Mercurial > jdk8-mips64-public > hotspot / file revision

 /*

  * 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

  * under the terms of the GNU General Public License version 2 only, as

  * published by the Free Software Foundation.

  *

  * This code is distributed in the hope that it will be useful, but WITHOUT

  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

  * version 2 for more details (a copy is included in the LICENSE file that

  * accompanied this code).

  *

  * You should have received a copy of the GNU General Public License version

  * 2 along with this work; if not, write to the Free Software Foundation,

  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

  *

  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

  * or visit www.oracle.com if you need additional information or have any

  * questions.

  *

  */

 #include "precompiled.hpp"

 #include "asm/macroAssembler.hpp"

 #include "asm/macroAssembler.inline.hpp"

 #include "code/debugInfoRec.hpp"

 #include "code/icBuffer.hpp"

 #include "code/vtableStubs.hpp"

 #include "interpreter/interpreter.hpp"

 #include "oops/compiledICHolder.hpp"

 #include "prims/jvmtiRedefineClassesTrace.hpp"

 #include "runtime/sharedRuntime.hpp"

 #include "runtime/vframeArray.hpp"

 #include "vmreg_x86.inline.hpp"

 #ifdef COMPILER1

 #include "c1/c1_Runtime1.hpp"

 #endif

 #ifdef COMPILER2

 #include "opto/runtime.hpp"

 #endif

 #define __ masm->

 const int StackAlignmentInSlots = StackAlignmentInBytes / VMRegImpl::stack_slot_size;

 class RegisterSaver {

   // Capture info about frame layout

 #define DEF_XMM_OFFS(regnum) xmm ## regnum ## _off = xmm_off + (regnum)*16/BytesPerInt, xmm ## regnum ## H_off

   enum layout {

                 fpu_state_off = 0,

                 fpu_state_end = fpu_state_off+FPUStateSizeInWords,

                 st0_off, st0H_off,

                 st1_off, st1H_off,

                 st2_off, st2H_off,

                 st3_off, st3H_off,

                 st4_off, st4H_off,

                 st5_off, st5H_off,

                 st6_off, st6H_off,

                 st7_off, st7H_off,

                 xmm_off,

                 DEF_XMM_OFFS(0),

                 DEF_XMM_OFFS(1),

                 DEF_XMM_OFFS(2),

                 DEF_XMM_OFFS(3),

                 DEF_XMM_OFFS(4),

                 DEF_XMM_OFFS(5),

                 DEF_XMM_OFFS(6),

                 DEF_XMM_OFFS(7),

                 flags_off = xmm7_off + 16/BytesPerInt + 1, // 16-byte stack alignment fill word

                 rdi_off,

                 rsi_off,

                 ignore_off,  // extra copy of rbp,

                 rsp_off,

                 rbx_off,

                 rdx_off,

                 rcx_off,

                 rax_off,

                 // The frame sender code expects that rbp will be in the "natural" place and

                 // will override any oopMap setting for it. We must therefore force the layout

                 // so that it agrees with the frame sender code.

                 rbp_off,

                 return_off,      // slot for return address

                 reg_save_size };

   enum { FPU_regs_live = flags_off - fpu_state_end };

   public:

   static OopMap* save_live_registers(MacroAssembler* masm, int additional_frame_words,

                                      int* total_frame_words, bool verify_fpu = true, bool save_vectors = false);

   static void restore_live_registers(MacroAssembler* masm, bool restore_vectors = false);

   static int rax_offset() { return rax_off; }

   static int rbx_offset() { return rbx_off; }

   // Offsets into the register save area

   // Used by deoptimization when it is managing result register

   // values on its own

   static int raxOffset(void) { return rax_off; }

   static int rdxOffset(void) { return rdx_off; }

   static int rbxOffset(void) { return rbx_off; }

   static int xmm0Offset(void) { return xmm0_off; }

   // This really returns a slot in the fp save area, which one is not important

   static int fpResultOffset(void) { return st0_off; }

   // During deoptimization only the result register need to be restored

   // all the other values have already been extracted.

   static void restore_result_registers(MacroAssembler* masm);

 };

 OopMap* RegisterSaver::save_live_registers(MacroAssembler* masm, int additional_frame_words,

                                            int* total_frame_words, bool verify_fpu, bool save_vectors) {

   int vect_words = 0;

 #ifdef COMPILER2

   if (save_vectors) {

     assert(UseAVX > 0, "256bit vectors are supported only with AVX");

     assert(MaxVectorSize == 32, "only 256bit vectors are supported now");

     // Save upper half of YMM registes

     vect_words = 8 * 16 / wordSize;

     additional_frame_words += vect_words;

   }

 #else

   assert(!save_vectors, "vectors are generated only by C2");

 #endif

   int frame_size_in_bytes = (reg_save_size + additional_frame_words) * wordSize;

   int frame_words = frame_size_in_bytes / wordSize;

   *total_frame_words = frame_words;

   assert(FPUStateSizeInWords == 27, "update stack layout");

   // save registers, fpu state, and flags

   // We assume caller has already has return address slot on the stack

   // We push epb twice in this sequence because we want the real rbp,

   // to be under the return like a normal enter and we want to use pusha

   // We push by hand instead of pusing push

   __ enter();

   __ pusha();

   __ pushf();

   __ subptr(rsp,FPU_regs_live*wordSize); // Push FPU registers space

   __ push_FPU_state();          // Save FPU state & init

   if (verify_fpu) {

     // Some stubs may have non standard FPU control word settings so

     // only check and reset the value when it required to be the

     // standard value.  The safepoint blob in particular can be used

     // in methods which are using the 24 bit control word for

     // optimized float math.

 #ifdef ASSERT

     // Make sure the control word has the expected value

     Label ok;

     __ cmpw(Address(rsp, 0), StubRoutines::fpu_cntrl_wrd_std());

     __ jccb(Assembler::equal, ok);

     __ stop("corrupted control word detected");

     __ bind(ok);

 #endif

     // Reset the control word to guard against exceptions being unmasked

     // since fstp_d can cause FPU stack underflow exceptions.  Write it

     // into the on stack copy and then reload that to make sure that the

     // current and future values are correct.

     __ movw(Address(rsp, 0), StubRoutines::fpu_cntrl_wrd_std());

   }

   __ frstor(Address(rsp, 0));

   if (!verify_fpu) {

     // Set the control word so that exceptions are masked for the

     // following code.

     __ fldcw(ExternalAddress(StubRoutines::addr_fpu_cntrl_wrd_std()));

   }

   // Save the FPU registers in de-opt-able form

   __ fstp_d(Address(rsp, st0_off*wordSize)); // st(0)

   __ fstp_d(Address(rsp, st1_off*wordSize)); // st(1)

   __ fstp_d(Address(rsp, st2_off*wordSize)); // st(2)

   __ fstp_d(Address(rsp, st3_off*wordSize)); // st(3)

   __ fstp_d(Address(rsp, st4_off*wordSize)); // st(4)

   __ fstp_d(Address(rsp, st5_off*wordSize)); // st(5)

   __ fstp_d(Address(rsp, st6_off*wordSize)); // st(6)

   __ fstp_d(Address(rsp, st7_off*wordSize)); // st(7)

   if( UseSSE == 1 ) {           // Save the XMM state

     __ movflt(Address(rsp,xmm0_off*wordSize),xmm0);

     __ movflt(Address(rsp,xmm1_off*wordSize),xmm1);

     __ movflt(Address(rsp,xmm2_off*wordSize),xmm2);

     __ movflt(Address(rsp,xmm3_off*wordSize),xmm3);

     __ movflt(Address(rsp,xmm4_off*wordSize),xmm4);

     __ movflt(Address(rsp,xmm5_off*wordSize),xmm5);

     __ movflt(Address(rsp,xmm6_off*wordSize),xmm6);

     __ movflt(Address(rsp,xmm7_off*wordSize),xmm7);

   } else if( UseSSE >= 2 ) {

     // Save whole 128bit (16 bytes) XMM regiters

     __ movdqu(Address(rsp,xmm0_off*wordSize),xmm0);

     __ movdqu(Address(rsp,xmm1_off*wordSize),xmm1);

     __ movdqu(Address(rsp,xmm2_off*wordSize),xmm2);

     __ movdqu(Address(rsp,xmm3_off*wordSize),xmm3);

     __ movdqu(Address(rsp,xmm4_off*wordSize),xmm4);

     __ movdqu(Address(rsp,xmm5_off*wordSize),xmm5);

     __ movdqu(Address(rsp,xmm6_off*wordSize),xmm6);

     __ movdqu(Address(rsp,xmm7_off*wordSize),xmm7);

   }

   if (vect_words > 0) {

     assert(vect_words*wordSize == 128, "");

     __ subptr(rsp, 128); // Save upper half of YMM registes

     __ vextractf128h(Address(rsp,  0),xmm0);

     __ vextractf128h(Address(rsp, 16),xmm1);

     __ vextractf128h(Address(rsp, 32),xmm2);

     __ vextractf128h(Address(rsp, 48),xmm3);

     __ vextractf128h(Address(rsp, 64),xmm4);

     __ vextractf128h(Address(rsp, 80),xmm5);

     __ vextractf128h(Address(rsp, 96),xmm6);

     __ vextractf128h(Address(rsp,112),xmm7);

   }

   // Set an oopmap for the call site.  This oopmap will map all

   // oop-registers and debug-info registers as callee-saved.  This

   // will allow deoptimization at this safepoint to find all possible

   // debug-info recordings, as well as let GC find all oops.

   OopMapSet *oop_maps = new OopMapSet();

   OopMap* map =  new OopMap( frame_words, 0 );

 #define STACK_OFFSET(x) VMRegImpl::stack2reg((x) + additional_frame_words)

   map->set_callee_saved(STACK_OFFSET( rax_off), rax->as_VMReg());

   map->set_callee_saved(STACK_OFFSET( rcx_off), rcx->as_VMReg());

   map->set_callee_saved(STACK_OFFSET( rdx_off), rdx->as_VMReg());

   map->set_callee_saved(STACK_OFFSET( rbx_off), rbx->as_VMReg());

   // rbp, location is known implicitly, no oopMap

   map->set_callee_saved(STACK_OFFSET( rsi_off), rsi->as_VMReg());

   map->set_callee_saved(STACK_OFFSET( rdi_off), rdi->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st0_off), as_FloatRegister(0)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st1_off), as_FloatRegister(1)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st2_off), as_FloatRegister(2)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st3_off), as_FloatRegister(3)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st4_off), as_FloatRegister(4)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st5_off), as_FloatRegister(5)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st6_off), as_FloatRegister(6)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(st7_off), as_FloatRegister(7)->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm0_off), xmm0->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm1_off), xmm1->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm2_off), xmm2->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm3_off), xmm3->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm4_off), xmm4->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm5_off), xmm5->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm6_off), xmm6->as_VMReg());

   map->set_callee_saved(STACK_OFFSET(xmm7_off), xmm7->as_VMReg());

   // %%% This is really a waste but we'll keep things as they were for now

   if (true) {

 #define NEXTREG(x) (x)->as_VMReg()->next()

     map->set_callee_saved(STACK_OFFSET(st0H_off), NEXTREG(as_FloatRegister(0)));

     map->set_callee_saved(STACK_OFFSET(st1H_off), NEXTREG(as_FloatRegister(1)));

     map->set_callee_saved(STACK_OFFSET(st2H_off), NEXTREG(as_FloatRegister(2)));

     map->set_callee_saved(STACK_OFFSET(st3H_off), NEXTREG(as_FloatRegister(3)));

     map->set_callee_saved(STACK_OFFSET(st4H_off), NEXTREG(as_FloatRegister(4)));

     map->set_callee_saved(STACK_OFFSET(st5H_off), NEXTREG(as_FloatRegister(5)));

     map->set_callee_saved(STACK_OFFSET(st6H_off), NEXTREG(as_FloatRegister(6)));

     map->set_callee_saved(STACK_OFFSET(st7H_off), NEXTREG(as_FloatRegister(7)));

     map->set_callee_saved(STACK_OFFSET(xmm0H_off), NEXTREG(xmm0));

     map->set_callee_saved(STACK_OFFSET(xmm1H_off), NEXTREG(xmm1));

     map->set_callee_saved(STACK_OFFSET(xmm2H_off), NEXTREG(xmm2));

     map->set_callee_saved(STACK_OFFSET(xmm3H_off), NEXTREG(xmm3));

     map->set_callee_saved(STACK_OFFSET(xmm4H_off), NEXTREG(xmm4));

     map->set_callee_saved(STACK_OFFSET(xmm5H_off), NEXTREG(xmm5));

     map->set_callee_saved(STACK_OFFSET(xmm6H_off), NEXTREG(xmm6));

     map->set_callee_saved(STACK_OFFSET(xmm7H_off), NEXTREG(xmm7));

 #undef NEXTREG

 #undef STACK_OFFSET

   }

   return map;

 }

 void RegisterSaver::restore_live_registers(MacroAssembler* masm, bool restore_vectors) {

   // Recover XMM & FPU state

   int additional_frame_bytes = 0;

 #ifdef COMPILER2

   if (restore_vectors) {

     assert(UseAVX > 0, "256bit vectors are supported only with AVX");

     assert(MaxVectorSize == 32, "only 256bit vectors are supported now");

     additional_frame_bytes = 128;

   }

 #else

   assert(!restore_vectors, "vectors are generated only by C2");

 #endif

   if (UseSSE == 1) {

     assert(additional_frame_bytes == 0, "");

     __ movflt(xmm0,Address(rsp,xmm0_off*wordSize));

     __ movflt(xmm1,Address(rsp,xmm1_off*wordSize));

     __ movflt(xmm2,Address(rsp,xmm2_off*wordSize));

     __ movflt(xmm3,Address(rsp,xmm3_off*wordSize));

     __ movflt(xmm4,Address(rsp,xmm4_off*wordSize));

     __ movflt(xmm5,Address(rsp,xmm5_off*wordSize));

     __ movflt(xmm6,Address(rsp,xmm6_off*wordSize));

     __ movflt(xmm7,Address(rsp,xmm7_off*wordSize));

   } else if (UseSSE >= 2) {

 #define STACK_ADDRESS(x) Address(rsp,(x)*wordSize + additional_frame_bytes)

     __ movdqu(xmm0,STACK_ADDRESS(xmm0_off));

     __ movdqu(xmm1,STACK_ADDRESS(xmm1_off));

     __ movdqu(xmm2,STACK_ADDRESS(xmm2_off));

     __ movdqu(xmm3,STACK_ADDRESS(xmm3_off));

     __ movdqu(xmm4,STACK_ADDRESS(xmm4_off));

     __ movdqu(xmm5,STACK_ADDRESS(xmm5_off));

     __ movdqu(xmm6,STACK_ADDRESS(xmm6_off));

     __ movdqu(xmm7,STACK_ADDRESS(xmm7_off));

 #undef STACK_ADDRESS

   }

   if (restore_vectors) {

     // Restore upper half of YMM registes.

     assert(additional_frame_bytes == 128, "");

     __ vinsertf128h(xmm0, Address(rsp,  0));

     __ vinsertf128h(xmm1, Address(rsp, 16));

     __ vinsertf128h(xmm2, Address(rsp, 32));

     __ vinsertf128h(xmm3, Address(rsp, 48));

     __ vinsertf128h(xmm4, Address(rsp, 64));

     __ vinsertf128h(xmm5, Address(rsp, 80));

     __ vinsertf128h(xmm6, Address(rsp, 96));

     __ vinsertf128h(xmm7, Address(rsp,112));

     __ addptr(rsp, additional_frame_bytes);

   }

   __ pop_FPU_state();

   __ addptr(rsp, FPU_regs_live*wordSize); // Pop FPU registers

   __ popf();

   __ popa();

   // Get the rbp, described implicitly by the frame sender code (no oopMap)

   __ pop(rbp);

 }

 void RegisterSaver::restore_result_registers(MacroAssembler* masm) {

   // Just restore result register. Only used by deoptimization. By

   // now any callee save register that needs to be restore to a c2

   // caller of the deoptee has been extracted into the vframeArray

   // and will be stuffed into the c2i adapter we create for later

   // restoration so only result registers need to be restored here.

   //

   __ frstor(Address(rsp, 0));      // Restore fpu state

   // Recover XMM & FPU state

   if( UseSSE == 1 ) {

     __ movflt(xmm0, Address(rsp, xmm0_off*wordSize));

   } else if( UseSSE >= 2 ) {

     __ movdbl(xmm0, Address(rsp, xmm0_off*wordSize));

   }

   __ movptr(rax, Address(rsp, rax_off*wordSize));

   __ movptr(rdx, Address(rsp, rdx_off*wordSize));

   // Pop all of the register save are off the stack except the return address

   __ addptr(rsp, return_off * wordSize);

 }

 // Is vector's size (in bytes) bigger than a size saved by default?

 // 16 bytes XMM registers are saved by default using SSE2 movdqu instructions.

 // Note, MaxVectorSize == 0 with UseSSE < 2 and vectors are not generated.

 bool SharedRuntime::is_wide_vector(int size) {

   return size > 16;

 }

 // The java_calling_convention describes stack locations as ideal slots on

 // a frame with no abi restrictions. Since we must observe abi restrictions

 // (like the placement of the register window) the slots must be biased by

 // the following value.

 static int reg2offset_in(VMReg r) {

   // Account for saved rbp, and return address

   // This should really be in_preserve_stack_slots

   return (r->reg2stack() + 2) * VMRegImpl::stack_slot_size;

 }

 static int reg2offset_out(VMReg r) {

   return (r->reg2stack() + SharedRuntime::out_preserve_stack_slots()) * VMRegImpl::stack_slot_size;

 }

 // ---------------------------------------------------------------------------

 // Read the array of BasicTypes from a signature, and compute where the

 // arguments should go.  Values in the VMRegPair regs array refer to 4-byte

 // quantities.  Values less than SharedInfo::stack0 are registers, those above

 // refer to 4-byte stack slots.  All stack slots are based off of the stack pointer

 // as framesizes are fixed.

 // VMRegImpl::stack0 refers to the first slot 0(sp).

 // and VMRegImpl::stack0+1 refers to the memory word 4-byes higher.  Register

 // up to RegisterImpl::number_of_registers) are the 32-bit

 // integer registers.

 // Pass first two oop/int args in registers ECX and EDX.

 // Pass first two float/double args in registers XMM0 and XMM1.

 // Doubles have precedence, so if you pass a mix of floats and doubles

 // the doubles will grab the registers before the floats will.

 // Note: the INPUTS in sig_bt are in units of Java argument words, which are

 // either 32-bit or 64-bit depending on the build.  The OUTPUTS are in 32-bit

 // units regardless of build. Of course for i486 there is no 64 bit build

 // ---------------------------------------------------------------------------

 // The compiled Java calling convention.

 // Pass first two oop/int args in registers ECX and EDX.

 // Pass first two float/double args in registers XMM0 and XMM1.

 // Doubles have precedence, so if you pass a mix of floats and doubles

 // the doubles will grab the registers before the floats will.

 int SharedRuntime::java_calling_convention(const BasicType *sig_bt,

                                            VMRegPair *regs,

                                            int total_args_passed,

                                            int is_outgoing) {

   uint    stack = 0;          // Starting stack position for args on stack

   // Pass first two oop/int args in registers ECX and EDX.

   uint reg_arg0 = 9999;

   uint reg_arg1 = 9999;

   // Pass first two float/double args in registers XMM0 and XMM1.

   // Doubles have precedence, so if you pass a mix of floats and doubles

   // the doubles will grab the registers before the floats will.

   // CNC - TURNED OFF FOR non-SSE.

   //       On Intel we have to round all doubles (and most floats) at

   //       call sites by storing to the stack in any case.

   // UseSSE=0 ==> Don't Use ==> 9999+0

   // UseSSE=1 ==> Floats only ==> 9999+1

   // UseSSE>=2 ==> Floats or doubles ==> 9999+2

   enum { fltarg_dontuse = 9999+0, fltarg_float_only = 9999+1, fltarg_flt_dbl = 9999+2 };

   uint fargs = (UseSSE>=2) ? 2 : UseSSE;

   uint freg_arg0 = 9999+fargs;

   uint freg_arg1 = 9999+fargs;

   // Pass doubles & longs aligned on the stack.  First count stack slots for doubles

   int i;

   for( i = 0; i < total_args_passed; i++) {

     if( sig_bt[i] == T_DOUBLE ) {

       // first 2 doubles go in registers

       if( freg_arg0 == fltarg_flt_dbl ) freg_arg0 = i;

       else if( freg_arg1 == fltarg_flt_dbl ) freg_arg1 = i;

       else // Else double is passed low on the stack to be aligned.

         stack += 2;

     } else if( sig_bt[i] == T_LONG ) {

       stack += 2;

     }

   }

   int dstack = 0;             // Separate counter for placing doubles

   // Now pick where all else goes.

   for( i = 0; i < total_args_passed; i++) {

     // From the type and the argument number (count) compute the location

     switch( sig_bt[i] ) {

     case T_SHORT:

     case T_CHAR:

     case T_BYTE:

     case T_BOOLEAN:

     case T_INT:

     case T_ARRAY:

     case T_OBJECT:

     case T_ADDRESS:

       if( reg_arg0 == 9999 )  {

         reg_arg0 = i;

         regs[i].set1(rcx->as_VMReg());

       } else if( reg_arg1 == 9999 )  {

         reg_arg1 = i;

         regs[i].set1(rdx->as_VMReg());

       } else {

         regs[i].set1(VMRegImpl::stack2reg(stack++));

       }

       break;

     case T_FLOAT:

       if( freg_arg0 == fltarg_flt_dbl || freg_arg0 == fltarg_float_only ) {

         freg_arg0 = i;

         regs[i].set1(xmm0->as_VMReg());

       } else if( freg_arg1 == fltarg_flt_dbl || freg_arg1 == fltarg_float_only ) {

         freg_arg1 = i;

         regs[i].set1(xmm1->as_VMReg());

       } else {

         regs[i].set1(VMRegImpl::stack2reg(stack++));

       }

       break;

     case T_LONG:

       assert(sig_bt[i+1] == T_VOID, "missing Half" );

       regs[i].set2(VMRegImpl::stack2reg(dstack));

       dstack += 2;

       break;

     case T_DOUBLE:

       assert(sig_bt[i+1] == T_VOID, "missing Half" );

       if( freg_arg0 == (uint)i ) {

         regs[i].set2(xmm0->as_VMReg());

       } else if( freg_arg1 == (uint)i ) {

         regs[i].set2(xmm1->as_VMReg());

       } else {

         regs[i].set2(VMRegImpl::stack2reg(dstack));

         dstack += 2;

       }

       break;

     case T_VOID: regs[i].set_bad(); break;

       break;

     default:

       ShouldNotReachHere();

       break;

     }

   }

   // return value can be odd number of VMRegImpl stack slots make multiple of 2

   return round_to(stack, 2);

 }

 // Patch the callers callsite with entry to compiled code if it exists.

 static void patch_callers_callsite(MacroAssembler *masm) {

   Label L;

   __ cmpptr(Address(rbx, in_bytes(Method::code_offset())), (int32_t)NULL_WORD);

   __ jcc(Assembler::equal, L);

   // Schedule the branch target address early.

   // Call into the VM to patch the caller, then jump to compiled callee

   // rax, isn't live so capture return address while we easily can

   __ movptr(rax, Address(rsp, 0));

   __ pusha();

   __ pushf();

   if (UseSSE == 1) {

     __ subptr(rsp, 2*wordSize);

     __ movflt(Address(rsp, 0), xmm0);

     __ movflt(Address(rsp, wordSize), xmm1);

   }

   if (UseSSE >= 2) {

     __ subptr(rsp, 4*wordSize);

     __ movdbl(Address(rsp, 0), xmm0);

     __ movdbl(Address(rsp, 2*wordSize), xmm1);

   }

 #ifdef COMPILER2

   // C2 may leave the stack dirty if not in SSE2+ mode

   if (UseSSE >= 2) {

     __ verify_FPU(0, "c2i transition should have clean FPU stack");

   } else {

     __ empty_FPU_stack();

   }

 #endif /* COMPILER2 */

   // VM needs caller's callsite

   __ push(rax);

   // VM needs target method

   __ push(rbx);

   __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::fixup_callers_callsite)));

   __ addptr(rsp, 2*wordSize);

   if (UseSSE == 1) {

     __ movflt(xmm0, Address(rsp, 0));

     __ movflt(xmm1, Address(rsp, wordSize));

     __ addptr(rsp, 2*wordSize);

   }

   if (UseSSE >= 2) {

     __ movdbl(xmm0, Address(rsp, 0));

     __ movdbl(xmm1, Address(rsp, 2*wordSize));

     __ addptr(rsp, 4*wordSize);

   }

   __ popf();

   __ popa();

   __ bind(L);

 }

 static void move_c2i_double(MacroAssembler *masm, XMMRegister r, int st_off) {

   int next_off = st_off - Interpreter::stackElementSize;

   __ movdbl(Address(rsp, next_off), r);

 }

 static void gen_c2i_adapter(MacroAssembler *masm,

                             int total_args_passed,

                             int comp_args_on_stack,

                             const BasicType *sig_bt,

                             const VMRegPair *regs,

                             Label& skip_fixup) {

   // Before we get into the guts of the C2I adapter, see if we should be here

   // at all.  We've come from compiled code and are attempting to jump to the

   // interpreter, which means the caller made a static call to get here

   // (vcalls always get a compiled target if there is one).  Check for a

   // compiled target.  If there is one, we need to patch the caller's call.

   patch_callers_callsite(masm);

   __ bind(skip_fixup);

 #ifdef COMPILER2

   // C2 may leave the stack dirty if not in SSE2+ mode

   if (UseSSE >= 2) {

     __ verify_FPU(0, "c2i transition should have clean FPU stack");

   } else {

     __ empty_FPU_stack();

   }

 #endif /* COMPILER2 */

   // Since all args are passed on the stack, total_args_passed * interpreter_

   // stack_element_size  is the

   // space we need.

   int extraspace = total_args_passed * Interpreter::stackElementSize;

   // Get return address

   __ pop(rax);

   // set senderSP value

   __ movptr(rsi, rsp);

   __ subptr(rsp, extraspace);

   // Now write the args into the outgoing interpreter space

   for (int i = 0; i < total_args_passed; i++) {

     if (sig_bt[i] == T_VOID) {

       assert(i > 0 && (sig_bt[i-1] == T_LONG || sig_bt[i-1] == T_DOUBLE), "missing half");

       continue;

     }

     // st_off points to lowest address on stack.

     int st_off = ((total_args_passed - 1) - i) * Interpreter::stackElementSize;

     int next_off = st_off - Interpreter::stackElementSize;

     // Say 4 args:

     // i   st_off

     // 0   12 T_LONG

     // 1    8 T_VOID

     // 2    4 T_OBJECT

     // 3    0 T_BOOL

     VMReg r_1 = regs[i].first();

     VMReg r_2 = regs[i].second();

     if (!r_1->is_valid()) {

       assert(!r_2->is_valid(), "");

       continue;

     }

     if (r_1->is_stack()) {

       // memory to memory use fpu stack top

       int ld_off = r_1->reg2stack() * VMRegImpl::stack_slot_size + extraspace;

       if (!r_2->is_valid()) {

         __ movl(rdi, Address(rsp, ld_off));

         __ movptr(Address(rsp, st_off), rdi);

       } else {

         // ld_off == LSW, ld_off+VMRegImpl::stack_slot_size == MSW

         // st_off == MSW, st_off-wordSize == LSW

         __ movptr(rdi, Address(rsp, ld_off));

         __ movptr(Address(rsp, next_off), rdi);

 #ifndef _LP64

         __ movptr(rdi, Address(rsp, ld_off + wordSize));

         __ movptr(Address(rsp, st_off), rdi);

 #else

 #ifdef ASSERT

         // Overwrite the unused slot with known junk

         __ mov64(rax, CONST64(0xdeadffffdeadaaaa));

         __ movptr(Address(rsp, st_off), rax);

 #endif /* ASSERT */

 #endif // _LP64

       }

     } else if (r_1->is_Register()) {

       Register r = r_1->as_Register();

       if (!r_2->is_valid()) {

         __ movl(Address(rsp, st_off), r);

       } else {

         // long/double in gpr

         NOT_LP64(ShouldNotReachHere());

         // Two VMRegs can be T_OBJECT, T_ADDRESS, T_DOUBLE, T_LONG

         // T_DOUBLE and T_LONG use two slots in the interpreter

         if ( sig_bt[i] == T_LONG || sig_bt[i] == T_DOUBLE) {

           // long/double in gpr

 #ifdef ASSERT

           // Overwrite the unused slot with known junk

           LP64_ONLY(__ mov64(rax, CONST64(0xdeadffffdeadaaab)));

           __ movptr(Address(rsp, st_off), rax);

 #endif /* ASSERT */

           __ movptr(Address(rsp, next_off), r);

         } else {

           __ movptr(Address(rsp, st_off), r);

         }

       }

     } else {

       assert(r_1->is_XMMRegister(), "");

       if (!r_2->is_valid()) {

         __ movflt(Address(rsp, st_off), r_1->as_XMMRegister());

       } else {

         assert(sig_bt[i] == T_DOUBLE || sig_bt[i] == T_LONG, "wrong type");

         move_c2i_double(masm, r_1->as_XMMRegister(), st_off);

       }

     }

   }

   // Schedule the branch target address early.

   __ movptr(rcx, Address(rbx, in_bytes(Method::interpreter_entry_offset())));

   // And repush original return address

   __ push(rax);

   __ jmp(rcx);

 }

 static void move_i2c_double(MacroAssembler *masm, XMMRegister r, Register saved_sp, int ld_off) {

   int next_val_off = ld_off - Interpreter::stackElementSize;

   __ movdbl(r, Address(saved_sp, next_val_off));

 }

 static void range_check(MacroAssembler* masm, Register pc_reg, Register temp_reg,

                         address code_start, address code_end,

                         Label& L_ok) {

   Label L_fail;

   __ lea(temp_reg, ExternalAddress(code_start));

   __ cmpptr(pc_reg, temp_reg);

   __ jcc(Assembler::belowEqual, L_fail);

   __ lea(temp_reg, ExternalAddress(code_end));

   __ cmpptr(pc_reg, temp_reg);

   __ jcc(Assembler::below, L_ok);

   __ bind(L_fail);

 }

 static void gen_i2c_adapter(MacroAssembler *masm,

                             int total_args_passed,

                             int comp_args_on_stack,

                             const BasicType *sig_bt,

                             const VMRegPair *regs) {

   // Note: rsi contains the senderSP on entry. We must preserve it since

   // we may do a i2c -> c2i transition if we lose a race where compiled

   // code goes non-entrant while we get args ready.

   // Adapters can be frameless because they do not require the caller

   // to perform additional cleanup work, such as correcting the stack pointer.

   // An i2c adapter is frameless because the *caller* frame, which is interpreted,

   // routinely repairs its own stack pointer (from interpreter_frame_last_sp),

   // even if a callee has modified the stack pointer.

   // A c2i adapter is frameless because the *callee* frame, which is interpreted,

   // routinely repairs its caller's stack pointer (from sender_sp, which is set

   // up via the senderSP register).

   // In other words, if *either* the caller or callee is interpreted, we can

   // get the stack pointer repaired after a call.

   // This is why c2i and i2c adapters cannot be indefinitely composed.

   // In particular, if a c2i adapter were to somehow call an i2c adapter,

   // both caller and callee would be compiled methods, and neither would

   // clean up the stack pointer changes performed by the two adapters.

   // If this happens, control eventually transfers back to the compiled

   // caller, but with an uncorrected stack, causing delayed havoc.

   // Pick up the return address

   __ movptr(rax, Address(rsp, 0));

   if (VerifyAdapterCalls &&

       (Interpreter::code() != NULL || StubRoutines::code1() != NULL)) {

     // So, let's test for cascading c2i/i2c adapters right now.

     //  assert(Interpreter::contains($return_addr) ||

     //         StubRoutines::contains($return_addr),

     //         "i2c adapter must return to an interpreter frame");

     __ block_comment("verify_i2c { ");

     Label L_ok;

     if (Interpreter::code() != NULL)

       range_check(masm, rax, rdi,

                   Interpreter::code()->code_start(), Interpreter::code()->code_end(),

                   L_ok);

     if (StubRoutines::code1() != NULL)

       range_check(masm, rax, rdi,

                   StubRoutines::code1()->code_begin(), StubRoutines::code1()->code_end(),

                   L_ok);

     if (StubRoutines::code2() != NULL)

       range_check(masm, rax, rdi,

                   StubRoutines::code2()->code_begin(), StubRoutines::code2()->code_end(),

                   L_ok);

     const char* msg = "i2c adapter must return to an interpreter frame";

     __ block_comment(msg);

     __ stop(msg);

     __ bind(L_ok);

     __ block_comment("} verify_i2ce ");

   }

   // Must preserve original SP for loading incoming arguments because

   // we need to align the outgoing SP for compiled code.

   __ movptr(rdi, rsp);

   // Cut-out for having no stack args.  Since up to 2 int/oop args are passed

   // in registers, we will occasionally have no stack args.

   int comp_words_on_stack = 0;

   if (comp_args_on_stack) {

     // Sig words on the stack are greater-than VMRegImpl::stack0.  Those in

     // registers are below.  By subtracting stack0, we either get a negative

     // number (all values in registers) or the maximum stack slot accessed.

     // int comp_args_on_stack = VMRegImpl::reg2stack(max_arg);

     // Convert 4-byte stack slots to words.

     comp_words_on_stack = round_to(comp_args_on_stack*4, wordSize)>>LogBytesPerWord;

     // Round up to miminum stack alignment, in wordSize

     comp_words_on_stack = round_to(comp_words_on_stack, 2);

     __ subptr(rsp, comp_words_on_stack * wordSize);

   }

   // Align the outgoing SP

   __ andptr(rsp, -(StackAlignmentInBytes));

   // push the return address on the stack (note that pushing, rather

   // than storing it, yields the correct frame alignment for the callee)

   __ push(rax);

   // Put saved SP in another register

   const Register saved_sp = rax;

   __ movptr(saved_sp, rdi);

   // Will jump to the compiled code just as if compiled code was doing it.

   // Pre-load the register-jump target early, to schedule it better.

   __ movptr(rdi, Address(rbx, in_bytes(Method::from_compiled_offset())));

   // Now generate the shuffle code.  Pick up all register args and move the

   // rest through the floating point stack top.

   for (int i = 0; i < total_args_passed; i++) {

     if (sig_bt[i] == T_VOID) {

       // Longs and doubles are passed in native word order, but misaligned

       // in the 32-bit build.

       assert(i > 0 && (sig_bt[i-1] == T_LONG || sig_bt[i-1] == T_DOUBLE), "missing half");

       continue;

     }

     // Pick up 0, 1 or 2 words from SP+offset.

     assert(!regs[i].second()->is_valid() || regs[i].first()->next() == regs[i].second(),

             "scrambled load targets?");

     // Load in argument order going down.

     int ld_off = (total_args_passed - i) * Interpreter::stackElementSize;

     // Point to interpreter value (vs. tag)

     int next_off = ld_off - Interpreter::stackElementSize;

     //

     //

     //

     VMReg r_1 = regs[i].first();

     VMReg r_2 = regs[i].second();

     if (!r_1->is_valid()) {

       assert(!r_2->is_valid(), "");

       continue;

     }

     if (r_1->is_stack()) {

       // Convert stack slot to an SP offset (+ wordSize to account for return address )

       int st_off = regs[i].first()->reg2stack()*VMRegImpl::stack_slot_size + wordSize;

       // We can use rsi as a temp here because compiled code doesn't need rsi as an input

       // and if we end up going thru a c2i because of a miss a reasonable value of rsi

       // we be generated.

       if (!r_2->is_valid()) {

         // __ fld_s(Address(saved_sp, ld_off));

         // __ fstp_s(Address(rsp, st_off));

         __ movl(rsi, Address(saved_sp, ld_off));

         __ movptr(Address(rsp, st_off), rsi);

       } else {

         // Interpreter local[n] == MSW, local[n+1] == LSW however locals

         // are accessed as negative so LSW is at LOW address

         // ld_off is MSW so get LSW

         // st_off is LSW (i.e. reg.first())

         // __ fld_d(Address(saved_sp, next_off));

         // __ fstp_d(Address(rsp, st_off));

         //

         // We are using two VMRegs. This can be either T_OBJECT, T_ADDRESS, T_LONG, or T_DOUBLE

         // the interpreter allocates two slots but only uses one for thr T_LONG or T_DOUBLE case

         // So we must adjust where to pick up the data to match the interpreter.

         //

         // Interpreter local[n] == MSW, local[n+1] == LSW however locals

         // are accessed as negative so LSW is at LOW address

         // ld_off is MSW so get LSW

         const int offset = (NOT_LP64(true ||) sig_bt[i]==T_LONG||sig_bt[i]==T_DOUBLE)?

                            next_off : ld_off;

         __ movptr(rsi, Address(saved_sp, offset));

         __ movptr(Address(rsp, st_off), rsi);

 #ifndef _LP64

         __ movptr(rsi, Address(saved_sp, ld_off));

         __ movptr(Address(rsp, st_off + wordSize), rsi);

 #endif // _LP64

       }

     } else if (r_1->is_Register()) {  // Register argument

       Register r = r_1->as_Register();

       assert(r != rax, "must be different");

       if (r_2->is_valid()) {

         //

         // We are using two VMRegs. This can be either T_OBJECT, T_ADDRESS, T_LONG, or T_DOUBLE

         // the interpreter allocates two slots but only uses one for thr T_LONG or T_DOUBLE case

         // So we must adjust where to pick up the data to match the interpreter.

         const int offset = (NOT_LP64(true ||) sig_bt[i]==T_LONG||sig_bt[i]==T_DOUBLE)?

                            next_off : ld_off;

         // this can be a misaligned move

         __ movptr(r, Address(saved_sp, offset));

 #ifndef _LP64

         assert(r_2->as_Register() != rax, "need another temporary register");

         // Remember r_1 is low address (and LSB on x86)

         // So r_2 gets loaded from high address regardless of the platform

         __ movptr(r_2->as_Register(), Address(saved_sp, ld_off));

 #endif // _LP64

       } else {

         __ movl(r, Address(saved_sp, ld_off));

       }

     } else {

       assert(r_1->is_XMMRegister(), "");

       if (!r_2->is_valid()) {

         __ movflt(r_1->as_XMMRegister(), Address(saved_sp, ld_off));

       } else {

         move_i2c_double(masm, r_1->as_XMMRegister(), saved_sp, ld_off);

       }

     }

   }

   // 6243940 We might end up in handle_wrong_method if

   // the callee is deoptimized as we race thru here. If that

   // happens we don't want to take a safepoint because the

   // caller frame will look interpreted and arguments are now

   // "compiled" so it is much better to make this transition

   // invisible to the stack walking code. Unfortunately if

   // we try and find the callee by normal means a safepoint

   // is possible. So we stash the desired callee in the thread

   // and the vm will find there should this case occur.

   __ get_thread(rax);

   __ movptr(Address(rax, JavaThread::callee_target_offset()), rbx);

   // move Method* to rax, in case we end up in an c2i adapter.

   // the c2i adapters expect Method* in rax, (c2) because c2's

   // resolve stubs return the result (the method) in rax,.

   // I'd love to fix this.

   __ mov(rax, rbx);

   __ jmp(rdi);

 }

 // ---------------------------------------------------------------

 AdapterHandlerEntry* SharedRuntime::generate_i2c2i_adapters(MacroAssembler *masm,

                                                             int total_args_passed,

                                                             int comp_args_on_stack,

                                                             const BasicType *sig_bt,

                                                             const VMRegPair *regs,

                                                             AdapterFingerPrint* fingerprint) {

   address i2c_entry = __ pc();

   gen_i2c_adapter(masm, total_args_passed, comp_args_on_stack, sig_bt, regs);

   // -------------------------------------------------------------------------

   // Generate a C2I adapter.  On entry we know rbx, holds the Method* during calls

   // to the interpreter.  The args start out packed in the compiled layout.  They

   // need to be unpacked into the interpreter layout.  This will almost always

   // require some stack space.  We grow the current (compiled) stack, then repack

   // the args.  We  finally end in a jump to the generic interpreter entry point.

   // On exit from the interpreter, the interpreter will restore our SP (lest the

   // compiled code, which relys solely on SP and not EBP, get sick).

   address c2i_unverified_entry = __ pc();

   Label skip_fixup;

   Register holder = rax;

   Register receiver = rcx;

   Register temp = rbx;

   {

     Label missed;

     __ movptr(temp, Address(receiver, oopDesc::klass_offset_in_bytes()));

     __ cmpptr(temp, Address(holder, CompiledICHolder::holder_klass_offset()));

     __ movptr(rbx, Address(holder, CompiledICHolder::holder_method_offset()));

     __ jcc(Assembler::notEqual, missed);

     // Method might have been compiled since the call site was patched to

     // interpreted if that is the case treat it as a miss so we can get

     // the call site corrected.

     __ cmpptr(Address(rbx, in_bytes(Method::code_offset())), (int32_t)NULL_WORD);

     __ jcc(Assembler::equal, skip_fixup);

     __ bind(missed);

     __ jump(RuntimeAddress(SharedRuntime::get_ic_miss_stub()));

   }

   address c2i_entry = __ pc();

   gen_c2i_adapter(masm, total_args_passed, comp_args_on_stack, sig_bt, regs, skip_fixup);

   __ flush();

   return AdapterHandlerLibrary::new_entry(fingerprint, i2c_entry, c2i_entry, c2i_unverified_entry);

 }

 int SharedRuntime::c_calling_convention(const BasicType *sig_bt,

                                          VMRegPair *regs,

                                          int total_args_passed) {

 // We return the amount of VMRegImpl stack slots we need to reserve for all

 // the arguments NOT counting out_preserve_stack_slots.

   uint    stack = 0;        // All arguments on stack

   for( int i = 0; i < total_args_passed; i++) {

     // From the type and the argument number (count) compute the location

     switch( sig_bt[i] ) {

     case T_BOOLEAN:

     case T_CHAR:

     case T_FLOAT:

     case T_BYTE:

     case T_SHORT:

     case T_INT:

     case T_OBJECT:

     case T_ARRAY:

     case T_ADDRESS:

     case T_METADATA:

       regs[i].set1(VMRegImpl::stack2reg(stack++));

       break;

     case T_LONG:

     case T_DOUBLE: // The stack numbering is reversed from Java

       // Since C arguments do not get reversed, the ordering for

       // doubles on the stack must be opposite the Java convention

       assert(sig_bt[i+1] == T_VOID, "missing Half" );

       regs[i].set2(VMRegImpl::stack2reg(stack));

       stack += 2;

       break;

     case T_VOID: regs[i].set_bad(); break;

     default:

       ShouldNotReachHere();

       break;

     }

   }

   return stack;

 }

 // A simple move of integer like type

 static void simple_move32(MacroAssembler* masm, VMRegPair src, VMRegPair dst) {

   if (src.first()->is_stack()) {

     if (dst.first()->is_stack()) {

       // stack to stack

       // __ ld(FP, reg2offset(src.first()) + STACK_BIAS, L5);

       // __ st(L5, SP, reg2offset(dst.first()) + STACK_BIAS);

       __ movl2ptr(rax, Address(rbp, reg2offset_in(src.first())));

       __ movptr(Address(rsp, reg2offset_out(dst.first())), rax);

     } else {

       // stack to reg

       __ movl2ptr(dst.first()->as_Register(),  Address(rbp, reg2offset_in(src.first())));

     }

   } else if (dst.first()->is_stack()) {

     // reg to stack

     // no need to sign extend on 64bit

     __ movptr(Address(rsp, reg2offset_out(dst.first())), src.first()->as_Register());

   } else {

     if (dst.first() != src.first()) {

       __ mov(dst.first()->as_Register(), src.first()->as_Register());

     }

   }

 }

 // An oop arg. Must pass a handle not the oop itself

 static void object_move(MacroAssembler* masm,

                         OopMap* map,

                         int oop_handle_offset,

                         int framesize_in_slots,

                         VMRegPair src,

                         VMRegPair dst,

                         bool is_receiver,

                         int* receiver_offset) {

   // Because of the calling conventions we know that src can be a

   // register or a stack location. dst can only be a stack location.

   assert(dst.first()->is_stack(), "must be stack");

   // must pass a handle. First figure out the location we use as a handle

   if (src.first()->is_stack()) {

     // Oop is already on the stack as an argument

     Register rHandle = rax;

     Label nil;

     __ xorptr(rHandle, rHandle);

     __ cmpptr(Address(rbp, reg2offset_in(src.first())), (int32_t)NULL_WORD);

     __ jcc(Assembler::equal, nil);

     __ lea(rHandle, Address(rbp, reg2offset_in(src.first())));

     __ bind(nil);

     __ movptr(Address(rsp, reg2offset_out(dst.first())), rHandle);

     int offset_in_older_frame = src.first()->reg2stack() + SharedRuntime::out_preserve_stack_slots();

     map->set_oop(VMRegImpl::stack2reg(offset_in_older_frame + framesize_in_slots));

     if (is_receiver) {

       *receiver_offset = (offset_in_older_frame + framesize_in_slots) * VMRegImpl::stack_slot_size;

     }

   } else {

     // Oop is in an a register we must store it to the space we reserve

     // on the stack for oop_handles

     const Register rOop = src.first()->as_Register();

     const Register rHandle = rax;

     int oop_slot = (rOop == rcx ? 0 : 1) * VMRegImpl::slots_per_word + oop_handle_offset;

     int offset = oop_slot*VMRegImpl::stack_slot_size;

     Label skip;

     __ movptr(Address(rsp, offset), rOop);

     map->set_oop(VMRegImpl::stack2reg(oop_slot));

     __ xorptr(rHandle, rHandle);

     __ cmpptr(rOop, (int32_t)NULL_WORD);

     __ jcc(Assembler::equal, skip);

     __ lea(rHandle, Address(rsp, offset));

     __ bind(skip);

     // Store the handle parameter

     __ movptr(Address(rsp, reg2offset_out(dst.first())), rHandle);

     if (is_receiver) {

       *receiver_offset = offset;

     }

   }

 }

 // A float arg may have to do float reg int reg conversion

 static void float_move(MacroAssembler* masm, VMRegPair src, VMRegPair dst) {

   assert(!src.second()->is_valid() && !dst.second()->is_valid(), "bad float_move");

   // Because of the calling convention we know that src is either a stack location

   // or an xmm register. dst can only be a stack location.

   assert(dst.first()->is_stack() && ( src.first()->is_stack() || src.first()->is_XMMRegister()), "bad parameters");

   if (src.first()->is_stack()) {

     __ movl(rax, Address(rbp, reg2offset_in(src.first())));

     __ movptr(Address(rsp, reg2offset_out(dst.first())), rax);

   } else {

     // reg to stack

     __ movflt(Address(rsp, reg2offset_out(dst.first())), src.first()->as_XMMRegister());

   }

 }

 // A long move

 static void long_move(MacroAssembler* masm, VMRegPair src, VMRegPair dst) {

   // The only legal possibility for a long_move VMRegPair is:

   // 1: two stack slots (possibly unaligned)

   // as neither the java  or C calling convention will use registers

   // for longs.

   if (src.first()->is_stack() && dst.first()->is_stack()) {

     assert(src.second()->is_stack() && dst.second()->is_stack(), "must be all stack");

     __ movptr(rax, Address(rbp, reg2offset_in(src.first())));

     NOT_LP64(__ movptr(rbx, Address(rbp, reg2offset_in(src.second()))));

     __ movptr(Address(rsp, reg2offset_out(dst.first())), rax);

     NOT_LP64(__ movptr(Address(rsp, reg2offset_out(dst.second())), rbx));

   } else {

     ShouldNotReachHere();

   }

 }

 // A double move

 static void double_move(MacroAssembler* masm, VMRegPair src, VMRegPair dst) {

   // The only legal possibilities for a double_move VMRegPair are:

   // The painful thing here is that like long_move a VMRegPair might be

   // Because of the calling convention we know that src is either

   //   1: a single physical register (xmm registers only)

   //   2: two stack slots (possibly unaligned)

   // dst can only be a pair of stack slots.

   assert(dst.first()->is_stack() && (src.first()->is_XMMRegister() || src.first()->is_stack()), "bad args");

   if (src.first()->is_stack()) {

     // source is all stack

     __ movptr(rax, Address(rbp, reg2offset_in(src.first())));

     NOT_LP64(__ movptr(rbx, Address(rbp, reg2offset_in(src.second()))));

     __ movptr(Address(rsp, reg2offset_out(dst.first())), rax);

     NOT_LP64(__ movptr(Address(rsp, reg2offset_out(dst.second())), rbx));

   } else {

     // reg to stack

     // No worries about stack alignment

     __ movdbl(Address(rsp, reg2offset_out(dst.first())), src.first()->as_XMMRegister());

   }

 }

 void SharedRuntime::save_native_result(MacroAssembler *masm, BasicType ret_type, int frame_slots) {

   // We always ignore the frame_slots arg and just use the space just below frame pointer

   // which by this time is free to use

   switch (ret_type) {

   case T_FLOAT:

     __ fstp_s(Address(rbp, -wordSize));

     break;

   case T_DOUBLE:

     __ fstp_d(Address(rbp, -2*wordSize));

     break;

   case T_VOID:  break;

   case T_LONG:

     __ movptr(Address(rbp, -wordSize), rax);

     NOT_LP64(__ movptr(Address(rbp, -2*wordSize), rdx));

     break;

   default: {

     __ movptr(Address(rbp, -wordSize), rax);

     }

   }

 }

 void SharedRuntime::restore_native_result(MacroAssembler *masm, BasicType ret_type, int frame_slots) {

   // We always ignore the frame_slots arg and just use the space just below frame pointer

   // which by this time is free to use

   switch (ret_type) {

   case T_FLOAT:

     __ fld_s(Address(rbp, -wordSize));

     break;

   case T_DOUBLE:

     __ fld_d(Address(rbp, -2*wordSize));

     break;

   case T_LONG:

     __ movptr(rax, Address(rbp, -wordSize));

     NOT_LP64(__ movptr(rdx, Address(rbp, -2*wordSize)));

     break;

   case T_VOID:  break;

   default: {

     __ movptr(rax, Address(rbp, -wordSize));

     }

   }

 }

 static void save_or_restore_arguments(MacroAssembler* masm,

                                       const int stack_slots,

                                       const int total_in_args,

                                       const int arg_save_area,

                                       OopMap* map,

                                       VMRegPair* in_regs,

                                       BasicType* in_sig_bt) {

   // if map is non-NULL then the code should store the values,

   // otherwise it should load them.

   int handle_index = 0;

   // Save down double word first

   for ( int i = 0; i < total_in_args; i++) {

     if (in_regs[i].first()->is_XMMRegister() && in_sig_bt[i] == T_DOUBLE) {

       int slot = handle_index * VMRegImpl::slots_per_word + arg_save_area;

       int offset = slot * VMRegImpl::stack_slot_size;

       handle_index += 2;

       assert(handle_index <= stack_slots, "overflow");

       if (map != NULL) {

         __ movdbl(Address(rsp, offset), in_regs[i].first()->as_XMMRegister());

       } else {

         __ movdbl(in_regs[i].first()->as_XMMRegister(), Address(rsp, offset));

       }

     }

     if (in_regs[i].first()->is_Register() && in_sig_bt[i] == T_LONG) {

       int slot = handle_index * VMRegImpl::slots_per_word + arg_save_area;

       int offset = slot * VMRegImpl::stack_slot_size;

       handle_index += 2;

       assert(handle_index <= stack_slots, "overflow");

       if (map != NULL) {

         __ movl(Address(rsp, offset), in_regs[i].first()->as_Register());

         if (in_regs[i].second()->is_Register()) {

           __ movl(Address(rsp, offset + 4), in_regs[i].second()->as_Register());

         }

       } else {

         __ movl(in_regs[i].first()->as_Register(), Address(rsp, offset));

         if (in_regs[i].second()->is_Register()) {

           __ movl(in_regs[i].second()->as_Register(), Address(rsp, offset + 4));

         }

       }

     }

   }

   // Save or restore single word registers

   for ( int i = 0; i < total_in_args; i++) {

     if (in_regs[i].first()->is_Register()) {

       int slot = handle_index++ * VMRegImpl::slots_per_word + arg_save_area;

       int offset = slot * VMRegImpl::stack_slot_size;

       assert(handle_index <= stack_slots, "overflow");

       if (in_sig_bt[i] == T_ARRAY && map != NULL) {

         map->set_oop(VMRegImpl::stack2reg(slot));;

       }

       // Value is in an input register pass we must flush it to the stack

       const Register reg = in_regs[i].first()->as_Register();

       switch (in_sig_bt[i]) {

         case T_ARRAY:

           if (map != NULL) {

             __ movptr(Address(rsp, offset), reg);

           } else {

             __ movptr(reg, Address(rsp, offset));

           }

           break;

         case T_BOOLEAN:

         case T_CHAR:

         case T_BYTE:

         case T_SHORT:

         case T_INT:

           if (map != NULL) {

             __ movl(Address(rsp, offset), reg);

           } else {

             __ movl(reg, Address(rsp, offset));

           }

           break;

         case T_OBJECT:

         default: ShouldNotReachHere();

       }

     } else if (in_regs[i].first()->is_XMMRegister()) {

       if (in_sig_bt[i] == T_FLOAT) {

         int slot = handle_index++ * VMRegImpl::slots_per_word + arg_save_area;

         int offset = slot * VMRegImpl::stack_slot_size;

         assert(handle_index <= stack_slots, "overflow");

         if (map != NULL) {

           __ movflt(Address(rsp, offset), in_regs[i].first()->as_XMMRegister());

         } else {

           __ movflt(in_regs[i].first()->as_XMMRegister(), Address(rsp, offset));

         }

       }

     } else if (in_regs[i].first()->is_stack()) {

       if (in_sig_bt[i] == T_ARRAY && map != NULL) {

         int offset_in_older_frame = in_regs[i].first()->reg2stack() + SharedRuntime::out_preserve_stack_slots();

         map->set_oop(VMRegImpl::stack2reg(offset_in_older_frame + stack_slots));

       }

     }

   }

 }

 // Check GC_locker::needs_gc and enter the runtime if it's true.  This

 // keeps a new JNI critical region from starting until a GC has been

 // forced.  Save down any oops in registers and describe them in an

 // OopMap.

 static void check_needs_gc_for_critical_native(MacroAssembler* masm,

                                                Register thread,

                                                int stack_slots,

                                                int total_c_args,

                                                int total_in_args,

                                                int arg_save_area,

                                                OopMapSet* oop_maps,

                                                VMRegPair* in_regs,

                                                BasicType* in_sig_bt) {

   __ block_comment("check GC_locker::needs_gc");

   Label cont;

   __ cmp8(ExternalAddress((address)GC_locker::needs_gc_address()), false);

   __ jcc(Assembler::equal, cont);

   // Save down any incoming oops and call into the runtime to halt for a GC

   OopMap* map = new OopMap(stack_slots * 2, 0 /* arg_slots*/);

   save_or_restore_arguments(masm, stack_slots, total_in_args,

                             arg_save_area, map, in_regs, in_sig_bt);

   address the_pc = __ pc();

   oop_maps->add_gc_map( __ offset(), map);

   __ set_last_Java_frame(thread, rsp, noreg, the_pc);

   __ block_comment("block_for_jni_critical");

   __ push(thread);

   __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::block_for_jni_critical)));

   __ increment(rsp, wordSize);

   __ get_thread(thread);

   __ reset_last_Java_frame(thread, false, true);

   save_or_restore_arguments(masm, stack_slots, total_in_args,

                             arg_save_area, NULL, in_regs, in_sig_bt);

   __ bind(cont);

 #ifdef ASSERT

   if (StressCriticalJNINatives) {

     // Stress register saving

     OopMap* map = new OopMap(stack_slots * 2, 0 /* arg_slots*/);

     save_or_restore_arguments(masm, stack_slots, total_in_args,

                               arg_save_area, map, in_regs, in_sig_bt);

     // Destroy argument registers

     for (int i = 0; i < total_in_args - 1; i++) {

       if (in_regs[i].first()->is_Register()) {

         const Register reg = in_regs[i].first()->as_Register();

         __ xorptr(reg, reg);

       } else if (in_regs[i].first()->is_XMMRegister()) {

         __ xorpd(in_regs[i].first()->as_XMMRegister(), in_regs[i].first()->as_XMMRegister());

       } else if (in_regs[i].first()->is_FloatRegister()) {

         ShouldNotReachHere();

       } else if (in_regs[i].first()->is_stack()) {

         // Nothing to do

       } else {

         ShouldNotReachHere();

       }

       if (in_sig_bt[i] == T_LONG || in_sig_bt[i] == T_DOUBLE) {

         i++;

       }

     }

     save_or_restore_arguments(masm, stack_slots, total_in_args,

                               arg_save_area, NULL, in_regs, in_sig_bt);

   }

 #endif

 }

 // Unpack an array argument into a pointer to the body and the length

 // if the array is non-null, otherwise pass 0 for both.

 static void unpack_array_argument(MacroAssembler* masm, VMRegPair reg, BasicType in_elem_type, VMRegPair body_arg, VMRegPair length_arg) {

   Register tmp_reg = rax;

   assert(!body_arg.first()->is_Register() || body_arg.first()->as_Register() != tmp_reg,

          "possible collision");

   assert(!length_arg.first()->is_Register() || length_arg.first()->as_Register() != tmp_reg,

          "possible collision");

   // Pass the length, ptr pair

   Label is_null, done;

   VMRegPair tmp(tmp_reg->as_VMReg());

   if (reg.first()->is_stack()) {

     // Load the arg up from the stack

     simple_move32(masm, reg, tmp);

     reg = tmp;

   }

   __ testptr(reg.first()->as_Register(), reg.first()->as_Register());

   __ jccb(Assembler::equal, is_null);

   __ lea(tmp_reg, Address(reg.first()->as_Register(), arrayOopDesc::base_offset_in_bytes(in_elem_type)));

   simple_move32(masm, tmp, body_arg);

   // load the length relative to the body.

   __ movl(tmp_reg, Address(tmp_reg, arrayOopDesc::length_offset_in_bytes() -

                            arrayOopDesc::base_offset_in_bytes(in_elem_type)));

   simple_move32(masm, tmp, length_arg);

   __ jmpb(done);

   __ bind(is_null);

   // Pass zeros

   __ xorptr(tmp_reg, tmp_reg);

   simple_move32(masm, tmp, body_arg);

   simple_move32(masm, tmp, length_arg);

   __ bind(done);

 }

 static void verify_oop_args(MacroAssembler* masm,

                             methodHandle method,

                             const BasicType* sig_bt,

                             const VMRegPair* regs) {

   Register temp_reg = rbx;  // not part of any compiled calling seq

   if (VerifyOops) {

     for (int i = 0; i < method->size_of_parameters(); i++) {

       if (sig_bt[i] == T_OBJECT ||

           sig_bt[i] == T_ARRAY) {

         VMReg r = regs[i].first();

         assert(r->is_valid(), "bad oop arg");

         if (r->is_stack()) {

           __ movptr(temp_reg, Address(rsp, r->reg2stack() * VMRegImpl::stack_slot_size + wordSize));

           __ verify_oop(temp_reg);

         } else {

           __ verify_oop(r->as_Register());

         }

       }

     }

   }

 }

 static void gen_special_dispatch(MacroAssembler* masm,

                                  methodHandle method,

                                  const BasicType* sig_bt,

                                  const VMRegPair* regs) {

   verify_oop_args(masm, method, sig_bt, regs);

   vmIntrinsics::ID iid = method->intrinsic_id();

   // Now write the args into the outgoing interpreter space

   bool     has_receiver   = false;

   Register receiver_reg   = noreg;

   int      member_arg_pos = -1;

   Register member_reg     = noreg;

   int      ref_kind       = MethodHandles::signature_polymorphic_intrinsic_ref_kind(iid);

   if (ref_kind != 0) {

     member_arg_pos = method->size_of_parameters() - 1;  // trailing MemberName argument

     member_reg = rbx;  // known to be free at this point

     has_receiver = MethodHandles::ref_kind_has_receiver(ref_kind);

   } else if (iid == vmIntrinsics::_invokeBasic) {

     has_receiver = true;

   } else {

     fatal(err_msg_res("unexpected intrinsic id %d", iid));

   }

   if (member_reg != noreg) {

     // Load the member_arg into register, if necessary.

     SharedRuntime::check_member_name_argument_is_last_argument(method, sig_bt, regs);

     VMReg r = regs[member_arg_pos].first();

     if (r->is_stack()) {

       __ movptr(member_reg, Address(rsp, r->reg2stack() * VMRegImpl::stack_slot_size + wordSize));

     } else {

       // no data motion is needed

       member_reg = r->as_Register();

     }

   }

   if (has_receiver) {

     // Make sure the receiver is loaded into a register.

     assert(method->size_of_parameters() > 0, "oob");

     assert(sig_bt[0] == T_OBJECT, "receiver argument must be an object");

     VMReg r = regs[0].first();

     assert(r->is_valid(), "bad receiver arg");

     if (r->is_stack()) {

       // Porting note:  This assumes that compiled calling conventions always

       // pass the receiver oop in a register.  If this is not true on some

       // platform, pick a temp and load the receiver from stack.

       fatal("receiver always in a register");

       receiver_reg = rcx;  // known to be free at this point

       __ movptr(receiver_reg, Address(rsp, r->reg2stack() * VMRegImpl::stack_slot_size + wordSize));

     } else {

       // no data motion is needed

       receiver_reg = r->as_Register();

     }

   }

   // Figure out which address we are really jumping to:

   MethodHandles::generate_method_handle_dispatch(masm, iid,

                                                  receiver_reg, member_reg, /*for_compiler_entry:*/ true);

 }

 // ---------------------------------------------------------------------------

 // Generate a native wrapper for a given method.  The method takes arguments

 // in the Java compiled code convention, marshals them to the native

 // convention (handlizes oops, etc), transitions to native, makes the call,

 // returns to java state (possibly blocking), unhandlizes any result and

 // returns.

 //

 // Critical native functions are a shorthand for the use of

 // GetPrimtiveArrayCritical and disallow the use of any other JNI

 // functions.  The wrapper is expected to unpack the arguments before

 // passing them to the callee and perform checks before and after the

 // native call to ensure that they GC_locker

 // lock_critical/unlock_critical semantics are followed.  Some other

 // parts of JNI setup are skipped like the tear down of the JNI handle

 // block and the check for pending exceptions it's impossible for them

 // to be thrown.

 //

 // They are roughly structured like this:

 //    if (GC_locker::needs_gc())

 //      SharedRuntime::block_for_jni_critical();

 //    tranistion to thread_in_native

 //    unpack arrray arguments and call native entry point

 //    check for safepoint in progress

 //    check if any thread suspend flags are set

 //      call into JVM and possible unlock the JNI critical

 //      if a GC was suppressed while in the critical native.

 //    transition back to thread_in_Java

 //    return to caller

 //

 nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,

                                                 methodHandle method,

                                                 int compile_id,

                                                 BasicType* in_sig_bt,

                                                 VMRegPair* in_regs,

                                                 BasicType ret_type) {

   if (method->is_method_handle_intrinsic()) {

     vmIntrinsics::ID iid = method->intrinsic_id();

     intptr_t start = (intptr_t)__ pc();

     int vep_offset = ((intptr_t)__ pc()) - start;

     gen_special_dispatch(masm,

                          method,

                          in_sig_bt,

                          in_regs);

     int frame_complete = ((intptr_t)__ pc()) - start;  // not complete, period

     __ flush();

     int stack_slots = SharedRuntime::out_preserve_stack_slots();  // no out slots at all, actually

     return nmethod::new_native_nmethod(method,

                                        compile_id,

                                        masm->code(),

                                        vep_offset,

                                        frame_complete,

                                        stack_slots / VMRegImpl::slots_per_word,

                                        in_ByteSize(-1),

                                        in_ByteSize(-1),

                                        (OopMapSet*)NULL);

   }

   bool is_critical_native = true;

   address native_func = method->critical_native_function();

   if (native_func == NULL) {

     native_func = method->native_function();

     is_critical_native = false;

   }

   assert(native_func != NULL, "must have function");

   // An OopMap for lock (and class if static)

   OopMapSet *oop_maps = new OopMapSet();

   // We have received a description of where all the java arg are located

   // on entry to the wrapper. We need to convert these args to where

   // the jni function will expect them. To figure out where they go

   // we convert the java signature to a C signature by inserting

   // the hidden arguments as arg[0] and possibly arg[1] (static method)

   const int total_in_args = method->size_of_parameters();

   int total_c_args = total_in_args;

   if (!is_critical_native) {

     total_c_args += 1;

     if (method->is_static()) {

       total_c_args++;

     }

   } else {

     for (int i = 0; i < total_in_args; i++) {

       if (in_sig_bt[i] == T_ARRAY) {

         total_c_args++;

       }

     }

   }

   BasicType* out_sig_bt = NEW_RESOURCE_ARRAY(BasicType, total_c_args);

   VMRegPair* out_regs   = NEW_RESOURCE_ARRAY(VMRegPair, total_c_args);

   BasicType* in_elem_bt = NULL;

   int argc = 0;

   if (!is_critical_native) {

     out_sig_bt[argc++] = T_ADDRESS;

     if (method->is_static()) {

       out_sig_bt[argc++] = T_OBJECT;

     }

     for (int i = 0; i < total_in_args ; i++ ) {

       out_sig_bt[argc++] = in_sig_bt[i];

     }

   } else {

     Thread* THREAD = Thread::current();

     in_elem_bt = NEW_RESOURCE_ARRAY(BasicType, total_in_args);

     SignatureStream ss(method->signature());

     for (int i = 0; i < total_in_args ; i++ ) {

       if (in_sig_bt[i] == T_ARRAY) {

         // Arrays are passed as int, elem* pair

         out_sig_bt[argc++] = T_INT;

         out_sig_bt[argc++] = T_ADDRESS;

         Symbol* atype = ss.as_symbol(CHECK_NULL);

         const char* at = atype->as_C_string();

         if (strlen(at) == 2) {

           assert(at[0] == '[', "must be");

           switch (at[1]) {

             case 'B': in_elem_bt[i]  = T_BYTE; break;

             case 'C': in_elem_bt[i]  = T_CHAR; break;

             case 'D': in_elem_bt[i]  = T_DOUBLE; break;

             case 'F': in_elem_bt[i]  = T_FLOAT; break;

             case 'I': in_elem_bt[i]  = T_INT; break;

             case 'J': in_elem_bt[i]  = T_LONG; break;

             case 'S': in_elem_bt[i]  = T_SHORT; break;

             case 'Z': in_elem_bt[i]  = T_BOOLEAN; break;

             default: ShouldNotReachHere();

           }

         }

       } else {

         out_sig_bt[argc++] = in_sig_bt[i];

         in_elem_bt[i] = T_VOID;

       }

       if (in_sig_bt[i] != T_VOID) {

         assert(in_sig_bt[i] == ss.type(), "must match");

         ss.next();

       }

     }

   }

   // Now figure out where the args must be stored and how much stack space

   // they require.

   int out_arg_slots;

   out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);

   // Compute framesize for the wrapper.  We need to handlize all oops in

   // registers a max of 2 on x86.

   // Calculate the total number of stack slots we will need.

   // First count the abi requirement plus all of the outgoing args

   int stack_slots = SharedRuntime::out_preserve_stack_slots() + out_arg_slots;

   // Now the space for the inbound oop handle area

   int total_save_slots = 2 * VMRegImpl::slots_per_word; // 2 arguments passed in registers

   if (is_critical_native) {

     // Critical natives may have to call out so they need a save area

     // for register arguments.

     int double_slots = 0;

     int single_slots = 0;

     for ( int i = 0; i < total_in_args; i++) {

       if (in_regs[i].first()->is_Register()) {

         const Register reg = in_regs[i].first()->as_Register();

         switch (in_sig_bt[i]) {

           case T_ARRAY:  // critical array (uses 2 slots on LP64)

           case T_BOOLEAN:

           case T_BYTE:

           case T_SHORT:

           case T_CHAR:

           case T_INT:  single_slots++; break;

           case T_LONG: double_slots++; break;

           default:  ShouldNotReachHere();

         }

       } else if (in_regs[i].first()->is_XMMRegister()) {

         switch (in_sig_bt[i]) {

           case T_FLOAT:  single_slots++; break;

           case T_DOUBLE: double_slots++; break;

           default:  ShouldNotReachHere();

         }

       } else if (in_regs[i].first()->is_FloatRegister()) {

         ShouldNotReachHere();

       }

     }

     total_save_slots = double_slots * 2 + single_slots;

     // align the save area

     if (double_slots != 0) {

       stack_slots = round_to(stack_slots, 2);

     }

   }

   int oop_handle_offset = stack_slots;

   stack_slots += total_save_slots;

   // Now any space we need for handlizing a klass if static method

   int klass_slot_offset = 0;

   int klass_offset = -1;

   int lock_slot_offset = 0;

   bool is_static = false;

   if (method->is_static()) {

     klass_slot_offset = stack_slots;

     stack_slots += VMRegImpl::slots_per_word;

     klass_offset = klass_slot_offset * VMRegImpl::stack_slot_size;

     is_static = true;

   }

   // Plus a lock if needed

   if (method->is_synchronized()) {

     lock_slot_offset = stack_slots;

     stack_slots += VMRegImpl::slots_per_word;

   }

   // Now a place (+2) to save return values or temp during shuffling

   // + 2 for return address (which we own) and saved rbp,

   stack_slots += 4;

   // Ok The space we have allocated will look like:

   //

   //

   // FP-> |                     |

   //      |---------------------|

   //      | 2 slots for moves   |

   //      |---------------------|

   //      | lock box (if sync)  |

   //      |---------------------| <- lock_slot_offset  (-lock_slot_rbp_offset)

   //      | klass (if static)   |

   //      |---------------------| <- klass_slot_offset

   //      | oopHandle area      |

   //      |---------------------| <- oop_handle_offset (a max of 2 registers)

   //      | outbound memory     |

   //      | based arguments     |

   //      |                     |

   //      |---------------------|

   //      |                     |

   // SP-> | out_preserved_slots |

   //

   //

   // ****************************************************************************

   // WARNING - on Windows Java Natives use pascal calling convention and pop the

   // arguments off of the stack after the jni call. Before the call we can use

   // instructions that are SP relative. After the jni call we switch to FP

   // relative instructions instead of re-adjusting the stack on windows.

   // ****************************************************************************

   // Now compute actual number of stack words we need rounding to make

   // stack properly aligned.

   stack_slots = round_to(stack_slots, StackAlignmentInSlots);

   int stack_size = stack_slots * VMRegImpl::stack_slot_size;

   intptr_t start = (intptr_t)__ pc();

   // First thing make an ic check to see if we should even be here

   // We are free to use all registers as temps without saving them and

   // restoring them except rbp. rbp is the only callee save register

   // as far as the interpreter and the compiler(s) are concerned.

   const Register ic_reg = rax;

   const Register receiver = rcx;

   Label hit;

   Label exception_pending;

   __ verify_oop(receiver);

   __ cmpptr(ic_reg, Address(receiver, oopDesc::klass_offset_in_bytes()));

   __ jcc(Assembler::equal, hit);

   __ jump(RuntimeAddress(SharedRuntime::get_ic_miss_stub()));

   // verified entry must be aligned for code patching.

   // and the first 5 bytes must be in the same cache line

   // if we align at 8 then we will be sure 5 bytes are in the same line

   __ align(8);

   __ bind(hit);

   int vep_offset = ((intptr_t)__ pc()) - start;

 #ifdef COMPILER1

   if (InlineObjectHash && method->intrinsic_id() == vmIntrinsics::_hashCode) {

     // Object.hashCode can pull the hashCode from the header word

     // instead of doing a full VM transition once it's been computed.

     // Since hashCode is usually polymorphic at call sites we can't do

     // this optimization at the call site without a lot of work.

     Label slowCase;

     Register receiver = rcx;

     Register result = rax;

     __ movptr(result, Address(receiver, oopDesc::mark_offset_in_bytes()));

     // check if locked

     __ testptr(result, markOopDesc::unlocked_value);

     __ jcc (Assembler::zero, slowCase);

     if (UseBiasedLocking) {

       // Check if biased and fall through to runtime if so

       __ testptr(result, markOopDesc::biased_lock_bit_in_place);

       __ jcc (Assembler::notZero, slowCase);

     }

     // get hash

     __ andptr(result, markOopDesc::hash_mask_in_place);

     // test if hashCode exists

     __ jcc  (Assembler::zero, slowCase);

     __ shrptr(result, markOopDesc::hash_shift);

     __ ret(0);

     __ bind (slowCase);

   }

 #endif // COMPILER1

   // The instruction at the verified entry point must be 5 bytes or longer

   // because it can be patched on the fly by make_non_entrant. The stack bang

   // instruction fits that requirement.

   // Generate stack overflow check

   if (UseStackBanging) {

     __ bang_stack_with_offset(StackShadowPages*os::vm_page_size());

   } else {

     // need a 5 byte instruction to allow MT safe patching to non-entrant

     __ fat_nop();

   }

   // Generate a new frame for the wrapper.

   __ enter();

   // -2 because return address is already present and so is saved rbp

   __ subptr(rsp, stack_size - 2*wordSize);

   // Frame is now completed as far as size and linkage.

   int frame_complete = ((intptr_t)__ pc()) - start;

   // Calculate the difference between rsp and rbp,. We need to know it

   // after the native call because on windows Java Natives will pop

   // the arguments and it is painful to do rsp relative addressing

   // in a platform independent way. So after the call we switch to

   // rbp, relative addressing.

   int fp_adjustment = stack_size - 2*wordSize;

 #ifdef COMPILER2

   // C2 may leave the stack dirty if not in SSE2+ mode

   if (UseSSE >= 2) {

     __ verify_FPU(0, "c2i transition should have clean FPU stack");

   } else {

     __ empty_FPU_stack();

   }

 #endif /* COMPILER2 */

   // Compute the rbp, offset for any slots used after the jni call

   int lock_slot_rbp_offset = (lock_slot_offset*VMRegImpl::stack_slot_size) - fp_adjustment;

   // We use rdi as a thread pointer because it is callee save and

   // if we load it once it is usable thru the entire wrapper

   const Register thread = rdi;

   // We use rsi as the oop handle for the receiver/klass

   // It is callee save so it survives the call to native

   const Register oop_handle_reg = rsi;

   __ get_thread(thread);

   if (is_critical_native) {

     check_needs_gc_for_critical_native(masm, thread, stack_slots, total_c_args, total_in_args,

                                        oop_handle_offset, oop_maps, in_regs, in_sig_bt);

   }

   //

   // We immediately shuffle the arguments so that any vm call we have to

   // make from here on out (sync slow path, jvmti, etc.) we will have

   // captured the oops from our caller and have a valid oopMap for

   // them.

   // -----------------

   // The Grand Shuffle

   //

   // Natives require 1 or 2 extra arguments over the normal ones: the JNIEnv*

   // and, if static, the class mirror instead of a receiver.  This pretty much

   // guarantees that register layout will not match (and x86 doesn't use reg

   // parms though amd does).  Since the native abi doesn't use register args

   // and the java conventions does we don't have to worry about collisions.

   // All of our moved are reg->stack or stack->stack.

   // We ignore the extra arguments during the shuffle and handle them at the

   // last moment. The shuffle is described by the two calling convention

   // vectors we have in our possession. We simply walk the java vector to

   // get the source locations and the c vector to get the destinations.

   int c_arg = is_critical_native ? 0 : (method->is_static() ? 2 : 1 );

   // Record rsp-based slot for receiver on stack for non-static methods

   int receiver_offset = -1;

   // This is a trick. We double the stack slots so we can claim

   // the oops in the caller's frame. Since we are sure to have

   // more args than the caller doubling is enough to make

   // sure we can capture all the incoming oop args from the

   // caller.

   //

   OopMap* map = new OopMap(stack_slots * 2, 0 /* arg_slots*/);

   // Mark location of rbp,

   // map->set_callee_saved(VMRegImpl::stack2reg( stack_slots - 2), stack_slots * 2, 0, rbp->as_VMReg());

   // We know that we only have args in at most two integer registers (rcx, rdx). So rax, rbx

   // Are free to temporaries if we have to do  stack to steck moves.

   // All inbound args are referenced based on rbp, and all outbound args via rsp.

   for (int i = 0; i < total_in_args ; i++, c_arg++ ) {

     switch (in_sig_bt[i]) {

       case T_ARRAY:

         if (is_critical_native) {

           unpack_array_argument(masm, in_regs[i], in_elem_bt[i], out_regs[c_arg + 1], out_regs[c_arg]);

           c_arg++;

           break;

         }

       case T_OBJECT:

         assert(!is_critical_native, "no oop arguments");

         object_move(masm, map, oop_handle_offset, stack_slots, in_regs[i], out_regs[c_arg],

                     ((i == 0) && (!is_static)),

                     &receiver_offset);

         break;

       case T_VOID:

         break;

       case T_FLOAT:

         float_move(masm, in_regs[i], out_regs[c_arg]);

           break;

       case T_DOUBLE:

         assert( i + 1 < total_in_args &&

                 in_sig_bt[i + 1] == T_VOID &&

                 out_sig_bt[c_arg+1] == T_VOID, "bad arg list");

         double_move(masm, in_regs[i], out_regs[c_arg]);

         break;

       case T_LONG :

         long_move(masm, in_regs[i], out_regs[c_arg]);

         break;

       case T_ADDRESS: assert(false, "found T_ADDRESS in java args");

       default:

         simple_move32(masm, in_regs[i], out_regs[c_arg]);

     }

   }

   // Pre-load a static method's oop into rsi.  Used both by locking code and

   // the normal JNI call code.

   if (method->is_static() && !is_critical_native) {

     //  load opp into a register

     __ movoop(oop_handle_reg, JNIHandles::make_local(method->method_holder()->java_mirror()));

     // Now handlize the static class mirror it's known not-null.

     __ movptr(Address(rsp, klass_offset), oop_handle_reg);

     map->set_oop(VMRegImpl::stack2reg(klass_slot_offset));

     // Now get the handle

     __ lea(oop_handle_reg, Address(rsp, klass_offset));

     // store the klass handle as second argument

     __ movptr(Address(rsp, wordSize), oop_handle_reg);

   }

   // Change state to native (we save the return address in the thread, since it might not

   // be pushed on the stack when we do a a stack traversal). It is enough that the pc()

   // points into the right code segment. It does not have to be the correct return pc.

   // We use the same pc/oopMap repeatedly when we call out

   intptr_t the_pc = (intptr_t) __ pc();

   oop_maps->add_gc_map(the_pc - start, map);

   __ set_last_Java_frame(thread, rsp, noreg, (address)the_pc);

   // We have all of the arguments setup at this point. We must not touch any register

   // argument registers at this point (what if we save/restore them there are no oop?

   {

     SkipIfEqual skip_if(masm, &DTraceMethodProbes, 0);

     __ mov_metadata(rax, method());

     __ call_VM_leaf(

          CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_method_entry),

          thread, rax);

   }

   // RedefineClasses() tracing support for obsolete method entry

   if (RC_TRACE_IN_RANGE(0x00001000, 0x00002000)) {

     __ mov_metadata(rax, method());

     __ call_VM_leaf(

          CAST_FROM_FN_PTR(address, SharedRuntime::rc_trace_method_entry),

          thread, rax);

   }

   // These are register definitions we need for locking/unlocking

   const Register swap_reg = rax;  // Must use rax, for cmpxchg instruction

   const Register obj_reg  = rcx;  // Will contain the oop

   const Register lock_reg = rdx;  // Address of compiler lock object (BasicLock)

   Label slow_path_lock;

   Label lock_done;

   // Lock a synchronized method

   if (method->is_synchronized()) {

     assert(!is_critical_native, "unhandled");

     const int mark_word_offset = BasicLock::displaced_header_offset_in_bytes();

     // Get the handle (the 2nd argument)

     __ movptr(oop_handle_reg, Address(rsp, wordSize));

     // Get address of the box

     __ lea(lock_reg, Address(rbp, lock_slot_rbp_offset));

     // Load the oop from the handle

     __ movptr(obj_reg, Address(oop_handle_reg, 0));

     if (UseBiasedLocking) {

       // Note that oop_handle_reg is trashed during this call

       __ biased_locking_enter(lock_reg, obj_reg, swap_reg, oop_handle_reg, false, lock_done, &slow_path_lock);

     }

     // Load immediate 1 into swap_reg %rax,

     __ movptr(swap_reg, 1);

     // Load (object->mark() | 1) into swap_reg %rax,

     __ orptr(swap_reg, Address(obj_reg, 0));

     // Save (object->mark() | 1) into BasicLock's displaced header

     __ movptr(Address(lock_reg, mark_word_offset), swap_reg);

     if (os::is_MP()) {

       __ lock();

     }

     // src -> dest iff dest == rax, else rax, <- dest

     // *obj_reg = lock_reg iff *obj_reg == rax, else rax, = *(obj_reg)

     __ cmpxchgptr(lock_reg, Address(obj_reg, 0));

     __ jcc(Assembler::equal, lock_done);

     // Test if the oopMark is an obvious stack pointer, i.e.,

     //  1) (mark & 3) == 0, and

     //  2) rsp <= mark < mark + os::pagesize()

     // These 3 tests can be done by evaluating the following

     // expression: ((mark - rsp) & (3 - os::vm_page_size())),

     // assuming both stack pointer and pagesize have their

     // least significant 2 bits clear.

     // NOTE: the oopMark is in swap_reg %rax, as the result of cmpxchg

     __ subptr(swap_reg, rsp);

     __ andptr(swap_reg, 3 - os::vm_page_size());

     // Save the test result, for recursive case, the result is zero

     __ movptr(Address(lock_reg, mark_word_offset), swap_reg);

     __ jcc(Assembler::notEqual, slow_path_lock);

     // Slow path will re-enter here

     __ bind(lock_done);

     if (UseBiasedLocking) {

       // Re-fetch oop_handle_reg as we trashed it above

       __ movptr(oop_handle_reg, Address(rsp, wordSize));

     }

   }

   // Finally just about ready to make the JNI call

   // get JNIEnv* which is first argument to native

   if (!is_critical_native) {

     __ lea(rdx, Address(thread, in_bytes(JavaThread::jni_environment_offset())));

     __ movptr(Address(rsp, 0), rdx);

   }

   // Now set thread in native

   __ movl(Address(thread, JavaThread::thread_state_offset()), _thread_in_native);

   __ call(RuntimeAddress(native_func));

   // WARNING - on Windows Java Natives use pascal calling convention and pop the

   // arguments off of the stack. We could just re-adjust the stack pointer here

   // and continue to do SP relative addressing but we instead switch to FP

   // relative addressing.

   // Unpack native results.

   switch (ret_type) {

   case T_BOOLEAN: __ c2bool(rax);            break;

   case T_CHAR   : __ andptr(rax, 0xFFFF);    break;

   case T_BYTE   : __ sign_extend_byte (rax); break;

   case T_SHORT  : __ sign_extend_short(rax); break;

   case T_INT    : /* nothing to do */        break;

   case T_DOUBLE :

   case T_FLOAT  :

     // Result is in st0 we'll save as needed

     break;

   case T_ARRAY:                 // Really a handle

   case T_OBJECT:                // Really a handle

       break; // can't de-handlize until after safepoint check

   case T_VOID: break;

   case T_LONG: break;

   default       : ShouldNotReachHere();

   }

   // Switch thread to "native transition" state before reading the synchronization state.

   // This additional state is necessary because reading and testing the synchronization

   // state is not atomic w.r.t. GC, as this scenario demonstrates:

   //     Java thread A, in _thread_in_native state, loads _not_synchronized and is preempted.

   //     VM thread changes sync state to synchronizing and suspends threads for GC.

   //     Thread A is resumed to finish this native method, but doesn't block here since it

   //     didn't see any synchronization is progress, and escapes.

   __ movl(Address(thread, JavaThread::thread_state_offset()), _thread_in_native_trans);

   if(os::is_MP()) {

     if (UseMembar) {

       // Force this write out before the read below

       __ membar(Assembler::Membar_mask_bits(

            Assembler::LoadLoad | Assembler::LoadStore |

            Assembler::StoreLoad | Assembler::StoreStore));

     } else {

       // Write serialization page so VM thread can do a pseudo remote membar.

       // We use the current thread pointer to calculate a thread specific

       // offset to write to within the page. This minimizes bus traffic

       // due to cache line collision.

       __ serialize_memory(thread, rcx);

     }

   }

   if (AlwaysRestoreFPU) {

     // Make sure the control word is correct.

     __ fldcw(ExternalAddress(StubRoutines::addr_fpu_cntrl_wrd_std()));

   }

   Label after_transition;

   // check for safepoint operation in progress and/or pending suspend requests

   { Label Continue;

     __ cmp32(ExternalAddress((address)SafepointSynchronize::address_of_state()),

              SafepointSynchronize::_not_synchronized);

     Label L;

     __ jcc(Assembler::notEqual, L);

     __ cmpl(Address(thread, JavaThread::suspend_flags_offset()), 0);

     __ jcc(Assembler::equal, Continue);

     __ bind(L);

     // Don't use call_VM as it will see a possible pending exception and forward it

     // and never return here preventing us from clearing _last_native_pc down below.

     // Also can't use call_VM_leaf either as it will check to see if rsi & rdi are

     // preserved and correspond to the bcp/locals pointers. So we do a runtime call

     // by hand.

     //

     save_native_result(masm, ret_type, stack_slots);

     __ push(thread);

     if (!is_critical_native) {

       __ call(RuntimeAddress(CAST_FROM_FN_PTR(address,

                                               JavaThread::check_special_condition_for_native_trans)));

     } else {

       __ call(RuntimeAddress(CAST_FROM_FN_PTR(address,

                                               JavaThread::check_special_condition_for_native_trans_and_transition)));

     }

     __ increment(rsp, wordSize);

     // Restore any method result value

     restore_native_result(masm, ret_type, stack_slots);

     if (is_critical_native) {

       // The call above performed the transition to thread_in_Java so

       // skip the transition logic below.

       __ jmpb(after_transition);

     }

     __ bind(Continue);

   }

   // change thread state

   __ movl(Address(thread, JavaThread::thread_state_offset()), _thread_in_Java);

   __ bind(after_transition);

   Label reguard;

   Label reguard_done;

   __ cmpl(Address(thread, JavaThread::stack_guard_state_offset()), JavaThread::stack_guard_yellow_disabled);

   __ jcc(Assembler::equal, reguard);

   // slow path reguard  re-enters here

   __ bind(reguard_done);

   // Handle possible exception (will unlock if necessary)

   // native result if any is live

   // Unlock

   Label slow_path_unlock;

   Label unlock_done;

   if (method->is_synchronized()) {

     Label done;

     // Get locked oop from the handle we passed to jni

     __ movptr(obj_reg, Address(oop_handle_reg, 0));

     if (UseBiasedLocking) {

       __ biased_locking_exit(obj_reg, rbx, done);

     }

     // Simple recursive lock?

     __ cmpptr(Address(rbp, lock_slot_rbp_offset), (int32_t)NULL_WORD);

     __ jcc(Assembler::equal, done);

     // Must save rax, if if it is live now because cmpxchg must use it

     if (ret_type != T_FLOAT && ret_type != T_DOUBLE && ret_type != T_VOID) {

       save_native_result(masm, ret_type, stack_slots);

     }

     //  get old displaced header

     __ movptr(rbx, Address(rbp, lock_slot_rbp_offset));

     // get address of the stack lock

     __ lea(rax, Address(rbp, lock_slot_rbp_offset));

     // Atomic swap old header if oop still contains the stack lock

     if (os::is_MP()) {

     __ lock();

     }

     // src -> dest iff dest == rax, else rax, <- dest

     // *obj_reg = rbx, iff *obj_reg == rax, else rax, = *(obj_reg)

     __ cmpxchgptr(rbx, Address(obj_reg, 0));

     __ jcc(Assembler::notEqual, slow_path_unlock);

     // slow path re-enters here

     __ bind(unlock_done);

     if (ret_type != T_FLOAT && ret_type != T_DOUBLE && ret_type != T_VOID) {

       restore_native_result(masm, ret_type, stack_slots);

     }

     __ bind(done);

   }

   {

     SkipIfEqual skip_if(masm, &DTraceMethodProbes, 0);

     // Tell dtrace about this method exit

     save_native_result(masm, ret_type, stack_slots);

     __ mov_metadata(rax, method());

     __ call_VM_leaf(

          CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_method_exit),

          thread, rax);

     restore_native_result(masm, ret_type, stack_slots);

   }

   // We can finally stop using that last_Java_frame we setup ages ago

   __ reset_last_Java_frame(thread, false, true);

   // Unpack oop result

   if (ret_type == T_OBJECT || ret_type == T_ARRAY) {

       Label L;

       __ cmpptr(rax, (int32_t)NULL_WORD);

       __ jcc(Assembler::equal, L);

       __ movptr(rax, Address(rax, 0));

       __ bind(L);

       __ verify_oop(rax);

   }

   if (!is_critical_native) {

     // reset handle block

     __ movptr(rcx, Address(thread, JavaThread::active_handles_offset()));

     __ movptr(Address(rcx, JNIHandleBlock::top_offset_in_bytes()), NULL_WORD);

     // Any exception pending?

     __ cmpptr(Address(thread, in_bytes(Thread::pending_exception_offset())), (int32_t)NULL_WORD);

     __ jcc(Assembler::notEqual, exception_pending);

   }

   // no exception, we're almost done

   // check that only result value is on FPU stack

   __ verify_FPU(ret_type == T_FLOAT || ret_type == T_DOUBLE ? 1 : 0, "native_wrapper normal exit");

   // Fixup floating pointer results so that result looks like a return from a compiled method

   if (ret_type == T_FLOAT) {

     if (UseSSE >= 1) {

       // Pop st0 and store as float and reload into xmm register

       __ fstp_s(Address(rbp, -4));

       __ movflt(xmm0, Address(rbp, -4));

     }

   } else if (ret_type == T_DOUBLE) {

     if (UseSSE >= 2) {

       // Pop st0 and store as double and reload into xmm register

       __ fstp_d(Address(rbp, -8));

       __ movdbl(xmm0, Address(rbp, -8));

     }

   }

   // Return

   __ leave();

   __ ret(0);

   // Unexpected paths are out of line and go here

   // Slow path locking & unlocking

   if (method->is_synchronized()) {

     // BEGIN Slow path lock

     __ bind(slow_path_lock);

     // has last_Java_frame setup. No exceptions so do vanilla call not call_VM

     // args are (oop obj, BasicLock* lock, JavaThread* thread)

     __ push(thread);

     __ push(lock_reg);

     __ push(obj_reg);

     __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::complete_monitor_locking_C)));

     __ addptr(rsp, 3*wordSize);

 #ifdef ASSERT

     { Label L;

     __ cmpptr(Address(thread, in_bytes(Thread::pending_exception_offset())), (int)NULL_WORD);

     __ jcc(Assembler::equal, L);

     __ stop("no pending exception allowed on exit from monitorenter");

     __ bind(L);

     }

 #endif

     __ jmp(lock_done);

     // END Slow path lock

     // BEGIN Slow path unlock

     __ bind(slow_path_unlock);

     // Slow path unlock

     if (ret_type == T_FLOAT || ret_type == T_DOUBLE ) {

       save_native_result(masm, ret_type, stack_slots);

     }

     // Save pending exception around call to VM (which contains an EXCEPTION_MARK)

     __ pushptr(Address(thread, in_bytes(Thread::pending_exception_offset())));

     __ movptr(Address(thread, in_bytes(Thread::pending_exception_offset())), NULL_WORD);

     // should be a peal

     // +wordSize because of the push above

     __ lea(rax, Address(rbp, lock_slot_rbp_offset));

     __ push(rax);

     __ push(obj_reg);

     __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::complete_monitor_unlocking_C)));

     __ addptr(rsp, 2*wordSize);

 #ifdef ASSERT

     {

       Label L;

       __ cmpptr(Address(thread, in_bytes(Thread::pending_exception_offset())), (int32_t)NULL_WORD);

       __ jcc(Assembler::equal, L);

       __ stop("no pending exception allowed on exit complete_monitor_unlocking_C");

       __ bind(L);

     }

 #endif /* ASSERT */

     __ popptr(Address(thread, in_bytes(Thread::pending_exception_offset())));

     if (ret_type == T_FLOAT || ret_type == T_DOUBLE ) {

       restore_native_result(masm, ret_type, stack_slots);

     }

     __ jmp(unlock_done);

     // END Slow path unlock

   }

   // SLOW PATH Reguard the stack if needed

   __ bind(reguard);

   save_native_result(masm, ret_type, stack_slots);

   {

     __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, SharedRuntime::reguard_yellow_pages)));

   }

   restore_native_result(masm, ret_type, stack_slots);

   __ jmp(reguard_done);

   // BEGIN EXCEPTION PROCESSING

   if (!is_critical_native) {

     // Forward  the exception

     __ bind(exception_pending);

     // remove possible return value from FPU register stack

     __ empty_FPU_stack();

     // pop our frame

     __ leave();

     // and forward the exception

     __ jump(RuntimeAddress(StubRoutines::forward_exception_entry()));

   }

   __ flush();

   nmethod *nm = nmethod::new_native_nmethod(method,

                                             compile_id,

                                             masm->code(),

                                             vep_offset,

                                             frame_complete,

                                             stack_slots / VMRegImpl::slots_per_word,

                                             (is_static ? in_ByteSize(klass_offset) : in_ByteSize(receiver_offset)),

                                             in_ByteSize(lock_slot_offset*VMRegImpl::stack_slot_size),

                                             oop_maps);

   if (is_critical_native) {

     nm->set_lazy_critical_native(true);

   }

   return nm;

 }

 #ifdef HAVE_DTRACE_H

 // ---------------------------------------------------------------------------

 // Generate a dtrace nmethod for a given signature.  The method takes arguments

 // in the Java compiled code convention, marshals them to the native

 // abi and then leaves nops at the position you would expect to call a native

 // function. When the probe is enabled the nops are replaced with a trap

 // instruction that dtrace inserts and the trace will cause a notification

 // to dtrace.

 //

 // The probes are only able to take primitive types and java/lang/String as

 // arguments.  No other java types are allowed. Strings are converted to utf8

 // strings so that from dtrace point of view java strings are converted to C

 // strings. There is an arbitrary fixed limit on the total space that a method

 // can use for converting the strings. (256 chars per string in the signature).

 // So any java string larger then this is truncated.

 nmethod *SharedRuntime::generate_dtrace_nmethod(

     MacroAssembler *masm, methodHandle method) {

   // generate_dtrace_nmethod is guarded by a mutex so we are sure to

   // be single threaded in this method.

   assert(AdapterHandlerLibrary_lock->owned_by_self(), "must be");

   // Fill in the signature array, for the calling-convention call.

   int total_args_passed = method->size_of_parameters();

   BasicType* in_sig_bt  = NEW_RESOURCE_ARRAY(BasicType, total_args_passed);

   VMRegPair  *in_regs   = NEW_RESOURCE_ARRAY(VMRegPair, total_args_passed);

   // The signature we are going to use for the trap that dtrace will see

   // java/lang/String is converted. We drop "this" and any other object

   // is converted to NULL.  (A one-slot java/lang/Long object reference

   // is converted to a two-slot long, which is why we double the allocation).

   BasicType* out_sig_bt = NEW_RESOURCE_ARRAY(BasicType, total_args_passed * 2);

   VMRegPair* out_regs   = NEW_RESOURCE_ARRAY(VMRegPair, total_args_passed * 2);

   int i=0;

   int total_strings = 0;

   int first_arg_to_pass = 0;

   int total_c_args = 0;

   if( !method->is_static() ) {  // Pass in receiver first

     in_sig_bt[i++] = T_OBJECT;

     first_arg_to_pass = 1;

   }

   // We need to convert the java args to where a native (non-jni) function

   // would expect them. To figure out where they go we convert the java

   // signature to a C signature.

   SignatureStream ss(method->signature());

   for ( ; !ss.at_return_type(); ss.next()) {

     BasicType bt = ss.type();

     in_sig_bt[i++] = bt;  // Collect remaining bits of signature

     out_sig_bt[total_c_args++] = bt;

     if( bt == T_OBJECT) {

       Symbol* s = ss.as_symbol_or_null();   // symbol is created

       if (s == vmSymbols::java_lang_String()) {

         total_strings++;

         out_sig_bt[total_c_args-1] = T_ADDRESS;

       } else if (s == vmSymbols::java_lang_Boolean() ||

                  s == vmSymbols::java_lang_Character() ||

                  s == vmSymbols::java_lang_Byte() ||

                  s == vmSymbols::java_lang_Short() ||

                  s == vmSymbols::java_lang_Integer() ||

                  s == vmSymbols::java_lang_Float()) {

         out_sig_bt[total_c_args-1] = T_INT;

       } else if (s == vmSymbols::java_lang_Long() ||

                  s == vmSymbols::java_lang_Double()) {

         out_sig_bt[total_c_args-1] = T_LONG;

         out_sig_bt[total_c_args++] = T_VOID;

       }

     } else if ( bt == T_LONG || bt == T_DOUBLE ) {

       in_sig_bt[i++] = T_VOID;   // Longs & doubles take 2 Java slots

       out_sig_bt[total_c_args++] = T_VOID;

     }

   }

   assert(i==total_args_passed, "validly parsed signature");

   // Now get the compiled-Java layout as input arguments

   int comp_args_on_stack;

   comp_args_on_stack = SharedRuntime::java_calling_convention(

       in_sig_bt, in_regs, total_args_passed, false);

   // Now figure out where the args must be stored and how much stack space

   // they require (neglecting out_preserve_stack_slots).

   int out_arg_slots;

   out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);

   // Calculate the total number of stack slots we will need.

   // First count the abi requirement plus all of the outgoing args

   int stack_slots = SharedRuntime::out_preserve_stack_slots() + out_arg_slots;

   // Now space for the string(s) we must convert

   int* string_locs   = NEW_RESOURCE_ARRAY(int, total_strings + 1);

   for (i = 0; i < total_strings ; i++) {

     string_locs[i] = stack_slots;

     stack_slots += max_dtrace_string_size / VMRegImpl::stack_slot_size;

   }

   // + 2 for return address (which we own) and saved rbp,

   stack_slots += 2;

   // Ok The space we have allocated will look like:

   //

   //

   // FP-> |                     |

   //      |---------------------|

   //      | string[n]           |

   //      |---------------------| <- string_locs[n]

   //      | string[n-1]         |

   //      |---------------------| <- string_locs[n-1]

   //      | ...                 |

   //      | ...                 |

   //      |---------------------| <- string_locs[1]

   //      | string[0]           |

   //      |---------------------| <- string_locs[0]

   //      | outbound memory     |

   //      | based arguments     |

   //      |                     |

   //      |---------------------|

   //      |                     |

   // SP-> | out_preserved_slots |

   //

   //

   // Now compute actual number of stack words we need rounding to make

   // stack properly aligned.

   stack_slots = round_to(stack_slots, 2 * VMRegImpl::slots_per_word);

   int stack_size = stack_slots * VMRegImpl::stack_slot_size;

   intptr_t start = (intptr_t)__ pc();

   // First thing make an ic check to see if we should even be here

   // We are free to use all registers as temps without saving them and

   // restoring them except rbp. rbp, is the only callee save register

   // as far as the interpreter and the compiler(s) are concerned.

   const Register ic_reg = rax;

   const Register receiver = rcx;

   Label hit;

   Label exception_pending;

   __ verify_oop(receiver);

   __ cmpl(ic_reg, Address(receiver, oopDesc::klass_offset_in_bytes()));

   __ jcc(Assembler::equal, hit);

   __ jump(RuntimeAddress(SharedRuntime::get_ic_miss_stub()));

   // verified entry must be aligned for code patching.

   // and the first 5 bytes must be in the same cache line

   // if we align at 8 then we will be sure 5 bytes are in the same line

   __ align(8);

   __ bind(hit);

   int vep_offset = ((intptr_t)__ pc()) - start;

   // The instruction at the verified entry point must be 5 bytes or longer

   // because it can be patched on the fly by make_non_entrant. The stack bang

   // instruction fits that requirement.

   // Generate stack overflow check

   if (UseStackBanging) {

     if (stack_size <= StackShadowPages*os::vm_page_size()) {

       __ bang_stack_with_offset(StackShadowPages*os::vm_page_size());

     } else {

       __ movl(rax, stack_size);

       __ bang_stack_size(rax, rbx);

     }

   } else {

     // need a 5 byte instruction to allow MT safe patching to non-entrant

     __ fat_nop();

   }

   assert(((int)__ pc() - start - vep_offset) >= 5,

          "valid size for make_non_entrant");

   // Generate a new frame for the wrapper.

   __ enter();

   // -2 because return address is already present and so is saved rbp,

   if (stack_size - 2*wordSize != 0) {

     __ subl(rsp, stack_size - 2*wordSize);

   }

   // Frame is now completed as far a size and linkage.

   int frame_complete = ((intptr_t)__ pc()) - start;

   // First thing we do store all the args as if we are doing the call.

   // Since the C calling convention is stack based that ensures that

   // all the Java register args are stored before we need to convert any

   // string we might have.

   int sid = 0;

   int c_arg, j_arg;

   int string_reg = 0;

   for (j_arg = first_arg_to_pass, c_arg = 0 ;

        j_arg < total_args_passed ; j_arg++, c_arg++ ) {

     VMRegPair src = in_regs[j_arg];

     VMRegPair dst = out_regs[c_arg];

     assert(dst.first()->is_stack() || in_sig_bt[j_arg] == T_VOID,

            "stack based abi assumed");

     switch (in_sig_bt[j_arg]) {

       case T_ARRAY:

       case T_OBJECT:

         if (out_sig_bt[c_arg] == T_ADDRESS) {

           // Any register based arg for a java string after the first

           // will be destroyed by the call to get_utf so we store

           // the original value in the location the utf string address

           // will eventually be stored.

           if (src.first()->is_reg()) {

             if (string_reg++ != 0) {

               simple_move32(masm, src, dst);

             }

           }

         } else if (out_sig_bt[c_arg] == T_INT || out_sig_bt[c_arg] == T_LONG) {

           // need to unbox a one-word value

           Register in_reg = rax;

           if ( src.first()->is_reg() ) {

             in_reg = src.first()->as_Register();

           } else {

             simple_move32(masm, src, in_reg->as_VMReg());