1 /*
   2  * Copyright (c) 1997, 2013, 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 
  30 class BiasedLockingCounters;
  31 
  32 // Contains all the definitions needed for x86 assembly code generation.
  33 
  34 // Calling convention
  35 class Argument VALUE_OBJ_CLASS_SPEC {
  36  public:
  37   enum {
  38 #ifdef _LP64
  39 #ifdef _WIN64
  40     n_int_register_parameters_c   = 4, // rcx, rdx, r8, r9 (c_rarg0, c_rarg1, ...)
  41     n_float_register_parameters_c = 4,  // xmm0 - xmm3 (c_farg0, c_farg1, ... )
  42 #else
  43     n_int_register_parameters_c   = 6, // rdi, rsi, rdx, rcx, r8, r9 (c_rarg0, c_rarg1, ...)
  44     n_float_register_parameters_c = 8,  // xmm0 - xmm7 (c_farg0, c_farg1, ... )
  45 #endif // _WIN64
  46     n_int_register_parameters_j   = 6, // j_rarg0, j_rarg1, ...
  47     n_float_register_parameters_j = 8  // j_farg0, j_farg1, ...
  48 #else
  49     n_register_parameters = 0   // 0 registers used to pass arguments
  50 #endif // _LP64
  51   };
  52 };
  53 
  54 
  55 #ifdef _LP64
  56 // Symbolically name the register arguments used by the c calling convention.
  57 // Windows is different from linux/solaris. So much for standards...
  58 
  59 #ifdef _WIN64
  60 
  61 REGISTER_DECLARATION(Register, c_rarg0, rcx);
  62 REGISTER_DECLARATION(Register, c_rarg1, rdx);
  63 REGISTER_DECLARATION(Register, c_rarg2, r8);
  64 REGISTER_DECLARATION(Register, c_rarg3, r9);
  65 
  66 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  67 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  68 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  69 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  70 
  71 #else
  72 
  73 REGISTER_DECLARATION(Register, c_rarg0, rdi);
  74 REGISTER_DECLARATION(Register, c_rarg1, rsi);
  75 REGISTER_DECLARATION(Register, c_rarg2, rdx);
  76 REGISTER_DECLARATION(Register, c_rarg3, rcx);
  77 REGISTER_DECLARATION(Register, c_rarg4, r8);
  78 REGISTER_DECLARATION(Register, c_rarg5, r9);
  79 
  80 REGISTER_DECLARATION(XMMRegister, c_farg0, xmm0);
  81 REGISTER_DECLARATION(XMMRegister, c_farg1, xmm1);
  82 REGISTER_DECLARATION(XMMRegister, c_farg2, xmm2);
  83 REGISTER_DECLARATION(XMMRegister, c_farg3, xmm3);
  84 REGISTER_DECLARATION(XMMRegister, c_farg4, xmm4);
  85 REGISTER_DECLARATION(XMMRegister, c_farg5, xmm5);
  86 REGISTER_DECLARATION(XMMRegister, c_farg6, xmm6);
  87 REGISTER_DECLARATION(XMMRegister, c_farg7, xmm7);
  88 
  89 #endif // _WIN64
  90 
  91 // Symbolically name the register arguments used by the Java calling convention.
  92 // We have control over the convention for java so we can do what we please.
  93 // What pleases us is to offset the java calling convention so that when
  94 // we call a suitable jni method the arguments are lined up and we don't
  95 // have to do little shuffling. A suitable jni method is non-static and a
  96 // small number of arguments (two fewer args on windows)
  97 //
  98 //        |-------------------------------------------------------|
  99 //        | c_rarg0   c_rarg1  c_rarg2 c_rarg3 c_rarg4 c_rarg5    |
 100 //        |-------------------------------------------------------|
 101 //        | rcx       rdx      r8      r9      rdi*    rsi*       | windows (* not a c_rarg)
 102 //        | rdi       rsi      rdx     rcx     r8      r9         | solaris/linux
 103 //        |-------------------------------------------------------|
 104 //        | j_rarg5   j_rarg0  j_rarg1 j_rarg2 j_rarg3 j_rarg4    |
 105 //        |-------------------------------------------------------|
 106 
 107 REGISTER_DECLARATION(Register, j_rarg0, c_rarg1);
 108 REGISTER_DECLARATION(Register, j_rarg1, c_rarg2);
 109 REGISTER_DECLARATION(Register, j_rarg2, c_rarg3);
 110 // Windows runs out of register args here
 111 #ifdef _WIN64
 112 REGISTER_DECLARATION(Register, j_rarg3, rdi);
 113 REGISTER_DECLARATION(Register, j_rarg4, rsi);
 114 #else
 115 REGISTER_DECLARATION(Register, j_rarg3, c_rarg4);
 116 REGISTER_DECLARATION(Register, j_rarg4, c_rarg5);
 117 #endif /* _WIN64 */
 118 REGISTER_DECLARATION(Register, j_rarg5, c_rarg0);
 119 
 120 REGISTER_DECLARATION(XMMRegister, j_farg0, xmm0);
 121 REGISTER_DECLARATION(XMMRegister, j_farg1, xmm1);
 122 REGISTER_DECLARATION(XMMRegister, j_farg2, xmm2);
 123 REGISTER_DECLARATION(XMMRegister, j_farg3, xmm3);
 124 REGISTER_DECLARATION(XMMRegister, j_farg4, xmm4);
 125 REGISTER_DECLARATION(XMMRegister, j_farg5, xmm5);
 126 REGISTER_DECLARATION(XMMRegister, j_farg6, xmm6);
 127 REGISTER_DECLARATION(XMMRegister, j_farg7, xmm7);
 128 
 129 REGISTER_DECLARATION(Register, rscratch1, r10);  // volatile
 130 REGISTER_DECLARATION(Register, rscratch2, r11);  // volatile
 131 
 132 REGISTER_DECLARATION(Register, r12_heapbase, r12); // callee-saved
 133 REGISTER_DECLARATION(Register, r15_thread, r15); // callee-saved
 134 
 135 #else
 136 // rscratch1 will apear in 32bit code that is dead but of course must compile
 137 // Using noreg ensures if the dead code is incorrectly live and executed it
 138 // will cause an assertion failure
 139 #define rscratch1 noreg
 140 #define rscratch2 noreg
 141 
 142 #endif // _LP64
 143 
 144 // JSR 292 fixed register usages:
 145 REGISTER_DECLARATION(Register, rbp_mh_SP_save, rbp);
 146 
 147 // Address is an abstraction used to represent a memory location
 148 // using any of the amd64 addressing modes with one object.
 149 //
 150 // Note: A register location is represented via a Register, not
 151 //       via an address for efficiency & simplicity reasons.
 152 
 153 class ArrayAddress;
 154 
 155 class Address VALUE_OBJ_CLASS_SPEC {
 156  public:
 157   enum ScaleFactor {
 158     no_scale = -1,
 159     times_1  =  0,
 160     times_2  =  1,
 161     times_4  =  2,
 162     times_8  =  3,
 163     times_ptr = LP64_ONLY(times_8) NOT_LP64(times_4)
 164   };
 165   static ScaleFactor times(int size) {
 166     assert(size >= 1 && size <= 8 && is_power_of_2(size), "bad scale size");
 167     if (size == 8)  return times_8;
 168     if (size == 4)  return times_4;
 169     if (size == 2)  return times_2;
 170     return times_1;
 171   }
 172   static int scale_size(ScaleFactor scale) {
 173     assert(scale != no_scale, "");
 174     assert(((1 << (int)times_1) == 1 &&
 175             (1 << (int)times_2) == 2 &&
 176             (1 << (int)times_4) == 4 &&
 177             (1 << (int)times_8) == 8), "");
 178     return (1 << (int)scale);
 179   }
 180 
 181  private:
 182   Register         _base;
 183   Register         _index;
 184   ScaleFactor      _scale;
 185   int              _disp;
 186   RelocationHolder _rspec;
 187 
 188   // Easily misused constructors make them private
 189   // %%% can we make these go away?
 190   NOT_LP64(Address(address loc, RelocationHolder spec);)
 191   Address(int disp, address loc, relocInfo::relocType rtype);
 192   Address(int disp, address loc, RelocationHolder spec);
 193 
 194  public:
 195 
 196  int disp() { return _disp; }
 197   // creation
 198   Address()
 199     : _base(noreg),
 200       _index(noreg),
 201       _scale(no_scale),
 202       _disp(0) {
 203   }
 204 
 205   // No default displacement otherwise Register can be implicitly
 206   // converted to 0(Register) which is quite a different animal.
 207 
 208   Address(Register base, int disp)
 209     : _base(base),
 210       _index(noreg),
 211       _scale(no_scale),
 212       _disp(disp) {
 213   }
 214 
 215   Address(Register base, Register index, ScaleFactor scale, int disp = 0)
 216     : _base (base),
 217       _index(index),
 218       _scale(scale),
 219       _disp (disp) {
 220     assert(!index->is_valid() == (scale == Address::no_scale),
 221            "inconsistent address");
 222   }
 223 
 224   Address(Register base, RegisterOrConstant index, ScaleFactor scale = times_1, int disp = 0)
 225     : _base (base),
 226       _index(index.register_or_noreg()),
 227       _scale(scale),
 228       _disp (disp + (index.constant_or_zero() * scale_size(scale))) {
 229     if (!index.is_register())  scale = Address::no_scale;
 230     assert(!_index->is_valid() == (scale == Address::no_scale),
 231            "inconsistent address");
 232   }
 233 
 234   Address plus_disp(int disp) const {
 235     Address a = (*this);
 236     a._disp += disp;
 237     return a;
 238   }
 239   Address plus_disp(RegisterOrConstant disp, ScaleFactor scale = times_1) const {
 240     Address a = (*this);
 241     a._disp += disp.constant_or_zero() * scale_size(scale);
 242     if (disp.is_register()) {
 243       assert(!a.index()->is_valid(), "competing indexes");
 244       a._index = disp.as_register();
 245       a._scale = scale;
 246     }
 247     return a;
 248   }
 249   bool is_same_address(Address a) const {
 250     // disregard _rspec
 251     return _base == a._base && _disp == a._disp && _index == a._index && _scale == a._scale;
 252   }
 253 
 254   // The following two overloads are used in connection with the
 255   // ByteSize type (see sizes.hpp).  They simplify the use of
 256   // ByteSize'd arguments in assembly code. Note that their equivalent
 257   // for the optimized build are the member functions with int disp
 258   // argument since ByteSize is mapped to an int type in that case.
 259   //
 260   // Note: DO NOT introduce similar overloaded functions for WordSize
 261   // arguments as in the optimized mode, both ByteSize and WordSize
 262   // are mapped to the same type and thus the compiler cannot make a
 263   // distinction anymore (=> compiler errors).
 264 
 265 #ifdef ASSERT
 266   Address(Register base, ByteSize disp)
 267     : _base(base),
 268       _index(noreg),
 269       _scale(no_scale),
 270       _disp(in_bytes(disp)) {
 271   }
 272 
 273   Address(Register base, Register index, ScaleFactor scale, ByteSize disp)
 274     : _base(base),
 275       _index(index),
 276       _scale(scale),
 277       _disp(in_bytes(disp)) {
 278     assert(!index->is_valid() == (scale == Address::no_scale),
 279            "inconsistent address");
 280   }
 281 
 282   Address(Register base, RegisterOrConstant index, ScaleFactor scale, ByteSize disp)
 283     : _base (base),
 284       _index(index.register_or_noreg()),
 285       _scale(scale),
 286       _disp (in_bytes(disp) + (index.constant_or_zero() * scale_size(scale))) {
 287     if (!index.is_register())  scale = Address::no_scale;
 288     assert(!_index->is_valid() == (scale == Address::no_scale),
 289            "inconsistent address");
 290   }
 291 
 292 #endif // ASSERT
 293 
 294   // accessors
 295   bool        uses(Register reg) const { return _base == reg || _index == reg; }
 296   Register    base()             const { return _base;  }
 297   Register    index()            const { return _index; }
 298   ScaleFactor scale()            const { return _scale; }
 299   int         disp()             const { return _disp;  }
 300 
 301   // Convert the raw encoding form into the form expected by the constructor for
 302   // Address.  An index of 4 (rsp) corresponds to having no index, so convert
 303   // that to noreg for the Address constructor.
 304   static Address make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc);
 305 
 306   static Address make_array(ArrayAddress);
 307 
 308  private:
 309   bool base_needs_rex() const {
 310     return _base != noreg && _base->encoding() >= 8;
 311   }
 312 
 313   bool index_needs_rex() const {
 314     return _index != noreg &&_index->encoding() >= 8;
 315   }
 316 
 317   relocInfo::relocType reloc() const { return _rspec.type(); }
 318 
 319   friend class Assembler;
 320   friend class MacroAssembler;
 321   friend class LIR_Assembler; // base/index/scale/disp
 322 };
 323 
 324 //
 325 // AddressLiteral has been split out from Address because operands of this type
 326 // need to be treated specially on 32bit vs. 64bit platforms. By splitting it out
 327 // the few instructions that need to deal with address literals are unique and the
 328 // MacroAssembler does not have to implement every instruction in the Assembler
 329 // in order to search for address literals that may need special handling depending
 330 // on the instruction and the platform. As small step on the way to merging i486/amd64
 331 // directories.
 332 //
 333 class AddressLiteral VALUE_OBJ_CLASS_SPEC {
 334   friend class ArrayAddress;
 335   RelocationHolder _rspec;
 336   // Typically we use AddressLiterals we want to use their rval
 337   // However in some situations we want the lval (effect address) of the item.
 338   // We provide a special factory for making those lvals.
 339   bool _is_lval;
 340 
 341   // If the target is far we'll need to load the ea of this to
 342   // a register to reach it. Otherwise if near we can do rip
 343   // relative addressing.
 344 
 345   address          _target;
 346 
 347  protected:
 348   // creation
 349   AddressLiteral()
 350     : _is_lval(false),
 351       _target(NULL)
 352   {}
 353 
 354   public:
 355 
 356 
 357   AddressLiteral(address target, relocInfo::relocType rtype);
 358 
 359   AddressLiteral(address target, RelocationHolder const& rspec)
 360     : _rspec(rspec),
 361       _is_lval(false),
 362       _target(target)
 363   {}
 364 
 365   AddressLiteral addr() {
 366     AddressLiteral ret = *this;
 367     ret._is_lval = true;
 368     return ret;
 369   }
 370 
 371 
 372  private:
 373 
 374   address target() { return _target; }
 375   bool is_lval() { return _is_lval; }
 376 
 377   relocInfo::relocType reloc() const { return _rspec.type(); }
 378   const RelocationHolder& rspec() const { return _rspec; }
 379 
 380   friend class Assembler;
 381   friend class MacroAssembler;
 382   friend class Address;
 383   friend class LIR_Assembler;
 384 };
 385 
 386 // Convience classes
 387 class RuntimeAddress: public AddressLiteral {
 388 
 389   public:
 390 
 391   RuntimeAddress(address target) : AddressLiteral(target, relocInfo::runtime_call_type) {}
 392 
 393 };
 394 
 395 class ExternalAddress: public AddressLiteral {
 396  private:
 397   static relocInfo::relocType reloc_for_target(address target) {
 398     // Sometimes ExternalAddress is used for values which aren't
 399     // exactly addresses, like the card table base.
 400     // external_word_type can't be used for values in the first page
 401     // so just skip the reloc in that case.
 402     return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none;
 403   }
 404 
 405  public:
 406 
 407   ExternalAddress(address target) : AddressLiteral(target, reloc_for_target(target)) {}
 408 
 409 };
 410 
 411 class InternalAddress: public AddressLiteral {
 412 
 413   public:
 414 
 415   InternalAddress(address target) : AddressLiteral(target, relocInfo::internal_word_type) {}
 416 
 417 };
 418 
 419 // x86 can do array addressing as a single operation since disp can be an absolute
 420 // address amd64 can't. We create a class that expresses the concept but does extra
 421 // magic on amd64 to get the final result
 422 
 423 class ArrayAddress VALUE_OBJ_CLASS_SPEC {
 424   private:
 425 
 426   AddressLiteral _base;
 427   Address        _index;
 428 
 429   public:
 430 
 431   ArrayAddress() {};
 432   ArrayAddress(AddressLiteral base, Address index): _base(base), _index(index) {};
 433   AddressLiteral base() { return _base; }
 434   Address index() { return _index; }
 435 
 436 };
 437 
 438 const int FPUStateSizeInWords = NOT_LP64(27) LP64_ONLY( 512 / wordSize);
 439 
 440 // The Intel x86/Amd64 Assembler: Pure assembler doing NO optimizations on the instruction
 441 // level (e.g. mov rax, 0 is not translated into xor rax, rax!); i.e., what you write
 442 // is what you get. The Assembler is generating code into a CodeBuffer.
 443 
 444 class Assembler : public AbstractAssembler  {
 445   friend class AbstractAssembler; // for the non-virtual hack
 446   friend class LIR_Assembler; // as_Address()
 447   friend class StubGenerator;
 448 
 449  public:
 450   enum Condition {                     // The x86 condition codes used for conditional jumps/moves.
 451     zero          = 0x4,
 452     notZero       = 0x5,
 453     equal         = 0x4,
 454     notEqual      = 0x5,
 455     less          = 0xc,
 456     lessEqual     = 0xe,
 457     greater       = 0xf,
 458     greaterEqual  = 0xd,
 459     below         = 0x2,
 460     belowEqual    = 0x6,
 461     above         = 0x7,
 462     aboveEqual    = 0x3,
 463     overflow      = 0x0,
 464     noOverflow    = 0x1,
 465     carrySet      = 0x2,
 466     carryClear    = 0x3,
 467     negative      = 0x8,
 468     positive      = 0x9,
 469     parity        = 0xa,
 470     noParity      = 0xb
 471   };
 472 
 473   enum Prefix {
 474     // segment overrides
 475     CS_segment = 0x2e,
 476     SS_segment = 0x36,
 477     DS_segment = 0x3e,
 478     ES_segment = 0x26,
 479     FS_segment = 0x64,
 480     GS_segment = 0x65,
 481 
 482     REX        = 0x40,
 483 
 484     REX_B      = 0x41,
 485     REX_X      = 0x42,
 486     REX_XB     = 0x43,
 487     REX_R      = 0x44,
 488     REX_RB     = 0x45,
 489     REX_RX     = 0x46,
 490     REX_RXB    = 0x47,
 491 
 492     REX_W      = 0x48,
 493 
 494     REX_WB     = 0x49,
 495     REX_WX     = 0x4A,
 496     REX_WXB    = 0x4B,
 497     REX_WR     = 0x4C,
 498     REX_WRB    = 0x4D,
 499     REX_WRX    = 0x4E,
 500     REX_WRXB   = 0x4F,
 501 
 502     VEX_3bytes = 0xC4,
 503     VEX_2bytes = 0xC5
 504   };
 505 
 506   enum VexPrefix {
 507     VEX_B = 0x20,
 508     VEX_X = 0x40,
 509     VEX_R = 0x80,
 510     VEX_W = 0x80
 511   };
 512 
 513   enum VexSimdPrefix {
 514     VEX_SIMD_NONE = 0x0,
 515     VEX_SIMD_66   = 0x1,
 516     VEX_SIMD_F3   = 0x2,
 517     VEX_SIMD_F2   = 0x3
 518   };
 519 
 520   enum VexOpcode {
 521     VEX_OPCODE_NONE  = 0x0,
 522     VEX_OPCODE_0F    = 0x1,
 523     VEX_OPCODE_0F_38 = 0x2,
 524     VEX_OPCODE_0F_3A = 0x3
 525   };
 526 
 527   enum WhichOperand {
 528     // input to locate_operand, and format code for relocations
 529     imm_operand  = 0,            // embedded 32-bit|64-bit immediate operand
 530     disp32_operand = 1,          // embedded 32-bit displacement or address
 531     call32_operand = 2,          // embedded 32-bit self-relative displacement
 532 #ifndef _LP64
 533     _WhichOperand_limit = 3
 534 #else
 535      narrow_oop_operand = 3,     // embedded 32-bit immediate narrow oop
 536     _WhichOperand_limit = 4
 537 #endif
 538   };
 539 
 540 
 541 
 542   // NOTE: The general philopsophy of the declarations here is that 64bit versions
 543   // of instructions are freely declared without the need for wrapping them an ifdef.
 544   // (Some dangerous instructions are ifdef's out of inappropriate jvm's.)
 545   // In the .cpp file the implementations are wrapped so that they are dropped out
 546   // of the resulting jvm. This is done mostly to keep the footprint of MINIMAL
 547   // to the size it was prior to merging up the 32bit and 64bit assemblers.
 548   //
 549   // This does mean you'll get a linker/runtime error if you use a 64bit only instruction
 550   // in a 32bit vm. This is somewhat unfortunate but keeps the ifdef noise down.
 551 
 552 private:
 553 
 554 
 555   // 64bit prefixes
 556   int prefix_and_encode(int reg_enc, bool byteinst = false);
 557   int prefixq_and_encode(int reg_enc);
 558 
 559   int prefix_and_encode(int dst_enc, int src_enc, bool byteinst = false);
 560   int prefixq_and_encode(int dst_enc, int src_enc);
 561 
 562   void prefix(Register reg);
 563   void prefix(Address adr);
 564   void prefixq(Address adr);
 565 
 566   void prefix(Address adr, Register reg,  bool byteinst = false);
 567   void prefix(Address adr, XMMRegister reg);
 568   void prefixq(Address adr, Register reg);
 569   void prefixq(Address adr, XMMRegister reg);
 570 
 571   void prefetch_prefix(Address src);
 572 
 573   void rex_prefix(Address adr, XMMRegister xreg,
 574                   VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 575   int  rex_prefix_and_encode(int dst_enc, int src_enc,
 576                              VexSimdPrefix pre, VexOpcode opc, bool rex_w);
 577 
 578   void vex_prefix(bool vex_r, bool vex_b, bool vex_x, bool vex_w,
 579                   int nds_enc, VexSimdPrefix pre, VexOpcode opc,
 580                   bool vector256);
 581 
 582   void vex_prefix(Address adr, int nds_enc, int xreg_enc,
 583                   VexSimdPrefix pre, VexOpcode opc,
 584                   bool vex_w, bool vector256);
 585 
 586   void vex_prefix(XMMRegister dst, XMMRegister nds, Address src,
 587                   VexSimdPrefix pre, bool vector256 = false) {
 588     int dst_enc = dst->encoding();
 589     int nds_enc = nds->is_valid() ? nds->encoding() : 0;
 590     vex_prefix(src, nds_enc, dst_enc, pre, VEX_OPCODE_0F, false, vector256);
 591   }
 592 
 593   void vex_prefix_0F38(Register dst, Register nds, Address src) {
 594     bool vex_w = false;
 595     bool vector256 = false;
 596     vex_prefix(src, nds->encoding(), dst->encoding(),
 597                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector256);
 598   }
 599 
 600   void vex_prefix_0F38_q(Register dst, Register nds, Address src) {
 601     bool vex_w = true;
 602     bool vector256 = false;
 603     vex_prefix(src, nds->encoding(), dst->encoding(),
 604                VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector256);
 605   }
 606   int  vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc,
 607                              VexSimdPrefix pre, VexOpcode opc,
 608                              bool vex_w, bool vector256);
 609 
 610   int  vex_prefix_0F38_and_encode(Register dst, Register nds, Register src) {
 611     bool vex_w = false;
 612     bool vector256 = false;
 613     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 614                                  VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector256);
 615   }
 616   int  vex_prefix_0F38_and_encode_q(Register dst, Register nds, Register src) {
 617     bool vex_w = true;
 618     bool vector256 = false;
 619     return vex_prefix_and_encode(dst->encoding(), nds->encoding(), src->encoding(),
 620                                  VEX_SIMD_NONE, VEX_OPCODE_0F_38, vex_w, vector256);
 621   }
 622   int  vex_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src,
 623                              VexSimdPrefix pre, bool vector256 = false,
 624                              VexOpcode opc = VEX_OPCODE_0F) {
 625     int src_enc = src->encoding();
 626     int dst_enc = dst->encoding();
 627     int nds_enc = nds->is_valid() ? nds->encoding() : 0;
 628     return vex_prefix_and_encode(dst_enc, nds_enc, src_enc, pre, opc, false, vector256);
 629   }
 630 
 631   void simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr,
 632                    VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F,
 633                    bool rex_w = false, bool vector256 = false);
 634 
 635   void simd_prefix(XMMRegister dst, Address src,
 636                    VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F) {
 637     simd_prefix(dst, xnoreg, src, pre, opc);
 638   }
 639 
 640   void simd_prefix(Address dst, XMMRegister src, VexSimdPrefix pre) {
 641     simd_prefix(src, dst, pre);
 642   }
 643   void simd_prefix_q(XMMRegister dst, XMMRegister nds, Address src,
 644                      VexSimdPrefix pre) {
 645     bool rex_w = true;
 646     simd_prefix(dst, nds, src, pre, VEX_OPCODE_0F, rex_w);
 647   }
 648 
 649   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src,
 650                              VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F,
 651                              bool rex_w = false, bool vector256 = false);
 652 
 653   // Move/convert 32-bit integer value.
 654   int simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, Register src,
 655                              VexSimdPrefix pre) {
 656     // It is OK to cast from Register to XMMRegister to pass argument here
 657     // since only encoding is used in simd_prefix_and_encode() and number of
 658     // Gen and Xmm registers are the same.
 659     return simd_prefix_and_encode(dst, nds, as_XMMRegister(src->encoding()), pre);
 660   }
 661   int simd_prefix_and_encode(XMMRegister dst, Register src, VexSimdPrefix pre) {
 662     return simd_prefix_and_encode(dst, xnoreg, src, pre);
 663   }
 664   int simd_prefix_and_encode(Register dst, XMMRegister src,
 665                              VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F) {
 666     return simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, pre, opc);
 667   }
 668 
 669   // Move/convert 64-bit integer value.
 670   int simd_prefix_and_encode_q(XMMRegister dst, XMMRegister nds, Register src,
 671                                VexSimdPrefix pre) {
 672     bool rex_w = true;
 673     return simd_prefix_and_encode(dst, nds, as_XMMRegister(src->encoding()), pre, VEX_OPCODE_0F, rex_w);
 674   }
 675   int simd_prefix_and_encode_q(XMMRegister dst, Register src, VexSimdPrefix pre) {
 676     return simd_prefix_and_encode_q(dst, xnoreg, src, pre);
 677   }
 678   int simd_prefix_and_encode_q(Register dst, XMMRegister src,
 679                              VexSimdPrefix pre, VexOpcode opc = VEX_OPCODE_0F) {
 680     bool rex_w = true;
 681     return simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, pre, opc, rex_w);
 682   }
 683 
 684   // Helper functions for groups of instructions
 685   void emit_arith_b(int op1, int op2, Register dst, int imm8);
 686 
 687   void emit_arith(int op1, int op2, Register dst, int32_t imm32);
 688   // Force generation of a 4 byte immediate value even if it fits into 8bit
 689   void emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32);
 690   void emit_arith(int op1, int op2, Register dst, Register src);
 691 
 692   void emit_simd_arith(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre);
 693   void emit_simd_arith(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre);
 694   void emit_simd_arith_nonds(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre);
 695   void emit_simd_arith_nonds(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre);
 696   void emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
 697                       Address src, VexSimdPrefix pre, bool vector256);
 698   void emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
 699                       XMMRegister src, VexSimdPrefix pre, bool vector256);
 700 
 701   void emit_operand(Register reg,
 702                     Register base, Register index, Address::ScaleFactor scale,
 703                     int disp,
 704                     RelocationHolder const& rspec,
 705                     int rip_relative_correction = 0);
 706 
 707   void emit_operand(Register reg, Address adr, int rip_relative_correction = 0);
 708 
 709   // operands that only take the original 32bit registers
 710   void emit_operand32(Register reg, Address adr);
 711 
 712   void emit_operand(XMMRegister reg,
 713                     Register base, Register index, Address::ScaleFactor scale,
 714                     int disp,
 715                     RelocationHolder const& rspec);
 716 
 717   void emit_operand(XMMRegister reg, Address adr);
 718 
 719   void emit_operand(MMXRegister reg, Address adr);
 720 
 721   // workaround gcc (3.2.1-7) bug
 722   void emit_operand(Address adr, MMXRegister reg);
 723 
 724 
 725   // Immediate-to-memory forms
 726   void emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32);
 727 
 728   void emit_farith(int b1, int b2, int i);
 729 
 730 
 731  protected:
 732   #ifdef ASSERT
 733   void check_relocation(RelocationHolder const& rspec, int format);
 734   #endif
 735 
 736   void emit_data(jint data, relocInfo::relocType    rtype, int format);
 737   void emit_data(jint data, RelocationHolder const& rspec, int format);
 738   void emit_data64(jlong data, relocInfo::relocType rtype, int format = 0);
 739   void emit_data64(jlong data, RelocationHolder const& rspec, int format = 0);
 740 
 741   bool reachable(AddressLiteral adr) NOT_LP64({ return true;});
 742 
 743   // These are all easily abused and hence protected
 744 
 745   // 32BIT ONLY SECTION
 746 #ifndef _LP64
 747   // Make these disappear in 64bit mode since they would never be correct
 748   void cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec);   // 32BIT ONLY
 749   void cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 750 
 751   void mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec);    // 32BIT ONLY
 752   void mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec);     // 32BIT ONLY
 753 
 754   void push_literal32(int32_t imm32, RelocationHolder const& rspec);                 // 32BIT ONLY
 755 #else
 756   // 64BIT ONLY SECTION
 757   void mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec);   // 64BIT ONLY
 758 
 759   void cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec);
 760   void cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec);
 761 
 762   void mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec);
 763   void mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec);
 764 #endif // _LP64
 765 
 766   // These are unique in that we are ensured by the caller that the 32bit
 767   // relative in these instructions will always be able to reach the potentially
 768   // 64bit address described by entry. Since they can take a 64bit address they
 769   // don't have the 32 suffix like the other instructions in this class.
 770 
 771   void call_literal(address entry, RelocationHolder const& rspec);
 772   void jmp_literal(address entry, RelocationHolder const& rspec);
 773 
 774   // Avoid using directly section
 775   // Instructions in this section are actually usable by anyone without danger
 776   // of failure but have performance issues that are addressed my enhanced
 777   // instructions which will do the proper thing base on the particular cpu.
 778   // We protect them because we don't trust you...
 779 
 780   // Don't use next inc() and dec() methods directly. INC & DEC instructions
 781   // could cause a partial flag stall since they don't set CF flag.
 782   // Use MacroAssembler::decrement() & MacroAssembler::increment() methods
 783   // which call inc() & dec() or add() & sub() in accordance with
 784   // the product flag UseIncDec value.
 785 
 786   void decl(Register dst);
 787   void decl(Address dst);
 788   void decq(Register dst);
 789   void decq(Address dst);
 790 
 791   void incl(Register dst);
 792   void incl(Address dst);
 793   void incq(Register dst);
 794   void incq(Address dst);
 795 
 796   // New cpus require use of movsd and movss to avoid partial register stall
 797   // when loading from memory. But for old Opteron use movlpd instead of movsd.
 798   // The selection is done in MacroAssembler::movdbl() and movflt().
 799 
 800   // Move Scalar Single-Precision Floating-Point Values
 801   void movss(XMMRegister dst, Address src);
 802   void movss(XMMRegister dst, XMMRegister src);
 803   void movss(Address dst, XMMRegister src);
 804 
 805   // Move Scalar Double-Precision Floating-Point Values
 806   void movsd(XMMRegister dst, Address src);
 807   void movsd(XMMRegister dst, XMMRegister src);
 808   void movsd(Address dst, XMMRegister src);
 809   void movlpd(XMMRegister dst, Address src);
 810 
 811   // New cpus require use of movaps and movapd to avoid partial register stall
 812   // when moving between registers.
 813   void movaps(XMMRegister dst, XMMRegister src);
 814   void movapd(XMMRegister dst, XMMRegister src);
 815 
 816   // End avoid using directly
 817 
 818 
 819   // Instruction prefixes
 820   void prefix(Prefix p);
 821 
 822   public:
 823 
 824   // Creation
 825   Assembler(CodeBuffer* code) : AbstractAssembler(code) {}
 826 
 827   // Decoding
 828   static address locate_operand(address inst, WhichOperand which);
 829   static address locate_next_instruction(address inst);
 830 
 831   // Utilities
 832   static bool is_polling_page_far() NOT_LP64({ return false;});
 833 
 834   // Generic instructions
 835   // Does 32bit or 64bit as needed for the platform. In some sense these
 836   // belong in macro assembler but there is no need for both varieties to exist
 837 
 838   void lea(Register dst, Address src);
 839 
 840   void mov(Register dst, Register src);
 841 
 842   void pusha();
 843   void popa();
 844 
 845   void pushf();
 846   void popf();
 847 
 848   void push(int32_t imm32);
 849 
 850   void push(Register src);
 851 
 852   void pop(Register dst);
 853 
 854   // These are dummies to prevent surprise implicit conversions to Register
 855   void push(void* v);
 856   void pop(void* v);
 857 
 858   // These do register sized moves/scans
 859   void rep_mov();
 860   void rep_stos();
 861   void rep_stosb();
 862   void repne_scan();
 863 #ifdef _LP64
 864   void repne_scanl();
 865 #endif
 866 
 867   // Vanilla instructions in lexical order
 868 
 869   void adcl(Address dst, int32_t imm32);
 870   void adcl(Address dst, Register src);
 871   void adcl(Register dst, int32_t imm32);
 872   void adcl(Register dst, Address src);
 873   void adcl(Register dst, Register src);
 874 
 875   void adcq(Register dst, int32_t imm32);
 876   void adcq(Register dst, Address src);
 877   void adcq(Register dst, Register src);
 878 
 879   void addl(Address dst, int32_t imm32);
 880   void addl(Address dst, Register src);
 881   void addl(Register dst, int32_t imm32);
 882   void addl(Register dst, Address src);
 883   void addl(Register dst, Register src);
 884 
 885   void addq(Address dst, int32_t imm32);
 886   void addq(Address dst, Register src);
 887   void addq(Register dst, int32_t imm32);
 888   void addq(Register dst, Address src);
 889   void addq(Register dst, Register src);
 890 
 891   void addr_nop_4();
 892   void addr_nop_5();
 893   void addr_nop_7();
 894   void addr_nop_8();
 895 
 896   // Add Scalar Double-Precision Floating-Point Values
 897   void addsd(XMMRegister dst, Address src);
 898   void addsd(XMMRegister dst, XMMRegister src);
 899 
 900   // Add Scalar Single-Precision Floating-Point Values
 901   void addss(XMMRegister dst, Address src);
 902   void addss(XMMRegister dst, XMMRegister src);
 903 
 904   // AES instructions
 905   void aesdec(XMMRegister dst, Address src);
 906   void aesdec(XMMRegister dst, XMMRegister src);
 907   void aesdeclast(XMMRegister dst, Address src);
 908   void aesdeclast(XMMRegister dst, XMMRegister src);
 909   void aesenc(XMMRegister dst, Address src);
 910   void aesenc(XMMRegister dst, XMMRegister src);
 911   void aesenclast(XMMRegister dst, Address src);
 912   void aesenclast(XMMRegister dst, XMMRegister src);
 913 
 914 
 915   void andl(Address  dst, int32_t imm32);
 916   void andl(Register dst, int32_t imm32);
 917   void andl(Register dst, Address src);
 918   void andl(Register dst, Register src);
 919 
 920   void andq(Address  dst, int32_t imm32);
 921   void andq(Register dst, int32_t imm32);
 922   void andq(Register dst, Address src);
 923   void andq(Register dst, Register src);
 924 
 925   // BMI instructions
 926   void andnl(Register dst, Register src1, Register src2);
 927   void andnl(Register dst, Register src1, Address src2);
 928   void andnq(Register dst, Register src1, Register src2);
 929   void andnq(Register dst, Register src1, Address src2);
 930 
 931   void blsil(Register dst, Register src);
 932   void blsil(Register dst, Address src);
 933   void blsiq(Register dst, Register src);
 934   void blsiq(Register dst, Address src);
 935 
 936   void blsmskl(Register dst, Register src);
 937   void blsmskl(Register dst, Address src);
 938   void blsmskq(Register dst, Register src);
 939   void blsmskq(Register dst, Address src);
 940 
 941   void blsrl(Register dst, Register src);
 942   void blsrl(Register dst, Address src);
 943   void blsrq(Register dst, Register src);
 944   void blsrq(Register dst, Address src);
 945 
 946   void bsfl(Register dst, Register src);
 947   void bsrl(Register dst, Register src);
 948 
 949 #ifdef _LP64
 950   void bsfq(Register dst, Register src);
 951   void bsrq(Register dst, Register src);
 952 #endif
 953 
 954   void bswapl(Register reg);
 955 
 956   void bswapq(Register reg);
 957 
 958   void call(Label& L, relocInfo::relocType rtype);
 959   void call(Register reg);  // push pc; pc <- reg
 960   void call(Address adr);   // push pc; pc <- adr
 961 
 962   void cdql();
 963 
 964   void cdqq();
 965 
 966   void cld();
 967 
 968   void clflush(Address adr);
 969 
 970   void cmovl(Condition cc, Register dst, Register src);
 971   void cmovl(Condition cc, Register dst, Address src);
 972 
 973   void cmovq(Condition cc, Register dst, Register src);
 974   void cmovq(Condition cc, Register dst, Address src);
 975 
 976 
 977   void cmpb(Address dst, int imm8);
 978 
 979   void cmpl(Address dst, int32_t imm32);
 980 
 981   void cmpl(Register dst, int32_t imm32);
 982   void cmpl(Register dst, Register src);
 983   void cmpl(Register dst, Address src);
 984 
 985   void cmpq(Address dst, int32_t imm32);
 986   void cmpq(Address dst, Register src);
 987 
 988   void cmpq(Register dst, int32_t imm32);
 989   void cmpq(Register dst, Register src);
 990   void cmpq(Register dst, Address src);
 991 
 992   // these are dummies used to catch attempting to convert NULL to Register
 993   void cmpl(Register dst, void* junk); // dummy
 994   void cmpq(Register dst, void* junk); // dummy
 995 
 996   void cmpw(Address dst, int imm16);
 997 
 998   void cmpxchg8 (Address adr);
 999 
1000   void cmpxchgl(Register reg, Address adr);
1001 
1002   void cmpxchgq(Register reg, Address adr);
1003 
1004   // Ordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1005   void comisd(XMMRegister dst, Address src);
1006   void comisd(XMMRegister dst, XMMRegister src);
1007 
1008   // Ordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1009   void comiss(XMMRegister dst, Address src);
1010   void comiss(XMMRegister dst, XMMRegister src);
1011 
1012   // Identify processor type and features
1013   void cpuid();
1014 
1015   // Convert Scalar Double-Precision Floating-Point Value to Scalar Single-Precision Floating-Point Value
1016   void cvtsd2ss(XMMRegister dst, XMMRegister src);
1017   void cvtsd2ss(XMMRegister dst, Address src);
1018 
1019   // Convert Doubleword Integer to Scalar Double-Precision Floating-Point Value
1020   void cvtsi2sdl(XMMRegister dst, Register src);
1021   void cvtsi2sdl(XMMRegister dst, Address src);
1022   void cvtsi2sdq(XMMRegister dst, Register src);
1023   void cvtsi2sdq(XMMRegister dst, Address src);
1024 
1025   // Convert Doubleword Integer to Scalar Single-Precision Floating-Point Value
1026   void cvtsi2ssl(XMMRegister dst, Register src);
1027   void cvtsi2ssl(XMMRegister dst, Address src);
1028   void cvtsi2ssq(XMMRegister dst, Register src);
1029   void cvtsi2ssq(XMMRegister dst, Address src);
1030 
1031   // Convert Packed Signed Doubleword Integers to Packed Double-Precision Floating-Point Value
1032   void cvtdq2pd(XMMRegister dst, XMMRegister src);
1033 
1034   // Convert Packed Signed Doubleword Integers to Packed Single-Precision Floating-Point Value
1035   void cvtdq2ps(XMMRegister dst, XMMRegister src);
1036 
1037   // Convert Scalar Single-Precision Floating-Point Value to Scalar Double-Precision Floating-Point Value
1038   void cvtss2sd(XMMRegister dst, XMMRegister src);
1039   void cvtss2sd(XMMRegister dst, Address src);
1040 
1041   // Convert with Truncation Scalar Double-Precision Floating-Point Value to Doubleword Integer
1042   void cvttsd2sil(Register dst, Address src);
1043   void cvttsd2sil(Register dst, XMMRegister src);
1044   void cvttsd2siq(Register dst, XMMRegister src);
1045 
1046   // Convert with Truncation Scalar Single-Precision Floating-Point Value to Doubleword Integer
1047   void cvttss2sil(Register dst, XMMRegister src);
1048   void cvttss2siq(Register dst, XMMRegister src);
1049 
1050   // Divide Scalar Double-Precision Floating-Point Values
1051   void divsd(XMMRegister dst, Address src);
1052   void divsd(XMMRegister dst, XMMRegister src);
1053 
1054   // Divide Scalar Single-Precision Floating-Point Values
1055   void divss(XMMRegister dst, Address src);
1056   void divss(XMMRegister dst, XMMRegister src);
1057 
1058   void emms();
1059 
1060   void fabs();
1061 
1062   void fadd(int i);
1063 
1064   void fadd_d(Address src);
1065   void fadd_s(Address src);
1066 
1067   // "Alternate" versions of x87 instructions place result down in FPU
1068   // stack instead of on TOS
1069 
1070   void fadda(int i); // "alternate" fadd
1071   void faddp(int i = 1);
1072 
1073   void fchs();
1074 
1075   void fcom(int i);
1076 
1077   void fcomp(int i = 1);
1078   void fcomp_d(Address src);
1079   void fcomp_s(Address src);
1080 
1081   void fcompp();
1082 
1083   void fcos();
1084 
1085   void fdecstp();
1086 
1087   void fdiv(int i);
1088   void fdiv_d(Address src);
1089   void fdivr_s(Address src);
1090   void fdiva(int i);  // "alternate" fdiv
1091   void fdivp(int i = 1);
1092 
1093   void fdivr(int i);
1094   void fdivr_d(Address src);
1095   void fdiv_s(Address src);
1096 
1097   void fdivra(int i); // "alternate" reversed fdiv
1098 
1099   void fdivrp(int i = 1);
1100 
1101   void ffree(int i = 0);
1102 
1103   void fild_d(Address adr);
1104   void fild_s(Address adr);
1105 
1106   void fincstp();
1107 
1108   void finit();
1109 
1110   void fist_s (Address adr);
1111   void fistp_d(Address adr);
1112   void fistp_s(Address adr);
1113 
1114   void fld1();
1115 
1116   void fld_d(Address adr);
1117   void fld_s(Address adr);
1118   void fld_s(int index);
1119   void fld_x(Address adr);  // extended-precision (80-bit) format
1120 
1121   void fldcw(Address src);
1122 
1123   void fldenv(Address src);
1124 
1125   void fldlg2();
1126 
1127   void fldln2();
1128 
1129   void fldz();
1130 
1131   void flog();
1132   void flog10();
1133 
1134   void fmul(int i);
1135 
1136   void fmul_d(Address src);
1137   void fmul_s(Address src);
1138 
1139   void fmula(int i);  // "alternate" fmul
1140 
1141   void fmulp(int i = 1);
1142 
1143   void fnsave(Address dst);
1144 
1145   void fnstcw(Address src);
1146 
1147   void fnstsw_ax();
1148 
1149   void fprem();
1150   void fprem1();
1151 
1152   void frstor(Address src);
1153 
1154   void fsin();
1155 
1156   void fsqrt();
1157 
1158   void fst_d(Address adr);
1159   void fst_s(Address adr);
1160 
1161   void fstp_d(Address adr);
1162   void fstp_d(int index);
1163   void fstp_s(Address adr);
1164   void fstp_x(Address adr); // extended-precision (80-bit) format
1165 
1166   void fsub(int i);
1167   void fsub_d(Address src);
1168   void fsub_s(Address src);
1169 
1170   void fsuba(int i);  // "alternate" fsub
1171 
1172   void fsubp(int i = 1);
1173 
1174   void fsubr(int i);
1175   void fsubr_d(Address src);
1176   void fsubr_s(Address src);
1177 
1178   void fsubra(int i); // "alternate" reversed fsub
1179 
1180   void fsubrp(int i = 1);
1181 
1182   void ftan();
1183 
1184   void ftst();
1185 
1186   void fucomi(int i = 1);
1187   void fucomip(int i = 1);
1188 
1189   void fwait();
1190 
1191   void fxch(int i = 1);
1192 
1193   void fxrstor(Address src);
1194 
1195   void fxsave(Address dst);
1196 
1197   void fyl2x();
1198   void frndint();
1199   void f2xm1();
1200   void fldl2e();
1201 
1202   void hlt();
1203 
1204   void idivl(Register src);
1205   void divl(Register src); // Unsigned division
1206 
1207   void idivq(Register src);
1208 
1209   void imull(Register dst, Register src);
1210   void imull(Register dst, Register src, int value);
1211   void imull(Register dst, Address src);
1212 
1213   void imulq(Register dst, Register src);
1214   void imulq(Register dst, Register src, int value);
1215 #ifdef _LP64
1216   void imulq(Register dst, Address src);
1217 #endif
1218 
1219 
1220   // jcc is the generic conditional branch generator to run-
1221   // time routines, jcc is used for branches to labels. jcc
1222   // takes a branch opcode (cc) and a label (L) and generates
1223   // either a backward branch or a forward branch and links it
1224   // to the label fixup chain. Usage:
1225   //
1226   // Label L;      // unbound label
1227   // jcc(cc, L);   // forward branch to unbound label
1228   // bind(L);      // bind label to the current pc
1229   // jcc(cc, L);   // backward branch to bound label
1230   // bind(L);      // illegal: a label may be bound only once
1231   //
1232   // Note: The same Label can be used for forward and backward branches
1233   // but it may be bound only once.
1234 
1235   void jcc(Condition cc, Label& L, bool maybe_short = true);
1236 
1237   // Conditional jump to a 8-bit offset to L.
1238   // WARNING: be very careful using this for forward jumps.  If the label is
1239   // not bound within an 8-bit offset of this instruction, a run-time error
1240   // will occur.
1241   void jccb(Condition cc, Label& L);
1242 
1243   void jmp(Address entry);    // pc <- entry
1244 
1245   // Label operations & relative jumps (PPUM Appendix D)
1246   void jmp(Label& L, bool maybe_short = true);   // unconditional jump to L
1247 
1248   void jmp(Register entry); // pc <- entry
1249 
1250   // Unconditional 8-bit offset jump to L.
1251   // WARNING: be very careful using this for forward jumps.  If the label is
1252   // not bound within an 8-bit offset of this instruction, a run-time error
1253   // will occur.
1254   void jmpb(Label& L);
1255 
1256   void ldmxcsr( Address src );
1257 
1258   void leal(Register dst, Address src);
1259 
1260   void leaq(Register dst, Address src);
1261 
1262   void lfence();
1263 
1264   void lock();
1265 
1266   void lzcntl(Register dst, Register src);
1267 
1268 #ifdef _LP64
1269   void lzcntq(Register dst, Register src);
1270 #endif
1271 
1272   enum Membar_mask_bits {
1273     StoreStore = 1 << 3,
1274     LoadStore  = 1 << 2,
1275     StoreLoad  = 1 << 1,
1276     LoadLoad   = 1 << 0
1277   };
1278 
1279   // Serializes memory and blows flags
1280   void membar(Membar_mask_bits order_constraint) {
1281     if (os::is_MP()) {
1282       // We only have to handle StoreLoad
1283       if (order_constraint & StoreLoad) {
1284         // All usable chips support "locked" instructions which suffice
1285         // as barriers, and are much faster than the alternative of
1286         // using cpuid instruction. We use here a locked add [esp],0.
1287         // This is conveniently otherwise a no-op except for blowing
1288         // flags.
1289         // Any change to this code may need to revisit other places in
1290         // the code where this idiom is used, in particular the
1291         // orderAccess code.
1292         lock();
1293         addl(Address(rsp, 0), 0);// Assert the lock# signal here
1294       }
1295     }
1296   }
1297 
1298   void mfence();
1299 
1300   // Moves
1301 
1302   void mov64(Register dst, int64_t imm64);
1303 
1304   void movb(Address dst, Register src);
1305   void movb(Address dst, int imm8);
1306   void movb(Register dst, Address src);
1307 
1308   void movdl(XMMRegister dst, Register src);
1309   void movdl(Register dst, XMMRegister src);
1310   void movdl(XMMRegister dst, Address src);
1311   void movdl(Address dst, XMMRegister src);
1312 
1313   // Move Double Quadword
1314   void movdq(XMMRegister dst, Register src);
1315   void movdq(Register dst, XMMRegister src);
1316 
1317   // Move Aligned Double Quadword
1318   void movdqa(XMMRegister dst, XMMRegister src);
1319   void movdqa(XMMRegister dst, Address src);
1320 
1321   // Move Unaligned Double Quadword
1322   void movdqu(Address     dst, XMMRegister src);
1323   void movdqu(XMMRegister dst, Address src);
1324   void movdqu(XMMRegister dst, XMMRegister src);
1325 
1326   // Move Unaligned 256bit Vector
1327   void vmovdqu(Address dst, XMMRegister src);
1328   void vmovdqu(XMMRegister dst, Address src);
1329   void vmovdqu(XMMRegister dst, XMMRegister src);
1330 
1331   // Move lower 64bit to high 64bit in 128bit register
1332   void movlhps(XMMRegister dst, XMMRegister src);
1333 
1334   void movl(Register dst, int32_t imm32);
1335   void movl(Address dst, int32_t imm32);
1336   void movl(Register dst, Register src);
1337   void movl(Register dst, Address src);
1338   void movl(Address dst, Register src);
1339 
1340   // These dummies prevent using movl from converting a zero (like NULL) into Register
1341   // by giving the compiler two choices it can't resolve
1342 
1343   void movl(Address  dst, void* junk);
1344   void movl(Register dst, void* junk);
1345 
1346 #ifdef _LP64
1347   void movq(Register dst, Register src);
1348   void movq(Register dst, Address src);
1349   void movq(Address  dst, Register src);
1350 #endif
1351 
1352   void movq(Address     dst, MMXRegister src );
1353   void movq(MMXRegister dst, Address src );
1354 
1355 #ifdef _LP64
1356   // These dummies prevent using movq from converting a zero (like NULL) into Register
1357   // by giving the compiler two choices it can't resolve
1358 
1359   void movq(Address  dst, void* dummy);
1360   void movq(Register dst, void* dummy);
1361 #endif
1362 
1363   // Move Quadword
1364   void movq(Address     dst, XMMRegister src);
1365   void movq(XMMRegister dst, Address src);
1366 
1367   void movsbl(Register dst, Address src);
1368   void movsbl(Register dst, Register src);
1369 
1370 #ifdef _LP64
1371   void movsbq(Register dst, Address src);
1372   void movsbq(Register dst, Register src);
1373 
1374   // Move signed 32bit immediate to 64bit extending sign
1375   void movslq(Address  dst, int32_t imm64);
1376   void movslq(Register dst, int32_t imm64);
1377 
1378   void movslq(Register dst, Address src);
1379   void movslq(Register dst, Register src);
1380   void movslq(Register dst, void* src); // Dummy declaration to cause NULL to be ambiguous
1381 #endif
1382 
1383   void movswl(Register dst, Address src);
1384   void movswl(Register dst, Register src);
1385 
1386 #ifdef _LP64
1387   void movswq(Register dst, Address src);
1388   void movswq(Register dst, Register src);
1389 #endif
1390 
1391   void movw(Address dst, int imm16);
1392   void movw(Register dst, Address src);
1393   void movw(Address dst, Register src);
1394 
1395   void movzbl(Register dst, Address src);
1396   void movzbl(Register dst, Register src);
1397 
1398 #ifdef _LP64
1399   void movzbq(Register dst, Address src);
1400   void movzbq(Register dst, Register src);
1401 #endif
1402 
1403   void movzwl(Register dst, Address src);
1404   void movzwl(Register dst, Register src);
1405 
1406 #ifdef _LP64
1407   void movzwq(Register dst, Address src);
1408   void movzwq(Register dst, Register src);
1409 #endif
1410 
1411   void mull(Address src);
1412   void mull(Register src);
1413 
1414   // Multiply Scalar Double-Precision Floating-Point Values
1415   void mulsd(XMMRegister dst, Address src);
1416   void mulsd(XMMRegister dst, XMMRegister src);
1417 
1418   // Multiply Scalar Single-Precision Floating-Point Values
1419   void mulss(XMMRegister dst, Address src);
1420   void mulss(XMMRegister dst, XMMRegister src);
1421 
1422   void negl(Register dst);
1423 
1424 #ifdef _LP64
1425   void negq(Register dst);
1426 #endif
1427 
1428   void nop(int i = 1);
1429 
1430   void notl(Register dst);
1431 
1432 #ifdef _LP64
1433   void notq(Register dst);
1434 #endif
1435 
1436   void orl(Address dst, int32_t imm32);
1437   void orl(Register dst, int32_t imm32);
1438   void orl(Register dst, Address src);
1439   void orl(Register dst, Register src);
1440 
1441   void orq(Address dst, int32_t imm32);
1442   void orq(Register dst, int32_t imm32);
1443   void orq(Register dst, Address src);
1444   void orq(Register dst, Register src);
1445 
1446   // Pack with unsigned saturation
1447   void packuswb(XMMRegister dst, XMMRegister src);
1448   void packuswb(XMMRegister dst, Address src);
1449   void vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1450 
1451   // Pemutation of 64bit words
1452   void vpermq(XMMRegister dst, XMMRegister src, int imm8, bool vector256);
1453 
1454   void pause();
1455 
1456   // SSE4.2 string instructions
1457   void pcmpestri(XMMRegister xmm1, XMMRegister xmm2, int imm8);
1458   void pcmpestri(XMMRegister xmm1, Address src, int imm8);
1459 
1460   // SSE 4.1 extract
1461   void pextrd(Register dst, XMMRegister src, int imm8);
1462   void pextrq(Register dst, XMMRegister src, int imm8);
1463 
1464   // SSE 4.1 insert
1465   void pinsrd(XMMRegister dst, Register src, int imm8);
1466   void pinsrq(XMMRegister dst, Register src, int imm8);
1467 
1468   // SSE4.1 packed move
1469   void pmovzxbw(XMMRegister dst, XMMRegister src);
1470   void pmovzxbw(XMMRegister dst, Address src);
1471 
1472 #ifndef _LP64 // no 32bit push/pop on amd64
1473   void popl(Address dst);
1474 #endif
1475 
1476 #ifdef _LP64
1477   void popq(Address dst);
1478 #endif
1479 
1480   void popcntl(Register dst, Address src);
1481   void popcntl(Register dst, Register src);
1482 
1483 #ifdef _LP64
1484   void popcntq(Register dst, Address src);
1485   void popcntq(Register dst, Register src);
1486 #endif
1487 
1488   // Prefetches (SSE, SSE2, 3DNOW only)
1489 
1490   void prefetchnta(Address src);
1491   void prefetchr(Address src);
1492   void prefetcht0(Address src);
1493   void prefetcht1(Address src);
1494   void prefetcht2(Address src);
1495   void prefetchw(Address src);
1496 
1497   // Shuffle Bytes
1498   void pshufb(XMMRegister dst, XMMRegister src);
1499   void pshufb(XMMRegister dst, Address src);
1500 
1501   // Shuffle Packed Doublewords
1502   void pshufd(XMMRegister dst, XMMRegister src, int mode);
1503   void pshufd(XMMRegister dst, Address src,     int mode);
1504 
1505   // Shuffle Packed Low Words
1506   void pshuflw(XMMRegister dst, XMMRegister src, int mode);
1507   void pshuflw(XMMRegister dst, Address src,     int mode);
1508 
1509   // Shift Right by bytes Logical DoubleQuadword Immediate
1510   void psrldq(XMMRegister dst, int shift);
1511 
1512   // Logical Compare 128bit
1513   void ptest(XMMRegister dst, XMMRegister src);
1514   void ptest(XMMRegister dst, Address src);
1515   // Logical Compare 256bit
1516   void vptest(XMMRegister dst, XMMRegister src);
1517   void vptest(XMMRegister dst, Address src);
1518 
1519   // Interleave Low Bytes
1520   void punpcklbw(XMMRegister dst, XMMRegister src);
1521   void punpcklbw(XMMRegister dst, Address src);
1522 
1523   // Interleave Low Doublewords
1524   void punpckldq(XMMRegister dst, XMMRegister src);
1525   void punpckldq(XMMRegister dst, Address src);
1526 
1527   // Interleave Low Quadwords
1528   void punpcklqdq(XMMRegister dst, XMMRegister src);
1529 
1530 #ifndef _LP64 // no 32bit push/pop on amd64
1531   void pushl(Address src);
1532 #endif
1533 
1534   void pushq(Address src);
1535 
1536   void rcll(Register dst, int imm8);
1537 
1538   void rclq(Register dst, int imm8);
1539 
1540   void rdtsc();
1541 
1542   void ret(int imm16);
1543 
1544   void sahf();
1545 
1546   void sarl(Register dst, int imm8);
1547   void sarl(Register dst);
1548 
1549   void sarq(Register dst, int imm8);
1550   void sarq(Register dst);
1551 
1552   void sbbl(Address dst, int32_t imm32);
1553   void sbbl(Register dst, int32_t imm32);
1554   void sbbl(Register dst, Address src);
1555   void sbbl(Register dst, Register src);
1556 
1557   void sbbq(Address dst, int32_t imm32);
1558   void sbbq(Register dst, int32_t imm32);
1559   void sbbq(Register dst, Address src);
1560   void sbbq(Register dst, Register src);
1561 
1562   void setb(Condition cc, Register dst);
1563 
1564   void shldl(Register dst, Register src);
1565 
1566   void shll(Register dst, int imm8);
1567   void shll(Register dst);
1568 
1569   void shlq(Register dst, int imm8);
1570   void shlq(Register dst);
1571 
1572   void shrdl(Register dst, Register src);
1573 
1574   void shrl(Register dst, int imm8);
1575   void shrl(Register dst);
1576 
1577   void shrq(Register dst, int imm8);
1578   void shrq(Register dst);
1579 
1580   void smovl(); // QQQ generic?
1581 
1582   // Compute Square Root of Scalar Double-Precision Floating-Point Value
1583   void sqrtsd(XMMRegister dst, Address src);
1584   void sqrtsd(XMMRegister dst, XMMRegister src);
1585 
1586   // Compute Square Root of Scalar Single-Precision Floating-Point Value
1587   void sqrtss(XMMRegister dst, Address src);
1588   void sqrtss(XMMRegister dst, XMMRegister src);
1589 
1590   void std();
1591 
1592   void stmxcsr( Address dst );
1593 
1594   void subl(Address dst, int32_t imm32);
1595   void subl(Address dst, Register src);
1596   void subl(Register dst, int32_t imm32);
1597   void subl(Register dst, Address src);
1598   void subl(Register dst, Register src);
1599 
1600   void subq(Address dst, int32_t imm32);
1601   void subq(Address dst, Register src);
1602   void subq(Register dst, int32_t imm32);
1603   void subq(Register dst, Address src);
1604   void subq(Register dst, Register src);
1605 
1606   // Force generation of a 4 byte immediate value even if it fits into 8bit
1607   void subl_imm32(Register dst, int32_t imm32);
1608   void subq_imm32(Register dst, int32_t imm32);
1609 
1610   // Subtract Scalar Double-Precision Floating-Point Values
1611   void subsd(XMMRegister dst, Address src);
1612   void subsd(XMMRegister dst, XMMRegister src);
1613 
1614   // Subtract Scalar Single-Precision Floating-Point Values
1615   void subss(XMMRegister dst, Address src);
1616   void subss(XMMRegister dst, XMMRegister src);
1617 
1618   void testb(Register dst, int imm8);
1619 
1620   void testl(Register dst, int32_t imm32);
1621   void testl(Register dst, Register src);
1622   void testl(Register dst, Address src);
1623 
1624   void testq(Register dst, int32_t imm32);
1625   void testq(Register dst, Register src);
1626 
1627   // BMI - count trailing zeros
1628   void tzcntl(Register dst, Register src);
1629   void tzcntq(Register dst, Register src);
1630 
1631   // Unordered Compare Scalar Double-Precision Floating-Point Values and set EFLAGS
1632   void ucomisd(XMMRegister dst, Address src);
1633   void ucomisd(XMMRegister dst, XMMRegister src);
1634 
1635   // Unordered Compare Scalar Single-Precision Floating-Point Values and set EFLAGS
1636   void ucomiss(XMMRegister dst, Address src);
1637   void ucomiss(XMMRegister dst, XMMRegister src);
1638 
1639   void xabort(int8_t imm8);
1640 
1641   void xaddl(Address dst, Register src);
1642 
1643   void xaddq(Address dst, Register src);
1644 
1645   void xbegin(Label& abort, relocInfo::relocType rtype = relocInfo::none);
1646 
1647   void xchgl(Register reg, Address adr);
1648   void xchgl(Register dst, Register src);
1649 
1650   void xchgq(Register reg, Address adr);
1651   void xchgq(Register dst, Register src);
1652 
1653   void xend();
1654 
1655   // Get Value of Extended Control Register
1656   void xgetbv();
1657 
1658   void xorl(Register dst, int32_t imm32);
1659   void xorl(Register dst, Address src);
1660   void xorl(Register dst, Register src);
1661 
1662   void xorq(Register dst, Address src);
1663   void xorq(Register dst, Register src);
1664 
1665   void set_byte_if_not_zero(Register dst); // sets reg to 1 if not zero, otherwise 0
1666 
1667   // AVX 3-operands scalar instructions (encoded with VEX prefix)
1668 
1669   void vaddsd(XMMRegister dst, XMMRegister nds, Address src);
1670   void vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1671   void vaddss(XMMRegister dst, XMMRegister nds, Address src);
1672   void vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1673   void vdivsd(XMMRegister dst, XMMRegister nds, Address src);
1674   void vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1675   void vdivss(XMMRegister dst, XMMRegister nds, Address src);
1676   void vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1677   void vmulsd(XMMRegister dst, XMMRegister nds, Address src);
1678   void vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1679   void vmulss(XMMRegister dst, XMMRegister nds, Address src);
1680   void vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1681   void vsubsd(XMMRegister dst, XMMRegister nds, Address src);
1682   void vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src);
1683   void vsubss(XMMRegister dst, XMMRegister nds, Address src);
1684   void vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src);
1685 
1686 
1687   //====================VECTOR ARITHMETIC=====================================
1688 
1689   // Add Packed Floating-Point Values
1690   void addpd(XMMRegister dst, XMMRegister src);
1691   void addps(XMMRegister dst, XMMRegister src);
1692   void vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1693   void vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1694   void vaddpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1695   void vaddps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1696 
1697   // Subtract Packed Floating-Point Values
1698   void subpd(XMMRegister dst, XMMRegister src);
1699   void subps(XMMRegister dst, XMMRegister src);
1700   void vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1701   void vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1702   void vsubpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1703   void vsubps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1704 
1705   // Multiply Packed Floating-Point Values
1706   void mulpd(XMMRegister dst, XMMRegister src);
1707   void mulps(XMMRegister dst, XMMRegister src);
1708   void vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1709   void vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1710   void vmulpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1711   void vmulps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1712 
1713   // Divide Packed Floating-Point Values
1714   void divpd(XMMRegister dst, XMMRegister src);
1715   void divps(XMMRegister dst, XMMRegister src);
1716   void vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1717   void vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1718   void vdivpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1719   void vdivps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1720 
1721   // Bitwise Logical AND of Packed Floating-Point Values
1722   void andpd(XMMRegister dst, XMMRegister src);
1723   void andps(XMMRegister dst, XMMRegister src);
1724   void vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1725   void vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1726   void vandpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1727   void vandps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1728 
1729   // Bitwise Logical XOR of Packed Floating-Point Values
1730   void xorpd(XMMRegister dst, XMMRegister src);
1731   void xorps(XMMRegister dst, XMMRegister src);
1732   void vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1733   void vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1734   void vxorpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1735   void vxorps(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1736 
1737   // Add packed integers
1738   void paddb(XMMRegister dst, XMMRegister src);
1739   void paddw(XMMRegister dst, XMMRegister src);
1740   void paddd(XMMRegister dst, XMMRegister src);
1741   void paddq(XMMRegister dst, XMMRegister src);
1742   void vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1743   void vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1744   void vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1745   void vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1746   void vpaddb(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1747   void vpaddw(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1748   void vpaddd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1749   void vpaddq(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1750 
1751   // Sub packed integers
1752   void psubb(XMMRegister dst, XMMRegister src);
1753   void psubw(XMMRegister dst, XMMRegister src);
1754   void psubd(XMMRegister dst, XMMRegister src);
1755   void psubq(XMMRegister dst, XMMRegister src);
1756   void vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1757   void vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1758   void vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1759   void vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1760   void vpsubb(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1761   void vpsubw(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1762   void vpsubd(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1763   void vpsubq(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1764 
1765   // Multiply packed integers (only shorts and ints)
1766   void pmullw(XMMRegister dst, XMMRegister src);
1767   void pmulld(XMMRegister dst, XMMRegister src);
1768   void vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1769   void vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1770   void vpmullw(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1771   void vpmulld(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1772 
1773   // Shift left packed integers
1774   void psllw(XMMRegister dst, int shift);
1775   void pslld(XMMRegister dst, int shift);
1776   void psllq(XMMRegister dst, int shift);
1777   void psllw(XMMRegister dst, XMMRegister shift);
1778   void pslld(XMMRegister dst, XMMRegister shift);
1779   void psllq(XMMRegister dst, XMMRegister shift);
1780   void vpsllw(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1781   void vpslld(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1782   void vpsllq(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1783   void vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1784   void vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1785   void vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1786 
1787   // Logical shift right packed integers
1788   void psrlw(XMMRegister dst, int shift);
1789   void psrld(XMMRegister dst, int shift);
1790   void psrlq(XMMRegister dst, int shift);
1791   void psrlw(XMMRegister dst, XMMRegister shift);
1792   void psrld(XMMRegister dst, XMMRegister shift);
1793   void psrlq(XMMRegister dst, XMMRegister shift);
1794   void vpsrlw(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1795   void vpsrld(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1796   void vpsrlq(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1797   void vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1798   void vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1799   void vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1800 
1801   // Arithmetic shift right packed integers (only shorts and ints, no instructions for longs)
1802   void psraw(XMMRegister dst, int shift);
1803   void psrad(XMMRegister dst, int shift);
1804   void psraw(XMMRegister dst, XMMRegister shift);
1805   void psrad(XMMRegister dst, XMMRegister shift);
1806   void vpsraw(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1807   void vpsrad(XMMRegister dst, XMMRegister src, int shift, bool vector256);
1808   void vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1809   void vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256);
1810 
1811   // And packed integers
1812   void pand(XMMRegister dst, XMMRegister src);
1813   void vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1814   void vpand(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1815 
1816   // Or packed integers
1817   void por(XMMRegister dst, XMMRegister src);
1818   void vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1819   void vpor(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1820 
1821   // Xor packed integers
1822   void pxor(XMMRegister dst, XMMRegister src);
1823   void vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256);
1824   void vpxor(XMMRegister dst, XMMRegister nds, Address src, bool vector256);
1825 
1826   // Copy low 128bit into high 128bit of YMM registers.
1827   void vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
1828   void vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src);
1829 
1830   // Load/store high 128bit of YMM registers which does not destroy other half.
1831   void vinsertf128h(XMMRegister dst, Address src);
1832   void vinserti128h(XMMRegister dst, Address src);
1833   void vextractf128h(Address dst, XMMRegister src);
1834   void vextracti128h(Address dst, XMMRegister src);
1835 
1836   // duplicate 4-bytes integer data from src into 8 locations in dest
1837   void vpbroadcastd(XMMRegister dst, XMMRegister src);
1838 
1839   // Carry-Less Multiplication Quadword
1840   void pclmulqdq(XMMRegister dst, XMMRegister src, int mask);
1841   void vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask);
1842 
1843   // AVX instruction which is used to clear upper 128 bits of YMM registers and
1844   // to avoid transaction penalty between AVX and SSE states. There is no
1845   // penalty if legacy SSE instructions are encoded using VEX prefix because
1846   // they always clear upper 128 bits. It should be used before calling
1847   // runtime code and native libraries.
1848   void vzeroupper();
1849 
1850  protected:
1851   // Next instructions require address alignment 16 bytes SSE mode.
1852   // They should be called only from corresponding MacroAssembler instructions.
1853   void andpd(XMMRegister dst, Address src);
1854   void andps(XMMRegister dst, Address src);
1855   void xorpd(XMMRegister dst, Address src);
1856   void xorps(XMMRegister dst, Address src);
1857 
1858 };
1859 
1860 #endif // CPU_X86_VM_ASSEMBLER_X86_HPP