1 /*
   2  * Copyright (c) 1997, 2018, 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 #define __ _masm->
  42 
  43 //----------------------------------------------------------------------------------------------------
  44 // Platform-dependent initialization
  45 
  46 void TemplateTable::pd_initialize() {
  47   // No i486 specific initialization
  48 }
  49 
  50 //----------------------------------------------------------------------------------------------------
  51 // Address computation
  52 
  53 // local variables
  54 static inline Address iaddress(int n)            {
  55   return Address(rdi, Interpreter::local_offset_in_bytes(n));
  56 }
  57 
  58 static inline Address laddress(int n)            { return iaddress(n + 1); }
  59 static inline Address haddress(int n)            { return iaddress(n + 0); }
  60 static inline Address faddress(int n)            { return iaddress(n); }
  61 static inline Address daddress(int n)            { return laddress(n); }
  62 static inline Address aaddress(int n)            { return iaddress(n); }
  63 
  64 static inline Address iaddress(Register r)       {
  65   return Address(rdi, r, Interpreter::stackElementScale());
  66 }
  67 static inline Address laddress(Register r)       {
  68   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(1));
  69 }
  70 static inline Address haddress(Register r)       {
  71   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(0));
  72 }
  73 
  74 static inline Address faddress(Register r)       { return iaddress(r); }
  75 static inline Address daddress(Register r)       { return laddress(r); }
  76 static inline Address aaddress(Register r)       { return iaddress(r); }
  77 
  78 // expression stack
  79 // (Note: Must not use symmetric equivalents at_rsp_m1/2 since they store
  80 // data beyond the rsp which is potentially unsafe in an MT environment;
  81 // an interrupt may overwrite that data.)
  82 static inline Address at_rsp   () {
  83   return Address(rsp, 0);
  84 }
  85 
  86 // At top of Java expression stack which may be different than rsp().  It
  87 // isn't for category 1 objects.
  88 static inline Address at_tos   () {
  89   Address tos = Address(rsp,  Interpreter::expr_offset_in_bytes(0));
  90   return tos;
  91 }
  92 
  93 static inline Address at_tos_p1() {
  94   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
  95 }
  96 
  97 static inline Address at_tos_p2() {
  98   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
  99 }
 100 
 101 // Condition conversion
 102 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 103   switch (cc) {
 104     case TemplateTable::equal        : return Assembler::notEqual;
 105     case TemplateTable::not_equal    : return Assembler::equal;
 106     case TemplateTable::less         : return Assembler::greaterEqual;
 107     case TemplateTable::less_equal   : return Assembler::greater;
 108     case TemplateTable::greater      : return Assembler::lessEqual;
 109     case TemplateTable::greater_equal: return Assembler::less;
 110   }
 111   ShouldNotReachHere();
 112   return Assembler::zero;
 113 }
 114 
 115 
 116 //----------------------------------------------------------------------------------------------------
 117 // Miscelaneous helper routines
 118 
 119 // Store an oop (or NULL) at the address described by obj.
 120 // If val == noreg this means store a NULL
 121 
 122 static void do_oop_store(InterpreterMacroAssembler* _masm,
 123                          Address obj,
 124                          Register val,
 125                          BarrierSet::Name barrier,
 126                          bool precise) {
 127   assert(val == noreg || val == rax, "parameter is just for looks");
 128   switch (barrier) {
 129 #if INCLUDE_ALL_GCS
 130     case BarrierSet::G1SATBCT:
 131     case BarrierSet::G1SATBCTLogging:
 132       {
 133         // flatten object address if needed
 134         // We do it regardless of precise because we need the registers
 135         if (obj.index() == noreg && obj.disp() == 0) {
 136           if (obj.base() != rdx) {
 137             __ movl(rdx, obj.base());
 138           }
 139         } else {
 140           __ leal(rdx, obj);
 141         }
 142         __ get_thread(rcx);
 143         __ save_bcp();
 144         __ g1_write_barrier_pre(rdx /* obj */,
 145                                 rbx /* pre_val */,
 146                                 rcx /* thread */,
 147                                 rsi /* tmp */,
 148                                 val != noreg /* tosca_live */,
 149                                 false /* expand_call */);
 150 
 151         // Do the actual store
 152         // noreg means NULL
 153         if (val == noreg) {
 154           __ movptr(Address(rdx, 0), NULL_WORD);
 155           // No post barrier for NULL
 156         } else {
 157           __ movl(Address(rdx, 0), val);
 158           __ g1_write_barrier_post(rdx /* store_adr */,
 159                                    val /* new_val */,
 160                                    rcx /* thread */,
 161                                    rbx /* tmp */,
 162                                    rsi /* tmp2 */);
 163         }
 164         __ restore_bcp();
 165 
 166       }
 167       break;
 168 #endif // INCLUDE_ALL_GCS
 169     case BarrierSet::CardTableModRef:
 170     case BarrierSet::CardTableExtension:
 171       {
 172         if (val == noreg) {
 173           __ movptr(obj, NULL_WORD);
 174         } else {
 175           __ movl(obj, val);
 176           // flatten object address if needed
 177           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 178             __ store_check(obj.base());
 179           } else {
 180             __ leal(rdx, obj);
 181             __ store_check(rdx);
 182           }
 183         }
 184       }
 185       break;
 186     case BarrierSet::ModRef:
 187     case BarrierSet::Other:
 188       if (val == noreg) {
 189         __ movptr(obj, NULL_WORD);
 190       } else {
 191         __ movl(obj, val);
 192       }
 193       break;
 194     default      :
 195       ShouldNotReachHere();
 196 
 197   }
 198 }
 199 
 200 Address TemplateTable::at_bcp(int offset) {
 201   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 202   return Address(rsi, offset);
 203 }
 204 
 205 
 206 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register bc_reg,
 207                                    Register temp_reg, bool load_bc_into_bc_reg/*=true*/,
 208                                    int byte_no) {
 209   if (!RewriteBytecodes)  return;
 210   Label L_patch_done;
 211 
 212   switch (bc) {
 213   case Bytecodes::_fast_aputfield:
 214   case Bytecodes::_fast_bputfield:
 215   case Bytecodes::_fast_zputfield:
 216   case Bytecodes::_fast_cputfield:
 217   case Bytecodes::_fast_dputfield:
 218   case Bytecodes::_fast_fputfield:
 219   case Bytecodes::_fast_iputfield:
 220   case Bytecodes::_fast_lputfield:
 221   case Bytecodes::_fast_sputfield:
 222     {
 223       // We skip bytecode quickening for putfield instructions when
 224       // the put_code written to the constant pool cache is zero.
 225       // This is required so that every execution of this instruction
 226       // calls out to InterpreterRuntime::resolve_get_put to do
 227       // additional, required work.
 228       assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
 229       assert(load_bc_into_bc_reg, "we use bc_reg as temp");
 230       __ get_cache_and_index_and_bytecode_at_bcp(bc_reg, temp_reg, temp_reg, byte_no, 1);
 231       __ movl(bc_reg, bc);
 232       __ cmpl(temp_reg, (int) 0);
 233       __ jcc(Assembler::zero, L_patch_done);  // don't patch
 234     }
 235     break;
 236   default:
 237     assert(byte_no == -1, "sanity");
 238     // the pair bytecodes have already done the load.
 239     if (load_bc_into_bc_reg) {
 240       __ movl(bc_reg, bc);
 241     }
 242   }
 243 
 244   if (JvmtiExport::can_post_breakpoint()) {
 245     Label L_fast_patch;
 246     // if a breakpoint is present we can't rewrite the stream directly
 247     __ movzbl(temp_reg, at_bcp(0));
 248     __ cmpl(temp_reg, Bytecodes::_breakpoint);
 249     __ jcc(Assembler::notEqual, L_fast_patch);
 250     __ get_method(temp_reg);
 251     // Let breakpoint table handling rewrite to quicker bytecode
 252     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), temp_reg, rsi, bc_reg);
 253 #ifndef ASSERT
 254     __ jmpb(L_patch_done);
 255 #else
 256     __ jmp(L_patch_done);
 257 #endif
 258     __ bind(L_fast_patch);
 259   }
 260 
 261 #ifdef ASSERT
 262   Label L_okay;
 263   __ load_unsigned_byte(temp_reg, at_bcp(0));
 264   __ cmpl(temp_reg, (int)Bytecodes::java_code(bc));
 265   __ jccb(Assembler::equal, L_okay);
 266   __ cmpl(temp_reg, bc_reg);
 267   __ jcc(Assembler::equal, L_okay);
 268   __ stop("patching the wrong bytecode");
 269   __ bind(L_okay);
 270 #endif
 271 
 272   // patch bytecode
 273   __ movb(at_bcp(0), bc_reg);
 274   __ bind(L_patch_done);
 275 }
 276 
 277 //----------------------------------------------------------------------------------------------------
 278 // Individual instructions
 279 
 280 void TemplateTable::nop() {
 281   transition(vtos, vtos);
 282   // nothing to do
 283 }
 284 
 285 void TemplateTable::shouldnotreachhere() {
 286   transition(vtos, vtos);
 287   __ stop("shouldnotreachhere bytecode");
 288 }
 289 
 290 
 291 
 292 void TemplateTable::aconst_null() {
 293   transition(vtos, atos);
 294   __ xorptr(rax, rax);
 295 }
 296 
 297 
 298 void TemplateTable::iconst(int value) {
 299   transition(vtos, itos);
 300   if (value == 0) {
 301     __ xorptr(rax, rax);
 302   } else {
 303     __ movptr(rax, value);
 304   }
 305 }
 306 
 307 
 308 void TemplateTable::lconst(int value) {
 309   transition(vtos, ltos);
 310   if (value == 0) {
 311     __ xorptr(rax, rax);
 312   } else {
 313     __ movptr(rax, value);
 314   }
 315   assert(value >= 0, "check this code");
 316   __ xorptr(rdx, rdx);
 317 }
 318 
 319 
 320 void TemplateTable::fconst(int value) {
 321   transition(vtos, ftos);
 322          if (value == 0) { __ fldz();
 323   } else if (value == 1) { __ fld1();
 324   } else if (value == 2) { __ fld1(); __ fld1(); __ faddp(); // should do a better solution here
 325   } else                 { ShouldNotReachHere();
 326   }
 327 }
 328 
 329 
 330 void TemplateTable::dconst(int value) {
 331   transition(vtos, dtos);
 332          if (value == 0) { __ fldz();
 333   } else if (value == 1) { __ fld1();
 334   } else                 { ShouldNotReachHere();
 335   }
 336 }
 337 
 338 
 339 void TemplateTable::bipush() {
 340   transition(vtos, itos);
 341   __ load_signed_byte(rax, at_bcp(1));
 342 }
 343 
 344 
 345 void TemplateTable::sipush() {
 346   transition(vtos, itos);
 347   __ load_unsigned_short(rax, at_bcp(1));
 348   __ bswapl(rax);
 349   __ sarl(rax, 16);
 350 }
 351 
 352 void TemplateTable::ldc(bool wide) {
 353   transition(vtos, vtos);
 354   Label call_ldc, notFloat, notClass, Done;
 355 
 356   if (wide) {
 357     __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 358   } else {
 359     __ load_unsigned_byte(rbx, at_bcp(1));
 360   }
 361   __ get_cpool_and_tags(rcx, rax);
 362   const int base_offset = ConstantPool::header_size() * wordSize;
 363   const int tags_offset = Array<u1>::base_offset_in_bytes();
 364 
 365   // get type
 366   __ xorptr(rdx, rdx);
 367   __ movb(rdx, Address(rax, rbx, Address::times_1, tags_offset));
 368 
 369   // unresolved class - get the resolved class
 370   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClass);
 371   __ jccb(Assembler::equal, call_ldc);
 372 
 373   // unresolved class in error (resolution failed) - call into runtime
 374   // so that the same error from first resolution attempt is thrown.
 375   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClassInError);
 376   __ jccb(Assembler::equal, call_ldc);
 377 
 378   // resolved class - need to call vm to get java mirror of the class
 379   __ cmpl(rdx, JVM_CONSTANT_Class);
 380   __ jcc(Assembler::notEqual, notClass);
 381 
 382   __ bind(call_ldc);
 383   __ movl(rcx, wide);
 384   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), rcx);
 385   __ push(atos);
 386   __ jmp(Done);
 387 
 388   __ bind(notClass);
 389   __ cmpl(rdx, JVM_CONSTANT_Float);
 390   __ jccb(Assembler::notEqual, notFloat);
 391   // ftos
 392   __ fld_s(    Address(rcx, rbx, Address::times_ptr, base_offset));
 393   __ push(ftos);
 394   __ jmp(Done);
 395 
 396   __ bind(notFloat);
 397 #ifdef ASSERT
 398   { Label L;
 399     __ cmpl(rdx, JVM_CONSTANT_Integer);
 400     __ jcc(Assembler::equal, L);
 401     // String and Object are rewritten to fast_aldc
 402     __ stop("unexpected tag type in ldc");
 403     __ bind(L);
 404   }
 405 #endif
 406   // itos JVM_CONSTANT_Integer only
 407   __ movl(rax, Address(rcx, rbx, Address::times_ptr, base_offset));
 408   __ push(itos);
 409   __ bind(Done);
 410 }
 411 
 412 // Fast path for caching oop constants.
 413 void TemplateTable::fast_aldc(bool wide) {
 414   transition(vtos, atos);
 415 
 416   Register result = rax;
 417   Register tmp = rdx;
 418   int index_size = wide ? sizeof(u2) : sizeof(u1);
 419 
 420   Label resolved;
 421 
 422   // We are resolved if the resolved reference cache entry contains a
 423   // non-null object (String, MethodType, etc.)
 424   assert_different_registers(result, tmp);
 425   __ get_cache_index_at_bcp(tmp, 1, index_size);
 426   __ load_resolved_reference_at_index(result, tmp);
 427   __ testl(result, result);
 428   __ jcc(Assembler::notZero, resolved);
 429 
 430   address entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_ldc);
 431 
 432   // first time invocation - must resolve first
 433   __ movl(tmp, (int)bytecode());
 434   __ call_VM(result, entry, tmp);
 435 
 436   __ bind(resolved);
 437 
 438   if (VerifyOops) {
 439     __ verify_oop(result);
 440   }
 441 }
 442 
 443 void TemplateTable::ldc2_w() {
 444   transition(vtos, vtos);
 445   Label Long, Done;
 446   __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 447 
 448   __ get_cpool_and_tags(rcx, rax);
 449   const int base_offset = ConstantPool::header_size() * wordSize;
 450   const int tags_offset = Array<u1>::base_offset_in_bytes();
 451 
 452   // get type
 453   __ cmpb(Address(rax, rbx, Address::times_1, tags_offset), JVM_CONSTANT_Double);
 454   __ jccb(Assembler::notEqual, Long);
 455   // dtos
 456   __ fld_d(    Address(rcx, rbx, Address::times_ptr, base_offset));
 457   __ push(dtos);
 458   __ jmpb(Done);
 459 
 460   __ bind(Long);
 461   // ltos
 462   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, base_offset + 0 * wordSize));
 463   NOT_LP64(__ movptr(rdx, Address(rcx, rbx, Address::times_ptr, base_offset + 1 * wordSize)));
 464 
 465   __ push(ltos);
 466 
 467   __ bind(Done);
 468 }
 469 
 470 
 471 void TemplateTable::locals_index(Register reg, int offset) {
 472   __ load_unsigned_byte(reg, at_bcp(offset));
 473   __ negptr(reg);
 474 }
 475 
 476 
 477 void TemplateTable::iload() {
 478   transition(vtos, itos);
 479   if (RewriteFrequentPairs) {
 480     Label rewrite, done;
 481 
 482     // get next byte
 483     __ load_unsigned_byte(rbx, at_bcp(Bytecodes::length_for(Bytecodes::_iload)));
 484     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 485     // last two iloads in a pair.  Comparing against fast_iload means that
 486     // the next bytecode is neither an iload or a caload, and therefore
 487     // an iload pair.
 488     __ cmpl(rbx, Bytecodes::_iload);
 489     __ jcc(Assembler::equal, done);
 490 
 491     __ cmpl(rbx, Bytecodes::_fast_iload);
 492     __ movl(rcx, Bytecodes::_fast_iload2);
 493     __ jccb(Assembler::equal, rewrite);
 494 
 495     // if _caload, rewrite to fast_icaload
 496     __ cmpl(rbx, Bytecodes::_caload);
 497     __ movl(rcx, Bytecodes::_fast_icaload);
 498     __ jccb(Assembler::equal, rewrite);
 499 
 500     // rewrite so iload doesn't check again.
 501     __ movl(rcx, Bytecodes::_fast_iload);
 502 
 503     // rewrite
 504     // rcx: fast bytecode
 505     __ bind(rewrite);
 506     patch_bytecode(Bytecodes::_iload, rcx, rbx, false);
 507     __ bind(done);
 508   }
 509 
 510   // Get the local value into tos
 511   locals_index(rbx);
 512   __ movl(rax, iaddress(rbx));
 513 }
 514 
 515 
 516 void TemplateTable::fast_iload2() {
 517   transition(vtos, itos);
 518   locals_index(rbx);
 519   __ movl(rax, iaddress(rbx));
 520   __ push(itos);
 521   locals_index(rbx, 3);
 522   __ movl(rax, iaddress(rbx));
 523 }
 524 
 525 void TemplateTable::fast_iload() {
 526   transition(vtos, itos);
 527   locals_index(rbx);
 528   __ movl(rax, iaddress(rbx));
 529 }
 530 
 531 
 532 void TemplateTable::lload() {
 533   transition(vtos, ltos);
 534   locals_index(rbx);
 535   __ movptr(rax, laddress(rbx));
 536   NOT_LP64(__ movl(rdx, haddress(rbx)));
 537 }
 538 
 539 
 540 void TemplateTable::fload() {
 541   transition(vtos, ftos);
 542   locals_index(rbx);
 543   __ fld_s(faddress(rbx));
 544 }
 545 
 546 
 547 void TemplateTable::dload() {
 548   transition(vtos, dtos);
 549   locals_index(rbx);
 550   __ fld_d(daddress(rbx));
 551 }
 552 
 553 
 554 void TemplateTable::aload() {
 555   transition(vtos, atos);
 556   locals_index(rbx);
 557   __ movptr(rax, aaddress(rbx));
 558 }
 559 
 560 
 561 void TemplateTable::locals_index_wide(Register reg) {
 562   __ load_unsigned_short(reg, at_bcp(2));
 563   __ bswapl(reg);
 564   __ shrl(reg, 16);
 565   __ negptr(reg);
 566 }
 567 
 568 
 569 void TemplateTable::wide_iload() {
 570   transition(vtos, itos);
 571   locals_index_wide(rbx);
 572   __ movl(rax, iaddress(rbx));
 573 }
 574 
 575 
 576 void TemplateTable::wide_lload() {
 577   transition(vtos, ltos);
 578   locals_index_wide(rbx);
 579   __ movptr(rax, laddress(rbx));
 580   NOT_LP64(__ movl(rdx, haddress(rbx)));
 581 }
 582 
 583 
 584 void TemplateTable::wide_fload() {
 585   transition(vtos, ftos);
 586   locals_index_wide(rbx);
 587   __ fld_s(faddress(rbx));
 588 }
 589 
 590 
 591 void TemplateTable::wide_dload() {
 592   transition(vtos, dtos);
 593   locals_index_wide(rbx);
 594   __ fld_d(daddress(rbx));
 595 }
 596 
 597 
 598 void TemplateTable::wide_aload() {
 599   transition(vtos, atos);
 600   locals_index_wide(rbx);
 601   __ movptr(rax, aaddress(rbx));
 602 }
 603 
 604 void TemplateTable::index_check(Register array, Register index) {
 605   // Pop ptr into array
 606   __ pop_ptr(array);
 607   index_check_without_pop(array, index);
 608 }
 609 
 610 void TemplateTable::index_check_without_pop(Register array, Register index) {
 611   // destroys rbx,
 612   // check array
 613   __ null_check(array, arrayOopDesc::length_offset_in_bytes());
 614   LP64_ONLY(__ movslq(index, index));
 615   // check index
 616   __ cmpl(index, Address(array, arrayOopDesc::length_offset_in_bytes()));
 617   if (index != rbx) {
 618     // ??? convention: move aberrant index into rbx, for exception message
 619     assert(rbx != array, "different registers");
 620     __ mov(rbx, index);
 621   }
 622   __ jump_cc(Assembler::aboveEqual,
 623              ExternalAddress(Interpreter::_throw_ArrayIndexOutOfBoundsException_entry));
 624 }
 625 
 626 
 627 void TemplateTable::iaload() {
 628   transition(itos, itos);
 629   // rdx: array
 630   index_check(rdx, rax);  // kills rbx,
 631   // rax,: index
 632   __ movl(rax, Address(rdx, rax, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_INT)));
 633 }
 634 
 635 
 636 void TemplateTable::laload() {
 637   transition(itos, ltos);
 638   // rax,: index
 639   // rdx: array
 640   index_check(rdx, rax);
 641   __ mov(rbx, rax);
 642   // rbx,: index
 643   __ movptr(rax, Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 0 * wordSize));
 644   NOT_LP64(__ movl(rdx, Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 1 * wordSize)));
 645 }
 646 
 647 
 648 void TemplateTable::faload() {
 649   transition(itos, ftos);
 650   // rdx: array
 651   index_check(rdx, rax);  // kills rbx,
 652   // rax,: index
 653   __ fld_s(Address(rdx, rax, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 654 }
 655 
 656 
 657 void TemplateTable::daload() {
 658   transition(itos, dtos);
 659   // rdx: array
 660   index_check(rdx, rax);  // kills rbx,
 661   // rax,: index
 662   __ fld_d(Address(rdx, rax, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 663 }
 664 
 665 
 666 void TemplateTable::aaload() {
 667   transition(itos, atos);
 668   // rdx: array
 669   index_check(rdx, rax);  // kills rbx,
 670   // rax,: index
 671   __ movptr(rax, Address(rdx, rax, Address::times_ptr, arrayOopDesc::base_offset_in_bytes(T_OBJECT)));
 672 }
 673 
 674 
 675 void TemplateTable::baload() {
 676   transition(itos, itos);
 677   // rdx: array
 678   index_check(rdx, rax);  // kills rbx,
 679   // rax,: index
 680   // can do better code for P5 - fix this at some point
 681   __ load_signed_byte(rbx, Address(rdx, rax, Address::times_1, arrayOopDesc::base_offset_in_bytes(T_BYTE)));
 682   __ mov(rax, rbx);
 683 }
 684 
 685 
 686 void TemplateTable::caload() {
 687   transition(itos, itos);
 688   // rdx: array
 689   index_check(rdx, rax);  // kills rbx,
 690   // rax,: index
 691   // can do better code for P5 - may want to improve this at some point
 692   __ load_unsigned_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 693   __ mov(rax, rbx);
 694 }
 695 
 696 // iload followed by caload frequent pair
 697 void TemplateTable::fast_icaload() {
 698   transition(vtos, itos);
 699   // load index out of locals
 700   locals_index(rbx);
 701   __ movl(rax, iaddress(rbx));
 702 
 703   // rdx: array
 704   index_check(rdx, rax);
 705   // rax,: index
 706   __ load_unsigned_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 707   __ mov(rax, rbx);
 708 }
 709 
 710 void TemplateTable::saload() {
 711   transition(itos, itos);
 712   // rdx: array
 713   index_check(rdx, rax);  // kills rbx,
 714   // rax,: index
 715   // can do better code for P5 - may want to improve this at some point
 716   __ load_signed_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_SHORT)));
 717   __ mov(rax, rbx);
 718 }
 719 
 720 
 721 void TemplateTable::iload(int n) {
 722   transition(vtos, itos);
 723   __ movl(rax, iaddress(n));
 724 }
 725 
 726 
 727 void TemplateTable::lload(int n) {
 728   transition(vtos, ltos);
 729   __ movptr(rax, laddress(n));
 730   NOT_LP64(__ movptr(rdx, haddress(n)));
 731 }
 732 
 733 
 734 void TemplateTable::fload(int n) {
 735   transition(vtos, ftos);
 736   __ fld_s(faddress(n));
 737 }
 738 
 739 
 740 void TemplateTable::dload(int n) {
 741   transition(vtos, dtos);
 742   __ fld_d(daddress(n));
 743 }
 744 
 745 
 746 void TemplateTable::aload(int n) {
 747   transition(vtos, atos);
 748   __ movptr(rax, aaddress(n));
 749 }
 750 
 751 
 752 void TemplateTable::aload_0() {
 753   transition(vtos, atos);
 754   // According to bytecode histograms, the pairs:
 755   //
 756   // _aload_0, _fast_igetfield
 757   // _aload_0, _fast_agetfield
 758   // _aload_0, _fast_fgetfield
 759   //
 760   // occur frequently. If RewriteFrequentPairs is set, the (slow) _aload_0
 761   // bytecode checks if the next bytecode is either _fast_igetfield,
 762   // _fast_agetfield or _fast_fgetfield and then rewrites the
 763   // current bytecode into a pair bytecode; otherwise it rewrites the current
 764   // bytecode into _fast_aload_0 that doesn't do the pair check anymore.
 765   //
 766   // Note: If the next bytecode is _getfield, the rewrite must be delayed,
 767   //       otherwise we may miss an opportunity for a pair.
 768   //
 769   // Also rewrite frequent pairs
 770   //   aload_0, aload_1
 771   //   aload_0, iload_1
 772   // These bytecodes with a small amount of code are most profitable to rewrite
 773   if (RewriteFrequentPairs) {
 774     Label rewrite, done;
 775     // get next byte
 776     __ load_unsigned_byte(rbx, at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)));
 777 
 778     // do actual aload_0
 779     aload(0);
 780 
 781     // if _getfield then wait with rewrite
 782     __ cmpl(rbx, Bytecodes::_getfield);
 783     __ jcc(Assembler::equal, done);
 784 
 785     // if _igetfield then reqrite to _fast_iaccess_0
 786     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 787     __ cmpl(rbx, Bytecodes::_fast_igetfield);
 788     __ movl(rcx, Bytecodes::_fast_iaccess_0);
 789     __ jccb(Assembler::equal, rewrite);
 790 
 791     // if _agetfield then reqrite to _fast_aaccess_0
 792     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 793     __ cmpl(rbx, Bytecodes::_fast_agetfield);
 794     __ movl(rcx, Bytecodes::_fast_aaccess_0);
 795     __ jccb(Assembler::equal, rewrite);
 796 
 797     // if _fgetfield then reqrite to _fast_faccess_0
 798     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 799     __ cmpl(rbx, Bytecodes::_fast_fgetfield);
 800     __ movl(rcx, Bytecodes::_fast_faccess_0);
 801     __ jccb(Assembler::equal, rewrite);
 802 
 803     // else rewrite to _fast_aload0
 804     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) == Bytecodes::_aload_0, "fix bytecode definition");
 805     __ movl(rcx, Bytecodes::_fast_aload_0);
 806 
 807     // rewrite
 808     // rcx: fast bytecode
 809     __ bind(rewrite);
 810     patch_bytecode(Bytecodes::_aload_0, rcx, rbx, false);
 811 
 812     __ bind(done);
 813   } else {
 814     aload(0);
 815   }
 816 }
 817 
 818 void TemplateTable::istore() {
 819   transition(itos, vtos);
 820   locals_index(rbx);
 821   __ movl(iaddress(rbx), rax);
 822 }
 823 
 824 
 825 void TemplateTable::lstore() {
 826   transition(ltos, vtos);
 827   locals_index(rbx);
 828   __ movptr(laddress(rbx), rax);
 829   NOT_LP64(__ movptr(haddress(rbx), rdx));
 830 }
 831 
 832 
 833 void TemplateTable::fstore() {
 834   transition(ftos, vtos);
 835   locals_index(rbx);
 836   __ fstp_s(faddress(rbx));
 837 }
 838 
 839 
 840 void TemplateTable::dstore() {
 841   transition(dtos, vtos);
 842   locals_index(rbx);
 843   __ fstp_d(daddress(rbx));
 844 }
 845 
 846 
 847 void TemplateTable::astore() {
 848   transition(vtos, vtos);
 849   __ pop_ptr(rax);
 850   locals_index(rbx);
 851   __ movptr(aaddress(rbx), rax);
 852 }
 853 
 854 
 855 void TemplateTable::wide_istore() {
 856   transition(vtos, vtos);
 857   __ pop_i(rax);
 858   locals_index_wide(rbx);
 859   __ movl(iaddress(rbx), rax);
 860 }
 861 
 862 
 863 void TemplateTable::wide_lstore() {
 864   transition(vtos, vtos);
 865   __ pop_l(rax, rdx);
 866   locals_index_wide(rbx);
 867   __ movptr(laddress(rbx), rax);
 868   NOT_LP64(__ movl(haddress(rbx), rdx));
 869 }
 870 
 871 
 872 void TemplateTable::wide_fstore() {
 873   wide_istore();
 874 }
 875 
 876 
 877 void TemplateTable::wide_dstore() {
 878   wide_lstore();
 879 }
 880 
 881 
 882 void TemplateTable::wide_astore() {
 883   transition(vtos, vtos);
 884   __ pop_ptr(rax);
 885   locals_index_wide(rbx);
 886   __ movptr(aaddress(rbx), rax);
 887 }
 888 
 889 
 890 void TemplateTable::iastore() {
 891   transition(itos, vtos);
 892   __ pop_i(rbx);
 893   // rax,: value
 894   // rdx: array
 895   index_check(rdx, rbx);  // prefer index in rbx,
 896   // rbx,: index
 897   __ movl(Address(rdx, rbx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_INT)), rax);
 898 }
 899 
 900 
 901 void TemplateTable::lastore() {
 902   transition(ltos, vtos);
 903   __ pop_i(rbx);
 904   // rax,: low(value)
 905   // rcx: array
 906   // rdx: high(value)
 907   index_check(rcx, rbx);  // prefer index in rbx,
 908   // rbx,: index
 909   __ movptr(Address(rcx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 0 * wordSize), rax);
 910   NOT_LP64(__ movl(Address(rcx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 1 * wordSize), rdx));
 911 }
 912 
 913 
 914 void TemplateTable::fastore() {
 915   transition(ftos, vtos);
 916   __ pop_i(rbx);
 917   // rdx: array
 918   // st0: value
 919   index_check(rdx, rbx);  // prefer index in rbx,
 920   // rbx,: index
 921   __ fstp_s(Address(rdx, rbx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 922 }
 923 
 924 
 925 void TemplateTable::dastore() {
 926   transition(dtos, vtos);
 927   __ pop_i(rbx);
 928   // rdx: array
 929   // st0: value
 930   index_check(rdx, rbx);  // prefer index in rbx,
 931   // rbx,: index
 932   __ fstp_d(Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 933 }
 934 
 935 
 936 void TemplateTable::aastore() {
 937   Label is_null, ok_is_subtype, done;
 938   transition(vtos, vtos);
 939   // stack: ..., array, index, value
 940   __ movptr(rax, at_tos());     // Value
 941   __ movl(rcx, at_tos_p1());  // Index
 942   __ movptr(rdx, at_tos_p2());  // Array
 943 
 944   Address element_address(rdx, rcx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 945   index_check_without_pop(rdx, rcx);      // kills rbx,
 946   // do array store check - check for NULL value first
 947   __ testptr(rax, rax);
 948   __ jcc(Assembler::zero, is_null);
 949 
 950   // Move subklass into EBX
 951   __ load_klass(rbx, rax);
 952   // Move superklass into EAX
 953   __ load_klass(rax, rdx);
 954   __ movptr(rax, Address(rax, ObjArrayKlass::element_klass_offset()));
 955   // Compress array+index*wordSize+12 into a single register.  Frees ECX.
 956   __ lea(rdx, element_address);
 957 
 958   // Generate subtype check.  Blows ECX.  Resets EDI to locals.
 959   // Superklass in EAX.  Subklass in EBX.
 960   __ gen_subtype_check( rbx, ok_is_subtype );
 961 
 962   // Come here on failure
 963   // object is at TOS
 964   __ jump(ExternalAddress(Interpreter::_throw_ArrayStoreException_entry));
 965 
 966   // Come here on success
 967   __ bind(ok_is_subtype);
 968 
 969   // Get the value to store
 970   __ movptr(rax, at_rsp());
 971   // and store it with appropriate barrier
 972   do_oop_store(_masm, Address(rdx, 0), rax, _bs->kind(), true);
 973 
 974   __ jmp(done);
 975 
 976   // Have a NULL in EAX, EDX=array, ECX=index.  Store NULL at ary[idx]
 977   __ bind(is_null);
 978   __ profile_null_seen(rbx);
 979 
 980   // Store NULL, (noreg means NULL to do_oop_store)
 981   do_oop_store(_masm, element_address, noreg, _bs->kind(), true);
 982 
 983   // Pop stack arguments
 984   __ bind(done);
 985   __ addptr(rsp, 3 * Interpreter::stackElementSize);
 986 }
 987 
 988 
 989 void TemplateTable::bastore() {
 990   transition(itos, vtos);
 991   __ pop_i(rbx);
 992   // rax: value
 993   // rbx: index
 994   // rdx: array
 995   index_check(rdx, rbx);  // prefer index in rbx
 996   // Need to check whether array is boolean or byte
 997   // since both types share the bastore bytecode.
 998   __ load_klass(rcx, rdx);
 999   __ movl(rcx, Address(rcx, Klass::layout_helper_offset()));
1000   int diffbit = Klass::layout_helper_boolean_diffbit();
1001   __ testl(rcx, diffbit);
1002   Label L_skip;
1003   __ jccb(Assembler::zero, L_skip);
1004   __ andl(rax, 1);  // if it is a T_BOOLEAN array, mask the stored value to 0/1
1005   __ bind(L_skip);
1006  __ movb(Address(rdx, rbx, Address::times_1, arrayOopDesc::base_offset_in_bytes(T_BYTE)), rax);
1007 }
1008 
1009 
1010 void TemplateTable::castore() {
1011   transition(itos, vtos);
1012   __ pop_i(rbx);
1013   // rax,: value
1014   // rdx: array
1015   index_check(rdx, rbx);  // prefer index in rbx,
1016   // rbx,: index
1017   __ movw(Address(rdx, rbx, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)), rax);
1018 }
1019 
1020 
1021 void TemplateTable::sastore() {
1022   castore();
1023 }
1024 
1025 
1026 void TemplateTable::istore(int n) {
1027   transition(itos, vtos);
1028   __ movl(iaddress(n), rax);
1029 }
1030 
1031 
1032 void TemplateTable::lstore(int n) {
1033   transition(ltos, vtos);
1034   __ movptr(laddress(n), rax);
1035   NOT_LP64(__ movptr(haddress(n), rdx));
1036 }
1037 
1038 
1039 void TemplateTable::fstore(int n) {
1040   transition(ftos, vtos);
1041   __ fstp_s(faddress(n));
1042 }
1043 
1044 
1045 void TemplateTable::dstore(int n) {
1046   transition(dtos, vtos);
1047   __ fstp_d(daddress(n));
1048 }
1049 
1050 
1051 void TemplateTable::astore(int n) {
1052   transition(vtos, vtos);
1053   __ pop_ptr(rax);
1054   __ movptr(aaddress(n), rax);
1055 }
1056 
1057 
1058 void TemplateTable::pop() {
1059   transition(vtos, vtos);
1060   __ addptr(rsp, Interpreter::stackElementSize);
1061 }
1062 
1063 
1064 void TemplateTable::pop2() {
1065   transition(vtos, vtos);
1066   __ addptr(rsp, 2*Interpreter::stackElementSize);
1067 }
1068 
1069 
1070 void TemplateTable::dup() {
1071   transition(vtos, vtos);
1072   // stack: ..., a
1073   __ load_ptr(0, rax);
1074   __ push_ptr(rax);
1075   // stack: ..., a, a
1076 }
1077 
1078 
1079 void TemplateTable::dup_x1() {
1080   transition(vtos, vtos);
1081   // stack: ..., a, b
1082   __ load_ptr( 0, rax);  // load b
1083   __ load_ptr( 1, rcx);  // load a
1084   __ store_ptr(1, rax);  // store b
1085   __ store_ptr(0, rcx);  // store a
1086   __ push_ptr(rax);      // push b
1087   // stack: ..., b, a, b
1088 }
1089 
1090 
1091 void TemplateTable::dup_x2() {
1092   transition(vtos, vtos);
1093   // stack: ..., a, b, c
1094   __ load_ptr( 0, rax);  // load c
1095   __ load_ptr( 2, rcx);  // load a
1096   __ store_ptr(2, rax);  // store c in a
1097   __ push_ptr(rax);      // push c
1098   // stack: ..., c, b, c, c
1099   __ load_ptr( 2, rax);  // load b
1100   __ store_ptr(2, rcx);  // store a in b
1101   // stack: ..., c, a, c, c
1102   __ store_ptr(1, rax);  // store b in c
1103   // stack: ..., c, a, b, c
1104 }
1105 
1106 
1107 void TemplateTable::dup2() {
1108   transition(vtos, vtos);
1109   // stack: ..., a, b
1110   __ load_ptr(1, rax);  // load a
1111   __ push_ptr(rax);     // push a
1112   __ load_ptr(1, rax);  // load b
1113   __ push_ptr(rax);     // push b
1114   // stack: ..., a, b, a, b
1115 }
1116 
1117 
1118 void TemplateTable::dup2_x1() {
1119   transition(vtos, vtos);
1120   // stack: ..., a, b, c
1121   __ load_ptr( 0, rcx);  // load c
1122   __ load_ptr( 1, rax);  // load b
1123   __ push_ptr(rax);      // push b
1124   __ push_ptr(rcx);      // push c
1125   // stack: ..., a, b, c, b, c
1126   __ store_ptr(3, rcx);  // store c in b
1127   // stack: ..., a, c, c, b, c
1128   __ load_ptr( 4, rcx);  // load a
1129   __ store_ptr(2, rcx);  // store a in 2nd c
1130   // stack: ..., a, c, a, b, c
1131   __ store_ptr(4, rax);  // store b in a
1132   // stack: ..., b, c, a, b, c
1133   // stack: ..., b, c, a, b, c
1134 }
1135 
1136 
1137 void TemplateTable::dup2_x2() {
1138   transition(vtos, vtos);
1139   // stack: ..., a, b, c, d
1140   __ load_ptr( 0, rcx);  // load d
1141   __ load_ptr( 1, rax);  // load c
1142   __ push_ptr(rax);      // push c
1143   __ push_ptr(rcx);      // push d
1144   // stack: ..., a, b, c, d, c, d
1145   __ load_ptr( 4, rax);  // load b
1146   __ store_ptr(2, rax);  // store b in d
1147   __ store_ptr(4, rcx);  // store d in b
1148   // stack: ..., a, d, c, b, c, d
1149   __ load_ptr( 5, rcx);  // load a
1150   __ load_ptr( 3, rax);  // load c
1151   __ store_ptr(3, rcx);  // store a in c
1152   __ store_ptr(5, rax);  // store c in a
1153   // stack: ..., c, d, a, b, c, d
1154   // stack: ..., c, d, a, b, c, d
1155 }
1156 
1157 
1158 void TemplateTable::swap() {
1159   transition(vtos, vtos);
1160   // stack: ..., a, b
1161   __ load_ptr( 1, rcx);  // load a
1162   __ load_ptr( 0, rax);  // load b
1163   __ store_ptr(0, rcx);  // store a in b
1164   __ store_ptr(1, rax);  // store b in a
1165   // stack: ..., b, a
1166 }
1167 
1168 
1169 void TemplateTable::iop2(Operation op) {
1170   transition(itos, itos);
1171   switch (op) {
1172     case add  :                   __ pop_i(rdx); __ addl (rax, rdx); break;
1173     case sub  : __ mov(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1174     case mul  :                   __ pop_i(rdx); __ imull(rax, rdx); break;
1175     case _and :                   __ pop_i(rdx); __ andl (rax, rdx); break;
1176     case _or  :                   __ pop_i(rdx); __ orl  (rax, rdx); break;
1177     case _xor :                   __ pop_i(rdx); __ xorl (rax, rdx); break;
1178     case shl  : __ mov(rcx, rax); __ pop_i(rax); __ shll (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1179     case shr  : __ mov(rcx, rax); __ pop_i(rax); __ sarl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1180     case ushr : __ mov(rcx, rax); __ pop_i(rax); __ shrl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1181     default   : ShouldNotReachHere();
1182   }
1183 }
1184 
1185 
1186 void TemplateTable::lop2(Operation op) {
1187   transition(ltos, ltos);
1188   __ pop_l(rbx, rcx);
1189   switch (op) {
1190     case add  : __ addl(rax, rbx); __ adcl(rdx, rcx); break;
1191     case sub  : __ subl(rbx, rax); __ sbbl(rcx, rdx);
1192                 __ mov (rax, rbx); __ mov (rdx, rcx); break;
1193     case _and : __ andl(rax, rbx); __ andl(rdx, rcx); break;
1194     case _or  : __ orl (rax, rbx); __ orl (rdx, rcx); break;
1195     case _xor : __ xorl(rax, rbx); __ xorl(rdx, rcx); break;
1196     default   : ShouldNotReachHere();
1197   }
1198 }
1199 
1200 
1201 void TemplateTable::idiv() {
1202   transition(itos, itos);
1203   __ mov(rcx, rax);
1204   __ pop_i(rax);
1205   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1206   //       they are not equal, one could do a normal division (no correction
1207   //       needed), which may speed up this implementation for the common case.
1208   //       (see also JVM spec., p.243 & p.271)
1209   __ corrected_idivl(rcx);
1210 }
1211 
1212 
1213 void TemplateTable::irem() {
1214   transition(itos, itos);
1215   __ mov(rcx, rax);
1216   __ pop_i(rax);
1217   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1218   //       they are not equal, one could do a normal division (no correction
1219   //       needed), which may speed up this implementation for the common case.
1220   //       (see also JVM spec., p.243 & p.271)
1221   __ corrected_idivl(rcx);
1222   __ mov(rax, rdx);
1223 }
1224 
1225 
1226 void TemplateTable::lmul() {
1227   transition(ltos, ltos);
1228   __ pop_l(rbx, rcx);
1229   __ push(rcx); __ push(rbx);
1230   __ push(rdx); __ push(rax);
1231   __ lmul(2 * wordSize, 0);
1232   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1233 }
1234 
1235 
1236 void TemplateTable::ldiv() {
1237   transition(ltos, ltos);
1238   __ pop_l(rbx, rcx);
1239   __ push(rcx); __ push(rbx);
1240   __ push(rdx); __ push(rax);
1241   // check if y = 0
1242   __ orl(rax, rdx);
1243   __ jump_cc(Assembler::zero,
1244              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1245   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::ldiv));
1246   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1247 }
1248 
1249 
1250 void TemplateTable::lrem() {
1251   transition(ltos, ltos);
1252   __ pop_l(rbx, rcx);
1253   __ push(rcx); __ push(rbx);
1254   __ push(rdx); __ push(rax);
1255   // check if y = 0
1256   __ orl(rax, rdx);
1257   __ jump_cc(Assembler::zero,
1258              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1259   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::lrem));
1260   __ addptr(rsp, 4 * wordSize);
1261 }
1262 
1263 
1264 void TemplateTable::lshl() {
1265   transition(itos, ltos);
1266   __ movl(rcx, rax);                             // get shift count
1267   __ pop_l(rax, rdx);                            // get shift value
1268   __ lshl(rdx, rax);
1269 }
1270 
1271 
1272 void TemplateTable::lshr() {
1273   transition(itos, ltos);
1274   __ mov(rcx, rax);                              // get shift count
1275   __ pop_l(rax, rdx);                            // get shift value
1276   __ lshr(rdx, rax, true);
1277 }
1278 
1279 
1280 void TemplateTable::lushr() {
1281   transition(itos, ltos);
1282   __ mov(rcx, rax);                              // get shift count
1283   __ pop_l(rax, rdx);                            // get shift value
1284   __ lshr(rdx, rax);
1285 }
1286 
1287 
1288 void TemplateTable::fop2(Operation op) {
1289   transition(ftos, ftos);
1290   switch (op) {
1291     case add: __ fadd_s (at_rsp());                break;
1292     case sub: __ fsubr_s(at_rsp());                break;
1293     case mul: __ fmul_s (at_rsp());                break;
1294     case div: __ fdivr_s(at_rsp());                break;
1295     case rem: __ fld_s  (at_rsp()); __ fremr(rax); break;
1296     default : ShouldNotReachHere();
1297   }
1298   __ f2ieee();
1299   __ pop(rax);  // pop float thing off
1300 }
1301 
1302 
1303 void TemplateTable::dop2(Operation op) {
1304   transition(dtos, dtos);
1305 
1306   switch (op) {
1307     case add: __ fadd_d (at_rsp());                break;
1308     case sub: __ fsubr_d(at_rsp());                break;
1309     case mul: {
1310       Label L_strict;
1311       Label L_join;
1312       const Address access_flags      (rcx, Method::access_flags_offset());
1313       __ get_method(rcx);
1314       __ movl(rcx, access_flags);
1315       __ testl(rcx, JVM_ACC_STRICT);
1316       __ jccb(Assembler::notZero, L_strict);
1317       __ fmul_d (at_rsp());
1318       __ jmpb(L_join);
1319       __ bind(L_strict);
1320       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1321       __ fmulp();
1322       __ fmul_d (at_rsp());
1323       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1324       __ fmulp();
1325       __ bind(L_join);
1326       break;
1327     }
1328     case div: {
1329       Label L_strict;
1330       Label L_join;
1331       const Address access_flags      (rcx, Method::access_flags_offset());
1332       __ get_method(rcx);
1333       __ movl(rcx, access_flags);
1334       __ testl(rcx, JVM_ACC_STRICT);
1335       __ jccb(Assembler::notZero, L_strict);
1336       __ fdivr_d(at_rsp());
1337       __ jmp(L_join);
1338       __ bind(L_strict);
1339       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1340       __ fmul_d (at_rsp());
1341       __ fdivrp();
1342       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1343       __ fmulp();
1344       __ bind(L_join);
1345       break;
1346     }
1347     case rem: __ fld_d  (at_rsp()); __ fremr(rax); break;
1348     default : ShouldNotReachHere();
1349   }
1350   __ d2ieee();
1351   // Pop double precision number from rsp.
1352   __ pop(rax);
1353   __ pop(rdx);
1354 }
1355 
1356 
1357 void TemplateTable::ineg() {
1358   transition(itos, itos);
1359   __ negl(rax);
1360 }
1361 
1362 
1363 void TemplateTable::lneg() {
1364   transition(ltos, ltos);
1365   __ lneg(rdx, rax);
1366 }
1367 
1368 
1369 void TemplateTable::fneg() {
1370   transition(ftos, ftos);
1371   __ fchs();
1372 }
1373 
1374 
1375 void TemplateTable::dneg() {
1376   transition(dtos, dtos);
1377   __ fchs();
1378 }
1379 
1380 
1381 void TemplateTable::iinc() {
1382   transition(vtos, vtos);
1383   __ load_signed_byte(rdx, at_bcp(2));           // get constant
1384   locals_index(rbx);
1385   __ addl(iaddress(rbx), rdx);
1386 }
1387 
1388 
1389 void TemplateTable::wide_iinc() {
1390   transition(vtos, vtos);
1391   __ movl(rdx, at_bcp(4));                       // get constant
1392   locals_index_wide(rbx);
1393   __ bswapl(rdx);                                 // swap bytes & sign-extend constant
1394   __ sarl(rdx, 16);
1395   __ addl(iaddress(rbx), rdx);
1396   // Note: should probably use only one movl to get both
1397   //       the index and the constant -> fix this
1398 }
1399 
1400 
1401 void TemplateTable::convert() {
1402   // Checking
1403 #ifdef ASSERT
1404   { TosState tos_in  = ilgl;
1405     TosState tos_out = ilgl;
1406     switch (bytecode()) {
1407       case Bytecodes::_i2l: // fall through
1408       case Bytecodes::_i2f: // fall through
1409       case Bytecodes::_i2d: // fall through
1410       case Bytecodes::_i2b: // fall through
1411       case Bytecodes::_i2c: // fall through
1412       case Bytecodes::_i2s: tos_in = itos; break;
1413       case Bytecodes::_l2i: // fall through
1414       case Bytecodes::_l2f: // fall through
1415       case Bytecodes::_l2d: tos_in = ltos; break;
1416       case Bytecodes::_f2i: // fall through
1417       case Bytecodes::_f2l: // fall through
1418       case Bytecodes::_f2d: tos_in = ftos; break;
1419       case Bytecodes::_d2i: // fall through
1420       case Bytecodes::_d2l: // fall through
1421       case Bytecodes::_d2f: tos_in = dtos; break;
1422       default             : ShouldNotReachHere();
1423     }
1424     switch (bytecode()) {
1425       case Bytecodes::_l2i: // fall through
1426       case Bytecodes::_f2i: // fall through
1427       case Bytecodes::_d2i: // fall through
1428       case Bytecodes::_i2b: // fall through
1429       case Bytecodes::_i2c: // fall through
1430       case Bytecodes::_i2s: tos_out = itos; break;
1431       case Bytecodes::_i2l: // fall through
1432       case Bytecodes::_f2l: // fall through
1433       case Bytecodes::_d2l: tos_out = ltos; break;
1434       case Bytecodes::_i2f: // fall through
1435       case Bytecodes::_l2f: // fall through
1436       case Bytecodes::_d2f: tos_out = ftos; break;
1437       case Bytecodes::_i2d: // fall through
1438       case Bytecodes::_l2d: // fall through
1439       case Bytecodes::_f2d: tos_out = dtos; break;
1440       default             : ShouldNotReachHere();
1441     }
1442     transition(tos_in, tos_out);
1443   }
1444 #endif // ASSERT
1445 
1446   // Conversion
1447   // (Note: use push(rcx)/pop(rcx) for 1/2-word stack-ptr manipulation)
1448   switch (bytecode()) {
1449     case Bytecodes::_i2l:
1450       __ extend_sign(rdx, rax);
1451       break;
1452     case Bytecodes::_i2f:
1453       __ push(rax);          // store int on tos
1454       __ fild_s(at_rsp());   // load int to ST0
1455       __ f2ieee();           // truncate to float size
1456       __ pop(rcx);           // adjust rsp
1457       break;
1458     case Bytecodes::_i2d:
1459       __ push(rax);          // add one slot for d2ieee()
1460       __ push(rax);          // store int on tos
1461       __ fild_s(at_rsp());   // load int to ST0
1462       __ d2ieee();           // truncate to double size
1463       __ pop(rcx);           // adjust rsp
1464       __ pop(rcx);
1465       break;
1466     case Bytecodes::_i2b:
1467       __ shll(rax, 24);      // truncate upper 24 bits
1468       __ sarl(rax, 24);      // and sign-extend byte
1469       LP64_ONLY(__ movsbl(rax, rax));
1470       break;
1471     case Bytecodes::_i2c:
1472       __ andl(rax, 0xFFFF);  // truncate upper 16 bits
1473       LP64_ONLY(__ movzwl(rax, rax));
1474       break;
1475     case Bytecodes::_i2s:
1476       __ shll(rax, 16);      // truncate upper 16 bits
1477       __ sarl(rax, 16);      // and sign-extend short
1478       LP64_ONLY(__ movswl(rax, rax));
1479       break;
1480     case Bytecodes::_l2i:
1481       /* nothing to do */
1482       break;
1483     case Bytecodes::_l2f:
1484       __ push(rdx);          // store long on tos
1485       __ push(rax);
1486       __ fild_d(at_rsp());   // load long to ST0
1487       __ f2ieee();           // truncate to float size
1488       __ pop(rcx);           // adjust rsp
1489       __ pop(rcx);
1490       break;
1491     case Bytecodes::_l2d:
1492       __ push(rdx);          // store long on tos
1493       __ push(rax);
1494       __ fild_d(at_rsp());   // load long to ST0
1495       __ d2ieee();           // truncate to double size
1496       __ pop(rcx);           // adjust rsp
1497       __ pop(rcx);
1498       break;
1499     case Bytecodes::_f2i:
1500       __ push(rcx);          // reserve space for argument
1501       __ fstp_s(at_rsp());   // pass float argument on stack
1502       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1503       break;
1504     case Bytecodes::_f2l:
1505       __ push(rcx);          // reserve space for argument
1506       __ fstp_s(at_rsp());   // pass float argument on stack
1507       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1508       break;
1509     case Bytecodes::_f2d:
1510       /* nothing to do */
1511       break;
1512     case Bytecodes::_d2i:
1513       __ push(rcx);          // reserve space for argument
1514       __ push(rcx);
1515       __ fstp_d(at_rsp());   // pass double argument on stack
1516       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 2);
1517       break;
1518     case Bytecodes::_d2l:
1519       __ push(rcx);          // reserve space for argument
1520       __ push(rcx);
1521       __ fstp_d(at_rsp());   // pass double argument on stack
1522       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 2);
1523       break;
1524     case Bytecodes::_d2f:
1525       __ push(rcx);          // reserve space for f2ieee()
1526       __ f2ieee();           // truncate to float size
1527       __ pop(rcx);           // adjust rsp
1528       break;
1529     default             :
1530       ShouldNotReachHere();
1531   }
1532 }
1533 
1534 
1535 void TemplateTable::lcmp() {
1536   transition(ltos, itos);
1537   // y = rdx:rax
1538   __ pop_l(rbx, rcx);             // get x = rcx:rbx
1539   __ lcmp2int(rcx, rbx, rdx, rax);// rcx := cmp(x, y)
1540   __ mov(rax, rcx);
1541 }
1542 
1543 
1544 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1545   if (is_float) {
1546     __ fld_s(at_rsp());
1547   } else {
1548     __ fld_d(at_rsp());
1549     __ pop(rdx);
1550   }
1551   __ pop(rcx);
1552   __ fcmp2int(rax, unordered_result < 0);
1553 }
1554 
1555 
1556 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1557   __ get_method(rcx);           // ECX holds method
1558   __ profile_taken_branch(rax,rbx); // EAX holds updated MDP, EBX holds bumped taken count
1559 
1560   const ByteSize be_offset = MethodCounters::backedge_counter_offset() +
1561                              InvocationCounter::counter_offset();
1562   const ByteSize inv_offset = MethodCounters::invocation_counter_offset() +
1563                               InvocationCounter::counter_offset();
1564 
1565   // Load up EDX with the branch displacement
1566   if (is_wide) {
1567     __ movl(rdx, at_bcp(1));
1568   } else {
1569     __ load_signed_short(rdx, at_bcp(1));
1570   }
1571   __ bswapl(rdx);
1572   if (!is_wide) __ sarl(rdx, 16);
1573   LP64_ONLY(__ movslq(rdx, rdx));
1574 
1575 
1576   // Handle all the JSR stuff here, then exit.
1577   // It's much shorter and cleaner than intermingling with the
1578   // non-JSR normal-branch stuff occurring below.
1579   if (is_jsr) {
1580     // Pre-load the next target bytecode into EBX
1581     __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1, 0));
1582 
1583     // compute return address as bci in rax,
1584     __ lea(rax, at_bcp((is_wide ? 5 : 3) - in_bytes(ConstMethod::codes_offset())));
1585     __ subptr(rax, Address(rcx, Method::const_offset()));
1586     // Adjust the bcp in RSI by the displacement in EDX
1587     __ addptr(rsi, rdx);
1588     // Push return address
1589     __ push_i(rax);
1590     // jsr returns vtos
1591     __ dispatch_only_noverify(vtos);
1592     return;
1593   }
1594 
1595   // Normal (non-jsr) branch handling
1596 
1597   // Adjust the bcp in RSI by the displacement in EDX
1598   __ addptr(rsi, rdx);
1599 
1600   assert(UseLoopCounter || !UseOnStackReplacement, "on-stack-replacement requires loop counters");
1601   Label backedge_counter_overflow;
1602   Label profile_method;
1603   Label dispatch;
1604   if (UseLoopCounter) {
1605     // increment backedge counter for backward branches
1606     // rax,: MDO
1607     // rbx,: MDO bumped taken-count
1608     // rcx: method
1609     // rdx: target offset
1610     // rsi: target bcp
1611     // rdi: locals pointer
1612     __ testl(rdx, rdx);             // check if forward or backward branch
1613     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1614 
1615     // check if MethodCounters exists
1616     Label has_counters;
1617     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1618     __ testptr(rax, rax);
1619     __ jcc(Assembler::notZero, has_counters);
1620     __ push(rdx);
1621     __ push(rcx);
1622     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::build_method_counters),
1623                rcx);
1624     __ pop(rcx);
1625     __ pop(rdx);
1626     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1627     __ testptr(rax, rax);
1628     __ jcc(Assembler::zero, dispatch);
1629     __ bind(has_counters);
1630 
1631     if (TieredCompilation) {
1632       Label no_mdo;
1633       int increment = InvocationCounter::count_increment;
1634       int mask = ((1 << Tier0BackedgeNotifyFreqLog) - 1) << InvocationCounter::count_shift;
1635       if (ProfileInterpreter) {
1636         // Are we profiling?
1637         __ movptr(rbx, Address(rcx, in_bytes(Method::method_data_offset())));
1638         __ testptr(rbx, rbx);
1639         __ jccb(Assembler::zero, no_mdo);
1640         // Increment the MDO backedge counter
1641         const Address mdo_backedge_counter(rbx, in_bytes(MethodData::backedge_counter_offset()) +
1642                                                 in_bytes(InvocationCounter::counter_offset()));
1643         __ increment_mask_and_jump(mdo_backedge_counter, increment, mask, rax, false, Assembler::zero,
1644                                    UseOnStackReplacement ? &backedge_counter_overflow : NULL);
1645         __ jmp(dispatch);
1646       }
1647       __ bind(no_mdo);
1648       // Increment backedge counter in MethodCounters*
1649       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1650       __ increment_mask_and_jump(Address(rcx, be_offset), increment, mask,
1651                                  rax, false, Assembler::zero,
1652                                  UseOnStackReplacement ? &backedge_counter_overflow : NULL);
1653     } else {
1654       // increment counter
1655       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1656       __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1657       __ incrementl(rax, InvocationCounter::count_increment); // increment counter
1658       __ movl(Address(rcx, be_offset), rax);        // store counter
1659 
1660       __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1661 
1662       __ andl(rax, InvocationCounter::count_mask_value);     // and the status bits
1663       __ addl(rax, Address(rcx, be_offset));        // add both counters
1664 
1665       if (ProfileInterpreter) {
1666         // Test to see if we should create a method data oop
1667         __ cmp32(rax,
1668                  ExternalAddress((address) &InvocationCounter::InterpreterProfileLimit));
1669         __ jcc(Assembler::less, dispatch);
1670 
1671         // if no method data exists, go to profile method
1672         __ test_method_data_pointer(rax, profile_method);
1673 
1674         if (UseOnStackReplacement) {
1675           // check for overflow against rbx, which is the MDO taken count
1676           __ cmp32(rbx,
1677                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1678           __ jcc(Assembler::below, dispatch);
1679 
1680           // When ProfileInterpreter is on, the backedge_count comes from the
1681           // MethodData*, which value does not get reset on the call to
1682           // frequency_counter_overflow().  To avoid excessive calls to the overflow
1683           // routine while the method is being compiled, add a second test to make
1684           // sure the overflow function is called only once every overflow_frequency.
1685           const int overflow_frequency = 1024;
1686           __ andptr(rbx, overflow_frequency-1);
1687           __ jcc(Assembler::zero, backedge_counter_overflow);
1688         }
1689       } else {
1690         if (UseOnStackReplacement) {
1691           // check for overflow against rax, which is the sum of the counters
1692           __ cmp32(rax,
1693                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1694           __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1695 
1696         }
1697       }
1698     }
1699     __ bind(dispatch);
1700   }
1701 
1702   // Pre-load the next target bytecode into EBX
1703   __ load_unsigned_byte(rbx, Address(rsi, 0));
1704 
1705   // continue with the bytecode @ target
1706   // rax,: return bci for jsr's, unused otherwise
1707   // rbx,: target bytecode
1708   // rsi: target bcp
1709   __ dispatch_only(vtos);
1710 
1711   if (UseLoopCounter) {
1712     if (ProfileInterpreter) {
1713       // Out-of-line code to allocate method data oop.
1714       __ bind(profile_method);
1715       __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::profile_method));
1716       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1717       __ set_method_data_pointer_for_bcp();
1718       __ jmp(dispatch);
1719     }
1720 
1721     if (UseOnStackReplacement) {
1722 
1723       // invocation counter overflow
1724       __ bind(backedge_counter_overflow);
1725       __ negptr(rdx);
1726       __ addptr(rdx, rsi);        // branch bcp
1727       call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::frequency_counter_overflow), rdx);
1728       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1729 
1730       // rax,: osr nmethod (osr ok) or NULL (osr not possible)
1731       // rbx,: target bytecode
1732       // rdx: scratch
1733       // rdi: locals pointer
1734       // rsi: bcp
1735       __ testptr(rax, rax);                      // test result
1736       __ jcc(Assembler::zero, dispatch);         // no osr if null
1737       // nmethod may have been invalidated (VM may block upon call_VM return)
1738       __ movl(rcx, Address(rax, nmethod::entry_bci_offset()));
1739       __ cmpl(rcx, InvalidOSREntryBci);
1740       __ jcc(Assembler::equal, dispatch);
1741 
1742       // We have the address of an on stack replacement routine in rax,
1743       // We need to prepare to execute the OSR method. First we must
1744       // migrate the locals and monitors off of the stack.
1745 
1746       __ mov(rbx, rax);                             // save the nmethod
1747 
1748       const Register thread = rcx;
1749       __ get_thread(thread);
1750       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1751       // rax, is OSR buffer, move it to expected parameter location
1752       __ mov(rcx, rax);
1753 
1754       // pop the interpreter frame
1755       __ movptr(rdx, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1756       __ leave();                                // remove frame anchor
1757       __ pop(rdi);                               // get return address
1758       __ mov(rsp, rdx);                          // set sp to sender sp
1759 
1760       // Align stack pointer for compiled code (note that caller is
1761       // responsible for undoing this fixup by remembering the old SP
1762       // in an rbp,-relative location)
1763       __ andptr(rsp, -(StackAlignmentInBytes));
1764 
1765       // push the (possibly adjusted) return address
1766       __ push(rdi);
1767 
1768       // and begin the OSR nmethod
1769       __ jmp(Address(rbx, nmethod::osr_entry_point_offset()));
1770     }
1771   }
1772 }
1773 
1774 
1775 void TemplateTable::if_0cmp(Condition cc) {
1776   transition(itos, vtos);
1777   // assume branch is more often taken than not (loops use backward branches)
1778   Label not_taken;
1779   __ testl(rax, rax);
1780   __ jcc(j_not(cc), not_taken);
1781   branch(false, false);
1782   __ bind(not_taken);
1783   __ profile_not_taken_branch(rax);
1784 }
1785 
1786 
1787 void TemplateTable::if_icmp(Condition cc) {
1788   transition(itos, vtos);
1789   // assume branch is more often taken than not (loops use backward branches)
1790   Label not_taken;
1791   __ pop_i(rdx);
1792   __ cmpl(rdx, rax);
1793   __ jcc(j_not(cc), not_taken);
1794   branch(false, false);
1795   __ bind(not_taken);
1796   __ profile_not_taken_branch(rax);
1797 }
1798 
1799 
1800 void TemplateTable::if_nullcmp(Condition cc) {
1801   transition(atos, vtos);
1802   // assume branch is more often taken than not (loops use backward branches)
1803   Label not_taken;
1804   __ testptr(rax, rax);
1805   __ jcc(j_not(cc), not_taken);
1806   branch(false, false);
1807   __ bind(not_taken);
1808   __ profile_not_taken_branch(rax);
1809 }
1810 
1811 
1812 void TemplateTable::if_acmp(Condition cc) {
1813   transition(atos, vtos);
1814   // assume branch is more often taken than not (loops use backward branches)
1815   Label not_taken;
1816   __ pop_ptr(rdx);
1817   __ cmpptr(rdx, rax);
1818   __ jcc(j_not(cc), not_taken);
1819   branch(false, false);
1820   __ bind(not_taken);
1821   __ profile_not_taken_branch(rax);
1822 }
1823 
1824 
1825 void TemplateTable::ret() {
1826   transition(vtos, vtos);
1827   locals_index(rbx);
1828   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1829   __ profile_ret(rbx, rcx);
1830   __ get_method(rax);
1831   __ movptr(rsi, Address(rax, Method::const_offset()));
1832   __ lea(rsi, Address(rsi, rbx, Address::times_1,
1833                       ConstMethod::codes_offset()));
1834   __ dispatch_next(vtos);
1835 }
1836 
1837 
1838 void TemplateTable::wide_ret() {
1839   transition(vtos, vtos);
1840   locals_index_wide(rbx);
1841   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1842   __ profile_ret(rbx, rcx);
1843   __ get_method(rax);
1844   __ movptr(rsi, Address(rax, Method::const_offset()));
1845   __ lea(rsi, Address(rsi, rbx, Address::times_1, ConstMethod::codes_offset()));
1846   __ dispatch_next(vtos);
1847 }
1848 
1849 
1850 void TemplateTable::tableswitch() {
1851   Label default_case, continue_execution;
1852   transition(itos, vtos);
1853   // align rsi
1854   __ lea(rbx, at_bcp(wordSize));
1855   __ andptr(rbx, -wordSize);
1856   // load lo & hi
1857   __ movl(rcx, Address(rbx, 1 * wordSize));
1858   __ movl(rdx, Address(rbx, 2 * wordSize));
1859   __ bswapl(rcx);
1860   __ bswapl(rdx);
1861   // check against lo & hi
1862   __ cmpl(rax, rcx);
1863   __ jccb(Assembler::less, default_case);
1864   __ cmpl(rax, rdx);
1865   __ jccb(Assembler::greater, default_case);
1866   // lookup dispatch offset
1867   __ subl(rax, rcx);
1868   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1869   __ profile_switch_case(rax, rbx, rcx);
1870   // continue execution
1871   __ bind(continue_execution);
1872   __ bswapl(rdx);
1873   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1874   __ addptr(rsi, rdx);
1875   __ dispatch_only(vtos);
1876   // handle default
1877   __ bind(default_case);
1878   __ profile_switch_default(rax);
1879   __ movl(rdx, Address(rbx, 0));
1880   __ jmp(continue_execution);
1881 }
1882 
1883 
1884 void TemplateTable::lookupswitch() {
1885   transition(itos, itos);
1886   __ stop("lookupswitch bytecode should have been rewritten");
1887 }
1888 
1889 
1890 void TemplateTable::fast_linearswitch() {
1891   transition(itos, vtos);
1892   Label loop_entry, loop, found, continue_execution;
1893   // bswapl rax, so we can avoid bswapping the table entries
1894   __ bswapl(rax);
1895   // align rsi
1896   __ lea(rbx, at_bcp(wordSize));                // btw: should be able to get rid of this instruction (change offsets below)
1897   __ andptr(rbx, -wordSize);
1898   // set counter
1899   __ movl(rcx, Address(rbx, wordSize));
1900   __ bswapl(rcx);
1901   __ jmpb(loop_entry);
1902   // table search
1903   __ bind(loop);
1904   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * wordSize));
1905   __ jccb(Assembler::equal, found);
1906   __ bind(loop_entry);
1907   __ decrementl(rcx);
1908   __ jcc(Assembler::greaterEqual, loop);
1909   // default case
1910   __ profile_switch_default(rax);
1911   __ movl(rdx, Address(rbx, 0));
1912   __ jmpb(continue_execution);
1913   // entry found -> get offset
1914   __ bind(found);
1915   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * wordSize));
1916   __ profile_switch_case(rcx, rax, rbx);
1917   // continue execution
1918   __ bind(continue_execution);
1919   __ bswapl(rdx);
1920   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1921   __ addptr(rsi, rdx);
1922   __ dispatch_only(vtos);
1923 }
1924 
1925 
1926 void TemplateTable::fast_binaryswitch() {
1927   transition(itos, vtos);
1928   // Implementation using the following core algorithm:
1929   //
1930   // int binary_search(int key, LookupswitchPair* array, int n) {
1931   //   // Binary search according to "Methodik des Programmierens" by
1932   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1933   //   int i = 0;
1934   //   int j = n;
1935   //   while (i+1 < j) {
1936   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1937   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1938   //     // where a stands for the array and assuming that the (inexisting)
1939   //     // element a[n] is infinitely big.
1940   //     int h = (i + j) >> 1;
1941   //     // i < h < j
1942   //     if (key < array[h].fast_match()) {
1943   //       j = h;
1944   //     } else {
1945   //       i = h;
1946   //     }
1947   //   }
1948   //   // R: a[i] <= key < a[i+1] or Q
1949   //   // (i.e., if key is within array, i is the correct index)
1950   //   return i;
1951   // }
1952 
1953   // register allocation
1954   const Register key   = rax;                    // already set (tosca)
1955   const Register array = rbx;
1956   const Register i     = rcx;
1957   const Register j     = rdx;
1958   const Register h     = rdi;                    // needs to be restored
1959   const Register temp  = rsi;
1960   // setup array
1961   __ save_bcp();
1962 
1963   __ lea(array, at_bcp(3*wordSize));             // btw: should be able to get rid of this instruction (change offsets below)
1964   __ andptr(array, -wordSize);
1965   // initialize i & j
1966   __ xorl(i, i);                                 // i = 0;
1967   __ movl(j, Address(array, -wordSize));         // j = length(array);
1968   // Convert j into native byteordering
1969   __ bswapl(j);
1970   // and start
1971   Label entry;
1972   __ jmp(entry);
1973 
1974   // binary search loop
1975   { Label loop;
1976     __ bind(loop);
1977     // int h = (i + j) >> 1;
1978     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
1979     __ sarl(h, 1);                               // h = (i + j) >> 1;
1980     // if (key < array[h].fast_match()) {
1981     //   j = h;
1982     // } else {
1983     //   i = h;
1984     // }
1985     // Convert array[h].match to native byte-ordering before compare
1986     __ movl(temp, Address(array, h, Address::times_8, 0*wordSize));
1987     __ bswapl(temp);
1988     __ cmpl(key, temp);
1989     // j = h if (key <  array[h].fast_match())
1990     __ cmov32(Assembler::less        , j, h);
1991     // i = h if (key >= array[h].fast_match())
1992     __ cmov32(Assembler::greaterEqual, i, h);
1993     // while (i+1 < j)
1994     __ bind(entry);
1995     __ leal(h, Address(i, 1));                   // i+1
1996     __ cmpl(h, j);                               // i+1 < j
1997     __ jcc(Assembler::less, loop);
1998   }
1999 
2000   // end of binary search, result index is i (must check again!)
2001   Label default_case;
2002   // Convert array[i].match to native byte-ordering before compare
2003   __ movl(temp, Address(array, i, Address::times_8, 0*wordSize));
2004   __ bswapl(temp);
2005   __ cmpl(key, temp);
2006   __ jcc(Assembler::notEqual, default_case);
2007 
2008   // entry found -> j = offset
2009   __ movl(j , Address(array, i, Address::times_8, 1*wordSize));
2010   __ profile_switch_case(i, key, array);
2011   __ bswapl(j);
2012   LP64_ONLY(__ movslq(j, j));
2013   __ restore_bcp();
2014   __ restore_locals();                           // restore rdi
2015   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2016 
2017   __ addptr(rsi, j);
2018   __ dispatch_only(vtos);
2019 
2020   // default case -> j = default offset
2021   __ bind(default_case);
2022   __ profile_switch_default(i);
2023   __ movl(j, Address(array, -2*wordSize));
2024   __ bswapl(j);
2025   LP64_ONLY(__ movslq(j, j));
2026   __ restore_bcp();
2027   __ restore_locals();                           // restore rdi
2028   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2029   __ addptr(rsi, j);
2030   __ dispatch_only(vtos);
2031 }
2032 
2033 
2034 void TemplateTable::_return(TosState state) {
2035   transition(state, state);
2036   assert(_desc->calls_vm(), "inconsistent calls_vm information"); // call in remove_activation
2037 
2038   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2039     assert(state == vtos, "only valid state");
2040     __ movptr(rax, aaddress(0));
2041     __ load_klass(rdi, rax);
2042     __ movl(rdi, Address(rdi, Klass::access_flags_offset()));
2043     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2044     Label skip_register_finalizer;
2045     __ jcc(Assembler::zero, skip_register_finalizer);
2046 
2047     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), rax);
2048 
2049     __ bind(skip_register_finalizer);
2050   }
2051 
2052   // Narrow result if state is itos but result type is smaller.
2053   // Need to narrow in the return bytecode rather than in generate_return_entry
2054   // since compiled code callers expect the result to already be narrowed.
2055   if (state == itos) {
2056     __ narrow(rax);
2057   }
2058   __ remove_activation(state, rsi);
2059 
2060   __ jmp(rsi);
2061 }
2062 
2063 
2064 // ----------------------------------------------------------------------------
2065 // Volatile variables demand their effects be made known to all CPU's in
2066 // order.  Store buffers on most chips allow reads & writes to reorder; the
2067 // JMM's ReadAfterWrite.java test fails in -Xint mode without some kind of
2068 // memory barrier (i.e., it's not sufficient that the interpreter does not
2069 // reorder volatile references, the hardware also must not reorder them).
2070 //
2071 // According to the new Java Memory Model (JMM):
2072 // (1) All volatiles are serialized wrt to each other.
2073 // ALSO reads & writes act as aquire & release, so:
2074 // (2) A read cannot let unrelated NON-volatile memory refs that happen after
2075 // the read float up to before the read.  It's OK for non-volatile memory refs
2076 // that happen before the volatile read to float down below it.
2077 // (3) Similar a volatile write cannot let unrelated NON-volatile memory refs
2078 // that happen BEFORE the write float down to after the write.  It's OK for
2079 // non-volatile memory refs that happen after the volatile write to float up
2080 // before it.
2081 //
2082 // We only put in barriers around volatile refs (they are expensive), not
2083 // _between_ memory refs (that would require us to track the flavor of the
2084 // previous memory refs).  Requirements (2) and (3) require some barriers
2085 // before volatile stores and after volatile loads.  These nearly cover
2086 // requirement (1) but miss the volatile-store-volatile-load case.  This final
2087 // case is placed after volatile-stores although it could just as well go
2088 // before volatile-loads.
2089 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits order_constraint ) {
2090   // Helper function to insert a is-volatile test and memory barrier
2091   if( !os::is_MP() ) return;    // Not needed on single CPU
2092   __ membar(order_constraint);
2093 }
2094 
2095 void TemplateTable::resolve_cache_and_index(int byte_no,
2096                                             Register Rcache,
2097                                             Register index,
2098                                             size_t index_size) {
2099   const Register temp = rbx;
2100   assert_different_registers(Rcache, index, temp);
2101 
2102   Label resolved;
2103     assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
2104     __ get_cache_and_index_and_bytecode_at_bcp(Rcache, index, temp, byte_no, 1, index_size);
2105     __ cmpl(temp, (int) bytecode());  // have we resolved this bytecode?
2106     __ jcc(Assembler::equal, resolved);
2107 
2108   // resolve first time through
2109   address entry;
2110   switch (bytecode()) {
2111     case Bytecodes::_getstatic      : // fall through
2112     case Bytecodes::_putstatic      : // fall through
2113     case Bytecodes::_getfield       : // fall through
2114     case Bytecodes::_putfield       : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);        break;
2115     case Bytecodes::_invokevirtual  : // fall through
2116     case Bytecodes::_invokespecial  : // fall through
2117     case Bytecodes::_invokestatic   : // fall through
2118     case Bytecodes::_invokeinterface: entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);         break;
2119     case Bytecodes::_invokehandle   : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokehandle);   break;
2120     case Bytecodes::_invokedynamic  : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);  break;
2121     default:
2122       fatal(err_msg("unexpected bytecode: %s", Bytecodes::name(bytecode())));
2123       break;
2124   }
2125   __ movl(temp, (int)bytecode());
2126   __ call_VM(noreg, entry, temp);
2127   // Update registers with resolved info
2128   __ get_cache_and_index_at_bcp(Rcache, index, 1, index_size);
2129   __ bind(resolved);
2130 }
2131 
2132 
2133 // The cache and index registers must be set before call
2134 void TemplateTable::load_field_cp_cache_entry(Register obj,
2135                                               Register cache,
2136                                               Register index,
2137                                               Register off,
2138                                               Register flags,
2139                                               bool is_static = false) {
2140   assert_different_registers(cache, index, flags, off);
2141 
2142   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2143   // Field offset
2144   __ movptr(off, Address(cache, index, Address::times_ptr,
2145                          in_bytes(cp_base_offset + ConstantPoolCacheEntry::f2_offset())));
2146   // Flags
2147   __ movl(flags, Address(cache, index, Address::times_ptr,
2148            in_bytes(cp_base_offset + ConstantPoolCacheEntry::flags_offset())));
2149 
2150   // klass overwrite register
2151   if (is_static) {
2152     __ movptr(obj, Address(cache, index, Address::times_ptr,
2153                            in_bytes(cp_base_offset + ConstantPoolCacheEntry::f1_offset())));
2154     const int mirror_offset = in_bytes(Klass::java_mirror_offset());
2155     __ movptr(obj, Address(obj, mirror_offset));
2156   }
2157 }
2158 
2159 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2160                                                Register method,
2161                                                Register itable_index,
2162                                                Register flags,
2163                                                bool is_invokevirtual,
2164                                                bool is_invokevfinal, /*unused*/
2165                                                bool is_invokedynamic) {
2166   // setup registers
2167   const Register cache = rcx;
2168   const Register index = rdx;
2169   assert_different_registers(method, flags);
2170   assert_different_registers(method, cache, index);
2171   assert_different_registers(itable_index, flags);
2172   assert_different_registers(itable_index, cache, index);
2173   // determine constant pool cache field offsets
2174   assert(is_invokevirtual == (byte_no == f2_byte), "is_invokevirtual flag redundant");
2175   const int method_offset = in_bytes(
2176     ConstantPoolCache::base_offset() +
2177       ((byte_no == f2_byte)
2178        ? ConstantPoolCacheEntry::f2_offset()
2179        : ConstantPoolCacheEntry::f1_offset()));
2180   const int flags_offset = in_bytes(ConstantPoolCache::base_offset() +
2181                                     ConstantPoolCacheEntry::flags_offset());
2182   // access constant pool cache fields
2183   const int index_offset = in_bytes(ConstantPoolCache::base_offset() +
2184                                     ConstantPoolCacheEntry::f2_offset());
2185 
2186   size_t index_size = (is_invokedynamic ? sizeof(u4) : sizeof(u2));
2187   resolve_cache_and_index(byte_no, cache, index, index_size);
2188     __ movptr(method, Address(cache, index, Address::times_ptr, method_offset));
2189 
2190   if (itable_index != noreg) {
2191     __ movptr(itable_index, Address(cache, index, Address::times_ptr, index_offset));
2192   }
2193   __ movl(flags, Address(cache, index, Address::times_ptr, flags_offset));
2194 }
2195 
2196 
2197 // The registers cache and index expected to be set before call.
2198 // Correct values of the cache and index registers are preserved.
2199 void TemplateTable::jvmti_post_field_access(Register cache,
2200                                             Register index,
2201                                             bool is_static,
2202                                             bool has_tos) {
2203   if (JvmtiExport::can_post_field_access()) {
2204     // Check to see if a field access watch has been set before we take
2205     // the time to call into the VM.
2206     Label L1;
2207     assert_different_registers(cache, index, rax);
2208     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2209     __ testl(rax,rax);
2210     __ jcc(Assembler::zero, L1);
2211 
2212     // cache entry pointer
2213     __ addptr(cache, in_bytes(ConstantPoolCache::base_offset()));
2214     __ shll(index, LogBytesPerWord);
2215     __ addptr(cache, index);
2216     if (is_static) {
2217       __ xorptr(rax, rax);      // NULL object reference
2218     } else {
2219       __ pop(atos);         // Get the object
2220       __ verify_oop(rax);
2221       __ push(atos);        // Restore stack state
2222     }
2223     // rax,:   object pointer or NULL
2224     // cache: cache entry pointer
2225     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access),
2226                rax, cache);
2227     __ get_cache_and_index_at_bcp(cache, index, 1);
2228     __ bind(L1);
2229   }
2230 }
2231 
2232 void TemplateTable::pop_and_check_object(Register r) {
2233   __ pop_ptr(r);
2234   __ null_check(r);  // for field access must check obj.
2235   __ verify_oop(r);
2236 }
2237 
2238 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2239   transition(vtos, vtos);
2240 
2241   const Register cache = rcx;
2242   const Register index = rdx;
2243   const Register obj   = rcx;
2244   const Register off   = rbx;
2245   const Register flags = rax;
2246 
2247   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2248   jvmti_post_field_access(cache, index, is_static, false);
2249   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2250 
2251   if (!is_static) pop_and_check_object(obj);
2252 
2253   const Address lo(obj, off, Address::times_1, 0*wordSize);
2254   const Address hi(obj, off, Address::times_1, 1*wordSize);
2255 
2256   Label Done, notByte, notBool, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2257 
2258   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2259   assert(btos == 0, "change code, btos != 0");
2260   // btos
2261   __ andptr(flags, ConstantPoolCacheEntry::tos_state_mask);
2262   __ jcc(Assembler::notZero, notByte);
2263 
2264   __ load_signed_byte(rax, lo );
2265   __ push(btos);
2266   // Rewrite bytecode to be faster
2267   if (!is_static) {
2268     patch_bytecode(Bytecodes::_fast_bgetfield, rcx, rbx);
2269   }
2270   __ jmp(Done);
2271 
2272   __ bind(notByte);
2273 
2274   __ cmpl(flags, ztos);
2275   __ jcc(Assembler::notEqual, notBool);
2276 
2277   // ztos (same code as btos)
2278   __ load_signed_byte(rax, lo);
2279   __ push(ztos);
2280   // Rewrite bytecode to be faster
2281   if (!is_static) {
2282     // use btos rewriting, no truncating to t/f bit is needed for getfield.
2283     patch_bytecode(Bytecodes::_fast_bgetfield, rcx, rbx);
2284   }
2285   __ jmp(Done);
2286 
2287   __ bind(notBool);
2288 
2289   // itos
2290   __ cmpl(flags, itos );
2291   __ jcc(Assembler::notEqual, notInt);
2292 
2293   __ movl(rax, lo );
2294   __ push(itos);
2295   // Rewrite bytecode to be faster
2296   if (!is_static) {
2297     patch_bytecode(Bytecodes::_fast_igetfield, rcx, rbx);
2298   }
2299   __ jmp(Done);
2300 
2301   __ bind(notInt);
2302   // atos
2303   __ cmpl(flags, atos );
2304   __ jcc(Assembler::notEqual, notObj);
2305 
2306   __ movl(rax, lo );
2307   __ push(atos);
2308   if (!is_static) {
2309     patch_bytecode(Bytecodes::_fast_agetfield, rcx, rbx);
2310   }
2311   __ jmp(Done);
2312 
2313   __ bind(notObj);
2314   // ctos
2315   __ cmpl(flags, ctos );
2316   __ jcc(Assembler::notEqual, notChar);
2317 
2318   __ load_unsigned_short(rax, lo );
2319   __ push(ctos);
2320   if (!is_static) {
2321     patch_bytecode(Bytecodes::_fast_cgetfield, rcx, rbx);
2322   }
2323   __ jmp(Done);
2324 
2325   __ bind(notChar);
2326   // stos
2327   __ cmpl(flags, stos );
2328   __ jcc(Assembler::notEqual, notShort);
2329 
2330   __ load_signed_short(rax, lo );
2331   __ push(stos);
2332   if (!is_static) {
2333     patch_bytecode(Bytecodes::_fast_sgetfield, rcx, rbx);
2334   }
2335   __ jmp(Done);
2336 
2337   __ bind(notShort);
2338   // ltos
2339   __ cmpl(flags, ltos );
2340   __ jcc(Assembler::notEqual, notLong);
2341 
2342   // Generate code as if volatile.  There just aren't enough registers to
2343   // save that information and this code is faster than the test.
2344   __ fild_d(lo);                // Must load atomically
2345   __ subptr(rsp,2*wordSize);    // Make space for store
2346   __ fistp_d(Address(rsp,0));
2347   __ pop(rax);
2348   __ pop(rdx);
2349 
2350   __ push(ltos);
2351   // Don't rewrite to _fast_lgetfield for potential volatile case.
2352   __ jmp(Done);
2353 
2354   __ bind(notLong);
2355   // ftos
2356   __ cmpl(flags, ftos );
2357   __ jcc(Assembler::notEqual, notFloat);
2358 
2359   __ fld_s(lo);
2360   __ push(ftos);
2361   if (!is_static) {
2362     patch_bytecode(Bytecodes::_fast_fgetfield, rcx, rbx);
2363   }
2364   __ jmp(Done);
2365 
2366   __ bind(notFloat);
2367   // dtos
2368   __ cmpl(flags, dtos );
2369   __ jcc(Assembler::notEqual, notDouble);
2370 
2371   __ fld_d(lo);
2372   __ push(dtos);
2373   if (!is_static) {
2374     patch_bytecode(Bytecodes::_fast_dgetfield, rcx, rbx);
2375   }
2376   __ jmpb(Done);
2377 
2378   __ bind(notDouble);
2379 
2380   __ stop("Bad state");
2381 
2382   __ bind(Done);
2383   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2384   // volatile_barrier( );
2385 }
2386 
2387 
2388 void TemplateTable::getfield(int byte_no) {
2389   getfield_or_static(byte_no, false);
2390 }
2391 
2392 
2393 void TemplateTable::getstatic(int byte_no) {
2394   getfield_or_static(byte_no, true);
2395 }
2396 
2397 // The registers cache and index expected to be set before call.
2398 // The function may destroy various registers, just not the cache and index registers.
2399 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2400 
2401   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2402 
2403   if (JvmtiExport::can_post_field_modification()) {
2404     // Check to see if a field modification watch has been set before we take
2405     // the time to call into the VM.
2406     Label L1;
2407     assert_different_registers(cache, index, rax);
2408     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2409     __ testl(rax, rax);
2410     __ jcc(Assembler::zero, L1);
2411 
2412     // The cache and index registers have been already set.
2413     // This allows to eliminate this call but the cache and index
2414     // registers have to be correspondingly used after this line.
2415     __ get_cache_and_index_at_bcp(rax, rdx, 1);
2416 
2417     if (is_static) {
2418       // Life is simple.  Null out the object pointer.
2419       __ xorptr(rbx, rbx);
2420     } else {
2421       // Life is harder. The stack holds the value on top, followed by the object.
2422       // We don't know the size of the value, though; it could be one or two words
2423       // depending on its type. As a result, we must find the type to determine where
2424       // the object is.
2425       Label two_word, valsize_known;
2426       __ movl(rcx, Address(rax, rdx, Address::times_ptr, in_bytes(cp_base_offset +
2427                                    ConstantPoolCacheEntry::flags_offset())));
2428       __ mov(rbx, rsp);
2429       __ shrl(rcx, ConstantPoolCacheEntry::tos_state_shift);
2430       // Make sure we don't need to mask rcx after the above shift
2431       ConstantPoolCacheEntry::verify_tos_state_shift();
2432       __ cmpl(rcx, ltos);
2433       __ jccb(Assembler::equal, two_word);
2434       __ cmpl(rcx, dtos);
2435       __ jccb(Assembler::equal, two_word);
2436       __ addptr(rbx, Interpreter::expr_offset_in_bytes(1)); // one word jvalue (not ltos, dtos)
2437       __ jmpb(valsize_known);
2438 
2439       __ bind(two_word);
2440       __ addptr(rbx, Interpreter::expr_offset_in_bytes(2)); // two words jvalue
2441 
2442       __ bind(valsize_known);
2443       // setup object pointer
2444       __ movptr(rbx, Address(rbx, 0));
2445     }
2446     // cache entry pointer
2447     __ addptr(rax, in_bytes(cp_base_offset));
2448     __ shll(rdx, LogBytesPerWord);
2449     __ addptr(rax, rdx);
2450     // object (tos)
2451     __ mov(rcx, rsp);
2452     // rbx,: object pointer set up above (NULL if static)
2453     // rax,: cache entry pointer
2454     // rcx: jvalue object on the stack
2455     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification),
2456                rbx, rax, rcx);
2457     __ get_cache_and_index_at_bcp(cache, index, 1);
2458     __ bind(L1);
2459   }
2460 }
2461 
2462 
2463 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2464   transition(vtos, vtos);
2465 
2466   const Register cache = rcx;
2467   const Register index = rdx;
2468   const Register obj   = rcx;
2469   const Register off   = rbx;
2470   const Register flags = rax;
2471 
2472   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2473   jvmti_post_field_mod(cache, index, is_static);
2474   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2475 
2476   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2477   // volatile_barrier( );
2478 
2479   Label notVolatile, Done;
2480   __ movl(rdx, flags);
2481   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2482   __ andl(rdx, 0x1);
2483 
2484   // field addresses
2485   const Address lo(obj, off, Address::times_1, 0*wordSize);
2486   const Address hi(obj, off, Address::times_1, 1*wordSize);
2487 
2488   Label notByte, notBool, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2489 
2490   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2491   assert(btos == 0, "change code, btos != 0");
2492   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2493   __ jcc(Assembler::notZero, notByte);
2494 
2495   // btos
2496   {
2497     __ pop(btos);
2498     if (!is_static) pop_and_check_object(obj);
2499     __ movb(lo, rax);
2500     if (!is_static) {
2501       patch_bytecode(Bytecodes::_fast_bputfield, rcx, rbx, true, byte_no);
2502     }
2503     __ jmp(Done);
2504   }
2505 
2506   __ bind(notByte);
2507   __ cmpl(flags, ztos);
2508   __ jcc(Assembler::notEqual, notBool);
2509 
2510   // ztos
2511   {
2512     __ pop(ztos);
2513     if (!is_static) pop_and_check_object(obj);
2514     __ andl(rax, 0x1);
2515     __ movb(lo, rax);
2516     if (!is_static) {
2517       patch_bytecode(Bytecodes::_fast_zputfield, rcx, rbx, true, byte_no);
2518     }
2519     __ jmp(Done);
2520   }
2521 
2522   __ bind(notBool);
2523   __ cmpl(flags, itos);
2524   __ jcc(Assembler::notEqual, notInt);
2525 
2526   // itos
2527   {
2528     __ pop(itos);
2529     if (!is_static) pop_and_check_object(obj);
2530     __ movl(lo, rax);
2531     if (!is_static) {
2532       patch_bytecode(Bytecodes::_fast_iputfield, rcx, rbx, true, byte_no);
2533     }
2534     __ jmp(Done);
2535   }
2536 
2537   __ bind(notInt);
2538   __ cmpl(flags, atos);
2539   __ jcc(Assembler::notEqual, notObj);
2540 
2541   // atos
2542   {
2543     __ pop(atos);
2544     if (!is_static) pop_and_check_object(obj);
2545     do_oop_store(_masm, lo, rax, _bs->kind(), false);
2546     if (!is_static) {
2547       patch_bytecode(Bytecodes::_fast_aputfield, rcx, rbx, true, byte_no);
2548     }
2549     __ jmp(Done);
2550   }
2551 
2552   __ bind(notObj);
2553   __ cmpl(flags, ctos);
2554   __ jcc(Assembler::notEqual, notChar);
2555 
2556   // ctos
2557   {
2558     __ pop(ctos);
2559     if (!is_static) pop_and_check_object(obj);
2560     __ movw(lo, rax);
2561     if (!is_static) {
2562       patch_bytecode(Bytecodes::_fast_cputfield, rcx, rbx, true, byte_no);
2563     }
2564     __ jmp(Done);
2565   }
2566 
2567   __ bind(notChar);
2568   __ cmpl(flags, stos);
2569   __ jcc(Assembler::notEqual, notShort);
2570 
2571   // stos
2572   {
2573     __ pop(stos);
2574     if (!is_static) pop_and_check_object(obj);
2575     __ movw(lo, rax);
2576     if (!is_static) {
2577       patch_bytecode(Bytecodes::_fast_sputfield, rcx, rbx, true, byte_no);
2578     }
2579     __ jmp(Done);
2580   }
2581 
2582   __ bind(notShort);
2583   __ cmpl(flags, ltos);
2584   __ jcc(Assembler::notEqual, notLong);
2585 
2586   // ltos
2587   {
2588     Label notVolatileLong;
2589     __ testl(rdx, rdx);
2590     __ jcc(Assembler::zero, notVolatileLong);
2591 
2592     __ pop(ltos);  // overwrites rdx, do this after testing volatile.
2593     if (!is_static) pop_and_check_object(obj);
2594 
2595     // Replace with real volatile test
2596     __ push(rdx);
2597     __ push(rax);                 // Must update atomically with FIST
2598     __ fild_d(Address(rsp,0));    // So load into FPU register
2599     __ fistp_d(lo);               // and put into memory atomically
2600     __ addptr(rsp, 2*wordSize);
2601     // volatile_barrier();
2602     volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2603                                                  Assembler::StoreStore));
2604     // Don't rewrite volatile version
2605     __ jmp(notVolatile);
2606 
2607     __ bind(notVolatileLong);
2608 
2609     __ pop(ltos);  // overwrites rdx
2610     if (!is_static) pop_and_check_object(obj);
2611     NOT_LP64(__ movptr(hi, rdx));
2612     __ movptr(lo, rax);
2613     if (!is_static) {
2614       patch_bytecode(Bytecodes::_fast_lputfield, rcx, rbx, true, byte_no);
2615     }
2616     __ jmp(notVolatile);
2617   }
2618 
2619   __ bind(notLong);
2620   __ cmpl(flags, ftos);
2621   __ jcc(Assembler::notEqual, notFloat);
2622 
2623   // ftos
2624   {
2625     __ pop(ftos);
2626     if (!is_static) pop_and_check_object(obj);
2627     __ fstp_s(lo);
2628     if (!is_static) {
2629       patch_bytecode(Bytecodes::_fast_fputfield, rcx, rbx, true, byte_no);
2630     }
2631     __ jmp(Done);
2632   }
2633 
2634   __ bind(notFloat);
2635 #ifdef ASSERT
2636   __ cmpl(flags, dtos);
2637   __ jcc(Assembler::notEqual, notDouble);
2638 #endif
2639 
2640   // dtos
2641   {
2642     __ pop(dtos);
2643     if (!is_static) pop_and_check_object(obj);
2644     __ fstp_d(lo);
2645     if (!is_static) {
2646       patch_bytecode(Bytecodes::_fast_dputfield, rcx, rbx, true, byte_no);
2647     }
2648     __ jmp(Done);
2649   }
2650 
2651 #ifdef ASSERT
2652   __ bind(notDouble);
2653   __ stop("Bad state");
2654 #endif
2655 
2656   __ bind(Done);
2657 
2658   // Check for volatile store
2659   __ testl(rdx, rdx);
2660   __ jcc(Assembler::zero, notVolatile);
2661   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2662                                                Assembler::StoreStore));
2663   __ bind(notVolatile);
2664 }
2665 
2666 
2667 void TemplateTable::putfield(int byte_no) {
2668   putfield_or_static(byte_no, false);
2669 }
2670 
2671 
2672 void TemplateTable::putstatic(int byte_no) {
2673   putfield_or_static(byte_no, true);
2674 }
2675 
2676 void TemplateTable::jvmti_post_fast_field_mod() {
2677   if (JvmtiExport::can_post_field_modification()) {
2678     // Check to see if a field modification watch has been set before we take
2679     // the time to call into the VM.
2680     Label L2;
2681      __ mov32(rcx, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2682      __ testl(rcx,rcx);
2683      __ jcc(Assembler::zero, L2);
2684      __ pop_ptr(rbx);               // copy the object pointer from tos
2685      __ verify_oop(rbx);
2686      __ push_ptr(rbx);              // put the object pointer back on tos
2687 
2688      // Save tos values before call_VM() clobbers them. Since we have
2689      // to do it for every data type, we use the saved values as the
2690      // jvalue object.
2691      switch (bytecode()) {          // load values into the jvalue object
2692      case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
2693      case Bytecodes::_fast_bputfield: // fall through
2694      case Bytecodes::_fast_zputfield: // fall through
2695      case Bytecodes::_fast_sputfield: // fall through
2696      case Bytecodes::_fast_cputfield: // fall through
2697      case Bytecodes::_fast_iputfield: __ push_i(rax); break;
2698      case Bytecodes::_fast_dputfield: __ push_d(); break;
2699      case Bytecodes::_fast_fputfield: __ push_f(); break;
2700      case Bytecodes::_fast_lputfield: __ push_l(rax); break;
2701 
2702      default:
2703        ShouldNotReachHere();
2704      }
2705      __ mov(rcx, rsp);              // points to jvalue on the stack
2706      // access constant pool cache entry
2707      __ get_cache_entry_pointer_at_bcp(rax, rdx, 1);
2708      __ verify_oop(rbx);
2709      // rbx,: object pointer copied above
2710      // rax,: cache entry pointer
2711      // rcx: jvalue object on the stack
2712      __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification), rbx, rax, rcx);
2713 
2714      switch (bytecode()) {             // restore tos values
2715      case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
2716      case Bytecodes::_fast_bputfield: // fall through
2717      case Bytecodes::_fast_zputfield: // fall through
2718      case Bytecodes::_fast_sputfield: // fall through
2719      case Bytecodes::_fast_cputfield: // fall through
2720      case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
2721      case Bytecodes::_fast_dputfield: __ pop_d(); break;
2722      case Bytecodes::_fast_fputfield: __ pop_f(); break;
2723      case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
2724      }
2725      __ bind(L2);
2726   }
2727 }
2728 
2729 void TemplateTable::fast_storefield(TosState state) {
2730   transition(state, vtos);
2731 
2732   ByteSize base = ConstantPoolCache::base_offset();
2733 
2734   jvmti_post_fast_field_mod();
2735 
2736   // access constant pool cache
2737   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2738 
2739   // test for volatile with rdx but rdx is tos register for lputfield.
2740   if (bytecode() == Bytecodes::_fast_lputfield) __ push(rdx);
2741   __ movl(rdx, Address(rcx, rbx, Address::times_ptr, in_bytes(base +
2742                        ConstantPoolCacheEntry::flags_offset())));
2743 
2744   // replace index with field offset from cache entry
2745   __ movptr(rbx, Address(rcx, rbx, Address::times_ptr, in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2746 
2747   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2748   // volatile_barrier( );
2749 
2750   Label notVolatile, Done;
2751   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2752   __ andl(rdx, 0x1);
2753   // Check for volatile store
2754   __ testl(rdx, rdx);
2755   __ jcc(Assembler::zero, notVolatile);
2756 
2757   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2758 
2759   // Get object from stack
2760   pop_and_check_object(rcx);
2761 
2762   // field addresses
2763   const Address lo(rcx, rbx, Address::times_1, 0*wordSize);
2764   const Address hi(rcx, rbx, Address::times_1, 1*wordSize);
2765 
2766   // access field
2767   switch (bytecode()) {
2768     case Bytecodes::_fast_zputfield: __ andl(rax, 0x1);  // boolean is true if LSB is 1
2769     // fall through to bputfield
2770     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2771     case Bytecodes::_fast_sputfield: // fall through
2772     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2773     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2774     case Bytecodes::_fast_lputfield:
2775       NOT_LP64(__ movptr(hi, rdx));
2776       __ movptr(lo, rax);
2777       break;
2778     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2779     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2780     case Bytecodes::_fast_aputfield: {
2781       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2782       break;
2783     }
2784     default:
2785       ShouldNotReachHere();
2786   }
2787 
2788   Label done;
2789   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2790                                                Assembler::StoreStore));
2791   // Barriers are so large that short branch doesn't reach!
2792   __ jmp(done);
2793 
2794   // Same code as above, but don't need rdx to test for volatile.
2795   __ bind(notVolatile);
2796 
2797   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2798 
2799   // Get object from stack
2800   pop_and_check_object(rcx);
2801 
2802   // access field
2803   switch (bytecode()) {
2804     case Bytecodes::_fast_zputfield: __ andl(rax, 0x1);  // boolean is true if LSB is 1
2805     // fall through to bputfield
2806     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2807     case Bytecodes::_fast_sputfield: // fall through
2808     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2809     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2810     case Bytecodes::_fast_lputfield:
2811       NOT_LP64(__ movptr(hi, rdx));
2812       __ movptr(lo, rax);
2813       break;
2814     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2815     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2816     case Bytecodes::_fast_aputfield: {
2817       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2818       break;
2819     }
2820     default:
2821       ShouldNotReachHere();
2822   }
2823   __ bind(done);
2824 }
2825 
2826 
2827 void TemplateTable::fast_accessfield(TosState state) {
2828   transition(atos, state);
2829 
2830   // do the JVMTI work here to avoid disturbing the register state below
2831   if (JvmtiExport::can_post_field_access()) {
2832     // Check to see if a field access watch has been set before we take
2833     // the time to call into the VM.
2834     Label L1;
2835     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2836     __ testl(rcx,rcx);
2837     __ jcc(Assembler::zero, L1);
2838     // access constant pool cache entry
2839     __ get_cache_entry_pointer_at_bcp(rcx, rdx, 1);
2840     __ push_ptr(rax);  // save object pointer before call_VM() clobbers it
2841     __ verify_oop(rax);
2842     // rax,: object pointer copied above
2843     // rcx: cache entry pointer
2844     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access), rax, rcx);
2845     __ pop_ptr(rax);   // restore object pointer
2846     __ bind(L1);
2847   }
2848 
2849   // access constant pool cache
2850   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2851   // replace index with field offset from cache entry
2852   __ movptr(rbx, Address(rcx,
2853                          rbx,
2854                          Address::times_ptr,
2855                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2856 
2857 
2858   // rax,: object
2859   __ verify_oop(rax);
2860   __ null_check(rax);
2861   // field addresses
2862   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2863   const Address hi = Address(rax, rbx, Address::times_1, 1*wordSize);
2864 
2865   // access field
2866   switch (bytecode()) {
2867     case Bytecodes::_fast_bgetfield: __ movsbl(rax, lo );                 break;
2868     case Bytecodes::_fast_sgetfield: __ load_signed_short(rax, lo );      break;
2869     case Bytecodes::_fast_cgetfield: __ load_unsigned_short(rax, lo );    break;
2870     case Bytecodes::_fast_igetfield: __ movl(rax, lo);                    break;
2871     case Bytecodes::_fast_lgetfield: __ stop("should not be rewritten");  break;
2872     case Bytecodes::_fast_fgetfield: __ fld_s(lo);                        break;
2873     case Bytecodes::_fast_dgetfield: __ fld_d(lo);                        break;
2874     case Bytecodes::_fast_agetfield: __ movptr(rax, lo); __ verify_oop(rax); break;
2875     default:
2876       ShouldNotReachHere();
2877   }
2878 
2879   // Doug Lea believes this is not needed with current Sparcs(TSO) and Intel(PSO)
2880   // volatile_barrier( );
2881 }
2882 
2883 void TemplateTable::fast_xaccess(TosState state) {
2884   transition(vtos, state);
2885   // get receiver
2886   __ movptr(rax, aaddress(0));
2887   // access constant pool cache
2888   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2889   __ movptr(rbx, Address(rcx,
2890                          rdx,
2891                          Address::times_ptr,
2892                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2893   // make sure exception is reported in correct bcp range (getfield is next instruction)
2894   __ increment(rsi);
2895   __ null_check(rax);
2896   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2897   if (state == itos) {
2898     __ movl(rax, lo);
2899   } else if (state == atos) {
2900     __ movptr(rax, lo);
2901     __ verify_oop(rax);
2902   } else if (state == ftos) {
2903     __ fld_s(lo);
2904   } else {
2905     ShouldNotReachHere();
2906   }
2907   __ decrement(rsi);
2908 }
2909 
2910 
2911 
2912 //----------------------------------------------------------------------------------------------------
2913 // Calls
2914 
2915 void TemplateTable::count_calls(Register method, Register temp) {
2916   // implemented elsewhere
2917   ShouldNotReachHere();
2918 }
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     __ jccb(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 (note: no return address pushed yet)
2969   if (load_receiver) {
2970     __ movl(recv, flags);
2971     __ andl(recv, ConstantPoolCacheEntry::parameter_size_mask);
2972     const int no_return_pc_pushed_yet = -1;  // argument slot correction before we push return address
2973     const int receiver_is_at_end      = -1;  // back off one slot to get receiver
2974     Address recv_addr = __ argument_address(recv, no_return_pc_pushed_yet + receiver_is_at_end);
2975     __ movptr(recv, recv_addr);
2976     __ verify_oop(recv);
2977   }
2978 
2979   if (save_flags) {
2980     __ mov(rsi, flags);
2981   }
2982 
2983   // compute return type
2984   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2985   // Make sure we don't need to mask flags after the above shift
2986   ConstantPoolCacheEntry::verify_tos_state_shift();
2987   // load return address
2988   {
2989     const address table_addr = (address) Interpreter::invoke_return_entry_table_for(code);
2990     ExternalAddress table(table_addr);
2991     __ movptr(flags, ArrayAddress(table, Address(noreg, flags, Address::times_ptr)));
2992   }
2993 
2994   // push return address
2995   __ push(flags);
2996 
2997   // Restore flags value from the constant pool cache, and restore rsi
2998   // for later null checks.  rsi is the bytecode pointer
2999   if (save_flags) {
3000     __ mov(flags, rsi);
3001     __ restore_bcp();
3002   }
3003 }
3004 
3005 
3006 void TemplateTable::invokevirtual_helper(Register index,
3007                                          Register recv,
3008                                          Register flags) {
3009   // Uses temporary registers rax, rdx
3010   assert_different_registers(index, recv, rax, rdx);
3011   assert(index == rbx, "");
3012   assert(recv  == rcx, "");
3013 
3014   // Test for an invoke of a final method
3015   Label notFinal;
3016   __ movl(rax, flags);
3017   __ andl(rax, (1 << ConstantPoolCacheEntry::is_vfinal_shift));
3018   __ jcc(Assembler::zero, notFinal);
3019 
3020   const Register method = index;  // method must be rbx
3021   assert(method == rbx,
3022          "Method* must be rbx for interpreter calling convention");
3023 
3024   // do the call - the index is actually the method to call
3025   // that is, f2 is a vtable index if !is_vfinal, else f2 is a Method*
3026 
3027   // It's final, need a null check here!
3028   __ null_check(recv);
3029 
3030   // profile this call
3031   __ profile_final_call(rax);
3032   __ profile_arguments_type(rax, method, rsi, true);
3033 
3034   __ jump_from_interpreted(method, rax);
3035 
3036   __ bind(notFinal);
3037 
3038   // get receiver klass
3039   __ null_check(recv, oopDesc::klass_offset_in_bytes());
3040   __ load_klass(rax, recv);
3041 
3042   // profile this call
3043   __ profile_virtual_call(rax, rdi, rdx);
3044 
3045   // get target Method* & entry point
3046   __ lookup_virtual_method(rax, index, method);
3047   __ profile_arguments_type(rdx, method, rsi, true);
3048   __ jump_from_interpreted(method, rdx);
3049 }
3050 
3051 
3052 void TemplateTable::invokevirtual(int byte_no) {
3053   transition(vtos, vtos);
3054   assert(byte_no == f2_byte, "use this argument");
3055   prepare_invoke(byte_no,
3056                  rbx,    // method or vtable index
3057                  noreg,  // unused itable index
3058                  rcx, rdx); // recv, flags
3059 
3060   // rbx: index
3061   // rcx: receiver
3062   // rdx: flags
3063 
3064   invokevirtual_helper(rbx, rcx, rdx);
3065 }
3066 
3067 
3068 void TemplateTable::invokespecial(int byte_no) {
3069   transition(vtos, vtos);
3070   assert(byte_no == f1_byte, "use this argument");
3071   prepare_invoke(byte_no, rbx, noreg,  // get f1 Method*
3072                  rcx);  // get receiver also for null check
3073   __ verify_oop(rcx);
3074   __ null_check(rcx);
3075   // do the call
3076   __ profile_call(rax);
3077   __ profile_arguments_type(rax, rbx, rsi, false);
3078   __ jump_from_interpreted(rbx, rax);
3079 }
3080 
3081 
3082 void TemplateTable::invokestatic(int byte_no) {
3083   transition(vtos, vtos);
3084   assert(byte_no == f1_byte, "use this argument");
3085   prepare_invoke(byte_no, rbx);  // get f1 Method*
3086   // do the call
3087   __ profile_call(rax);
3088   __ profile_arguments_type(rax, rbx, rsi, false);
3089   __ jump_from_interpreted(rbx, rax);
3090 }
3091 
3092 
3093 void TemplateTable::fast_invokevfinal(int byte_no) {
3094   transition(vtos, vtos);
3095   assert(byte_no == f2_byte, "use this argument");
3096   __ stop("fast_invokevfinal not used on x86");
3097 }
3098 
3099 
3100 void TemplateTable::invokeinterface(int byte_no) {
3101   transition(vtos, vtos);
3102   assert(byte_no == f1_byte, "use this argument");
3103   prepare_invoke(byte_no, rax, rbx,  // get f1 Klass*, f2 Method*
3104                  rcx, rdx); // recv, flags
3105 
3106   // rax: reference klass (from f1)
3107   // rbx: method (from f2)
3108   // rcx: receiver
3109   // rdx: flags
3110 
3111   // Special case of invokeinterface called for virtual method of
3112   // java.lang.Object.  See cpCacheOop.cpp for details.
3113   // This code isn't produced by javac, but could be produced by
3114   // another compliant java compiler.
3115   Label notMethod;
3116   __ movl(rdi, rdx);
3117   __ andl(rdi, (1 << ConstantPoolCacheEntry::is_forced_virtual_shift));
3118   __ jcc(Assembler::zero, notMethod);
3119 
3120   invokevirtual_helper(rbx, rcx, rdx);
3121   __ bind(notMethod);
3122 
3123   // Get receiver klass into rdx - also a null check
3124   __ restore_locals();  // restore rdi
3125   __ null_check(rcx, oopDesc::klass_offset_in_bytes());
3126   __ load_klass(rdx, rcx);
3127 
3128   Label no_such_interface, no_such_method;
3129 
3130   // Receiver subtype check against REFC.
3131   // Superklass in rax. Subklass in rdx. Blows rcx, rdi.
3132   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3133                              rdx, rax, noreg,
3134                              // outputs: scan temp. reg, scan temp. reg
3135                              rsi, rdi,
3136                              no_such_interface,
3137                              /*return_method=*/false);
3138 
3139 
3140   // profile this call
3141   __ restore_bcp(); // rbcp was destroyed by receiver type check
3142   __ profile_virtual_call(rdx, rsi, rdi);
3143 
3144   // Get declaring interface class from method, and itable index
3145   __ movptr(rax, Address(rbx, Method::const_offset()));
3146   __ movptr(rax, Address(rax, ConstMethod::constants_offset()));
3147   __ movptr(rax, Address(rax, ConstantPool::pool_holder_offset_in_bytes()));
3148   __ movl(rbx, Address(rbx, Method::itable_index_offset()));
3149   __ subl(rbx, Method::itable_index_max);
3150   __ negl(rbx);
3151 
3152   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3153                              rdx, rax, rbx,
3154                              // outputs: method, scan temp. reg
3155                              rbx, rsi,
3156                              no_such_interface);
3157 
3158   // rbx: Method* to call
3159   // rcx: receiver
3160   // Check for abstract method error
3161   // Note: This should be done more efficiently via a throw_abstract_method_error
3162   //       interpreter entry point and a conditional jump to it in case of a null
3163   //       method.
3164   __ testptr(rbx, rbx);
3165   __ jcc(Assembler::zero, no_such_method);
3166 
3167   __ profile_arguments_type(rdx, rbx, rsi, true);
3168 
3169   // do the call
3170   // rcx: receiver
3171   // rbx,: Method*
3172   __ jump_from_interpreted(rbx, rdx);
3173   __ should_not_reach_here();
3174 
3175   // exception handling code follows...
3176   // note: must restore interpreter registers to canonical
3177   //       state for exception handling to work correctly!
3178 
3179   __ bind(no_such_method);
3180   // throw exception
3181   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3182   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3183   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3184   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3185   // the call_VM checks for exception, so we should never return here.
3186   __ should_not_reach_here();
3187 
3188   __ bind(no_such_interface);
3189   // throw exception
3190   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3191   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3192   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3193   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3194                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3195   // the call_VM checks for exception, so we should never return here.
3196   __ should_not_reach_here();
3197 }
3198 
3199 void TemplateTable::invokehandle(int byte_no) {
3200   transition(vtos, vtos);
3201   assert(byte_no == f1_byte, "use this argument");
3202   const Register rbx_method = rbx;
3203   const Register rax_mtype  = rax;
3204   const Register rcx_recv   = rcx;
3205   const Register rdx_flags  = rdx;
3206 
3207   if (!EnableInvokeDynamic) {
3208     // rewriter does not generate this bytecode
3209     __ should_not_reach_here();
3210     return;
3211   }
3212 
3213   prepare_invoke(byte_no, rbx_method, rax_mtype, rcx_recv);
3214   __ verify_method_ptr(rbx_method);
3215   __ verify_oop(rcx_recv);
3216   __ null_check(rcx_recv);
3217 
3218   // rax: MethodType object (from cpool->resolved_references[f1], if necessary)
3219   // rbx: MH.invokeExact_MT method (from f2)
3220 
3221   // Note:  rax_mtype is already pushed (if necessary) by prepare_invoke
3222 
3223   // FIXME: profile the LambdaForm also
3224   __ profile_final_call(rax);
3225   __ profile_arguments_type(rdx, rbx_method, rsi, true);
3226 
3227   __ jump_from_interpreted(rbx_method, rdx);
3228 }
3229 
3230 
3231 void TemplateTable::invokedynamic(int byte_no) {
3232   transition(vtos, vtos);
3233   assert(byte_no == f1_byte, "use this argument");
3234 
3235   if (!EnableInvokeDynamic) {
3236     // We should not encounter this bytecode if !EnableInvokeDynamic.
3237     // The verifier will stop it.  However, if we get past the verifier,
3238     // this will stop the thread in a reasonable way, without crashing the JVM.
3239     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3240                      InterpreterRuntime::throw_IncompatibleClassChangeError));
3241     // the call_VM checks for exception, so we should never return here.
3242     __ should_not_reach_here();
3243     return;
3244   }
3245 
3246   const Register rbx_method   = rbx;
3247   const Register rax_callsite = rax;
3248 
3249   prepare_invoke(byte_no, rbx_method, rax_callsite);
3250 
3251   // rax: CallSite object (from cpool->resolved_references[f1])
3252   // rbx: MH.linkToCallSite method (from f2)
3253 
3254   // Note:  rax_callsite is already pushed by prepare_invoke
3255 
3256   // %%% should make a type profile for any invokedynamic that takes a ref argument
3257   // profile this call
3258   __ profile_call(rsi);
3259   __ profile_arguments_type(rdx, rbx, rsi, false);
3260 
3261   __ verify_oop(rax_callsite);
3262 
3263   __ jump_from_interpreted(rbx_method, rdx);
3264 }
3265 
3266 //----------------------------------------------------------------------------------------------------
3267 // Allocation
3268 
3269 void TemplateTable::_new() {
3270   transition(vtos, atos);
3271   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3272   Label slow_case;
3273   Label slow_case_no_pop;
3274   Label done;
3275   Label initialize_header;
3276   Label initialize_object;  // including clearing the fields
3277   Label allocate_shared;
3278 
3279   __ get_cpool_and_tags(rcx, rax);
3280 
3281   // Make sure the class we're about to instantiate has been resolved.
3282   // This is done before loading InstanceKlass to be consistent with the order
3283   // how Constant Pool is updated (see ConstantPool::klass_at_put)
3284   const int tags_offset = Array<u1>::base_offset_in_bytes();
3285   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset), JVM_CONSTANT_Class);
3286   __ jcc(Assembler::notEqual, slow_case_no_pop);
3287 
3288   // get InstanceKlass
3289   __ movptr(rcx, Address(rcx, rdx, Address::times_ptr, sizeof(ConstantPool)));
3290   __ push(rcx);  // save the contexts of klass for initializing the header
3291 
3292   // make sure klass is initialized & doesn't have finalizer
3293   // make sure klass is fully initialized
3294   __ cmpb(Address(rcx, InstanceKlass::init_state_offset()), InstanceKlass::fully_initialized);
3295   __ jcc(Assembler::notEqual, slow_case);
3296 
3297   // get instance_size in InstanceKlass (scaled to a count of bytes)
3298   __ movl(rdx, Address(rcx, Klass::layout_helper_offset()));
3299   // test to see if it has a finalizer or is malformed in some way
3300   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3301   __ jcc(Assembler::notZero, slow_case);
3302 
3303   //
3304   // Allocate the instance
3305   // 1) Try to allocate in the TLAB
3306   // 2) if fail and the object is large allocate in the shared Eden
3307   // 3) if the above fails (or is not applicable), go to a slow case
3308   // (creates a new TLAB, etc.)
3309 
3310   const bool allow_shared_alloc =
3311     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3312 
3313   const Register thread = rcx;
3314   if (UseTLAB || allow_shared_alloc) {
3315     __ get_thread(thread);
3316   }
3317 
3318   if (UseTLAB) {
3319     __ movptr(rax, Address(thread, in_bytes(JavaThread::tlab_top_offset())));
3320     __ lea(rbx, Address(rax, rdx, Address::times_1));
3321     __ cmpptr(rbx, Address(thread, in_bytes(JavaThread::tlab_end_offset())));
3322     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3323     __ movptr(Address(thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3324     if (ZeroTLAB) {
3325       // the fields have been already cleared
3326       __ jmp(initialize_header);
3327     } else {
3328       // initialize both the header and fields
3329       __ jmp(initialize_object);
3330     }
3331   }
3332 
3333   // Allocation in the shared Eden, if allowed.
3334   //
3335   // rdx: instance size in bytes
3336   if (allow_shared_alloc) {
3337     __ bind(allocate_shared);
3338 
3339     ExternalAddress heap_top((address)Universe::heap()->top_addr());
3340 
3341     Label retry;
3342     __ bind(retry);
3343     __ movptr(rax, heap_top);
3344     __ lea(rbx, Address(rax, rdx, Address::times_1));
3345     __ cmpptr(rbx, ExternalAddress((address)Universe::heap()->end_addr()));
3346     __ jcc(Assembler::above, slow_case);
3347 
3348     // Compare rax, with the top addr, and if still equal, store the new
3349     // top addr in rbx, at the address of the top addr pointer. Sets ZF if was
3350     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3351     //
3352     // rax,: object begin
3353     // rbx,: object end
3354     // rdx: instance size in bytes
3355     __ locked_cmpxchgptr(rbx, heap_top);
3356 
3357     // if someone beat us on the allocation, try again, otherwise continue
3358     __ jcc(Assembler::notEqual, retry);
3359 
3360     __ incr_allocated_bytes(thread, rdx, 0);
3361   }
3362 
3363   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3364     // The object is initialized before the header.  If the object size is
3365     // zero, go directly to the header initialization.
3366     __ bind(initialize_object);
3367     __ decrement(rdx, sizeof(oopDesc));
3368     __ jcc(Assembler::zero, initialize_header);
3369 
3370     // Initialize topmost object field, divide rdx by 8, check if odd and
3371     // test if zero.
3372     __ xorl(rcx, rcx);    // use zero reg to clear memory (shorter code)
3373     __ shrl(rdx, LogBytesPerLong); // divide by 2*oopSize and set carry flag if odd
3374 
3375     // rdx must have been multiple of 8
3376 #ifdef ASSERT
3377     // make sure rdx was multiple of 8
3378     Label L;
3379     // Ignore partial flag stall after shrl() since it is debug VM
3380     __ jccb(Assembler::carryClear, L);
3381     __ stop("object size is not multiple of 2 - adjust this code");
3382     __ bind(L);
3383     // rdx must be > 0, no extra check needed here
3384 #endif
3385 
3386     // initialize remaining object fields: rdx was a multiple of 8
3387     { Label loop;
3388     __ bind(loop);
3389     __ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 1*oopSize), rcx);
3390     NOT_LP64(__ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 2*oopSize), rcx));
3391     __ decrement(rdx);
3392     __ jcc(Assembler::notZero, loop);
3393     }
3394 
3395     // initialize object header only.
3396     __ bind(initialize_header);
3397     if (UseBiasedLocking) {
3398       __ pop(rcx);   // get saved klass back in the register.
3399       __ movptr(rbx, Address(rcx, Klass::prototype_header_offset()));
3400       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()), rbx);
3401     } else {
3402       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()),
3403                 (int32_t)markOopDesc::prototype()); // header
3404       __ pop(rcx);   // get saved klass back in the register.
3405     }
3406     __ store_klass(rax, rcx);  // klass
3407 
3408     {
3409       SkipIfEqual skip_if(_masm, &DTraceAllocProbes, 0);
3410       // Trigger dtrace event for fastpath
3411       __ push(atos);
3412       __ call_VM_leaf(
3413            CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3414       __ pop(atos);
3415     }
3416 
3417     __ jmp(done);
3418   }
3419 
3420   // slow case
3421   __ bind(slow_case);
3422   __ pop(rcx);   // restore stack pointer to what it was when we came in.
3423   __ bind(slow_case_no_pop);
3424   __ get_constant_pool(rax);
3425   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3426   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), rax, rdx);
3427 
3428   // continue
3429   __ bind(done);
3430 }
3431 
3432 
3433 void TemplateTable::newarray() {
3434   transition(itos, atos);
3435   __ push_i(rax);                                 // make sure everything is on the stack
3436   __ load_unsigned_byte(rdx, at_bcp(1));
3437   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray), rdx, rax);
3438   __ pop_i(rdx);                                  // discard size
3439 }
3440 
3441 
3442 void TemplateTable::anewarray() {
3443   transition(itos, atos);
3444   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3445   __ get_constant_pool(rcx);
3446   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray), rcx, rdx, rax);
3447 }
3448 
3449 
3450 void TemplateTable::arraylength() {
3451   transition(atos, itos);
3452   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3453   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3454 }
3455 
3456 
3457 void TemplateTable::checkcast() {
3458   transition(atos, atos);
3459   Label done, is_null, ok_is_subtype, quicked, resolved;
3460   __ testptr(rax, rax);   // Object is in EAX
3461   __ jcc(Assembler::zero, is_null);
3462 
3463   // Get cpool & tags index
3464   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3465   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3466   // See if bytecode has already been quicked
3467   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3468   __ jcc(Assembler::equal, quicked);
3469 
3470   __ push(atos);
3471   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3472   // vm_result_2 has metadata result
3473   // borrow rdi from locals
3474   __ get_thread(rdi);
3475   __ get_vm_result_2(rax, rdi);
3476   __ restore_locals();
3477   __ pop_ptr(rdx);
3478   __ jmpb(resolved);
3479 
3480   // Get superklass in EAX and subklass in EBX
3481   __ bind(quicked);
3482   __ mov(rdx, rax);          // Save object in EDX; EAX needed for subtype check
3483   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3484 
3485   __ bind(resolved);
3486   __ load_klass(rbx, rdx);
3487 
3488   // Generate subtype check.  Blows ECX.  Resets EDI.  Object in EDX.
3489   // Superklass in EAX.  Subklass in EBX.
3490   __ gen_subtype_check( rbx, ok_is_subtype );
3491 
3492   // Come here on failure
3493   __ push(rdx);
3494   // object is at TOS
3495   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3496 
3497   // Come here on success
3498   __ bind(ok_is_subtype);
3499   __ mov(rax,rdx);           // Restore object in EDX
3500 
3501   // Collect counts on whether this check-cast sees NULLs a lot or not.
3502   if (ProfileInterpreter) {
3503     __ jmp(done);
3504     __ bind(is_null);
3505     __ profile_null_seen(rcx);
3506   } else {
3507     __ bind(is_null);   // same as 'done'
3508   }
3509   __ bind(done);
3510 }
3511 
3512 
3513 void TemplateTable::instanceof() {
3514   transition(atos, itos);
3515   Label done, is_null, ok_is_subtype, quicked, resolved;
3516   __ testptr(rax, rax);
3517   __ jcc(Assembler::zero, is_null);
3518 
3519   // Get cpool & tags index
3520   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3521   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3522   // See if bytecode has already been quicked
3523   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3524   __ jcc(Assembler::equal, quicked);
3525 
3526   __ push(atos);
3527   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3528   // vm_result_2 has metadata result
3529   // borrow rdi from locals
3530   __ get_thread(rdi);
3531   __ get_vm_result_2(rax, rdi);
3532   __ restore_locals();
3533   __ pop_ptr(rdx);
3534   __ load_klass(rdx, rdx);
3535   __ jmp(resolved);
3536 
3537   // Get superklass in EAX and subklass in EDX
3538   __ bind(quicked);
3539   __ load_klass(rdx, rax);
3540   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3541 
3542   __ bind(resolved);
3543 
3544   // Generate subtype check.  Blows ECX.  Resets EDI.
3545   // Superklass in EAX.  Subklass in EDX.
3546   __ gen_subtype_check( rdx, ok_is_subtype );
3547 
3548   // Come here on failure
3549   __ xorl(rax,rax);
3550   __ jmpb(done);
3551   // Come here on success
3552   __ bind(ok_is_subtype);
3553   __ movl(rax, 1);
3554 
3555   // Collect counts on whether this test sees NULLs a lot or not.
3556   if (ProfileInterpreter) {
3557     __ jmp(done);
3558     __ bind(is_null);
3559     __ profile_null_seen(rcx);
3560   } else {
3561     __ bind(is_null);   // same as 'done'
3562   }
3563   __ bind(done);
3564   // rax, = 0: obj == NULL or  obj is not an instanceof the specified klass
3565   // rax, = 1: obj != NULL and obj is     an instanceof the specified klass
3566 }
3567 
3568 
3569 //----------------------------------------------------------------------------------------------------
3570 // Breakpoints
3571 void TemplateTable::_breakpoint() {
3572 
3573   // Note: We get here even if we are single stepping..
3574   // jbug inists on setting breakpoints at every bytecode
3575   // even if we are in single step mode.
3576 
3577   transition(vtos, vtos);
3578 
3579   // get the unpatched byte code
3580   __ get_method(rcx);
3581   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::get_original_bytecode_at), rcx, rsi);
3582   __ mov(rbx, rax);
3583 
3584   // post the breakpoint event
3585   __ get_method(rcx);
3586   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint), rcx, rsi);
3587 
3588   // complete the execution of original bytecode
3589   __ dispatch_only_normal(vtos);
3590 }
3591 
3592 
3593 //----------------------------------------------------------------------------------------------------
3594 // Exceptions
3595 
3596 void TemplateTable::athrow() {
3597   transition(atos, vtos);
3598   __ null_check(rax);
3599   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3600 }
3601 
3602 
3603 //----------------------------------------------------------------------------------------------------
3604 // Synchronization
3605 //
3606 // Note: monitorenter & exit are symmetric routines; which is reflected
3607 //       in the assembly code structure as well
3608 //
3609 // Stack layout:
3610 //
3611 // [expressions  ] <--- rsp               = expression stack top
3612 // ..
3613 // [expressions  ]
3614 // [monitor entry] <--- monitor block top = expression stack bot
3615 // ..
3616 // [monitor entry]
3617 // [frame data   ] <--- monitor block bot
3618 // ...
3619 // [saved rbp,    ] <--- rbp,
3620 
3621 
3622 void TemplateTable::monitorenter() {
3623   transition(atos, vtos);
3624 
3625   // check for NULL object
3626   __ null_check(rax);
3627 
3628   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3629   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3630   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3631   Label allocated;
3632 
3633   // initialize entry pointer
3634   __ xorl(rdx, rdx);                             // points to free slot or NULL
3635 
3636   // find a free slot in the monitor block (result in rdx)
3637   { Label entry, loop, exit;
3638     __ movptr(rcx, monitor_block_top);           // points to current entry, starting with top-most entry
3639 
3640     __ lea(rbx, monitor_block_bot);              // points to word before bottom of monitor block
3641     __ jmpb(entry);
3642 
3643     __ bind(loop);
3644     __ cmpptr(Address(rcx, BasicObjectLock::obj_offset_in_bytes()), (int32_t)NULL_WORD);  // check if current entry is used
3645     __ cmovptr(Assembler::equal, rdx, rcx);      // if not used then remember entry in rdx
3646     __ cmpptr(rax, Address(rcx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3647     __ jccb(Assembler::equal, exit);             // if same object then stop searching
3648     __ addptr(rcx, entry_size);                  // otherwise advance to next entry
3649     __ bind(entry);
3650     __ cmpptr(rcx, rbx);                         // check if bottom reached
3651     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3652     __ bind(exit);
3653   }
3654 
3655   __ testptr(rdx, rdx);                          // check if a slot has been found
3656   __ jccb(Assembler::notZero, allocated);        // if found, continue with that one
3657 
3658   // allocate one if there's no free slot
3659   { Label entry, loop;
3660     // 1. compute new pointers                   // rsp: old expression stack top
3661     __ movptr(rdx, monitor_block_bot);           // rdx: old expression stack bottom
3662     __ subptr(rsp, entry_size);                  // move expression stack top
3663     __ subptr(rdx, entry_size);                  // move expression stack bottom
3664     __ mov(rcx, rsp);                            // set start value for copy loop
3665     __ movptr(monitor_block_bot, rdx);           // set new monitor block top
3666     __ jmp(entry);
3667     // 2. move expression stack contents
3668     __ bind(loop);
3669     __ movptr(rbx, Address(rcx, entry_size));    // load expression stack word from old location
3670     __ movptr(Address(rcx, 0), rbx);             // and store it at new location
3671     __ addptr(rcx, wordSize);                    // advance to next word
3672     __ bind(entry);
3673     __ cmpptr(rcx, rdx);                         // check if bottom reached
3674     __ jcc(Assembler::notEqual, loop);           // if not at bottom then copy next word
3675   }
3676 
3677   // call run-time routine
3678   // rdx: points to monitor entry
3679   __ bind(allocated);
3680 
3681   // Increment bcp to point to the next bytecode, so exception handling for async. exceptions work correctly.
3682   // The object has already been poped from the stack, so the expression stack looks correct.
3683   __ increment(rsi);
3684 
3685   __ movptr(Address(rdx, BasicObjectLock::obj_offset_in_bytes()), rax);     // store object
3686   __ lock_object(rdx);
3687 
3688   // check to make sure this monitor doesn't cause stack overflow after locking
3689   __ save_bcp();  // in case of exception
3690   __ generate_stack_overflow_check(0);
3691 
3692   // The bcp has already been incremented. Just need to dispatch to next instruction.
3693   __ dispatch_next(vtos);
3694 }
3695 
3696 
3697 void TemplateTable::monitorexit() {
3698   transition(atos, vtos);
3699 
3700   // check for NULL object
3701   __ null_check(rax);
3702 
3703   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3704   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3705   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3706   Label found;
3707 
3708   // find matching slot
3709   { Label entry, loop;
3710     __ movptr(rdx, monitor_block_top);           // points to current entry, starting with top-most entry
3711     __ lea(rbx, monitor_block_bot);             // points to word before bottom of monitor block
3712     __ jmpb(entry);
3713 
3714     __ bind(loop);
3715     __ cmpptr(rax, Address(rdx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3716     __ jcc(Assembler::equal, found);             // if same object then stop searching
3717     __ addptr(rdx, entry_size);                  // otherwise advance to next entry
3718     __ bind(entry);
3719     __ cmpptr(rdx, rbx);                         // check if bottom reached
3720     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3721   }
3722 
3723   // error handling. Unlocking was not block-structured
3724   Label end;
3725   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_illegal_monitor_state_exception));
3726   __ should_not_reach_here();
3727 
3728   // call run-time routine
3729   // rcx: points to monitor entry
3730   __ bind(found);
3731   __ push_ptr(rax);                                 // make sure object is on stack (contract with oopMaps)
3732   __ unlock_object(rdx);
3733   __ pop_ptr(rax);                                  // discard object
3734   __ bind(end);
3735 }
3736 
3737 
3738 //----------------------------------------------------------------------------------------------------
3739 // Wide instructions
3740 
3741 void TemplateTable::wide() {
3742   transition(vtos, vtos);
3743   __ load_unsigned_byte(rbx, at_bcp(1));
3744   ExternalAddress wtable((address)Interpreter::_wentry_point);
3745   __ jump(ArrayAddress(wtable, Address(noreg, rbx, Address::times_ptr)));
3746   // Note: the rsi increment step is part of the individual wide bytecode implementations
3747 }
3748 
3749 
3750 //----------------------------------------------------------------------------------------------------
3751 // Multi arrays
3752 
3753 void TemplateTable::multianewarray() {
3754   transition(vtos, atos);
3755   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3756   // last dim is on top of stack; we want address of first one:
3757   // first_addr = last_addr + (ndims - 1) * stackElementSize - 1*wordsize
3758   // the latter wordSize to point to the beginning of the array.
3759   __ lea(  rax, Address(rsp, rax, Interpreter::stackElementScale(), -wordSize));
3760   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray), rax);     // pass in rax,
3761   __ load_unsigned_byte(rbx, at_bcp(3));
3762   __ lea(rsp, Address(rsp, rbx, Interpreter::stackElementScale()));  // get rid of counts
3763 }
3764 
3765 #endif /* !CC_INTERP */