cpu/x86/x86_64.ad

//
// Copyright (c) 2003, 2019, 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.
//
//

// AMD64 Architecture Description File

//----------REGISTER DEFINITION BLOCK------------------------------------------
// This information is used by the matcher and the register allocator to
// describe individual registers and classes of registers within the target
// archtecture.

register %{
//----------Architecture Description Register Definitions----------------------
// General Registers
// "reg_def"  name ( register save type, C convention save type,
//                   ideal register type, encoding );
// Register Save Types:
//
// NS  = No-Save:       The register allocator assumes that these registers
//                      can be used without saving upon entry to the method, &
//                      that they do not need to be saved at call sites.
//
// SOC = Save-On-Call:  The register allocator assumes that these registers
//                      can be used without saving upon entry to the method,
//                      but that they must be saved at call sites.
//
// SOE = Save-On-Entry: The register allocator assumes that these registers
//                      must be saved before using them upon entry to the
//                      method, but they do not need to be saved at call
//                      sites.
//
// AS  = Always-Save:   The register allocator assumes that these registers
//                      must be saved before using them upon entry to the
//                      method, & that they must be saved at call sites.
//
// Ideal Register Type is used to determine how to save & restore a
// register.  Op_RegI will get spilled with LoadI/StoreI, Op_RegP will get
// spilled with LoadP/StoreP.  If the register supports both, use Op_RegI.
//
// The encoding number is the actual bit-pattern placed into the opcodes.

// General Registers
// R8-R15 must be encoded with REX.  (RSP, RBP, RSI, RDI need REX when
// used as byte registers)

// Previously set RBX, RSI, and RDI as save-on-entry for java code
// Turn off SOE in java-code due to frequent use of uncommon-traps.
// Now that allocator is better, turn on RSI and RDI as SOE registers.

reg_def RAX  (SOC, SOC, Op_RegI,  0, rax->as_VMReg());
reg_def RAX_H(SOC, SOC, Op_RegI,  0, rax->as_VMReg()->next());

reg_def RCX  (SOC, SOC, Op_RegI,  1, rcx->as_VMReg());
reg_def RCX_H(SOC, SOC, Op_RegI,  1, rcx->as_VMReg()->next());

reg_def RDX  (SOC, SOC, Op_RegI,  2, rdx->as_VMReg());
reg_def RDX_H(SOC, SOC, Op_RegI,  2, rdx->as_VMReg()->next());

reg_def RBX  (SOC, SOE, Op_RegI,  3, rbx->as_VMReg());
reg_def RBX_H(SOC, SOE, Op_RegI,  3, rbx->as_VMReg()->next());

reg_def RSP  (NS,  NS,  Op_RegI,  4, rsp->as_VMReg());
reg_def RSP_H(NS,  NS,  Op_RegI,  4, rsp->as_VMReg()->next());

// now that adapter frames are gone RBP is always saved and restored by the prolog/epilog code
reg_def RBP  (NS, SOE, Op_RegI,  5, rbp->as_VMReg());
reg_def RBP_H(NS, SOE, Op_RegI,  5, rbp->as_VMReg()->next());

#ifdef _WIN64

reg_def RSI  (SOC, SOE, Op_RegI,  6, rsi->as_VMReg());
reg_def RSI_H(SOC, SOE, Op_RegI,  6, rsi->as_VMReg()->next());

reg_def RDI  (SOC, SOE, Op_RegI,  7, rdi->as_VMReg());
reg_def RDI_H(SOC, SOE, Op_RegI,  7, rdi->as_VMReg()->next());

#else

reg_def RSI  (SOC, SOC, Op_RegI,  6, rsi->as_VMReg());
reg_def RSI_H(SOC, SOC, Op_RegI,  6, rsi->as_VMReg()->next());

reg_def RDI  (SOC, SOC, Op_RegI,  7, rdi->as_VMReg());
reg_def RDI_H(SOC, SOC, Op_RegI,  7, rdi->as_VMReg()->next());

#endif

reg_def R8   (SOC, SOC, Op_RegI,  8, r8->as_VMReg());
reg_def R8_H (SOC, SOC, Op_RegI,  8, r8->as_VMReg()->next());

reg_def R9   (SOC, SOC, Op_RegI,  9, r9->as_VMReg());
reg_def R9_H (SOC, SOC, Op_RegI,  9, r9->as_VMReg()->next());

reg_def R10  (SOC, SOC, Op_RegI, 10, r10->as_VMReg());
reg_def R10_H(SOC, SOC, Op_RegI, 10, r10->as_VMReg()->next());

reg_def R11  (SOC, SOC, Op_RegI, 11, r11->as_VMReg());
reg_def R11_H(SOC, SOC, Op_RegI, 11, r11->as_VMReg()->next());

reg_def R12  (SOC, SOE, Op_RegI, 12, r12->as_VMReg());
reg_def R12_H(SOC, SOE, Op_RegI, 12, r12->as_VMReg()->next());

reg_def R13  (SOC, SOE, Op_RegI, 13, r13->as_VMReg());
reg_def R13_H(SOC, SOE, Op_RegI, 13, r13->as_VMReg()->next());

reg_def R14  (SOC, SOE, Op_RegI, 14, r14->as_VMReg());
reg_def R14_H(SOC, SOE, Op_RegI, 14, r14->as_VMReg()->next());

reg_def R15  (SOC, SOE, Op_RegI, 15, r15->as_VMReg());
reg_def R15_H(SOC, SOE, Op_RegI, 15, r15->as_VMReg()->next());


// Floating Point Registers

// Specify priority of register selection within phases of register
// allocation.  Highest priority is first.  A useful heuristic is to
// give registers a low priority when they are required by machine
// instructions, like EAX and EDX on I486, and choose no-save registers
// before save-on-call, & save-on-call before save-on-entry.  Registers
// which participate in fixed calling sequences should come last.
// Registers which are used as pairs must fall on an even boundary.

alloc_class chunk0(R10,         R10_H,
                   R11,         R11_H,
                   R8,          R8_H,
                   R9,          R9_H,
                   R12,         R12_H,
                   RCX,         RCX_H,
                   RBX,         RBX_H,
                   RDI,         RDI_H,
                   RDX,         RDX_H,
                   RSI,         RSI_H,
                   RAX,         RAX_H,
                   RBP,         RBP_H,
                   R13,         R13_H,
                   R14,         R14_H,
                   R15,         R15_H,
                   RSP,         RSP_H);


//----------Architecture Description Register Classes--------------------------
// Several register classes are automatically defined based upon information in
// this architecture description.
// 1) reg_class inline_cache_reg           ( /* as def'd in frame section */ )
// 2) reg_class compiler_method_oop_reg    ( /* as def'd in frame section */ )
// 2) reg_class interpreter_method_oop_reg ( /* as def'd in frame section */ )
// 3) reg_class stack_slots( /* one chunk of stack-based "registers" */ )
//

// Empty register class.
reg_class no_reg();

// Class for all pointer/long registers
reg_class all_reg(RAX, RAX_H,
                  RDX, RDX_H,
                  RBP, RBP_H,
                  RDI, RDI_H,
                  RSI, RSI_H,
                  RCX, RCX_H,
                  RBX, RBX_H,
                  RSP, RSP_H,
                  R8,  R8_H,
                  R9,  R9_H,
                  R10, R10_H,
                  R11, R11_H,
                  R12, R12_H,
                  R13, R13_H,
                  R14, R14_H,
                  R15, R15_H);

// Class for all int registers
reg_class all_int_reg(RAX
                      RDX,
                      RBP,
                      RDI,
                      RSI,
                      RCX,
                      RBX,
                      R8,
                      R9,
                      R10,
                      R11,
                      R12,
                      R13,
                      R14);

// Class for all pointer registers
reg_class any_reg %{
  return _ANY_REG_mask;
%}

// Class for all pointer registers (excluding RSP)
reg_class ptr_reg %{
  return _PTR_REG_mask;
%}

// Class for all pointer registers (excluding RSP and RBP)
reg_class ptr_reg_no_rbp %{
  return _PTR_REG_NO_RBP_mask;
%}

// Class for all pointer registers (excluding RAX and RSP)
reg_class ptr_no_rax_reg %{
  return _PTR_NO_RAX_REG_mask;
%}

// Class for all pointer registers (excluding RAX, RBX, and RSP)
reg_class ptr_no_rax_rbx_reg %{
  return _PTR_NO_RAX_RBX_REG_mask;
%}

// Class for all long registers (excluding RSP)
reg_class long_reg %{
  return _LONG_REG_mask;
%}

// Class for all long registers (excluding RAX, RDX and RSP)
reg_class long_no_rax_rdx_reg %{
  return _LONG_NO_RAX_RDX_REG_mask;
%}

// Class for all long registers (excluding RCX and RSP)
reg_class long_no_rcx_reg %{
  return _LONG_NO_RCX_REG_mask;
%}

// Class for all int registers (excluding RSP)
reg_class int_reg %{
  return _INT_REG_mask;
%}

// Class for all int registers (excluding RAX, RDX, and RSP)
reg_class int_no_rax_rdx_reg %{
  return _INT_NO_RAX_RDX_REG_mask;
%}

// Class for all int registers (excluding RCX and RSP)
reg_class int_no_rcx_reg %{
  return _INT_NO_RCX_REG_mask;
%}

// Singleton class for RAX pointer register
reg_class ptr_rax_reg(RAX, RAX_H);

// Singleton class for RBX pointer register
reg_class ptr_rbx_reg(RBX, RBX_H);

// Singleton class for RSI pointer register
reg_class ptr_rsi_reg(RSI, RSI_H);

// Singleton class for RDI pointer register
reg_class ptr_rdi_reg(RDI, RDI_H);

// Singleton class for stack pointer
reg_class ptr_rsp_reg(RSP, RSP_H);

// Singleton class for TLS pointer
reg_class ptr_r15_reg(R15, R15_H);

// Singleton class for RAX long register
reg_class long_rax_reg(RAX, RAX_H);

// Singleton class for RCX long register
reg_class long_rcx_reg(RCX, RCX_H);

// Singleton class for RDX long register
reg_class long_rdx_reg(RDX, RDX_H);

// Singleton class for RAX int register
reg_class int_rax_reg(RAX);

// Singleton class for RBX int register
reg_class int_rbx_reg(RBX);

// Singleton class for RCX int register
reg_class int_rcx_reg(RCX);

// Singleton class for RCX int register
reg_class int_rdx_reg(RDX);

// Singleton class for RCX int register
reg_class int_rdi_reg(RDI);

// Singleton class for instruction pointer
// reg_class ip_reg(RIP);

%}

//----------SOURCE BLOCK-------------------------------------------------------
// This is a block of C++ code which provides values, functions, and
// definitions necessary in the rest of the architecture description
source_hpp %{

extern RegMask _ANY_REG_mask;
extern RegMask _PTR_REG_mask;
extern RegMask _PTR_REG_NO_RBP_mask;
extern RegMask _PTR_NO_RAX_REG_mask;
extern RegMask _PTR_NO_RAX_RBX_REG_mask;
extern RegMask _LONG_REG_mask;
extern RegMask _LONG_NO_RAX_RDX_REG_mask;
extern RegMask _LONG_NO_RCX_REG_mask;
extern RegMask _INT_REG_mask;
extern RegMask _INT_NO_RAX_RDX_REG_mask;
extern RegMask _INT_NO_RCX_REG_mask;

extern RegMask _STACK_OR_PTR_REG_mask;
extern RegMask _STACK_OR_LONG_REG_mask;
extern RegMask _STACK_OR_INT_REG_mask;

inline const RegMask& STACK_OR_PTR_REG_mask()  { return _STACK_OR_PTR_REG_mask;  }
inline const RegMask& STACK_OR_LONG_REG_mask() { return _STACK_OR_LONG_REG_mask; }
inline const RegMask& STACK_OR_INT_REG_mask()  { return _STACK_OR_INT_REG_mask;  }

%}

source %{
#define   RELOC_IMM64    Assembler::imm_operand
#define   RELOC_DISP32   Assembler::disp32_operand

#define __ _masm.

RegMask _ANY_REG_mask;
RegMask _PTR_REG_mask;
RegMask _PTR_REG_NO_RBP_mask;
RegMask _PTR_NO_RAX_REG_mask;
RegMask _PTR_NO_RAX_RBX_REG_mask;
RegMask _LONG_REG_mask;
RegMask _LONG_NO_RAX_RDX_REG_mask;
RegMask _LONG_NO_RCX_REG_mask;
RegMask _INT_REG_mask;
RegMask _INT_NO_RAX_RDX_REG_mask;
RegMask _INT_NO_RCX_REG_mask;
RegMask _STACK_OR_PTR_REG_mask;
RegMask _STACK_OR_LONG_REG_mask;
RegMask _STACK_OR_INT_REG_mask;

static bool need_r12_heapbase() {
  return UseCompressedOops || UseCompressedClassPointers;
}

void reg_mask_init() {
  // _ALL_REG_mask is generated by adlc from the all_reg register class below.
  // We derive a number of subsets from it.
  _ANY_REG_mask = _ALL_REG_mask;

  if (PreserveFramePointer) {
    _ANY_REG_mask.Remove(OptoReg::as_OptoReg(rbp->as_VMReg()));
    _ANY_REG_mask.Remove(OptoReg::as_OptoReg(rbp->as_VMReg()->next()));
  }
  if (need_r12_heapbase()) {
    _ANY_REG_mask.Remove(OptoReg::as_OptoReg(r12->as_VMReg()));
    _ANY_REG_mask.Remove(OptoReg::as_OptoReg(r12->as_VMReg()->next()));
  }

  _PTR_REG_mask = _ANY_REG_mask;
  _PTR_REG_mask.Remove(OptoReg::as_OptoReg(rsp->as_VMReg()));
  _PTR_REG_mask.Remove(OptoReg::as_OptoReg(rsp->as_VMReg()->next()));
  _PTR_REG_mask.Remove(OptoReg::as_OptoReg(r15->as_VMReg()));
  _PTR_REG_mask.Remove(OptoReg::as_OptoReg(r15->as_VMReg()->next()));

  _STACK_OR_PTR_REG_mask = _PTR_REG_mask;
  _STACK_OR_PTR_REG_mask.OR(STACK_OR_STACK_SLOTS_mask());

  _PTR_REG_NO_RBP_mask = _PTR_REG_mask;
  _PTR_REG_NO_RBP_mask.Remove(OptoReg::as_OptoReg(rbp->as_VMReg()));
  _PTR_REG_NO_RBP_mask.Remove(OptoReg::as_OptoReg(rbp->as_VMReg()->next()));

  _PTR_NO_RAX_REG_mask = _PTR_REG_mask;
  _PTR_NO_RAX_REG_mask.Remove(OptoReg::as_OptoReg(rax->as_VMReg()));
  _PTR_NO_RAX_REG_mask.Remove(OptoReg::as_OptoReg(rax->as_VMReg()->next()));

  _PTR_NO_RAX_RBX_REG_mask = _PTR_NO_RAX_REG_mask;
  _PTR_NO_RAX_RBX_REG_mask.Remove(OptoReg::as_OptoReg(rbx->as_VMReg()));
  _PTR_NO_RAX_RBX_REG_mask.Remove(OptoReg::as_OptoReg(rbx->as_VMReg()->next()));

  _LONG_REG_mask = _PTR_REG_mask;
  _STACK_OR_LONG_REG_mask = _LONG_REG_mask;
  _STACK_OR_LONG_REG_mask.OR(STACK_OR_STACK_SLOTS_mask());

  _LONG_NO_RAX_RDX_REG_mask = _LONG_REG_mask;
  _LONG_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rax->as_VMReg()));
  _LONG_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rax->as_VMReg()->next()));
  _LONG_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rdx->as_VMReg()));
  _LONG_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rdx->as_VMReg()->next()));

  _LONG_NO_RCX_REG_mask = _LONG_REG_mask;
  _LONG_NO_RCX_REG_mask.Remove(OptoReg::as_OptoReg(rcx->as_VMReg()));
  _LONG_NO_RCX_REG_mask.Remove(OptoReg::as_OptoReg(rcx->as_VMReg()->next()));

  _INT_REG_mask = _ALL_INT_REG_mask;
  if (PreserveFramePointer) {
    _INT_REG_mask.Remove(OptoReg::as_OptoReg(rbp->as_VMReg()));
  }
  if (need_r12_heapbase()) {
    _INT_REG_mask.Remove(OptoReg::as_OptoReg(r12->as_VMReg()));
  }

  _STACK_OR_INT_REG_mask = _INT_REG_mask;
  _STACK_OR_INT_REG_mask.OR(STACK_OR_STACK_SLOTS_mask());

  _INT_NO_RAX_RDX_REG_mask = _INT_REG_mask;
  _INT_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rax->as_VMReg()));
  _INT_NO_RAX_RDX_REG_mask.Remove(OptoReg::as_OptoReg(rdx->as_VMReg()));

  _INT_NO_RCX_REG_mask = _INT_REG_mask;
  _INT_NO_RCX_REG_mask.Remove(OptoReg::as_OptoReg(rcx->as_VMReg()));
}

static bool generate_vzeroupper(Compile* C) {
  return (VM_Version::supports_vzeroupper() && (C->max_vector_size() > 16 || C->clear_upper_avx() == true)) ? true: false;  // Generate vzeroupper
}

static int clear_avx_size() {
  return generate_vzeroupper(Compile::current()) ? 3: 0;  // vzeroupper
}

// !!!!! Special hack to get all types of calls to specify the byte offset
//       from the start of the call to the point where the return address
//       will point.
int MachCallStaticJavaNode::ret_addr_offset()
{
  int offset = 5; // 5 bytes from start of call to where return address points
  offset += clear_avx_size();
  return offset;
}

int MachCallDynamicJavaNode::ret_addr_offset()
{
  int offset = 15; // 15 bytes from start of call to where return address points
  offset += clear_avx_size();
  return offset;
}

int MachCallRuntimeNode::ret_addr_offset() {
  int offset = 13; // movq r10,#addr; callq (r10)
  offset += clear_avx_size();
  return offset;
}

// Indicate if the safepoint node needs the polling page as an input,
// it does if the polling page is more than disp32 away.
bool SafePointNode::needs_polling_address_input()
{
  return SafepointMechanism::uses_thread_local_poll() || Assembler::is_polling_page_far();
}

//
// Compute padding required for nodes which need alignment
//

// The address of the call instruction needs to be 4-byte aligned to
// ensure that it does not span a cache line so that it can be patched.
int CallStaticJavaDirectNode::compute_padding(int current_offset) const
{
  current_offset += clear_avx_size(); // skip vzeroupper
  current_offset += 1; // skip call opcode byte
  return align_up(current_offset, alignment_required()) - current_offset;
}

// The address of the call instruction needs to be 4-byte aligned to
// ensure that it does not span a cache line so that it can be patched.
int CallDynamicJavaDirectNode::compute_padding(int current_offset) const
{
  current_offset += clear_avx_size(); // skip vzeroupper
  current_offset += 11; // skip movq instruction + call opcode byte
  return align_up(current_offset, alignment_required()) - current_offset;
}

// EMIT_RM()
void emit_rm(CodeBuffer &cbuf, int f1, int f2, int f3) {
  unsigned char c = (unsigned char) ((f1 << 6) | (f2 << 3) | f3);
  cbuf.insts()->emit_int8(c);
}

// EMIT_CC()
void emit_cc(CodeBuffer &cbuf, int f1, int f2) {
  unsigned char c = (unsigned char) (f1 | f2);
  cbuf.insts()->emit_int8(c);
}

// EMIT_OPCODE()
void emit_opcode(CodeBuffer &cbuf, int code) {
  cbuf.insts()->emit_int8((unsigned char) code);
}

// EMIT_OPCODE() w/ relocation information
void emit_opcode(CodeBuffer &cbuf,
                 int code, relocInfo::relocType reloc, int offset, int format)
{
  cbuf.relocate(cbuf.insts_mark() + offset, reloc, format);
  emit_opcode(cbuf, code);
}

// EMIT_D8()
void emit_d8(CodeBuffer &cbuf, int d8) {
  cbuf.insts()->emit_int8((unsigned char) d8);
}

// EMIT_D16()
void emit_d16(CodeBuffer &cbuf, int d16) {
  cbuf.insts()->emit_int16(d16);
}

// EMIT_D32()
void emit_d32(CodeBuffer &cbuf, int d32) {
  cbuf.insts()->emit_int32(d32);
}

// EMIT_D64()
void emit_d64(CodeBuffer &cbuf, int64_t d64) {
  cbuf.insts()->emit_int64(d64);
}

// emit 32 bit value and construct relocation entry from relocInfo::relocType
void emit_d32_reloc(CodeBuffer& cbuf,
                    int d32,
                    relocInfo::relocType reloc,
                    int format)
{
  assert(reloc != relocInfo::external_word_type, "use 2-arg emit_d32_reloc");
  cbuf.relocate(cbuf.insts_mark(), reloc, format);
  cbuf.insts()->emit_int32(d32);
}

// emit 32 bit value and construct relocation entry from RelocationHolder
void emit_d32_reloc(CodeBuffer& cbuf, int d32, RelocationHolder const& rspec, int format) {
#ifdef ASSERT
  if (rspec.reloc()->type() == relocInfo::oop_type &&
      d32 != 0 && d32 != (intptr_t) Universe::non_oop_word()) {
    assert(Universe::heap()->is_in_reserved((address)(intptr_t)d32), "should be real oop");
    assert(oopDesc::is_oop(cast_to_oop((intptr_t)d32)), "cannot embed broken oops in code");
  }
#endif
  cbuf.relocate(cbuf.insts_mark(), rspec, format);
  cbuf.insts()->emit_int32(d32);
}

void emit_d32_reloc(CodeBuffer& cbuf, address addr) {
  address next_ip = cbuf.insts_end() + 4;
  emit_d32_reloc(cbuf, (int) (addr - next_ip),
                 external_word_Relocation::spec(addr),
                 RELOC_DISP32);
}


// emit 64 bit value and construct relocation entry from relocInfo::relocType
void emit_d64_reloc(CodeBuffer& cbuf, int64_t d64, relocInfo::relocType reloc, int format) {
  cbuf.relocate(cbuf.insts_mark(), reloc, format);
  cbuf.insts()->emit_int64(d64);
}

// emit 64 bit value and construct relocation entry from RelocationHolder
void emit_d64_reloc(CodeBuffer& cbuf, int64_t d64, RelocationHolder const& rspec, int format) {
#ifdef ASSERT
  if (rspec.reloc()->type() == relocInfo::oop_type &&
      d64 != 0 && d64 != (int64_t) Universe::non_oop_word()) {
    assert(Universe::heap()->is_in_reserved((address)d64), "should be real oop");
    assert(oopDesc::is_oop(cast_to_oop(d64)), "cannot embed broken oops in code");
  }
#endif
  cbuf.relocate(cbuf.insts_mark(), rspec, format);
  cbuf.insts()->emit_int64(d64);
}

// Access stack slot for load or store
void store_to_stackslot(CodeBuffer &cbuf, int opcode, int rm_field, int disp)
{
  emit_opcode(cbuf, opcode);                  // (e.g., FILD   [RSP+src])
  if (-0x80 <= disp && disp < 0x80) {
    emit_rm(cbuf, 0x01, rm_field, RSP_enc);   // R/M byte
    emit_rm(cbuf, 0x00, RSP_enc, RSP_enc);    // SIB byte
    emit_d8(cbuf, disp);     // Displacement  // R/M byte
  } else {
    emit_rm(cbuf, 0x02, rm_field, RSP_enc);   // R/M byte
    emit_rm(cbuf, 0x00, RSP_enc, RSP_enc);    // SIB byte
    emit_d32(cbuf, disp);     // Displacement // R/M byte
  }
}

   // rRegI ereg, memory mem) %{    // emit_reg_mem
void encode_RegMem(CodeBuffer &cbuf,
                   int reg,
                   int base, int index, int scale, int disp, relocInfo::relocType disp_reloc)
{
  assert(disp_reloc == relocInfo::none, "cannot have disp");
  int regenc = reg & 7;
  int baseenc = base & 7;
  int indexenc = index & 7;

  // There is no index & no scale, use form without SIB byte
  if (index == 0x4 && scale == 0 && base != RSP_enc && base != R12_enc) {
    // If no displacement, mode is 0x0; unless base is [RBP] or [R13]
    if (disp == 0 && base != RBP_enc && base != R13_enc) {
      emit_rm(cbuf, 0x0, regenc, baseenc); // *
    } else if (-0x80 <= disp && disp < 0x80 && disp_reloc == relocInfo::none) {
      // If 8-bit displacement, mode 0x1
      emit_rm(cbuf, 0x1, regenc, baseenc); // *
      emit_d8(cbuf, disp);
    } else {
      // If 32-bit displacement
      if (base == -1) { // Special flag for absolute address
        emit_rm(cbuf, 0x0, regenc, 0x5); // *
        if (disp_reloc != relocInfo::none) {
          emit_d32_reloc(cbuf, disp, relocInfo::oop_type, RELOC_DISP32);
        } else {
          emit_d32(cbuf, disp);
        }
      } else {
        // Normal base + offset
        emit_rm(cbuf, 0x2, regenc, baseenc); // *
        if (disp_reloc != relocInfo::none) {
          emit_d32_reloc(cbuf, disp, relocInfo::oop_type, RELOC_DISP32);
        } else {
          emit_d32(cbuf, disp);
        }
      }
    }
  } else {
    // Else, encode with the SIB byte
    // If no displacement, mode is 0x0; unless base is [RBP] or [R13]
    if (disp == 0 && base != RBP_enc && base != R13_enc) {
      // If no displacement
      emit_rm(cbuf, 0x0, regenc, 0x4); // *
      emit_rm(cbuf, scale, indexenc, baseenc);
    } else {
      if (-0x80 <= disp && disp < 0x80 && disp_reloc == relocInfo::none) {
        // If 8-bit displacement, mode 0x1
        emit_rm(cbuf, 0x1, regenc, 0x4); // *
        emit_rm(cbuf, scale, indexenc, baseenc);
        emit_d8(cbuf, disp);
      } else {
        // If 32-bit displacement
        if (base == 0x04 ) {
          emit_rm(cbuf, 0x2, regenc, 0x4);
          emit_rm(cbuf, scale, indexenc, 0x04); // XXX is this valid???
        } else {
          emit_rm(cbuf, 0x2, regenc, 0x4);
          emit_rm(cbuf, scale, indexenc, baseenc); // *
        }
        if (disp_reloc != relocInfo::none) {
          emit_d32_reloc(cbuf, disp, relocInfo::oop_type, RELOC_DISP32);
        } else {
          emit_d32(cbuf, disp);
        }
      }
    }
  }
}

// This could be in MacroAssembler but it's fairly C2 specific
void emit_cmpfp_fixup(MacroAssembler& _masm) {
  Label exit;
  __ jccb(Assembler::noParity, exit);
  __ pushf();
  //
  // comiss/ucomiss instructions set ZF,PF,CF flags and
  // zero OF,AF,SF for NaN values.
  // Fixup flags by zeroing ZF,PF so that compare of NaN
  // values returns 'less than' result (CF is set).
  // Leave the rest of flags unchanged.
  //
  //    7 6 5 4 3 2 1 0
  //   |S|Z|r|A|r|P|r|C|  (r - reserved bit)
  //    0 0 1 0 1 0 1 1   (0x2B)
  //
  __ andq(Address(rsp, 0), 0xffffff2b);
  __ popf();
  __ bind(exit);
}

void emit_cmpfp3(MacroAssembler& _masm, Register dst) {
  Label done;
  __ movl(dst, -1);
  __ jcc(Assembler::parity, done);
  __ jcc(Assembler::below, done);
  __ setb(Assembler::notEqual, dst);
  __ movzbl(dst, dst);
  __ bind(done);
}

// Math.min()    # Math.max()
// --------------------------
// ucomis[s/d]   #
// ja   -> b     # a
// jp   -> NaN   # NaN
// jb   -> a     # b
// je            #
// |-jz -> a | b # a & b
// |    -> a     #
void emit_fp_min_max(MacroAssembler& _masm, XMMRegister dst,
                     XMMRegister a, XMMRegister b,
                     XMMRegister xmmt, Register rt,
                     bool min, bool single) {

  Label nan, zero, below, above, done;

  if (single)
    __ ucomiss(a, b);
  else
    __ ucomisd(a, b);

  if (dst->encoding() != (min ? b : a)->encoding())
    __ jccb(Assembler::above, above); // CF=0 & ZF=0
  else
    __ jccb(Assembler::above, done);

  __ jccb(Assembler::parity, nan);  // PF=1
  __ jccb(Assembler::below, below); // CF=1

  // equal
  __ vpxor(xmmt, xmmt, xmmt, Assembler::AVX_128bit);
  if (single) {
    __ ucomiss(a, xmmt);
    __ jccb(Assembler::equal, zero);

    __ movflt(dst, a);
    __ jmp(done);
  }
  else {
    __ ucomisd(a, xmmt);
    __ jccb(Assembler::equal, zero);

    __ movdbl(dst, a);
    __ jmp(done);
  }

  __ bind(zero);
  if (min)
    __ vpor(dst, a, b, Assembler::AVX_128bit);
  else
    __ vpand(dst, a, b, Assembler::AVX_128bit);

  __ jmp(done);

  __ bind(above);
  if (single)
    __ movflt(dst, min ? b : a);
  else
    __ movdbl(dst, min ? b : a);

  __ jmp(done);

  __ bind(nan);
  if (single) {
    __ movl(rt, 0x7fc00000); // Float.NaN
    __ movdl(dst, rt);
  }
  else {
    __ mov64(rt, 0x7ff8000000000000L); // Double.NaN
    __ movdq(dst, rt);
  }
  __ jmp(done);

  __ bind(below);
  if (single)
    __ movflt(dst, min ? a : b);
  else
    __ movdbl(dst, min ? a : b);

  __ bind(done);
}

//=============================================================================
const RegMask& MachConstantBaseNode::_out_RegMask = RegMask::Empty;

int Compile::ConstantTable::calculate_table_base_offset() const {
  return 0;  // absolute addressing, no offset
}

bool MachConstantBaseNode::requires_postalloc_expand() const { return false; }
void MachConstantBaseNode::postalloc_expand(GrowableArray <Node *> *nodes, PhaseRegAlloc *ra_) {
  ShouldNotReachHere();
}

void MachConstantBaseNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const {
  // Empty encoding
}

uint MachConstantBaseNode::size(PhaseRegAlloc* ra_) const {
  return 0;
}

#ifndef PRODUCT
void MachConstantBaseNode::format(PhaseRegAlloc* ra_, outputStream* st) const {
  st->print("# MachConstantBaseNode (empty encoding)");
}
#endif


//=============================================================================
#ifndef PRODUCT
void MachPrologNode::format(PhaseRegAlloc* ra_, outputStream* st) const {
  Compile* C = ra_->C;

  int framesize = C->frame_size_in_bytes();
  int bangsize = C->bang_size_in_bytes();
  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
  // Remove wordSize for return addr which is already pushed.
  framesize -= wordSize;

  if (C->need_stack_bang(bangsize)) {
    framesize -= wordSize;
    st->print("# stack bang (%d bytes)", bangsize);
    st->print("\n\t");
    st->print("pushq   rbp\t# Save rbp");
    if (PreserveFramePointer) {
        st->print("\n\t");
        st->print("movq    rbp, rsp\t# Save the caller's SP into rbp");
    }
    if (framesize) {
      st->print("\n\t");
      st->print("subq    rsp, #%d\t# Create frame",framesize);
    }
  } else {
    st->print("subq    rsp, #%d\t# Create frame",framesize);
    st->print("\n\t");
    framesize -= wordSize;
    st->print("movq    [rsp + #%d], rbp\t# Save rbp",framesize);
    if (PreserveFramePointer) {
      st->print("\n\t");
      st->print("movq    rbp, rsp\t# Save the caller's SP into rbp");
      if (framesize > 0) {
        st->print("\n\t");
        st->print("addq    rbp, #%d", framesize);
      }
    }
  }

  if (VerifyStackAtCalls) {
    st->print("\n\t");
    framesize -= wordSize;
    st->print("movq    [rsp + #%d], 0xbadb100d\t# Majik cookie for stack depth check",framesize);
#ifdef ASSERT
    st->print("\n\t");
    st->print("# stack alignment check");
#endif
  }
  if (C->stub_function() != NULL && BarrierSet::barrier_set()->barrier_set_nmethod() != NULL) {
    st->print("\n\t");
    st->print("cmpl    [r15_thread + #disarmed_offset], #disarmed_value\t");
    st->print("\n\t");
    st->print("je      fast_entry\t");
    st->print("\n\t");
    st->print("call    #nmethod_entry_barrier_stub\t");
    st->print("\n\tfast_entry:");
  }
  st->cr();
}
#endif

void MachPrologNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  Compile* C = ra_->C;
  MacroAssembler _masm(&cbuf);

  int framesize = C->frame_size_in_bytes();
  int bangsize = C->bang_size_in_bytes();

  if (C->clinit_barrier_on_entry()) {
    assert(VM_Version::supports_fast_class_init_checks(), "sanity");
    assert(!C->method()->holder()->is_not_initialized(), "initialization should have been started");

    Label L_skip_barrier;
    Register klass = rscratch1;

    __ mov_metadata(klass, C->method()->holder()->constant_encoding());
    __ clinit_barrier(klass, r15_thread, &L_skip_barrier /*L_fast_path*/);

    __ jump(RuntimeAddress(SharedRuntime::get_handle_wrong_method_stub())); // slow path

    __ bind(L_skip_barrier);
  }

  __ verified_entry(framesize, C->need_stack_bang(bangsize)?bangsize:0, false, C->stub_function() != NULL);

  C->set_frame_complete(cbuf.insts_size());

  if (C->has_mach_constant_base_node()) {
    // NOTE: We set the table base offset here because users might be
    // emitted before MachConstantBaseNode.
    Compile::ConstantTable& constant_table = C->constant_table();
    constant_table.set_table_base_offset(constant_table.calculate_table_base_offset());
  }
}

uint MachPrologNode::size(PhaseRegAlloc* ra_) const
{
  return MachNode::size(ra_); // too many variables; just compute it
                              // the hard way
}

int MachPrologNode::reloc() const
{
  return 0; // a large enough number
}

//=============================================================================
#ifndef PRODUCT
void MachEpilogNode::format(PhaseRegAlloc* ra_, outputStream* st) const
{
  Compile* C = ra_->C;
  if (generate_vzeroupper(C)) {
    st->print("vzeroupper");
    st->cr(); st->print("\t");
  }

  int framesize = C->frame_size_in_bytes();
  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
  // Remove word for return adr already pushed
  // and RBP
  framesize -= 2*wordSize;

  if (framesize) {
    st->print_cr("addq    rsp, %d\t# Destroy frame", framesize);
    st->print("\t");
  }

  st->print_cr("popq    rbp");
  if (do_polling() && C->is_method_compilation()) {
    st->print("\t");
    if (SafepointMechanism::uses_thread_local_poll()) {
      st->print_cr("movq    rscratch1, poll_offset[r15_thread] #polling_page_address\n\t"
                   "testl   rax, [rscratch1]\t"
                   "# Safepoint: poll for GC");
    } else if (Assembler::is_polling_page_far()) {
      st->print_cr("movq    rscratch1, #polling_page_address\n\t"
                   "testl   rax, [rscratch1]\t"
                   "# Safepoint: poll for GC");
    } else {
      st->print_cr("testl   rax, [rip + #offset_to_poll_page]\t"
                   "# Safepoint: poll for GC");
    }
  }
}
#endif

void MachEpilogNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const
{
  Compile* C = ra_->C;
  MacroAssembler _masm(&cbuf);

  if (generate_vzeroupper(C)) {
    // Clear upper bits of YMM registers when current compiled code uses
    // wide vectors to avoid AVX <-> SSE transition penalty during call.
    __ vzeroupper();
  }

  int framesize = C->frame_size_in_bytes();
  assert((framesize & (StackAlignmentInBytes-1)) == 0, "frame size not aligned");
  // Remove word for return adr already pushed
  // and RBP
  framesize -= 2*wordSize;

  // Note that VerifyStackAtCalls' Majik cookie does not change the frame size popped here

  if (framesize) {
    emit_opcode(cbuf, Assembler::REX_W);
    if (framesize < 0x80) {
      emit_opcode(cbuf, 0x83); // addq rsp, #framesize
      emit_rm(cbuf, 0x3, 0x00, RSP_enc);
      emit_d8(cbuf, framesize);
    } else {
      emit_opcode(cbuf, 0x81); // addq rsp, #framesize
      emit_rm(cbuf, 0x3, 0x00, RSP_enc);
      emit_d32(cbuf, framesize);
    }
  }

  // popq rbp
  emit_opcode(cbuf, 0x58 | RBP_enc);

  if (StackReservedPages > 0 && C->has_reserved_stack_access()) {
    __ reserved_stack_check();
  }

  if (do_polling() && C->is_method_compilation()) {
    MacroAssembler _masm(&cbuf);
    if (SafepointMechanism::uses_thread_local_poll()) {
      __ movq(rscratch1, Address(r15_thread, Thread::polling_page_offset()));
      __ relocate(relocInfo::poll_return_type);
      __ testl(rax, Address(rscratch1, 0));
    } else {
      AddressLiteral polling_page(os::get_polling_page(), relocInfo::poll_return_type);
      if (Assembler::is_polling_page_far()) {
        __ lea(rscratch1, polling_page);
        __ relocate(relocInfo::poll_return_type);
        __ testl(rax, Address(rscratch1, 0));
      } else {
        __ testl(rax, polling_page);
      }
    }
  }
}

uint MachEpilogNode::size(PhaseRegAlloc* ra_) const
{
  return MachNode::size(ra_); // too many variables; just compute it
                              // the hard way
}

int MachEpilogNode::reloc() const
{
  return 2; // a large enough number
}

const Pipeline* MachEpilogNode::pipeline() const
{
  return MachNode::pipeline_class();
}

int MachEpilogNode::safepoint_offset() const
{
  return 0;
}

//=============================================================================

enum RC {
  rc_bad,
  rc_int,
  rc_float,
  rc_stack
};

static enum RC rc_class(OptoReg::Name reg)
{
  if( !OptoReg::is_valid(reg)  ) return rc_bad;

  if (OptoReg::is_stack(reg)) return rc_stack;

  VMReg r = OptoReg::as_VMReg(reg);

  if (r->is_Register()) return rc_int;

  assert(r->is_XMMRegister(), "must be");
  return rc_float;
}

// Next two methods are shared by 32- and 64-bit VM. They are defined in x86.ad.
static int vec_mov_helper(CodeBuffer *cbuf, bool do_size, int src_lo, int dst_lo,
                          int src_hi, int dst_hi, uint ireg, outputStream* st);

int vec_spill_helper(CodeBuffer *cbuf, bool do_size, bool is_load,
                     int stack_offset, int reg, uint ireg, outputStream* st);

static void vec_stack_to_stack_helper(CodeBuffer *cbuf, int src_offset,
                                      int dst_offset, uint ireg, outputStream* st) {
  if (cbuf) {
    MacroAssembler _masm(cbuf);
    switch (ireg) {
    case Op_VecS:
      __ movq(Address(rsp, -8), rax);
      __ movl(rax, Address(rsp, src_offset));
      __ movl(Address(rsp, dst_offset), rax);
      __ movq(rax, Address(rsp, -8));
      break;
    case Op_VecD:
      __ pushq(Address(rsp, src_offset));
      __ popq (Address(rsp, dst_offset));
      break;
    case Op_VecX:
      __ pushq(Address(rsp, src_offset));
      __ popq (Address(rsp, dst_offset));
      __ pushq(Address(rsp, src_offset+8));
      __ popq (Address(rsp, dst_offset+8));
      break;
    case Op_VecY:
      __ vmovdqu(Address(rsp, -32), xmm0);
      __ vmovdqu(xmm0, Address(rsp, src_offset));
      __ vmovdqu(Address(rsp, dst_offset), xmm0);
      __ vmovdqu(xmm0, Address(rsp, -32));
      break;
    case Op_VecZ:
      __ evmovdquq(Address(rsp, -64), xmm0, 2);
      __ evmovdquq(xmm0, Address(rsp, src_offset), 2);
      __ evmovdquq(Address(rsp, dst_offset), xmm0, 2);
      __ evmovdquq(xmm0, Address(rsp, -64), 2);
      break;
    default:
      ShouldNotReachHere();
    }
#ifndef PRODUCT
  } else {
    switch (ireg) {
    case Op_VecS:
      st->print("movq    [rsp - #8], rax\t# 32-bit mem-mem spill\n\t"
                "movl    rax, [rsp + #%d]\n\t"
                "movl    [rsp + #%d], rax\n\t"
                "movq    rax, [rsp - #8]",
                src_offset, dst_offset);
      break;
    case Op_VecD:
      st->print("pushq   [rsp + #%d]\t# 64-bit mem-mem spill\n\t"
                "popq    [rsp + #%d]",
                src_offset, dst_offset);
      break;
     case Op_VecX:
      st->print("pushq   [rsp + #%d]\t# 128-bit mem-mem spill\n\t"
                "popq    [rsp + #%d]\n\t"
                "pushq   [rsp + #%d]\n\t"
                "popq    [rsp + #%d]",
                src_offset, dst_offset, src_offset+8, dst_offset+8);
      break;
    case Op_VecY:
      st->print("vmovdqu [rsp - #32], xmm0\t# 256-bit mem-mem spill\n\t"
                "vmovdqu xmm0, [rsp + #%d]\n\t"
                "vmovdqu [rsp + #%d], xmm0\n\t"
                "vmovdqu xmm0, [rsp - #32]",
                src_offset, dst_offset);
      break;
    case Op_VecZ:
      st->print("vmovdqu [rsp - #64], xmm0\t# 512-bit mem-mem spill\n\t"
                "vmovdqu xmm0, [rsp + #%d]\n\t"
                "vmovdqu [rsp + #%d], xmm0\n\t"
                "vmovdqu xmm0, [rsp - #64]",
                src_offset, dst_offset);
      break;
    default:
      ShouldNotReachHere();
    }
#endif
  }
}

uint MachSpillCopyNode::implementation(CodeBuffer* cbuf,
                                       PhaseRegAlloc* ra_,
                                       bool do_size,
                                       outputStream* st) const {
  assert(cbuf != NULL || st  != NULL, "sanity");
  // Get registers to move
  OptoReg::Name src_second = ra_->get_reg_second(in(1));
  OptoReg::Name src_first = ra_->get_reg_first(in(1));
  OptoReg::Name dst_second = ra_->get_reg_second(this);
  OptoReg::Name dst_first = ra_->get_reg_first(this);

  enum RC src_second_rc = rc_class(src_second);
  enum RC src_first_rc = rc_class(src_first);
  enum RC dst_second_rc = rc_class(dst_second);
  enum RC dst_first_rc = rc_class(dst_first);

  assert(OptoReg::is_valid(src_first) && OptoReg::is_valid(dst_first),
         "must move at least 1 register" );

  if (src_first == dst_first && src_second == dst_second) {
    // Self copy, no move
    return 0;
  }
  if (bottom_type()->isa_vect() != NULL) {
    uint ireg = ideal_reg();
    assert((src_first_rc != rc_int && dst_first_rc != rc_int), "sanity");
    assert((ireg == Op_VecS || ireg == Op_VecD || ireg == Op_VecX || ireg == Op_VecY || ireg == Op_VecZ ), "sanity");
    if( src_first_rc == rc_stack && dst_first_rc == rc_stack ) {
      // mem -> mem
      int src_offset = ra_->reg2offset(src_first);
      int dst_offset = ra_->reg2offset(dst_first);
      vec_stack_to_stack_helper(cbuf, src_offset, dst_offset, ireg, st);
    } else if (src_first_rc == rc_float && dst_first_rc == rc_float ) {
      vec_mov_helper(cbuf, false, src_first, dst_first, src_second, dst_second, ireg, st);
    } else if (src_first_rc == rc_float && dst_first_rc == rc_stack ) {
      int stack_offset = ra_->reg2offset(dst_first);
      vec_spill_helper(cbuf, false, false, stack_offset, src_first, ireg, st);
    } else if (src_first_rc == rc_stack && dst_first_rc == rc_float ) {
      int stack_offset = ra_->reg2offset(src_first);
      vec_spill_helper(cbuf, false, true,  stack_offset, dst_first, ireg, st);
    } else {
      ShouldNotReachHere();
    }
    return 0;
  }
  if (src_first_rc == rc_stack) {
    // mem ->
    if (dst_first_rc == rc_stack) {
      // mem -> mem
      assert(src_second != dst_first, "overlap");
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        int src_offset = ra_->reg2offset(src_first);
        int dst_offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ pushq(Address(rsp, src_offset));
          __ popq (Address(rsp, dst_offset));
#ifndef PRODUCT
        } else {
          st->print("pushq   [rsp + #%d]\t# 64-bit mem-mem spill\n\t"
                    "popq    [rsp + #%d]",
                     src_offset, dst_offset);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        // No pushl/popl, so:
        int src_offset = ra_->reg2offset(src_first);
        int dst_offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movq(Address(rsp, -8), rax);
          __ movl(rax, Address(rsp, src_offset));
          __ movl(Address(rsp, dst_offset), rax);
          __ movq(rax, Address(rsp, -8));
#ifndef PRODUCT
        } else {
          st->print("movq    [rsp - #8], rax\t# 32-bit mem-mem spill\n\t"
                    "movl    rax, [rsp + #%d]\n\t"
                    "movl    [rsp + #%d], rax\n\t"
                    "movq    rax, [rsp - #8]",
                     src_offset, dst_offset);
#endif
        }
      }
      return 0;
    } else if (dst_first_rc == rc_int) {
      // mem -> gpr
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        int offset = ra_->reg2offset(src_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movq(as_Register(Matcher::_regEncode[dst_first]), Address(rsp, offset));
#ifndef PRODUCT
        } else {
          st->print("movq    %s, [rsp + #%d]\t# spill",
                     Matcher::regName[dst_first],
                     offset);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        int offset = ra_->reg2offset(src_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movl(as_Register(Matcher::_regEncode[dst_first]), Address(rsp, offset));
#ifndef PRODUCT
        } else {
          st->print("movl    %s, [rsp + #%d]\t# spill",
                     Matcher::regName[dst_first],
                     offset);
#endif
        }
      }
      return 0;
    } else if (dst_first_rc == rc_float) {
      // mem-> xmm
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        int offset = ra_->reg2offset(src_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdbl( as_XMMRegister(Matcher::_regEncode[dst_first]), Address(rsp, offset));
#ifndef PRODUCT
        } else {
          st->print("%s  %s, [rsp + #%d]\t# spill",
                     UseXmmLoadAndClearUpper ? "movsd " : "movlpd",
                     Matcher::regName[dst_first],
                     offset);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        int offset = ra_->reg2offset(src_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movflt( as_XMMRegister(Matcher::_regEncode[dst_first]), Address(rsp, offset));
#ifndef PRODUCT
        } else {
          st->print("movss   %s, [rsp + #%d]\t# spill",
                     Matcher::regName[dst_first],
                     offset);
#endif
        }
      }
      return 0;
    }
  } else if (src_first_rc == rc_int) {
    // gpr ->
    if (dst_first_rc == rc_stack) {
      // gpr -> mem
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        int offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movq(Address(rsp, offset), as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movq    [rsp + #%d], %s\t# spill",
                     offset,
                     Matcher::regName[src_first]);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        int offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movl(Address(rsp, offset), as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movl    [rsp + #%d], %s\t# spill",
                     offset,
                     Matcher::regName[src_first]);
#endif
        }
      }
      return 0;
    } else if (dst_first_rc == rc_int) {
      // gpr -> gpr
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movq(as_Register(Matcher::_regEncode[dst_first]),
                  as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movq    %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
        return 0;
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movl(as_Register(Matcher::_regEncode[dst_first]),
                  as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movl    %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
        return 0;
      }
    } else if (dst_first_rc == rc_float) {
      // gpr -> xmm
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdq( as_XMMRegister(Matcher::_regEncode[dst_first]), as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movdq   %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdl( as_XMMRegister(Matcher::_regEncode[dst_first]), as_Register(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movdl   %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      }
      return 0;
    }
  } else if (src_first_rc == rc_float) {
    // xmm ->
    if (dst_first_rc == rc_stack) {
      // xmm -> mem
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        int offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdbl( Address(rsp, offset), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movsd   [rsp + #%d], %s\t# spill",
                     offset,
                     Matcher::regName[src_first]);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        int offset = ra_->reg2offset(dst_first);
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movflt(Address(rsp, offset), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movss   [rsp + #%d], %s\t# spill",
                     offset,
                     Matcher::regName[src_first]);
#endif
        }
      }
      return 0;
    } else if (dst_first_rc == rc_int) {
      // xmm -> gpr
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdq( as_Register(Matcher::_regEncode[dst_first]), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movdq   %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdl( as_Register(Matcher::_regEncode[dst_first]), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("movdl   %s, %s\t# spill",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      }
      return 0;
    } else if (dst_first_rc == rc_float) {
      // xmm -> xmm
      if ((src_first & 1) == 0 && src_first + 1 == src_second &&
          (dst_first & 1) == 0 && dst_first + 1 == dst_second) {
        // 64-bit
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movdbl( as_XMMRegister(Matcher::_regEncode[dst_first]), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("%s  %s, %s\t# spill",
                     UseXmmRegToRegMoveAll ? "movapd" : "movsd ",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      } else {
        // 32-bit
        assert(!((src_first & 1) == 0 && src_first + 1 == src_second), "no transform");
        assert(!((dst_first & 1) == 0 && dst_first + 1 == dst_second), "no transform");
        if (cbuf) {
          MacroAssembler _masm(cbuf);
          __ movflt( as_XMMRegister(Matcher::_regEncode[dst_first]), as_XMMRegister(Matcher::_regEncode[src_first]));
#ifndef PRODUCT
        } else {
          st->print("%s  %s, %s\t# spill",
                     UseXmmRegToRegMoveAll ? "movaps" : "movss ",
                     Matcher::regName[dst_first],
                     Matcher::regName[src_first]);
#endif
        }
      }
      return 0;
    }
  }

  assert(0," foo ");
  Unimplemented();
  return 0;
}

#ifndef PRODUCT
void MachSpillCopyNode::format(PhaseRegAlloc *ra_, outputStream* st) const {
  implementation(NULL, ra_, false, st);
}
#endif

void MachSpillCopyNode::emit(CodeBuffer &cbuf, PhaseRegAlloc *ra_) const {
  implementation(&cbuf, ra_, false, NULL);
}

uint MachSpillCopyNode::size(PhaseRegAlloc *ra_) const {
  return MachNode::size(ra_);
}

//=============================================================================
#ifndef PRODUCT
void BoxLockNode::format(PhaseRegAlloc* ra_, outputStream* st) const
{
  int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
  int reg = ra_->get_reg_first(this);
  st->print("leaq    %s, [rsp + #%d]\t# box lock",
            Matcher::regName[reg], offset);
}
#endif

void BoxLockNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const
{
  int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
  int reg = ra_->get_encode(this);
  if (offset >= 0x80) {
    emit_opcode(cbuf, reg < 8 ? Assembler::REX_W : Assembler::REX_WR);
    emit_opcode(cbuf, 0x8D); // LEA  reg,[SP+offset]
    emit_rm(cbuf, 0x2, reg & 7, 0x04);
    emit_rm(cbuf, 0x0, 0x04, RSP_enc);
    emit_d32(cbuf, offset);
  } else {
    emit_opcode(cbuf, reg < 8 ? Assembler::REX_W : Assembler::REX_WR);
    emit_opcode(cbuf, 0x8D); // LEA  reg,[SP+offset]
    emit_rm(cbuf, 0x1, reg & 7, 0x04);
    emit_rm(cbuf, 0x0, 0x04, RSP_enc);
    emit_d8(cbuf, offset);
  }
}

uint BoxLockNode::size(PhaseRegAlloc *ra_) const
{
  int offset = ra_->reg2offset(in_RegMask(0).find_first_elem());
  return (offset < 0x80) ? 5 : 8; // REX
}

//=============================================================================
#ifndef PRODUCT
void MachUEPNode::format(PhaseRegAlloc* ra_, outputStream* st) const
{
  if (UseCompressedClassPointers) {
    st->print_cr("movl    rscratch1, [j_rarg0 + oopDesc::klass_offset_in_bytes()]\t# compressed klass");
    st->print_cr("\tdecode_klass_not_null rscratch1, rscratch1");
    st->print_cr("\tcmpq    rax, rscratch1\t # Inline cache check");
  } else {
    st->print_cr("\tcmpq    rax, [j_rarg0 + oopDesc::klass_offset_in_bytes()]\t"
                 "# Inline cache check");
  }
  st->print_cr("\tjne     SharedRuntime::_ic_miss_stub");
  st->print_cr("\tnop\t# nops to align entry point");
}
#endif

void MachUEPNode::emit(CodeBuffer& cbuf, PhaseRegAlloc* ra_) const
{
  MacroAssembler masm(&cbuf);
  uint insts_size = cbuf.insts_size();
  if (UseCompressedClassPointers) {
    masm.load_klass(rscratch1, j_rarg0);
    masm.cmpptr(rax, rscratch1);
  } else {
    masm.cmpptr(rax, Address(j_rarg0, oopDesc::klass_offset_in_bytes()));
  }

  masm.jump_cc(Assembler::notEqual, RuntimeAddress(SharedRuntime::get_ic_miss_stub()));

  /* WARNING these NOPs are critical so that verified entry point is properly
     4 bytes aligned for patching by NativeJump::patch_verified_entry() */
  int nops_cnt = 4 - ((cbuf.insts_size() - insts_size) & 0x3);
  if (OptoBreakpoint) {
    // Leave space for int3
    nops_cnt -= 1;
  }
  nops_cnt &= 0x3; // Do not add nops if code is aligned.
  if (nops_cnt > 0)
    masm.nop(nops_cnt);
}

uint MachUEPNode::size(PhaseRegAlloc* ra_) const
{
  return MachNode::size(ra_); // too many variables; just compute it
                              // the hard way
}


//=============================================================================

int Matcher::regnum_to_fpu_offset(int regnum)
{
  return regnum - 32; // The FP registers are in the second chunk
}

// This is UltraSparc specific, true just means we have fast l2f conversion
const bool Matcher::convL2FSupported(void) {
  return true;
}

// Is this branch offset short enough that a short branch can be used?
//
// NOTE: If the platform does not provide any short branch variants, then
//       this method should return false for offset 0.
bool Matcher::is_short_branch_offset(int rule, int br_size, int offset) {
  // The passed offset is relative to address of the branch.
  // On 86 a branch displacement is calculated relative to address
  // of a next instruction.
  offset -= br_size;

  // the short version of jmpConUCF2 contains multiple branches,
  // making the reach slightly less
  if (rule == jmpConUCF2_rule)
    return (-126 <= offset && offset <= 125);
  return (-128 <= offset && offset <= 127);
}

const bool Matcher::isSimpleConstant64(jlong value) {
  // Will one (StoreL ConL) be cheaper than two (StoreI ConI)?.
  //return value == (int) value;  // Cf. storeImmL and immL32.

  // Probably always true, even if a temp register is required.
  return true;
}

// The ecx parameter to rep stosq for the ClearArray node is in words.
const bool Matcher::init_array_count_is_in_bytes = false;

// No additional cost for CMOVL.
const int Matcher::long_cmove_cost() { return 0; }

// No CMOVF/CMOVD with SSE2
const int Matcher::float_cmove_cost() { return ConditionalMoveLimit; }

// Does the CPU require late expand (see block.cpp for description of late expand)?
const bool Matcher::require_postalloc_expand = false;

// Do we need to mask the count passed to shift instructions or does
// the cpu only look at the lower 5/6 bits anyway?
const bool Matcher::need_masked_shift_count = false;

bool Matcher::narrow_oop_use_complex_address() {
  assert(UseCompressedOops, "only for compressed oops code");
  return (LogMinObjAlignmentInBytes <= 3);
}

bool Matcher::narrow_klass_use_complex_address() {
  assert(UseCompressedClassPointers, "only for compressed klass code");
  return (LogKlassAlignmentInBytes <= 3);
}

bool Matcher::const_oop_prefer_decode() {
  // Prefer ConN+DecodeN over ConP.
  return true;
}

bool Matcher::const_klass_prefer_decode() {
  // TODO: Either support matching DecodeNKlass (heap-based) in operand
  //       or condisider the following:
  // Prefer ConNKlass+DecodeNKlass over ConP in simple compressed klass mode.
  //return CompressedKlassPointers::base() == NULL;
  return true;
}

// Is it better to copy float constants, or load them directly from
// memory?  Intel can load a float constant from a direct address,
// requiring no extra registers.  Most RISCs will have to materialize
// an address into a register first, so they would do better to copy
// the constant from stack.
const bool Matcher::rematerialize_float_constants = true; // XXX

// If CPU can load and store mis-aligned doubles directly then no
// fixup is needed.  Else we split the double into 2 integer pieces
// and move it piece-by-piece.  Only happens when passing doubles into
// C code as the Java calling convention forces doubles to be aligned.
const bool Matcher::misaligned_doubles_ok = true;

// No-op on amd64
void Matcher::pd_implicit_null_fixup(MachNode *node, uint idx) {}

// Advertise here if the CPU requires explicit rounding operations to
// implement the UseStrictFP mode.
const bool Matcher::strict_fp_requires_explicit_rounding = true;

// Are floats conerted to double when stored to stack during deoptimization?
// On x64 it is stored without convertion so we can use normal access.
bool Matcher::float_in_double() { return false; }

// Do ints take an entire long register or just half?
const bool Matcher::int_in_long = true;

// Return whether or not this register is ever used as an argument.
// This function is used on startup to build the trampoline stubs in
// generateOptoStub.  Registers not mentioned will be killed by the VM
// call in the trampoline, and arguments in those registers not be
// available to the callee.
bool Matcher::can_be_java_arg(int reg)
{
  return
    reg ==  RDI_num || reg == RDI_H_num ||
    reg ==  RSI_num || reg == RSI_H_num ||
    reg ==  RDX_num || reg == RDX_H_num ||
    reg ==  RCX_num || reg == RCX_H_num ||
    reg ==   R8_num || reg ==  R8_H_num ||
    reg ==   R9_num || reg ==  R9_H_num ||
    reg ==  R12_num || reg == R12_H_num ||
    reg == XMM0_num || reg == XMM0b_num ||
    reg == XMM1_num || reg == XMM1b_num ||
    reg == XMM2_num || reg == XMM2b_num ||
    reg == XMM3_num || reg == XMM3b_num ||
    reg == XMM4_num || reg == XMM4b_num ||
    reg == XMM5_num || reg == XMM5b_num ||
    reg == XMM6_num || reg == XMM6b_num ||
    reg == XMM7_num || reg == XMM7b_num;
}

bool Matcher::is_spillable_arg(int reg)
{
  return can_be_java_arg(reg);
}

bool Matcher::use_asm_for_ldiv_by_con( jlong divisor ) {
  // In 64 bit mode a code which use multiply when
  // devisor is constant is faster than hardware
  // DIV instruction (it uses MulHiL).
  return false;
}

// Register for DIVI projection of divmodI
RegMask Matcher::divI_proj_mask() {
  return INT_RAX_REG_mask();
}

// Register for MODI projection of divmodI
RegMask Matcher::modI_proj_mask() {
  return INT_RDX_REG_mask();
}

// Register for DIVL projection of divmodL
RegMask Matcher::divL_proj_mask() {
  return LONG_RAX_REG_mask();
}

// Register for MODL projection of divmodL
RegMask Matcher::modL_proj_mask() {
  return LONG_RDX_REG_mask();
}

// Register for saving SP into on method handle invokes. Not used on x86_64.
const RegMask Matcher::method_handle_invoke_SP_save_mask() {
    return NO_REG_mask();
}

%}

//----------ENCODING BLOCK-----------------------------------------------------
// This block specifies the encoding classes used by the compiler to
// output byte streams.  Encoding classes are parameterized macros
// used by Machine Instruction Nodes in order to generate the bit
// encoding of the instruction.  Operands specify their base encoding
// interface with the interface keyword.  There are currently
// supported four interfaces, REG_INTER, CONST_INTER, MEMORY_INTER, &
// COND_INTER.  REG_INTER causes an operand to generate a function
// which returns its register number when queried.  CONST_INTER causes
// an operand to generate a function which returns the value of the
// constant when queried.  MEMORY_INTER causes an operand to generate
// four functions which return the Base Register, the Index Register,
// the Scale Value, and the Offset Value of the operand when queried.
// COND_INTER causes an operand to generate six functions which return
// the encoding code (ie - encoding bits for the instruction)
// associated with each basic boolean condition for a conditional
// instruction.
//
// Instructions specify two basic values for encoding.  Again, a
// function is available to check if the constant displacement is an
// oop. They use the ins_encode keyword to specify their encoding
// classes (which must be a sequence of enc_class names, and their
// parameters, specified in the encoding block), and they use the
// opcode keyword to specify, in order, their primary, secondary, and
// tertiary opcode.  Only the opcode sections which a particular
// instruction needs for encoding need to be specified.
encode %{
  // Build emit functions for each basic byte or larger field in the
  // intel encoding scheme (opcode, rm, sib, immediate), and call them
  // from C++ code in the enc_class source block.  Emit functions will
  // live in the main source block for now.  In future, we can
  // generalize this by adding a syntax that specifies the sizes of
  // fields in an order, so that the adlc can build the emit functions
  // automagically

  // Emit primary opcode
  enc_class OpcP
  %{
    emit_opcode(cbuf, $primary);
  %}

  // Emit secondary opcode
  enc_class OpcS
  %{
    emit_opcode(cbuf, $secondary);
  %}

  // Emit tertiary opcode
  enc_class OpcT
  %{
    emit_opcode(cbuf, $tertiary);
  %}

  // Emit opcode directly
  enc_class Opcode(immI d8)
  %{
    emit_opcode(cbuf, $d8$$constant);
  %}

  // Emit size prefix
  enc_class SizePrefix
  %{
    emit_opcode(cbuf, 0x66);
  %}

  enc_class reg(rRegI reg)
  %{
    emit_rm(cbuf, 0x3, 0, $reg$$reg & 7);
  %}

  enc_class reg_reg(rRegI dst, rRegI src)
  %{
    emit_rm(cbuf, 0x3, $dst$$reg & 7, $src$$reg & 7);
  %}

  enc_class opc_reg_reg(immI opcode, rRegI dst, rRegI src)
  %{
    emit_opcode(cbuf, $opcode$$constant);
    emit_rm(cbuf, 0x3, $dst$$reg & 7, $src$$reg & 7);
  %}

  enc_class cdql_enc(no_rax_rdx_RegI div)
  %{
    // Full implementation of Java idiv and irem; checks for
    // special case as described in JVM spec., p.243 & p.271.
    //
    //         normal case                           special case
    //
    // input : rax: dividend                         min_int
    //         reg: divisor                          -1
    //
    // output: rax: quotient  (= rax idiv reg)       min_int
    //         rdx: remainder (= rax irem reg)       0
    //
    //  Code sequnce:
    //
    //    0:   3d 00 00 00 80          cmp    $0x80000000,%eax
    //    5:   75 07/08                jne    e <normal>
    //    7:   33 d2                   xor    %edx,%edx
    //  [div >= 8 -> offset + 1]
    //  [REX_B]
    //    9:   83 f9 ff                cmp    $0xffffffffffffffff,$div
    //    c:   74 03/04                je     11 <done>
    // 000000000000000e <normal>:
    //    e:   99                      cltd
    //  [div >= 8 -> offset + 1]
    //  [REX_B]
    //    f:   f7 f9                   idiv   $div
    // 0000000000000011 <done>:

    // cmp    $0x80000000,%eax
    emit_opcode(cbuf, 0x3d);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x80);

    // jne    e <normal>
    emit_opcode(cbuf, 0x75);
    emit_d8(cbuf, $div$$reg < 8 ? 0x07 : 0x08);

    // xor    %edx,%edx
    emit_opcode(cbuf, 0x33);
    emit_d8(cbuf, 0xD2);

    // cmp    $0xffffffffffffffff,%ecx
    if ($div$$reg >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
    }
    emit_opcode(cbuf, 0x83);
    emit_rm(cbuf, 0x3, 0x7, $div$$reg & 7);
    emit_d8(cbuf, 0xFF);

    // je     11 <done>
    emit_opcode(cbuf, 0x74);
    emit_d8(cbuf, $div$$reg < 8 ? 0x03 : 0x04);

    // <normal>
    // cltd
    emit_opcode(cbuf, 0x99);

    // idivl (note: must be emitted by the user of this rule)
    // <done>
  %}

  enc_class cdqq_enc(no_rax_rdx_RegL div)
  %{
    // Full implementation of Java ldiv and lrem; checks for
    // special case as described in JVM spec., p.243 & p.271.
    //
    //         normal case                           special case
    //
    // input : rax: dividend                         min_long
    //         reg: divisor                          -1
    //
    // output: rax: quotient  (= rax idiv reg)       min_long
    //         rdx: remainder (= rax irem reg)       0
    //
    //  Code sequnce:
    //
    //    0:   48 ba 00 00 00 00 00    mov    $0x8000000000000000,%rdx
    //    7:   00 00 80
    //    a:   48 39 d0                cmp    %rdx,%rax
    //    d:   75 08                   jne    17 <normal>
    //    f:   33 d2                   xor    %edx,%edx
    //   11:   48 83 f9 ff             cmp    $0xffffffffffffffff,$div
    //   15:   74 05                   je     1c <done>
    // 0000000000000017 <normal>:
    //   17:   48 99                   cqto
    //   19:   48 f7 f9                idiv   $div
    // 000000000000001c <done>:

    // mov    $0x8000000000000000,%rdx
    emit_opcode(cbuf, Assembler::REX_W);
    emit_opcode(cbuf, 0xBA);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x00);
    emit_d8(cbuf, 0x80);

    // cmp    %rdx,%rax
    emit_opcode(cbuf, Assembler::REX_W);
    emit_opcode(cbuf, 0x39);
    emit_d8(cbuf, 0xD0);

    // jne    17 <normal>
    emit_opcode(cbuf, 0x75);
    emit_d8(cbuf, 0x08);

    // xor    %edx,%edx
    emit_opcode(cbuf, 0x33);
    emit_d8(cbuf, 0xD2);

    // cmp    $0xffffffffffffffff,$div
    emit_opcode(cbuf, $div$$reg < 8 ? Assembler::REX_W : Assembler::REX_WB);
    emit_opcode(cbuf, 0x83);
    emit_rm(cbuf, 0x3, 0x7, $div$$reg & 7);
    emit_d8(cbuf, 0xFF);

    // je     1e <done>
    emit_opcode(cbuf, 0x74);
    emit_d8(cbuf, 0x05);

    // <normal>
    // cqto
    emit_opcode(cbuf, Assembler::REX_W);
    emit_opcode(cbuf, 0x99);

    // idivq (note: must be emitted by the user of this rule)
    // <done>
  %}

  // Opcde enc_class for 8/32 bit immediate instructions with sign-extension
  enc_class OpcSE(immI imm)
  %{
    // Emit primary opcode and set sign-extend bit
    // Check for 8-bit immediate, and set sign extend bit in opcode
    if (-0x80 <= $imm$$constant && $imm$$constant < 0x80) {
      emit_opcode(cbuf, $primary | 0x02);
    } else {
      // 32-bit immediate
      emit_opcode(cbuf, $primary);
    }
  %}

  enc_class OpcSErm(rRegI dst, immI imm)
  %{
    // OpcSEr/m
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    // Emit primary opcode and set sign-extend bit
    // Check for 8-bit immediate, and set sign extend bit in opcode
    if (-0x80 <= $imm$$constant && $imm$$constant < 0x80) {
      emit_opcode(cbuf, $primary | 0x02);
    } else {
      // 32-bit immediate
      emit_opcode(cbuf, $primary);
    }
    // Emit r/m byte with secondary opcode, after primary opcode.
    emit_rm(cbuf, 0x3, $secondary, dstenc);
  %}

  enc_class OpcSErm_wide(rRegL dst, immI imm)
  %{
    // OpcSEr/m
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    // Emit primary opcode and set sign-extend bit
    // Check for 8-bit immediate, and set sign extend bit in opcode
    if (-0x80 <= $imm$$constant && $imm$$constant < 0x80) {
      emit_opcode(cbuf, $primary | 0x02);
    } else {
      // 32-bit immediate
      emit_opcode(cbuf, $primary);
    }
    // Emit r/m byte with secondary opcode, after primary opcode.
    emit_rm(cbuf, 0x3, $secondary, dstenc);
  %}

  enc_class Con8or32(immI imm)
  %{
    // Check for 8-bit immediate, and set sign extend bit in opcode
    if (-0x80 <= $imm$$constant && $imm$$constant < 0x80) {
      $$$emit8$imm$$constant;
    } else {
      // 32-bit immediate
      $$$emit32$imm$$constant;
    }
  %}

  enc_class opc2_reg(rRegI dst)
  %{
    // BSWAP
    emit_cc(cbuf, $secondary, $dst$$reg);
  %}

  enc_class opc3_reg(rRegI dst)
  %{
    // BSWAP
    emit_cc(cbuf, $tertiary, $dst$$reg);
  %}

  enc_class reg_opc(rRegI div)
  %{
    // INC, DEC, IDIV, IMOD, JMP indirect, ...
    emit_rm(cbuf, 0x3, $secondary, $div$$reg & 7);
  %}

  enc_class enc_cmov(cmpOp cop)
  %{
    // CMOV
    $$$emit8$primary;
    emit_cc(cbuf, $secondary, $cop$$cmpcode);
  %}

  enc_class enc_PartialSubtypeCheck()
  %{
    Register Rrdi = as_Register(RDI_enc); // result register
    Register Rrax = as_Register(RAX_enc); // super class
    Register Rrcx = as_Register(RCX_enc); // killed
    Register Rrsi = as_Register(RSI_enc); // sub class
    Label miss;
    const bool set_cond_codes = true;

    MacroAssembler _masm(&cbuf);
    __ check_klass_subtype_slow_path(Rrsi, Rrax, Rrcx, Rrdi,
                                     NULL, &miss,
                                     /*set_cond_codes:*/ true);
    if ($primary) {
      __ xorptr(Rrdi, Rrdi);
    }
    __ bind(miss);
  %}

  enc_class clear_avx %{
    debug_only(int off0 = cbuf.insts_size());
    if (generate_vzeroupper(Compile::current())) {
      // Clear upper bits of YMM registers to avoid AVX <-> SSE transition penalty
      // Clear upper bits of YMM registers when current compiled code uses
      // wide vectors to avoid AVX <-> SSE transition penalty during call.
      MacroAssembler _masm(&cbuf);
      __ vzeroupper();
    }
    debug_only(int off1 = cbuf.insts_size());
    assert(off1 - off0 == clear_avx_size(), "correct size prediction");
  %}

  enc_class Java_To_Runtime(method meth) %{
    // No relocation needed
    MacroAssembler _masm(&cbuf);
    __ mov64(r10, (int64_t) $meth$$method);
    __ call(r10);
  %}

  enc_class Java_To_Interpreter(method meth)
  %{
    // CALL Java_To_Interpreter
    // This is the instruction starting address for relocation info.
    cbuf.set_insts_mark();
    $$$emit8$primary;
    // CALL directly to the runtime
    emit_d32_reloc(cbuf,
                   (int) ($meth$$method - ((intptr_t) cbuf.insts_end()) - 4),
                   runtime_call_Relocation::spec(),
                   RELOC_DISP32);
  %}

  enc_class Java_Static_Call(method meth)
  %{
    // JAVA STATIC CALL
    // CALL to fixup routine.  Fixup routine uses ScopeDesc info to
    // determine who we intended to call.
    cbuf.set_insts_mark();
    $$$emit8$primary;

    if (!_method) {
      emit_d32_reloc(cbuf, (int) ($meth$$method - ((intptr_t) cbuf.insts_end()) - 4),
                     runtime_call_Relocation::spec(),
                     RELOC_DISP32);
    } else {
      int method_index = resolved_method_index(cbuf);
      RelocationHolder rspec = _optimized_virtual ? opt_virtual_call_Relocation::spec(method_index)
                                                  : static_call_Relocation::spec(method_index);
      emit_d32_reloc(cbuf, (int) ($meth$$method - ((intptr_t) cbuf.insts_end()) - 4),
                     rspec, RELOC_DISP32);
      // Emit stubs for static call.
      address mark = cbuf.insts_mark();
      address stub = CompiledStaticCall::emit_to_interp_stub(cbuf, mark);
      if (stub == NULL) {
        ciEnv::current()->record_failure("CodeCache is full");
        return;
      }
#if INCLUDE_AOT
      CompiledStaticCall::emit_to_aot_stub(cbuf, mark);
#endif
    }
  %}

  enc_class Java_Dynamic_Call(method meth) %{
    MacroAssembler _masm(&cbuf);
    __ ic_call((address)$meth$$method, resolved_method_index(cbuf));
  %}

  enc_class Java_Compiled_Call(method meth)
  %{
    // JAVA COMPILED CALL
    int disp = in_bytes(Method:: from_compiled_offset());

    // XXX XXX offset is 128 is 1.5 NON-PRODUCT !!!
    // assert(-0x80 <= disp && disp < 0x80, "compiled_code_offset isn't small");

    // callq *disp(%rax)
    cbuf.set_insts_mark();
    $$$emit8$primary;
    if (disp < 0x80) {
      emit_rm(cbuf, 0x01, $secondary, RAX_enc); // R/M byte
      emit_d8(cbuf, disp); // Displacement
    } else {
      emit_rm(cbuf, 0x02, $secondary, RAX_enc); // R/M byte
      emit_d32(cbuf, disp); // Displacement
    }
  %}

  enc_class reg_opc_imm(rRegI dst, immI8 shift)
  %{
    // SAL, SAR, SHR
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    $$$emit8$primary;
    emit_rm(cbuf, 0x3, $secondary, dstenc);
    $$$emit8$shift$$constant;
  %}

  enc_class reg_opc_imm_wide(rRegL dst, immI8 shift)
  %{
    // SAL, SAR, SHR
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    $$$emit8$primary;
    emit_rm(cbuf, 0x3, $secondary, dstenc);
    $$$emit8$shift$$constant;
  %}

  enc_class load_immI(rRegI dst, immI src)
  %{
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xB8 | dstenc);
    $$$emit32$src$$constant;
  %}

  enc_class load_immL(rRegL dst, immL src)
  %{
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xB8 | dstenc);
    emit_d64(cbuf, $src$$constant);
  %}

  enc_class load_immUL32(rRegL dst, immUL32 src)
  %{
    // same as load_immI, but this time we care about zeroes in the high word
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xB8 | dstenc);
    $$$emit32$src$$constant;
  %}

  enc_class load_immL32(rRegL dst, immL32 src)
  %{
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xC7);
    emit_rm(cbuf, 0x03, 0x00, dstenc);
    $$$emit32$src$$constant;
  %}

  enc_class load_immP31(rRegP dst, immP32 src)
  %{
    // same as load_immI, but this time we care about zeroes in the high word
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xB8 | dstenc);
    $$$emit32$src$$constant;
  %}

  enc_class load_immP(rRegP dst, immP src)
  %{
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    emit_opcode(cbuf, 0xB8 | dstenc);
    // This next line should be generated from ADLC
    if ($src->constant_reloc() != relocInfo::none) {
      emit_d64_reloc(cbuf, $src$$constant, $src->constant_reloc(), RELOC_IMM64);
    } else {
      emit_d64(cbuf, $src$$constant);
    }
  %}

  enc_class Con32(immI src)
  %{
    // Output immediate
    $$$emit32$src$$constant;
  %}

  enc_class Con32F_as_bits(immF src)
  %{
    // Output Float immediate bits
    jfloat jf = $src$$constant;
    jint jf_as_bits = jint_cast(jf);
    emit_d32(cbuf, jf_as_bits);
  %}

  enc_class Con16(immI src)
  %{
    // Output immediate
    $$$emit16$src$$constant;
  %}

  // How is this different from Con32??? XXX
  enc_class Con_d32(immI src)
  %{
    emit_d32(cbuf,$src$$constant);
  %}

  enc_class conmemref (rRegP t1) %{    // Con32(storeImmI)
    // Output immediate memory reference
    emit_rm(cbuf, 0x00, $t1$$reg, 0x05 );
    emit_d32(cbuf, 0x00);
  %}

  enc_class lock_prefix()
  %{
    emit_opcode(cbuf, 0xF0); // lock
  %}

  enc_class REX_mem(memory mem)
  %{
    if ($mem$$base >= 8) {
      if ($mem$$index < 8) {
        emit_opcode(cbuf, Assembler::REX_B);
      } else {
        emit_opcode(cbuf, Assembler::REX_XB);
      }
    } else {
      if ($mem$$index >= 8) {
        emit_opcode(cbuf, Assembler::REX_X);
      }
    }
  %}

  enc_class REX_mem_wide(memory mem)
  %{
    if ($mem$$base >= 8) {
      if ($mem$$index < 8) {
        emit_opcode(cbuf, Assembler::REX_WB);
      } else {
        emit_opcode(cbuf, Assembler::REX_WXB);
      }
    } else {
      if ($mem$$index < 8) {
        emit_opcode(cbuf, Assembler::REX_W);
      } else {
        emit_opcode(cbuf, Assembler::REX_WX);
      }
    }
  %}

  // for byte regs
  enc_class REX_breg(rRegI reg)
  %{
    if ($reg$$reg >= 4) {
      emit_opcode(cbuf, $reg$$reg < 8 ? Assembler::REX : Assembler::REX_B);
    }
  %}

  // for byte regs
  enc_class REX_reg_breg(rRegI dst, rRegI src)
  %{
    if ($dst$$reg < 8) {
      if ($src$$reg >= 4) {
        emit_opcode(cbuf, $src$$reg < 8 ? Assembler::REX : Assembler::REX_B);
      }
    } else {
      if ($src$$reg < 8) {
        emit_opcode(cbuf, Assembler::REX_R);
      } else {
        emit_opcode(cbuf, Assembler::REX_RB);
      }
    }
  %}

  // for byte regs
  enc_class REX_breg_mem(rRegI reg, memory mem)
  %{
    if ($reg$$reg < 8) {
      if ($mem$$base < 8) {
        if ($mem$$index >= 8) {
          emit_opcode(cbuf, Assembler::REX_X);
        } else if ($reg$$reg >= 4) {
          emit_opcode(cbuf, Assembler::REX);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_B);
        } else {
          emit_opcode(cbuf, Assembler::REX_XB);
        }
      }
    } else {
      if ($mem$$base < 8) {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_R);
        } else {
          emit_opcode(cbuf, Assembler::REX_RX);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_RB);
        } else {
          emit_opcode(cbuf, Assembler::REX_RXB);
        }
      }
    }
  %}

  enc_class REX_reg(rRegI reg)
  %{
    if ($reg$$reg >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
    }
  %}

  enc_class REX_reg_wide(rRegI reg)
  %{
    if ($reg$$reg < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
    }
  %}

  enc_class REX_reg_reg(rRegI dst, rRegI src)
  %{
    if ($dst$$reg < 8) {
      if ($src$$reg >= 8) {
        emit_opcode(cbuf, Assembler::REX_B);
      }
    } else {
      if ($src$$reg < 8) {
        emit_opcode(cbuf, Assembler::REX_R);
      } else {
        emit_opcode(cbuf, Assembler::REX_RB);
      }
    }
  %}

  enc_class REX_reg_reg_wide(rRegI dst, rRegI src)
  %{
    if ($dst$$reg < 8) {
      if ($src$$reg < 8) {
        emit_opcode(cbuf, Assembler::REX_W);
      } else {
        emit_opcode(cbuf, Assembler::REX_WB);
      }
    } else {
      if ($src$$reg < 8) {
        emit_opcode(cbuf, Assembler::REX_WR);
      } else {
        emit_opcode(cbuf, Assembler::REX_WRB);
      }
    }
  %}

  enc_class REX_reg_mem(rRegI reg, memory mem)
  %{
    if ($reg$$reg < 8) {
      if ($mem$$base < 8) {
        if ($mem$$index >= 8) {
          emit_opcode(cbuf, Assembler::REX_X);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_B);
        } else {
          emit_opcode(cbuf, Assembler::REX_XB);
        }
      }
    } else {
      if ($mem$$base < 8) {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_R);
        } else {
          emit_opcode(cbuf, Assembler::REX_RX);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_RB);
        } else {
          emit_opcode(cbuf, Assembler::REX_RXB);
        }
      }
    }
  %}

  enc_class REX_reg_mem_wide(rRegL reg, memory mem)
  %{
    if ($reg$$reg < 8) {
      if ($mem$$base < 8) {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_W);
        } else {
          emit_opcode(cbuf, Assembler::REX_WX);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_WB);
        } else {
          emit_opcode(cbuf, Assembler::REX_WXB);
        }
      }
    } else {
      if ($mem$$base < 8) {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_WR);
        } else {
          emit_opcode(cbuf, Assembler::REX_WRX);
        }
      } else {
        if ($mem$$index < 8) {
          emit_opcode(cbuf, Assembler::REX_WRB);
        } else {
          emit_opcode(cbuf, Assembler::REX_WRXB);
        }
      }
    }
  %}

  enc_class reg_mem(rRegI ereg, memory mem)
  %{
    // High registers handle in encode_RegMem
    int reg = $ereg$$reg;
    int base = $mem$$base;
    int index = $mem$$index;
    int scale = $mem$$scale;
    int disp = $mem$$disp;
    relocInfo::relocType disp_reloc = $mem->disp_reloc();

    encode_RegMem(cbuf, reg, base, index, scale, disp, disp_reloc);
  %}

  enc_class RM_opc_mem(immI rm_opcode, memory mem)
  %{
    int rm_byte_opcode = $rm_opcode$$constant;

    // High registers handle in encode_RegMem
    int base = $mem$$base;
    int index = $mem$$index;
    int scale = $mem$$scale;
    int displace = $mem$$disp;

    relocInfo::relocType disp_reloc = $mem->disp_reloc();       // disp-as-oop when
                                            // working with static
                                            // globals
    encode_RegMem(cbuf, rm_byte_opcode, base, index, scale, displace,
                  disp_reloc);
  %}

  enc_class reg_lea(rRegI dst, rRegI src0, immI src1)
  %{
    int reg_encoding = $dst$$reg;
    int base         = $src0$$reg;      // 0xFFFFFFFF indicates no base
    int index        = 0x04;            // 0x04 indicates no index
    int scale        = 0x00;            // 0x00 indicates no scale
    int displace     = $src1$$constant; // 0x00 indicates no displacement
    relocInfo::relocType disp_reloc = relocInfo::none;
    encode_RegMem(cbuf, reg_encoding, base, index, scale, displace,
                  disp_reloc);
  %}

  enc_class neg_reg(rRegI dst)
  %{
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    }
    // NEG $dst
    emit_opcode(cbuf, 0xF7);
    emit_rm(cbuf, 0x3, 0x03, dstenc);
  %}

  enc_class neg_reg_wide(rRegI dst)
  %{
    int dstenc = $dst$$reg;
    if (dstenc < 8) {
      emit_opcode(cbuf, Assembler::REX_W);
    } else {
      emit_opcode(cbuf, Assembler::REX_WB);
      dstenc -= 8;
    }
    // NEG $dst
    emit_opcode(cbuf, 0xF7);
    emit_rm(cbuf, 0x3, 0x03, dstenc);
  %}

  enc_class setLT_reg(rRegI dst)
  %{
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    } else if (dstenc >= 4) {
      emit_opcode(cbuf, Assembler::REX);
    }
    // SETLT $dst
    emit_opcode(cbuf, 0x0F);
    emit_opcode(cbuf, 0x9C);
    emit_rm(cbuf, 0x3, 0x0, dstenc);
  %}

  enc_class setNZ_reg(rRegI dst)
  %{
    int dstenc = $dst$$reg;
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
      dstenc -= 8;
    } else if (dstenc >= 4) {
      emit_opcode(cbuf, Assembler::REX);
    }
    // SETNZ $dst
    emit_opcode(cbuf, 0x0F);
    emit_opcode(cbuf, 0x95);
    emit_rm(cbuf, 0x3, 0x0, dstenc);
  %}


  // Compare the lonogs and set -1, 0, or 1 into dst
  enc_class cmpl3_flag(rRegL src1, rRegL src2, rRegI dst)
  %{
    int src1enc = $src1$$reg;
    int src2enc = $src2$$reg;
    int dstenc = $dst$$reg;

    // cmpq $src1, $src2
    if (src1enc < 8) {
      if (src2enc < 8) {
        emit_opcode(cbuf, Assembler::REX_W);
      } else {
        emit_opcode(cbuf, Assembler::REX_WB);
      }
    } else {
      if (src2enc < 8) {
        emit_opcode(cbuf, Assembler::REX_WR);
      } else {
        emit_opcode(cbuf, Assembler::REX_WRB);
      }
    }
    emit_opcode(cbuf, 0x3B);
    emit_rm(cbuf, 0x3, src1enc & 7, src2enc & 7);

    // movl $dst, -1
    if (dstenc >= 8) {
      emit_opcode(cbuf, Assembler::REX_B);
    }
    emit_opcode(cbuf, 0xB8 | (dstenc & 7));
    emit_d32(cbuf, -1);

    // jl,s done
    emit_opcode(cbuf, 0x7C);
    emit_d8(cbuf, dstenc < 4 ? 0x06 : 0x08);

    // setne $dst
    if (dstenc >= 4) {
      emit_opcode(cbuf, dstenc < 8 ? Assembler::REX : Assembler::REX_B);
    }
    emit_opcode(cbuf, 0x0F);
    emit_opcode(cbuf, 0x95);
    emit_opcode(cbuf, 0xC0 | (dstenc & 7));

    // movzbl $dst, $dst
    if (dstenc >= 4) {
      emit_opcode(cbuf, dstenc < 8 ? Assembler::REX : Assembler::REX_RB);
    }
    emit_opcode(cbuf, 0x0F);
    emit_opcode(cbuf, 0xB6);
    emit_rm(cbuf, 0x3, dstenc & 7, dstenc & 7);
  %}

  enc_class Push_ResultXD(regD dst) %{
    MacroAssembler _masm(&cbuf);
    __ fstp_d(Address(rsp, 0));
    __ movdbl($dst$$XMMRegister, Address(rsp, 0));
    __ addptr(rsp, 8);
  %}

  enc_class Push_SrcXD(regD src) %{
    MacroAssembler _masm(&cbuf);
    __ subptr(rsp, 8);
    __ movdbl(Address(rsp, 0), $src$$XMMRegister);
    __ fld_d(Address(rsp, 0));
  %}


  enc_class enc_rethrow()
  %{
    cbuf.set_insts_mark();
    emit_opcode(cbuf, 0xE9); // jmp entry
    emit_d32_reloc(cbuf,
                   (int) (OptoRuntime::rethrow_stub() - cbuf.insts_end() - 4),
                   runtime_call_Relocation::spec(),
                   RELOC_DISP32);
  %}

%}


//----------FRAME--------------------------------------------------------------
// Definition of frame structure and management information.
//
//  S T A C K   L A Y O U T    Allocators stack-slot number
//                             |   (to get allocators register number
//  G  Owned by    |        |  v    add OptoReg::stack0())
//  r   CALLER     |        |
//  o     |        +--------+      pad to even-align allocators stack-slot
//  w     V        |  pad0  |        numbers; owned by CALLER
//  t   -----------+--------+----> Matcher::_in_arg_limit, unaligned
//  h     ^        |   in   |  5
//        |        |  args  |  4   Holes in incoming args owned by SELF
//  |     |        |        |  3
//  |     |        +--------+
//  V     |        | old out|      Empty on Intel, window on Sparc
//        |    old |preserve|      Must be even aligned.
//        |     SP-+--------+----> Matcher::_old_SP, even aligned
//        |        |   in   |  3   area for Intel ret address
//     Owned by    |preserve|      Empty on Sparc.
//       SELF      +--------+
//        |        |  pad2  |  2   pad to align old SP
//        |        +--------+  1
//        |        | locks  |  0
//        |        +--------+----> OptoReg::stack0(), even aligned
//        |        |  pad1  | 11   pad to align new SP
//        |        +--------+
//        |        |        | 10
//        |        | spills |  9   spills
//        V        |        |  8   (pad0 slot for callee)
//      -----------+--------+----> Matcher::_out_arg_limit, unaligned
//        ^        |  out   |  7
//        |        |  args  |  6   Holes in outgoing args owned by CALLEE
//     Owned by    +--------+
//      CALLEE     | new out|  6   Empty on Intel, window on Sparc
//        |    new |preserve|      Must be even-aligned.
//        |     SP-+--------+----> Matcher::_new_SP, even aligned
//        |        |        |
//
// Note 1: Only region 8-11 is determined by the allocator.  Region 0-5 is
//         known from SELF's arguments and the Java calling convention.
//         Region 6-7 is determined per call site.
// Note 2: If the calling convention leaves holes in the incoming argument
//         area, those holes are owned by SELF.  Holes in the outgoing area
//         are owned by the CALLEE.  Holes should not be nessecary in the
//         incoming area, as the Java calling convention is completely under
//         the control of the AD file.  Doubles can be sorted and packed to
//         avoid holes.  Holes in the outgoing arguments may be nessecary for
//         varargs C calling conventions.
// Note 3: Region 0-3 is even aligned, with pad2 as needed.  Region 3-5 is
//         even aligned with pad0 as needed.
//         Region 6 is even aligned.  Region 6-7 is NOT even aligned;
//         region 6-11 is even aligned; it may be padded out more so that
//         the region from SP to FP meets the minimum stack alignment.
// Note 4: For I2C adapters, the incoming FP may not meet the minimum stack
//         alignment.  Region 11, pad1, may be dynamically extended so that
//         SP meets the minimum alignment.

frame
%{
  // What direction does stack grow in (assumed to be same for C & Java)
  stack_direction(TOWARDS_LOW);

  // These three registers define part of the calling convention
  // between compiled code and the interpreter.
  inline_cache_reg(RAX);                // Inline Cache Register
  interpreter_method_oop_reg(RBX);      // Method Oop Register when
                                        // calling interpreter

  // Optional: name the operand used by cisc-spilling to access
  // [stack_pointer + offset]
  cisc_spilling_operand_name(indOffset32);

  // Number of stack slots consumed by locking an object
  sync_stack_slots(2);

  // Compiled code's Frame Pointer
  frame_pointer(RSP);

  // Interpreter stores its frame pointer in a register which is
  // stored to the stack by I2CAdaptors.
  // I2CAdaptors convert from interpreted java to compiled java.
  interpreter_frame_pointer(RBP);

  // Stack alignment requirement
  stack_alignment(StackAlignmentInBytes); // Alignment size in bytes (128-bit -> 16 bytes)

  // Number of stack slots between incoming argument block and the start of
  // a new frame.  The PROLOG must add this many slots to the stack.  The
  // EPILOG must remove this many slots.  amd64 needs two slots for
  // return address.
  in_preserve_stack_slots(4 + 2 * VerifyStackAtCalls);

  // Number of outgoing stack slots killed above the out_preserve_stack_slots
  // for calls to C.  Supports the var-args backing area for register parms.
  varargs_C_out_slots_killed(frame::arg_reg_save_area_bytes/BytesPerInt);

  // The after-PROLOG location of the return address.  Location of
  // return address specifies a type (REG or STACK) and a number
  // representing the register number (i.e. - use a register name) or
  // stack slot.
  // Ret Addr is on stack in slot 0 if no locks or verification or alignment.
  // Otherwise, it is above the locks and verification slot and alignment word
  return_addr(STACK - 2 +
              align_up((Compile::current()->in_preserve_stack_slots() +
                        Compile::current()->fixed_slots()),
                       stack_alignment_in_slots()));

  // Body of function which returns an integer array locating
  // arguments either in registers or in stack slots.  Passed an array
  // of ideal registers called "sig" and a "length" count.  Stack-slot
  // offsets are based on outgoing arguments, i.e. a CALLER setting up
  // arguments for a CALLEE.  Incoming stack arguments are
  // automatically biased by the preserve_stack_slots field above.

  calling_convention
  %{
    // No difference between ingoing/outgoing just pass false
    SharedRuntime::java_calling_convention(sig_bt, regs, length, false);
  %}

  c_calling_convention
  %{
    // This is obviously always outgoing
    (void) SharedRuntime::c_calling_convention(sig_bt, regs, /*regs2=*/NULL, length);
  %}

  // Location of compiled Java return values.  Same as C for now.
  return_value
  %{
    assert(ideal_reg >= Op_RegI && ideal_reg <= Op_RegL,
           "only return normal values");

    static const int lo[Op_RegL + 1] = {
      0,
      0,
      RAX_num,  // Op_RegN
      RAX_num,  // Op_RegI
      RAX_num,  // Op_RegP
      XMM0_num, // Op_RegF
      XMM0_num, // Op_RegD
      RAX_num   // Op_RegL
    };
    static const int hi[Op_RegL + 1] = {
      0,
      0,
      OptoReg::Bad, // Op_RegN
      OptoReg::Bad, // Op_RegI
      RAX_H_num,    // Op_RegP
      OptoReg::Bad, // Op_RegF
      XMM0b_num,    // Op_RegD
      RAX_H_num     // Op_RegL
    };
    // Excluded flags and vector registers.
    assert(ARRAY_SIZE(hi) == _last_machine_leaf - 6, "missing type");
    return OptoRegPair(hi[ideal_reg], lo[ideal_reg]);
  %}
%}

//----------ATTRIBUTES---------------------------------------------------------
//----------Operand Attributes-------------------------------------------------
op_attrib op_cost(0);        // Required cost attribute

//----------Instruction Attributes---------------------------------------------
ins_attrib ins_cost(100);       // Required cost attribute
ins_attrib ins_size(8);         // Required size attribute (in bits)
ins_attrib ins_short_branch(0); // Required flag: is this instruction
                                // a non-matching short branch variant
                                // of some long branch?
ins_attrib ins_alignment(1);    // Required alignment attribute (must
                                // be a power of 2) specifies the
                                // alignment that some part of the
                                // instruction (not necessarily the
                                // start) requires.  If > 1, a
                                // compute_padding() function must be
                                // provided for the instruction

//----------OPERANDS-----------------------------------------------------------
// Operand definitions must precede instruction definitions for correct parsing
// in the ADLC because operands constitute user defined types which are used in
// instruction definitions.

//----------Simple Operands----------------------------------------------------
// Immediate Operands
// Integer Immediate
operand immI()
%{
  match(ConI);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for test vs zero
operand immI0()
%{
  predicate(n->get_int() == 0);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for increment
operand immI1()
%{
  predicate(n->get_int() == 1);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for decrement
operand immI_M1()
%{
  predicate(n->get_int() == -1);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Valid scale values for addressing modes
operand immI2()
%{
  predicate(0 <= n->get_int() && (n->get_int() <= 3));
  match(ConI);

  format %{ %}
  interface(CONST_INTER);
%}

operand immI8()
%{
  predicate((-0x80 <= n->get_int()) && (n->get_int() < 0x80));
  match(ConI);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

operand immU8()
%{
  predicate((0 <= n->get_int()) && (n->get_int() <= 255));
  match(ConI);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

operand immI16()
%{
  predicate((-32768 <= n->get_int()) && (n->get_int() <= 32767));
  match(ConI);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// Int Immediate non-negative
operand immU31()
%{
  predicate(n->get_int() >= 0);
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for long shifts
operand immI_32()
%{
  predicate( n->get_int() == 32 );
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for long shifts
operand immI_64()
%{
  predicate( n->get_int() == 64 );
  match(ConI);

  op_cost(0);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer Immediate
operand immP()
%{
  match(ConP);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// NULL Pointer Immediate
operand immP0()
%{
  predicate(n->get_ptr() == 0);
  match(ConP);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

// Pointer Immediate
operand immN() %{
  match(ConN);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

operand immNKlass() %{
  match(ConNKlass);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// NULL Pointer Immediate
operand immN0() %{
  predicate(n->get_narrowcon() == 0);
  match(ConN);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

operand immP31()
%{
  predicate(n->as_Type()->type()->reloc() == relocInfo::none
            && (n->get_ptr() >> 31) == 0);
  match(ConP);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}


// Long Immediate
operand immL()
%{
  match(ConL);

  op_cost(20);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate 8-bit
operand immL8()
%{
  predicate(-0x80L <= n->get_long() && n->get_long() < 0x80L);
  match(ConL);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate 32-bit unsigned
operand immUL32()
%{
  predicate(n->get_long() == (unsigned int) (n->get_long()));
  match(ConL);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate 32-bit signed
operand immL32()
%{
  predicate(n->get_long() == (int) (n->get_long()));
  match(ConL);

  op_cost(15);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate zero
operand immL0()
%{
  predicate(n->get_long() == 0L);
  match(ConL);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// Constant for increment
operand immL1()
%{
  predicate(n->get_long() == 1);
  match(ConL);

  format %{ %}
  interface(CONST_INTER);
%}

// Constant for decrement
operand immL_M1()
%{
  predicate(n->get_long() == -1);
  match(ConL);

  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate: the value 10
operand immL10()
%{
  predicate(n->get_long() == 10);
  match(ConL);

  format %{ %}
  interface(CONST_INTER);
%}

// Long immediate from 0 to 127.
// Used for a shorter form of long mul by 10.
operand immL_127()
%{
  predicate(0 <= n->get_long() && n->get_long() < 0x80);
  match(ConL);

  op_cost(10);
  format %{ %}
  interface(CONST_INTER);
%}

// Long Immediate: low 32-bit mask
operand immL_32bits()
%{
  predicate(n->get_long() == 0xFFFFFFFFL);
  match(ConL);
  op_cost(20);

  format %{ %}
  interface(CONST_INTER);
%}

// Float Immediate zero
operand immF0()
%{
  predicate(jint_cast(n->getf()) == 0);
  match(ConF);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

// Float Immediate
operand immF()
%{
  match(ConF);

  op_cost(15);
  format %{ %}
  interface(CONST_INTER);
%}

// Double Immediate zero
operand immD0()
%{
  predicate(jlong_cast(n->getd()) == 0);
  match(ConD);

  op_cost(5);
  format %{ %}
  interface(CONST_INTER);
%}

// Double Immediate
operand immD()
%{
  match(ConD);

  op_cost(15);
  format %{ %}
  interface(CONST_INTER);
%}

// Immediates for special shifts (sign extend)

// Constants for increment
operand immI_16()
%{
  predicate(n->get_int() == 16);
  match(ConI);

  format %{ %}
  interface(CONST_INTER);
%}

operand immI_24()
%{
  predicate(n->get_int() == 24);
  match(ConI);

  format %{ %}
  interface(CONST_INTER);
%}

// Constant for byte-wide masking
operand immI_255()
%{
  predicate(n->get_int() == 255);
  match(ConI);

  format %{ %}
  interface(CONST_INTER);
%}

// Constant for short-wide masking
operand immI_65535()
%{
  predicate(n->get_int() == 65535);
  match(ConI);

  format %{ %}
  interface(CONST_INTER);
%}

// Constant for byte-wide masking
operand immL_255()
%{
  predicate(n->get_long() == 255);
  match(ConL);

  format %{ %}
  interface(CONST_INTER);
%}

// Constant for short-wide masking
operand immL_65535()
%{
  predicate(n->get_long() == 65535);
  match(ConL);

  format %{ %}
  interface(CONST_INTER);
%}

// Register Operands
// Integer Register
operand rRegI()
%{
  constraint(ALLOC_IN_RC(int_reg));
  match(RegI);

  match(rax_RegI);
  match(rbx_RegI);
  match(rcx_RegI);
  match(rdx_RegI);
  match(rdi_RegI);

  format %{ %}
  interface(REG_INTER);
%}

// Special Registers
operand rax_RegI()
%{
  constraint(ALLOC_IN_RC(int_rax_reg));
  match(RegI);
  match(rRegI);

  format %{ "RAX" %}
  interface(REG_INTER);
%}

// Special Registers
operand rbx_RegI()
%{
  constraint(ALLOC_IN_RC(int_rbx_reg));
  match(RegI);
  match(rRegI);

  format %{ "RBX" %}
  interface(REG_INTER);
%}

operand rcx_RegI()
%{
  constraint(ALLOC_IN_RC(int_rcx_reg));
  match(RegI);
  match(rRegI);

  format %{ "RCX" %}
  interface(REG_INTER);
%}

operand rdx_RegI()
%{
  constraint(ALLOC_IN_RC(int_rdx_reg));
  match(RegI);
  match(rRegI);

  format %{ "RDX" %}
  interface(REG_INTER);
%}

operand rdi_RegI()
%{
  constraint(ALLOC_IN_RC(int_rdi_reg));
  match(RegI);
  match(rRegI);

  format %{ "RDI" %}
  interface(REG_INTER);
%}

operand no_rcx_RegI()
%{
  constraint(ALLOC_IN_RC(int_no_rcx_reg));
  match(RegI);
  match(rax_RegI);
  match(rbx_RegI);
  match(rdx_RegI);
  match(rdi_RegI);

  format %{ %}
  interface(REG_INTER);
%}

operand no_rax_rdx_RegI()
%{
  constraint(ALLOC_IN_RC(int_no_rax_rdx_reg));
  match(RegI);
  match(rbx_RegI);
  match(rcx_RegI);
  match(rdi_RegI);

  format %{ %}
  interface(REG_INTER);
%}

// Pointer Register
operand any_RegP()
%{
  constraint(ALLOC_IN_RC(any_reg));
  match(RegP);
  match(rax_RegP);
  match(rbx_RegP);
  match(rdi_RegP);
  match(rsi_RegP);
  match(rbp_RegP);
  match(r15_RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

operand rRegP()
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(RegP);
  match(rax_RegP);
  match(rbx_RegP);
  match(rdi_RegP);
  match(rsi_RegP);
  match(rbp_RegP);  // See Q&A below about
  match(r15_RegP);  // r15_RegP and rbp_RegP.

  format %{ %}
  interface(REG_INTER);
%}

operand rRegN() %{
  constraint(ALLOC_IN_RC(int_reg));
  match(RegN);

  format %{ %}
  interface(REG_INTER);
%}

// Question: Why is r15_RegP (the read-only TLS register) a match for rRegP?
// Answer: Operand match rules govern the DFA as it processes instruction inputs.
// It's fine for an instruction input that expects rRegP to match a r15_RegP.
// The output of an instruction is controlled by the allocator, which respects
// register class masks, not match rules.  Unless an instruction mentions
// r15_RegP or any_RegP explicitly as its output, r15 will not be considered
// by the allocator as an input.
// The same logic applies to rbp_RegP being a match for rRegP: If PreserveFramePointer==true,
// the RBP is used as a proper frame pointer and is not included in ptr_reg. As a
// result, RBP is not included in the output of the instruction either.

operand no_rax_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_no_rax_reg));
  match(RegP);
  match(rbx_RegP);
  match(rsi_RegP);
  match(rdi_RegP);

  format %{ %}
  interface(REG_INTER);
%}

// This operand is not allowed to use RBP even if
// RBP is not used to hold the frame pointer.
operand no_rbp_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_reg_no_rbp));
  match(RegP);
  match(rbx_RegP);
  match(rsi_RegP);
  match(rdi_RegP);

  format %{ %}
  interface(REG_INTER);
%}

operand no_rax_rbx_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_no_rax_rbx_reg));
  match(RegP);
  match(rsi_RegP);
  match(rdi_RegP);

  format %{ %}
  interface(REG_INTER);
%}

// Special Registers
// Return a pointer value
operand rax_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_rax_reg));
  match(RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

// Special Registers
// Return a compressed pointer value
operand rax_RegN()
%{
  constraint(ALLOC_IN_RC(int_rax_reg));
  match(RegN);
  match(rRegN);

  format %{ %}
  interface(REG_INTER);
%}

// Used in AtomicAdd
operand rbx_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_rbx_reg));
  match(RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

operand rsi_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_rsi_reg));
  match(RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

// Used in rep stosq
operand rdi_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_rdi_reg));
  match(RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

operand r15_RegP()
%{
  constraint(ALLOC_IN_RC(ptr_r15_reg));
  match(RegP);
  match(rRegP);

  format %{ %}
  interface(REG_INTER);
%}

operand rRegL()
%{
  constraint(ALLOC_IN_RC(long_reg));
  match(RegL);
  match(rax_RegL);
  match(rdx_RegL);

  format %{ %}
  interface(REG_INTER);
%}

// Special Registers
operand no_rax_rdx_RegL()
%{
  constraint(ALLOC_IN_RC(long_no_rax_rdx_reg));
  match(RegL);
  match(rRegL);

  format %{ %}
  interface(REG_INTER);
%}

operand no_rax_RegL()
%{
  constraint(ALLOC_IN_RC(long_no_rax_rdx_reg));
  match(RegL);
  match(rRegL);
  match(rdx_RegL);

  format %{ %}
  interface(REG_INTER);
%}

operand no_rcx_RegL()
%{
  constraint(ALLOC_IN_RC(long_no_rcx_reg));
  match(RegL);
  match(rRegL);

  format %{ %}
  interface(REG_INTER);
%}

operand rax_RegL()
%{
  constraint(ALLOC_IN_RC(long_rax_reg));
  match(RegL);
  match(rRegL);

  format %{ "RAX" %}
  interface(REG_INTER);
%}

operand rcx_RegL()
%{
  constraint(ALLOC_IN_RC(long_rcx_reg));
  match(RegL);
  match(rRegL);

  format %{ %}
  interface(REG_INTER);
%}

operand rdx_RegL()
%{
  constraint(ALLOC_IN_RC(long_rdx_reg));
  match(RegL);
  match(rRegL);

  format %{ %}
  interface(REG_INTER);
%}

// Flags register, used as output of compare instructions
operand rFlagsReg()
%{
  constraint(ALLOC_IN_RC(int_flags));
  match(RegFlags);

  format %{ "RFLAGS" %}
  interface(REG_INTER);
%}

// Flags register, used as output of FLOATING POINT compare instructions
operand rFlagsRegU()
%{
  constraint(ALLOC_IN_RC(int_flags));
  match(RegFlags);

  format %{ "RFLAGS_U" %}
  interface(REG_INTER);
%}

operand rFlagsRegUCF() %{
  constraint(ALLOC_IN_RC(int_flags));
  match(RegFlags);
  predicate(false);

  format %{ "RFLAGS_U_CF" %}
  interface(REG_INTER);
%}

// Float register operands
operand regF() %{
   constraint(ALLOC_IN_RC(float_reg));
   match(RegF);

   format %{ %}
   interface(REG_INTER);
%}

// Float register operands
operand legRegF() %{
   constraint(ALLOC_IN_RC(float_reg_legacy));
   match(RegF);

   format %{ %}
   interface(REG_INTER);
%}

// Float register operands
operand vlRegF() %{
   constraint(ALLOC_IN_RC(float_reg_vl));
   match(RegF);

   format %{ %}
   interface(REG_INTER);
%}

// Double register operands
operand regD() %{
   constraint(ALLOC_IN_RC(double_reg));
   match(RegD);

   format %{ %}
   interface(REG_INTER);
%}

// Double register operands
operand legRegD() %{
   constraint(ALLOC_IN_RC(double_reg_legacy));
   match(RegD);

   format %{ %}
   interface(REG_INTER);
%}

// Double register operands
operand vlRegD() %{
   constraint(ALLOC_IN_RC(double_reg_vl));
   match(RegD);

   format %{ %}
   interface(REG_INTER);
%}

// Vectors
operand vecS() %{
  constraint(ALLOC_IN_RC(vectors_reg_vlbwdq));
  match(VecS);

  format %{ %}
  interface(REG_INTER);
%}

// Vectors
operand legVecS() %{
  constraint(ALLOC_IN_RC(vectors_reg_legacy));
  match(VecS);

  format %{ %}
  interface(REG_INTER);
%}

operand vecD() %{
  constraint(ALLOC_IN_RC(vectord_reg_vlbwdq));
  match(VecD);

  format %{ %}
  interface(REG_INTER);
%}

operand legVecD() %{
  constraint(ALLOC_IN_RC(vectord_reg_legacy));
  match(VecD);

  format %{ %}
  interface(REG_INTER);
%}

operand vecX() %{
  constraint(ALLOC_IN_RC(vectorx_reg_vlbwdq));
  match(VecX);

  format %{ %}
  interface(REG_INTER);
%}

operand legVecX() %{
  constraint(ALLOC_IN_RC(vectorx_reg_legacy));
  match(VecX);

  format %{ %}
  interface(REG_INTER);
%}

operand vecY() %{
  constraint(ALLOC_IN_RC(vectory_reg_vlbwdq));
  match(VecY);

  format %{ %}
  interface(REG_INTER);
%}

operand legVecY() %{
  constraint(ALLOC_IN_RC(vectory_reg_legacy));
  match(VecY);

  format %{ %}
  interface(REG_INTER);
%}

//----------Memory Operands----------------------------------------------------
// Direct Memory Operand
// operand direct(immP addr)
// %{
//   match(addr);

//   format %{ "[$addr]" %}
//   interface(MEMORY_INTER) %{
//     base(0xFFFFFFFF);
//     index(0x4);
//     scale(0x0);
//     disp($addr);
//   %}
// %}

// Indirect Memory Operand
operand indirect(any_RegP reg)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(reg);

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp(0x0);
  %}
%}

// Indirect Memory Plus Short Offset Operand
operand indOffset8(any_RegP reg, immL8 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);

  format %{ "[$reg + $off (8-bit)]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Long Offset Operand
operand indOffset32(any_RegP reg, immL32 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg off);

  format %{ "[$reg + $off (32-bit)]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Index Register Plus Offset Operand
operand indIndexOffset(any_RegP reg, rRegL lreg, immL32 off)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (AddP reg lreg) off);

  op_cost(10);
  format %{"[$reg + $off + $lreg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Index Register Plus Offset Operand
operand indIndex(any_RegP reg, rRegL lreg)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg lreg);

  op_cost(10);
  format %{"[$reg + $lreg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp(0x0);
  %}
%}

// Indirect Memory Times Scale Plus Index Register
operand indIndexScale(any_RegP reg, rRegL lreg, immI2 scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP reg (LShiftL lreg scale));

  op_cost(10);
  format %{"[$reg + $lreg << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp(0x0);
  %}
%}

operand indPosIndexScale(any_RegP reg, rRegI idx, immI2 scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(n->in(3)->in(1)->as_Type()->type()->is_long()->_lo >= 0);
  match(AddP reg (LShiftL (ConvI2L idx) scale));

  op_cost(10);
  format %{"[$reg + pos $idx << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($idx);
    scale($scale);
    disp(0x0);
  %}
%}

// Indirect Memory Times Scale Plus Index Register Plus Offset Operand
operand indIndexScaleOffset(any_RegP reg, immL32 off, rRegL lreg, immI2 scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (AddP reg (LShiftL lreg scale)) off);

  op_cost(10);
  format %{"[$reg + $off + $lreg << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp($off);
  %}
%}

// Indirect Memory Plus Positive Index Register Plus Offset Operand
operand indPosIndexOffset(any_RegP reg, immL32 off, rRegI idx)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(n->in(2)->in(3)->as_Type()->type()->is_long()->_lo >= 0);
  match(AddP (AddP reg (ConvI2L idx)) off);

  op_cost(10);
  format %{"[$reg + $off + $idx]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($idx);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Times Scale Plus Positive Index Register Plus Offset Operand
operand indPosIndexScaleOffset(any_RegP reg, immL32 off, rRegI idx, immI2 scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(n->in(2)->in(3)->in(1)->as_Type()->type()->is_long()->_lo >= 0);
  match(AddP (AddP reg (LShiftL (ConvI2L idx) scale)) off);

  op_cost(10);
  format %{"[$reg + $off + $idx << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($idx);
    scale($scale);
    disp($off);
  %}
%}

// Indirect Narrow Oop Plus Offset Operand
// Note: x86 architecture doesn't support "scale * index + offset" without a base
// we can't free r12 even with CompressedOops::base() == NULL.
operand indCompressedOopOffset(rRegN reg, immL32 off) %{
  predicate(UseCompressedOops && (CompressedOops::shift() == Address::times_8));
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) off);

  op_cost(10);
  format %{"[R12 + $reg << 3 + $off] (compressed oop addressing)" %}
  interface(MEMORY_INTER) %{
    base(0xc); // R12
    index($reg);
    scale(0x3);
    disp($off);
  %}
%}

// Indirect Memory Operand
operand indirectNarrow(rRegN reg)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(DecodeN reg);

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp(0x0);
  %}
%}

// Indirect Memory Plus Short Offset Operand
operand indOffset8Narrow(rRegN reg, immL8 off)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) off);

  format %{ "[$reg + $off (8-bit)]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Long Offset Operand
operand indOffset32Narrow(rRegN reg, immL32 off)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) off);

  format %{ "[$reg + $off (32-bit)]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index(0x4);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Index Register Plus Offset Operand
operand indIndexOffsetNarrow(rRegN reg, rRegL lreg, immL32 off)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (AddP (DecodeN reg) lreg) off);

  op_cost(10);
  format %{"[$reg + $off + $lreg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Plus Index Register Plus Offset Operand
operand indIndexNarrow(rRegN reg, rRegL lreg)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) lreg);

  op_cost(10);
  format %{"[$reg + $lreg]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale(0x0);
    disp(0x0);
  %}
%}

// Indirect Memory Times Scale Plus Index Register
operand indIndexScaleNarrow(rRegN reg, rRegL lreg, immI2 scale)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (DecodeN reg) (LShiftL lreg scale));

  op_cost(10);
  format %{"[$reg + $lreg << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp(0x0);
  %}
%}

// Indirect Memory Times Scale Plus Index Register Plus Offset Operand
operand indIndexScaleOffsetNarrow(rRegN reg, immL32 off, rRegL lreg, immI2 scale)
%{
  predicate(CompressedOops::shift() == 0);
  constraint(ALLOC_IN_RC(ptr_reg));
  match(AddP (AddP (DecodeN reg) (LShiftL lreg scale)) off);

  op_cost(10);
  format %{"[$reg + $off + $lreg << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($lreg);
    scale($scale);
    disp($off);
  %}
%}

// Indirect Memory Times Plus Positive Index Register Plus Offset Operand
operand indPosIndexOffsetNarrow(rRegN reg, immL32 off, rRegI idx)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(CompressedOops::shift() == 0 && n->in(2)->in(3)->as_Type()->type()->is_long()->_lo >= 0);
  match(AddP (AddP (DecodeN reg) (ConvI2L idx)) off);

  op_cost(10);
  format %{"[$reg + $off + $idx]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($idx);
    scale(0x0);
    disp($off);
  %}
%}

// Indirect Memory Times Scale Plus Positive Index Register Plus Offset Operand
operand indPosIndexScaleOffsetNarrow(rRegN reg, immL32 off, rRegI idx, immI2 scale)
%{
  constraint(ALLOC_IN_RC(ptr_reg));
  predicate(CompressedOops::shift() == 0 && n->in(2)->in(3)->in(1)->as_Type()->type()->is_long()->_lo >= 0);
  match(AddP (AddP (DecodeN reg) (LShiftL (ConvI2L idx) scale)) off);

  op_cost(10);
  format %{"[$reg + $off + $idx << $scale]" %}
  interface(MEMORY_INTER) %{
    base($reg);
    index($idx);
    scale($scale);
    disp($off);
  %}
%}

//----------Special Memory Operands--------------------------------------------
// Stack Slot Operand - This operand is used for loading and storing temporary
//                      values on the stack where a match requires a value to
//                      flow through memory.
operand stackSlotP(sRegP reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x4);   // RSP
    index(0x4);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotI(sRegI reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x4);   // RSP
    index(0x4);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotF(sRegF reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x4);   // RSP
    index(0x4);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

operand stackSlotD(sRegD reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x4);   // RSP
    index(0x4);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}
operand stackSlotL(sRegL reg)
%{
  constraint(ALLOC_IN_RC(stack_slots));
  // No match rule because this operand is only generated in matching

  format %{ "[$reg]" %}
  interface(MEMORY_INTER) %{
    base(0x4);   // RSP
    index(0x4);  // No Index
    scale(0x0);  // No Scale
    disp($reg);  // Stack Offset
  %}
%}

//----------Conditional Branch Operands----------------------------------------
// Comparison Op  - This is the operation of the comparison, and is limited to
//                  the following set of codes:
//                  L (<), LE (<=), G (>), GE (>=), E (==), NE (!=)
//
// Other attributes of the comparison, such as unsignedness, are specified
// by the comparison instruction that sets a condition code flags register.
// That result is represented by a flags operand whose subtype is appropriate
// to the unsignedness (etc.) of the comparison.
//
// Later, the instruction which matches both the Comparison Op (a Bool) and
// the flags (produced by the Cmp) specifies the coding of the comparison op
// by matching a specific subtype of Bool operand below, such as cmpOpU.

// Comparision Code
operand cmpOp()
%{
  match(Bool);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x4, "e");
    not_equal(0x5, "ne");
    less(0xC, "l");
    greater_equal(0xD, "ge");
    less_equal(0xE, "le");
    greater(0xF, "g");
    overflow(0x0, "o");
    no_overflow(0x1, "no");
  %}
%}

// Comparison Code, unsigned compare.  Used by FP also, with
// C2 (unordered) turned into GT or LT already.  The other bits
// C0 and C3 are turned into Carry & Zero flags.
operand cmpOpU()
%{
  match(Bool);

  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x4, "e");
    not_equal(0x5, "ne");
    less(0x2, "b");
    greater_equal(0x3, "nb");
    less_equal(0x6, "be");
    greater(0x7, "nbe");
    overflow(0x0, "o");
    no_overflow(0x1, "no");
  %}
%}


// Floating comparisons that don't require any fixup for the unordered case
operand cmpOpUCF() %{
  match(Bool);
  predicate(n->as_Bool()->_test._test == BoolTest::lt ||
            n->as_Bool()->_test._test == BoolTest::ge ||
            n->as_Bool()->_test._test == BoolTest::le ||
            n->as_Bool()->_test._test == BoolTest::gt);
  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x4, "e");
    not_equal(0x5, "ne");
    less(0x2, "b");
    greater_equal(0x3, "nb");
    less_equal(0x6, "be");
    greater(0x7, "nbe");
    overflow(0x0, "o");
    no_overflow(0x1, "no");
  %}
%}


// Floating comparisons that can be fixed up with extra conditional jumps
operand cmpOpUCF2() %{
  match(Bool);
  predicate(n->as_Bool()->_test._test == BoolTest::ne ||
            n->as_Bool()->_test._test == BoolTest::eq);
  format %{ "" %}
  interface(COND_INTER) %{
    equal(0x4, "e");
    not_equal(0x5, "ne");
    less(0x2, "b");
    greater_equal(0x3, "nb");
    less_equal(0x6, "be");
    greater(0x7, "nbe");
    overflow(0x0, "o");
    no_overflow(0x1, "no");
  %}
%}

//----------OPERAND CLASSES----------------------------------------------------
// Operand Classes are groups of operands that are used as to simplify
// instruction definitions by not requiring the AD writer to specify separate
// instructions for every form of operand when the instruction accepts
// multiple operand types with the same basic encoding and format.  The classic
// case of this is memory operands.

opclass memory(indirect, indOffset8, indOffset32, indIndexOffset, indIndex,
               indIndexScale, indPosIndexScale, indIndexScaleOffset, indPosIndexOffset, indPosIndexScaleOffset,
               indCompressedOopOffset,
               indirectNarrow, indOffset8Narrow, indOffset32Narrow,
               indIndexOffsetNarrow, indIndexNarrow, indIndexScaleNarrow,
               indIndexScaleOffsetNarrow, indPosIndexOffsetNarrow, indPosIndexScaleOffsetNarrow);

//----------PIPELINE-----------------------------------------------------------
// Rules which define the behavior of the target architectures pipeline.
pipeline %{

//----------ATTRIBUTES---------------------------------------------------------
attributes %{
  variable_size_instructions;        // Fixed size instructions
  max_instructions_per_bundle = 3;   // Up to 3 instructions per bundle
  instruction_unit_size = 1;         // An instruction is 1 bytes long
  instruction_fetch_unit_size = 16;  // The processor fetches one line
  instruction_fetch_units = 1;       // of 16 bytes

  // List of nop instructions
  nops( MachNop );
%}

//----------RESOURCES----------------------------------------------------------
// Resources are the functional units available to the machine

// Generic P2/P3 pipeline
// 3 decoders, only D0 handles big operands; a "bundle" is the limit of
// 3 instructions decoded per cycle.
// 2 load/store ops per cycle, 1 branch, 1 FPU,
// 3 ALU op, only ALU0 handles mul instructions.
resources( D0, D1, D2, DECODE = D0 | D1 | D2,
           MS0, MS1, MS2, MEM = MS0 | MS1 | MS2,
           BR, FPU,
           ALU0, ALU1, ALU2, ALU = ALU0 | ALU1 | ALU2);

//----------PIPELINE DESCRIPTION-----------------------------------------------
// Pipeline Description specifies the stages in the machine's pipeline

// Generic P2/P3 pipeline
pipe_desc(S0, S1, S2, S3, S4, S5);

//----------PIPELINE CLASSES---------------------------------------------------
// Pipeline Classes describe the stages in which input and output are
// referenced by the hardware pipeline.

// Naming convention: ialu or fpu
// Then: _reg
// Then: _reg if there is a 2nd register
// Then: _long if it's a pair of instructions implementing a long
// Then: _fat if it requires the big decoder
//   Or: _mem if it requires the big decoder and a memory unit.

// Integer ALU reg operation
pipe_class ialu_reg(rRegI dst)
%{
    single_instruction;
    dst    : S4(write);
    dst    : S3(read);
    DECODE : S0;        // any decoder
    ALU    : S3;        // any alu
%}

// Long ALU reg operation
pipe_class ialu_reg_long(rRegL dst)
%{
    instruction_count(2);
    dst    : S4(write);
    dst    : S3(read);
    DECODE : S0(2);     // any 2 decoders
    ALU    : S3(2);     // both alus
%}

// Integer ALU reg operation using big decoder
pipe_class ialu_reg_fat(rRegI dst)
%{
    single_instruction;
    dst    : S4(write);
    dst    : S3(read);
    D0     : S0;        // big decoder only
    ALU    : S3;        // any alu
%}

// Long ALU reg operation using big decoder
pipe_class ialu_reg_long_fat(rRegL dst)
%{
    instruction_count(2);
    dst    : S4(write);
    dst    : S3(read);
    D0     : S0(2);     // big decoder only; twice
    ALU    : S3(2);     // any 2 alus
%}

// Integer ALU reg-reg operation
pipe_class ialu_reg_reg(rRegI dst, rRegI src)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    DECODE : S0;        // any decoder
    ALU    : S3;        // any alu
%}

// Long ALU reg-reg operation
pipe_class ialu_reg_reg_long(rRegL dst, rRegL src)
%{
    instruction_count(2);
    dst    : S4(write);
    src    : S3(read);
    DECODE : S0(2);     // any 2 decoders
    ALU    : S3(2);     // both alus
%}

// Integer ALU reg-reg operation
pipe_class ialu_reg_reg_fat(rRegI dst, memory src)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    D0     : S0;        // big decoder only
    ALU    : S3;        // any alu
%}

// Long ALU reg-reg operation
pipe_class ialu_reg_reg_long_fat(rRegL dst, rRegL src)
%{
    instruction_count(2);
    dst    : S4(write);
    src    : S3(read);
    D0     : S0(2);     // big decoder only; twice
    ALU    : S3(2);     // both alus
%}

// Integer ALU reg-mem operation
pipe_class ialu_reg_mem(rRegI dst, memory mem)
%{
    single_instruction;
    dst    : S5(write);
    mem    : S3(read);
    D0     : S0;        // big decoder only
    ALU    : S4;        // any alu
    MEM    : S3;        // any mem
%}

// Integer mem operation (prefetch)
pipe_class ialu_mem(memory mem)
%{
    single_instruction;
    mem    : S3(read);
    D0     : S0;        // big decoder only
    MEM    : S3;        // any mem
%}

// Integer Store to Memory
pipe_class ialu_mem_reg(memory mem, rRegI src)
%{
    single_instruction;
    mem    : S3(read);
    src    : S5(read);
    D0     : S0;        // big decoder only
    ALU    : S4;        // any alu
    MEM    : S3;
%}

// // Long Store to Memory
// pipe_class ialu_mem_long_reg(memory mem, rRegL src)
// %{
//     instruction_count(2);
//     mem    : S3(read);
//     src    : S5(read);
//     D0     : S0(2);          // big decoder only; twice
//     ALU    : S4(2);     // any 2 alus
//     MEM    : S3(2);  // Both mems
// %}

// Integer Store to Memory
pipe_class ialu_mem_imm(memory mem)
%{
    single_instruction;
    mem    : S3(read);
    D0     : S0;        // big decoder only
    ALU    : S4;        // any alu
    MEM    : S3;
%}

// Integer ALU0 reg-reg operation
pipe_class ialu_reg_reg_alu0(rRegI dst, rRegI src)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    D0     : S0;        // Big decoder only
    ALU0   : S3;        // only alu0
%}

// Integer ALU0 reg-mem operation
pipe_class ialu_reg_mem_alu0(rRegI dst, memory mem)
%{
    single_instruction;
    dst    : S5(write);
    mem    : S3(read);
    D0     : S0;        // big decoder only
    ALU0   : S4;        // ALU0 only
    MEM    : S3;        // any mem
%}

// Integer ALU reg-reg operation
pipe_class ialu_cr_reg_reg(rFlagsReg cr, rRegI src1, rRegI src2)
%{
    single_instruction;
    cr     : S4(write);
    src1   : S3(read);
    src2   : S3(read);
    DECODE : S0;        // any decoder
    ALU    : S3;        // any alu
%}

// Integer ALU reg-imm operation
pipe_class ialu_cr_reg_imm(rFlagsReg cr, rRegI src1)
%{
    single_instruction;
    cr     : S4(write);
    src1   : S3(read);
    DECODE : S0;        // any decoder
    ALU    : S3;        // any alu
%}

// Integer ALU reg-mem operation
pipe_class ialu_cr_reg_mem(rFlagsReg cr, rRegI src1, memory src2)
%{
    single_instruction;
    cr     : S4(write);
    src1   : S3(read);
    src2   : S3(read);
    D0     : S0;        // big decoder only
    ALU    : S4;        // any alu
    MEM    : S3;
%}

// Conditional move reg-reg
pipe_class pipe_cmplt( rRegI p, rRegI q, rRegI y)
%{
    instruction_count(4);
    y      : S4(read);
    q      : S3(read);
    p      : S3(read);
    DECODE : S0(4);     // any decoder
%}

// Conditional move reg-reg
pipe_class pipe_cmov_reg( rRegI dst, rRegI src, rFlagsReg cr)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    cr     : S3(read);
    DECODE : S0;        // any decoder
%}

// Conditional move reg-mem
pipe_class pipe_cmov_mem( rFlagsReg cr, rRegI dst, memory src)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    cr     : S3(read);
    DECODE : S0;        // any decoder
    MEM    : S3;
%}

// Conditional move reg-reg long
pipe_class pipe_cmov_reg_long( rFlagsReg cr, rRegL dst, rRegL src)
%{
    single_instruction;
    dst    : S4(write);
    src    : S3(read);
    cr     : S3(read);
    DECODE : S0(2);     // any 2 decoders
%}

// XXX
// // Conditional move double reg-reg
// pipe_class pipe_cmovD_reg( rFlagsReg cr, regDPR1 dst, regD src)
// %{
//     single_instruction;
//     dst    : S4(write);
//     src    : S3(read);
//     cr     : S3(read);
//     DECODE : S0;     // any decoder
// %}

// Float reg-reg operation
pipe_class fpu_reg(regD dst)
%{
    instruction_count(2);
    dst    : S3(read);
    DECODE : S0(2);     // any 2 decoders
    FPU    : S3;
%}

// Float reg-reg operation
pipe_class fpu_reg_reg(regD dst, regD src)
%{
    instruction_count(2);
    dst    : S4(write);
    src    : S3(read);
    DECODE : S0(2);     // any 2 decoders
    FPU    : S3;
%}

// Float reg-reg operation
pipe_class fpu_reg_reg_reg(regD dst, regD src1, regD src2)
%{
    instruction_count(3);
    dst    : S4(write);
    src1   : S3(read);
    src2   : S3(read);
    DECODE : S0(3);     // any 3 decoders
    FPU    : S3(2);
%}

// Float reg-reg operation
pipe_class fpu_reg_reg_reg_reg(regD dst, regD src1, regD src2, regD src3)
%{
    instruction_count(4);
    dst    : S4(write);
    src1   : S3(read);
    src2   : S3(read);
    src3   : S3(read);
    DECODE : S0(4);     // any 3 decoders
    FPU    : S3(2);
%}

// Float reg-reg operation
pipe_class fpu_reg_mem_reg_reg(regD dst, memory src1, regD src2, regD src3)
%{
    instruction_count(4);
    dst    : S4(write);
    src1   : S3(read);
    src2   : S3(read);
    src3   : S3(read);
    DECODE : S1(3);     // any 3 decoders
    D0     : S0;        // Big decoder only
    FPU    : S3(2);
    MEM    : S3;
%}

// Float reg-mem operation
pipe_class fpu_reg_mem(regD dst, memory mem)
%{
    instruction_count(2);
    dst    : S5(write);
    mem    : S3(read);
    D0     : S0;        // big decoder only
    DECODE : S1;        // any decoder for FPU POP
    FPU    : S4;
    MEM    : S3;        // any mem
%}

// Float reg-mem operation
pipe_class fpu_reg_reg_mem(regD dst, regD src1, memory mem)
%{
    instruction_count(3);
    dst    : S5(write);
    src1   : S3(read);
    mem    : S3(read);
    D0     : S0;        // big decoder only
    DECODE : S1(2);     // any decoder for FPU POP
    FPU    : S4;
    MEM    : S3;        // any mem
%}

// Float mem-reg operation
pipe_class fpu_mem_reg(memory mem, regD src)
%{
    instruction_count(2);
    src    : S5(read);
    mem    : S3(read);
    DECODE : S0;        // any decoder for FPU PUSH
    D0     : S1;        // big decoder only
    FPU    : S4;
    MEM    : S3;        // any mem
%}

pipe_class fpu_mem_reg_reg(memory mem, regD src1, regD src2)
%{
    instruction_count(3);
    src1   : S3(read);
    src2   : S3(read);
    mem    : S3(read);
    DECODE : S0(2);     // any decoder for FPU PUSH
    D0     : S1;        // big decoder only
    FPU    : S4;
    MEM    : S3;        // any mem
%}

pipe_class fpu_mem_reg_mem(memory mem, regD src1, memory src2)
%{
    instruction_count(3);
    src1   : S3(read);
    src2   : S3(read);
    mem    : S4(read);
    DECODE : S0;        // any decoder for FPU PUSH
    D0     : S0(2);     // big decoder only
    FPU    : S4;
    MEM    : S3(2);     // any mem
%}

pipe_class fpu_mem_mem(memory dst, memory src1)
%{
    instruction_count(2);
    src1   : S3(read);
    dst    : S4(read);
    D0     : S0(2);     // big decoder only
    MEM    : S3(2);     // any mem
%}

pipe_class fpu_mem_mem_mem(memory dst, memory src1, memory src2)
%{
    instruction_count(3);
    src1   : S3(read);
    src2   : S3(read);
    dst    : S4(read);
    D0     : S0(3);     // big decoder only
    FPU    : S4;
    MEM    : S3(3);     // any mem
%}

pipe_class fpu_mem_reg_con(memory mem, regD src1)
%{
    instruction_count(3);
    src1   : S4(read);
    mem    : S4(read);
    DECODE : S0;        // any decoder for FPU PUSH
    D0     : S0(2);     // big decoder only
    FPU    : S4;
    MEM    : S3(2);     // any mem
%}

// Float load constant
pipe_class fpu_reg_con(regD dst)
%{
    instruction_count(2);
    dst    : S5(write);
    D0     : S0;        // big decoder only for the load
    DECODE : S1;        // any decoder for FPU POP
    FPU    : S4;
    MEM    : S3;        // any mem
%}

// Float load constant
pipe_class fpu_reg_reg_con(regD dst, regD src)
%{
    instruction_count(3);
    dst    : S5(write);
    src    : S3(read);
    D0     : S0;        // big decoder only for the load
    DECODE : S1(2);     // any decoder for FPU POP
    FPU    : S4;
    MEM    : S3;        // any mem
%}

// UnConditional branch
pipe_class pipe_jmp(label labl)
%{
    single_instruction;
    BR   : S3;
%}

// Conditional branch
pipe_class pipe_jcc(cmpOp cmp, rFlagsReg cr, label labl)
%{
    single_instruction;
    cr    : S1(read);
    BR    : S3;
%}

// Allocation idiom
pipe_class pipe_cmpxchg(rRegP dst, rRegP heap_ptr)
%{
    instruction_count(1); force_serialization;
    fixed_latency(6);
    heap_ptr : S3(read);
    DECODE   : S0(3);
    D0       : S2;
    MEM      : S3;
    ALU      : S3(2);
    dst      : S5(write);
    BR       : S5;
%}

// Generic big/slow expanded idiom
pipe_class pipe_slow()
%{
    instruction_count(10); multiple_bundles; force_serialization;
    fixed_latency(100);
    D0  : S0(2);
    MEM : S3(2);
%}

// The real do-nothing guy
pipe_class empty()
%{
    instruction_count(0);
%}

// Define the class for the Nop node
define
%{
   MachNop = empty;
%}

%}

//----------INSTRUCTIONS-------------------------------------------------------
//
// match      -- States which machine-independent subtree may be replaced
//               by this instruction.
// ins_cost   -- The estimated cost of this instruction is used by instruction
//               selection to identify a minimum cost tree of machine
//               instructions that matches a tree of machine-independent
//               instructions.
// format     -- A string providing the disassembly for this instruction.
//               The value of an instruction's operand may be inserted
//               by referring to it with a '$' prefix.
// opcode     -- Three instruction opcodes may be provided.  These are referred
//               to within an encode class as $primary, $secondary, and $tertiary
//               rrspectively.  The primary opcode is commonly used to
//               indicate the type of machine instruction, while secondary
//               and tertiary are often used for prefix options or addressing
//               modes.
// ins_encode -- A list of encode classes with parameters. The encode class
//               name must have been defined in an 'enc_class' specification
//               in the encode section of the architecture description.


//----------Load/Store/Move Instructions---------------------------------------
//----------Load Instructions--------------------------------------------------

// Load Byte (8 bit signed)
instruct loadB(rRegI dst, memory mem)
%{
  match(Set dst (LoadB mem));

  ins_cost(125);
  format %{ "movsbl  $dst, $mem\t# byte" %}

  ins_encode %{
    __ movsbl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Byte (8 bit signed) into Long Register
instruct loadB2L(rRegL dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadB mem)));

  ins_cost(125);
  format %{ "movsbq  $dst, $mem\t# byte -> long" %}

  ins_encode %{
    __ movsbq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Byte (8 bit UNsigned)
instruct loadUB(rRegI dst, memory mem)
%{
  match(Set dst (LoadUB mem));

  ins_cost(125);
  format %{ "movzbl  $dst, $mem\t# ubyte" %}

  ins_encode %{
    __ movzbl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Byte (8 bit UNsigned) into Long Register
instruct loadUB2L(rRegL dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadUB mem)));

  ins_cost(125);
  format %{ "movzbq  $dst, $mem\t# ubyte -> long" %}

  ins_encode %{
    __ movzbq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Byte (8 bit UNsigned) with 32-bit mask into Long Register
instruct loadUB2L_immI(rRegL dst, memory mem, immI mask, rFlagsReg cr) %{
  match(Set dst (ConvI2L (AndI (LoadUB mem) mask)));
  effect(KILL cr);

  format %{ "movzbq  $dst, $mem\t# ubyte & 32-bit mask -> long\n\t"
            "andl    $dst, right_n_bits($mask, 8)" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    __ movzbq(Rdst, $mem$$Address);
    __ andl(Rdst, $mask$$constant & right_n_bits(8));
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Short (16 bit signed)
instruct loadS(rRegI dst, memory mem)
%{
  match(Set dst (LoadS mem));

  ins_cost(125);
  format %{ "movswl $dst, $mem\t# short" %}

  ins_encode %{
    __ movswl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Short (16 bit signed) to Byte (8 bit signed)
instruct loadS2B(rRegI dst, memory mem, immI_24 twentyfour) %{
  match(Set dst (RShiftI (LShiftI (LoadS mem) twentyfour) twentyfour));

  ins_cost(125);
  format %{ "movsbl $dst, $mem\t# short -> byte" %}
  ins_encode %{
    __ movsbl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Short (16 bit signed) into Long Register
instruct loadS2L(rRegL dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadS mem)));

  ins_cost(125);
  format %{ "movswq $dst, $mem\t# short -> long" %}

  ins_encode %{
    __ movswq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Short/Char (16 bit UNsigned)
instruct loadUS(rRegI dst, memory mem)
%{
  match(Set dst (LoadUS mem));

  ins_cost(125);
  format %{ "movzwl  $dst, $mem\t# ushort/char" %}

  ins_encode %{
    __ movzwl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Short/Char (16 bit UNsigned) to Byte (8 bit signed)
instruct loadUS2B(rRegI dst, memory mem, immI_24 twentyfour) %{
  match(Set dst (RShiftI (LShiftI (LoadUS mem) twentyfour) twentyfour));

  ins_cost(125);
  format %{ "movsbl $dst, $mem\t# ushort -> byte" %}
  ins_encode %{
    __ movsbl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Short/Char (16 bit UNsigned) into Long Register
instruct loadUS2L(rRegL dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadUS mem)));

  ins_cost(125);
  format %{ "movzwq  $dst, $mem\t# ushort/char -> long" %}

  ins_encode %{
    __ movzwq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Short/Char (16 bit UNsigned) with mask 0xFF into Long Register
instruct loadUS2L_immI_255(rRegL dst, memory mem, immI_255 mask) %{
  match(Set dst (ConvI2L (AndI (LoadUS mem) mask)));

  format %{ "movzbq  $dst, $mem\t# ushort/char & 0xFF -> long" %}
  ins_encode %{
    __ movzbq($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Short/Char (16 bit UNsigned) with 32-bit mask into Long Register
instruct loadUS2L_immI(rRegL dst, memory mem, immI mask, rFlagsReg cr) %{
  match(Set dst (ConvI2L (AndI (LoadUS mem) mask)));
  effect(KILL cr);

  format %{ "movzwq  $dst, $mem\t# ushort/char & 32-bit mask -> long\n\t"
            "andl    $dst, right_n_bits($mask, 16)" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    __ movzwq(Rdst, $mem$$Address);
    __ andl(Rdst, $mask$$constant & right_n_bits(16));
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer
instruct loadI(rRegI dst, memory mem)
%{
  match(Set dst (LoadI mem));

  ins_cost(125);
  format %{ "movl    $dst, $mem\t# int" %}

  ins_encode %{
    __ movl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Integer (32 bit signed) to Byte (8 bit signed)
instruct loadI2B(rRegI dst, memory mem, immI_24 twentyfour) %{
  match(Set dst (RShiftI (LShiftI (LoadI mem) twentyfour) twentyfour));

  ins_cost(125);
  format %{ "movsbl  $dst, $mem\t# int -> byte" %}
  ins_encode %{
    __ movsbl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer (32 bit signed) to Unsigned Byte (8 bit UNsigned)
instruct loadI2UB(rRegI dst, memory mem, immI_255 mask) %{
  match(Set dst (AndI (LoadI mem) mask));

  ins_cost(125);
  format %{ "movzbl  $dst, $mem\t# int -> ubyte" %}
  ins_encode %{
    __ movzbl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer (32 bit signed) to Short (16 bit signed)
instruct loadI2S(rRegI dst, memory mem, immI_16 sixteen) %{
  match(Set dst (RShiftI (LShiftI (LoadI mem) sixteen) sixteen));

  ins_cost(125);
  format %{ "movswl  $dst, $mem\t# int -> short" %}
  ins_encode %{
    __ movswl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer (32 bit signed) to Unsigned Short/Char (16 bit UNsigned)
instruct loadI2US(rRegI dst, memory mem, immI_65535 mask) %{
  match(Set dst (AndI (LoadI mem) mask));

  ins_cost(125);
  format %{ "movzwl  $dst, $mem\t# int -> ushort/char" %}
  ins_encode %{
    __ movzwl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer into Long Register
instruct loadI2L(rRegL dst, memory mem)
%{
  match(Set dst (ConvI2L (LoadI mem)));

  ins_cost(125);
  format %{ "movslq  $dst, $mem\t# int -> long" %}

  ins_encode %{
    __ movslq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Integer with mask 0xFF into Long Register
instruct loadI2L_immI_255(rRegL dst, memory mem, immI_255 mask) %{
  match(Set dst (ConvI2L (AndI (LoadI mem) mask)));

  format %{ "movzbq  $dst, $mem\t# int & 0xFF -> long" %}
  ins_encode %{
    __ movzbq($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer with mask 0xFFFF into Long Register
instruct loadI2L_immI_65535(rRegL dst, memory mem, immI_65535 mask) %{
  match(Set dst (ConvI2L (AndI (LoadI mem) mask)));

  format %{ "movzwq  $dst, $mem\t# int & 0xFFFF -> long" %}
  ins_encode %{
    __ movzwq($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Integer with a 31-bit mask into Long Register
instruct loadI2L_immU31(rRegL dst, memory mem, immU31 mask, rFlagsReg cr) %{
  match(Set dst (ConvI2L (AndI (LoadI mem) mask)));
  effect(KILL cr);

  format %{ "movl    $dst, $mem\t# int & 31-bit mask -> long\n\t"
            "andl    $dst, $mask" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    __ movl(Rdst, $mem$$Address);
    __ andl(Rdst, $mask$$constant);
  %}
  ins_pipe(ialu_reg_mem);
%}

// Load Unsigned Integer into Long Register
instruct loadUI2L(rRegL dst, memory mem, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L (LoadI mem)) mask));

  ins_cost(125);
  format %{ "movl    $dst, $mem\t# uint -> long" %}

  ins_encode %{
    __ movl($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem);
%}

// Load Long
instruct loadL(rRegL dst, memory mem)
%{
  match(Set dst (LoadL mem));

  ins_cost(125);
  format %{ "movq    $dst, $mem\t# long" %}

  ins_encode %{
    __ movq($dst$$Register, $mem$$Address);
  %}

  ins_pipe(ialu_reg_mem); // XXX
%}

// Load Range
instruct loadRange(rRegI dst, memory mem)
%{
  match(Set dst (LoadRange mem));

  ins_cost(125); // XXX
  format %{ "movl    $dst, $mem\t# range" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_mem);
%}

// Load Pointer
instruct loadP(rRegP dst, memory mem)
%{
  match(Set dst (LoadP mem));
  predicate(n->as_Load()->barrier_data() == 0);

  ins_cost(125); // XXX
  format %{ "movq    $dst, $mem\t# ptr" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_mem); // XXX
%}

// Load Compressed Pointer
instruct loadN(rRegN dst, memory mem)
%{
   match(Set dst (LoadN mem));

   ins_cost(125); // XXX
   format %{ "movl    $dst, $mem\t# compressed ptr" %}
   ins_encode %{
     __ movl($dst$$Register, $mem$$Address);
   %}
   ins_pipe(ialu_reg_mem); // XXX
%}


// Load Klass Pointer
instruct loadKlass(rRegP dst, memory mem)
%{
  match(Set dst (LoadKlass mem));

  ins_cost(125); // XXX
  format %{ "movq    $dst, $mem\t# class" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_mem); // XXX
%}

// Load narrow Klass Pointer
instruct loadNKlass(rRegN dst, memory mem)
%{
  match(Set dst (LoadNKlass mem));

  ins_cost(125); // XXX
  format %{ "movl    $dst, $mem\t# compressed klass ptr" %}
  ins_encode %{
    __ movl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg_mem); // XXX
%}

// Load Float
instruct loadF(regF dst, memory mem)
%{
  match(Set dst (LoadF mem));

  ins_cost(145); // XXX
  format %{ "movss   $dst, $mem\t# float" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $mem$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Load Float
instruct MoveF2VL(vlRegF dst, regF src) %{
  match(Set dst src);
  format %{ "movss $dst,$src\t! load float (4 bytes)" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Float
instruct MoveF2LEG(legRegF dst, regF src) %{
  match(Set dst src);
  format %{ "movss $dst,$src\t# if src != dst load float (4 bytes)" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Float
instruct MoveVL2F(regF dst, vlRegF src) %{
  match(Set dst src);
  format %{ "movss $dst,$src\t! load float (4 bytes)" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Float
instruct MoveLEG2F(regF dst, legRegF src) %{
  match(Set dst src);
  format %{ "movss $dst,$src\t# if src != dst load float (4 bytes)" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Double
instruct loadD_partial(regD dst, memory mem)
%{
  predicate(!UseXmmLoadAndClearUpper);
  match(Set dst (LoadD mem));

  ins_cost(145); // XXX
  format %{ "movlpd  $dst, $mem\t# double" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $mem$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct loadD(regD dst, memory mem)
%{
  predicate(UseXmmLoadAndClearUpper);
  match(Set dst (LoadD mem));

  ins_cost(145); // XXX
  format %{ "movsd   $dst, $mem\t# double" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $mem$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Load Double
instruct MoveD2VL(vlRegD dst, regD src) %{
  match(Set dst src);
  format %{ "movsd $dst,$src\t! load double (8 bytes)" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Double
instruct MoveD2LEG(legRegD dst, regD src) %{
  match(Set dst src);
  format %{ "movsd $dst,$src\t# if src != dst load double (8 bytes)" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Double
instruct MoveVL2D(regD dst, vlRegD src) %{
  match(Set dst src);
  format %{ "movsd $dst,$src\t! load double (8 bytes)" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Load Double
instruct MoveLEG2D(regD dst, legRegD src) %{
  match(Set dst src);
  format %{ "movsd $dst,$src\t# if src != dst load double (8 bytes)" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe( fpu_reg_reg );
%}

// Following pseudo code describes the algorithm for max[FD]:
// Min algorithm is on similar lines
//  btmp = (b < +0.0) ? a : b
//  atmp = (b < +0.0) ? b : a
//  Tmp  = Max_Float(atmp , btmp)
//  Res  = (atmp == NaN) ? atmp : Tmp

// max = java.lang.Math.max(float a, float b)
instruct maxF_reg(legRegF dst, legRegF a, legRegF b, legRegF tmp, legRegF atmp, legRegF btmp) %{
  predicate(UseAVX > 0 && !n->is_reduction());
  match(Set dst (MaxF a b));
  effect(USE a, USE b, TEMP tmp, TEMP atmp, TEMP btmp);
  format %{
     "blendvps         $btmp,$b,$a,$b           \n\t"
     "blendvps         $atmp,$a,$b,$b           \n\t"
     "vmaxss           $tmp,$atmp,$btmp         \n\t"
     "cmpps.unordered  $btmp,$atmp,$atmp        \n\t"
     "blendvps         $dst,$tmp,$atmp,$btmp    \n\t"
  %}
  ins_encode %{
    int vector_len = Assembler::AVX_128bit;
    __ blendvps($btmp$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, vector_len);
    __ blendvps($atmp$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $b$$XMMRegister, vector_len);
    __ vmaxss($tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister);
    __ cmpps($btmp$$XMMRegister, $atmp$$XMMRegister, $atmp$$XMMRegister, Assembler::_false, vector_len);
    __ blendvps($dst$$XMMRegister, $tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister, vector_len);
 %}
  ins_pipe( pipe_slow );
%}

instruct maxF_reduction_reg(legRegF dst, legRegF a, legRegF b, legRegF xmmt, rRegI tmp, rFlagsReg cr) %{
  predicate(UseAVX > 0 && n->is_reduction());
  match(Set dst (MaxF a b));
  effect(USE a, USE b, TEMP xmmt, TEMP tmp, KILL cr);

  format %{ "$dst = max($a, $b)\t# intrinsic (float)" %}
  ins_encode %{
    emit_fp_min_max(_masm, $dst$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $xmmt$$XMMRegister, $tmp$$Register,
                    false /*min*/, true /*single*/);
  %}
  ins_pipe( pipe_slow );
%}

// max = java.lang.Math.max(double a, double b)
instruct maxD_reg(legRegD dst, legRegD a, legRegD b, legRegD tmp, legRegD atmp, legRegD btmp) %{
  predicate(UseAVX > 0 && !n->is_reduction());
  match(Set dst (MaxD a b));
  effect(USE a, USE b, TEMP atmp, TEMP btmp, TEMP tmp);
  format %{
     "blendvpd         $btmp,$b,$a,$b            \n\t"
     "blendvpd         $atmp,$a,$b,$b            \n\t"
     "vmaxsd           $tmp,$atmp,$btmp          \n\t"
     "cmppd.unordered  $btmp,$atmp,$atmp         \n\t"
     "blendvpd         $dst,$tmp,$atmp,$btmp     \n\t"
  %}
  ins_encode %{
    int vector_len = Assembler::AVX_128bit;
    __ blendvpd($btmp$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, vector_len);
    __ blendvpd($atmp$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $b$$XMMRegister, vector_len);
    __ vmaxsd($tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister);
    __ cmppd($btmp$$XMMRegister, $atmp$$XMMRegister, $atmp$$XMMRegister, Assembler::_false, vector_len);
    __ blendvpd($dst$$XMMRegister, $tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister, vector_len);
  %}
  ins_pipe( pipe_slow );
%}

instruct maxD_reduction_reg(legRegD dst, legRegD a, legRegD b, legRegD xmmt, rRegL tmp, rFlagsReg cr) %{
  predicate(UseAVX > 0 && n->is_reduction());
  match(Set dst (MaxD a b));
  effect(USE a, USE b, TEMP xmmt, TEMP tmp, KILL cr);

  format %{ "$dst = max($a, $b)\t# intrinsic (double)" %}
  ins_encode %{
    emit_fp_min_max(_masm, $dst$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $xmmt$$XMMRegister, $tmp$$Register,
                    false /*min*/, false /*single*/);
  %}
  ins_pipe( pipe_slow );
%}

// min = java.lang.Math.min(float a, float b)
instruct minF_reg(legRegF dst, legRegF a, legRegF b, legRegF tmp, legRegF atmp, legRegF btmp) %{
  predicate(UseAVX > 0 && !n->is_reduction());
  match(Set dst (MinF a b));
  effect(USE a, USE b, TEMP tmp, TEMP atmp, TEMP btmp);
  format %{
     "blendvps         $atmp,$a,$b,$a             \n\t"
     "blendvps         $btmp,$b,$a,$a             \n\t"
     "vminss           $tmp,$atmp,$btmp           \n\t"
     "cmpps.unordered  $btmp,$atmp,$atmp          \n\t"
     "blendvps         $dst,$tmp,$atmp,$btmp      \n\t"
  %}
  ins_encode %{
    int vector_len = Assembler::AVX_128bit;
    __ blendvps($atmp$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, vector_len);
    __ blendvps($btmp$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, $a$$XMMRegister, vector_len);
    __ vminss($tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister);
    __ cmpps($btmp$$XMMRegister, $atmp$$XMMRegister, $atmp$$XMMRegister, Assembler::_false, vector_len);
    __ blendvps($dst$$XMMRegister, $tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister, vector_len);
  %}
  ins_pipe( pipe_slow );
%}

instruct minF_reduction_reg(legRegF dst, legRegF a, legRegF b, legRegF xmmt, rRegI tmp, rFlagsReg cr) %{
  predicate(UseAVX > 0 && n->is_reduction());
  match(Set dst (MinF a b));
  effect(USE a, USE b, TEMP xmmt, TEMP tmp, KILL cr);

  format %{ "$dst = min($a, $b)\t# intrinsic (float)" %}
  ins_encode %{
    emit_fp_min_max(_masm, $dst$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $xmmt$$XMMRegister, $tmp$$Register,
                    true /*min*/, true /*single*/);
  %}
  ins_pipe( pipe_slow );
%}

// min = java.lang.Math.min(double a, double b)
instruct minD_reg(legRegD dst, legRegD a, legRegD b, legRegD tmp, legRegD atmp, legRegD btmp) %{
  predicate(UseAVX > 0 && !n->is_reduction());
  match(Set dst (MinD a b));
  effect(USE a, USE b, TEMP tmp, TEMP atmp, TEMP btmp);
  format %{
     "blendvpd         $atmp,$a,$b,$a           \n\t"
     "blendvpd         $btmp,$b,$a,$a           \n\t"
     "vminsd           $tmp,$atmp,$btmp         \n\t"
     "cmppd.unordered  $btmp,$atmp,$atmp        \n\t"
     "blendvpd         $dst,$tmp,$atmp,$btmp    \n\t"
  %}
  ins_encode %{
    int vector_len = Assembler::AVX_128bit;
    __ blendvpd($atmp$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, vector_len);
    __ blendvpd($btmp$$XMMRegister, $b$$XMMRegister, $a$$XMMRegister, $a$$XMMRegister, vector_len);
    __ vminsd($tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister);
    __ cmppd($btmp$$XMMRegister, $atmp$$XMMRegister, $atmp$$XMMRegister, Assembler::_false, vector_len);
    __ blendvpd($dst$$XMMRegister, $tmp$$XMMRegister, $atmp$$XMMRegister, $btmp$$XMMRegister, vector_len);
  %}
  ins_pipe( pipe_slow );
%}

instruct minD_reduction_reg(legRegD dst, legRegD a, legRegD b, legRegD xmmt, rRegL tmp, rFlagsReg cr) %{
  predicate(UseAVX > 0 && n->is_reduction());
  match(Set dst (MinD a b));
  effect(USE a, USE b, TEMP xmmt, TEMP tmp, KILL cr);

  format %{ "$dst = min($a, $b)\t# intrinsic (double)" %}
  ins_encode %{
    emit_fp_min_max(_masm, $dst$$XMMRegister, $a$$XMMRegister, $b$$XMMRegister, $xmmt$$XMMRegister, $tmp$$Register,
                    true /*min*/, false /*single*/);
  %}
  ins_pipe( pipe_slow );
%}

// Load Effective Address
instruct leaP8(rRegP dst, indOffset8 mem)
%{
  match(Set dst mem);

  ins_cost(110); // XXX
  format %{ "leaq    $dst, $mem\t# ptr 8" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaP32(rRegP dst, indOffset32 mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr 32" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

// instruct leaPIdx(rRegP dst, indIndex mem)
// %{
//   match(Set dst mem);

//   ins_cost(110);
//   format %{ "leaq    $dst, $mem\t# ptr idx" %}
//   opcode(0x8D);
//   ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
//   ins_pipe(ialu_reg_reg_fat);
// %}

instruct leaPIdxOff(rRegP dst, indIndexOffset mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxoff" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPIdxScale(rRegP dst, indIndexScale mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxscale" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPPosIdxScale(rRegP dst, indPosIndexScale mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxscale" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPIdxScaleOff(rRegP dst, indIndexScaleOffset mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxscaleoff" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPPosIdxOff(rRegP dst, indPosIndexOffset mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr posidxoff" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPPosIdxScaleOff(rRegP dst, indPosIndexScaleOffset mem)
%{
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr posidxscaleoff" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

// Load Effective Address which uses Narrow (32-bits) oop
instruct leaPCompressedOopOffset(rRegP dst, indCompressedOopOffset mem)
%{
  predicate(UseCompressedOops && (CompressedOops::shift() != 0));
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr compressedoopoff32" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaP8Narrow(rRegP dst, indOffset8Narrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110); // XXX
  format %{ "leaq    $dst, $mem\t# ptr off8narrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaP32Narrow(rRegP dst, indOffset32Narrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr off32narrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPIdxOffNarrow(rRegP dst, indIndexOffsetNarrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxoffnarrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPIdxScaleNarrow(rRegP dst, indIndexScaleNarrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxscalenarrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPIdxScaleOffNarrow(rRegP dst, indIndexScaleOffsetNarrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr idxscaleoffnarrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPPosIdxOffNarrow(rRegP dst, indPosIndexOffsetNarrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr posidxoffnarrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct leaPPosIdxScaleOffNarrow(rRegP dst, indPosIndexScaleOffsetNarrow mem)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst mem);

  ins_cost(110);
  format %{ "leaq    $dst, $mem\t# ptr posidxscaleoffnarrow" %}
  opcode(0x8D);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_reg_fat);
%}

instruct loadConI(rRegI dst, immI src)
%{
  match(Set dst src);

  format %{ "movl    $dst, $src\t# int" %}
  ins_encode(load_immI(dst, src));
  ins_pipe(ialu_reg_fat); // XXX
%}

instruct loadConI0(rRegI dst, immI0 src, rFlagsReg cr)
%{
  match(Set dst src);
  effect(KILL cr);

  ins_cost(50);
  format %{ "xorl    $dst, $dst\t# int" %}
  opcode(0x33); /* + rd */
  ins_encode(REX_reg_reg(dst, dst), OpcP, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

instruct loadConL(rRegL dst, immL src)
%{
  match(Set dst src);

  ins_cost(150);
  format %{ "movq    $dst, $src\t# long" %}
  ins_encode(load_immL(dst, src));
  ins_pipe(ialu_reg);
%}

instruct loadConL0(rRegL dst, immL0 src, rFlagsReg cr)
%{
  match(Set dst src);
  effect(KILL cr);

  ins_cost(50);
  format %{ "xorl    $dst, $dst\t# long" %}
  opcode(0x33); /* + rd */
  ins_encode(REX_reg_reg(dst, dst), OpcP, reg_reg(dst, dst));
  ins_pipe(ialu_reg); // XXX
%}

instruct loadConUL32(rRegL dst, immUL32 src)
%{
  match(Set dst src);

  ins_cost(60);
  format %{ "movl    $dst, $src\t# long (unsigned 32-bit)" %}
  ins_encode(load_immUL32(dst, src));
  ins_pipe(ialu_reg);
%}

instruct loadConL32(rRegL dst, immL32 src)
%{
  match(Set dst src);

  ins_cost(70);
  format %{ "movq    $dst, $src\t# long (32-bit)" %}
  ins_encode(load_immL32(dst, src));
  ins_pipe(ialu_reg);
%}

instruct loadConP(rRegP dst, immP con) %{
  match(Set dst con);

  format %{ "movq    $dst, $con\t# ptr" %}
  ins_encode(load_immP(dst, con));
  ins_pipe(ialu_reg_fat); // XXX
%}

instruct loadConP0(rRegP dst, immP0 src, rFlagsReg cr)
%{
  match(Set dst src);
  effect(KILL cr);

  ins_cost(50);
  format %{ "xorl    $dst, $dst\t# ptr" %}
  opcode(0x33); /* + rd */
  ins_encode(REX_reg_reg(dst, dst), OpcP, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

instruct loadConP31(rRegP dst, immP31 src, rFlagsReg cr)
%{
  match(Set dst src);
  effect(KILL cr);

  ins_cost(60);
  format %{ "movl    $dst, $src\t# ptr (positive 32-bit)" %}
  ins_encode(load_immP31(dst, src));
  ins_pipe(ialu_reg);
%}

instruct loadConF(regF dst, immF con) %{
  match(Set dst con);
  ins_cost(125);
  format %{ "movss   $dst, [$constantaddress]\t# load from constant table: float=$con" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, $constantaddress($con));
  %}
  ins_pipe(pipe_slow);
%}

instruct loadConN0(rRegN dst, immN0 src, rFlagsReg cr) %{
  match(Set dst src);
  effect(KILL cr);
  format %{ "xorq    $dst, $src\t# compressed NULL ptr" %}
  ins_encode %{
    __ xorq($dst$$Register, $dst$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct loadConN(rRegN dst, immN src) %{
  match(Set dst src);

  ins_cost(125);
  format %{ "movl    $dst, $src\t# compressed ptr" %}
  ins_encode %{
    address con = (address)$src$$constant;
    if (con == NULL) {
      ShouldNotReachHere();
    } else {
      __ set_narrow_oop($dst$$Register, (jobject)$src$$constant);
    }
  %}
  ins_pipe(ialu_reg_fat); // XXX
%}

instruct loadConNKlass(rRegN dst, immNKlass src) %{
  match(Set dst src);

  ins_cost(125);
  format %{ "movl    $dst, $src\t# compressed klass ptr" %}
  ins_encode %{
    address con = (address)$src$$constant;
    if (con == NULL) {
      ShouldNotReachHere();
    } else {
      __ set_narrow_klass($dst$$Register, (Klass*)$src$$constant);
    }
  %}
  ins_pipe(ialu_reg_fat); // XXX
%}

instruct loadConF0(regF dst, immF0 src)
%{
  match(Set dst src);
  ins_cost(100);

  format %{ "xorps   $dst, $dst\t# float 0.0" %}
  ins_encode %{
    __ xorps($dst$$XMMRegister, $dst$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

// Use the same format since predicate() can not be used here.
instruct loadConD(regD dst, immD con) %{
  match(Set dst con);
  ins_cost(125);
  format %{ "movsd   $dst, [$constantaddress]\t# load from constant table: double=$con" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, $constantaddress($con));
  %}
  ins_pipe(pipe_slow);
%}

instruct loadConD0(regD dst, immD0 src)
%{
  match(Set dst src);
  ins_cost(100);

  format %{ "xorpd   $dst, $dst\t# double 0.0" %}
  ins_encode %{
    __ xorpd ($dst$$XMMRegister, $dst$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

instruct loadSSI(rRegI dst, stackSlotI src)
%{
  match(Set dst src);

  ins_cost(125);
  format %{ "movl    $dst, $src\t# int stk" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct loadSSL(rRegL dst, stackSlotL src)
%{
  match(Set dst src);

  ins_cost(125);
  format %{ "movq    $dst, $src\t# long stk" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct loadSSP(rRegP dst, stackSlotP src)
%{
  match(Set dst src);

  ins_cost(125);
  format %{ "movq    $dst, $src\t# ptr stk" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct loadSSF(regF dst, stackSlotF src)
%{
  match(Set dst src);

  ins_cost(125);
  format %{ "movss   $dst, $src\t# float stk" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, Address(rsp, $src$$disp));
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Use the same format since predicate() can not be used here.
instruct loadSSD(regD dst, stackSlotD src)
%{
  match(Set dst src);

  ins_cost(125);
  format %{ "movsd   $dst, $src\t# double stk" %}
  ins_encode  %{
    __ movdbl($dst$$XMMRegister, Address(rsp, $src$$disp));
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Prefetch instructions for allocation.
// Must be safe to execute with invalid address (cannot fault).

instruct prefetchAlloc( memory mem ) %{
  predicate(AllocatePrefetchInstr==3);
  match(PrefetchAllocation mem);
  ins_cost(125);

  format %{ "PREFETCHW $mem\t# Prefetch allocation into level 1 cache and mark modified" %}
  ins_encode %{
    __ prefetchw($mem$$Address);
  %}
  ins_pipe(ialu_mem);
%}

instruct prefetchAllocNTA( memory mem ) %{
  predicate(AllocatePrefetchInstr==0);
  match(PrefetchAllocation mem);
  ins_cost(125);

  format %{ "PREFETCHNTA $mem\t# Prefetch allocation to non-temporal cache for write" %}
  ins_encode %{
    __ prefetchnta($mem$$Address);
  %}
  ins_pipe(ialu_mem);
%}

instruct prefetchAllocT0( memory mem ) %{
  predicate(AllocatePrefetchInstr==1);
  match(PrefetchAllocation mem);
  ins_cost(125);

  format %{ "PREFETCHT0 $mem\t# Prefetch allocation to level 1 and 2 caches for write" %}
  ins_encode %{
    __ prefetcht0($mem$$Address);
  %}
  ins_pipe(ialu_mem);
%}

instruct prefetchAllocT2( memory mem ) %{
  predicate(AllocatePrefetchInstr==2);
  match(PrefetchAllocation mem);
  ins_cost(125);

  format %{ "PREFETCHT2 $mem\t# Prefetch allocation to level 2 cache for write" %}
  ins_encode %{
    __ prefetcht2($mem$$Address);
  %}
  ins_pipe(ialu_mem);
%}

//----------Store Instructions-------------------------------------------------

// Store Byte
instruct storeB(memory mem, rRegI src)
%{
  match(Set mem (StoreB mem src));

  ins_cost(125); // XXX
  format %{ "movb    $mem, $src\t# byte" %}
  opcode(0x88);
  ins_encode(REX_breg_mem(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_mem_reg);
%}

// Store Char/Short
instruct storeC(memory mem, rRegI src)
%{
  match(Set mem (StoreC mem src));

  ins_cost(125); // XXX
  format %{ "movw    $mem, $src\t# char/short" %}
  opcode(0x89);
  ins_encode(SizePrefix, REX_reg_mem(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_mem_reg);
%}

// Store Integer
instruct storeI(memory mem, rRegI src)
%{
  match(Set mem (StoreI mem src));

  ins_cost(125); // XXX
  format %{ "movl    $mem, $src\t# int" %}
  opcode(0x89);
  ins_encode(REX_reg_mem(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_mem_reg);
%}

// Store Long
instruct storeL(memory mem, rRegL src)
%{
  match(Set mem (StoreL mem src));

  ins_cost(125); // XXX
  format %{ "movq    $mem, $src\t# long" %}
  opcode(0x89);
  ins_encode(REX_reg_mem_wide(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_mem_reg); // XXX
%}

// Store Pointer
instruct storeP(memory mem, any_RegP src)
%{
  match(Set mem (StoreP mem src));

  ins_cost(125); // XXX
  format %{ "movq    $mem, $src\t# ptr" %}
  opcode(0x89);
  ins_encode(REX_reg_mem_wide(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmP0(memory mem, immP0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreP mem zero));

  ins_cost(125); // XXX
  format %{ "movq    $mem, R12\t# ptr (R12_heapbase==0)" %}
  ins_encode %{
    __ movq($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

// Store NULL Pointer, mark word, or other simple pointer constant.
instruct storeImmP(memory mem, immP31 src)
%{
  match(Set mem (StoreP mem src));

  ins_cost(150); // XXX
  format %{ "movq    $mem, $src\t# ptr" %}
  opcode(0xC7); /* C7 /0 */
  ins_encode(REX_mem_wide(mem), OpcP, RM_opc_mem(0x00, mem), Con32(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Compressed Pointer
instruct storeN(memory mem, rRegN src)
%{
  match(Set mem (StoreN mem src));

  ins_cost(125); // XXX
  format %{ "movl    $mem, $src\t# compressed ptr" %}
  ins_encode %{
    __ movl($mem$$Address, $src$$Register);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeNKlass(memory mem, rRegN src)
%{
  match(Set mem (StoreNKlass mem src));

  ins_cost(125); // XXX
  format %{ "movl    $mem, $src\t# compressed klass ptr" %}
  ins_encode %{
    __ movl($mem$$Address, $src$$Register);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmN0(memory mem, immN0 zero)
%{
  predicate(CompressedOops::base() == NULL && CompressedKlassPointers::base() == NULL);
  match(Set mem (StoreN mem zero));

  ins_cost(125); // XXX
  format %{ "movl    $mem, R12\t# compressed ptr (R12_heapbase==0)" %}
  ins_encode %{
    __ movl($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmN(memory mem, immN src)
%{
  match(Set mem (StoreN mem src));

  ins_cost(150); // XXX
  format %{ "movl    $mem, $src\t# compressed ptr" %}
  ins_encode %{
    address con = (address)$src$$constant;
    if (con == NULL) {
      __ movl($mem$$Address, (int32_t)0);
    } else {
      __ set_narrow_oop($mem$$Address, (jobject)$src$$constant);
    }
  %}
  ins_pipe(ialu_mem_imm);
%}

instruct storeImmNKlass(memory mem, immNKlass src)
%{
  match(Set mem (StoreNKlass mem src));

  ins_cost(150); // XXX
  format %{ "movl    $mem, $src\t# compressed klass ptr" %}
  ins_encode %{
    __ set_narrow_klass($mem$$Address, (Klass*)$src$$constant);
  %}
  ins_pipe(ialu_mem_imm);
%}

// Store Integer Immediate
instruct storeImmI0(memory mem, immI0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreI mem zero));

  ins_cost(125); // XXX
  format %{ "movl    $mem, R12\t# int (R12_heapbase==0)" %}
  ins_encode %{
    __ movl($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmI(memory mem, immI src)
%{
  match(Set mem (StoreI mem src));

  ins_cost(150);
  format %{ "movl    $mem, $src\t# int" %}
  opcode(0xC7); /* C7 /0 */
  ins_encode(REX_mem(mem), OpcP, RM_opc_mem(0x00, mem), Con32(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Long Immediate
instruct storeImmL0(memory mem, immL0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreL mem zero));

  ins_cost(125); // XXX
  format %{ "movq    $mem, R12\t# long (R12_heapbase==0)" %}
  ins_encode %{
    __ movq($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmL(memory mem, immL32 src)
%{
  match(Set mem (StoreL mem src));

  ins_cost(150);
  format %{ "movq    $mem, $src\t# long" %}
  opcode(0xC7); /* C7 /0 */
  ins_encode(REX_mem_wide(mem), OpcP, RM_opc_mem(0x00, mem), Con32(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Short/Char Immediate
instruct storeImmC0(memory mem, immI0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreC mem zero));

  ins_cost(125); // XXX
  format %{ "movw    $mem, R12\t# short/char (R12_heapbase==0)" %}
  ins_encode %{
    __ movw($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmI16(memory mem, immI16 src)
%{
  predicate(UseStoreImmI16);
  match(Set mem (StoreC mem src));

  ins_cost(150);
  format %{ "movw    $mem, $src\t# short/char" %}
  opcode(0xC7); /* C7 /0 Same as 32 store immediate with prefix */
  ins_encode(SizePrefix, REX_mem(mem), OpcP, RM_opc_mem(0x00, mem),Con16(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Byte Immediate
instruct storeImmB0(memory mem, immI0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreB mem zero));

  ins_cost(125); // XXX
  format %{ "movb    $mem, R12\t# short/char (R12_heapbase==0)" %}
  ins_encode %{
    __ movb($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmB(memory mem, immI8 src)
%{
  match(Set mem (StoreB mem src));

  ins_cost(150); // XXX
  format %{ "movb    $mem, $src\t# byte" %}
  opcode(0xC6); /* C6 /0 */
  ins_encode(REX_mem(mem), OpcP, RM_opc_mem(0x00, mem), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Store CMS card-mark Immediate
instruct storeImmCM0_reg(memory mem, immI0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreCM mem zero));

  ins_cost(125); // XXX
  format %{ "movb    $mem, R12\t# CMS card-mark byte 0 (R12_heapbase==0)" %}
  ins_encode %{
    __ movb($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeImmCM0(memory mem, immI0 src)
%{
  match(Set mem (StoreCM mem src));

  ins_cost(150); // XXX
  format %{ "movb    $mem, $src\t# CMS card-mark byte 0" %}
  opcode(0xC6); /* C6 /0 */
  ins_encode(REX_mem(mem), OpcP, RM_opc_mem(0x00, mem), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Float
instruct storeF(memory mem, regF src)
%{
  match(Set mem (StoreF mem src));

  ins_cost(95); // XXX
  format %{ "movss   $mem, $src\t# float" %}
  ins_encode %{
    __ movflt($mem$$Address, $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Store immediate Float value (it is faster than store from XMM register)
instruct storeF0(memory mem, immF0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreF mem zero));

  ins_cost(25); // XXX
  format %{ "movl    $mem, R12\t# float 0. (R12_heapbase==0)" %}
  ins_encode %{
    __ movl($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeF_imm(memory mem, immF src)
%{
  match(Set mem (StoreF mem src));

  ins_cost(50);
  format %{ "movl    $mem, $src\t# float" %}
  opcode(0xC7); /* C7 /0 */
  ins_encode(REX_mem(mem), OpcP, RM_opc_mem(0x00, mem), Con32F_as_bits(src));
  ins_pipe(ialu_mem_imm);
%}

// Store Double
instruct storeD(memory mem, regD src)
%{
  match(Set mem (StoreD mem src));

  ins_cost(95); // XXX
  format %{ "movsd   $mem, $src\t# double" %}
  ins_encode %{
    __ movdbl($mem$$Address, $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

// Store immediate double 0.0 (it is faster than store from XMM register)
instruct storeD0_imm(memory mem, immD0 src)
%{
  predicate(!UseCompressedOops || (CompressedOops::base() != NULL));
  match(Set mem (StoreD mem src));

  ins_cost(50);
  format %{ "movq    $mem, $src\t# double 0." %}
  opcode(0xC7); /* C7 /0 */
  ins_encode(REX_mem_wide(mem), OpcP, RM_opc_mem(0x00, mem), Con32F_as_bits(src));
  ins_pipe(ialu_mem_imm);
%}

instruct storeD0(memory mem, immD0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) && (CompressedKlassPointers::base() == NULL));
  match(Set mem (StoreD mem zero));

  ins_cost(25); // XXX
  format %{ "movq    $mem, R12\t# double 0. (R12_heapbase==0)" %}
  ins_encode %{
    __ movq($mem$$Address, r12);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct storeSSI(stackSlotI dst, rRegI src)
%{
  match(Set dst src);

  ins_cost(100);
  format %{ "movl    $dst, $src\t# int stk" %}
  opcode(0x89);
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe( ialu_mem_reg );
%}

instruct storeSSL(stackSlotL dst, rRegL src)
%{
  match(Set dst src);

  ins_cost(100);
  format %{ "movq    $dst, $src\t# long stk" %}
  opcode(0x89);
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct storeSSP(stackSlotP dst, rRegP src)
%{
  match(Set dst src);

  ins_cost(100);
  format %{ "movq    $dst, $src\t# ptr stk" %}
  opcode(0x89);
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct storeSSF(stackSlotF dst, regF src)
%{
  match(Set dst src);

  ins_cost(95); // XXX
  format %{ "movss   $dst, $src\t# float stk" %}
  ins_encode %{
    __ movflt(Address(rsp, $dst$$disp), $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct storeSSD(stackSlotD dst, regD src)
%{
  match(Set dst src);

  ins_cost(95); // XXX
  format %{ "movsd   $dst, $src\t# double stk" %}
  ins_encode %{
    __ movdbl(Address(rsp, $dst$$disp), $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

//----------BSWAP Instructions-------------------------------------------------
instruct bytes_reverse_int(rRegI dst) %{
  match(Set dst (ReverseBytesI dst));

  format %{ "bswapl  $dst" %}
  opcode(0x0F, 0xC8);  /*Opcode 0F /C8 */
  ins_encode( REX_reg(dst), OpcP, opc2_reg(dst) );
  ins_pipe( ialu_reg );
%}

instruct bytes_reverse_long(rRegL dst) %{
  match(Set dst (ReverseBytesL dst));

  format %{ "bswapq  $dst" %}
  opcode(0x0F, 0xC8); /* Opcode 0F /C8 */
  ins_encode( REX_reg_wide(dst), OpcP, opc2_reg(dst) );
  ins_pipe( ialu_reg);
%}

instruct bytes_reverse_unsigned_short(rRegI dst, rFlagsReg cr) %{
  match(Set dst (ReverseBytesUS dst));
  effect(KILL cr);

  format %{ "bswapl  $dst\n\t"
            "shrl    $dst,16\n\t" %}
  ins_encode %{
    __ bswapl($dst$$Register);
    __ shrl($dst$$Register, 16);
  %}
  ins_pipe( ialu_reg );
%}

instruct bytes_reverse_short(rRegI dst, rFlagsReg cr) %{
  match(Set dst (ReverseBytesS dst));
  effect(KILL cr);

  format %{ "bswapl  $dst\n\t"
            "sar     $dst,16\n\t" %}
  ins_encode %{
    __ bswapl($dst$$Register);
    __ sarl($dst$$Register, 16);
  %}
  ins_pipe( ialu_reg );
%}

//---------- Zeros Count Instructions ------------------------------------------

instruct countLeadingZerosI(rRegI dst, rRegI src, rFlagsReg cr) %{
  predicate(UseCountLeadingZerosInstruction);
  match(Set dst (CountLeadingZerosI src));
  effect(KILL cr);

  format %{ "lzcntl  $dst, $src\t# count leading zeros (int)" %}
  ins_encode %{
    __ lzcntl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct countLeadingZerosI_bsr(rRegI dst, rRegI src, rFlagsReg cr) %{
  predicate(!UseCountLeadingZerosInstruction);
  match(Set dst (CountLeadingZerosI src));
  effect(KILL cr);

  format %{ "bsrl    $dst, $src\t# count leading zeros (int)\n\t"
            "jnz     skip\n\t"
            "movl    $dst, -1\n"
      "skip:\n\t"
            "negl    $dst\n\t"
            "addl    $dst, 31" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    Register Rsrc = $src$$Register;
    Label skip;
    __ bsrl(Rdst, Rsrc);
    __ jccb(Assembler::notZero, skip);
    __ movl(Rdst, -1);
    __ bind(skip);
    __ negl(Rdst);
    __ addl(Rdst, BitsPerInt - 1);
  %}
  ins_pipe(ialu_reg);
%}

instruct countLeadingZerosL(rRegI dst, rRegL src, rFlagsReg cr) %{
  predicate(UseCountLeadingZerosInstruction);
  match(Set dst (CountLeadingZerosL src));
  effect(KILL cr);

  format %{ "lzcntq  $dst, $src\t# count leading zeros (long)" %}
  ins_encode %{
    __ lzcntq($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct countLeadingZerosL_bsr(rRegI dst, rRegL src, rFlagsReg cr) %{
  predicate(!UseCountLeadingZerosInstruction);
  match(Set dst (CountLeadingZerosL src));
  effect(KILL cr);

  format %{ "bsrq    $dst, $src\t# count leading zeros (long)\n\t"
            "jnz     skip\n\t"
            "movl    $dst, -1\n"
      "skip:\n\t"
            "negl    $dst\n\t"
            "addl    $dst, 63" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    Register Rsrc = $src$$Register;
    Label skip;
    __ bsrq(Rdst, Rsrc);
    __ jccb(Assembler::notZero, skip);
    __ movl(Rdst, -1);
    __ bind(skip);
    __ negl(Rdst);
    __ addl(Rdst, BitsPerLong - 1);
  %}
  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosI(rRegI dst, rRegI src, rFlagsReg cr) %{
  predicate(UseCountTrailingZerosInstruction);
  match(Set dst (CountTrailingZerosI src));
  effect(KILL cr);

  format %{ "tzcntl    $dst, $src\t# count trailing zeros (int)" %}
  ins_encode %{
    __ tzcntl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosI_bsf(rRegI dst, rRegI src, rFlagsReg cr) %{
  predicate(!UseCountTrailingZerosInstruction);
  match(Set dst (CountTrailingZerosI src));
  effect(KILL cr);

  format %{ "bsfl    $dst, $src\t# count trailing zeros (int)\n\t"
            "jnz     done\n\t"
            "movl    $dst, 32\n"
      "done:" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    Label done;
    __ bsfl(Rdst, $src$$Register);
    __ jccb(Assembler::notZero, done);
    __ movl(Rdst, BitsPerInt);
    __ bind(done);
  %}
  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosL(rRegI dst, rRegL src, rFlagsReg cr) %{
  predicate(UseCountTrailingZerosInstruction);
  match(Set dst (CountTrailingZerosL src));
  effect(KILL cr);

  format %{ "tzcntq    $dst, $src\t# count trailing zeros (long)" %}
  ins_encode %{
    __ tzcntq($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct countTrailingZerosL_bsf(rRegI dst, rRegL src, rFlagsReg cr) %{
  predicate(!UseCountTrailingZerosInstruction);
  match(Set dst (CountTrailingZerosL src));
  effect(KILL cr);

  format %{ "bsfq    $dst, $src\t# count trailing zeros (long)\n\t"
            "jnz     done\n\t"
            "movl    $dst, 64\n"
      "done:" %}
  ins_encode %{
    Register Rdst = $dst$$Register;
    Label done;
    __ bsfq(Rdst, $src$$Register);
    __ jccb(Assembler::notZero, done);
    __ movl(Rdst, BitsPerLong);
    __ bind(done);
  %}
  ins_pipe(ialu_reg);
%}


//---------- Population Count Instructions -------------------------------------

instruct popCountI(rRegI dst, rRegI src, rFlagsReg cr) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountI src));
  effect(KILL cr);

  format %{ "popcnt  $dst, $src" %}
  ins_encode %{
    __ popcntl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct popCountI_mem(rRegI dst, memory mem, rFlagsReg cr) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountI (LoadI mem)));
  effect(KILL cr);

  format %{ "popcnt  $dst, $mem" %}
  ins_encode %{
    __ popcntl($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg);
%}

// Note: Long.bitCount(long) returns an int.
instruct popCountL(rRegI dst, rRegL src, rFlagsReg cr) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountL src));
  effect(KILL cr);

  format %{ "popcnt  $dst, $src" %}
  ins_encode %{
    __ popcntq($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

// Note: Long.bitCount(long) returns an int.
instruct popCountL_mem(rRegI dst, memory mem, rFlagsReg cr) %{
  predicate(UsePopCountInstruction);
  match(Set dst (PopCountL (LoadL mem)));
  effect(KILL cr);

  format %{ "popcnt  $dst, $mem" %}
  ins_encode %{
    __ popcntq($dst$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_reg);
%}


//----------MemBar Instructions-----------------------------------------------
// Memory barrier flavors

instruct membar_acquire()
%{
  match(MemBarAcquire);
  match(LoadFence);
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-acquire ! (empty encoding)" %}
  ins_encode();
  ins_pipe(empty);
%}

instruct membar_acquire_lock()
%{
  match(MemBarAcquireLock);
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-acquire (prior CMPXCHG in FastLock so empty encoding)" %}
  ins_encode();
  ins_pipe(empty);
%}

instruct membar_release()
%{
  match(MemBarRelease);
  match(StoreFence);
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-release ! (empty encoding)" %}
  ins_encode();
  ins_pipe(empty);
%}

instruct membar_release_lock()
%{
  match(MemBarReleaseLock);
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-release (a FastUnlock follows so empty encoding)" %}
  ins_encode();
  ins_pipe(empty);
%}

instruct membar_volatile(rFlagsReg cr) %{
  match(MemBarVolatile);
  effect(KILL cr);
  ins_cost(400);

  format %{
    $$template
    $$emit$$"lock addl [rsp + #0], 0\t! membar_volatile"
  %}
  ins_encode %{
    __ membar(Assembler::StoreLoad);
  %}
  ins_pipe(pipe_slow);
%}

instruct unnecessary_membar_volatile()
%{
  match(MemBarVolatile);
  predicate(Matcher::post_store_load_barrier(n));
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-volatile (unnecessary so empty encoding)" %}
  ins_encode();
  ins_pipe(empty);
%}

instruct membar_storestore() %{
  match(MemBarStoreStore);
  ins_cost(0);

  size(0);
  format %{ "MEMBAR-storestore (empty encoding)" %}
  ins_encode( );
  ins_pipe(empty);
%}

//----------Move Instructions--------------------------------------------------

instruct castX2P(rRegP dst, rRegL src)
%{
  match(Set dst (CastX2P src));

  format %{ "movq    $dst, $src\t# long->ptr" %}
  ins_encode %{
    if ($dst$$reg != $src$$reg) {
      __ movptr($dst$$Register, $src$$Register);
    }
  %}
  ins_pipe(ialu_reg_reg); // XXX
%}

instruct castP2X(rRegL dst, rRegP src)
%{
  match(Set dst (CastP2X src));

  format %{ "movq    $dst, $src\t# ptr -> long" %}
  ins_encode %{
    if ($dst$$reg != $src$$reg) {
      __ movptr($dst$$Register, $src$$Register);
    }
  %}
  ins_pipe(ialu_reg_reg); // XXX
%}

// Convert oop into int for vectors alignment masking
instruct convP2I(rRegI dst, rRegP src)
%{
  match(Set dst (ConvL2I (CastP2X src)));

  format %{ "movl    $dst, $src\t# ptr -> int" %}
  ins_encode %{
    __ movl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_reg); // XXX
%}

// Convert compressed oop into int for vectors alignment masking
// in case of 32bit oops (heap < 4Gb).
instruct convN2I(rRegI dst, rRegN src)
%{
  predicate(CompressedOops::shift() == 0);
  match(Set dst (ConvL2I (CastP2X (DecodeN src))));

  format %{ "movl    $dst, $src\t# compressed ptr -> int" %}
  ins_encode %{
    __ movl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_reg); // XXX
%}

// Convert oop pointer into compressed form
instruct encodeHeapOop(rRegN dst, rRegP src, rFlagsReg cr) %{
  predicate(n->bottom_type()->make_ptr()->ptr() != TypePtr::NotNull);
  match(Set dst (EncodeP src));
  effect(KILL cr);
  format %{ "encode_heap_oop $dst,$src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    if (s != d) {
      __ movq(d, s);
    }
    __ encode_heap_oop(d);
  %}
  ins_pipe(ialu_reg_long);
%}

instruct encodeHeapOop_not_null(rRegN dst, rRegP src, rFlagsReg cr) %{
  predicate(n->bottom_type()->make_ptr()->ptr() == TypePtr::NotNull);
  match(Set dst (EncodeP src));
  effect(KILL cr);
  format %{ "encode_heap_oop_not_null $dst,$src" %}
  ins_encode %{
    __ encode_heap_oop_not_null($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_long);
%}

instruct decodeHeapOop(rRegP dst, rRegN src, rFlagsReg cr) %{
  predicate(n->bottom_type()->is_ptr()->ptr() != TypePtr::NotNull &&
            n->bottom_type()->is_ptr()->ptr() != TypePtr::Constant);
  match(Set dst (DecodeN src));
  effect(KILL cr);
  format %{ "decode_heap_oop $dst,$src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    if (s != d) {
      __ movq(d, s);
    }
    __ decode_heap_oop(d);
  %}
  ins_pipe(ialu_reg_long);
%}

instruct decodeHeapOop_not_null(rRegP dst, rRegN src, rFlagsReg cr) %{
  predicate(n->bottom_type()->is_ptr()->ptr() == TypePtr::NotNull ||
            n->bottom_type()->is_ptr()->ptr() == TypePtr::Constant);
  match(Set dst (DecodeN src));
  effect(KILL cr);
  format %{ "decode_heap_oop_not_null $dst,$src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    if (s != d) {
      __ decode_heap_oop_not_null(d, s);
    } else {
      __ decode_heap_oop_not_null(d);
    }
  %}
  ins_pipe(ialu_reg_long);
%}

instruct encodeKlass_not_null(rRegN dst, rRegP src, rFlagsReg cr) %{
  match(Set dst (EncodePKlass src));
  effect(KILL cr);
  format %{ "encode_klass_not_null $dst,$src" %}
  ins_encode %{
    __ encode_klass_not_null($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_long);
%}

instruct decodeKlass_not_null(rRegP dst, rRegN src, rFlagsReg cr) %{
  match(Set dst (DecodeNKlass src));
  effect(KILL cr);
  format %{ "decode_klass_not_null $dst,$src" %}
  ins_encode %{
    Register s = $src$$Register;
    Register d = $dst$$Register;
    if (s != d) {
      __ decode_klass_not_null(d, s);
    } else {
      __ decode_klass_not_null(d);
    }
  %}
  ins_pipe(ialu_reg_long);
%}


//----------Conditional Move---------------------------------------------------
// Jump
// dummy instruction for generating temp registers
instruct jumpXtnd_offset(rRegL switch_val, immI2 shift, rRegI dest) %{
  match(Jump (LShiftL switch_val shift));
  ins_cost(350);
  predicate(false);
  effect(TEMP dest);

  format %{ "leaq    $dest, [$constantaddress]\n\t"
            "jmp     [$dest + $switch_val << $shift]\n\t" %}
  ins_encode %{
    // We could use jump(ArrayAddress) except that the macro assembler needs to use r10
    // to do that and the compiler is using that register as one it can allocate.
    // So we build it all by hand.
    // Address index(noreg, switch_reg, (Address::ScaleFactor)$shift$$constant);
    // ArrayAddress dispatch(table, index);
    Address dispatch($dest$$Register, $switch_val$$Register, (Address::ScaleFactor) $shift$$constant);
    __ lea($dest$$Register, $constantaddress);
    __ jmp(dispatch);
  %}
  ins_pipe(pipe_jmp);
%}

instruct jumpXtnd_addr(rRegL switch_val, immI2 shift, immL32 offset, rRegI dest) %{
  match(Jump (AddL (LShiftL switch_val shift) offset));
  ins_cost(350);
  effect(TEMP dest);

  format %{ "leaq    $dest, [$constantaddress]\n\t"
            "jmp     [$dest + $switch_val << $shift + $offset]\n\t" %}
  ins_encode %{
    // We could use jump(ArrayAddress) except that the macro assembler needs to use r10
    // to do that and the compiler is using that register as one it can allocate.
    // So we build it all by hand.
    // Address index(noreg, switch_reg, (Address::ScaleFactor) $shift$$constant, (int) $offset$$constant);
    // ArrayAddress dispatch(table, index);
    Address dispatch($dest$$Register, $switch_val$$Register, (Address::ScaleFactor) $shift$$constant, (int) $offset$$constant);
    __ lea($dest$$Register, $constantaddress);
    __ jmp(dispatch);
  %}
  ins_pipe(pipe_jmp);
%}

instruct jumpXtnd(rRegL switch_val, rRegI dest) %{
  match(Jump switch_val);
  ins_cost(350);
  effect(TEMP dest);

  format %{ "leaq    $dest, [$constantaddress]\n\t"
            "jmp     [$dest + $switch_val]\n\t" %}
  ins_encode %{
    // We could use jump(ArrayAddress) except that the macro assembler needs to use r10
    // to do that and the compiler is using that register as one it can allocate.
    // So we build it all by hand.
    // Address index(noreg, switch_reg, Address::times_1);
    // ArrayAddress dispatch(table, index);
    Address dispatch($dest$$Register, $switch_val$$Register, Address::times_1);
    __ lea($dest$$Register, $constantaddress);
    __ jmp(dispatch);
  %}
  ins_pipe(pipe_jmp);
%}

// Conditional move
instruct cmovI_reg(rRegI dst, rRegI src, rFlagsReg cr, cmpOp cop)
%{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovl$cop $dst, $src\t# signed, int" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}

instruct cmovI_regU(cmpOpU cop, rFlagsRegU cr, rRegI dst, rRegI src) %{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovl$cop $dst, $src\t# unsigned, int" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}

instruct cmovI_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegI dst, rRegI src) %{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovI_regU(cop, cr, dst, src);
  %}
%}

// Conditional move
instruct cmovI_mem(cmpOp cop, rFlagsReg cr, rRegI dst, memory src) %{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst (LoadI src))));

  ins_cost(250); // XXX
  format %{ "cmovl$cop $dst, $src\t# signed, int" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_mem(dst, src), enc_cmov(cop), reg_mem(dst, src));
  ins_pipe(pipe_cmov_mem);
%}

// Conditional move
instruct cmovI_memU(cmpOpU cop, rFlagsRegU cr, rRegI dst, memory src)
%{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst (LoadI src))));

  ins_cost(250); // XXX
  format %{ "cmovl$cop $dst, $src\t# unsigned, int" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_mem(dst, src), enc_cmov(cop), reg_mem(dst, src));
  ins_pipe(pipe_cmov_mem);
%}

instruct cmovI_memUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegI dst, memory src) %{
  match(Set dst (CMoveI (Binary cop cr) (Binary dst (LoadI src))));
  ins_cost(250);
  expand %{
    cmovI_memU(cop, cr, dst, src);
  %}
%}

// Conditional move
instruct cmovN_reg(rRegN dst, rRegN src, rFlagsReg cr, cmpOp cop)
%{
  match(Set dst (CMoveN (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovl$cop $dst, $src\t# signed, compressed ptr" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}

// Conditional move
instruct cmovN_regU(cmpOpU cop, rFlagsRegU cr, rRegN dst, rRegN src)
%{
  match(Set dst (CMoveN (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovl$cop $dst, $src\t# unsigned, compressed ptr" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}

instruct cmovN_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegN dst, rRegN src) %{
  match(Set dst (CMoveN (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovN_regU(cop, cr, dst, src);
  %}
%}

// Conditional move
instruct cmovP_reg(rRegP dst, rRegP src, rFlagsReg cr, cmpOp cop)
%{
  match(Set dst (CMoveP (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# signed, ptr" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg_wide(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);  // XXX
%}

// Conditional move
instruct cmovP_regU(cmpOpU cop, rFlagsRegU cr, rRegP dst, rRegP src)
%{
  match(Set dst (CMoveP (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# unsigned, ptr" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg_wide(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg); // XXX
%}

instruct cmovP_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegP dst, rRegP src) %{
  match(Set dst (CMoveP (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovP_regU(cop, cr, dst, src);
  %}
%}

// DISABLED: Requires the ADLC to emit a bottom_type call that
// correctly meets the two pointer arguments; one is an incoming
// register but the other is a memory operand.  ALSO appears to
// be buggy with implicit null checks.
//
//// Conditional move
//instruct cmovP_mem(cmpOp cop, rFlagsReg cr, rRegP dst, memory src)
//%{
//  match(Set dst (CMoveP (Binary cop cr) (Binary dst (LoadP src))));
//  ins_cost(250);
//  format %{ "CMOV$cop $dst,$src\t# ptr" %}
//  opcode(0x0F,0x40);
//  ins_encode( enc_cmov(cop), reg_mem( dst, src ) );
//  ins_pipe( pipe_cmov_mem );
//%}
//
//// Conditional move
//instruct cmovP_memU(cmpOpU cop, rFlagsRegU cr, rRegP dst, memory src)
//%{
//  match(Set dst (CMoveP (Binary cop cr) (Binary dst (LoadP src))));
//  ins_cost(250);
//  format %{ "CMOV$cop $dst,$src\t# ptr" %}
//  opcode(0x0F,0x40);
//  ins_encode( enc_cmov(cop), reg_mem( dst, src ) );
//  ins_pipe( pipe_cmov_mem );
//%}

instruct cmovL_reg(cmpOp cop, rFlagsReg cr, rRegL dst, rRegL src)
%{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# signed, long" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg_wide(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);  // XXX
%}

instruct cmovL_mem(cmpOp cop, rFlagsReg cr, rRegL dst, memory src)
%{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst (LoadL src))));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# signed, long" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_mem_wide(dst, src), enc_cmov(cop), reg_mem(dst, src));
  ins_pipe(pipe_cmov_mem);  // XXX
%}

instruct cmovL_regU(cmpOpU cop, rFlagsRegU cr, rRegL dst, rRegL src)
%{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# unsigned, long" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_reg_wide(dst, src), enc_cmov(cop), reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg); // XXX
%}

instruct cmovL_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegL dst, rRegL src) %{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovL_regU(cop, cr, dst, src);
  %}
%}

instruct cmovL_memU(cmpOpU cop, rFlagsRegU cr, rRegL dst, memory src)
%{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst (LoadL src))));

  ins_cost(200); // XXX
  format %{ "cmovq$cop $dst, $src\t# unsigned, long" %}
  opcode(0x0F, 0x40);
  ins_encode(REX_reg_mem_wide(dst, src), enc_cmov(cop), reg_mem(dst, src));
  ins_pipe(pipe_cmov_mem); // XXX
%}

instruct cmovL_memUCF(cmpOpUCF cop, rFlagsRegUCF cr, rRegL dst, memory src) %{
  match(Set dst (CMoveL (Binary cop cr) (Binary dst (LoadL src))));
  ins_cost(200);
  expand %{
    cmovL_memU(cop, cr, dst, src);
  %}
%}

instruct cmovF_reg(cmpOp cop, rFlagsReg cr, regF dst, regF src)
%{
  match(Set dst (CMoveF (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "jn$cop    skip\t# signed cmove float\n\t"
            "movss     $dst, $src\n"
    "skip:" %}
  ins_encode %{
    Label Lskip;
    // Invert sense of branch from sense of CMOV
    __ jccb((Assembler::Condition)($cop$$cmpcode^1), Lskip);
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
    __ bind(Lskip);
  %}
  ins_pipe(pipe_slow);
%}

// instruct cmovF_mem(cmpOp cop, rFlagsReg cr, regF dst, memory src)
// %{
//   match(Set dst (CMoveF (Binary cop cr) (Binary dst (LoadL src))));

//   ins_cost(200); // XXX
//   format %{ "jn$cop    skip\t# signed cmove float\n\t"
//             "movss     $dst, $src\n"
//     "skip:" %}
//   ins_encode(enc_cmovf_mem_branch(cop, dst, src));
//   ins_pipe(pipe_slow);
// %}

instruct cmovF_regU(cmpOpU cop, rFlagsRegU cr, regF dst, regF src)
%{
  match(Set dst (CMoveF (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "jn$cop    skip\t# unsigned cmove float\n\t"
            "movss     $dst, $src\n"
    "skip:" %}
  ins_encode %{
    Label Lskip;
    // Invert sense of branch from sense of CMOV
    __ jccb((Assembler::Condition)($cop$$cmpcode^1), Lskip);
    __ movflt($dst$$XMMRegister, $src$$XMMRegister);
    __ bind(Lskip);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmovF_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, regF dst, regF src) %{
  match(Set dst (CMoveF (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovF_regU(cop, cr, dst, src);
  %}
%}

instruct cmovD_reg(cmpOp cop, rFlagsReg cr, regD dst, regD src)
%{
  match(Set dst (CMoveD (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "jn$cop    skip\t# signed cmove double\n\t"
            "movsd     $dst, $src\n"
    "skip:" %}
  ins_encode %{
    Label Lskip;
    // Invert sense of branch from sense of CMOV
    __ jccb((Assembler::Condition)($cop$$cmpcode^1), Lskip);
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
    __ bind(Lskip);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmovD_regU(cmpOpU cop, rFlagsRegU cr, regD dst, regD src)
%{
  match(Set dst (CMoveD (Binary cop cr) (Binary dst src)));

  ins_cost(200); // XXX
  format %{ "jn$cop    skip\t# unsigned cmove double\n\t"
            "movsd     $dst, $src\n"
    "skip:" %}
  ins_encode %{
    Label Lskip;
    // Invert sense of branch from sense of CMOV
    __ jccb((Assembler::Condition)($cop$$cmpcode^1), Lskip);
    __ movdbl($dst$$XMMRegister, $src$$XMMRegister);
    __ bind(Lskip);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmovD_regUCF(cmpOpUCF cop, rFlagsRegUCF cr, regD dst, regD src) %{
  match(Set dst (CMoveD (Binary cop cr) (Binary dst src)));
  ins_cost(200);
  expand %{
    cmovD_regU(cop, cr, dst, src);
  %}
%}

//----------Arithmetic Instructions--------------------------------------------
//----------Addition Instructions----------------------------------------------

instruct addI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (AddI dst src));
  effect(KILL cr);

  format %{ "addl    $dst, $src\t# int" %}
  opcode(0x03);
  ins_encode(REX_reg_reg(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct addI_rReg_imm(rRegI dst, immI src, rFlagsReg cr)
%{
  match(Set dst (AddI dst src));
  effect(KILL cr);

  format %{ "addl    $dst, $src\t# int" %}
  opcode(0x81, 0x00); /* /0 id */
  ins_encode(OpcSErm(dst, src), Con8or32(src));
  ins_pipe( ialu_reg );
%}

instruct addI_rReg_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (AddI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "addl    $dst, $src\t# int" %}
  opcode(0x03);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct addI_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (AddI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(150); // XXX
  format %{ "addl    $dst, $src\t# int" %}
  opcode(0x01); /* Opcode 01 /r */
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct addI_mem_imm(memory dst, immI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (AddI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "addl    $dst, $src\t# int" %}
  opcode(0x81); /* Opcode 81 /0 id */
  ins_encode(REX_mem(dst), OpcSE(src), RM_opc_mem(0x00, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

instruct incI_rReg(rRegI dst, immI1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (AddI dst src));
  effect(KILL cr);

  format %{ "incl    $dst\t# int" %}
  opcode(0xFF, 0x00); // FF /0
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct incI_mem(memory dst, immI1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (StoreI dst (AddI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "incl    $dst\t# int" %}
  opcode(0xFF); /* Opcode FF /0 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(0x00, dst));
  ins_pipe(ialu_mem_imm);
%}

// XXX why does that use AddI
instruct decI_rReg(rRegI dst, immI_M1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (AddI dst src));
  effect(KILL cr);

  format %{ "decl    $dst\t# int" %}
  opcode(0xFF, 0x01); // FF /1
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// XXX why does that use AddI
instruct decI_mem(memory dst, immI_M1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (StoreI dst (AddI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "decl    $dst\t# int" %}
  opcode(0xFF); /* Opcode FF /1 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(0x01, dst));
  ins_pipe(ialu_mem_imm);
%}

instruct leaI_rReg_immI(rRegI dst, rRegI src0, immI src1)
%{
  match(Set dst (AddI src0 src1));

  ins_cost(110);
  format %{ "addr32 leal $dst, [$src0 + $src1]\t# int" %}
  opcode(0x8D); /* 0x8D /r */
  ins_encode(Opcode(0x67), REX_reg_reg(dst, src0), OpcP, reg_lea(dst, src0, src1)); // XXX
  ins_pipe(ialu_reg_reg);
%}

instruct addL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (AddL dst src));
  effect(KILL cr);

  format %{ "addq    $dst, $src\t# long" %}
  opcode(0x03);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct addL_rReg_imm(rRegL dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (AddL dst src));
  effect(KILL cr);

  format %{ "addq    $dst, $src\t# long" %}
  opcode(0x81, 0x00); /* /0 id */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe( ialu_reg );
%}

instruct addL_rReg_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (AddL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "addq    $dst, $src\t# long" %}
  opcode(0x03);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct addL_mem_rReg(memory dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (AddL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(150); // XXX
  format %{ "addq    $dst, $src\t# long" %}
  opcode(0x01); /* Opcode 01 /r */
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct addL_mem_imm(memory dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (AddL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "addq    $dst, $src\t# long" %}
  opcode(0x81); /* Opcode 81 /0 id */
  ins_encode(REX_mem_wide(dst),
             OpcSE(src), RM_opc_mem(0x00, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

instruct incL_rReg(rRegI dst, immL1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (AddL dst src));
  effect(KILL cr);

  format %{ "incq    $dst\t# long" %}
  opcode(0xFF, 0x00); // FF /0
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct incL_mem(memory dst, immL1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (StoreL dst (AddL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "incq    $dst\t# long" %}
  opcode(0xFF); /* Opcode FF /0 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(0x00, dst));
  ins_pipe(ialu_mem_imm);
%}

// XXX why does that use AddL
instruct decL_rReg(rRegL dst, immL_M1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (AddL dst src));
  effect(KILL cr);

  format %{ "decq    $dst\t# long" %}
  opcode(0xFF, 0x01); // FF /1
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// XXX why does that use AddL
instruct decL_mem(memory dst, immL_M1 src, rFlagsReg cr)
%{
  predicate(UseIncDec);
  match(Set dst (StoreL dst (AddL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "decq    $dst\t# long" %}
  opcode(0xFF); /* Opcode FF /1 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(0x01, dst));
  ins_pipe(ialu_mem_imm);
%}

instruct leaL_rReg_immL(rRegL dst, rRegL src0, immL32 src1)
%{
  match(Set dst (AddL src0 src1));

  ins_cost(110);
  format %{ "leaq    $dst, [$src0 + $src1]\t# long" %}
  opcode(0x8D); /* 0x8D /r */
  ins_encode(REX_reg_reg_wide(dst, src0), OpcP, reg_lea(dst, src0, src1)); // XXX
  ins_pipe(ialu_reg_reg);
%}

instruct addP_rReg(rRegP dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (AddP dst src));
  effect(KILL cr);

  format %{ "addq    $dst, $src\t# ptr" %}
  opcode(0x03);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct addP_rReg_imm(rRegP dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (AddP dst src));
  effect(KILL cr);

  format %{ "addq    $dst, $src\t# ptr" %}
  opcode(0x81, 0x00); /* /0 id */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe( ialu_reg );
%}

// XXX addP mem ops ????

instruct leaP_rReg_imm(rRegP dst, rRegP src0, immL32 src1)
%{
  match(Set dst (AddP src0 src1));

  ins_cost(110);
  format %{ "leaq    $dst, [$src0 + $src1]\t# ptr" %}
  opcode(0x8D); /* 0x8D /r */
  ins_encode(REX_reg_reg_wide(dst, src0), OpcP, reg_lea(dst, src0, src1));// XXX
  ins_pipe(ialu_reg_reg);
%}

instruct checkCastPP(rRegP dst)
%{
  match(Set dst (CheckCastPP dst));

  size(0);
  format %{ "# checkcastPP of $dst" %}
  ins_encode(/* empty encoding */);
  ins_pipe(empty);
%}

instruct castPP(rRegP dst)
%{
  match(Set dst (CastPP dst));

  size(0);
  format %{ "# castPP of $dst" %}
  ins_encode(/* empty encoding */);
  ins_pipe(empty);
%}

instruct castII(rRegI dst)
%{
  match(Set dst (CastII dst));

  size(0);
  format %{ "# castII of $dst" %}
  ins_encode(/* empty encoding */);
  ins_cost(0);
  ins_pipe(empty);
%}

// LoadP-locked same as a regular LoadP when used with compare-swap
instruct loadPLocked(rRegP dst, memory mem)
%{
  match(Set dst (LoadPLocked mem));

  ins_cost(125); // XXX
  format %{ "movq    $dst, $mem\t# ptr locked" %}
  opcode(0x8B);
  ins_encode(REX_reg_mem_wide(dst, mem), OpcP, reg_mem(dst, mem));
  ins_pipe(ialu_reg_mem); // XXX
%}

// Conditional-store of the updated heap-top.
// Used during allocation of the shared heap.
// Sets flags (EQ) on success.  Implemented with a CMPXCHG on Intel.

instruct storePConditional(memory heap_top_ptr,
                           rax_RegP oldval, rRegP newval,
                           rFlagsReg cr)
%{
  predicate(n->as_LoadStore()->barrier_data() == 0);
  match(Set cr (StorePConditional heap_top_ptr (Binary oldval newval)));

  format %{ "cmpxchgq $heap_top_ptr, $newval\t# (ptr) "
            "If rax == $heap_top_ptr then store $newval into $heap_top_ptr" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, heap_top_ptr),
             OpcP, OpcS,
             reg_mem(newval, heap_top_ptr));
  ins_pipe(pipe_cmpxchg);
%}

// Conditional-store of an int value.
// ZF flag is set on success, reset otherwise.  Implemented with a CMPXCHG.
instruct storeIConditional(memory mem, rax_RegI oldval, rRegI newval, rFlagsReg cr)
%{
  match(Set cr (StoreIConditional mem (Binary oldval newval)));
  effect(KILL oldval);

  format %{ "cmpxchgl $mem, $newval\t# If rax == $mem then store $newval into $mem" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem(newval, mem),
             OpcP, OpcS,
             reg_mem(newval, mem));
  ins_pipe(pipe_cmpxchg);
%}

// Conditional-store of a long value.
// ZF flag is set on success, reset otherwise.  Implemented with a CMPXCHG.
instruct storeLConditional(memory mem, rax_RegL oldval, rRegL newval, rFlagsReg cr)
%{
  match(Set cr (StoreLConditional mem (Binary oldval newval)));
  effect(KILL oldval);

  format %{ "cmpxchgq $mem, $newval\t# If rax == $mem then store $newval into $mem" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, mem),
             OpcP, OpcS,
             reg_mem(newval, mem));
  ins_pipe(pipe_cmpxchg);
%}


// XXX No flag versions for CompareAndSwap{P,I,L} because matcher can't match them
instruct compareAndSwapP(rRegI res,
                         memory mem_ptr,
                         rax_RegP oldval, rRegP newval,
                         rFlagsReg cr)
%{
  predicate(VM_Version::supports_cx8() && n->as_LoadStore()->barrier_data() == 0);
  match(Set res (CompareAndSwapP mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapP mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgq $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndSwapL(rRegI res,
                         memory mem_ptr,
                         rax_RegL oldval, rRegL newval,
                         rFlagsReg cr)
%{
  predicate(VM_Version::supports_cx8());
  match(Set res (CompareAndSwapL mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapL mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgq $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndSwapI(rRegI res,
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set res (CompareAndSwapI mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapI mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgl $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndSwapB(rRegI res,
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set res (CompareAndSwapB mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapB mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgb $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB0);
  ins_encode(lock_prefix,
             REX_breg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndSwapS(rRegI res,
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set res (CompareAndSwapS mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapS mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgw $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             SizePrefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndSwapN(rRegI res,
                          memory mem_ptr,
                          rax_RegN oldval, rRegN newval,
                          rFlagsReg cr) %{
  match(Set res (CompareAndSwapN mem_ptr (Binary oldval newval)));
  match(Set res (WeakCompareAndSwapN mem_ptr (Binary oldval newval)));
  effect(KILL cr, KILL oldval);

  format %{ "cmpxchgl $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"
            "sete    $res\n\t"
            "movzbl  $res, $res" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr),
             REX_breg(res), Opcode(0x0F), Opcode(0x94), reg(res), // sete
             REX_reg_breg(res, res), // movzbl
             Opcode(0xF), Opcode(0xB6), reg_reg(res, res));
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeB(
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set oldval (CompareAndExchangeB mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgb $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"  %}
  opcode(0x0F, 0xB0);
  ins_encode(lock_prefix,
             REX_breg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr) // lock cmpxchg
             );
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeS(
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set oldval (CompareAndExchangeS mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgw $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"  %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             SizePrefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr) // lock cmpxchg
             );
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeI(
                         memory mem_ptr,
                         rax_RegI oldval, rRegI newval,
                         rFlagsReg cr)
%{
  match(Set oldval (CompareAndExchangeI mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgl $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"  %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr) // lock cmpxchg
             );
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeL(
                         memory mem_ptr,
                         rax_RegL oldval, rRegL newval,
                         rFlagsReg cr)
%{
  predicate(VM_Version::supports_cx8());
  match(Set oldval (CompareAndExchangeL mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgq $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t"  %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr)  // lock cmpxchg
            );
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeN(
                          memory mem_ptr,
                          rax_RegN oldval, rRegN newval,
                          rFlagsReg cr) %{
  match(Set oldval (CompareAndExchangeN mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgl $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr)  // lock cmpxchg
          );
  ins_pipe( pipe_cmpxchg );
%}

instruct compareAndExchangeP(
                         memory mem_ptr,
                         rax_RegP oldval, rRegP newval,
                         rFlagsReg cr)
%{
  predicate(VM_Version::supports_cx8() && n->as_LoadStore()->barrier_data() == 0);
  match(Set oldval (CompareAndExchangeP mem_ptr (Binary oldval newval)));
  effect(KILL cr);

  format %{ "cmpxchgq $mem_ptr,$newval\t# "
            "If rax == $mem_ptr then store $newval into $mem_ptr\n\t" %}
  opcode(0x0F, 0xB1);
  ins_encode(lock_prefix,
             REX_reg_mem_wide(newval, mem_ptr),
             OpcP, OpcS,
             reg_mem(newval, mem_ptr)  // lock cmpxchg
          );
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddB_no_res( memory mem, Universe dummy, immI add, rFlagsReg cr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddB mem add));
  effect(KILL cr);
  format %{ "ADDB  [$mem],$add" %}
  ins_encode %{
    __ lock();
    __ addb($mem$$Address, $add$$constant);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddB( memory mem, rRegI newval, rFlagsReg cr) %{
  match(Set newval (GetAndAddB mem newval));
  effect(KILL cr);
  format %{ "XADDB  [$mem],$newval" %}
  ins_encode %{
    __ lock();
    __ xaddb($mem$$Address, $newval$$Register);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddS_no_res( memory mem, Universe dummy, immI add, rFlagsReg cr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddS mem add));
  effect(KILL cr);
  format %{ "ADDW  [$mem],$add" %}
  ins_encode %{
    __ lock();
    __ addw($mem$$Address, $add$$constant);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddS( memory mem, rRegI newval, rFlagsReg cr) %{
  match(Set newval (GetAndAddS mem newval));
  effect(KILL cr);
  format %{ "XADDW  [$mem],$newval" %}
  ins_encode %{
    __ lock();
    __ xaddw($mem$$Address, $newval$$Register);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddI_no_res( memory mem, Universe dummy, immI add, rFlagsReg cr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddI mem add));
  effect(KILL cr);
  format %{ "ADDL  [$mem],$add" %}
  ins_encode %{
    __ lock();
    __ addl($mem$$Address, $add$$constant);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddI( memory mem, rRegI newval, rFlagsReg cr) %{
  match(Set newval (GetAndAddI mem newval));
  effect(KILL cr);
  format %{ "XADDL  [$mem],$newval" %}
  ins_encode %{
    __ lock();
    __ xaddl($mem$$Address, $newval$$Register);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddL_no_res( memory mem, Universe dummy, immL32 add, rFlagsReg cr) %{
  predicate(n->as_LoadStore()->result_not_used());
  match(Set dummy (GetAndAddL mem add));
  effect(KILL cr);
  format %{ "ADDQ  [$mem],$add" %}
  ins_encode %{
    __ lock();
    __ addq($mem$$Address, $add$$constant);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xaddL( memory mem, rRegL newval, rFlagsReg cr) %{
  match(Set newval (GetAndAddL mem newval));
  effect(KILL cr);
  format %{ "XADDQ  [$mem],$newval" %}
  ins_encode %{
    __ lock();
    __ xaddq($mem$$Address, $newval$$Register);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgB( memory mem, rRegI newval) %{
  match(Set newval (GetAndSetB mem newval));
  format %{ "XCHGB  $newval,[$mem]" %}
  ins_encode %{
    __ xchgb($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgS( memory mem, rRegI newval) %{
  match(Set newval (GetAndSetS mem newval));
  format %{ "XCHGW  $newval,[$mem]" %}
  ins_encode %{
    __ xchgw($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgI( memory mem, rRegI newval) %{
  match(Set newval (GetAndSetI mem newval));
  format %{ "XCHGL  $newval,[$mem]" %}
  ins_encode %{
    __ xchgl($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgL( memory mem, rRegL newval) %{
  match(Set newval (GetAndSetL mem newval));
  format %{ "XCHGL  $newval,[$mem]" %}
  ins_encode %{
    __ xchgq($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgP( memory mem, rRegP newval) %{
  match(Set newval (GetAndSetP mem newval));
  predicate(n->as_LoadStore()->barrier_data() == 0);
  format %{ "XCHGQ  $newval,[$mem]" %}
  ins_encode %{
    __ xchgq($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

instruct xchgN( memory mem, rRegN newval) %{
  match(Set newval (GetAndSetN mem newval));
  format %{ "XCHGL  $newval,$mem]" %}
  ins_encode %{
    __ xchgl($newval$$Register, $mem$$Address);
  %}
  ins_pipe( pipe_cmpxchg );
%}

//----------Abs Instructions-------------------------------------------

// Integer Absolute Instructions
instruct absI_rReg(rRegI dst, rRegI src, rRegI tmp, rFlagsReg cr)
%{
  match(Set dst (AbsI src));
  effect(TEMP dst, TEMP tmp, KILL cr);
  format %{ "movl $tmp, $src\n\t"
            "sarl $tmp, 31\n\t"
            "movl $dst, $src\n\t"
            "xorl $dst, $tmp\n\t"
            "subl $dst, $tmp\n"
          %}
  ins_encode %{
    __ movl($tmp$$Register, $src$$Register);
    __ sarl($tmp$$Register, 31);
    __ movl($dst$$Register, $src$$Register);
    __ xorl($dst$$Register, $tmp$$Register);
    __ subl($dst$$Register, $tmp$$Register);
  %}

  ins_pipe(ialu_reg_reg);
%}

// Long Absolute Instructions
instruct absL_rReg(rRegL dst, rRegL src, rRegL tmp, rFlagsReg cr)
%{
  match(Set dst (AbsL src));
  effect(TEMP dst, TEMP tmp, KILL cr);
  format %{ "movq $tmp, $src\n\t"
            "sarq $tmp, 63\n\t"
            "movq $dst, $src\n\t"
            "xorq $dst, $tmp\n\t"
            "subq $dst, $tmp\n"
          %}
  ins_encode %{
    __ movq($tmp$$Register, $src$$Register);
    __ sarq($tmp$$Register, 63);
    __ movq($dst$$Register, $src$$Register);
    __ xorq($dst$$Register, $tmp$$Register);
    __ subq($dst$$Register, $tmp$$Register);
  %}

  ins_pipe(ialu_reg_reg);
%}

//----------Subtraction Instructions-------------------------------------------

// Integer Subtraction Instructions
instruct subI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (SubI dst src));
  effect(KILL cr);

  format %{ "subl    $dst, $src\t# int" %}
  opcode(0x2B);
  ins_encode(REX_reg_reg(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct subI_rReg_imm(rRegI dst, immI src, rFlagsReg cr)
%{
  match(Set dst (SubI dst src));
  effect(KILL cr);

  format %{ "subl    $dst, $src\t# int" %}
  opcode(0x81, 0x05);  /* Opcode 81 /5 */
  ins_encode(OpcSErm(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

instruct subI_rReg_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (SubI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "subl    $dst, $src\t# int" %}
  opcode(0x2B);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct subI_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (SubI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "subl    $dst, $src\t# int" %}
  opcode(0x29); /* Opcode 29 /r */
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct subI_mem_imm(memory dst, immI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (SubI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "subl    $dst, $src\t# int" %}
  opcode(0x81); /* Opcode 81 /5 id */
  ins_encode(REX_mem(dst), OpcSE(src), RM_opc_mem(0x05, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

instruct subL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (SubL dst src));
  effect(KILL cr);

  format %{ "subq    $dst, $src\t# long" %}
  opcode(0x2B);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct subL_rReg_imm(rRegI dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (SubL dst src));
  effect(KILL cr);

  format %{ "subq    $dst, $src\t# long" %}
  opcode(0x81, 0x05);  /* Opcode 81 /5 */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

instruct subL_rReg_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (SubL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "subq    $dst, $src\t# long" %}
  opcode(0x2B);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

instruct subL_mem_rReg(memory dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (SubL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "subq    $dst, $src\t# long" %}
  opcode(0x29); /* Opcode 29 /r */
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct subL_mem_imm(memory dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (SubL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125); // XXX
  format %{ "subq    $dst, $src\t# long" %}
  opcode(0x81); /* Opcode 81 /5 id */
  ins_encode(REX_mem_wide(dst),
             OpcSE(src), RM_opc_mem(0x05, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Subtract from a pointer
// XXX hmpf???
instruct subP_rReg(rRegP dst, rRegI src, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (AddP dst (SubI zero src)));
  effect(KILL cr);

  format %{ "subq    $dst, $src\t# ptr - int" %}
  opcode(0x2B);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

instruct negI_rReg(rRegI dst, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (SubI zero dst));
  effect(KILL cr);

  format %{ "negl    $dst\t# int" %}
  opcode(0xF7, 0x03);  // Opcode F7 /3
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct negI_mem(memory dst, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (SubI zero (LoadI dst))));
  effect(KILL cr);

  format %{ "negl    $dst\t# int" %}
  opcode(0xF7, 0x03);  // Opcode F7 /3
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_reg);
%}

instruct negL_rReg(rRegL dst, immL0 zero, rFlagsReg cr)
%{
  match(Set dst (SubL zero dst));
  effect(KILL cr);

  format %{ "negq    $dst\t# long" %}
  opcode(0xF7, 0x03);  // Opcode F7 /3
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct negL_mem(memory dst, immL0 zero, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (SubL zero (LoadL dst))));
  effect(KILL cr);

  format %{ "negq    $dst\t# long" %}
  opcode(0xF7, 0x03);  // Opcode F7 /3
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_reg);
%}

//----------Multiplication/Division Instructions-------------------------------
// Integer Multiplication Instructions
// Multiply Register

instruct mulI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (MulI dst src));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imull   $dst, $src\t# int" %}
  opcode(0x0F, 0xAF);
  ins_encode(REX_reg_reg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct mulI_rReg_imm(rRegI dst, rRegI src, immI imm, rFlagsReg cr)
%{
  match(Set dst (MulI src imm));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imull   $dst, $src, $imm\t# int" %}
  opcode(0x69); /* 69 /r id */
  ins_encode(REX_reg_reg(dst, src),
             OpcSE(imm), reg_reg(dst, src), Con8or32(imm));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct mulI_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (MulI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(350);
  format %{ "imull   $dst, $src\t# int" %}
  opcode(0x0F, 0xAF);
  ins_encode(REX_reg_mem(dst, src), OpcP, OpcS, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem_alu0);
%}

instruct mulI_mem_imm(rRegI dst, memory src, immI imm, rFlagsReg cr)
%{
  match(Set dst (MulI (LoadI src) imm));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imull   $dst, $src, $imm\t# int" %}
  opcode(0x69); /* 69 /r id */
  ins_encode(REX_reg_mem(dst, src),
             OpcSE(imm), reg_mem(dst, src), Con8or32(imm));
  ins_pipe(ialu_reg_mem_alu0);
%}

instruct mulAddS2I_rReg(rRegI dst, rRegI src1, rRegI src2, rRegI src3, rFlagsReg cr)
%{
  match(Set dst (MulAddS2I (Binary dst src1) (Binary src2 src3)));
  effect(KILL cr, KILL src2);

  expand %{ mulI_rReg(dst, src1, cr);
           mulI_rReg(src2, src3, cr);
           addI_rReg(dst, src2, cr); %}
%}

instruct mulL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (MulL dst src));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imulq   $dst, $src\t# long" %}
  opcode(0x0F, 0xAF);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct mulL_rReg_imm(rRegL dst, rRegL src, immL32 imm, rFlagsReg cr)
%{
  match(Set dst (MulL src imm));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imulq   $dst, $src, $imm\t# long" %}
  opcode(0x69); /* 69 /r id */
  ins_encode(REX_reg_reg_wide(dst, src),
             OpcSE(imm), reg_reg(dst, src), Con8or32(imm));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct mulL_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (MulL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(350);
  format %{ "imulq   $dst, $src\t# long" %}
  opcode(0x0F, 0xAF);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, OpcS, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem_alu0);
%}

instruct mulL_mem_imm(rRegL dst, memory src, immL32 imm, rFlagsReg cr)
%{
  match(Set dst (MulL (LoadL src) imm));
  effect(KILL cr);

  ins_cost(300);
  format %{ "imulq   $dst, $src, $imm\t# long" %}
  opcode(0x69); /* 69 /r id */
  ins_encode(REX_reg_mem_wide(dst, src),
             OpcSE(imm), reg_mem(dst, src), Con8or32(imm));
  ins_pipe(ialu_reg_mem_alu0);
%}

instruct mulHiL_rReg(rdx_RegL dst, no_rax_RegL src, rax_RegL rax, rFlagsReg cr)
%{
  match(Set dst (MulHiL src rax));
  effect(USE_KILL rax, KILL cr);

  ins_cost(300);
  format %{ "imulq   RDX:RAX, RAX, $src\t# mulhi" %}
  opcode(0xF7, 0x5); /* Opcode F7 /5 */
  ins_encode(REX_reg_wide(src), OpcP, reg_opc(src));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct divI_rReg(rax_RegI rax, rdx_RegI rdx, no_rax_rdx_RegI div,
                   rFlagsReg cr)
%{
  match(Set rax (DivI rax div));
  effect(KILL rdx, KILL cr);

  ins_cost(30*100+10*100); // XXX
  format %{ "cmpl    rax, 0x80000000\t# idiv\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpl    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdql\n\t"
            "idivl   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdql_enc(div), REX_reg(div), OpcP, reg_opc(div));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct divL_rReg(rax_RegL rax, rdx_RegL rdx, no_rax_rdx_RegL div,
                   rFlagsReg cr)
%{
  match(Set rax (DivL rax div));
  effect(KILL rdx, KILL cr);

  ins_cost(30*100+10*100); // XXX
  format %{ "movq    rdx, 0x8000000000000000\t# ldiv\n\t"
            "cmpq    rax, rdx\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpq    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdqq\n\t"
            "idivq   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdqq_enc(div), REX_reg_wide(div), OpcP, reg_opc(div));
  ins_pipe(ialu_reg_reg_alu0);
%}

// Integer DIVMOD with Register, both quotient and mod results
instruct divModI_rReg_divmod(rax_RegI rax, rdx_RegI rdx, no_rax_rdx_RegI div,
                             rFlagsReg cr)
%{
  match(DivModI rax div);
  effect(KILL cr);

  ins_cost(30*100+10*100); // XXX
  format %{ "cmpl    rax, 0x80000000\t# idiv\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpl    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdql\n\t"
            "idivl   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdql_enc(div), REX_reg(div), OpcP, reg_opc(div));
  ins_pipe(pipe_slow);
%}

// Long DIVMOD with Register, both quotient and mod results
instruct divModL_rReg_divmod(rax_RegL rax, rdx_RegL rdx, no_rax_rdx_RegL div,
                             rFlagsReg cr)
%{
  match(DivModL rax div);
  effect(KILL cr);

  ins_cost(30*100+10*100); // XXX
  format %{ "movq    rdx, 0x8000000000000000\t# ldiv\n\t"
            "cmpq    rax, rdx\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpq    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdqq\n\t"
            "idivq   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdqq_enc(div), REX_reg_wide(div), OpcP, reg_opc(div));
  ins_pipe(pipe_slow);
%}

//----------- DivL-By-Constant-Expansions--------------------------------------
// DivI cases are handled by the compiler

// Magic constant, reciprocal of 10
instruct loadConL_0x6666666666666667(rRegL dst)
%{
  effect(DEF dst);

  format %{ "movq    $dst, #0x666666666666667\t# Used in div-by-10" %}
  ins_encode(load_immL(dst, 0x6666666666666667));
  ins_pipe(ialu_reg);
%}

instruct mul_hi(rdx_RegL dst, no_rax_RegL src, rax_RegL rax, rFlagsReg cr)
%{
  effect(DEF dst, USE src, USE_KILL rax, KILL cr);

  format %{ "imulq   rdx:rax, rax, $src\t# Used in div-by-10" %}
  opcode(0xF7, 0x5); /* Opcode F7 /5 */
  ins_encode(REX_reg_wide(src), OpcP, reg_opc(src));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct sarL_rReg_63(rRegL dst, rFlagsReg cr)
%{
  effect(USE_DEF dst, KILL cr);

  format %{ "sarq    $dst, #63\t# Used in div-by-10" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(reg_opc_imm_wide(dst, 0x3F));
  ins_pipe(ialu_reg);
%}

instruct sarL_rReg_2(rRegL dst, rFlagsReg cr)
%{
  effect(USE_DEF dst, KILL cr);

  format %{ "sarq    $dst, #2\t# Used in div-by-10" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(reg_opc_imm_wide(dst, 0x2));
  ins_pipe(ialu_reg);
%}

instruct divL_10(rdx_RegL dst, no_rax_RegL src, immL10 div)
%{
  match(Set dst (DivL src div));

  ins_cost((5+8)*100);
  expand %{
    rax_RegL rax;                     // Killed temp
    rFlagsReg cr;                     // Killed
    loadConL_0x6666666666666667(rax); // movq  rax, 0x6666666666666667
    mul_hi(dst, src, rax, cr);        // mulq  rdx:rax <= rax * $src
    sarL_rReg_63(src, cr);            // sarq  src, 63
    sarL_rReg_2(dst, cr);             // sarq  rdx, 2
    subL_rReg(dst, src, cr);          // subl  rdx, src
  %}
%}

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

instruct modI_rReg(rdx_RegI rdx, rax_RegI rax, no_rax_rdx_RegI div,
                   rFlagsReg cr)
%{
  match(Set rdx (ModI rax div));
  effect(KILL rax, KILL cr);

  ins_cost(300); // XXX
  format %{ "cmpl    rax, 0x80000000\t# irem\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpl    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdql\n\t"
            "idivl   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdql_enc(div), REX_reg(div), OpcP, reg_opc(div));
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct modL_rReg(rdx_RegL rdx, rax_RegL rax, no_rax_rdx_RegL div,
                   rFlagsReg cr)
%{
  match(Set rdx (ModL rax div));
  effect(KILL rax, KILL cr);

  ins_cost(300); // XXX
  format %{ "movq    rdx, 0x8000000000000000\t# lrem\n\t"
            "cmpq    rax, rdx\n\t"
            "jne,s   normal\n\t"
            "xorl    rdx, rdx\n\t"
            "cmpq    $div, -1\n\t"
            "je,s    done\n"
    "normal: cdqq\n\t"
            "idivq   $div\n"
    "done:"        %}
  opcode(0xF7, 0x7);  /* Opcode F7 /7 */
  ins_encode(cdqq_enc(div), REX_reg_wide(div), OpcP, reg_opc(div));
  ins_pipe(ialu_reg_reg_alu0);
%}

// Integer Shift Instructions
// Shift Left by one
instruct salI_rReg_1(rRegI dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (LShiftI dst shift));
  effect(KILL cr);

  format %{ "sall    $dst, $shift" %}
  opcode(0xD1, 0x4); /* D1 /4 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// Shift Left by one
instruct salI_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (LShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sall    $dst, $shift\t" %}
  opcode(0xD1, 0x4); /* D1 /4 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Shift Left by 8-bit immediate
instruct salI_rReg_imm(rRegI dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (LShiftI dst shift));
  effect(KILL cr);

  format %{ "sall    $dst, $shift" %}
  opcode(0xC1, 0x4); /* C1 /4 ib */
  ins_encode(reg_opc_imm(dst, shift));
  ins_pipe(ialu_reg);
%}

// Shift Left by 8-bit immediate
instruct salI_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (LShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sall    $dst, $shift" %}
  opcode(0xC1, 0x4); /* C1 /4 ib */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Shift Left by variable
instruct salI_rReg_CL(rRegI dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (LShiftI dst shift));
  effect(KILL cr);

  format %{ "sall    $dst, $shift" %}
  opcode(0xD3, 0x4); /* D3 /4 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Shift Left by variable
instruct salI_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (LShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sall    $dst, $shift" %}
  opcode(0xD3, 0x4); /* D3 /4 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Arithmetic shift right by one
instruct sarI_rReg_1(rRegI dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (RShiftI dst shift));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xD1, 0x7); /* D1 /7 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// Arithmetic shift right by one
instruct sarI_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (RShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xD1, 0x7); /* D1 /7 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by 8-bit immediate
instruct sarI_rReg_imm(rRegI dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (RShiftI dst shift));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(reg_opc_imm(dst, shift));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by 8-bit immediate
instruct sarI_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (RShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by variable
instruct sarI_rReg_CL(rRegI dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (RShiftI dst shift));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xD3, 0x7); /* D3 /7 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Arithmetic Shift Right by variable
instruct sarI_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (RShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "sarl    $dst, $shift" %}
  opcode(0xD3, 0x7); /* D3 /7 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Logical shift right by one
instruct shrI_rReg_1(rRegI dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (URShiftI dst shift));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xD1, 0x5); /* D1 /5 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// Logical shift right by one
instruct shrI_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (URShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xD1, 0x5); /* D1 /5 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Logical Shift Right by 8-bit immediate
instruct shrI_rReg_imm(rRegI dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (URShiftI dst shift));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xC1, 0x5); /* C1 /5 ib */
  ins_encode(reg_opc_imm(dst, shift));
  ins_pipe(ialu_reg);
%}

// Logical Shift Right by 8-bit immediate
instruct shrI_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (URShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xC1, 0x5); /* C1 /5 ib */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Logical Shift Right by variable
instruct shrI_rReg_CL(rRegI dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (URShiftI dst shift));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xD3, 0x5); /* D3 /5 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Logical Shift Right by variable
instruct shrI_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (URShiftI (LoadI dst) shift)));
  effect(KILL cr);

  format %{ "shrl    $dst, $shift" %}
  opcode(0xD3, 0x5); /* D3 /5 */
  ins_encode(REX_mem(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Long Shift Instructions
// Shift Left by one
instruct salL_rReg_1(rRegL dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (LShiftL dst shift));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xD1, 0x4); /* D1 /4 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// Shift Left by one
instruct salL_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (LShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xD1, 0x4); /* D1 /4 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Shift Left by 8-bit immediate
instruct salL_rReg_imm(rRegL dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (LShiftL dst shift));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xC1, 0x4); /* C1 /4 ib */
  ins_encode(reg_opc_imm_wide(dst, shift));
  ins_pipe(ialu_reg);
%}

// Shift Left by 8-bit immediate
instruct salL_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (LShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xC1, 0x4); /* C1 /4 ib */
  ins_encode(REX_mem_wide(dst), OpcP,
             RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Shift Left by variable
instruct salL_rReg_CL(rRegL dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (LShiftL dst shift));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xD3, 0x4); /* D3 /4 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Shift Left by variable
instruct salL_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (LShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "salq    $dst, $shift" %}
  opcode(0xD3, 0x4); /* D3 /4 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Arithmetic shift right by one
instruct sarL_rReg_1(rRegL dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (RShiftL dst shift));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xD1, 0x7); /* D1 /7 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

// Arithmetic shift right by one
instruct sarL_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (RShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xD1, 0x7); /* D1 /7 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by 8-bit immediate
instruct sarL_rReg_imm(rRegL dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (RShiftL dst shift));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(reg_opc_imm_wide(dst, shift));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by 8-bit immediate
instruct sarL_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (RShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xC1, 0x7); /* C1 /7 ib */
  ins_encode(REX_mem_wide(dst), OpcP,
             RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Arithmetic Shift Right by variable
instruct sarL_rReg_CL(rRegL dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (RShiftL dst shift));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xD3, 0x7); /* D3 /7 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Arithmetic Shift Right by variable
instruct sarL_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (RShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "sarq    $dst, $shift" %}
  opcode(0xD3, 0x7); /* D3 /7 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Logical shift right by one
instruct shrL_rReg_1(rRegL dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (URShiftL dst shift));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xD1, 0x5); /* D1 /5 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst ));
  ins_pipe(ialu_reg);
%}

// Logical shift right by one
instruct shrL_mem_1(memory dst, immI1 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (URShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xD1, 0x5); /* D1 /5 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_imm);
%}

// Logical Shift Right by 8-bit immediate
instruct shrL_rReg_imm(rRegL dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (URShiftL dst shift));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xC1, 0x5); /* C1 /5 ib */
  ins_encode(reg_opc_imm_wide(dst, shift));
  ins_pipe(ialu_reg);
%}


// Logical Shift Right by 8-bit immediate
instruct shrL_mem_imm(memory dst, immI8 shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (URShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xC1, 0x5); /* C1 /5 ib */
  ins_encode(REX_mem_wide(dst), OpcP,
             RM_opc_mem(secondary, dst), Con8or32(shift));
  ins_pipe(ialu_mem_imm);
%}

// Logical Shift Right by variable
instruct shrL_rReg_CL(rRegL dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (URShiftL dst shift));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xD3, 0x5); /* D3 /5 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}

// Logical Shift Right by variable
instruct shrL_mem_CL(memory dst, rcx_RegI shift, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (URShiftL (LoadL dst) shift)));
  effect(KILL cr);

  format %{ "shrq    $dst, $shift" %}
  opcode(0xD3, 0x5); /* D3 /5 */
  ins_encode(REX_mem_wide(dst), OpcP, RM_opc_mem(secondary, dst));
  ins_pipe(ialu_mem_reg);
%}

// Logical Shift Right by 24, followed by Arithmetic Shift Left by 24.
// This idiom is used by the compiler for the i2b bytecode.
instruct i2b(rRegI dst, rRegI src, immI_24 twentyfour)
%{
  match(Set dst (RShiftI (LShiftI src twentyfour) twentyfour));

  format %{ "movsbl  $dst, $src\t# i2b" %}
  opcode(0x0F, 0xBE);
  ins_encode(REX_reg_breg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// Logical Shift Right by 16, followed by Arithmetic Shift Left by 16.
// This idiom is used by the compiler the i2s bytecode.
instruct i2s(rRegI dst, rRegI src, immI_16 sixteen)
%{
  match(Set dst (RShiftI (LShiftI src sixteen) sixteen));

  format %{ "movswl  $dst, $src\t# i2s" %}
  opcode(0x0F, 0xBF);
  ins_encode(REX_reg_reg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// ROL/ROR instructions

// ROL expand
instruct rolI_rReg_imm1(rRegI dst, rFlagsReg cr) %{
  effect(KILL cr, USE_DEF dst);

  format %{ "roll    $dst" %}
  opcode(0xD1, 0x0); /* Opcode  D1 /0 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct rolI_rReg_imm8(rRegI dst, immI8 shift, rFlagsReg cr) %{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "roll    $dst, $shift" %}
  opcode(0xC1, 0x0); /* Opcode C1 /0 ib */
  ins_encode( reg_opc_imm(dst, shift) );
  ins_pipe(ialu_reg);
%}

instruct rolI_rReg_CL(no_rcx_RegI dst, rcx_RegI shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "roll    $dst, $shift" %}
  opcode(0xD3, 0x0); /* Opcode D3 /0 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}
// end of ROL expand

// Rotate Left by one
instruct rolI_rReg_i1(rRegI dst, immI1 lshift, immI_M1 rshift, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI dst lshift) (URShiftI dst rshift)));

  expand %{
    rolI_rReg_imm1(dst, cr);
  %}
%}

// Rotate Left by 8-bit immediate
instruct rolI_rReg_i8(rRegI dst, immI8 lshift, immI8 rshift, rFlagsReg cr)
%{
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 0x1f));
  match(Set dst (OrI (LShiftI dst lshift) (URShiftI dst rshift)));

  expand %{
    rolI_rReg_imm8(dst, lshift, cr);
  %}
%}

// Rotate Left by variable
instruct rolI_rReg_Var_C0(no_rcx_RegI dst, rcx_RegI shift, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI dst shift) (URShiftI dst (SubI zero shift))));

  expand %{
    rolI_rReg_CL(dst, shift, cr);
  %}
%}

// Rotate Left by variable
instruct rolI_rReg_Var_C32(no_rcx_RegI dst, rcx_RegI shift, immI_32 c32, rFlagsReg cr)
%{
  match(Set dst (OrI (LShiftI dst shift) (URShiftI dst (SubI c32 shift))));

  expand %{
    rolI_rReg_CL(dst, shift, cr);
  %}
%}

// ROR expand
instruct rorI_rReg_imm1(rRegI dst, rFlagsReg cr)
%{
  effect(USE_DEF dst, KILL cr);

  format %{ "rorl    $dst" %}
  opcode(0xD1, 0x1); /* D1 /1 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct rorI_rReg_imm8(rRegI dst, immI8 shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rorl    $dst, $shift" %}
  opcode(0xC1, 0x1); /* C1 /1 ib */
  ins_encode(reg_opc_imm(dst, shift));
  ins_pipe(ialu_reg);
%}

instruct rorI_rReg_CL(no_rcx_RegI dst, rcx_RegI shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rorl    $dst, $shift" %}
  opcode(0xD3, 0x1); /* D3 /1 */
  ins_encode(REX_reg(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}
// end of ROR expand

// Rotate Right by one
instruct rorI_rReg_i1(rRegI dst, immI1 rshift, immI_M1 lshift, rFlagsReg cr)
%{
  match(Set dst (OrI (URShiftI dst rshift) (LShiftI dst lshift)));

  expand %{
    rorI_rReg_imm1(dst, cr);
  %}
%}

// Rotate Right by 8-bit immediate
instruct rorI_rReg_i8(rRegI dst, immI8 rshift, immI8 lshift, rFlagsReg cr)
%{
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 0x1f));
  match(Set dst (OrI (URShiftI dst rshift) (LShiftI dst lshift)));

  expand %{
    rorI_rReg_imm8(dst, rshift, cr);
  %}
%}

// Rotate Right by variable
instruct rorI_rReg_Var_C0(no_rcx_RegI dst, rcx_RegI shift, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (OrI (URShiftI dst shift) (LShiftI dst (SubI zero shift))));

  expand %{
    rorI_rReg_CL(dst, shift, cr);
  %}
%}

// Rotate Right by variable
instruct rorI_rReg_Var_C32(no_rcx_RegI dst, rcx_RegI shift, immI_32 c32, rFlagsReg cr)
%{
  match(Set dst (OrI (URShiftI dst shift) (LShiftI dst (SubI c32 shift))));

  expand %{
    rorI_rReg_CL(dst, shift, cr);
  %}
%}

// for long rotate
// ROL expand
instruct rolL_rReg_imm1(rRegL dst, rFlagsReg cr) %{
  effect(USE_DEF dst, KILL cr);

  format %{ "rolq    $dst" %}
  opcode(0xD1, 0x0); /* Opcode  D1 /0 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct rolL_rReg_imm8(rRegL dst, immI8 shift, rFlagsReg cr) %{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rolq    $dst, $shift" %}
  opcode(0xC1, 0x0); /* Opcode C1 /0 ib */
  ins_encode( reg_opc_imm_wide(dst, shift) );
  ins_pipe(ialu_reg);
%}

instruct rolL_rReg_CL(no_rcx_RegL dst, rcx_RegI shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rolq    $dst, $shift" %}
  opcode(0xD3, 0x0); /* Opcode D3 /0 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}
// end of ROL expand

// Rotate Left by one
instruct rolL_rReg_i1(rRegL dst, immI1 lshift, immI_M1 rshift, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL dst lshift) (URShiftL dst rshift)));

  expand %{
    rolL_rReg_imm1(dst, cr);
  %}
%}

// Rotate Left by 8-bit immediate
instruct rolL_rReg_i8(rRegL dst, immI8 lshift, immI8 rshift, rFlagsReg cr)
%{
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 0x3f));
  match(Set dst (OrL (LShiftL dst lshift) (URShiftL dst rshift)));

  expand %{
    rolL_rReg_imm8(dst, lshift, cr);
  %}
%}

// Rotate Left by variable
instruct rolL_rReg_Var_C0(no_rcx_RegL dst, rcx_RegI shift, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL dst shift) (URShiftL dst (SubI zero shift))));

  expand %{
    rolL_rReg_CL(dst, shift, cr);
  %}
%}

// Rotate Left by variable
instruct rolL_rReg_Var_C64(no_rcx_RegL dst, rcx_RegI shift, immI_64 c64, rFlagsReg cr)
%{
  match(Set dst (OrL (LShiftL dst shift) (URShiftL dst (SubI c64 shift))));

  expand %{
    rolL_rReg_CL(dst, shift, cr);
  %}
%}

// ROR expand
instruct rorL_rReg_imm1(rRegL dst, rFlagsReg cr)
%{
  effect(USE_DEF dst, KILL cr);

  format %{ "rorq    $dst" %}
  opcode(0xD1, 0x1); /* D1 /1 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg);
%}

instruct rorL_rReg_imm8(rRegL dst, immI8 shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rorq    $dst, $shift" %}
  opcode(0xC1, 0x1); /* C1 /1 ib */
  ins_encode(reg_opc_imm_wide(dst, shift));
  ins_pipe(ialu_reg);
%}

instruct rorL_rReg_CL(no_rcx_RegL dst, rcx_RegI shift, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE shift, KILL cr);

  format %{ "rorq    $dst, $shift" %}
  opcode(0xD3, 0x1); /* D3 /1 */
  ins_encode(REX_reg_wide(dst), OpcP, reg_opc(dst));
  ins_pipe(ialu_reg_reg);
%}
// end of ROR expand

// Rotate Right by one
instruct rorL_rReg_i1(rRegL dst, immI1 rshift, immI_M1 lshift, rFlagsReg cr)
%{
  match(Set dst (OrL (URShiftL dst rshift) (LShiftL dst lshift)));

  expand %{
    rorL_rReg_imm1(dst, cr);
  %}
%}

// Rotate Right by 8-bit immediate
instruct rorL_rReg_i8(rRegL dst, immI8 rshift, immI8 lshift, rFlagsReg cr)
%{
  predicate(0 == ((n->in(1)->in(2)->get_int() + n->in(2)->in(2)->get_int()) & 0x3f));
  match(Set dst (OrL (URShiftL dst rshift) (LShiftL dst lshift)));

  expand %{
    rorL_rReg_imm8(dst, rshift, cr);
  %}
%}

// Rotate Right by variable
instruct rorL_rReg_Var_C0(no_rcx_RegL dst, rcx_RegI shift, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (OrL (URShiftL dst shift) (LShiftL dst (SubI zero shift))));

  expand %{
    rorL_rReg_CL(dst, shift, cr);
  %}
%}

// Rotate Right by variable
instruct rorL_rReg_Var_C64(no_rcx_RegL dst, rcx_RegI shift, immI_64 c64, rFlagsReg cr)
%{
  match(Set dst (OrL (URShiftL dst shift) (LShiftL dst (SubI c64 shift))));

  expand %{
    rorL_rReg_CL(dst, shift, cr);
  %}
%}

// Logical Instructions

// Integer Logical Instructions

// And Instructions
// And Register with Register
instruct andI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (AndI dst src));
  effect(KILL cr);

  format %{ "andl    $dst, $src\t# int" %}
  opcode(0x23);
  ins_encode(REX_reg_reg(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// And Register with Immediate 255
instruct andI_rReg_imm255(rRegI dst, immI_255 src)
%{
  match(Set dst (AndI dst src));

  format %{ "movzbl  $dst, $dst\t# int & 0xFF" %}
  opcode(0x0F, 0xB6);
  ins_encode(REX_reg_breg(dst, dst), OpcP, OpcS, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate 255 and promote to long
instruct andI2L_rReg_imm255(rRegL dst, rRegI src, immI_255 mask)
%{
  match(Set dst (ConvI2L (AndI src mask)));

  format %{ "movzbl  $dst, $src\t# int & 0xFF -> long" %}
  opcode(0x0F, 0xB6);
  ins_encode(REX_reg_breg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate 65535
instruct andI_rReg_imm65535(rRegI dst, immI_65535 src)
%{
  match(Set dst (AndI dst src));

  format %{ "movzwl  $dst, $dst\t# int & 0xFFFF" %}
  opcode(0x0F, 0xB7);
  ins_encode(REX_reg_reg(dst, dst), OpcP, OpcS, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate 65535 and promote to long
instruct andI2L_rReg_imm65535(rRegL dst, rRegI src, immI_65535 mask)
%{
  match(Set dst (ConvI2L (AndI src mask)));

  format %{ "movzwl  $dst, $src\t# int & 0xFFFF -> long" %}
  opcode(0x0F, 0xB7);
  ins_encode(REX_reg_reg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate
instruct andI_rReg_imm(rRegI dst, immI src, rFlagsReg cr)
%{
  match(Set dst (AndI dst src));
  effect(KILL cr);

  format %{ "andl    $dst, $src\t# int" %}
  opcode(0x81, 0x04); /* Opcode 81 /4 */
  ins_encode(OpcSErm(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// And Register with Memory
instruct andI_rReg_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (AndI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "andl    $dst, $src\t# int" %}
  opcode(0x23);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// And Memory with Register
instruct andB_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreB dst (AndI (LoadB dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "andb    $dst, $src\t# byte" %}
  opcode(0x20);
  ins_encode(REX_breg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct andI_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (AndI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "andl    $dst, $src\t# int" %}
  opcode(0x21); /* Opcode 21 /r */
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// And Memory with Immediate
instruct andI_mem_imm(memory dst, immI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (AndI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "andl    $dst, $src\t# int" %}
  opcode(0x81, 0x4); /* Opcode 81 /4 id */
  ins_encode(REX_mem(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// BMI1 instructions
instruct andnI_rReg_rReg_mem(rRegI dst, rRegI src1, memory src2, immI_M1 minus_1, rFlagsReg cr) %{
  match(Set dst (AndI (XorI src1 minus_1) (LoadI src2)));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "andnl  $dst, $src1, $src2" %}

  ins_encode %{
    __ andnl($dst$$Register, $src1$$Register, $src2$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct andnI_rReg_rReg_rReg(rRegI dst, rRegI src1, rRegI src2, immI_M1 minus_1, rFlagsReg cr) %{
  match(Set dst (AndI (XorI src1 minus_1) src2));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "andnl  $dst, $src1, $src2" %}

  ins_encode %{
    __ andnl($dst$$Register, $src1$$Register, $src2$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct blsiI_rReg_rReg(rRegI dst, rRegI src, immI0 imm_zero, rFlagsReg cr) %{
  match(Set dst (AndI (SubI imm_zero src) src));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsil  $dst, $src" %}

  ins_encode %{
    __ blsil($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct blsiI_rReg_mem(rRegI dst, memory src, immI0 imm_zero, rFlagsReg cr) %{
  match(Set dst (AndI (SubI imm_zero (LoadI src) ) (LoadI src) ));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsil  $dst, $src" %}

  ins_encode %{
    __ blsil($dst$$Register, $src$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct blsmskI_rReg_mem(rRegI dst, memory src, immI_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (XorI (AddI (LoadI src) minus_1) (LoadI src) ) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsmskl $dst, $src" %}

  ins_encode %{
    __ blsmskl($dst$$Register, $src$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct blsmskI_rReg_rReg(rRegI dst, rRegI src, immI_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (XorI (AddI src minus_1) src));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsmskl $dst, $src" %}

  ins_encode %{
    __ blsmskl($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg);
%}

instruct blsrI_rReg_rReg(rRegI dst, rRegI src, immI_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (AndI (AddI src minus_1) src) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsrl  $dst, $src" %}

  ins_encode %{
    __ blsrl($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg_mem);
%}

instruct blsrI_rReg_mem(rRegI dst, memory src, immI_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (AndI (AddI (LoadI src) minus_1) (LoadI src) ) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsrl  $dst, $src" %}

  ins_encode %{
    __ blsrl($dst$$Register, $src$$Address);
  %}

  ins_pipe(ialu_reg);
%}

// Or Instructions
// Or Register with Register
instruct orI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (OrI dst src));
  effect(KILL cr);

  format %{ "orl     $dst, $src\t# int" %}
  opcode(0x0B);
  ins_encode(REX_reg_reg(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// Or Register with Immediate
instruct orI_rReg_imm(rRegI dst, immI src, rFlagsReg cr)
%{
  match(Set dst (OrI dst src));
  effect(KILL cr);

  format %{ "orl     $dst, $src\t# int" %}
  opcode(0x81, 0x01); /* Opcode 81 /1 id */
  ins_encode(OpcSErm(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// Or Register with Memory
instruct orI_rReg_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (OrI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "orl     $dst, $src\t# int" %}
  opcode(0x0B);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// Or Memory with Register
instruct orB_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreB dst (OrI (LoadB dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "orb    $dst, $src\t# byte" %}
  opcode(0x08);
  ins_encode(REX_breg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct orI_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (OrI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "orl     $dst, $src\t# int" %}
  opcode(0x09); /* Opcode 09 /r */
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// Or Memory with Immediate
instruct orI_mem_imm(memory dst, immI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (OrI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "orl     $dst, $src\t# int" %}
  opcode(0x81, 0x1); /* Opcode 81 /1 id */
  ins_encode(REX_mem(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Xor Instructions
// Xor Register with Register
instruct xorI_rReg(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (XorI dst src));
  effect(KILL cr);

  format %{ "xorl    $dst, $src\t# int" %}
  opcode(0x33);
  ins_encode(REX_reg_reg(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// Xor Register with Immediate -1
instruct xorI_rReg_im1(rRegI dst, immI_M1 imm) %{
  match(Set dst (XorI dst imm));

  format %{ "not    $dst" %}
  ins_encode %{
     __ notl($dst$$Register);
  %}
  ins_pipe(ialu_reg);
%}

// Xor Register with Immediate
instruct xorI_rReg_imm(rRegI dst, immI src, rFlagsReg cr)
%{
  match(Set dst (XorI dst src));
  effect(KILL cr);

  format %{ "xorl    $dst, $src\t# int" %}
  opcode(0x81, 0x06); /* Opcode 81 /6 id */
  ins_encode(OpcSErm(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// Xor Register with Memory
instruct xorI_rReg_mem(rRegI dst, memory src, rFlagsReg cr)
%{
  match(Set dst (XorI dst (LoadI src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "xorl    $dst, $src\t# int" %}
  opcode(0x33);
  ins_encode(REX_reg_mem(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// Xor Memory with Register
instruct xorB_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreB dst (XorI (LoadB dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "xorb    $dst, $src\t# byte" %}
  opcode(0x30);
  ins_encode(REX_breg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

instruct xorI_mem_rReg(memory dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (XorI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "xorl    $dst, $src\t# int" %}
  opcode(0x31); /* Opcode 31 /r */
  ins_encode(REX_reg_mem(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// Xor Memory with Immediate
instruct xorI_mem_imm(memory dst, immI src, rFlagsReg cr)
%{
  match(Set dst (StoreI dst (XorI (LoadI dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "xorl    $dst, $src\t# int" %}
  opcode(0x81, 0x6); /* Opcode 81 /6 id */
  ins_encode(REX_mem(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}


// Long Logical Instructions

// And Instructions
// And Register with Register
instruct andL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (AndL dst src));
  effect(KILL cr);

  format %{ "andq    $dst, $src\t# long" %}
  opcode(0x23);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// And Register with Immediate 255
instruct andL_rReg_imm255(rRegL dst, immL_255 src)
%{
  match(Set dst (AndL dst src));

  format %{ "movzbq  $dst, $dst\t# long & 0xFF" %}
  opcode(0x0F, 0xB6);
  ins_encode(REX_reg_reg_wide(dst, dst), OpcP, OpcS, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate 65535
instruct andL_rReg_imm65535(rRegL dst, immL_65535 src)
%{
  match(Set dst (AndL dst src));

  format %{ "movzwq  $dst, $dst\t# long & 0xFFFF" %}
  opcode(0x0F, 0xB7);
  ins_encode(REX_reg_reg_wide(dst, dst), OpcP, OpcS, reg_reg(dst, dst));
  ins_pipe(ialu_reg);
%}

// And Register with Immediate
instruct andL_rReg_imm(rRegL dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (AndL dst src));
  effect(KILL cr);

  format %{ "andq    $dst, $src\t# long" %}
  opcode(0x81, 0x04); /* Opcode 81 /4 */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// And Register with Memory
instruct andL_rReg_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (AndL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "andq    $dst, $src\t# long" %}
  opcode(0x23);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// And Memory with Register
instruct andL_mem_rReg(memory dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (AndL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "andq    $dst, $src\t# long" %}
  opcode(0x21); /* Opcode 21 /r */
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// And Memory with Immediate
instruct andL_mem_imm(memory dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (AndL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "andq    $dst, $src\t# long" %}
  opcode(0x81, 0x4); /* Opcode 81 /4 id */
  ins_encode(REX_mem_wide(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// BMI1 instructions
instruct andnL_rReg_rReg_mem(rRegL dst, rRegL src1, memory src2, immL_M1 minus_1, rFlagsReg cr) %{
  match(Set dst (AndL (XorL src1 minus_1) (LoadL src2)));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "andnq  $dst, $src1, $src2" %}

  ins_encode %{
    __ andnq($dst$$Register, $src1$$Register, $src2$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct andnL_rReg_rReg_rReg(rRegL dst, rRegL src1, rRegL src2, immL_M1 minus_1, rFlagsReg cr) %{
  match(Set dst (AndL (XorL src1 minus_1) src2));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "andnq  $dst, $src1, $src2" %}

  ins_encode %{
  __ andnq($dst$$Register, $src1$$Register, $src2$$Register);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct blsiL_rReg_rReg(rRegL dst, rRegL src, immL0 imm_zero, rFlagsReg cr) %{
  match(Set dst (AndL (SubL imm_zero src) src));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsiq  $dst, $src" %}

  ins_encode %{
    __ blsiq($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg);
%}

instruct blsiL_rReg_mem(rRegL dst, memory src, immL0 imm_zero, rFlagsReg cr) %{
  match(Set dst (AndL (SubL imm_zero (LoadL src) ) (LoadL src) ));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsiq  $dst, $src" %}

  ins_encode %{
    __ blsiq($dst$$Register, $src$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct blsmskL_rReg_mem(rRegL dst, memory src, immL_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (XorL (AddL (LoadL src) minus_1) (LoadL src) ) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsmskq $dst, $src" %}

  ins_encode %{
    __ blsmskq($dst$$Register, $src$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct blsmskL_rReg_rReg(rRegL dst, rRegL src, immL_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (XorL (AddL src minus_1) src));
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsmskq $dst, $src" %}

  ins_encode %{
    __ blsmskq($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg);
%}

instruct blsrL_rReg_rReg(rRegL dst, rRegL src, immL_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (AndL (AddL src minus_1) src) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  format %{ "blsrq  $dst, $src" %}

  ins_encode %{
    __ blsrq($dst$$Register, $src$$Register);
  %}

  ins_pipe(ialu_reg);
%}

instruct blsrL_rReg_mem(rRegL dst, memory src, immL_M1 minus_1, rFlagsReg cr)
%{
  match(Set dst (AndL (AddL (LoadL src) minus_1) (LoadL src)) );
  predicate(UseBMI1Instructions);
  effect(KILL cr);

  ins_cost(125);
  format %{ "blsrq  $dst, $src" %}

  ins_encode %{
    __ blsrq($dst$$Register, $src$$Address);
  %}

  ins_pipe(ialu_reg);
%}

// Or Instructions
// Or Register with Register
instruct orL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (OrL dst src));
  effect(KILL cr);

  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x0B);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// Use any_RegP to match R15 (TLS register) without spilling.
instruct orL_rReg_castP2X(rRegL dst, any_RegP src, rFlagsReg cr) %{
  match(Set dst (OrL dst (CastP2X src)));
  effect(KILL cr);

  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x0B);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}


// Or Register with Immediate
instruct orL_rReg_imm(rRegL dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (OrL dst src));
  effect(KILL cr);

  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x81, 0x01); /* Opcode 81 /1 id */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// Or Register with Memory
instruct orL_rReg_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (OrL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x0B);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// Or Memory with Register
instruct orL_mem_rReg(memory dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (OrL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x09); /* Opcode 09 /r */
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// Or Memory with Immediate
instruct orL_mem_imm(memory dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (OrL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "orq     $dst, $src\t# long" %}
  opcode(0x81, 0x1); /* Opcode 81 /1 id */
  ins_encode(REX_mem_wide(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Xor Instructions
// Xor Register with Register
instruct xorL_rReg(rRegL dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (XorL dst src));
  effect(KILL cr);

  format %{ "xorq    $dst, $src\t# long" %}
  opcode(0x33);
  ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst, src));
  ins_pipe(ialu_reg_reg);
%}

// Xor Register with Immediate -1
instruct xorL_rReg_im1(rRegL dst, immL_M1 imm) %{
  match(Set dst (XorL dst imm));

  format %{ "notq   $dst" %}
  ins_encode %{
     __ notq($dst$$Register);
  %}
  ins_pipe(ialu_reg);
%}

// Xor Register with Immediate
instruct xorL_rReg_imm(rRegL dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (XorL dst src));
  effect(KILL cr);

  format %{ "xorq    $dst, $src\t# long" %}
  opcode(0x81, 0x06); /* Opcode 81 /6 id */
  ins_encode(OpcSErm_wide(dst, src), Con8or32(src));
  ins_pipe(ialu_reg);
%}

// Xor Register with Memory
instruct xorL_rReg_mem(rRegL dst, memory src, rFlagsReg cr)
%{
  match(Set dst (XorL dst (LoadL src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "xorq    $dst, $src\t# long" %}
  opcode(0x33);
  ins_encode(REX_reg_mem_wide(dst, src), OpcP, reg_mem(dst, src));
  ins_pipe(ialu_reg_mem);
%}

// Xor Memory with Register
instruct xorL_mem_rReg(memory dst, rRegL src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (XorL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(150);
  format %{ "xorq    $dst, $src\t# long" %}
  opcode(0x31); /* Opcode 31 /r */
  ins_encode(REX_reg_mem_wide(src, dst), OpcP, reg_mem(src, dst));
  ins_pipe(ialu_mem_reg);
%}

// Xor Memory with Immediate
instruct xorL_mem_imm(memory dst, immL32 src, rFlagsReg cr)
%{
  match(Set dst (StoreL dst (XorL (LoadL dst) src)));
  effect(KILL cr);

  ins_cost(125);
  format %{ "xorq    $dst, $src\t# long" %}
  opcode(0x81, 0x6); /* Opcode 81 /6 id */
  ins_encode(REX_mem_wide(dst), OpcSE(src),
             RM_opc_mem(secondary, dst), Con8or32(src));
  ins_pipe(ialu_mem_imm);
%}

// Convert Int to Boolean
instruct convI2B(rRegI dst, rRegI src, rFlagsReg cr)
%{
  match(Set dst (Conv2B src));
  effect(KILL cr);

  format %{ "testl   $src, $src\t# ci2b\n\t"
            "setnz   $dst\n\t"
            "movzbl  $dst, $dst" %}
  ins_encode(REX_reg_reg(src, src), opc_reg_reg(0x85, src, src), // testl
             setNZ_reg(dst),
             REX_reg_breg(dst, dst), // movzbl
             Opcode(0x0F), Opcode(0xB6), reg_reg(dst, dst));
  ins_pipe(pipe_slow); // XXX
%}

// Convert Pointer to Boolean
instruct convP2B(rRegI dst, rRegP src, rFlagsReg cr)
%{
  match(Set dst (Conv2B src));
  effect(KILL cr);

  format %{ "testq   $src, $src\t# cp2b\n\t"
            "setnz   $dst\n\t"
            "movzbl  $dst, $dst" %}
  ins_encode(REX_reg_reg_wide(src, src), opc_reg_reg(0x85, src, src), // testq
             setNZ_reg(dst),
             REX_reg_breg(dst, dst), // movzbl
             Opcode(0x0F), Opcode(0xB6), reg_reg(dst, dst));
  ins_pipe(pipe_slow); // XXX
%}

instruct cmpLTMask(rRegI dst, rRegI p, rRegI q, rFlagsReg cr)
%{
  match(Set dst (CmpLTMask p q));
  effect(KILL cr);

  ins_cost(400);
  format %{ "cmpl    $p, $q\t# cmpLTMask\n\t"
            "setlt   $dst\n\t"
            "movzbl  $dst, $dst\n\t"
            "negl    $dst" %}
  ins_encode(REX_reg_reg(p, q), opc_reg_reg(0x3B, p, q), // cmpl
             setLT_reg(dst),
             REX_reg_breg(dst, dst), // movzbl
             Opcode(0x0F), Opcode(0xB6), reg_reg(dst, dst),
             neg_reg(dst));
  ins_pipe(pipe_slow);
%}

instruct cmpLTMask0(rRegI dst, immI0 zero, rFlagsReg cr)
%{
  match(Set dst (CmpLTMask dst zero));
  effect(KILL cr);

  ins_cost(100);
  format %{ "sarl    $dst, #31\t# cmpLTMask0" %}
  ins_encode %{
  __ sarl($dst$$Register, 31);
  %}
  ins_pipe(ialu_reg);
%}

/* Better to save a register than avoid a branch */
instruct cadd_cmpLTMask(rRegI p, rRegI q, rRegI y, rFlagsReg cr)
%{
  match(Set p (AddI (AndI (CmpLTMask p q) y) (SubI p q)));
  effect(KILL cr);
  ins_cost(300);
  format %{ "subl    $p,$q\t# cadd_cmpLTMask\n\t"
            "jge     done\n\t"
            "addl    $p,$y\n"
            "done:   " %}
  ins_encode %{
    Register Rp = $p$$Register;
    Register Rq = $q$$Register;
    Register Ry = $y$$Register;
    Label done;
    __ subl(Rp, Rq);
    __ jccb(Assembler::greaterEqual, done);
    __ addl(Rp, Ry);
    __ bind(done);
  %}
  ins_pipe(pipe_cmplt);
%}

/* Better to save a register than avoid a branch */
instruct and_cmpLTMask(rRegI p, rRegI q, rRegI y, rFlagsReg cr)
%{
  match(Set y (AndI (CmpLTMask p q) y));
  effect(KILL cr);

  ins_cost(300);

  format %{ "cmpl    $p, $q\t# and_cmpLTMask\n\t"
            "jlt     done\n\t"
            "xorl    $y, $y\n"
            "done:   " %}
  ins_encode %{
    Register Rp = $p$$Register;
    Register Rq = $q$$Register;
    Register Ry = $y$$Register;
    Label done;
    __ cmpl(Rp, Rq);
    __ jccb(Assembler::less, done);
    __ xorl(Ry, Ry);
    __ bind(done);
  %}
  ins_pipe(pipe_cmplt);
%}


//---------- FP Instructions------------------------------------------------

instruct cmpF_cc_reg(rFlagsRegU cr, regF src1, regF src2)
%{
  match(Set cr (CmpF src1 src2));

  ins_cost(145);
  format %{ "ucomiss $src1, $src2\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$XMMRegister);
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpF_cc_reg_CF(rFlagsRegUCF cr, regF src1, regF src2) %{
  match(Set cr (CmpF src1 src2));

  ins_cost(100);
  format %{ "ucomiss $src1, $src2" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpF_cc_mem(rFlagsRegU cr, regF src1, memory src2)
%{
  match(Set cr (CmpF src1 (LoadF src2)));

  ins_cost(145);
  format %{ "ucomiss $src1, $src2\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$Address);
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpF_cc_memCF(rFlagsRegUCF cr, regF src1, memory src2) %{
  match(Set cr (CmpF src1 (LoadF src2)));

  ins_cost(100);
  format %{ "ucomiss $src1, $src2" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$Address);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpF_cc_imm(rFlagsRegU cr, regF src, immF con) %{
  match(Set cr (CmpF src con));

  ins_cost(145);
  format %{ "ucomiss $src, [$constantaddress]\t# load from constant table: float=$con\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomiss($src$$XMMRegister, $constantaddress($con));
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpF_cc_immCF(rFlagsRegUCF cr, regF src, immF con) %{
  match(Set cr (CmpF src con));
  ins_cost(100);
  format %{ "ucomiss $src, [$constantaddress]\t# load from constant table: float=$con" %}
  ins_encode %{
    __ ucomiss($src$$XMMRegister, $constantaddress($con));
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_reg(rFlagsRegU cr, regD src1, regD src2)
%{
  match(Set cr (CmpD src1 src2));

  ins_cost(145);
  format %{ "ucomisd $src1, $src2\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$XMMRegister);
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_reg_CF(rFlagsRegUCF cr, regD src1, regD src2) %{
  match(Set cr (CmpD src1 src2));

  ins_cost(100);
  format %{ "ucomisd $src1, $src2 test" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_mem(rFlagsRegU cr, regD src1, memory src2)
%{
  match(Set cr (CmpD src1 (LoadD src2)));

  ins_cost(145);
  format %{ "ucomisd $src1, $src2\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$Address);
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_memCF(rFlagsRegUCF cr, regD src1, memory src2) %{
  match(Set cr (CmpD src1 (LoadD src2)));

  ins_cost(100);
  format %{ "ucomisd $src1, $src2" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$Address);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_imm(rFlagsRegU cr, regD src, immD con) %{
  match(Set cr (CmpD src con));

  ins_cost(145);
  format %{ "ucomisd $src, [$constantaddress]\t# load from constant table: double=$con\n\t"
            "jnp,s   exit\n\t"
            "pushfq\t# saw NaN, set CF\n\t"
            "andq    [rsp], #0xffffff2b\n\t"
            "popfq\n"
    "exit:" %}
  ins_encode %{
    __ ucomisd($src$$XMMRegister, $constantaddress($con));
    emit_cmpfp_fixup(_masm);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpD_cc_immCF(rFlagsRegUCF cr, regD src, immD con) %{
  match(Set cr (CmpD src con));
  ins_cost(100);
  format %{ "ucomisd $src, [$constantaddress]\t# load from constant table: double=$con" %}
  ins_encode %{
    __ ucomisd($src$$XMMRegister, $constantaddress($con));
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpF_reg(rRegI dst, regF src1, regF src2, rFlagsReg cr)
%{
  match(Set dst (CmpF3 src1 src2));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomiss $src1, $src2\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$XMMRegister);
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpF_mem(rRegI dst, regF src1, memory src2, rFlagsReg cr)
%{
  match(Set dst (CmpF3 src1 (LoadF src2)));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomiss $src1, $src2\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomiss($src1$$XMMRegister, $src2$$Address);
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpF_imm(rRegI dst, regF src, immF con, rFlagsReg cr) %{
  match(Set dst (CmpF3 src con));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomiss $src, [$constantaddress]\t# load from constant table: float=$con\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomiss($src$$XMMRegister, $constantaddress($con));
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpD_reg(rRegI dst, regD src1, regD src2, rFlagsReg cr)
%{
  match(Set dst (CmpD3 src1 src2));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomisd $src1, $src2\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$XMMRegister);
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpD_mem(rRegI dst, regD src1, memory src2, rFlagsReg cr)
%{
  match(Set dst (CmpD3 src1 (LoadD src2)));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomisd $src1, $src2\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomisd($src1$$XMMRegister, $src2$$Address);
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

// Compare into -1,0,1
instruct cmpD_imm(rRegI dst, regD src, immD con, rFlagsReg cr) %{
  match(Set dst (CmpD3 src con));
  effect(KILL cr);

  ins_cost(275);
  format %{ "ucomisd $src, [$constantaddress]\t# load from constant table: double=$con\n\t"
            "movl    $dst, #-1\n\t"
            "jp,s    done\n\t"
            "jb,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n"
    "done:" %}
  ins_encode %{
    __ ucomisd($src$$XMMRegister, $constantaddress($con));
    emit_cmpfp3(_masm, $dst$$Register);
  %}
  ins_pipe(pipe_slow);
%}

//----------Arithmetic Conversion Instructions---------------------------------

instruct roundFloat_nop(regF dst)
%{
  match(Set dst (RoundFloat dst));

  ins_cost(0);
  ins_encode();
  ins_pipe(empty);
%}

instruct roundDouble_nop(regD dst)
%{
  match(Set dst (RoundDouble dst));

  ins_cost(0);
  ins_encode();
  ins_pipe(empty);
%}

instruct convF2D_reg_reg(regD dst, regF src)
%{
  match(Set dst (ConvF2D src));

  format %{ "cvtss2sd $dst, $src" %}
  ins_encode %{
    __ cvtss2sd ($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convF2D_reg_mem(regD dst, memory src)
%{
  match(Set dst (ConvF2D (LoadF src)));

  format %{ "cvtss2sd $dst, $src" %}
  ins_encode %{
    __ cvtss2sd ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convD2F_reg_reg(regF dst, regD src)
%{
  match(Set dst (ConvD2F src));

  format %{ "cvtsd2ss $dst, $src" %}
  ins_encode %{
    __ cvtsd2ss ($dst$$XMMRegister, $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convD2F_reg_mem(regF dst, memory src)
%{
  match(Set dst (ConvD2F (LoadD src)));

  format %{ "cvtsd2ss $dst, $src" %}
  ins_encode %{
    __ cvtsd2ss ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

// XXX do mem variants
instruct convF2I_reg_reg(rRegI dst, regF src, rFlagsReg cr)
%{
  match(Set dst (ConvF2I src));
  effect(KILL cr);

  format %{ "cvttss2sil $dst, $src\t# f2i\n\t"
            "cmpl    $dst, #0x80000000\n\t"
            "jne,s   done\n\t"
            "subq    rsp, #8\n\t"
            "movss   [rsp], $src\n\t"
            "call    f2i_fixup\n\t"
            "popq    $dst\n"
    "done:   "%}
  ins_encode %{
    Label done;
    __ cvttss2sil($dst$$Register, $src$$XMMRegister);
    __ cmpl($dst$$Register, 0x80000000);
    __ jccb(Assembler::notEqual, done);
    __ subptr(rsp, 8);
    __ movflt(Address(rsp, 0), $src$$XMMRegister);
    __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::x86::f2i_fixup())));
    __ pop($dst$$Register);
    __ bind(done);
  %}
  ins_pipe(pipe_slow);
%}

instruct convF2L_reg_reg(rRegL dst, regF src, rFlagsReg cr)
%{
  match(Set dst (ConvF2L src));
  effect(KILL cr);

  format %{ "cvttss2siq $dst, $src\t# f2l\n\t"
            "cmpq    $dst, [0x8000000000000000]\n\t"
            "jne,s   done\n\t"
            "subq    rsp, #8\n\t"
            "movss   [rsp], $src\n\t"
            "call    f2l_fixup\n\t"
            "popq    $dst\n"
    "done:   "%}
  ins_encode %{
    Label done;
    __ cvttss2siq($dst$$Register, $src$$XMMRegister);
    __ cmp64($dst$$Register,
             ExternalAddress((address) StubRoutines::x86::double_sign_flip()));
    __ jccb(Assembler::notEqual, done);
    __ subptr(rsp, 8);
    __ movflt(Address(rsp, 0), $src$$XMMRegister);
    __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::x86::f2l_fixup())));
    __ pop($dst$$Register);
    __ bind(done);
  %}
  ins_pipe(pipe_slow);
%}

instruct convD2I_reg_reg(rRegI dst, regD src, rFlagsReg cr)
%{
  match(Set dst (ConvD2I src));
  effect(KILL cr);

  format %{ "cvttsd2sil $dst, $src\t# d2i\n\t"
            "cmpl    $dst, #0x80000000\n\t"
            "jne,s   done\n\t"
            "subq    rsp, #8\n\t"
            "movsd   [rsp], $src\n\t"
            "call    d2i_fixup\n\t"
            "popq    $dst\n"
    "done:   "%}
  ins_encode %{
    Label done;
    __ cvttsd2sil($dst$$Register, $src$$XMMRegister);
    __ cmpl($dst$$Register, 0x80000000);
    __ jccb(Assembler::notEqual, done);
    __ subptr(rsp, 8);
    __ movdbl(Address(rsp, 0), $src$$XMMRegister);
    __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::x86::d2i_fixup())));
    __ pop($dst$$Register);
    __ bind(done);
  %}
  ins_pipe(pipe_slow);
%}

instruct convD2L_reg_reg(rRegL dst, regD src, rFlagsReg cr)
%{
  match(Set dst (ConvD2L src));
  effect(KILL cr);

  format %{ "cvttsd2siq $dst, $src\t# d2l\n\t"
            "cmpq    $dst, [0x8000000000000000]\n\t"
            "jne,s   done\n\t"
            "subq    rsp, #8\n\t"
            "movsd   [rsp], $src\n\t"
            "call    d2l_fixup\n\t"
            "popq    $dst\n"
    "done:   "%}
  ins_encode %{
    Label done;
    __ cvttsd2siq($dst$$Register, $src$$XMMRegister);
    __ cmp64($dst$$Register,
             ExternalAddress((address) StubRoutines::x86::double_sign_flip()));
    __ jccb(Assembler::notEqual, done);
    __ subptr(rsp, 8);
    __ movdbl(Address(rsp, 0), $src$$XMMRegister);
    __ call(RuntimeAddress(CAST_FROM_FN_PTR(address, StubRoutines::x86::d2l_fixup())));
    __ pop($dst$$Register);
    __ bind(done);
  %}
  ins_pipe(pipe_slow);
%}

instruct convI2F_reg_reg(regF dst, rRegI src)
%{
  predicate(!UseXmmI2F);
  match(Set dst (ConvI2F src));

  format %{ "cvtsi2ssl $dst, $src\t# i2f" %}
  ins_encode %{
    __ cvtsi2ssl ($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convI2F_reg_mem(regF dst, memory src)
%{
  match(Set dst (ConvI2F (LoadI src)));

  format %{ "cvtsi2ssl $dst, $src\t# i2f" %}
  ins_encode %{
    __ cvtsi2ssl ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convI2D_reg_reg(regD dst, rRegI src)
%{
  predicate(!UseXmmI2D);
  match(Set dst (ConvI2D src));

  format %{ "cvtsi2sdl $dst, $src\t# i2d" %}
  ins_encode %{
    __ cvtsi2sdl ($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convI2D_reg_mem(regD dst, memory src)
%{
  match(Set dst (ConvI2D (LoadI src)));

  format %{ "cvtsi2sdl $dst, $src\t# i2d" %}
  ins_encode %{
    __ cvtsi2sdl ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convXI2F_reg(regF dst, rRegI src)
%{
  predicate(UseXmmI2F);
  match(Set dst (ConvI2F src));

  format %{ "movdl $dst, $src\n\t"
            "cvtdq2psl $dst, $dst\t# i2f" %}
  ins_encode %{
    __ movdl($dst$$XMMRegister, $src$$Register);
    __ cvtdq2ps($dst$$XMMRegister, $dst$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convXI2D_reg(regD dst, rRegI src)
%{
  predicate(UseXmmI2D);
  match(Set dst (ConvI2D src));

  format %{ "movdl $dst, $src\n\t"
            "cvtdq2pdl $dst, $dst\t# i2d" %}
  ins_encode %{
    __ movdl($dst$$XMMRegister, $src$$Register);
    __ cvtdq2pd($dst$$XMMRegister, $dst$$XMMRegister);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convL2F_reg_reg(regF dst, rRegL src)
%{
  match(Set dst (ConvL2F src));

  format %{ "cvtsi2ssq $dst, $src\t# l2f" %}
  ins_encode %{
    __ cvtsi2ssq ($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convL2F_reg_mem(regF dst, memory src)
%{
  match(Set dst (ConvL2F (LoadL src)));

  format %{ "cvtsi2ssq $dst, $src\t# l2f" %}
  ins_encode %{
    __ cvtsi2ssq ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convL2D_reg_reg(regD dst, rRegL src)
%{
  match(Set dst (ConvL2D src));

  format %{ "cvtsi2sdq $dst, $src\t# l2d" %}
  ins_encode %{
    __ cvtsi2sdq ($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convL2D_reg_mem(regD dst, memory src)
%{
  match(Set dst (ConvL2D (LoadL src)));

  format %{ "cvtsi2sdq $dst, $src\t# l2d" %}
  ins_encode %{
    __ cvtsi2sdq ($dst$$XMMRegister, $src$$Address);
  %}
  ins_pipe(pipe_slow); // XXX
%}

instruct convI2L_reg_reg(rRegL dst, rRegI src)
%{
  match(Set dst (ConvI2L src));

  ins_cost(125);
  format %{ "movslq  $dst, $src\t# i2l" %}
  ins_encode %{
    __ movslq($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

// instruct convI2L_reg_reg_foo(rRegL dst, rRegI src)
// %{
//   match(Set dst (ConvI2L src));
// //   predicate(_kids[0]->_leaf->as_Type()->type()->is_int()->_lo >= 0 &&
// //             _kids[0]->_leaf->as_Type()->type()->is_int()->_hi >= 0);
//   predicate(((const TypeNode*) n)->type()->is_long()->_hi ==
//             (unsigned int) ((const TypeNode*) n)->type()->is_long()->_hi &&
//             ((const TypeNode*) n)->type()->is_long()->_lo ==
//             (unsigned int) ((const TypeNode*) n)->type()->is_long()->_lo);

//   format %{ "movl    $dst, $src\t# unsigned i2l" %}
//   ins_encode(enc_copy(dst, src));
// //   opcode(0x63); // needs REX.W
// //   ins_encode(REX_reg_reg_wide(dst, src), OpcP, reg_reg(dst,src));
//   ins_pipe(ialu_reg_reg);
// %}

// Zero-extend convert int to long
instruct convI2L_reg_reg_zex(rRegL dst, rRegI src, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L src) mask));

  format %{ "movl    $dst, $src\t# i2l zero-extend\n\t" %}
  ins_encode %{
    if ($dst$$reg != $src$$reg) {
      __ movl($dst$$Register, $src$$Register);
    }
  %}
  ins_pipe(ialu_reg_reg);
%}

// Zero-extend convert int to long
instruct convI2L_reg_mem_zex(rRegL dst, memory src, immL_32bits mask)
%{
  match(Set dst (AndL (ConvI2L (LoadI src)) mask));

  format %{ "movl    $dst, $src\t# i2l zero-extend\n\t" %}
  ins_encode %{
    __ movl($dst$$Register, $src$$Address);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct zerox_long_reg_reg(rRegL dst, rRegL src, immL_32bits mask)
%{
  match(Set dst (AndL src mask));

  format %{ "movl    $dst, $src\t# zero-extend long" %}
  ins_encode %{
    __ movl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct convL2I_reg_reg(rRegI dst, rRegL src)
%{
  match(Set dst (ConvL2I src));

  format %{ "movl    $dst, $src\t# l2i" %}
  ins_encode %{
    __ movl($dst$$Register, $src$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}


instruct MoveF2I_stack_reg(rRegI dst, stackSlotF src) %{
  match(Set dst (MoveF2I src));
  effect(DEF dst, USE src);

  ins_cost(125);
  format %{ "movl    $dst, $src\t# MoveF2I_stack_reg" %}
  ins_encode %{
    __ movl($dst$$Register, Address(rsp, $src$$disp));
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct MoveI2F_stack_reg(regF dst, stackSlotI src) %{
  match(Set dst (MoveI2F src));
  effect(DEF dst, USE src);

  ins_cost(125);
  format %{ "movss   $dst, $src\t# MoveI2F_stack_reg" %}
  ins_encode %{
    __ movflt($dst$$XMMRegister, Address(rsp, $src$$disp));
  %}
  ins_pipe(pipe_slow);
%}

instruct MoveD2L_stack_reg(rRegL dst, stackSlotD src) %{
  match(Set dst (MoveD2L src));
  effect(DEF dst, USE src);

  ins_cost(125);
  format %{ "movq    $dst, $src\t# MoveD2L_stack_reg" %}
  ins_encode %{
    __ movq($dst$$Register, Address(rsp, $src$$disp));
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct MoveL2D_stack_reg_partial(regD dst, stackSlotL src) %{
  predicate(!UseXmmLoadAndClearUpper);
  match(Set dst (MoveL2D src));
  effect(DEF dst, USE src);

  ins_cost(125);
  format %{ "movlpd  $dst, $src\t# MoveL2D_stack_reg" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, Address(rsp, $src$$disp));
  %}
  ins_pipe(pipe_slow);
%}

instruct MoveL2D_stack_reg(regD dst, stackSlotL src) %{
  predicate(UseXmmLoadAndClearUpper);
  match(Set dst (MoveL2D src));
  effect(DEF dst, USE src);

  ins_cost(125);
  format %{ "movsd   $dst, $src\t# MoveL2D_stack_reg" %}
  ins_encode %{
    __ movdbl($dst$$XMMRegister, Address(rsp, $src$$disp));
  %}
  ins_pipe(pipe_slow);
%}


instruct MoveF2I_reg_stack(stackSlotI dst, regF src) %{
  match(Set dst (MoveF2I src));
  effect(DEF dst, USE src);

  ins_cost(95); // XXX
  format %{ "movss   $dst, $src\t# MoveF2I_reg_stack" %}
  ins_encode %{
    __ movflt(Address(rsp, $dst$$disp), $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

instruct MoveI2F_reg_stack(stackSlotF dst, rRegI src) %{
  match(Set dst (MoveI2F src));
  effect(DEF dst, USE src);

  ins_cost(100);
  format %{ "movl    $dst, $src\t# MoveI2F_reg_stack" %}
  ins_encode %{
    __ movl(Address(rsp, $dst$$disp), $src$$Register);
  %}
  ins_pipe( ialu_mem_reg );
%}

instruct MoveD2L_reg_stack(stackSlotL dst, regD src) %{
  match(Set dst (MoveD2L src));
  effect(DEF dst, USE src);

  ins_cost(95); // XXX
  format %{ "movsd   $dst, $src\t# MoveL2D_reg_stack" %}
  ins_encode %{
    __ movdbl(Address(rsp, $dst$$disp), $src$$XMMRegister);
  %}
  ins_pipe(pipe_slow);
%}

instruct MoveL2D_reg_stack(stackSlotD dst, rRegL src) %{
  match(Set dst (MoveL2D src));
  effect(DEF dst, USE src);

  ins_cost(100);
  format %{ "movq    $dst, $src\t# MoveL2D_reg_stack" %}
  ins_encode %{
    __ movq(Address(rsp, $dst$$disp), $src$$Register);
  %}
  ins_pipe(ialu_mem_reg);
%}

instruct MoveF2I_reg_reg(rRegI dst, regF src) %{
  match(Set dst (MoveF2I src));
  effect(DEF dst, USE src);
  ins_cost(85);
  format %{ "movd    $dst,$src\t# MoveF2I" %}
  ins_encode %{
    __ movdl($dst$$Register, $src$$XMMRegister);
  %}
  ins_pipe( pipe_slow );
%}

instruct MoveD2L_reg_reg(rRegL dst, regD src) %{
  match(Set dst (MoveD2L src));
  effect(DEF dst, USE src);
  ins_cost(85);
  format %{ "movd    $dst,$src\t# MoveD2L" %}
  ins_encode %{
    __ movdq($dst$$Register, $src$$XMMRegister);
  %}
  ins_pipe( pipe_slow );
%}

instruct MoveI2F_reg_reg(regF dst, rRegI src) %{
  match(Set dst (MoveI2F src));
  effect(DEF dst, USE src);
  ins_cost(100);
  format %{ "movd    $dst,$src\t# MoveI2F" %}
  ins_encode %{
    __ movdl($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe( pipe_slow );
%}

instruct MoveL2D_reg_reg(regD dst, rRegL src) %{
  match(Set dst (MoveL2D src));
  effect(DEF dst, USE src);
  ins_cost(100);
  format %{ "movd    $dst,$src\t# MoveL2D" %}
  ins_encode %{
     __ movdq($dst$$XMMRegister, $src$$Register);
  %}
  ins_pipe( pipe_slow );
%}


// =======================================================================
// fast clearing of an array
instruct rep_stos(rcx_RegL cnt, rdi_RegP base, regD tmp, rax_RegI zero,
                  Universe dummy, rFlagsReg cr)
%{
  predicate(!((ClearArrayNode*)n)->is_large());
  match(Set dummy (ClearArray cnt base));
  effect(USE_KILL cnt, USE_KILL base, TEMP tmp, KILL zero, KILL cr);

  format %{ $$template
    $$emit$$"xorq    rax, rax\t# ClearArray:\n\t"
    $$emit$$"cmp     InitArrayShortSize,rcx\n\t"
    $$emit$$"jg      LARGE\n\t"
    $$emit$$"dec     rcx\n\t"
    $$emit$$"js      DONE\t# Zero length\n\t"
    $$emit$$"mov     rax,(rdi,rcx,8)\t# LOOP\n\t"
    $$emit$$"dec     rcx\n\t"
    $$emit$$"jge     LOOP\n\t"
    $$emit$$"jmp     DONE\n\t"
    $$emit$$"# LARGE:\n\t"
    if (UseFastStosb) {
       $$emit$$"shlq    rcx,3\t# Convert doublewords to bytes\n\t"
       $$emit$$"rep     stosb\t# Store rax to *rdi++ while rcx--\n\t"
    } else if (UseXMMForObjInit) {
       $$emit$$"mov     rdi,rax\n\t"
       $$emit$$"vpxor   ymm0,ymm0,ymm0\n\t"
       $$emit$$"jmpq    L_zero_64_bytes\n\t"
       $$emit$$"# L_loop:\t# 64-byte LOOP\n\t"
       $$emit$$"vmovdqu ymm0,(rax)\n\t"
       $$emit$$"vmovdqu ymm0,0x20(rax)\n\t"
       $$emit$$"add     0x40,rax\n\t"
       $$emit$$"# L_zero_64_bytes:\n\t"
       $$emit$$"sub     0x8,rcx\n\t"
       $$emit$$"jge     L_loop\n\t"
       $$emit$$"add     0x4,rcx\n\t"
       $$emit$$"jl      L_tail\n\t"
       $$emit$$"vmovdqu ymm0,(rax)\n\t"
       $$emit$$"add     0x20,rax\n\t"
       $$emit$$"sub     0x4,rcx\n\t"
       $$emit$$"# L_tail:\t# Clearing tail bytes\n\t"
       $$emit$$"add     0x4,rcx\n\t"
       $$emit$$"jle     L_end\n\t"
       $$emit$$"dec     rcx\n\t"
       $$emit$$"# L_sloop:\t# 8-byte short loop\n\t"
       $$emit$$"vmovq   xmm0,(rax)\n\t"
       $$emit$$"add     0x8,rax\n\t"
       $$emit$$"dec     rcx\n\t"
       $$emit$$"jge     L_sloop\n\t"
       $$emit$$"# L_end:\n\t"
    } else {
       $$emit$$"rep     stosq\t# Store rax to *rdi++ while rcx--\n\t"
    }
    $$emit$$"# DONE"
  %}
  ins_encode %{
    __ clear_mem($base$$Register, $cnt$$Register, $zero$$Register,
                 $tmp$$XMMRegister, false);
  %}
  ins_pipe(pipe_slow);
%}

instruct rep_stos_large(rcx_RegL cnt, rdi_RegP base, regD tmp, rax_RegI zero,
                        Universe dummy, rFlagsReg cr)
%{
  predicate(((ClearArrayNode*)n)->is_large());
  match(Set dummy (ClearArray cnt base));
  effect(USE_KILL cnt, USE_KILL base, TEMP tmp, KILL zero, KILL cr);

  format %{ $$template
    if (UseFastStosb) {
       $$emit$$"xorq    rax, rax\t# ClearArray:\n\t"
       $$emit$$"shlq    rcx,3\t# Convert doublewords to bytes\n\t"
       $$emit$$"rep     stosb\t# Store rax to *rdi++ while rcx--"
    } else if (UseXMMForObjInit) {
       $$emit$$"mov     rdi,rax\t# ClearArray:\n\t"
       $$emit$$"vpxor   ymm0,ymm0,ymm0\n\t"
       $$emit$$"jmpq    L_zero_64_bytes\n\t"
       $$emit$$"# L_loop:\t# 64-byte LOOP\n\t"
       $$emit$$"vmovdqu ymm0,(rax)\n\t"
       $$emit$$"vmovdqu ymm0,0x20(rax)\n\t"
       $$emit$$"add     0x40,rax\n\t"
       $$emit$$"# L_zero_64_bytes:\n\t"
       $$emit$$"sub     0x8,rcx\n\t"
       $$emit$$"jge     L_loop\n\t"
       $$emit$$"add     0x4,rcx\n\t"
       $$emit$$"jl      L_tail\n\t"
       $$emit$$"vmovdqu ymm0,(rax)\n\t"
       $$emit$$"add     0x20,rax\n\t"
       $$emit$$"sub     0x4,rcx\n\t"
       $$emit$$"# L_tail:\t# Clearing tail bytes\n\t"
       $$emit$$"add     0x4,rcx\n\t"
       $$emit$$"jle     L_end\n\t"
       $$emit$$"dec     rcx\n\t"
       $$emit$$"# L_sloop:\t# 8-byte short loop\n\t"
       $$emit$$"vmovq   xmm0,(rax)\n\t"
       $$emit$$"add     0x8,rax\n\t"
       $$emit$$"dec     rcx\n\t"
       $$emit$$"jge     L_sloop\n\t"
       $$emit$$"# L_end:\n\t"
    } else {
       $$emit$$"xorq    rax, rax\t# ClearArray:\n\t"
       $$emit$$"rep     stosq\t# Store rax to *rdi++ while rcx--"
    }
  %}
  ins_encode %{
    __ clear_mem($base$$Register, $cnt$$Register, $zero$$Register,
                 $tmp$$XMMRegister, true);
  %}
  ins_pipe(pipe_slow);
%}

instruct string_compareL(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
                         rax_RegI result, legVecS tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$XMMRegister, StrIntrinsicNode::LL);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_compareU(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
                         rax_RegI result, legVecS tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$XMMRegister, StrIntrinsicNode::UU);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_compareLU(rdi_RegP str1, rcx_RegI cnt1, rsi_RegP str2, rdx_RegI cnt2,
                          rax_RegI result, legVecS tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::LU);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register, $result$$Register,
                      $tmp1$$XMMRegister, StrIntrinsicNode::LU);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_compareUL(rsi_RegP str1, rdx_RegI cnt1, rdi_RegP str2, rcx_RegI cnt2,
                          rax_RegI result, legVecS tmp1, rFlagsReg cr)
%{
  predicate(((StrCompNode*)n)->encoding() == StrIntrinsicNode::UL);
  match(Set result (StrComp (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP tmp1, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL cr);

  format %{ "String Compare byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL $tmp1" %}
  ins_encode %{
    __ string_compare($str2$$Register, $str1$$Register,
                      $cnt2$$Register, $cnt1$$Register, $result$$Register,
                      $tmp1$$XMMRegister, StrIntrinsicNode::UL);
  %}
  ins_pipe( pipe_slow );
%}

// fast search of substring with known size.
instruct string_indexof_conL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
                             rbx_RegI result, legVecS vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    if (icnt2 >= 16) {
      // IndexOf for constant substrings with size >= 16 elements
      // which don't need to be loaded through stack.
      __ string_indexofC8($str1$$Register, $str2$$Register,
                          $cnt1$$Register, $cnt2$$Register,
                          icnt2, $result$$Register,
                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
    } else {
      // Small strings are loaded through stack if they cross page boundary.
      __ string_indexof($str1$$Register, $str2$$Register,
                        $cnt1$$Register, $cnt2$$Register,
                        icnt2, $result$$Register,
                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
    }
  %}
  ins_pipe( pipe_slow );
%}

// fast search of substring with known size.
instruct string_indexof_conU(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
                             rbx_RegI result, legVecS vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    if (icnt2 >= 8) {
      // IndexOf for constant substrings with size >= 8 elements
      // which don't need to be loaded through stack.
      __ string_indexofC8($str1$$Register, $str2$$Register,
                          $cnt1$$Register, $cnt2$$Register,
                          icnt2, $result$$Register,
                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
    } else {
      // Small strings are loaded through stack if they cross page boundary.
      __ string_indexof($str1$$Register, $str2$$Register,
                        $cnt1$$Register, $cnt2$$Register,
                        icnt2, $result$$Register,
                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
    }
  %}
  ins_pipe( pipe_slow );
%}

// fast search of substring with known size.
instruct string_indexof_conUL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, immI int_cnt2,
                             rbx_RegI result, legVecS vec, rax_RegI cnt2, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 int_cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$int_cnt2 -> $result   // KILL $vec, $cnt1, $cnt2, $tmp" %}
  ins_encode %{
    int icnt2 = (int)$int_cnt2$$constant;
    if (icnt2 >= 8) {
      // IndexOf for constant substrings with size >= 8 elements
      // which don't need to be loaded through stack.
      __ string_indexofC8($str1$$Register, $str2$$Register,
                          $cnt1$$Register, $cnt2$$Register,
                          icnt2, $result$$Register,
                          $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
    } else {
      // Small strings are loaded through stack if they cross page boundary.
      __ string_indexof($str1$$Register, $str2$$Register,
                        $cnt1$$Register, $cnt2$$Register,
                        icnt2, $result$$Register,
                        $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
    }
  %}
  ins_pipe( pipe_slow );
%}

instruct string_indexofL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
                         rbx_RegI result, legVecS vec, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::LL));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf byte[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      (-1), $result$$Register,
                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::LL);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_indexofU(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
                         rbx_RegI result, legVecS vec, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UU));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      (-1), $result$$Register,
                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UU);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_indexofUL(rdi_RegP str1, rdx_RegI cnt1, rsi_RegP str2, rax_RegI cnt2,
                         rbx_RegI result, legVecS vec, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics && (((StrIndexOfNode*)n)->encoding() == StrIntrinsicNode::UL));
  match(Set result (StrIndexOf (Binary str1 cnt1) (Binary str2 cnt2)));
  effect(TEMP vec, USE_KILL str1, USE_KILL str2, USE_KILL cnt1, USE_KILL cnt2, KILL tmp, KILL cr);

  format %{ "String IndexOf char[] $str1,$cnt1,$str2,$cnt2 -> $result   // KILL all" %}
  ins_encode %{
    __ string_indexof($str1$$Register, $str2$$Register,
                      $cnt1$$Register, $cnt2$$Register,
                      (-1), $result$$Register,
                      $vec$$XMMRegister, $tmp$$Register, StrIntrinsicNode::UL);
  %}
  ins_pipe( pipe_slow );
%}

instruct string_indexofU_char(rdi_RegP str1, rdx_RegI cnt1, rax_RegI ch,
                              rbx_RegI result, legVecS vec1, legVecS vec2, legVecS vec3, rcx_RegI tmp, rFlagsReg cr)
%{
  predicate(UseSSE42Intrinsics);
  match(Set result (StrIndexOfChar (Binary str1 cnt1) ch));
  effect(TEMP vec1, TEMP vec2, TEMP vec3, USE_KILL str1, USE_KILL cnt1, USE_KILL ch, TEMP tmp, KILL cr);
  format %{ "String IndexOf char[] $str1,$cnt1,$ch -> $result   // KILL all" %}
  ins_encode %{
    __ string_indexof_char($str1$$Register, $cnt1$$Register, $ch$$Register, $result$$Register,
                           $vec1$$XMMRegister, $vec2$$XMMRegister, $vec3$$XMMRegister, $tmp$$Register);
  %}
  ins_pipe( pipe_slow );
%}

// fast string equals
instruct string_equals(rdi_RegP str1, rsi_RegP str2, rcx_RegI cnt, rax_RegI result,
                       legVecS tmp1, legVecS tmp2, rbx_RegI tmp3, rFlagsReg cr)
%{
  match(Set result (StrEquals (Binary str1 str2) cnt));
  effect(TEMP tmp1, TEMP tmp2, USE_KILL str1, USE_KILL str2, USE_KILL cnt, KILL tmp3, KILL cr);

  format %{ "String Equals $str1,$str2,$cnt -> $result    // KILL $tmp1, $tmp2, $tmp3" %}
  ins_encode %{
    __ arrays_equals(false, $str1$$Register, $str2$$Register,
                     $cnt$$Register, $result$$Register, $tmp3$$Register,
                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
  %}
  ins_pipe( pipe_slow );
%}

// fast array equals
instruct array_equalsB(rdi_RegP ary1, rsi_RegP ary2, rax_RegI result,
                       legVecS tmp1, legVecS tmp2, rcx_RegI tmp3, rbx_RegI tmp4, rFlagsReg cr)
%{
  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::LL);
  match(Set result (AryEq ary1 ary2));
  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);

  format %{ "Array Equals byte[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
  ins_encode %{
    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, false /* char */);
  %}
  ins_pipe( pipe_slow );
%}

instruct array_equalsC(rdi_RegP ary1, rsi_RegP ary2, rax_RegI result,
                      legVecS tmp1, legVecS tmp2, rcx_RegI tmp3, rbx_RegI tmp4, rFlagsReg cr)
%{
  predicate(((AryEqNode*)n)->encoding() == StrIntrinsicNode::UU);
  match(Set result (AryEq ary1 ary2));
  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL ary2, KILL tmp3, KILL tmp4, KILL cr);

  format %{ "Array Equals char[] $ary1,$ary2 -> $result   // KILL $tmp1, $tmp2, $tmp3, $tmp4" %}
  ins_encode %{
    __ arrays_equals(true, $ary1$$Register, $ary2$$Register,
                     $tmp3$$Register, $result$$Register, $tmp4$$Register,
                     $tmp1$$XMMRegister, $tmp2$$XMMRegister, true /* char */);
  %}
  ins_pipe( pipe_slow );
%}

instruct has_negatives(rsi_RegP ary1, rcx_RegI len, rax_RegI result,
                      legVecS tmp1, legVecS tmp2, rbx_RegI tmp3, rFlagsReg cr)
%{
  match(Set result (HasNegatives ary1 len));
  effect(TEMP tmp1, TEMP tmp2, USE_KILL ary1, USE_KILL len, KILL tmp3, KILL cr);

  format %{ "has negatives byte[] $ary1,$len -> $result   // KILL $tmp1, $tmp2, $tmp3" %}
  ins_encode %{
    __ has_negatives($ary1$$Register, $len$$Register,
                     $result$$Register, $tmp3$$Register,
                     $tmp1$$XMMRegister, $tmp2$$XMMRegister);
  %}
  ins_pipe( pipe_slow );
%}

// fast char[] to byte[] compression
instruct string_compress(rsi_RegP src, rdi_RegP dst, rdx_RegI len, legVecS tmp1, legVecS tmp2, legVecS tmp3, legVecS tmp4,
                         rcx_RegI tmp5, rax_RegI result, rFlagsReg cr) %{
  match(Set result (StrCompressedCopy src (Binary dst len)));
  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL tmp5, KILL cr);

  format %{ "String Compress $src,$dst -> $result    // KILL RAX, RCX, RDX" %}
  ins_encode %{
    __ char_array_compress($src$$Register, $dst$$Register, $len$$Register,
                           $tmp1$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister,
                           $tmp4$$XMMRegister, $tmp5$$Register, $result$$Register);
  %}
  ins_pipe( pipe_slow );
%}

// fast byte[] to char[] inflation
instruct string_inflate(Universe dummy, rsi_RegP src, rdi_RegP dst, rdx_RegI len,
                        legVecS tmp1, rcx_RegI tmp2, rFlagsReg cr) %{
  match(Set dummy (StrInflatedCopy src (Binary dst len)));
  effect(TEMP tmp1, TEMP tmp2, USE_KILL src, USE_KILL dst, USE_KILL len, KILL cr);

  format %{ "String Inflate $src,$dst    // KILL $tmp1, $tmp2" %}
  ins_encode %{
    __ byte_array_inflate($src$$Register, $dst$$Register, $len$$Register,
                          $tmp1$$XMMRegister, $tmp2$$Register);
  %}
  ins_pipe( pipe_slow );
%}

// encode char[] to byte[] in ISO_8859_1
instruct encode_iso_array(rsi_RegP src, rdi_RegP dst, rdx_RegI len,
                          legVecS tmp1, legVecS tmp2, legVecS tmp3, legVecS tmp4,
                          rcx_RegI tmp5, rax_RegI result, rFlagsReg cr) %{
  match(Set result (EncodeISOArray src (Binary dst len)));
  effect(TEMP tmp1, TEMP tmp2, TEMP tmp3, TEMP tmp4, USE_KILL src, USE_KILL dst, USE_KILL len, KILL tmp5, KILL cr);

  format %{ "Encode array $src,$dst,$len -> $result    // KILL RCX, RDX, $tmp1, $tmp2, $tmp3, $tmp4, RSI, RDI " %}
  ins_encode %{
    __ encode_iso_array($src$$Register, $dst$$Register, $len$$Register,
                        $tmp1$$XMMRegister, $tmp2$$XMMRegister, $tmp3$$XMMRegister,
                        $tmp4$$XMMRegister, $tmp5$$Register, $result$$Register);
  %}
  ins_pipe( pipe_slow );
%}

//----------Overflow Math Instructions-----------------------------------------

instruct overflowAddI_rReg(rFlagsReg cr, rax_RegI op1, rRegI op2)
%{
  match(Set cr (OverflowAddI op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "addl    $op1, $op2\t# overflow check int" %}

  ins_encode %{
    __ addl($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowAddI_rReg_imm(rFlagsReg cr, rax_RegI op1, immI op2)
%{
  match(Set cr (OverflowAddI op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "addl    $op1, $op2\t# overflow check int" %}

  ins_encode %{
    __ addl($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowAddL_rReg(rFlagsReg cr, rax_RegL op1, rRegL op2)
%{
  match(Set cr (OverflowAddL op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "addq    $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ addq($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowAddL_rReg_imm(rFlagsReg cr, rax_RegL op1, immL32 op2)
%{
  match(Set cr (OverflowAddL op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "addq    $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ addq($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowSubI_rReg(rFlagsReg cr, rRegI op1, rRegI op2)
%{
  match(Set cr (OverflowSubI op1 op2));

  format %{ "cmpl    $op1, $op2\t# overflow check int" %}
  ins_encode %{
    __ cmpl($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowSubI_rReg_imm(rFlagsReg cr, rRegI op1, immI op2)
%{
  match(Set cr (OverflowSubI op1 op2));

  format %{ "cmpl    $op1, $op2\t# overflow check int" %}
  ins_encode %{
    __ cmpl($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowSubL_rReg(rFlagsReg cr, rRegL op1, rRegL op2)
%{
  match(Set cr (OverflowSubL op1 op2));

  format %{ "cmpq    $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ cmpq($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowSubL_rReg_imm(rFlagsReg cr, rRegL op1, immL32 op2)
%{
  match(Set cr (OverflowSubL op1 op2));

  format %{ "cmpq    $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ cmpq($op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowNegI_rReg(rFlagsReg cr, immI0 zero, rax_RegI op2)
%{
  match(Set cr (OverflowSubI zero op2));
  effect(DEF cr, USE_KILL op2);

  format %{ "negl    $op2\t# overflow check int" %}
  ins_encode %{
    __ negl($op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowNegL_rReg(rFlagsReg cr, immL0 zero, rax_RegL op2)
%{
  match(Set cr (OverflowSubL zero op2));
  effect(DEF cr, USE_KILL op2);

  format %{ "negq    $op2\t# overflow check long" %}
  ins_encode %{
    __ negq($op2$$Register);
  %}
  ins_pipe(ialu_reg_reg);
%}

instruct overflowMulI_rReg(rFlagsReg cr, rax_RegI op1, rRegI op2)
%{
  match(Set cr (OverflowMulI op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "imull    $op1, $op2\t# overflow check int" %}
  ins_encode %{
    __ imull($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct overflowMulI_rReg_imm(rFlagsReg cr, rRegI op1, immI op2, rRegI tmp)
%{
  match(Set cr (OverflowMulI op1 op2));
  effect(DEF cr, TEMP tmp, USE op1, USE op2);

  format %{ "imull    $tmp, $op1, $op2\t# overflow check int" %}
  ins_encode %{
    __ imull($tmp$$Register, $op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct overflowMulL_rReg(rFlagsReg cr, rax_RegL op1, rRegL op2)
%{
  match(Set cr (OverflowMulL op1 op2));
  effect(DEF cr, USE_KILL op1, USE op2);

  format %{ "imulq    $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ imulq($op1$$Register, $op2$$Register);
  %}
  ins_pipe(ialu_reg_reg_alu0);
%}

instruct overflowMulL_rReg_imm(rFlagsReg cr, rRegL op1, immL32 op2, rRegL tmp)
%{
  match(Set cr (OverflowMulL op1 op2));
  effect(DEF cr, TEMP tmp, USE op1, USE op2);

  format %{ "imulq    $tmp, $op1, $op2\t# overflow check long" %}
  ins_encode %{
    __ imulq($tmp$$Register, $op1$$Register, $op2$$constant);
  %}
  ins_pipe(ialu_reg_reg_alu0);
%}


//----------Control Flow Instructions------------------------------------------
// Signed compare Instructions

// XXX more variants!!
instruct compI_rReg(rFlagsReg cr, rRegI op1, rRegI op2)
%{
  match(Set cr (CmpI op1 op2));
  effect(DEF cr, USE op1, USE op2);

  format %{ "cmpl    $op1, $op2" %}
  opcode(0x3B);  /* Opcode 3B /r */
  ins_encode(REX_reg_reg(op1, op2), OpcP, reg_reg(op1, op2));
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compI_rReg_imm(rFlagsReg cr, rRegI op1, immI op2)
%{
  match(Set cr (CmpI op1 op2));

  format %{ "cmpl    $op1, $op2" %}
  opcode(0x81, 0x07); /* Opcode 81 /7 */
  ins_encode(OpcSErm(op1, op2), Con8or32(op2));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compI_rReg_mem(rFlagsReg cr, rRegI op1, memory op2)
%{
  match(Set cr (CmpI op1 (LoadI op2)));

  ins_cost(500); // XXX
  format %{ "cmpl    $op1, $op2" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testI_reg(rFlagsReg cr, rRegI src, immI0 zero)
%{
  match(Set cr (CmpI src zero));

  format %{ "testl   $src, $src" %}
  opcode(0x85);
  ins_encode(REX_reg_reg(src, src), OpcP, reg_reg(src, src));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testI_reg_imm(rFlagsReg cr, rRegI src, immI con, immI0 zero)
%{
  match(Set cr (CmpI (AndI src con) zero));

  format %{ "testl   $src, $con" %}
  opcode(0xF7, 0x00);
  ins_encode(REX_reg(src), OpcP, reg_opc(src), Con32(con));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testI_reg_mem(rFlagsReg cr, rRegI src, memory mem, immI0 zero)
%{
  match(Set cr (CmpI (AndI src (LoadI mem)) zero));

  format %{ "testl   $src, $mem" %}
  opcode(0x85);
  ins_encode(REX_reg_mem(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_cr_reg_mem);
%}

// Unsigned compare Instructions; really, same as signed except they
// produce an rFlagsRegU instead of rFlagsReg.
instruct compU_rReg(rFlagsRegU cr, rRegI op1, rRegI op2)
%{
  match(Set cr (CmpU op1 op2));

  format %{ "cmpl    $op1, $op2\t# unsigned" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_reg(op1, op2), OpcP, reg_reg(op1, op2));
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compU_rReg_imm(rFlagsRegU cr, rRegI op1, immI op2)
%{
  match(Set cr (CmpU op1 op2));

  format %{ "cmpl    $op1, $op2\t# unsigned" %}
  opcode(0x81,0x07); /* Opcode 81 /7 */
  ins_encode(OpcSErm(op1, op2), Con8or32(op2));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compU_rReg_mem(rFlagsRegU cr, rRegI op1, memory op2)
%{
  match(Set cr (CmpU op1 (LoadI op2)));

  ins_cost(500); // XXX
  format %{ "cmpl    $op1, $op2\t# unsigned" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

// // // Cisc-spilled version of cmpU_rReg
// //instruct compU_mem_rReg(rFlagsRegU cr, memory op1, rRegI op2)
// //%{
// //  match(Set cr (CmpU (LoadI op1) op2));
// //
// //  format %{ "CMPu   $op1,$op2" %}
// //  ins_cost(500);
// //  opcode(0x39);  /* Opcode 39 /r */
// //  ins_encode( OpcP, reg_mem( op1, op2) );
// //%}

instruct testU_reg(rFlagsRegU cr, rRegI src, immI0 zero)
%{
  match(Set cr (CmpU src zero));

  format %{ "testl   $src, $src\t# unsigned" %}
  opcode(0x85);
  ins_encode(REX_reg_reg(src, src), OpcP, reg_reg(src, src));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compP_rReg(rFlagsRegU cr, rRegP op1, rRegP op2)
%{
  match(Set cr (CmpP op1 op2));

  format %{ "cmpq    $op1, $op2\t# ptr" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_reg_wide(op1, op2), OpcP, reg_reg(op1, op2));
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compP_rReg_mem(rFlagsRegU cr, rRegP op1, memory op2)
%{
  match(Set cr (CmpP op1 (LoadP op2)));
  predicate(n->in(2)->as_Load()->barrier_data() == 0);

  ins_cost(500); // XXX
  format %{ "cmpq    $op1, $op2\t# ptr" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem_wide(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

// // // Cisc-spilled version of cmpP_rReg
// //instruct compP_mem_rReg(rFlagsRegU cr, memory op1, rRegP op2)
// //%{
// //  match(Set cr (CmpP (LoadP op1) op2));
// //
// //  format %{ "CMPu   $op1,$op2" %}
// //  ins_cost(500);
// //  opcode(0x39);  /* Opcode 39 /r */
// //  ins_encode( OpcP, reg_mem( op1, op2) );
// //%}

// XXX this is generalized by compP_rReg_mem???
// Compare raw pointer (used in out-of-heap check).
// Only works because non-oop pointers must be raw pointers
// and raw pointers have no anti-dependencies.
instruct compP_mem_rReg(rFlagsRegU cr, rRegP op1, memory op2)
%{
  predicate(n->in(2)->in(2)->bottom_type()->reloc() == relocInfo::none &&
            n->in(2)->as_Load()->barrier_data() == 0);
  match(Set cr (CmpP op1 (LoadP op2)));

  format %{ "cmpq    $op1, $op2\t# raw ptr" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem_wide(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

// This will generate a signed flags result. This should be OK since
// any compare to a zero should be eq/neq.
instruct testP_reg(rFlagsReg cr, rRegP src, immP0 zero)
%{
  match(Set cr (CmpP src zero));

  format %{ "testq   $src, $src\t# ptr" %}
  opcode(0x85);
  ins_encode(REX_reg_reg_wide(src, src), OpcP, reg_reg(src, src));
  ins_pipe(ialu_cr_reg_imm);
%}

// This will generate a signed flags result. This should be OK since
// any compare to a zero should be eq/neq.
instruct testP_mem(rFlagsReg cr, memory op, immP0 zero)
%{
  predicate((!UseCompressedOops || (CompressedOops::base() != NULL)) &&
            n->in(1)->as_Load()->barrier_data() == 0);
  match(Set cr (CmpP (LoadP op) zero));

  ins_cost(500); // XXX
  format %{ "testq   $op, 0xffffffffffffffff\t# ptr" %}
  opcode(0xF7); /* Opcode F7 /0 */
  ins_encode(REX_mem_wide(op),
             OpcP, RM_opc_mem(0x00, op), Con_d32(0xFFFFFFFF));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testP_mem_reg0(rFlagsReg cr, memory mem, immP0 zero)
%{
  predicate(UseCompressedOops && (CompressedOops::base() == NULL) &&
            (CompressedKlassPointers::base() == NULL) &&
            n->in(1)->as_Load()->barrier_data() == 0);
  match(Set cr (CmpP (LoadP mem) zero));

  format %{ "cmpq    R12, $mem\t# ptr (R12_heapbase==0)" %}
  ins_encode %{
    __ cmpq(r12, $mem$$Address);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct compN_rReg(rFlagsRegU cr, rRegN op1, rRegN op2)
%{
  match(Set cr (CmpN op1 op2));

  format %{ "cmpl    $op1, $op2\t# compressed ptr" %}
  ins_encode %{ __ cmpl($op1$$Register, $op2$$Register); %}
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compN_rReg_mem(rFlagsRegU cr, rRegN src, memory mem)
%{
  match(Set cr (CmpN src (LoadN mem)));

  format %{ "cmpl    $src, $mem\t# compressed ptr" %}
  ins_encode %{
    __ cmpl($src$$Register, $mem$$Address);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct compN_rReg_imm(rFlagsRegU cr, rRegN op1, immN op2) %{
  match(Set cr (CmpN op1 op2));

  format %{ "cmpl    $op1, $op2\t# compressed ptr" %}
  ins_encode %{
    __ cmp_narrow_oop($op1$$Register, (jobject)$op2$$constant);
  %}
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compN_mem_imm(rFlagsRegU cr, memory mem, immN src)
%{
  match(Set cr (CmpN src (LoadN mem)));

  format %{ "cmpl    $mem, $src\t# compressed ptr" %}
  ins_encode %{
    __ cmp_narrow_oop($mem$$Address, (jobject)$src$$constant);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct compN_rReg_imm_klass(rFlagsRegU cr, rRegN op1, immNKlass op2) %{
  match(Set cr (CmpN op1 op2));

  format %{ "cmpl    $op1, $op2\t# compressed klass ptr" %}
  ins_encode %{
    __ cmp_narrow_klass($op1$$Register, (Klass*)$op2$$constant);
  %}
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compN_mem_imm_klass(rFlagsRegU cr, memory mem, immNKlass src)
%{
  match(Set cr (CmpN src (LoadNKlass mem)));

  format %{ "cmpl    $mem, $src\t# compressed klass ptr" %}
  ins_encode %{
    __ cmp_narrow_klass($mem$$Address, (Klass*)$src$$constant);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testN_reg(rFlagsReg cr, rRegN src, immN0 zero) %{
  match(Set cr (CmpN src zero));

  format %{ "testl   $src, $src\t# compressed ptr" %}
  ins_encode %{ __ testl($src$$Register, $src$$Register); %}
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testN_mem(rFlagsReg cr, memory mem, immN0 zero)
%{
  predicate(CompressedOops::base() != NULL);
  match(Set cr (CmpN (LoadN mem) zero));

  ins_cost(500); // XXX
  format %{ "testl   $mem, 0xffffffff\t# compressed ptr" %}
  ins_encode %{
    __ cmpl($mem$$Address, (int)0xFFFFFFFF);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testN_mem_reg0(rFlagsReg cr, memory mem, immN0 zero)
%{
  predicate(CompressedOops::base() == NULL && (CompressedKlassPointers::base() == NULL));
  match(Set cr (CmpN (LoadN mem) zero));

  format %{ "cmpl    R12, $mem\t# compressed ptr (R12_heapbase==0)" %}
  ins_encode %{
    __ cmpl(r12, $mem$$Address);
  %}
  ins_pipe(ialu_cr_reg_mem);
%}

// Yanked all unsigned pointer compare operations.
// Pointer compares are done with CmpP which is already unsigned.

instruct compL_rReg(rFlagsReg cr, rRegL op1, rRegL op2)
%{
  match(Set cr (CmpL op1 op2));

  format %{ "cmpq    $op1, $op2" %}
  opcode(0x3B);  /* Opcode 3B /r */
  ins_encode(REX_reg_reg_wide(op1, op2), OpcP, reg_reg(op1, op2));
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compL_rReg_imm(rFlagsReg cr, rRegL op1, immL32 op2)
%{
  match(Set cr (CmpL op1 op2));

  format %{ "cmpq    $op1, $op2" %}
  opcode(0x81, 0x07); /* Opcode 81 /7 */
  ins_encode(OpcSErm_wide(op1, op2), Con8or32(op2));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compL_rReg_mem(rFlagsReg cr, rRegL op1, memory op2)
%{
  match(Set cr (CmpL op1 (LoadL op2)));

  format %{ "cmpq    $op1, $op2" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem_wide(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testL_reg(rFlagsReg cr, rRegL src, immL0 zero)
%{
  match(Set cr (CmpL src zero));

  format %{ "testq   $src, $src" %}
  opcode(0x85);
  ins_encode(REX_reg_reg_wide(src, src), OpcP, reg_reg(src, src));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testL_reg_imm(rFlagsReg cr, rRegL src, immL32 con, immL0 zero)
%{
  match(Set cr (CmpL (AndL src con) zero));

  format %{ "testq   $src, $con\t# long" %}
  opcode(0xF7, 0x00);
  ins_encode(REX_reg_wide(src), OpcP, reg_opc(src), Con32(con));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct testL_reg_mem(rFlagsReg cr, rRegL src, memory mem, immL0 zero)
%{
  match(Set cr (CmpL (AndL src (LoadL mem)) zero));

  format %{ "testq   $src, $mem" %}
  opcode(0x85);
  ins_encode(REX_reg_mem_wide(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testL_reg_mem2(rFlagsReg cr, rRegP src, memory mem, immL0 zero)
%{
  match(Set cr (CmpL (AndL (CastP2X src) (LoadL mem)) zero));

  format %{ "testq   $src, $mem" %}
  opcode(0x85);
  ins_encode(REX_reg_mem_wide(src, mem), OpcP, reg_mem(src, mem));
  ins_pipe(ialu_cr_reg_mem);
%}

// Manifest a CmpL result in an integer register.  Very painful.
// This is the test to avoid.
instruct cmpL3_reg_reg(rRegI dst, rRegL src1, rRegL src2, rFlagsReg flags)
%{
  match(Set dst (CmpL3 src1 src2));
  effect(KILL flags);

  ins_cost(275); // XXX
  format %{ "cmpq    $src1, $src2\t# CmpL3\n\t"
            "movl    $dst, -1\n\t"
            "jl,s    done\n\t"
            "setne   $dst\n\t"
            "movzbl  $dst, $dst\n\t"
    "done:" %}
  ins_encode(cmpl3_flag(src1, src2, dst));
  ins_pipe(pipe_slow);
%}

// Unsigned long compare Instructions; really, same as signed long except they
// produce an rFlagsRegU instead of rFlagsReg.
instruct compUL_rReg(rFlagsRegU cr, rRegL op1, rRegL op2)
%{
  match(Set cr (CmpUL op1 op2));

  format %{ "cmpq    $op1, $op2\t# unsigned" %}
  opcode(0x3B);  /* Opcode 3B /r */
  ins_encode(REX_reg_reg_wide(op1, op2), OpcP, reg_reg(op1, op2));
  ins_pipe(ialu_cr_reg_reg);
%}

instruct compUL_rReg_imm(rFlagsRegU cr, rRegL op1, immL32 op2)
%{
  match(Set cr (CmpUL op1 op2));

  format %{ "cmpq    $op1, $op2\t# unsigned" %}
  opcode(0x81, 0x07); /* Opcode 81 /7 */
  ins_encode(OpcSErm_wide(op1, op2), Con8or32(op2));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compUL_rReg_mem(rFlagsRegU cr, rRegL op1, memory op2)
%{
  match(Set cr (CmpUL op1 (LoadL op2)));

  format %{ "cmpq    $op1, $op2\t# unsigned" %}
  opcode(0x3B); /* Opcode 3B /r */
  ins_encode(REX_reg_mem_wide(op1, op2), OpcP, reg_mem(op1, op2));
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testUL_reg(rFlagsRegU cr, rRegL src, immL0 zero)
%{
  match(Set cr (CmpUL src zero));

  format %{ "testq   $src, $src\t# unsigned" %}
  opcode(0x85);
  ins_encode(REX_reg_reg_wide(src, src), OpcP, reg_reg(src, src));
  ins_pipe(ialu_cr_reg_imm);
%}

instruct compB_mem_imm(rFlagsReg cr, memory mem, immI8 imm)
%{
  match(Set cr (CmpI (LoadB mem) imm));

  ins_cost(125);
  format %{ "cmpb    $mem, $imm" %}
  ins_encode %{ __ cmpb($mem$$Address, $imm$$constant); %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testUB_mem_imm(rFlagsReg cr, memory mem, immU8 imm, immI0 zero)
%{
  match(Set cr (CmpI (AndI (LoadUB mem) imm) zero));

  ins_cost(125);
  format %{ "testb   $mem, $imm\t# ubyte" %}
  ins_encode %{ __ testb($mem$$Address, $imm$$constant); %}
  ins_pipe(ialu_cr_reg_mem);
%}

instruct testB_mem_imm(rFlagsReg cr, memory mem, immI8 imm, immI0 zero)
%{
  match(Set cr (CmpI (AndI (LoadB mem) imm) zero));

  ins_cost(125);
  format %{ "testb   $mem, $imm\t# byte" %}
  ins_encode %{ __ testb($mem$$Address, $imm$$constant); %}
  ins_pipe(ialu_cr_reg_mem);
%}

//----------Max and Min--------------------------------------------------------
// Min Instructions

instruct cmovI_reg_g(rRegI dst, rRegI src, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE src, USE cr);

  format %{ "cmovlgt $dst, $src\t# min" %}
  opcode(0x0F, 0x4F);
  ins_encode(REX_reg_reg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}


instruct minI_rReg(rRegI dst, rRegI src)
%{
  match(Set dst (MinI dst src));

  ins_cost(200);
  expand %{
    rFlagsReg cr;
    compI_rReg(cr, dst, src);
    cmovI_reg_g(dst, src, cr);
  %}
%}

instruct cmovI_reg_l(rRegI dst, rRegI src, rFlagsReg cr)
%{
  effect(USE_DEF dst, USE src, USE cr);

  format %{ "cmovllt $dst, $src\t# max" %}
  opcode(0x0F, 0x4C);
  ins_encode(REX_reg_reg(dst, src), OpcP, OpcS, reg_reg(dst, src));
  ins_pipe(pipe_cmov_reg);
%}


instruct maxI_rReg(rRegI dst, rRegI src)
%{
  match(Set dst (MaxI dst src));

  ins_cost(200);
  expand %{
    rFlagsReg cr;
    compI_rReg(cr, dst, src);
    cmovI_reg_l(dst, src, cr);
  %}
%}

// ============================================================================
// Branch Instructions

// Jump Direct - Label defines a relative address from JMP+1
instruct jmpDir(label labl)
%{
  match(Goto);
  effect(USE labl);

  ins_cost(300);
  format %{ "jmp     $labl" %}
  size(5);
  ins_encode %{
    Label* L = $labl$$label;
    __ jmp(*L, false); // Always long jump
  %}
  ins_pipe(pipe_jmp);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpCon(cmpOp cop, rFlagsReg cr, label labl)
%{
  match(If cop cr);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop     $labl" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEnd(cmpOp cop, rFlagsReg cr, label labl)
%{
  predicate(!n->has_vector_mask_set());
  match(CountedLoopEnd cop cr);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop     $labl\t# loop end" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEndU(cmpOpU cop, rFlagsRegU cmp, label labl) %{
  predicate(!n->has_vector_mask_set());
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,u   $labl\t# loop end" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

instruct jmpLoopEndUCF(cmpOpUCF cop, rFlagsRegUCF cmp, label labl) %{
  predicate(!n->has_vector_mask_set());
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(200);
  format %{ "j$cop,u   $labl\t# loop end" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

// mask version
// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEnd_and_restoreMask(cmpOp cop, rFlagsReg cr, label labl)
%{
  predicate(n->has_vector_mask_set());
  match(CountedLoopEnd cop cr);
  effect(USE labl);

  ins_cost(400);
  format %{ "j$cop     $labl\t# loop end\n\t"
            "restorevectmask \t# vector mask restore for loops" %}
  size(10);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
    __ restorevectmask();
  %}
  ins_pipe(pipe_jcc);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEndU_and_restoreMask(cmpOpU cop, rFlagsRegU cmp, label labl) %{
  predicate(n->has_vector_mask_set());
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(400);
  format %{ "j$cop,u   $labl\t# loop end\n\t"
            "restorevectmask \t# vector mask restore for loops" %}
  size(10);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
    __ restorevectmask();
  %}
  ins_pipe(pipe_jcc);
%}

instruct jmpLoopEndUCF_and_restoreMask(cmpOpUCF cop, rFlagsRegUCF cmp, label labl) %{
  predicate(n->has_vector_mask_set());
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,u   $labl\t# loop end\n\t"
            "restorevectmask \t# vector mask restore for loops" %}
  size(10);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
    __ restorevectmask();
  %}
  ins_pipe(pipe_jcc);
%}

// Jump Direct Conditional - using unsigned comparison
instruct jmpConU(cmpOpU cop, rFlagsRegU cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,u   $labl" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

instruct jmpConUCF(cmpOpUCF cop, rFlagsRegUCF cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(200);
  format %{ "j$cop,u   $labl" %}
  size(6);
  ins_encode %{
    Label* L = $labl$$label;
    __ jcc((Assembler::Condition)($cop$$cmpcode), *L, false); // Always long jump
  %}
  ins_pipe(pipe_jcc);
%}

instruct jmpConUCF2(cmpOpUCF2 cop, rFlagsRegUCF cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(200);
  format %{ $$template
    if ($cop$$cmpcode == Assembler::notEqual) {
      $$emit$$"jp,u    $labl\n\t"
      $$emit$$"j$cop,u   $labl"
    } else {
      $$emit$$"jp,u    done\n\t"
      $$emit$$"j$cop,u   $labl\n\t"
      $$emit$$"done:"
    }
  %}
  ins_encode %{
    Label* l = $labl$$label;
    if ($cop$$cmpcode == Assembler::notEqual) {
      __ jcc(Assembler::parity, *l, false);
      __ jcc(Assembler::notEqual, *l, false);
    } else if ($cop$$cmpcode == Assembler::equal) {
      Label done;
      __ jccb(Assembler::parity, done);
      __ jcc(Assembler::equal, *l, false);
      __ bind(done);
    } else {
       ShouldNotReachHere();
    }
  %}
  ins_pipe(pipe_jcc);
%}

// ============================================================================
// The 2nd slow-half of a subtype check.  Scan the subklass's 2ndary
// superklass array for an instance of the superklass.  Set a hidden
// internal cache on a hit (cache is checked with exposed code in
// gen_subtype_check()).  Return NZ for a miss or zero for a hit.  The
// encoding ALSO sets flags.

instruct partialSubtypeCheck(rdi_RegP result,
                             rsi_RegP sub, rax_RegP super, rcx_RegI rcx,
                             rFlagsReg cr)
%{
  match(Set result (PartialSubtypeCheck sub super));
  effect(KILL rcx, KILL cr);

  ins_cost(1100);  // slightly larger than the next version
  format %{ "movq    rdi, [$sub + in_bytes(Klass::secondary_supers_offset())]\n\t"
            "movl    rcx, [rdi + Array<Klass*>::length_offset_in_bytes()]\t# length to scan\n\t"
            "addq    rdi, Array<Klass*>::base_offset_in_bytes()\t# Skip to start of data; set NZ in case count is zero\n\t"
            "repne   scasq\t# Scan *rdi++ for a match with rax while rcx--\n\t"
            "jne,s   miss\t\t# Missed: rdi not-zero\n\t"
            "movq    [$sub + in_bytes(Klass::secondary_super_cache_offset())], $super\t# Hit: update cache\n\t"
            "xorq    $result, $result\t\t Hit: rdi zero\n\t"
    "miss:\t" %}

  opcode(0x1); // Force a XOR of RDI
  ins_encode(enc_PartialSubtypeCheck());
  ins_pipe(pipe_slow);
%}

instruct partialSubtypeCheck_vs_Zero(rFlagsReg cr,
                                     rsi_RegP sub, rax_RegP super, rcx_RegI rcx,
                                     immP0 zero,
                                     rdi_RegP result)
%{
  match(Set cr (CmpP (PartialSubtypeCheck sub super) zero));
  effect(KILL rcx, KILL result);

  ins_cost(1000);
  format %{ "movq    rdi, [$sub + in_bytes(Klass::secondary_supers_offset())]\n\t"
            "movl    rcx, [rdi + Array<Klass*>::length_offset_in_bytes()]\t# length to scan\n\t"
            "addq    rdi, Array<Klass*>::base_offset_in_bytes()\t# Skip to start of data; set NZ in case count is zero\n\t"
            "repne   scasq\t# Scan *rdi++ for a match with rax while cx-- != 0\n\t"
            "jne,s   miss\t\t# Missed: flags nz\n\t"
            "movq    [$sub + in_bytes(Klass::secondary_super_cache_offset())], $super\t# Hit: update cache\n\t"
    "miss:\t" %}

  opcode(0x0); // No need to XOR RDI
  ins_encode(enc_PartialSubtypeCheck());
  ins_pipe(pipe_slow);
%}

// ============================================================================
// Branch Instructions -- short offset versions
//
// These instructions are used to replace jumps of a long offset (the default
// match) with jumps of a shorter offset.  These instructions are all tagged
// with the ins_short_branch attribute, which causes the ADLC to suppress the
// match rules in general matching.  Instead, the ADLC generates a conversion
// method in the MachNode which can be used to do in-place replacement of the
// long variant with the shorter variant.  The compiler will determine if a
// branch can be taken by the is_short_branch_offset() predicate in the machine
// specific code section of the file.

// Jump Direct - Label defines a relative address from JMP+1
instruct jmpDir_short(label labl) %{
  match(Goto);
  effect(USE labl);

  ins_cost(300);
  format %{ "jmp,s   $labl" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jmpb(*L);
  %}
  ins_pipe(pipe_jmp);
  ins_short_branch(1);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpCon_short(cmpOp cop, rFlagsReg cr, label labl) %{
  match(If cop cr);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,s   $labl" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEnd_short(cmpOp cop, rFlagsReg cr, label labl) %{
  match(CountedLoopEnd cop cr);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,s   $labl\t# loop end" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

// Jump Direct Conditional - Label defines a relative address from Jcc+1
instruct jmpLoopEndU_short(cmpOpU cop, rFlagsRegU cmp, label labl) %{
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,us  $labl\t# loop end" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

instruct jmpLoopEndUCF_short(cmpOpUCF cop, rFlagsRegUCF cmp, label labl) %{
  match(CountedLoopEnd cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,us  $labl\t# loop end" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

// Jump Direct Conditional - using unsigned comparison
instruct jmpConU_short(cmpOpU cop, rFlagsRegU cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,us  $labl" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

instruct jmpConUCF_short(cmpOpUCF cop, rFlagsRegUCF cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ "j$cop,us  $labl" %}
  size(2);
  ins_encode %{
    Label* L = $labl$$label;
    __ jccb((Assembler::Condition)($cop$$cmpcode), *L);
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

instruct jmpConUCF2_short(cmpOpUCF2 cop, rFlagsRegUCF cmp, label labl) %{
  match(If cop cmp);
  effect(USE labl);

  ins_cost(300);
  format %{ $$template
    if ($cop$$cmpcode == Assembler::notEqual) {
      $$emit$$"jp,u,s  $labl\n\t"
      $$emit$$"j$cop,u,s  $labl"
    } else {
      $$emit$$"jp,u,s  done\n\t"
      $$emit$$"j$cop,u,s  $labl\n\t"
      $$emit$$"done:"
    }
  %}
  size(4);
  ins_encode %{
    Label* l = $labl$$label;
    if ($cop$$cmpcode == Assembler::notEqual) {
      __ jccb(Assembler::parity, *l);
      __ jccb(Assembler::notEqual, *l);
    } else if ($cop$$cmpcode == Assembler::equal) {
      Label done;
      __ jccb(Assembler::parity, done);
      __ jccb(Assembler::equal, *l);
      __ bind(done);
    } else {
       ShouldNotReachHere();
    }
  %}
  ins_pipe(pipe_jcc);
  ins_short_branch(1);
%}

// ============================================================================
// inlined locking and unlocking

instruct cmpFastLockRTM(rFlagsReg cr, rRegP object, rbx_RegP box, rax_RegI tmp, rdx_RegI scr, rRegI cx1, rRegI cx2) %{
  predicate(Compile::current()->use_rtm());
  match(Set cr (FastLock object box));
  effect(TEMP tmp, TEMP scr, TEMP cx1, TEMP cx2, USE_KILL box);
  ins_cost(300);
  format %{ "fastlock $object,$box\t! kills $box,$tmp,$scr,$cx1,$cx2" %}
  ins_encode %{
    __ fast_lock($object$$Register, $box$$Register, $tmp$$Register,
                 $scr$$Register, $cx1$$Register, $cx2$$Register,
                 _counters, _rtm_counters, _stack_rtm_counters,
                 ((Method*)(ra_->C->method()->constant_encoding()))->method_data(),
                 true, ra_->C->profile_rtm());
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpFastLock(rFlagsReg cr, rRegP object, rbx_RegP box, rax_RegI tmp, rRegP scr) %{
  predicate(!Compile::current()->use_rtm());
  match(Set cr (FastLock object box));
  effect(TEMP tmp, TEMP scr, USE_KILL box);
  ins_cost(300);
  format %{ "fastlock $object,$box\t! kills $box,$tmp,$scr" %}
  ins_encode %{
    __ fast_lock($object$$Register, $box$$Register, $tmp$$Register,
                 $scr$$Register, noreg, noreg, _counters, NULL, NULL, NULL, false, false);
  %}
  ins_pipe(pipe_slow);
%}

instruct cmpFastUnlock(rFlagsReg cr, rRegP object, rax_RegP box, rRegP tmp) %{
  match(Set cr (FastUnlock object box));
  effect(TEMP tmp, USE_KILL box);
  ins_cost(300);
  format %{ "fastunlock $object,$box\t! kills $box,$tmp" %}
  ins_encode %{
    __ fast_unlock($object$$Register, $box$$Register, $tmp$$Register, ra_->C->use_rtm());
  %}
  ins_pipe(pipe_slow);
%}


// ============================================================================
// Safepoint Instructions
instruct safePoint_poll(rFlagsReg cr)
%{
  predicate(!Assembler::is_polling_page_far() && SafepointMechanism::uses_global_page_poll());
  match(SafePoint);
  effect(KILL cr);

  format %{ "testl   rax, [rip + #offset_to_poll_page]\t"
            "# Safepoint: poll for GC" %}
  ins_cost(125);
  ins_encode %{
    AddressLiteral addr(os::get_polling_page(), relocInfo::poll_type);
    __ testl(rax, addr);
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct safePoint_poll_far(rFlagsReg cr, rRegP poll)
%{
  predicate(Assembler::is_polling_page_far() && SafepointMechanism::uses_global_page_poll());
  match(SafePoint poll);
  effect(KILL cr, USE poll);

  format %{ "testl   rax, [$poll]\t"
            "# Safepoint: poll for GC" %}
  ins_cost(125);
  ins_encode %{
    __ relocate(relocInfo::poll_type);
    __ testl(rax, Address($poll$$Register, 0));
  %}
  ins_pipe(ialu_reg_mem);
%}

instruct safePoint_poll_tls(rFlagsReg cr, rRegP poll)
%{
  predicate(SafepointMechanism::uses_thread_local_poll());
  match(SafePoint poll);
  effect(KILL cr, USE poll);

  format %{ "testl   rax, [$poll]\t"
            "# Safepoint: poll for GC" %}
  ins_cost(125);
  size(4); /* setting an explicit size will cause debug builds to assert if size is incorrect */
  ins_encode %{
    __ relocate(relocInfo::poll_type);
    address pre_pc = __ pc();
    __ testl(rax, Address($poll$$Register, 0));
    assert(nativeInstruction_at(pre_pc)->is_safepoint_poll(), "must emit test %%eax [reg]");
  %}
  ins_pipe(ialu_reg_mem);
%}

// ============================================================================
// Procedure Call/Return Instructions
// Call Java Static Instruction
// Note: If this code changes, the corresponding ret_addr_offset() and
//       compute_padding() functions will have to be adjusted.
instruct CallStaticJavaDirect(method meth) %{
  match(CallStaticJava);
  effect(USE meth);

  ins_cost(300);
  format %{ "call,static " %}
  opcode(0xE8); /* E8 cd */
  ins_encode(clear_avx, Java_Static_Call(meth), call_epilog);
  ins_pipe(pipe_slow);
  ins_alignment(4);
%}

// Call Java Dynamic Instruction
// Note: If this code changes, the corresponding ret_addr_offset() and
//       compute_padding() functions will have to be adjusted.
instruct CallDynamicJavaDirect(method meth)
%{
  match(CallDynamicJava);
  effect(USE meth);

  ins_cost(300);
  format %{ "movq    rax, #Universe::non_oop_word()\n\t"
            "call,dynamic " %}
  ins_encode(clear_avx, Java_Dynamic_Call(meth), call_epilog);
  ins_pipe(pipe_slow);
  ins_alignment(4);
%}

// Call Runtime Instruction
instruct CallRuntimeDirect(method meth)
%{
  match(CallRuntime);
  effect(USE meth);

  ins_cost(300);
  format %{ "call,runtime " %}
  ins_encode(clear_avx, Java_To_Runtime(meth));
  ins_pipe(pipe_slow);
%}

// Call runtime without safepoint
instruct CallLeafDirect(method meth)
%{
  match(CallLeaf);
  effect(USE meth);

  ins_cost(300);
  format %{ "call_leaf,runtime " %}
  ins_encode(clear_avx, Java_To_Runtime(meth));
  ins_pipe(pipe_slow);
%}

// Call runtime without safepoint
instruct CallLeafNoFPDirect(method meth)
%{
  match(CallLeafNoFP);
  effect(USE meth);

  ins_cost(300);
  format %{ "call_leaf_nofp,runtime " %}
  ins_encode(clear_avx, Java_To_Runtime(meth));
  ins_pipe(pipe_slow);
%}

// Return Instruction
// Remove the return address & jump to it.
// Notice: We always emit a nop after a ret to make sure there is room
// for safepoint patching
instruct Ret()
%{
  match(Return);

  format %{ "ret" %}
  opcode(0xC3);
  ins_encode(OpcP);
  ins_pipe(pipe_jmp);
%}

// Tail Call; Jump from runtime stub to Java code.
// Also known as an 'interprocedural jump'.
// Target of jump will eventually return to caller.
// TailJump below removes the return address.
instruct TailCalljmpInd(no_rbp_RegP jump_target, rbx_RegP method_oop)
%{
  match(TailCall jump_target method_oop);

  ins_cost(300);
  format %{ "jmp     $jump_target\t# rbx holds method oop" %}
  opcode(0xFF, 0x4); /* Opcode FF /4 */
  ins_encode(REX_reg(jump_target), OpcP, reg_opc(jump_target));
  ins_pipe(pipe_jmp);
%}

// Tail Jump; remove the return address; jump to target.
// TailCall above leaves the return address around.
instruct tailjmpInd(no_rbp_RegP jump_target, rax_RegP ex_oop)
%{
  match(TailJump jump_target ex_oop);

  ins_cost(300);
  format %{ "popq    rdx\t# pop return address\n\t"
            "jmp     $jump_target" %}
  opcode(0xFF, 0x4); /* Opcode FF /4 */
  ins_encode(Opcode(0x5a), // popq rdx
             REX_reg(jump_target), OpcP, reg_opc(jump_target));
  ins_pipe(pipe_jmp);
%}

// Create exception oop: created by stack-crawling runtime code.
// Created exception is now available to this handler, and is setup
// just prior to jumping to this handler.  No code emitted.
instruct CreateException(rax_RegP ex_oop)
%{
  match(Set ex_oop (CreateEx));

  size(0);
  // use the following format syntax
  format %{ "# exception oop is in rax; no code emitted" %}
  ins_encode();
  ins_pipe(empty);
%}

// Rethrow exception:
// The exception oop will come in the first argument position.
// Then JUMP (not call) to the rethrow stub code.
instruct RethrowException()
%{
  match(Rethrow);

  // use the following format syntax
  format %{ "jmp     rethrow_stub" %}
  ins_encode(enc_rethrow);
  ins_pipe(pipe_jmp);
%}

// ============================================================================
// This name is KNOWN by the ADLC and cannot be changed.
// The ADLC forces a 'TypeRawPtr::BOTTOM' output type
// for this guy.
instruct tlsLoadP(r15_RegP dst) %{
  match(Set dst (ThreadLocal));
  effect(DEF dst);

  size(0);
  format %{ "# TLS is in R15" %}
  ins_encode( /*empty encoding*/ );
  ins_pipe(ialu_reg_reg);
%}


//----------PEEPHOLE RULES-----------------------------------------------------
// These must follow all instruction definitions as they use the names
// defined in the instructions definitions.
//
// peepmatch ( root_instr_name [preceding_instruction]* );
//
// peepconstraint %{
// (instruction_number.operand_name relational_op instruction_number.operand_name
//  [, ...] );
// // instruction numbers are zero-based using left to right order in peepmatch
//
// peepreplace ( instr_name  ( [instruction_number.operand_name]* ) );
// // provide an instruction_number.operand_name for each operand that appears
// // in the replacement instruction's match rule
//
// ---------VM FLAGS---------------------------------------------------------
//
// All peephole optimizations can be turned off using -XX:-OptoPeephole
//
// Each peephole rule is given an identifying number starting with zero and
// increasing by one in the order seen by the parser.  An individual peephole
// can be enabled, and all others disabled, by using -XX:OptoPeepholeAt=#
// on the command-line.
//
// ---------CURRENT LIMITATIONS----------------------------------------------
//
// Only match adjacent instructions in same basic block
// Only equality constraints
// Only constraints between operands, not (0.dest_reg == RAX_enc)
// Only one replacement instruction
//
// ---------EXAMPLE----------------------------------------------------------
//
// // pertinent parts of existing instructions in architecture description
// instruct movI(rRegI dst, rRegI src)
// %{
//   match(Set dst (CopyI src));
// %}
//
// instruct incI_rReg(rRegI dst, immI1 src, rFlagsReg cr)
// %{
//   match(Set dst (AddI dst src));
//   effect(KILL cr);
// %}
//
// // Change (inc mov) to lea
// peephole %{
//   // increment preceeded by register-register move
//   peepmatch ( incI_rReg movI );
//   // require that the destination register of the increment
//   // match the destination register of the move
//   peepconstraint ( 0.dst == 1.dst );
//   // construct a replacement instruction that sets
//   // the destination to ( move's source register + one )
//   peepreplace ( leaI_rReg_immI( 0.dst 1.src 0.src ) );
// %}
//

// Implementation no longer uses movX instructions since
// machine-independent system no longer uses CopyX nodes.
//
// peephole
// %{
//   peepmatch (incI_rReg movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_rReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (decI_rReg movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_rReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addI_rReg_imm movI);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaI_rReg_immI(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (incL_rReg movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_rReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (decL_rReg movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_rReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addL_rReg_imm movL);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaL_rReg_immL(0.dst 1.src 0.src));
// %}

// peephole
// %{
//   peepmatch (addP_rReg_imm movP);
//   peepconstraint (0.dst == 1.dst);
//   peepreplace (leaP_rReg_imm(0.dst 1.src 0.src));
// %}

// // Change load of spilled value to only a spill
// instruct storeI(memory mem, rRegI src)
// %{
//   match(Set mem (StoreI mem src));
// %}
//
// instruct loadI(rRegI dst, memory mem)
// %{
//   match(Set dst (LoadI mem));
// %}
//

peephole
%{
  peepmatch (loadI storeI);
  peepconstraint (1.src == 0.dst, 1.mem == 0.mem);
  peepreplace (storeI(1.mem 1.mem 1.src));
%}

peephole
%{
  peepmatch (loadL storeL);
  peepconstraint (1.src == 0.dst, 1.mem == 0.mem);
  peepreplace (storeL(1.mem 1.mem 1.src));
%}

//----------SMARTSPILL RULES---------------------------------------------------
// These must follow all instruction definitions as they use the names
// defined in the instructions definitions.