1 /*
   2  * Copyright (c) 1997, 2014, 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/interp_masm.hpp"
  30 #include "interpreter/templateTable.hpp"
  31 #include "memory/universe.inline.hpp"
  32 #include "oops/methodData.hpp"
  33 #include "oops/objArrayKlass.hpp"
  34 #include "oops/oop.inline.hpp"
  35 #include "prims/methodHandles.hpp"
  36 #include "runtime/sharedRuntime.hpp"
  37 #include "runtime/stubRoutines.hpp"
  38 #include "runtime/synchronizer.hpp"
  39 #include "utilities/macros.hpp"
  40 
  41 #ifndef CC_INTERP
  42 #define __ _masm->
  43 
  44 //----------------------------------------------------------------------------------------------------
  45 // Platform-dependent initialization
  46 
  47 void TemplateTable::pd_initialize() {
  48   // No i486 specific initialization
  49 }
  50 
  51 //----------------------------------------------------------------------------------------------------
  52 // Address computation
  53 
  54 // local variables
  55 static inline Address iaddress(int n)            {
  56   return Address(rdi, Interpreter::local_offset_in_bytes(n));
  57 }
  58 
  59 static inline Address laddress(int n)            { return iaddress(n + 1); }
  60 static inline Address haddress(int n)            { return iaddress(n + 0); }
  61 static inline Address faddress(int n)            { return iaddress(n); }
  62 static inline Address daddress(int n)            { return laddress(n); }
  63 static inline Address aaddress(int n)            { return iaddress(n); }
  64 
  65 static inline Address iaddress(Register r)       {
  66   return Address(rdi, r, Interpreter::stackElementScale());
  67 }
  68 static inline Address laddress(Register r)       {
  69   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(1));
  70 }
  71 static inline Address haddress(Register r)       {
  72   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(0));
  73 }
  74 
  75 static inline Address faddress(Register r)       { return iaddress(r); }
  76 static inline Address daddress(Register r)       { return laddress(r); }
  77 static inline Address aaddress(Register r)       { return iaddress(r); }
  78 
  79 // expression stack
  80 // (Note: Must not use symmetric equivalents at_rsp_m1/2 since they store
  81 // data beyond the rsp which is potentially unsafe in an MT environment;
  82 // an interrupt may overwrite that data.)
  83 static inline Address at_rsp   () {
  84   return Address(rsp, 0);
  85 }
  86 
  87 // At top of Java expression stack which may be different than rsp().  It
  88 // isn't for category 1 objects.
  89 static inline Address at_tos   () {
  90   Address tos = Address(rsp,  Interpreter::expr_offset_in_bytes(0));
  91   return tos;
  92 }
  93 
  94 static inline Address at_tos_p1() {
  95   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
  96 }
  97 
  98 static inline Address at_tos_p2() {
  99   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
 100 }
 101 
 102 // Condition conversion
 103 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 104   switch (cc) {
 105     case TemplateTable::equal        : return Assembler::notEqual;
 106     case TemplateTable::not_equal    : return Assembler::equal;
 107     case TemplateTable::less         : return Assembler::greaterEqual;
 108     case TemplateTable::less_equal   : return Assembler::greater;
 109     case TemplateTable::greater      : return Assembler::lessEqual;
 110     case TemplateTable::greater_equal: return Assembler::less;
 111   }
 112   ShouldNotReachHere();
 113   return Assembler::zero;
 114 }
 115 
 116 
 117 //----------------------------------------------------------------------------------------------------
 118 // Miscelaneous helper routines
 119 
 120 // Store an oop (or NULL) at the address described by obj.
 121 // If val == noreg this means store a NULL
 122 
 123 static void do_oop_store(InterpreterMacroAssembler* _masm,
 124                          Address obj,
 125                          Register val,
 126                          BarrierSet::Name barrier,
 127                          bool precise) {
 128   assert(val == noreg || val == rax, "parameter is just for looks");
 129   switch (barrier) {
 130 #if INCLUDE_ALL_GCS
 131     case BarrierSet::G1SATBCT:
 132     case BarrierSet::G1SATBCTLogging:
 133       {
 134         // flatten object address if needed
 135         // We do it regardless of precise because we need the registers
 136         if (obj.index() == noreg && obj.disp() == 0) {
 137           if (obj.base() != rdx) {
 138             __ movl(rdx, obj.base());
 139           }
 140         } else {
 141           __ leal(rdx, obj);
 142         }
 143         __ get_thread(rcx);
 144         __ save_bcp();
 145         __ g1_write_barrier_pre(rdx /* obj */,
 146                                 rbx /* pre_val */,
 147                                 rcx /* thread */,
 148                                 rsi /* tmp */,
 149                                 val != noreg /* tosca_live */,
 150                                 false /* expand_call */);
 151 
 152         // Do the actual store
 153         // noreg means NULL
 154         if (val == noreg) {
 155           __ movptr(Address(rdx, 0), NULL_WORD);
 156           // No post barrier for NULL
 157         } else {
 158           __ movl(Address(rdx, 0), val);
 159           __ g1_write_barrier_post(rdx /* store_adr */,
 160                                    val /* new_val */,
 161                                    rcx /* thread */,
 162                                    rbx /* tmp */,
 163                                    rsi /* tmp2 */);
 164         }
 165         __ restore_bcp();
 166 
 167       }
 168       break;
 169 #endif // INCLUDE_ALL_GCS
 170     case BarrierSet::CardTableModRef:
 171     case BarrierSet::CardTableExtension:
 172       {
 173         if (val == noreg) {
 174           __ movptr(obj, NULL_WORD);
 175         } else {
 176           __ movl(obj, val);
 177           // flatten object address if needed
 178           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 179             __ store_check(obj.base());
 180           } else {
 181             __ leal(rdx, obj);
 182             __ store_check(rdx);
 183           }
 184         }
 185       }
 186       break;
 187     case BarrierSet::ModRef:
 188     case BarrierSet::Other:
 189       if (val == noreg) {
 190         __ movptr(obj, NULL_WORD);
 191       } else {
 192         __ movl(obj, val);
 193       }
 194       break;
 195     default      :
 196       ShouldNotReachHere();
 197 
 198   }
 199 }
 200 
 201 Address TemplateTable::at_bcp(int offset) {
 202   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 203   return Address(rsi, offset);
 204 }
 205 
 206 
 207 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register bc_reg,
 208                                    Register temp_reg, bool load_bc_into_bc_reg/*=true*/,
 209                                    int byte_no) {
 210   if (!RewriteBytecodes)  return;
 211   Label L_patch_done;
 212 
 213   switch (bc) {
 214   case Bytecodes::_fast_aputfield:
 215   case Bytecodes::_fast_bputfield:
 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   // rdx: array
 994   index_check(rdx, rbx);  // prefer index in rbx,
 995   // rbx,: index
 996   __ movb(Address(rdx, rbx, Address::times_1, arrayOopDesc::base_offset_in_bytes(T_BYTE)), rax);
 997 }
 998 
 999 
1000 void TemplateTable::castore() {
1001   transition(itos, vtos);
1002   __ pop_i(rbx);
1003   // rax,: value
1004   // rdx: array
1005   index_check(rdx, rbx);  // prefer index in rbx,
1006   // rbx,: index
1007   __ movw(Address(rdx, rbx, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)), rax);
1008 }
1009 
1010 
1011 void TemplateTable::sastore() {
1012   castore();
1013 }
1014 
1015 
1016 void TemplateTable::istore(int n) {
1017   transition(itos, vtos);
1018   __ movl(iaddress(n), rax);
1019 }
1020 
1021 
1022 void TemplateTable::lstore(int n) {
1023   transition(ltos, vtos);
1024   __ movptr(laddress(n), rax);
1025   NOT_LP64(__ movptr(haddress(n), rdx));
1026 }
1027 
1028 
1029 void TemplateTable::fstore(int n) {
1030   transition(ftos, vtos);
1031   __ fstp_s(faddress(n));
1032 }
1033 
1034 
1035 void TemplateTable::dstore(int n) {
1036   transition(dtos, vtos);
1037   __ fstp_d(daddress(n));
1038 }
1039 
1040 
1041 void TemplateTable::astore(int n) {
1042   transition(vtos, vtos);
1043   __ pop_ptr(rax);
1044   __ movptr(aaddress(n), rax);
1045 }
1046 
1047 
1048 void TemplateTable::pop() {
1049   transition(vtos, vtos);
1050   __ addptr(rsp, Interpreter::stackElementSize);
1051 }
1052 
1053 
1054 void TemplateTable::pop2() {
1055   transition(vtos, vtos);
1056   __ addptr(rsp, 2*Interpreter::stackElementSize);
1057 }
1058 
1059 
1060 void TemplateTable::dup() {
1061   transition(vtos, vtos);
1062   // stack: ..., a
1063   __ load_ptr(0, rax);
1064   __ push_ptr(rax);
1065   // stack: ..., a, a
1066 }
1067 
1068 
1069 void TemplateTable::dup_x1() {
1070   transition(vtos, vtos);
1071   // stack: ..., a, b
1072   __ load_ptr( 0, rax);  // load b
1073   __ load_ptr( 1, rcx);  // load a
1074   __ store_ptr(1, rax);  // store b
1075   __ store_ptr(0, rcx);  // store a
1076   __ push_ptr(rax);      // push b
1077   // stack: ..., b, a, b
1078 }
1079 
1080 
1081 void TemplateTable::dup_x2() {
1082   transition(vtos, vtos);
1083   // stack: ..., a, b, c
1084   __ load_ptr( 0, rax);  // load c
1085   __ load_ptr( 2, rcx);  // load a
1086   __ store_ptr(2, rax);  // store c in a
1087   __ push_ptr(rax);      // push c
1088   // stack: ..., c, b, c, c
1089   __ load_ptr( 2, rax);  // load b
1090   __ store_ptr(2, rcx);  // store a in b
1091   // stack: ..., c, a, c, c
1092   __ store_ptr(1, rax);  // store b in c
1093   // stack: ..., c, a, b, c
1094 }
1095 
1096 
1097 void TemplateTable::dup2() {
1098   transition(vtos, vtos);
1099   // stack: ..., a, b
1100   __ load_ptr(1, rax);  // load a
1101   __ push_ptr(rax);     // push a
1102   __ load_ptr(1, rax);  // load b
1103   __ push_ptr(rax);     // push b
1104   // stack: ..., a, b, a, b
1105 }
1106 
1107 
1108 void TemplateTable::dup2_x1() {
1109   transition(vtos, vtos);
1110   // stack: ..., a, b, c
1111   __ load_ptr( 0, rcx);  // load c
1112   __ load_ptr( 1, rax);  // load b
1113   __ push_ptr(rax);      // push b
1114   __ push_ptr(rcx);      // push c
1115   // stack: ..., a, b, c, b, c
1116   __ store_ptr(3, rcx);  // store c in b
1117   // stack: ..., a, c, c, b, c
1118   __ load_ptr( 4, rcx);  // load a
1119   __ store_ptr(2, rcx);  // store a in 2nd c
1120   // stack: ..., a, c, a, b, c
1121   __ store_ptr(4, rax);  // store b in a
1122   // stack: ..., b, c, a, b, c
1123   // stack: ..., b, c, a, b, c
1124 }
1125 
1126 
1127 void TemplateTable::dup2_x2() {
1128   transition(vtos, vtos);
1129   // stack: ..., a, b, c, d
1130   __ load_ptr( 0, rcx);  // load d
1131   __ load_ptr( 1, rax);  // load c
1132   __ push_ptr(rax);      // push c
1133   __ push_ptr(rcx);      // push d
1134   // stack: ..., a, b, c, d, c, d
1135   __ load_ptr( 4, rax);  // load b
1136   __ store_ptr(2, rax);  // store b in d
1137   __ store_ptr(4, rcx);  // store d in b
1138   // stack: ..., a, d, c, b, c, d
1139   __ load_ptr( 5, rcx);  // load a
1140   __ load_ptr( 3, rax);  // load c
1141   __ store_ptr(3, rcx);  // store a in c
1142   __ store_ptr(5, rax);  // store c in a
1143   // stack: ..., c, d, a, b, c, d
1144   // stack: ..., c, d, a, b, c, d
1145 }
1146 
1147 
1148 void TemplateTable::swap() {
1149   transition(vtos, vtos);
1150   // stack: ..., a, b
1151   __ load_ptr( 1, rcx);  // load a
1152   __ load_ptr( 0, rax);  // load b
1153   __ store_ptr(0, rcx);  // store a in b
1154   __ store_ptr(1, rax);  // store b in a
1155   // stack: ..., b, a
1156 }
1157 
1158 
1159 void TemplateTable::iop2(Operation op) {
1160   transition(itos, itos);
1161   switch (op) {
1162     case add  :                   __ pop_i(rdx); __ addl (rax, rdx); break;
1163     case sub  : __ mov(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1164     case mul  :                   __ pop_i(rdx); __ imull(rax, rdx); break;
1165     case _and :                   __ pop_i(rdx); __ andl (rax, rdx); break;
1166     case _or  :                   __ pop_i(rdx); __ orl  (rax, rdx); break;
1167     case _xor :                   __ pop_i(rdx); __ xorl (rax, rdx); break;
1168     case shl  : __ mov(rcx, rax); __ pop_i(rax); __ shll (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1169     case shr  : __ mov(rcx, rax); __ pop_i(rax); __ sarl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1170     case ushr : __ mov(rcx, rax); __ pop_i(rax); __ shrl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1171     default   : ShouldNotReachHere();
1172   }
1173 }
1174 
1175 
1176 void TemplateTable::lop2(Operation op) {
1177   transition(ltos, ltos);
1178   __ pop_l(rbx, rcx);
1179   switch (op) {
1180     case add  : __ addl(rax, rbx); __ adcl(rdx, rcx); break;
1181     case sub  : __ subl(rbx, rax); __ sbbl(rcx, rdx);
1182                 __ mov (rax, rbx); __ mov (rdx, rcx); break;
1183     case _and : __ andl(rax, rbx); __ andl(rdx, rcx); break;
1184     case _or  : __ orl (rax, rbx); __ orl (rdx, rcx); break;
1185     case _xor : __ xorl(rax, rbx); __ xorl(rdx, rcx); break;
1186     default   : ShouldNotReachHere();
1187   }
1188 }
1189 
1190 
1191 void TemplateTable::idiv() {
1192   transition(itos, itos);
1193   __ mov(rcx, rax);
1194   __ pop_i(rax);
1195   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1196   //       they are not equal, one could do a normal division (no correction
1197   //       needed), which may speed up this implementation for the common case.
1198   //       (see also JVM spec., p.243 & p.271)
1199   __ corrected_idivl(rcx);
1200 }
1201 
1202 
1203 void TemplateTable::irem() {
1204   transition(itos, itos);
1205   __ mov(rcx, rax);
1206   __ pop_i(rax);
1207   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1208   //       they are not equal, one could do a normal division (no correction
1209   //       needed), which may speed up this implementation for the common case.
1210   //       (see also JVM spec., p.243 & p.271)
1211   __ corrected_idivl(rcx);
1212   __ mov(rax, rdx);
1213 }
1214 
1215 
1216 void TemplateTable::lmul() {
1217   transition(ltos, ltos);
1218   __ pop_l(rbx, rcx);
1219   __ push(rcx); __ push(rbx);
1220   __ push(rdx); __ push(rax);
1221   __ lmul(2 * wordSize, 0);
1222   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1223 }
1224 
1225 
1226 void TemplateTable::ldiv() {
1227   transition(ltos, ltos);
1228   __ pop_l(rbx, rcx);
1229   __ push(rcx); __ push(rbx);
1230   __ push(rdx); __ push(rax);
1231   // check if y = 0
1232   __ orl(rax, rdx);
1233   __ jump_cc(Assembler::zero,
1234              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1235   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::ldiv));
1236   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1237 }
1238 
1239 
1240 void TemplateTable::lrem() {
1241   transition(ltos, ltos);
1242   __ pop_l(rbx, rcx);
1243   __ push(rcx); __ push(rbx);
1244   __ push(rdx); __ push(rax);
1245   // check if y = 0
1246   __ orl(rax, rdx);
1247   __ jump_cc(Assembler::zero,
1248              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1249   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::lrem));
1250   __ addptr(rsp, 4 * wordSize);
1251 }
1252 
1253 
1254 void TemplateTable::lshl() {
1255   transition(itos, ltos);
1256   __ movl(rcx, rax);                             // get shift count
1257   __ pop_l(rax, rdx);                            // get shift value
1258   __ lshl(rdx, rax);
1259 }
1260 
1261 
1262 void TemplateTable::lshr() {
1263   transition(itos, ltos);
1264   __ mov(rcx, rax);                              // get shift count
1265   __ pop_l(rax, rdx);                            // get shift value
1266   __ lshr(rdx, rax, true);
1267 }
1268 
1269 
1270 void TemplateTable::lushr() {
1271   transition(itos, ltos);
1272   __ mov(rcx, rax);                              // get shift count
1273   __ pop_l(rax, rdx);                            // get shift value
1274   __ lshr(rdx, rax);
1275 }
1276 
1277 
1278 void TemplateTable::fop2(Operation op) {
1279   transition(ftos, ftos);
1280   switch (op) {
1281     case add: __ fadd_s (at_rsp());                break;
1282     case sub: __ fsubr_s(at_rsp());                break;
1283     case mul: __ fmul_s (at_rsp());                break;
1284     case div: __ fdivr_s(at_rsp());                break;
1285     case rem: __ fld_s  (at_rsp()); __ fremr(rax); break;
1286     default : ShouldNotReachHere();
1287   }
1288   __ f2ieee();
1289   __ pop(rax);  // pop float thing off
1290 }
1291 
1292 
1293 void TemplateTable::dop2(Operation op) {
1294   transition(dtos, dtos);
1295 
1296   switch (op) {
1297     case add: __ fadd_d (at_rsp());                break;
1298     case sub: __ fsubr_d(at_rsp());                break;
1299     case mul: {
1300       Label L_strict;
1301       Label L_join;
1302       const Address access_flags      (rcx, Method::access_flags_offset());
1303       __ get_method(rcx);
1304       __ movl(rcx, access_flags);
1305       __ testl(rcx, JVM_ACC_STRICT);
1306       __ jccb(Assembler::notZero, L_strict);
1307       __ fmul_d (at_rsp());
1308       __ jmpb(L_join);
1309       __ bind(L_strict);
1310       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1311       __ fmulp();
1312       __ fmul_d (at_rsp());
1313       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1314       __ fmulp();
1315       __ bind(L_join);
1316       break;
1317     }
1318     case div: {
1319       Label L_strict;
1320       Label L_join;
1321       const Address access_flags      (rcx, Method::access_flags_offset());
1322       __ get_method(rcx);
1323       __ movl(rcx, access_flags);
1324       __ testl(rcx, JVM_ACC_STRICT);
1325       __ jccb(Assembler::notZero, L_strict);
1326       __ fdivr_d(at_rsp());
1327       __ jmp(L_join);
1328       __ bind(L_strict);
1329       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1330       __ fmul_d (at_rsp());
1331       __ fdivrp();
1332       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1333       __ fmulp();
1334       __ bind(L_join);
1335       break;
1336     }
1337     case rem: __ fld_d  (at_rsp()); __ fremr(rax); break;
1338     default : ShouldNotReachHere();
1339   }
1340   __ d2ieee();
1341   // Pop double precision number from rsp.
1342   __ pop(rax);
1343   __ pop(rdx);
1344 }
1345 
1346 
1347 void TemplateTable::ineg() {
1348   transition(itos, itos);
1349   __ negl(rax);
1350 }
1351 
1352 
1353 void TemplateTable::lneg() {
1354   transition(ltos, ltos);
1355   __ lneg(rdx, rax);
1356 }
1357 
1358 
1359 void TemplateTable::fneg() {
1360   transition(ftos, ftos);
1361   __ fchs();
1362 }
1363 
1364 
1365 void TemplateTable::dneg() {
1366   transition(dtos, dtos);
1367   __ fchs();
1368 }
1369 
1370 
1371 void TemplateTable::iinc() {
1372   transition(vtos, vtos);
1373   __ load_signed_byte(rdx, at_bcp(2));           // get constant
1374   locals_index(rbx);
1375   __ addl(iaddress(rbx), rdx);
1376 }
1377 
1378 
1379 void TemplateTable::wide_iinc() {
1380   transition(vtos, vtos);
1381   __ movl(rdx, at_bcp(4));                       // get constant
1382   locals_index_wide(rbx);
1383   __ bswapl(rdx);                                 // swap bytes & sign-extend constant
1384   __ sarl(rdx, 16);
1385   __ addl(iaddress(rbx), rdx);
1386   // Note: should probably use only one movl to get both
1387   //       the index and the constant -> fix this
1388 }
1389 
1390 
1391 void TemplateTable::convert() {
1392   // Checking
1393 #ifdef ASSERT
1394   { TosState tos_in  = ilgl;
1395     TosState tos_out = ilgl;
1396     switch (bytecode()) {
1397       case Bytecodes::_i2l: // fall through
1398       case Bytecodes::_i2f: // fall through
1399       case Bytecodes::_i2d: // fall through
1400       case Bytecodes::_i2b: // fall through
1401       case Bytecodes::_i2c: // fall through
1402       case Bytecodes::_i2s: tos_in = itos; break;
1403       case Bytecodes::_l2i: // fall through
1404       case Bytecodes::_l2f: // fall through
1405       case Bytecodes::_l2d: tos_in = ltos; break;
1406       case Bytecodes::_f2i: // fall through
1407       case Bytecodes::_f2l: // fall through
1408       case Bytecodes::_f2d: tos_in = ftos; break;
1409       case Bytecodes::_d2i: // fall through
1410       case Bytecodes::_d2l: // fall through
1411       case Bytecodes::_d2f: tos_in = dtos; break;
1412       default             : ShouldNotReachHere();
1413     }
1414     switch (bytecode()) {
1415       case Bytecodes::_l2i: // fall through
1416       case Bytecodes::_f2i: // fall through
1417       case Bytecodes::_d2i: // fall through
1418       case Bytecodes::_i2b: // fall through
1419       case Bytecodes::_i2c: // fall through
1420       case Bytecodes::_i2s: tos_out = itos; break;
1421       case Bytecodes::_i2l: // fall through
1422       case Bytecodes::_f2l: // fall through
1423       case Bytecodes::_d2l: tos_out = ltos; break;
1424       case Bytecodes::_i2f: // fall through
1425       case Bytecodes::_l2f: // fall through
1426       case Bytecodes::_d2f: tos_out = ftos; break;
1427       case Bytecodes::_i2d: // fall through
1428       case Bytecodes::_l2d: // fall through
1429       case Bytecodes::_f2d: tos_out = dtos; break;
1430       default             : ShouldNotReachHere();
1431     }
1432     transition(tos_in, tos_out);
1433   }
1434 #endif // ASSERT
1435 
1436   // Conversion
1437   // (Note: use push(rcx)/pop(rcx) for 1/2-word stack-ptr manipulation)
1438   switch (bytecode()) {
1439     case Bytecodes::_i2l:
1440       __ extend_sign(rdx, rax);
1441       break;
1442     case Bytecodes::_i2f:
1443       __ push(rax);          // store int on tos
1444       __ fild_s(at_rsp());   // load int to ST0
1445       __ f2ieee();           // truncate to float size
1446       __ pop(rcx);           // adjust rsp
1447       break;
1448     case Bytecodes::_i2d:
1449       __ push(rax);          // add one slot for d2ieee()
1450       __ push(rax);          // store int on tos
1451       __ fild_s(at_rsp());   // load int to ST0
1452       __ d2ieee();           // truncate to double size
1453       __ pop(rcx);           // adjust rsp
1454       __ pop(rcx);
1455       break;
1456     case Bytecodes::_i2b:
1457       __ shll(rax, 24);      // truncate upper 24 bits
1458       __ sarl(rax, 24);      // and sign-extend byte
1459       LP64_ONLY(__ movsbl(rax, rax));
1460       break;
1461     case Bytecodes::_i2c:
1462       __ andl(rax, 0xFFFF);  // truncate upper 16 bits
1463       LP64_ONLY(__ movzwl(rax, rax));
1464       break;
1465     case Bytecodes::_i2s:
1466       __ shll(rax, 16);      // truncate upper 16 bits
1467       __ sarl(rax, 16);      // and sign-extend short
1468       LP64_ONLY(__ movswl(rax, rax));
1469       break;
1470     case Bytecodes::_l2i:
1471       /* nothing to do */
1472       break;
1473     case Bytecodes::_l2f:
1474       __ push(rdx);          // store long on tos
1475       __ push(rax);
1476       __ fild_d(at_rsp());   // load long to ST0
1477       __ f2ieee();           // truncate to float size
1478       __ pop(rcx);           // adjust rsp
1479       __ pop(rcx);
1480       break;
1481     case Bytecodes::_l2d:
1482       __ push(rdx);          // store long on tos
1483       __ push(rax);
1484       __ fild_d(at_rsp());   // load long to ST0
1485       __ d2ieee();           // truncate to double size
1486       __ pop(rcx);           // adjust rsp
1487       __ pop(rcx);
1488       break;
1489     case Bytecodes::_f2i:
1490       __ push(rcx);          // reserve space for argument
1491       __ fstp_s(at_rsp());   // pass float argument on stack
1492       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1493       break;
1494     case Bytecodes::_f2l:
1495       __ push(rcx);          // reserve space for argument
1496       __ fstp_s(at_rsp());   // pass float argument on stack
1497       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1498       break;
1499     case Bytecodes::_f2d:
1500       /* nothing to do */
1501       break;
1502     case Bytecodes::_d2i:
1503       __ push(rcx);          // reserve space for argument
1504       __ push(rcx);
1505       __ fstp_d(at_rsp());   // pass double argument on stack
1506       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 2);
1507       break;
1508     case Bytecodes::_d2l:
1509       __ push(rcx);          // reserve space for argument
1510       __ push(rcx);
1511       __ fstp_d(at_rsp());   // pass double argument on stack
1512       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 2);
1513       break;
1514     case Bytecodes::_d2f:
1515       __ push(rcx);          // reserve space for f2ieee()
1516       __ f2ieee();           // truncate to float size
1517       __ pop(rcx);           // adjust rsp
1518       break;
1519     default             :
1520       ShouldNotReachHere();
1521   }
1522 }
1523 
1524 
1525 void TemplateTable::lcmp() {
1526   transition(ltos, itos);
1527   // y = rdx:rax
1528   __ pop_l(rbx, rcx);             // get x = rcx:rbx
1529   __ lcmp2int(rcx, rbx, rdx, rax);// rcx := cmp(x, y)
1530   __ mov(rax, rcx);
1531 }
1532 
1533 
1534 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1535   if (is_float) {
1536     __ fld_s(at_rsp());
1537   } else {
1538     __ fld_d(at_rsp());
1539     __ pop(rdx);
1540   }
1541   __ pop(rcx);
1542   __ fcmp2int(rax, unordered_result < 0);
1543 }
1544 
1545 
1546 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1547   __ get_method(rcx);           // ECX holds method
1548   __ profile_taken_branch(rax,rbx); // EAX holds updated MDP, EBX holds bumped taken count
1549 
1550   const ByteSize be_offset = MethodCounters::backedge_counter_offset() +
1551                              InvocationCounter::counter_offset();
1552   const ByteSize inv_offset = MethodCounters::invocation_counter_offset() +
1553                               InvocationCounter::counter_offset();
1554 
1555   // Load up EDX with the branch displacement
1556   if (is_wide) {
1557     __ movl(rdx, at_bcp(1));
1558   } else {
1559     __ load_signed_short(rdx, at_bcp(1));
1560   }
1561   __ bswapl(rdx);
1562   if (!is_wide) __ sarl(rdx, 16);
1563   LP64_ONLY(__ movslq(rdx, rdx));
1564 
1565 
1566   // Handle all the JSR stuff here, then exit.
1567   // It's much shorter and cleaner than intermingling with the
1568   // non-JSR normal-branch stuff occurring below.
1569   if (is_jsr) {
1570     // Pre-load the next target bytecode into EBX
1571     __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1, 0));
1572 
1573     // compute return address as bci in rax,
1574     __ lea(rax, at_bcp((is_wide ? 5 : 3) - in_bytes(ConstMethod::codes_offset())));
1575     __ subptr(rax, Address(rcx, Method::const_offset()));
1576     // Adjust the bcp in RSI by the displacement in EDX
1577     __ addptr(rsi, rdx);
1578     // Push return address
1579     __ push_i(rax);
1580     // jsr returns vtos
1581     __ dispatch_only_noverify(vtos);
1582     return;
1583   }
1584 
1585   // Normal (non-jsr) branch handling
1586 
1587   // Adjust the bcp in RSI by the displacement in EDX
1588   __ addptr(rsi, rdx);
1589 
1590   assert(UseLoopCounter || !UseOnStackReplacement, "on-stack-replacement requires loop counters");
1591   Label backedge_counter_overflow;
1592   Label profile_method;
1593   Label dispatch;
1594   if (UseLoopCounter) {
1595     // increment backedge counter for backward branches
1596     // rax,: MDO
1597     // rbx,: MDO bumped taken-count
1598     // rcx: method
1599     // rdx: target offset
1600     // rsi: target bcp
1601     // rdi: locals pointer
1602     __ testl(rdx, rdx);             // check if forward or backward branch
1603     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1604 
1605     // check if MethodCounters exists
1606     Label has_counters;
1607     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1608     __ testptr(rax, rax);
1609     __ jcc(Assembler::notZero, has_counters);
1610     __ push(rdx);
1611     __ push(rcx);
1612     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::build_method_counters),
1613                rcx);
1614     __ pop(rcx);
1615     __ pop(rdx);
1616     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1617     __ testptr(rax, rax);
1618     __ jcc(Assembler::zero, dispatch);
1619     __ bind(has_counters);
1620 
1621     if (TieredCompilation) {
1622       Label no_mdo;
1623       int increment = InvocationCounter::count_increment;
1624       if (ProfileInterpreter) {
1625         // Are we profiling?
1626         __ movptr(rbx, Address(rcx, in_bytes(Method::method_data_offset())));
1627         __ testptr(rbx, rbx);
1628         __ jccb(Assembler::zero, no_mdo);
1629         // Increment the MDO backedge counter
1630         const Address mdo_backedge_counter(rbx, in_bytes(MethodData::backedge_counter_offset()) +
1631                                                 in_bytes(InvocationCounter::counter_offset()));
1632         const Address mask(rbx, in_bytes(MethodData::backedge_mask_offset()));
1633         __ increment_mask_and_jump(mdo_backedge_counter, increment, mask,
1634                                    rax, false, Assembler::zero, &backedge_counter_overflow);
1635         __ jmp(dispatch);
1636       }
1637       __ bind(no_mdo);
1638       // Increment backedge counter in MethodCounters*
1639       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1640       const Address mask(rcx, in_bytes(MethodCounters::backedge_mask_offset()));
1641       __ increment_mask_and_jump(Address(rcx, be_offset), increment, mask,
1642                                  rax, false, Assembler::zero, &backedge_counter_overflow);
1643     } else { // not TieredCompilation
1644       // increment counter
1645       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1646       __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1647       __ incrementl(rax, InvocationCounter::count_increment); // increment counter
1648       __ movl(Address(rcx, be_offset), rax);        // store counter
1649 
1650       __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1651 
1652       __ andl(rax, InvocationCounter::count_mask_value);     // and the status bits
1653       __ addl(rax, Address(rcx, be_offset));        // add both counters
1654 
1655       if (ProfileInterpreter) {
1656         // Test to see if we should create a method data oop
1657         __ cmp32(rax, Address(rcx, in_bytes(MethodCounters::interpreter_profile_limit_offset())));
1658         __ jcc(Assembler::less, dispatch);
1659 
1660         // if no method data exists, go to profile method
1661         __ test_method_data_pointer(rax, profile_method);
1662 
1663         if (UseOnStackReplacement) {
1664           // check for overflow against rbx, which is the MDO taken count
1665           __ cmp32(rbx, Address(rcx, in_bytes(MethodCounters::interpreter_backward_branch_limit_offset())));
1666           __ jcc(Assembler::below, dispatch);
1667 
1668           // When ProfileInterpreter is on, the backedge_count comes from the
1669           // MethodData*, which value does not get reset on the call to
1670           // frequency_counter_overflow().  To avoid excessive calls to the overflow
1671           // routine while the method is being compiled, add a second test to make
1672           // sure the overflow function is called only once every overflow_frequency.
1673           const int overflow_frequency = 1024;
1674           __ andptr(rbx, overflow_frequency-1);
1675           __ jcc(Assembler::zero, backedge_counter_overflow);
1676         }
1677       } else {
1678         if (UseOnStackReplacement) {
1679           // check for overflow against rax, which is the sum of the counters
1680           __ cmp32(rax, Address(rcx, in_bytes(MethodCounters::interpreter_backward_branch_limit_offset())));
1681           __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1682 
1683         }
1684       }
1685     }
1686     __ bind(dispatch);
1687   }
1688 
1689   // Pre-load the next target bytecode into EBX
1690   __ load_unsigned_byte(rbx, Address(rsi, 0));
1691 
1692   // continue with the bytecode @ target
1693   // rax,: return bci for jsr's, unused otherwise
1694   // rbx,: target bytecode
1695   // rsi: target bcp
1696   __ dispatch_only(vtos);
1697 
1698   if (UseLoopCounter) {
1699     if (ProfileInterpreter) {
1700       // Out-of-line code to allocate method data oop.
1701       __ bind(profile_method);
1702       __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::profile_method));
1703       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1704       __ set_method_data_pointer_for_bcp();
1705       __ jmp(dispatch);
1706     }
1707 
1708     if (UseOnStackReplacement) {
1709 
1710       // invocation counter overflow
1711       __ bind(backedge_counter_overflow);
1712       __ negptr(rdx);
1713       __ addptr(rdx, rsi);        // branch bcp
1714       call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::frequency_counter_overflow), rdx);
1715       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1716 
1717       // rax,: osr nmethod (osr ok) or NULL (osr not possible)
1718       // rbx,: target bytecode
1719       // rdx: scratch
1720       // rdi: locals pointer
1721       // rsi: bcp
1722       __ testptr(rax, rax);                      // test result
1723       __ jcc(Assembler::zero, dispatch);         // no osr if null
1724       // nmethod may have been invalidated (VM may block upon call_VM return)
1725       __ cmpb(Address(rax, nmethod::state_offset()), nmethod::in_use);
1726       __ jcc(Assembler::notEqual, dispatch);
1727 
1728       // We have the address of an on stack replacement routine in rax,
1729       // We need to prepare to execute the OSR method. First we must
1730       // migrate the locals and monitors off of the stack.
1731 
1732       __ mov(rbx, rax);                             // save the nmethod
1733 
1734       __ get_thread(rcx);
1735       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1736       // rax, is OSR buffer, move it to expected parameter location
1737       __ mov(rcx, rax);
1738 
1739       // pop the interpreter frame
1740       __ movptr(rdx, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1741       __ leave();                                // remove frame anchor
1742       __ pop(rdi);                               // get return address
1743       __ mov(rsp, rdx);                          // set sp to sender sp
1744 
1745       // Align stack pointer for compiled code (note that caller is
1746       // responsible for undoing this fixup by remembering the old SP
1747       // in an rbp,-relative location)
1748       __ andptr(rsp, -(StackAlignmentInBytes));
1749 
1750       // push the (possibly adjusted) return address
1751       __ push(rdi);
1752 
1753       // and begin the OSR nmethod
1754       __ jmp(Address(rbx, nmethod::osr_entry_point_offset()));
1755     }
1756   }
1757 }
1758 
1759 
1760 void TemplateTable::if_0cmp(Condition cc) {
1761   transition(itos, vtos);
1762   // assume branch is more often taken than not (loops use backward branches)
1763   Label not_taken;
1764   __ testl(rax, rax);
1765   __ jcc(j_not(cc), not_taken);
1766   branch(false, false);
1767   __ bind(not_taken);
1768   __ profile_not_taken_branch(rax);
1769 }
1770 
1771 
1772 void TemplateTable::if_icmp(Condition cc) {
1773   transition(itos, vtos);
1774   // assume branch is more often taken than not (loops use backward branches)
1775   Label not_taken;
1776   __ pop_i(rdx);
1777   __ cmpl(rdx, rax);
1778   __ jcc(j_not(cc), not_taken);
1779   branch(false, false);
1780   __ bind(not_taken);
1781   __ profile_not_taken_branch(rax);
1782 }
1783 
1784 
1785 void TemplateTable::if_nullcmp(Condition cc) {
1786   transition(atos, vtos);
1787   // assume branch is more often taken than not (loops use backward branches)
1788   Label not_taken;
1789   __ testptr(rax, rax);
1790   __ jcc(j_not(cc), not_taken);
1791   branch(false, false);
1792   __ bind(not_taken);
1793   __ profile_not_taken_branch(rax);
1794 }
1795 
1796 
1797 void TemplateTable::if_acmp(Condition cc) {
1798   transition(atos, vtos);
1799   // assume branch is more often taken than not (loops use backward branches)
1800   Label not_taken;
1801   __ pop_ptr(rdx);
1802   __ cmpptr(rdx, rax);
1803   __ jcc(j_not(cc), not_taken);
1804   branch(false, false);
1805   __ bind(not_taken);
1806   __ profile_not_taken_branch(rax);
1807 }
1808 
1809 
1810 void TemplateTable::ret() {
1811   transition(vtos, vtos);
1812   locals_index(rbx);
1813   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1814   __ profile_ret(rbx, rcx);
1815   __ get_method(rax);
1816   __ movptr(rsi, Address(rax, Method::const_offset()));
1817   __ lea(rsi, Address(rsi, rbx, Address::times_1,
1818                       ConstMethod::codes_offset()));
1819   __ dispatch_next(vtos);
1820 }
1821 
1822 
1823 void TemplateTable::wide_ret() {
1824   transition(vtos, vtos);
1825   locals_index_wide(rbx);
1826   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1827   __ profile_ret(rbx, rcx);
1828   __ get_method(rax);
1829   __ movptr(rsi, Address(rax, Method::const_offset()));
1830   __ lea(rsi, Address(rsi, rbx, Address::times_1, ConstMethod::codes_offset()));
1831   __ dispatch_next(vtos);
1832 }
1833 
1834 
1835 void TemplateTable::tableswitch() {
1836   Label default_case, continue_execution;
1837   transition(itos, vtos);
1838   // align rsi
1839   __ lea(rbx, at_bcp(wordSize));
1840   __ andptr(rbx, -wordSize);
1841   // load lo & hi
1842   __ movl(rcx, Address(rbx, 1 * wordSize));
1843   __ movl(rdx, Address(rbx, 2 * wordSize));
1844   __ bswapl(rcx);
1845   __ bswapl(rdx);
1846   // check against lo & hi
1847   __ cmpl(rax, rcx);
1848   __ jccb(Assembler::less, default_case);
1849   __ cmpl(rax, rdx);
1850   __ jccb(Assembler::greater, default_case);
1851   // lookup dispatch offset
1852   __ subl(rax, rcx);
1853   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1854   __ profile_switch_case(rax, rbx, rcx);
1855   // continue execution
1856   __ bind(continue_execution);
1857   __ bswapl(rdx);
1858   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1859   __ addptr(rsi, rdx);
1860   __ dispatch_only(vtos);
1861   // handle default
1862   __ bind(default_case);
1863   __ profile_switch_default(rax);
1864   __ movl(rdx, Address(rbx, 0));
1865   __ jmp(continue_execution);
1866 }
1867 
1868 
1869 void TemplateTable::lookupswitch() {
1870   transition(itos, itos);
1871   __ stop("lookupswitch bytecode should have been rewritten");
1872 }
1873 
1874 
1875 void TemplateTable::fast_linearswitch() {
1876   transition(itos, vtos);
1877   Label loop_entry, loop, found, continue_execution;
1878   // bswapl rax, so we can avoid bswapping the table entries
1879   __ bswapl(rax);
1880   // align rsi
1881   __ lea(rbx, at_bcp(wordSize));                // btw: should be able to get rid of this instruction (change offsets below)
1882   __ andptr(rbx, -wordSize);
1883   // set counter
1884   __ movl(rcx, Address(rbx, wordSize));
1885   __ bswapl(rcx);
1886   __ jmpb(loop_entry);
1887   // table search
1888   __ bind(loop);
1889   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * wordSize));
1890   __ jccb(Assembler::equal, found);
1891   __ bind(loop_entry);
1892   __ decrementl(rcx);
1893   __ jcc(Assembler::greaterEqual, loop);
1894   // default case
1895   __ profile_switch_default(rax);
1896   __ movl(rdx, Address(rbx, 0));
1897   __ jmpb(continue_execution);
1898   // entry found -> get offset
1899   __ bind(found);
1900   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * wordSize));
1901   __ profile_switch_case(rcx, rax, rbx);
1902   // continue execution
1903   __ bind(continue_execution);
1904   __ bswapl(rdx);
1905   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1906   __ addptr(rsi, rdx);
1907   __ dispatch_only(vtos);
1908 }
1909 
1910 
1911 void TemplateTable::fast_binaryswitch() {
1912   transition(itos, vtos);
1913   // Implementation using the following core algorithm:
1914   //
1915   // int binary_search(int key, LookupswitchPair* array, int n) {
1916   //   // Binary search according to "Methodik des Programmierens" by
1917   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1918   //   int i = 0;
1919   //   int j = n;
1920   //   while (i+1 < j) {
1921   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1922   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1923   //     // where a stands for the array and assuming that the (inexisting)
1924   //     // element a[n] is infinitely big.
1925   //     int h = (i + j) >> 1;
1926   //     // i < h < j
1927   //     if (key < array[h].fast_match()) {
1928   //       j = h;
1929   //     } else {
1930   //       i = h;
1931   //     }
1932   //   }
1933   //   // R: a[i] <= key < a[i+1] or Q
1934   //   // (i.e., if key is within array, i is the correct index)
1935   //   return i;
1936   // }
1937 
1938   // register allocation
1939   const Register key   = rax;                    // already set (tosca)
1940   const Register array = rbx;
1941   const Register i     = rcx;
1942   const Register j     = rdx;
1943   const Register h     = rdi;                    // needs to be restored
1944   const Register temp  = rsi;
1945   // setup array
1946   __ save_bcp();
1947 
1948   __ lea(array, at_bcp(3*wordSize));             // btw: should be able to get rid of this instruction (change offsets below)
1949   __ andptr(array, -wordSize);
1950   // initialize i & j
1951   __ xorl(i, i);                                 // i = 0;
1952   __ movl(j, Address(array, -wordSize));         // j = length(array);
1953   // Convert j into native byteordering
1954   __ bswapl(j);
1955   // and start
1956   Label entry;
1957   __ jmp(entry);
1958 
1959   // binary search loop
1960   { Label loop;
1961     __ bind(loop);
1962     // int h = (i + j) >> 1;
1963     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
1964     __ sarl(h, 1);                               // h = (i + j) >> 1;
1965     // if (key < array[h].fast_match()) {
1966     //   j = h;
1967     // } else {
1968     //   i = h;
1969     // }
1970     // Convert array[h].match to native byte-ordering before compare
1971     __ movl(temp, Address(array, h, Address::times_8, 0*wordSize));
1972     __ bswapl(temp);
1973     __ cmpl(key, temp);
1974     // j = h if (key <  array[h].fast_match())
1975     __ cmov32(Assembler::less        , j, h);
1976     // i = h if (key >= array[h].fast_match())
1977     __ cmov32(Assembler::greaterEqual, i, h);
1978     // while (i+1 < j)
1979     __ bind(entry);
1980     __ leal(h, Address(i, 1));                   // i+1
1981     __ cmpl(h, j);                               // i+1 < j
1982     __ jcc(Assembler::less, loop);
1983   }
1984 
1985   // end of binary search, result index is i (must check again!)
1986   Label default_case;
1987   // Convert array[i].match to native byte-ordering before compare
1988   __ movl(temp, Address(array, i, Address::times_8, 0*wordSize));
1989   __ bswapl(temp);
1990   __ cmpl(key, temp);
1991   __ jcc(Assembler::notEqual, default_case);
1992 
1993   // entry found -> j = offset
1994   __ movl(j , Address(array, i, Address::times_8, 1*wordSize));
1995   __ profile_switch_case(i, key, array);
1996   __ bswapl(j);
1997   LP64_ONLY(__ movslq(j, j));
1998   __ restore_bcp();
1999   __ restore_locals();                           // restore rdi
2000   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2001 
2002   __ addptr(rsi, j);
2003   __ dispatch_only(vtos);
2004 
2005   // default case -> j = default offset
2006   __ bind(default_case);
2007   __ profile_switch_default(i);
2008   __ movl(j, Address(array, -2*wordSize));
2009   __ bswapl(j);
2010   LP64_ONLY(__ movslq(j, j));
2011   __ restore_bcp();
2012   __ restore_locals();                           // restore rdi
2013   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2014   __ addptr(rsi, j);
2015   __ dispatch_only(vtos);
2016 }
2017 
2018 
2019 void TemplateTable::_return(TosState state) {
2020   transition(state, state);
2021   assert(_desc->calls_vm(), "inconsistent calls_vm information"); // call in remove_activation
2022 
2023   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2024     assert(state == vtos, "only valid state");
2025     __ movptr(rax, aaddress(0));
2026     __ load_klass(rdi, rax);
2027     __ movl(rdi, Address(rdi, Klass::access_flags_offset()));
2028     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2029     Label skip_register_finalizer;
2030     __ jcc(Assembler::zero, skip_register_finalizer);
2031 
2032     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), rax);
2033 
2034     __ bind(skip_register_finalizer);
2035   }
2036 
2037   __ remove_activation(state, rsi);
2038   __ jmp(rsi);
2039 }
2040 
2041 
2042 // ----------------------------------------------------------------------------
2043 // Volatile variables demand their effects be made known to all CPU's in
2044 // order.  Store buffers on most chips allow reads & writes to reorder; the
2045 // JMM's ReadAfterWrite.java test fails in -Xint mode without some kind of
2046 // memory barrier (i.e., it's not sufficient that the interpreter does not
2047 // reorder volatile references, the hardware also must not reorder them).
2048 //
2049 // According to the new Java Memory Model (JMM):
2050 // (1) All volatiles are serialized wrt to each other.
2051 // ALSO reads & writes act as aquire & release, so:
2052 // (2) A read cannot let unrelated NON-volatile memory refs that happen after
2053 // the read float up to before the read.  It's OK for non-volatile memory refs
2054 // that happen before the volatile read to float down below it.
2055 // (3) Similar a volatile write cannot let unrelated NON-volatile memory refs
2056 // that happen BEFORE the write float down to after the write.  It's OK for
2057 // non-volatile memory refs that happen after the volatile write to float up
2058 // before it.
2059 //
2060 // We only put in barriers around volatile refs (they are expensive), not
2061 // _between_ memory refs (that would require us to track the flavor of the
2062 // previous memory refs).  Requirements (2) and (3) require some barriers
2063 // before volatile stores and after volatile loads.  These nearly cover
2064 // requirement (1) but miss the volatile-store-volatile-load case.  This final
2065 // case is placed after volatile-stores although it could just as well go
2066 // before volatile-loads.
2067 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits order_constraint ) {
2068   // Helper function to insert a is-volatile test and memory barrier
2069   if( !os::is_MP() ) return;    // Not needed on single CPU
2070   __ membar(order_constraint);
2071 }
2072 
2073 void TemplateTable::resolve_cache_and_index(int byte_no,
2074                                             Register Rcache,
2075                                             Register index,
2076                                             size_t index_size) {
2077   const Register temp = rbx;
2078   assert_different_registers(Rcache, index, temp);
2079 
2080   Label resolved;
2081     assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
2082     __ get_cache_and_index_and_bytecode_at_bcp(Rcache, index, temp, byte_no, 1, index_size);
2083     __ cmpl(temp, (int) bytecode());  // have we resolved this bytecode?
2084     __ jcc(Assembler::equal, resolved);
2085 
2086   // resolve first time through
2087   address entry;
2088   switch (bytecode()) {
2089     case Bytecodes::_getstatic      : // fall through
2090     case Bytecodes::_putstatic      : // fall through
2091     case Bytecodes::_getfield       : // fall through
2092     case Bytecodes::_putfield       : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);        break;
2093     case Bytecodes::_invokevirtual  : // fall through
2094     case Bytecodes::_invokespecial  : // fall through
2095     case Bytecodes::_invokestatic   : // fall through
2096     case Bytecodes::_invokeinterface: entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);         break;
2097     case Bytecodes::_invokehandle   : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokehandle);   break;
2098     case Bytecodes::_invokedynamic  : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);  break;
2099     default:
2100       fatal(err_msg("unexpected bytecode: %s", Bytecodes::name(bytecode())));
2101       break;
2102   }
2103   __ movl(temp, (int)bytecode());
2104   __ call_VM(noreg, entry, temp);
2105   // Update registers with resolved info
2106   __ get_cache_and_index_at_bcp(Rcache, index, 1, index_size);
2107   __ bind(resolved);
2108 }
2109 
2110 
2111 // The cache and index registers must be set before call
2112 void TemplateTable::load_field_cp_cache_entry(Register obj,
2113                                               Register cache,
2114                                               Register index,
2115                                               Register off,
2116                                               Register flags,
2117                                               bool is_static = false) {
2118   assert_different_registers(cache, index, flags, off);
2119 
2120   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2121   // Field offset
2122   __ movptr(off, Address(cache, index, Address::times_ptr,
2123                          in_bytes(cp_base_offset + ConstantPoolCacheEntry::f2_offset())));
2124   // Flags
2125   __ movl(flags, Address(cache, index, Address::times_ptr,
2126            in_bytes(cp_base_offset + ConstantPoolCacheEntry::flags_offset())));
2127 
2128   // klass overwrite register
2129   if (is_static) {
2130     __ movptr(obj, Address(cache, index, Address::times_ptr,
2131                            in_bytes(cp_base_offset + ConstantPoolCacheEntry::f1_offset())));
2132     const int mirror_offset = in_bytes(Klass::java_mirror_offset());
2133     __ movptr(obj, Address(obj, mirror_offset));
2134   }
2135 }
2136 
2137 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2138                                                Register method,
2139                                                Register itable_index,
2140                                                Register flags,
2141                                                bool is_invokevirtual,
2142                                                bool is_invokevfinal, /*unused*/
2143                                                bool is_invokedynamic) {
2144   // setup registers
2145   const Register cache = rcx;
2146   const Register index = rdx;
2147   assert_different_registers(method, flags);
2148   assert_different_registers(method, cache, index);
2149   assert_different_registers(itable_index, flags);
2150   assert_different_registers(itable_index, cache, index);
2151   // determine constant pool cache field offsets
2152   assert(is_invokevirtual == (byte_no == f2_byte), "is_invokevirtual flag redundant");
2153   const int method_offset = in_bytes(
2154     ConstantPoolCache::base_offset() +
2155       ((byte_no == f2_byte)
2156        ? ConstantPoolCacheEntry::f2_offset()
2157        : ConstantPoolCacheEntry::f1_offset()));
2158   const int flags_offset = in_bytes(ConstantPoolCache::base_offset() +
2159                                     ConstantPoolCacheEntry::flags_offset());
2160   // access constant pool cache fields
2161   const int index_offset = in_bytes(ConstantPoolCache::base_offset() +
2162                                     ConstantPoolCacheEntry::f2_offset());
2163 
2164   size_t index_size = (is_invokedynamic ? sizeof(u4) : sizeof(u2));
2165   resolve_cache_and_index(byte_no, cache, index, index_size);
2166     __ movptr(method, Address(cache, index, Address::times_ptr, method_offset));
2167 
2168   if (itable_index != noreg) {
2169     __ movptr(itable_index, Address(cache, index, Address::times_ptr, index_offset));
2170   }
2171   __ movl(flags, Address(cache, index, Address::times_ptr, flags_offset));
2172 }
2173 
2174 
2175 // The registers cache and index expected to be set before call.
2176 // Correct values of the cache and index registers are preserved.
2177 void TemplateTable::jvmti_post_field_access(Register cache,
2178                                             Register index,
2179                                             bool is_static,
2180                                             bool has_tos) {
2181   if (JvmtiExport::can_post_field_access()) {
2182     // Check to see if a field access watch has been set before we take
2183     // the time to call into the VM.
2184     Label L1;
2185     assert_different_registers(cache, index, rax);
2186     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2187     __ testl(rax,rax);
2188     __ jcc(Assembler::zero, L1);
2189 
2190     // cache entry pointer
2191     __ addptr(cache, in_bytes(ConstantPoolCache::base_offset()));
2192     __ shll(index, LogBytesPerWord);
2193     __ addptr(cache, index);
2194     if (is_static) {
2195       __ xorptr(rax, rax);      // NULL object reference
2196     } else {
2197       __ pop(atos);         // Get the object
2198       __ verify_oop(rax);
2199       __ push(atos);        // Restore stack state
2200     }
2201     // rax,:   object pointer or NULL
2202     // cache: cache entry pointer
2203     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access),
2204                rax, cache);
2205     __ get_cache_and_index_at_bcp(cache, index, 1);
2206     __ bind(L1);
2207   }
2208 }
2209 
2210 void TemplateTable::pop_and_check_object(Register r) {
2211   __ pop_ptr(r);
2212   __ null_check(r);  // for field access must check obj.
2213   __ verify_oop(r);
2214 }
2215 
2216 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2217   transition(vtos, vtos);
2218 
2219   const Register cache = rcx;
2220   const Register index = rdx;
2221   const Register obj   = rcx;
2222   const Register off   = rbx;
2223   const Register flags = rax;
2224 
2225   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2226   jvmti_post_field_access(cache, index, is_static, false);
2227   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2228 
2229   if (!is_static) pop_and_check_object(obj);
2230 
2231   const Address lo(obj, off, Address::times_1, 0*wordSize);
2232   const Address hi(obj, off, Address::times_1, 1*wordSize);
2233 
2234   Label Done, notByte, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2235 
2236   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2237   assert(btos == 0, "change code, btos != 0");
2238   // btos
2239   __ andptr(flags, ConstantPoolCacheEntry::tos_state_mask);
2240   __ jcc(Assembler::notZero, notByte);
2241 
2242   __ load_signed_byte(rax, lo );
2243   __ push(btos);
2244   // Rewrite bytecode to be faster
2245   if (!is_static) {
2246     patch_bytecode(Bytecodes::_fast_bgetfield, rcx, rbx);
2247   }
2248   __ jmp(Done);
2249 
2250   __ bind(notByte);
2251   // itos
2252   __ cmpl(flags, itos );
2253   __ jcc(Assembler::notEqual, notInt);
2254 
2255   __ movl(rax, lo );
2256   __ push(itos);
2257   // Rewrite bytecode to be faster
2258   if (!is_static) {
2259     patch_bytecode(Bytecodes::_fast_igetfield, rcx, rbx);
2260   }
2261   __ jmp(Done);
2262 
2263   __ bind(notInt);
2264   // atos
2265   __ cmpl(flags, atos );
2266   __ jcc(Assembler::notEqual, notObj);
2267 
2268   __ movl(rax, lo );
2269   __ push(atos);
2270   if (!is_static) {
2271     patch_bytecode(Bytecodes::_fast_agetfield, rcx, rbx);
2272   }
2273   __ jmp(Done);
2274 
2275   __ bind(notObj);
2276   // ctos
2277   __ cmpl(flags, ctos );
2278   __ jcc(Assembler::notEqual, notChar);
2279 
2280   __ load_unsigned_short(rax, lo );
2281   __ push(ctos);
2282   if (!is_static) {
2283     patch_bytecode(Bytecodes::_fast_cgetfield, rcx, rbx);
2284   }
2285   __ jmp(Done);
2286 
2287   __ bind(notChar);
2288   // stos
2289   __ cmpl(flags, stos );
2290   __ jcc(Assembler::notEqual, notShort);
2291 
2292   __ load_signed_short(rax, lo );
2293   __ push(stos);
2294   if (!is_static) {
2295     patch_bytecode(Bytecodes::_fast_sgetfield, rcx, rbx);
2296   }
2297   __ jmp(Done);
2298 
2299   __ bind(notShort);
2300   // ltos
2301   __ cmpl(flags, ltos );
2302   __ jcc(Assembler::notEqual, notLong);
2303 
2304   // Generate code as if volatile.  There just aren't enough registers to
2305   // save that information and this code is faster than the test.
2306   __ fild_d(lo);                // Must load atomically
2307   __ subptr(rsp,2*wordSize);    // Make space for store
2308   __ fistp_d(Address(rsp,0));
2309   __ pop(rax);
2310   __ pop(rdx);
2311 
2312   __ push(ltos);
2313   // Don't rewrite to _fast_lgetfield for potential volatile case.
2314   __ jmp(Done);
2315 
2316   __ bind(notLong);
2317   // ftos
2318   __ cmpl(flags, ftos );
2319   __ jcc(Assembler::notEqual, notFloat);
2320 
2321   __ fld_s(lo);
2322   __ push(ftos);
2323   if (!is_static) {
2324     patch_bytecode(Bytecodes::_fast_fgetfield, rcx, rbx);
2325   }
2326   __ jmp(Done);
2327 
2328   __ bind(notFloat);
2329   // dtos
2330   __ cmpl(flags, dtos );
2331   __ jcc(Assembler::notEqual, notDouble);
2332 
2333   __ fld_d(lo);
2334   __ push(dtos);
2335   if (!is_static) {
2336     patch_bytecode(Bytecodes::_fast_dgetfield, rcx, rbx);
2337   }
2338   __ jmpb(Done);
2339 
2340   __ bind(notDouble);
2341 
2342   __ stop("Bad state");
2343 
2344   __ bind(Done);
2345   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2346   // volatile_barrier( );
2347 }
2348 
2349 
2350 void TemplateTable::getfield(int byte_no) {
2351   getfield_or_static(byte_no, false);
2352 }
2353 
2354 
2355 void TemplateTable::getstatic(int byte_no) {
2356   getfield_or_static(byte_no, true);
2357 }
2358 
2359 // The registers cache and index expected to be set before call.
2360 // The function may destroy various registers, just not the cache and index registers.
2361 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2362 
2363   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2364 
2365   if (JvmtiExport::can_post_field_modification()) {
2366     // Check to see if a field modification watch has been set before we take
2367     // the time to call into the VM.
2368     Label L1;
2369     assert_different_registers(cache, index, rax);
2370     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2371     __ testl(rax, rax);
2372     __ jcc(Assembler::zero, L1);
2373 
2374     // The cache and index registers have been already set.
2375     // This allows to eliminate this call but the cache and index
2376     // registers have to be correspondingly used after this line.
2377     __ get_cache_and_index_at_bcp(rax, rdx, 1);
2378 
2379     if (is_static) {
2380       // Life is simple.  Null out the object pointer.
2381       __ xorptr(rbx, rbx);
2382     } else {
2383       // Life is harder. The stack holds the value on top, followed by the object.
2384       // We don't know the size of the value, though; it could be one or two words
2385       // depending on its type. As a result, we must find the type to determine where
2386       // the object is.
2387       Label two_word, valsize_known;
2388       __ movl(rcx, Address(rax, rdx, Address::times_ptr, in_bytes(cp_base_offset +
2389                                    ConstantPoolCacheEntry::flags_offset())));
2390       __ mov(rbx, rsp);
2391       __ shrl(rcx, ConstantPoolCacheEntry::tos_state_shift);
2392       // Make sure we don't need to mask rcx after the above shift
2393       ConstantPoolCacheEntry::verify_tos_state_shift();
2394       __ cmpl(rcx, ltos);
2395       __ jccb(Assembler::equal, two_word);
2396       __ cmpl(rcx, dtos);
2397       __ jccb(Assembler::equal, two_word);
2398       __ addptr(rbx, Interpreter::expr_offset_in_bytes(1)); // one word jvalue (not ltos, dtos)
2399       __ jmpb(valsize_known);
2400 
2401       __ bind(two_word);
2402       __ addptr(rbx, Interpreter::expr_offset_in_bytes(2)); // two words jvalue
2403 
2404       __ bind(valsize_known);
2405       // setup object pointer
2406       __ movptr(rbx, Address(rbx, 0));
2407     }
2408     // cache entry pointer
2409     __ addptr(rax, in_bytes(cp_base_offset));
2410     __ shll(rdx, LogBytesPerWord);
2411     __ addptr(rax, rdx);
2412     // object (tos)
2413     __ mov(rcx, rsp);
2414     // rbx,: object pointer set up above (NULL if static)
2415     // rax,: cache entry pointer
2416     // rcx: jvalue object on the stack
2417     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification),
2418                rbx, rax, rcx);
2419     __ get_cache_and_index_at_bcp(cache, index, 1);
2420     __ bind(L1);
2421   }
2422 }
2423 
2424 
2425 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2426   transition(vtos, vtos);
2427 
2428   const Register cache = rcx;
2429   const Register index = rdx;
2430   const Register obj   = rcx;
2431   const Register off   = rbx;
2432   const Register flags = rax;
2433 
2434   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2435   jvmti_post_field_mod(cache, index, is_static);
2436   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2437 
2438   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2439   // volatile_barrier( );
2440 
2441   Label notVolatile, Done;
2442   __ movl(rdx, flags);
2443   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2444   __ andl(rdx, 0x1);
2445 
2446   // field addresses
2447   const Address lo(obj, off, Address::times_1, 0*wordSize);
2448   const Address hi(obj, off, Address::times_1, 1*wordSize);
2449 
2450   Label notByte, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2451 
2452   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2453   assert(btos == 0, "change code, btos != 0");
2454   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2455   __ jcc(Assembler::notZero, notByte);
2456 
2457   // btos
2458   {
2459     __ pop(btos);
2460     if (!is_static) pop_and_check_object(obj);
2461     __ movb(lo, rax);
2462     if (!is_static) {
2463       patch_bytecode(Bytecodes::_fast_bputfield, rcx, rbx, true, byte_no);
2464     }
2465     __ jmp(Done);
2466   }
2467 
2468   __ bind(notByte);
2469   __ cmpl(flags, itos);
2470   __ jcc(Assembler::notEqual, notInt);
2471 
2472   // itos
2473   {
2474     __ pop(itos);
2475     if (!is_static) pop_and_check_object(obj);
2476     __ movl(lo, rax);
2477     if (!is_static) {
2478       patch_bytecode(Bytecodes::_fast_iputfield, rcx, rbx, true, byte_no);
2479     }
2480     __ jmp(Done);
2481   }
2482 
2483   __ bind(notInt);
2484   __ cmpl(flags, atos);
2485   __ jcc(Assembler::notEqual, notObj);
2486 
2487   // atos
2488   {
2489     __ pop(atos);
2490     if (!is_static) pop_and_check_object(obj);
2491     do_oop_store(_masm, lo, rax, _bs->kind(), false);
2492     if (!is_static) {
2493       patch_bytecode(Bytecodes::_fast_aputfield, rcx, rbx, true, byte_no);
2494     }
2495     __ jmp(Done);
2496   }
2497 
2498   __ bind(notObj);
2499   __ cmpl(flags, ctos);
2500   __ jcc(Assembler::notEqual, notChar);
2501 
2502   // ctos
2503   {
2504     __ pop(ctos);
2505     if (!is_static) pop_and_check_object(obj);
2506     __ movw(lo, rax);
2507     if (!is_static) {
2508       patch_bytecode(Bytecodes::_fast_cputfield, rcx, rbx, true, byte_no);
2509     }
2510     __ jmp(Done);
2511   }
2512 
2513   __ bind(notChar);
2514   __ cmpl(flags, stos);
2515   __ jcc(Assembler::notEqual, notShort);
2516 
2517   // stos
2518   {
2519     __ pop(stos);
2520     if (!is_static) pop_and_check_object(obj);
2521     __ movw(lo, rax);
2522     if (!is_static) {
2523       patch_bytecode(Bytecodes::_fast_sputfield, rcx, rbx, true, byte_no);
2524     }
2525     __ jmp(Done);
2526   }
2527 
2528   __ bind(notShort);
2529   __ cmpl(flags, ltos);
2530   __ jcc(Assembler::notEqual, notLong);
2531 
2532   // ltos
2533   {
2534     Label notVolatileLong;
2535     __ testl(rdx, rdx);
2536     __ jcc(Assembler::zero, notVolatileLong);
2537 
2538     __ pop(ltos);  // overwrites rdx, do this after testing volatile.
2539     if (!is_static) pop_and_check_object(obj);
2540 
2541     // Replace with real volatile test
2542     __ push(rdx);
2543     __ push(rax);                 // Must update atomically with FIST
2544     __ fild_d(Address(rsp,0));    // So load into FPU register
2545     __ fistp_d(lo);               // and put into memory atomically
2546     __ addptr(rsp, 2*wordSize);
2547     // volatile_barrier();
2548     volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2549                                                  Assembler::StoreStore));
2550     // Don't rewrite volatile version
2551     __ jmp(notVolatile);
2552 
2553     __ bind(notVolatileLong);
2554 
2555     __ pop(ltos);  // overwrites rdx
2556     if (!is_static) pop_and_check_object(obj);
2557     NOT_LP64(__ movptr(hi, rdx));
2558     __ movptr(lo, rax);
2559     if (!is_static) {
2560       patch_bytecode(Bytecodes::_fast_lputfield, rcx, rbx, true, byte_no);
2561     }
2562     __ jmp(notVolatile);
2563   }
2564 
2565   __ bind(notLong);
2566   __ cmpl(flags, ftos);
2567   __ jcc(Assembler::notEqual, notFloat);
2568 
2569   // ftos
2570   {
2571     __ pop(ftos);
2572     if (!is_static) pop_and_check_object(obj);
2573     __ fstp_s(lo);
2574     if (!is_static) {
2575       patch_bytecode(Bytecodes::_fast_fputfield, rcx, rbx, true, byte_no);
2576     }
2577     __ jmp(Done);
2578   }
2579 
2580   __ bind(notFloat);
2581 #ifdef ASSERT
2582   __ cmpl(flags, dtos);
2583   __ jcc(Assembler::notEqual, notDouble);
2584 #endif
2585 
2586   // dtos
2587   {
2588     __ pop(dtos);
2589     if (!is_static) pop_and_check_object(obj);
2590     __ fstp_d(lo);
2591     if (!is_static) {
2592       patch_bytecode(Bytecodes::_fast_dputfield, rcx, rbx, true, byte_no);
2593     }
2594     __ jmp(Done);
2595   }
2596 
2597 #ifdef ASSERT
2598   __ bind(notDouble);
2599   __ stop("Bad state");
2600 #endif
2601 
2602   __ bind(Done);
2603 
2604   // Check for volatile store
2605   __ testl(rdx, rdx);
2606   __ jcc(Assembler::zero, notVolatile);
2607   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2608                                                Assembler::StoreStore));
2609   __ bind(notVolatile);
2610 }
2611 
2612 
2613 void TemplateTable::putfield(int byte_no) {
2614   putfield_or_static(byte_no, false);
2615 }
2616 
2617 
2618 void TemplateTable::putstatic(int byte_no) {
2619   putfield_or_static(byte_no, true);
2620 }
2621 
2622 void TemplateTable::jvmti_post_fast_field_mod() {
2623   if (JvmtiExport::can_post_field_modification()) {
2624     // Check to see if a field modification watch has been set before we take
2625     // the time to call into the VM.
2626     Label L2;
2627      __ mov32(rcx, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2628      __ testl(rcx,rcx);
2629      __ jcc(Assembler::zero, L2);
2630      __ pop_ptr(rbx);               // copy the object pointer from tos
2631      __ verify_oop(rbx);
2632      __ push_ptr(rbx);              // put the object pointer back on tos
2633 
2634      // Save tos values before call_VM() clobbers them. Since we have
2635      // to do it for every data type, we use the saved values as the
2636      // jvalue object.
2637      switch (bytecode()) {          // load values into the jvalue object
2638      case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
2639      case Bytecodes::_fast_bputfield: // fall through
2640      case Bytecodes::_fast_sputfield: // fall through
2641      case Bytecodes::_fast_cputfield: // fall through
2642      case Bytecodes::_fast_iputfield: __ push_i(rax); break;
2643      case Bytecodes::_fast_dputfield: __ push_d(); break;
2644      case Bytecodes::_fast_fputfield: __ push_f(); break;
2645      case Bytecodes::_fast_lputfield: __ push_l(rax); break;
2646 
2647      default:
2648        ShouldNotReachHere();
2649      }
2650      __ mov(rcx, rsp);              // points to jvalue on the stack
2651      // access constant pool cache entry
2652      __ get_cache_entry_pointer_at_bcp(rax, rdx, 1);
2653      __ verify_oop(rbx);
2654      // rbx,: object pointer copied above
2655      // rax,: cache entry pointer
2656      // rcx: jvalue object on the stack
2657      __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification), rbx, rax, rcx);
2658 
2659      switch (bytecode()) {             // restore tos values
2660      case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
2661      case Bytecodes::_fast_bputfield: // fall through
2662      case Bytecodes::_fast_sputfield: // fall through
2663      case Bytecodes::_fast_cputfield: // fall through
2664      case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
2665      case Bytecodes::_fast_dputfield: __ pop_d(); break;
2666      case Bytecodes::_fast_fputfield: __ pop_f(); break;
2667      case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
2668      }
2669      __ bind(L2);
2670   }
2671 }
2672 
2673 void TemplateTable::fast_storefield(TosState state) {
2674   transition(state, vtos);
2675 
2676   ByteSize base = ConstantPoolCache::base_offset();
2677 
2678   jvmti_post_fast_field_mod();
2679 
2680   // access constant pool cache
2681   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2682 
2683   // test for volatile with rdx but rdx is tos register for lputfield.
2684   if (bytecode() == Bytecodes::_fast_lputfield) __ push(rdx);
2685   __ movl(rdx, Address(rcx, rbx, Address::times_ptr, in_bytes(base +
2686                        ConstantPoolCacheEntry::flags_offset())));
2687 
2688   // replace index with field offset from cache entry
2689   __ movptr(rbx, Address(rcx, rbx, Address::times_ptr, in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2690 
2691   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2692   // volatile_barrier( );
2693 
2694   Label notVolatile, Done;
2695   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2696   __ andl(rdx, 0x1);
2697   // Check for volatile store
2698   __ testl(rdx, rdx);
2699   __ jcc(Assembler::zero, notVolatile);
2700 
2701   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2702 
2703   // Get object from stack
2704   pop_and_check_object(rcx);
2705 
2706   // field addresses
2707   const Address lo(rcx, rbx, Address::times_1, 0*wordSize);
2708   const Address hi(rcx, rbx, Address::times_1, 1*wordSize);
2709 
2710   // access field
2711   switch (bytecode()) {
2712     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2713     case Bytecodes::_fast_sputfield: // fall through
2714     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2715     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2716     case Bytecodes::_fast_lputfield:
2717       NOT_LP64(__ movptr(hi, rdx));
2718       __ movptr(lo, rax);
2719       break;
2720     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2721     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2722     case Bytecodes::_fast_aputfield: {
2723       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2724       break;
2725     }
2726     default:
2727       ShouldNotReachHere();
2728   }
2729 
2730   Label done;
2731   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2732                                                Assembler::StoreStore));
2733   // Barriers are so large that short branch doesn't reach!
2734   __ jmp(done);
2735 
2736   // Same code as above, but don't need rdx to test for volatile.
2737   __ bind(notVolatile);
2738 
2739   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2740 
2741   // Get object from stack
2742   pop_and_check_object(rcx);
2743 
2744   // access field
2745   switch (bytecode()) {
2746     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2747     case Bytecodes::_fast_sputfield: // fall through
2748     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2749     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2750     case Bytecodes::_fast_lputfield:
2751       NOT_LP64(__ movptr(hi, rdx));
2752       __ movptr(lo, rax);
2753       break;
2754     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2755     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2756     case Bytecodes::_fast_aputfield: {
2757       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2758       break;
2759     }
2760     default:
2761       ShouldNotReachHere();
2762   }
2763   __ bind(done);
2764 }
2765 
2766 
2767 void TemplateTable::fast_accessfield(TosState state) {
2768   transition(atos, state);
2769 
2770   // do the JVMTI work here to avoid disturbing the register state below
2771   if (JvmtiExport::can_post_field_access()) {
2772     // Check to see if a field access watch has been set before we take
2773     // the time to call into the VM.
2774     Label L1;
2775     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2776     __ testl(rcx,rcx);
2777     __ jcc(Assembler::zero, L1);
2778     // access constant pool cache entry
2779     __ get_cache_entry_pointer_at_bcp(rcx, rdx, 1);
2780     __ push_ptr(rax);  // save object pointer before call_VM() clobbers it
2781     __ verify_oop(rax);
2782     // rax,: object pointer copied above
2783     // rcx: cache entry pointer
2784     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access), rax, rcx);
2785     __ pop_ptr(rax);   // restore object pointer
2786     __ bind(L1);
2787   }
2788 
2789   // access constant pool cache
2790   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2791   // replace index with field offset from cache entry
2792   __ movptr(rbx, Address(rcx,
2793                          rbx,
2794                          Address::times_ptr,
2795                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2796 
2797 
2798   // rax,: object
2799   __ verify_oop(rax);
2800   __ null_check(rax);
2801   // field addresses
2802   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2803   const Address hi = Address(rax, rbx, Address::times_1, 1*wordSize);
2804 
2805   // access field
2806   switch (bytecode()) {
2807     case Bytecodes::_fast_bgetfield: __ movsbl(rax, lo );                 break;
2808     case Bytecodes::_fast_sgetfield: __ load_signed_short(rax, lo );      break;
2809     case Bytecodes::_fast_cgetfield: __ load_unsigned_short(rax, lo );    break;
2810     case Bytecodes::_fast_igetfield: __ movl(rax, lo);                    break;
2811     case Bytecodes::_fast_lgetfield: __ stop("should not be rewritten");  break;
2812     case Bytecodes::_fast_fgetfield: __ fld_s(lo);                        break;
2813     case Bytecodes::_fast_dgetfield: __ fld_d(lo);                        break;
2814     case Bytecodes::_fast_agetfield: __ movptr(rax, lo); __ verify_oop(rax); break;
2815     default:
2816       ShouldNotReachHere();
2817   }
2818 
2819   // Doug Lea believes this is not needed with current Sparcs(TSO) and Intel(PSO)
2820   // volatile_barrier( );
2821 }
2822 
2823 void TemplateTable::fast_xaccess(TosState state) {
2824   transition(vtos, state);
2825   // get receiver
2826   __ movptr(rax, aaddress(0));
2827   // access constant pool cache
2828   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2829   __ movptr(rbx, Address(rcx,
2830                          rdx,
2831                          Address::times_ptr,
2832                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2833   // make sure exception is reported in correct bcp range (getfield is next instruction)
2834   __ increment(rsi);
2835   __ null_check(rax);
2836   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2837   if (state == itos) {
2838     __ movl(rax, lo);
2839   } else if (state == atos) {
2840     __ movptr(rax, lo);
2841     __ verify_oop(rax);
2842   } else if (state == ftos) {
2843     __ fld_s(lo);
2844   } else {
2845     ShouldNotReachHere();
2846   }
2847   __ decrement(rsi);
2848 }
2849 
2850 
2851 
2852 //----------------------------------------------------------------------------------------------------
2853 // Calls
2854 
2855 void TemplateTable::count_calls(Register method, Register temp) {
2856   // implemented elsewhere
2857   ShouldNotReachHere();
2858 }
2859 
2860 
2861 void TemplateTable::prepare_invoke(int byte_no,
2862                                    Register method,  // linked method (or i-klass)
2863                                    Register index,   // itable index, MethodType, etc.
2864                                    Register recv,    // if caller wants to see it
2865                                    Register flags    // if caller wants to test it
2866                                    ) {
2867   // determine flags
2868   const Bytecodes::Code code = bytecode();
2869   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
2870   const bool is_invokedynamic    = code == Bytecodes::_invokedynamic;
2871   const bool is_invokehandle     = code == Bytecodes::_invokehandle;
2872   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
2873   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
2874   const bool load_receiver       = (recv  != noreg);
2875   const bool save_flags          = (flags != noreg);
2876   assert(load_receiver == (code != Bytecodes::_invokestatic && code != Bytecodes::_invokedynamic), "");
2877   assert(save_flags    == (is_invokeinterface || is_invokevirtual), "need flags for vfinal");
2878   assert(flags == noreg || flags == rdx, "");
2879   assert(recv  == noreg || recv  == rcx, "");
2880 
2881   // setup registers & access constant pool cache
2882   if (recv  == noreg)  recv  = rcx;
2883   if (flags == noreg)  flags = rdx;
2884   assert_different_registers(method, index, recv, flags);
2885 
2886   // save 'interpreter return address'
2887   __ save_bcp();
2888 
2889   load_invoke_cp_cache_entry(byte_no, method, index, flags, is_invokevirtual, false, is_invokedynamic);
2890 
2891   // maybe push appendix to arguments (just before return address)
2892   if (is_invokedynamic || is_invokehandle) {
2893     Label L_no_push;
2894     __ testl(flags, (1 << ConstantPoolCacheEntry::has_appendix_shift));
2895     __ jccb(Assembler::zero, L_no_push);
2896     // Push the appendix as a trailing parameter.
2897     // This must be done before we get the receiver,
2898     // since the parameter_size includes it.
2899     __ push(rbx);
2900     __ mov(rbx, index);
2901     assert(ConstantPoolCacheEntry::_indy_resolved_references_appendix_offset == 0, "appendix expected at index+0");
2902     __ load_resolved_reference_at_index(index, rbx);
2903     __ pop(rbx);
2904     __ push(index);  // push appendix (MethodType, CallSite, etc.)
2905     __ bind(L_no_push);
2906   }
2907 
2908   // load receiver if needed (note: no return address pushed yet)
2909   if (load_receiver) {
2910     __ movl(recv, flags);
2911     __ andl(recv, ConstantPoolCacheEntry::parameter_size_mask);
2912     const int no_return_pc_pushed_yet = -1;  // argument slot correction before we push return address
2913     const int receiver_is_at_end      = -1;  // back off one slot to get receiver
2914     Address recv_addr = __ argument_address(recv, no_return_pc_pushed_yet + receiver_is_at_end);
2915     __ movptr(recv, recv_addr);
2916     __ verify_oop(recv);
2917   }
2918 
2919   if (save_flags) {
2920     __ mov(rsi, flags);
2921   }
2922 
2923   // compute return type
2924   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2925   // Make sure we don't need to mask flags after the above shift
2926   ConstantPoolCacheEntry::verify_tos_state_shift();
2927   // load return address
2928   {
2929     const address table_addr = (address) Interpreter::invoke_return_entry_table_for(code);
2930     ExternalAddress table(table_addr);
2931     __ movptr(flags, ArrayAddress(table, Address(noreg, flags, Address::times_ptr)));
2932   }
2933 
2934   // push return address
2935   __ push(flags);
2936 
2937   // Restore flags value from the constant pool cache, and restore rsi
2938   // for later null checks.  rsi is the bytecode pointer
2939   if (save_flags) {
2940     __ mov(flags, rsi);
2941     __ restore_bcp();
2942   }
2943 }
2944 
2945 
2946 void TemplateTable::invokevirtual_helper(Register index,
2947                                          Register recv,
2948                                          Register flags) {
2949   // Uses temporary registers rax, rdx
2950   assert_different_registers(index, recv, rax, rdx);
2951   assert(index == rbx, "");
2952   assert(recv  == rcx, "");
2953 
2954   // Test for an invoke of a final method
2955   Label notFinal;
2956   __ movl(rax, flags);
2957   __ andl(rax, (1 << ConstantPoolCacheEntry::is_vfinal_shift));
2958   __ jcc(Assembler::zero, notFinal);
2959 
2960   const Register method = index;  // method must be rbx
2961   assert(method == rbx,
2962          "Method* must be rbx for interpreter calling convention");
2963 
2964   // do the call - the index is actually the method to call
2965   // that is, f2 is a vtable index if !is_vfinal, else f2 is a Method*
2966 
2967   // It's final, need a null check here!
2968   __ null_check(recv);
2969 
2970   // profile this call
2971   __ profile_final_call(rax);
2972   __ profile_arguments_type(rax, method, rsi, true);
2973 
2974   __ jump_from_interpreted(method, rax);
2975 
2976   __ bind(notFinal);
2977 
2978   // get receiver klass
2979   __ null_check(recv, oopDesc::klass_offset_in_bytes());
2980   __ load_klass(rax, recv);
2981 
2982   // profile this call
2983   __ profile_virtual_call(rax, rdi, rdx);
2984 
2985   // get target Method* & entry point
2986   __ lookup_virtual_method(rax, index, method);
2987   __ profile_arguments_type(rdx, method, rsi, true);
2988   __ jump_from_interpreted(method, rdx);
2989 }
2990 
2991 
2992 void TemplateTable::invokevirtual(int byte_no) {
2993   transition(vtos, vtos);
2994   assert(byte_no == f2_byte, "use this argument");
2995   prepare_invoke(byte_no,
2996                  rbx,    // method or vtable index
2997                  noreg,  // unused itable index
2998                  rcx, rdx); // recv, flags
2999 
3000   // rbx: index
3001   // rcx: receiver
3002   // rdx: flags
3003 
3004   invokevirtual_helper(rbx, rcx, rdx);
3005 }
3006 
3007 
3008 void TemplateTable::invokespecial(int byte_no) {
3009   transition(vtos, vtos);
3010   assert(byte_no == f1_byte, "use this argument");
3011   prepare_invoke(byte_no, rbx, noreg,  // get f1 Method*
3012                  rcx);  // get receiver also for null check
3013   __ verify_oop(rcx);
3014   __ null_check(rcx);
3015   // do the call
3016   __ profile_call(rax);
3017   __ profile_arguments_type(rax, rbx, rsi, false);
3018   __ jump_from_interpreted(rbx, rax);
3019 }
3020 
3021 
3022 void TemplateTable::invokestatic(int byte_no) {
3023   transition(vtos, vtos);
3024   assert(byte_no == f1_byte, "use this argument");
3025   prepare_invoke(byte_no, rbx);  // get f1 Method*
3026   // do the call
3027   __ profile_call(rax);
3028   __ profile_arguments_type(rax, rbx, rsi, false);
3029   __ jump_from_interpreted(rbx, rax);
3030 }
3031 
3032 
3033 void TemplateTable::fast_invokevfinal(int byte_no) {
3034   transition(vtos, vtos);
3035   assert(byte_no == f2_byte, "use this argument");
3036   __ stop("fast_invokevfinal not used on x86");
3037 }
3038 
3039 
3040 void TemplateTable::invokeinterface(int byte_no) {
3041   transition(vtos, vtos);
3042   assert(byte_no == f1_byte, "use this argument");
3043   prepare_invoke(byte_no, rax, rbx,  // get f1 Klass*, f2 itable index
3044                  rcx, rdx); // recv, flags
3045 
3046   // rax: interface klass (from f1)
3047   // rbx: itable index (from f2)
3048   // rcx: receiver
3049   // rdx: flags
3050 
3051   // Special case of invokeinterface called for virtual method of
3052   // java.lang.Object.  See cpCacheOop.cpp for details.
3053   // This code isn't produced by javac, but could be produced by
3054   // another compliant java compiler.
3055   Label notMethod;
3056   __ movl(rdi, rdx);
3057   __ andl(rdi, (1 << ConstantPoolCacheEntry::is_forced_virtual_shift));
3058   __ jcc(Assembler::zero, notMethod);
3059 
3060   invokevirtual_helper(rbx, rcx, rdx);
3061   __ bind(notMethod);
3062 
3063   // Get receiver klass into rdx - also a null check
3064   __ restore_locals();  // restore rdi
3065   __ null_check(rcx, oopDesc::klass_offset_in_bytes());
3066   __ load_klass(rdx, rcx);
3067 
3068   // profile this call
3069   __ profile_virtual_call(rdx, rsi, rdi);
3070 
3071   Label no_such_interface, no_such_method;
3072 
3073   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3074                              rdx, rax, rbx,
3075                              // outputs: method, scan temp. reg
3076                              rbx, rsi,
3077                              no_such_interface);
3078 
3079   // rbx: Method* to call
3080   // rcx: receiver
3081   // Check for abstract method error
3082   // Note: This should be done more efficiently via a throw_abstract_method_error
3083   //       interpreter entry point and a conditional jump to it in case of a null
3084   //       method.
3085   __ testptr(rbx, rbx);
3086   __ jcc(Assembler::zero, no_such_method);
3087 
3088   __ profile_arguments_type(rdx, rbx, rsi, true);
3089 
3090   // do the call
3091   // rcx: receiver
3092   // rbx,: Method*
3093   __ jump_from_interpreted(rbx, rdx);
3094   __ should_not_reach_here();
3095 
3096   // exception handling code follows...
3097   // note: must restore interpreter registers to canonical
3098   //       state for exception handling to work correctly!
3099 
3100   __ bind(no_such_method);
3101   // throw exception
3102   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3103   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3104   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3105   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3106   // the call_VM checks for exception, so we should never return here.
3107   __ should_not_reach_here();
3108 
3109   __ bind(no_such_interface);
3110   // throw exception
3111   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3112   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3113   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3114   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3115                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3116   // the call_VM checks for exception, so we should never return here.
3117   __ should_not_reach_here();
3118 }
3119 
3120 void TemplateTable::invokehandle(int byte_no) {
3121   transition(vtos, vtos);
3122   assert(byte_no == f1_byte, "use this argument");
3123   const Register rbx_method = rbx;
3124   const Register rax_mtype  = rax;
3125   const Register rcx_recv   = rcx;
3126   const Register rdx_flags  = rdx;
3127 
3128   prepare_invoke(byte_no, rbx_method, rax_mtype, rcx_recv);
3129   __ verify_method_ptr(rbx_method);
3130   __ verify_oop(rcx_recv);
3131   __ null_check(rcx_recv);
3132 
3133   // rax: MethodType object (from cpool->resolved_references[f1], if necessary)
3134   // rbx: MH.invokeExact_MT method (from f2)
3135 
3136   // Note:  rax_mtype is already pushed (if necessary) by prepare_invoke
3137 
3138   // FIXME: profile the LambdaForm also
3139   __ profile_final_call(rax);
3140   __ profile_arguments_type(rdx, rbx_method, rsi, true);
3141 
3142   __ jump_from_interpreted(rbx_method, rdx);
3143 }
3144 
3145 
3146 void TemplateTable::invokedynamic(int byte_no) {
3147   transition(vtos, vtos);
3148   assert(byte_no == f1_byte, "use this argument");
3149 
3150   const Register rbx_method   = rbx;
3151   const Register rax_callsite = rax;
3152 
3153   prepare_invoke(byte_no, rbx_method, rax_callsite);
3154 
3155   // rax: CallSite object (from cpool->resolved_references[f1])
3156   // rbx: MH.linkToCallSite method (from f2)
3157 
3158   // Note:  rax_callsite is already pushed by prepare_invoke
3159 
3160   // %%% should make a type profile for any invokedynamic that takes a ref argument
3161   // profile this call
3162   __ profile_call(rsi);
3163   __ profile_arguments_type(rdx, rbx, rsi, false);
3164 
3165   __ verify_oop(rax_callsite);
3166 
3167   __ jump_from_interpreted(rbx_method, rdx);
3168 }
3169 
3170 //----------------------------------------------------------------------------------------------------
3171 // Allocation
3172 
3173 void TemplateTable::_new() {
3174   transition(vtos, atos);
3175   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3176   Label slow_case;
3177   Label slow_case_no_pop;
3178   Label done;
3179   Label initialize_header;
3180   Label initialize_object;  // including clearing the fields
3181   Label allocate_shared;
3182 
3183   __ get_cpool_and_tags(rcx, rax);
3184 
3185   // Make sure the class we're about to instantiate has been resolved.
3186   // This is done before loading InstanceKlass to be consistent with the order
3187   // how Constant Pool is updated (see ConstantPool::klass_at_put)
3188   const int tags_offset = Array<u1>::base_offset_in_bytes();
3189   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset), JVM_CONSTANT_Class);
3190   __ jcc(Assembler::notEqual, slow_case_no_pop);
3191 
3192   // get InstanceKlass
3193   __ movptr(rcx, Address(rcx, rdx, Address::times_ptr, sizeof(ConstantPool)));
3194   __ push(rcx);  // save the contexts of klass for initializing the header
3195 
3196   // make sure klass is initialized & doesn't have finalizer
3197   // make sure klass is fully initialized
3198   __ cmpb(Address(rcx, InstanceKlass::init_state_offset()), InstanceKlass::fully_initialized);
3199   __ jcc(Assembler::notEqual, slow_case);
3200 
3201   // get instance_size in InstanceKlass (scaled to a count of bytes)
3202   __ movl(rdx, Address(rcx, Klass::layout_helper_offset()));
3203   // test to see if it has a finalizer or is malformed in some way
3204   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3205   __ jcc(Assembler::notZero, slow_case);
3206 
3207   //
3208   // Allocate the instance
3209   // 1) Try to allocate in the TLAB
3210   // 2) if fail and the object is large allocate in the shared Eden
3211   // 3) if the above fails (or is not applicable), go to a slow case
3212   // (creates a new TLAB, etc.)
3213 
3214   const bool allow_shared_alloc =
3215     Universe::heap()->supports_inline_contig_alloc();
3216 
3217   const Register thread = rcx;
3218   if (UseTLAB || allow_shared_alloc) {
3219     __ get_thread(thread);
3220   }
3221 
3222   if (UseTLAB) {
3223     __ movptr(rax, Address(thread, in_bytes(JavaThread::tlab_top_offset())));
3224     __ lea(rbx, Address(rax, rdx, Address::times_1));
3225     __ cmpptr(rbx, Address(thread, in_bytes(JavaThread::tlab_end_offset())));
3226     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3227     __ movptr(Address(thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3228     if (ZeroTLAB) {
3229       // the fields have been already cleared
3230       __ jmp(initialize_header);
3231     } else {
3232       // initialize both the header and fields
3233       __ jmp(initialize_object);
3234     }
3235   }
3236 
3237   // Allocation in the shared Eden, if allowed.
3238   //
3239   // rdx: instance size in bytes
3240   if (allow_shared_alloc) {
3241     __ bind(allocate_shared);
3242 
3243     ExternalAddress heap_top((address)Universe::heap()->top_addr());
3244 
3245     Label retry;
3246     __ bind(retry);
3247     __ movptr(rax, heap_top);
3248     __ lea(rbx, Address(rax, rdx, Address::times_1));
3249     __ cmpptr(rbx, ExternalAddress((address)Universe::heap()->end_addr()));
3250     __ jcc(Assembler::above, slow_case);
3251 
3252     // Compare rax, with the top addr, and if still equal, store the new
3253     // top addr in rbx, at the address of the top addr pointer. Sets ZF if was
3254     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3255     //
3256     // rax,: object begin
3257     // rbx,: object end
3258     // rdx: instance size in bytes
3259     __ locked_cmpxchgptr(rbx, heap_top);
3260 
3261     // if someone beat us on the allocation, try again, otherwise continue
3262     __ jcc(Assembler::notEqual, retry);
3263 
3264     __ incr_allocated_bytes(thread, rdx, 0);
3265   }
3266 
3267   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3268     // The object is initialized before the header.  If the object size is
3269     // zero, go directly to the header initialization.
3270     __ bind(initialize_object);
3271     __ decrement(rdx, sizeof(oopDesc));
3272     __ jcc(Assembler::zero, initialize_header);
3273 
3274     // Initialize topmost object field, divide rdx by 8, check if odd and
3275     // test if zero.
3276     __ xorl(rcx, rcx);    // use zero reg to clear memory (shorter code)
3277     __ shrl(rdx, LogBytesPerLong); // divide by 2*oopSize and set carry flag if odd
3278 
3279     // rdx must have been multiple of 8
3280 #ifdef ASSERT
3281     // make sure rdx was multiple of 8
3282     Label L;
3283     // Ignore partial flag stall after shrl() since it is debug VM
3284     __ jccb(Assembler::carryClear, L);
3285     __ stop("object size is not multiple of 2 - adjust this code");
3286     __ bind(L);
3287     // rdx must be > 0, no extra check needed here
3288 #endif
3289 
3290     // initialize remaining object fields: rdx was a multiple of 8
3291     { Label loop;
3292     __ bind(loop);
3293     __ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 1*oopSize), rcx);
3294     NOT_LP64(__ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 2*oopSize), rcx));
3295     __ decrement(rdx);
3296     __ jcc(Assembler::notZero, loop);
3297     }
3298 
3299     // initialize object header only.
3300     __ bind(initialize_header);
3301     if (UseBiasedLocking) {
3302       __ pop(rcx);   // get saved klass back in the register.
3303       __ movptr(rbx, Address(rcx, Klass::prototype_header_offset()));
3304       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()), rbx);
3305     } else {
3306       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()),
3307                 (int32_t)markOopDesc::prototype()); // header
3308       __ pop(rcx);   // get saved klass back in the register.
3309     }
3310     __ store_klass(rax, rcx);  // klass
3311 
3312     {
3313       SkipIfEqual skip_if(_masm, &DTraceAllocProbes, 0);
3314       // Trigger dtrace event for fastpath
3315       __ push(atos);
3316       __ call_VM_leaf(
3317            CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3318       __ pop(atos);
3319     }
3320 
3321     __ jmp(done);
3322   }
3323 
3324   // slow case
3325   __ bind(slow_case);
3326   __ pop(rcx);   // restore stack pointer to what it was when we came in.
3327   __ bind(slow_case_no_pop);
3328   __ get_constant_pool(rax);
3329   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3330   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), rax, rdx);
3331 
3332   // continue
3333   __ bind(done);
3334 }
3335 
3336 
3337 void TemplateTable::newarray() {
3338   transition(itos, atos);
3339   __ push_i(rax);                                 // make sure everything is on the stack
3340   __ load_unsigned_byte(rdx, at_bcp(1));
3341   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray), rdx, rax);
3342   __ pop_i(rdx);                                  // discard size
3343 }
3344 
3345 
3346 void TemplateTable::anewarray() {
3347   transition(itos, atos);
3348   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3349   __ get_constant_pool(rcx);
3350   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray), rcx, rdx, rax);
3351 }
3352 
3353 
3354 void TemplateTable::arraylength() {
3355   transition(atos, itos);
3356   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3357   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3358 }
3359 
3360 
3361 void TemplateTable::checkcast() {
3362   transition(atos, atos);
3363   Label done, is_null, ok_is_subtype, quicked, resolved;
3364   __ testptr(rax, rax);   // Object is in EAX
3365   __ jcc(Assembler::zero, is_null);
3366 
3367   // Get cpool & tags index
3368   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3369   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3370   // See if bytecode has already been quicked
3371   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3372   __ jcc(Assembler::equal, quicked);
3373 
3374   __ push(atos);
3375   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3376   // vm_result_2 has metadata result
3377   // borrow rdi from locals
3378   __ get_thread(rdi);
3379   __ get_vm_result_2(rax, rdi);
3380   __ restore_locals();
3381   __ pop_ptr(rdx);
3382   __ jmpb(resolved);
3383 
3384   // Get superklass in EAX and subklass in EBX
3385   __ bind(quicked);
3386   __ mov(rdx, rax);          // Save object in EDX; EAX needed for subtype check
3387   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3388 
3389   __ bind(resolved);
3390   __ load_klass(rbx, rdx);
3391 
3392   // Generate subtype check.  Blows ECX.  Resets EDI.  Object in EDX.
3393   // Superklass in EAX.  Subklass in EBX.
3394   __ gen_subtype_check( rbx, ok_is_subtype );
3395 
3396   // Come here on failure
3397   __ push(rdx);
3398   // object is at TOS
3399   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3400 
3401   // Come here on success
3402   __ bind(ok_is_subtype);
3403   __ mov(rax,rdx);           // Restore object in EDX
3404 
3405   // Collect counts on whether this check-cast sees NULLs a lot or not.
3406   if (ProfileInterpreter) {
3407     __ jmp(done);
3408     __ bind(is_null);
3409     __ profile_null_seen(rcx);
3410   } else {
3411     __ bind(is_null);   // same as 'done'
3412   }
3413   __ bind(done);
3414 }
3415 
3416 
3417 void TemplateTable::instanceof() {
3418   transition(atos, itos);
3419   Label done, is_null, ok_is_subtype, quicked, resolved;
3420   __ testptr(rax, rax);
3421   __ jcc(Assembler::zero, is_null);
3422 
3423   // Get cpool & tags index
3424   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3425   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3426   // See if bytecode has already been quicked
3427   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3428   __ jcc(Assembler::equal, quicked);
3429 
3430   __ push(atos);
3431   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3432   // vm_result_2 has metadata result
3433   // borrow rdi from locals
3434   __ get_thread(rdi);
3435   __ get_vm_result_2(rax, rdi);
3436   __ restore_locals();
3437   __ pop_ptr(rdx);
3438   __ load_klass(rdx, rdx);
3439   __ jmp(resolved);
3440 
3441   // Get superklass in EAX and subklass in EDX
3442   __ bind(quicked);
3443   __ load_klass(rdx, rax);
3444   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3445 
3446   __ bind(resolved);
3447 
3448   // Generate subtype check.  Blows ECX.  Resets EDI.
3449   // Superklass in EAX.  Subklass in EDX.
3450   __ gen_subtype_check( rdx, ok_is_subtype );
3451 
3452   // Come here on failure
3453   __ xorl(rax,rax);
3454   __ jmpb(done);
3455   // Come here on success
3456   __ bind(ok_is_subtype);
3457   __ movl(rax, 1);
3458 
3459   // Collect counts on whether this test sees NULLs a lot or not.
3460   if (ProfileInterpreter) {
3461     __ jmp(done);
3462     __ bind(is_null);
3463     __ profile_null_seen(rcx);
3464   } else {
3465     __ bind(is_null);   // same as 'done'
3466   }
3467   __ bind(done);
3468   // rax, = 0: obj == NULL or  obj is not an instanceof the specified klass
3469   // rax, = 1: obj != NULL and obj is     an instanceof the specified klass
3470 }
3471 
3472 
3473 //----------------------------------------------------------------------------------------------------
3474 // Breakpoints
3475 void TemplateTable::_breakpoint() {
3476 
3477   // Note: We get here even if we are single stepping..
3478   // jbug inists on setting breakpoints at every bytecode
3479   // even if we are in single step mode.
3480 
3481   transition(vtos, vtos);
3482 
3483   // get the unpatched byte code
3484   __ get_method(rcx);
3485   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::get_original_bytecode_at), rcx, rsi);
3486   __ mov(rbx, rax);
3487 
3488   // post the breakpoint event
3489   __ get_method(rcx);
3490   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint), rcx, rsi);
3491 
3492   // complete the execution of original bytecode
3493   __ dispatch_only_normal(vtos);
3494 }
3495 
3496 
3497 //----------------------------------------------------------------------------------------------------
3498 // Exceptions
3499 
3500 void TemplateTable::athrow() {
3501   transition(atos, vtos);
3502   __ null_check(rax);
3503   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3504 }
3505 
3506 
3507 //----------------------------------------------------------------------------------------------------
3508 // Synchronization
3509 //
3510 // Note: monitorenter & exit are symmetric routines; which is reflected
3511 //       in the assembly code structure as well
3512 //
3513 // Stack layout:
3514 //
3515 // [expressions  ] <--- rsp               = expression stack top
3516 // ..
3517 // [expressions  ]
3518 // [monitor entry] <--- monitor block top = expression stack bot
3519 // ..
3520 // [monitor entry]
3521 // [frame data   ] <--- monitor block bot
3522 // ...
3523 // [saved rbp,    ] <--- rbp,
3524 
3525 
3526 void TemplateTable::monitorenter() {
3527   transition(atos, vtos);
3528 
3529   // check for NULL object
3530   __ null_check(rax);
3531 
3532   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3533   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3534   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3535   Label allocated;
3536 
3537   // initialize entry pointer
3538   __ xorl(rdx, rdx);                             // points to free slot or NULL
3539 
3540   // find a free slot in the monitor block (result in rdx)
3541   { Label entry, loop, exit;
3542     __ movptr(rcx, monitor_block_top);           // points to current entry, starting with top-most entry
3543 
3544     __ lea(rbx, monitor_block_bot);              // points to word before bottom of monitor block
3545     __ jmpb(entry);
3546 
3547     __ bind(loop);
3548     __ cmpptr(Address(rcx, BasicObjectLock::obj_offset_in_bytes()), (int32_t)NULL_WORD);  // check if current entry is used
3549     __ cmovptr(Assembler::equal, rdx, rcx);      // if not used then remember entry in rdx
3550     __ cmpptr(rax, Address(rcx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3551     __ jccb(Assembler::equal, exit);             // if same object then stop searching
3552     __ addptr(rcx, entry_size);                  // otherwise advance to next entry
3553     __ bind(entry);
3554     __ cmpptr(rcx, rbx);                         // check if bottom reached
3555     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3556     __ bind(exit);
3557   }
3558 
3559   __ testptr(rdx, rdx);                          // check if a slot has been found
3560   __ jccb(Assembler::notZero, allocated);        // if found, continue with that one
3561 
3562   // allocate one if there's no free slot
3563   { Label entry, loop;
3564     // 1. compute new pointers                   // rsp: old expression stack top
3565     __ movptr(rdx, monitor_block_bot);           // rdx: old expression stack bottom
3566     __ subptr(rsp, entry_size);                  // move expression stack top
3567     __ subptr(rdx, entry_size);                  // move expression stack bottom
3568     __ mov(rcx, rsp);                            // set start value for copy loop
3569     __ movptr(monitor_block_bot, rdx);           // set new monitor block top
3570     __ jmp(entry);
3571     // 2. move expression stack contents
3572     __ bind(loop);
3573     __ movptr(rbx, Address(rcx, entry_size));    // load expression stack word from old location
3574     __ movptr(Address(rcx, 0), rbx);             // and store it at new location
3575     __ addptr(rcx, wordSize);                    // advance to next word
3576     __ bind(entry);
3577     __ cmpptr(rcx, rdx);                         // check if bottom reached
3578     __ jcc(Assembler::notEqual, loop);           // if not at bottom then copy next word
3579   }
3580 
3581   // call run-time routine
3582   // rdx: points to monitor entry
3583   __ bind(allocated);
3584 
3585   // Increment bcp to point to the next bytecode, so exception handling for async. exceptions work correctly.
3586   // The object has already been poped from the stack, so the expression stack looks correct.
3587   __ increment(rsi);
3588 
3589   __ movptr(Address(rdx, BasicObjectLock::obj_offset_in_bytes()), rax);     // store object
3590   __ lock_object(rdx);
3591 
3592   // check to make sure this monitor doesn't cause stack overflow after locking
3593   __ save_bcp();  // in case of exception
3594   __ generate_stack_overflow_check(0);
3595 
3596   // The bcp has already been incremented. Just need to dispatch to next instruction.
3597   __ dispatch_next(vtos);
3598 }
3599 
3600 
3601 void TemplateTable::monitorexit() {
3602   transition(atos, vtos);
3603 
3604   // check for NULL object
3605   __ null_check(rax);
3606 
3607   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3608   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3609   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3610   Label found;
3611 
3612   // find matching slot
3613   { Label entry, loop;
3614     __ movptr(rdx, monitor_block_top);           // points to current entry, starting with top-most entry
3615     __ lea(rbx, monitor_block_bot);             // points to word before bottom of monitor block
3616     __ jmpb(entry);
3617 
3618     __ bind(loop);
3619     __ cmpptr(rax, Address(rdx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3620     __ jcc(Assembler::equal, found);             // if same object then stop searching
3621     __ addptr(rdx, entry_size);                  // otherwise advance to next entry
3622     __ bind(entry);
3623     __ cmpptr(rdx, rbx);                         // check if bottom reached
3624     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3625   }
3626 
3627   // error handling. Unlocking was not block-structured
3628   Label end;
3629   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_illegal_monitor_state_exception));
3630   __ should_not_reach_here();
3631 
3632   // call run-time routine
3633   // rcx: points to monitor entry
3634   __ bind(found);
3635   __ push_ptr(rax);                                 // make sure object is on stack (contract with oopMaps)
3636   __ unlock_object(rdx);
3637   __ pop_ptr(rax);                                  // discard object
3638   __ bind(end);
3639 }
3640 
3641 
3642 //----------------------------------------------------------------------------------------------------
3643 // Wide instructions
3644 
3645 void TemplateTable::wide() {
3646   transition(vtos, vtos);
3647   __ load_unsigned_byte(rbx, at_bcp(1));
3648   ExternalAddress wtable((address)Interpreter::_wentry_point);
3649   __ jump(ArrayAddress(wtable, Address(noreg, rbx, Address::times_ptr)));
3650   // Note: the rsi increment step is part of the individual wide bytecode implementations
3651 }
3652 
3653 
3654 //----------------------------------------------------------------------------------------------------
3655 // Multi arrays
3656 
3657 void TemplateTable::multianewarray() {
3658   transition(vtos, atos);
3659   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3660   // last dim is on top of stack; we want address of first one:
3661   // first_addr = last_addr + (ndims - 1) * stackElementSize - 1*wordsize
3662   // the latter wordSize to point to the beginning of the array.
3663   __ lea(  rax, Address(rsp, rax, Interpreter::stackElementScale(), -wordSize));
3664   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray), rax);     // pass in rax,
3665   __ load_unsigned_byte(rbx, at_bcp(3));
3666   __ lea(rsp, Address(rsp, rbx, Interpreter::stackElementScale()));  // get rid of counts
3667 }
3668 
3669 #endif /* !CC_INTERP */