1 /*
   2  * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #include "precompiled.hpp"
  26 #include "asm/macroAssembler.hpp"
  27 #include "interpreter/interpreter.hpp"
  28 #include "interpreter/interpreterRuntime.hpp"
  29 #include "interpreter/templateTable.hpp"
  30 #include "memory/universe.inline.hpp"
  31 #include "oops/methodData.hpp"
  32 #include "oops/objArrayKlass.hpp"
  33 #include "oops/oop.inline.hpp"
  34 #include "prims/methodHandles.hpp"
  35 #include "runtime/sharedRuntime.hpp"
  36 #include "runtime/stubRoutines.hpp"
  37 #include "runtime/synchronizer.hpp"
  38 #include "utilities/macros.hpp"
  39 
  40 #ifndef CC_INTERP
  41 #define __ _masm->
  42 
  43 //----------------------------------------------------------------------------------------------------
  44 // Platform-dependent initialization
  45 
  46 void TemplateTable::pd_initialize() {
  47   // No i486 specific initialization
  48 }
  49 
  50 //----------------------------------------------------------------------------------------------------
  51 // Address computation
  52 
  53 // local variables
  54 static inline Address iaddress(int n)            {
  55   return Address(rdi, Interpreter::local_offset_in_bytes(n));
  56 }
  57 
  58 static inline Address laddress(int n)            { return iaddress(n + 1); }
  59 static inline Address haddress(int n)            { return iaddress(n + 0); }
  60 static inline Address faddress(int n)            { return iaddress(n); }
  61 static inline Address daddress(int n)            { return laddress(n); }
  62 static inline Address aaddress(int n)            { return iaddress(n); }
  63 
  64 static inline Address iaddress(Register r)       {
  65   return Address(rdi, r, Interpreter::stackElementScale());
  66 }
  67 static inline Address laddress(Register r)       {
  68   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(1));
  69 }
  70 static inline Address haddress(Register r)       {
  71   return Address(rdi, r, Interpreter::stackElementScale(), Interpreter::local_offset_in_bytes(0));
  72 }
  73 
  74 static inline Address faddress(Register r)       { return iaddress(r); }
  75 static inline Address daddress(Register r)       { return laddress(r); }
  76 static inline Address aaddress(Register r)       { return iaddress(r); }
  77 
  78 // expression stack
  79 // (Note: Must not use symmetric equivalents at_rsp_m1/2 since they store
  80 // data beyond the rsp which is potentially unsafe in an MT environment;
  81 // an interrupt may overwrite that data.)
  82 static inline Address at_rsp   () {
  83   return Address(rsp, 0);
  84 }
  85 
  86 // At top of Java expression stack which may be different than rsp().  It
  87 // isn't for category 1 objects.
  88 static inline Address at_tos   () {
  89   Address tos = Address(rsp,  Interpreter::expr_offset_in_bytes(0));
  90   return tos;
  91 }
  92 
  93 static inline Address at_tos_p1() {
  94   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
  95 }
  96 
  97 static inline Address at_tos_p2() {
  98   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
  99 }
 100 
 101 // Condition conversion
 102 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 103   switch (cc) {
 104     case TemplateTable::equal        : return Assembler::notEqual;
 105     case TemplateTable::not_equal    : return Assembler::equal;
 106     case TemplateTable::less         : return Assembler::greaterEqual;
 107     case TemplateTable::less_equal   : return Assembler::greater;
 108     case TemplateTable::greater      : return Assembler::lessEqual;
 109     case TemplateTable::greater_equal: return Assembler::less;
 110   }
 111   ShouldNotReachHere();
 112   return Assembler::zero;
 113 }
 114 
 115 
 116 //----------------------------------------------------------------------------------------------------
 117 // Miscelaneous helper routines
 118 
 119 // Store an oop (or NULL) at the address described by obj.
 120 // If val == noreg this means store a NULL
 121 
 122 static void do_oop_store(InterpreterMacroAssembler* _masm,
 123                          Address obj,
 124                          Register val,
 125                          BarrierSet::Name barrier,
 126                          bool precise) {
 127   assert(val == noreg || val == rax, "parameter is just for looks");
 128   switch (barrier) {
 129 #if INCLUDE_ALL_GCS
 130     case BarrierSet::G1SATBCT:
 131     case BarrierSet::G1SATBCTLogging:
 132       {
 133         // flatten object address if needed
 134         // We do it regardless of precise because we need the registers
 135         if (obj.index() == noreg && obj.disp() == 0) {
 136           if (obj.base() != rdx) {
 137             __ movl(rdx, obj.base());
 138           }
 139         } else {
 140           __ leal(rdx, obj);
 141         }
 142         __ get_thread(rcx);
 143         __ save_bcp();
 144         __ g1_write_barrier_pre(rdx /* obj */,
 145                                 rbx /* pre_val */,
 146                                 rcx /* thread */,
 147                                 rsi /* tmp */,
 148                                 val != noreg /* tosca_live */,
 149                                 false /* expand_call */);
 150 
 151         // Do the actual store
 152         // noreg means NULL
 153         if (val == noreg) {
 154           __ movptr(Address(rdx, 0), NULL_WORD);
 155           // No post barrier for NULL
 156         } else {
 157           __ movl(Address(rdx, 0), val);
 158           __ g1_write_barrier_post(rdx /* store_adr */,
 159                                    val /* new_val */,
 160                                    rcx /* thread */,
 161                                    rbx /* tmp */,
 162                                    rsi /* tmp2 */);
 163         }
 164         __ restore_bcp();
 165 
 166       }
 167       break;
 168 #endif // INCLUDE_ALL_GCS
 169     case BarrierSet::CardTableModRef:
 170     case BarrierSet::CardTableExtension:
 171       {
 172         if (val == noreg) {
 173           __ movptr(obj, NULL_WORD);
 174         } else {
 175           __ movl(obj, val);
 176           // flatten object address if needed
 177           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 178             __ store_check(obj.base());
 179           } else {
 180             __ leal(rdx, obj);
 181             __ store_check(rdx);
 182           }
 183         }
 184       }
 185       break;
 186     case BarrierSet::ModRef:
 187     case BarrierSet::Other:
 188       if (val == noreg) {
 189         __ movptr(obj, NULL_WORD);
 190       } else {
 191         __ movl(obj, val);
 192       }
 193       break;
 194     default      :
 195       ShouldNotReachHere();
 196 
 197   }
 198 }
 199 
 200 Address TemplateTable::at_bcp(int offset) {
 201   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 202   return Address(rsi, offset);
 203 }
 204 
 205 
 206 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register bc_reg,
 207                                    Register temp_reg, bool load_bc_into_bc_reg/*=true*/,
 208                                    int byte_no) {
 209   if (!RewriteBytecodes)  return;
 210   Label L_patch_done;
 211 
 212   switch (bc) {
 213   case Bytecodes::_fast_aputfield:
 214   case Bytecodes::_fast_bputfield:
 215   case Bytecodes::_fast_cputfield:
 216   case Bytecodes::_fast_dputfield:
 217   case Bytecodes::_fast_fputfield:
 218   case Bytecodes::_fast_iputfield:
 219   case Bytecodes::_fast_lputfield:
 220   case Bytecodes::_fast_sputfield:
 221     {
 222       // We skip bytecode quickening for putfield instructions when
 223       // the put_code written to the constant pool cache is zero.
 224       // This is required so that every execution of this instruction
 225       // calls out to InterpreterRuntime::resolve_get_put to do
 226       // additional, required work.
 227       assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
 228       assert(load_bc_into_bc_reg, "we use bc_reg as temp");
 229       __ get_cache_and_index_and_bytecode_at_bcp(bc_reg, temp_reg, temp_reg, byte_no, 1);
 230       __ movl(bc_reg, bc);
 231       __ cmpl(temp_reg, (int) 0);
 232       __ jcc(Assembler::zero, L_patch_done);  // don't patch
 233     }
 234     break;
 235   default:
 236     assert(byte_no == -1, "sanity");
 237     // the pair bytecodes have already done the load.
 238     if (load_bc_into_bc_reg) {
 239       __ movl(bc_reg, bc);
 240     }
 241   }
 242 
 243   if (JvmtiExport::can_post_breakpoint()) {
 244     Label L_fast_patch;
 245     // if a breakpoint is present we can't rewrite the stream directly
 246     __ movzbl(temp_reg, at_bcp(0));
 247     __ cmpl(temp_reg, Bytecodes::_breakpoint);
 248     __ jcc(Assembler::notEqual, L_fast_patch);
 249     __ get_method(temp_reg);
 250     // Let breakpoint table handling rewrite to quicker bytecode
 251     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), temp_reg, rsi, bc_reg);
 252 #ifndef ASSERT
 253     __ jmpb(L_patch_done);
 254 #else
 255     __ jmp(L_patch_done);
 256 #endif
 257     __ bind(L_fast_patch);
 258   }
 259 
 260 #ifdef ASSERT
 261   Label L_okay;
 262   __ load_unsigned_byte(temp_reg, at_bcp(0));
 263   __ cmpl(temp_reg, (int)Bytecodes::java_code(bc));
 264   __ jccb(Assembler::equal, L_okay);
 265   __ cmpl(temp_reg, bc_reg);
 266   __ jcc(Assembler::equal, L_okay);
 267   __ stop("patching the wrong bytecode");
 268   __ bind(L_okay);
 269 #endif
 270 
 271   // patch bytecode
 272   __ movb(at_bcp(0), bc_reg);
 273   __ bind(L_patch_done);
 274 }
 275 
 276 //----------------------------------------------------------------------------------------------------
 277 // Individual instructions
 278 
 279 void TemplateTable::nop() {
 280   transition(vtos, vtos);
 281   // nothing to do
 282 }
 283 
 284 void TemplateTable::shouldnotreachhere() {
 285   transition(vtos, vtos);
 286   __ stop("shouldnotreachhere bytecode");
 287 }
 288 
 289 
 290 
 291 void TemplateTable::aconst_null() {
 292   transition(vtos, atos);
 293   __ xorptr(rax, rax);
 294 }
 295 
 296 
 297 void TemplateTable::iconst(int value) {
 298   transition(vtos, itos);
 299   if (value == 0) {
 300     __ xorptr(rax, rax);
 301   } else {
 302     __ movptr(rax, value);
 303   }
 304 }
 305 
 306 
 307 void TemplateTable::lconst(int value) {
 308   transition(vtos, ltos);
 309   if (value == 0) {
 310     __ xorptr(rax, rax);
 311   } else {
 312     __ movptr(rax, value);
 313   }
 314   assert(value >= 0, "check this code");
 315   __ xorptr(rdx, rdx);
 316 }
 317 
 318 
 319 void TemplateTable::fconst(int value) {
 320   transition(vtos, ftos);
 321          if (value == 0) { __ fldz();
 322   } else if (value == 1) { __ fld1();
 323   } else if (value == 2) { __ fld1(); __ fld1(); __ faddp(); // should do a better solution here
 324   } else                 { ShouldNotReachHere();
 325   }
 326 }
 327 
 328 
 329 void TemplateTable::dconst(int value) {
 330   transition(vtos, dtos);
 331          if (value == 0) { __ fldz();
 332   } else if (value == 1) { __ fld1();
 333   } else                 { ShouldNotReachHere();
 334   }
 335 }
 336 
 337 
 338 void TemplateTable::bipush() {
 339   transition(vtos, itos);
 340   __ load_signed_byte(rax, at_bcp(1));
 341 }
 342 
 343 
 344 void TemplateTable::sipush() {
 345   transition(vtos, itos);
 346   __ load_unsigned_short(rax, at_bcp(1));
 347   __ bswapl(rax);
 348   __ sarl(rax, 16);
 349 }
 350 
 351 void TemplateTable::ldc(bool wide) {
 352   transition(vtos, vtos);
 353   Label call_ldc, notFloat, notClass, Done;
 354 
 355   if (wide) {
 356     __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 357   } else {
 358     __ load_unsigned_byte(rbx, at_bcp(1));
 359   }
 360   __ get_cpool_and_tags(rcx, rax);
 361   const int base_offset = ConstantPool::header_size() * wordSize;
 362   const int tags_offset = Array<u1>::base_offset_in_bytes();
 363 
 364   // get type
 365   __ xorptr(rdx, rdx);
 366   __ movb(rdx, Address(rax, rbx, Address::times_1, tags_offset));
 367 
 368   // unresolved class - get the resolved class
 369   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClass);
 370   __ jccb(Assembler::equal, call_ldc);
 371 
 372   // unresolved class in error (resolution failed) - call into runtime
 373   // so that the same error from first resolution attempt is thrown.
 374   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClassInError);
 375   __ jccb(Assembler::equal, call_ldc);
 376 
 377   // resolved class - need to call vm to get java mirror of the class
 378   __ cmpl(rdx, JVM_CONSTANT_Class);
 379   __ jcc(Assembler::notEqual, notClass);
 380 
 381   __ bind(call_ldc);
 382   __ movl(rcx, wide);
 383   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), rcx);
 384   __ push(atos);
 385   __ jmp(Done);
 386 
 387   __ bind(notClass);
 388   __ cmpl(rdx, JVM_CONSTANT_Float);
 389   __ jccb(Assembler::notEqual, notFloat);
 390   // ftos
 391   __ fld_s(    Address(rcx, rbx, Address::times_ptr, base_offset));
 392   __ push(ftos);
 393   __ jmp(Done);
 394 
 395   __ bind(notFloat);
 396 #ifdef ASSERT
 397   { Label L;
 398     __ cmpl(rdx, JVM_CONSTANT_Integer);
 399     __ jcc(Assembler::equal, L);
 400     // String and Object are rewritten to fast_aldc
 401     __ stop("unexpected tag type in ldc");
 402     __ bind(L);
 403   }
 404 #endif
 405   // itos JVM_CONSTANT_Integer only
 406   __ movl(rax, Address(rcx, rbx, Address::times_ptr, base_offset));
 407   __ push(itos);
 408   __ bind(Done);
 409 }
 410 
 411 // Fast path for caching oop constants.
 412 void TemplateTable::fast_aldc(bool wide) {
 413   transition(vtos, atos);
 414 
 415   Register result = rax;
 416   Register tmp = rdx;
 417   int index_size = wide ? sizeof(u2) : sizeof(u1);
 418 
 419   Label resolved;
 420 
 421   // We are resolved if the resolved reference cache entry contains a
 422   // non-null object (String, MethodType, etc.)
 423   assert_different_registers(result, tmp);
 424   __ get_cache_index_at_bcp(tmp, 1, index_size);
 425   __ load_resolved_reference_at_index(result, tmp);
 426   __ testl(result, result);
 427   __ jcc(Assembler::notZero, resolved);
 428 
 429   address entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_ldc);
 430 
 431   // first time invocation - must resolve first
 432   __ movl(tmp, (int)bytecode());
 433   __ call_VM(result, entry, tmp);
 434 
 435   __ bind(resolved);
 436 
 437   if (VerifyOops) {
 438     __ verify_oop(result);
 439   }
 440 }
 441 
 442 void TemplateTable::ldc2_w() {
 443   transition(vtos, vtos);
 444   Label Long, Done;
 445   __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 446 
 447   __ get_cpool_and_tags(rcx, rax);
 448   const int base_offset = ConstantPool::header_size() * wordSize;
 449   const int tags_offset = Array<u1>::base_offset_in_bytes();
 450 
 451   // get type
 452   __ cmpb(Address(rax, rbx, Address::times_1, tags_offset), JVM_CONSTANT_Double);
 453   __ jccb(Assembler::notEqual, Long);
 454   // dtos
 455   __ fld_d(    Address(rcx, rbx, Address::times_ptr, base_offset));
 456   __ push(dtos);
 457   __ jmpb(Done);
 458 
 459   __ bind(Long);
 460   // ltos
 461   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, base_offset + 0 * wordSize));
 462   NOT_LP64(__ movptr(rdx, Address(rcx, rbx, Address::times_ptr, base_offset + 1 * wordSize)));
 463 
 464   __ push(ltos);
 465 
 466   __ bind(Done);
 467 }
 468 
 469 
 470 void TemplateTable::locals_index(Register reg, int offset) {
 471   __ load_unsigned_byte(reg, at_bcp(offset));
 472   __ negptr(reg);
 473 }
 474 
 475 
 476 void TemplateTable::iload() {
 477   transition(vtos, itos);
 478   if (RewriteFrequentPairs) {
 479     Label rewrite, done;
 480 
 481     // get next byte
 482     __ load_unsigned_byte(rbx, at_bcp(Bytecodes::length_for(Bytecodes::_iload)));
 483     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 484     // last two iloads in a pair.  Comparing against fast_iload means that
 485     // the next bytecode is neither an iload or a caload, and therefore
 486     // an iload pair.
 487     __ cmpl(rbx, Bytecodes::_iload);
 488     __ jcc(Assembler::equal, done);
 489 
 490     __ cmpl(rbx, Bytecodes::_fast_iload);
 491     __ movl(rcx, Bytecodes::_fast_iload2);
 492     __ jccb(Assembler::equal, rewrite);
 493 
 494     // if _caload, rewrite to fast_icaload
 495     __ cmpl(rbx, Bytecodes::_caload);
 496     __ movl(rcx, Bytecodes::_fast_icaload);
 497     __ jccb(Assembler::equal, rewrite);
 498 
 499     // rewrite so iload doesn't check again.
 500     __ movl(rcx, Bytecodes::_fast_iload);
 501 
 502     // rewrite
 503     // rcx: fast bytecode
 504     __ bind(rewrite);
 505     patch_bytecode(Bytecodes::_iload, rcx, rbx, false);
 506     __ bind(done);
 507   }
 508 
 509   // Get the local value into tos
 510   locals_index(rbx);
 511   __ movl(rax, iaddress(rbx));
 512 }
 513 
 514 
 515 void TemplateTable::fast_iload2() {
 516   transition(vtos, itos);
 517   locals_index(rbx);
 518   __ movl(rax, iaddress(rbx));
 519   __ push(itos);
 520   locals_index(rbx, 3);
 521   __ movl(rax, iaddress(rbx));
 522 }
 523 
 524 void TemplateTable::fast_iload() {
 525   transition(vtos, itos);
 526   locals_index(rbx);
 527   __ movl(rax, iaddress(rbx));
 528 }
 529 
 530 
 531 void TemplateTable::lload() {
 532   transition(vtos, ltos);
 533   locals_index(rbx);
 534   __ movptr(rax, laddress(rbx));
 535   NOT_LP64(__ movl(rdx, haddress(rbx)));
 536 }
 537 
 538 
 539 void TemplateTable::fload() {
 540   transition(vtos, ftos);
 541   locals_index(rbx);
 542   __ fld_s(faddress(rbx));
 543 }
 544 
 545 
 546 void TemplateTable::dload() {
 547   transition(vtos, dtos);
 548   locals_index(rbx);
 549   __ fld_d(daddress(rbx));
 550 }
 551 
 552 
 553 void TemplateTable::aload() {
 554   transition(vtos, atos);
 555   locals_index(rbx);
 556   __ movptr(rax, aaddress(rbx));
 557 }
 558 
 559 
 560 void TemplateTable::locals_index_wide(Register reg) {
 561   __ movl(reg, at_bcp(2));
 562   __ bswapl(reg);
 563   __ shrl(reg, 16);
 564   __ negptr(reg);
 565 }
 566 
 567 
 568 void TemplateTable::wide_iload() {
 569   transition(vtos, itos);
 570   locals_index_wide(rbx);
 571   __ movl(rax, iaddress(rbx));
 572 }
 573 
 574 
 575 void TemplateTable::wide_lload() {
 576   transition(vtos, ltos);
 577   locals_index_wide(rbx);
 578   __ movptr(rax, laddress(rbx));
 579   NOT_LP64(__ movl(rdx, haddress(rbx)));
 580 }
 581 
 582 
 583 void TemplateTable::wide_fload() {
 584   transition(vtos, ftos);
 585   locals_index_wide(rbx);
 586   __ fld_s(faddress(rbx));
 587 }
 588 
 589 
 590 void TemplateTable::wide_dload() {
 591   transition(vtos, dtos);
 592   locals_index_wide(rbx);
 593   __ fld_d(daddress(rbx));
 594 }
 595 
 596 
 597 void TemplateTable::wide_aload() {
 598   transition(vtos, atos);
 599   locals_index_wide(rbx);
 600   __ movptr(rax, aaddress(rbx));
 601 }
 602 
 603 void TemplateTable::index_check(Register array, Register index) {
 604   // Pop ptr into array
 605   __ pop_ptr(array);
 606   index_check_without_pop(array, index);
 607 }
 608 
 609 void TemplateTable::index_check_without_pop(Register array, Register index) {
 610   // destroys rbx,
 611   // check array
 612   __ null_check(array, arrayOopDesc::length_offset_in_bytes());
 613   LP64_ONLY(__ movslq(index, index));
 614   // check index
 615   __ cmpl(index, Address(array, arrayOopDesc::length_offset_in_bytes()));
 616   if (index != rbx) {
 617     // ??? convention: move aberrant index into rbx, for exception message
 618     assert(rbx != array, "different registers");
 619     __ mov(rbx, index);
 620   }
 621   __ jump_cc(Assembler::aboveEqual,
 622              ExternalAddress(Interpreter::_throw_ArrayIndexOutOfBoundsException_entry));
 623 }
 624 
 625 
 626 void TemplateTable::iaload() {
 627   transition(itos, itos);
 628   // rdx: array
 629   index_check(rdx, rax);  // kills rbx,
 630   // rax,: index
 631   __ movl(rax, Address(rdx, rax, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_INT)));
 632 }
 633 
 634 
 635 void TemplateTable::laload() {
 636   transition(itos, ltos);
 637   // rax,: index
 638   // rdx: array
 639   index_check(rdx, rax);
 640   __ mov(rbx, rax);
 641   // rbx,: index
 642   __ movptr(rax, Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 0 * wordSize));
 643   NOT_LP64(__ movl(rdx, Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 1 * wordSize)));
 644 }
 645 
 646 
 647 void TemplateTable::faload() {
 648   transition(itos, ftos);
 649   // rdx: array
 650   index_check(rdx, rax);  // kills rbx,
 651   // rax,: index
 652   __ fld_s(Address(rdx, rax, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 653 }
 654 
 655 
 656 void TemplateTable::daload() {
 657   transition(itos, dtos);
 658   // rdx: array
 659   index_check(rdx, rax);  // kills rbx,
 660   // rax,: index
 661   __ fld_d(Address(rdx, rax, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 662 }
 663 
 664 
 665 void TemplateTable::aaload() {
 666   transition(itos, atos);
 667   // rdx: array
 668   index_check(rdx, rax);  // kills rbx,
 669   // rax,: index
 670   __ movptr(rax, Address(rdx, rax, Address::times_ptr, arrayOopDesc::base_offset_in_bytes(T_OBJECT)));
 671 }
 672 
 673 
 674 void TemplateTable::baload() {
 675   transition(itos, itos);
 676   // rdx: array
 677   index_check(rdx, rax);  // kills rbx,
 678   // rax,: index
 679   // can do better code for P5 - fix this at some point
 680   __ load_signed_byte(rbx, Address(rdx, rax, Address::times_1, arrayOopDesc::base_offset_in_bytes(T_BYTE)));
 681   __ mov(rax, rbx);
 682 }
 683 
 684 
 685 void TemplateTable::caload() {
 686   transition(itos, itos);
 687   // rdx: array
 688   index_check(rdx, rax);  // kills rbx,
 689   // rax,: index
 690   // can do better code for P5 - may want to improve this at some point
 691   __ load_unsigned_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 692   __ mov(rax, rbx);
 693 }
 694 
 695 // iload followed by caload frequent pair
 696 void TemplateTable::fast_icaload() {
 697   transition(vtos, itos);
 698   // load index out of locals
 699   locals_index(rbx);
 700   __ movl(rax, iaddress(rbx));
 701 
 702   // rdx: array
 703   index_check(rdx, rax);
 704   // rax,: index
 705   __ load_unsigned_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 706   __ mov(rax, rbx);
 707 }
 708 
 709 void TemplateTable::saload() {
 710   transition(itos, itos);
 711   // rdx: array
 712   index_check(rdx, rax);  // kills rbx,
 713   // rax,: index
 714   // can do better code for P5 - may want to improve this at some point
 715   __ load_signed_short(rbx, Address(rdx, rax, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_SHORT)));
 716   __ mov(rax, rbx);
 717 }
 718 
 719 
 720 void TemplateTable::iload(int n) {
 721   transition(vtos, itos);
 722   __ movl(rax, iaddress(n));
 723 }
 724 
 725 
 726 void TemplateTable::lload(int n) {
 727   transition(vtos, ltos);
 728   __ movptr(rax, laddress(n));
 729   NOT_LP64(__ movptr(rdx, haddress(n)));
 730 }
 731 
 732 
 733 void TemplateTable::fload(int n) {
 734   transition(vtos, ftos);
 735   __ fld_s(faddress(n));
 736 }
 737 
 738 
 739 void TemplateTable::dload(int n) {
 740   transition(vtos, dtos);
 741   __ fld_d(daddress(n));
 742 }
 743 
 744 
 745 void TemplateTable::aload(int n) {
 746   transition(vtos, atos);
 747   __ movptr(rax, aaddress(n));
 748 }
 749 
 750 
 751 void TemplateTable::aload_0() {
 752   transition(vtos, atos);
 753   // According to bytecode histograms, the pairs:
 754   //
 755   // _aload_0, _fast_igetfield
 756   // _aload_0, _fast_agetfield
 757   // _aload_0, _fast_fgetfield
 758   //
 759   // occur frequently. If RewriteFrequentPairs is set, the (slow) _aload_0
 760   // bytecode checks if the next bytecode is either _fast_igetfield,
 761   // _fast_agetfield or _fast_fgetfield and then rewrites the
 762   // current bytecode into a pair bytecode; otherwise it rewrites the current
 763   // bytecode into _fast_aload_0 that doesn't do the pair check anymore.
 764   //
 765   // Note: If the next bytecode is _getfield, the rewrite must be delayed,
 766   //       otherwise we may miss an opportunity for a pair.
 767   //
 768   // Also rewrite frequent pairs
 769   //   aload_0, aload_1
 770   //   aload_0, iload_1
 771   // These bytecodes with a small amount of code are most profitable to rewrite
 772   if (RewriteFrequentPairs) {
 773     Label rewrite, done;
 774     // get next byte
 775     __ load_unsigned_byte(rbx, at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)));
 776 
 777     // do actual aload_0
 778     aload(0);
 779 
 780     // if _getfield then wait with rewrite
 781     __ cmpl(rbx, Bytecodes::_getfield);
 782     __ jcc(Assembler::equal, done);
 783 
 784     // if _igetfield then reqrite to _fast_iaccess_0
 785     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 786     __ cmpl(rbx, Bytecodes::_fast_igetfield);
 787     __ movl(rcx, Bytecodes::_fast_iaccess_0);
 788     __ jccb(Assembler::equal, rewrite);
 789 
 790     // if _agetfield then reqrite to _fast_aaccess_0
 791     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 792     __ cmpl(rbx, Bytecodes::_fast_agetfield);
 793     __ movl(rcx, Bytecodes::_fast_aaccess_0);
 794     __ jccb(Assembler::equal, rewrite);
 795 
 796     // if _fgetfield then reqrite to _fast_faccess_0
 797     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) == Bytecodes::_aload_0, "fix bytecode definition");
 798     __ cmpl(rbx, Bytecodes::_fast_fgetfield);
 799     __ movl(rcx, Bytecodes::_fast_faccess_0);
 800     __ jccb(Assembler::equal, rewrite);
 801 
 802     // else rewrite to _fast_aload0
 803     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) == Bytecodes::_aload_0, "fix bytecode definition");
 804     __ movl(rcx, Bytecodes::_fast_aload_0);
 805 
 806     // rewrite
 807     // rcx: fast bytecode
 808     __ bind(rewrite);
 809     patch_bytecode(Bytecodes::_aload_0, rcx, rbx, false);
 810 
 811     __ bind(done);
 812   } else {
 813     aload(0);
 814   }
 815 }
 816 
 817 void TemplateTable::istore() {
 818   transition(itos, vtos);
 819   locals_index(rbx);
 820   __ movl(iaddress(rbx), rax);
 821 }
 822 
 823 
 824 void TemplateTable::lstore() {
 825   transition(ltos, vtos);
 826   locals_index(rbx);
 827   __ movptr(laddress(rbx), rax);
 828   NOT_LP64(__ movptr(haddress(rbx), rdx));
 829 }
 830 
 831 
 832 void TemplateTable::fstore() {
 833   transition(ftos, vtos);
 834   locals_index(rbx);
 835   __ fstp_s(faddress(rbx));
 836 }
 837 
 838 
 839 void TemplateTable::dstore() {
 840   transition(dtos, vtos);
 841   locals_index(rbx);
 842   __ fstp_d(daddress(rbx));
 843 }
 844 
 845 
 846 void TemplateTable::astore() {
 847   transition(vtos, vtos);
 848   __ pop_ptr(rax);
 849   locals_index(rbx);
 850   __ movptr(aaddress(rbx), rax);
 851 }
 852 
 853 
 854 void TemplateTable::wide_istore() {
 855   transition(vtos, vtos);
 856   __ pop_i(rax);
 857   locals_index_wide(rbx);
 858   __ movl(iaddress(rbx), rax);
 859 }
 860 
 861 
 862 void TemplateTable::wide_lstore() {
 863   transition(vtos, vtos);
 864   __ pop_l(rax, rdx);
 865   locals_index_wide(rbx);
 866   __ movptr(laddress(rbx), rax);
 867   NOT_LP64(__ movl(haddress(rbx), rdx));
 868 }
 869 
 870 
 871 void TemplateTable::wide_fstore() {
 872   wide_istore();
 873 }
 874 
 875 
 876 void TemplateTable::wide_dstore() {
 877   wide_lstore();
 878 }
 879 
 880 
 881 void TemplateTable::wide_astore() {
 882   transition(vtos, vtos);
 883   __ pop_ptr(rax);
 884   locals_index_wide(rbx);
 885   __ movptr(aaddress(rbx), rax);
 886 }
 887 
 888 
 889 void TemplateTable::iastore() {
 890   transition(itos, vtos);
 891   __ pop_i(rbx);
 892   // rax,: value
 893   // rdx: array
 894   index_check(rdx, rbx);  // prefer index in rbx,
 895   // rbx,: index
 896   __ movl(Address(rdx, rbx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_INT)), rax);
 897 }
 898 
 899 
 900 void TemplateTable::lastore() {
 901   transition(ltos, vtos);
 902   __ pop_i(rbx);
 903   // rax,: low(value)
 904   // rcx: array
 905   // rdx: high(value)
 906   index_check(rcx, rbx);  // prefer index in rbx,
 907   // rbx,: index
 908   __ movptr(Address(rcx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 0 * wordSize), rax);
 909   NOT_LP64(__ movl(Address(rcx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_LONG) + 1 * wordSize), rdx));
 910 }
 911 
 912 
 913 void TemplateTable::fastore() {
 914   transition(ftos, vtos);
 915   __ pop_i(rbx);
 916   // rdx: array
 917   // st0: value
 918   index_check(rdx, rbx);  // prefer index in rbx,
 919   // rbx,: index
 920   __ fstp_s(Address(rdx, rbx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 921 }
 922 
 923 
 924 void TemplateTable::dastore() {
 925   transition(dtos, vtos);
 926   __ pop_i(rbx);
 927   // rdx: array
 928   // st0: value
 929   index_check(rdx, rbx);  // prefer index in rbx,
 930   // rbx,: index
 931   __ fstp_d(Address(rdx, rbx, Address::times_8, arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 932 }
 933 
 934 
 935 void TemplateTable::aastore() {
 936   Label is_null, ok_is_subtype, done;
 937   transition(vtos, vtos);
 938   // stack: ..., array, index, value
 939   __ movptr(rax, at_tos());     // Value
 940   __ movl(rcx, at_tos_p1());  // Index
 941   __ movptr(rdx, at_tos_p2());  // Array
 942 
 943   Address element_address(rdx, rcx, Address::times_4, arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 944   index_check_without_pop(rdx, rcx);      // kills rbx,
 945   // do array store check - check for NULL value first
 946   __ testptr(rax, rax);
 947   __ jcc(Assembler::zero, is_null);
 948 
 949   // Move subklass into EBX
 950   __ load_klass(rbx, rax);
 951   // Move superklass into EAX
 952   __ load_klass(rax, rdx);
 953   __ movptr(rax, Address(rax, ObjArrayKlass::element_klass_offset()));
 954   // Compress array+index*wordSize+12 into a single register.  Frees ECX.
 955   __ lea(rdx, element_address);
 956 
 957   // Generate subtype check.  Blows ECX.  Resets EDI to locals.
 958   // Superklass in EAX.  Subklass in EBX.
 959   __ gen_subtype_check( rbx, ok_is_subtype );
 960 
 961   // Come here on failure
 962   // object is at TOS
 963   __ jump(ExternalAddress(Interpreter::_throw_ArrayStoreException_entry));
 964 
 965   // Come here on success
 966   __ bind(ok_is_subtype);
 967 
 968   // Get the value to store
 969   __ movptr(rax, at_rsp());
 970   // and store it with appropriate barrier
 971   do_oop_store(_masm, Address(rdx, 0), rax, _bs->kind(), true);
 972 
 973   __ jmp(done);
 974 
 975   // Have a NULL in EAX, EDX=array, ECX=index.  Store NULL at ary[idx]
 976   __ bind(is_null);
 977   __ profile_null_seen(rbx);
 978 
 979   // Store NULL, (noreg means NULL to do_oop_store)
 980   do_oop_store(_masm, element_address, noreg, _bs->kind(), true);
 981 
 982   // Pop stack arguments
 983   __ bind(done);
 984   __ addptr(rsp, 3 * Interpreter::stackElementSize);
 985 }
 986 
 987 
 988 void TemplateTable::bastore() {
 989   transition(itos, vtos);
 990   __ pop_i(rbx);
 991   // rax,: value
 992   // rdx: array
 993   index_check(rdx, rbx);  // prefer index in rbx,
 994   // rbx,: index
 995   __ movb(Address(rdx, rbx, Address::times_1, arrayOopDesc::base_offset_in_bytes(T_BYTE)), rax);
 996 }
 997 
 998 
 999 void TemplateTable::castore() {
1000   transition(itos, vtos);
1001   __ pop_i(rbx);
1002   // rax,: value
1003   // rdx: array
1004   index_check(rdx, rbx);  // prefer index in rbx,
1005   // rbx,: index
1006   __ movw(Address(rdx, rbx, Address::times_2, arrayOopDesc::base_offset_in_bytes(T_CHAR)), rax);
1007 }
1008 
1009 
1010 void TemplateTable::sastore() {
1011   castore();
1012 }
1013 
1014 
1015 void TemplateTable::istore(int n) {
1016   transition(itos, vtos);
1017   __ movl(iaddress(n), rax);
1018 }
1019 
1020 
1021 void TemplateTable::lstore(int n) {
1022   transition(ltos, vtos);
1023   __ movptr(laddress(n), rax);
1024   NOT_LP64(__ movptr(haddress(n), rdx));
1025 }
1026 
1027 
1028 void TemplateTable::fstore(int n) {
1029   transition(ftos, vtos);
1030   __ fstp_s(faddress(n));
1031 }
1032 
1033 
1034 void TemplateTable::dstore(int n) {
1035   transition(dtos, vtos);
1036   __ fstp_d(daddress(n));
1037 }
1038 
1039 
1040 void TemplateTable::astore(int n) {
1041   transition(vtos, vtos);
1042   __ pop_ptr(rax);
1043   __ movptr(aaddress(n), rax);
1044 }
1045 
1046 
1047 void TemplateTable::pop() {
1048   transition(vtos, vtos);
1049   __ addptr(rsp, Interpreter::stackElementSize);
1050 }
1051 
1052 
1053 void TemplateTable::pop2() {
1054   transition(vtos, vtos);
1055   __ addptr(rsp, 2*Interpreter::stackElementSize);
1056 }
1057 
1058 
1059 void TemplateTable::dup() {
1060   transition(vtos, vtos);
1061   // stack: ..., a
1062   __ load_ptr(0, rax);
1063   __ push_ptr(rax);
1064   // stack: ..., a, a
1065 }
1066 
1067 
1068 void TemplateTable::dup_x1() {
1069   transition(vtos, vtos);
1070   // stack: ..., a, b
1071   __ load_ptr( 0, rax);  // load b
1072   __ load_ptr( 1, rcx);  // load a
1073   __ store_ptr(1, rax);  // store b
1074   __ store_ptr(0, rcx);  // store a
1075   __ push_ptr(rax);      // push b
1076   // stack: ..., b, a, b
1077 }
1078 
1079 
1080 void TemplateTable::dup_x2() {
1081   transition(vtos, vtos);
1082   // stack: ..., a, b, c
1083   __ load_ptr( 0, rax);  // load c
1084   __ load_ptr( 2, rcx);  // load a
1085   __ store_ptr(2, rax);  // store c in a
1086   __ push_ptr(rax);      // push c
1087   // stack: ..., c, b, c, c
1088   __ load_ptr( 2, rax);  // load b
1089   __ store_ptr(2, rcx);  // store a in b
1090   // stack: ..., c, a, c, c
1091   __ store_ptr(1, rax);  // store b in c
1092   // stack: ..., c, a, b, c
1093 }
1094 
1095 
1096 void TemplateTable::dup2() {
1097   transition(vtos, vtos);
1098   // stack: ..., a, b
1099   __ load_ptr(1, rax);  // load a
1100   __ push_ptr(rax);     // push a
1101   __ load_ptr(1, rax);  // load b
1102   __ push_ptr(rax);     // push b
1103   // stack: ..., a, b, a, b
1104 }
1105 
1106 
1107 void TemplateTable::dup2_x1() {
1108   transition(vtos, vtos);
1109   // stack: ..., a, b, c
1110   __ load_ptr( 0, rcx);  // load c
1111   __ load_ptr( 1, rax);  // load b
1112   __ push_ptr(rax);      // push b
1113   __ push_ptr(rcx);      // push c
1114   // stack: ..., a, b, c, b, c
1115   __ store_ptr(3, rcx);  // store c in b
1116   // stack: ..., a, c, c, b, c
1117   __ load_ptr( 4, rcx);  // load a
1118   __ store_ptr(2, rcx);  // store a in 2nd c
1119   // stack: ..., a, c, a, b, c
1120   __ store_ptr(4, rax);  // store b in a
1121   // stack: ..., b, c, a, b, c
1122   // stack: ..., b, c, a, b, c
1123 }
1124 
1125 
1126 void TemplateTable::dup2_x2() {
1127   transition(vtos, vtos);
1128   // stack: ..., a, b, c, d
1129   __ load_ptr( 0, rcx);  // load d
1130   __ load_ptr( 1, rax);  // load c
1131   __ push_ptr(rax);      // push c
1132   __ push_ptr(rcx);      // push d
1133   // stack: ..., a, b, c, d, c, d
1134   __ load_ptr( 4, rax);  // load b
1135   __ store_ptr(2, rax);  // store b in d
1136   __ store_ptr(4, rcx);  // store d in b
1137   // stack: ..., a, d, c, b, c, d
1138   __ load_ptr( 5, rcx);  // load a
1139   __ load_ptr( 3, rax);  // load c
1140   __ store_ptr(3, rcx);  // store a in c
1141   __ store_ptr(5, rax);  // store c in a
1142   // stack: ..., c, d, a, b, c, d
1143   // stack: ..., c, d, a, b, c, d
1144 }
1145 
1146 
1147 void TemplateTable::swap() {
1148   transition(vtos, vtos);
1149   // stack: ..., a, b
1150   __ load_ptr( 1, rcx);  // load a
1151   __ load_ptr( 0, rax);  // load b
1152   __ store_ptr(0, rcx);  // store a in b
1153   __ store_ptr(1, rax);  // store b in a
1154   // stack: ..., b, a
1155 }
1156 
1157 
1158 void TemplateTable::iop2(Operation op) {
1159   transition(itos, itos);
1160   switch (op) {
1161     case add  :                   __ pop_i(rdx); __ addl (rax, rdx); break;
1162     case sub  : __ mov(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1163     case mul  :                   __ pop_i(rdx); __ imull(rax, rdx); break;
1164     case _and :                   __ pop_i(rdx); __ andl (rax, rdx); break;
1165     case _or  :                   __ pop_i(rdx); __ orl  (rax, rdx); break;
1166     case _xor :                   __ pop_i(rdx); __ xorl (rax, rdx); break;
1167     case shl  : __ mov(rcx, rax); __ pop_i(rax); __ shll (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1168     case shr  : __ mov(rcx, rax); __ pop_i(rax); __ sarl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1169     case ushr : __ mov(rcx, rax); __ pop_i(rax); __ shrl (rax);      break; // implicit masking of lower 5 bits by Intel shift instr.
1170     default   : ShouldNotReachHere();
1171   }
1172 }
1173 
1174 
1175 void TemplateTable::lop2(Operation op) {
1176   transition(ltos, ltos);
1177   __ pop_l(rbx, rcx);
1178   switch (op) {
1179     case add  : __ addl(rax, rbx); __ adcl(rdx, rcx); break;
1180     case sub  : __ subl(rbx, rax); __ sbbl(rcx, rdx);
1181                 __ mov (rax, rbx); __ mov (rdx, rcx); break;
1182     case _and : __ andl(rax, rbx); __ andl(rdx, rcx); break;
1183     case _or  : __ orl (rax, rbx); __ orl (rdx, rcx); break;
1184     case _xor : __ xorl(rax, rbx); __ xorl(rdx, rcx); break;
1185     default   : ShouldNotReachHere();
1186   }
1187 }
1188 
1189 
1190 void TemplateTable::idiv() {
1191   transition(itos, itos);
1192   __ mov(rcx, rax);
1193   __ pop_i(rax);
1194   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1195   //       they are not equal, one could do a normal division (no correction
1196   //       needed), which may speed up this implementation for the common case.
1197   //       (see also JVM spec., p.243 & p.271)
1198   __ corrected_idivl(rcx);
1199 }
1200 
1201 
1202 void TemplateTable::irem() {
1203   transition(itos, itos);
1204   __ mov(rcx, rax);
1205   __ pop_i(rax);
1206   // Note: could xor rax, and rcx and compare with (-1 ^ min_int). If
1207   //       they are not equal, one could do a normal division (no correction
1208   //       needed), which may speed up this implementation for the common case.
1209   //       (see also JVM spec., p.243 & p.271)
1210   __ corrected_idivl(rcx);
1211   __ mov(rax, rdx);
1212 }
1213 
1214 
1215 void TemplateTable::lmul() {
1216   transition(ltos, ltos);
1217   __ pop_l(rbx, rcx);
1218   __ push(rcx); __ push(rbx);
1219   __ push(rdx); __ push(rax);
1220   __ lmul(2 * wordSize, 0);
1221   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1222 }
1223 
1224 
1225 void TemplateTable::ldiv() {
1226   transition(ltos, ltos);
1227   __ pop_l(rbx, rcx);
1228   __ push(rcx); __ push(rbx);
1229   __ push(rdx); __ push(rax);
1230   // check if y = 0
1231   __ orl(rax, rdx);
1232   __ jump_cc(Assembler::zero,
1233              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1234   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::ldiv));
1235   __ addptr(rsp, 4 * wordSize);  // take off temporaries
1236 }
1237 
1238 
1239 void TemplateTable::lrem() {
1240   transition(ltos, ltos);
1241   __ pop_l(rbx, rcx);
1242   __ push(rcx); __ push(rbx);
1243   __ push(rdx); __ push(rax);
1244   // check if y = 0
1245   __ orl(rax, rdx);
1246   __ jump_cc(Assembler::zero,
1247              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1248   __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::lrem));
1249   __ addptr(rsp, 4 * wordSize);
1250 }
1251 
1252 
1253 void TemplateTable::lshl() {
1254   transition(itos, ltos);
1255   __ movl(rcx, rax);                             // get shift count
1256   __ pop_l(rax, rdx);                            // get shift value
1257   __ lshl(rdx, rax);
1258 }
1259 
1260 
1261 void TemplateTable::lshr() {
1262   transition(itos, ltos);
1263   __ mov(rcx, rax);                              // get shift count
1264   __ pop_l(rax, rdx);                            // get shift value
1265   __ lshr(rdx, rax, true);
1266 }
1267 
1268 
1269 void TemplateTable::lushr() {
1270   transition(itos, ltos);
1271   __ mov(rcx, rax);                              // get shift count
1272   __ pop_l(rax, rdx);                            // get shift value
1273   __ lshr(rdx, rax);
1274 }
1275 
1276 
1277 void TemplateTable::fop2(Operation op) {
1278   transition(ftos, ftos);
1279   switch (op) {
1280     case add: __ fadd_s (at_rsp());                break;
1281     case sub: __ fsubr_s(at_rsp());                break;
1282     case mul: __ fmul_s (at_rsp());                break;
1283     case div: __ fdivr_s(at_rsp());                break;
1284     case rem: __ fld_s  (at_rsp()); __ fremr(rax); break;
1285     default : ShouldNotReachHere();
1286   }
1287   __ f2ieee();
1288   __ pop(rax);  // pop float thing off
1289 }
1290 
1291 
1292 void TemplateTable::dop2(Operation op) {
1293   transition(dtos, dtos);
1294 
1295   switch (op) {
1296     case add: __ fadd_d (at_rsp());                break;
1297     case sub: __ fsubr_d(at_rsp());                break;
1298     case mul: {
1299       Label L_strict;
1300       Label L_join;
1301       const Address access_flags      (rcx, Method::access_flags_offset());
1302       __ get_method(rcx);
1303       __ movl(rcx, access_flags);
1304       __ testl(rcx, JVM_ACC_STRICT);
1305       __ jccb(Assembler::notZero, L_strict);
1306       __ fmul_d (at_rsp());
1307       __ jmpb(L_join);
1308       __ bind(L_strict);
1309       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1310       __ fmulp();
1311       __ fmul_d (at_rsp());
1312       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1313       __ fmulp();
1314       __ bind(L_join);
1315       break;
1316     }
1317     case div: {
1318       Label L_strict;
1319       Label L_join;
1320       const Address access_flags      (rcx, Method::access_flags_offset());
1321       __ get_method(rcx);
1322       __ movl(rcx, access_flags);
1323       __ testl(rcx, JVM_ACC_STRICT);
1324       __ jccb(Assembler::notZero, L_strict);
1325       __ fdivr_d(at_rsp());
1326       __ jmp(L_join);
1327       __ bind(L_strict);
1328       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias1()));
1329       __ fmul_d (at_rsp());
1330       __ fdivrp();
1331       __ fld_x(ExternalAddress(StubRoutines::addr_fpu_subnormal_bias2()));
1332       __ fmulp();
1333       __ bind(L_join);
1334       break;
1335     }
1336     case rem: __ fld_d  (at_rsp()); __ fremr(rax); break;
1337     default : ShouldNotReachHere();
1338   }
1339   __ d2ieee();
1340   // Pop double precision number from rsp.
1341   __ pop(rax);
1342   __ pop(rdx);
1343 }
1344 
1345 
1346 void TemplateTable::ineg() {
1347   transition(itos, itos);
1348   __ negl(rax);
1349 }
1350 
1351 
1352 void TemplateTable::lneg() {
1353   transition(ltos, ltos);
1354   __ lneg(rdx, rax);
1355 }
1356 
1357 
1358 void TemplateTable::fneg() {
1359   transition(ftos, ftos);
1360   __ fchs();
1361 }
1362 
1363 
1364 void TemplateTable::dneg() {
1365   transition(dtos, dtos);
1366   __ fchs();
1367 }
1368 
1369 
1370 void TemplateTable::iinc() {
1371   transition(vtos, vtos);
1372   __ load_signed_byte(rdx, at_bcp(2));           // get constant
1373   locals_index(rbx);
1374   __ addl(iaddress(rbx), rdx);
1375 }
1376 
1377 
1378 void TemplateTable::wide_iinc() {
1379   transition(vtos, vtos);
1380   __ movl(rdx, at_bcp(4));                       // get constant
1381   locals_index_wide(rbx);
1382   __ bswapl(rdx);                                 // swap bytes & sign-extend constant
1383   __ sarl(rdx, 16);
1384   __ addl(iaddress(rbx), rdx);
1385   // Note: should probably use only one movl to get both
1386   //       the index and the constant -> fix this
1387 }
1388 
1389 
1390 void TemplateTable::convert() {
1391   // Checking
1392 #ifdef ASSERT
1393   { TosState tos_in  = ilgl;
1394     TosState tos_out = ilgl;
1395     switch (bytecode()) {
1396       case Bytecodes::_i2l: // fall through
1397       case Bytecodes::_i2f: // fall through
1398       case Bytecodes::_i2d: // fall through
1399       case Bytecodes::_i2b: // fall through
1400       case Bytecodes::_i2c: // fall through
1401       case Bytecodes::_i2s: tos_in = itos; break;
1402       case Bytecodes::_l2i: // fall through
1403       case Bytecodes::_l2f: // fall through
1404       case Bytecodes::_l2d: tos_in = ltos; break;
1405       case Bytecodes::_f2i: // fall through
1406       case Bytecodes::_f2l: // fall through
1407       case Bytecodes::_f2d: tos_in = ftos; break;
1408       case Bytecodes::_d2i: // fall through
1409       case Bytecodes::_d2l: // fall through
1410       case Bytecodes::_d2f: tos_in = dtos; break;
1411       default             : ShouldNotReachHere();
1412     }
1413     switch (bytecode()) {
1414       case Bytecodes::_l2i: // fall through
1415       case Bytecodes::_f2i: // fall through
1416       case Bytecodes::_d2i: // fall through
1417       case Bytecodes::_i2b: // fall through
1418       case Bytecodes::_i2c: // fall through
1419       case Bytecodes::_i2s: tos_out = itos; break;
1420       case Bytecodes::_i2l: // fall through
1421       case Bytecodes::_f2l: // fall through
1422       case Bytecodes::_d2l: tos_out = ltos; break;
1423       case Bytecodes::_i2f: // fall through
1424       case Bytecodes::_l2f: // fall through
1425       case Bytecodes::_d2f: tos_out = ftos; break;
1426       case Bytecodes::_i2d: // fall through
1427       case Bytecodes::_l2d: // fall through
1428       case Bytecodes::_f2d: tos_out = dtos; break;
1429       default             : ShouldNotReachHere();
1430     }
1431     transition(tos_in, tos_out);
1432   }
1433 #endif // ASSERT
1434 
1435   // Conversion
1436   // (Note: use push(rcx)/pop(rcx) for 1/2-word stack-ptr manipulation)
1437   switch (bytecode()) {
1438     case Bytecodes::_i2l:
1439       __ extend_sign(rdx, rax);
1440       break;
1441     case Bytecodes::_i2f:
1442       __ push(rax);          // store int on tos
1443       __ fild_s(at_rsp());   // load int to ST0
1444       __ f2ieee();           // truncate to float size
1445       __ pop(rcx);           // adjust rsp
1446       break;
1447     case Bytecodes::_i2d:
1448       __ push(rax);          // add one slot for d2ieee()
1449       __ push(rax);          // store int on tos
1450       __ fild_s(at_rsp());   // load int to ST0
1451       __ d2ieee();           // truncate to double size
1452       __ pop(rcx);           // adjust rsp
1453       __ pop(rcx);
1454       break;
1455     case Bytecodes::_i2b:
1456       __ shll(rax, 24);      // truncate upper 24 bits
1457       __ sarl(rax, 24);      // and sign-extend byte
1458       LP64_ONLY(__ movsbl(rax, rax));
1459       break;
1460     case Bytecodes::_i2c:
1461       __ andl(rax, 0xFFFF);  // truncate upper 16 bits
1462       LP64_ONLY(__ movzwl(rax, rax));
1463       break;
1464     case Bytecodes::_i2s:
1465       __ shll(rax, 16);      // truncate upper 16 bits
1466       __ sarl(rax, 16);      // and sign-extend short
1467       LP64_ONLY(__ movswl(rax, rax));
1468       break;
1469     case Bytecodes::_l2i:
1470       /* nothing to do */
1471       break;
1472     case Bytecodes::_l2f:
1473       __ push(rdx);          // store long on tos
1474       __ push(rax);
1475       __ fild_d(at_rsp());   // load long to ST0
1476       __ f2ieee();           // truncate to float size
1477       __ pop(rcx);           // adjust rsp
1478       __ pop(rcx);
1479       break;
1480     case Bytecodes::_l2d:
1481       __ push(rdx);          // store long on tos
1482       __ push(rax);
1483       __ fild_d(at_rsp());   // load long to ST0
1484       __ d2ieee();           // truncate to double size
1485       __ pop(rcx);           // adjust rsp
1486       __ pop(rcx);
1487       break;
1488     case Bytecodes::_f2i:
1489       __ push(rcx);          // reserve space for argument
1490       __ fstp_s(at_rsp());   // pass float argument on stack
1491       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1492       break;
1493     case Bytecodes::_f2l:
1494       __ push(rcx);          // reserve space for argument
1495       __ fstp_s(at_rsp());   // pass float argument on stack
1496       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1497       break;
1498     case Bytecodes::_f2d:
1499       /* nothing to do */
1500       break;
1501     case Bytecodes::_d2i:
1502       __ push(rcx);          // reserve space for argument
1503       __ push(rcx);
1504       __ fstp_d(at_rsp());   // pass double argument on stack
1505       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 2);
1506       break;
1507     case Bytecodes::_d2l:
1508       __ push(rcx);          // reserve space for argument
1509       __ push(rcx);
1510       __ fstp_d(at_rsp());   // pass double argument on stack
1511       __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 2);
1512       break;
1513     case Bytecodes::_d2f:
1514       __ push(rcx);          // reserve space for f2ieee()
1515       __ f2ieee();           // truncate to float size
1516       __ pop(rcx);           // adjust rsp
1517       break;
1518     default             :
1519       ShouldNotReachHere();
1520   }
1521 }
1522 
1523 
1524 void TemplateTable::lcmp() {
1525   transition(ltos, itos);
1526   // y = rdx:rax
1527   __ pop_l(rbx, rcx);             // get x = rcx:rbx
1528   __ lcmp2int(rcx, rbx, rdx, rax);// rcx := cmp(x, y)
1529   __ mov(rax, rcx);
1530 }
1531 
1532 
1533 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1534   if (is_float) {
1535     __ fld_s(at_rsp());
1536   } else {
1537     __ fld_d(at_rsp());
1538     __ pop(rdx);
1539   }
1540   __ pop(rcx);
1541   __ fcmp2int(rax, unordered_result < 0);
1542 }
1543 
1544 
1545 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1546   __ get_method(rcx);           // ECX holds method
1547   __ profile_taken_branch(rax,rbx); // EAX holds updated MDP, EBX holds bumped taken count
1548 
1549   const ByteSize be_offset = MethodCounters::backedge_counter_offset() +
1550                              InvocationCounter::counter_offset();
1551   const ByteSize inv_offset = MethodCounters::invocation_counter_offset() +
1552                               InvocationCounter::counter_offset();
1553 
1554   // Load up EDX with the branch displacement
1555   __ movl(rdx, at_bcp(1));
1556   __ bswapl(rdx);
1557   if (!is_wide) __ sarl(rdx, 16);
1558   LP64_ONLY(__ movslq(rdx, rdx));
1559 
1560 
1561   // Handle all the JSR stuff here, then exit.
1562   // It's much shorter and cleaner than intermingling with the
1563   // non-JSR normal-branch stuff occurring below.
1564   if (is_jsr) {
1565     // Pre-load the next target bytecode into EBX
1566     __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1, 0));
1567 
1568     // compute return address as bci in rax,
1569     __ lea(rax, at_bcp((is_wide ? 5 : 3) - in_bytes(ConstMethod::codes_offset())));
1570     __ subptr(rax, Address(rcx, Method::const_offset()));
1571     // Adjust the bcp in RSI by the displacement in EDX
1572     __ addptr(rsi, rdx);
1573     // Push return address
1574     __ push_i(rax);
1575     // jsr returns vtos
1576     __ dispatch_only_noverify(vtos);
1577     return;
1578   }
1579 
1580   // Normal (non-jsr) branch handling
1581 
1582   // Adjust the bcp in RSI by the displacement in EDX
1583   __ addptr(rsi, rdx);
1584 
1585   assert(UseLoopCounter || !UseOnStackReplacement, "on-stack-replacement requires loop counters");
1586   Label backedge_counter_overflow;
1587   Label profile_method;
1588   Label dispatch;
1589   if (UseLoopCounter) {
1590     // increment backedge counter for backward branches
1591     // rax,: MDO
1592     // rbx,: MDO bumped taken-count
1593     // rcx: method
1594     // rdx: target offset
1595     // rsi: target bcp
1596     // rdi: locals pointer
1597     __ testl(rdx, rdx);             // check if forward or backward branch
1598     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1599 
1600     // check if MethodCounters exists
1601     Label has_counters;
1602     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1603     __ testptr(rax, rax);
1604     __ jcc(Assembler::notZero, has_counters);
1605     __ push(rdx);
1606     __ push(rcx);
1607     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::build_method_counters),
1608                rcx);
1609     __ pop(rcx);
1610     __ pop(rdx);
1611     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1612     __ testptr(rax, rax);
1613     __ jcc(Assembler::zero, dispatch);
1614     __ bind(has_counters);
1615 
1616     if (TieredCompilation) {
1617       Label no_mdo;
1618       int increment = InvocationCounter::count_increment;
1619       int mask = ((1 << Tier0BackedgeNotifyFreqLog) - 1) << InvocationCounter::count_shift;
1620       if (ProfileInterpreter) {
1621         // Are we profiling?
1622         __ movptr(rbx, Address(rcx, in_bytes(Method::method_data_offset())));
1623         __ testptr(rbx, rbx);
1624         __ jccb(Assembler::zero, no_mdo);
1625         // Increment the MDO backedge counter
1626         const Address mdo_backedge_counter(rbx, in_bytes(MethodData::backedge_counter_offset()) +
1627                                                 in_bytes(InvocationCounter::counter_offset()));
1628         __ increment_mask_and_jump(mdo_backedge_counter, increment, mask,
1629                                    rax, false, Assembler::zero, &backedge_counter_overflow);
1630         __ jmp(dispatch);
1631       }
1632       __ bind(no_mdo);
1633       // Increment backedge counter in MethodCounters*
1634       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1635       __ increment_mask_and_jump(Address(rcx, be_offset), increment, mask,
1636                                  rax, false, Assembler::zero, &backedge_counter_overflow);
1637     } else {
1638       // increment counter
1639       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1640       __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1641       __ incrementl(rax, InvocationCounter::count_increment); // increment counter
1642       __ movl(Address(rcx, be_offset), rax);        // store counter
1643 
1644       __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1645 
1646       __ andl(rax, InvocationCounter::count_mask_value);     // and the status bits
1647       __ addl(rax, Address(rcx, be_offset));        // add both counters
1648 
1649       if (ProfileInterpreter) {
1650         // Test to see if we should create a method data oop
1651         __ cmp32(rax,
1652                  ExternalAddress((address) &InvocationCounter::InterpreterProfileLimit));
1653         __ jcc(Assembler::less, dispatch);
1654 
1655         // if no method data exists, go to profile method
1656         __ test_method_data_pointer(rax, profile_method);
1657 
1658         if (UseOnStackReplacement) {
1659           // check for overflow against rbx, which is the MDO taken count
1660           __ cmp32(rbx,
1661                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1662           __ jcc(Assembler::below, dispatch);
1663 
1664           // When ProfileInterpreter is on, the backedge_count comes from the
1665           // MethodData*, which value does not get reset on the call to
1666           // frequency_counter_overflow().  To avoid excessive calls to the overflow
1667           // routine while the method is being compiled, add a second test to make
1668           // sure the overflow function is called only once every overflow_frequency.
1669           const int overflow_frequency = 1024;
1670           __ andptr(rbx, overflow_frequency-1);
1671           __ jcc(Assembler::zero, backedge_counter_overflow);
1672         }
1673       } else {
1674         if (UseOnStackReplacement) {
1675           // check for overflow against rax, which is the sum of the counters
1676           __ cmp32(rax,
1677                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1678           __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1679 
1680         }
1681       }
1682     }
1683     __ bind(dispatch);
1684   }
1685 
1686   // Pre-load the next target bytecode into EBX
1687   __ load_unsigned_byte(rbx, Address(rsi, 0));
1688 
1689   // continue with the bytecode @ target
1690   // rax,: return bci for jsr's, unused otherwise
1691   // rbx,: target bytecode
1692   // rsi: target bcp
1693   __ dispatch_only(vtos);
1694 
1695   if (UseLoopCounter) {
1696     if (ProfileInterpreter) {
1697       // Out-of-line code to allocate method data oop.
1698       __ bind(profile_method);
1699       __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::profile_method));
1700       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1701       __ set_method_data_pointer_for_bcp();
1702       __ jmp(dispatch);
1703     }
1704 
1705     if (UseOnStackReplacement) {
1706 
1707       // invocation counter overflow
1708       __ bind(backedge_counter_overflow);
1709       __ negptr(rdx);
1710       __ addptr(rdx, rsi);        // branch bcp
1711       call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::frequency_counter_overflow), rdx);
1712       __ load_unsigned_byte(rbx, Address(rsi, 0));  // restore target bytecode
1713 
1714       // rax,: osr nmethod (osr ok) or NULL (osr not possible)
1715       // rbx,: target bytecode
1716       // rdx: scratch
1717       // rdi: locals pointer
1718       // rsi: bcp
1719       __ testptr(rax, rax);                      // test result
1720       __ jcc(Assembler::zero, dispatch);         // no osr if null
1721       // nmethod may have been invalidated (VM may block upon call_VM return)
1722       __ movl(rcx, Address(rax, nmethod::entry_bci_offset()));
1723       __ cmpl(rcx, InvalidOSREntryBci);
1724       __ jcc(Assembler::equal, dispatch);
1725 
1726       // We have the address of an on stack replacement routine in rax,
1727       // We need to prepare to execute the OSR method. First we must
1728       // migrate the locals and monitors off of the stack.
1729 
1730       __ mov(rbx, rax);                             // save the nmethod
1731 
1732       const Register thread = rcx;
1733       __ get_thread(thread);
1734       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1735       // rax, is OSR buffer, move it to expected parameter location
1736       __ mov(rcx, rax);
1737 
1738       // pop the interpreter frame
1739       __ movptr(rdx, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1740       __ leave();                                // remove frame anchor
1741       __ pop(rdi);                               // get return address
1742       __ mov(rsp, rdx);                          // set sp to sender sp
1743 
1744       // Align stack pointer for compiled code (note that caller is
1745       // responsible for undoing this fixup by remembering the old SP
1746       // in an rbp,-relative location)
1747       __ andptr(rsp, -(StackAlignmentInBytes));
1748 
1749       // push the (possibly adjusted) return address
1750       __ push(rdi);
1751 
1752       // and begin the OSR nmethod
1753       __ jmp(Address(rbx, nmethod::osr_entry_point_offset()));
1754     }
1755   }
1756 }
1757 
1758 
1759 void TemplateTable::if_0cmp(Condition cc) {
1760   transition(itos, vtos);
1761   // assume branch is more often taken than not (loops use backward branches)
1762   Label not_taken;
1763   __ testl(rax, rax);
1764   __ jcc(j_not(cc), not_taken);
1765   branch(false, false);
1766   __ bind(not_taken);
1767   __ profile_not_taken_branch(rax);
1768 }
1769 
1770 
1771 void TemplateTable::if_icmp(Condition cc) {
1772   transition(itos, vtos);
1773   // assume branch is more often taken than not (loops use backward branches)
1774   Label not_taken;
1775   __ pop_i(rdx);
1776   __ cmpl(rdx, rax);
1777   __ jcc(j_not(cc), not_taken);
1778   branch(false, false);
1779   __ bind(not_taken);
1780   __ profile_not_taken_branch(rax);
1781 }
1782 
1783 
1784 void TemplateTable::if_nullcmp(Condition cc) {
1785   transition(atos, vtos);
1786   // assume branch is more often taken than not (loops use backward branches)
1787   Label not_taken;
1788   __ testptr(rax, rax);
1789   __ jcc(j_not(cc), not_taken);
1790   branch(false, false);
1791   __ bind(not_taken);
1792   __ profile_not_taken_branch(rax);
1793 }
1794 
1795 
1796 void TemplateTable::if_acmp(Condition cc) {
1797   transition(atos, vtos);
1798   // assume branch is more often taken than not (loops use backward branches)
1799   Label not_taken;
1800   __ pop_ptr(rdx);
1801   __ cmpptr(rdx, rax);
1802   __ jcc(j_not(cc), not_taken);
1803   branch(false, false);
1804   __ bind(not_taken);
1805   __ profile_not_taken_branch(rax);
1806 }
1807 
1808 
1809 void TemplateTable::ret() {
1810   transition(vtos, vtos);
1811   locals_index(rbx);
1812   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1813   __ profile_ret(rbx, rcx);
1814   __ get_method(rax);
1815   __ movptr(rsi, Address(rax, Method::const_offset()));
1816   __ lea(rsi, Address(rsi, rbx, Address::times_1,
1817                       ConstMethod::codes_offset()));
1818   __ dispatch_next(vtos);
1819 }
1820 
1821 
1822 void TemplateTable::wide_ret() {
1823   transition(vtos, vtos);
1824   locals_index_wide(rbx);
1825   __ movptr(rbx, iaddress(rbx));                   // get return bci, compute return bcp
1826   __ profile_ret(rbx, rcx);
1827   __ get_method(rax);
1828   __ movptr(rsi, Address(rax, Method::const_offset()));
1829   __ lea(rsi, Address(rsi, rbx, Address::times_1, ConstMethod::codes_offset()));
1830   __ dispatch_next(vtos);
1831 }
1832 
1833 
1834 void TemplateTable::tableswitch() {
1835   Label default_case, continue_execution;
1836   transition(itos, vtos);
1837   // align rsi
1838   __ lea(rbx, at_bcp(wordSize));
1839   __ andptr(rbx, -wordSize);
1840   // load lo & hi
1841   __ movl(rcx, Address(rbx, 1 * wordSize));
1842   __ movl(rdx, Address(rbx, 2 * wordSize));
1843   __ bswapl(rcx);
1844   __ bswapl(rdx);
1845   // check against lo & hi
1846   __ cmpl(rax, rcx);
1847   __ jccb(Assembler::less, default_case);
1848   __ cmpl(rax, rdx);
1849   __ jccb(Assembler::greater, default_case);
1850   // lookup dispatch offset
1851   __ subl(rax, rcx);
1852   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1853   __ profile_switch_case(rax, rbx, rcx);
1854   // continue execution
1855   __ bind(continue_execution);
1856   __ bswapl(rdx);
1857   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1858   __ addptr(rsi, rdx);
1859   __ dispatch_only(vtos);
1860   // handle default
1861   __ bind(default_case);
1862   __ profile_switch_default(rax);
1863   __ movl(rdx, Address(rbx, 0));
1864   __ jmp(continue_execution);
1865 }
1866 
1867 
1868 void TemplateTable::lookupswitch() {
1869   transition(itos, itos);
1870   __ stop("lookupswitch bytecode should have been rewritten");
1871 }
1872 
1873 
1874 void TemplateTable::fast_linearswitch() {
1875   transition(itos, vtos);
1876   Label loop_entry, loop, found, continue_execution;
1877   // bswapl rax, so we can avoid bswapping the table entries
1878   __ bswapl(rax);
1879   // align rsi
1880   __ lea(rbx, at_bcp(wordSize));                // btw: should be able to get rid of this instruction (change offsets below)
1881   __ andptr(rbx, -wordSize);
1882   // set counter
1883   __ movl(rcx, Address(rbx, wordSize));
1884   __ bswapl(rcx);
1885   __ jmpb(loop_entry);
1886   // table search
1887   __ bind(loop);
1888   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * wordSize));
1889   __ jccb(Assembler::equal, found);
1890   __ bind(loop_entry);
1891   __ decrementl(rcx);
1892   __ jcc(Assembler::greaterEqual, loop);
1893   // default case
1894   __ profile_switch_default(rax);
1895   __ movl(rdx, Address(rbx, 0));
1896   __ jmpb(continue_execution);
1897   // entry found -> get offset
1898   __ bind(found);
1899   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * wordSize));
1900   __ profile_switch_case(rcx, rax, rbx);
1901   // continue execution
1902   __ bind(continue_execution);
1903   __ bswapl(rdx);
1904   __ load_unsigned_byte(rbx, Address(rsi, rdx, Address::times_1));
1905   __ addptr(rsi, rdx);
1906   __ dispatch_only(vtos);
1907 }
1908 
1909 
1910 void TemplateTable::fast_binaryswitch() {
1911   transition(itos, vtos);
1912   // Implementation using the following core algorithm:
1913   //
1914   // int binary_search(int key, LookupswitchPair* array, int n) {
1915   //   // Binary search according to "Methodik des Programmierens" by
1916   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1917   //   int i = 0;
1918   //   int j = n;
1919   //   while (i+1 < j) {
1920   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1921   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1922   //     // where a stands for the array and assuming that the (inexisting)
1923   //     // element a[n] is infinitely big.
1924   //     int h = (i + j) >> 1;
1925   //     // i < h < j
1926   //     if (key < array[h].fast_match()) {
1927   //       j = h;
1928   //     } else {
1929   //       i = h;
1930   //     }
1931   //   }
1932   //   // R: a[i] <= key < a[i+1] or Q
1933   //   // (i.e., if key is within array, i is the correct index)
1934   //   return i;
1935   // }
1936 
1937   // register allocation
1938   const Register key   = rax;                    // already set (tosca)
1939   const Register array = rbx;
1940   const Register i     = rcx;
1941   const Register j     = rdx;
1942   const Register h     = rdi;                    // needs to be restored
1943   const Register temp  = rsi;
1944   // setup array
1945   __ save_bcp();
1946 
1947   __ lea(array, at_bcp(3*wordSize));             // btw: should be able to get rid of this instruction (change offsets below)
1948   __ andptr(array, -wordSize);
1949   // initialize i & j
1950   __ xorl(i, i);                                 // i = 0;
1951   __ movl(j, Address(array, -wordSize));         // j = length(array);
1952   // Convert j into native byteordering
1953   __ bswapl(j);
1954   // and start
1955   Label entry;
1956   __ jmp(entry);
1957 
1958   // binary search loop
1959   { Label loop;
1960     __ bind(loop);
1961     // int h = (i + j) >> 1;
1962     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
1963     __ sarl(h, 1);                               // h = (i + j) >> 1;
1964     // if (key < array[h].fast_match()) {
1965     //   j = h;
1966     // } else {
1967     //   i = h;
1968     // }
1969     // Convert array[h].match to native byte-ordering before compare
1970     __ movl(temp, Address(array, h, Address::times_8, 0*wordSize));
1971     __ bswapl(temp);
1972     __ cmpl(key, temp);
1973     // j = h if (key <  array[h].fast_match())
1974     __ cmov32(Assembler::less        , j, h);
1975     // i = h if (key >= array[h].fast_match())
1976     __ cmov32(Assembler::greaterEqual, i, h);
1977     // while (i+1 < j)
1978     __ bind(entry);
1979     __ leal(h, Address(i, 1));                   // i+1
1980     __ cmpl(h, j);                               // i+1 < j
1981     __ jcc(Assembler::less, loop);
1982   }
1983 
1984   // end of binary search, result index is i (must check again!)
1985   Label default_case;
1986   // Convert array[i].match to native byte-ordering before compare
1987   __ movl(temp, Address(array, i, Address::times_8, 0*wordSize));
1988   __ bswapl(temp);
1989   __ cmpl(key, temp);
1990   __ jcc(Assembler::notEqual, default_case);
1991 
1992   // entry found -> j = offset
1993   __ movl(j , Address(array, i, Address::times_8, 1*wordSize));
1994   __ profile_switch_case(i, key, array);
1995   __ bswapl(j);
1996   LP64_ONLY(__ movslq(j, j));
1997   __ restore_bcp();
1998   __ restore_locals();                           // restore rdi
1999   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2000 
2001   __ addptr(rsi, j);
2002   __ dispatch_only(vtos);
2003 
2004   // default case -> j = default offset
2005   __ bind(default_case);
2006   __ profile_switch_default(i);
2007   __ movl(j, Address(array, -2*wordSize));
2008   __ bswapl(j);
2009   LP64_ONLY(__ movslq(j, j));
2010   __ restore_bcp();
2011   __ restore_locals();                           // restore rdi
2012   __ load_unsigned_byte(rbx, Address(rsi, j, Address::times_1));
2013   __ addptr(rsi, j);
2014   __ dispatch_only(vtos);
2015 }
2016 
2017 
2018 void TemplateTable::_return(TosState state) {
2019   transition(state, state);
2020   assert(_desc->calls_vm(), "inconsistent calls_vm information"); // call in remove_activation
2021 
2022   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2023     assert(state == vtos, "only valid state");
2024     __ movptr(rax, aaddress(0));
2025     __ load_klass(rdi, rax);
2026     __ movl(rdi, Address(rdi, Klass::access_flags_offset()));
2027     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2028     Label skip_register_finalizer;
2029     __ jcc(Assembler::zero, skip_register_finalizer);
2030 
2031     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), rax);
2032 
2033     __ bind(skip_register_finalizer);
2034   }
2035 
2036   __ remove_activation(state, rsi);
2037   __ jmp(rsi);
2038 }
2039 
2040 
2041 // ----------------------------------------------------------------------------
2042 // Volatile variables demand their effects be made known to all CPU's in
2043 // order.  Store buffers on most chips allow reads & writes to reorder; the
2044 // JMM's ReadAfterWrite.java test fails in -Xint mode without some kind of
2045 // memory barrier (i.e., it's not sufficient that the interpreter does not
2046 // reorder volatile references, the hardware also must not reorder them).
2047 //
2048 // According to the new Java Memory Model (JMM):
2049 // (1) All volatiles are serialized wrt to each other.
2050 // ALSO reads & writes act as aquire & release, so:
2051 // (2) A read cannot let unrelated NON-volatile memory refs that happen after
2052 // the read float up to before the read.  It's OK for non-volatile memory refs
2053 // that happen before the volatile read to float down below it.
2054 // (3) Similar a volatile write cannot let unrelated NON-volatile memory refs
2055 // that happen BEFORE the write float down to after the write.  It's OK for
2056 // non-volatile memory refs that happen after the volatile write to float up
2057 // before it.
2058 //
2059 // We only put in barriers around volatile refs (they are expensive), not
2060 // _between_ memory refs (that would require us to track the flavor of the
2061 // previous memory refs).  Requirements (2) and (3) require some barriers
2062 // before volatile stores and after volatile loads.  These nearly cover
2063 // requirement (1) but miss the volatile-store-volatile-load case.  This final
2064 // case is placed after volatile-stores although it could just as well go
2065 // before volatile-loads.
2066 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits order_constraint ) {
2067   // Helper function to insert a is-volatile test and memory barrier
2068   if( !os::is_MP() ) return;    // Not needed on single CPU
2069   __ membar(order_constraint);
2070 }
2071 
2072 void TemplateTable::resolve_cache_and_index(int byte_no,
2073                                             Register Rcache,
2074                                             Register index,
2075                                             size_t index_size) {
2076   const Register temp = rbx;
2077   assert_different_registers(Rcache, index, temp);
2078 
2079   Label resolved;
2080     assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
2081     __ get_cache_and_index_and_bytecode_at_bcp(Rcache, index, temp, byte_no, 1, index_size);
2082     __ cmpl(temp, (int) bytecode());  // have we resolved this bytecode?
2083     __ jcc(Assembler::equal, resolved);
2084 
2085   // resolve first time through
2086   address entry;
2087   switch (bytecode()) {
2088     case Bytecodes::_getstatic      : // fall through
2089     case Bytecodes::_putstatic      : // fall through
2090     case Bytecodes::_getfield       : // fall through
2091     case Bytecodes::_putfield       : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);        break;
2092     case Bytecodes::_invokevirtual  : // fall through
2093     case Bytecodes::_invokespecial  : // fall through
2094     case Bytecodes::_invokestatic   : // fall through
2095     case Bytecodes::_invokeinterface: entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);         break;
2096     case Bytecodes::_invokehandle   : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokehandle);   break;
2097     case Bytecodes::_invokedynamic  : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);  break;
2098     default:
2099       fatal(err_msg("unexpected bytecode: %s", Bytecodes::name(bytecode())));
2100       break;
2101   }
2102   __ movl(temp, (int)bytecode());
2103   __ call_VM(noreg, entry, temp);
2104   // Update registers with resolved info
2105   __ get_cache_and_index_at_bcp(Rcache, index, 1, index_size);
2106   __ bind(resolved);
2107 }
2108 
2109 
2110 // The cache and index registers must be set before call
2111 void TemplateTable::load_field_cp_cache_entry(Register obj,
2112                                               Register cache,
2113                                               Register index,
2114                                               Register off,
2115                                               Register flags,
2116                                               bool is_static = false) {
2117   assert_different_registers(cache, index, flags, off);
2118 
2119   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2120   // Field offset
2121   __ movptr(off, Address(cache, index, Address::times_ptr,
2122                          in_bytes(cp_base_offset + ConstantPoolCacheEntry::f2_offset())));
2123   // Flags
2124   __ movl(flags, Address(cache, index, Address::times_ptr,
2125            in_bytes(cp_base_offset + ConstantPoolCacheEntry::flags_offset())));
2126 
2127   // klass overwrite register
2128   if (is_static) {
2129     __ movptr(obj, Address(cache, index, Address::times_ptr,
2130                            in_bytes(cp_base_offset + ConstantPoolCacheEntry::f1_offset())));
2131     const int mirror_offset = in_bytes(Klass::java_mirror_offset());
2132     __ movptr(obj, Address(obj, mirror_offset));
2133   }
2134 }
2135 
2136 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2137                                                Register method,
2138                                                Register itable_index,
2139                                                Register flags,
2140                                                bool is_invokevirtual,
2141                                                bool is_invokevfinal, /*unused*/
2142                                                bool is_invokedynamic) {
2143   // setup registers
2144   const Register cache = rcx;
2145   const Register index = rdx;
2146   assert_different_registers(method, flags);
2147   assert_different_registers(method, cache, index);
2148   assert_different_registers(itable_index, flags);
2149   assert_different_registers(itable_index, cache, index);
2150   // determine constant pool cache field offsets
2151   assert(is_invokevirtual == (byte_no == f2_byte), "is_invokevirtual flag redundant");
2152   const int method_offset = in_bytes(
2153     ConstantPoolCache::base_offset() +
2154       ((byte_no == f2_byte)
2155        ? ConstantPoolCacheEntry::f2_offset()
2156        : ConstantPoolCacheEntry::f1_offset()));
2157   const int flags_offset = in_bytes(ConstantPoolCache::base_offset() +
2158                                     ConstantPoolCacheEntry::flags_offset());
2159   // access constant pool cache fields
2160   const int index_offset = in_bytes(ConstantPoolCache::base_offset() +
2161                                     ConstantPoolCacheEntry::f2_offset());
2162 
2163   size_t index_size = (is_invokedynamic ? sizeof(u4) : sizeof(u2));
2164   resolve_cache_and_index(byte_no, cache, index, index_size);
2165     __ movptr(method, Address(cache, index, Address::times_ptr, method_offset));
2166 
2167   if (itable_index != noreg) {
2168     __ movptr(itable_index, Address(cache, index, Address::times_ptr, index_offset));
2169   }
2170   __ movl(flags, Address(cache, index, Address::times_ptr, flags_offset));
2171 }
2172 
2173 
2174 // The registers cache and index expected to be set before call.
2175 // Correct values of the cache and index registers are preserved.
2176 void TemplateTable::jvmti_post_field_access(Register cache,
2177                                             Register index,
2178                                             bool is_static,
2179                                             bool has_tos) {
2180   if (JvmtiExport::can_post_field_access()) {
2181     // Check to see if a field access watch has been set before we take
2182     // the time to call into the VM.
2183     Label L1;
2184     assert_different_registers(cache, index, rax);
2185     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2186     __ testl(rax,rax);
2187     __ jcc(Assembler::zero, L1);
2188 
2189     // cache entry pointer
2190     __ addptr(cache, in_bytes(ConstantPoolCache::base_offset()));
2191     __ shll(index, LogBytesPerWord);
2192     __ addptr(cache, index);
2193     if (is_static) {
2194       __ xorptr(rax, rax);      // NULL object reference
2195     } else {
2196       __ pop(atos);         // Get the object
2197       __ verify_oop(rax);
2198       __ push(atos);        // Restore stack state
2199     }
2200     // rax,:   object pointer or NULL
2201     // cache: cache entry pointer
2202     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access),
2203                rax, cache);
2204     __ get_cache_and_index_at_bcp(cache, index, 1);
2205     __ bind(L1);
2206   }
2207 }
2208 
2209 void TemplateTable::pop_and_check_object(Register r) {
2210   __ pop_ptr(r);
2211   __ null_check(r);  // for field access must check obj.
2212   __ verify_oop(r);
2213 }
2214 
2215 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2216   transition(vtos, vtos);
2217 
2218   const Register cache = rcx;
2219   const Register index = rdx;
2220   const Register obj   = rcx;
2221   const Register off   = rbx;
2222   const Register flags = rax;
2223 
2224   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2225   jvmti_post_field_access(cache, index, is_static, false);
2226   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2227 
2228   if (!is_static) pop_and_check_object(obj);
2229 
2230   const Address lo(obj, off, Address::times_1, 0*wordSize);
2231   const Address hi(obj, off, Address::times_1, 1*wordSize);
2232 
2233   Label Done, notByte, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2234 
2235   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2236   assert(btos == 0, "change code, btos != 0");
2237   // btos
2238   __ andptr(flags, ConstantPoolCacheEntry::tos_state_mask);
2239   __ jcc(Assembler::notZero, notByte);
2240 
2241   __ load_signed_byte(rax, lo );
2242   __ push(btos);
2243   // Rewrite bytecode to be faster
2244   if (!is_static) {
2245     patch_bytecode(Bytecodes::_fast_bgetfield, rcx, rbx);
2246   }
2247   __ jmp(Done);
2248 
2249   __ bind(notByte);
2250   // itos
2251   __ cmpl(flags, itos );
2252   __ jcc(Assembler::notEqual, notInt);
2253 
2254   __ movl(rax, lo );
2255   __ push(itos);
2256   // Rewrite bytecode to be faster
2257   if (!is_static) {
2258     patch_bytecode(Bytecodes::_fast_igetfield, rcx, rbx);
2259   }
2260   __ jmp(Done);
2261 
2262   __ bind(notInt);
2263   // atos
2264   __ cmpl(flags, atos );
2265   __ jcc(Assembler::notEqual, notObj);
2266 
2267   __ movl(rax, lo );
2268   __ push(atos);
2269   if (!is_static) {
2270     patch_bytecode(Bytecodes::_fast_agetfield, rcx, rbx);
2271   }
2272   __ jmp(Done);
2273 
2274   __ bind(notObj);
2275   // ctos
2276   __ cmpl(flags, ctos );
2277   __ jcc(Assembler::notEqual, notChar);
2278 
2279   __ load_unsigned_short(rax, lo );
2280   __ push(ctos);
2281   if (!is_static) {
2282     patch_bytecode(Bytecodes::_fast_cgetfield, rcx, rbx);
2283   }
2284   __ jmp(Done);
2285 
2286   __ bind(notChar);
2287   // stos
2288   __ cmpl(flags, stos );
2289   __ jcc(Assembler::notEqual, notShort);
2290 
2291   __ load_signed_short(rax, lo );
2292   __ push(stos);
2293   if (!is_static) {
2294     patch_bytecode(Bytecodes::_fast_sgetfield, rcx, rbx);
2295   }
2296   __ jmp(Done);
2297 
2298   __ bind(notShort);
2299   // ltos
2300   __ cmpl(flags, ltos );
2301   __ jcc(Assembler::notEqual, notLong);
2302 
2303   // Generate code as if volatile.  There just aren't enough registers to
2304   // save that information and this code is faster than the test.
2305   __ fild_d(lo);                // Must load atomically
2306   __ subptr(rsp,2*wordSize);    // Make space for store
2307   __ fistp_d(Address(rsp,0));
2308   __ pop(rax);
2309   __ pop(rdx);
2310 
2311   __ push(ltos);
2312   // Don't rewrite to _fast_lgetfield for potential volatile case.
2313   __ jmp(Done);
2314 
2315   __ bind(notLong);
2316   // ftos
2317   __ cmpl(flags, ftos );
2318   __ jcc(Assembler::notEqual, notFloat);
2319 
2320   __ fld_s(lo);
2321   __ push(ftos);
2322   if (!is_static) {
2323     patch_bytecode(Bytecodes::_fast_fgetfield, rcx, rbx);
2324   }
2325   __ jmp(Done);
2326 
2327   __ bind(notFloat);
2328   // dtos
2329   __ cmpl(flags, dtos );
2330   __ jcc(Assembler::notEqual, notDouble);
2331 
2332   __ fld_d(lo);
2333   __ push(dtos);
2334   if (!is_static) {
2335     patch_bytecode(Bytecodes::_fast_dgetfield, rcx, rbx);
2336   }
2337   __ jmpb(Done);
2338 
2339   __ bind(notDouble);
2340 
2341   __ stop("Bad state");
2342 
2343   __ bind(Done);
2344   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2345   // volatile_barrier( );
2346 }
2347 
2348 
2349 void TemplateTable::getfield(int byte_no) {
2350   getfield_or_static(byte_no, false);
2351 }
2352 
2353 
2354 void TemplateTable::getstatic(int byte_no) {
2355   getfield_or_static(byte_no, true);
2356 }
2357 
2358 // The registers cache and index expected to be set before call.
2359 // The function may destroy various registers, just not the cache and index registers.
2360 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2361 
2362   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2363 
2364   if (JvmtiExport::can_post_field_modification()) {
2365     // Check to see if a field modification watch has been set before we take
2366     // the time to call into the VM.
2367     Label L1;
2368     assert_different_registers(cache, index, rax);
2369     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2370     __ testl(rax, rax);
2371     __ jcc(Assembler::zero, L1);
2372 
2373     // The cache and index registers have been already set.
2374     // This allows to eliminate this call but the cache and index
2375     // registers have to be correspondingly used after this line.
2376     __ get_cache_and_index_at_bcp(rax, rdx, 1);
2377 
2378     if (is_static) {
2379       // Life is simple.  Null out the object pointer.
2380       __ xorptr(rbx, rbx);
2381     } else {
2382       // Life is harder. The stack holds the value on top, followed by the object.
2383       // We don't know the size of the value, though; it could be one or two words
2384       // depending on its type. As a result, we must find the type to determine where
2385       // the object is.
2386       Label two_word, valsize_known;
2387       __ movl(rcx, Address(rax, rdx, Address::times_ptr, in_bytes(cp_base_offset +
2388                                    ConstantPoolCacheEntry::flags_offset())));
2389       __ mov(rbx, rsp);
2390       __ shrl(rcx, ConstantPoolCacheEntry::tos_state_shift);
2391       // Make sure we don't need to mask rcx after the above shift
2392       ConstantPoolCacheEntry::verify_tos_state_shift();
2393       __ cmpl(rcx, ltos);
2394       __ jccb(Assembler::equal, two_word);
2395       __ cmpl(rcx, dtos);
2396       __ jccb(Assembler::equal, two_word);
2397       __ addptr(rbx, Interpreter::expr_offset_in_bytes(1)); // one word jvalue (not ltos, dtos)
2398       __ jmpb(valsize_known);
2399 
2400       __ bind(two_word);
2401       __ addptr(rbx, Interpreter::expr_offset_in_bytes(2)); // two words jvalue
2402 
2403       __ bind(valsize_known);
2404       // setup object pointer
2405       __ movptr(rbx, Address(rbx, 0));
2406     }
2407     // cache entry pointer
2408     __ addptr(rax, in_bytes(cp_base_offset));
2409     __ shll(rdx, LogBytesPerWord);
2410     __ addptr(rax, rdx);
2411     // object (tos)
2412     __ mov(rcx, rsp);
2413     // rbx,: object pointer set up above (NULL if static)
2414     // rax,: cache entry pointer
2415     // rcx: jvalue object on the stack
2416     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification),
2417                rbx, rax, rcx);
2418     __ get_cache_and_index_at_bcp(cache, index, 1);
2419     __ bind(L1);
2420   }
2421 }
2422 
2423 
2424 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2425   transition(vtos, vtos);
2426 
2427   const Register cache = rcx;
2428   const Register index = rdx;
2429   const Register obj   = rcx;
2430   const Register off   = rbx;
2431   const Register flags = rax;
2432 
2433   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2434   jvmti_post_field_mod(cache, index, is_static);
2435   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2436 
2437   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2438   // volatile_barrier( );
2439 
2440   Label notVolatile, Done;
2441   __ movl(rdx, flags);
2442   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2443   __ andl(rdx, 0x1);
2444 
2445   // field addresses
2446   const Address lo(obj, off, Address::times_1, 0*wordSize);
2447   const Address hi(obj, off, Address::times_1, 1*wordSize);
2448 
2449   Label notByte, notInt, notShort, notChar, notLong, notFloat, notObj, notDouble;
2450 
2451   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2452   assert(btos == 0, "change code, btos != 0");
2453   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2454   __ jcc(Assembler::notZero, notByte);
2455 
2456   // btos
2457   {
2458     __ pop(btos);
2459     if (!is_static) pop_and_check_object(obj);
2460     __ movb(lo, rax);
2461     if (!is_static) {
2462       patch_bytecode(Bytecodes::_fast_bputfield, rcx, rbx, true, byte_no);
2463     }
2464     __ jmp(Done);
2465   }
2466 
2467   __ bind(notByte);
2468   __ cmpl(flags, itos);
2469   __ jcc(Assembler::notEqual, notInt);
2470 
2471   // itos
2472   {
2473     __ pop(itos);
2474     if (!is_static) pop_and_check_object(obj);
2475     __ movl(lo, rax);
2476     if (!is_static) {
2477       patch_bytecode(Bytecodes::_fast_iputfield, rcx, rbx, true, byte_no);
2478     }
2479     __ jmp(Done);
2480   }
2481 
2482   __ bind(notInt);
2483   __ cmpl(flags, atos);
2484   __ jcc(Assembler::notEqual, notObj);
2485 
2486   // atos
2487   {
2488     __ pop(atos);
2489     if (!is_static) pop_and_check_object(obj);
2490     do_oop_store(_masm, lo, rax, _bs->kind(), false);
2491     if (!is_static) {
2492       patch_bytecode(Bytecodes::_fast_aputfield, rcx, rbx, true, byte_no);
2493     }
2494     __ jmp(Done);
2495   }
2496 
2497   __ bind(notObj);
2498   __ cmpl(flags, ctos);
2499   __ jcc(Assembler::notEqual, notChar);
2500 
2501   // ctos
2502   {
2503     __ pop(ctos);
2504     if (!is_static) pop_and_check_object(obj);
2505     __ movw(lo, rax);
2506     if (!is_static) {
2507       patch_bytecode(Bytecodes::_fast_cputfield, rcx, rbx, true, byte_no);
2508     }
2509     __ jmp(Done);
2510   }
2511 
2512   __ bind(notChar);
2513   __ cmpl(flags, stos);
2514   __ jcc(Assembler::notEqual, notShort);
2515 
2516   // stos
2517   {
2518     __ pop(stos);
2519     if (!is_static) pop_and_check_object(obj);
2520     __ movw(lo, rax);
2521     if (!is_static) {
2522       patch_bytecode(Bytecodes::_fast_sputfield, rcx, rbx, true, byte_no);
2523     }
2524     __ jmp(Done);
2525   }
2526 
2527   __ bind(notShort);
2528   __ cmpl(flags, ltos);
2529   __ jcc(Assembler::notEqual, notLong);
2530 
2531   // ltos
2532   {
2533     Label notVolatileLong;
2534     __ testl(rdx, rdx);
2535     __ jcc(Assembler::zero, notVolatileLong);
2536 
2537     __ pop(ltos);  // overwrites rdx, do this after testing volatile.
2538     if (!is_static) pop_and_check_object(obj);
2539 
2540     // Replace with real volatile test
2541     __ push(rdx);
2542     __ push(rax);                 // Must update atomically with FIST
2543     __ fild_d(Address(rsp,0));    // So load into FPU register
2544     __ fistp_d(lo);               // and put into memory atomically
2545     __ addptr(rsp, 2*wordSize);
2546     // volatile_barrier();
2547     volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2548                                                  Assembler::StoreStore));
2549     // Don't rewrite volatile version
2550     __ jmp(notVolatile);
2551 
2552     __ bind(notVolatileLong);
2553 
2554     __ pop(ltos);  // overwrites rdx
2555     if (!is_static) pop_and_check_object(obj);
2556     NOT_LP64(__ movptr(hi, rdx));
2557     __ movptr(lo, rax);
2558     if (!is_static) {
2559       patch_bytecode(Bytecodes::_fast_lputfield, rcx, rbx, true, byte_no);
2560     }
2561     __ jmp(notVolatile);
2562   }
2563 
2564   __ bind(notLong);
2565   __ cmpl(flags, ftos);
2566   __ jcc(Assembler::notEqual, notFloat);
2567 
2568   // ftos
2569   {
2570     __ pop(ftos);
2571     if (!is_static) pop_and_check_object(obj);
2572     __ fstp_s(lo);
2573     if (!is_static) {
2574       patch_bytecode(Bytecodes::_fast_fputfield, rcx, rbx, true, byte_no);
2575     }
2576     __ jmp(Done);
2577   }
2578 
2579   __ bind(notFloat);
2580 #ifdef ASSERT
2581   __ cmpl(flags, dtos);
2582   __ jcc(Assembler::notEqual, notDouble);
2583 #endif
2584 
2585   // dtos
2586   {
2587     __ pop(dtos);
2588     if (!is_static) pop_and_check_object(obj);
2589     __ fstp_d(lo);
2590     if (!is_static) {
2591       patch_bytecode(Bytecodes::_fast_dputfield, rcx, rbx, true, byte_no);
2592     }
2593     __ jmp(Done);
2594   }
2595 
2596 #ifdef ASSERT
2597   __ bind(notDouble);
2598   __ stop("Bad state");
2599 #endif
2600 
2601   __ bind(Done);
2602 
2603   // Check for volatile store
2604   __ testl(rdx, rdx);
2605   __ jcc(Assembler::zero, notVolatile);
2606   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2607                                                Assembler::StoreStore));
2608   __ bind(notVolatile);
2609 }
2610 
2611 
2612 void TemplateTable::putfield(int byte_no) {
2613   putfield_or_static(byte_no, false);
2614 }
2615 
2616 
2617 void TemplateTable::putstatic(int byte_no) {
2618   putfield_or_static(byte_no, true);
2619 }
2620 
2621 void TemplateTable::jvmti_post_fast_field_mod() {
2622   if (JvmtiExport::can_post_field_modification()) {
2623     // Check to see if a field modification watch has been set before we take
2624     // the time to call into the VM.
2625     Label L2;
2626      __ mov32(rcx, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2627      __ testl(rcx,rcx);
2628      __ jcc(Assembler::zero, L2);
2629      __ pop_ptr(rbx);               // copy the object pointer from tos
2630      __ verify_oop(rbx);
2631      __ push_ptr(rbx);              // put the object pointer back on tos
2632 
2633      // Save tos values before call_VM() clobbers them. Since we have
2634      // to do it for every data type, we use the saved values as the
2635      // jvalue object.
2636      switch (bytecode()) {          // load values into the jvalue object
2637      case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
2638      case Bytecodes::_fast_bputfield: // fall through
2639      case Bytecodes::_fast_sputfield: // fall through
2640      case Bytecodes::_fast_cputfield: // fall through
2641      case Bytecodes::_fast_iputfield: __ push_i(rax); break;
2642      case Bytecodes::_fast_dputfield: __ push_d(); break;
2643      case Bytecodes::_fast_fputfield: __ push_f(); break;
2644      case Bytecodes::_fast_lputfield: __ push_l(rax); break;
2645 
2646      default:
2647        ShouldNotReachHere();
2648      }
2649      __ mov(rcx, rsp);              // points to jvalue on the stack
2650      // access constant pool cache entry
2651      __ get_cache_entry_pointer_at_bcp(rax, rdx, 1);
2652      __ verify_oop(rbx);
2653      // rbx,: object pointer copied above
2654      // rax,: cache entry pointer
2655      // rcx: jvalue object on the stack
2656      __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification), rbx, rax, rcx);
2657 
2658      switch (bytecode()) {             // restore tos values
2659      case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
2660      case Bytecodes::_fast_bputfield: // fall through
2661      case Bytecodes::_fast_sputfield: // fall through
2662      case Bytecodes::_fast_cputfield: // fall through
2663      case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
2664      case Bytecodes::_fast_dputfield: __ pop_d(); break;
2665      case Bytecodes::_fast_fputfield: __ pop_f(); break;
2666      case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
2667      }
2668      __ bind(L2);
2669   }
2670 }
2671 
2672 void TemplateTable::fast_storefield(TosState state) {
2673   transition(state, vtos);
2674 
2675   ByteSize base = ConstantPoolCache::base_offset();
2676 
2677   jvmti_post_fast_field_mod();
2678 
2679   // access constant pool cache
2680   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2681 
2682   // test for volatile with rdx but rdx is tos register for lputfield.
2683   if (bytecode() == Bytecodes::_fast_lputfield) __ push(rdx);
2684   __ movl(rdx, Address(rcx, rbx, Address::times_ptr, in_bytes(base +
2685                        ConstantPoolCacheEntry::flags_offset())));
2686 
2687   // replace index with field offset from cache entry
2688   __ movptr(rbx, Address(rcx, rbx, Address::times_ptr, in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2689 
2690   // Doug Lea believes this is not needed with current Sparcs (TSO) and Intel (PSO).
2691   // volatile_barrier( );
2692 
2693   Label notVolatile, Done;
2694   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2695   __ andl(rdx, 0x1);
2696   // Check for volatile store
2697   __ testl(rdx, rdx);
2698   __ jcc(Assembler::zero, notVolatile);
2699 
2700   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2701 
2702   // Get object from stack
2703   pop_and_check_object(rcx);
2704 
2705   // field addresses
2706   const Address lo(rcx, rbx, Address::times_1, 0*wordSize);
2707   const Address hi(rcx, rbx, Address::times_1, 1*wordSize);
2708 
2709   // access field
2710   switch (bytecode()) {
2711     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2712     case Bytecodes::_fast_sputfield: // fall through
2713     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2714     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2715     case Bytecodes::_fast_lputfield:
2716       NOT_LP64(__ movptr(hi, rdx));
2717       __ movptr(lo, rax);
2718       break;
2719     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2720     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2721     case Bytecodes::_fast_aputfield: {
2722       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2723       break;
2724     }
2725     default:
2726       ShouldNotReachHere();
2727   }
2728 
2729   Label done;
2730   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2731                                                Assembler::StoreStore));
2732   // Barriers are so large that short branch doesn't reach!
2733   __ jmp(done);
2734 
2735   // Same code as above, but don't need rdx to test for volatile.
2736   __ bind(notVolatile);
2737 
2738   if (bytecode() == Bytecodes::_fast_lputfield) __ pop(rdx);
2739 
2740   // Get object from stack
2741   pop_and_check_object(rcx);
2742 
2743   // access field
2744   switch (bytecode()) {
2745     case Bytecodes::_fast_bputfield: __ movb(lo, rax); break;
2746     case Bytecodes::_fast_sputfield: // fall through
2747     case Bytecodes::_fast_cputfield: __ movw(lo, rax); break;
2748     case Bytecodes::_fast_iputfield: __ movl(lo, rax); break;
2749     case Bytecodes::_fast_lputfield:
2750       NOT_LP64(__ movptr(hi, rdx));
2751       __ movptr(lo, rax);
2752       break;
2753     case Bytecodes::_fast_fputfield: __ fstp_s(lo); break;
2754     case Bytecodes::_fast_dputfield: __ fstp_d(lo); break;
2755     case Bytecodes::_fast_aputfield: {
2756       do_oop_store(_masm, lo, rax, _bs->kind(), false);
2757       break;
2758     }
2759     default:
2760       ShouldNotReachHere();
2761   }
2762   __ bind(done);
2763 }
2764 
2765 
2766 void TemplateTable::fast_accessfield(TosState state) {
2767   transition(atos, state);
2768 
2769   // do the JVMTI work here to avoid disturbing the register state below
2770   if (JvmtiExport::can_post_field_access()) {
2771     // Check to see if a field access watch has been set before we take
2772     // the time to call into the VM.
2773     Label L1;
2774     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2775     __ testl(rcx,rcx);
2776     __ jcc(Assembler::zero, L1);
2777     // access constant pool cache entry
2778     __ get_cache_entry_pointer_at_bcp(rcx, rdx, 1);
2779     __ push_ptr(rax);  // save object pointer before call_VM() clobbers it
2780     __ verify_oop(rax);
2781     // rax,: object pointer copied above
2782     // rcx: cache entry pointer
2783     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access), rax, rcx);
2784     __ pop_ptr(rax);   // restore object pointer
2785     __ bind(L1);
2786   }
2787 
2788   // access constant pool cache
2789   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2790   // replace index with field offset from cache entry
2791   __ movptr(rbx, Address(rcx,
2792                          rbx,
2793                          Address::times_ptr,
2794                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2795 
2796 
2797   // rax,: object
2798   __ verify_oop(rax);
2799   __ null_check(rax);
2800   // field addresses
2801   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2802   const Address hi = Address(rax, rbx, Address::times_1, 1*wordSize);
2803 
2804   // access field
2805   switch (bytecode()) {
2806     case Bytecodes::_fast_bgetfield: __ movsbl(rax, lo );                 break;
2807     case Bytecodes::_fast_sgetfield: __ load_signed_short(rax, lo );      break;
2808     case Bytecodes::_fast_cgetfield: __ load_unsigned_short(rax, lo );    break;
2809     case Bytecodes::_fast_igetfield: __ movl(rax, lo);                    break;
2810     case Bytecodes::_fast_lgetfield: __ stop("should not be rewritten");  break;
2811     case Bytecodes::_fast_fgetfield: __ fld_s(lo);                        break;
2812     case Bytecodes::_fast_dgetfield: __ fld_d(lo);                        break;
2813     case Bytecodes::_fast_agetfield: __ movptr(rax, lo); __ verify_oop(rax); break;
2814     default:
2815       ShouldNotReachHere();
2816   }
2817 
2818   // Doug Lea believes this is not needed with current Sparcs(TSO) and Intel(PSO)
2819   // volatile_barrier( );
2820 }
2821 
2822 void TemplateTable::fast_xaccess(TosState state) {
2823   transition(vtos, state);
2824   // get receiver
2825   __ movptr(rax, aaddress(0));
2826   // access constant pool cache
2827   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2828   __ movptr(rbx, Address(rcx,
2829                          rdx,
2830                          Address::times_ptr,
2831                          in_bytes(ConstantPoolCache::base_offset() + ConstantPoolCacheEntry::f2_offset())));
2832   // make sure exception is reported in correct bcp range (getfield is next instruction)
2833   __ increment(rsi);
2834   __ null_check(rax);
2835   const Address lo = Address(rax, rbx, Address::times_1, 0*wordSize);
2836   if (state == itos) {
2837     __ movl(rax, lo);
2838   } else if (state == atos) {
2839     __ movptr(rax, lo);
2840     __ verify_oop(rax);
2841   } else if (state == ftos) {
2842     __ fld_s(lo);
2843   } else {
2844     ShouldNotReachHere();
2845   }
2846   __ decrement(rsi);
2847 }
2848 
2849 
2850 
2851 //----------------------------------------------------------------------------------------------------
2852 // Calls
2853 
2854 void TemplateTable::count_calls(Register method, Register temp) {
2855   // implemented elsewhere
2856   ShouldNotReachHere();
2857 }
2858 
2859 
2860 void TemplateTable::prepare_invoke(int byte_no,
2861                                    Register method,  // linked method (or i-klass)
2862                                    Register index,   // itable index, MethodType, etc.
2863                                    Register recv,    // if caller wants to see it
2864                                    Register flags    // if caller wants to test it
2865                                    ) {
2866   // determine flags
2867   const Bytecodes::Code code = bytecode();
2868   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
2869   const bool is_invokedynamic    = code == Bytecodes::_invokedynamic;
2870   const bool is_invokehandle     = code == Bytecodes::_invokehandle;
2871   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
2872   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
2873   const bool load_receiver       = (recv  != noreg);
2874   const bool save_flags          = (flags != noreg);
2875   assert(load_receiver == (code != Bytecodes::_invokestatic && code != Bytecodes::_invokedynamic), "");
2876   assert(save_flags    == (is_invokeinterface || is_invokevirtual), "need flags for vfinal");
2877   assert(flags == noreg || flags == rdx, "");
2878   assert(recv  == noreg || recv  == rcx, "");
2879 
2880   // setup registers & access constant pool cache
2881   if (recv  == noreg)  recv  = rcx;
2882   if (flags == noreg)  flags = rdx;
2883   assert_different_registers(method, index, recv, flags);
2884 
2885   // save 'interpreter return address'
2886   __ save_bcp();
2887 
2888   load_invoke_cp_cache_entry(byte_no, method, index, flags, is_invokevirtual, false, is_invokedynamic);
2889 
2890   // maybe push appendix to arguments (just before return address)
2891   if (is_invokedynamic || is_invokehandle) {
2892     Label L_no_push;
2893     __ testl(flags, (1 << ConstantPoolCacheEntry::has_appendix_shift));
2894     __ jccb(Assembler::zero, L_no_push);
2895     // Push the appendix as a trailing parameter.
2896     // This must be done before we get the receiver,
2897     // since the parameter_size includes it.
2898     __ push(rbx);
2899     __ mov(rbx, index);
2900     assert(ConstantPoolCacheEntry::_indy_resolved_references_appendix_offset == 0, "appendix expected at index+0");
2901     __ load_resolved_reference_at_index(index, rbx);
2902     __ pop(rbx);
2903     __ push(index);  // push appendix (MethodType, CallSite, etc.)
2904     __ bind(L_no_push);
2905   }
2906 
2907   // load receiver if needed (note: no return address pushed yet)
2908   if (load_receiver) {
2909     __ movl(recv, flags);
2910     __ andl(recv, ConstantPoolCacheEntry::parameter_size_mask);
2911     const int no_return_pc_pushed_yet = -1;  // argument slot correction before we push return address
2912     const int receiver_is_at_end      = -1;  // back off one slot to get receiver
2913     Address recv_addr = __ argument_address(recv, no_return_pc_pushed_yet + receiver_is_at_end);
2914     __ movptr(recv, recv_addr);
2915     __ verify_oop(recv);
2916   }
2917 
2918   if (save_flags) {
2919     __ mov(rsi, flags);
2920   }
2921 
2922   // compute return type
2923   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2924   // Make sure we don't need to mask flags after the above shift
2925   ConstantPoolCacheEntry::verify_tos_state_shift();
2926   // load return address
2927   {
2928     const address table_addr = (is_invokeinterface || is_invokedynamic) ?
2929         (address)Interpreter::return_5_addrs_by_index_table() :
2930         (address)Interpreter::return_3_addrs_by_index_table();
2931     ExternalAddress table(table_addr);
2932     __ movptr(flags, ArrayAddress(table, Address(noreg, flags, Address::times_ptr)));
2933   }
2934 
2935   // push return address
2936   __ push(flags);
2937 
2938   // Restore flags value from the constant pool cache, and restore rsi
2939   // for later null checks.  rsi is the bytecode pointer
2940   if (save_flags) {
2941     __ mov(flags, rsi);
2942     __ restore_bcp();
2943   }
2944 }
2945 
2946 
2947 void TemplateTable::invokevirtual_helper(Register index,
2948                                          Register recv,
2949                                          Register flags) {
2950   // Uses temporary registers rax, rdx
2951   assert_different_registers(index, recv, rax, rdx);
2952   assert(index == rbx, "");
2953   assert(recv  == rcx, "");
2954 
2955   // Test for an invoke of a final method
2956   Label notFinal;
2957   __ movl(rax, flags);
2958   __ andl(rax, (1 << ConstantPoolCacheEntry::is_vfinal_shift));
2959   __ jcc(Assembler::zero, notFinal);
2960 
2961   const Register method = index;  // method must be rbx
2962   assert(method == rbx,
2963          "Method* must be rbx for interpreter calling convention");
2964 
2965   // do the call - the index is actually the method to call
2966   // that is, f2 is a vtable index if !is_vfinal, else f2 is a Method*
2967 
2968   // It's final, need a null check here!
2969   __ null_check(recv);
2970 
2971   // profile this call
2972   __ profile_final_call(rax);
2973   __ profile_arguments_type(rax, method, rsi, true);
2974 
2975   __ jump_from_interpreted(method, rax);
2976 
2977   __ bind(notFinal);
2978 
2979   // get receiver klass
2980   __ null_check(recv, oopDesc::klass_offset_in_bytes());
2981   __ load_klass(rax, recv);
2982 
2983   // profile this call
2984   __ profile_virtual_call(rax, rdi, rdx);
2985 
2986   // get target Method* & entry point
2987   __ lookup_virtual_method(rax, index, method);
2988   __ profile_arguments_type(rdx, method, rsi, true);
2989   __ jump_from_interpreted(method, rdx);
2990 }
2991 
2992 
2993 void TemplateTable::invokevirtual(int byte_no) {
2994   transition(vtos, vtos);
2995   assert(byte_no == f2_byte, "use this argument");
2996   prepare_invoke(byte_no,
2997                  rbx,    // method or vtable index
2998                  noreg,  // unused itable index
2999                  rcx, rdx); // recv, flags
3000 
3001   // rbx: index
3002   // rcx: receiver
3003   // rdx: flags
3004 
3005   invokevirtual_helper(rbx, rcx, rdx);
3006 }
3007 
3008 
3009 void TemplateTable::invokespecial(int byte_no) {
3010   transition(vtos, vtos);
3011   assert(byte_no == f1_byte, "use this argument");
3012   prepare_invoke(byte_no, rbx, noreg,  // get f1 Method*
3013                  rcx);  // get receiver also for null check
3014   __ verify_oop(rcx);
3015   __ null_check(rcx);
3016   // do the call
3017   __ profile_call(rax);
3018   __ profile_arguments_type(rax, rbx, rsi, false);
3019   __ jump_from_interpreted(rbx, rax);
3020 }
3021 
3022 
3023 void TemplateTable::invokestatic(int byte_no) {
3024   transition(vtos, vtos);
3025   assert(byte_no == f1_byte, "use this argument");
3026   prepare_invoke(byte_no, rbx);  // get f1 Method*
3027   // do the call
3028   __ profile_call(rax);
3029   __ profile_arguments_type(rax, rbx, rsi, false);
3030   __ jump_from_interpreted(rbx, rax);
3031 }
3032 
3033 
3034 void TemplateTable::fast_invokevfinal(int byte_no) {
3035   transition(vtos, vtos);
3036   assert(byte_no == f2_byte, "use this argument");
3037   __ stop("fast_invokevfinal not used on x86");
3038 }
3039 
3040 
3041 void TemplateTable::invokeinterface(int byte_no) {
3042   transition(vtos, vtos);
3043   assert(byte_no == f1_byte, "use this argument");
3044   prepare_invoke(byte_no, rax, rbx,  // get f1 Klass*, f2 itable index
3045                  rcx, rdx); // recv, flags
3046 
3047   // rax: interface klass (from f1)
3048   // rbx: itable index (from f2)
3049   // rcx: receiver
3050   // rdx: flags
3051 
3052   // Special case of invokeinterface called for virtual method of
3053   // java.lang.Object.  See cpCacheOop.cpp for details.
3054   // This code isn't produced by javac, but could be produced by
3055   // another compliant java compiler.
3056   Label notMethod;
3057   __ movl(rdi, rdx);
3058   __ andl(rdi, (1 << ConstantPoolCacheEntry::is_forced_virtual_shift));
3059   __ jcc(Assembler::zero, notMethod);
3060 
3061   invokevirtual_helper(rbx, rcx, rdx);
3062   __ bind(notMethod);
3063 
3064   // Get receiver klass into rdx - also a null check
3065   __ restore_locals();  // restore rdi
3066   __ null_check(rcx, oopDesc::klass_offset_in_bytes());
3067   __ load_klass(rdx, rcx);
3068 
3069   // profile this call
3070   __ profile_virtual_call(rdx, rsi, rdi);
3071 
3072   Label no_such_interface, no_such_method;
3073 
3074   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3075                              rdx, rax, rbx,
3076                              // outputs: method, scan temp. reg
3077                              rbx, rsi,
3078                              no_such_interface);
3079 
3080   // rbx: Method* to call
3081   // rcx: receiver
3082   // Check for abstract method error
3083   // Note: This should be done more efficiently via a throw_abstract_method_error
3084   //       interpreter entry point and a conditional jump to it in case of a null
3085   //       method.
3086   __ testptr(rbx, rbx);
3087   __ jcc(Assembler::zero, no_such_method);
3088 
3089   __ profile_arguments_type(rdx, rbx, rsi, true);
3090 
3091   // do the call
3092   // rcx: receiver
3093   // rbx,: Method*
3094   __ jump_from_interpreted(rbx, rdx);
3095   __ should_not_reach_here();
3096 
3097   // exception handling code follows...
3098   // note: must restore interpreter registers to canonical
3099   //       state for exception handling to work correctly!
3100 
3101   __ bind(no_such_method);
3102   // throw exception
3103   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3104   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3105   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3106   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3107   // the call_VM checks for exception, so we should never return here.
3108   __ should_not_reach_here();
3109 
3110   __ bind(no_such_interface);
3111   // throw exception
3112   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3113   __ restore_bcp();      // rsi must be correct for exception handler   (was destroyed)
3114   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3115   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3116                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3117   // the call_VM checks for exception, so we should never return here.
3118   __ should_not_reach_here();
3119 }
3120 
3121 void TemplateTable::invokehandle(int byte_no) {
3122   transition(vtos, vtos);
3123   assert(byte_no == f1_byte, "use this argument");
3124   const Register rbx_method = rbx;
3125   const Register rax_mtype  = rax;
3126   const Register rcx_recv   = rcx;
3127   const Register rdx_flags  = rdx;
3128 
3129   if (!EnableInvokeDynamic) {
3130     // rewriter does not generate this bytecode
3131     __ should_not_reach_here();
3132     return;
3133   }
3134 
3135   prepare_invoke(byte_no, rbx_method, rax_mtype, rcx_recv);
3136   __ verify_method_ptr(rbx_method);
3137   __ verify_oop(rcx_recv);
3138   __ null_check(rcx_recv);
3139 
3140   // rax: MethodType object (from cpool->resolved_references[f1], if necessary)
3141   // rbx: MH.invokeExact_MT method (from f2)
3142 
3143   // Note:  rax_mtype is already pushed (if necessary) by prepare_invoke
3144 
3145   // FIXME: profile the LambdaForm also
3146   __ profile_final_call(rax);
3147   __ profile_arguments_type(rdx, rbx_method, rsi, true);
3148 
3149   __ jump_from_interpreted(rbx_method, rdx);
3150 }
3151 
3152 
3153 void TemplateTable::invokedynamic(int byte_no) {
3154   transition(vtos, vtos);
3155   assert(byte_no == f1_byte, "use this argument");
3156 
3157   if (!EnableInvokeDynamic) {
3158     // We should not encounter this bytecode if !EnableInvokeDynamic.
3159     // The verifier will stop it.  However, if we get past the verifier,
3160     // this will stop the thread in a reasonable way, without crashing the JVM.
3161     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3162                      InterpreterRuntime::throw_IncompatibleClassChangeError));
3163     // the call_VM checks for exception, so we should never return here.
3164     __ should_not_reach_here();
3165     return;
3166   }
3167 
3168   const Register rbx_method   = rbx;
3169   const Register rax_callsite = rax;
3170 
3171   prepare_invoke(byte_no, rbx_method, rax_callsite);
3172 
3173   // rax: CallSite object (from cpool->resolved_references[f1])
3174   // rbx: MH.linkToCallSite method (from f2)
3175 
3176   // Note:  rax_callsite is already pushed by prepare_invoke
3177 
3178   // %%% should make a type profile for any invokedynamic that takes a ref argument
3179   // profile this call
3180   __ profile_call(rsi);
3181   __ profile_arguments_type(rdx, rbx, rsi, false);
3182 
3183   __ verify_oop(rax_callsite);
3184 
3185   __ jump_from_interpreted(rbx_method, rdx);
3186 }
3187 
3188 //----------------------------------------------------------------------------------------------------
3189 // Allocation
3190 
3191 void TemplateTable::_new() {
3192   transition(vtos, atos);
3193   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3194   Label slow_case;
3195   Label slow_case_no_pop;
3196   Label done;
3197   Label initialize_header;
3198   Label initialize_object;  // including clearing the fields
3199   Label allocate_shared;
3200 
3201   __ get_cpool_and_tags(rcx, rax);
3202 
3203   // Make sure the class we're about to instantiate has been resolved.
3204   // This is done before loading InstanceKlass to be consistent with the order
3205   // how Constant Pool is updated (see ConstantPool::klass_at_put)
3206   const int tags_offset = Array<u1>::base_offset_in_bytes();
3207   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset), JVM_CONSTANT_Class);
3208   __ jcc(Assembler::notEqual, slow_case_no_pop);
3209 
3210   // get InstanceKlass
3211   __ movptr(rcx, Address(rcx, rdx, Address::times_ptr, sizeof(ConstantPool)));
3212   __ push(rcx);  // save the contexts of klass for initializing the header
3213 
3214   // make sure klass is initialized & doesn't have finalizer
3215   // make sure klass is fully initialized
3216   __ cmpb(Address(rcx, InstanceKlass::init_state_offset()), InstanceKlass::fully_initialized);
3217   __ jcc(Assembler::notEqual, slow_case);
3218 
3219   // get instance_size in InstanceKlass (scaled to a count of bytes)
3220   __ movl(rdx, Address(rcx, Klass::layout_helper_offset()));
3221   // test to see if it has a finalizer or is malformed in some way
3222   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3223   __ jcc(Assembler::notZero, slow_case);
3224 
3225   //
3226   // Allocate the instance
3227   // 1) Try to allocate in the TLAB
3228   // 2) if fail and the object is large allocate in the shared Eden
3229   // 3) if the above fails (or is not applicable), go to a slow case
3230   // (creates a new TLAB, etc.)
3231 
3232   const bool allow_shared_alloc =
3233     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3234 
3235   const Register thread = rcx;
3236   if (UseTLAB || allow_shared_alloc) {
3237     __ get_thread(thread);
3238   }
3239 
3240   if (UseTLAB) {
3241     __ movptr(rax, Address(thread, in_bytes(JavaThread::tlab_top_offset())));
3242     __ lea(rbx, Address(rax, rdx, Address::times_1));
3243     __ cmpptr(rbx, Address(thread, in_bytes(JavaThread::tlab_end_offset())));
3244     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3245     __ movptr(Address(thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3246     if (ZeroTLAB) {
3247       // the fields have been already cleared
3248       __ jmp(initialize_header);
3249     } else {
3250       // initialize both the header and fields
3251       __ jmp(initialize_object);
3252     }
3253   }
3254 
3255   // Allocation in the shared Eden, if allowed.
3256   //
3257   // rdx: instance size in bytes
3258   if (allow_shared_alloc) {
3259     __ bind(allocate_shared);
3260 
3261     ExternalAddress heap_top((address)Universe::heap()->top_addr());
3262 
3263     Label retry;
3264     __ bind(retry);
3265     __ movptr(rax, heap_top);
3266     __ lea(rbx, Address(rax, rdx, Address::times_1));
3267     __ cmpptr(rbx, ExternalAddress((address)Universe::heap()->end_addr()));
3268     __ jcc(Assembler::above, slow_case);
3269 
3270     // Compare rax, with the top addr, and if still equal, store the new
3271     // top addr in rbx, at the address of the top addr pointer. Sets ZF if was
3272     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3273     //
3274     // rax,: object begin
3275     // rbx,: object end
3276     // rdx: instance size in bytes
3277     __ locked_cmpxchgptr(rbx, heap_top);
3278 
3279     // if someone beat us on the allocation, try again, otherwise continue
3280     __ jcc(Assembler::notEqual, retry);
3281 
3282     __ incr_allocated_bytes(thread, rdx, 0);
3283   }
3284 
3285   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3286     // The object is initialized before the header.  If the object size is
3287     // zero, go directly to the header initialization.
3288     __ bind(initialize_object);
3289     __ decrement(rdx, sizeof(oopDesc));
3290     __ jcc(Assembler::zero, initialize_header);
3291 
3292     // Initialize topmost object field, divide rdx by 8, check if odd and
3293     // test if zero.
3294     __ xorl(rcx, rcx);    // use zero reg to clear memory (shorter code)
3295     __ shrl(rdx, LogBytesPerLong); // divide by 2*oopSize and set carry flag if odd
3296 
3297     // rdx must have been multiple of 8
3298 #ifdef ASSERT
3299     // make sure rdx was multiple of 8
3300     Label L;
3301     // Ignore partial flag stall after shrl() since it is debug VM
3302     __ jccb(Assembler::carryClear, L);
3303     __ stop("object size is not multiple of 2 - adjust this code");
3304     __ bind(L);
3305     // rdx must be > 0, no extra check needed here
3306 #endif
3307 
3308     // initialize remaining object fields: rdx was a multiple of 8
3309     { Label loop;
3310     __ bind(loop);
3311     __ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 1*oopSize), rcx);
3312     NOT_LP64(__ movptr(Address(rax, rdx, Address::times_8, sizeof(oopDesc) - 2*oopSize), rcx));
3313     __ decrement(rdx);
3314     __ jcc(Assembler::notZero, loop);
3315     }
3316 
3317     // initialize object header only.
3318     __ bind(initialize_header);
3319     if (UseBiasedLocking) {
3320       __ pop(rcx);   // get saved klass back in the register.
3321       __ movptr(rbx, Address(rcx, Klass::prototype_header_offset()));
3322       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()), rbx);
3323     } else {
3324       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes ()),
3325                 (int32_t)markOopDesc::prototype()); // header
3326       __ pop(rcx);   // get saved klass back in the register.
3327     }
3328     __ store_klass(rax, rcx);  // klass
3329 
3330     {
3331       SkipIfEqual skip_if(_masm, &DTraceAllocProbes, 0);
3332       // Trigger dtrace event for fastpath
3333       __ push(atos);
3334       __ call_VM_leaf(
3335            CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3336       __ pop(atos);
3337     }
3338 
3339     __ jmp(done);
3340   }
3341 
3342   // slow case
3343   __ bind(slow_case);
3344   __ pop(rcx);   // restore stack pointer to what it was when we came in.
3345   __ bind(slow_case_no_pop);
3346   __ get_constant_pool(rax);
3347   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3348   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), rax, rdx);
3349 
3350   // continue
3351   __ bind(done);
3352 }
3353 
3354 
3355 void TemplateTable::newarray() {
3356   transition(itos, atos);
3357   __ push_i(rax);                                 // make sure everything is on the stack
3358   __ load_unsigned_byte(rdx, at_bcp(1));
3359   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray), rdx, rax);
3360   __ pop_i(rdx);                                  // discard size
3361 }
3362 
3363 
3364 void TemplateTable::anewarray() {
3365   transition(itos, atos);
3366   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3367   __ get_constant_pool(rcx);
3368   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray), rcx, rdx, rax);
3369 }
3370 
3371 
3372 void TemplateTable::arraylength() {
3373   transition(atos, itos);
3374   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3375   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3376 }
3377 
3378 
3379 void TemplateTable::checkcast() {
3380   transition(atos, atos);
3381   Label done, is_null, ok_is_subtype, quicked, resolved;
3382   __ testptr(rax, rax);   // Object is in EAX
3383   __ jcc(Assembler::zero, is_null);
3384 
3385   // Get cpool & tags index
3386   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3387   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3388   // See if bytecode has already been quicked
3389   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3390   __ jcc(Assembler::equal, quicked);
3391 
3392   __ push(atos);
3393   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3394   // vm_result_2 has metadata result
3395   // borrow rdi from locals
3396   __ get_thread(rdi);
3397   __ get_vm_result_2(rax, rdi);
3398   __ restore_locals();
3399   __ pop_ptr(rdx);
3400   __ jmpb(resolved);
3401 
3402   // Get superklass in EAX and subklass in EBX
3403   __ bind(quicked);
3404   __ mov(rdx, rax);          // Save object in EDX; EAX needed for subtype check
3405   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3406 
3407   __ bind(resolved);
3408   __ load_klass(rbx, rdx);
3409 
3410   // Generate subtype check.  Blows ECX.  Resets EDI.  Object in EDX.
3411   // Superklass in EAX.  Subklass in EBX.
3412   __ gen_subtype_check( rbx, ok_is_subtype );
3413 
3414   // Come here on failure
3415   __ push(rdx);
3416   // object is at TOS
3417   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3418 
3419   // Come here on success
3420   __ bind(ok_is_subtype);
3421   __ mov(rax,rdx);           // Restore object in EDX
3422 
3423   // Collect counts on whether this check-cast sees NULLs a lot or not.
3424   if (ProfileInterpreter) {
3425     __ jmp(done);
3426     __ bind(is_null);
3427     __ profile_null_seen(rcx);
3428   } else {
3429     __ bind(is_null);   // same as 'done'
3430   }
3431   __ bind(done);
3432 }
3433 
3434 
3435 void TemplateTable::instanceof() {
3436   transition(atos, itos);
3437   Label done, is_null, ok_is_subtype, quicked, resolved;
3438   __ testptr(rax, rax);
3439   __ jcc(Assembler::zero, is_null);
3440 
3441   // Get cpool & tags index
3442   __ get_cpool_and_tags(rcx, rdx); // ECX=cpool, EDX=tags array
3443   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // EBX=index
3444   // See if bytecode has already been quicked
3445   __ cmpb(Address(rdx, rbx, Address::times_1, Array<u1>::base_offset_in_bytes()), JVM_CONSTANT_Class);
3446   __ jcc(Assembler::equal, quicked);
3447 
3448   __ push(atos);
3449   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3450   // vm_result_2 has metadata result
3451   // borrow rdi from locals
3452   __ get_thread(rdi);
3453   __ get_vm_result_2(rax, rdi);
3454   __ restore_locals();
3455   __ pop_ptr(rdx);
3456   __ load_klass(rdx, rdx);
3457   __ jmp(resolved);
3458 
3459   // Get superklass in EAX and subklass in EDX
3460   __ bind(quicked);
3461   __ load_klass(rdx, rax);
3462   __ movptr(rax, Address(rcx, rbx, Address::times_ptr, sizeof(ConstantPool)));
3463 
3464   __ bind(resolved);
3465 
3466   // Generate subtype check.  Blows ECX.  Resets EDI.
3467   // Superklass in EAX.  Subklass in EDX.
3468   __ gen_subtype_check( rdx, ok_is_subtype );
3469 
3470   // Come here on failure
3471   __ xorl(rax,rax);
3472   __ jmpb(done);
3473   // Come here on success
3474   __ bind(ok_is_subtype);
3475   __ movl(rax, 1);
3476 
3477   // Collect counts on whether this test sees NULLs a lot or not.
3478   if (ProfileInterpreter) {
3479     __ jmp(done);
3480     __ bind(is_null);
3481     __ profile_null_seen(rcx);
3482   } else {
3483     __ bind(is_null);   // same as 'done'
3484   }
3485   __ bind(done);
3486   // rax, = 0: obj == NULL or  obj is not an instanceof the specified klass
3487   // rax, = 1: obj != NULL and obj is     an instanceof the specified klass
3488 }
3489 
3490 
3491 //----------------------------------------------------------------------------------------------------
3492 // Breakpoints
3493 void TemplateTable::_breakpoint() {
3494 
3495   // Note: We get here even if we are single stepping..
3496   // jbug inists on setting breakpoints at every bytecode
3497   // even if we are in single step mode.
3498 
3499   transition(vtos, vtos);
3500 
3501   // get the unpatched byte code
3502   __ get_method(rcx);
3503   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::get_original_bytecode_at), rcx, rsi);
3504   __ mov(rbx, rax);
3505 
3506   // post the breakpoint event
3507   __ get_method(rcx);
3508   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint), rcx, rsi);
3509 
3510   // complete the execution of original bytecode
3511   __ dispatch_only_normal(vtos);
3512 }
3513 
3514 
3515 //----------------------------------------------------------------------------------------------------
3516 // Exceptions
3517 
3518 void TemplateTable::athrow() {
3519   transition(atos, vtos);
3520   __ null_check(rax);
3521   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3522 }
3523 
3524 
3525 //----------------------------------------------------------------------------------------------------
3526 // Synchronization
3527 //
3528 // Note: monitorenter & exit are symmetric routines; which is reflected
3529 //       in the assembly code structure as well
3530 //
3531 // Stack layout:
3532 //
3533 // [expressions  ] <--- rsp               = expression stack top
3534 // ..
3535 // [expressions  ]
3536 // [monitor entry] <--- monitor block top = expression stack bot
3537 // ..
3538 // [monitor entry]
3539 // [frame data   ] <--- monitor block bot
3540 // ...
3541 // [saved rbp,    ] <--- rbp,
3542 
3543 
3544 void TemplateTable::monitorenter() {
3545   transition(atos, vtos);
3546 
3547   // check for NULL object
3548   __ null_check(rax);
3549 
3550   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3551   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3552   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3553   Label allocated;
3554 
3555   // initialize entry pointer
3556   __ xorl(rdx, rdx);                             // points to free slot or NULL
3557 
3558   // find a free slot in the monitor block (result in rdx)
3559   { Label entry, loop, exit;
3560     __ movptr(rcx, monitor_block_top);           // points to current entry, starting with top-most entry
3561 
3562     __ lea(rbx, monitor_block_bot);              // points to word before bottom of monitor block
3563     __ jmpb(entry);
3564 
3565     __ bind(loop);
3566     __ cmpptr(Address(rcx, BasicObjectLock::obj_offset_in_bytes()), (int32_t)NULL_WORD);  // check if current entry is used
3567     __ cmovptr(Assembler::equal, rdx, rcx);      // if not used then remember entry in rdx
3568     __ cmpptr(rax, Address(rcx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3569     __ jccb(Assembler::equal, exit);             // if same object then stop searching
3570     __ addptr(rcx, entry_size);                  // otherwise advance to next entry
3571     __ bind(entry);
3572     __ cmpptr(rcx, rbx);                         // check if bottom reached
3573     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3574     __ bind(exit);
3575   }
3576 
3577   __ testptr(rdx, rdx);                          // check if a slot has been found
3578   __ jccb(Assembler::notZero, allocated);        // if found, continue with that one
3579 
3580   // allocate one if there's no free slot
3581   { Label entry, loop;
3582     // 1. compute new pointers                   // rsp: old expression stack top
3583     __ movptr(rdx, monitor_block_bot);           // rdx: old expression stack bottom
3584     __ subptr(rsp, entry_size);                  // move expression stack top
3585     __ subptr(rdx, entry_size);                  // move expression stack bottom
3586     __ mov(rcx, rsp);                            // set start value for copy loop
3587     __ movptr(monitor_block_bot, rdx);           // set new monitor block top
3588     __ jmp(entry);
3589     // 2. move expression stack contents
3590     __ bind(loop);
3591     __ movptr(rbx, Address(rcx, entry_size));    // load expression stack word from old location
3592     __ movptr(Address(rcx, 0), rbx);             // and store it at new location
3593     __ addptr(rcx, wordSize);                    // advance to next word
3594     __ bind(entry);
3595     __ cmpptr(rcx, rdx);                         // check if bottom reached
3596     __ jcc(Assembler::notEqual, loop);           // if not at bottom then copy next word
3597   }
3598 
3599   // call run-time routine
3600   // rdx: points to monitor entry
3601   __ bind(allocated);
3602 
3603   // Increment bcp to point to the next bytecode, so exception handling for async. exceptions work correctly.
3604   // The object has already been poped from the stack, so the expression stack looks correct.
3605   __ increment(rsi);
3606 
3607   __ movptr(Address(rdx, BasicObjectLock::obj_offset_in_bytes()), rax);     // store object
3608   __ lock_object(rdx);
3609 
3610   // check to make sure this monitor doesn't cause stack overflow after locking
3611   __ save_bcp();  // in case of exception
3612   __ generate_stack_overflow_check(0);
3613 
3614   // The bcp has already been incremented. Just need to dispatch to next instruction.
3615   __ dispatch_next(vtos);
3616 }
3617 
3618 
3619 void TemplateTable::monitorexit() {
3620   transition(atos, vtos);
3621 
3622   // check for NULL object
3623   __ null_check(rax);
3624 
3625   const Address monitor_block_top(rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3626   const Address monitor_block_bot(rbp, frame::interpreter_frame_initial_sp_offset        * wordSize);
3627   const int entry_size =         (     frame::interpreter_frame_monitor_size()           * wordSize);
3628   Label found;
3629 
3630   // find matching slot
3631   { Label entry, loop;
3632     __ movptr(rdx, monitor_block_top);           // points to current entry, starting with top-most entry
3633     __ lea(rbx, monitor_block_bot);             // points to word before bottom of monitor block
3634     __ jmpb(entry);
3635 
3636     __ bind(loop);
3637     __ cmpptr(rax, Address(rdx, BasicObjectLock::obj_offset_in_bytes()));   // check if current entry is for same object
3638     __ jcc(Assembler::equal, found);             // if same object then stop searching
3639     __ addptr(rdx, entry_size);                  // otherwise advance to next entry
3640     __ bind(entry);
3641     __ cmpptr(rdx, rbx);                         // check if bottom reached
3642     __ jcc(Assembler::notEqual, loop);           // if not at bottom then check this entry
3643   }
3644 
3645   // error handling. Unlocking was not block-structured
3646   Label end;
3647   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_illegal_monitor_state_exception));
3648   __ should_not_reach_here();
3649 
3650   // call run-time routine
3651   // rcx: points to monitor entry
3652   __ bind(found);
3653   __ push_ptr(rax);                                 // make sure object is on stack (contract with oopMaps)
3654   __ unlock_object(rdx);
3655   __ pop_ptr(rax);                                  // discard object
3656   __ bind(end);
3657 }
3658 
3659 
3660 //----------------------------------------------------------------------------------------------------
3661 // Wide instructions
3662 
3663 void TemplateTable::wide() {
3664   transition(vtos, vtos);
3665   __ load_unsigned_byte(rbx, at_bcp(1));
3666   ExternalAddress wtable((address)Interpreter::_wentry_point);
3667   __ jump(ArrayAddress(wtable, Address(noreg, rbx, Address::times_ptr)));
3668   // Note: the rsi increment step is part of the individual wide bytecode implementations
3669 }
3670 
3671 
3672 //----------------------------------------------------------------------------------------------------
3673 // Multi arrays
3674 
3675 void TemplateTable::multianewarray() {
3676   transition(vtos, atos);
3677   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3678   // last dim is on top of stack; we want address of first one:
3679   // first_addr = last_addr + (ndims - 1) * stackElementSize - 1*wordsize
3680   // the latter wordSize to point to the beginning of the array.
3681   __ lea(  rax, Address(rsp, rax, Interpreter::stackElementScale(), -wordSize));
3682   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray), rax);     // pass in rax,
3683   __ load_unsigned_byte(rbx, at_bcp(3));
3684   __ lea(rsp, Address(rsp, rbx, Interpreter::stackElementScale()));  // get rid of counts
3685 }
3686 
3687 #endif /* !CC_INTERP */