1 /*
   2  * Copyright (c) 1997, 2016, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #ifndef CPU_X86_VM_ASSEMBLER_X86_HPP
  26 #define CPU_X86_VM_ASSEMBLER_X86_HPP
  27 
  28 #include "asm/register.hpp"
  29 #include "vm_version_x86.hpp"
  30 
  31 class BiasedLockingCounters;
  32 
  33 // Contains all the definitions needed for x86 assembly code generation.
  34 
  35 // Calling convention
  36 class Argument VALUE_OBJ_CLASS_SPEC {
  37  public:
  38   enum {
  39 #ifdef _LP64
  40 #ifdef _WIN64
  41     n_int_register_parameters_c   = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...)
  42     n_float_register_parameters_c = 4,  // xmm0 - xmm3 (c_farg0, c_farg1, ... )
  43 #else
  44     n_int_register_parameters_c   = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...)
  45     n_float_register_parameters_c = 8,  // xmm0 - xmm7 (c_farg0, c_farg1, ... )
  46 #endif // _WIN64
  47     n_int_register_parameters_j   = 6, // j_rarg0, j_rarg1, ...
  48     n_float_register_parameters_j = 8  // j_farg0, j_farg1, ...
  49 #else
  50     n_register_parameters = 0   // 0 registers used to pass arguments
  51 #endif // _LP64
  52   };
  53 };
  54 
  55 
  56 #ifdef _LP64
  57 // Symbolically name the register arguments used by the c calling convention.
  58 // Windows is different from linux/solaris. So much for standards...
  59 
  60 #ifdef _WIN64
  61 
  62 REGISTER_DECLARATION(Register, c_rarg0, rcx);
  63 REGISTER_DECLARATION(Register, c_rarg1, rdx);
  64 REGISTER_DECLARATION(Register, c_rarg2, r8);
  65 REGISTER_DECLARATION(Register, c_rarg3, r9);
  66 
  67 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  68 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  69 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  70 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  71 
  72 #else
  73 
  74 REGISTER_DECLARATION(Register, c_rarg0, rdi);
  75 REGISTER_DECLARATION(Register, c_rarg1, rsi);
  76 REGISTER_DECLARATION(Register, c_rarg2, rdx);
  77 REGISTER_DECLARATION(Register, c_rarg3, rcx);
  78 REGISTER_DECLARATION(Register, c_rarg4, r8);
  79 REGISTER_DECLARATION(Register, c_rarg5, r9);
  80 
  81 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  82 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  83 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  84 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  85 REGISTER_DECLARATION(XMMRegister, c_farg4, xmm4);
  86 REGISTER_DECLARATION(XMMRegister, c_farg5, xmm5);
  87 REGISTER_DECLARATION(XMMRegister, c_farg6, xmm6);
  88 REGISTER_DECLARATION(XMMRegister, c_farg7, xmm7);
  89 
  90 #endif // _WIN64
  91 
  92 // Symbolically name the register arguments used by the Java calling convention.
  93 // We have control over the convention for java so we can do what we please.
  94 // What pleases us is to offset the java calling convention so that when
  95 // we call a suitable jni method the arguments are lined up and we don't
  96 // have to do little shuffling. A suitable jni method is non-static and a
  97 // small number of arguments (two fewer args on windows)
  98 //
  99 //        |-------------------------------------------------------|
 100 //        | c_rarg0   c_rarg1  c_rarg2 c_rarg3 c_rarg4 c_rarg5    |
 101 //        |-------------------------------------------------------|
 102 //        | rcx       rdx      r8      r9      rdi*    rsi*       | windows (* not a c_rarg)
 103 //        | rdi       rsi      rdx     rcx     r8      r9         | solaris/linux
 104 //        |-------------------------------------------------------|
 105 //        | j_rarg5   j_rarg0  j_rarg1 j_rarg2 j_rarg3 j_rarg4    |
 106 //        |-------------------------------------------------------|
 107 
 108 REGISTER_DECLARATION(Register, j_rarg0, c_rarg1);
 109 REGISTER_DECLARATION(Register, j_rarg1, c_rarg2);
 110 REGISTER_DECLARATION(Register, j_rarg2, c_rarg3);
 111 // Windows runs out of register args here
 112 #ifdef _WIN64
 113 REGISTER_DECLARATION(Register, j_rarg3, rdi);
 114 REGISTER_DECLARATION(Register, j_rarg4, rsi);
 115 #else
 116 REGISTER_DECLARATION(Register, j_rarg3, c_rarg4);
 117 REGISTER_DECLARATION(Register, j_rarg4, c_rarg5);
 118 #endif /* _WIN64 */
 119 REGISTER_DECLARATION(Register, j_rarg5, c_rarg0);
 120 
 121 REGISTER_DECLARATION(XMMRegister, j_farg0, xmm0);
 122 REGISTER_DECLARATION(XMMRegister, j_farg1, xmm1);
 123 REGISTER_DECLARATION(XMMRegister, j_farg2, xmm2);
 124 REGISTER_DECLARATION(XMMRegister, j_farg3, xmm3);
 125 REGISTER_DECLARATION(XMMRegister, j_farg4, xmm4);
 126 REGISTER_DECLARATION(XMMRegister, j_farg5, xmm5);
 127 REGISTER_DECLARATION(XMMRegister, j_farg6, xmm6);
 128 REGISTER_DECLARATION(XMMRegister, j_farg7, xmm7);
 129 
 130 REGISTER_DECLARATION(Register, rscratch1, r10);  // volatile
 131 REGISTER_DECLARATION(Register, rscratch2, r11);  // volatile
 132 
 133 REGISTER_DECLARATION(Register, r12_heapbase, r12); // callee-saved
 134 REGISTER_DECLARATION(Register, r15_thread, r15); // callee-saved
 135 
 136 #else
 137 // rscratch1 will apear in 32bit code that is dead but of course must compile
 138 // Using noreg ensures if the dead code is incorrectly live and executed it
 139 // will cause an assertion failure
 140 #define rscratch1 noreg
 141 #define rscratch2 noreg
 142 
 143 #endif // _LP64
 144 
 145 // JSR 292
 146 // On x86, the SP does not have to be saved when invoking method handle intrinsics
 147 // or compiled lambda forms. We indicate that by setting rbp_mh_SP_save to noreg.
 148 REGISTER_DECLARATION(Register, rbp_mh_SP_save, noreg);
 149 
 150 // Address is an abstraction used to represent a memory location
 151 // using any of the amd64 addressing modes with one object.
 152 //
 153 // Note: A register location is represented via a Register, not
 154 //       via an address for efficiency & simplicity reasons.
 155 
 156 class ArrayAddress;
 157 
 158 class Address VALUE_OBJ_CLASS_SPEC {
 159  public:
 160   enum ScaleFactor {
 161     no_scale = -1,
 162     times_1  =  0,
 163     times_2  =  1,
 164     times_4  =  2,
 165     times_8  =  3,
 166     times_ptr = LP64_ONLY(times_8) NOT_LP64(times_4)
 167   };
 168   static ScaleFactor times(int size) {
 169     assert(size >= 1 && size <= 8 && is_power_of_2(size), "bad scale size");
 170     if (size == 8)  return times_8;
 171     if (size == 4)  return times_4;
 172     if (size == 2)  return times_2;
 173     return times_1;
 174   }
 175   static int scale_size(ScaleFactor scale) {
 176     assert(scale != no_scale, "");
 177     assert(((1 << (int)times_1) == 1 &&
 178             (1 << (int)times_2) == 2 &&
 179             (1 << (int)times_4) == 4 &&
 180             (1 << (int)times_8) == 8), "");
 181     return (1 << (int)scale);
 182   }
 183 
 184  private:
 185   Register         _base;
 186   Register         _index;
 187   ScaleFactor      _scale;
 188   int              _disp;
 189   RelocationHolder _rspec;
 190 
 191   // Easily misused constructors make them private
 192   // %%% can we make these go away?
 193   NOT_LP64(Address(address loc, RelocationHolder spec);)
 194   Address(int disp, address loc, relocInfo::relocType rtype);
 195   Address(int disp, address loc, RelocationHolder spec);
 196 
 197  public:
 198 
 199  int disp() { return _disp; }
 200   // creation
 201   Address()
 202     : _base(noreg),
 203       _index(noreg),
 204       _scale(no_scale),
 205       _disp(0) {
 206   }
 207 
 208   // No default displacement otherwise Register can be implicitly
 209   // converted to 0(Register) which is quite a different animal.
 210 
 211   Address(Register base, int disp)
 212     : _base(base),
 213       _index(noreg),
 214       _scale(no_scale),
 215       _disp(disp) {
 216   }
 217 
 218   Address(Register base, Register index, ScaleFactor scale, int disp = 0)
 219     : _base (base),
 220       _index(index),
 221       _scale(scale),
 222       _disp (disp) {
 223     assert(!index->is_valid() == (scale == Address::no_scale),
 224            "inconsistent address");
 225   }
 226 
 227   Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0)
 228     : _base (base),
 229       _index(index.register_or_noreg()),
 230       _scale(scale),
 231       _disp (disp + (index.constant_or_zero() * scale_size(scale))) {
 232     if (!index.is_register())  scale = Address::no_scale;
 233     assert(!_index->is_valid() == (scale == Address::no_scale),
 234            "inconsistent address");
 235   }
 236 
 237   Address plus_disp(int disp) const {
 238     Address a = (*this);
 239     a._disp += disp;
 240     return a;
 241   }
 242   Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const {
 243     Address a = (*this);
 244     a._disp += disp.constant_or_zero() * scale_size(scale);
 245     if (disp.is_register()) {
 246       assert(!a.index()->is_valid(), "competing indexes");
 247       a._index = disp.as_register();
 248       a._scale = scale;
 249     }
 250     return a;
 251   }
 252   bool is_same_address(Address a) const {
 253     // disregard _rspec
 254     return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale;
 255   }
 256 
 257   // The following two overloads are used in connection with the
 258   // ByteSize type (see sizes.hpp).  They simplify the use of
 259   // ByteSize'd arguments in assembly code. Note that their equivalent
 260   // for the optimized build are the member functions with int disp
 261   // argument since ByteSize is mapped to an int type in that case.
 262   //
 263   // Note: DO NOT introduce similar overloaded functions for WordSize
 264   // arguments as in the optimized mode, both ByteSize and WordSize
 265   // are mapped to the same type and thus the compiler cannot make a
 266   // distinction anymore (=> compiler errors).
 267 
 268 #ifdef ASSERT
 269   Address(Register base, ByteSize disp)
 270     : _base(base),
 271       _index(noreg),
 272       _scale(no_scale),
 273       _disp(in_bytes(disp)) {
 274   }
 275 
 276   Address(Register base, Register index, ScaleFactor scale, ByteSize disp)
 277     : _base(base),
 278       _index(index),
 279       _scale(scale),
 280       _disp(in_bytes(disp)) {
 281     assert(!index->is_valid() == (scale == Address::no_scale),
 282            "inconsistent address");
 283   }
 284 
 285   Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp)
 286     : _base (base),
 287       _index(index.register_or_noreg()),
 288       _scale(scale),
 289       _disp (in_bytes(disp) + (index.constant_or_zero() * scale_size(scale))) {
 290     if (!index.is_register())  scale = Address::no_scale;
 291     assert(!_index->is_valid() == (scale == Address::no_scale),
 292            "inconsistent address");
 293   }
 294 
 295 #endif // ASSERT
 296 
 297   // accessors
 298   bool        uses(Register reg) const { return _base == reg || _index == reg; }
 299   Register    base()             const { return _base;  }
 300   Register    index()            const { return _index; }
 301   ScaleFactor scale()            const { return _scale; }
 302   int         disp()             const { return _disp;  }
 303 
 304   // Convert the raw encoding form into the form expected by the constructor for
 305   // Address.  An index of 4 (rsp) corresponds to having no index, so convert
 306   // that to noreg for the Address constructor.
 307   static Address make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc);
 308 
 309   static Address make_array(ArrayAddress);
 310 
 311  private:
 312   bool base_needs_rex() const {
 313     return _base != noreg && _base->encoding() >= 8;
 314   }
 315 
 316   bool index_needs_rex() const {
 317     return _index != noreg &&_index->encoding() >= 8;
 318   }
 319 
 320   relocInfo::relocType reloc() const { return _rspec.type(); }
 321 
 322   friend class Assembler;
 323   friend class MacroAssembler;
 324   friend class LIR_Assembler; // base/index/scale/disp
 325 };
 326 
 327 //
 328 // AddressLiteral has been split out from Address because operands of this type
 329 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out
 330 // the few instructions that need to deal with address literals are unique and the
 331 // MacroAssembler does not have to implement every instruction in the Assembler
 332 // in order to search for address literals that may need special handling depending
 333 // on the instruction and the platform. As small step on the way to merging i486/amd64
 334 // directories.
 335 //
 336 class AddressLiteral VALUE_OBJ_CLASS_SPEC {
 337   friend class ArrayAddress;
 338   RelocationHolder _rspec;
 339   // Typically we use AddressLiterals we want to use their rval
 340   // However in some situations we want the lval (effect address) of the item.
 341   // We provide a special factory for making those lvals.
 342   bool _is_lval;
 343 
 344   // If the target is far we'll need to load the ea of this to
 345   // a register to reach it. Otherwise if near we can do rip
 346   // relative addressing.
 347 
 348   address          _target;
 349 
 350  protected:
 351   // creation
 352   AddressLiteral()
 353     : _is_lval(false),
 354       _target(NULL)
 355   {}
 356 
 357   public:
 358 
 359 
 360   AddressLiteral(address target, relocInfo::relocType rtype);
 361 
 362   AddressLiteral(address target, RelocationHolder const& rspec)
 363     : _rspec(rspec),
 364       _is_lval(false),
 365       _target(target)
 366   {}
 367 
 368   AddressLiteral addr() {
 369     AddressLiteral ret = *this;
 370     ret._is_lval = true;
 371     return ret;
 372   }
 373 
 374 
 375  private:
 376 
 377   address target() { return _target; }
 378   bool is_lval() { return _is_lval; }
 379 
 380   relocInfo::relocType reloc() const { return _rspec.type(); }
 381   const RelocationHolder& rspec() const { return _rspec; }
 382 
 383   friend class Assembler;
 384   friend class MacroAssembler;
 385   friend class Address;
 386   friend class LIR_Assembler;
 387 };
 388 
 389 // Convience classes
 390 class RuntimeAddress: public AddressLiteral {
 391 
 392   public:
 393 
 394   RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {}
 395 
 396 };
 397 
 398 class ExternalAddress: public AddressLiteral {
 399  private:
 400   static relocInfo::relocType reloc_for_target(address target) {
 401     // Sometimes ExternalAddress is used for values which aren't
 402     // exactly addresses, like the card table base.
 403     // external_word_type can't be used for values in the first page
 404     // so just skip the reloc in that case.
 405     return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none;
 406   }
 407 
 408  public:
 409 
 410   ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {}
 411 
 412 };
 413 
 414 class InternalAddress: public AddressLiteral {
 415 
 416   public:
 417 
 418   InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {}
 419 
 420 };
 421 
 422 // x86 can do array addressing as a single operation since disp can be an absolute
 423 // address amd64 can't. We create a class that expresses the concept but does extra
 424 // magic on amd64 to get the final result
 425 
 426 class ArrayAddress VALUE_OBJ_CLASS_SPEC {
 427   private:
 428 
 429   AddressLiteral _base;
 430   Address        _index;
 431 
 432   public:
 433 
 434   ArrayAddress() {};
 435   ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {};
 436   AddressLiteral base() { return _base; }
 437   Address index() { return _index; }
 438 
 439 };
 440 
 441 class InstructionAttr;
 442 
 443 // 64-bit refect the fxsave size which is 512 bytes and the new xsave area on EVEX which is another 2176 bytes
 444 // See fxsave and xsave(EVEX enabled) documentation for layout
 445 const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY(2688 / wordSize);
 446 
 447 // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction
 448 // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write
 449 // is what you get. The Assembler is generating code into a CodeBuffer.
 450 
 451 class Assembler : public AbstractAssembler  {
 452   friend class AbstractAssembler; // for the non-virtual hack
 453   friend class LIR_Assembler; // as_Address()
 454   friend class StubGenerator;
 455 
 456  public:
 457   enum Condition {                     // The x86 condition codes used for conditional jumps/moves.
 458     zero          = 0x4,
 459     notZero       = 0x5,
 460     equal         = 0x4,
 461     notEqual      = 0x5,
 462     less          = 0xc,
 463     lessEqual     = 0xe,
 464     greater       = 0xf,
 465     greaterEqual  = 0xd,
 466     below         = 0x2,
 467     belowEqual    = 0x6,
 468     above         = 0x7,
 469     aboveEqual    = 0x3,
 470     overflow      = 0x0,
 471     noOverflow    = 0x1,
 472     carrySet      = 0x2,
 473     carryClear    = 0x3,
 474     negative      = 0x8,
 475     positive      = 0x9,
 476     parity        = 0xa,
 477     noParity      = 0xb
 478   };
 479 
 480   enum Prefix {
 481     // segment overrides
 482     CS_segment = 0x2e,
 483     SS_segment = 0x36,
 484     DS_segment = 0x3e,
 485     ES_segment = 0x26,
 486     FS_segment = 0x64,
 487     GS_segment = 0x65,
 488 
 489     REX        = 0x40,
 490 
 491     REX_B      = 0x41,
 492     REX_X      = 0x42,
 493     REX_XB     = 0x43,
 494     REX_R      = 0x44,
 495     REX_RB     = 0x45,
 496     REX_RX     = 0x46,
 497     REX_RXB    = 0x47,
 498 
 499     REX_W      = 0x48,
 500 
 501     REX_WB     = 0x49,
 502     REX_WX     = 0x4A,
 503     REX_WXB    = 0x4B,
 504     REX_WR     = 0x4C,
 505     REX_WRB    = 0x4D,
 506     REX_WRX    = 0x4E,
 507     REX_WRXB   = 0x4F,
 508 
 509     VEX_3bytes = 0xC4,
 510     VEX_2bytes = 0xC5,
 511     EVEX_4bytes = 0x62,
 512     Prefix_EMPTY = 0x0
 513   };
 514 
 515   enum VexPrefix {
 516     VEX_B = 0x20,
 517     VEX_X = 0x40,
 518     VEX_R = 0x80,
 519     VEX_W = 0x80
 520   };
 521 
 522   enum ExexPrefix {
 523     EVEX_F  = 0x04,
 524     EVEX_V  = 0x08,
 525     EVEX_Rb = 0x10,
 526     EVEX_X  = 0x40,
 527     EVEX_Z  = 0x80
 528   };
 529 
 530   enum VexSimdPrefix {
 531     VEX_SIMD_NONE = 0x0,
 532     VEX_SIMD_66   = 0x1,
 533     VEX_SIMD_F3   = 0x2,
 534     VEX_SIMD_F2   = 0x3
 535   };
 536 
 537   enum VexOpcode {
 538     VEX_OPCODE_NONE  = 0x0,
 539     VEX_OPCODE_0F    = 0x1,
 540     VEX_OPCODE_0F_38 = 0x2,
 541     VEX_OPCODE_0F_3A = 0x3,
 542     VEX_OPCODE_MASK  = 0x1F
 543   };
 544 
 545   enum AvxVectorLen {
 546     AVX_128bit = 0x0,
 547     AVX_256bit = 0x1,
 548     AVX_512bit = 0x2,
 549     AVX_NoVec  = 0x4
 550   };
 551 
 552   enum EvexTupleType {
 553     EVEX_FV   = 0,
 554     EVEX_HV   = 4,
 555     EVEX_FVM  = 6,
 556     EVEX_T1S  = 7,
 557     EVEX_T1F  = 11,
 558     EVEX_T2   = 13,
 559     EVEX_T4   = 15,
 560     EVEX_T8   = 17,
 561     EVEX_HVM  = 18,
 562     EVEX_QVM  = 19,
 563     EVEX_OVM  = 20,
 564     EVEX_M128 = 21,
 565     EVEX_DUP  = 22,
 566     EVEX_ETUP = 23
 567   };
 568 
 569   enum EvexInputSizeInBits {
 570     EVEX_8bit  = 0,
 571     EVEX_16bit = 1,
 572     EVEX_32bit = 2,
 573     EVEX_64bit = 3,
 574     EVEX_NObit = 4
 575   };
 576 
 577   enum WhichOperand {
 578     // input to locate_operand, and format code for relocations
 579     imm_operand  = 0,            // embedded 32-bit|64-bit immediate operand
 580     disp32_operand = 1,          // embedded 32-bit displacement or address
 581     call32_operand = 2,          // embedded 32-bit self-relative displacement
 582 #ifndef _LP64
 583     _WhichOperand_limit = 3
 584 #else
 585      narrow_oop_operand = 3,     // embedded 32-bit immediate narrow oop
 586     _WhichOperand_limit = 4
 587 #endif
 588   };
 589 
 590   enum ComparisonPredicate {
 591     eq = 0,
 592     lt = 1,
 593     le = 2,
 594     _false = 3,
 595     neq = 4,
 596     nlt = 5,
 597     nle = 6,
 598     _true = 7
 599   };
 600 
 601 
 602   // NOTE: The general philopsophy of the declarations here is that 64bit versions
 603   // of instructions are freely declared without the need for wrapping them an ifdef.
 604   // (Some dangerous instructions are ifdef's out of inappropriate jvm's.)
 605   // In the .cpp file the implementations are wrapped so that they are dropped out
 606   // of the resulting jvm. This is done mostly to keep the footprint of MINIMAL
 607   // to the size it was prior to merging up the 32bit and 64bit assemblers.
 608   //
 609   // This does mean you'll get a linker/runtime error if you use a 64bit only instruction
 610   // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down.
 611 
 612 private:
 613 
 614   bool _legacy_mode_bw;
 615   bool _legacy_mode_dq;
 616   bool _legacy_mode_vl;
 617   bool _legacy_mode_vlbw;
 618   bool _is_managed;
 619   bool _vector_masking;    // For stub code use only
 620 
 621   class InstructionAttr *_attributes;
 622 
 623   // 64bit prefixes
 624   int prefix_and_encode(int reg_enc, bool byteinst = false);
 625   int prefixq_and_encode(int reg_enc);
 626 
 627   int prefix_and_encode(int dst_enc, int src_enc) {
 628     return prefix_and_encode(dst_enc, false, src_enc, false);
 629   }
 630   int prefix_and_encode(int dst_enc, bool dst_is_byte, int src_enc, bool src_is_byte);
 631   int prefixq_and_encode(int dst_enc, int src_enc);
 632 
 633   void prefix(Register reg);
 634   void prefix(Register dst, Register src, Prefix p);
 635   void prefix(Register dst, Address adr, Prefix p);
 636   void prefix(Address adr);
 637   void prefixq(Address adr);
 638 
 639   void prefix(Address adr, Register reg,  bool byteinst = false);
 640   void prefix(Address adr, XMMRegister reg);
 641   void prefixq(Address adr, Register reg);
 642   void prefixq(Address adr, XMMRegister reg);
 643 
 644   void prefetch_prefix(Address src);
 645 
 646   void rex_prefix(Address adr, XMMRegister xreg,
 647                   VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 648   int  rex_prefix_and_encode(int dst_enc, int src_enc,
 649                              VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 650 
 651   void vex_prefix(bool vex_r, bool vex_b, bool vex_x, int nds_enc, VexSimdPrefix pre, VexOpcode opc);
 652 
 653   void evex_prefix(bool vex_r, bool vex_b, bool vex_x, bool evex_r, bool evex_v,
 654                    int nds_enc, VexSimdPrefix pre, VexOpcode opc);
 655 
 656   void vex_prefix(Address adr, int nds_enc, int xreg_enc,
 657                   VexSimdPrefix pre, VexOpcode opc,
 658                   InstructionAttr *attributes);
 659 
 660   int  vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc,
 661                              VexSimdPrefix pre, VexOpcode opc,
 662                              InstructionAttr *attributes);
 663 
 664   void simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre,
 665                    VexOpcode opc, InstructionAttr *attributes);
 666 
 667   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPrefix pre,
 668                              VexOpcode opc, InstructionAttr *attributes);
 669 
 670   // Helper functions for groups of instructions
 671   void emit_arith_b(int op1, int op2, Register dst, int imm8);
 672 
 673   void emit_arith(int op1, int op2, Register dst, int32_t imm32);
 674   // Force generation of a 4 byte immediate value even if it fits into 8bit
 675   void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32);
 676   void emit_arith(int op1, int op2, Register dst, Register src);
 677 
 678   bool emit_compressed_disp_byte(int &disp);
 679 
 680   void emit_operand(Register reg,
 681                     Register base, Register index, Address::ScaleFactor scale,
 682                     int disp,
 683                     RelocationHolder const& rspec,
 684                     int rip_relative_correction = 0);
 685 
 686   void emit_operand(Register reg, Address adr, int rip_relative_correction = 0);
 687 
 688   // operands that only take the original 32bit registers
 689   void emit_operand32(Register reg, Address adr);
 690 
 691   void emit_operand(XMMRegister reg,
 692                     Register base, Register index, Address::ScaleFactor scale,
 693                     int disp,
 694                     RelocationHolder const& rspec);
 695 
 696   void emit_operand(XMMRegister reg, Address adr);
 697 
 698   void emit_operand(MMXRegister reg, Address adr);
 699 
 700   // workaround gcc (3.2.1-7) bug
 701   void emit_operand(Address adr, MMXRegister reg);
 702 
 703 
 704   // Immediate-to-memory forms
 705   void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32);
 706 
 707   void emit_farith(int b1, int b2, int i);
 708 
 709 
 710  protected:
 711   #ifdef ASSERT
 712   void check_relocation(RelocationHolder const& rspec, int format);
 713   #endif
 714 
 715   void emit_data(jint data, relocInfo::relocType    rtype, int format);
 716   void emit_data(jint data, RelocationHolder const& rspec, int format);
 717   void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0);
 718   void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0);
 719 
 720   bool reachable(AddressLiteral adr) NOT_LP64({ return true;});
 721 
 722   // These are all easily abused and hence protected
 723 
 724   // 32BIT ONLY SECTION
 725 #ifndef _LP64
 726   // Make these disappear in 64bit mode since they would never be correct
 727   void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec);   // 32BIT ONLY
 728   void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 729 
 730   void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 731   void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec);     // 32BIT ONLY
 732 
 733   void push_literal32(int32_t imm32, RelocationHolder const& rspec);                 // 32BIT ONLY
 734 #else
 735   // 64BIT ONLY SECTION
 736   void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec);   // 64BIT ONLY
 737 
 738   void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec);
 739   void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec);
 740 
 741   void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec);
 742   void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec);
 743 #endif // _LP64
 744 
 745   // These are unique in that we are ensured by the caller that the 32bit
 746   // relative in these instructions will always be able to reach the potentially
 747   // 64bit address described by entry. Since they can take a 64bit address they
 748   // don't have the 32 suffix like the other instructions in this class.
 749 
 750   void call_literal(address entry, RelocationHolder const& rspec);
 751   void jmp_literal(address entry, RelocationHolder const& rspec);
 752 
 753   // Avoid using directly section
 754   // Instructions in this section are actually usable by anyone without danger
 755   // of failure but have performance issues that are addressed my enhanced
 756   // instructions which will do the proper thing base on the particular cpu.
 757   // We protect them because we don't trust you...
 758 
 759   // Don't use next inc() and dec() methods directly. INC & DEC instructions
 760   // could cause a partial flag stall since they don't set CF flag.
 761   // Use MacroAssembler::decrement() & MacroAssembler::increment() methods
 762   // which call inc() & dec() or add() & sub() in accordance with
 763   // the product flag UseIncDec value.
 764 
 765   void decl(Register dst);
 766   void decl(Address dst);
 767   void decq(Register dst);
 768   void decq(Address dst);
 769 
 770   void incl(Register dst);
 771   void incl(Address dst);
 772   void incq(Register dst);
 773   void incq(Address dst);
 774 
 775   // New cpus require use of movsd and movss to avoid partial register stall
 776   // when loading from memory. But for old Opteron use movlpd instead of movsd.
 777   // The selection is done in MacroAssembler::movdbl() and movflt().
 778 
 779   // Move Scalar Single-Precision Floating-Point Values
 780   void movss(XMMRegister dst, Address src);
 781   void movss(XMMRegister dst, XMMRegister src);
 782   void movss(Address dst, XMMRegister src);
 783 
 784   // Move Scalar Double-Precision Floating-Point Values
 785   void movsd(XMMRegister dst, Address src);
 786   void movsd(XMMRegister dst, XMMRegister src);
 787   void movsd(Address dst, XMMRegister src);
 788   void movlpd(XMMRegister dst, Address src);
 789 
 790   // New cpus require use of movaps and movapd to avoid partial register stall
 791   // when moving between registers.
 792   void movaps(XMMRegister dst, XMMRegister src);
 793   void movapd(XMMRegister dst, XMMRegister src);
 794 
 795   // End avoid using directly
 796 
 797 
 798   // Instruction prefixes
 799   void prefix(Prefix p);
 800 
 801   public:
 802 
 803   // Creation
 804   Assembler(CodeBuffer* code) : AbstractAssembler(code) {
 805     init_attributes();
 806   }
 807 
 808   // Decoding
 809   static address locate_operand(address inst, WhichOperand which);
 810   static address locate_next_instruction(address inst);
 811 
 812   // Utilities
 813   static bool is_polling_page_far() NOT_LP64({ return false;});
 814   static bool query_compressed_disp_byte(int disp, bool is_evex_inst, int vector_len,
 815                                          int cur_tuple_type, int in_size_in_bits, int cur_encoding);
 816 
 817   // Generic instructions
 818   // Does 32bit or 64bit as needed for the platform. In some sense these
 819   // belong in macro assembler but there is no need for both varieties to exist
 820 
 821   void init_attributes(void) {
 822     _legacy_mode_bw = (VM_Version::supports_avx512bw() == false);
 823     _legacy_mode_dq = (VM_Version::supports_avx512dq() == false);
 824     _legacy_mode_vl = (VM_Version::supports_avx512vl() == false);
 825     _legacy_mode_vlbw = (VM_Version::supports_avx512vlbw() == false);
 826     _is_managed = false;
 827     _vector_masking = false;
 828     _attributes = NULL;
 829   }
 830 
 831   void set_attributes(InstructionAttr *attributes) { _attributes = attributes; }
 832   void clear_attributes(void) { _attributes = NULL; }
 833 
 834   void set_managed(void) { _is_managed = true; }
 835   void clear_managed(void) { _is_managed = false; }
 836   bool is_managed(void) { return _is_managed; }
 837 
 838   // Following functions are for stub code use only
 839   void set_vector_masking(void) { _vector_masking = true; }
 840   void clear_vector_masking(void) { _vector_masking = false; }
 841   bool is_vector_masking(void) { return _vector_masking; }
 842 
 843   void lea(Register dst, Address src);
 844 
 845   void mov(Register dst, Register src);
 846 
 847   void pusha();
 848   void popa();
 849 
 850   void pushf();
 851   void popf();
 852 
 853   void push(int32_t imm32);
 854 
 855   void push(Register src);
 856 
 857   void pop(Register dst);
 858 
 859   // These are dummies to prevent surprise implicit conversions to Register
 860   void push(void* v);
 861   void pop(void* v);
 862 
 863   // These do register sized moves/scans
 864   void rep_mov();
 865   void rep_stos();
 866   void rep_stosb();
 867   void repne_scan();
 868 #ifdef _LP64
 869   void repne_scanl();
 870 #endif
 871 
 872   // Vanilla instructions in lexical order
 873 
 874   void adcl(Address dst, int32_t imm32);
 875   void adcl(Address dst, Register src);
 876   void adcl(Register dst, int32_t imm32);
 877   void adcl(Register dst, Address src);
 878   void adcl(Register dst, Register src);
 879 
 880   void adcq(Register dst, int32_t imm32);
 881   void adcq(Register dst, Address src);
 882   void adcq(Register dst, Register src);
 883 
 884   void addb(Address dst, int imm8);
 885   void addw(Address dst, int imm16);
 886 
 887   void addl(Address dst, int32_t imm32);
 888   void addl(Address dst, Register src);
 889   void addl(Register dst, int32_t imm32);
 890   void addl(Register dst, Address src);
 891   void addl(Register dst, Register src);
 892 
 893   void addq(Address dst, int32_t imm32);
 894   void addq(Address dst, Register src);
 895   void addq(Register dst, int32_t imm32);
 896   void addq(Register dst, Address src);
 897   void addq(Register dst, Register src);
 898 
 899 #ifdef _LP64
 900  //Add Unsigned Integers with Carry Flag
 901   void adcxq(Register dst, Register src);
 902 
 903  //Add Unsigned Integers with Overflow Flag
 904   void adoxq(Register dst, Register src);
 905 #endif
 906 
 907   void addr_nop_4();
 908   void addr_nop_5();
 909   void addr_nop_7();
 910   void addr_nop_8();
 911 
 912   // Add Scalar Double-Precision Floating-Point Values
 913   void addsd(XMMRegister dst, Address src);
 914   void addsd(XMMRegister dst, XMMRegister src);
 915 
 916   // Add Scalar Single-Precision Floating-Point Values
 917   void addss(XMMRegister dst, Address src);
 918   void addss(XMMRegister dst, XMMRegister src);
 919 
 920   // AES instructions
 921   void aesdec(XMMRegister dst, Address src);
 922   void aesdec(XMMRegister dst, XMMRegister src);
 923   void aesdeclast(XMMRegister dst, Address src);
 924   void aesdeclast(XMMRegister dst, XMMRegister src);
 925   void aesenc(XMMRegister dst, Address src);
 926   void aesenc(XMMRegister dst, XMMRegister src);
 927   void aesenclast(XMMRegister dst, Address src);
 928   void aesenclast(XMMRegister dst, XMMRegister src);
 929 
 930 
 931   void andl(Address  dst, int32_t imm32);
 932   void andl(Register dst, int32_t imm32);
 933   void andl(Register dst, Address src);
 934   void andl(Register dst, Register src);
 935 
 936   void andq(Address  dst, int32_t imm32);
 937   void andq(Register dst, int32_t imm32);
 938   void andq(Register dst, Address src);
 939   void andq(Register dst, Register src);
 940 
 941   // BMI instructions
 942   void andnl(Register dst, Register src1, Register src2);
 943   void andnl(Register dst, Register src1, Address src2);
 944   void andnq(Register dst, Register src1, Register src2);
 945   void andnq(Register dst, Register src1, Address src2);
 946 
 947   void blsil(Register dst, Register src);
 948   void blsil(Register dst, Address src);
 949   void blsiq(Register dst, Register src);
 950   void blsiq(Register dst, Address src);
 951 
 952   void blsmskl(Register dst, Register src);
 953   void blsmskl(Register dst, Address src);
 954   void blsmskq(Register dst, Register src);
 955   void blsmskq(Register dst, Address src);
 956 
 957   void blsrl(Register dst, Register src);
 958   void blsrl(Register dst, Address src);
 959   void blsrq(Register dst, Register src);
 960   void blsrq(Register dst, Address src);
 961 
 962   void bsfl(Register dst, Register src);
 963   void bsrl(Register dst, Register src);
 964 
 965 #ifdef _LP64
 966   void bsfq(Register dst, Register src);
 967   void bsrq(Register dst, Register src);
 968 #endif
 969 
 970   void bswapl(Register reg);
 971 
 972   void bswapq(Register reg);
 973 
 974   void call(Label& L, relocInfo::relocType rtype);
 975   void call(Register reg);  // push pc; pc <- reg
 976   void call(Address adr);   // push pc; pc <- adr
 977 
 978   void cdql();
 979 
 980   void cdqq();
 981 
 982   void cld();
 983 
 984   void clflush(Address adr);
 985 
 986   void cmovl(Condition cc, Register dst, Register src);
 987   void cmovl(Condition cc, Register dst, Address src);
 988 
 989   void cmovq(Condition cc, Register dst, Register src);
 990   void cmovq(Condition cc, Register dst, Address src);
 991 
 992 
 993   void cmpb(Address dst, int imm8);
 994 
 995   void cmpl(Address dst, int32_t imm32);
 996 
 997   void cmpl(Register dst, int32_t imm32);
 998   void cmpl(Register dst, Register src);
 999   void cmpl(Register dst, Address src);
1000 
1001   void cmpq(Address dst, int32_t imm32);
1002   void cmpq(Address dst, Register src);
1003 
1004   void cmpq(Register dst, int32_t imm32);
1005   void cmpq(Register dst, Register src);
1006   void cmpq(Register dst, Address src);
1007 
1008   // these are dummies used to catch attempting to convert NULL to Register
1009   void cmpl(Register dst, void* junk); // dummy
1010   void cmpq(Register dst, void* junk); // dummy
1011 
1012   void cmpw(Address dst, int imm16);
1013 
1014   void cmpxchg8 (Address adr);
1015 
1016   void cmpxchgb(Register reg, Address adr);
1017   void cmpxchgl(Register reg, Address adr);
1018 
1019   void cmpxchgq(Register reg, Address adr);
1020 
1021   // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1022   void comisd(XMMRegister dst, Address src);
1023   void comisd(XMMRegister dst, XMMRegister src);
1024 
1025   // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1026   void comiss(XMMRegister dst, Address src);
1027   void comiss(XMMRegister dst, XMMRegister src);
1028 
1029   // Identify processor type and features
1030   void cpuid();
1031 
1032   // CRC32C
1033   void crc32(Register crc, Register v, int8_t sizeInBytes);
1034   void crc32(Register crc, Address adr, int8_t sizeInBytes);
1035 
1036   // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floating-Point Value
1037   void cvtsd2ss(XMMRegister dst, XMMRegister src);
1038   void cvtsd2ss(XMMRegister dst, Address src);
1039 
1040   // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value
1041   void cvtsi2sdl(XMMRegister dst, Register src);
1042   void cvtsi2sdl(XMMRegister dst, Address src);
1043   void cvtsi2sdq(XMMRegister dst, Register src);
1044   void cvtsi2sdq(XMMRegister dst, Address src);
1045 
1046   // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value
1047   void cvtsi2ssl(XMMRegister dst, Register src);
1048   void cvtsi2ssl(XMMRegister dst, Address src);
1049   void cvtsi2ssq(XMMRegister dst, Register src);
1050   void cvtsi2ssq(XMMRegister dst, Address src);
1051 
1052   // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Value
1053   void cvtdq2pd(XMMRegister dst, XMMRegister src);
1054 
1055   // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Value
1056   void cvtdq2ps(XMMRegister dst, XMMRegister src);
1057 
1058   // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floating-Point Value
1059   void cvtss2sd(XMMRegister dst, XMMRegister src);
1060   void cvtss2sd(XMMRegister dst, Address src);
1061 
1062   // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Integer
1063   void cvttsd2sil(Register dst, Address src);
1064   void cvttsd2sil(Register dst, XMMRegister src);
1065   void cvttsd2siq(Register dst, XMMRegister src);
1066 
1067   // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Integer
1068   void cvttss2sil(Register dst, XMMRegister src);
1069   void cvttss2siq(Register dst, XMMRegister src);
1070 
1071   void cvttpd2dq(XMMRegister dst, XMMRegister src);
1072 
1073   // Divide Scalar Double-Precision Floating-Point Values
1074   void divsd(XMMRegister dst, Address src);
1075   void divsd(XMMRegister dst, XMMRegister src);
1076 
1077   // Divide Scalar Single-Precision Floating-Point Values
1078   void divss(XMMRegister dst, Address src);
1079   void divss(XMMRegister dst, XMMRegister src);
1080 
1081   void emms();
1082 
1083   void fabs();
1084 
1085   void fadd(int i);
1086 
1087   void fadd_d(Address src);
1088   void fadd_s(Address src);
1089 
1090   // "Alternate" versions of x87 instructions place result down in FPU
1091   // stack instead of on TOS
1092 
1093   void fadda(int i); // "alternate" fadd
1094   void faddp(int i = 1);
1095 
1096   void fchs();
1097 
1098   void fcom(int i);
1099 
1100   void fcomp(int i = 1);
1101   void fcomp_d(Address src);
1102   void fcomp_s(Address src);
1103 
1104   void fcompp();
1105 
1106   void fcos();
1107 
1108   void fdecstp();
1109 
1110   void fdiv(int i);
1111   void fdiv_d(Address src);
1112   void fdivr_s(Address src);
1113   void fdiva(int i);  // "alternate" fdiv
1114   void fdivp(int i = 1);
1115 
1116   void fdivr(int i);
1117   void fdivr_d(Address src);
1118   void fdiv_s(Address src);
1119 
1120   void fdivra(int i); // "alternate" reversed fdiv
1121 
1122   void fdivrp(int i = 1);
1123 
1124   void ffree(int i = 0);
1125 
1126   void fild_d(Address adr);
1127   void fild_s(Address adr);
1128 
1129   void fincstp();
1130 
1131   void finit();
1132 
1133   void fist_s (Address adr);
1134   void fistp_d(Address adr);
1135   void fistp_s(Address adr);
1136 
1137   void fld1();
1138 
1139   void fld_d(Address adr);
1140   void fld_s(Address adr);
1141   void fld_s(int index);
1142   void fld_x(Address adr);  // extended-precision (80-bit) format
1143 
1144   void fldcw(Address src);
1145 
1146   void fldenv(Address src);
1147 
1148   void fldlg2();
1149 
1150   void fldln2();
1151 
1152   void fldz();
1153 
1154   void flog();
1155   void flog10();
1156 
1157   void fmul(int i);
1158 
1159   void fmul_d(Address src);
1160   void fmul_s(Address src);
1161 
1162   void fmula(int i);  // "alternate" fmul
1163 
1164   void fmulp(int i = 1);
1165 
1166   void fnsave(Address dst);
1167 
1168   void fnstcw(Address src);
1169 
1170   void fnstsw_ax();
1171 
1172   void fprem();
1173   void fprem1();
1174 
1175   void frstor(Address src);
1176 
1177   void fsin();
1178 
1179   void fsqrt();
1180 
1181   void fst_d(Address adr);
1182   void fst_s(Address adr);
1183 
1184   void fstp_d(Address adr);
1185   void fstp_d(int index);
1186   void fstp_s(Address adr);
1187   void fstp_x(Address adr); // extended-precision (80-bit) format
1188 
1189   void fsub(int i);
1190   void fsub_d(Address src);
1191   void fsub_s(Address src);
1192 
1193   void fsuba(int i);  // "alternate" fsub
1194 
1195   void fsubp(int i = 1);
1196 
1197   void fsubr(int i);
1198   void fsubr_d(Address src);
1199   void fsubr_s(Address src);
1200 
1201   void fsubra(int i); // "alternate" reversed fsub
1202 
1203   void fsubrp(int i = 1);
1204 
1205   void ftan();
1206 
1207   void ftst();
1208 
1209   void fucomi(int i = 1);
1210   void fucomip(int i = 1);
1211 
1212   void fwait();
1213 
1214   void fxch(int i = 1);
1215 
1216   void fxrstor(Address src);
1217   void xrstor(Address src);
1218 
1219   void fxsave(Address dst);
1220   void xsave(Address dst);
1221 
1222   void fyl2x();
1223   void frndint();
1224   void f2xm1();
1225   void fldl2e();
1226 
1227   void hlt();
1228 
1229   void idivl(Register src);
1230   void divl(Register src); // Unsigned division
1231 
1232 #ifdef _LP64
1233   void idivq(Register src);
1234 #endif
1235 
1236   void imull(Register src);
1237   void imull(Register dst, Register src);
1238   void imull(Register dst, Register src, int value);
1239   void imull(Register dst, Address src);
1240 
1241 #ifdef _LP64
1242   void imulq(Register dst, Register src);
1243   void imulq(Register dst, Register src, int value);
1244   void imulq(Register dst, Address src);
1245 #endif
1246 
1247   // jcc is the generic conditional branch generator to run-
1248   // time routines, jcc is used for branches to labels. jcc
1249   // takes a branch opcode (cc) and a label (L) and generates
1250   // either a backward branch or a forward branch and links it
1251   // to the label fixup chain. Usage:
1252   //
1253   // Label L;      // unbound label
1254   // jcc(cc, L);   // forward branch to unbound label
1255   // bind(L);      // bind label to the current pc
1256   // jcc(cc, L);   // backward branch to bound label
1257   // bind(L);      // illegal: a label may be bound only once
1258   //
1259   // Note: The same Label can be used for forward and backward branches
1260   // but it may be bound only once.
1261 
1262   void jcc(Condition cc, Label& L, bool maybe_short = true);
1263 
1264   // Conditional jump to a 8-bit offset to L.
1265   // WARNING: be very careful using this for forward jumps.  If the label is
1266   // not bound within an 8-bit offset of this instruction, a run-time error
1267   // will occur.
1268   void jccb(Condition cc, Label& L);
1269 
1270   void jmp(Address entry);    // pc <- entry
1271 
1272   // Label operations & relative jumps (PPUM Appendix D)
1273   void jmp(Label& L, bool maybe_short = true);   // unconditional jump to L
1274 
1275   void jmp(Register entry); // pc <- entry
1276 
1277   // Unconditional 8-bit offset jump to L.
1278   // WARNING: be very careful using this for forward jumps.  If the label is
1279   // not bound within an 8-bit offset of this instruction, a run-time error
1280   // will occur.
1281   void jmpb(Label& L);
1282 
1283   void ldmxcsr( Address src );
1284 
1285   void leal(Register dst, Address src);
1286 
1287   void leaq(Register dst, Address src);
1288 
1289   void lfence();
1290 
1291   void lock();
1292 
1293   void lzcntl(Register dst, Register src);
1294 
1295 #ifdef _LP64
1296   void lzcntq(Register dst, Register src);
1297 #endif
1298 
1299   enum Membar_mask_bits {
1300     StoreStore = 1 << 3,
1301     LoadStore  = 1 << 2,
1302     StoreLoad  = 1 << 1,
1303     LoadLoad   = 1 << 0
1304   };
1305 
1306   // Serializes memory and blows flags
1307   void membar(Membar_mask_bits order_constraint) {
1308     if (os::is_MP()) {
1309       // We only have to handle StoreLoad
1310       if (order_constraint & StoreLoad) {
1311         // All usable chips support "locked" instructions which suffice
1312         // as barriers, and are much faster than the alternative of
1313         // using cpuid instruction. We use here a locked add [esp-C],0.
1314         // This is conveniently otherwise a no-op except for blowing
1315         // flags, and introducing a false dependency on target memory
1316         // location. We can't do anything with flags, but we can avoid
1317         // memory dependencies in the current method by locked-adding
1318         // somewhere else on the stack. Doing [esp+C] will collide with
1319         // something on stack in current method, hence we go for [esp-C].
1320         // It is convenient since it is almost always in data cache, for
1321         // any small C.  We need to step back from SP to avoid data
1322         // dependencies with other things on below SP (callee-saves, for
1323         // example). Without a clear way to figure out the minimal safe
1324         // distance from SP, it makes sense to step back the complete
1325         // cache line, as this will also avoid possible second-order effects
1326         // with locked ops against the cache line. Our choice of offset
1327         // is bounded by x86 operand encoding, which should stay within
1328         // [-128; +127] to have the 8-byte displacement encoding.
1329         //
1330         // Any change to this code may need to revisit other places in
1331         // the code where this idiom is used, in particular the
1332         // orderAccess code.
1333 
1334         int offset = -VM_Version::L1_line_size();
1335         if (offset < -128) {
1336           offset = -128;
1337         }
1338 
1339         lock();
1340         addl(Address(rsp, offset), 0);// Assert the lock# signal here
1341       }
1342     }
1343   }
1344 
1345   void mfence();
1346 
1347   // Moves
1348 
1349   void mov64(Register dst, int64_t imm64);
1350 
1351   void movb(Address dst, Register src);
1352   void movb(Address dst, int imm8);
1353   void movb(Register dst, Address src);
1354 
1355   void movddup(XMMRegister dst, XMMRegister src);
1356 
1357   void kmovbl(KRegister dst, Register src);
1358   void kmovbl(Register dst, KRegister src);
1359   void kmovwl(KRegister dst, Register src);
1360   void kmovwl(KRegister dst, Address src);
1361   void kmovwl(Register dst, KRegister src);
1362   void kmovdl(KRegister dst, Register src);
1363   void kmovdl(Register dst, KRegister src);
1364   void kmovql(KRegister dst, KRegister src);
1365   void kmovql(Address dst, KRegister src);
1366   void kmovql(KRegister dst, Address src);
1367   void kmovql(KRegister dst, Register src);
1368   void kmovql(Register dst, KRegister src);
1369 
1370   void knotwl(KRegister dst, KRegister src);
1371 
1372   void kortestbl(KRegister dst, KRegister src);
1373   void kortestwl(KRegister dst, KRegister src);
1374   void kortestdl(KRegister dst, KRegister src);
1375   void kortestql(KRegister dst, KRegister src);
1376 
1377   void ktestq(KRegister src1, KRegister src2);
1378   void ktestd(KRegister src1, KRegister src2);
1379 
1380   void ktestql(KRegister dst, KRegister src);
1381 
1382   void movdl(XMMRegister dst, Register src);
1383   void movdl(Register dst, XMMRegister src);
1384   void movdl(XMMRegister dst, Address src);
1385   void movdl(Address dst, XMMRegister src);
1386 
1387   // Move Double Quadword
1388   void movdq(XMMRegister dst, Register src);
1389   void movdq(Register dst, XMMRegister src);
1390 
1391   // Move Aligned Double Quadword
1392   void movdqa(XMMRegister dst, XMMRegister src);
1393   void movdqa(XMMRegister dst, Address src);
1394 
1395   // Move Unaligned Double Quadword
1396   void movdqu(Address     dst, XMMRegister src);
1397   void movdqu(XMMRegister dst, Address src);
1398   void movdqu(XMMRegister dst, XMMRegister src);
1399 
1400   // Move Unaligned 256bit Vector
1401   void vmovdqu(Address dst, XMMRegister src);
1402   void vmovdqu(XMMRegister dst, Address src);
1403   void vmovdqu(XMMRegister dst, XMMRegister src);
1404 
1405    // Move Unaligned 512bit Vector
1406   void evmovdqub(Address dst, XMMRegister src, int vector_len);
1407   void evmovdqub(XMMRegister dst, Address src, int vector_len);
1408   void evmovdqub(XMMRegister dst, XMMRegister src, int vector_len);
1409   void evmovdqub(XMMRegister dst, KRegister mask, Address src, int vector_len);
1410   void evmovdquw(Address dst, XMMRegister src, int vector_len);
1411   void evmovdquw(Address dst, KRegister mask, XMMRegister src, int vector_len);
1412   void evmovdquw(XMMRegister dst, Address src, int vector_len);
1413   void evmovdquw(XMMRegister dst, KRegister mask, Address src, int vector_len);
1414   void evmovdqul(Address dst, XMMRegister src, int vector_len);
1415   void evmovdqul(XMMRegister dst, Address src, int vector_len);
1416   void evmovdqul(XMMRegister dst, XMMRegister src, int vector_len);
1417   void evmovdquq(Address dst, XMMRegister src, int vector_len);
1418   void evmovdquq(XMMRegister dst, Address src, int vector_len);
1419   void evmovdquq(XMMRegister dst, XMMRegister src, int vector_len);
1420 
1421   // Move lower 64bit to high 64bit in 128bit register
1422   void movlhps(XMMRegister dst, XMMRegister src);
1423 
1424   void movl(Register dst, int32_t imm32);
1425   void movl(Address dst, int32_t imm32);
1426   void movl(Register dst, Register src);
1427   void movl(Register dst, Address src);
1428   void movl(Address dst, Register src);
1429 
1430   // These dummies prevent using movl from converting a zero (like NULL) into Register
1431   // by giving the compiler two choices it can't resolve
1432 
1433   void movl(Address  dst, void* junk);
1434   void movl(Register dst, void* junk);
1435 
1436 #ifdef _LP64
1437   void movq(Register dst, Register src);
1438   void movq(Register dst, Address src);
1439   void movq(Address  dst, Register src);
1440 #endif
1441 
1442   void movq(Address     dst, MMXRegister src );
1443   void movq(MMXRegister dst, Address src );
1444 
1445 #ifdef _LP64
1446   // These dummies prevent using movq from converting a zero (like NULL) into Register
1447   // by giving the compiler two choices it can't resolve
1448 
1449   void movq(Address  dst, void* dummy);
1450   void movq(Register dst, void* dummy);
1451 #endif
1452 
1453   // Move Quadword
1454   void movq(Address     dst, XMMRegister src);
1455   void movq(XMMRegister dst, Address src);
1456 
1457   void movsbl(Register dst, Address src);
1458   void movsbl(Register dst, Register src);
1459 
1460 #ifdef _LP64
1461   void movsbq(Register dst, Address src);
1462   void movsbq(Register dst, Register src);
1463 
1464   // Move signed 32bit immediate to 64bit extending sign
1465   void movslq(Address  dst, int32_t imm64);
1466   void movslq(Register dst, int32_t imm64);
1467 
1468   void movslq(Register dst, Address src);
1469   void movslq(Register dst, Register src);
1470   void movslq(Register dst, void* src); // Dummy declaration to cause NULL to be ambiguous
1471 #endif
1472 
1473   void movswl(Register dst, Address src);
1474   void movswl(Register dst, Register src);
1475 
1476 #ifdef _LP64
1477   void movswq(Register dst, Address src);
1478   void movswq(Register dst, Register src);
1479 #endif
1480 
1481   void movw(Address dst, int imm16);
1482   void movw(Register dst, Address src);
1483   void movw(Address dst, Register src);
1484 
1485   void movzbl(Register dst, Address src);
1486   void movzbl(Register dst, Register src);
1487 
1488 #ifdef _LP64
1489   void movzbq(Register dst, Address src);
1490   void movzbq(Register dst, Register src);
1491 #endif
1492 
1493   void movzwl(Register dst, Address src);
1494   void movzwl(Register dst, Register src);
1495 
1496 #ifdef _LP64
1497   void movzwq(Register dst, Address src);
1498   void movzwq(Register dst, Register src);
1499 #endif
1500 
1501   // Unsigned multiply with RAX destination register
1502   void mull(Address src);
1503   void mull(Register src);
1504 
1505 #ifdef _LP64
1506   void mulq(Address src);
1507   void mulq(Register src);
1508   void mulxq(Register dst1, Register dst2, Register src);
1509 #endif
1510 
1511   // Multiply Scalar Double-Precision Floating-Point Values
1512   void mulsd(XMMRegister dst, Address src);
1513   void mulsd(XMMRegister dst, XMMRegister src);
1514 
1515   // Multiply Scalar Single-Precision Floating-Point Values
1516   void mulss(XMMRegister dst, Address src);
1517   void mulss(XMMRegister dst, XMMRegister src);
1518 
1519   void negl(Register dst);
1520 
1521 #ifdef _LP64
1522   void negq(Register dst);
1523 #endif
1524 
1525   void nop(int i = 1);
1526 
1527   void notl(Register dst);
1528 
1529 #ifdef _LP64
1530   void notq(Register dst);
1531 #endif
1532 
1533   void orl(Address dst, int32_t imm32);
1534   void orl(Register dst, int32_t imm32);
1535   void orl(Register dst, Address src);
1536   void orl(Register dst, Register src);
1537   void orl(Address dst, Register src);
1538 
1539   void orq(Address dst, int32_t imm32);
1540   void orq(Register dst, int32_t imm32);
1541   void orq(Register dst, Address src);
1542   void orq(Register dst, Register src);
1543 
1544   // Pack with unsigned saturation
1545   void packuswb(XMMRegister dst, XMMRegister src);
1546   void packuswb(XMMRegister dst, Address src);
1547   void vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1548 
1549   // Pemutation of 64bit words
1550   void vpermq(XMMRegister dst, XMMRegister src, int imm8, int vector_len);
1551   void vpermq(XMMRegister dst, XMMRegister src, int imm8);
1552   void vperm2i128(XMMRegister dst,  XMMRegister nds, XMMRegister src, int imm8);
1553   void vperm2f128(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8);
1554 
1555   void pause();
1556 
1557   // SSE4.2 string instructions
1558   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
1559   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
1560 
1561   void pcmpeqb(XMMRegister dst, XMMRegister src);
1562   void vpcmpeqb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1563   void evpcmpeqb(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1564   void evpcmpeqb(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1565   void evpcmpeqb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_len);
1566 
1567   void evpcmpgtb(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1568   void evpcmpgtb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_len);
1569 
1570   void evpcmpuw(KRegister kdst, XMMRegister nds, XMMRegister src, ComparisonPredicate vcc, int vector_len);
1571   void evpcmpuw(KRegister kdst, KRegister mask, XMMRegister nds, XMMRegister src, ComparisonPredicate of, int vector_len);
1572   void evpcmpuw(KRegister kdst, XMMRegister nds, Address src, ComparisonPredicate vcc, int vector_len);
1573 
1574   void pcmpeqw(XMMRegister dst, XMMRegister src);
1575   void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1576   void evpcmpeqw(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1577   void evpcmpeqw(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1578 
1579   void pcmpeqd(XMMRegister dst, XMMRegister src);
1580   void vpcmpeqd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1581   void evpcmpeqd(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1582   void evpcmpeqd(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1583 
1584   void pcmpeqq(XMMRegister dst, XMMRegister src);
1585   void vpcmpeqq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1586   void evpcmpeqq(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1587   void evpcmpeqq(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1588 
1589   void pmovmskb(Register dst, XMMRegister src);
1590   void vpmovmskb(Register dst, XMMRegister src);
1591 
1592   // SSE 4.1 extract
1593   void pextrd(Register dst, XMMRegister src, int imm8);
1594   void pextrq(Register dst, XMMRegister src, int imm8);
1595   void pextrd(Address dst, XMMRegister src, int imm8);
1596   void pextrq(Address dst, XMMRegister src, int imm8);
1597   void pextrb(Address dst, XMMRegister src, int imm8);
1598   // SSE 2 extract
1599   void pextrw(Register dst, XMMRegister src, int imm8);
1600   void pextrw(Address dst, XMMRegister src, int imm8);
1601 
1602   // SSE 4.1 insert
1603   void pinsrd(XMMRegister dst, Register src, int imm8);
1604   void pinsrq(XMMRegister dst, Register src, int imm8);
1605   void pinsrd(XMMRegister dst, Address src, int imm8);
1606   void pinsrq(XMMRegister dst, Address src, int imm8);
1607   void pinsrb(XMMRegister dst, Address src, int imm8);
1608   // SSE 2 insert
1609   void pinsrw(XMMRegister dst, Register src, int imm8);
1610   void pinsrw(XMMRegister dst, Address src, int imm8);
1611 
1612   // SSE4.1 packed move
1613   void pmovzxbw(XMMRegister dst, XMMRegister src);
1614   void pmovzxbw(XMMRegister dst, Address src);
1615 
1616   void vpmovzxbw( XMMRegister dst, Address src, int vector_len);
1617   void evpmovzxbw(XMMRegister dst, KRegister mask, Address src, int vector_len);
1618 
1619   void evpmovwb(Address dst, XMMRegister src, int vector_len);
1620   void evpmovwb(Address dst, KRegister mask, XMMRegister src, int vector_len);
1621 
1622 #ifndef _LP64 // no 32bit push/pop on amd64
1623   void popl(Address dst);
1624 #endif
1625 
1626 #ifdef _LP64
1627   void popq(Address dst);
1628 #endif
1629 
1630   void popcntl(Register dst, Address src);
1631   void popcntl(Register dst, Register src);
1632 
1633 #ifdef _LP64
1634   void popcntq(Register dst, Address src);
1635   void popcntq(Register dst, Register src);
1636 #endif
1637 
1638   // Prefetches (SSE, SSE2, 3DNOW only)
1639 
1640   void prefetchnta(Address src);
1641   void prefetchr(Address src);
1642   void prefetcht0(Address src);
1643   void prefetcht1(Address src);
1644   void prefetcht2(Address src);
1645   void prefetchw(Address src);
1646 
1647   // Shuffle Bytes
1648   void pshufb(XMMRegister dst, XMMRegister src);
1649   void pshufb(XMMRegister dst, Address src);
1650   void vpshufb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1651 
1652   // Shuffle Packed Doublewords
1653   void pshufd(XMMRegister dst, XMMRegister src, int mode);
1654   void pshufd(XMMRegister dst, Address src,     int mode);
1655   void vpshufd(XMMRegister dst, XMMRegister src, int mode, int vector_len);
1656 
1657   // Shuffle Packed Low Words
1658   void pshuflw(XMMRegister dst, XMMRegister src, int mode);
1659   void pshuflw(XMMRegister dst, Address src,     int mode);
1660 
1661   // Shift Right by bytes Logical DoubleQuadword Immediate
1662   void psrldq(XMMRegister dst, int shift);
1663   // Shift Left by bytes Logical DoubleQuadword Immediate
1664   void pslldq(XMMRegister dst, int shift);
1665 
1666   // Logical Compare 128bit
1667   void ptest(XMMRegister dst, XMMRegister src);
1668   void ptest(XMMRegister dst, Address src);
1669   // Logical Compare 256bit
1670   void vptest(XMMRegister dst, XMMRegister src);
1671   void vptest(XMMRegister dst, Address src);
1672 
1673   // Interleave Low Bytes
1674   void punpcklbw(XMMRegister dst, XMMRegister src);
1675   void punpcklbw(XMMRegister dst, Address src);
1676 
1677   // Interleave Low Doublewords
1678   void punpckldq(XMMRegister dst, XMMRegister src);
1679   void punpckldq(XMMRegister dst, Address src);
1680 
1681   // Interleave Low Quadwords
1682   void punpcklqdq(XMMRegister dst, XMMRegister src);
1683 
1684 #ifndef _LP64 // no 32bit push/pop on amd64
1685   void pushl(Address src);
1686 #endif
1687 
1688   void pushq(Address src);
1689 
1690   void rcll(Register dst, int imm8);
1691 
1692   void rclq(Register dst, int imm8);
1693 
1694   void rcrq(Register dst, int imm8);
1695 
1696   void rcpps(XMMRegister dst, XMMRegister src);
1697 
1698   void rcpss(XMMRegister dst, XMMRegister src);
1699 
1700   void rdtsc();
1701 
1702   void ret(int imm16);
1703 
1704 #ifdef _LP64
1705   void rorq(Register dst, int imm8);
1706   void rorxq(Register dst, Register src, int imm8);
1707   void rorxd(Register dst, Register src, int imm8);
1708 #endif
1709 
1710   void sahf();
1711 
1712   void sarl(Register dst, int imm8);
1713   void sarl(Register dst);
1714 
1715   void sarq(Register dst, int imm8);
1716   void sarq(Register dst);
1717 
1718   void sbbl(Address dst, int32_t imm32);
1719   void sbbl(Register dst, int32_t imm32);
1720   void sbbl(Register dst, Address src);
1721   void sbbl(Register dst, Register src);
1722 
1723   void sbbq(Address dst, int32_t imm32);
1724   void sbbq(Register dst, int32_t imm32);
1725   void sbbq(Register dst, Address src);
1726   void sbbq(Register dst, Register src);
1727 
1728   void setb(Condition cc, Register dst);
1729 
1730   void palignr(XMMRegister dst, XMMRegister src, int imm8);
1731   void vpalignr(XMMRegister dst, XMMRegister src1, XMMRegister src2, int imm8, int vector_len);
1732 
1733   void pblendw(XMMRegister dst, XMMRegister src, int imm8);
1734 
1735   void sha1rnds4(XMMRegister dst, XMMRegister src, int imm8);
1736   void sha1nexte(XMMRegister dst, XMMRegister src);
1737   void sha1msg1(XMMRegister dst, XMMRegister src);
1738   void sha1msg2(XMMRegister dst, XMMRegister src);
1739   // xmm0 is implicit additional source to the following instruction.
1740   void sha256rnds2(XMMRegister dst, XMMRegister src);
1741   void sha256msg1(XMMRegister dst, XMMRegister src);
1742   void sha256msg2(XMMRegister dst, XMMRegister src);
1743 
1744   void shldl(Register dst, Register src);
1745   void shldl(Register dst, Register src, int8_t imm8);
1746 
1747   void shll(Register dst, int imm8);
1748   void shll(Register dst);
1749 
1750   void shlq(Register dst, int imm8);
1751   void shlq(Register dst);
1752 
1753   void shrdl(Register dst, Register src);
1754 
1755   void shrl(Register dst, int imm8);
1756   void shrl(Register dst);
1757 
1758   void shrq(Register dst, int imm8);
1759   void shrq(Register dst);
1760 
1761   void smovl(); // QQQ generic?
1762 
1763   // Compute Square Root of Scalar Double-Precision Floating-Point Value
1764   void sqrtsd(XMMRegister dst, Address src);
1765   void sqrtsd(XMMRegister dst, XMMRegister src);
1766 
1767   // Compute Square Root of Scalar Single-Precision Floating-Point Value
1768   void sqrtss(XMMRegister dst, Address src);
1769   void sqrtss(XMMRegister dst, XMMRegister src);
1770 
1771   void std();
1772 
1773   void stmxcsr( Address dst );
1774 
1775   void subl(Address dst, int32_t imm32);
1776   void subl(Address dst, Register src);
1777   void subl(Register dst, int32_t imm32);
1778   void subl(Register dst, Address src);
1779   void subl(Register dst, Register src);
1780 
1781   void subq(Address dst, int32_t imm32);
1782   void subq(Address dst, Register src);
1783   void subq(Register dst, int32_t imm32);
1784   void subq(Register dst, Address src);
1785   void subq(Register dst, Register src);
1786 
1787   // Force generation of a 4 byte immediate value even if it fits into 8bit
1788   void subl_imm32(Register dst, int32_t imm32);
1789   void subq_imm32(Register dst, int32_t imm32);
1790 
1791   // Subtract Scalar Double-Precision Floating-Point Values
1792   void subsd(XMMRegister dst, Address src);
1793   void subsd(XMMRegister dst, XMMRegister src);
1794 
1795   // Subtract Scalar Single-Precision Floating-Point Values
1796   void subss(XMMRegister dst, Address src);
1797   void subss(XMMRegister dst, XMMRegister src);
1798 
1799   void testb(Register dst, int imm8);
1800   void testb(Address dst, int imm8);
1801 
1802   void testl(Register dst, int32_t imm32);
1803   void testl(Register dst, Register src);
1804   void testl(Register dst, Address src);
1805 
1806   void testq(Register dst, int32_t imm32);
1807   void testq(Register dst, Register src);
1808 
1809   // BMI - count trailing zeros
1810   void tzcntl(Register dst, Register src);
1811   void tzcntq(Register dst, Register src);
1812 
1813   // Unordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1814   void ucomisd(XMMRegister dst, Address src);
1815   void ucomisd(XMMRegister dst, XMMRegister src);
1816 
1817   // Unordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1818   void ucomiss(XMMRegister dst, Address src);
1819   void ucomiss(XMMRegister dst, XMMRegister src);
1820 
1821   void xabort(int8_t imm8);
1822 
1823   void xaddb(Address dst, Register src);
1824   void xaddw(Address dst, Register src);
1825   void xaddl(Address dst, Register src);
1826   void xaddq(Address dst, Register src);
1827 
1828   void xbegin(Label& abort, relocInfo::relocType rtype = relocInfo::none);
1829 
1830   void xchgb(Register reg, Address adr);
1831   void xchgw(Register reg, Address adr);
1832   void xchgl(Register reg, Address adr);
1833   void xchgl(Register dst, Register src);
1834 
1835   void xchgq(Register reg, Address adr);
1836   void xchgq(Register dst, Register src);
1837 
1838   void xend();
1839 
1840   // Get Value of Extended Control Register
1841   void xgetbv();
1842 
1843   void xorl(Register dst, int32_t imm32);
1844   void xorl(Register dst, Address src);
1845   void xorl(Register dst, Register src);
1846 
1847   void xorb(Register dst, Address src);
1848 
1849   void xorq(Register dst, Address src);
1850   void xorq(Register dst, Register src);
1851 
1852   void set_byte_if_not_zero(Register dst); // sets reg to 1 if not zero, otherwise 0
1853 
1854   // AVX 3-operands scalar instructions (encoded with VEX prefix)
1855 
1856   void vaddsd(XMMRegister dst, XMMRegister nds, Address src);
1857   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1858   void vaddss(XMMRegister dst, XMMRegister nds, Address src);
1859   void vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1860   void vdivsd(XMMRegister dst, XMMRegister nds, Address src);
1861   void vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1862   void vdivss(XMMRegister dst, XMMRegister nds, Address src);
1863   void vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1864   void vfmadd231sd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1865   void vfmadd231ss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1866   void vmulsd(XMMRegister dst, XMMRegister nds, Address src);
1867   void vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1868   void vmulss(XMMRegister dst, XMMRegister nds, Address src);
1869   void vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1870   void vsubsd(XMMRegister dst, XMMRegister nds, Address src);
1871   void vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1872   void vsubss(XMMRegister dst, XMMRegister nds, Address src);
1873   void vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1874 
1875   void shlxl(Register dst, Register src1, Register src2);
1876   void shlxq(Register dst, Register src1, Register src2);
1877 
1878   //====================VECTOR ARITHMETIC=====================================
1879 
1880   // Add Packed Floating-Point Values
1881   void addpd(XMMRegister dst, XMMRegister src);
1882   void addpd(XMMRegister dst, Address src);
1883   void addps(XMMRegister dst, XMMRegister src);
1884   void vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1885   void vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1886   void vaddpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1887   void vaddps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1888 
1889   // Subtract Packed Floating-Point Values
1890   void subpd(XMMRegister dst, XMMRegister src);
1891   void subps(XMMRegister dst, XMMRegister src);
1892   void vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1893   void vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1894   void vsubpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1895   void vsubps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1896 
1897   // Multiply Packed Floating-Point Values
1898   void mulpd(XMMRegister dst, XMMRegister src);
1899   void mulpd(XMMRegister dst, Address src);
1900   void mulps(XMMRegister dst, XMMRegister src);
1901   void vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1902   void vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1903   void vmulpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1904   void vmulps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1905 
1906   // Divide Packed Floating-Point Values
1907   void divpd(XMMRegister dst, XMMRegister src);
1908   void divps(XMMRegister dst, XMMRegister src);
1909   void vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1910   void vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1911   void vdivpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1912   void vdivps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1913 
1914   // Sqrt Packed Floating-Point Values - Double precision only
1915   void vsqrtpd(XMMRegister dst, XMMRegister src, int vector_len);
1916   void vsqrtpd(XMMRegister dst, Address src, int vector_len);
1917 
1918   // Bitwise Logical AND of Packed Floating-Point Values
1919   void andpd(XMMRegister dst, XMMRegister src);
1920   void andps(XMMRegister dst, XMMRegister src);
1921   void vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1922   void vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1923   void vandpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1924   void vandps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1925 
1926   void unpckhpd(XMMRegister dst, XMMRegister src);
1927   void unpcklpd(XMMRegister dst, XMMRegister src);
1928 
1929   // Bitwise Logical XOR of Packed Floating-Point Values
1930   void xorpd(XMMRegister dst, XMMRegister src);
1931   void xorps(XMMRegister dst, XMMRegister src);
1932   void vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1933   void vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1934   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1935   void vxorps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1936 
1937   // Add horizontal packed integers
1938   void vphaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1939   void vphaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1940   void phaddw(XMMRegister dst, XMMRegister src);
1941   void phaddd(XMMRegister dst, XMMRegister src);
1942 
1943   // Add packed integers
1944   void paddb(XMMRegister dst, XMMRegister src);
1945   void paddw(XMMRegister dst, XMMRegister src);
1946   void paddd(XMMRegister dst, XMMRegister src);
1947   void paddd(XMMRegister dst, Address src);
1948   void paddq(XMMRegister dst, XMMRegister src);
1949   void vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1950   void vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1951   void vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1952   void vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1953   void vpaddb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1954   void vpaddw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1955   void vpaddd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1956   void vpaddq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1957 
1958   // Sub packed integers
1959   void psubb(XMMRegister dst, XMMRegister src);
1960   void psubw(XMMRegister dst, XMMRegister src);
1961   void psubd(XMMRegister dst, XMMRegister src);
1962   void psubq(XMMRegister dst, XMMRegister src);
1963   void vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1964   void vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1965   void vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1966   void vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1967   void vpsubb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1968   void vpsubw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1969   void vpsubd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1970   void vpsubq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1971 
1972   // Multiply packed integers (only shorts and ints)
1973   void pmullw(XMMRegister dst, XMMRegister src);
1974   void pmulld(XMMRegister dst, XMMRegister src);
1975   void vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1976   void vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1977   void vpmullq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1978   void vpmullw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1979   void vpmulld(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1980   void vpmullq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1981 
1982   // Shift left packed integers
1983   void psllw(XMMRegister dst, int shift);
1984   void pslld(XMMRegister dst, int shift);
1985   void psllq(XMMRegister dst, int shift);
1986   void psllw(XMMRegister dst, XMMRegister shift);
1987   void pslld(XMMRegister dst, XMMRegister shift);
1988   void psllq(XMMRegister dst, XMMRegister shift);
1989   void vpsllw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
1990   void vpslld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
1991   void vpsllq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
1992   void vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
1993   void vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
1994   void vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
1995 
1996   // Logical shift right packed integers
1997   void psrlw(XMMRegister dst, int shift);
1998   void psrld(XMMRegister dst, int shift);
1999   void psrlq(XMMRegister dst, int shift);
2000   void psrlw(XMMRegister dst, XMMRegister shift);
2001   void psrld(XMMRegister dst, XMMRegister shift);
2002   void psrlq(XMMRegister dst, XMMRegister shift);
2003   void vpsrlw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2004   void vpsrld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2005   void vpsrlq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2006   void vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2007   void vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2008   void vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2009 
2010   // Arithmetic shift right packed integers (only shorts and ints, no instructions for longs)
2011   void psraw(XMMRegister dst, int shift);
2012   void psrad(XMMRegister dst, int shift);
2013   void psraw(XMMRegister dst, XMMRegister shift);
2014   void psrad(XMMRegister dst, XMMRegister shift);
2015   void vpsraw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2016   void vpsrad(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2017   void vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2018   void vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2019 
2020   // And packed integers
2021   void pand(XMMRegister dst, XMMRegister src);
2022   void vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2023   void vpand(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2024 
2025   // Andn packed integers
2026   void pandn(XMMRegister dst, XMMRegister src);
2027 
2028   // Or packed integers
2029   void por(XMMRegister dst, XMMRegister src);
2030   void vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2031   void vpor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2032 
2033   // Xor packed integers
2034   void pxor(XMMRegister dst, XMMRegister src);
2035   void vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2036   void vpxor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2037 
2038   // vinserti forms
2039   void vinserti128(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2040   void vinserti128(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2041   void vinserti32x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2042   void vinserti32x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2043   void vinserti64x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2044 
2045   // vinsertf forms
2046   void vinsertf128(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2047   void vinsertf128(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2048   void vinsertf32x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2049   void vinsertf32x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2050   void vinsertf64x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2051   void vinsertf64x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2052 
2053   // vextracti forms
2054   void vextracti128(XMMRegister dst, XMMRegister src, uint8_t imm8);
2055   void vextracti128(Address dst, XMMRegister src, uint8_t imm8);
2056   void vextracti32x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2057   void vextracti32x4(Address dst, XMMRegister src, uint8_t imm8);
2058   void vextracti64x2(XMMRegister dst, XMMRegister src, uint8_t imm8);
2059   void vextracti64x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2060 
2061   // vextractf forms
2062   void vextractf128(XMMRegister dst, XMMRegister src, uint8_t imm8);
2063   void vextractf128(Address dst, XMMRegister src, uint8_t imm8);
2064   void vextractf32x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2065   void vextractf32x4(Address dst, XMMRegister src, uint8_t imm8);
2066   void vextractf64x2(XMMRegister dst, XMMRegister src, uint8_t imm8);
2067   void vextractf64x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2068   void vextractf64x4(Address dst, XMMRegister src, uint8_t imm8);
2069 
2070   // legacy xmm sourced word/dword replicate
2071   void vpbroadcastw(XMMRegister dst, XMMRegister src);
2072   void vpbroadcastd(XMMRegister dst, XMMRegister src);
2073 
2074   // xmm/mem sourced byte/word/dword/qword replicate
2075   void evpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len);
2076   void evpbroadcastb(XMMRegister dst, Address src, int vector_len);
2077   void evpbroadcastw(XMMRegister dst, XMMRegister src, int vector_len);
2078   void evpbroadcastw(XMMRegister dst, Address src, int vector_len);
2079   void evpbroadcastd(XMMRegister dst, XMMRegister src, int vector_len);
2080   void evpbroadcastd(XMMRegister dst, Address src, int vector_len);
2081   void evpbroadcastq(XMMRegister dst, XMMRegister src, int vector_len);
2082   void evpbroadcastq(XMMRegister dst, Address src, int vector_len);
2083 
2084   // scalar single/double precision replicate
2085   void evpbroadcastss(XMMRegister dst, XMMRegister src, int vector_len);
2086   void evpbroadcastss(XMMRegister dst, Address src, int vector_len);
2087   void evpbroadcastsd(XMMRegister dst, XMMRegister src, int vector_len);
2088   void evpbroadcastsd(XMMRegister dst, Address src, int vector_len);
2089 
2090   // gpr sourced byte/word/dword/qword replicate
2091   void evpbroadcastb(XMMRegister dst, Register src, int vector_len);
2092   void evpbroadcastw(XMMRegister dst, Register src, int vector_len);
2093   void evpbroadcastd(XMMRegister dst, Register src, int vector_len);
2094   void evpbroadcastq(XMMRegister dst, Register src, int vector_len);
2095 
2096   // Carry-Less Multiplication Quadword
2097   void pclmulqdq(XMMRegister dst, XMMRegister src, int mask);
2098   void vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask);
2099 
2100   // AVX instruction which is used to clear upper 128 bits of YMM registers and
2101   // to avoid transaction penalty between AVX and SSE states. There is no
2102   // penalty if legacy SSE instructions are encoded using VEX prefix because
2103   // they always clear upper 128 bits. It should be used before calling
2104   // runtime code and native libraries.
2105   void vzeroupper();
2106 
2107   // AVX support for vectorized conditional move (double). The following two instructions used only coupled.
2108   void cmppd(XMMRegister dst, XMMRegister nds, XMMRegister src, int cop, int vector_len);
2109   void blendvpd(XMMRegister dst, XMMRegister nds, XMMRegister src1, XMMRegister src2, int vector_len);
2110   void vpblendd(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8, int vector_len);
2111 
2112  protected:
2113   // Next instructions require address alignment 16 bytes SSE mode.
2114   // They should be called only from corresponding MacroAssembler instructions.
2115   void andpd(XMMRegister dst, Address src);
2116   void andps(XMMRegister dst, Address src);
2117   void xorpd(XMMRegister dst, Address src);
2118   void xorps(XMMRegister dst, Address src);
2119 
2120 };
2121 
2122 // The Intel x86/Amd64 Assembler attributes: All fields enclosed here are to guide encoding level decisions.
2123 // Specific set functions are for specialized use, else defaults or whatever was supplied to object construction
2124 // are applied.
2125 class InstructionAttr {
2126 public:
2127   InstructionAttr(
2128     int vector_len,     // The length of vector to be applied in encoding - for both AVX and EVEX
2129     bool rex_vex_w,     // Width of data: if 32-bits or less, false, else if 64-bit or specially defined, true
2130     bool legacy_mode,   // Details if either this instruction is conditionally encoded to AVX or earlier if true else possibly EVEX
2131     bool no_reg_mask,   // when true, k0 is used when EVEX encoding is chosen, else k1 is used under the same condition
2132     bool uses_vl)       // This instruction may have legacy constraints based on vector length for EVEX
2133     :
2134       _avx_vector_len(vector_len),
2135       _rex_vex_w(rex_vex_w),
2136       _rex_vex_w_reverted(false),
2137       _legacy_mode(legacy_mode),
2138       _no_reg_mask(no_reg_mask),
2139       _uses_vl(uses_vl),
2140       _tuple_type(Assembler::EVEX_ETUP),
2141       _input_size_in_bits(Assembler::EVEX_NObit),
2142       _is_evex_instruction(false),
2143       _evex_encoding(0),
2144       _is_clear_context(true),
2145       _is_extended_context(false),
2146       _current_assembler(NULL),
2147       _embedded_opmask_register_specifier(1) { // hard code k1, it will be initialized for now
2148     if (UseAVX < 3) _legacy_mode = true;
2149   }
2150 
2151   ~InstructionAttr() {
2152     if (_current_assembler != NULL) {
2153       _current_assembler->clear_attributes();
2154     }
2155     _current_assembler = NULL;
2156   }
2157 
2158 private:
2159   int  _avx_vector_len;
2160   bool _rex_vex_w;
2161   bool _rex_vex_w_reverted;
2162   bool _legacy_mode;
2163   bool _no_reg_mask;
2164   bool _uses_vl;
2165   int  _tuple_type;
2166   int  _input_size_in_bits;
2167   bool _is_evex_instruction;
2168   int  _evex_encoding;
2169   bool _is_clear_context;
2170   bool _is_extended_context;
2171   int _embedded_opmask_register_specifier;
2172 
2173   Assembler *_current_assembler;
2174 
2175 public:
2176   // query functions for field accessors
2177   int  get_vector_len(void) const { return _avx_vector_len; }
2178   bool is_rex_vex_w(void) const { return _rex_vex_w; }
2179   bool is_rex_vex_w_reverted(void) { return _rex_vex_w_reverted; }
2180   bool is_legacy_mode(void) const { return _legacy_mode; }
2181   bool is_no_reg_mask(void) const { return _no_reg_mask; }
2182   bool uses_vl(void) const { return _uses_vl; }
2183   int  get_tuple_type(void) const { return _tuple_type; }
2184   int  get_input_size(void) const { return _input_size_in_bits; }
2185   int  is_evex_instruction(void) const { return _is_evex_instruction; }
2186   int  get_evex_encoding(void) const { return _evex_encoding; }
2187   bool is_clear_context(void) const { return _is_clear_context; }
2188   bool is_extended_context(void) const { return _is_extended_context; }
2189   int get_embedded_opmask_register_specifier(void) const { return _embedded_opmask_register_specifier; }
2190 
2191   // Set the vector len manually
2192   void set_vector_len(int vector_len) { _avx_vector_len = vector_len; }
2193 
2194   // Set revert rex_vex_w for avx encoding
2195   void set_rex_vex_w_reverted(void) { _rex_vex_w_reverted = true; }
2196 
2197   // Set rex_vex_w based on state
2198   void set_rex_vex_w(bool state) { _rex_vex_w = state; }
2199 
2200   // Set the instruction to be encoded in AVX mode
2201   void set_is_legacy_mode(void) { _legacy_mode = true; }
2202 
2203   // Set the current instuction to be encoded as an EVEX instuction
2204   void set_is_evex_instruction(void) { _is_evex_instruction = true; }
2205 
2206   // Internal encoding data used in compressed immediate offset programming
2207   void set_evex_encoding(int value) { _evex_encoding = value; }
2208 
2209   // Set the Evex.Z field to be used to clear all non directed XMM/YMM/ZMM components
2210   void reset_is_clear_context(void) { _is_clear_context = false; }
2211 
2212   // Map back to current asembler so that we can manage object level assocation
2213   void set_current_assembler(Assembler *current_assembler) { _current_assembler = current_assembler; }
2214 
2215   // Address modifiers used for compressed displacement calculation
2216   void set_address_attributes(int tuple_type, int input_size_in_bits) {
2217     if (VM_Version::supports_evex()) {
2218       _tuple_type = tuple_type;
2219       _input_size_in_bits = input_size_in_bits;
2220     }
2221   }
2222 
2223   // Set embedded opmask register specifier.
2224   void set_embedded_opmask_register_specifier(KRegister mask) {
2225     _embedded_opmask_register_specifier = (*mask).encoding() & 0x7;
2226   }
2227 
2228 };
2229 
2230 #endif // CPU_X86_VM_ASSEMBLER_X86_HPP