1 /*
   2  * Copyright (c) 1997, 2015, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #ifndef CPU_X86_VM_ASSEMBLER_X86_HPP
  26 #define CPU_X86_VM_ASSEMBLER_X86_HPP
  27 
  28 #include "asm/register.hpp"
  29 #include "vm_version_x86.hpp"
  30 
  31 class BiasedLockingCounters;
  32 
  33 // Contains all the definitions needed for x86 assembly code generation.
  34 
  35 // Calling convention
  36 class Argument VALUE_OBJ_CLASS_SPEC {
  37  public:
  38   enum {
  39 #ifdef _LP64
  40 #ifdef _WIN64
  41     n_int_register_parameters_c   = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...)
  42     n_float_register_parameters_c = 4,  // xmm0 - xmm3 (c_farg0, c_farg1, ... )
  43 #else
  44     n_int_register_parameters_c   = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...)
  45     n_float_register_parameters_c = 8,  // xmm0 - xmm7 (c_farg0, c_farg1, ... )
  46 #endif // _WIN64
  47     n_int_register_parameters_j   = 6, // j_rarg0, j_rarg1, ...
  48     n_float_register_parameters_j = 8  // j_farg0, j_farg1, ...
  49 #else
  50     n_register_parameters = 0   // 0 registers used to pass arguments
  51 #endif // _LP64
  52   };
  53 };
  54 
  55 
  56 #ifdef _LP64
  57 // Symbolically name the register arguments used by the c calling convention.
  58 // Windows is different from linux/solaris. So much for standards...
  59 
  60 #ifdef _WIN64
  61 
  62 REGISTER_DECLARATION(Register, c_rarg0, rcx);
  63 REGISTER_DECLARATION(Register, c_rarg1, rdx);
  64 REGISTER_DECLARATION(Register, c_rarg2, r8);
  65 REGISTER_DECLARATION(Register, c_rarg3, r9);
  66 
  67 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  68 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  69 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  70 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  71 
  72 #else
  73 
  74 REGISTER_DECLARATION(Register, c_rarg0, rdi);
  75 REGISTER_DECLARATION(Register, c_rarg1, rsi);
  76 REGISTER_DECLARATION(Register, c_rarg2, rdx);
  77 REGISTER_DECLARATION(Register, c_rarg3, rcx);
  78 REGISTER_DECLARATION(Register, c_rarg4, r8);
  79 REGISTER_DECLARATION(Register, c_rarg5, r9);
  80 
  81 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  82 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  83 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  84 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  85 REGISTER_DECLARATION(XMMRegister, c_farg4, xmm4);
  86 REGISTER_DECLARATION(XMMRegister, c_farg5, xmm5);
  87 REGISTER_DECLARATION(XMMRegister, c_farg6, xmm6);
  88 REGISTER_DECLARATION(XMMRegister, c_farg7, xmm7);
  89 
  90 #endif // _WIN64
  91 
  92 // Symbolically name the register arguments used by the Java calling convention.
  93 // We have control over the convention for java so we can do what we please.
  94 // What pleases us is to offset the java calling convention so that when
  95 // we call a suitable jni method the arguments are lined up and we don't
  96 // have to do little shuffling. A suitable jni method is non-static and a
  97 // small number of arguments (two fewer args on windows)
  98 //
  99 //        |-------------------------------------------------------|
 100 //        | c_rarg0   c_rarg1  c_rarg2 c_rarg3 c_rarg4 c_rarg5    |
 101 //        |-------------------------------------------------------|
 102 //        | rcx       rdx      r8      r9      rdi*    rsi*       | windows (* not a c_rarg)
 103 //        | rdi       rsi      rdx     rcx     r8      r9         | solaris/linux
 104 //        |-------------------------------------------------------|
 105 //        | j_rarg5   j_rarg0  j_rarg1 j_rarg2 j_rarg3 j_rarg4    |
 106 //        |-------------------------------------------------------|
 107 
 108 REGISTER_DECLARATION(Register, j_rarg0, c_rarg1);
 109 REGISTER_DECLARATION(Register, j_rarg1, c_rarg2);
 110 REGISTER_DECLARATION(Register, j_rarg2, c_rarg3);
 111 // Windows runs out of register args here
 112 #ifdef _WIN64
 113 REGISTER_DECLARATION(Register, j_rarg3, rdi);
 114 REGISTER_DECLARATION(Register, j_rarg4, rsi);
 115 #else
 116 REGISTER_DECLARATION(Register, j_rarg3, c_rarg4);
 117 REGISTER_DECLARATION(Register, j_rarg4, c_rarg5);
 118 #endif /* _WIN64 */
 119 REGISTER_DECLARATION(Register, j_rarg5, c_rarg0);
 120 
 121 REGISTER_DECLARATION(XMMRegister, j_farg0, xmm0);
 122 REGISTER_DECLARATION(XMMRegister, j_farg1, xmm1);
 123 REGISTER_DECLARATION(XMMRegister, j_farg2, xmm2);
 124 REGISTER_DECLARATION(XMMRegister, j_farg3, xmm3);
 125 REGISTER_DECLARATION(XMMRegister, j_farg4, xmm4);
 126 REGISTER_DECLARATION(XMMRegister, j_farg5, xmm5);
 127 REGISTER_DECLARATION(XMMRegister, j_farg6, xmm6);
 128 REGISTER_DECLARATION(XMMRegister, j_farg7, xmm7);
 129 
 130 REGISTER_DECLARATION(Register, rscratch1, r10);  // volatile
 131 REGISTER_DECLARATION(Register, rscratch2, r11);  // volatile
 132 
 133 REGISTER_DECLARATION(Register, r12_heapbase, r12); // callee-saved
 134 REGISTER_DECLARATION(Register, r15_thread, r15); // callee-saved
 135 
 136 #else
 137 // rscratch1 will apear in 32bit code that is dead but of course must compile
 138 // Using noreg ensures if the dead code is incorrectly live and executed it
 139 // will cause an assertion failure
 140 #define rscratch1 noreg
 141 #define rscratch2 noreg
 142 
 143 #endif // _LP64
 144 
 145 // JSR 292
 146 // On x86, the SP does not have to be saved when invoking method handle intrinsics
 147 // or compiled lambda forms. We indicate that by setting rbp_mh_SP_save to noreg.
 148 REGISTER_DECLARATION(Register, rbp_mh_SP_save, noreg);
 149 
 150 // Address is an abstraction used to represent a memory location
 151 // using any of the amd64 addressing modes with one object.
 152 //
 153 // Note: A register location is represented via a Register, not
 154 //       via an address for efficiency & simplicity reasons.
 155 
 156 class ArrayAddress;
 157 
 158 class Address VALUE_OBJ_CLASS_SPEC {
 159  public:
 160   enum ScaleFactor {
 161     no_scale = -1,
 162     times_1  =  0,
 163     times_2  =  1,
 164     times_4  =  2,
 165     times_8  =  3,
 166     times_ptr = LP64_ONLY(times_8) NOT_LP64(times_4)
 167   };
 168   static ScaleFactor times(int size) {
 169     assert(size >= 1 && size <= 8 && is_power_of_2(size), "bad scale size");
 170     if (size == 8)  return times_8;
 171     if (size == 4)  return times_4;
 172     if (size == 2)  return times_2;
 173     return times_1;
 174   }
 175   static int scale_size(ScaleFactor scale) {
 176     assert(scale != no_scale, "");
 177     assert(((1 << (int)times_1) == 1 &&
 178             (1 << (int)times_2) == 2 &&
 179             (1 << (int)times_4) == 4 &&
 180             (1 << (int)times_8) == 8), "");
 181     return (1 << (int)scale);
 182   }
 183 
 184  private:
 185   Register         _base;
 186   Register         _index;
 187   ScaleFactor      _scale;
 188   int              _disp;
 189   RelocationHolder _rspec;
 190 
 191   // Easily misused constructors make them private
 192   // %%% can we make these go away?
 193   NOT_LP64(Address(address loc, RelocationHolder spec);)
 194   Address(int disp, address loc, relocInfo::relocType rtype);
 195   Address(int disp, address loc, RelocationHolder spec);
 196 
 197  public:
 198 
 199  int disp() { return _disp; }
 200   // creation
 201   Address()
 202     : _base(noreg),
 203       _index(noreg),
 204       _scale(no_scale),
 205       _disp(0) {
 206   }
 207 
 208   // No default displacement otherwise Register can be implicitly
 209   // converted to 0(Register) which is quite a different animal.
 210 
 211   Address(Register base, int disp)
 212     : _base(base),
 213       _index(noreg),
 214       _scale(no_scale),
 215       _disp(disp) {
 216   }
 217 
 218   Address(Register base, Register index, ScaleFactor scale, int disp = 0)
 219     : _base (base),
 220       _index(index),
 221       _scale(scale),
 222       _disp (disp) {
 223     assert(!index->is_valid() == (scale == Address::no_scale),
 224            "inconsistent address");
 225   }
 226 
 227   Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0)
 228     : _base (base),
 229       _index(index.register_or_noreg()),
 230       _scale(scale),
 231       _disp (disp + (index.constant_or_zero() * scale_size(scale))) {
 232     if (!index.is_register())  scale = Address::no_scale;
 233     assert(!_index->is_valid() == (scale == Address::no_scale),
 234            "inconsistent address");
 235   }
 236 
 237   Address plus_disp(int disp) const {
 238     Address a = (*this);
 239     a._disp += disp;
 240     return a;
 241   }
 242   Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const {
 243     Address a = (*this);
 244     a._disp += disp.constant_or_zero() * scale_size(scale);
 245     if (disp.is_register()) {
 246       assert(!a.index()->is_valid(), "competing indexes");
 247       a._index = disp.as_register();
 248       a._scale = scale;
 249     }
 250     return a;
 251   }
 252   bool is_same_address(Address a) const {
 253     // disregard _rspec
 254     return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale;
 255   }
 256 
 257   // The following two overloads are used in connection with the
 258   // ByteSize type (see sizes.hpp).  They simplify the use of
 259   // ByteSize'd arguments in assembly code. Note that their equivalent
 260   // for the optimized build are the member functions with int disp
 261   // argument since ByteSize is mapped to an int type in that case.
 262   //
 263   // Note: DO NOT introduce similar overloaded functions for WordSize
 264   // arguments as in the optimized mode, both ByteSize and WordSize
 265   // are mapped to the same type and thus the compiler cannot make a
 266   // distinction anymore (=> compiler errors).
 267 
 268 #ifdef ASSERT
 269   Address(Register base, ByteSize disp)
 270     : _base(base),
 271       _index(noreg),
 272       _scale(no_scale),
 273       _disp(in_bytes(disp)) {
 274   }
 275 
 276   Address(Register base, Register index, ScaleFactor scale, ByteSize disp)
 277     : _base(base),
 278       _index(index),
 279       _scale(scale),
 280       _disp(in_bytes(disp)) {
 281     assert(!index->is_valid() == (scale == Address::no_scale),
 282            "inconsistent address");
 283   }
 284 
 285   Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp)
 286     : _base (base),
 287       _index(index.register_or_noreg()),
 288       _scale(scale),
 289       _disp (in_bytes(disp) + (index.constant_or_zero() * scale_size(scale))) {
 290     if (!index.is_register())  scale = Address::no_scale;
 291     assert(!_index->is_valid() == (scale == Address::no_scale),
 292            "inconsistent address");
 293   }
 294 
 295 #endif // ASSERT
 296 
 297   // accessors
 298   bool        uses(Register reg) const { return _base == reg || _index == reg; }
 299   Register    base()             const { return _base;  }
 300   Register    index()            const { return _index; }
 301   ScaleFactor scale()            const { return _scale; }
 302   int         disp()             const { return _disp;  }
 303 
 304   // Convert the raw encoding form into the form expected by the constructor for
 305   // Address.  An index of 4 (rsp) corresponds to having no index, so convert
 306   // that to noreg for the Address constructor.
 307   static Address make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc);
 308 
 309   static Address make_array(ArrayAddress);
 310 
 311  private:
 312   bool base_needs_rex() const {
 313     return _base != noreg && _base->encoding() >= 8;
 314   }
 315 
 316   bool index_needs_rex() const {
 317     return _index != noreg &&_index->encoding() >= 8;
 318   }
 319 
 320   relocInfo::relocType reloc() const { return _rspec.type(); }
 321 
 322   friend class Assembler;
 323   friend class MacroAssembler;
 324   friend class LIR_Assembler; // base/index/scale/disp
 325 };
 326 
 327 //
 328 // AddressLiteral has been split out from Address because operands of this type
 329 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out
 330 // the few instructions that need to deal with address literals are unique and the
 331 // MacroAssembler does not have to implement every instruction in the Assembler
 332 // in order to search for address literals that may need special handling depending
 333 // on the instruction and the platform. As small step on the way to merging i486/amd64
 334 // directories.
 335 //
 336 class AddressLiteral VALUE_OBJ_CLASS_SPEC {
 337   friend class ArrayAddress;
 338   RelocationHolder _rspec;
 339   // Typically we use AddressLiterals we want to use their rval
 340   // However in some situations we want the lval (effect address) of the item.
 341   // We provide a special factory for making those lvals.
 342   bool _is_lval;
 343 
 344   // If the target is far we'll need to load the ea of this to
 345   // a register to reach it. Otherwise if near we can do rip
 346   // relative addressing.
 347 
 348   address          _target;
 349 
 350  protected:
 351   // creation
 352   AddressLiteral()
 353     : _is_lval(false),
 354       _target(NULL)
 355   {}
 356 
 357   public:
 358 
 359 
 360   AddressLiteral(address target, relocInfo::relocType rtype);
 361 
 362   AddressLiteral(address target, RelocationHolder const& rspec)
 363     : _rspec(rspec),
 364       _is_lval(false),
 365       _target(target)
 366   {}
 367 
 368   AddressLiteral addr() {
 369     AddressLiteral ret = *this;
 370     ret._is_lval = true;
 371     return ret;
 372   }
 373 
 374 
 375  private:
 376 
 377   address target() { return _target; }
 378   bool is_lval() { return _is_lval; }
 379 
 380   relocInfo::relocType reloc() const { return _rspec.type(); }
 381   const RelocationHolder& rspec() const { return _rspec; }
 382 
 383   friend class Assembler;
 384   friend class MacroAssembler;
 385   friend class Address;
 386   friend class LIR_Assembler;
 387 };
 388 
 389 // Convience classes
 390 class RuntimeAddress: public AddressLiteral {
 391 
 392   public:
 393 
 394   RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {}
 395 
 396 };
 397 
 398 class ExternalAddress: public AddressLiteral {
 399  private:
 400   static relocInfo::relocType reloc_for_target(address target) {
 401     // Sometimes ExternalAddress is used for values which aren't
 402     // exactly addresses, like the card table base.
 403     // external_word_type can't be used for values in the first page
 404     // so just skip the reloc in that case.
 405     return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none;
 406   }
 407 
 408  public:
 409 
 410   ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {}
 411 
 412 };
 413 
 414 class InternalAddress: public AddressLiteral {
 415 
 416   public:
 417 
 418   InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {}
 419 
 420 };
 421 
 422 // x86 can do array addressing as a single operation since disp can be an absolute
 423 // address amd64 can't. We create a class that expresses the concept but does extra
 424 // magic on amd64 to get the final result
 425 
 426 class ArrayAddress VALUE_OBJ_CLASS_SPEC {
 427   private:
 428 
 429   AddressLiteral _base;
 430   Address        _index;
 431 
 432   public:
 433 
 434   ArrayAddress() {};
 435   ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {};
 436   AddressLiteral base() { return _base; }
 437   Address index() { return _index; }
 438 
 439 };
 440 
 441 // 64-bit refect the fxsave size which is 512 bytes and the new xsave area on EVEX which is another 2176 bytes
 442 // See fxsave and xsave(EVEX enabled) documentation for layout
 443 const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY(2688 / wordSize);
 444 
 445 // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction
 446 // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write
 447 // is what you get. The Assembler is generating code into a CodeBuffer.
 448 
 449 class Assembler : public AbstractAssembler  {
 450   friend class AbstractAssembler; // for the non-virtual hack
 451   friend class LIR_Assembler; // as_Address()
 452   friend class StubGenerator;
 453 
 454  public:
 455   enum Condition {                     // The x86 condition codes used for conditional jumps/moves.
 456     zero          = 0x4,
 457     notZero       = 0x5,
 458     equal         = 0x4,
 459     notEqual      = 0x5,
 460     less          = 0xc,
 461     lessEqual     = 0xe,
 462     greater       = 0xf,
 463     greaterEqual  = 0xd,
 464     below         = 0x2,
 465     belowEqual    = 0x6,
 466     above         = 0x7,
 467     aboveEqual    = 0x3,
 468     overflow      = 0x0,
 469     noOverflow    = 0x1,
 470     carrySet      = 0x2,
 471     carryClear    = 0x3,
 472     negative      = 0x8,
 473     positive      = 0x9,
 474     parity        = 0xa,
 475     noParity      = 0xb
 476   };
 477 
 478   enum Prefix {
 479     // segment overrides
 480     CS_segment = 0x2e,
 481     SS_segment = 0x36,
 482     DS_segment = 0x3e,
 483     ES_segment = 0x26,
 484     FS_segment = 0x64,
 485     GS_segment = 0x65,
 486 
 487     REX        = 0x40,
 488 
 489     REX_B      = 0x41,
 490     REX_X      = 0x42,
 491     REX_XB     = 0x43,
 492     REX_R      = 0x44,
 493     REX_RB     = 0x45,
 494     REX_RX     = 0x46,
 495     REX_RXB    = 0x47,
 496 
 497     REX_W      = 0x48,
 498 
 499     REX_WB     = 0x49,
 500     REX_WX     = 0x4A,
 501     REX_WXB    = 0x4B,
 502     REX_WR     = 0x4C,
 503     REX_WRB    = 0x4D,
 504     REX_WRX    = 0x4E,
 505     REX_WRXB   = 0x4F,
 506 
 507     VEX_3bytes = 0xC4,
 508     VEX_2bytes = 0xC5,
 509     EVEX_4bytes = 0x62,
 510     Prefix_EMPTY = 0x0
 511   };
 512 
 513   enum VexPrefix {
 514     VEX_B = 0x20,
 515     VEX_X = 0x40,
 516     VEX_R = 0x80,
 517     VEX_W = 0x80
 518   };
 519 
 520   enum ExexPrefix {
 521     EVEX_F  = 0x04,
 522     EVEX_V  = 0x08,
 523     EVEX_Rb = 0x10,
 524     EVEX_X  = 0x40,
 525     EVEX_Z  = 0x80
 526   };
 527 
 528   enum VexSimdPrefix {
 529     VEX_SIMD_NONE = 0x0,
 530     VEX_SIMD_66   = 0x1,
 531     VEX_SIMD_F3   = 0x2,
 532     VEX_SIMD_F2   = 0x3
 533   };
 534 
 535   enum VexOpcode {
 536     VEX_OPCODE_NONE  = 0x0,
 537     VEX_OPCODE_0F    = 0x1,
 538     VEX_OPCODE_0F_38 = 0x2,
 539     VEX_OPCODE_0F_3A = 0x3,
 540     VEX_OPCODE_MASK  = 0x1F
 541   };
 542 
 543   enum AvxVectorLen {
 544     AVX_128bit = 0x0,
 545     AVX_256bit = 0x1,
 546     AVX_512bit = 0x2,
 547     AVX_NoVec  = 0x4
 548   };
 549 
 550   enum EvexTupleType {
 551     EVEX_FV   = 0,
 552     EVEX_HV   = 4,
 553     EVEX_FVM  = 6,
 554     EVEX_T1S  = 7,
 555     EVEX_T1F  = 11,
 556     EVEX_T2   = 13,
 557     EVEX_T4   = 15,
 558     EVEX_T8   = 17,
 559     EVEX_HVM  = 18,
 560     EVEX_QVM  = 19,
 561     EVEX_OVM  = 20,
 562     EVEX_M128 = 21,
 563     EVEX_DUP  = 22,
 564     EVEX_ETUP = 23
 565   };
 566 
 567   enum EvexInputSizeInBits {
 568     EVEX_8bit  = 0,
 569     EVEX_16bit = 1,
 570     EVEX_32bit = 2,
 571     EVEX_64bit = 3
 572   };
 573 
 574   enum WhichOperand {
 575     // input to locate_operand, and format code for relocations
 576     imm_operand  = 0,            // embedded 32-bit|64-bit immediate operand
 577     disp32_operand = 1,          // embedded 32-bit displacement or address
 578     call32_operand = 2,          // embedded 32-bit self-relative displacement
 579 #ifndef _LP64
 580     _WhichOperand_limit = 3
 581 #else
 582      narrow_oop_operand = 3,     // embedded 32-bit immediate narrow oop
 583     _WhichOperand_limit = 4
 584 #endif
 585   };
 586 
 587 
 588 
 589   // NOTE: The general philopsophy of the declarations here is that 64bit versions
 590   // of instructions are freely declared without the need for wrapping them an ifdef.
 591   // (Some dangerous instructions are ifdef's out of inappropriate jvm's.)
 592   // In the .cpp file the implementations are wrapped so that they are dropped out
 593   // of the resulting jvm. This is done mostly to keep the footprint of MINIMAL
 594   // to the size it was prior to merging up the 32bit and 64bit assemblers.
 595   //
 596   // This does mean you'll get a linker/runtime error if you use a 64bit only instruction
 597   // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down.
 598 
 599 private:
 600 
 601   int _evex_encoding;
 602   int _input_size_in_bits;
 603   int _avx_vector_len;
 604   int _tuple_type;
 605   bool _is_evex_instruction;
 606   bool _legacy_mode_bw;
 607   bool _legacy_mode_dq;
 608   bool _legacy_mode_vl;
 609   bool _legacy_mode_vlbw;
 610   bool _instruction_uses_vl;
 611 
 612   // 64bit prefixes
 613   int prefix_and_encode(int reg_enc, bool byteinst = false);
 614   int prefixq_and_encode(int reg_enc);
 615 
 616   int prefix_and_encode(int dst_enc, int src_enc) {
 617     return prefix_and_encode(dst_enc, false, src_enc, false);
 618   }
 619   int prefix_and_encode(int dst_enc, bool dst_is_byte, int src_enc, bool src_is_byte);
 620   int prefixq_and_encode(int dst_enc, int src_enc);
 621 
 622   void prefix(Register reg);
 623   void prefix(Register dst, Register src, Prefix p);
 624   void prefix(Register dst, Address adr, Prefix p);
 625   void prefix(Address adr);
 626   void prefixq(Address adr);
 627 
 628   void prefix(Address adr, Register reg,  bool byteinst = false);
 629   void prefix(Address adr, XMMRegister reg);
 630   void prefixq(Address adr, Register reg);
 631   void prefixq(Address adr, XMMRegister reg);
 632 
 633   void prefetch_prefix(Address src);
 634 
 635   void rex_prefix(Address adr, XMMRegister xreg,
 636                   VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 637   int  rex_prefix_and_encode(int dst_enc, int src_enc,
 638                              VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 639 
 640   void vex_prefix(bool vex_r, bool vex_b, bool vex_x, bool vex_w,
 641                   int nds_enc, VexSimdPrefix pre, VexOpcode opc,
 642                   int vector_len);
 643 
 644   void evex_prefix(bool vex_r, bool vex_b, bool vex_x, bool vex_w, bool evex_r, bool evex_v,
 645                    int nds_enc, VexSimdPrefix pre, VexOpcode opc,
 646                    bool is_extended_context, bool is_merge_context,
 647                    int vector_len, bool no_mask_reg );
 648 
 649   void vex_prefix(Address adr, int nds_enc, int xreg_enc,
 650                   VexSimdPrefix pre, VexOpcode opc,
 651                   bool vex_w, int vector_len,
 652                   bool legacy_mode = false, bool no_mask_reg = false);
 653 
 654   void vex_prefix(XMMRegister dst, XMMRegister nds, Address src,
 655                   VexSimdPrefix pre, int vector_len = AVX_128bit,
 656                   bool no_mask_reg = false, bool legacy_mode = false) {
 657     int dst_enc = dst->encoding();
 658     int nds_enc = nds->is_valid() ? nds->encoding() : 0;
 659     vex_prefix(src, nds_enc, dst_enc, pre, VEX_OPCODE_0F, false, vector_len, legacy_mode, no_mask_reg);
 660   }
 661 
 662   void vex_prefix_q(XMMRegister dst, XMMRegister nds, Address src,
 663                     VexSimdPrefix pre, int vector_len = AVX_128bit,
 664                     bool no_mask_reg = false) {
 665     int dst_enc = dst->encoding();
 666     int nds_enc = nds->is_valid() ? nds->encoding() : 0;
 667     vex_prefix(src, nds_enc, dst_enc, pre, VEX_OPCODE_0F, true, vector_len, false, no_mask_reg);
 668   }
 669 
 670   void vex_prefix_0F38(Register dst, Register nds, Address src, bool no_mask_reg = false) {
 671     bool vex_w = false;
 672     int vector_len = AVX_128bit;
 673     vex_prefix(src, nds->encoding(), dst->encoding(),
 674                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w,
 675                vector_len, no_mask_reg);
 676   }
 677 
 678   void vex_prefix_0F38_legacy(Register dst, Register nds, Address src, bool no_mask_reg = false) {
 679     bool vex_w = false;
 680     int vector_len = AVX_128bit;
 681     vex_prefix(src, nds->encoding(), dst->encoding(),
 682                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w,
 683                vector_len, true, no_mask_reg);
 684   }
 685 
 686   void vex_prefix_0F38_q(Register dst, Register nds, Address src, bool no_mask_reg = false) {
 687     bool vex_w = true;
 688     int vector_len = AVX_128bit;
 689     vex_prefix(src, nds->encoding(), dst->encoding(),
 690                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w,
 691                vector_len, no_mask_reg);
 692   }
 693 
 694   void vex_prefix_0F38_q_legacy(Register dst, Register nds, Address src, bool no_mask_reg = false) {
 695     bool vex_w = true;
 696     int vector_len = AVX_128bit;
 697     vex_prefix(src, nds->encoding(), dst->encoding(),
 698                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w,
 699                vector_len, true, no_mask_reg);
 700   }
 701 
 702   int  vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc,
 703                              VexSimdPrefix pre, VexOpcode opc,
 704                              bool vex_w, int vector_len,
 705                              bool legacy_mode, bool no_mask_reg);
 706 
 707   int  vex_prefix_0F38_and_encode(Register dst, Register nds, Register src, bool no_mask_reg = false) {
 708     bool vex_w = false;
 709     int vector_len = AVX_128bit;
 710     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 711                                  VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector_len,
 712                                  false, no_mask_reg);
 713   }
 714 
 715   int  vex_prefix_0F38_and_encode_legacy(Register dst, Register nds, Register src, bool no_mask_reg = false) {
 716     bool vex_w = false;
 717     int vector_len = AVX_128bit;
 718     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 719       VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector_len,
 720       true, no_mask_reg);
 721   }
 722 
 723   int  vex_prefix_0F38_and_encode_q(Register dst, Register nds, Register src, bool no_mask_reg = false) {
 724     bool vex_w = true;
 725     int vector_len = AVX_128bit;
 726     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 727                                  VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector_len,
 728                                  false, no_mask_reg);
 729   }
 730 
 731   int  vex_prefix_0F38_and_encode_q_legacy(Register dst, Register nds, Register src, bool no_mask_reg = false) {
 732     bool vex_w = true;
 733     int vector_len = AVX_128bit;
 734     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 735                                  VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector_len,
 736                                  true, no_mask_reg);
 737   }
 738 
 739   int  vex_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src,
 740                              VexSimdPrefix pre, int vector_len = AVX_128bit,
 741                              VexOpcode opc = VEX_OPCODE_0F, bool legacy_mode = false,
 742                              bool no_mask_reg = false) {
 743     int src_enc = src->encoding();
 744     int dst_enc = dst->encoding();
 745     int nds_enc = nds->is_valid() ? nds->encoding() : 0;
 746     return vex_prefix_and_encode(dst_enc, nds_enc, src_enc, pre, opc, false, vector_len, legacy_mode, no_mask_reg);
 747   }
 748 
 749   void simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr,
 750                    VexSimdPrefix pre, bool no_mask_reg, VexOpcode opc = VEX_OPCODE_0F,
 751                    bool rex_w = false, int vector_len = AVX_128bit, bool legacy_mode = false);
 752 
 753   void simd_prefix(XMMRegister dst, Address src, VexSimdPrefix pre,
 754                    bool no_mask_reg, VexOpcode opc = VEX_OPCODE_0F) {
 755     simd_prefix(dst, xnoreg, src, pre, no_mask_reg, opc);
 756   }
 757 
 758   void simd_prefix(Address dst, XMMRegister src, VexSimdPrefix pre, bool no_mask_reg) {
 759     simd_prefix(src, dst, pre, no_mask_reg);
 760   }
 761   void simd_prefix_q(XMMRegister dst, XMMRegister nds, Address src,
 762                      VexSimdPrefix pre, bool no_mask_reg = false) {
 763     bool rex_w = true;
 764     simd_prefix(dst, nds, src, pre, no_mask_reg, VEX_OPCODE_0F, rex_w);
 765   }
 766 
 767   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src,
 768                              VexSimdPrefix pre, bool no_mask_reg,
 769                              VexOpcode opc = VEX_OPCODE_0F,
 770                              bool rex_w = false, int vector_len = AVX_128bit,
 771                              bool legacy_mode = false);
 772 
 773   int kreg_prefix_and_encode(KRegister dst, KRegister nds, KRegister src,
 774                              VexSimdPrefix pre, bool no_mask_reg,
 775                              VexOpcode opc = VEX_OPCODE_0F,
 776                              bool rex_w = false, int vector_len = AVX_128bit);
 777 
 778   int kreg_prefix_and_encode(KRegister dst, KRegister nds, Register src,
 779                              VexSimdPrefix pre, bool no_mask_reg,
 780                              VexOpcode opc = VEX_OPCODE_0F,
 781                              bool rex_w = false, int vector_len = AVX_128bit);
 782 
 783   // Move/convert 32-bit integer value.
 784   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, Register src,
 785                              VexSimdPrefix pre, bool no_mask_reg) {
 786     // It is OK to cast from Register to XMMRegister to pass argument here
 787     // since only encoding is used in simd_prefix_and_encode() and number of
 788     // Gen and Xmm registers are the same.
 789     return simd_prefix_and_encode(dst, nds, as_XMMRegister(src->encoding()), pre, no_mask_reg, VEX_OPCODE_0F);
 790   }
 791   int simd_prefix_and_encode(XMMRegister dst, Register src, VexSimdPrefix pre, bool no_mask_reg) {
 792     return simd_prefix_and_encode(dst, xnoreg, src, pre, no_mask_reg);
 793   }
 794   int simd_prefix_and_encode(Register dst, XMMRegister src,
 795                              VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F,
 796                              bool no_mask_reg = false) {
 797     return simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, pre, no_mask_reg, opc);
 798   }
 799 
 800   // Move/convert 64-bit integer value.
 801   int simd_prefix_and_encode_q(XMMRegister dst, XMMRegister nds, Register src,
 802                                VexSimdPrefix pre, bool no_mask_reg = false) {
 803     bool rex_w = true;
 804     return simd_prefix_and_encode(dst, nds, as_XMMRegister(src->encoding()), pre, no_mask_reg, VEX_OPCODE_0F, rex_w);
 805   }
 806   int simd_prefix_and_encode_q(XMMRegister dst, Register src, VexSimdPrefix pre, bool no_mask_reg) {
 807     return simd_prefix_and_encode_q(dst, xnoreg, src, pre, no_mask_reg);
 808   }
 809   int simd_prefix_and_encode_q(Register dst, XMMRegister src,
 810                                VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F,
 811                                bool no_mask_reg = false) {
 812     bool rex_w = true;
 813     return simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, pre, no_mask_reg, opc, rex_w);
 814   }
 815 
 816   // Helper functions for groups of instructions
 817   void emit_arith_b(int op1, int op2, Register dst, int imm8);
 818 
 819   void emit_arith(int op1, int op2, Register dst, int32_t imm32);
 820   // Force generation of a 4 byte immediate value even if it fits into 8bit
 821   void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32);
 822   void emit_arith(int op1, int op2, Register dst, Register src);
 823 
 824   void emit_simd_arith(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre, bool no_mask_reg = false, bool legacy_mode = false);
 825   void emit_simd_arith_q(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre, bool no_mask_reg = false);
 826   void emit_simd_arith(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre, bool no_mask_reg = false, bool legacy_mode = false);
 827   void emit_simd_arith_q(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre, bool no_mask_reg = false);
 828   void emit_simd_arith_nonds(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre, bool no_mask_reg = false);
 829   void emit_simd_arith_nonds_q(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre, bool no_mask_reg = false);
 830   void emit_simd_arith_nonds(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre, bool no_mask_reg = false, bool legacy_mode = false);
 831   void emit_simd_arith_nonds_q(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre, bool no_mask_reg = false);
 832   void emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
 833                       Address src, VexSimdPrefix pre, int vector_len,
 834                       bool no_mask_reg = false, bool legacy_mode = false);
 835   void emit_vex_arith_q(int opcode, XMMRegister dst, XMMRegister nds,
 836                         Address src, VexSimdPrefix pre, int vector_len,
 837                         bool no_mask_reg = false);
 838   void emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
 839                       XMMRegister src, VexSimdPrefix pre, int vector_len,
 840                       bool no_mask_reg = false, bool legacy_mode = false);
 841   void emit_vex_arith_q(int opcode, XMMRegister dst, XMMRegister nds,
 842                         XMMRegister src, VexSimdPrefix pre, int vector_len,
 843                         bool no_mask_reg = false);
 844 
 845   bool emit_compressed_disp_byte(int &disp);
 846 
 847   void emit_operand(Register reg,
 848                     Register base, Register index, Address::ScaleFactor scale,
 849                     int disp,
 850                     RelocationHolder const& rspec,
 851                     int rip_relative_correction = 0);
 852 
 853   void emit_operand(Register reg, Address adr, int rip_relative_correction = 0);
 854 
 855   // operands that only take the original 32bit registers
 856   void emit_operand32(Register reg, Address adr);
 857 
 858   void emit_operand(XMMRegister reg,
 859                     Register base, Register index, Address::ScaleFactor scale,
 860                     int disp,
 861                     RelocationHolder const& rspec);
 862 
 863   void emit_operand(XMMRegister reg, Address adr);
 864 
 865   void emit_operand(MMXRegister reg, Address adr);
 866 
 867   // workaround gcc (3.2.1-7) bug
 868   void emit_operand(Address adr, MMXRegister reg);
 869 
 870 
 871   // Immediate-to-memory forms
 872   void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32);
 873 
 874   void emit_farith(int b1, int b2, int i);
 875 
 876 
 877  protected:
 878   #ifdef ASSERT
 879   void check_relocation(RelocationHolder const& rspec, int format);
 880   #endif
 881 
 882   void emit_data(jint data, relocInfo::relocType    rtype, int format);
 883   void emit_data(jint data, RelocationHolder const& rspec, int format);
 884   void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0);
 885   void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0);
 886 
 887   bool reachable(AddressLiteral adr) NOT_LP64({ return true;});
 888 
 889   // These are all easily abused and hence protected
 890 
 891   // 32BIT ONLY SECTION
 892 #ifndef _LP64
 893   // Make these disappear in 64bit mode since they would never be correct
 894   void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec);   // 32BIT ONLY
 895   void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 896 
 897   void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 898   void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec);     // 32BIT ONLY
 899 
 900   void push_literal32(int32_t imm32, RelocationHolder const& rspec);                 // 32BIT ONLY
 901 #else
 902   // 64BIT ONLY SECTION
 903   void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec);   // 64BIT ONLY
 904 
 905   void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec);
 906   void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec);
 907 
 908   void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec);
 909   void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec);
 910 #endif // _LP64
 911 
 912   // These are unique in that we are ensured by the caller that the 32bit
 913   // relative in these instructions will always be able to reach the potentially
 914   // 64bit address described by entry. Since they can take a 64bit address they
 915   // don't have the 32 suffix like the other instructions in this class.
 916 
 917   void call_literal(address entry, RelocationHolder const& rspec);
 918   void jmp_literal(address entry, RelocationHolder const& rspec);
 919 
 920   // Avoid using directly section
 921   // Instructions in this section are actually usable by anyone without danger
 922   // of failure but have performance issues that are addressed my enhanced
 923   // instructions which will do the proper thing base on the particular cpu.
 924   // We protect them because we don't trust you...
 925 
 926   // Don't use next inc() and dec() methods directly. INC & DEC instructions
 927   // could cause a partial flag stall since they don't set CF flag.
 928   // Use MacroAssembler::decrement() & MacroAssembler::increment() methods
 929   // which call inc() & dec() or add() & sub() in accordance with
 930   // the product flag UseIncDec value.
 931 
 932   void decl(Register dst);
 933   void decl(Address dst);
 934   void decq(Register dst);
 935   void decq(Address dst);
 936 
 937   void incl(Register dst);
 938   void incl(Address dst);
 939   void incq(Register dst);
 940   void incq(Address dst);
 941 
 942   // New cpus require use of movsd and movss to avoid partial register stall
 943   // when loading from memory. But for old Opteron use movlpd instead of movsd.
 944   // The selection is done in MacroAssembler::movdbl() and movflt().
 945 
 946   // Move Scalar Single-Precision Floating-Point Values
 947   void movss(XMMRegister dst, Address src);
 948   void movss(XMMRegister dst, XMMRegister src);
 949   void movss(Address dst, XMMRegister src);
 950 
 951   // Move Scalar Double-Precision Floating-Point Values
 952   void movsd(XMMRegister dst, Address src);
 953   void movsd(XMMRegister dst, XMMRegister src);
 954   void movsd(Address dst, XMMRegister src);
 955   void movlpd(XMMRegister dst, Address src);
 956 
 957   // New cpus require use of movaps and movapd to avoid partial register stall
 958   // when moving between registers.
 959   void movaps(XMMRegister dst, XMMRegister src);
 960   void movapd(XMMRegister dst, XMMRegister src);
 961 
 962   // End avoid using directly
 963 
 964 
 965   // Instruction prefixes
 966   void prefix(Prefix p);
 967 
 968   public:
 969 
 970   // Creation
 971   Assembler(CodeBuffer* code) : AbstractAssembler(code) {
 972     init_attributes();
 973   }
 974 
 975   // Decoding
 976   static address locate_operand(address inst, WhichOperand which);
 977   static address locate_next_instruction(address inst);
 978 
 979   // Utilities
 980   static bool is_polling_page_far() NOT_LP64({ return false;});
 981   static bool query_compressed_disp_byte(int disp, bool is_evex_inst, int vector_len,
 982                                          int cur_tuple_type, int in_size_in_bits, int cur_encoding);
 983 
 984   // Generic instructions
 985   // Does 32bit or 64bit as needed for the platform. In some sense these
 986   // belong in macro assembler but there is no need for both varieties to exist
 987 
 988   void init_attributes(void) {
 989     _evex_encoding = 0;
 990     _input_size_in_bits = 0;
 991     _avx_vector_len = AVX_NoVec;
 992     _tuple_type = EVEX_ETUP;
 993     _is_evex_instruction = false;
 994     _legacy_mode_bw = (VM_Version::supports_avx512bw() == false);
 995     _legacy_mode_dq = (VM_Version::supports_avx512dq() == false);
 996     _legacy_mode_vl = (VM_Version::supports_avx512vl() == false);
 997     _legacy_mode_vlbw = (VM_Version::supports_avx512vlbw() == false);
 998     _instruction_uses_vl = false;
 999   }
1000 
1001   void lea(Register dst, Address src);
1002 
1003   void mov(Register dst, Register src);
1004 
1005   void pusha();
1006   void popa();
1007 
1008   void pushf();
1009   void popf();
1010 
1011   void push(int32_t imm32);
1012 
1013   void push(Register src);
1014 
1015   void pop(Register dst);
1016 
1017   // These are dummies to prevent surprise implicit conversions to Register
1018   void push(void* v);
1019   void pop(void* v);
1020 
1021   // These do register sized moves/scans
1022   void rep_mov();
1023   void rep_stos();
1024   void rep_stosb();
1025   void repne_scan();
1026 #ifdef _LP64
1027   void repne_scanl();
1028 #endif
1029 
1030   // Vanilla instructions in lexical order
1031 
1032   void adcl(Address dst, int32_t imm32);
1033   void adcl(Address dst, Register src);
1034   void adcl(Register dst, int32_t imm32);
1035   void adcl(Register dst, Address src);
1036   void adcl(Register dst, Register src);
1037 
1038   void adcq(Register dst, int32_t imm32);
1039   void adcq(Register dst, Address src);
1040   void adcq(Register dst, Register src);
1041 
1042   void addl(Address dst, int32_t imm32);
1043   void addl(Address dst, Register src);
1044   void addl(Register dst, int32_t imm32);
1045   void addl(Register dst, Address src);
1046   void addl(Register dst, Register src);
1047 
1048   void addq(Address dst, int32_t imm32);
1049   void addq(Address dst, Register src);
1050   void addq(Register dst, int32_t imm32);
1051   void addq(Register dst, Address src);
1052   void addq(Register dst, Register src);
1053 
1054 #ifdef _LP64
1055  //Add Unsigned Integers with Carry Flag
1056   void adcxq(Register dst, Register src);
1057 
1058  //Add Unsigned Integers with Overflow Flag
1059   void adoxq(Register dst, Register src);
1060 #endif
1061 
1062   void addr_nop_4();
1063   void addr_nop_5();
1064   void addr_nop_7();
1065   void addr_nop_8();
1066 
1067   // Add Scalar Double-Precision Floating-Point Values
1068   void addsd(XMMRegister dst, Address src);
1069   void addsd(XMMRegister dst, XMMRegister src);
1070 
1071   // Add Scalar Single-Precision Floating-Point Values
1072   void addss(XMMRegister dst, Address src);
1073   void addss(XMMRegister dst, XMMRegister src);
1074 
1075   // AES instructions
1076   void aesdec(XMMRegister dst, Address src);
1077   void aesdec(XMMRegister dst, XMMRegister src);
1078   void aesdeclast(XMMRegister dst, Address src);
1079   void aesdeclast(XMMRegister dst, XMMRegister src);
1080   void aesenc(XMMRegister dst, Address src);
1081   void aesenc(XMMRegister dst, XMMRegister src);
1082   void aesenclast(XMMRegister dst, Address src);
1083   void aesenclast(XMMRegister dst, XMMRegister src);
1084 
1085 
1086   void andl(Address  dst, int32_t imm32);
1087   void andl(Register dst, int32_t imm32);
1088   void andl(Register dst, Address src);
1089   void andl(Register dst, Register src);
1090 
1091   void andq(Address  dst, int32_t imm32);
1092   void andq(Register dst, int32_t imm32);
1093   void andq(Register dst, Address src);
1094   void andq(Register dst, Register src);
1095 
1096   // BMI instructions
1097   void andnl(Register dst, Register src1, Register src2);
1098   void andnl(Register dst, Register src1, Address src2);
1099   void andnq(Register dst, Register src1, Register src2);
1100   void andnq(Register dst, Register src1, Address src2);
1101 
1102   void blsil(Register dst, Register src);
1103   void blsil(Register dst, Address src);
1104   void blsiq(Register dst, Register src);
1105   void blsiq(Register dst, Address src);
1106 
1107   void blsmskl(Register dst, Register src);
1108   void blsmskl(Register dst, Address src);
1109   void blsmskq(Register dst, Register src);
1110   void blsmskq(Register dst, Address src);
1111 
1112   void blsrl(Register dst, Register src);
1113   void blsrl(Register dst, Address src);
1114   void blsrq(Register dst, Register src);
1115   void blsrq(Register dst, Address src);
1116 
1117   void bsfl(Register dst, Register src);
1118   void bsrl(Register dst, Register src);
1119 
1120 #ifdef _LP64
1121   void bsfq(Register dst, Register src);
1122   void bsrq(Register dst, Register src);
1123 #endif
1124 
1125   void bswapl(Register reg);
1126 
1127   void bswapq(Register reg);
1128 
1129   void call(Label& L, relocInfo::relocType rtype);
1130   void call(Register reg);  // push pc; pc <- reg
1131   void call(Address adr);   // push pc; pc <- adr
1132 
1133   void cdql();
1134 
1135   void cdqq();
1136 
1137   void cld();
1138 
1139   void clflush(Address adr);
1140 
1141   void cmovl(Condition cc, Register dst, Register src);
1142   void cmovl(Condition cc, Register dst, Address src);
1143 
1144   void cmovq(Condition cc, Register dst, Register src);
1145   void cmovq(Condition cc, Register dst, Address src);
1146 
1147 
1148   void cmpb(Address dst, int imm8);
1149 
1150   void cmpl(Address dst, int32_t imm32);
1151 
1152   void cmpl(Register dst, int32_t imm32);
1153   void cmpl(Register dst, Register src);
1154   void cmpl(Register dst, Address src);
1155 
1156   void cmpq(Address dst, int32_t imm32);
1157   void cmpq(Address dst, Register src);
1158 
1159   void cmpq(Register dst, int32_t imm32);
1160   void cmpq(Register dst, Register src);
1161   void cmpq(Register dst, Address src);
1162 
1163   // these are dummies used to catch attempting to convert NULL to Register
1164   void cmpl(Register dst, void* junk); // dummy
1165   void cmpq(Register dst, void* junk); // dummy
1166 
1167   void cmpw(Address dst, int imm16);
1168 
1169   void cmpxchg8 (Address adr);
1170 
1171   void cmpxchgb(Register reg, Address adr);
1172   void cmpxchgl(Register reg, Address adr);
1173 
1174   void cmpxchgq(Register reg, Address adr);
1175 
1176   // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1177   void comisd(XMMRegister dst, Address src);
1178   void comisd(XMMRegister dst, XMMRegister src);
1179 
1180   // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1181   void comiss(XMMRegister dst, Address src);
1182   void comiss(XMMRegister dst, XMMRegister src);
1183 
1184   // Identify processor type and features
1185   void cpuid();
1186 
1187   // CRC32C
1188   void crc32(Register crc, Register v, int8_t sizeInBytes);
1189   void crc32(Register crc, Address adr, int8_t sizeInBytes);
1190 
1191   // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floating-Point Value
1192   void cvtsd2ss(XMMRegister dst, XMMRegister src);
1193   void cvtsd2ss(XMMRegister dst, Address src);
1194 
1195   // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value
1196   void cvtsi2sdl(XMMRegister dst, Register src);
1197   void cvtsi2sdl(XMMRegister dst, Address src);
1198   void cvtsi2sdq(XMMRegister dst, Register src);
1199   void cvtsi2sdq(XMMRegister dst, Address src);
1200 
1201   // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value
1202   void cvtsi2ssl(XMMRegister dst, Register src);
1203   void cvtsi2ssl(XMMRegister dst, Address src);
1204   void cvtsi2ssq(XMMRegister dst, Register src);
1205   void cvtsi2ssq(XMMRegister dst, Address src);
1206 
1207   // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Value
1208   void cvtdq2pd(XMMRegister dst, XMMRegister src);
1209 
1210   // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Value
1211   void cvtdq2ps(XMMRegister dst, XMMRegister src);
1212 
1213   // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floating-Point Value
1214   void cvtss2sd(XMMRegister dst, XMMRegister src);
1215   void cvtss2sd(XMMRegister dst, Address src);
1216 
1217   // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Integer
1218   void cvttsd2sil(Register dst, Address src);
1219   void cvttsd2sil(Register dst, XMMRegister src);
1220   void cvttsd2siq(Register dst, XMMRegister src);
1221 
1222   // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Integer
1223   void cvttss2sil(Register dst, XMMRegister src);
1224   void cvttss2siq(Register dst, XMMRegister src);
1225 
1226   // Divide Scalar Double-Precision Floating-Point Values
1227   void divsd(XMMRegister dst, Address src);
1228   void divsd(XMMRegister dst, XMMRegister src);
1229 
1230   // Divide Scalar Single-Precision Floating-Point Values
1231   void divss(XMMRegister dst, Address src);
1232   void divss(XMMRegister dst, XMMRegister src);
1233 
1234   void emms();
1235 
1236   void fabs();
1237 
1238   void fadd(int i);
1239 
1240   void fadd_d(Address src);
1241   void fadd_s(Address src);
1242 
1243   // "Alternate" versions of x87 instructions place result down in FPU
1244   // stack instead of on TOS
1245 
1246   void fadda(int i); // "alternate" fadd
1247   void faddp(int i = 1);
1248 
1249   void fchs();
1250 
1251   void fcom(int i);
1252 
1253   void fcomp(int i = 1);
1254   void fcomp_d(Address src);
1255   void fcomp_s(Address src);
1256 
1257   void fcompp();
1258 
1259   void fcos();
1260 
1261   void fdecstp();
1262 
1263   void fdiv(int i);
1264   void fdiv_d(Address src);
1265   void fdivr_s(Address src);
1266   void fdiva(int i);  // "alternate" fdiv
1267   void fdivp(int i = 1);
1268 
1269   void fdivr(int i);
1270   void fdivr_d(Address src);
1271   void fdiv_s(Address src);
1272 
1273   void fdivra(int i); // "alternate" reversed fdiv
1274 
1275   void fdivrp(int i = 1);
1276 
1277   void ffree(int i = 0);
1278 
1279   void fild_d(Address adr);
1280   void fild_s(Address adr);
1281 
1282   void fincstp();
1283 
1284   void finit();
1285 
1286   void fist_s (Address adr);
1287   void fistp_d(Address adr);
1288   void fistp_s(Address adr);
1289 
1290   void fld1();
1291 
1292   void fld_d(Address adr);
1293   void fld_s(Address adr);
1294   void fld_s(int index);
1295   void fld_x(Address adr);  // extended-precision (80-bit) format
1296 
1297   void fldcw(Address src);
1298 
1299   void fldenv(Address src);
1300 
1301   void fldlg2();
1302 
1303   void fldln2();
1304 
1305   void fldz();
1306 
1307   void flog();
1308   void flog10();
1309 
1310   void fmul(int i);
1311 
1312   void fmul_d(Address src);
1313   void fmul_s(Address src);
1314 
1315   void fmula(int i);  // "alternate" fmul
1316 
1317   void fmulp(int i = 1);
1318 
1319   void fnsave(Address dst);
1320 
1321   void fnstcw(Address src);
1322 
1323   void fnstsw_ax();
1324 
1325   void fprem();
1326   void fprem1();
1327 
1328   void frstor(Address src);
1329 
1330   void fsin();
1331 
1332   void fsqrt();
1333 
1334   void fst_d(Address adr);
1335   void fst_s(Address adr);
1336 
1337   void fstp_d(Address adr);
1338   void fstp_d(int index);
1339   void fstp_s(Address adr);
1340   void fstp_x(Address adr); // extended-precision (80-bit) format
1341 
1342   void fsub(int i);
1343   void fsub_d(Address src);
1344   void fsub_s(Address src);
1345 
1346   void fsuba(int i);  // "alternate" fsub
1347 
1348   void fsubp(int i = 1);
1349 
1350   void fsubr(int i);
1351   void fsubr_d(Address src);
1352   void fsubr_s(Address src);
1353 
1354   void fsubra(int i); // "alternate" reversed fsub
1355 
1356   void fsubrp(int i = 1);
1357 
1358   void ftan();
1359 
1360   void ftst();
1361 
1362   void fucomi(int i = 1);
1363   void fucomip(int i = 1);
1364 
1365   void fwait();
1366 
1367   void fxch(int i = 1);
1368 
1369   void fxrstor(Address src);
1370   void xrstor(Address src);
1371 
1372   void fxsave(Address dst);
1373   void xsave(Address dst);
1374 
1375   void fyl2x();
1376   void frndint();
1377   void f2xm1();
1378   void fldl2e();
1379 
1380   void hlt();
1381 
1382   void idivl(Register src);
1383   void divl(Register src); // Unsigned division
1384 
1385 #ifdef _LP64
1386   void idivq(Register src);
1387 #endif
1388 
1389   void imull(Register dst, Register src);
1390   void imull(Register dst, Register src, int value);
1391   void imull(Register dst, Address src);
1392 
1393 #ifdef _LP64
1394   void imulq(Register dst, Register src);
1395   void imulq(Register dst, Register src, int value);
1396   void imulq(Register dst, Address src);
1397 #endif
1398 
1399   // jcc is the generic conditional branch generator to run-
1400   // time routines, jcc is used for branches to labels. jcc
1401   // takes a branch opcode (cc) and a label (L) and generates
1402   // either a backward branch or a forward branch and links it
1403   // to the label fixup chain. Usage:
1404   //
1405   // Label L;      // unbound label
1406   // jcc(cc, L);   // forward branch to unbound label
1407   // bind(L);      // bind label to the current pc
1408   // jcc(cc, L);   // backward branch to bound label
1409   // bind(L);      // illegal: a label may be bound only once
1410   //
1411   // Note: The same Label can be used for forward and backward branches
1412   // but it may be bound only once.
1413 
1414   void jcc(Condition cc, Label& L, bool maybe_short = true);
1415 
1416   // Conditional jump to a 8-bit offset to L.
1417   // WARNING: be very careful using this for forward jumps.  If the label is
1418   // not bound within an 8-bit offset of this instruction, a run-time error
1419   // will occur.
1420   void jccb(Condition cc, Label& L);
1421 
1422   void jmp(Address entry);    // pc <- entry
1423 
1424   // Label operations & relative jumps (PPUM Appendix D)
1425   void jmp(Label& L, bool maybe_short = true);   // unconditional jump to L
1426 
1427   void jmp(Register entry); // pc <- entry
1428 
1429   // Unconditional 8-bit offset jump to L.
1430   // WARNING: be very careful using this for forward jumps.  If the label is
1431   // not bound within an 8-bit offset of this instruction, a run-time error
1432   // will occur.
1433   void jmpb(Label& L);
1434 
1435   void ldmxcsr( Address src );
1436 
1437   void leal(Register dst, Address src);
1438 
1439   void leaq(Register dst, Address src);
1440 
1441   void lfence();
1442 
1443   void lock();
1444 
1445   void lzcntl(Register dst, Register src);
1446 
1447 #ifdef _LP64
1448   void lzcntq(Register dst, Register src);
1449 #endif
1450 
1451   enum Membar_mask_bits {
1452     StoreStore = 1 << 3,
1453     LoadStore  = 1 << 2,
1454     StoreLoad  = 1 << 1,
1455     LoadLoad   = 1 << 0
1456   };
1457 
1458   // Serializes memory and blows flags
1459   void membar(Membar_mask_bits order_constraint) {
1460     if (os::is_MP()) {
1461       // We only have to handle StoreLoad
1462       if (order_constraint & StoreLoad) {
1463         // All usable chips support "locked" instructions which suffice
1464         // as barriers, and are much faster than the alternative of
1465         // using cpuid instruction. We use here a locked add [esp-C],0.
1466         // This is conveniently otherwise a no-op except for blowing
1467         // flags, and introducing a false dependency on target memory
1468         // location. We can't do anything with flags, but we can avoid
1469         // memory dependencies in the current method by locked-adding
1470         // somewhere else on the stack. Doing [esp+C] will collide with
1471         // something on stack in current method, hence we go for [esp-C].
1472         // It is convenient since it is almost always in data cache, for
1473         // any small C.  We need to step back from SP to avoid data
1474         // dependencies with other things on below SP (callee-saves, for
1475         // example). Without a clear way to figure out the minimal safe
1476         // distance from SP, it makes sense to step back the complete
1477         // cache line, as this will also avoid possible second-order effects
1478         // with locked ops against the cache line. Our choice of offset
1479         // is bounded by x86 operand encoding, which should stay within
1480         // [-128; +127] to have the 8-byte displacement encoding.
1481         //
1482         // Any change to this code may need to revisit other places in
1483         // the code where this idiom is used, in particular the
1484         // orderAccess code.
1485 
1486         int offset = -VM_Version::L1_line_size();
1487         if (offset < -128) {
1488           offset = -128;
1489         }
1490 
1491         lock();
1492         addl(Address(rsp, offset), 0);// Assert the lock# signal here
1493       }
1494     }
1495   }
1496 
1497   void mfence();
1498 
1499   // Moves
1500 
1501   void mov64(Register dst, int64_t imm64);
1502 
1503   void movb(Address dst, Register src);
1504   void movb(Address dst, int imm8);
1505   void movb(Register dst, Address src);
1506 
1507   void movddup(XMMRegister dst, XMMRegister src);
1508 
1509   void kmovql(KRegister dst, KRegister src);
1510   void kmovql(KRegister dst, Register src);
1511   void kmovdl(KRegister dst, Register src);
1512   void kmovwl(KRegister dst, Register src);
1513   void kmovql(Address dst, KRegister src);
1514   void kmovql(KRegister dst, Address src);
1515 
1516   void movdl(XMMRegister dst, Register src);
1517   void movdl(Register dst, XMMRegister src);
1518   void movdl(XMMRegister dst, Address src);
1519   void movdl(Address dst, XMMRegister src);
1520 
1521   // Move Double Quadword
1522   void movdq(XMMRegister dst, Register src);
1523   void movdq(Register dst, XMMRegister src);
1524 
1525   // Move Aligned Double Quadword
1526   void movdqa(XMMRegister dst, XMMRegister src);
1527   void movdqa(XMMRegister dst, Address src);
1528 
1529   // Move Unaligned Double Quadword
1530   void movdqu(Address     dst, XMMRegister src);
1531   void movdqu(XMMRegister dst, Address src);
1532   void movdqu(XMMRegister dst, XMMRegister src);
1533 
1534   // Move Unaligned 256bit Vector
1535   void vmovdqu(Address dst, XMMRegister src);
1536   void vmovdqu(XMMRegister dst, Address src);
1537   void vmovdqu(XMMRegister dst, XMMRegister src);
1538 
1539    // Move Unaligned 512bit Vector
1540   void evmovdqul(Address dst, XMMRegister src, int vector_len);
1541   void evmovdqul(XMMRegister dst, Address src, int vector_len);
1542   void evmovdqul(XMMRegister dst, XMMRegister src, int vector_len);
1543   void evmovdquq(Address dst, XMMRegister src, int vector_len);
1544   void evmovdquq(XMMRegister dst, Address src, int vector_len);
1545   void evmovdquq(XMMRegister dst, XMMRegister src, int vector_len);
1546 
1547   // Move lower 64bit to high 64bit in 128bit register
1548   void movlhps(XMMRegister dst, XMMRegister src);
1549 
1550   void movl(Register dst, int32_t imm32);
1551   void movl(Address dst, int32_t imm32);
1552   void movl(Register dst, Register src);
1553   void movl(Register dst, Address src);
1554   void movl(Address dst, Register src);
1555 
1556   // These dummies prevent using movl from converting a zero (like NULL) into Register
1557   // by giving the compiler two choices it can't resolve
1558 
1559   void movl(Address  dst, void* junk);
1560   void movl(Register dst, void* junk);
1561 
1562 #ifdef _LP64
1563   void movq(Register dst, Register src);
1564   void movq(Register dst, Address src);
1565   void movq(Address  dst, Register src);
1566 #endif
1567 
1568   void movq(Address     dst, MMXRegister src );
1569   void movq(MMXRegister dst, Address src );
1570 
1571 #ifdef _LP64
1572   // These dummies prevent using movq from converting a zero (like NULL) into Register
1573   // by giving the compiler two choices it can't resolve
1574 
1575   void movq(Address  dst, void* dummy);
1576   void movq(Register dst, void* dummy);
1577 #endif
1578 
1579   // Move Quadword
1580   void movq(Address     dst, XMMRegister src);
1581   void movq(XMMRegister dst, Address src);
1582 
1583   void movsbl(Register dst, Address src);
1584   void movsbl(Register dst, Register src);
1585 
1586 #ifdef _LP64
1587   void movsbq(Register dst, Address src);
1588   void movsbq(Register dst, Register src);
1589 
1590   // Move signed 32bit immediate to 64bit extending sign
1591   void movslq(Address  dst, int32_t imm64);
1592   void movslq(Register dst, int32_t imm64);
1593 
1594   void movslq(Register dst, Address src);
1595   void movslq(Register dst, Register src);
1596   void movslq(Register dst, void* src); // Dummy declaration to cause NULL to be ambiguous
1597 #endif
1598 
1599   void movswl(Register dst, Address src);
1600   void movswl(Register dst, Register src);
1601 
1602 #ifdef _LP64
1603   void movswq(Register dst, Address src);
1604   void movswq(Register dst, Register src);
1605 #endif
1606 
1607   void movw(Address dst, int imm16);
1608   void movw(Register dst, Address src);
1609   void movw(Address dst, Register src);
1610 
1611   void movzbl(Register dst, Address src);
1612   void movzbl(Register dst, Register src);
1613 
1614 #ifdef _LP64
1615   void movzbq(Register dst, Address src);
1616   void movzbq(Register dst, Register src);
1617 #endif
1618 
1619   void movzwl(Register dst, Address src);
1620   void movzwl(Register dst, Register src);
1621 
1622 #ifdef _LP64
1623   void movzwq(Register dst, Address src);
1624   void movzwq(Register dst, Register src);
1625 #endif
1626 
1627   // Unsigned multiply with RAX destination register
1628   void mull(Address src);
1629   void mull(Register src);
1630 
1631 #ifdef _LP64
1632   void mulq(Address src);
1633   void mulq(Register src);
1634   void mulxq(Register dst1, Register dst2, Register src);
1635 #endif
1636 
1637   // Multiply Scalar Double-Precision Floating-Point Values
1638   void mulsd(XMMRegister dst, Address src);
1639   void mulsd(XMMRegister dst, XMMRegister src);
1640 
1641   // Multiply Scalar Single-Precision Floating-Point Values
1642   void mulss(XMMRegister dst, Address src);
1643   void mulss(XMMRegister dst, XMMRegister src);
1644 
1645   void negl(Register dst);
1646 
1647 #ifdef _LP64
1648   void negq(Register dst);
1649 #endif
1650 
1651   void nop(int i = 1);
1652 
1653   void notl(Register dst);
1654 
1655 #ifdef _LP64
1656   void notq(Register dst);
1657 #endif
1658 
1659   void orl(Address dst, int32_t imm32);
1660   void orl(Register dst, int32_t imm32);
1661   void orl(Register dst, Address src);
1662   void orl(Register dst, Register src);
1663   void orl(Address dst, Register src);
1664 
1665   void orq(Address dst, int32_t imm32);
1666   void orq(Register dst, int32_t imm32);
1667   void orq(Register dst, Address src);
1668   void orq(Register dst, Register src);
1669 
1670   // Pack with unsigned saturation
1671   void packuswb(XMMRegister dst, XMMRegister src);
1672   void packuswb(XMMRegister dst, Address src);
1673   void vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1674 
1675   // Pemutation of 64bit words
1676   void vpermq(XMMRegister dst, XMMRegister src, int imm8, int vector_len);
1677   void vpermq(XMMRegister dst, XMMRegister src, int imm8);
1678 
1679   void pause();
1680 
1681   // SSE4.2 string instructions
1682   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
1683   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
1684 
1685   void pcmpeqw(XMMRegister dst, XMMRegister src);
1686   void vpcmpeqw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1687 
1688   void pmovmskb(Register dst, XMMRegister src);
1689   void vpmovmskb(Register dst, XMMRegister src);
1690 
1691   // SSE 4.1 extract
1692   void pextrd(Register dst, XMMRegister src, int imm8);
1693   void pextrq(Register dst, XMMRegister src, int imm8);
1694   // SSE 2 extract
1695   void pextrw(Register dst, XMMRegister src, int imm8);
1696 
1697   // SSE 4.1 insert
1698   void pinsrd(XMMRegister dst, Register src, int imm8);
1699   void pinsrq(XMMRegister dst, Register src, int imm8);
1700   // SSE 2 insert
1701   void pinsrw(XMMRegister dst, Register src, int imm8);
1702 
1703   // SSE4.1 packed move
1704   void pmovzxbw(XMMRegister dst, XMMRegister src);
1705   void pmovzxbw(XMMRegister dst, Address src);
1706 
1707   void vpmovzxbw(XMMRegister dst, Address src);
1708 
1709 #ifndef _LP64 // no 32bit push/pop on amd64
1710   void popl(Address dst);
1711 #endif
1712 
1713 #ifdef _LP64
1714   void popq(Address dst);
1715 #endif
1716 
1717   void popcntl(Register dst, Address src);
1718   void popcntl(Register dst, Register src);
1719 
1720 #ifdef _LP64
1721   void popcntq(Register dst, Address src);
1722   void popcntq(Register dst, Register src);
1723 #endif
1724 
1725   // Prefetches (SSE, SSE2, 3DNOW only)
1726 
1727   void prefetchnta(Address src);
1728   void prefetchr(Address src);
1729   void prefetcht0(Address src);
1730   void prefetcht1(Address src);
1731   void prefetcht2(Address src);
1732   void prefetchw(Address src);
1733 
1734   // Shuffle Bytes
1735   void pshufb(XMMRegister dst, XMMRegister src);
1736   void pshufb(XMMRegister dst, Address src);
1737 
1738   // Shuffle Packed Doublewords
1739   void pshufd(XMMRegister dst, XMMRegister src, int mode);
1740   void pshufd(XMMRegister dst, Address src,     int mode);
1741 
1742   // Shuffle Packed Low Words
1743   void pshuflw(XMMRegister dst, XMMRegister src, int mode);
1744   void pshuflw(XMMRegister dst, Address src,     int mode);
1745 
1746   // Shift Right by bytes Logical DoubleQuadword Immediate
1747   void psrldq(XMMRegister dst, int shift);
1748   // Shift Left by bytes Logical DoubleQuadword Immediate
1749   void pslldq(XMMRegister dst, int shift);
1750 
1751   // Logical Compare 128bit
1752   void ptest(XMMRegister dst, XMMRegister src);
1753   void ptest(XMMRegister dst, Address src);
1754   // Logical Compare 256bit
1755   void vptest(XMMRegister dst, XMMRegister src);
1756   void vptest(XMMRegister dst, Address src);
1757 
1758   // Interleave Low Bytes
1759   void punpcklbw(XMMRegister dst, XMMRegister src);
1760   void punpcklbw(XMMRegister dst, Address src);
1761 
1762   // Interleave Low Doublewords
1763   void punpckldq(XMMRegister dst, XMMRegister src);
1764   void punpckldq(XMMRegister dst, Address src);
1765 
1766   // Interleave Low Quadwords
1767   void punpcklqdq(XMMRegister dst, XMMRegister src);
1768 
1769 #ifndef _LP64 // no 32bit push/pop on amd64
1770   void pushl(Address src);
1771 #endif
1772 
1773   void pushq(Address src);
1774 
1775   void rcll(Register dst, int imm8);
1776 
1777   void rclq(Register dst, int imm8);
1778 
1779   void rcrq(Register dst, int imm8);
1780 
1781   void rcpps(XMMRegister dst, XMMRegister src);
1782 
1783   void rcpss(XMMRegister dst, XMMRegister src);
1784 
1785   void rdtsc();
1786 
1787   void ret(int imm16);
1788 
1789 #ifdef _LP64
1790   void rorq(Register dst, int imm8);
1791   void rorxq(Register dst, Register src, int imm8);
1792 #endif
1793 
1794   void sahf();
1795 
1796   void sarl(Register dst, int imm8);
1797   void sarl(Register dst);
1798 
1799   void sarq(Register dst, int imm8);
1800   void sarq(Register dst);
1801 
1802   void sbbl(Address dst, int32_t imm32);
1803   void sbbl(Register dst, int32_t imm32);
1804   void sbbl(Register dst, Address src);
1805   void sbbl(Register dst, Register src);
1806 
1807   void sbbq(Address dst, int32_t imm32);
1808   void sbbq(Register dst, int32_t imm32);
1809   void sbbq(Register dst, Address src);
1810   void sbbq(Register dst, Register src);
1811 
1812   void setb(Condition cc, Register dst);
1813 
1814   void shldl(Register dst, Register src);
1815   void shldl(Register dst, Register src, int8_t imm8);
1816 
1817   void shll(Register dst, int imm8);
1818   void shll(Register dst);
1819 
1820   void shlq(Register dst, int imm8);
1821   void shlq(Register dst);
1822 
1823   void shrdl(Register dst, Register src);
1824 
1825   void shrl(Register dst, int imm8);
1826   void shrl(Register dst);
1827 
1828   void shrq(Register dst, int imm8);
1829   void shrq(Register dst);
1830 
1831   void smovl(); // QQQ generic?
1832 
1833   // Compute Square Root of Scalar Double-Precision Floating-Point Value
1834   void sqrtsd(XMMRegister dst, Address src);
1835   void sqrtsd(XMMRegister dst, XMMRegister src);
1836 
1837   // Compute Square Root of Scalar Single-Precision Floating-Point Value
1838   void sqrtss(XMMRegister dst, Address src);
1839   void sqrtss(XMMRegister dst, XMMRegister src);
1840 
1841   void std();
1842 
1843   void stmxcsr( Address dst );
1844 
1845   void subl(Address dst, int32_t imm32);
1846   void subl(Address dst, Register src);
1847   void subl(Register dst, int32_t imm32);
1848   void subl(Register dst, Address src);
1849   void subl(Register dst, Register src);
1850 
1851   void subq(Address dst, int32_t imm32);
1852   void subq(Address dst, Register src);
1853   void subq(Register dst, int32_t imm32);
1854   void subq(Register dst, Address src);
1855   void subq(Register dst, Register src);
1856 
1857   // Force generation of a 4 byte immediate value even if it fits into 8bit
1858   void subl_imm32(Register dst, int32_t imm32);
1859   void subq_imm32(Register dst, int32_t imm32);
1860 
1861   // Subtract Scalar Double-Precision Floating-Point Values
1862   void subsd(XMMRegister dst, Address src);
1863   void subsd(XMMRegister dst, XMMRegister src);
1864 
1865   // Subtract Scalar Single-Precision Floating-Point Values
1866   void subss(XMMRegister dst, Address src);
1867   void subss(XMMRegister dst, XMMRegister src);
1868 
1869   void testb(Register dst, int imm8);
1870 
1871   void testl(Register dst, int32_t imm32);
1872   void testl(Register dst, Register src);
1873   void testl(Register dst, Address src);
1874 
1875   void testq(Register dst, int32_t imm32);
1876   void testq(Register dst, Register src);
1877 
1878   // BMI - count trailing zeros
1879   void tzcntl(Register dst, Register src);
1880   void tzcntq(Register dst, Register src);
1881 
1882   // Unordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1883   void ucomisd(XMMRegister dst, Address src);
1884   void ucomisd(XMMRegister dst, XMMRegister src);
1885 
1886   // Unordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1887   void ucomiss(XMMRegister dst, Address src);
1888   void ucomiss(XMMRegister dst, XMMRegister src);
1889 
1890   void xabort(int8_t imm8);
1891 
1892   void xaddl(Address dst, Register src);
1893 
1894   void xaddq(Address dst, Register src);
1895 
1896   void xbegin(Label& abort, relocInfo::relocType rtype = relocInfo::none);
1897 
1898   void xchgl(Register reg, Address adr);
1899   void xchgl(Register dst, Register src);
1900 
1901   void xchgq(Register reg, Address adr);
1902   void xchgq(Register dst, Register src);
1903 
1904   void xend();
1905 
1906   // Get Value of Extended Control Register
1907   void xgetbv();
1908 
1909   void xorl(Register dst, int32_t imm32);
1910   void xorl(Register dst, Address src);
1911   void xorl(Register dst, Register src);
1912 
1913   void xorq(Register dst, Address src);
1914   void xorq(Register dst, Register src);
1915 
1916   void set_byte_if_not_zero(Register dst); // sets reg to 1 if not zero, otherwise 0
1917 
1918   // AVX 3-operands scalar instructions (encoded with VEX prefix)
1919 
1920   void vaddsd(XMMRegister dst, XMMRegister nds, Address src);
1921   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1922   void vaddss(XMMRegister dst, XMMRegister nds, Address src);
1923   void vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1924   void vdivsd(XMMRegister dst, XMMRegister nds, Address src);
1925   void vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1926   void vdivss(XMMRegister dst, XMMRegister nds, Address src);
1927   void vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1928   void vmulsd(XMMRegister dst, XMMRegister nds, Address src);
1929   void vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1930   void vmulss(XMMRegister dst, XMMRegister nds, Address src);
1931   void vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1932   void vsubsd(XMMRegister dst, XMMRegister nds, Address src);
1933   void vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1934   void vsubss(XMMRegister dst, XMMRegister nds, Address src);
1935   void vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1936 
1937 
1938   //====================VECTOR ARITHMETIC=====================================
1939 
1940   // Add Packed Floating-Point Values
1941   void addpd(XMMRegister dst, XMMRegister src);
1942   void addps(XMMRegister dst, XMMRegister src);
1943   void vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1944   void vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1945   void vaddpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1946   void vaddps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1947 
1948   // Subtract Packed Floating-Point Values
1949   void subpd(XMMRegister dst, XMMRegister src);
1950   void subps(XMMRegister dst, XMMRegister src);
1951   void vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1952   void vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1953   void vsubpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1954   void vsubps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1955 
1956   // Multiply Packed Floating-Point Values
1957   void mulpd(XMMRegister dst, XMMRegister src);
1958   void mulpd(XMMRegister dst, Address src);
1959   void mulps(XMMRegister dst, XMMRegister src);
1960   void vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1961   void vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1962   void vmulpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1963   void vmulps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1964 
1965   // Divide Packed Floating-Point Values
1966   void divpd(XMMRegister dst, XMMRegister src);
1967   void divps(XMMRegister dst, XMMRegister src);
1968   void vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1969   void vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1970   void vdivpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1971   void vdivps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1972 
1973   // Sqrt Packed Floating-Point Values - Double precision only
1974   void vsqrtpd(XMMRegister dst, XMMRegister src, int vector_len);
1975   void vsqrtpd(XMMRegister dst, Address src, int vector_len);
1976 
1977   // Bitwise Logical AND of Packed Floating-Point Values
1978   void andpd(XMMRegister dst, XMMRegister src);
1979   void andps(XMMRegister dst, XMMRegister src);
1980   void vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1981   void vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1982   void vandpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1983   void vandps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1984 
1985   void unpckhpd(XMMRegister dst, XMMRegister src);
1986   void unpcklpd(XMMRegister dst, XMMRegister src);
1987 
1988   // Bitwise Logical XOR of Packed Floating-Point Values
1989   void xorpd(XMMRegister dst, XMMRegister src);
1990   void xorps(XMMRegister dst, XMMRegister src);
1991   void vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1992   void vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1993   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1994   void vxorps(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
1995 
1996   // Add horizontal packed integers
1997   void vphaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1998   void vphaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
1999   void phaddw(XMMRegister dst, XMMRegister src);
2000   void phaddd(XMMRegister dst, XMMRegister src);
2001 
2002   // Add packed integers
2003   void paddb(XMMRegister dst, XMMRegister src);
2004   void paddw(XMMRegister dst, XMMRegister src);
2005   void paddd(XMMRegister dst, XMMRegister src);
2006   void paddq(XMMRegister dst, XMMRegister src);
2007   void vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2008   void vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2009   void vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2010   void vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2011   void vpaddb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2012   void vpaddw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2013   void vpaddd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2014   void vpaddq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2015 
2016   // Sub packed integers
2017   void psubb(XMMRegister dst, XMMRegister src);
2018   void psubw(XMMRegister dst, XMMRegister src);
2019   void psubd(XMMRegister dst, XMMRegister src);
2020   void psubq(XMMRegister dst, XMMRegister src);
2021   void vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2022   void vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2023   void vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2024   void vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2025   void vpsubb(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2026   void vpsubw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2027   void vpsubd(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2028   void vpsubq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2029 
2030   // Multiply packed integers (only shorts and ints)
2031   void pmullw(XMMRegister dst, XMMRegister src);
2032   void pmulld(XMMRegister dst, XMMRegister src);
2033   void vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2034   void vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2035   void vpmullq(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2036   void vpmullw(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2037   void vpmulld(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2038   void vpmullq(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2039 
2040   // Shift left packed integers
2041   void psllw(XMMRegister dst, int shift);
2042   void pslld(XMMRegister dst, int shift);
2043   void psllq(XMMRegister dst, int shift);
2044   void psllw(XMMRegister dst, XMMRegister shift);
2045   void pslld(XMMRegister dst, XMMRegister shift);
2046   void psllq(XMMRegister dst, XMMRegister shift);
2047   void vpsllw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2048   void vpslld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2049   void vpsllq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2050   void vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2051   void vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2052   void vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2053 
2054   // Logical shift right packed integers
2055   void psrlw(XMMRegister dst, int shift);
2056   void psrld(XMMRegister dst, int shift);
2057   void psrlq(XMMRegister dst, int shift);
2058   void psrlw(XMMRegister dst, XMMRegister shift);
2059   void psrld(XMMRegister dst, XMMRegister shift);
2060   void psrlq(XMMRegister dst, XMMRegister shift);
2061   void vpsrlw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2062   void vpsrld(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2063   void vpsrlq(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2064   void vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2065   void vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2066   void vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2067 
2068   // Arithmetic shift right packed integers (only shorts and ints, no instructions for longs)
2069   void psraw(XMMRegister dst, int shift);
2070   void psrad(XMMRegister dst, int shift);
2071   void psraw(XMMRegister dst, XMMRegister shift);
2072   void psrad(XMMRegister dst, XMMRegister shift);
2073   void vpsraw(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2074   void vpsrad(XMMRegister dst, XMMRegister src, int shift, int vector_len);
2075   void vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2076   void vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, int vector_len);
2077 
2078   // And packed integers
2079   void pand(XMMRegister dst, XMMRegister src);
2080   void vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2081   void vpand(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2082 
2083   // Andn packed integers
2084   void pandn(XMMRegister dst, XMMRegister src);
2085 
2086   // Or packed integers
2087   void por(XMMRegister dst, XMMRegister src);
2088   void vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2089   void vpor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2090 
2091   // Xor packed integers
2092   void pxor(XMMRegister dst, XMMRegister src);
2093   void vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, int vector_len);
2094   void vpxor(XMMRegister dst, XMMRegister nds, Address src, int vector_len);
2095 
2096   // Copy low 128bit into high 128bit of YMM registers.
2097   void vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
2098   void vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
2099   void vextractf128h(XMMRegister dst, XMMRegister src);
2100   void vextracti128h(XMMRegister dst, XMMRegister src);
2101 
2102   // Load/store high 128bit of YMM registers which does not destroy other half.
2103   void vinsertf128h(XMMRegister dst, Address src);
2104   void vinserti128h(XMMRegister dst, Address src);
2105   void vextractf128h(Address dst, XMMRegister src);
2106   void vextracti128h(Address dst, XMMRegister src);
2107 
2108   // Copy low 256bit into high 256bit of ZMM registers.
2109   void vinserti64x4h(XMMRegister dst, XMMRegister nds, XMMRegister src);
2110   void vinsertf64x4h(XMMRegister dst, XMMRegister nds, XMMRegister src);
2111   void vextracti64x4h(XMMRegister dst, XMMRegister src);
2112   void vextractf64x4h(XMMRegister dst, XMMRegister src);
2113   void vextractf64x4h(Address dst, XMMRegister src);
2114   void vinsertf64x4h(XMMRegister dst, Address src);
2115 
2116   // Copy targeted 128bit segments of the ZMM registers
2117   void vextracti64x2h(XMMRegister dst, XMMRegister src, int value);
2118   void vextractf64x2h(XMMRegister dst, XMMRegister src, int value);
2119   void vextractf32x4h(XMMRegister dst, XMMRegister src, int value);
2120   void vextractf32x4h(Address dst, XMMRegister src, int value);
2121   void vinsertf32x4h(XMMRegister dst, XMMRegister nds, XMMRegister src, int value);
2122   void vinsertf32x4h(XMMRegister dst, Address src, int value);
2123 
2124   // duplicate 4-bytes integer data from src into 8 locations in dest
2125   void vpbroadcastd(XMMRegister dst, XMMRegister src);
2126 
2127   // duplicate 2-bytes integer data from src into 16 locations in dest
2128   void vpbroadcastw(XMMRegister dst, XMMRegister src);
2129 
2130   // duplicate n-bytes integer data from src into vector_len locations in dest
2131   void evpbroadcastb(XMMRegister dst, XMMRegister src, int vector_len);
2132   void evpbroadcastb(XMMRegister dst, Address src, int vector_len);
2133   void evpbroadcastw(XMMRegister dst, XMMRegister src, int vector_len);
2134   void evpbroadcastw(XMMRegister dst, Address src, int vector_len);
2135   void evpbroadcastd(XMMRegister dst, XMMRegister src, int vector_len);
2136   void evpbroadcastd(XMMRegister dst, Address src, int vector_len);
2137   void evpbroadcastq(XMMRegister dst, XMMRegister src, int vector_len);
2138   void evpbroadcastq(XMMRegister dst, Address src, int vector_len);
2139 
2140   void evpbroadcastss(XMMRegister dst, XMMRegister src, int vector_len);
2141   void evpbroadcastss(XMMRegister dst, Address src, int vector_len);
2142   void evpbroadcastsd(XMMRegister dst, XMMRegister src, int vector_len);
2143   void evpbroadcastsd(XMMRegister dst, Address src, int vector_len);
2144 
2145   void evpbroadcastb(XMMRegister dst, Register src, int vector_len);
2146   void evpbroadcastw(XMMRegister dst, Register src, int vector_len);
2147   void evpbroadcastd(XMMRegister dst, Register src, int vector_len);
2148   void evpbroadcastq(XMMRegister dst, Register src, int vector_len);
2149 
2150   // Carry-Less Multiplication Quadword
2151   void pclmulqdq(XMMRegister dst, XMMRegister src, int mask);
2152   void vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask);
2153 
2154   // AVX instruction which is used to clear upper 128 bits of YMM registers and
2155   // to avoid transaction penalty between AVX and SSE states. There is no
2156   // penalty if legacy SSE instructions are encoded using VEX prefix because
2157   // they always clear upper 128 bits. It should be used before calling
2158   // runtime code and native libraries.
2159   void vzeroupper();
2160 
2161   // AVX support for vectorized conditional move (double). The following two instructions used only coupled.
2162   void cmppd(XMMRegister dst, XMMRegister nds, XMMRegister src, int cop, int vector_len);
2163   void vpblendd(XMMRegister dst, XMMRegister nds, XMMRegister src1, XMMRegister src2, int vector_len);
2164 
2165 
2166  protected:
2167   // Next instructions require address alignment 16 bytes SSE mode.
2168   // They should be called only from corresponding MacroAssembler instructions.
2169   void andpd(XMMRegister dst, Address src);
2170   void andps(XMMRegister dst, Address src);
2171   void xorpd(XMMRegister dst, Address src);
2172   void xorps(XMMRegister dst, Address src);
2173 
2174 };
2175 
2176 #endif // CPU_X86_VM_ASSEMBLER_X86_HPP