1 /*
   2  * Copyright (c) 1997, 2019, 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_ASSEMBLER_X86_HPP
  26 #define CPU_X86_ASSEMBLER_X86_HPP
  27 
  28 #include "asm/register.hpp"
  29 #include "runtime/vm_version.hpp"
  30 
  31 class BiasedLockingCounters;
  32 
  33 // Contains all the definitions needed for x86 assembly code generation.
  34 
  35 // Calling convention
  36 class Argument {
  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 {
 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   XMMRegister      _xmmindex;
 188   ScaleFactor      _scale;
 189   int              _disp;
 190   bool             _isxmmindex;
 191   RelocationHolder _rspec;
 192 
 193   // Easily misused constructors make them private
 194   // %%% can we make these go away?
 195   NOT_LP64(Address(address loc, RelocationHolder spec);)
 196   Address(int disp, address loc, relocInfo::relocType rtype);
 197   Address(int disp, address loc, RelocationHolder spec);
 198 
 199  public:
 200 
 201  int disp() { return _disp; }
 202   // creation
 203   Address()
 204     : _base(noreg),
 205       _index(noreg),
 206       _xmmindex(xnoreg),
 207       _scale(no_scale),
 208       _disp(0),
 209       _isxmmindex(false){
 210   }
 211 
 212   // No default displacement otherwise Register can be implicitly
 213   // converted to 0(Register) which is quite a different animal.
 214 
 215   Address(Register base, int disp)
 216     : _base(base),
 217       _index(noreg),
 218       _xmmindex(xnoreg),
 219       _scale(no_scale),
 220       _disp(disp),
 221       _isxmmindex(false){
 222   }
 223 
 224   Address(Register base, Register index, ScaleFactor scale, int disp = 0)
 225     : _base (base),
 226       _index(index),
 227       _xmmindex(xnoreg),
 228       _scale(scale),
 229       _disp (disp),
 230       _isxmmindex(false) {
 231     assert(!index->is_valid() == (scale == Address::no_scale),
 232            "inconsistent address");
 233   }
 234 
 235   Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0)
 236     : _base (base),
 237       _index(index.register_or_noreg()),
 238       _xmmindex(xnoreg),
 239       _scale(scale),
 240       _disp (disp + (index.constant_or_zero() * scale_size(scale))),
 241       _isxmmindex(false){
 242     if (!index.is_register())  scale = Address::no_scale;
 243     assert(!_index->is_valid() == (scale == Address::no_scale),
 244            "inconsistent address");
 245   }
 246 
 247   Address(Register base, XMMRegister index, ScaleFactor scale, int disp = 0)
 248     : _base (base),
 249       _index(noreg),
 250       _xmmindex(index),
 251       _scale(scale),
 252       _disp(disp),
 253       _isxmmindex(true) {
 254       assert(!index->is_valid() == (scale == Address::no_scale),
 255              "inconsistent address");
 256   }
 257 
 258   Address plus_disp(int disp) const {
 259     Address a = (*this);
 260     a._disp += disp;
 261     return a;
 262   }
 263   Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const {
 264     Address a = (*this);
 265     a._disp += disp.constant_or_zero() * scale_size(scale);
 266     if (disp.is_register()) {
 267       assert(!a.index()->is_valid(), "competing indexes");
 268       a._index = disp.as_register();
 269       a._scale = scale;
 270     }
 271     return a;
 272   }
 273   bool is_same_address(Address a) const {
 274     // disregard _rspec
 275     return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale;
 276   }
 277 
 278   // The following two overloads are used in connection with the
 279   // ByteSize type (see sizes.hpp).  They simplify the use of
 280   // ByteSize'd arguments in assembly code. Note that their equivalent
 281   // for the optimized build are the member functions with int disp
 282   // argument since ByteSize is mapped to an int type in that case.
 283   //
 284   // Note: DO NOT introduce similar overloaded functions for WordSize
 285   // arguments as in the optimized mode, both ByteSize and WordSize
 286   // are mapped to the same type and thus the compiler cannot make a
 287   // distinction anymore (=> compiler errors).
 288 
 289 #ifdef ASSERT
 290   Address(Register base, ByteSize disp)
 291     : _base(base),
 292       _index(noreg),
 293       _xmmindex(xnoreg),
 294       _scale(no_scale),
 295       _disp(in_bytes(disp)),
 296       _isxmmindex(false){
 297   }
 298 
 299   Address(Register base, Register index, ScaleFactor scale, ByteSize disp)
 300     : _base(base),
 301       _index(index),
 302       _xmmindex(xnoreg),
 303       _scale(scale),
 304       _disp(in_bytes(disp)),
 305       _isxmmindex(false){
 306     assert(!index->is_valid() == (scale == Address::no_scale),
 307            "inconsistent address");
 308   }
 309   Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp)
 310     : _base (base),
 311       _index(index.register_or_noreg()),
 312       _xmmindex(xnoreg),
 313       _scale(scale),
 314       _disp (in_bytes(disp) + (index.constant_or_zero() * scale_size(scale))),
 315       _isxmmindex(false) {
 316     if (!index.is_register())  scale = Address::no_scale;
 317     assert(!_index->is_valid() == (scale == Address::no_scale),
 318            "inconsistent address");
 319   }
 320 
 321 #endif // ASSERT
 322 
 323   // accessors
 324   bool        uses(Register reg) const { return _base == reg || _index == reg; }
 325   Register    base()             const { return _base;  }
 326   Register    index()            const { return _index; }
 327   XMMRegister xmmindex()         const { return _xmmindex; }
 328   ScaleFactor scale()            const { return _scale; }
 329   int         disp()             const { return _disp;  }
 330   bool        isxmmindex()       const { return _isxmmindex; }
 331 
 332   // Convert the raw encoding form into the form expected by the constructor for
 333   // Address.  An index of 4 (rsp) corresponds to having no index, so convert
 334   // that to noreg for the Address constructor.
 335   static Address make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc);
 336 
 337   static Address make_array(ArrayAddress);
 338 
 339  private:
 340   bool base_needs_rex() const {
 341     return _base != noreg && _base->encoding() >= 8;
 342   }
 343 
 344   bool index_needs_rex() const {
 345     return _index != noreg &&_index->encoding() >= 8;
 346   }
 347 
 348   bool xmmindex_needs_rex() const {
 349     return _xmmindex != xnoreg && _xmmindex->encoding() >= 8;
 350   }
 351 
 352   relocInfo::relocType reloc() const { return _rspec.type(); }
 353 
 354   friend class Assembler;
 355   friend class MacroAssembler;
 356   friend class LIR_Assembler; // base/index/scale/disp
 357 };
 358 
 359 //
 360 // AddressLiteral has been split out from Address because operands of this type
 361 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out
 362 // the few instructions that need to deal with address literals are unique and the
 363 // MacroAssembler does not have to implement every instruction in the Assembler
 364 // in order to search for address literals that may need special handling depending
 365 // on the instruction and the platform. As small step on the way to merging i486/amd64
 366 // directories.
 367 //
 368 class AddressLiteral {
 369   friend class ArrayAddress;
 370   RelocationHolder _rspec;
 371   // Typically we use AddressLiterals we want to use their rval
 372   // However in some situations we want the lval (effect address) of the item.
 373   // We provide a special factory for making those lvals.
 374   bool _is_lval;
 375 
 376   // If the target is far we'll need to load the ea of this to
 377   // a register to reach it. Otherwise if near we can do rip
 378   // relative addressing.
 379 
 380   address          _target;
 381 
 382  protected:
 383   // creation
 384   AddressLiteral()
 385     : _is_lval(false),
 386       _target(NULL)
 387   {}
 388 
 389   public:
 390 
 391 
 392   AddressLiteral(address target, relocInfo::relocType rtype);
 393 
 394   AddressLiteral(address target, RelocationHolder const& rspec)
 395     : _rspec(rspec),
 396       _is_lval(false),
 397       _target(target)
 398   {}
 399 
 400   AddressLiteral addr() {
 401     AddressLiteral ret = *this;
 402     ret._is_lval = true;
 403     return ret;
 404   }
 405 
 406 
 407  private:
 408 
 409   address target() { return _target; }
 410   bool is_lval() { return _is_lval; }
 411 
 412   relocInfo::relocType reloc() const { return _rspec.type(); }
 413   const RelocationHolder& rspec() const { return _rspec; }
 414 
 415   friend class Assembler;
 416   friend class MacroAssembler;
 417   friend class Address;
 418   friend class LIR_Assembler;
 419 };
 420 
 421 // Convience classes
 422 class RuntimeAddress: public AddressLiteral {
 423 
 424   public:
 425 
 426   RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {}
 427 
 428 };
 429 
 430 class ExternalAddress: public AddressLiteral {
 431  private:
 432   static relocInfo::relocType reloc_for_target(address target) {
 433     // Sometimes ExternalAddress is used for values which aren't
 434     // exactly addresses, like the card table base.
 435     // external_word_type can't be used for values in the first page
 436     // so just skip the reloc in that case.
 437     return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none;
 438   }
 439 
 440  public:
 441 
 442   ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {}
 443 
 444 };
 445 
 446 class InternalAddress: public AddressLiteral {
 447 
 448   public:
 449 
 450   InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {}
 451 
 452 };
 453 
 454 // x86 can do array addressing as a single operation since disp can be an absolute
 455 // address amd64 can't. We create a class that expresses the concept but does extra
 456 // magic on amd64 to get the final result
 457 
 458 class ArrayAddress {
 459   private:
 460 
 461   AddressLiteral _base;
 462   Address        _index;
 463 
 464   public:
 465 
 466   ArrayAddress() {};
 467   ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {};
 468   AddressLiteral base() { return _base; }
 469   Address index() { return _index; }
 470 
 471 };
 472 
 473 class InstructionAttr;
 474 
 475 // 64-bit refect the fxsave size which is 512 bytes and the new xsave area on EVEX which is another 2176 bytes
 476 // See fxsave and xsave(EVEX enabled) documentation for layout
 477 const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY(2688 / wordSize);
 478 
 479 // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction
 480 // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write
 481 // is what you get. The Assembler is generating code into a CodeBuffer.
 482 
 483 class Assembler : public AbstractAssembler  {
 484   friend class AbstractAssembler; // for the non-virtual hack
 485   friend class LIR_Assembler; // as_Address()
 486   friend class StubGenerator;
 487 
 488  public:
 489   enum Condition {                     // The x86 condition codes used for conditional jumps/moves.
 490     zero          = 0x4,
 491     notZero       = 0x5,
 492     equal         = 0x4,
 493     notEqual      = 0x5,
 494     less          = 0xc,
 495     lessEqual     = 0xe,
 496     greater       = 0xf,
 497     greaterEqual  = 0xd,
 498     below         = 0x2,
 499     belowEqual    = 0x6,
 500     above         = 0x7,
 501     aboveEqual    = 0x3,
 502     overflow      = 0x0,
 503     noOverflow    = 0x1,
 504     carrySet      = 0x2,
 505     carryClear    = 0x3,
 506     negative      = 0x8,
 507     positive      = 0x9,
 508     parity        = 0xa,
 509     noParity      = 0xb
 510   };
 511 
 512   enum Prefix {
 513     // segment overrides
 514     CS_segment = 0x2e,
 515     SS_segment = 0x36,
 516     DS_segment = 0x3e,
 517     ES_segment = 0x26,
 518     FS_segment = 0x64,
 519     GS_segment = 0x65,
 520 
 521     REX        = 0x40,
 522 
 523     REX_B      = 0x41,
 524     REX_X      = 0x42,
 525     REX_XB     = 0x43,
 526     REX_R      = 0x44,
 527     REX_RB     = 0x45,
 528     REX_RX     = 0x46,
 529     REX_RXB    = 0x47,
 530 
 531     REX_W      = 0x48,
 532 
 533     REX_WB     = 0x49,
 534     REX_WX     = 0x4A,
 535     REX_WXB    = 0x4B,
 536     REX_WR     = 0x4C,
 537     REX_WRB    = 0x4D,
 538     REX_WRX    = 0x4E,
 539     REX_WRXB   = 0x4F,
 540 
 541     VEX_3bytes = 0xC4,
 542     VEX_2bytes = 0xC5,
 543     EVEX_4bytes = 0x62,
 544     Prefix_EMPTY = 0x0
 545   };
 546 
 547   enum VexPrefix {
 548     VEX_B = 0x20,
 549     VEX_X = 0x40,
 550     VEX_R = 0x80,
 551     VEX_W = 0x80
 552   };
 553 
 554   enum ExexPrefix {
 555     EVEX_F  = 0x04,
 556     EVEX_V  = 0x08,
 557     EVEX_Rb = 0x10,
 558     EVEX_X  = 0x40,
 559     EVEX_Z  = 0x80
 560   };
 561 
 562   enum VexSimdPrefix {
 563     VEX_SIMD_NONE = 0x0,
 564     VEX_SIMD_66   = 0x1,
 565     VEX_SIMD_F3   = 0x2,
 566     VEX_SIMD_F2   = 0x3
 567   };
 568 
 569   enum VexOpcode {
 570     VEX_OPCODE_NONE  = 0x0,
 571     VEX_OPCODE_0F    = 0x1,
 572     VEX_OPCODE_0F_38 = 0x2,
 573     VEX_OPCODE_0F_3A = 0x3,
 574     VEX_OPCODE_MASK  = 0x1F
 575   };
 576 
 577   enum AvxVectorLen {
 578     AVX_128bit = 0x0,
 579     AVX_256bit = 0x1,
 580     AVX_512bit = 0x2,
 581     AVX_NoVec  = 0x4
 582   };
 583 
 584   enum EvexTupleType {
 585     EVEX_FV   = 0,
 586     EVEX_HV   = 4,
 587     EVEX_FVM  = 6,
 588     EVEX_T1S  = 7,
 589     EVEX_T1F  = 11,
 590     EVEX_T2   = 13,
 591     EVEX_T4   = 15,
 592     EVEX_T8   = 17,
 593     EVEX_HVM  = 18,
 594     EVEX_QVM  = 19,
 595     EVEX_OVM  = 20,
 596     EVEX_M128 = 21,
 597     EVEX_DUP  = 22,
 598     EVEX_ETUP = 23
 599   };
 600 
 601   enum EvexInputSizeInBits {
 602     EVEX_8bit  = 0,
 603     EVEX_16bit = 1,
 604     EVEX_32bit = 2,
 605     EVEX_64bit = 3,
 606     EVEX_NObit = 4
 607   };
 608 
 609   enum WhichOperand {
 610     // input to locate_operand, and format code for relocations
 611     imm_operand  = 0,            // embedded 32-bit|64-bit immediate operand
 612     disp32_operand = 1,          // embedded 32-bit displacement or address
 613     call32_operand = 2,          // embedded 32-bit self-relative displacement
 614 #ifndef _LP64
 615     _WhichOperand_limit = 3
 616 #else
 617      narrow_oop_operand = 3,     // embedded 32-bit immediate narrow oop
 618     _WhichOperand_limit = 4
 619 #endif
 620   };
 621 
 622   enum ComparisonPredicate {
 623     eq = 0,
 624     lt = 1,
 625     le = 2,
 626     _false = 3,
 627     neq = 4,
 628     nlt = 5,
 629     nle = 6,
 630     _true = 7
 631   };
 632 
 633   //---<  calculate length of instruction  >---
 634   // As instruction size can't be found out easily on x86/x64,
 635   // we just use '4' for len and maxlen.
 636   // instruction must start at passed address
 637   static unsigned int instr_len(unsigned char *instr) { return 4; }
 638 
 639   //---<  longest instructions  >---
 640   // Max instruction length is not specified in architecture documentation.
 641   // We could use a "safe enough" estimate (15), but just default to
 642   // instruction length guess from above.
 643   static unsigned int instr_maxlen() { return 4; }
 644 
 645   // NOTE: The general philopsophy of the declarations here is that 64bit versions
 646   // of instructions are freely declared without the need for wrapping them an ifdef.
 647   // (Some dangerous instructions are ifdef's out of inappropriate jvm's.)
 648   // In the .cpp file the implementations are wrapped so that they are dropped out
 649   // of the resulting jvm. This is done mostly to keep the footprint of MINIMAL
 650   // to the size it was prior to merging up the 32bit and 64bit assemblers.
 651   //
 652   // This does mean you'll get a linker/runtime error if you use a 64bit only instruction
 653   // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down.
 654 
 655 private:
 656 
 657   bool _legacy_mode_bw;
 658   bool _legacy_mode_dq;
 659   bool _legacy_mode_vl;
 660   bool _legacy_mode_vlbw;
 661   bool _is_managed;
 662   bool _vector_masking;    // For stub code use only
 663 
 664   class InstructionAttr *_attributes;
 665 
 666   // 64bit prefixes
 667   int prefix_and_encode(int reg_enc, bool byteinst = false);
 668   int prefixq_and_encode(int reg_enc);
 669 
 670   int prefix_and_encode(int dst_enc, int src_enc) {
 671     return prefix_and_encode(dst_enc, false, src_enc, false);
 672   }
 673   int prefix_and_encode(int dst_enc, bool dst_is_byte, int src_enc, bool src_is_byte);
 674   int prefixq_and_encode(int dst_enc, int src_enc);
 675 
 676   void prefix(Register reg);
 677   void prefix(Register dst, Register src, Prefix p);
 678   void prefix(Register dst, Address adr, Prefix p);
 679   void prefix(Address adr);
 680   void prefixq(Address adr);
 681 
 682   void prefix(Address adr, Register reg,  bool byteinst = false);
 683   void prefix(Address adr, XMMRegister reg);
 684   void prefixq(Address adr, Register reg);
 685   void prefixq(Address adr, XMMRegister reg);
 686 
 687   void prefetch_prefix(Address src);
 688 
 689   void rex_prefix(Address adr, XMMRegister xreg,
 690                   VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 691   int  rex_prefix_and_encode(int dst_enc, int src_enc,
 692                              VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 693 
 694   void vex_prefix(bool vex_r, bool vex_b, bool vex_x, int nds_enc, VexSimdPrefix pre, VexOpcode opc);
 695 
 696   void evex_prefix(bool vex_r, bool vex_b, bool vex_x, bool evex_r, bool evex_v,
 697                    int nds_enc, VexSimdPrefix pre, VexOpcode opc);
 698 
 699   void vex_prefix(Address adr, int nds_enc, int xreg_enc,
 700                   VexSimdPrefix pre, VexOpcode opc,
 701                   InstructionAttr *attributes);
 702 
 703   int  vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc,
 704                              VexSimdPrefix pre, VexOpcode opc,
 705                              InstructionAttr *attributes);
 706 
 707   void simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre,
 708                    VexOpcode opc, InstructionAttr *attributes);
 709 
 710   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPrefix pre,
 711                              VexOpcode opc, InstructionAttr *attributes);
 712 
 713   // Helper functions for groups of instructions
 714   void emit_arith_b(int op1, int op2, Register dst, int imm8);
 715 
 716   void emit_arith(int op1, int op2, Register dst, int32_t imm32);
 717   // Force generation of a 4 byte immediate value even if it fits into 8bit
 718   void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32);
 719   void emit_arith(int op1, int op2, Register dst, Register src);
 720 
 721   bool emit_compressed_disp_byte(int &disp);
 722 
 723   void emit_operand(Register reg,
 724                     Register base, Register index, Address::ScaleFactor scale,
 725                     int disp,
 726                     RelocationHolder const& rspec,
 727                     int rip_relative_correction = 0);
 728 
 729   void emit_operand(XMMRegister reg, Register base, XMMRegister index,
 730                     Address::ScaleFactor scale,
 731                     int disp, RelocationHolder const& rspec);
 732 
 733   void emit_operand(Register reg, Address adr, int rip_relative_correction = 0);
 734 
 735   // operands that only take the original 32bit registers
 736   void emit_operand32(Register reg, Address adr);
 737 
 738   void emit_operand(XMMRegister reg,
 739                     Register base, Register index, Address::ScaleFactor scale,
 740                     int disp,
 741                     RelocationHolder const& rspec);
 742 
 743   void emit_operand(XMMRegister reg, Address adr);
 744 
 745   void emit_operand(MMXRegister reg, Address adr);
 746 
 747   // workaround gcc (3.2.1-7) bug
 748   void emit_operand(Address adr, MMXRegister reg);
 749 
 750 
 751   // Immediate-to-memory forms
 752   void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32);
 753 
 754   void emit_farith(int b1, int b2, int i);
 755 
 756 
 757  protected:
 758   #ifdef ASSERT
 759   void check_relocation(RelocationHolder const& rspec, int format);
 760   #endif
 761 
 762   void emit_data(jint data, relocInfo::relocType    rtype, int format);
 763   void emit_data(jint data, RelocationHolder const& rspec, int format);
 764   void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0);
 765   void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0);
 766 
 767   bool reachable(AddressLiteral adr) NOT_LP64({ return true;});
 768 
 769   // These are all easily abused and hence protected
 770 
 771   // 32BIT ONLY SECTION
 772 #ifndef _LP64
 773   // Make these disappear in 64bit mode since they would never be correct
 774   void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec);   // 32BIT ONLY
 775   void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 776 
 777   void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 778   void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec);     // 32BIT ONLY
 779 
 780   void push_literal32(int32_t imm32, RelocationHolder const& rspec);                 // 32BIT ONLY
 781 #else
 782   // 64BIT ONLY SECTION
 783   void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec);   // 64BIT ONLY
 784 
 785   void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec);
 786   void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec);
 787 
 788   void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec);
 789   void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec);
 790 #endif // _LP64
 791 
 792   // These are unique in that we are ensured by the caller that the 32bit
 793   // relative in these instructions will always be able to reach the potentially
 794   // 64bit address described by entry. Since they can take a 64bit address they
 795   // don't have the 32 suffix like the other instructions in this class.
 796 
 797   void call_literal(address entry, RelocationHolder const& rspec);
 798   void jmp_literal(address entry, RelocationHolder const& rspec);
 799 
 800   // Avoid using directly section
 801   // Instructions in this section are actually usable by anyone without danger
 802   // of failure but have performance issues that are addressed my enhanced
 803   // instructions which will do the proper thing base on the particular cpu.
 804   // We protect them because we don't trust you...
 805 
 806   // Don't use next inc() and dec() methods directly. INC & DEC instructions
 807   // could cause a partial flag stall since they don't set CF flag.
 808   // Use MacroAssembler::decrement() & MacroAssembler::increment() methods
 809   // which call inc() & dec() or add() & sub() in accordance with
 810   // the product flag UseIncDec value.
 811 
 812   void decl(Register dst);
 813   void decl(Address dst);
 814   void decq(Register dst);
 815   void decq(Address dst);
 816 
 817   void incl(Register dst);
 818   void incl(Address dst);
 819   void incq(Register dst);
 820   void incq(Address dst);
 821 
 822   // New cpus require use of movsd and movss to avoid partial register stall
 823   // when loading from memory. But for old Opteron use movlpd instead of movsd.
 824   // The selection is done in MacroAssembler::movdbl() and movflt().
 825 
 826   // Move Scalar Single-Precision Floating-Point Values
 827   void movss(XMMRegister dst, Address src);
 828   void movss(XMMRegister dst, XMMRegister src);
 829   void movss(Address dst, XMMRegister src);
 830 
 831   // Move Scalar Double-Precision Floating-Point Values
 832   void movsd(XMMRegister dst, Address src);
 833   void movsd(XMMRegister dst, XMMRegister src);
 834   void movsd(Address dst, XMMRegister src);
 835   void movlpd(XMMRegister dst, Address src);
 836 
 837   // New cpus require use of movaps and movapd to avoid partial register stall
 838   // when moving between registers.
 839   void movaps(XMMRegister dst, XMMRegister src);
 840   void movapd(XMMRegister dst, XMMRegister src);
 841 
 842   // End avoid using directly
 843 
 844 
 845   // Instruction prefixes
 846   void prefix(Prefix p);
 847 
 848   public:
 849 
 850   // Creation
 851   Assembler(CodeBuffer* code) : AbstractAssembler(code) {
 852     init_attributes();
 853   }
 854 
 855   // Decoding
 856   static address locate_operand(address inst, WhichOperand which);
 857   static address locate_next_instruction(address inst);
 858 
 859   // Utilities
 860   static bool is_polling_page_far() NOT_LP64({ return false;});
 861   static bool query_compressed_disp_byte(int disp, bool is_evex_inst, int vector_len,
 862                                          int cur_tuple_type, int in_size_in_bits, int cur_encoding);
 863 
 864   // Generic instructions
 865   // Does 32bit or 64bit as needed for the platform. In some sense these
 866   // belong in macro assembler but there is no need for both varieties to exist
 867 
 868   void init_attributes(void) {
 869     _legacy_mode_bw = (VM_Version::supports_avx512bw() == false);
 870     _legacy_mode_dq = (VM_Version::supports_avx512dq() == false);
 871     _legacy_mode_vl = (VM_Version::supports_avx512vl() == false);
 872     _legacy_mode_vlbw = (VM_Version::supports_avx512vlbw() == false);
 873     _is_managed = false;
 874     _vector_masking = false;
 875     _attributes = NULL;
 876   }
 877 
 878   void set_attributes(InstructionAttr *attributes) { _attributes = attributes; }
 879   void clear_attributes(void) { _attributes = NULL; }
 880 
 881   void set_managed(void) { _is_managed = true; }
 882   void clear_managed(void) { _is_managed = false; }
 883   bool is_managed(void) { return _is_managed; }
 884 
 885   void lea(Register dst, Address src);
 886 
 887   void mov(Register dst, Register src);
 888 
 889   void pusha();
 890   void popa();
 891 
 892   void pushf();
 893   void popf();
 894 
 895   void push(int32_t imm32);
 896 
 897   void push(Register src);
 898 
 899   void pop(Register dst);
 900 
 901   // These are dummies to prevent surprise implicit conversions to Register
 902   void push(void* v);
 903   void pop(void* v);
 904 
 905   // These do register sized moves/scans
 906   void rep_mov();
 907   void rep_stos();
 908   void rep_stosb();
 909   void repne_scan();
 910 #ifdef _LP64
 911   void repne_scanl();
 912 #endif
 913 
 914   // Vanilla instructions in lexical order
 915 
 916   void adcl(Address dst, int32_t imm32);
 917   void adcl(Address dst, Register src);
 918   void adcl(Register dst, int32_t imm32);
 919   void adcl(Register dst, Address src);
 920   void adcl(Register dst, Register src);
 921 
 922   void adcq(Register dst, int32_t imm32);
 923   void adcq(Register dst, Address src);
 924   void adcq(Register dst, Register src);
 925 
 926   void addb(Address dst, int imm8);
 927   void addw(Address dst, int imm16);
 928 
 929   void addl(Address dst, int32_t imm32);
 930   void addl(Address dst, Register src);
 931   void addl(Register dst, int32_t imm32);
 932   void addl(Register dst, Address src);
 933   void addl(Register dst, Register src);
 934 
 935   void addq(Address dst, int32_t imm32);
 936   void addq(Address dst, Register src);
 937   void addq(Register dst, int32_t imm32);
 938   void addq(Register dst, Address src);
 939   void addq(Register dst, Register src);
 940 
 941 #ifdef _LP64
 942  //Add Unsigned Integers with Carry Flag
 943   void adcxq(Register dst, Register src);
 944 
 945  //Add Unsigned Integers with Overflow Flag
 946   void adoxq(Register dst, Register src);
 947 #endif
 948 
 949   void addr_nop_4();
 950   void addr_nop_5();
 951   void addr_nop_7();
 952   void addr_nop_8();
 953 
 954   // Add Scalar Double-Precision Floating-Point Values
 955   void addsd(XMMRegister dst, Address src);
 956   void addsd(XMMRegister dst, XMMRegister src);
 957 
 958   // Add Scalar Single-Precision Floating-Point Values
 959   void addss(XMMRegister dst, Address src);
 960   void addss(XMMRegister dst, XMMRegister src);
 961 
 962   // AES instructions
 963   void aesdec(XMMRegister dst, Address src);
 964   void aesdec(XMMRegister dst, XMMRegister src);
 965   void aesdeclast(XMMRegister dst, Address src);
 966   void aesdeclast(XMMRegister dst, XMMRegister src);
 967   void aesenc(XMMRegister dst, Address src);
 968   void aesenc(XMMRegister dst, XMMRegister src);
 969   void aesenclast(XMMRegister dst, Address src);
 970   void aesenclast(XMMRegister dst, XMMRegister src);
 971   // Vector AES instructions
 972   void vaesenc(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
 973   void vaesenclast(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
 974   void vaesdec(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
 975   void vaesdeclast(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
 976 
 977   void andl(Address  dst, int32_t imm32);
 978   void andl(Register dst, int32_t imm32);
 979   void andl(Register dst, Address src);
 980   void andl(Register dst, Register src);
 981 
 982   void andq(Address  dst, int32_t imm32);
 983   void andq(Register dst, int32_t imm32);
 984   void andq(Register dst, Address src);
 985   void andq(Register dst, Register src);
 986 
 987   // BMI instructions
 988   void andnl(Register dst, Register src1, Register src2);
 989   void andnl(Register dst, Register src1, Address src2);
 990   void andnq(Register dst, Register src1, Register src2);
 991   void andnq(Register dst, Register src1, Address src2);
 992 
 993   void blsil(Register dst, Register src);
 994   void blsil(Register dst, Address src);
 995   void blsiq(Register dst, Register src);
 996   void blsiq(Register dst, Address src);
 997 
 998   void blsmskl(Register dst, Register src);
 999   void blsmskl(Register dst, Address src);
1000   void blsmskq(Register dst, Register src);
1001   void blsmskq(Register dst, Address src);
1002 
1003   void blsrl(Register dst, Register src);
1004   void blsrl(Register dst, Address src);
1005   void blsrq(Register dst, Register src);
1006   void blsrq(Register dst, Address src);
1007 
1008   void bsfl(Register dst, Register src);
1009   void bsrl(Register dst, Register src);
1010 
1011 #ifdef _LP64
1012   void bsfq(Register dst, Register src);
1013   void bsrq(Register dst, Register src);
1014 #endif
1015 
1016   void bswapl(Register reg);
1017 
1018   void bswapq(Register reg);
1019 
1020   void call(Label& L, relocInfo::relocType rtype);
1021   void call(Register reg);  // push pc; pc <- reg
1022   void call(Address adr);   // push pc; pc <- adr
1023 
1024   void cdql();
1025 
1026   void cdqq();
1027 
1028   void cld();
1029 
1030   void clflush(Address adr);
1031   void clflushopt(Address adr);
1032   void clwb(Address adr);
1033 
1034   void cmovl(Condition cc, Register dst, Register src);
1035   void cmovl(Condition cc, Register dst, Address src);
1036 
1037   void cmovq(Condition cc, Register dst, Register src);
1038   void cmovq(Condition cc, Register dst, Address src);
1039 
1040 
1041   void cmpb(Address dst, int imm8);
1042 
1043   void cmpl(Address dst, int32_t imm32);
1044 
1045   void cmpl(Register dst, int32_t imm32);
1046   void cmpl(Register dst, Register src);
1047   void cmpl(Register dst, Address src);
1048 
1049   void cmpq(Address dst, int32_t imm32);
1050   void cmpq(Address dst, Register src);
1051 
1052   void cmpq(Register dst, int32_t imm32);
1053   void cmpq(Register dst, Register src);
1054   void cmpq(Register dst, Address src);
1055 
1056   // these are dummies used to catch attempting to convert NULL to Register
1057   void cmpl(Register dst, void* junk); // dummy
1058   void cmpq(Register dst, void* junk); // dummy
1059 
1060   void cmpw(Address dst, int imm16);
1061 
1062   void cmpxchg8 (Address adr);
1063 
1064   void cmpxchgb(Register reg, Address adr);
1065   void cmpxchgl(Register reg, Address adr);
1066 
1067   void cmpxchgq(Register reg, Address adr);
1068 
1069   // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1070   void comisd(XMMRegister dst, Address src);
1071   void comisd(XMMRegister dst, XMMRegister src);
1072 
1073   // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1074   void comiss(XMMRegister dst, Address src);
1075   void comiss(XMMRegister dst, XMMRegister src);
1076 
1077   // Identify processor type and features
1078   void cpuid();
1079 
1080   // CRC32C
1081   void crc32(Register crc, Register v, int8_t sizeInBytes);
1082   void crc32(Register crc, Address adr, int8_t sizeInBytes);
1083 
1084   // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floating-Point Value
1085   void cvtsd2ss(XMMRegister dst, XMMRegister src);
1086   void cvtsd2ss(XMMRegister dst, Address src);
1087 
1088   // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value
1089   void cvtsi2sdl(XMMRegister dst, Register src);
1090   void cvtsi2sdl(XMMRegister dst, Address src);
1091   void cvtsi2sdq(XMMRegister dst, Register src);
1092   void cvtsi2sdq(XMMRegister dst, Address src);
1093 
1094   // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value
1095   void cvtsi2ssl(XMMRegister dst, Register src);
1096   void cvtsi2ssl(XMMRegister dst, Address src);
1097   void cvtsi2ssq(XMMRegister dst, Register src);
1098   void cvtsi2ssq(XMMRegister dst, Address src);
1099 
1100   // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Value
1101   void cvtdq2pd(XMMRegister dst, XMMRegister src);
1102 
1103   // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Value
1104   void cvtdq2ps(XMMRegister dst, XMMRegister src);
1105 
1106   // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floating-Point Value
1107   void cvtss2sd(XMMRegister dst, XMMRegister src);
1108   void cvtss2sd(XMMRegister dst, Address src);
1109 
1110   // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Integer
1111   void cvttsd2sil(Register dst, Address src);
1112   void cvttsd2sil(Register dst, XMMRegister src);
1113   void cvttsd2siq(Register dst, XMMRegister src);
1114 
1115   // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Integer
1116   void cvttss2sil(Register dst, XMMRegister src);
1117   void cvttss2siq(Register dst, XMMRegister src);
1118 
1119   void cvttpd2dq(XMMRegister dst, XMMRegister src);
1120 
1121   //Abs of packed Integer values
1122   void pabsb(XMMRegister dst, XMMRegister src);
1123   void pabsw(XMMRegister dst, XMMRegister src);
1124   void pabsd(XMMRegister dst, XMMRegister src);
1125   void vpabsb(XMMRegister dst, XMMRegister src, int vector_len);
1126   void vpabsw(XMMRegister dst, XMMRegister src, int vector_len);
1127   void vpabsd(XMMRegister dst, XMMRegister src, int vector_len);
1128   void evpabsq(XMMRegister dst, XMMRegister src, int vector_len);
1129 
1130   // Divide Scalar Double-Precision Floating-Point Values
1131   void divsd(XMMRegister dst, Address src);
1132   void divsd(XMMRegister dst, XMMRegister src);
1133 
1134   // Divide Scalar Single-Precision Floating-Point Values
1135   void divss(XMMRegister dst, Address src);
1136   void divss(XMMRegister dst, XMMRegister src);
1137 
1138   void emms();
1139 
1140   void fabs();
1141 
1142   void fadd(int i);
1143 
1144   void fadd_d(Address src);
1145   void fadd_s(Address src);
1146 
1147   // "Alternate" versions of x87 instructions place result down in FPU
1148   // stack instead of on TOS
1149 
1150   void fadda(int i); // "alternate" fadd
1151   void faddp(int i = 1);
1152 
1153   void fchs();
1154 
1155   void fcom(int i);
1156 
1157   void fcomp(int i = 1);
1158   void fcomp_d(Address src);
1159   void fcomp_s(Address src);
1160 
1161   void fcompp();
1162 
1163   void fcos();
1164 
1165   void fdecstp();
1166 
1167   void fdiv(int i);
1168   void fdiv_d(Address src);
1169   void fdivr_s(Address src);
1170   void fdiva(int i);  // "alternate" fdiv
1171   void fdivp(int i = 1);
1172 
1173   void fdivr(int i);
1174   void fdivr_d(Address src);
1175   void fdiv_s(Address src);
1176 
1177   void fdivra(int i); // "alternate" reversed fdiv
1178 
1179   void fdivrp(int i = 1);
1180 
1181   void ffree(int i = 0);
1182 
1183   void fild_d(Address adr);
1184   void fild_s(Address adr);
1185 
1186   void fincstp();
1187 
1188   void finit();
1189 
1190   void fist_s (Address adr);
1191   void fistp_d(Address adr);
1192   void fistp_s(Address adr);
1193 
1194   void fld1();
1195 
1196   void fld_d(Address adr);
1197   void fld_s(Address adr);
1198   void fld_s(int index);
1199   void fld_x(Address adr);  // extended-precision (80-bit) format
1200 
1201   void fldcw(Address src);
1202 
1203   void fldenv(Address src);
1204 
1205   void fldlg2();
1206 
1207   void fldln2();
1208 
1209   void fldz();
1210 
1211   void flog();
1212   void flog10();
1213 
1214   void fmul(int i);
1215 
1216   void fmul_d(Address src);
1217   void fmul_s(Address src);
1218 
1219   void fmula(int i);  // "alternate" fmul
1220 
1221   void fmulp(int i = 1);
1222 
1223   void fnsave(Address dst);
1224 
1225   void fnstcw(Address src);
1226 
1227   void fnstsw_ax();
1228 
1229   void fprem();
1230   void fprem1();
1231 
1232   void frstor(Address src);
1233 
1234   void fsin();
1235 
1236   void fsqrt();
1237 
1238   void fst_d(Address adr);
1239   void fst_s(Address adr);
1240 
1241   void fstp_d(Address adr);
1242   void fstp_d(int index);
1243   void fstp_s(Address adr);
1244   void fstp_x(Address adr); // extended-precision (80-bit) format
1245 
1246   void fsub(int i);
1247   void fsub_d(Address src);
1248   void fsub_s(Address src);
1249 
1250   void fsuba(int i);  // "alternate" fsub
1251 
1252   void fsubp(int i = 1);
1253 
1254   void fsubr(int i);
1255   void fsubr_d(Address src);
1256   void fsubr_s(Address src);
1257 
1258   void fsubra(int i); // "alternate" reversed fsub
1259 
1260   void fsubrp(int i = 1);
1261 
1262   void ftan();
1263 
1264   void ftst();
1265 
1266   void fucomi(int i = 1);
1267   void fucomip(int i = 1);
1268 
1269   void fwait();
1270 
1271   void fxch(int i = 1);
1272 
1273   void fxrstor(Address src);
1274   void xrstor(Address src);
1275 
1276   void fxsave(Address dst);
1277   void xsave(Address dst);
1278 
1279   void fyl2x();
1280   void frndint();
1281   void f2xm1();
1282   void fldl2e();
1283 
1284   void hlt();
1285 
1286   void idivl(Register src);
1287   void divl(Register src); // Unsigned division
1288 
1289 #ifdef _LP64
1290   void idivq(Register src);
1291 #endif
1292 
1293   void imull(Register src);
1294   void imull(Register dst, Register src);
1295   void imull(Register dst, Register src, int value);
1296   void imull(Register dst, Address src);
1297 
1298 #ifdef _LP64
1299   void imulq(Register dst, Register src);
1300   void imulq(Register dst, Register src, int value);
1301   void imulq(Register dst, Address src);
1302 #endif
1303 
1304   // jcc is the generic conditional branch generator to run-
1305   // time routines, jcc is used for branches to labels. jcc
1306   // takes a branch opcode (cc) and a label (L) and generates
1307   // either a backward branch or a forward branch and links it
1308   // to the label fixup chain. Usage:
1309   //
1310   // Label L;      // unbound label
1311   // jcc(cc, L);   // forward branch to unbound label
1312   // bind(L);      // bind label to the current pc
1313   // jcc(cc, L);   // backward branch to bound label
1314   // bind(L);      // illegal: a label may be bound only once
1315   //
1316   // Note: The same Label can be used for forward and backward branches
1317   // but it may be bound only once.
1318 
1319   void jcc(Condition cc, Label& L, bool maybe_short = true);
1320 
1321   // Conditional jump to a 8-bit offset to L.
1322   // WARNING: be very careful using this for forward jumps.  If the label is
1323   // not bound within an 8-bit offset of this instruction, a run-time error
1324   // will occur.
1325 
1326   // Use macro to record file and line number.
1327   #define jccb(cc, L) jccb_0(cc, L, __FILE__, __LINE__)
1328 
1329   void jccb_0(Condition cc, Label& L, const char* file, int line);
1330 
1331   void jmp(Address entry);    // pc <- entry
1332 
1333   // Label operations & relative jumps (PPUM Appendix D)
1334   void jmp(Label& L, bool maybe_short = true);   // unconditional jump to L
1335 
1336   void jmp(Register entry); // pc <- entry
1337 
1338   // Unconditional 8-bit offset jump to L.
1339   // WARNING: be very careful using this for forward jumps.  If the label is
1340   // not bound within an 8-bit offset of this instruction, a run-time error
1341   // will occur.
1342 
1343   // Use macro to record file and line number.
1344   #define jmpb(L) jmpb_0(L, __FILE__, __LINE__)
1345 
1346   void jmpb_0(Label& L, const char* file, int line);
1347 
1348   void ldmxcsr( Address src );
1349 
1350   void leal(Register dst, Address src);
1351 
1352   void leaq(Register dst, Address src);
1353 
1354   void lfence();
1355 
1356   void lock();
1357 
1358   void lzcntl(Register dst, Register src);
1359 
1360 #ifdef _LP64
1361   void lzcntq(Register dst, Register src);
1362 #endif
1363 
1364   enum Membar_mask_bits {
1365     StoreStore = 1 << 3,
1366     LoadStore  = 1 << 2,
1367     StoreLoad  = 1 << 1,
1368     LoadLoad   = 1 << 0
1369   };
1370 
1371   // Serializes memory and blows flags
1372   void membar(Membar_mask_bits order_constraint) {
1373     // We only have to handle StoreLoad
1374     if (order_constraint & StoreLoad) {
1375       // All usable chips support "locked" instructions which suffice
1376       // as barriers, and are much faster than the alternative of
1377       // using cpuid instruction. We use here a locked add [esp-C],0.
1378       // This is conveniently otherwise a no-op except for blowing
1379       // flags, and introducing a false dependency on target memory
1380       // location. We can't do anything with flags, but we can avoid
1381       // memory dependencies in the current method by locked-adding
1382       // somewhere else on the stack. Doing [esp+C] will collide with
1383       // something on stack in current method, hence we go for [esp-C].
1384       // It is convenient since it is almost always in data cache, for
1385       // any small C.  We need to step back from SP to avoid data
1386       // dependencies with other things on below SP (callee-saves, for
1387       // example). Without a clear way to figure out the minimal safe
1388       // distance from SP, it makes sense to step back the complete
1389       // cache line, as this will also avoid possible second-order effects
1390       // with locked ops against the cache line. Our choice of offset
1391       // is bounded by x86 operand encoding, which should stay within
1392       // [-128; +127] to have the 8-byte displacement encoding.
1393       //
1394       // Any change to this code may need to revisit other places in
1395       // the code where this idiom is used, in particular the
1396       // orderAccess code.
1397 
1398       int offset = -VM_Version::L1_line_size();
1399       if (offset < -128) {
1400         offset = -128;
1401       }
1402 
1403       lock();
1404       addl(Address(rsp, offset), 0);// Assert the lock# signal here
1405     }
1406   }
1407 
1408   void mfence();
1409   void sfence();
1410 
1411   // Moves
1412 
1413   void mov64(Register dst, int64_t imm64);
1414 
1415   void movb(Address dst, Register src);
1416   void movb(Address dst, int imm8);
1417   void movb(Register dst, Address src);
1418 
1419   void movddup(XMMRegister dst, XMMRegister src);
1420 
1421   void kmovbl(KRegister dst, Register src);
1422   void kmovbl(Register dst, KRegister src);
1423   void kmovwl(KRegister dst, Register src);
1424   void kmovwl(KRegister dst, Address src);
1425   void kmovwl(Register dst, KRegister src);
1426   void kmovdl(KRegister dst, Register src);
1427   void kmovdl(Register dst, KRegister src);
1428   void kmovql(KRegister dst, KRegister src);
1429   void kmovql(Address dst, KRegister src);
1430   void kmovql(KRegister dst, Address src);
1431   void kmovql(KRegister dst, Register src);
1432   void kmovql(Register dst, KRegister src);
1433 
1434   void knotwl(KRegister dst, KRegister src);
1435 
1436   void kortestbl(KRegister dst, KRegister src);
1437   void kortestwl(KRegister dst, KRegister src);
1438   void kortestdl(KRegister dst, KRegister src);
1439   void kortestql(KRegister dst, KRegister src);
1440 
1441   void ktestq(KRegister src1, KRegister src2);
1442   void ktestd(KRegister src1, KRegister src2);
1443 
1444   void ktestql(KRegister dst, KRegister src);
1445 
1446   void movdl(XMMRegister dst, Register src);
1447   void movdl(Register dst, XMMRegister src);
1448   void movdl(XMMRegister dst, Address src);
1449   void movdl(Address dst, XMMRegister src);
1450 
1451   // Move Double Quadword
1452   void movdq(XMMRegister dst, Register src);
1453   void movdq(Register dst, XMMRegister src);
1454 
1455   // Move Aligned Double Quadword
1456   void movdqa(XMMRegister dst, XMMRegister src);
1457   void movdqa(XMMRegister dst, Address src);
1458 
1459   // Move Unaligned Double Quadword
1460   void movdqu(Address     dst, XMMRegister src);
1461   void movdqu(XMMRegister dst, Address src);
1462   void movdqu(XMMRegister dst, XMMRegister src);
1463 
1464   // Move Unaligned 256bit Vector
1465   void vmovdqu(Address dst, XMMRegister src);
1466   void vmovdqu(XMMRegister dst, Address src);
1467   void vmovdqu(XMMRegister dst, XMMRegister src);
1468 
1469    // Move Unaligned 512bit Vector
1470   void evmovdqub(Address dst, XMMRegister src, int vector_len);
1471   void evmovdqub(XMMRegister dst, Address src, int vector_len);
1472   void evmovdqub(XMMRegister dst, XMMRegister src, int vector_len);
1473   void evmovdqub(XMMRegister dst, KRegister mask, Address src, int vector_len);
1474   void evmovdquw(Address dst, XMMRegister src, int vector_len);
1475   void evmovdquw(Address dst, KRegister mask, XMMRegister src, int vector_len);
1476   void evmovdquw(XMMRegister dst, Address src, int vector_len);
1477   void evmovdquw(XMMRegister dst, KRegister mask, Address src, int vector_len);
1478   void evmovdqul(Address dst, XMMRegister src, int vector_len);
1479   void evmovdqul(XMMRegister dst, Address src, int vector_len);
1480   void evmovdqul(XMMRegister dst, XMMRegister src, int vector_len);
1481   void evmovdquq(Address dst, XMMRegister src, int vector_len);
1482   void evmovdquq(XMMRegister dst, Address src, int vector_len);
1483   void evmovdquq(XMMRegister dst, XMMRegister src, int vector_len);
1484 
1485   // Move lower 64bit to high 64bit in 128bit register
1486   void movlhps(XMMRegister dst, XMMRegister src);
1487 
1488   void movl(Register dst, int32_t imm32);
1489   void movl(Address dst, int32_t imm32);
1490   void movl(Register dst, Register src);
1491   void movl(Register dst, Address src);
1492   void movl(Address dst, Register src);
1493 
1494   // These dummies prevent using movl from converting a zero (like NULL) into Register
1495   // by giving the compiler two choices it can't resolve
1496 
1497   void movl(Address  dst, void* junk);
1498   void movl(Register dst, void* junk);
1499 
1500 #ifdef _LP64
1501   void movq(Register dst, Register src);
1502   void movq(Register dst, Address src);
1503   void movq(Address  dst, Register src);
1504 #endif
1505 
1506   void movq(Address     dst, MMXRegister src );
1507   void movq(MMXRegister dst, Address src );
1508 
1509 #ifdef _LP64
1510   // These dummies prevent using movq from converting a zero (like NULL) into Register
1511   // by giving the compiler two choices it can't resolve
1512 
1513   void movq(Address  dst, void* dummy);
1514   void movq(Register dst, void* dummy);
1515 #endif
1516 
1517   // Move Quadword
1518   void movq(Address     dst, XMMRegister src);
1519   void movq(XMMRegister dst, Address src);
1520 
1521   void movsbl(Register dst, Address src);
1522   void movsbl(Register dst, Register src);
1523 
1524 #ifdef _LP64
1525   void movsbq(Register dst, Address src);
1526   void movsbq(Register dst, Register src);
1527 
1528   // Move signed 32bit immediate to 64bit extending sign
1529   void movslq(Address  dst, int32_t imm64);
1530   void movslq(Register dst, int32_t imm64);
1531 
1532   void movslq(Register dst, Address src);
1533   void movslq(Register dst, Register src);
1534   void movslq(Register dst, void* src); // Dummy declaration to cause NULL to be ambiguous
1535 #endif
1536 
1537   void movswl(Register dst, Address src);
1538   void movswl(Register dst, Register src);
1539 
1540 #ifdef _LP64
1541   void movswq(Register dst, Address src);
1542   void movswq(Register dst, Register src);
1543 #endif
1544 
1545   void movw(Address dst, int imm16);
1546   void movw(Register dst, Address src);
1547   void movw(Address dst, Register src);
1548 
1549   void movzbl(Register dst, Address src);
1550   void movzbl(Register dst, Register src);
1551 
1552 #ifdef _LP64
1553   void movzbq(Register dst, Address src);
1554   void movzbq(Register dst, Register src);
1555 #endif
1556 
1557   void movzwl(Register dst, Address src);
1558   void movzwl(Register dst, Register src);
1559 
1560 #ifdef _LP64
1561   void movzwq(Register dst, Address src);
1562   void movzwq(Register dst, Register src);
1563 #endif
1564 
1565   // Unsigned multiply with RAX destination register
1566   void mull(Address src);
1567   void mull(Register src);
1568 
1569 #ifdef _LP64
1570   void mulq(Address src);
1571   void mulq(Register src);
1572   void mulxq(Register dst1, Register dst2, Register src);
1573 #endif
1574 
1575   // Multiply Scalar Double-Precision Floating-Point Values
1576   void mulsd(XMMRegister dst, Address src);
1577   void mulsd(XMMRegister dst, XMMRegister src);
1578 
1579   // Multiply Scalar Single-Precision Floating-Point Values
1580   void mulss(XMMRegister dst, Address src);
1581   void mulss(XMMRegister dst, XMMRegister src);
1582 
1583   void negl(Register dst);
1584 
1585 #ifdef _LP64
1586   void negq(Register dst);
1587 #endif
1588 
1589   void nop(int i = 1);
1590 
1591   void notl(Register dst);
1592 
1593 #ifdef _LP64
1594   void notq(Register dst);
1595 
1596   void btsq(Address dst, int imm8);
1597   void btrq(Address dst, int imm8);
1598 #endif
1599 
1600   void orl(Address dst, int32_t imm32);
1601   void orl(Register dst, int32_t imm32);
1602   void orl(Register dst, Address src);
1603   void orl(Register dst, Register src);
1604   void orl(Address dst, Register src);
1605 
1606   void orb(Address dst, int imm8);
1607 
1608   void orq(Address dst, int32_t imm32);
1609   void orq(Register dst, int32_t imm32);
1610   void orq(Register dst, Address src);
1611   void orq(Register dst, Register src);
1612 
1613   // Pack with unsigned saturation
1614   void packuswb(XMMRegister dst, XMMRegister src);
1615   void packuswb(XMMRegister dst, Address src);
1616   void vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1617 
1618   // Pemutation of 64bit words
1619   void vpermq(XMMRegister dst, XMMRegister src, int imm8, int vector_len);
1620   void vpermq(XMMRegister dst, XMMRegister src, int imm8);
1621   void vpermq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1622   void vperm2i128(XMMRegister dst,  XMMRegister nds, XMMRegister src, int imm8);
1623   void vperm2f128(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8);
1624   void evpermi2q(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1625 
1626   void pause();
1627 
1628   // Undefined Instruction
1629   void ud2();
1630 
1631   // SSE4.2 string instructions
1632   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
1633   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
1634 
1635   void pcmpeqb(XMMRegister dst, XMMRegister src);
1636   void vpcmpeqb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1637   void evpcmpeqb(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1638   void evpcmpeqb(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1639   void evpcmpeqb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_len);
1640 
1641   void evpcmpgtb(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1642   void evpcmpgtb(KRegister kdst, KRegister mask, XMMRegister nds, Address src, int vector_len);
1643 
1644   void evpcmpuw(KRegister kdst, XMMRegister nds, XMMRegister src, ComparisonPredicate vcc, int vector_len);
1645   void evpcmpuw(KRegister kdst, KRegister mask, XMMRegister nds, XMMRegister src, ComparisonPredicate of, int vector_len);
1646   void evpcmpuw(KRegister kdst, XMMRegister nds, Address src, ComparisonPredicate vcc, int vector_len);
1647 
1648   void pcmpeqw(XMMRegister dst, XMMRegister src);
1649   void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1650   void evpcmpeqw(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1651   void evpcmpeqw(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1652 
1653   void pcmpeqd(XMMRegister dst, XMMRegister src);
1654   void vpcmpeqd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1655   void evpcmpeqd(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1656   void evpcmpeqd(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1657 
1658   void pcmpeqq(XMMRegister dst, XMMRegister src);
1659   void vpcmpeqq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1660   void evpcmpeqq(KRegister kdst, XMMRegister nds, XMMRegister src, int vector_len);
1661   void evpcmpeqq(KRegister kdst, XMMRegister nds, Address src, int vector_len);
1662 
1663   void pmovmskb(Register dst, XMMRegister src);
1664   void vpmovmskb(Register dst, XMMRegister src);
1665 
1666   // SSE 4.1 extract
1667   void pextrd(Register dst, XMMRegister src, int imm8);
1668   void pextrq(Register dst, XMMRegister src, int imm8);
1669   void pextrd(Address dst, XMMRegister src, int imm8);
1670   void pextrq(Address dst, XMMRegister src, int imm8);
1671   void pextrb(Address dst, XMMRegister src, int imm8);
1672   // SSE 2 extract
1673   void pextrw(Register dst, XMMRegister src, int imm8);
1674   void pextrw(Address dst, XMMRegister src, int imm8);
1675 
1676   // SSE 4.1 insert
1677   void pinsrd(XMMRegister dst, Register src, int imm8);
1678   void pinsrq(XMMRegister dst, Register src, int imm8);
1679   void pinsrd(XMMRegister dst, Address src, int imm8);
1680   void pinsrq(XMMRegister dst, Address src, int imm8);
1681   void pinsrb(XMMRegister dst, Address src, int imm8);
1682   // SSE 2 insert
1683   void pinsrw(XMMRegister dst, Register src, int imm8);
1684   void pinsrw(XMMRegister dst, Address src, int imm8);
1685 
1686   // SSE4.1 packed move
1687   void pmovzxbw(XMMRegister dst, XMMRegister src);
1688   void pmovzxbw(XMMRegister dst, Address src);
1689 
1690   void vpmovzxbw( XMMRegister dst, Address src, int vector_len);
1691   void vpmovzxbw(XMMRegister dst, XMMRegister src, int vector_len);
1692   void evpmovzxbw(XMMRegister dst, KRegister mask, Address src, int vector_len);
1693 
1694   void evpmovwb(Address dst, XMMRegister src, int vector_len);
1695   void evpmovwb(Address dst, KRegister mask, XMMRegister src, int vector_len);
1696 
1697   void vpmovzxwd(XMMRegister dst, XMMRegister src, int vector_len);
1698 
1699   void evpmovdb(Address dst, XMMRegister src, int vector_len);
1700 
1701   // Sign extend moves
1702   void pmovsxbw(XMMRegister dst, XMMRegister src);
1703   void vpmovsxbw(XMMRegister dst, XMMRegister src, int vector_len);
1704 
1705   // Multiply add
1706   void pmaddwd(XMMRegister dst, XMMRegister src);
1707   void vpmaddwd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1708   // Multiply add accumulate
1709   void evpdpwssd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1710 
1711 #ifndef _LP64 // no 32bit push/pop on amd64
1712   void popl(Address dst);
1713 #endif
1714 
1715 #ifdef _LP64
1716   void popq(Address dst);
1717 #endif
1718 
1719   void popcntl(Register dst, Address src);
1720   void popcntl(Register dst, Register src);
1721 
1722   void vpopcntd(XMMRegister dst, XMMRegister src, int vector_len);
1723 
1724 #ifdef _LP64
1725   void popcntq(Register dst, Address src);
1726   void popcntq(Register dst, Register src);
1727 #endif
1728 
1729   // Prefetches (SSE, SSE2, 3DNOW only)
1730 
1731   void prefetchnta(Address src);
1732   void prefetchr(Address src);
1733   void prefetcht0(Address src);
1734   void prefetcht1(Address src);
1735   void prefetcht2(Address src);
1736   void prefetchw(Address src);
1737 
1738   // Shuffle Bytes
1739   void pshufb(XMMRegister dst, XMMRegister src);
1740   void pshufb(XMMRegister dst, Address src);
1741   void vpshufb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1742 
1743   // Shuffle Packed Doublewords
1744   void pshufd(XMMRegister dst, XMMRegister src, int mode);
1745   void pshufd(XMMRegister dst, Address src,     int mode);
1746   void vpshufd(XMMRegister dst, XMMRegister src, int mode, int vector_len);
1747 
1748   // Shuffle Packed Low Words
1749   void pshuflw(XMMRegister dst, XMMRegister src, int mode);
1750   void pshuflw(XMMRegister dst, Address src,     int mode);
1751 
1752   // Shuffle packed values at 128 bit granularity
1753   void evshufi64x2(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8, int vector_len);
1754 
1755   // Shift Right by bytes Logical DoubleQuadword Immediate
1756   void psrldq(XMMRegister dst, int shift);
1757   // Shift Left by bytes Logical DoubleQuadword Immediate
1758   void pslldq(XMMRegister dst, int shift);
1759 
1760   // Logical Compare 128bit
1761   void ptest(XMMRegister dst, XMMRegister src);
1762   void ptest(XMMRegister dst, Address src);
1763   // Logical Compare 256bit
1764   void vptest(XMMRegister dst, XMMRegister src);
1765   void vptest(XMMRegister dst, Address src);
1766 
1767   // Interleave Low Bytes
1768   void punpcklbw(XMMRegister dst, XMMRegister src);
1769   void punpcklbw(XMMRegister dst, Address src);
1770 
1771   // Interleave Low Doublewords
1772   void punpckldq(XMMRegister dst, XMMRegister src);
1773   void punpckldq(XMMRegister dst, Address src);
1774 
1775   // Interleave Low Quadwords
1776   void punpcklqdq(XMMRegister dst, XMMRegister src);
1777 
1778 #ifndef _LP64 // no 32bit push/pop on amd64
1779   void pushl(Address src);
1780 #endif
1781 
1782   void pushq(Address src);
1783 
1784   void rcll(Register dst, int imm8);
1785 
1786   void rclq(Register dst, int imm8);
1787 
1788   void rcrq(Register dst, int imm8);
1789 
1790   void rcpps(XMMRegister dst, XMMRegister src);
1791 
1792   void rcpss(XMMRegister dst, XMMRegister src);
1793 
1794   void rdtsc();
1795 
1796   void ret(int imm16);
1797 
1798 #ifdef _LP64
1799   void rorq(Register dst, int imm8);
1800   void rorxq(Register dst, Register src, int imm8);
1801   void rorxd(Register dst, Register src, int imm8);
1802 #endif
1803 
1804   void sahf();
1805 
1806   void sarl(Register dst, int imm8);
1807   void sarl(Register dst);
1808 
1809   void sarq(Register dst, int imm8);
1810   void sarq(Register dst);
1811 
1812   void sbbl(Address dst, int32_t imm32);
1813   void sbbl(Register dst, int32_t imm32);
1814   void sbbl(Register dst, Address src);
1815   void sbbl(Register dst, Register src);
1816 
1817   void sbbq(Address dst, int32_t imm32);
1818   void sbbq(Register dst, int32_t imm32);
1819   void sbbq(Register dst, Address src);
1820   void sbbq(Register dst, Register src);
1821 
1822   void setb(Condition cc, Register dst);
1823 
1824   void palignr(XMMRegister dst, XMMRegister src, int imm8);
1825   void vpalignr(XMMRegister dst, XMMRegister src1, XMMRegister src2, int imm8, int vector_len);
1826   void evalignq(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
1827 
1828   void pblendw(XMMRegister dst, XMMRegister src, int imm8);
1829 
1830   void sha1rnds4(XMMRegister dst, XMMRegister src, int imm8);
1831   void sha1nexte(XMMRegister dst, XMMRegister src);
1832   void sha1msg1(XMMRegister dst, XMMRegister src);
1833   void sha1msg2(XMMRegister dst, XMMRegister src);
1834   // xmm0 is implicit additional source to the following instruction.
1835   void sha256rnds2(XMMRegister dst, XMMRegister src);
1836   void sha256msg1(XMMRegister dst, XMMRegister src);
1837   void sha256msg2(XMMRegister dst, XMMRegister src);
1838 
1839   void shldl(Register dst, Register src);
1840   void shldl(Register dst, Register src, int8_t imm8);
1841 
1842   void shll(Register dst, int imm8);
1843   void shll(Register dst);
1844 
1845   void shlq(Register dst, int imm8);
1846   void shlq(Register dst);
1847 
1848   void shrdl(Register dst, Register src);
1849 
1850   void shrl(Register dst, int imm8);
1851   void shrl(Register dst);
1852 
1853   void shrq(Register dst, int imm8);
1854   void shrq(Register dst);
1855 
1856   void smovl(); // QQQ generic?
1857 
1858   // Compute Square Root of Scalar Double-Precision Floating-Point Value
1859   void sqrtsd(XMMRegister dst, Address src);
1860   void sqrtsd(XMMRegister dst, XMMRegister src);
1861 
1862   void roundsd(XMMRegister dst, Address src, int32_t rmode);
1863   void roundsd(XMMRegister dst, XMMRegister src, int32_t rmode);
1864 
1865   // Compute Square Root of Scalar Single-Precision Floating-Point Value
1866   void sqrtss(XMMRegister dst, Address src);
1867   void sqrtss(XMMRegister dst, XMMRegister src);
1868 
1869   void std();
1870 
1871   void stmxcsr( Address dst );
1872 
1873   void subl(Address dst, int32_t imm32);
1874   void subl(Address dst, Register src);
1875   void subl(Register dst, int32_t imm32);
1876   void subl(Register dst, Address src);
1877   void subl(Register dst, Register src);
1878 
1879   void subq(Address dst, int32_t imm32);
1880   void subq(Address dst, Register src);
1881   void subq(Register dst, int32_t imm32);
1882   void subq(Register dst, Address src);
1883   void subq(Register dst, Register src);
1884 
1885   // Force generation of a 4 byte immediate value even if it fits into 8bit
1886   void subl_imm32(Register dst, int32_t imm32);
1887   void subq_imm32(Register dst, int32_t imm32);
1888 
1889   // Subtract Scalar Double-Precision Floating-Point Values
1890   void subsd(XMMRegister dst, Address src);
1891   void subsd(XMMRegister dst, XMMRegister src);
1892 
1893   // Subtract Scalar Single-Precision Floating-Point Values
1894   void subss(XMMRegister dst, Address src);
1895   void subss(XMMRegister dst, XMMRegister src);
1896 
1897   void testb(Register dst, int imm8);
1898   void testb(Address dst, int imm8);
1899 
1900   void testl(Register dst, int32_t imm32);
1901   void testl(Register dst, Register src);
1902   void testl(Register dst, Address src);
1903 
1904   void testq(Register dst, int32_t imm32);
1905   void testq(Register dst, Register src);
1906   void testq(Register dst, Address src);
1907 
1908   // BMI - count trailing zeros
1909   void tzcntl(Register dst, Register src);
1910   void tzcntq(Register dst, Register src);
1911 
1912   // Unordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1913   void ucomisd(XMMRegister dst, Address src);
1914   void ucomisd(XMMRegister dst, XMMRegister src);
1915 
1916   // Unordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1917   void ucomiss(XMMRegister dst, Address src);
1918   void ucomiss(XMMRegister dst, XMMRegister src);
1919 
1920   void xabort(int8_t imm8);
1921 
1922   void xaddb(Address dst, Register src);
1923   void xaddw(Address dst, Register src);
1924   void xaddl(Address dst, Register src);
1925   void xaddq(Address dst, Register src);
1926 
1927   void xbegin(Label& abort, relocInfo::relocType rtype = relocInfo::none);
1928 
1929   void xchgb(Register reg, Address adr);
1930   void xchgw(Register reg, Address adr);
1931   void xchgl(Register reg, Address adr);
1932   void xchgl(Register dst, Register src);
1933 
1934   void xchgq(Register reg, Address adr);
1935   void xchgq(Register dst, Register src);
1936 
1937   void xend();
1938 
1939   // Get Value of Extended Control Register
1940   void xgetbv();
1941 
1942   void xorl(Register dst, int32_t imm32);
1943   void xorl(Register dst, Address src);
1944   void xorl(Register dst, Register src);
1945 
1946   void xorb(Register dst, Address src);
1947 
1948   void xorq(Register dst, Address src);
1949   void xorq(Register dst, Register src);
1950 
1951   void set_byte_if_not_zero(Register dst); // sets reg to 1 if not zero, otherwise 0
1952 
1953   // AVX 3-operands scalar instructions (encoded with VEX prefix)
1954 
1955   void vaddsd(XMMRegister dst, XMMRegister nds, Address src);
1956   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1957   void vaddss(XMMRegister dst, XMMRegister nds, Address src);
1958   void vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1959   void vdivsd(XMMRegister dst, XMMRegister nds, Address src);
1960   void vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1961   void vdivss(XMMRegister dst, XMMRegister nds, Address src);
1962   void vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1963   void vfmadd231sd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1964   void vfmadd231ss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1965   void vmulsd(XMMRegister dst, XMMRegister nds, Address src);
1966   void vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1967   void vmulss(XMMRegister dst, XMMRegister nds, Address src);
1968   void vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1969   void vsubsd(XMMRegister dst, XMMRegister nds, Address src);
1970   void vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1971   void vsubss(XMMRegister dst, XMMRegister nds, Address src);
1972   void vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1973 
1974   void vmaxss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1975   void vmaxsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1976   void vminss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1977   void vminsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1978 
1979   void shlxl(Register dst, Register src1, Register src2);
1980   void shlxq(Register dst, Register src1, Register src2);
1981 
1982   //====================VECTOR ARITHMETIC=====================================
1983 
1984   // Add Packed Floating-Point Values
1985   void addpd(XMMRegister dst, XMMRegister src);
1986   void addpd(XMMRegister dst, Address src);
1987   void addps(XMMRegister dst, XMMRegister src);
1988   void vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1989   void vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1990   void vaddpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1991   void vaddps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1992 
1993   // Subtract Packed Floating-Point Values
1994   void subpd(XMMRegister dst, XMMRegister src);
1995   void subps(XMMRegister dst, XMMRegister src);
1996   void vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1997   void vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1998   void vsubpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1999   void vsubps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2000 
2001   // Multiply Packed Floating-Point Values
2002   void mulpd(XMMRegister dst, XMMRegister src);
2003   void mulpd(XMMRegister dst, Address src);
2004   void mulps(XMMRegister dst, XMMRegister src);
2005   void vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2006   void vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2007   void vmulpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2008   void vmulps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2009 
2010   void vfmadd231pd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2011   void vfmadd231ps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2012   void vfmadd231pd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2013   void vfmadd231ps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2014 
2015   // Divide Packed Floating-Point Values
2016   void divpd(XMMRegister dst, XMMRegister src);
2017   void divps(XMMRegister dst, XMMRegister src);
2018   void vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2019   void vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2020   void vdivpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2021   void vdivps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2022 
2023   // Sqrt Packed Floating-Point Values
2024   void vsqrtpd(XMMRegister dst, XMMRegister src, int vector_len);
2025   void vsqrtpd(XMMRegister dst, Address src, int vector_len);
2026   void vsqrtps(XMMRegister dst, XMMRegister src, int vector_len);
2027   void vsqrtps(XMMRegister dst, Address src, int vector_len);
2028 
2029   // Round Packed Double precision value.
2030   void vroundpd(XMMRegister dst, XMMRegister src, int32_t rmode, int vector_len);
2031   void vroundpd(XMMRegister dst, Address src, int32_t rmode, int vector_len);
2032   void vrndscalepd(XMMRegister dst,  XMMRegister src,  int32_t rmode, int vector_len);
2033   void vrndscalepd(XMMRegister dst, Address src, int32_t rmode, int vector_len);
2034 
2035   // Bitwise Logical AND of Packed Floating-Point Values
2036   void andpd(XMMRegister dst, XMMRegister src);
2037   void andps(XMMRegister dst, XMMRegister src);
2038   void vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2039   void vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2040   void vandpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2041   void vandps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2042 
2043   void unpckhpd(XMMRegister dst, XMMRegister src);
2044   void unpcklpd(XMMRegister dst, XMMRegister src);
2045 
2046   // Bitwise Logical XOR of Packed Floating-Point Values
2047   void xorpd(XMMRegister dst, XMMRegister src);
2048   void xorps(XMMRegister dst, XMMRegister src);
2049   void vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2050   void vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2051   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2052   void vxorps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2053 
2054   // Add horizontal packed integers
2055   void vphaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2056   void vphaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2057   void phaddw(XMMRegister dst, XMMRegister src);
2058   void phaddd(XMMRegister dst, XMMRegister src);
2059 
2060   // Add packed integers
2061   void paddb(XMMRegister dst, XMMRegister src);
2062   void paddw(XMMRegister dst, XMMRegister src);
2063   void paddd(XMMRegister dst, XMMRegister src);
2064   void paddd(XMMRegister dst, Address src);
2065   void paddq(XMMRegister dst, XMMRegister src);
2066   void vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2067   void vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2068   void vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2069   void vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2070   void vpaddb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2071   void vpaddw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2072   void vpaddd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2073   void vpaddq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2074 
2075   // Sub packed integers
2076   void psubb(XMMRegister dst, XMMRegister src);
2077   void psubw(XMMRegister dst, XMMRegister src);
2078   void psubd(XMMRegister dst, XMMRegister src);
2079   void psubq(XMMRegister dst, XMMRegister src);
2080   void vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2081   void vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2082   void vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2083   void vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2084   void vpsubb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2085   void vpsubw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2086   void vpsubd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2087   void vpsubq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2088 
2089   // Multiply packed integers (only shorts and ints)
2090   void pmullw(XMMRegister dst, XMMRegister src);
2091   void pmulld(XMMRegister dst, XMMRegister src);
2092   void vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2093   void vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2094   void vpmullq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2095   void vpmullw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2096   void vpmulld(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2097   void vpmullq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2098 
2099   // Shift left packed integers
2100   void psllw(XMMRegister dst, int shift);
2101   void pslld(XMMRegister dst, int shift);
2102   void psllq(XMMRegister dst, int shift);
2103   void psllw(XMMRegister dst, XMMRegister shift);
2104   void pslld(XMMRegister dst, XMMRegister shift);
2105   void psllq(XMMRegister dst, XMMRegister shift);
2106   void vpsllw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2107   void vpslld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2108   void vpsllq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2109   void vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2110   void vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2111   void vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2112   void vpslldq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2113 
2114   // Logical shift right packed integers
2115   void psrlw(XMMRegister dst, int shift);
2116   void psrld(XMMRegister dst, int shift);
2117   void psrlq(XMMRegister dst, int shift);
2118   void psrlw(XMMRegister dst, XMMRegister shift);
2119   void psrld(XMMRegister dst, XMMRegister shift);
2120   void psrlq(XMMRegister dst, XMMRegister shift);
2121   void vpsrlw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2122   void vpsrld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2123   void vpsrlq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2124   void vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2125   void vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2126   void vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2127   void vpsrldq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2128   void evpsrlvw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2129   void evpsllvw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2130 
2131   // Arithmetic shift right packed integers (only shorts and ints, no instructions for longs)
2132   void psraw(XMMRegister dst, int shift);
2133   void psrad(XMMRegister dst, int shift);
2134   void psraw(XMMRegister dst, XMMRegister shift);
2135   void psrad(XMMRegister dst, XMMRegister shift);
2136   void vpsraw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2137   void vpsrad(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2138   void vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2139   void vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2140   void evpsraq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2141   void evpsraq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2142 
2143   // And packed integers
2144   void pand(XMMRegister dst, XMMRegister src);
2145   void vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2146   void vpand(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2147   void vpandq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2148 
2149   // Andn packed integers
2150   void pandn(XMMRegister dst, XMMRegister src);
2151   void vpandn(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2152 
2153   // Or packed integers
2154   void por(XMMRegister dst, XMMRegister src);
2155   void vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2156   void vpor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2157   void vporq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2158 
2159   // Xor packed integers
2160   void pxor(XMMRegister dst, XMMRegister src);
2161   void vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2162   void vpxor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2163   void evpxorq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2164   void evpxorq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2165 
2166 
2167   // vinserti forms
2168   void vinserti128(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2169   void vinserti128(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2170   void vinserti32x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2171   void vinserti32x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2172   void vinserti64x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2173 
2174   // vinsertf forms
2175   void vinsertf128(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2176   void vinsertf128(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2177   void vinsertf32x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2178   void vinsertf32x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2179   void vinsertf64x4(XMMRegister dst, XMMRegister nds, XMMRegister src, uint8_t imm8);
2180   void vinsertf64x4(XMMRegister dst, XMMRegister nds, Address src, uint8_t imm8);
2181 
2182   // vextracti forms
2183   void vextracti128(XMMRegister dst, XMMRegister src, uint8_t imm8);
2184   void vextracti128(Address dst, XMMRegister src, uint8_t imm8);
2185   void vextracti32x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2186   void vextracti32x4(Address dst, XMMRegister src, uint8_t imm8);
2187   void vextracti64x2(XMMRegister dst, XMMRegister src, uint8_t imm8);
2188   void vextracti64x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2189   void vextracti64x4(Address dst, XMMRegister src, uint8_t imm8);
2190 
2191   // vextractf forms
2192   void vextractf128(XMMRegister dst, XMMRegister src, uint8_t imm8);
2193   void vextractf128(Address dst, XMMRegister src, uint8_t imm8);
2194   void vextractf32x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2195   void vextractf32x4(Address dst, XMMRegister src, uint8_t imm8);
2196   void vextractf64x2(XMMRegister dst, XMMRegister src, uint8_t imm8);
2197   void vextractf64x4(XMMRegister dst, XMMRegister src, uint8_t imm8);
2198   void vextractf64x4(Address dst, XMMRegister src, uint8_t imm8);
2199 
2200   // xmm/mem sourced byte/word/dword/qword replicate
2201   void vpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len);
2202   void vpbroadcastb(XMMRegister dst, Address src, int vector_len);
2203   void vpbroadcastw(XMMRegister dst, XMMRegister src, int vector_len);
2204   void vpbroadcastw(XMMRegister dst, Address src, int vector_len);
2205   void vpbroadcastd(XMMRegister dst, XMMRegister src, int vector_len);
2206   void vpbroadcastd(XMMRegister dst, Address src, int vector_len);
2207   void vpbroadcastq(XMMRegister dst, XMMRegister src, int vector_len);
2208   void vpbroadcastq(XMMRegister dst, Address src, int vector_len);
2209 
2210   void evbroadcasti64x2(XMMRegister dst, XMMRegister src, int vector_len);
2211   void evbroadcasti64x2(XMMRegister dst, Address src, int vector_len);
2212 
2213   // scalar single/double precision replicate
2214   void vpbroadcastss(XMMRegister dst, XMMRegister src, int vector_len);
2215   void vpbroadcastss(XMMRegister dst, Address src, int vector_len);
2216   void vpbroadcastsd(XMMRegister dst, XMMRegister src, int vector_len);
2217   void vpbroadcastsd(XMMRegister dst, Address src, int vector_len);
2218 
2219   // gpr sourced byte/word/dword/qword replicate
2220   void evpbroadcastb(XMMRegister dst, Register src, int vector_len);
2221   void evpbroadcastw(XMMRegister dst, Register src, int vector_len);
2222   void evpbroadcastd(XMMRegister dst, Register src, int vector_len);
2223   void evpbroadcastq(XMMRegister dst, Register src, int vector_len);
2224 
2225   void evpgatherdd(XMMRegister dst, KRegister k1, Address src, int vector_len);
2226 
2227   // Carry-Less Multiplication Quadword
2228   void pclmulqdq(XMMRegister dst, XMMRegister src, int mask);
2229   void vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask);
2230   void evpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask, int vector_len);
2231   // AVX instruction which is used to clear upper 128 bits of YMM registers and
2232   // to avoid transaction penalty between AVX and SSE states. There is no
2233   // penalty if legacy SSE instructions are encoded using VEX prefix because
2234   // they always clear upper 128 bits. It should be used before calling
2235   // runtime code and native libraries.
2236   void vzeroupper();
2237 
2238   // AVX support for vectorized conditional move (float/double). The following two instructions used only coupled.
2239   void cmppd(XMMRegister dst, XMMRegister nds, XMMRegister src, int cop, int vector_len);
2240   void blendvpd(XMMRegister dst, XMMRegister nds, XMMRegister src1, XMMRegister src2, int vector_len);
2241   void cmpps(XMMRegister dst, XMMRegister nds, XMMRegister src, int cop, int vector_len);
2242   void blendvps(XMMRegister dst, XMMRegister nds, XMMRegister src1, XMMRegister src2, int vector_len);
2243   void vpblendd(XMMRegister dst, XMMRegister nds, XMMRegister src, int imm8, int vector_len);
2244 
2245  protected:
2246   // Next instructions require address alignment 16 bytes SSE mode.
2247   // They should be called only from corresponding MacroAssembler instructions.
2248   void andpd(XMMRegister dst, Address src);
2249   void andps(XMMRegister dst, Address src);
2250   void xorpd(XMMRegister dst, Address src);
2251   void xorps(XMMRegister dst, Address src);
2252 
2253 };
2254 
2255 // The Intel x86/Amd64 Assembler attributes: All fields enclosed here are to guide encoding level decisions.
2256 // Specific set functions are for specialized use, else defaults or whatever was supplied to object construction
2257 // are applied.
2258 class InstructionAttr {
2259 public:
2260   InstructionAttr(
2261     int vector_len,     // The length of vector to be applied in encoding - for both AVX and EVEX
2262     bool rex_vex_w,     // Width of data: if 32-bits or less, false, else if 64-bit or specially defined, true
2263     bool legacy_mode,   // Details if either this instruction is conditionally encoded to AVX or earlier if true else possibly EVEX
2264     bool no_reg_mask,   // when true, k0 is used when EVEX encoding is chosen, else embedded_opmask_register_specifier is used
2265     bool uses_vl)       // This instruction may have legacy constraints based on vector length for EVEX
2266     :
2267       _avx_vector_len(vector_len),
2268       _rex_vex_w(rex_vex_w),
2269       _rex_vex_w_reverted(false),
2270       _legacy_mode(legacy_mode),
2271       _no_reg_mask(no_reg_mask),
2272       _uses_vl(uses_vl),
2273       _tuple_type(Assembler::EVEX_ETUP),
2274       _input_size_in_bits(Assembler::EVEX_NObit),
2275       _is_evex_instruction(false),
2276       _evex_encoding(0),
2277       _is_clear_context(true),
2278       _is_extended_context(false),
2279       _embedded_opmask_register_specifier(0), // hard code k0
2280       _current_assembler(NULL) {
2281     if (UseAVX < 3) _legacy_mode = true;
2282   }
2283 
2284   ~InstructionAttr() {
2285     if (_current_assembler != NULL) {
2286       _current_assembler->clear_attributes();
2287     }
2288     _current_assembler = NULL;
2289   }
2290 
2291 private:
2292   int  _avx_vector_len;
2293   bool _rex_vex_w;
2294   bool _rex_vex_w_reverted;
2295   bool _legacy_mode;
2296   bool _no_reg_mask;
2297   bool _uses_vl;
2298   int  _tuple_type;
2299   int  _input_size_in_bits;
2300   bool _is_evex_instruction;
2301   int  _evex_encoding;
2302   bool _is_clear_context;
2303   bool _is_extended_context;
2304   int _embedded_opmask_register_specifier;
2305 
2306   Assembler *_current_assembler;
2307 
2308 public:
2309   // query functions for field accessors
2310   int  get_vector_len(void) const { return _avx_vector_len; }
2311   bool is_rex_vex_w(void) const { return _rex_vex_w; }
2312   bool is_rex_vex_w_reverted(void) { return _rex_vex_w_reverted; }
2313   bool is_legacy_mode(void) const { return _legacy_mode; }
2314   bool is_no_reg_mask(void) const { return _no_reg_mask; }
2315   bool uses_vl(void) const { return _uses_vl; }
2316   int  get_tuple_type(void) const { return _tuple_type; }
2317   int  get_input_size(void) const { return _input_size_in_bits; }
2318   int  is_evex_instruction(void) const { return _is_evex_instruction; }
2319   int  get_evex_encoding(void) const { return _evex_encoding; }
2320   bool is_clear_context(void) const { return _is_clear_context; }
2321   bool is_extended_context(void) const { return _is_extended_context; }
2322   int get_embedded_opmask_register_specifier(void) const { return _embedded_opmask_register_specifier; }
2323 
2324   // Set the vector len manually
2325   void set_vector_len(int vector_len) { _avx_vector_len = vector_len; }
2326 
2327   // Set revert rex_vex_w for avx encoding
2328   void set_rex_vex_w_reverted(void) { _rex_vex_w_reverted = true; }
2329 
2330   // Set rex_vex_w based on state
2331   void set_rex_vex_w(bool state) { _rex_vex_w = state; }
2332 
2333   // Set the instruction to be encoded in AVX mode
2334   void set_is_legacy_mode(void) { _legacy_mode = true; }
2335 
2336   // Set the current instuction to be encoded as an EVEX instuction
2337   void set_is_evex_instruction(void) { _is_evex_instruction = true; }
2338 
2339   // Internal encoding data used in compressed immediate offset programming
2340   void set_evex_encoding(int value) { _evex_encoding = value; }
2341 
2342   // Set the Evex.Z field to be used to clear all non directed XMM/YMM/ZMM components
2343   void reset_is_clear_context(void) { _is_clear_context = false; }
2344 
2345   // Map back to current asembler so that we can manage object level assocation
2346   void set_current_assembler(Assembler *current_assembler) { _current_assembler = current_assembler; }
2347 
2348   // Address modifiers used for compressed displacement calculation
2349   void set_address_attributes(int tuple_type, int input_size_in_bits) {
2350     if (VM_Version::supports_evex()) {
2351       _tuple_type = tuple_type;
2352       _input_size_in_bits = input_size_in_bits;
2353     }
2354   }
2355 
2356   // Set embedded opmask register specifier.
2357   void set_embedded_opmask_register_specifier(KRegister mask) {
2358     _embedded_opmask_register_specifier = (*mask).encoding() & 0x7;
2359   }
2360 
2361 };
2362 
2363 #endif // CPU_X86_ASSEMBLER_X86_HPP