.
@@ -2937,11 +2935,6 @@
   // Construct the path to the archive
   char jvm_path[JVM_MAXPATHLEN];
   os::jvm_path(jvm_path, sizeof(jvm_path));
-#ifdef TIERED
-  if (strstr(jvm_path, "client") != NULL) {
-    force_client_mode = true;
-  }
-#endif // TIERED
   char *end = strrchr(jvm_path, *os::file_separator());
   if (end != NULL) *end = '\0';
   char *shared_archive_path = NEW_C_HEAP_ARRAY(char, strlen(jvm_path) +
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/runtime/globals.hpp
--- a/src/share/vm/runtime/globals.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/runtime/globals.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -3013,7 +3013,7 @@
   product(intx, SafepointTimeoutDelay, 10000,                               \
           "Delay in milliseconds for option SafepointTimeout")              \
                                                                             \
-  product(intx, NmethodSweepFraction, 4,                                    \
+  product(intx, NmethodSweepFraction, 16,                                    \
           "Number of invocations of sweeper to cover all nmethods")         \
                                                                             \
   product(intx, NmethodSweepCheckInterval, 5,                               \
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/runtime/sharedRuntime.cpp
--- a/src/share/vm/runtime/sharedRuntime.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/runtime/sharedRuntime.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -804,6 +804,7 @@
         if (thread->deopt_mark() != NULL) {
           Deoptimization::cleanup_deopt_info(thread, NULL);
         }
+        Events::log_exception(thread, "StackOverflowError at " INTPTR_FORMAT, pc);
         return StubRoutines::throw_StackOverflowError_entry();
       }
 
@@ -820,8 +821,10 @@
 
           if (vt_stub->is_abstract_method_error(pc)) {
             assert(!vt_stub->is_vtable_stub(), "should never see AbstractMethodErrors from vtable-type VtableStubs");
+            Events::log_exception(thread, "AbstractMethodError at " INTPTR_FORMAT, pc);
             return StubRoutines::throw_AbstractMethodError_entry();
           } else {
+            Events::log_exception(thread, "NullPointerException at vtable entry " INTPTR_FORMAT, pc);
             return StubRoutines::throw_NullPointerException_at_call_entry();
           }
         } else {
@@ -838,6 +841,7 @@
           if (!cb->is_nmethod()) {
             guarantee(cb->is_adapter_blob() || cb->is_method_handles_adapter_blob(),
                       "exception happened outside interpreter, nmethods and vtable stubs (1)");
+            Events::log_exception(thread, "NullPointerException in code blob at " INTPTR_FORMAT, pc);
             // There is no handler here, so we will simply unwind.
             return StubRoutines::throw_NullPointerException_at_call_entry();
           }
@@ -849,6 +853,7 @@
             // => the nmethod is not yet active (i.e., the frame
             // is not set up yet) => use return address pushed by
             // caller => don't push another return address
+            Events::log_exception(thread, "NullPointerException in IC check " INTPTR_FORMAT, pc);
             return StubRoutines::throw_NullPointerException_at_call_entry();
           }
 
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/runtime/sweeper.cpp
--- a/src/share/vm/runtime/sweeper.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/runtime/sweeper.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -266,7 +266,17 @@
 
     // The last invocation iterates until there are no more nmethods
     for (int i = 0; (i < todo || _invocations == 1) && _current != NULL; i++) {
+      if (SafepointSynchronize::is_synchronizing()) { // Safepoint request
+        if (PrintMethodFlushing && Verbose) {
+          tty->print_cr("### Sweep at %d out of %d, invocation: %d, yielding to safepoint", _seen, CodeCache::nof_nmethods(), _invocations);
+        }
+        MutexUnlockerEx mu(CodeCache_lock, Mutex::_no_safepoint_check_flag);
 
+        assert(Thread::current()->is_Java_thread(), "should be java thread");
+        JavaThread* thread = (JavaThread*)Thread::current();
+        ThreadBlockInVM tbivm(thread);
+        thread->java_suspend_self();
+      }
       // Since we will give up the CodeCache_lock, always skip ahead
       // to the next nmethod.  Other blobs can be deleted by other
       // threads but nmethods are only reclaimed by the sweeper.
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/utilities/debug.cpp
--- a/src/share/vm/utilities/debug.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/utilities/debug.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -600,6 +600,10 @@
   tty->flush();
 }
 
+extern "C" void events() {
+  Command c("events");
+  Events::print();
+}
 
 // Given a heap address that was valid before the most recent GC, if
 // the oop that used to contain it is still live, prints the new
@@ -759,7 +763,7 @@
 
   tty->print_cr("misc.");
   tty->print_cr("  flush()       - flushes the log file");
-  tty->print_cr("  events()      - dump last 50 events");
+  tty->print_cr("  events()      - dump events from ring buffers");
 
 
   tty->print_cr("compiler debugging");
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/utilities/events.cpp
--- a/src/share/vm/utilities/events.cpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/utilities/events.cpp	Fri Feb 17 12:18:52 2012 -0800
@@ -68,6 +68,10 @@
   }
 }
 
+void Events::print() {
+  print_all(tty);
+}
+
 void Events::init() {
   if (LogEvents) {
     _messages = new StringEventLog("Events");
diff -r d3384450b649 -r 15085a6eb50c src/share/vm/utilities/events.hpp
--- a/src/share/vm/utilities/events.hpp	Fri Feb 17 06:34:49 2012 -0800
+++ b/src/share/vm/utilities/events.hpp	Fri Feb 17 12:18:52 2012 -0800
@@ -35,20 +35,12 @@
 // This facility is extremly useful for post-mortem debugging. The eventlog
 // often provides crucial information about events leading up to the crash.
 //
-// All arguments past the format string must be passed as an intptr_t.
-//
-// To log a single event use:
-//    Events::log("New nmethod has been created " INTPTR_FORMAT, nm);
-//
-// To log a block of events use:
-//    EventMark m("GarbageCollecting %d", (intptr_t)gc_number);
-//
-// The constructor to eventlog indents the eventlog until the
-// destructor has been executed.
-//
-// IMPLEMENTATION RESTRICTION:
-//   Max 3 arguments are saved for each logged event.
-//
+// Abstractly the logs can record whatever they way but normally they
+// would record at least a timestamp and the current Thread, along
+// with whatever data they need in a ring buffer.  Commonly fixed
+// length text messages are recorded for simplicity but other
+// strategies could be used.  Several logs are provided by default but
+// new instances can be created as needed.
 
 // The base event log dumping class that is registered for dumping at
 // crash time.  This is a very generic interface that is mainly here
@@ -79,7 +71,7 @@
 template <class T> class EventLogBase : public EventLog {
   template <class X> class EventRecord {
    public:
-    jlong   timestamp;
+    double  timestamp;
     Thread* thread;
     X       data;
   };
@@ -102,6 +94,10 @@
     _records = new EventRecord<T>[length];
   }
 
+  double fetch_timestamp() {
+    return os::elapsedTime();
+  }
+
   // move the ring buffer to next open slot and return the index of
   // the slot to use for the current message.  Should only be called
   // while mutex is held.
@@ -130,7 +126,7 @@
   void print(outputStream* out, T& e);
 
   void print(outputStream* out, EventRecord<T>& e) {
-    out->print("Event: " INT64_FORMAT " ", e.timestamp);
+    out->print("Event: %.3f ", e.timestamp);
     if (e.thread != NULL) {
       out->print("Thread " INTPTR_FORMAT " ", e.thread);
     }
@@ -155,7 +151,7 @@
   void logv(Thread* thread, const char* format, va_list ap) {
     if (!should_log()) return;
 
-    jlong timestamp = os::javaTimeNanos() / NANOSECS_PER_MILLISEC;
+    double timestamp = fetch_timestamp();
     MutexLockerEx ml(&_mutex, Mutex::_no_safepoint_check_flag);
     int index = compute_log_index();
     _records[index].thread = thread;
@@ -193,9 +189,8 @@
  public:
   static void print_all(outputStream* out);
 
-  static void print() {
-    print_all(tty);
-  }
+  // Dump all events to the tty
+  static void print();
 
   // Logs a generic message with timestamp and format as printf.
   static void log(Thread* thread, const char* format, ...);
@@ -255,6 +250,7 @@
   out->print_cr("%s (%d events):", _name, _count);
   if (_count == 0) {
     out->print_cr("No events");
+    out->cr();
     return;
   }