1 /*
   2  * Copyright (c) 2003, 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 #include "precompiled.hpp"
  26 #include "asm/macroAssembler.hpp"
  27 #include "interpreter/interpreter.hpp"
  28 #include "interpreter/interpreterRuntime.hpp"
  29 #include "interpreter/templateTable.hpp"
  30 #include "memory/universe.inline.hpp"
  31 #include "oops/methodData.hpp"
  32 #include "oops/objArrayKlass.hpp"
  33 #include "oops/oop.inline.hpp"
  34 #include "prims/methodHandles.hpp"
  35 #include "runtime/sharedRuntime.hpp"
  36 #include "runtime/stubRoutines.hpp"
  37 #include "runtime/synchronizer.hpp"
  38 #include "utilities/macros.hpp"
  39 
  40 #ifndef CC_INTERP
  41 
  42 #define __ _masm->
  43 
  44 // Platform-dependent initialization
  45 
  46 void TemplateTable::pd_initialize() {
  47   // No amd64 specific initialization
  48 }
  49 
  50 // Address computation: local variables
  51 
  52 static inline Address iaddress(int n) {
  53   return Address(r14, Interpreter::local_offset_in_bytes(n));
  54 }
  55 
  56 static inline Address laddress(int n) {
  57   return iaddress(n + 1);
  58 }
  59 
  60 static inline Address faddress(int n) {
  61   return iaddress(n);
  62 }
  63 
  64 static inline Address daddress(int n) {
  65   return laddress(n);
  66 }
  67 
  68 static inline Address aaddress(int n) {
  69   return iaddress(n);
  70 }
  71 
  72 static inline Address iaddress(Register r) {
  73   return Address(r14, r, Address::times_8);
  74 }
  75 
  76 static inline Address laddress(Register r) {
  77   return Address(r14, r, Address::times_8, Interpreter::local_offset_in_bytes(1));
  78 }
  79 
  80 static inline Address faddress(Register r) {
  81   return iaddress(r);
  82 }
  83 
  84 static inline Address daddress(Register r) {
  85   return laddress(r);
  86 }
  87 
  88 static inline Address aaddress(Register r) {
  89   return iaddress(r);
  90 }
  91 
  92 static inline Address at_rsp() {
  93   return Address(rsp, 0);
  94 }
  95 
  96 // At top of Java expression stack which may be different than esp().  It
  97 // isn't for category 1 objects.
  98 static inline Address at_tos   () {
  99   return Address(rsp,  Interpreter::expr_offset_in_bytes(0));
 100 }
 101 
 102 static inline Address at_tos_p1() {
 103   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
 104 }
 105 
 106 static inline Address at_tos_p2() {
 107   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
 108 }
 109 
 110 static inline Address at_tos_p3() {
 111   return Address(rsp,  Interpreter::expr_offset_in_bytes(3));
 112 }
 113 
 114 // Condition conversion
 115 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 116   switch (cc) {
 117   case TemplateTable::equal        : return Assembler::notEqual;
 118   case TemplateTable::not_equal    : return Assembler::equal;
 119   case TemplateTable::less         : return Assembler::greaterEqual;
 120   case TemplateTable::less_equal   : return Assembler::greater;
 121   case TemplateTable::greater      : return Assembler::lessEqual;
 122   case TemplateTable::greater_equal: return Assembler::less;
 123   }
 124   ShouldNotReachHere();
 125   return Assembler::zero;
 126 }
 127 
 128 
 129 // Miscelaneous helper routines
 130 // Store an oop (or NULL) at the address described by obj.
 131 // If val == noreg this means store a NULL
 132 
 133 static void do_oop_store(InterpreterMacroAssembler* _masm,
 134                          Address obj,
 135                          Register val,
 136                          BarrierSet::Name barrier,
 137                          bool precise) {
 138   assert(val == noreg || val == rax, "parameter is just for looks");
 139   switch (barrier) {
 140 #if INCLUDE_ALL_GCS
 141     case BarrierSet::G1SATBCT:
 142     case BarrierSet::G1SATBCTLogging:
 143       {
 144         // flatten object address if needed
 145         if (obj.index() == noreg && obj.disp() == 0) {
 146           if (obj.base() != rdx) {
 147             __ movq(rdx, obj.base());
 148           }
 149         } else {
 150           __ leaq(rdx, obj);
 151         }
 152         __ g1_write_barrier_pre(rdx /* obj */,
 153                                 rbx /* pre_val */,
 154                                 r15_thread /* thread */,
 155                                 r8  /* tmp */,
 156                                 val != noreg /* tosca_live */,
 157                                 false /* expand_call */);
 158         if (val == noreg) {
 159           __ store_heap_oop_null(Address(rdx, 0));
 160         } else {
 161           // G1 barrier needs uncompressed oop for region cross check.
 162           Register new_val = val;
 163           if (UseCompressedOops) {
 164             new_val = rbx;
 165             __ movptr(new_val, val);
 166           }
 167           __ store_heap_oop(Address(rdx, 0), val);
 168           __ g1_write_barrier_post(rdx /* store_adr */,
 169                                    new_val /* new_val */,
 170                                    r15_thread /* thread */,
 171                                    r8 /* tmp */,
 172                                    rbx /* tmp2 */);
 173         }
 174       }
 175       break;
 176 #endif // INCLUDE_ALL_GCS
 177     case BarrierSet::CardTableModRef:
 178     case BarrierSet::CardTableExtension:
 179       {
 180         if (val == noreg) {
 181           __ store_heap_oop_null(obj);
 182         } else {
 183           __ store_heap_oop(obj, val);
 184           // flatten object address if needed
 185           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 186             __ store_check(obj.base());
 187           } else {
 188             __ leaq(rdx, obj);
 189             __ store_check(rdx);
 190           }
 191         }
 192       }
 193       break;
 194     case BarrierSet::ModRef:
 195     case BarrierSet::Other:
 196       if (val == noreg) {
 197         __ store_heap_oop_null(obj);
 198       } else {
 199         __ store_heap_oop(obj, val);
 200       }
 201       break;
 202     default      :
 203       ShouldNotReachHere();
 204 
 205   }
 206 }
 207 
 208 Address TemplateTable::at_bcp(int offset) {
 209   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 210   return Address(r13, offset);
 211 }
 212 
 213 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register bc_reg,
 214                                    Register temp_reg, bool load_bc_into_bc_reg/*=true*/,
 215                                    int byte_no) {
 216   if (!RewriteBytecodes)  return;
 217   Label L_patch_done;
 218 
 219   switch (bc) {
 220   case Bytecodes::_fast_aputfield:
 221   case Bytecodes::_fast_bputfield:
 222   case Bytecodes::_fast_cputfield:
 223   case Bytecodes::_fast_dputfield:
 224   case Bytecodes::_fast_fputfield:
 225   case Bytecodes::_fast_iputfield:
 226   case Bytecodes::_fast_lputfield:
 227   case Bytecodes::_fast_sputfield:
 228     {
 229       // We skip bytecode quickening for putfield instructions when
 230       // the put_code written to the constant pool cache is zero.
 231       // This is required so that every execution of this instruction
 232       // calls out to InterpreterRuntime::resolve_get_put to do
 233       // additional, required work.
 234       assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
 235       assert(load_bc_into_bc_reg, "we use bc_reg as temp");
 236       __ get_cache_and_index_and_bytecode_at_bcp(temp_reg, bc_reg, temp_reg, byte_no, 1);
 237       __ movl(bc_reg, bc);
 238       __ cmpl(temp_reg, (int) 0);
 239       __ jcc(Assembler::zero, L_patch_done);  // don't patch
 240     }
 241     break;
 242   default:
 243     assert(byte_no == -1, "sanity");
 244     // the pair bytecodes have already done the load.
 245     if (load_bc_into_bc_reg) {
 246       __ movl(bc_reg, bc);
 247     }
 248   }
 249 
 250   if (JvmtiExport::can_post_breakpoint()) {
 251     Label L_fast_patch;
 252     // if a breakpoint is present we can't rewrite the stream directly
 253     __ movzbl(temp_reg, at_bcp(0));
 254     __ cmpl(temp_reg, Bytecodes::_breakpoint);
 255     __ jcc(Assembler::notEqual, L_fast_patch);
 256     __ get_method(temp_reg);
 257     // Let breakpoint table handling rewrite to quicker bytecode
 258     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), temp_reg, r13, bc_reg);
 259 #ifndef ASSERT
 260     __ jmpb(L_patch_done);
 261 #else
 262     __ jmp(L_patch_done);
 263 #endif
 264     __ bind(L_fast_patch);
 265   }
 266 
 267 #ifdef ASSERT
 268   Label L_okay;
 269   __ load_unsigned_byte(temp_reg, at_bcp(0));
 270   __ cmpl(temp_reg, (int) Bytecodes::java_code(bc));
 271   __ jcc(Assembler::equal, L_okay);
 272   __ cmpl(temp_reg, bc_reg);
 273   __ jcc(Assembler::equal, L_okay);
 274   __ stop("patching the wrong bytecode");
 275   __ bind(L_okay);
 276 #endif
 277 
 278   // patch bytecode
 279   __ movb(at_bcp(0), bc_reg);
 280   __ bind(L_patch_done);
 281 }
 282 
 283 
 284 // Individual instructions
 285 
 286 void TemplateTable::nop() {
 287   transition(vtos, vtos);
 288   // nothing to do
 289 }
 290 
 291 void TemplateTable::shouldnotreachhere() {
 292   transition(vtos, vtos);
 293   __ stop("shouldnotreachhere bytecode");
 294 }
 295 
 296 void TemplateTable::aconst_null() {
 297   transition(vtos, atos);
 298   __ xorl(rax, rax);
 299 }
 300 
 301 void TemplateTable::iconst(int value) {
 302   transition(vtos, itos);
 303   if (value == 0) {
 304     __ xorl(rax, rax);
 305   } else {
 306     __ movl(rax, value);
 307   }
 308 }
 309 
 310 void TemplateTable::lconst(int value) {
 311   transition(vtos, ltos);
 312   if (value == 0) {
 313     __ xorl(rax, rax);
 314   } else {
 315     __ movl(rax, value);
 316   }
 317 }
 318 
 319 void TemplateTable::fconst(int value) {
 320   transition(vtos, ftos);
 321   static float one = 1.0f, two = 2.0f;
 322   switch (value) {
 323   case 0:
 324     __ xorps(xmm0, xmm0);
 325     break;
 326   case 1:
 327     __ movflt(xmm0, ExternalAddress((address) &one));
 328     break;
 329   case 2:
 330     __ movflt(xmm0, ExternalAddress((address) &two));
 331     break;
 332   default:
 333     ShouldNotReachHere();
 334     break;
 335   }
 336 }
 337 
 338 void TemplateTable::dconst(int value) {
 339   transition(vtos, dtos);
 340   static double one = 1.0;
 341   switch (value) {
 342   case 0:
 343     __ xorpd(xmm0, xmm0);
 344     break;
 345   case 1:
 346     __ movdbl(xmm0, ExternalAddress((address) &one));
 347     break;
 348   default:
 349     ShouldNotReachHere();
 350     break;
 351   }
 352 }
 353 
 354 void TemplateTable::bipush() {
 355   transition(vtos, itos);
 356   __ load_signed_byte(rax, at_bcp(1));
 357 }
 358 
 359 void TemplateTable::sipush() {
 360   transition(vtos, itos);
 361   __ load_unsigned_short(rax, at_bcp(1));
 362   __ bswapl(rax);
 363   __ sarl(rax, 16);
 364 }
 365 
 366 void TemplateTable::ldc(bool wide) {
 367   transition(vtos, vtos);
 368   Label call_ldc, notFloat, notClass, Done;
 369 
 370   if (wide) {
 371     __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 372   } else {
 373     __ load_unsigned_byte(rbx, at_bcp(1));
 374   }
 375 
 376   __ get_cpool_and_tags(rcx, rax);
 377   const int base_offset = ConstantPool::header_size() * wordSize;
 378   const int tags_offset = Array<u1>::base_offset_in_bytes();
 379 
 380   // get type
 381   __ movzbl(rdx, Address(rax, rbx, Address::times_1, tags_offset));
 382 
 383   // unresolved class - get the resolved class
 384   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClass);
 385   __ jccb(Assembler::equal, call_ldc);
 386 
 387   // unresolved class in error state - call into runtime to throw the error
 388   // from the first resolution attempt
 389   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClassInError);
 390   __ jccb(Assembler::equal, call_ldc);
 391 
 392   // resolved class - need to call vm to get java mirror of the class
 393   __ cmpl(rdx, JVM_CONSTANT_Class);
 394   __ jcc(Assembler::notEqual, notClass);
 395 
 396   __ bind(call_ldc);
 397   __ movl(c_rarg1, wide);
 398   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), c_rarg1);
 399   __ push_ptr(rax);
 400   __ verify_oop(rax);
 401   __ jmp(Done);
 402 
 403   __ bind(notClass);
 404   __ cmpl(rdx, JVM_CONSTANT_Float);
 405   __ jccb(Assembler::notEqual, notFloat);
 406   // ftos
 407   __ movflt(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 408   __ push_f();
 409   __ jmp(Done);
 410 
 411   __ bind(notFloat);
 412 #ifdef ASSERT
 413   {
 414     Label L;
 415     __ cmpl(rdx, JVM_CONSTANT_Integer);
 416     __ jcc(Assembler::equal, L);
 417     // String and Object are rewritten to fast_aldc
 418     __ stop("unexpected tag type in ldc");
 419     __ bind(L);
 420   }
 421 #endif
 422   // itos JVM_CONSTANT_Integer only
 423   __ movl(rax, Address(rcx, rbx, Address::times_8, base_offset));
 424   __ push_i(rax);
 425   __ bind(Done);
 426 }
 427 
 428 // Fast path for caching oop constants.
 429 void TemplateTable::fast_aldc(bool wide) {
 430   transition(vtos, atos);
 431 
 432   Register result = rax;
 433   Register tmp = rdx;
 434   int index_size = wide ? sizeof(u2) : sizeof(u1);
 435 
 436   Label resolved;
 437 
 438   // We are resolved if the resolved reference cache entry contains a
 439   // non-null object (String, MethodType, etc.)
 440   assert_different_registers(result, tmp);
 441   __ get_cache_index_at_bcp(tmp, 1, index_size);
 442   __ load_resolved_reference_at_index(result, tmp);
 443   __ testl(result, result);
 444   __ jcc(Assembler::notZero, resolved);
 445 
 446   address entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_ldc);
 447 
 448   // first time invocation - must resolve first
 449   __ movl(tmp, (int)bytecode());
 450   __ call_VM(result, entry, tmp);
 451 
 452   __ bind(resolved);
 453 
 454   if (VerifyOops) {
 455     __ verify_oop(result);
 456   }
 457 }
 458 
 459 void TemplateTable::ldc2_w() {
 460   transition(vtos, vtos);
 461   Label Long, Done;
 462   __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 463 
 464   __ get_cpool_and_tags(rcx, rax);
 465   const int base_offset = ConstantPool::header_size() * wordSize;
 466   const int tags_offset = Array<u1>::base_offset_in_bytes();
 467 
 468   // get type
 469   __ cmpb(Address(rax, rbx, Address::times_1, tags_offset),
 470           JVM_CONSTANT_Double);
 471   __ jccb(Assembler::notEqual, Long);
 472   // dtos
 473   __ movdbl(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 474   __ push_d();
 475   __ jmpb(Done);
 476 
 477   __ bind(Long);
 478   // ltos
 479   __ movq(rax, Address(rcx, rbx, Address::times_8, base_offset));
 480   __ push_l();
 481 
 482   __ bind(Done);
 483 }
 484 
 485 void TemplateTable::locals_index(Register reg, int offset) {
 486   __ load_unsigned_byte(reg, at_bcp(offset));
 487   __ negptr(reg);
 488 }
 489 
 490 void TemplateTable::iload() {
 491   transition(vtos, itos);
 492   if (RewriteFrequentPairs) {
 493     Label rewrite, done;
 494     const Register bc = c_rarg3;
 495     assert(rbx != bc, "register damaged");
 496 
 497     // get next byte
 498     __ load_unsigned_byte(rbx,
 499                           at_bcp(Bytecodes::length_for(Bytecodes::_iload)));
 500     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 501     // last two iloads in a pair.  Comparing against fast_iload means that
 502     // the next bytecode is neither an iload or a caload, and therefore
 503     // an iload pair.
 504     __ cmpl(rbx, Bytecodes::_iload);
 505     __ jcc(Assembler::equal, done);
 506 
 507     __ cmpl(rbx, Bytecodes::_fast_iload);
 508     __ movl(bc, Bytecodes::_fast_iload2);
 509     __ jccb(Assembler::equal, rewrite);
 510 
 511     // if _caload, rewrite to fast_icaload
 512     __ cmpl(rbx, Bytecodes::_caload);
 513     __ movl(bc, Bytecodes::_fast_icaload);
 514     __ jccb(Assembler::equal, rewrite);
 515 
 516     // rewrite so iload doesn't check again.
 517     __ movl(bc, Bytecodes::_fast_iload);
 518 
 519     // rewrite
 520     // bc: fast bytecode
 521     __ bind(rewrite);
 522     patch_bytecode(Bytecodes::_iload, bc, rbx, false);
 523     __ bind(done);
 524   }
 525 
 526   // Get the local value into tos
 527   locals_index(rbx);
 528   __ movl(rax, iaddress(rbx));
 529 }
 530 
 531 void TemplateTable::fast_iload2() {
 532   transition(vtos, itos);
 533   locals_index(rbx);
 534   __ movl(rax, iaddress(rbx));
 535   __ push(itos);
 536   locals_index(rbx, 3);
 537   __ movl(rax, iaddress(rbx));
 538 }
 539 
 540 void TemplateTable::fast_iload() {
 541   transition(vtos, itos);
 542   locals_index(rbx);
 543   __ movl(rax, iaddress(rbx));
 544 }
 545 
 546 void TemplateTable::lload() {
 547   transition(vtos, ltos);
 548   locals_index(rbx);
 549   __ movq(rax, laddress(rbx));
 550 }
 551 
 552 void TemplateTable::fload() {
 553   transition(vtos, ftos);
 554   locals_index(rbx);
 555   __ movflt(xmm0, faddress(rbx));
 556 }
 557 
 558 void TemplateTable::dload() {
 559   transition(vtos, dtos);
 560   locals_index(rbx);
 561   __ movdbl(xmm0, daddress(rbx));
 562 }
 563 
 564 void TemplateTable::aload() {
 565   transition(vtos, atos);
 566   locals_index(rbx);
 567   __ movptr(rax, aaddress(rbx));
 568 }
 569 
 570 void TemplateTable::locals_index_wide(Register reg) {
 571   __ movl(reg, at_bcp(2));
 572   __ bswapl(reg);
 573   __ shrl(reg, 16);
 574   __ negptr(reg);
 575 }
 576 
 577 void TemplateTable::wide_iload() {
 578   transition(vtos, itos);
 579   locals_index_wide(rbx);
 580   __ movl(rax, iaddress(rbx));
 581 }
 582 
 583 void TemplateTable::wide_lload() {
 584   transition(vtos, ltos);
 585   locals_index_wide(rbx);
 586   __ movq(rax, laddress(rbx));
 587 }
 588 
 589 void TemplateTable::wide_fload() {
 590   transition(vtos, ftos);
 591   locals_index_wide(rbx);
 592   __ movflt(xmm0, faddress(rbx));
 593 }
 594 
 595 void TemplateTable::wide_dload() {
 596   transition(vtos, dtos);
 597   locals_index_wide(rbx);
 598   __ movdbl(xmm0, daddress(rbx));
 599 }
 600 
 601 void TemplateTable::wide_aload() {
 602   transition(vtos, atos);
 603   locals_index_wide(rbx);
 604   __ movptr(rax, aaddress(rbx));
 605 }
 606 
 607 void TemplateTable::index_check(Register array, Register index) {
 608   // destroys rbx
 609   // check array
 610   __ null_check(array, arrayOopDesc::length_offset_in_bytes());
 611   // sign extend index for use by indexed load
 612   __ movl2ptr(index, index);
 613   // check index
 614   __ cmpl(index, Address(array, arrayOopDesc::length_offset_in_bytes()));
 615   if (index != rbx) {
 616     // ??? convention: move aberrant index into ebx for exception message
 617     assert(rbx != array, "different registers");
 618     __ movl(rbx, index);
 619   }
 620   __ jump_cc(Assembler::aboveEqual,
 621              ExternalAddress(Interpreter::_throw_ArrayIndexOutOfBoundsException_entry));
 622 }
 623 
 624 void TemplateTable::iaload() {
 625   transition(itos, itos);
 626   __ pop_ptr(rdx);
 627   // eax: index
 628   // rdx: array
 629   index_check(rdx, rax); // kills rbx
 630   __ movl(rax, Address(rdx, rax,
 631                        Address::times_4,
 632                        arrayOopDesc::base_offset_in_bytes(T_INT)));
 633 }
 634 
 635 void TemplateTable::laload() {
 636   transition(itos, ltos);
 637   __ pop_ptr(rdx);
 638   // eax: index
 639   // rdx: array
 640   index_check(rdx, rax); // kills rbx
 641   __ movq(rax, Address(rdx, rbx,
 642                        Address::times_8,
 643                        arrayOopDesc::base_offset_in_bytes(T_LONG)));
 644 }
 645 
 646 void TemplateTable::faload() {
 647   transition(itos, ftos);
 648   __ pop_ptr(rdx);
 649   // eax: index
 650   // rdx: array
 651   index_check(rdx, rax); // kills rbx
 652   __ movflt(xmm0, Address(rdx, rax,
 653                          Address::times_4,
 654                          arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 655 }
 656 
 657 void TemplateTable::daload() {
 658   transition(itos, dtos);
 659   __ pop_ptr(rdx);
 660   // eax: index
 661   // rdx: array
 662   index_check(rdx, rax); // kills rbx
 663   __ movdbl(xmm0, Address(rdx, rax,
 664                           Address::times_8,
 665                           arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 666 }
 667 
 668 void TemplateTable::aaload() {
 669   transition(itos, atos);
 670   __ pop_ptr(rdx);
 671   // eax: index
 672   // rdx: array
 673   index_check(rdx, rax); // kills rbx
 674   __ load_heap_oop(rax, Address(rdx, rax,
 675                                 UseCompressedOops ? Address::times_4 : Address::times_8,
 676                                 arrayOopDesc::base_offset_in_bytes(T_OBJECT)));
 677 }
 678 
 679 void TemplateTable::baload() {
 680   transition(itos, itos);
 681   __ pop_ptr(rdx);
 682   // eax: index
 683   // rdx: array
 684   index_check(rdx, rax); // kills rbx
 685   __ load_signed_byte(rax,
 686                       Address(rdx, rax,
 687                               Address::times_1,
 688                               arrayOopDesc::base_offset_in_bytes(T_BYTE)));
 689 }
 690 
 691 void TemplateTable::caload() {
 692   transition(itos, itos);
 693   __ pop_ptr(rdx);
 694   // eax: index
 695   // rdx: array
 696   index_check(rdx, rax); // kills rbx
 697   __ load_unsigned_short(rax,
 698                          Address(rdx, rax,
 699                                  Address::times_2,
 700                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 701 }
 702 
 703 // iload followed by caload frequent pair
 704 void TemplateTable::fast_icaload() {
 705   transition(vtos, itos);
 706   // load index out of locals
 707   locals_index(rbx);
 708   __ movl(rax, iaddress(rbx));
 709 
 710   // eax: index
 711   // rdx: array
 712   __ pop_ptr(rdx);
 713   index_check(rdx, rax); // kills rbx
 714   __ load_unsigned_short(rax,
 715                          Address(rdx, rax,
 716                                  Address::times_2,
 717                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 718 }
 719 
 720 void TemplateTable::saload() {
 721   transition(itos, itos);
 722   __ pop_ptr(rdx);
 723   // eax: index
 724   // rdx: array
 725   index_check(rdx, rax); // kills rbx
 726   __ load_signed_short(rax,
 727                        Address(rdx, rax,
 728                                Address::times_2,
 729                                arrayOopDesc::base_offset_in_bytes(T_SHORT)));
 730 }
 731 
 732 void TemplateTable::iload(int n) {
 733   transition(vtos, itos);
 734   __ movl(rax, iaddress(n));
 735 }
 736 
 737 void TemplateTable::lload(int n) {
 738   transition(vtos, ltos);
 739   __ movq(rax, laddress(n));
 740 }
 741 
 742 void TemplateTable::fload(int n) {
 743   transition(vtos, ftos);
 744   __ movflt(xmm0, faddress(n));
 745 }
 746 
 747 void TemplateTable::dload(int n) {
 748   transition(vtos, dtos);
 749   __ movdbl(xmm0, daddress(n));
 750 }
 751 
 752 void TemplateTable::aload(int n) {
 753   transition(vtos, atos);
 754   __ movptr(rax, aaddress(n));
 755 }
 756 
 757 void TemplateTable::aload_0() {
 758   transition(vtos, atos);
 759   // According to bytecode histograms, the pairs:
 760   //
 761   // _aload_0, _fast_igetfield
 762   // _aload_0, _fast_agetfield
 763   // _aload_0, _fast_fgetfield
 764   //
 765   // occur frequently. If RewriteFrequentPairs is set, the (slow)
 766   // _aload_0 bytecode checks if the next bytecode is either
 767   // _fast_igetfield, _fast_agetfield or _fast_fgetfield and then
 768   // rewrites the current bytecode into a pair bytecode; otherwise it
 769   // rewrites the current bytecode into _fast_aload_0 that doesn't do
 770   // the pair check anymore.
 771   //
 772   // Note: If the next bytecode is _getfield, the rewrite must be
 773   //       delayed, otherwise we may miss an opportunity for a pair.
 774   //
 775   // Also rewrite frequent pairs
 776   //   aload_0, aload_1
 777   //   aload_0, iload_1
 778   // These bytecodes with a small amount of code are most profitable
 779   // to rewrite
 780   if (RewriteFrequentPairs) {
 781     Label rewrite, done;
 782     const Register bc = c_rarg3;
 783     assert(rbx != bc, "register damaged");
 784     // get next byte
 785     __ load_unsigned_byte(rbx,
 786                           at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)));
 787 
 788     // do actual aload_0
 789     aload(0);
 790 
 791     // if _getfield then wait with rewrite
 792     __ cmpl(rbx, Bytecodes::_getfield);
 793     __ jcc(Assembler::equal, done);
 794 
 795     // if _igetfield then reqrite to _fast_iaccess_0
 796     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) ==
 797            Bytecodes::_aload_0,
 798            "fix bytecode definition");
 799     __ cmpl(rbx, Bytecodes::_fast_igetfield);
 800     __ movl(bc, Bytecodes::_fast_iaccess_0);
 801     __ jccb(Assembler::equal, rewrite);
 802 
 803     // if _agetfield then reqrite to _fast_aaccess_0
 804     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) ==
 805            Bytecodes::_aload_0,
 806            "fix bytecode definition");
 807     __ cmpl(rbx, Bytecodes::_fast_agetfield);
 808     __ movl(bc, Bytecodes::_fast_aaccess_0);
 809     __ jccb(Assembler::equal, rewrite);
 810 
 811     // if _fgetfield then reqrite to _fast_faccess_0
 812     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) ==
 813            Bytecodes::_aload_0,
 814            "fix bytecode definition");
 815     __ cmpl(rbx, Bytecodes::_fast_fgetfield);
 816     __ movl(bc, Bytecodes::_fast_faccess_0);
 817     __ jccb(Assembler::equal, rewrite);
 818 
 819     // else rewrite to _fast_aload0
 820     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) ==
 821            Bytecodes::_aload_0,
 822            "fix bytecode definition");
 823     __ movl(bc, Bytecodes::_fast_aload_0);
 824 
 825     // rewrite
 826     // bc: fast bytecode
 827     __ bind(rewrite);
 828     patch_bytecode(Bytecodes::_aload_0, bc, rbx, false);
 829 
 830     __ bind(done);
 831   } else {
 832     aload(0);
 833   }
 834 }
 835 
 836 void TemplateTable::istore() {
 837   transition(itos, vtos);
 838   locals_index(rbx);
 839   __ movl(iaddress(rbx), rax);
 840 }
 841 
 842 void TemplateTable::lstore() {
 843   transition(ltos, vtos);
 844   locals_index(rbx);
 845   __ movq(laddress(rbx), rax);
 846 }
 847 
 848 void TemplateTable::fstore() {
 849   transition(ftos, vtos);
 850   locals_index(rbx);
 851   __ movflt(faddress(rbx), xmm0);
 852 }
 853 
 854 void TemplateTable::dstore() {
 855   transition(dtos, vtos);
 856   locals_index(rbx);
 857   __ movdbl(daddress(rbx), xmm0);
 858 }
 859 
 860 void TemplateTable::astore() {
 861   transition(vtos, vtos);
 862   __ pop_ptr(rax);
 863   locals_index(rbx);
 864   __ movptr(aaddress(rbx), rax);
 865 }
 866 
 867 void TemplateTable::wide_istore() {
 868   transition(vtos, vtos);
 869   __ pop_i();
 870   locals_index_wide(rbx);
 871   __ movl(iaddress(rbx), rax);
 872 }
 873 
 874 void TemplateTable::wide_lstore() {
 875   transition(vtos, vtos);
 876   __ pop_l();
 877   locals_index_wide(rbx);
 878   __ movq(laddress(rbx), rax);
 879 }
 880 
 881 void TemplateTable::wide_fstore() {
 882   transition(vtos, vtos);
 883   __ pop_f();
 884   locals_index_wide(rbx);
 885   __ movflt(faddress(rbx), xmm0);
 886 }
 887 
 888 void TemplateTable::wide_dstore() {
 889   transition(vtos, vtos);
 890   __ pop_d();
 891   locals_index_wide(rbx);
 892   __ movdbl(daddress(rbx), xmm0);
 893 }
 894 
 895 void TemplateTable::wide_astore() {
 896   transition(vtos, vtos);
 897   __ pop_ptr(rax);
 898   locals_index_wide(rbx);
 899   __ movptr(aaddress(rbx), rax);
 900 }
 901 
 902 void TemplateTable::iastore() {
 903   transition(itos, vtos);
 904   __ pop_i(rbx);
 905   __ pop_ptr(rdx);
 906   // eax: value
 907   // ebx: index
 908   // rdx: array
 909   index_check(rdx, rbx); // prefer index in ebx
 910   __ movl(Address(rdx, rbx,
 911                   Address::times_4,
 912                   arrayOopDesc::base_offset_in_bytes(T_INT)),
 913           rax);
 914 }
 915 
 916 void TemplateTable::lastore() {
 917   transition(ltos, vtos);
 918   __ pop_i(rbx);
 919   __ pop_ptr(rdx);
 920   // rax: value
 921   // ebx: index
 922   // rdx: array
 923   index_check(rdx, rbx); // prefer index in ebx
 924   __ movq(Address(rdx, rbx,
 925                   Address::times_8,
 926                   arrayOopDesc::base_offset_in_bytes(T_LONG)),
 927           rax);
 928 }
 929 
 930 void TemplateTable::fastore() {
 931   transition(ftos, vtos);
 932   __ pop_i(rbx);
 933   __ pop_ptr(rdx);
 934   // xmm0: value
 935   // ebx:  index
 936   // rdx:  array
 937   index_check(rdx, rbx); // prefer index in ebx
 938   __ movflt(Address(rdx, rbx,
 939                    Address::times_4,
 940                    arrayOopDesc::base_offset_in_bytes(T_FLOAT)),
 941            xmm0);
 942 }
 943 
 944 void TemplateTable::dastore() {
 945   transition(dtos, vtos);
 946   __ pop_i(rbx);
 947   __ pop_ptr(rdx);
 948   // xmm0: value
 949   // ebx:  index
 950   // rdx:  array
 951   index_check(rdx, rbx); // prefer index in ebx
 952   __ movdbl(Address(rdx, rbx,
 953                    Address::times_8,
 954                    arrayOopDesc::base_offset_in_bytes(T_DOUBLE)),
 955            xmm0);
 956 }
 957 
 958 void TemplateTable::aastore() {
 959   Label is_null, ok_is_subtype, done;
 960   transition(vtos, vtos);
 961   // stack: ..., array, index, value
 962   __ movptr(rax, at_tos());    // value
 963   __ movl(rcx, at_tos_p1()); // index
 964   __ movptr(rdx, at_tos_p2()); // array
 965 
 966   Address element_address(rdx, rcx,
 967                           UseCompressedOops? Address::times_4 : Address::times_8,
 968                           arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 969 
 970   index_check(rdx, rcx);     // kills rbx
 971   // do array store check - check for NULL value first
 972   __ testptr(rax, rax);
 973   __ jcc(Assembler::zero, is_null);
 974 
 975   // Move subklass into rbx
 976   __ load_klass(rbx, rax);
 977   // Move superklass into rax
 978   __ load_klass(rax, rdx);
 979   __ movptr(rax, Address(rax,
 980                          ObjArrayKlass::element_klass_offset()));
 981   // Compress array + index*oopSize + 12 into a single register.  Frees rcx.
 982   __ lea(rdx, element_address);
 983 
 984   // Generate subtype check.  Blows rcx, rdi
 985   // Superklass in rax.  Subklass in rbx.
 986   __ gen_subtype_check(rbx, ok_is_subtype);
 987 
 988   // Come here on failure
 989   // object is at TOS
 990   __ jump(ExternalAddress(Interpreter::_throw_ArrayStoreException_entry));
 991 
 992   // Come here on success
 993   __ bind(ok_is_subtype);
 994 
 995   // Get the value we will store
 996   __ movptr(rax, at_tos());
 997   // Now store using the appropriate barrier
 998   do_oop_store(_masm, Address(rdx, 0), rax, _bs->kind(), true);
 999   __ jmp(done);
1000 
1001   // Have a NULL in rax, rdx=array, ecx=index.  Store NULL at ary[idx]
1002   __ bind(is_null);
1003   __ profile_null_seen(rbx);
1004 
1005   // Store a NULL
1006   do_oop_store(_masm, element_address, noreg, _bs->kind(), true);
1007 
1008   // Pop stack arguments
1009   __ bind(done);
1010   __ addptr(rsp, 3 * Interpreter::stackElementSize);
1011 }
1012 
1013 void TemplateTable::bastore() {
1014   transition(itos, vtos);
1015   __ pop_i(rbx);
1016   __ pop_ptr(rdx);
1017   // eax: value
1018   // ebx: index
1019   // rdx: array
1020   index_check(rdx, rbx); // prefer index in ebx
1021   __ movb(Address(rdx, rbx,
1022                   Address::times_1,
1023                   arrayOopDesc::base_offset_in_bytes(T_BYTE)),
1024           rax);
1025 }
1026 
1027 void TemplateTable::castore() {
1028   transition(itos, vtos);
1029   __ pop_i(rbx);
1030   __ pop_ptr(rdx);
1031   // eax: value
1032   // ebx: index
1033   // rdx: array
1034   index_check(rdx, rbx);  // prefer index in ebx
1035   __ movw(Address(rdx, rbx,
1036                   Address::times_2,
1037                   arrayOopDesc::base_offset_in_bytes(T_CHAR)),
1038           rax);
1039 }
1040 
1041 void TemplateTable::sastore() {
1042   castore();
1043 }
1044 
1045 void TemplateTable::istore(int n) {
1046   transition(itos, vtos);
1047   __ movl(iaddress(n), rax);
1048 }
1049 
1050 void TemplateTable::lstore(int n) {
1051   transition(ltos, vtos);
1052   __ movq(laddress(n), rax);
1053 }
1054 
1055 void TemplateTable::fstore(int n) {
1056   transition(ftos, vtos);
1057   __ movflt(faddress(n), xmm0);
1058 }
1059 
1060 void TemplateTable::dstore(int n) {
1061   transition(dtos, vtos);
1062   __ movdbl(daddress(n), xmm0);
1063 }
1064 
1065 void TemplateTable::astore(int n) {
1066   transition(vtos, vtos);
1067   __ pop_ptr(rax);
1068   __ movptr(aaddress(n), rax);
1069 }
1070 
1071 void TemplateTable::pop() {
1072   transition(vtos, vtos);
1073   __ addptr(rsp, Interpreter::stackElementSize);
1074 }
1075 
1076 void TemplateTable::pop2() {
1077   transition(vtos, vtos);
1078   __ addptr(rsp, 2 * Interpreter::stackElementSize);
1079 }
1080 
1081 void TemplateTable::dup() {
1082   transition(vtos, vtos);
1083   __ load_ptr(0, rax);
1084   __ push_ptr(rax);
1085   // stack: ..., a, a
1086 }
1087 
1088 void TemplateTable::dup_x1() {
1089   transition(vtos, vtos);
1090   // stack: ..., a, b
1091   __ load_ptr( 0, rax);  // load b
1092   __ load_ptr( 1, rcx);  // load a
1093   __ store_ptr(1, rax);  // store b
1094   __ store_ptr(0, rcx);  // store a
1095   __ push_ptr(rax);      // push b
1096   // stack: ..., b, a, b
1097 }
1098 
1099 void TemplateTable::dup_x2() {
1100   transition(vtos, vtos);
1101   // stack: ..., a, b, c
1102   __ load_ptr( 0, rax);  // load c
1103   __ load_ptr( 2, rcx);  // load a
1104   __ store_ptr(2, rax);  // store c in a
1105   __ push_ptr(rax);      // push c
1106   // stack: ..., c, b, c, c
1107   __ load_ptr( 2, rax);  // load b
1108   __ store_ptr(2, rcx);  // store a in b
1109   // stack: ..., c, a, c, c
1110   __ store_ptr(1, rax);  // store b in c
1111   // stack: ..., c, a, b, c
1112 }
1113 
1114 void TemplateTable::dup2() {
1115   transition(vtos, vtos);
1116   // stack: ..., a, b
1117   __ load_ptr(1, rax);  // load a
1118   __ push_ptr(rax);     // push a
1119   __ load_ptr(1, rax);  // load b
1120   __ push_ptr(rax);     // push b
1121   // stack: ..., a, b, a, b
1122 }
1123 
1124 void TemplateTable::dup2_x1() {
1125   transition(vtos, vtos);
1126   // stack: ..., a, b, c
1127   __ load_ptr( 0, rcx);  // load c
1128   __ load_ptr( 1, rax);  // load b
1129   __ push_ptr(rax);      // push b
1130   __ push_ptr(rcx);      // push c
1131   // stack: ..., a, b, c, b, c
1132   __ store_ptr(3, rcx);  // store c in b
1133   // stack: ..., a, c, c, b, c
1134   __ load_ptr( 4, rcx);  // load a
1135   __ store_ptr(2, rcx);  // store a in 2nd c
1136   // stack: ..., a, c, a, b, c
1137   __ store_ptr(4, rax);  // store b in a
1138   // stack: ..., b, c, a, b, c
1139 }
1140 
1141 void TemplateTable::dup2_x2() {
1142   transition(vtos, vtos);
1143   // stack: ..., a, b, c, d
1144   __ load_ptr( 0, rcx);  // load d
1145   __ load_ptr( 1, rax);  // load c
1146   __ push_ptr(rax);      // push c
1147   __ push_ptr(rcx);      // push d
1148   // stack: ..., a, b, c, d, c, d
1149   __ load_ptr( 4, rax);  // load b
1150   __ store_ptr(2, rax);  // store b in d
1151   __ store_ptr(4, rcx);  // store d in b
1152   // stack: ..., a, d, c, b, c, d
1153   __ load_ptr( 5, rcx);  // load a
1154   __ load_ptr( 3, rax);  // load c
1155   __ store_ptr(3, rcx);  // store a in c
1156   __ store_ptr(5, rax);  // store c in a
1157   // stack: ..., c, d, a, b, c, d
1158 }
1159 
1160 void TemplateTable::swap() {
1161   transition(vtos, vtos);
1162   // stack: ..., a, b
1163   __ load_ptr( 1, rcx);  // load a
1164   __ load_ptr( 0, rax);  // load b
1165   __ store_ptr(0, rcx);  // store a in b
1166   __ store_ptr(1, rax);  // store b in a
1167   // stack: ..., b, a
1168 }
1169 
1170 void TemplateTable::iop2(Operation op) {
1171   transition(itos, itos);
1172   switch (op) {
1173   case add  :                    __ pop_i(rdx); __ addl (rax, rdx); break;
1174   case sub  : __ movl(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1175   case mul  :                    __ pop_i(rdx); __ imull(rax, rdx); break;
1176   case _and :                    __ pop_i(rdx); __ andl (rax, rdx); break;
1177   case _or  :                    __ pop_i(rdx); __ orl  (rax, rdx); break;
1178   case _xor :                    __ pop_i(rdx); __ xorl (rax, rdx); break;
1179   case shl  : __ movl(rcx, rax); __ pop_i(rax); __ shll (rax);      break;
1180   case shr  : __ movl(rcx, rax); __ pop_i(rax); __ sarl (rax);      break;
1181   case ushr : __ movl(rcx, rax); __ pop_i(rax); __ shrl (rax);      break;
1182   default   : ShouldNotReachHere();
1183   }
1184 }
1185 
1186 void TemplateTable::lop2(Operation op) {
1187   transition(ltos, ltos);
1188   switch (op) {
1189   case add  :                    __ pop_l(rdx); __ addptr(rax, rdx); break;
1190   case sub  : __ mov(rdx, rax);  __ pop_l(rax); __ subptr(rax, rdx); break;
1191   case _and :                    __ pop_l(rdx); __ andptr(rax, rdx); break;
1192   case _or  :                    __ pop_l(rdx); __ orptr (rax, rdx); break;
1193   case _xor :                    __ pop_l(rdx); __ xorptr(rax, rdx); break;
1194   default   : ShouldNotReachHere();
1195   }
1196 }
1197 
1198 void TemplateTable::idiv() {
1199   transition(itos, itos);
1200   __ movl(rcx, rax);
1201   __ pop_i(rax);
1202   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1203   //       they are not equal, one could do a normal division (no correction
1204   //       needed), which may speed up this implementation for the common case.
1205   //       (see also JVM spec., p.243 & p.271)
1206   __ corrected_idivl(rcx);
1207 }
1208 
1209 void TemplateTable::irem() {
1210   transition(itos, itos);
1211   __ movl(rcx, rax);
1212   __ pop_i(rax);
1213   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1214   //       they are not equal, one could do a normal division (no correction
1215   //       needed), which may speed up this implementation for the common case.
1216   //       (see also JVM spec., p.243 & p.271)
1217   __ corrected_idivl(rcx);
1218   __ movl(rax, rdx);
1219 }
1220 
1221 void TemplateTable::lmul() {
1222   transition(ltos, ltos);
1223   __ pop_l(rdx);
1224   __ imulq(rax, rdx);
1225 }
1226 
1227 void TemplateTable::ldiv() {
1228   transition(ltos, ltos);
1229   __ mov(rcx, rax);
1230   __ pop_l(rax);
1231   // generate explicit div0 check
1232   __ testq(rcx, rcx);
1233   __ jump_cc(Assembler::zero,
1234              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1235   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1236   //       they are not equal, one could do a normal division (no correction
1237   //       needed), which may speed up this implementation for the common case.
1238   //       (see also JVM spec., p.243 & p.271)
1239   __ corrected_idivq(rcx); // kills rbx
1240 }
1241 
1242 void TemplateTable::lrem() {
1243   transition(ltos, ltos);
1244   __ mov(rcx, rax);
1245   __ pop_l(rax);
1246   __ testq(rcx, rcx);
1247   __ jump_cc(Assembler::zero,
1248              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1249   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1250   //       they are not equal, one could do a normal division (no correction
1251   //       needed), which may speed up this implementation for the common case.
1252   //       (see also JVM spec., p.243 & p.271)
1253   __ corrected_idivq(rcx); // kills rbx
1254   __ mov(rax, rdx);
1255 }
1256 
1257 void TemplateTable::lshl() {
1258   transition(itos, ltos);
1259   __ movl(rcx, rax);                             // get shift count
1260   __ pop_l(rax);                                 // get shift value
1261   __ shlq(rax);
1262 }
1263 
1264 void TemplateTable::lshr() {
1265   transition(itos, ltos);
1266   __ movl(rcx, rax);                             // get shift count
1267   __ pop_l(rax);                                 // get shift value
1268   __ sarq(rax);
1269 }
1270 
1271 void TemplateTable::lushr() {
1272   transition(itos, ltos);
1273   __ movl(rcx, rax);                             // get shift count
1274   __ pop_l(rax);                                 // get shift value
1275   __ shrq(rax);
1276 }
1277 
1278 void TemplateTable::fop2(Operation op) {
1279   transition(ftos, ftos);
1280   switch (op) {
1281   case add:
1282     __ addss(xmm0, at_rsp());
1283     __ addptr(rsp, Interpreter::stackElementSize);
1284     break;
1285   case sub:
1286     __ movflt(xmm1, xmm0);
1287     __ pop_f(xmm0);
1288     __ subss(xmm0, xmm1);
1289     break;
1290   case mul:
1291     __ mulss(xmm0, at_rsp());
1292     __ addptr(rsp, Interpreter::stackElementSize);
1293     break;
1294   case div:
1295     __ movflt(xmm1, xmm0);
1296     __ pop_f(xmm0);
1297     __ divss(xmm0, xmm1);
1298     break;
1299   case rem:
1300     __ movflt(xmm1, xmm0);
1301     __ pop_f(xmm0);
1302     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::frem), 2);
1303     break;
1304   default:
1305     ShouldNotReachHere();
1306     break;
1307   }
1308 }
1309 
1310 void TemplateTable::dop2(Operation op) {
1311   transition(dtos, dtos);
1312   switch (op) {
1313   case add:
1314     __ addsd(xmm0, at_rsp());
1315     __ addptr(rsp, 2 * Interpreter::stackElementSize);
1316     break;
1317   case sub:
1318     __ movdbl(xmm1, xmm0);
1319     __ pop_d(xmm0);
1320     __ subsd(xmm0, xmm1);
1321     break;
1322   case mul:
1323     __ mulsd(xmm0, at_rsp());
1324     __ addptr(rsp, 2 * Interpreter::stackElementSize);
1325     break;
1326   case div:
1327     __ movdbl(xmm1, xmm0);
1328     __ pop_d(xmm0);
1329     __ divsd(xmm0, xmm1);
1330     break;
1331   case rem:
1332     __ movdbl(xmm1, xmm0);
1333     __ pop_d(xmm0);
1334     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::drem), 2);
1335     break;
1336   default:
1337     ShouldNotReachHere();
1338     break;
1339   }
1340 }
1341 
1342 void TemplateTable::ineg() {
1343   transition(itos, itos);
1344   __ negl(rax);
1345 }
1346 
1347 void TemplateTable::lneg() {
1348   transition(ltos, ltos);
1349   __ negq(rax);
1350 }
1351 
1352 // Note: 'double' and 'long long' have 32-bits alignment on x86.
1353 static jlong* double_quadword(jlong *adr, jlong lo, jlong hi) {
1354   // Use the expression (adr)&(~0xF) to provide 128-bits aligned address
1355   // of 128-bits operands for SSE instructions.
1356   jlong *operand = (jlong*)(((intptr_t)adr)&((intptr_t)(~0xF)));
1357   // Store the value to a 128-bits operand.
1358   operand[0] = lo;
1359   operand[1] = hi;
1360   return operand;
1361 }
1362 
1363 // Buffer for 128-bits masks used by SSE instructions.
1364 static jlong float_signflip_pool[2*2];
1365 static jlong double_signflip_pool[2*2];
1366 
1367 void TemplateTable::fneg() {
1368   transition(ftos, ftos);
1369   static jlong *float_signflip  = double_quadword(&float_signflip_pool[1], 0x8000000080000000, 0x8000000080000000);
1370   __ xorps(xmm0, ExternalAddress((address) float_signflip));
1371 }
1372 
1373 void TemplateTable::dneg() {
1374   transition(dtos, dtos);
1375   static jlong *double_signflip  = double_quadword(&double_signflip_pool[1], 0x8000000000000000, 0x8000000000000000);
1376   __ xorpd(xmm0, ExternalAddress((address) double_signflip));
1377 }
1378 
1379 void TemplateTable::iinc() {
1380   transition(vtos, vtos);
1381   __ load_signed_byte(rdx, at_bcp(2)); // get constant
1382   locals_index(rbx);
1383   __ addl(iaddress(rbx), rdx);
1384 }
1385 
1386 void TemplateTable::wide_iinc() {
1387   transition(vtos, vtos);
1388   __ movl(rdx, at_bcp(4)); // get constant
1389   locals_index_wide(rbx);
1390   __ bswapl(rdx); // swap bytes & sign-extend constant
1391   __ sarl(rdx, 16);
1392   __ addl(iaddress(rbx), rdx);
1393   // Note: should probably use only one movl to get both
1394   //       the index and the constant -> fix this
1395 }
1396 
1397 void TemplateTable::convert() {
1398   // Checking
1399 #ifdef ASSERT
1400   {
1401     TosState tos_in  = ilgl;
1402     TosState tos_out = ilgl;
1403     switch (bytecode()) {
1404     case Bytecodes::_i2l: // fall through
1405     case Bytecodes::_i2f: // fall through
1406     case Bytecodes::_i2d: // fall through
1407     case Bytecodes::_i2b: // fall through
1408     case Bytecodes::_i2c: // fall through
1409     case Bytecodes::_i2s: tos_in = itos; break;
1410     case Bytecodes::_l2i: // fall through
1411     case Bytecodes::_l2f: // fall through
1412     case Bytecodes::_l2d: tos_in = ltos; break;
1413     case Bytecodes::_f2i: // fall through
1414     case Bytecodes::_f2l: // fall through
1415     case Bytecodes::_f2d: tos_in = ftos; break;
1416     case Bytecodes::_d2i: // fall through
1417     case Bytecodes::_d2l: // fall through
1418     case Bytecodes::_d2f: tos_in = dtos; break;
1419     default             : ShouldNotReachHere();
1420     }
1421     switch (bytecode()) {
1422     case Bytecodes::_l2i: // fall through
1423     case Bytecodes::_f2i: // fall through
1424     case Bytecodes::_d2i: // fall through
1425     case Bytecodes::_i2b: // fall through
1426     case Bytecodes::_i2c: // fall through
1427     case Bytecodes::_i2s: tos_out = itos; break;
1428     case Bytecodes::_i2l: // fall through
1429     case Bytecodes::_f2l: // fall through
1430     case Bytecodes::_d2l: tos_out = ltos; break;
1431     case Bytecodes::_i2f: // fall through
1432     case Bytecodes::_l2f: // fall through
1433     case Bytecodes::_d2f: tos_out = ftos; break;
1434     case Bytecodes::_i2d: // fall through
1435     case Bytecodes::_l2d: // fall through
1436     case Bytecodes::_f2d: tos_out = dtos; break;
1437     default             : ShouldNotReachHere();
1438     }
1439     transition(tos_in, tos_out);
1440   }
1441 #endif // ASSERT
1442 
1443   static const int64_t is_nan = 0x8000000000000000L;
1444 
1445   // Conversion
1446   switch (bytecode()) {
1447   case Bytecodes::_i2l:
1448     __ movslq(rax, rax);
1449     break;
1450   case Bytecodes::_i2f:
1451     __ cvtsi2ssl(xmm0, rax);
1452     break;
1453   case Bytecodes::_i2d:
1454     __ cvtsi2sdl(xmm0, rax);
1455     break;
1456   case Bytecodes::_i2b:
1457     __ movsbl(rax, rax);
1458     break;
1459   case Bytecodes::_i2c:
1460     __ movzwl(rax, rax);
1461     break;
1462   case Bytecodes::_i2s:
1463     __ movswl(rax, rax);
1464     break;
1465   case Bytecodes::_l2i:
1466     __ movl(rax, rax);
1467     break;
1468   case Bytecodes::_l2f:
1469     __ cvtsi2ssq(xmm0, rax);
1470     break;
1471   case Bytecodes::_l2d:
1472     __ cvtsi2sdq(xmm0, rax);
1473     break;
1474   case Bytecodes::_f2i:
1475   {
1476     Label L;
1477     __ cvttss2sil(rax, xmm0);
1478     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1479     __ jcc(Assembler::notEqual, L);
1480     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1481     __ bind(L);
1482   }
1483     break;
1484   case Bytecodes::_f2l:
1485   {
1486     Label L;
1487     __ cvttss2siq(rax, xmm0);
1488     // NaN or overflow/underflow?
1489     __ cmp64(rax, ExternalAddress((address) &is_nan));
1490     __ jcc(Assembler::notEqual, L);
1491     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1492     __ bind(L);
1493   }
1494     break;
1495   case Bytecodes::_f2d:
1496     __ cvtss2sd(xmm0, xmm0);
1497     break;
1498   case Bytecodes::_d2i:
1499   {
1500     Label L;
1501     __ cvttsd2sil(rax, xmm0);
1502     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1503     __ jcc(Assembler::notEqual, L);
1504     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 1);
1505     __ bind(L);
1506   }
1507     break;
1508   case Bytecodes::_d2l:
1509   {
1510     Label L;
1511     __ cvttsd2siq(rax, xmm0);
1512     // NaN or overflow/underflow?
1513     __ cmp64(rax, ExternalAddress((address) &is_nan));
1514     __ jcc(Assembler::notEqual, L);
1515     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 1);
1516     __ bind(L);
1517   }
1518     break;
1519   case Bytecodes::_d2f:
1520     __ cvtsd2ss(xmm0, xmm0);
1521     break;
1522   default:
1523     ShouldNotReachHere();
1524   }
1525 }
1526 
1527 void TemplateTable::lcmp() {
1528   transition(ltos, itos);
1529   Label done;
1530   __ pop_l(rdx);
1531   __ cmpq(rdx, rax);
1532   __ movl(rax, -1);
1533   __ jccb(Assembler::less, done);
1534   __ setb(Assembler::notEqual, rax);
1535   __ movzbl(rax, rax);
1536   __ bind(done);
1537 }
1538 
1539 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1540   Label done;
1541   if (is_float) {
1542     // XXX get rid of pop here, use ... reg, mem32
1543     __ pop_f(xmm1);
1544     __ ucomiss(xmm1, xmm0);
1545   } else {
1546     // XXX get rid of pop here, use ... reg, mem64
1547     __ pop_d(xmm1);
1548     __ ucomisd(xmm1, xmm0);
1549   }
1550   if (unordered_result < 0) {
1551     __ movl(rax, -1);
1552     __ jccb(Assembler::parity, done);
1553     __ jccb(Assembler::below, done);
1554     __ setb(Assembler::notEqual, rdx);
1555     __ movzbl(rax, rdx);
1556   } else {
1557     __ movl(rax, 1);
1558     __ jccb(Assembler::parity, done);
1559     __ jccb(Assembler::above, done);
1560     __ movl(rax, 0);
1561     __ jccb(Assembler::equal, done);
1562     __ decrementl(rax);
1563   }
1564   __ bind(done);
1565 }
1566 
1567 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1568   __ get_method(rcx); // rcx holds method
1569   __ profile_taken_branch(rax, rbx); // rax holds updated MDP, rbx
1570                                      // holds bumped taken count
1571 
1572   const ByteSize be_offset = MethodCounters::backedge_counter_offset() +
1573                              InvocationCounter::counter_offset();
1574   const ByteSize inv_offset = MethodCounters::invocation_counter_offset() +
1575                               InvocationCounter::counter_offset();
1576 
1577   // Load up edx with the branch displacement
1578   __ movl(rdx, at_bcp(1));
1579   __ bswapl(rdx);
1580 
1581   if (!is_wide) {
1582     __ sarl(rdx, 16);
1583   }
1584   __ movl2ptr(rdx, rdx);
1585 
1586   // Handle all the JSR stuff here, then exit.
1587   // It's much shorter and cleaner than intermingling with the non-JSR
1588   // normal-branch stuff occurring below.
1589   if (is_jsr) {
1590     // Pre-load the next target bytecode into rbx
1591     __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1, 0));
1592 
1593     // compute return address as bci in rax
1594     __ lea(rax, at_bcp((is_wide ? 5 : 3) -
1595                         in_bytes(ConstMethod::codes_offset())));
1596     __ subptr(rax, Address(rcx, Method::const_offset()));
1597     // Adjust the bcp in r13 by the displacement in rdx
1598     __ addptr(r13, rdx);
1599     // jsr returns atos that is not an oop
1600     __ push_i(rax);
1601     __ dispatch_only(vtos);
1602     return;
1603   }
1604 
1605   // Normal (non-jsr) branch handling
1606 
1607   // Adjust the bcp in r13 by the displacement in rdx
1608   __ addptr(r13, rdx);
1609 
1610   assert(UseLoopCounter || !UseOnStackReplacement,
1611          "on-stack-replacement requires loop counters");
1612   Label backedge_counter_overflow;
1613   Label profile_method;
1614   Label dispatch;
1615   if (UseLoopCounter) {
1616     // increment backedge counter for backward branches
1617     // rax: MDO
1618     // ebx: MDO bumped taken-count
1619     // rcx: method
1620     // rdx: target offset
1621     // r13: target bcp
1622     // r14: locals pointer
1623     __ testl(rdx, rdx);             // check if forward or backward branch
1624     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1625 
1626     // check if MethodCounters exists
1627     Label has_counters;
1628     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1629     __ testptr(rax, rax);
1630     __ jcc(Assembler::notZero, has_counters);
1631     __ push(rdx);
1632     __ push(rcx);
1633     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::build_method_counters),
1634                rcx);
1635     __ pop(rcx);
1636     __ pop(rdx);
1637     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1638     __ jcc(Assembler::zero, dispatch);
1639     __ bind(has_counters);
1640 
1641     if (TieredCompilation) {
1642       Label no_mdo;
1643       int increment = InvocationCounter::count_increment;
1644       int mask = ((1 << Tier0BackedgeNotifyFreqLog) - 1) << InvocationCounter::count_shift;
1645       if (ProfileInterpreter) {
1646         // Are we profiling?
1647         __ movptr(rbx, Address(rcx, in_bytes(Method::method_data_offset())));
1648         __ testptr(rbx, rbx);
1649         __ jccb(Assembler::zero, no_mdo);
1650         // Increment the MDO backedge counter
1651         const Address mdo_backedge_counter(rbx, in_bytes(MethodData::backedge_counter_offset()) +
1652                                            in_bytes(InvocationCounter::counter_offset()));
1653         __ increment_mask_and_jump(mdo_backedge_counter, increment, mask,
1654                                    rax, false, Assembler::zero, &backedge_counter_overflow);
1655         __ jmp(dispatch);
1656       }
1657       __ bind(no_mdo);
1658       // Increment backedge counter in MethodCounters*
1659       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1660       __ increment_mask_and_jump(Address(rcx, be_offset), increment, mask,
1661                                  rax, false, Assembler::zero, &backedge_counter_overflow);
1662     } else {
1663       // increment counter
1664       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1665       __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1666       __ incrementl(rax, InvocationCounter::count_increment); // increment counter
1667       __ movl(Address(rcx, be_offset), rax);        // store counter
1668 
1669       __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1670 
1671       __ andl(rax, InvocationCounter::count_mask_value); // and the status bits
1672       __ addl(rax, Address(rcx, be_offset));        // add both counters
1673 
1674       if (ProfileInterpreter) {
1675         // Test to see if we should create a method data oop
1676         __ cmp32(rax,
1677                  ExternalAddress((address) &InvocationCounter::InterpreterProfileLimit));
1678         __ jcc(Assembler::less, dispatch);
1679 
1680         // if no method data exists, go to profile method
1681         __ test_method_data_pointer(rax, profile_method);
1682 
1683         if (UseOnStackReplacement) {
1684           // check for overflow against ebx which is the MDO taken count
1685           __ cmp32(rbx,
1686                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1687           __ jcc(Assembler::below, dispatch);
1688 
1689           // When ProfileInterpreter is on, the backedge_count comes
1690           // from the MethodData*, which value does not get reset on
1691           // the call to frequency_counter_overflow().  To avoid
1692           // excessive calls to the overflow routine while the method is
1693           // being compiled, add a second test to make sure the overflow
1694           // function is called only once every overflow_frequency.
1695           const int overflow_frequency = 1024;
1696           __ andl(rbx, overflow_frequency - 1);
1697           __ jcc(Assembler::zero, backedge_counter_overflow);
1698 
1699         }
1700       } else {
1701         if (UseOnStackReplacement) {
1702           // check for overflow against eax, which is the sum of the
1703           // counters
1704           __ cmp32(rax,
1705                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1706           __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1707 
1708         }
1709       }
1710     }
1711     __ bind(dispatch);
1712   }
1713 
1714   // Pre-load the next target bytecode into rbx
1715   __ load_unsigned_byte(rbx, Address(r13, 0));
1716 
1717   // continue with the bytecode @ target
1718   // eax: return bci for jsr's, unused otherwise
1719   // ebx: target bytecode
1720   // r13: target bcp
1721   __ dispatch_only(vtos);
1722 
1723   if (UseLoopCounter) {
1724     if (ProfileInterpreter) {
1725       // Out-of-line code to allocate method data oop.
1726       __ bind(profile_method);
1727       __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::profile_method));
1728       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1729       __ set_method_data_pointer_for_bcp();
1730       __ jmp(dispatch);
1731     }
1732 
1733     if (UseOnStackReplacement) {
1734       // invocation counter overflow
1735       __ bind(backedge_counter_overflow);
1736       __ negptr(rdx);
1737       __ addptr(rdx, r13); // branch bcp
1738       // IcoResult frequency_counter_overflow([JavaThread*], address branch_bcp)
1739       __ call_VM(noreg,
1740                  CAST_FROM_FN_PTR(address,
1741                                   InterpreterRuntime::frequency_counter_overflow),
1742                  rdx);
1743       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1744 
1745       // rax: osr nmethod (osr ok) or NULL (osr not possible)
1746       // ebx: target bytecode
1747       // rdx: scratch
1748       // r14: locals pointer
1749       // r13: bcp
1750       __ testptr(rax, rax);                        // test result
1751       __ jcc(Assembler::zero, dispatch);         // no osr if null
1752       // nmethod may have been invalidated (VM may block upon call_VM return)
1753       __ movl(rcx, Address(rax, nmethod::entry_bci_offset()));
1754       __ cmpl(rcx, InvalidOSREntryBci);
1755       __ jcc(Assembler::equal, dispatch);
1756 
1757       // We have the address of an on stack replacement routine in eax
1758       // We need to prepare to execute the OSR method. First we must
1759       // migrate the locals and monitors off of the stack.
1760 
1761       __ mov(r13, rax);                             // save the nmethod
1762 
1763       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1764 
1765       // eax is OSR buffer, move it to expected parameter location
1766       __ mov(j_rarg0, rax);
1767 
1768       // We use j_rarg definitions here so that registers don't conflict as parameter
1769       // registers change across platforms as we are in the midst of a calling
1770       // sequence to the OSR nmethod and we don't want collision. These are NOT parameters.
1771 
1772       const Register retaddr = j_rarg2;
1773       const Register sender_sp = j_rarg1;
1774 
1775       // pop the interpreter frame
1776       __ movptr(sender_sp, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1777       __ leave();                                // remove frame anchor
1778       __ pop(retaddr);                           // get return address
1779       __ mov(rsp, sender_sp);                   // set sp to sender sp
1780       // Ensure compiled code always sees stack at proper alignment
1781       __ andptr(rsp, -(StackAlignmentInBytes));
1782 
1783       // unlike x86 we need no specialized return from compiled code
1784       // to the interpreter or the call stub.
1785 
1786       // push the return address
1787       __ push(retaddr);
1788 
1789       // and begin the OSR nmethod
1790       __ jmp(Address(r13, nmethod::osr_entry_point_offset()));
1791     }
1792   }
1793 }
1794 
1795 
1796 void TemplateTable::if_0cmp(Condition cc) {
1797   transition(itos, vtos);
1798   // assume branch is more often taken than not (loops use backward branches)
1799   Label not_taken;
1800   __ testl(rax, rax);
1801   __ jcc(j_not(cc), not_taken);
1802   branch(false, false);
1803   __ bind(not_taken);
1804   __ profile_not_taken_branch(rax);
1805 }
1806 
1807 void TemplateTable::if_icmp(Condition cc) {
1808   transition(itos, vtos);
1809   // assume branch is more often taken than not (loops use backward branches)
1810   Label not_taken;
1811   __ pop_i(rdx);
1812   __ cmpl(rdx, rax);
1813   __ jcc(j_not(cc), not_taken);
1814   branch(false, false);
1815   __ bind(not_taken);
1816   __ profile_not_taken_branch(rax);
1817 }
1818 
1819 void TemplateTable::if_nullcmp(Condition cc) {
1820   transition(atos, vtos);
1821   // assume branch is more often taken than not (loops use backward branches)
1822   Label not_taken;
1823   __ testptr(rax, rax);
1824   __ jcc(j_not(cc), not_taken);
1825   branch(false, false);
1826   __ bind(not_taken);
1827   __ profile_not_taken_branch(rax);
1828 }
1829 
1830 void TemplateTable::if_acmp(Condition cc) {
1831   transition(atos, vtos);
1832   // assume branch is more often taken than not (loops use backward branches)
1833   Label not_taken;
1834   __ pop_ptr(rdx);
1835   __ cmpptr(rdx, rax);
1836   __ jcc(j_not(cc), not_taken);
1837   branch(false, false);
1838   __ bind(not_taken);
1839   __ profile_not_taken_branch(rax);
1840 }
1841 
1842 void TemplateTable::ret() {
1843   transition(vtos, vtos);
1844   locals_index(rbx);
1845   __ movslq(rbx, iaddress(rbx)); // get return bci, compute return bcp
1846   __ profile_ret(rbx, rcx);
1847   __ get_method(rax);
1848   __ movptr(r13, Address(rax, Method::const_offset()));
1849   __ lea(r13, Address(r13, rbx, Address::times_1,
1850                       ConstMethod::codes_offset()));
1851   __ dispatch_next(vtos);
1852 }
1853 
1854 void TemplateTable::wide_ret() {
1855   transition(vtos, vtos);
1856   locals_index_wide(rbx);
1857   __ movptr(rbx, aaddress(rbx)); // get return bci, compute return bcp
1858   __ profile_ret(rbx, rcx);
1859   __ get_method(rax);
1860   __ movptr(r13, Address(rax, Method::const_offset()));
1861   __ lea(r13, Address(r13, rbx, Address::times_1, ConstMethod::codes_offset()));
1862   __ dispatch_next(vtos);
1863 }
1864 
1865 void TemplateTable::tableswitch() {
1866   Label default_case, continue_execution;
1867   transition(itos, vtos);
1868   // align r13
1869   __ lea(rbx, at_bcp(BytesPerInt));
1870   __ andptr(rbx, -BytesPerInt);
1871   // load lo & hi
1872   __ movl(rcx, Address(rbx, BytesPerInt));
1873   __ movl(rdx, Address(rbx, 2 * BytesPerInt));
1874   __ bswapl(rcx);
1875   __ bswapl(rdx);
1876   // check against lo & hi
1877   __ cmpl(rax, rcx);
1878   __ jcc(Assembler::less, default_case);
1879   __ cmpl(rax, rdx);
1880   __ jcc(Assembler::greater, default_case);
1881   // lookup dispatch offset
1882   __ subl(rax, rcx);
1883   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1884   __ profile_switch_case(rax, rbx, rcx);
1885   // continue execution
1886   __ bind(continue_execution);
1887   __ bswapl(rdx);
1888   __ movl2ptr(rdx, rdx);
1889   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1890   __ addptr(r13, rdx);
1891   __ dispatch_only(vtos);
1892   // handle default
1893   __ bind(default_case);
1894   __ profile_switch_default(rax);
1895   __ movl(rdx, Address(rbx, 0));
1896   __ jmp(continue_execution);
1897 }
1898 
1899 void TemplateTable::lookupswitch() {
1900   transition(itos, itos);
1901   __ stop("lookupswitch bytecode should have been rewritten");
1902 }
1903 
1904 void TemplateTable::fast_linearswitch() {
1905   transition(itos, vtos);
1906   Label loop_entry, loop, found, continue_execution;
1907   // bswap rax so we can avoid bswapping the table entries
1908   __ bswapl(rax);
1909   // align r13
1910   __ lea(rbx, at_bcp(BytesPerInt)); // btw: should be able to get rid of
1911                                     // this instruction (change offsets
1912                                     // below)
1913   __ andptr(rbx, -BytesPerInt);
1914   // set counter
1915   __ movl(rcx, Address(rbx, BytesPerInt));
1916   __ bswapl(rcx);
1917   __ jmpb(loop_entry);
1918   // table search
1919   __ bind(loop);
1920   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * BytesPerInt));
1921   __ jcc(Assembler::equal, found);
1922   __ bind(loop_entry);
1923   __ decrementl(rcx);
1924   __ jcc(Assembler::greaterEqual, loop);
1925   // default case
1926   __ profile_switch_default(rax);
1927   __ movl(rdx, Address(rbx, 0));
1928   __ jmp(continue_execution);
1929   // entry found -> get offset
1930   __ bind(found);
1931   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * BytesPerInt));
1932   __ profile_switch_case(rcx, rax, rbx);
1933   // continue execution
1934   __ bind(continue_execution);
1935   __ bswapl(rdx);
1936   __ movl2ptr(rdx, rdx);
1937   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1938   __ addptr(r13, rdx);
1939   __ dispatch_only(vtos);
1940 }
1941 
1942 void TemplateTable::fast_binaryswitch() {
1943   transition(itos, vtos);
1944   // Implementation using the following core algorithm:
1945   //
1946   // int binary_search(int key, LookupswitchPair* array, int n) {
1947   //   // Binary search according to "Methodik des Programmierens" by
1948   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1949   //   int i = 0;
1950   //   int j = n;
1951   //   while (i+1 < j) {
1952   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1953   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1954   //     // where a stands for the array and assuming that the (inexisting)
1955   //     // element a[n] is infinitely big.
1956   //     int h = (i + j) >> 1;
1957   //     // i < h < j
1958   //     if (key < array[h].fast_match()) {
1959   //       j = h;
1960   //     } else {
1961   //       i = h;
1962   //     }
1963   //   }
1964   //   // R: a[i] <= key < a[i+1] or Q
1965   //   // (i.e., if key is within array, i is the correct index)
1966   //   return i;
1967   // }
1968 
1969   // Register allocation
1970   const Register key   = rax; // already set (tosca)
1971   const Register array = rbx;
1972   const Register i     = rcx;
1973   const Register j     = rdx;
1974   const Register h     = rdi;
1975   const Register temp  = rsi;
1976 
1977   // Find array start
1978   __ lea(array, at_bcp(3 * BytesPerInt)); // btw: should be able to
1979                                           // get rid of this
1980                                           // instruction (change
1981                                           // offsets below)
1982   __ andptr(array, -BytesPerInt);
1983 
1984   // Initialize i & j
1985   __ xorl(i, i);                            // i = 0;
1986   __ movl(j, Address(array, -BytesPerInt)); // j = length(array);
1987 
1988   // Convert j into native byteordering
1989   __ bswapl(j);
1990 
1991   // And start
1992   Label entry;
1993   __ jmp(entry);
1994 
1995   // binary search loop
1996   {
1997     Label loop;
1998     __ bind(loop);
1999     // int h = (i + j) >> 1;
2000     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
2001     __ sarl(h, 1);                               // h = (i + j) >> 1;
2002     // if (key < array[h].fast_match()) {
2003     //   j = h;
2004     // } else {
2005     //   i = h;
2006     // }
2007     // Convert array[h].match to native byte-ordering before compare
2008     __ movl(temp, Address(array, h, Address::times_8));
2009     __ bswapl(temp);
2010     __ cmpl(key, temp);
2011     // j = h if (key <  array[h].fast_match())
2012     __ cmovl(Assembler::less, j, h);
2013     // i = h if (key >= array[h].fast_match())
2014     __ cmovl(Assembler::greaterEqual, i, h);
2015     // while (i+1 < j)
2016     __ bind(entry);
2017     __ leal(h, Address(i, 1)); // i+1
2018     __ cmpl(h, j);             // i+1 < j
2019     __ jcc(Assembler::less, loop);
2020   }
2021 
2022   // end of binary search, result index is i (must check again!)
2023   Label default_case;
2024   // Convert array[i].match to native byte-ordering before compare
2025   __ movl(temp, Address(array, i, Address::times_8));
2026   __ bswapl(temp);
2027   __ cmpl(key, temp);
2028   __ jcc(Assembler::notEqual, default_case);
2029 
2030   // entry found -> j = offset
2031   __ movl(j , Address(array, i, Address::times_8, BytesPerInt));
2032   __ profile_switch_case(i, key, array);
2033   __ bswapl(j);
2034   __ movl2ptr(j, j);
2035   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
2036   __ addptr(r13, j);
2037   __ dispatch_only(vtos);
2038 
2039   // default case -> j = default offset
2040   __ bind(default_case);
2041   __ profile_switch_default(i);
2042   __ movl(j, Address(array, -2 * BytesPerInt));
2043   __ bswapl(j);
2044   __ movl2ptr(j, j);
2045   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
2046   __ addptr(r13, j);
2047   __ dispatch_only(vtos);
2048 }
2049 
2050 
2051 void TemplateTable::_return(TosState state) {
2052   transition(state, state);
2053   assert(_desc->calls_vm(),
2054          "inconsistent calls_vm information"); // call in remove_activation
2055 
2056   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2057     assert(state == vtos, "only valid state");
2058     __ movptr(c_rarg1, aaddress(0));
2059     __ load_klass(rdi, c_rarg1);
2060     __ movl(rdi, Address(rdi, Klass::access_flags_offset()));
2061     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2062     Label skip_register_finalizer;
2063     __ jcc(Assembler::zero, skip_register_finalizer);
2064 
2065     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), c_rarg1);
2066 
2067     __ bind(skip_register_finalizer);
2068   }
2069 
2070   __ remove_activation(state, r13);
2071   __ jmp(r13);
2072 }
2073 
2074 // ----------------------------------------------------------------------------
2075 // Volatile variables demand their effects be made known to all CPU's
2076 // in order.  Store buffers on most chips allow reads & writes to
2077 // reorder; the JMM's ReadAfterWrite.java test fails in -Xint mode
2078 // without some kind of memory barrier (i.e., it's not sufficient that
2079 // the interpreter does not reorder volatile references, the hardware
2080 // also must not reorder them).
2081 //
2082 // According to the new Java Memory Model (JMM):
2083 // (1) All volatiles are serialized wrt to each other.  ALSO reads &
2084 //     writes act as aquire & release, so:
2085 // (2) A read cannot let unrelated NON-volatile memory refs that
2086 //     happen after the read float up to before the read.  It's OK for
2087 //     non-volatile memory refs that happen before the volatile read to
2088 //     float down below it.
2089 // (3) Similar a volatile write cannot let unrelated NON-volatile
2090 //     memory refs that happen BEFORE the write float down to after the
2091 //     write.  It's OK for non-volatile memory refs that happen after the
2092 //     volatile write to float up before it.
2093 //
2094 // We only put in barriers around volatile refs (they are expensive),
2095 // not _between_ memory refs (that would require us to track the
2096 // flavor of the previous memory refs).  Requirements (2) and (3)
2097 // require some barriers before volatile stores and after volatile
2098 // loads.  These nearly cover requirement (1) but miss the
2099 // volatile-store-volatile-load case.  This final case is placed after
2100 // volatile-stores although it could just as well go before
2101 // volatile-loads.
2102 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits
2103                                      order_constraint) {
2104   // Helper function to insert a is-volatile test and memory barrier
2105   if (os::is_MP()) { // Not needed on single CPU
2106     __ membar(order_constraint);
2107   }
2108 }
2109 
2110 void TemplateTable::resolve_cache_and_index(int byte_no,
2111                                             Register Rcache,
2112                                             Register index,
2113                                             size_t index_size) {
2114   const Register temp = rbx;
2115   assert_different_registers(Rcache, index, temp);
2116 
2117   Label resolved;
2118     assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
2119     __ get_cache_and_index_and_bytecode_at_bcp(Rcache, index, temp, byte_no, 1, index_size);
2120     __ cmpl(temp, (int) bytecode());  // have we resolved this bytecode?
2121     __ jcc(Assembler::equal, resolved);
2122 
2123   // resolve first time through
2124   address entry;
2125   switch (bytecode()) {
2126   case Bytecodes::_getstatic:
2127   case Bytecodes::_putstatic:
2128   case Bytecodes::_getfield:
2129   case Bytecodes::_putfield:
2130     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);
2131     break;
2132   case Bytecodes::_invokevirtual:
2133   case Bytecodes::_invokespecial:
2134   case Bytecodes::_invokestatic:
2135   case Bytecodes::_invokeinterface:
2136     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);
2137     break;
2138   case Bytecodes::_invokehandle:
2139     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokehandle);
2140     break;
2141   case Bytecodes::_invokedynamic:
2142     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);
2143     break;
2144   default:
2145     fatal(err_msg("unexpected bytecode: %s", Bytecodes::name(bytecode())));
2146     break;
2147   }
2148   __ movl(temp, (int) bytecode());
2149   __ call_VM(noreg, entry, temp);
2150 
2151   // Update registers with resolved info
2152   __ get_cache_and_index_at_bcp(Rcache, index, 1, index_size);
2153   __ bind(resolved);
2154 }
2155 
2156 // The cache and index registers must be set before call
2157 void TemplateTable::load_field_cp_cache_entry(Register obj,
2158                                               Register cache,
2159                                               Register index,
2160                                               Register off,
2161                                               Register flags,
2162                                               bool is_static = false) {
2163   assert_different_registers(cache, index, flags, off);
2164 
2165   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2166   // Field offset
2167   __ movptr(off, Address(cache, index, Address::times_ptr,
2168                          in_bytes(cp_base_offset +
2169                                   ConstantPoolCacheEntry::f2_offset())));
2170   // Flags
2171   __ movl(flags, Address(cache, index, Address::times_ptr,
2172                          in_bytes(cp_base_offset +
2173                                   ConstantPoolCacheEntry::flags_offset())));
2174 
2175   // klass overwrite register
2176   if (is_static) {
2177     __ movptr(obj, Address(cache, index, Address::times_ptr,
2178                            in_bytes(cp_base_offset +
2179                                     ConstantPoolCacheEntry::f1_offset())));
2180     const int mirror_offset = in_bytes(Klass::java_mirror_offset());
2181     __ movptr(obj, Address(obj, mirror_offset));
2182   }
2183 }
2184 
2185 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2186                                                Register method,
2187                                                Register itable_index,
2188                                                Register flags,
2189                                                bool is_invokevirtual,
2190                                                bool is_invokevfinal, /*unused*/
2191                                                bool is_invokedynamic) {
2192   // setup registers
2193   const Register cache = rcx;
2194   const Register index = rdx;
2195   assert_different_registers(method, flags);
2196   assert_different_registers(method, cache, index);
2197   assert_different_registers(itable_index, flags);
2198   assert_different_registers(itable_index, cache, index);
2199   // determine constant pool cache field offsets
2200   assert(is_invokevirtual == (byte_no == f2_byte), "is_invokevirtual flag redundant");
2201   const int method_offset = in_bytes(
2202     ConstantPoolCache::base_offset() +
2203       ((byte_no == f2_byte)
2204        ? ConstantPoolCacheEntry::f2_offset()
2205        : ConstantPoolCacheEntry::f1_offset()));
2206   const int flags_offset = in_bytes(ConstantPoolCache::base_offset() +
2207                                     ConstantPoolCacheEntry::flags_offset());
2208   // access constant pool cache fields
2209   const int index_offset = in_bytes(ConstantPoolCache::base_offset() +
2210                                     ConstantPoolCacheEntry::f2_offset());
2211 
2212   size_t index_size = (is_invokedynamic ? sizeof(u4) : sizeof(u2));
2213   resolve_cache_and_index(byte_no, cache, index, index_size);
2214     __ movptr(method, Address(cache, index, Address::times_ptr, method_offset));
2215 
2216   if (itable_index != noreg) {
2217     // pick up itable or appendix index from f2 also:
2218     __ movptr(itable_index, Address(cache, index, Address::times_ptr, index_offset));
2219   }
2220   __ movl(flags, Address(cache, index, Address::times_ptr, flags_offset));
2221 }
2222 
2223 // Correct values of the cache and index registers are preserved.
2224 void TemplateTable::jvmti_post_field_access(Register cache, Register index,
2225                                             bool is_static, bool has_tos) {
2226   // do the JVMTI work here to avoid disturbing the register state below
2227   // We use c_rarg registers here because we want to use the register used in
2228   // the call to the VM
2229   if (JvmtiExport::can_post_field_access()) {
2230     // Check to see if a field access watch has been set before we
2231     // take the time to call into the VM.
2232     Label L1;
2233     assert_different_registers(cache, index, rax);
2234     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2235     __ testl(rax, rax);
2236     __ jcc(Assembler::zero, L1);
2237 
2238     __ get_cache_and_index_at_bcp(c_rarg2, c_rarg3, 1);
2239 
2240     // cache entry pointer
2241     __ addptr(c_rarg2, in_bytes(ConstantPoolCache::base_offset()));
2242     __ shll(c_rarg3, LogBytesPerWord);
2243     __ addptr(c_rarg2, c_rarg3);
2244     if (is_static) {
2245       __ xorl(c_rarg1, c_rarg1); // NULL object reference
2246     } else {
2247       __ movptr(c_rarg1, at_tos()); // get object pointer without popping it
2248       __ verify_oop(c_rarg1);
2249     }
2250     // c_rarg1: object pointer or NULL
2251     // c_rarg2: cache entry pointer
2252     // c_rarg3: jvalue object on the stack
2253     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
2254                                        InterpreterRuntime::post_field_access),
2255                c_rarg1, c_rarg2, c_rarg3);
2256     __ get_cache_and_index_at_bcp(cache, index, 1);
2257     __ bind(L1);
2258   }
2259 }
2260 
2261 void TemplateTable::pop_and_check_object(Register r) {
2262   __ pop_ptr(r);
2263   __ null_check(r);  // for field access must check obj.
2264   __ verify_oop(r);
2265 }
2266 
2267 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2268   transition(vtos, vtos);
2269 
2270   const Register cache = rcx;
2271   const Register index = rdx;
2272   const Register obj   = c_rarg3;
2273   const Register off   = rbx;
2274   const Register flags = rax;
2275   const Register bc = c_rarg3; // uses same reg as obj, so don't mix them
2276 
2277   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2278   jvmti_post_field_access(cache, index, is_static, false);
2279   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2280 
2281   if (!is_static) {
2282     // obj is on the stack
2283     pop_and_check_object(obj);
2284   }
2285 
2286   const Address field(obj, off, Address::times_1);
2287 
2288   Label Done, notByte, notInt, notShort, notChar,
2289               notLong, notFloat, notObj, notDouble;
2290 
2291   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2292   // Make sure we don't need to mask edx after the above shift
2293   assert(btos == 0, "change code, btos != 0");
2294 
2295   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2296   __ jcc(Assembler::notZero, notByte);
2297   // btos
2298   __ load_signed_byte(rax, field);
2299   __ push(btos);
2300   // Rewrite bytecode to be faster
2301   if (!is_static) {
2302     patch_bytecode(Bytecodes::_fast_bgetfield, bc, rbx);
2303   }
2304   __ jmp(Done);
2305 
2306   __ bind(notByte);
2307   __ cmpl(flags, atos);
2308   __ jcc(Assembler::notEqual, notObj);
2309   // atos
2310   __ load_heap_oop(rax, field);
2311   __ push(atos);
2312   if (!is_static) {
2313     patch_bytecode(Bytecodes::_fast_agetfield, bc, rbx);
2314   }
2315   __ jmp(Done);
2316 
2317   __ bind(notObj);
2318   __ cmpl(flags, itos);
2319   __ jcc(Assembler::notEqual, notInt);
2320   // itos
2321   __ movl(rax, field);
2322   __ push(itos);
2323   // Rewrite bytecode to be faster
2324   if (!is_static) {
2325     patch_bytecode(Bytecodes::_fast_igetfield, bc, rbx);
2326   }
2327   __ jmp(Done);
2328 
2329   __ bind(notInt);
2330   __ cmpl(flags, ctos);
2331   __ jcc(Assembler::notEqual, notChar);
2332   // ctos
2333   __ load_unsigned_short(rax, field);
2334   __ push(ctos);
2335   // Rewrite bytecode to be faster
2336   if (!is_static) {
2337     patch_bytecode(Bytecodes::_fast_cgetfield, bc, rbx);
2338   }
2339   __ jmp(Done);
2340 
2341   __ bind(notChar);
2342   __ cmpl(flags, stos);
2343   __ jcc(Assembler::notEqual, notShort);
2344   // stos
2345   __ load_signed_short(rax, field);
2346   __ push(stos);
2347   // Rewrite bytecode to be faster
2348   if (!is_static) {
2349     patch_bytecode(Bytecodes::_fast_sgetfield, bc, rbx);
2350   }
2351   __ jmp(Done);
2352 
2353   __ bind(notShort);
2354   __ cmpl(flags, ltos);
2355   __ jcc(Assembler::notEqual, notLong);
2356   // ltos
2357   __ movq(rax, field);
2358   __ push(ltos);
2359   // Rewrite bytecode to be faster
2360   if (!is_static) {
2361     patch_bytecode(Bytecodes::_fast_lgetfield, bc, rbx);
2362   }
2363   __ jmp(Done);
2364 
2365   __ bind(notLong);
2366   __ cmpl(flags, ftos);
2367   __ jcc(Assembler::notEqual, notFloat);
2368   // ftos
2369   __ movflt(xmm0, field);
2370   __ push(ftos);
2371   // Rewrite bytecode to be faster
2372   if (!is_static) {
2373     patch_bytecode(Bytecodes::_fast_fgetfield, bc, rbx);
2374   }
2375   __ jmp(Done);
2376 
2377   __ bind(notFloat);
2378 #ifdef ASSERT
2379   __ cmpl(flags, dtos);
2380   __ jcc(Assembler::notEqual, notDouble);
2381 #endif
2382   // dtos
2383   __ movdbl(xmm0, field);
2384   __ push(dtos);
2385   // Rewrite bytecode to be faster
2386   if (!is_static) {
2387     patch_bytecode(Bytecodes::_fast_dgetfield, bc, rbx);
2388   }
2389 #ifdef ASSERT
2390   __ jmp(Done);
2391 
2392   __ bind(notDouble);
2393   __ stop("Bad state");
2394 #endif
2395 
2396   __ bind(Done);
2397   // [jk] not needed currently
2398   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadLoad |
2399   //                                              Assembler::LoadStore));
2400 }
2401 
2402 
2403 void TemplateTable::getfield(int byte_no) {
2404   getfield_or_static(byte_no, false);
2405 }
2406 
2407 void TemplateTable::getstatic(int byte_no) {
2408   getfield_or_static(byte_no, true);
2409 }
2410 
2411 // The registers cache and index expected to be set before call.
2412 // The function may destroy various registers, just not the cache and index registers.
2413 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2414   transition(vtos, vtos);
2415 
2416   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2417 
2418   if (JvmtiExport::can_post_field_modification()) {
2419     // Check to see if a field modification watch has been set before
2420     // we take the time to call into the VM.
2421     Label L1;
2422     assert_different_registers(cache, index, rax);
2423     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2424     __ testl(rax, rax);
2425     __ jcc(Assembler::zero, L1);
2426 
2427     __ get_cache_and_index_at_bcp(c_rarg2, rscratch1, 1);
2428 
2429     if (is_static) {
2430       // Life is simple.  Null out the object pointer.
2431       __ xorl(c_rarg1, c_rarg1);
2432     } else {
2433       // Life is harder. The stack holds the value on top, followed by
2434       // the object.  We don't know the size of the value, though; it
2435       // could be one or two words depending on its type. As a result,
2436       // we must find the type to determine where the object is.
2437       __ movl(c_rarg3, Address(c_rarg2, rscratch1,
2438                            Address::times_8,
2439                            in_bytes(cp_base_offset +
2440                                      ConstantPoolCacheEntry::flags_offset())));
2441       __ shrl(c_rarg3, ConstantPoolCacheEntry::tos_state_shift);
2442       // Make sure we don't need to mask rcx after the above shift
2443       ConstantPoolCacheEntry::verify_tos_state_shift();
2444       __ movptr(c_rarg1, at_tos_p1());  // initially assume a one word jvalue
2445       __ cmpl(c_rarg3, ltos);
2446       __ cmovptr(Assembler::equal,
2447                  c_rarg1, at_tos_p2()); // ltos (two word jvalue)
2448       __ cmpl(c_rarg3, dtos);
2449       __ cmovptr(Assembler::equal,
2450                  c_rarg1, at_tos_p2()); // dtos (two word jvalue)
2451     }
2452     // cache entry pointer
2453     __ addptr(c_rarg2, in_bytes(cp_base_offset));
2454     __ shll(rscratch1, LogBytesPerWord);
2455     __ addptr(c_rarg2, rscratch1);
2456     // object (tos)
2457     __ mov(c_rarg3, rsp);
2458     // c_rarg1: object pointer set up above (NULL if static)
2459     // c_rarg2: cache entry pointer
2460     // c_rarg3: jvalue object on the stack
2461     __ call_VM(noreg,
2462                CAST_FROM_FN_PTR(address,
2463                                 InterpreterRuntime::post_field_modification),
2464                c_rarg1, c_rarg2, c_rarg3);
2465     __ get_cache_and_index_at_bcp(cache, index, 1);
2466     __ bind(L1);
2467   }
2468 }
2469 
2470 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2471   transition(vtos, vtos);
2472 
2473   const Register cache = rcx;
2474   const Register index = rdx;
2475   const Register obj   = rcx;
2476   const Register off   = rbx;
2477   const Register flags = rax;
2478   const Register bc    = c_rarg3;
2479 
2480   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2481   jvmti_post_field_mod(cache, index, is_static);
2482 #ifdef ASSERT
2483   verify_stable(cache, index, is_static, false);
2484 #endif //ASSERT
2485   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2486 
2487   // [jk] not needed currently
2488   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2489   //                                              Assembler::StoreStore));
2490 
2491   Label notVolatile, Done;
2492   __ movl(rdx, flags);
2493   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2494   __ andl(rdx, 0x1);
2495 
2496   // field address
2497   const Address field(obj, off, Address::times_1);
2498 
2499   Label notByte, notInt, notShort, notChar,
2500         notLong, notFloat, notObj, notDouble;
2501 
2502   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2503 
2504   assert(btos == 0, "change code, btos != 0");
2505   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2506   __ jcc(Assembler::notZero, notByte);
2507 
2508   // btos
2509   {
2510     __ pop(btos);
2511     if (!is_static) pop_and_check_object(obj);
2512     __ movb(field, rax);
2513     if (!is_static) {
2514       patch_bytecode(Bytecodes::_fast_bputfield, bc, rbx, true, byte_no);
2515     }
2516     __ jmp(Done);
2517   }
2518 
2519   __ bind(notByte);
2520   __ cmpl(flags, atos);
2521   __ jcc(Assembler::notEqual, notObj);
2522 
2523   // atos
2524   {
2525     __ pop(atos);
2526     if (!is_static) pop_and_check_object(obj);
2527     // Store into the field
2528     do_oop_store(_masm, field, rax, _bs->kind(), false);
2529     if (!is_static) {
2530       patch_bytecode(Bytecodes::_fast_aputfield, bc, rbx, true, byte_no);
2531     }
2532     __ jmp(Done);
2533   }
2534 
2535   __ bind(notObj);
2536   __ cmpl(flags, itos);
2537   __ jcc(Assembler::notEqual, notInt);
2538 
2539   // itos
2540   {
2541     __ pop(itos);
2542     if (!is_static) pop_and_check_object(obj);
2543     __ movl(field, rax);
2544     if (!is_static) {
2545       patch_bytecode(Bytecodes::_fast_iputfield, bc, rbx, true, byte_no);
2546     }
2547     __ jmp(Done);
2548   }
2549 
2550   __ bind(notInt);
2551   __ cmpl(flags, ctos);
2552   __ jcc(Assembler::notEqual, notChar);
2553 
2554   // ctos
2555   {
2556     __ pop(ctos);
2557     if (!is_static) pop_and_check_object(obj);
2558     __ movw(field, rax);
2559     if (!is_static) {
2560       patch_bytecode(Bytecodes::_fast_cputfield, bc, rbx, true, byte_no);
2561     }
2562     __ jmp(Done);
2563   }
2564 
2565   __ bind(notChar);
2566   __ cmpl(flags, stos);
2567   __ jcc(Assembler::notEqual, notShort);
2568 
2569   // stos
2570   {
2571     __ pop(stos);
2572     if (!is_static) pop_and_check_object(obj);
2573     __ movw(field, rax);
2574     if (!is_static) {
2575       patch_bytecode(Bytecodes::_fast_sputfield, bc, rbx, true, byte_no);
2576     }
2577     __ jmp(Done);
2578   }
2579 
2580   __ bind(notShort);
2581   __ cmpl(flags, ltos);
2582   __ jcc(Assembler::notEqual, notLong);
2583 
2584   // ltos
2585   {
2586     __ pop(ltos);
2587     if (!is_static) pop_and_check_object(obj);
2588     __ movq(field, rax);
2589     if (!is_static) {
2590       patch_bytecode(Bytecodes::_fast_lputfield, bc, rbx, true, byte_no);
2591     }
2592     __ jmp(Done);
2593   }
2594 
2595   __ bind(notLong);
2596   __ cmpl(flags, ftos);
2597   __ jcc(Assembler::notEqual, notFloat);
2598 
2599   // ftos
2600   {
2601     __ pop(ftos);
2602     if (!is_static) pop_and_check_object(obj);
2603     __ movflt(field, xmm0);
2604     if (!is_static) {
2605       patch_bytecode(Bytecodes::_fast_fputfield, bc, rbx, true, byte_no);
2606     }
2607     __ jmp(Done);
2608   }
2609 
2610   __ bind(notFloat);
2611 #ifdef ASSERT
2612   __ cmpl(flags, dtos);
2613   __ jcc(Assembler::notEqual, notDouble);
2614 #endif
2615 
2616   // dtos
2617   {
2618     __ pop(dtos);
2619     if (!is_static) pop_and_check_object(obj);
2620     __ movdbl(field, xmm0);
2621     if (!is_static) {
2622       patch_bytecode(Bytecodes::_fast_dputfield, bc, rbx, true, byte_no);
2623     }
2624   }
2625 
2626 #ifdef ASSERT
2627   __ jmp(Done);
2628 
2629   __ bind(notDouble);
2630   __ stop("Bad state");
2631 #endif
2632 
2633   __ bind(Done);
2634 
2635   // Check for volatile store
2636   __ testl(rdx, rdx);
2637   __ jcc(Assembler::zero, notVolatile);
2638   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2639                                                Assembler::StoreStore));
2640   __ bind(notVolatile);
2641 }
2642 
2643 void TemplateTable::putfield(int byte_no) {
2644   putfield_or_static(byte_no, false);
2645 }
2646 
2647 void TemplateTable::putstatic(int byte_no) {
2648   putfield_or_static(byte_no, true);
2649 }
2650 
2651 void TemplateTable::jvmti_post_fast_field_mod() {
2652   if (JvmtiExport::can_post_field_modification()) {
2653     // Check to see if a field modification watch has been set before
2654     // we take the time to call into the VM.
2655     Label L2;
2656     __ mov32(c_rarg3, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2657     __ testl(c_rarg3, c_rarg3);
2658     __ jcc(Assembler::zero, L2);
2659     __ pop_ptr(rbx);                  // copy the object pointer from tos
2660     __ verify_oop(rbx);
2661     __ push_ptr(rbx);                 // put the object pointer back on tos
2662     // Save tos values before call_VM() clobbers them. Since we have
2663     // to do it for every data type, we use the saved values as the
2664     // jvalue object.
2665     switch (bytecode()) {          // load values into the jvalue object
2666     case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
2667     case Bytecodes::_fast_bputfield: // fall through
2668     case Bytecodes::_fast_sputfield: // fall through
2669     case Bytecodes::_fast_cputfield: // fall through
2670     case Bytecodes::_fast_iputfield: __ push_i(rax); break;
2671     case Bytecodes::_fast_dputfield: __ push_d(); break;
2672     case Bytecodes::_fast_fputfield: __ push_f(); break;
2673     case Bytecodes::_fast_lputfield: __ push_l(rax); break;
2674 
2675     default:
2676       ShouldNotReachHere();
2677     }
2678     __ mov(c_rarg3, rsp);             // points to jvalue on the stack
2679     // access constant pool cache entry
2680     __ get_cache_entry_pointer_at_bcp(c_rarg2, rax, 1);
2681     __ verify_oop(rbx);
2682     // rbx: object pointer copied above
2683     // c_rarg2: cache entry pointer
2684     // c_rarg3: jvalue object on the stack
2685     __ call_VM(noreg,
2686                CAST_FROM_FN_PTR(address,
2687                                 InterpreterRuntime::post_field_modification),
2688                rbx, c_rarg2, c_rarg3);
2689 
2690     switch (bytecode()) {             // restore tos values
2691     case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
2692     case Bytecodes::_fast_bputfield: // fall through
2693     case Bytecodes::_fast_sputfield: // fall through
2694     case Bytecodes::_fast_cputfield: // fall through
2695     case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
2696     case Bytecodes::_fast_dputfield: __ pop_d(); break;
2697     case Bytecodes::_fast_fputfield: __ pop_f(); break;
2698     case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
2699     }
2700     __ bind(L2);
2701   }
2702 }
2703 
2704 void TemplateTable::fast_storefield(TosState state) {
2705   transition(state, vtos);
2706 
2707   ByteSize base = ConstantPoolCache::base_offset();
2708 
2709   jvmti_post_fast_field_mod();
2710 
2711 #ifdef ASSERT
2712   verify_stable_fast();
2713 #endif // ASSERT
2714 
2715   // access constant pool cache
2716   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2717 
2718   // test for volatile with rdx
2719   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2720                        in_bytes(base +
2721                                 ConstantPoolCacheEntry::flags_offset())));
2722 
2723   // replace index with field offset from cache entry
2724   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2725                          in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2726 
2727   // [jk] not needed currently
2728   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2729   //                                              Assembler::StoreStore));
2730 
2731   Label notVolatile;
2732   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2733   __ andl(rdx, 0x1);
2734 
2735   // Get object from stack
2736   pop_and_check_object(rcx);
2737 
2738   // field address
2739   const Address field(rcx, rbx, Address::times_1);
2740 
2741   // access field
2742   switch (bytecode()) {
2743   case Bytecodes::_fast_aputfield:
2744     do_oop_store(_masm, field, rax, _bs->kind(), false);
2745     break;
2746   case Bytecodes::_fast_lputfield:
2747     __ movq(field, rax);
2748     break;
2749   case Bytecodes::_fast_iputfield:
2750     __ movl(field, rax);
2751     break;
2752   case Bytecodes::_fast_bputfield:
2753     __ movb(field, rax);
2754     break;
2755   case Bytecodes::_fast_sputfield:
2756     // fall through
2757   case Bytecodes::_fast_cputfield:
2758     __ movw(field, rax);
2759     break;
2760   case Bytecodes::_fast_fputfield:
2761     __ movflt(field, xmm0);
2762     break;
2763   case Bytecodes::_fast_dputfield:
2764     __ movdbl(field, xmm0);
2765     break;
2766   default:
2767     ShouldNotReachHere();
2768   }
2769 
2770   // Check for volatile store
2771   __ testl(rdx, rdx);
2772   __ jcc(Assembler::zero, notVolatile);
2773   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2774                                                Assembler::StoreStore));
2775   __ bind(notVolatile);
2776 }
2777 
2778 
2779 void TemplateTable::fast_accessfield(TosState state) {
2780   transition(atos, state);
2781 
2782   // Do the JVMTI work here to avoid disturbing the register state below
2783   if (JvmtiExport::can_post_field_access()) {
2784     // Check to see if a field access watch has been set before we
2785     // take the time to call into the VM.
2786     Label L1;
2787     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2788     __ testl(rcx, rcx);
2789     __ jcc(Assembler::zero, L1);
2790     // access constant pool cache entry
2791     __ get_cache_entry_pointer_at_bcp(c_rarg2, rcx, 1);
2792     __ verify_oop(rax);
2793     __ push_ptr(rax);  // save object pointer before call_VM() clobbers it
2794     __ mov(c_rarg1, rax);
2795     // c_rarg1: object pointer copied above
2796     // c_rarg2: cache entry pointer
2797     __ call_VM(noreg,
2798                CAST_FROM_FN_PTR(address,
2799                                 InterpreterRuntime::post_field_access),
2800                c_rarg1, c_rarg2);
2801     __ pop_ptr(rax); // restore object pointer
2802     __ bind(L1);
2803   }
2804 
2805   // access constant pool cache
2806   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2807   // replace index with field offset from cache entry
2808   // [jk] not needed currently
2809   // if (os::is_MP()) {
2810   //   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2811   //                        in_bytes(ConstantPoolCache::base_offset() +
2812   //                                 ConstantPoolCacheEntry::flags_offset())));
2813   //   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2814   //   __ andl(rdx, 0x1);
2815   // }
2816   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2817                          in_bytes(ConstantPoolCache::base_offset() +
2818                                   ConstantPoolCacheEntry::f2_offset())));
2819 
2820   // rax: object
2821   __ verify_oop(rax);
2822   __ null_check(rax);
2823   Address field(rax, rbx, Address::times_1);
2824 
2825   // access field
2826   switch (bytecode()) {
2827   case Bytecodes::_fast_agetfield:
2828     __ load_heap_oop(rax, field);
2829     __ verify_oop(rax);
2830     break;
2831   case Bytecodes::_fast_lgetfield:
2832     __ movq(rax, field);
2833     break;
2834   case Bytecodes::_fast_igetfield:
2835     __ movl(rax, field);
2836     break;
2837   case Bytecodes::_fast_bgetfield:
2838     __ movsbl(rax, field);
2839     break;
2840   case Bytecodes::_fast_sgetfield:
2841     __ load_signed_short(rax, field);
2842     break;
2843   case Bytecodes::_fast_cgetfield:
2844     __ load_unsigned_short(rax, field);
2845     break;
2846   case Bytecodes::_fast_fgetfield:
2847     __ movflt(xmm0, field);
2848     break;
2849   case Bytecodes::_fast_dgetfield:
2850     __ movdbl(xmm0, field);
2851     break;
2852   default:
2853     ShouldNotReachHere();
2854   }
2855   // [jk] not needed currently
2856   // if (os::is_MP()) {
2857   //   Label notVolatile;
2858   //   __ testl(rdx, rdx);
2859   //   __ jcc(Assembler::zero, notVolatile);
2860   //   __ membar(Assembler::LoadLoad);
2861   //   __ bind(notVolatile);
2862   //};
2863 }
2864 
2865 void TemplateTable::fast_xaccess(TosState state) {
2866   transition(vtos, state);
2867 
2868   // get receiver
2869   __ movptr(rax, aaddress(0));
2870   // access constant pool cache
2871   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2872   __ movptr(rbx,
2873             Address(rcx, rdx, Address::times_8,
2874                     in_bytes(ConstantPoolCache::base_offset() +
2875                              ConstantPoolCacheEntry::f2_offset())));
2876   // make sure exception is reported in correct bcp range (getfield is
2877   // next instruction)
2878   __ increment(r13);
2879   __ null_check(rax);
2880   switch (state) {
2881   case itos:
2882     __ movl(rax, Address(rax, rbx, Address::times_1));
2883     break;
2884   case atos:
2885     __ load_heap_oop(rax, Address(rax, rbx, Address::times_1));
2886     __ verify_oop(rax);
2887     break;
2888   case ftos:
2889     __ movflt(xmm0, Address(rax, rbx, Address::times_1));
2890     break;
2891   default:
2892     ShouldNotReachHere();
2893   }
2894 
2895   // [jk] not needed currently
2896   // if (os::is_MP()) {
2897   //   Label notVolatile;
2898   //   __ movl(rdx, Address(rcx, rdx, Address::times_8,
2899   //                        in_bytes(ConstantPoolCache::base_offset() +
2900   //                                 ConstantPoolCacheEntry::flags_offset())));
2901   //   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2902   //   __ testl(rdx, 0x1);
2903   //   __ jcc(Assembler::zero, notVolatile);
2904   //   __ membar(Assembler::LoadLoad);
2905   //   __ bind(notVolatile);
2906   // }
2907 
2908   __ decrement(r13);
2909 }
2910 
2911 
2912 
2913 //-----------------------------------------------------------------------------
2914 // Calls
2915 
2916 void TemplateTable::count_calls(Register method, Register temp) {
2917   // implemented elsewhere
2918   ShouldNotReachHere();
2919 }
2920 
2921 void TemplateTable::prepare_invoke(int byte_no,
2922                                    Register method,  // linked method (or i-klass)
2923                                    Register index,   // itable index, MethodType, etc.
2924                                    Register recv,    // if caller wants to see it
2925                                    Register flags    // if caller wants to test it
2926                                    ) {
2927   // determine flags
2928   const Bytecodes::Code code = bytecode();
2929   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
2930   const bool is_invokedynamic    = code == Bytecodes::_invokedynamic;
2931   const bool is_invokehandle     = code == Bytecodes::_invokehandle;
2932   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
2933   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
2934   const bool load_receiver       = (recv  != noreg);
2935   const bool save_flags          = (flags != noreg);
2936   assert(load_receiver == (code != Bytecodes::_invokestatic && code != Bytecodes::_invokedynamic), "");
2937   assert(save_flags    == (is_invokeinterface || is_invokevirtual), "need flags for vfinal");
2938   assert(flags == noreg || flags == rdx, "");
2939   assert(recv  == noreg || recv  == rcx, "");
2940 
2941   // setup registers & access constant pool cache
2942   if (recv  == noreg)  recv  = rcx;
2943   if (flags == noreg)  flags = rdx;
2944   assert_different_registers(method, index, recv, flags);
2945 
2946   // save 'interpreter return address'
2947   __ save_bcp();
2948 
2949   load_invoke_cp_cache_entry(byte_no, method, index, flags, is_invokevirtual, false, is_invokedynamic);
2950 
2951   // maybe push appendix to arguments (just before return address)
2952   if (is_invokedynamic || is_invokehandle) {
2953     Label L_no_push;
2954     __ testl(flags, (1 << ConstantPoolCacheEntry::has_appendix_shift));
2955     __ jcc(Assembler::zero, L_no_push);
2956     // Push the appendix as a trailing parameter.
2957     // This must be done before we get the receiver,
2958     // since the parameter_size includes it.
2959     __ push(rbx);
2960     __ mov(rbx, index);
2961     assert(ConstantPoolCacheEntry::_indy_resolved_references_appendix_offset == 0, "appendix expected at index+0");
2962     __ load_resolved_reference_at_index(index, rbx);
2963     __ pop(rbx);
2964     __ push(index);  // push appendix (MethodType, CallSite, etc.)
2965     __ bind(L_no_push);
2966   }
2967 
2968   // load receiver if needed (after appendix is pushed so parameter size is correct)
2969   // Note: no return address pushed yet
2970   if (load_receiver) {
2971     __ movl(recv, flags);
2972     __ andl(recv, ConstantPoolCacheEntry::parameter_size_mask);
2973     const int no_return_pc_pushed_yet = -1;  // argument slot correction before we push return address
2974     const int receiver_is_at_end      = -1;  // back off one slot to get receiver
2975     Address recv_addr = __ argument_address(recv, no_return_pc_pushed_yet + receiver_is_at_end);
2976     __ movptr(recv, recv_addr);
2977     __ verify_oop(recv);
2978   }
2979 
2980   if (save_flags) {
2981     __ movl(r13, flags);
2982   }
2983 
2984   // compute return type
2985   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2986   // Make sure we don't need to mask flags after the above shift
2987   ConstantPoolCacheEntry::verify_tos_state_shift();
2988   // load return address
2989   {
2990     const address table_addr = (is_invokeinterface || is_invokedynamic) ?
2991         (address)Interpreter::return_5_addrs_by_index_table() :
2992         (address)Interpreter::return_3_addrs_by_index_table();
2993     ExternalAddress table(table_addr);
2994     __ lea(rscratch1, table);
2995     __ movptr(flags, Address(rscratch1, flags, Address::times_ptr));
2996   }
2997 
2998   // push return address
2999   __ push(flags);
3000 
3001   // Restore flags value from the constant pool cache, and restore rsi
3002   // for later null checks.  r13 is the bytecode pointer
3003   if (save_flags) {
3004     __ movl(flags, r13);
3005     __ restore_bcp();
3006   }
3007 }
3008 
3009 
3010 void TemplateTable::invokevirtual_helper(Register index,
3011                                          Register recv,
3012                                          Register flags) {
3013   // Uses temporary registers rax, rdx
3014   assert_different_registers(index, recv, rax, rdx);
3015   assert(index == rbx, "");
3016   assert(recv  == rcx, "");
3017 
3018   // Test for an invoke of a final method
3019   Label notFinal;
3020   __ movl(rax, flags);
3021   __ andl(rax, (1 << ConstantPoolCacheEntry::is_vfinal_shift));
3022   __ jcc(Assembler::zero, notFinal);
3023 
3024   const Register method = index;  // method must be rbx
3025   assert(method == rbx,
3026          "Method* must be rbx for interpreter calling convention");
3027 
3028   // do the call - the index is actually the method to call
3029   // that is, f2 is a vtable index if !is_vfinal, else f2 is a Method*
3030 
3031   // It's final, need a null check here!
3032   __ null_check(recv);
3033 
3034   // profile this call
3035   __ profile_final_call(rax);
3036 
3037   __ jump_from_interpreted(method, rax);
3038 
3039   __ bind(notFinal);
3040 
3041   // get receiver klass
3042   __ null_check(recv, oopDesc::klass_offset_in_bytes());
3043   __ load_klass(rax, recv);
3044 
3045   // profile this call
3046   __ profile_virtual_call(rax, r14, rdx);
3047 
3048   // get target Method* & entry point
3049   __ lookup_virtual_method(rax, index, method);
3050   __ jump_from_interpreted(method, rdx);
3051 }
3052 
3053 
3054 void TemplateTable::invokevirtual(int byte_no) {
3055   transition(vtos, vtos);
3056   assert(byte_no == f2_byte, "use this argument");
3057   prepare_invoke(byte_no,
3058                  rbx,    // method or vtable index
3059                  noreg,  // unused itable index
3060                  rcx, rdx); // recv, flags
3061 
3062   // rbx: index
3063   // rcx: receiver
3064   // rdx: flags
3065 
3066   invokevirtual_helper(rbx, rcx, rdx);
3067 }
3068 
3069 
3070 void TemplateTable::invokespecial(int byte_no) {
3071   transition(vtos, vtos);
3072   assert(byte_no == f1_byte, "use this argument");
3073   prepare_invoke(byte_no, rbx, noreg,  // get f1 Method*
3074                  rcx);  // get receiver also for null check
3075   __ verify_oop(rcx);
3076   __ null_check(rcx);
3077   // do the call
3078   __ profile_call(rax);
3079   __ jump_from_interpreted(rbx, rax);
3080 }
3081 
3082 
3083 void TemplateTable::invokestatic(int byte_no) {
3084   transition(vtos, vtos);
3085   assert(byte_no == f1_byte, "use this argument");
3086   prepare_invoke(byte_no, rbx);  // get f1 Method*
3087   // do the call
3088   __ profile_call(rax);
3089   __ jump_from_interpreted(rbx, rax);
3090 }
3091 
3092 void TemplateTable::fast_invokevfinal(int byte_no) {
3093   transition(vtos, vtos);
3094   assert(byte_no == f2_byte, "use this argument");
3095   __ stop("fast_invokevfinal not used on amd64");
3096 }
3097 
3098 void TemplateTable::invokeinterface(int byte_no) {
3099   transition(vtos, vtos);
3100   assert(byte_no == f1_byte, "use this argument");
3101   prepare_invoke(byte_no, rax, rbx,  // get f1 Klass*, f2 itable index
3102                  rcx, rdx); // recv, flags
3103 
3104   // rax: interface klass (from f1)
3105   // rbx: itable index (from f2)
3106   // rcx: receiver
3107   // rdx: flags
3108 
3109   // Special case of invokeinterface called for virtual method of
3110   // java.lang.Object.  See cpCacheOop.cpp for details.
3111   // This code isn't produced by javac, but could be produced by
3112   // another compliant java compiler.
3113   Label notMethod;
3114   __ movl(r14, rdx);
3115   __ andl(r14, (1 << ConstantPoolCacheEntry::is_forced_virtual_shift));
3116   __ jcc(Assembler::zero, notMethod);
3117 
3118   invokevirtual_helper(rbx, rcx, rdx);
3119   __ bind(notMethod);
3120 
3121   // Get receiver klass into rdx - also a null check
3122   __ restore_locals();  // restore r14
3123   __ null_check(rcx, oopDesc::klass_offset_in_bytes());
3124   __ load_klass(rdx, rcx);
3125 
3126   // profile this call
3127   __ profile_virtual_call(rdx, r13, r14);
3128 
3129   Label no_such_interface, no_such_method;
3130 
3131   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3132                              rdx, rax, rbx,
3133                              // outputs: method, scan temp. reg
3134                              rbx, r13,
3135                              no_such_interface);
3136 
3137   // rbx: Method* to call
3138   // rcx: receiver
3139   // Check for abstract method error
3140   // Note: This should be done more efficiently via a throw_abstract_method_error
3141   //       interpreter entry point and a conditional jump to it in case of a null
3142   //       method.
3143   __ testptr(rbx, rbx);
3144   __ jcc(Assembler::zero, no_such_method);
3145 
3146   // do the call
3147   // rcx: receiver
3148   // rbx,: Method*
3149   __ jump_from_interpreted(rbx, rdx);
3150   __ should_not_reach_here();
3151 
3152   // exception handling code follows...
3153   // note: must restore interpreter registers to canonical
3154   //       state for exception handling to work correctly!
3155 
3156   __ bind(no_such_method);
3157   // throw exception
3158   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3159   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3160   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3161   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3162   // the call_VM checks for exception, so we should never return here.
3163   __ should_not_reach_here();
3164 
3165   __ bind(no_such_interface);
3166   // throw exception
3167   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3168   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3169   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3170   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3171                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3172   // the call_VM checks for exception, so we should never return here.
3173   __ should_not_reach_here();
3174 }
3175 
3176 
3177 void TemplateTable::invokehandle(int byte_no) {
3178   transition(vtos, vtos);
3179   assert(byte_no == f1_byte, "use this argument");
3180   const Register rbx_method = rbx;
3181   const Register rax_mtype  = rax;
3182   const Register rcx_recv   = rcx;
3183   const Register rdx_flags  = rdx;
3184 
3185   if (!EnableInvokeDynamic) {
3186     // rewriter does not generate this bytecode
3187     __ should_not_reach_here();
3188     return;
3189   }
3190 
3191   prepare_invoke(byte_no, rbx_method, rax_mtype, rcx_recv);
3192   __ verify_method_ptr(rbx_method);
3193   __ verify_oop(rcx_recv);
3194   __ null_check(rcx_recv);
3195 
3196   // rax: MethodType object (from cpool->resolved_references[f1], if necessary)
3197   // rbx: MH.invokeExact_MT method (from f2)
3198 
3199   // Note:  rax_mtype is already pushed (if necessary) by prepare_invoke
3200 
3201   // FIXME: profile the LambdaForm also
3202   __ profile_final_call(rax);
3203 
3204   __ jump_from_interpreted(rbx_method, rdx);
3205 }
3206 
3207 
3208 void TemplateTable::invokedynamic(int byte_no) {
3209   transition(vtos, vtos);
3210   assert(byte_no == f1_byte, "use this argument");
3211 
3212   if (!EnableInvokeDynamic) {
3213     // We should not encounter this bytecode if !EnableInvokeDynamic.
3214     // The verifier will stop it.  However, if we get past the verifier,
3215     // this will stop the thread in a reasonable way, without crashing the JVM.
3216     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3217                      InterpreterRuntime::throw_IncompatibleClassChangeError));
3218     // the call_VM checks for exception, so we should never return here.
3219     __ should_not_reach_here();
3220     return;
3221   }
3222 
3223   const Register rbx_method   = rbx;
3224   const Register rax_callsite = rax;
3225 
3226   prepare_invoke(byte_no, rbx_method, rax_callsite);
3227 
3228   // rax: CallSite object (from cpool->resolved_references[f1])
3229   // rbx: MH.linkToCallSite method (from f2)
3230 
3231   // Note:  rax_callsite is already pushed by prepare_invoke
3232 
3233   // %%% should make a type profile for any invokedynamic that takes a ref argument
3234   // profile this call
3235   __ profile_call(r13);
3236 
3237   __ verify_oop(rax_callsite);
3238 
3239   __ jump_from_interpreted(rbx_method, rdx);
3240 }
3241 
3242 
3243 //-----------------------------------------------------------------------------
3244 // Allocation
3245 
3246 void TemplateTable::_new() {
3247   transition(vtos, atos);
3248   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3249   Label slow_case;
3250   Label done;
3251   Label initialize_header;
3252   Label initialize_object; // including clearing the fields
3253   Label allocate_shared;
3254 
3255   __ get_cpool_and_tags(rsi, rax);
3256   // Make sure the class we're about to instantiate has been resolved.
3257   // This is done before loading InstanceKlass to be consistent with the order
3258   // how Constant Pool is updated (see ConstantPool::klass_at_put)
3259   const int tags_offset = Array<u1>::base_offset_in_bytes();
3260   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset),
3261           JVM_CONSTANT_Class);
3262   __ jcc(Assembler::notEqual, slow_case);
3263 
3264   // get InstanceKlass
3265   __ movptr(rsi, Address(rsi, rdx,
3266             Address::times_8, sizeof(ConstantPool)));
3267 
3268   // make sure klass is initialized & doesn't have finalizer
3269   // make sure klass is fully initialized
3270   __ cmpb(Address(rsi,
3271                   InstanceKlass::init_state_offset()),
3272           InstanceKlass::fully_initialized);
3273   __ jcc(Assembler::notEqual, slow_case);
3274 
3275   // get instance_size in InstanceKlass (scaled to a count of bytes)
3276   __ movl(rdx,
3277           Address(rsi,
3278                   Klass::layout_helper_offset()));
3279   // test to see if it has a finalizer or is malformed in some way
3280   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3281   __ jcc(Assembler::notZero, slow_case);
3282 
3283   // Allocate the instance
3284   // 1) Try to allocate in the TLAB
3285   // 2) if fail and the object is large allocate in the shared Eden
3286   // 3) if the above fails (or is not applicable), go to a slow case
3287   // (creates a new TLAB, etc.)
3288 
3289   const bool allow_shared_alloc =
3290     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3291 
3292   if (UseTLAB) {
3293     __ movptr(rax, Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())));
3294     __ lea(rbx, Address(rax, rdx, Address::times_1));
3295     __ cmpptr(rbx, Address(r15_thread, in_bytes(JavaThread::tlab_end_offset())));
3296     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3297     __ movptr(Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3298     if (ZeroTLAB) {
3299       // the fields have been already cleared
3300       __ jmp(initialize_header);
3301     } else {
3302       // initialize both the header and fields
3303       __ jmp(initialize_object);
3304     }
3305   }
3306 
3307   // Allocation in the shared Eden, if allowed.
3308   //
3309   // rdx: instance size in bytes
3310   if (allow_shared_alloc) {
3311     __ bind(allocate_shared);
3312 
3313     ExternalAddress top((address)Universe::heap()->top_addr());
3314     ExternalAddress end((address)Universe::heap()->end_addr());
3315 
3316     const Register RtopAddr = rscratch1;
3317     const Register RendAddr = rscratch2;
3318 
3319     __ lea(RtopAddr, top);
3320     __ lea(RendAddr, end);
3321     __ movptr(rax, Address(RtopAddr, 0));
3322 
3323     // For retries rax gets set by cmpxchgq
3324     Label retry;
3325     __ bind(retry);
3326     __ lea(rbx, Address(rax, rdx, Address::times_1));
3327     __ cmpptr(rbx, Address(RendAddr, 0));
3328     __ jcc(Assembler::above, slow_case);
3329 
3330     // Compare rax with the top addr, and if still equal, store the new
3331     // top addr in rbx at the address of the top addr pointer. Sets ZF if was
3332     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3333     //
3334     // rax: object begin
3335     // rbx: object end
3336     // rdx: instance size in bytes
3337     if (os::is_MP()) {
3338       __ lock();
3339     }
3340     __ cmpxchgptr(rbx, Address(RtopAddr, 0));
3341 
3342     // if someone beat us on the allocation, try again, otherwise continue
3343     __ jcc(Assembler::notEqual, retry);
3344 
3345     __ incr_allocated_bytes(r15_thread, rdx, 0);
3346   }
3347 
3348   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3349     // The object is initialized before the header.  If the object size is
3350     // zero, go directly to the header initialization.
3351     __ bind(initialize_object);
3352     __ decrementl(rdx, sizeof(oopDesc));
3353     __ jcc(Assembler::zero, initialize_header);
3354 
3355     // Initialize object fields
3356     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3357     __ shrl(rdx, LogBytesPerLong);  // divide by oopSize to simplify the loop
3358     {
3359       Label loop;
3360       __ bind(loop);
3361       __ movq(Address(rax, rdx, Address::times_8,
3362                       sizeof(oopDesc) - oopSize),
3363               rcx);
3364       __ decrementl(rdx);
3365       __ jcc(Assembler::notZero, loop);
3366     }
3367 
3368     // initialize object header only.
3369     __ bind(initialize_header);
3370     if (UseBiasedLocking) {
3371       __ movptr(rscratch1, Address(rsi, Klass::prototype_header_offset()));
3372       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()), rscratch1);
3373     } else {
3374       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()),
3375                (intptr_t) markOopDesc::prototype()); // header (address 0x1)
3376     }
3377     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3378     __ store_klass_gap(rax, rcx);  // zero klass gap for compressed oops
3379     __ store_klass(rax, rsi);      // store klass last
3380 
3381     {
3382       SkipIfEqual skip(_masm, &DTraceAllocProbes, false);
3383       // Trigger dtrace event for fastpath
3384       __ push(atos); // save the return value
3385       __ call_VM_leaf(
3386            CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3387       __ pop(atos); // restore the return value
3388 
3389     }
3390     __ jmp(done);
3391   }
3392 
3393 
3394   // slow case
3395   __ bind(slow_case);
3396   __ get_constant_pool(c_rarg1);
3397   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3398   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), c_rarg1, c_rarg2);
3399   __ verify_oop(rax);
3400 
3401   // continue
3402   __ bind(done);
3403 }
3404 
3405 void TemplateTable::newarray() {
3406   transition(itos, atos);
3407   __ load_unsigned_byte(c_rarg1, at_bcp(1));
3408   __ movl(c_rarg2, rax);
3409   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray),
3410           c_rarg1, c_rarg2);
3411 }
3412 
3413 void TemplateTable::anewarray() {
3414   transition(itos, atos);
3415   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3416   __ get_constant_pool(c_rarg1);
3417   __ movl(c_rarg3, rax);
3418   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray),
3419           c_rarg1, c_rarg2, c_rarg3);
3420 }
3421 
3422 void TemplateTable::arraylength() {
3423   transition(atos, itos);
3424   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3425   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3426 }
3427 
3428 void TemplateTable::checkcast() {
3429   transition(atos, atos);
3430   Label done, is_null, ok_is_subtype, quicked, resolved;
3431   __ testptr(rax, rax); // object is in rax
3432   __ jcc(Assembler::zero, is_null);
3433 
3434   // Get cpool & tags index
3435   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3436   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3437   // See if bytecode has already been quicked
3438   __ cmpb(Address(rdx, rbx,
3439                   Address::times_1,
3440                   Array<u1>::base_offset_in_bytes()),
3441           JVM_CONSTANT_Class);
3442   __ jcc(Assembler::equal, quicked);
3443   __ push(atos); // save receiver for result, and for GC
3444   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3445   // vm_result_2 has metadata result
3446   __ get_vm_result_2(rax, r15_thread);
3447   __ pop_ptr(rdx); // restore receiver
3448   __ jmpb(resolved);
3449 
3450   // Get superklass in rax and subklass in rbx
3451   __ bind(quicked);
3452   __ mov(rdx, rax); // Save object in rdx; rax needed for subtype check
3453   __ movptr(rax, Address(rcx, rbx,
3454                        Address::times_8, sizeof(ConstantPool)));
3455 
3456   __ bind(resolved);
3457   __ load_klass(rbx, rdx);
3458 
3459   // Generate subtype check.  Blows rcx, rdi.  Object in rdx.
3460   // Superklass in rax.  Subklass in rbx.
3461   __ gen_subtype_check(rbx, ok_is_subtype);
3462 
3463   // Come here on failure
3464   __ push_ptr(rdx);
3465   // object is at TOS
3466   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3467 
3468   // Come here on success
3469   __ bind(ok_is_subtype);
3470   __ mov(rax, rdx); // Restore object in rdx
3471 
3472   // Collect counts on whether this check-cast sees NULLs a lot or not.
3473   if (ProfileInterpreter) {
3474     __ jmp(done);
3475     __ bind(is_null);
3476     __ profile_null_seen(rcx);
3477   } else {
3478     __ bind(is_null);   // same as 'done'
3479   }
3480   __ bind(done);
3481 }
3482 
3483 void TemplateTable::instanceof() {
3484   transition(atos, itos);
3485   Label done, is_null, ok_is_subtype, quicked, resolved;
3486   __ testptr(rax, rax);
3487   __ jcc(Assembler::zero, is_null);
3488 
3489   // Get cpool & tags index
3490   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3491   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3492   // See if bytecode has already been quicked
3493   __ cmpb(Address(rdx, rbx,
3494                   Address::times_1,
3495                   Array<u1>::base_offset_in_bytes()),
3496           JVM_CONSTANT_Class);
3497   __ jcc(Assembler::equal, quicked);
3498 
3499   __ push(atos); // save receiver for result, and for GC
3500   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3501   // vm_result_2 has metadata result
3502   __ get_vm_result_2(rax, r15_thread);
3503   __ pop_ptr(rdx); // restore receiver
3504   __ verify_oop(rdx);
3505   __ load_klass(rdx, rdx);
3506   __ jmpb(resolved);
3507 
3508   // Get superklass in rax and subklass in rdx
3509   __ bind(quicked);
3510   __ load_klass(rdx, rax);
3511   __ movptr(rax, Address(rcx, rbx,
3512                          Address::times_8, sizeof(ConstantPool)));
3513 
3514   __ bind(resolved);
3515 
3516   // Generate subtype check.  Blows rcx, rdi
3517   // Superklass in rax.  Subklass in rdx.
3518   __ gen_subtype_check(rdx, ok_is_subtype);
3519 
3520   // Come here on failure
3521   __ xorl(rax, rax);
3522   __ jmpb(done);
3523   // Come here on success
3524   __ bind(ok_is_subtype);
3525   __ movl(rax, 1);
3526 
3527   // Collect counts on whether this test sees NULLs a lot or not.
3528   if (ProfileInterpreter) {
3529     __ jmp(done);
3530     __ bind(is_null);
3531     __ profile_null_seen(rcx);
3532   } else {
3533     __ bind(is_null);   // same as 'done'
3534   }
3535   __ bind(done);
3536   // rax = 0: obj == NULL or  obj is not an instanceof the specified klass
3537   // rax = 1: obj != NULL and obj is     an instanceof the specified klass
3538 }
3539 
3540 //-----------------------------------------------------------------------------
3541 // Breakpoints
3542 void TemplateTable::_breakpoint() {
3543   // Note: We get here even if we are single stepping..
3544   // jbug inists on setting breakpoints at every bytecode
3545   // even if we are in single step mode.
3546 
3547   transition(vtos, vtos);
3548 
3549   // get the unpatched byte code
3550   __ get_method(c_rarg1);
3551   __ call_VM(noreg,
3552              CAST_FROM_FN_PTR(address,
3553                               InterpreterRuntime::get_original_bytecode_at),
3554              c_rarg1, r13);
3555   __ mov(rbx, rax);
3556 
3557   // post the breakpoint event
3558   __ get_method(c_rarg1);
3559   __ call_VM(noreg,
3560              CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint),
3561              c_rarg1, r13);
3562 
3563   // complete the execution of original bytecode
3564   __ dispatch_only_normal(vtos);
3565 }
3566 
3567 //-----------------------------------------------------------------------------
3568 // Exceptions
3569 
3570 void TemplateTable::athrow() {
3571   transition(atos, vtos);
3572   __ null_check(rax);
3573   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3574 }
3575 
3576 //-----------------------------------------------------------------------------
3577 // Synchronization
3578 //
3579 // Note: monitorenter & exit are symmetric routines; which is reflected
3580 //       in the assembly code structure as well
3581 //
3582 // Stack layout:
3583 //
3584 // [expressions  ] <--- rsp               = expression stack top
3585 // ..
3586 // [expressions  ]
3587 // [monitor entry] <--- monitor block top = expression stack bot
3588 // ..
3589 // [monitor entry]
3590 // [frame data   ] <--- monitor block bot
3591 // ...
3592 // [saved rbp    ] <--- rbp
3593 void TemplateTable::monitorenter() {
3594   transition(atos, vtos);
3595 
3596   // check for NULL object
3597   __ null_check(rax);
3598 
3599   const Address monitor_block_top(
3600         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3601   const Address monitor_block_bot(
3602         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3603   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3604 
3605   Label allocated;
3606 
3607   // initialize entry pointer
3608   __ xorl(c_rarg1, c_rarg1); // points to free slot or NULL
3609 
3610   // find a free slot in the monitor block (result in c_rarg1)
3611   {
3612     Label entry, loop, exit;
3613     __ movptr(c_rarg3, monitor_block_top); // points to current entry,
3614                                      // starting with top-most entry
3615     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3616                                      // of monitor block
3617     __ jmpb(entry);
3618 
3619     __ bind(loop);
3620     // check if current entry is used
3621     __ cmpptr(Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()), (int32_t) NULL_WORD);
3622     // if not used then remember entry in c_rarg1
3623     __ cmov(Assembler::equal, c_rarg1, c_rarg3);
3624     // check if current entry is for same object
3625     __ cmpptr(rax, Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()));
3626     // if same object then stop searching
3627     __ jccb(Assembler::equal, exit);
3628     // otherwise advance to next entry
3629     __ addptr(c_rarg3, entry_size);
3630     __ bind(entry);
3631     // check if bottom reached
3632     __ cmpptr(c_rarg3, c_rarg2);
3633     // if not at bottom then check this entry
3634     __ jcc(Assembler::notEqual, loop);
3635     __ bind(exit);
3636   }
3637 
3638   __ testptr(c_rarg1, c_rarg1); // check if a slot has been found
3639   __ jcc(Assembler::notZero, allocated); // if found, continue with that one
3640 
3641   // allocate one if there's no free slot
3642   {
3643     Label entry, loop;
3644     // 1. compute new pointers             // rsp: old expression stack top
3645     __ movptr(c_rarg1, monitor_block_bot); // c_rarg1: old expression stack bottom
3646     __ subptr(rsp, entry_size);            // move expression stack top
3647     __ subptr(c_rarg1, entry_size);        // move expression stack bottom
3648     __ mov(c_rarg3, rsp);                  // set start value for copy loop
3649     __ movptr(monitor_block_bot, c_rarg1); // set new monitor block bottom
3650     __ jmp(entry);
3651     // 2. move expression stack contents
3652     __ bind(loop);
3653     __ movptr(c_rarg2, Address(c_rarg3, entry_size)); // load expression stack
3654                                                       // word from old location
3655     __ movptr(Address(c_rarg3, 0), c_rarg2);          // and store it at new location
3656     __ addptr(c_rarg3, wordSize);                     // advance to next word
3657     __ bind(entry);
3658     __ cmpptr(c_rarg3, c_rarg1);            // check if bottom reached
3659     __ jcc(Assembler::notEqual, loop);      // if not at bottom then
3660                                             // copy next word
3661   }
3662 
3663   // call run-time routine
3664   // c_rarg1: points to monitor entry
3665   __ bind(allocated);
3666 
3667   // Increment bcp to point to the next bytecode, so exception
3668   // handling for async. exceptions work correctly.
3669   // The object has already been poped from the stack, so the
3670   // expression stack looks correct.
3671   __ increment(r13);
3672 
3673   // store object
3674   __ movptr(Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()), rax);
3675   __ lock_object(c_rarg1);
3676 
3677   // check to make sure this monitor doesn't cause stack overflow after locking
3678   __ save_bcp();  // in case of exception
3679   __ generate_stack_overflow_check(0);
3680 
3681   // The bcp has already been incremented. Just need to dispatch to
3682   // next instruction.
3683   __ dispatch_next(vtos);
3684 }
3685 
3686 
3687 void TemplateTable::monitorexit() {
3688   transition(atos, vtos);
3689 
3690   // check for NULL object
3691   __ null_check(rax);
3692 
3693   const Address monitor_block_top(
3694         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3695   const Address monitor_block_bot(
3696         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3697   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3698 
3699   Label found;
3700 
3701   // find matching slot
3702   {
3703     Label entry, loop;
3704     __ movptr(c_rarg1, monitor_block_top); // points to current entry,
3705                                      // starting with top-most entry
3706     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3707                                      // of monitor block
3708     __ jmpb(entry);
3709 
3710     __ bind(loop);
3711     // check if current entry is for same object
3712     __ cmpptr(rax, Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()));
3713     // if same object then stop searching
3714     __ jcc(Assembler::equal, found);
3715     // otherwise advance to next entry
3716     __ addptr(c_rarg1, entry_size);
3717     __ bind(entry);
3718     // check if bottom reached
3719     __ cmpptr(c_rarg1, c_rarg2);
3720     // if not at bottom then check this entry
3721     __ jcc(Assembler::notEqual, loop);
3722   }
3723 
3724   // error handling. Unlocking was not block-structured
3725   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3726                    InterpreterRuntime::throw_illegal_monitor_state_exception));
3727   __ should_not_reach_here();
3728 
3729   // call run-time routine
3730   // rsi: points to monitor entry
3731   __ bind(found);
3732   __ push_ptr(rax); // make sure object is on stack (contract with oopMaps)
3733   __ unlock_object(c_rarg1);
3734   __ pop_ptr(rax); // discard object
3735 }
3736 
3737 
3738 // Wide instructions
3739 void TemplateTable::wide() {
3740   transition(vtos, vtos);
3741   __ load_unsigned_byte(rbx, at_bcp(1));
3742   __ lea(rscratch1, ExternalAddress((address)Interpreter::_wentry_point));
3743   __ jmp(Address(rscratch1, rbx, Address::times_8));
3744   // Note: the r13 increment step is part of the individual wide
3745   // bytecode implementations
3746 }
3747 
3748 
3749 // Multi arrays
3750 void TemplateTable::multianewarray() {
3751   transition(vtos, atos);
3752   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3753   // last dim is on top of stack; we want address of first one:
3754   // first_addr = last_addr + (ndims - 1) * wordSize
3755   __ lea(c_rarg1, Address(rsp, rax, Address::times_8, -wordSize));
3756   call_VM(rax,
3757           CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray),
3758           c_rarg1);
3759   __ load_unsigned_byte(rbx, at_bcp(3));
3760   __ lea(rsp, Address(rsp, rbx, Address::times_8));
3761 }
3762 #endif // !CC_INTERP
3763 
3764 #ifdef ASSERT
3765 
3766 // Correct values of the cache and index registers are preserved.
3767 void TemplateTable::verify_stable(Register cache, Register index,
3768                                   bool is_static, bool has_tos) {
3769   if (FoldStableValues && (VerifyStable || TraceStableFieldUpdates)) {
3770     ByteSize cp_base_offset = ConstantPoolCache::base_offset();
3771     assert_different_registers(cache, index, rax);
3772     __ get_cache_and_index_at_bcp(c_rarg2, rscratch1, 1);
3773 
3774     if (is_static) {
3775       // Life is simple.  Null out the object pointer.
3776       __ xorl(c_rarg1, c_rarg1);
3777     } else {
3778       // Life is harder. The stack holds the value on top, followed by
3779       // the object.  We don't know the size of the value, though; it
3780       // could be one or two words depending on its type. As a result,
3781       // we must find the type to determine where the object is.
3782       __ movl(c_rarg3, Address(c_rarg2, rscratch1,
3783                            Address::times_8,
3784                            in_bytes(cp_base_offset +
3785                                      ConstantPoolCacheEntry::flags_offset())));
3786       __ shrl(c_rarg3, ConstantPoolCacheEntry::tos_state_shift);
3787       // Make sure we don't need to mask rcx after the above shift
3788       ConstantPoolCacheEntry::verify_tos_state_shift();
3789       __ movptr(c_rarg1, at_tos_p1());  // initially assume a one word jvalue
3790       __ cmpl(c_rarg3, ltos);
3791       __ cmovptr(Assembler::equal,
3792                  c_rarg1, at_tos_p2()); // ltos (two word jvalue)
3793       __ cmpl(c_rarg3, dtos);
3794       __ cmovptr(Assembler::equal,
3795                  c_rarg1, at_tos_p2()); // dtos (two word jvalue)
3796     }
3797     // cache entry pointer
3798     __ addptr(c_rarg2, in_bytes(cp_base_offset));
3799     __ shll(rscratch1, LogBytesPerWord);
3800     __ addptr(c_rarg2, rscratch1);
3801     // object (tos)
3802     __ mov(c_rarg3, rsp);
3803     // c_rarg1: object pointer set up above (NULL if static)
3804     // c_rarg2: cache entry pointer
3805     // c_rarg3: jvalue object on the stack
3806     __ call_VM(noreg,
3807                CAST_FROM_FN_PTR(address,
3808                                 InterpreterRuntime::verify_stable),
3809                c_rarg1, c_rarg2, c_rarg3);
3810     __ get_cache_and_index_at_bcp(cache, index, 1);
3811   }
3812 }
3813 
3814 void TemplateTable::verify_stable_fast() {
3815   if (FoldStableValues && (VerifyStable || TraceStableFieldUpdates)) {
3816     __ pop_ptr(rbx);                  // copy the object pointer from tos
3817     __ verify_oop(rbx);
3818     __ push_ptr(rbx);                 // put the object pointer back on tos
3819     // Save tos values before call_VM() clobbers them. Since we have
3820     // to do it for every data type, we use the saved values as the
3821     // jvalue object.
3822     switch (bytecode()) {          // load values into the jvalue object
3823     case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
3824     case Bytecodes::_fast_bputfield: // fall through
3825     case Bytecodes::_fast_sputfield: // fall through
3826     case Bytecodes::_fast_cputfield: // fall through
3827     case Bytecodes::_fast_iputfield: __ push_i(rax); break;
3828     case Bytecodes::_fast_dputfield: __ push_d(); break;
3829     case Bytecodes::_fast_fputfield: __ push_f(); break;
3830     case Bytecodes::_fast_lputfield: __ push_l(rax); break;
3831 
3832     default:
3833       ShouldNotReachHere();
3834     }
3835     __ mov(c_rarg3, rsp);             // points to jvalue on the stack
3836     // access constant pool cache entry
3837     __ get_cache_entry_pointer_at_bcp(c_rarg2, rax, 1);
3838     __ verify_oop(rbx);
3839     // rbx: object pointer copied above
3840     // c_rarg2: cache entry pointer
3841     // c_rarg3: jvalue object on the stack
3842     __ call_VM(noreg,
3843                CAST_FROM_FN_PTR(address,
3844                                 InterpreterRuntime::verify_stable),
3845                rbx, c_rarg2, c_rarg3);
3846 
3847     switch (bytecode()) {             // restore tos values
3848     case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
3849     case Bytecodes::_fast_bputfield: // fall through
3850     case Bytecodes::_fast_sputfield: // fall through
3851     case Bytecodes::_fast_cputfield: // fall through
3852     case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
3853     case Bytecodes::_fast_dputfield: __ pop_d(); break;
3854     case Bytecodes::_fast_fputfield: __ pop_f(); break;
3855     case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
3856     }
3857   }
3858 }
3859 
3860 #endif // ASSERT