1 /*
   2  * Copyright (c) 2003, 2018, Oracle and/or its affiliates. All rights reserved.
   3  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   4  *
   5  * This code is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This code is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  12  * version 2 for more details (a copy is included in the LICENSE file that
  13  * accompanied this code).
  14  *
  15  * You should have received a copy of the GNU General Public License version
  16  * 2 along with this work; if not, write to the Free Software Foundation,
  17  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  18  *
  19  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  20  * or visit www.oracle.com if you need additional information or have any
  21  * questions.
  22  *
  23  */
  24 
  25 #include "precompiled.hpp"
  26 #include "asm/macroAssembler.hpp"
  27 #include "interpreter/interpreter.hpp"
  28 #include "interpreter/interpreterRuntime.hpp"
  29 #include "interpreter/templateTable.hpp"
  30 #include "memory/universe.inline.hpp"
  31 #include "oops/methodData.hpp"
  32 #include "oops/objArrayKlass.hpp"
  33 #include "oops/oop.inline.hpp"
  34 #include "prims/methodHandles.hpp"
  35 #include "runtime/sharedRuntime.hpp"
  36 #include "runtime/stubRoutines.hpp"
  37 #include "runtime/synchronizer.hpp"
  38 #include "utilities/macros.hpp"
  39 
  40 #ifndef CC_INTERP
  41 
  42 #define __ _masm->
  43 
  44 // Platform-dependent initialization
  45 
  46 void TemplateTable::pd_initialize() {
  47   // No amd64 specific initialization
  48 }
  49 
  50 // Address computation: local variables
  51 
  52 static inline Address iaddress(int n) {
  53   return Address(r14, Interpreter::local_offset_in_bytes(n));
  54 }
  55 
  56 static inline Address laddress(int n) {
  57   return iaddress(n + 1);
  58 }
  59 
  60 static inline Address faddress(int n) {
  61   return iaddress(n);
  62 }
  63 
  64 static inline Address daddress(int n) {
  65   return laddress(n);
  66 }
  67 
  68 static inline Address aaddress(int n) {
  69   return iaddress(n);
  70 }
  71 
  72 static inline Address iaddress(Register r) {
  73   return Address(r14, r, Address::times_8);
  74 }
  75 
  76 static inline Address laddress(Register r) {
  77   return Address(r14, r, Address::times_8, Interpreter::local_offset_in_bytes(1));
  78 }
  79 
  80 static inline Address faddress(Register r) {
  81   return iaddress(r);
  82 }
  83 
  84 static inline Address daddress(Register r) {
  85   return laddress(r);
  86 }
  87 
  88 static inline Address aaddress(Register r) {
  89   return iaddress(r);
  90 }
  91 
  92 static inline Address at_rsp() {
  93   return Address(rsp, 0);
  94 }
  95 
  96 // At top of Java expression stack which may be different than esp().  It
  97 // isn't for category 1 objects.
  98 static inline Address at_tos   () {
  99   return Address(rsp,  Interpreter::expr_offset_in_bytes(0));
 100 }
 101 
 102 static inline Address at_tos_p1() {
 103   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
 104 }
 105 
 106 static inline Address at_tos_p2() {
 107   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
 108 }
 109 
 110 static inline Address at_tos_p3() {
 111   return Address(rsp,  Interpreter::expr_offset_in_bytes(3));
 112 }
 113 
 114 // Condition conversion
 115 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 116   switch (cc) {
 117   case TemplateTable::equal        : return Assembler::notEqual;
 118   case TemplateTable::not_equal    : return Assembler::equal;
 119   case TemplateTable::less         : return Assembler::greaterEqual;
 120   case TemplateTable::less_equal   : return Assembler::greater;
 121   case TemplateTable::greater      : return Assembler::lessEqual;
 122   case TemplateTable::greater_equal: return Assembler::less;
 123   }
 124   ShouldNotReachHere();
 125   return Assembler::zero;
 126 }
 127 
 128 
 129 // Miscelaneous helper routines
 130 // Store an oop (or NULL) at the address described by obj.
 131 // If val == noreg this means store a NULL
 132 
 133 static void do_oop_store(InterpreterMacroAssembler* _masm,
 134                          Address obj,
 135                          Register val,
 136                          BarrierSet::Name barrier,
 137                          bool precise) {
 138   assert(val == noreg || val == rax, "parameter is just for looks");
 139   switch (barrier) {
 140 #if INCLUDE_ALL_GCS
 141     case BarrierSet::G1SATBCT:
 142     case BarrierSet::G1SATBCTLogging:
 143       {
 144         // flatten object address if needed
 145         if (obj.index() == noreg && obj.disp() == 0) {
 146           if (obj.base() != rdx) {
 147             __ movq(rdx, obj.base());
 148           }
 149         } else {
 150           __ leaq(rdx, obj);
 151         }
 152         __ g1_write_barrier_pre(rdx /* obj */,
 153                                 rbx /* pre_val */,
 154                                 r15_thread /* thread */,
 155                                 r8  /* tmp */,
 156                                 val != noreg /* tosca_live */,
 157                                 false /* expand_call */);
 158         if (val == noreg) {
 159           __ store_heap_oop_null(Address(rdx, 0));
 160         } else {
 161           // G1 barrier needs uncompressed oop for region cross check.
 162           Register new_val = val;
 163           if (UseCompressedOops) {
 164             new_val = rbx;
 165             __ movptr(new_val, val);
 166           }
 167           __ store_heap_oop(Address(rdx, 0), val);
 168           __ g1_write_barrier_post(rdx /* store_adr */,
 169                                    new_val /* new_val */,
 170                                    r15_thread /* thread */,
 171                                    r8 /* tmp */,
 172                                    rbx /* tmp2 */);
 173         }
 174       }
 175       break;
 176 #endif // INCLUDE_ALL_GCS
 177     case BarrierSet::CardTableModRef:
 178     case BarrierSet::CardTableExtension:
 179       {
 180         if (val == noreg) {
 181           __ store_heap_oop_null(obj);
 182         } else {
 183           __ store_heap_oop(obj, val);
 184           // flatten object address if needed
 185           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 186             __ store_check(obj.base());
 187           } else {
 188             __ leaq(rdx, obj);
 189             __ store_check(rdx);
 190           }
 191         }
 192       }
 193       break;
 194     case BarrierSet::ModRef:
 195     case BarrierSet::Other:
 196       if (val == noreg) {
 197         __ store_heap_oop_null(obj);
 198       } else {
 199         __ store_heap_oop(obj, val);
 200       }
 201       break;
 202     default      :
 203       ShouldNotReachHere();
 204 
 205   }
 206 }
 207 
 208 Address TemplateTable::at_bcp(int offset) {
 209   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 210   return Address(r13, offset);
 211 }
 212 
 213 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register bc_reg,
 214                                    Register temp_reg, bool load_bc_into_bc_reg/*=true*/,
 215                                    int byte_no) {
 216   if (!RewriteBytecodes)  return;
 217   Label L_patch_done;
 218 
 219   switch (bc) {
 220   case Bytecodes::_fast_aputfield:
 221   case Bytecodes::_fast_bputfield:
 222   case Bytecodes::_fast_zputfield:
 223   case Bytecodes::_fast_cputfield:
 224   case Bytecodes::_fast_dputfield:
 225   case Bytecodes::_fast_fputfield:
 226   case Bytecodes::_fast_iputfield:
 227   case Bytecodes::_fast_lputfield:
 228   case Bytecodes::_fast_sputfield:
 229     {
 230       // We skip bytecode quickening for putfield instructions when
 231       // the put_code written to the constant pool cache is zero.
 232       // This is required so that every execution of this instruction
 233       // calls out to InterpreterRuntime::resolve_get_put to do
 234       // additional, required work.
 235       assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
 236       assert(load_bc_into_bc_reg, "we use bc_reg as temp");
 237       __ get_cache_and_index_and_bytecode_at_bcp(temp_reg, bc_reg, temp_reg, byte_no, 1);
 238       __ movl(bc_reg, bc);
 239       __ cmpl(temp_reg, (int) 0);
 240       __ jcc(Assembler::zero, L_patch_done);  // don't patch
 241     }
 242     break;
 243   default:
 244     assert(byte_no == -1, "sanity");
 245     // the pair bytecodes have already done the load.
 246     if (load_bc_into_bc_reg) {
 247       __ movl(bc_reg, bc);
 248     }
 249   }
 250 
 251   if (JvmtiExport::can_post_breakpoint()) {
 252     Label L_fast_patch;
 253     // if a breakpoint is present we can't rewrite the stream directly
 254     __ movzbl(temp_reg, at_bcp(0));
 255     __ cmpl(temp_reg, Bytecodes::_breakpoint);
 256     __ jcc(Assembler::notEqual, L_fast_patch);
 257     __ get_method(temp_reg);
 258     // Let breakpoint table handling rewrite to quicker bytecode
 259     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), temp_reg, r13, bc_reg);
 260 #ifndef ASSERT
 261     __ jmpb(L_patch_done);
 262 #else
 263     __ jmp(L_patch_done);
 264 #endif
 265     __ bind(L_fast_patch);
 266   }
 267 
 268 #ifdef ASSERT
 269   Label L_okay;
 270   __ load_unsigned_byte(temp_reg, at_bcp(0));
 271   __ cmpl(temp_reg, (int) Bytecodes::java_code(bc));
 272   __ jcc(Assembler::equal, L_okay);
 273   __ cmpl(temp_reg, bc_reg);
 274   __ jcc(Assembler::equal, L_okay);
 275   __ stop("patching the wrong bytecode");
 276   __ bind(L_okay);
 277 #endif
 278 
 279   // patch bytecode
 280   __ movb(at_bcp(0), bc_reg);
 281   __ bind(L_patch_done);
 282 }
 283 
 284 
 285 // Individual instructions
 286 
 287 void TemplateTable::nop() {
 288   transition(vtos, vtos);
 289   // nothing to do
 290 }
 291 
 292 void TemplateTable::shouldnotreachhere() {
 293   transition(vtos, vtos);
 294   __ stop("shouldnotreachhere bytecode");
 295 }
 296 
 297 void TemplateTable::aconst_null() {
 298   transition(vtos, atos);
 299   __ xorl(rax, rax);
 300 }
 301 
 302 void TemplateTable::iconst(int value) {
 303   transition(vtos, itos);
 304   if (value == 0) {
 305     __ xorl(rax, rax);
 306   } else {
 307     __ movl(rax, value);
 308   }
 309 }
 310 
 311 void TemplateTable::lconst(int value) {
 312   transition(vtos, ltos);
 313   if (value == 0) {
 314     __ xorl(rax, rax);
 315   } else {
 316     __ movl(rax, value);
 317   }
 318 }
 319 
 320 void TemplateTable::fconst(int value) {
 321   transition(vtos, ftos);
 322   static float one = 1.0f, two = 2.0f;
 323   switch (value) {
 324   case 0:
 325     __ xorps(xmm0, xmm0);
 326     break;
 327   case 1:
 328     __ movflt(xmm0, ExternalAddress((address) &one));
 329     break;
 330   case 2:
 331     __ movflt(xmm0, ExternalAddress((address) &two));
 332     break;
 333   default:
 334     ShouldNotReachHere();
 335     break;
 336   }
 337 }
 338 
 339 void TemplateTable::dconst(int value) {
 340   transition(vtos, dtos);
 341   static double one = 1.0;
 342   switch (value) {
 343   case 0:
 344     __ xorpd(xmm0, xmm0);
 345     break;
 346   case 1:
 347     __ movdbl(xmm0, ExternalAddress((address) &one));
 348     break;
 349   default:
 350     ShouldNotReachHere();
 351     break;
 352   }
 353 }
 354 
 355 void TemplateTable::bipush() {
 356   transition(vtos, itos);
 357   __ load_signed_byte(rax, at_bcp(1));
 358 }
 359 
 360 void TemplateTable::sipush() {
 361   transition(vtos, itos);
 362   __ load_unsigned_short(rax, at_bcp(1));
 363   __ bswapl(rax);
 364   __ sarl(rax, 16);
 365 }
 366 
 367 void TemplateTable::ldc(bool wide) {
 368   transition(vtos, vtos);
 369   Label call_ldc, notFloat, notClass, Done;
 370 
 371   if (wide) {
 372     __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 373   } else {
 374     __ load_unsigned_byte(rbx, at_bcp(1));
 375   }
 376 
 377   __ get_cpool_and_tags(rcx, rax);
 378   const int base_offset = ConstantPool::header_size() * wordSize;
 379   const int tags_offset = Array<u1>::base_offset_in_bytes();
 380 
 381   // get type
 382   __ movzbl(rdx, Address(rax, rbx, Address::times_1, tags_offset));
 383 
 384   // unresolved class - get the resolved class
 385   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClass);
 386   __ jccb(Assembler::equal, call_ldc);
 387 
 388   // unresolved class in error state - call into runtime to throw the error
 389   // from the first resolution attempt
 390   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClassInError);
 391   __ jccb(Assembler::equal, call_ldc);
 392 
 393   // resolved class - need to call vm to get java mirror of the class
 394   __ cmpl(rdx, JVM_CONSTANT_Class);
 395   __ jcc(Assembler::notEqual, notClass);
 396 
 397   __ bind(call_ldc);
 398   __ movl(c_rarg1, wide);
 399   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), c_rarg1);
 400   __ push_ptr(rax);
 401   __ verify_oop(rax);
 402   __ jmp(Done);
 403 
 404   __ bind(notClass);
 405   __ cmpl(rdx, JVM_CONSTANT_Float);
 406   __ jccb(Assembler::notEqual, notFloat);
 407   // ftos
 408   __ movflt(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 409   __ push_f();
 410   __ jmp(Done);
 411 
 412   __ bind(notFloat);
 413 #ifdef ASSERT
 414   {
 415     Label L;
 416     __ cmpl(rdx, JVM_CONSTANT_Integer);
 417     __ jcc(Assembler::equal, L);
 418     // String and Object are rewritten to fast_aldc
 419     __ stop("unexpected tag type in ldc");
 420     __ bind(L);
 421   }
 422 #endif
 423   // itos JVM_CONSTANT_Integer only
 424   __ movl(rax, Address(rcx, rbx, Address::times_8, base_offset));
 425   __ push_i(rax);
 426   __ bind(Done);
 427 }
 428 
 429 // Fast path for caching oop constants.
 430 void TemplateTable::fast_aldc(bool wide) {
 431   transition(vtos, atos);
 432 
 433   Register result = rax;
 434   Register tmp = rdx;
 435   int index_size = wide ? sizeof(u2) : sizeof(u1);
 436 
 437   Label resolved;
 438 
 439   // We are resolved if the resolved reference cache entry contains a
 440   // non-null object (String, MethodType, etc.)
 441   assert_different_registers(result, tmp);
 442   __ get_cache_index_at_bcp(tmp, 1, index_size);
 443   __ load_resolved_reference_at_index(result, tmp);
 444   __ testl(result, result);
 445   __ jcc(Assembler::notZero, resolved);
 446 
 447   address entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_ldc);
 448 
 449   // first time invocation - must resolve first
 450   __ movl(tmp, (int)bytecode());
 451   __ call_VM(result, entry, tmp);
 452 
 453   __ bind(resolved);
 454 
 455   if (VerifyOops) {
 456     __ verify_oop(result);
 457   }
 458 }
 459 
 460 void TemplateTable::ldc2_w() {
 461   transition(vtos, vtos);
 462   Label Long, Done;
 463   __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 464 
 465   __ get_cpool_and_tags(rcx, rax);
 466   const int base_offset = ConstantPool::header_size() * wordSize;
 467   const int tags_offset = Array<u1>::base_offset_in_bytes();
 468 
 469   // get type
 470   __ cmpb(Address(rax, rbx, Address::times_1, tags_offset),
 471           JVM_CONSTANT_Double);
 472   __ jccb(Assembler::notEqual, Long);
 473   // dtos
 474   __ movdbl(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 475   __ push_d();
 476   __ jmpb(Done);
 477 
 478   __ bind(Long);
 479   // ltos
 480   __ movq(rax, Address(rcx, rbx, Address::times_8, base_offset));
 481   __ push_l();
 482 
 483   __ bind(Done);
 484 }
 485 
 486 void TemplateTable::locals_index(Register reg, int offset) {
 487   __ load_unsigned_byte(reg, at_bcp(offset));
 488   __ negptr(reg);
 489 }
 490 
 491 void TemplateTable::iload() {
 492   transition(vtos, itos);
 493   if (RewriteFrequentPairs) {
 494     Label rewrite, done;
 495     const Register bc = c_rarg3;
 496     assert(rbx != bc, "register damaged");
 497 
 498     // get next byte
 499     __ load_unsigned_byte(rbx,
 500                           at_bcp(Bytecodes::length_for(Bytecodes::_iload)));
 501     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 502     // last two iloads in a pair.  Comparing against fast_iload means that
 503     // the next bytecode is neither an iload or a caload, and therefore
 504     // an iload pair.
 505     __ cmpl(rbx, Bytecodes::_iload);
 506     __ jcc(Assembler::equal, done);
 507 
 508     __ cmpl(rbx, Bytecodes::_fast_iload);
 509     __ movl(bc, Bytecodes::_fast_iload2);
 510     __ jccb(Assembler::equal, rewrite);
 511 
 512     // if _caload, rewrite to fast_icaload
 513     __ cmpl(rbx, Bytecodes::_caload);
 514     __ movl(bc, Bytecodes::_fast_icaload);
 515     __ jccb(Assembler::equal, rewrite);
 516 
 517     // rewrite so iload doesn't check again.
 518     __ movl(bc, Bytecodes::_fast_iload);
 519 
 520     // rewrite
 521     // bc: fast bytecode
 522     __ bind(rewrite);
 523     patch_bytecode(Bytecodes::_iload, bc, rbx, false);
 524     __ bind(done);
 525   }
 526 
 527   // Get the local value into tos
 528   locals_index(rbx);
 529   __ movl(rax, iaddress(rbx));
 530 }
 531 
 532 void TemplateTable::fast_iload2() {
 533   transition(vtos, itos);
 534   locals_index(rbx);
 535   __ movl(rax, iaddress(rbx));
 536   __ push(itos);
 537   locals_index(rbx, 3);
 538   __ movl(rax, iaddress(rbx));
 539 }
 540 
 541 void TemplateTable::fast_iload() {
 542   transition(vtos, itos);
 543   locals_index(rbx);
 544   __ movl(rax, iaddress(rbx));
 545 }
 546 
 547 void TemplateTable::lload() {
 548   transition(vtos, ltos);
 549   locals_index(rbx);
 550   __ movq(rax, laddress(rbx));
 551 }
 552 
 553 void TemplateTable::fload() {
 554   transition(vtos, ftos);
 555   locals_index(rbx);
 556   __ movflt(xmm0, faddress(rbx));
 557 }
 558 
 559 void TemplateTable::dload() {
 560   transition(vtos, dtos);
 561   locals_index(rbx);
 562   __ movdbl(xmm0, daddress(rbx));
 563 }
 564 
 565 void TemplateTable::aload() {
 566   transition(vtos, atos);
 567   locals_index(rbx);
 568   __ movptr(rax, aaddress(rbx));
 569 }
 570 
 571 void TemplateTable::locals_index_wide(Register reg) {
 572   __ load_unsigned_short(reg, at_bcp(2));
 573   __ bswapl(reg);
 574   __ shrl(reg, 16);
 575   __ negptr(reg);
 576 }
 577 
 578 void TemplateTable::wide_iload() {
 579   transition(vtos, itos);
 580   locals_index_wide(rbx);
 581   __ movl(rax, iaddress(rbx));
 582 }
 583 
 584 void TemplateTable::wide_lload() {
 585   transition(vtos, ltos);
 586   locals_index_wide(rbx);
 587   __ movq(rax, laddress(rbx));
 588 }
 589 
 590 void TemplateTable::wide_fload() {
 591   transition(vtos, ftos);
 592   locals_index_wide(rbx);
 593   __ movflt(xmm0, faddress(rbx));
 594 }
 595 
 596 void TemplateTable::wide_dload() {
 597   transition(vtos, dtos);
 598   locals_index_wide(rbx);
 599   __ movdbl(xmm0, daddress(rbx));
 600 }
 601 
 602 void TemplateTable::wide_aload() {
 603   transition(vtos, atos);
 604   locals_index_wide(rbx);
 605   __ movptr(rax, aaddress(rbx));
 606 }
 607 
 608 void TemplateTable::index_check(Register array, Register index) {
 609   // destroys rbx
 610   // check array
 611   __ null_check(array, arrayOopDesc::length_offset_in_bytes());
 612   // sign extend index for use by indexed load
 613   __ movl2ptr(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 ebx for exception message
 618     assert(rbx != array, "different registers");
 619     __ movl(rbx, index);
 620   }
 621   __ jump_cc(Assembler::aboveEqual,
 622              ExternalAddress(Interpreter::_throw_ArrayIndexOutOfBoundsException_entry));
 623 }
 624 
 625 void TemplateTable::iaload() {
 626   transition(itos, itos);
 627   __ pop_ptr(rdx);
 628   // eax: index
 629   // rdx: array
 630   index_check(rdx, rax); // kills rbx
 631   __ movl(rax, Address(rdx, rax,
 632                        Address::times_4,
 633                        arrayOopDesc::base_offset_in_bytes(T_INT)));
 634 }
 635 
 636 void TemplateTable::laload() {
 637   transition(itos, ltos);
 638   __ pop_ptr(rdx);
 639   // eax: index
 640   // rdx: array
 641   index_check(rdx, rax); // kills rbx
 642   __ movq(rax, Address(rdx, rbx,
 643                        Address::times_8,
 644                        arrayOopDesc::base_offset_in_bytes(T_LONG)));
 645 }
 646 
 647 void TemplateTable::faload() {
 648   transition(itos, ftos);
 649   __ pop_ptr(rdx);
 650   // eax: index
 651   // rdx: array
 652   index_check(rdx, rax); // kills rbx
 653   __ movflt(xmm0, Address(rdx, rax,
 654                          Address::times_4,
 655                          arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 656 }
 657 
 658 void TemplateTable::daload() {
 659   transition(itos, dtos);
 660   __ pop_ptr(rdx);
 661   // eax: index
 662   // rdx: array
 663   index_check(rdx, rax); // kills rbx
 664   __ movdbl(xmm0, Address(rdx, rax,
 665                           Address::times_8,
 666                           arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 667 }
 668 
 669 void TemplateTable::aaload() {
 670   transition(itos, atos);
 671   __ pop_ptr(rdx);
 672   // eax: index
 673   // rdx: array
 674   index_check(rdx, rax); // kills rbx
 675   __ load_heap_oop(rax, Address(rdx, rax,
 676                                 UseCompressedOops ? Address::times_4 : Address::times_8,
 677                                 arrayOopDesc::base_offset_in_bytes(T_OBJECT)));
 678 }
 679 
 680 void TemplateTable::baload() {
 681   transition(itos, itos);
 682   __ pop_ptr(rdx);
 683   // eax: index
 684   // rdx: array
 685   index_check(rdx, rax); // kills rbx
 686   __ load_signed_byte(rax,
 687                       Address(rdx, rax,
 688                               Address::times_1,
 689                               arrayOopDesc::base_offset_in_bytes(T_BYTE)));
 690 }
 691 
 692 void TemplateTable::caload() {
 693   transition(itos, itos);
 694   __ pop_ptr(rdx);
 695   // eax: index
 696   // rdx: array
 697   index_check(rdx, rax); // kills rbx
 698   __ load_unsigned_short(rax,
 699                          Address(rdx, rax,
 700                                  Address::times_2,
 701                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 702 }
 703 
 704 // iload followed by caload frequent pair
 705 void TemplateTable::fast_icaload() {
 706   transition(vtos, itos);
 707   // load index out of locals
 708   locals_index(rbx);
 709   __ movl(rax, iaddress(rbx));
 710 
 711   // eax: index
 712   // rdx: array
 713   __ pop_ptr(rdx);
 714   index_check(rdx, rax); // kills rbx
 715   __ load_unsigned_short(rax,
 716                          Address(rdx, rax,
 717                                  Address::times_2,
 718                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 719 }
 720 
 721 void TemplateTable::saload() {
 722   transition(itos, itos);
 723   __ pop_ptr(rdx);
 724   // eax: index
 725   // rdx: array
 726   index_check(rdx, rax); // kills rbx
 727   __ load_signed_short(rax,
 728                        Address(rdx, rax,
 729                                Address::times_2,
 730                                arrayOopDesc::base_offset_in_bytes(T_SHORT)));
 731 }
 732 
 733 void TemplateTable::iload(int n) {
 734   transition(vtos, itos);
 735   __ movl(rax, iaddress(n));
 736 }
 737 
 738 void TemplateTable::lload(int n) {
 739   transition(vtos, ltos);
 740   __ movq(rax, laddress(n));
 741 }
 742 
 743 void TemplateTable::fload(int n) {
 744   transition(vtos, ftos);
 745   __ movflt(xmm0, faddress(n));
 746 }
 747 
 748 void TemplateTable::dload(int n) {
 749   transition(vtos, dtos);
 750   __ movdbl(xmm0, daddress(n));
 751 }
 752 
 753 void TemplateTable::aload(int n) {
 754   transition(vtos, atos);
 755   __ movptr(rax, aaddress(n));
 756 }
 757 
 758 void TemplateTable::aload_0() {
 759   transition(vtos, atos);
 760   // According to bytecode histograms, the pairs:
 761   //
 762   // _aload_0, _fast_igetfield
 763   // _aload_0, _fast_agetfield
 764   // _aload_0, _fast_fgetfield
 765   //
 766   // occur frequently. If RewriteFrequentPairs is set, the (slow)
 767   // _aload_0 bytecode checks if the next bytecode is either
 768   // _fast_igetfield, _fast_agetfield or _fast_fgetfield and then
 769   // rewrites the current bytecode into a pair bytecode; otherwise it
 770   // rewrites the current bytecode into _fast_aload_0 that doesn't do
 771   // the pair check anymore.
 772   //
 773   // Note: If the next bytecode is _getfield, the rewrite must be
 774   //       delayed, otherwise we may miss an opportunity for a pair.
 775   //
 776   // Also rewrite frequent pairs
 777   //   aload_0, aload_1
 778   //   aload_0, iload_1
 779   // These bytecodes with a small amount of code are most profitable
 780   // to rewrite
 781   if (RewriteFrequentPairs) {
 782     Label rewrite, done;
 783     const Register bc = c_rarg3;
 784     assert(rbx != bc, "register damaged");
 785     // get next byte
 786     __ load_unsigned_byte(rbx,
 787                           at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)));
 788 
 789     // do actual aload_0
 790     aload(0);
 791 
 792     // if _getfield then wait with rewrite
 793     __ cmpl(rbx, Bytecodes::_getfield);
 794     __ jcc(Assembler::equal, done);
 795 
 796     // if _igetfield then reqrite to _fast_iaccess_0
 797     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) ==
 798            Bytecodes::_aload_0,
 799            "fix bytecode definition");
 800     __ cmpl(rbx, Bytecodes::_fast_igetfield);
 801     __ movl(bc, Bytecodes::_fast_iaccess_0);
 802     __ jccb(Assembler::equal, rewrite);
 803 
 804     // if _agetfield then reqrite to _fast_aaccess_0
 805     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) ==
 806            Bytecodes::_aload_0,
 807            "fix bytecode definition");
 808     __ cmpl(rbx, Bytecodes::_fast_agetfield);
 809     __ movl(bc, Bytecodes::_fast_aaccess_0);
 810     __ jccb(Assembler::equal, rewrite);
 811 
 812     // if _fgetfield then reqrite to _fast_faccess_0
 813     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) ==
 814            Bytecodes::_aload_0,
 815            "fix bytecode definition");
 816     __ cmpl(rbx, Bytecodes::_fast_fgetfield);
 817     __ movl(bc, Bytecodes::_fast_faccess_0);
 818     __ jccb(Assembler::equal, rewrite);
 819 
 820     // else rewrite to _fast_aload0
 821     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) ==
 822            Bytecodes::_aload_0,
 823            "fix bytecode definition");
 824     __ movl(bc, Bytecodes::_fast_aload_0);
 825 
 826     // rewrite
 827     // bc: fast bytecode
 828     __ bind(rewrite);
 829     patch_bytecode(Bytecodes::_aload_0, bc, rbx, false);
 830 
 831     __ bind(done);
 832   } else {
 833     aload(0);
 834   }
 835 }
 836 
 837 void TemplateTable::istore() {
 838   transition(itos, vtos);
 839   locals_index(rbx);
 840   __ movl(iaddress(rbx), rax);
 841 }
 842 
 843 void TemplateTable::lstore() {
 844   transition(ltos, vtos);
 845   locals_index(rbx);
 846   __ movq(laddress(rbx), rax);
 847 }
 848 
 849 void TemplateTable::fstore() {
 850   transition(ftos, vtos);
 851   locals_index(rbx);
 852   __ movflt(faddress(rbx), xmm0);
 853 }
 854 
 855 void TemplateTable::dstore() {
 856   transition(dtos, vtos);
 857   locals_index(rbx);
 858   __ movdbl(daddress(rbx), xmm0);
 859 }
 860 
 861 void TemplateTable::astore() {
 862   transition(vtos, vtos);
 863   __ pop_ptr(rax);
 864   locals_index(rbx);
 865   __ movptr(aaddress(rbx), rax);
 866 }
 867 
 868 void TemplateTable::wide_istore() {
 869   transition(vtos, vtos);
 870   __ pop_i();
 871   locals_index_wide(rbx);
 872   __ movl(iaddress(rbx), rax);
 873 }
 874 
 875 void TemplateTable::wide_lstore() {
 876   transition(vtos, vtos);
 877   __ pop_l();
 878   locals_index_wide(rbx);
 879   __ movq(laddress(rbx), rax);
 880 }
 881 
 882 void TemplateTable::wide_fstore() {
 883   transition(vtos, vtos);
 884   __ pop_f();
 885   locals_index_wide(rbx);
 886   __ movflt(faddress(rbx), xmm0);
 887 }
 888 
 889 void TemplateTable::wide_dstore() {
 890   transition(vtos, vtos);
 891   __ pop_d();
 892   locals_index_wide(rbx);
 893   __ movdbl(daddress(rbx), xmm0);
 894 }
 895 
 896 void TemplateTable::wide_astore() {
 897   transition(vtos, vtos);
 898   __ pop_ptr(rax);
 899   locals_index_wide(rbx);
 900   __ movptr(aaddress(rbx), rax);
 901 }
 902 
 903 void TemplateTable::iastore() {
 904   transition(itos, vtos);
 905   __ pop_i(rbx);
 906   __ pop_ptr(rdx);
 907   // eax: value
 908   // ebx: index
 909   // rdx: array
 910   index_check(rdx, rbx); // prefer index in ebx
 911   __ movl(Address(rdx, rbx,
 912                   Address::times_4,
 913                   arrayOopDesc::base_offset_in_bytes(T_INT)),
 914           rax);
 915 }
 916 
 917 void TemplateTable::lastore() {
 918   transition(ltos, vtos);
 919   __ pop_i(rbx);
 920   __ pop_ptr(rdx);
 921   // rax: value
 922   // ebx: index
 923   // rdx: array
 924   index_check(rdx, rbx); // prefer index in ebx
 925   __ movq(Address(rdx, rbx,
 926                   Address::times_8,
 927                   arrayOopDesc::base_offset_in_bytes(T_LONG)),
 928           rax);
 929 }
 930 
 931 void TemplateTable::fastore() {
 932   transition(ftos, vtos);
 933   __ pop_i(rbx);
 934   __ pop_ptr(rdx);
 935   // xmm0: value
 936   // ebx:  index
 937   // rdx:  array
 938   index_check(rdx, rbx); // prefer index in ebx
 939   __ movflt(Address(rdx, rbx,
 940                    Address::times_4,
 941                    arrayOopDesc::base_offset_in_bytes(T_FLOAT)),
 942            xmm0);
 943 }
 944 
 945 void TemplateTable::dastore() {
 946   transition(dtos, vtos);
 947   __ pop_i(rbx);
 948   __ pop_ptr(rdx);
 949   // xmm0: value
 950   // ebx:  index
 951   // rdx:  array
 952   index_check(rdx, rbx); // prefer index in ebx
 953   __ movdbl(Address(rdx, rbx,
 954                    Address::times_8,
 955                    arrayOopDesc::base_offset_in_bytes(T_DOUBLE)),
 956            xmm0);
 957 }
 958 
 959 void TemplateTable::aastore() {
 960   Label is_null, ok_is_subtype, done;
 961   transition(vtos, vtos);
 962   // stack: ..., array, index, value
 963   __ movptr(rax, at_tos());    // value
 964   __ movl(rcx, at_tos_p1()); // index
 965   __ movptr(rdx, at_tos_p2()); // array
 966 
 967   Address element_address(rdx, rcx,
 968                           UseCompressedOops? Address::times_4 : Address::times_8,
 969                           arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 970 
 971   index_check(rdx, rcx);     // kills rbx
 972   // do array store check - check for NULL value first
 973   __ testptr(rax, rax);
 974   __ jcc(Assembler::zero, is_null);
 975 
 976   // Move subklass into rbx
 977   __ load_klass(rbx, rax);
 978   // Move superklass into rax
 979   __ load_klass(rax, rdx);
 980   __ movptr(rax, Address(rax,
 981                          ObjArrayKlass::element_klass_offset()));
 982   // Compress array + index*oopSize + 12 into a single register.  Frees rcx.
 983   __ lea(rdx, element_address);
 984 
 985   // Generate subtype check.  Blows rcx, rdi
 986   // Superklass in rax.  Subklass in rbx.
 987   __ gen_subtype_check(rbx, ok_is_subtype);
 988 
 989   // Come here on failure
 990   // object is at TOS
 991   __ jump(ExternalAddress(Interpreter::_throw_ArrayStoreException_entry));
 992 
 993   // Come here on success
 994   __ bind(ok_is_subtype);
 995 
 996   // Get the value we will store
 997   __ movptr(rax, at_tos());
 998   // Now store using the appropriate barrier
 999   do_oop_store(_masm, Address(rdx, 0), rax, _bs->kind(), true);
1000   __ jmp(done);
1001 
1002   // Have a NULL in rax, rdx=array, ecx=index.  Store NULL at ary[idx]
1003   __ bind(is_null);
1004   __ profile_null_seen(rbx);
1005 
1006   // Store a NULL
1007   do_oop_store(_masm, element_address, noreg, _bs->kind(), true);
1008 
1009   // Pop stack arguments
1010   __ bind(done);
1011   __ addptr(rsp, 3 * Interpreter::stackElementSize);
1012 }
1013 
1014 void TemplateTable::bastore() {
1015   transition(itos, vtos);
1016   __ pop_i(rbx);
1017   __ pop_ptr(rdx);
1018   // eax: value
1019   // ebx: index
1020   // rdx: array
1021   index_check(rdx, rbx); // prefer index in ebx
1022   // Need to check whether array is boolean or byte
1023   // since both types share the bastore bytecode.
1024   __ load_klass(rcx, rdx);
1025   __ movl(rcx, Address(rcx, Klass::layout_helper_offset()));
1026   int diffbit = Klass::layout_helper_boolean_diffbit();
1027   __ testl(rcx, diffbit);
1028   Label L_skip;
1029   __ jccb(Assembler::zero, L_skip);
1030   __ andl(rax, 1);  // if it is a T_BOOLEAN array, mask the stored value to 0/1
1031   __ bind(L_skip);
1032   __ movb(Address(rdx, rbx,
1033                   Address::times_1,
1034                   arrayOopDesc::base_offset_in_bytes(T_BYTE)),
1035           rax);
1036 }
1037 
1038 void TemplateTable::castore() {
1039   transition(itos, vtos);
1040   __ pop_i(rbx);
1041   __ pop_ptr(rdx);
1042   // eax: value
1043   // ebx: index
1044   // rdx: array
1045   index_check(rdx, rbx);  // prefer index in ebx
1046   __ movw(Address(rdx, rbx,
1047                   Address::times_2,
1048                   arrayOopDesc::base_offset_in_bytes(T_CHAR)),
1049           rax);
1050 }
1051 
1052 void TemplateTable::sastore() {
1053   castore();
1054 }
1055 
1056 void TemplateTable::istore(int n) {
1057   transition(itos, vtos);
1058   __ movl(iaddress(n), rax);
1059 }
1060 
1061 void TemplateTable::lstore(int n) {
1062   transition(ltos, vtos);
1063   __ movq(laddress(n), rax);
1064 }
1065 
1066 void TemplateTable::fstore(int n) {
1067   transition(ftos, vtos);
1068   __ movflt(faddress(n), xmm0);
1069 }
1070 
1071 void TemplateTable::dstore(int n) {
1072   transition(dtos, vtos);
1073   __ movdbl(daddress(n), xmm0);
1074 }
1075 
1076 void TemplateTable::astore(int n) {
1077   transition(vtos, vtos);
1078   __ pop_ptr(rax);
1079   __ movptr(aaddress(n), rax);
1080 }
1081 
1082 void TemplateTable::pop() {
1083   transition(vtos, vtos);
1084   __ addptr(rsp, Interpreter::stackElementSize);
1085 }
1086 
1087 void TemplateTable::pop2() {
1088   transition(vtos, vtos);
1089   __ addptr(rsp, 2 * Interpreter::stackElementSize);
1090 }
1091 
1092 void TemplateTable::dup() {
1093   transition(vtos, vtos);
1094   __ load_ptr(0, rax);
1095   __ push_ptr(rax);
1096   // stack: ..., a, a
1097 }
1098 
1099 void TemplateTable::dup_x1() {
1100   transition(vtos, vtos);
1101   // stack: ..., a, b
1102   __ load_ptr( 0, rax);  // load b
1103   __ load_ptr( 1, rcx);  // load a
1104   __ store_ptr(1, rax);  // store b
1105   __ store_ptr(0, rcx);  // store a
1106   __ push_ptr(rax);      // push b
1107   // stack: ..., b, a, b
1108 }
1109 
1110 void TemplateTable::dup_x2() {
1111   transition(vtos, vtos);
1112   // stack: ..., a, b, c
1113   __ load_ptr( 0, rax);  // load c
1114   __ load_ptr( 2, rcx);  // load a
1115   __ store_ptr(2, rax);  // store c in a
1116   __ push_ptr(rax);      // push c
1117   // stack: ..., c, b, c, c
1118   __ load_ptr( 2, rax);  // load b
1119   __ store_ptr(2, rcx);  // store a in b
1120   // stack: ..., c, a, c, c
1121   __ store_ptr(1, rax);  // store b in c
1122   // stack: ..., c, a, b, c
1123 }
1124 
1125 void TemplateTable::dup2() {
1126   transition(vtos, vtos);
1127   // stack: ..., a, b
1128   __ load_ptr(1, rax);  // load a
1129   __ push_ptr(rax);     // push a
1130   __ load_ptr(1, rax);  // load b
1131   __ push_ptr(rax);     // push b
1132   // stack: ..., a, b, a, b
1133 }
1134 
1135 void TemplateTable::dup2_x1() {
1136   transition(vtos, vtos);
1137   // stack: ..., a, b, c
1138   __ load_ptr( 0, rcx);  // load c
1139   __ load_ptr( 1, rax);  // load b
1140   __ push_ptr(rax);      // push b
1141   __ push_ptr(rcx);      // push c
1142   // stack: ..., a, b, c, b, c
1143   __ store_ptr(3, rcx);  // store c in b
1144   // stack: ..., a, c, c, b, c
1145   __ load_ptr( 4, rcx);  // load a
1146   __ store_ptr(2, rcx);  // store a in 2nd c
1147   // stack: ..., a, c, a, b, c
1148   __ store_ptr(4, rax);  // store b in a
1149   // stack: ..., b, c, a, b, c
1150 }
1151 
1152 void TemplateTable::dup2_x2() {
1153   transition(vtos, vtos);
1154   // stack: ..., a, b, c, d
1155   __ load_ptr( 0, rcx);  // load d
1156   __ load_ptr( 1, rax);  // load c
1157   __ push_ptr(rax);      // push c
1158   __ push_ptr(rcx);      // push d
1159   // stack: ..., a, b, c, d, c, d
1160   __ load_ptr( 4, rax);  // load b
1161   __ store_ptr(2, rax);  // store b in d
1162   __ store_ptr(4, rcx);  // store d in b
1163   // stack: ..., a, d, c, b, c, d
1164   __ load_ptr( 5, rcx);  // load a
1165   __ load_ptr( 3, rax);  // load c
1166   __ store_ptr(3, rcx);  // store a in c
1167   __ store_ptr(5, rax);  // store c in a
1168   // stack: ..., c, d, a, b, c, d
1169 }
1170 
1171 void TemplateTable::swap() {
1172   transition(vtos, vtos);
1173   // stack: ..., a, b
1174   __ load_ptr( 1, rcx);  // load a
1175   __ load_ptr( 0, rax);  // load b
1176   __ store_ptr(0, rcx);  // store a in b
1177   __ store_ptr(1, rax);  // store b in a
1178   // stack: ..., b, a
1179 }
1180 
1181 void TemplateTable::iop2(Operation op) {
1182   transition(itos, itos);
1183   switch (op) {
1184   case add  :                    __ pop_i(rdx); __ addl (rax, rdx); break;
1185   case sub  : __ movl(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1186   case mul  :                    __ pop_i(rdx); __ imull(rax, rdx); break;
1187   case _and :                    __ pop_i(rdx); __ andl (rax, rdx); break;
1188   case _or  :                    __ pop_i(rdx); __ orl  (rax, rdx); break;
1189   case _xor :                    __ pop_i(rdx); __ xorl (rax, rdx); break;
1190   case shl  : __ movl(rcx, rax); __ pop_i(rax); __ shll (rax);      break;
1191   case shr  : __ movl(rcx, rax); __ pop_i(rax); __ sarl (rax);      break;
1192   case ushr : __ movl(rcx, rax); __ pop_i(rax); __ shrl (rax);      break;
1193   default   : ShouldNotReachHere();
1194   }
1195 }
1196 
1197 void TemplateTable::lop2(Operation op) {
1198   transition(ltos, ltos);
1199   switch (op) {
1200   case add  :                    __ pop_l(rdx); __ addptr(rax, rdx); break;
1201   case sub  : __ mov(rdx, rax);  __ pop_l(rax); __ subptr(rax, rdx); break;
1202   case _and :                    __ pop_l(rdx); __ andptr(rax, rdx); break;
1203   case _or  :                    __ pop_l(rdx); __ orptr (rax, rdx); break;
1204   case _xor :                    __ pop_l(rdx); __ xorptr(rax, rdx); break;
1205   default   : ShouldNotReachHere();
1206   }
1207 }
1208 
1209 void TemplateTable::idiv() {
1210   transition(itos, itos);
1211   __ movl(rcx, rax);
1212   __ pop_i(rax);
1213   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1214   //       they are not equal, one could do a normal division (no correction
1215   //       needed), which may speed up this implementation for the common case.
1216   //       (see also JVM spec., p.243 & p.271)
1217   __ corrected_idivl(rcx);
1218 }
1219 
1220 void TemplateTable::irem() {
1221   transition(itos, itos);
1222   __ movl(rcx, rax);
1223   __ pop_i(rax);
1224   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1225   //       they are not equal, one could do a normal division (no correction
1226   //       needed), which may speed up this implementation for the common case.
1227   //       (see also JVM spec., p.243 & p.271)
1228   __ corrected_idivl(rcx);
1229   __ movl(rax, rdx);
1230 }
1231 
1232 void TemplateTable::lmul() {
1233   transition(ltos, ltos);
1234   __ pop_l(rdx);
1235   __ imulq(rax, rdx);
1236 }
1237 
1238 void TemplateTable::ldiv() {
1239   transition(ltos, ltos);
1240   __ mov(rcx, rax);
1241   __ pop_l(rax);
1242   // generate explicit div0 check
1243   __ testq(rcx, rcx);
1244   __ jump_cc(Assembler::zero,
1245              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1246   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1247   //       they are not equal, one could do a normal division (no correction
1248   //       needed), which may speed up this implementation for the common case.
1249   //       (see also JVM spec., p.243 & p.271)
1250   __ corrected_idivq(rcx); // kills rbx
1251 }
1252 
1253 void TemplateTable::lrem() {
1254   transition(ltos, ltos);
1255   __ mov(rcx, rax);
1256   __ pop_l(rax);
1257   __ testq(rcx, rcx);
1258   __ jump_cc(Assembler::zero,
1259              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1260   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1261   //       they are not equal, one could do a normal division (no correction
1262   //       needed), which may speed up this implementation for the common case.
1263   //       (see also JVM spec., p.243 & p.271)
1264   __ corrected_idivq(rcx); // kills rbx
1265   __ mov(rax, rdx);
1266 }
1267 
1268 void TemplateTable::lshl() {
1269   transition(itos, ltos);
1270   __ movl(rcx, rax);                             // get shift count
1271   __ pop_l(rax);                                 // get shift value
1272   __ shlq(rax);
1273 }
1274 
1275 void TemplateTable::lshr() {
1276   transition(itos, ltos);
1277   __ movl(rcx, rax);                             // get shift count
1278   __ pop_l(rax);                                 // get shift value
1279   __ sarq(rax);
1280 }
1281 
1282 void TemplateTable::lushr() {
1283   transition(itos, ltos);
1284   __ movl(rcx, rax);                             // get shift count
1285   __ pop_l(rax);                                 // get shift value
1286   __ shrq(rax);
1287 }
1288 
1289 void TemplateTable::fop2(Operation op) {
1290   transition(ftos, ftos);
1291   switch (op) {
1292   case add:
1293     __ addss(xmm0, at_rsp());
1294     __ addptr(rsp, Interpreter::stackElementSize);
1295     break;
1296   case sub:
1297     __ movflt(xmm1, xmm0);
1298     __ pop_f(xmm0);
1299     __ subss(xmm0, xmm1);
1300     break;
1301   case mul:
1302     __ mulss(xmm0, at_rsp());
1303     __ addptr(rsp, Interpreter::stackElementSize);
1304     break;
1305   case div:
1306     __ movflt(xmm1, xmm0);
1307     __ pop_f(xmm0);
1308     __ divss(xmm0, xmm1);
1309     break;
1310   case rem:
1311     __ movflt(xmm1, xmm0);
1312     __ pop_f(xmm0);
1313     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::frem), 2);
1314     break;
1315   default:
1316     ShouldNotReachHere();
1317     break;
1318   }
1319 }
1320 
1321 void TemplateTable::dop2(Operation op) {
1322   transition(dtos, dtos);
1323   switch (op) {
1324   case add:
1325     __ addsd(xmm0, at_rsp());
1326     __ addptr(rsp, 2 * Interpreter::stackElementSize);
1327     break;
1328   case sub:
1329     __ movdbl(xmm1, xmm0);
1330     __ pop_d(xmm0);
1331     __ subsd(xmm0, xmm1);
1332     break;
1333   case mul:
1334     __ mulsd(xmm0, at_rsp());
1335     __ addptr(rsp, 2 * Interpreter::stackElementSize);
1336     break;
1337   case div:
1338     __ movdbl(xmm1, xmm0);
1339     __ pop_d(xmm0);
1340     __ divsd(xmm0, xmm1);
1341     break;
1342   case rem:
1343     __ movdbl(xmm1, xmm0);
1344     __ pop_d(xmm0);
1345     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::drem), 2);
1346     break;
1347   default:
1348     ShouldNotReachHere();
1349     break;
1350   }
1351 }
1352 
1353 void TemplateTable::ineg() {
1354   transition(itos, itos);
1355   __ negl(rax);
1356 }
1357 
1358 void TemplateTable::lneg() {
1359   transition(ltos, ltos);
1360   __ negq(rax);
1361 }
1362 
1363 // Note: 'double' and 'long long' have 32-bits alignment on x86.
1364 static jlong* double_quadword(jlong *adr, jlong lo, jlong hi) {
1365   // Use the expression (adr)&(~0xF) to provide 128-bits aligned address
1366   // of 128-bits operands for SSE instructions.
1367   jlong *operand = (jlong*)(((intptr_t)adr)&((intptr_t)(~0xF)));
1368   // Store the value to a 128-bits operand.
1369   operand[0] = lo;
1370   operand[1] = hi;
1371   return operand;
1372 }
1373 
1374 // Buffer for 128-bits masks used by SSE instructions.
1375 static jlong float_signflip_pool[2*2];
1376 static jlong double_signflip_pool[2*2];
1377 
1378 void TemplateTable::fneg() {
1379   transition(ftos, ftos);
1380   static jlong *float_signflip  = double_quadword(&float_signflip_pool[1], 0x8000000080000000, 0x8000000080000000);
1381   __ xorps(xmm0, ExternalAddress((address) float_signflip));
1382 }
1383 
1384 void TemplateTable::dneg() {
1385   transition(dtos, dtos);
1386   static jlong *double_signflip  = double_quadword(&double_signflip_pool[1], 0x8000000000000000, 0x8000000000000000);
1387   __ xorpd(xmm0, ExternalAddress((address) double_signflip));
1388 }
1389 
1390 void TemplateTable::iinc() {
1391   transition(vtos, vtos);
1392   __ load_signed_byte(rdx, at_bcp(2)); // get constant
1393   locals_index(rbx);
1394   __ addl(iaddress(rbx), rdx);
1395 }
1396 
1397 void TemplateTable::wide_iinc() {
1398   transition(vtos, vtos);
1399   __ movl(rdx, at_bcp(4)); // get constant
1400   locals_index_wide(rbx);
1401   __ bswapl(rdx); // swap bytes & sign-extend constant
1402   __ sarl(rdx, 16);
1403   __ addl(iaddress(rbx), rdx);
1404   // Note: should probably use only one movl to get both
1405   //       the index and the constant -> fix this
1406 }
1407 
1408 void TemplateTable::convert() {
1409   // Checking
1410 #ifdef ASSERT
1411   {
1412     TosState tos_in  = ilgl;
1413     TosState tos_out = ilgl;
1414     switch (bytecode()) {
1415     case Bytecodes::_i2l: // fall through
1416     case Bytecodes::_i2f: // fall through
1417     case Bytecodes::_i2d: // fall through
1418     case Bytecodes::_i2b: // fall through
1419     case Bytecodes::_i2c: // fall through
1420     case Bytecodes::_i2s: tos_in = itos; break;
1421     case Bytecodes::_l2i: // fall through
1422     case Bytecodes::_l2f: // fall through
1423     case Bytecodes::_l2d: tos_in = ltos; break;
1424     case Bytecodes::_f2i: // fall through
1425     case Bytecodes::_f2l: // fall through
1426     case Bytecodes::_f2d: tos_in = ftos; break;
1427     case Bytecodes::_d2i: // fall through
1428     case Bytecodes::_d2l: // fall through
1429     case Bytecodes::_d2f: tos_in = dtos; break;
1430     default             : ShouldNotReachHere();
1431     }
1432     switch (bytecode()) {
1433     case Bytecodes::_l2i: // fall through
1434     case Bytecodes::_f2i: // fall through
1435     case Bytecodes::_d2i: // fall through
1436     case Bytecodes::_i2b: // fall through
1437     case Bytecodes::_i2c: // fall through
1438     case Bytecodes::_i2s: tos_out = itos; break;
1439     case Bytecodes::_i2l: // fall through
1440     case Bytecodes::_f2l: // fall through
1441     case Bytecodes::_d2l: tos_out = ltos; break;
1442     case Bytecodes::_i2f: // fall through
1443     case Bytecodes::_l2f: // fall through
1444     case Bytecodes::_d2f: tos_out = ftos; break;
1445     case Bytecodes::_i2d: // fall through
1446     case Bytecodes::_l2d: // fall through
1447     case Bytecodes::_f2d: tos_out = dtos; break;
1448     default             : ShouldNotReachHere();
1449     }
1450     transition(tos_in, tos_out);
1451   }
1452 #endif // ASSERT
1453 
1454   static const int64_t is_nan = 0x8000000000000000L;
1455 
1456   // Conversion
1457   switch (bytecode()) {
1458   case Bytecodes::_i2l:
1459     __ movslq(rax, rax);
1460     break;
1461   case Bytecodes::_i2f:
1462     __ cvtsi2ssl(xmm0, rax);
1463     break;
1464   case Bytecodes::_i2d:
1465     __ cvtsi2sdl(xmm0, rax);
1466     break;
1467   case Bytecodes::_i2b:
1468     __ movsbl(rax, rax);
1469     break;
1470   case Bytecodes::_i2c:
1471     __ movzwl(rax, rax);
1472     break;
1473   case Bytecodes::_i2s:
1474     __ movswl(rax, rax);
1475     break;
1476   case Bytecodes::_l2i:
1477     __ movl(rax, rax);
1478     break;
1479   case Bytecodes::_l2f:
1480     __ cvtsi2ssq(xmm0, rax);
1481     break;
1482   case Bytecodes::_l2d:
1483     __ cvtsi2sdq(xmm0, rax);
1484     break;
1485   case Bytecodes::_f2i:
1486   {
1487     Label L;
1488     __ cvttss2sil(rax, xmm0);
1489     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1490     __ jcc(Assembler::notEqual, L);
1491     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1492     __ bind(L);
1493   }
1494     break;
1495   case Bytecodes::_f2l:
1496   {
1497     Label L;
1498     __ cvttss2siq(rax, xmm0);
1499     // NaN or overflow/underflow?
1500     __ cmp64(rax, ExternalAddress((address) &is_nan));
1501     __ jcc(Assembler::notEqual, L);
1502     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1503     __ bind(L);
1504   }
1505     break;
1506   case Bytecodes::_f2d:
1507     __ cvtss2sd(xmm0, xmm0);
1508     break;
1509   case Bytecodes::_d2i:
1510   {
1511     Label L;
1512     __ cvttsd2sil(rax, xmm0);
1513     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1514     __ jcc(Assembler::notEqual, L);
1515     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 1);
1516     __ bind(L);
1517   }
1518     break;
1519   case Bytecodes::_d2l:
1520   {
1521     Label L;
1522     __ cvttsd2siq(rax, xmm0);
1523     // NaN or overflow/underflow?
1524     __ cmp64(rax, ExternalAddress((address) &is_nan));
1525     __ jcc(Assembler::notEqual, L);
1526     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 1);
1527     __ bind(L);
1528   }
1529     break;
1530   case Bytecodes::_d2f:
1531     __ cvtsd2ss(xmm0, xmm0);
1532     break;
1533   default:
1534     ShouldNotReachHere();
1535   }
1536 }
1537 
1538 void TemplateTable::lcmp() {
1539   transition(ltos, itos);
1540   Label done;
1541   __ pop_l(rdx);
1542   __ cmpq(rdx, rax);
1543   __ movl(rax, -1);
1544   __ jccb(Assembler::less, done);
1545   __ setb(Assembler::notEqual, rax);
1546   __ movzbl(rax, rax);
1547   __ bind(done);
1548 }
1549 
1550 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1551   Label done;
1552   if (is_float) {
1553     // XXX get rid of pop here, use ... reg, mem32
1554     __ pop_f(xmm1);
1555     __ ucomiss(xmm1, xmm0);
1556   } else {
1557     // XXX get rid of pop here, use ... reg, mem64
1558     __ pop_d(xmm1);
1559     __ ucomisd(xmm1, xmm0);
1560   }
1561   if (unordered_result < 0) {
1562     __ movl(rax, -1);
1563     __ jccb(Assembler::parity, done);
1564     __ jccb(Assembler::below, done);
1565     __ setb(Assembler::notEqual, rdx);
1566     __ movzbl(rax, rdx);
1567   } else {
1568     __ movl(rax, 1);
1569     __ jccb(Assembler::parity, done);
1570     __ jccb(Assembler::above, done);
1571     __ movl(rax, 0);
1572     __ jccb(Assembler::equal, done);
1573     __ decrementl(rax);
1574   }
1575   __ bind(done);
1576 }
1577 
1578 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1579   __ get_method(rcx); // rcx holds method
1580   __ profile_taken_branch(rax, rbx); // rax holds updated MDP, rbx
1581                                      // holds bumped taken count
1582 
1583   const ByteSize be_offset = MethodCounters::backedge_counter_offset() +
1584                              InvocationCounter::counter_offset();
1585   const ByteSize inv_offset = MethodCounters::invocation_counter_offset() +
1586                               InvocationCounter::counter_offset();
1587 
1588   // Load up edx with the branch displacement
1589   if (is_wide) {
1590     __ movl(rdx, at_bcp(1));
1591   } else {
1592     __ load_signed_short(rdx, at_bcp(1));
1593   }
1594   __ bswapl(rdx);
1595 
1596   if (!is_wide) {
1597     __ sarl(rdx, 16);
1598   }
1599   __ movl2ptr(rdx, rdx);
1600 
1601   // Handle all the JSR stuff here, then exit.
1602   // It's much shorter and cleaner than intermingling with the non-JSR
1603   // normal-branch stuff occurring below.
1604   if (is_jsr) {
1605     // Pre-load the next target bytecode into rbx
1606     __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1, 0));
1607 
1608     // compute return address as bci in rax
1609     __ lea(rax, at_bcp((is_wide ? 5 : 3) -
1610                         in_bytes(ConstMethod::codes_offset())));
1611     __ subptr(rax, Address(rcx, Method::const_offset()));
1612     // Adjust the bcp in r13 by the displacement in rdx
1613     __ addptr(r13, rdx);
1614     // jsr returns atos that is not an oop
1615     __ push_i(rax);
1616     __ dispatch_only(vtos);
1617     return;
1618   }
1619 
1620   // Normal (non-jsr) branch handling
1621 
1622   // Adjust the bcp in r13 by the displacement in rdx
1623   __ addptr(r13, rdx);
1624 
1625   assert(UseLoopCounter || !UseOnStackReplacement,
1626          "on-stack-replacement requires loop counters");
1627   Label backedge_counter_overflow;
1628   Label profile_method;
1629   Label dispatch;
1630   if (UseLoopCounter) {
1631     // increment backedge counter for backward branches
1632     // rax: MDO
1633     // ebx: MDO bumped taken-count
1634     // rcx: method
1635     // rdx: target offset
1636     // r13: target bcp
1637     // r14: locals pointer
1638     __ testl(rdx, rdx);             // check if forward or backward branch
1639     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1640 
1641     // check if MethodCounters exists
1642     Label has_counters;
1643     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1644     __ testptr(rax, rax);
1645     __ jcc(Assembler::notZero, has_counters);
1646     __ push(rdx);
1647     __ push(rcx);
1648     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::build_method_counters),
1649                rcx);
1650     __ pop(rcx);
1651     __ pop(rdx);
1652     __ movptr(rax, Address(rcx, Method::method_counters_offset()));
1653     __ jcc(Assembler::zero, dispatch);
1654     __ bind(has_counters);
1655 
1656     if (TieredCompilation) {
1657       Label no_mdo;
1658       int increment = InvocationCounter::count_increment;
1659       int mask = ((1 << Tier0BackedgeNotifyFreqLog) - 1) << InvocationCounter::count_shift;
1660       if (ProfileInterpreter) {
1661         // Are we profiling?
1662         __ movptr(rbx, Address(rcx, in_bytes(Method::method_data_offset())));
1663         __ testptr(rbx, rbx);
1664         __ jccb(Assembler::zero, no_mdo);
1665         // Increment the MDO backedge counter
1666         const Address mdo_backedge_counter(rbx, in_bytes(MethodData::backedge_counter_offset()) +
1667                                            in_bytes(InvocationCounter::counter_offset()));
1668         __ increment_mask_and_jump(mdo_backedge_counter, increment, mask, rax, false, Assembler::zero,
1669                                    UseOnStackReplacement ? &backedge_counter_overflow : NULL);
1670         __ jmp(dispatch);
1671       }
1672       __ bind(no_mdo);
1673       // Increment backedge counter in MethodCounters*
1674       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1675       __ increment_mask_and_jump(Address(rcx, be_offset), increment, mask,
1676                                  rax, false, Assembler::zero,
1677                                  UseOnStackReplacement ? &backedge_counter_overflow : NULL);
1678     } else {
1679       // increment counter
1680       __ movptr(rcx, Address(rcx, Method::method_counters_offset()));
1681       __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1682       __ incrementl(rax, InvocationCounter::count_increment); // increment counter
1683       __ movl(Address(rcx, be_offset), rax);        // store counter
1684 
1685       __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1686 
1687       __ andl(rax, InvocationCounter::count_mask_value); // and the status bits
1688       __ addl(rax, Address(rcx, be_offset));        // add both counters
1689 
1690       if (ProfileInterpreter) {
1691         // Test to see if we should create a method data oop
1692         __ cmp32(rax,
1693                  ExternalAddress((address) &InvocationCounter::InterpreterProfileLimit));
1694         __ jcc(Assembler::less, dispatch);
1695 
1696         // if no method data exists, go to profile method
1697         __ test_method_data_pointer(rax, profile_method);
1698 
1699         if (UseOnStackReplacement) {
1700           // check for overflow against ebx which is the MDO taken count
1701           __ cmp32(rbx,
1702                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1703           __ jcc(Assembler::below, dispatch);
1704 
1705           // When ProfileInterpreter is on, the backedge_count comes
1706           // from the MethodData*, which value does not get reset on
1707           // the call to frequency_counter_overflow().  To avoid
1708           // excessive calls to the overflow routine while the method is
1709           // being compiled, add a second test to make sure the overflow
1710           // function is called only once every overflow_frequency.
1711           const int overflow_frequency = 1024;
1712           __ andl(rbx, overflow_frequency - 1);
1713           __ jcc(Assembler::zero, backedge_counter_overflow);
1714 
1715         }
1716       } else {
1717         if (UseOnStackReplacement) {
1718           // check for overflow against eax, which is the sum of the
1719           // counters
1720           __ cmp32(rax,
1721                    ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1722           __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1723 
1724         }
1725       }
1726     }
1727     __ bind(dispatch);
1728   }
1729 
1730   // Pre-load the next target bytecode into rbx
1731   __ load_unsigned_byte(rbx, Address(r13, 0));
1732 
1733   // continue with the bytecode @ target
1734   // eax: return bci for jsr's, unused otherwise
1735   // ebx: target bytecode
1736   // r13: target bcp
1737   __ dispatch_only(vtos);
1738 
1739   if (UseLoopCounter) {
1740     if (ProfileInterpreter) {
1741       // Out-of-line code to allocate method data oop.
1742       __ bind(profile_method);
1743       __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::profile_method));
1744       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1745       __ set_method_data_pointer_for_bcp();
1746       __ jmp(dispatch);
1747     }
1748 
1749     if (UseOnStackReplacement) {
1750       // invocation counter overflow
1751       __ bind(backedge_counter_overflow);
1752       __ negptr(rdx);
1753       __ addptr(rdx, r13); // branch bcp
1754       // IcoResult frequency_counter_overflow([JavaThread*], address branch_bcp)
1755       __ call_VM(noreg,
1756                  CAST_FROM_FN_PTR(address,
1757                                   InterpreterRuntime::frequency_counter_overflow),
1758                  rdx);
1759       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1760 
1761       // rax: osr nmethod (osr ok) or NULL (osr not possible)
1762       // ebx: target bytecode
1763       // rdx: scratch
1764       // r14: locals pointer
1765       // r13: bcp
1766       __ testptr(rax, rax);                        // test result
1767       __ jcc(Assembler::zero, dispatch);         // no osr if null
1768       // nmethod may have been invalidated (VM may block upon call_VM return)
1769       __ movl(rcx, Address(rax, nmethod::entry_bci_offset()));
1770       __ cmpl(rcx, InvalidOSREntryBci);
1771       __ jcc(Assembler::equal, dispatch);
1772 
1773       // We have the address of an on stack replacement routine in eax
1774       // We need to prepare to execute the OSR method. First we must
1775       // migrate the locals and monitors off of the stack.
1776 
1777       __ mov(r13, rax);                             // save the nmethod
1778 
1779       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1780 
1781       // eax is OSR buffer, move it to expected parameter location
1782       __ mov(j_rarg0, rax);
1783 
1784       // We use j_rarg definitions here so that registers don't conflict as parameter
1785       // registers change across platforms as we are in the midst of a calling
1786       // sequence to the OSR nmethod and we don't want collision. These are NOT parameters.
1787 
1788       const Register retaddr = j_rarg2;
1789       const Register sender_sp = j_rarg1;
1790 
1791       // pop the interpreter frame
1792       __ movptr(sender_sp, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1793       __ leave();                                // remove frame anchor
1794       __ pop(retaddr);                           // get return address
1795       __ mov(rsp, sender_sp);                   // set sp to sender sp
1796       // Ensure compiled code always sees stack at proper alignment
1797       __ andptr(rsp, -(StackAlignmentInBytes));
1798 
1799       // unlike x86 we need no specialized return from compiled code
1800       // to the interpreter or the call stub.
1801 
1802       // push the return address
1803       __ push(retaddr);
1804 
1805       // and begin the OSR nmethod
1806       __ jmp(Address(r13, nmethod::osr_entry_point_offset()));
1807     }
1808   }
1809 }
1810 
1811 
1812 void TemplateTable::if_0cmp(Condition cc) {
1813   transition(itos, vtos);
1814   // assume branch is more often taken than not (loops use backward branches)
1815   Label not_taken;
1816   __ testl(rax, rax);
1817   __ jcc(j_not(cc), not_taken);
1818   branch(false, false);
1819   __ bind(not_taken);
1820   __ profile_not_taken_branch(rax);
1821 }
1822 
1823 void TemplateTable::if_icmp(Condition cc) {
1824   transition(itos, vtos);
1825   // assume branch is more often taken than not (loops use backward branches)
1826   Label not_taken;
1827   __ pop_i(rdx);
1828   __ cmpl(rdx, rax);
1829   __ jcc(j_not(cc), not_taken);
1830   branch(false, false);
1831   __ bind(not_taken);
1832   __ profile_not_taken_branch(rax);
1833 }
1834 
1835 void TemplateTable::if_nullcmp(Condition cc) {
1836   transition(atos, vtos);
1837   // assume branch is more often taken than not (loops use backward branches)
1838   Label not_taken;
1839   __ testptr(rax, rax);
1840   __ jcc(j_not(cc), not_taken);
1841   branch(false, false);
1842   __ bind(not_taken);
1843   __ profile_not_taken_branch(rax);
1844 }
1845 
1846 void TemplateTable::if_acmp(Condition cc) {
1847   transition(atos, vtos);
1848   // assume branch is more often taken than not (loops use backward branches)
1849   Label not_taken;
1850   __ pop_ptr(rdx);
1851   __ cmpptr(rdx, rax);
1852   __ jcc(j_not(cc), not_taken);
1853   branch(false, false);
1854   __ bind(not_taken);
1855   __ profile_not_taken_branch(rax);
1856 }
1857 
1858 void TemplateTable::ret() {
1859   transition(vtos, vtos);
1860   locals_index(rbx);
1861   __ movslq(rbx, iaddress(rbx)); // get return bci, compute return bcp
1862   __ profile_ret(rbx, rcx);
1863   __ get_method(rax);
1864   __ movptr(r13, Address(rax, Method::const_offset()));
1865   __ lea(r13, Address(r13, rbx, Address::times_1,
1866                       ConstMethod::codes_offset()));
1867   __ dispatch_next(vtos);
1868 }
1869 
1870 void TemplateTable::wide_ret() {
1871   transition(vtos, vtos);
1872   locals_index_wide(rbx);
1873   __ movptr(rbx, aaddress(rbx)); // get return bci, compute return bcp
1874   __ profile_ret(rbx, rcx);
1875   __ get_method(rax);
1876   __ movptr(r13, Address(rax, Method::const_offset()));
1877   __ lea(r13, Address(r13, rbx, Address::times_1, ConstMethod::codes_offset()));
1878   __ dispatch_next(vtos);
1879 }
1880 
1881 void TemplateTable::tableswitch() {
1882   Label default_case, continue_execution;
1883   transition(itos, vtos);
1884   // align r13
1885   __ lea(rbx, at_bcp(BytesPerInt));
1886   __ andptr(rbx, -BytesPerInt);
1887   // load lo & hi
1888   __ movl(rcx, Address(rbx, BytesPerInt));
1889   __ movl(rdx, Address(rbx, 2 * BytesPerInt));
1890   __ bswapl(rcx);
1891   __ bswapl(rdx);
1892   // check against lo & hi
1893   __ cmpl(rax, rcx);
1894   __ jcc(Assembler::less, default_case);
1895   __ cmpl(rax, rdx);
1896   __ jcc(Assembler::greater, default_case);
1897   // lookup dispatch offset
1898   __ subl(rax, rcx);
1899   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1900   __ profile_switch_case(rax, rbx, rcx);
1901   // continue execution
1902   __ bind(continue_execution);
1903   __ bswapl(rdx);
1904   __ movl2ptr(rdx, rdx);
1905   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1906   __ addptr(r13, rdx);
1907   __ dispatch_only(vtos);
1908   // handle default
1909   __ bind(default_case);
1910   __ profile_switch_default(rax);
1911   __ movl(rdx, Address(rbx, 0));
1912   __ jmp(continue_execution);
1913 }
1914 
1915 void TemplateTable::lookupswitch() {
1916   transition(itos, itos);
1917   __ stop("lookupswitch bytecode should have been rewritten");
1918 }
1919 
1920 void TemplateTable::fast_linearswitch() {
1921   transition(itos, vtos);
1922   Label loop_entry, loop, found, continue_execution;
1923   // bswap rax so we can avoid bswapping the table entries
1924   __ bswapl(rax);
1925   // align r13
1926   __ lea(rbx, at_bcp(BytesPerInt)); // btw: should be able to get rid of
1927                                     // this instruction (change offsets
1928                                     // below)
1929   __ andptr(rbx, -BytesPerInt);
1930   // set counter
1931   __ movl(rcx, Address(rbx, BytesPerInt));
1932   __ bswapl(rcx);
1933   __ jmpb(loop_entry);
1934   // table search
1935   __ bind(loop);
1936   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * BytesPerInt));
1937   __ jcc(Assembler::equal, found);
1938   __ bind(loop_entry);
1939   __ decrementl(rcx);
1940   __ jcc(Assembler::greaterEqual, loop);
1941   // default case
1942   __ profile_switch_default(rax);
1943   __ movl(rdx, Address(rbx, 0));
1944   __ jmp(continue_execution);
1945   // entry found -> get offset
1946   __ bind(found);
1947   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * BytesPerInt));
1948   __ profile_switch_case(rcx, rax, rbx);
1949   // continue execution
1950   __ bind(continue_execution);
1951   __ bswapl(rdx);
1952   __ movl2ptr(rdx, rdx);
1953   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1954   __ addptr(r13, rdx);
1955   __ dispatch_only(vtos);
1956 }
1957 
1958 void TemplateTable::fast_binaryswitch() {
1959   transition(itos, vtos);
1960   // Implementation using the following core algorithm:
1961   //
1962   // int binary_search(int key, LookupswitchPair* array, int n) {
1963   //   // Binary search according to "Methodik des Programmierens" by
1964   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1965   //   int i = 0;
1966   //   int j = n;
1967   //   while (i+1 < j) {
1968   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1969   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1970   //     // where a stands for the array and assuming that the (inexisting)
1971   //     // element a[n] is infinitely big.
1972   //     int h = (i + j) >> 1;
1973   //     // i < h < j
1974   //     if (key < array[h].fast_match()) {
1975   //       j = h;
1976   //     } else {
1977   //       i = h;
1978   //     }
1979   //   }
1980   //   // R: a[i] <= key < a[i+1] or Q
1981   //   // (i.e., if key is within array, i is the correct index)
1982   //   return i;
1983   // }
1984 
1985   // Register allocation
1986   const Register key   = rax; // already set (tosca)
1987   const Register array = rbx;
1988   const Register i     = rcx;
1989   const Register j     = rdx;
1990   const Register h     = rdi;
1991   const Register temp  = rsi;
1992 
1993   // Find array start
1994   __ lea(array, at_bcp(3 * BytesPerInt)); // btw: should be able to
1995                                           // get rid of this
1996                                           // instruction (change
1997                                           // offsets below)
1998   __ andptr(array, -BytesPerInt);
1999 
2000   // Initialize i & j
2001   __ xorl(i, i);                            // i = 0;
2002   __ movl(j, Address(array, -BytesPerInt)); // j = length(array);
2003 
2004   // Convert j into native byteordering
2005   __ bswapl(j);
2006 
2007   // And start
2008   Label entry;
2009   __ jmp(entry);
2010 
2011   // binary search loop
2012   {
2013     Label loop;
2014     __ bind(loop);
2015     // int h = (i + j) >> 1;
2016     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
2017     __ sarl(h, 1);                               // h = (i + j) >> 1;
2018     // if (key < array[h].fast_match()) {
2019     //   j = h;
2020     // } else {
2021     //   i = h;
2022     // }
2023     // Convert array[h].match to native byte-ordering before compare
2024     __ movl(temp, Address(array, h, Address::times_8));
2025     __ bswapl(temp);
2026     __ cmpl(key, temp);
2027     // j = h if (key <  array[h].fast_match())
2028     __ cmovl(Assembler::less, j, h);
2029     // i = h if (key >= array[h].fast_match())
2030     __ cmovl(Assembler::greaterEqual, i, h);
2031     // while (i+1 < j)
2032     __ bind(entry);
2033     __ leal(h, Address(i, 1)); // i+1
2034     __ cmpl(h, j);             // i+1 < j
2035     __ jcc(Assembler::less, loop);
2036   }
2037 
2038   // end of binary search, result index is i (must check again!)
2039   Label default_case;
2040   // Convert array[i].match to native byte-ordering before compare
2041   __ movl(temp, Address(array, i, Address::times_8));
2042   __ bswapl(temp);
2043   __ cmpl(key, temp);
2044   __ jcc(Assembler::notEqual, default_case);
2045 
2046   // entry found -> j = offset
2047   __ movl(j , Address(array, i, Address::times_8, BytesPerInt));
2048   __ profile_switch_case(i, key, array);
2049   __ bswapl(j);
2050   __ movl2ptr(j, j);
2051   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
2052   __ addptr(r13, j);
2053   __ dispatch_only(vtos);
2054 
2055   // default case -> j = default offset
2056   __ bind(default_case);
2057   __ profile_switch_default(i);
2058   __ movl(j, Address(array, -2 * BytesPerInt));
2059   __ bswapl(j);
2060   __ movl2ptr(j, j);
2061   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
2062   __ addptr(r13, j);
2063   __ dispatch_only(vtos);
2064 }
2065 
2066 
2067 void TemplateTable::_return(TosState state) {
2068   transition(state, state);
2069   assert(_desc->calls_vm(),
2070          "inconsistent calls_vm information"); // call in remove_activation
2071 
2072   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2073     assert(state == vtos, "only valid state");
2074     __ movptr(c_rarg1, aaddress(0));
2075     __ load_klass(rdi, c_rarg1);
2076     __ movl(rdi, Address(rdi, Klass::access_flags_offset()));
2077     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2078     Label skip_register_finalizer;
2079     __ jcc(Assembler::zero, skip_register_finalizer);
2080 
2081     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), c_rarg1);
2082 
2083     __ bind(skip_register_finalizer);
2084   }
2085 
2086   // Narrow result if state is itos but result type is smaller.
2087   // Need to narrow in the return bytecode rather than in generate_return_entry
2088   // since compiled code callers expect the result to already be narrowed.
2089   if (state == itos) {
2090     __ narrow(rax);
2091   }
2092   __ remove_activation(state, r13);
2093 
2094   __ jmp(r13);
2095 }
2096 
2097 // ----------------------------------------------------------------------------
2098 // Volatile variables demand their effects be made known to all CPU's
2099 // in order.  Store buffers on most chips allow reads & writes to
2100 // reorder; the JMM's ReadAfterWrite.java test fails in -Xint mode
2101 // without some kind of memory barrier (i.e., it's not sufficient that
2102 // the interpreter does not reorder volatile references, the hardware
2103 // also must not reorder them).
2104 //
2105 // According to the new Java Memory Model (JMM):
2106 // (1) All volatiles are serialized wrt to each other.  ALSO reads &
2107 //     writes act as aquire & release, so:
2108 // (2) A read cannot let unrelated NON-volatile memory refs that
2109 //     happen after the read float up to before the read.  It's OK for
2110 //     non-volatile memory refs that happen before the volatile read to
2111 //     float down below it.
2112 // (3) Similar a volatile write cannot let unrelated NON-volatile
2113 //     memory refs that happen BEFORE the write float down to after the
2114 //     write.  It's OK for non-volatile memory refs that happen after the
2115 //     volatile write to float up before it.
2116 //
2117 // We only put in barriers around volatile refs (they are expensive),
2118 // not _between_ memory refs (that would require us to track the
2119 // flavor of the previous memory refs).  Requirements (2) and (3)
2120 // require some barriers before volatile stores and after volatile
2121 // loads.  These nearly cover requirement (1) but miss the
2122 // volatile-store-volatile-load case.  This final case is placed after
2123 // volatile-stores although it could just as well go before
2124 // volatile-loads.
2125 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits
2126                                      order_constraint) {
2127   // Helper function to insert a is-volatile test and memory barrier
2128   if (os::is_MP()) { // Not needed on single CPU
2129     __ membar(order_constraint);
2130   }
2131 }
2132 
2133 void TemplateTable::resolve_cache_and_index(int byte_no,
2134                                             Register Rcache,
2135                                             Register index,
2136                                             size_t index_size) {
2137   const Register temp = rbx;
2138   assert_different_registers(Rcache, index, temp);
2139 
2140   Label resolved;
2141     assert(byte_no == f1_byte || byte_no == f2_byte, "byte_no out of range");
2142     __ get_cache_and_index_and_bytecode_at_bcp(Rcache, index, temp, byte_no, 1, index_size);
2143     __ cmpl(temp, (int) bytecode());  // have we resolved this bytecode?
2144     __ jcc(Assembler::equal, resolved);
2145 
2146   // resolve first time through
2147   address entry;
2148   switch (bytecode()) {
2149   case Bytecodes::_getstatic:
2150   case Bytecodes::_putstatic:
2151   case Bytecodes::_getfield:
2152   case Bytecodes::_putfield:
2153     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);
2154     break;
2155   case Bytecodes::_invokevirtual:
2156   case Bytecodes::_invokespecial:
2157   case Bytecodes::_invokestatic:
2158   case Bytecodes::_invokeinterface:
2159     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);
2160     break;
2161   case Bytecodes::_invokehandle:
2162     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokehandle);
2163     break;
2164   case Bytecodes::_invokedynamic:
2165     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invokedynamic);
2166     break;
2167   default:
2168     fatal(err_msg("unexpected bytecode: %s", Bytecodes::name(bytecode())));
2169     break;
2170   }
2171   __ movl(temp, (int) bytecode());
2172   __ call_VM(noreg, entry, temp);
2173 
2174   // Update registers with resolved info
2175   __ get_cache_and_index_at_bcp(Rcache, index, 1, index_size);
2176   __ bind(resolved);
2177 }
2178 
2179 // The cache and index registers must be set before call
2180 void TemplateTable::load_field_cp_cache_entry(Register obj,
2181                                               Register cache,
2182                                               Register index,
2183                                               Register off,
2184                                               Register flags,
2185                                               bool is_static = false) {
2186   assert_different_registers(cache, index, flags, off);
2187 
2188   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2189   // Field offset
2190   __ movptr(off, Address(cache, index, Address::times_ptr,
2191                          in_bytes(cp_base_offset +
2192                                   ConstantPoolCacheEntry::f2_offset())));
2193   // Flags
2194   __ movl(flags, Address(cache, index, Address::times_ptr,
2195                          in_bytes(cp_base_offset +
2196                                   ConstantPoolCacheEntry::flags_offset())));
2197 
2198   // klass overwrite register
2199   if (is_static) {
2200     __ movptr(obj, Address(cache, index, Address::times_ptr,
2201                            in_bytes(cp_base_offset +
2202                                     ConstantPoolCacheEntry::f1_offset())));
2203     const int mirror_offset = in_bytes(Klass::java_mirror_offset());
2204     __ movptr(obj, Address(obj, mirror_offset));
2205   }
2206 }
2207 
2208 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2209                                                Register method,
2210                                                Register itable_index,
2211                                                Register flags,
2212                                                bool is_invokevirtual,
2213                                                bool is_invokevfinal, /*unused*/
2214                                                bool is_invokedynamic) {
2215   // setup registers
2216   const Register cache = rcx;
2217   const Register index = rdx;
2218   assert_different_registers(method, flags);
2219   assert_different_registers(method, cache, index);
2220   assert_different_registers(itable_index, flags);
2221   assert_different_registers(itable_index, cache, index);
2222   // determine constant pool cache field offsets
2223   assert(is_invokevirtual == (byte_no == f2_byte), "is_invokevirtual flag redundant");
2224   const int method_offset = in_bytes(
2225     ConstantPoolCache::base_offset() +
2226       ((byte_no == f2_byte)
2227        ? ConstantPoolCacheEntry::f2_offset()
2228        : ConstantPoolCacheEntry::f1_offset()));
2229   const int flags_offset = in_bytes(ConstantPoolCache::base_offset() +
2230                                     ConstantPoolCacheEntry::flags_offset());
2231   // access constant pool cache fields
2232   const int index_offset = in_bytes(ConstantPoolCache::base_offset() +
2233                                     ConstantPoolCacheEntry::f2_offset());
2234 
2235   size_t index_size = (is_invokedynamic ? sizeof(u4) : sizeof(u2));
2236   resolve_cache_and_index(byte_no, cache, index, index_size);
2237     __ movptr(method, Address(cache, index, Address::times_ptr, method_offset));
2238 
2239   if (itable_index != noreg) {
2240     // pick up itable or appendix index from f2 also:
2241     __ movptr(itable_index, Address(cache, index, Address::times_ptr, index_offset));
2242   }
2243   __ movl(flags, Address(cache, index, Address::times_ptr, flags_offset));
2244 }
2245 
2246 // Correct values of the cache and index registers are preserved.
2247 void TemplateTable::jvmti_post_field_access(Register cache, Register index,
2248                                             bool is_static, bool has_tos) {
2249   // do the JVMTI work here to avoid disturbing the register state below
2250   // We use c_rarg registers here because we want to use the register used in
2251   // the call to the VM
2252   if (JvmtiExport::can_post_field_access()) {
2253     // Check to see if a field access watch has been set before we
2254     // take the time to call into the VM.
2255     Label L1;
2256     assert_different_registers(cache, index, rax);
2257     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2258     __ testl(rax, rax);
2259     __ jcc(Assembler::zero, L1);
2260 
2261     __ get_cache_and_index_at_bcp(c_rarg2, c_rarg3, 1);
2262 
2263     // cache entry pointer
2264     __ addptr(c_rarg2, in_bytes(ConstantPoolCache::base_offset()));
2265     __ shll(c_rarg3, LogBytesPerWord);
2266     __ addptr(c_rarg2, c_rarg3);
2267     if (is_static) {
2268       __ xorl(c_rarg1, c_rarg1); // NULL object reference
2269     } else {
2270       __ movptr(c_rarg1, at_tos()); // get object pointer without popping it
2271       __ verify_oop(c_rarg1);
2272     }
2273     // c_rarg1: object pointer or NULL
2274     // c_rarg2: cache entry pointer
2275     // c_rarg3: jvalue object on the stack
2276     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
2277                                        InterpreterRuntime::post_field_access),
2278                c_rarg1, c_rarg2, c_rarg3);
2279     __ get_cache_and_index_at_bcp(cache, index, 1);
2280     __ bind(L1);
2281   }
2282 }
2283 
2284 void TemplateTable::pop_and_check_object(Register r) {
2285   __ pop_ptr(r);
2286   __ null_check(r);  // for field access must check obj.
2287   __ verify_oop(r);
2288 }
2289 
2290 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2291   transition(vtos, vtos);
2292 
2293   const Register cache = rcx;
2294   const Register index = rdx;
2295   const Register obj   = c_rarg3;
2296   const Register off   = rbx;
2297   const Register flags = rax;
2298   const Register bc = c_rarg3; // uses same reg as obj, so don't mix them
2299 
2300   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2301   jvmti_post_field_access(cache, index, is_static, false);
2302   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2303 
2304   if (!is_static) {
2305     // obj is on the stack
2306     pop_and_check_object(obj);
2307   }
2308 
2309   const Address field(obj, off, Address::times_1);
2310 
2311   Label Done, notByte, notBool, notInt, notShort, notChar,
2312               notLong, notFloat, notObj, notDouble;
2313 
2314   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2315   // Make sure we don't need to mask edx after the above shift
2316   assert(btos == 0, "change code, btos != 0");
2317 
2318   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2319   __ jcc(Assembler::notZero, notByte);
2320   // btos
2321   __ load_signed_byte(rax, field);
2322   __ push(btos);
2323   // Rewrite bytecode to be faster
2324   if (!is_static) {
2325     patch_bytecode(Bytecodes::_fast_bgetfield, bc, rbx);
2326   }
2327   __ jmp(Done);
2328 
2329   __ bind(notByte);
2330   __ cmpl(flags, ztos);
2331   __ jcc(Assembler::notEqual, notBool);
2332 
2333   // ztos (same code as btos)
2334   __ load_signed_byte(rax, field);
2335   __ push(ztos);
2336   // Rewrite bytecode to be faster
2337   if (!is_static) {
2338     // use btos rewriting, no truncating to t/f bit is needed for getfield.
2339     patch_bytecode(Bytecodes::_fast_bgetfield, bc, rbx);
2340   }
2341   __ jmp(Done);
2342 
2343   __ bind(notBool);
2344   __ cmpl(flags, atos);
2345   __ jcc(Assembler::notEqual, notObj);
2346   // atos
2347   __ load_heap_oop(rax, field);
2348   __ push(atos);
2349   if (!is_static) {
2350     patch_bytecode(Bytecodes::_fast_agetfield, bc, rbx);
2351   }
2352   __ jmp(Done);
2353 
2354   __ bind(notObj);
2355   __ cmpl(flags, itos);
2356   __ jcc(Assembler::notEqual, notInt);
2357   // itos
2358   __ movl(rax, field);
2359   __ push(itos);
2360   // Rewrite bytecode to be faster
2361   if (!is_static) {
2362     patch_bytecode(Bytecodes::_fast_igetfield, bc, rbx);
2363   }
2364   __ jmp(Done);
2365 
2366   __ bind(notInt);
2367   __ cmpl(flags, ctos);
2368   __ jcc(Assembler::notEqual, notChar);
2369   // ctos
2370   __ load_unsigned_short(rax, field);
2371   __ push(ctos);
2372   // Rewrite bytecode to be faster
2373   if (!is_static) {
2374     patch_bytecode(Bytecodes::_fast_cgetfield, bc, rbx);
2375   }
2376   __ jmp(Done);
2377 
2378   __ bind(notChar);
2379   __ cmpl(flags, stos);
2380   __ jcc(Assembler::notEqual, notShort);
2381   // stos
2382   __ load_signed_short(rax, field);
2383   __ push(stos);
2384   // Rewrite bytecode to be faster
2385   if (!is_static) {
2386     patch_bytecode(Bytecodes::_fast_sgetfield, bc, rbx);
2387   }
2388   __ jmp(Done);
2389 
2390   __ bind(notShort);
2391   __ cmpl(flags, ltos);
2392   __ jcc(Assembler::notEqual, notLong);
2393   // ltos
2394   __ movq(rax, field);
2395   __ push(ltos);
2396   // Rewrite bytecode to be faster
2397   if (!is_static) {
2398     patch_bytecode(Bytecodes::_fast_lgetfield, bc, rbx);
2399   }
2400   __ jmp(Done);
2401 
2402   __ bind(notLong);
2403   __ cmpl(flags, ftos);
2404   __ jcc(Assembler::notEqual, notFloat);
2405   // ftos
2406   __ movflt(xmm0, field);
2407   __ push(ftos);
2408   // Rewrite bytecode to be faster
2409   if (!is_static) {
2410     patch_bytecode(Bytecodes::_fast_fgetfield, bc, rbx);
2411   }
2412   __ jmp(Done);
2413 
2414   __ bind(notFloat);
2415 #ifdef ASSERT
2416   __ cmpl(flags, dtos);
2417   __ jcc(Assembler::notEqual, notDouble);
2418 #endif
2419   // dtos
2420   __ movdbl(xmm0, field);
2421   __ push(dtos);
2422   // Rewrite bytecode to be faster
2423   if (!is_static) {
2424     patch_bytecode(Bytecodes::_fast_dgetfield, bc, rbx);
2425   }
2426 #ifdef ASSERT
2427   __ jmp(Done);
2428 
2429   __ bind(notDouble);
2430   __ stop("Bad state");
2431 #endif
2432 
2433   __ bind(Done);
2434   // [jk] not needed currently
2435   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadLoad |
2436   //                                              Assembler::LoadStore));
2437 }
2438 
2439 
2440 void TemplateTable::getfield(int byte_no) {
2441   getfield_or_static(byte_no, false);
2442 }
2443 
2444 void TemplateTable::getstatic(int byte_no) {
2445   getfield_or_static(byte_no, true);
2446 }
2447 
2448 // The registers cache and index expected to be set before call.
2449 // The function may destroy various registers, just not the cache and index registers.
2450 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2451   transition(vtos, vtos);
2452 
2453   ByteSize cp_base_offset = ConstantPoolCache::base_offset();
2454 
2455   if (JvmtiExport::can_post_field_modification()) {
2456     // Check to see if a field modification watch has been set before
2457     // we take the time to call into the VM.
2458     Label L1;
2459     assert_different_registers(cache, index, rax);
2460     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2461     __ testl(rax, rax);
2462     __ jcc(Assembler::zero, L1);
2463 
2464     __ get_cache_and_index_at_bcp(c_rarg2, rscratch1, 1);
2465 
2466     if (is_static) {
2467       // Life is simple.  Null out the object pointer.
2468       __ xorl(c_rarg1, c_rarg1);
2469     } else {
2470       // Life is harder. The stack holds the value on top, followed by
2471       // the object.  We don't know the size of the value, though; it
2472       // could be one or two words depending on its type. As a result,
2473       // we must find the type to determine where the object is.
2474       __ movl(c_rarg3, Address(c_rarg2, rscratch1,
2475                            Address::times_8,
2476                            in_bytes(cp_base_offset +
2477                                      ConstantPoolCacheEntry::flags_offset())));
2478       __ shrl(c_rarg3, ConstantPoolCacheEntry::tos_state_shift);
2479       // Make sure we don't need to mask rcx after the above shift
2480       ConstantPoolCacheEntry::verify_tos_state_shift();
2481       __ movptr(c_rarg1, at_tos_p1());  // initially assume a one word jvalue
2482       __ cmpl(c_rarg3, ltos);
2483       __ cmovptr(Assembler::equal,
2484                  c_rarg1, at_tos_p2()); // ltos (two word jvalue)
2485       __ cmpl(c_rarg3, dtos);
2486       __ cmovptr(Assembler::equal,
2487                  c_rarg1, at_tos_p2()); // dtos (two word jvalue)
2488     }
2489     // cache entry pointer
2490     __ addptr(c_rarg2, in_bytes(cp_base_offset));
2491     __ shll(rscratch1, LogBytesPerWord);
2492     __ addptr(c_rarg2, rscratch1);
2493     // object (tos)
2494     __ mov(c_rarg3, rsp);
2495     // c_rarg1: object pointer set up above (NULL if static)
2496     // c_rarg2: cache entry pointer
2497     // c_rarg3: jvalue object on the stack
2498     __ call_VM(noreg,
2499                CAST_FROM_FN_PTR(address,
2500                                 InterpreterRuntime::post_field_modification),
2501                c_rarg1, c_rarg2, c_rarg3);
2502     __ get_cache_and_index_at_bcp(cache, index, 1);
2503     __ bind(L1);
2504   }
2505 }
2506 
2507 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2508   transition(vtos, vtos);
2509 
2510   const Register cache = rcx;
2511   const Register index = rdx;
2512   const Register obj   = rcx;
2513   const Register off   = rbx;
2514   const Register flags = rax;
2515   const Register bc    = c_rarg3;
2516 
2517   resolve_cache_and_index(byte_no, cache, index, sizeof(u2));
2518   jvmti_post_field_mod(cache, index, is_static);
2519   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2520 
2521   // [jk] not needed currently
2522   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2523   //                                              Assembler::StoreStore));
2524 
2525   Label notVolatile, Done;
2526   __ movl(rdx, flags);
2527   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2528   __ andl(rdx, 0x1);
2529 
2530   // field address
2531   const Address field(obj, off, Address::times_1);
2532 
2533   Label notByte, notBool, notInt, notShort, notChar,
2534         notLong, notFloat, notObj, notDouble;
2535 
2536   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
2537 
2538   assert(btos == 0, "change code, btos != 0");
2539   __ andl(flags, ConstantPoolCacheEntry::tos_state_mask);
2540   __ jcc(Assembler::notZero, notByte);
2541 
2542   // btos
2543   {
2544     __ pop(btos);
2545     if (!is_static) pop_and_check_object(obj);
2546     __ movb(field, rax);
2547     if (!is_static) {
2548       patch_bytecode(Bytecodes::_fast_bputfield, bc, rbx, true, byte_no);
2549     }
2550     __ jmp(Done);
2551   }
2552 
2553   __ bind(notByte);
2554   __ cmpl(flags, ztos);
2555   __ jcc(Assembler::notEqual, notBool);
2556 
2557   // ztos
2558   {
2559     __ pop(ztos);
2560     if (!is_static) pop_and_check_object(obj);
2561     __ andl(rax, 0x1);
2562     __ movb(field, rax);
2563     if (!is_static) {
2564       patch_bytecode(Bytecodes::_fast_zputfield, bc, rbx, true, byte_no);
2565     }
2566     __ jmp(Done);
2567   }
2568 
2569   __ bind(notBool);
2570   __ cmpl(flags, atos);
2571   __ jcc(Assembler::notEqual, notObj);
2572 
2573   // atos
2574   {
2575     __ pop(atos);
2576     if (!is_static) pop_and_check_object(obj);
2577     // Store into the field
2578     do_oop_store(_masm, field, rax, _bs->kind(), false);
2579     if (!is_static) {
2580       patch_bytecode(Bytecodes::_fast_aputfield, bc, rbx, true, byte_no);
2581     }
2582     __ jmp(Done);
2583   }
2584 
2585   __ bind(notObj);
2586   __ cmpl(flags, itos);
2587   __ jcc(Assembler::notEqual, notInt);
2588 
2589   // itos
2590   {
2591     __ pop(itos);
2592     if (!is_static) pop_and_check_object(obj);
2593     __ movl(field, rax);
2594     if (!is_static) {
2595       patch_bytecode(Bytecodes::_fast_iputfield, bc, rbx, true, byte_no);
2596     }
2597     __ jmp(Done);
2598   }
2599 
2600   __ bind(notInt);
2601   __ cmpl(flags, ctos);
2602   __ jcc(Assembler::notEqual, notChar);
2603 
2604   // ctos
2605   {
2606     __ pop(ctos);
2607     if (!is_static) pop_and_check_object(obj);
2608     __ movw(field, rax);
2609     if (!is_static) {
2610       patch_bytecode(Bytecodes::_fast_cputfield, bc, rbx, true, byte_no);
2611     }
2612     __ jmp(Done);
2613   }
2614 
2615   __ bind(notChar);
2616   __ cmpl(flags, stos);
2617   __ jcc(Assembler::notEqual, notShort);
2618 
2619   // stos
2620   {
2621     __ pop(stos);
2622     if (!is_static) pop_and_check_object(obj);
2623     __ movw(field, rax);
2624     if (!is_static) {
2625       patch_bytecode(Bytecodes::_fast_sputfield, bc, rbx, true, byte_no);
2626     }
2627     __ jmp(Done);
2628   }
2629 
2630   __ bind(notShort);
2631   __ cmpl(flags, ltos);
2632   __ jcc(Assembler::notEqual, notLong);
2633 
2634   // ltos
2635   {
2636     __ pop(ltos);
2637     if (!is_static) pop_and_check_object(obj);
2638     __ movq(field, rax);
2639     if (!is_static) {
2640       patch_bytecode(Bytecodes::_fast_lputfield, bc, rbx, true, byte_no);
2641     }
2642     __ jmp(Done);
2643   }
2644 
2645   __ bind(notLong);
2646   __ cmpl(flags, ftos);
2647   __ jcc(Assembler::notEqual, notFloat);
2648 
2649   // ftos
2650   {
2651     __ pop(ftos);
2652     if (!is_static) pop_and_check_object(obj);
2653     __ movflt(field, xmm0);
2654     if (!is_static) {
2655       patch_bytecode(Bytecodes::_fast_fputfield, bc, rbx, true, byte_no);
2656     }
2657     __ jmp(Done);
2658   }
2659 
2660   __ bind(notFloat);
2661 #ifdef ASSERT
2662   __ cmpl(flags, dtos);
2663   __ jcc(Assembler::notEqual, notDouble);
2664 #endif
2665 
2666   // dtos
2667   {
2668     __ pop(dtos);
2669     if (!is_static) pop_and_check_object(obj);
2670     __ movdbl(field, xmm0);
2671     if (!is_static) {
2672       patch_bytecode(Bytecodes::_fast_dputfield, bc, rbx, true, byte_no);
2673     }
2674   }
2675 
2676 #ifdef ASSERT
2677   __ jmp(Done);
2678 
2679   __ bind(notDouble);
2680   __ stop("Bad state");
2681 #endif
2682 
2683   __ bind(Done);
2684 
2685   // Check for volatile store
2686   __ testl(rdx, rdx);
2687   __ jcc(Assembler::zero, notVolatile);
2688   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2689                                                Assembler::StoreStore));
2690   __ bind(notVolatile);
2691 }
2692 
2693 void TemplateTable::putfield(int byte_no) {
2694   putfield_or_static(byte_no, false);
2695 }
2696 
2697 void TemplateTable::putstatic(int byte_no) {
2698   putfield_or_static(byte_no, true);
2699 }
2700 
2701 void TemplateTable::jvmti_post_fast_field_mod() {
2702   if (JvmtiExport::can_post_field_modification()) {
2703     // Check to see if a field modification watch has been set before
2704     // we take the time to call into the VM.
2705     Label L2;
2706     __ mov32(c_rarg3, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2707     __ testl(c_rarg3, c_rarg3);
2708     __ jcc(Assembler::zero, L2);
2709     __ pop_ptr(rbx);                  // copy the object pointer from tos
2710     __ verify_oop(rbx);
2711     __ push_ptr(rbx);                 // put the object pointer back on tos
2712     // Save tos values before call_VM() clobbers them. Since we have
2713     // to do it for every data type, we use the saved values as the
2714     // jvalue object.
2715     switch (bytecode()) {          // load values into the jvalue object
2716     case Bytecodes::_fast_aputfield: __ push_ptr(rax); break;
2717     case Bytecodes::_fast_bputfield: // fall through
2718     case Bytecodes::_fast_zputfield: // fall through
2719     case Bytecodes::_fast_sputfield: // fall through
2720     case Bytecodes::_fast_cputfield: // fall through
2721     case Bytecodes::_fast_iputfield: __ push_i(rax); break;
2722     case Bytecodes::_fast_dputfield: __ push_d(); break;
2723     case Bytecodes::_fast_fputfield: __ push_f(); break;
2724     case Bytecodes::_fast_lputfield: __ push_l(rax); break;
2725 
2726     default:
2727       ShouldNotReachHere();
2728     }
2729     __ mov(c_rarg3, rsp);             // points to jvalue on the stack
2730     // access constant pool cache entry
2731     __ get_cache_entry_pointer_at_bcp(c_rarg2, rax, 1);
2732     __ verify_oop(rbx);
2733     // rbx: object pointer copied above
2734     // c_rarg2: cache entry pointer
2735     // c_rarg3: jvalue object on the stack
2736     __ call_VM(noreg,
2737                CAST_FROM_FN_PTR(address,
2738                                 InterpreterRuntime::post_field_modification),
2739                rbx, c_rarg2, c_rarg3);
2740 
2741     switch (bytecode()) {             // restore tos values
2742     case Bytecodes::_fast_aputfield: __ pop_ptr(rax); break;
2743     case Bytecodes::_fast_bputfield: // fall through
2744     case Bytecodes::_fast_zputfield: // fall through
2745     case Bytecodes::_fast_sputfield: // fall through
2746     case Bytecodes::_fast_cputfield: // fall through
2747     case Bytecodes::_fast_iputfield: __ pop_i(rax); break;
2748     case Bytecodes::_fast_dputfield: __ pop_d(); break;
2749     case Bytecodes::_fast_fputfield: __ pop_f(); break;
2750     case Bytecodes::_fast_lputfield: __ pop_l(rax); break;
2751     }
2752     __ bind(L2);
2753   }
2754 }
2755 
2756 void TemplateTable::fast_storefield(TosState state) {
2757   transition(state, vtos);
2758 
2759   ByteSize base = ConstantPoolCache::base_offset();
2760 
2761   jvmti_post_fast_field_mod();
2762 
2763   // access constant pool cache
2764   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2765 
2766   // test for volatile with rdx
2767   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2768                        in_bytes(base +
2769                                 ConstantPoolCacheEntry::flags_offset())));
2770 
2771   // replace index with field offset from cache entry
2772   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2773                          in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2774 
2775   // [jk] not needed currently
2776   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2777   //                                              Assembler::StoreStore));
2778 
2779   Label notVolatile;
2780   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2781   __ andl(rdx, 0x1);
2782 
2783   // Get object from stack
2784   pop_and_check_object(rcx);
2785 
2786   // field address
2787   const Address field(rcx, rbx, Address::times_1);
2788 
2789   // access field
2790   switch (bytecode()) {
2791   case Bytecodes::_fast_aputfield:
2792     do_oop_store(_masm, field, rax, _bs->kind(), false);
2793     break;
2794   case Bytecodes::_fast_lputfield:
2795     __ movq(field, rax);
2796     break;
2797   case Bytecodes::_fast_iputfield:
2798     __ movl(field, rax);
2799     break;
2800   case Bytecodes::_fast_zputfield:
2801     __ andl(rax, 0x1);  // boolean is true if LSB is 1
2802     // fall through to bputfield
2803   case Bytecodes::_fast_bputfield:
2804     __ movb(field, rax);
2805     break;
2806   case Bytecodes::_fast_sputfield:
2807     // fall through
2808   case Bytecodes::_fast_cputfield:
2809     __ movw(field, rax);
2810     break;
2811   case Bytecodes::_fast_fputfield:
2812     __ movflt(field, xmm0);
2813     break;
2814   case Bytecodes::_fast_dputfield:
2815     __ movdbl(field, xmm0);
2816     break;
2817   default:
2818     ShouldNotReachHere();
2819   }
2820 
2821   // Check for volatile store
2822   __ testl(rdx, rdx);
2823   __ jcc(Assembler::zero, notVolatile);
2824   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2825                                                Assembler::StoreStore));
2826   __ bind(notVolatile);
2827 }
2828 
2829 
2830 void TemplateTable::fast_accessfield(TosState state) {
2831   transition(atos, state);
2832 
2833   // Do the JVMTI work here to avoid disturbing the register state below
2834   if (JvmtiExport::can_post_field_access()) {
2835     // Check to see if a field access watch has been set before we
2836     // take the time to call into the VM.
2837     Label L1;
2838     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2839     __ testl(rcx, rcx);
2840     __ jcc(Assembler::zero, L1);
2841     // access constant pool cache entry
2842     __ get_cache_entry_pointer_at_bcp(c_rarg2, rcx, 1);
2843     __ verify_oop(rax);
2844     __ push_ptr(rax);  // save object pointer before call_VM() clobbers it
2845     __ mov(c_rarg1, rax);
2846     // c_rarg1: object pointer copied above
2847     // c_rarg2: cache entry pointer
2848     __ call_VM(noreg,
2849                CAST_FROM_FN_PTR(address,
2850                                 InterpreterRuntime::post_field_access),
2851                c_rarg1, c_rarg2);
2852     __ pop_ptr(rax); // restore object pointer
2853     __ bind(L1);
2854   }
2855 
2856   // access constant pool cache
2857   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2858   // replace index with field offset from cache entry
2859   // [jk] not needed currently
2860   // if (os::is_MP()) {
2861   //   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2862   //                        in_bytes(ConstantPoolCache::base_offset() +
2863   //                                 ConstantPoolCacheEntry::flags_offset())));
2864   //   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2865   //   __ andl(rdx, 0x1);
2866   // }
2867   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2868                          in_bytes(ConstantPoolCache::base_offset() +
2869                                   ConstantPoolCacheEntry::f2_offset())));
2870 
2871   // rax: object
2872   __ verify_oop(rax);
2873   __ null_check(rax);
2874   Address field(rax, rbx, Address::times_1);
2875 
2876   // access field
2877   switch (bytecode()) {
2878   case Bytecodes::_fast_agetfield:
2879     __ load_heap_oop(rax, field);
2880     __ verify_oop(rax);
2881     break;
2882   case Bytecodes::_fast_lgetfield:
2883     __ movq(rax, field);
2884     break;
2885   case Bytecodes::_fast_igetfield:
2886     __ movl(rax, field);
2887     break;
2888   case Bytecodes::_fast_bgetfield:
2889     __ movsbl(rax, field);
2890     break;
2891   case Bytecodes::_fast_sgetfield:
2892     __ load_signed_short(rax, field);
2893     break;
2894   case Bytecodes::_fast_cgetfield:
2895     __ load_unsigned_short(rax, field);
2896     break;
2897   case Bytecodes::_fast_fgetfield:
2898     __ movflt(xmm0, field);
2899     break;
2900   case Bytecodes::_fast_dgetfield:
2901     __ movdbl(xmm0, field);
2902     break;
2903   default:
2904     ShouldNotReachHere();
2905   }
2906   // [jk] not needed currently
2907   // if (os::is_MP()) {
2908   //   Label notVolatile;
2909   //   __ testl(rdx, rdx);
2910   //   __ jcc(Assembler::zero, notVolatile);
2911   //   __ membar(Assembler::LoadLoad);
2912   //   __ bind(notVolatile);
2913   //};
2914 }
2915 
2916 void TemplateTable::fast_xaccess(TosState state) {
2917   transition(vtos, state);
2918 
2919   // get receiver
2920   __ movptr(rax, aaddress(0));
2921   // access constant pool cache
2922   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2923   __ movptr(rbx,
2924             Address(rcx, rdx, Address::times_8,
2925                     in_bytes(ConstantPoolCache::base_offset() +
2926                              ConstantPoolCacheEntry::f2_offset())));
2927   // make sure exception is reported in correct bcp range (getfield is
2928   // next instruction)
2929   __ increment(r13);
2930   __ null_check(rax);
2931   switch (state) {
2932   case itos:
2933     __ movl(rax, Address(rax, rbx, Address::times_1));
2934     break;
2935   case atos:
2936     __ load_heap_oop(rax, Address(rax, rbx, Address::times_1));
2937     __ verify_oop(rax);
2938     break;
2939   case ftos:
2940     __ movflt(xmm0, Address(rax, rbx, Address::times_1));
2941     break;
2942   default:
2943     ShouldNotReachHere();
2944   }
2945 
2946   // [jk] not needed currently
2947   // if (os::is_MP()) {
2948   //   Label notVolatile;
2949   //   __ movl(rdx, Address(rcx, rdx, Address::times_8,
2950   //                        in_bytes(ConstantPoolCache::base_offset() +
2951   //                                 ConstantPoolCacheEntry::flags_offset())));
2952   //   __ shrl(rdx, ConstantPoolCacheEntry::is_volatile_shift);
2953   //   __ testl(rdx, 0x1);
2954   //   __ jcc(Assembler::zero, notVolatile);
2955   //   __ membar(Assembler::LoadLoad);
2956   //   __ bind(notVolatile);
2957   // }
2958 
2959   __ decrement(r13);
2960 }
2961 
2962 
2963 
2964 //-----------------------------------------------------------------------------
2965 // Calls
2966 
2967 void TemplateTable::count_calls(Register method, Register temp) {
2968   // implemented elsewhere
2969   ShouldNotReachHere();
2970 }
2971 
2972 void TemplateTable::prepare_invoke(int byte_no,
2973                                    Register method,  // linked method (or i-klass)
2974                                    Register index,   // itable index, MethodType, etc.
2975                                    Register recv,    // if caller wants to see it
2976                                    Register flags    // if caller wants to test it
2977                                    ) {
2978   // determine flags
2979   const Bytecodes::Code code = bytecode();
2980   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
2981   const bool is_invokedynamic    = code == Bytecodes::_invokedynamic;
2982   const bool is_invokehandle     = code == Bytecodes::_invokehandle;
2983   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
2984   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
2985   const bool load_receiver       = (recv  != noreg);
2986   const bool save_flags          = (flags != noreg);
2987   assert(load_receiver == (code != Bytecodes::_invokestatic && code != Bytecodes::_invokedynamic), "");
2988   assert(save_flags    == (is_invokeinterface || is_invokevirtual), "need flags for vfinal");
2989   assert(flags == noreg || flags == rdx, "");
2990   assert(recv  == noreg || recv  == rcx, "");
2991 
2992   // setup registers & access constant pool cache
2993   if (recv  == noreg)  recv  = rcx;
2994   if (flags == noreg)  flags = rdx;
2995   assert_different_registers(method, index, recv, flags);
2996 
2997   // save 'interpreter return address'
2998   __ save_bcp();
2999 
3000   load_invoke_cp_cache_entry(byte_no, method, index, flags, is_invokevirtual, false, is_invokedynamic);
3001 
3002   // maybe push appendix to arguments (just before return address)
3003   if (is_invokedynamic || is_invokehandle) {
3004     Label L_no_push;
3005     __ testl(flags, (1 << ConstantPoolCacheEntry::has_appendix_shift));
3006     __ jcc(Assembler::zero, L_no_push);
3007     // Push the appendix as a trailing parameter.
3008     // This must be done before we get the receiver,
3009     // since the parameter_size includes it.
3010     __ push(rbx);
3011     __ mov(rbx, index);
3012     assert(ConstantPoolCacheEntry::_indy_resolved_references_appendix_offset == 0, "appendix expected at index+0");
3013     __ load_resolved_reference_at_index(index, rbx);
3014     __ pop(rbx);
3015     __ push(index);  // push appendix (MethodType, CallSite, etc.)
3016     __ bind(L_no_push);
3017   }
3018 
3019   // load receiver if needed (after appendix is pushed so parameter size is correct)
3020   // Note: no return address pushed yet
3021   if (load_receiver) {
3022     __ movl(recv, flags);
3023     __ andl(recv, ConstantPoolCacheEntry::parameter_size_mask);
3024     const int no_return_pc_pushed_yet = -1;  // argument slot correction before we push return address
3025     const int receiver_is_at_end      = -1;  // back off one slot to get receiver
3026     Address recv_addr = __ argument_address(recv, no_return_pc_pushed_yet + receiver_is_at_end);
3027     __ movptr(recv, recv_addr);
3028     __ verify_oop(recv);
3029   }
3030 
3031   if (save_flags) {
3032     __ movl(r13, flags);
3033   }
3034 
3035   // compute return type
3036   __ shrl(flags, ConstantPoolCacheEntry::tos_state_shift);
3037   // Make sure we don't need to mask flags after the above shift
3038   ConstantPoolCacheEntry::verify_tos_state_shift();
3039   // load return address
3040   {
3041     const address table_addr = (address) Interpreter::invoke_return_entry_table_for(code);
3042     ExternalAddress table(table_addr);
3043     __ lea(rscratch1, table);
3044     __ movptr(flags, Address(rscratch1, flags, Address::times_ptr));
3045   }
3046 
3047   // push return address
3048   __ push(flags);
3049 
3050   // Restore flags value from the constant pool cache, and restore rsi
3051   // for later null checks.  r13 is the bytecode pointer
3052   if (save_flags) {
3053     __ movl(flags, r13);
3054     __ restore_bcp();
3055   }
3056 }
3057 
3058 
3059 void TemplateTable::invokevirtual_helper(Register index,
3060                                          Register recv,
3061                                          Register flags) {
3062   // Uses temporary registers rax, rdx
3063   assert_different_registers(index, recv, rax, rdx);
3064   assert(index == rbx, "");
3065   assert(recv  == rcx, "");
3066 
3067   // Test for an invoke of a final method
3068   Label notFinal;
3069   __ movl(rax, flags);
3070   __ andl(rax, (1 << ConstantPoolCacheEntry::is_vfinal_shift));
3071   __ jcc(Assembler::zero, notFinal);
3072 
3073   const Register method = index;  // method must be rbx
3074   assert(method == rbx,
3075          "Method* must be rbx for interpreter calling convention");
3076 
3077   // do the call - the index is actually the method to call
3078   // that is, f2 is a vtable index if !is_vfinal, else f2 is a Method*
3079 
3080   // It's final, need a null check here!
3081   __ null_check(recv);
3082 
3083   // profile this call
3084   __ profile_final_call(rax);
3085   __ profile_arguments_type(rax, method, r13, true);
3086 
3087   __ jump_from_interpreted(method, rax);
3088 
3089   __ bind(notFinal);
3090 
3091   // get receiver klass
3092   __ null_check(recv, oopDesc::klass_offset_in_bytes());
3093   __ load_klass(rax, recv);
3094 
3095   // profile this call
3096   __ profile_virtual_call(rax, r14, rdx);
3097 
3098   // get target Method* & entry point
3099   __ lookup_virtual_method(rax, index, method);
3100   __ profile_arguments_type(rdx, method, r13, true);
3101   __ jump_from_interpreted(method, rdx);
3102 }
3103 
3104 
3105 void TemplateTable::invokevirtual(int byte_no) {
3106   transition(vtos, vtos);
3107   assert(byte_no == f2_byte, "use this argument");
3108   prepare_invoke(byte_no,
3109                  rbx,    // method or vtable index
3110                  noreg,  // unused itable index
3111                  rcx, rdx); // recv, flags
3112 
3113   // rbx: index
3114   // rcx: receiver
3115   // rdx: flags
3116 
3117   invokevirtual_helper(rbx, rcx, rdx);
3118 }
3119 
3120 
3121 void TemplateTable::invokespecial(int byte_no) {
3122   transition(vtos, vtos);
3123   assert(byte_no == f1_byte, "use this argument");
3124   prepare_invoke(byte_no, rbx, noreg,  // get f1 Method*
3125                  rcx);  // get receiver also for null check
3126   __ verify_oop(rcx);
3127   __ null_check(rcx);
3128   // do the call
3129   __ profile_call(rax);
3130   __ profile_arguments_type(rax, rbx, r13, false);
3131   __ jump_from_interpreted(rbx, rax);
3132 }
3133 
3134 
3135 void TemplateTable::invokestatic(int byte_no) {
3136   transition(vtos, vtos);
3137   assert(byte_no == f1_byte, "use this argument");
3138   prepare_invoke(byte_no, rbx);  // get f1 Method*
3139   // do the call
3140   __ profile_call(rax);
3141   __ profile_arguments_type(rax, rbx, r13, false);
3142   __ jump_from_interpreted(rbx, rax);
3143 }
3144 
3145 void TemplateTable::fast_invokevfinal(int byte_no) {
3146   transition(vtos, vtos);
3147   assert(byte_no == f2_byte, "use this argument");
3148   __ stop("fast_invokevfinal not used on amd64");
3149 }
3150 
3151 void TemplateTable::invokeinterface(int byte_no) {
3152   transition(vtos, vtos);
3153   assert(byte_no == f1_byte, "use this argument");
3154   prepare_invoke(byte_no, rax, rbx,  // get f1 Klass*, f2 Method*
3155                  rcx, rdx); // recv, flags
3156 
3157   // rax: reference klass (from f1)
3158   // rbx: method (from f2)
3159   // rcx: receiver
3160   // rdx: flags
3161 
3162   // Special case of invokeinterface called for virtual method of
3163   // java.lang.Object.  See cpCacheOop.cpp for details.
3164   // This code isn't produced by javac, but could be produced by
3165   // another compliant java compiler.
3166   Label notMethod;
3167   __ movl(r14, rdx);
3168   __ andl(r14, (1 << ConstantPoolCacheEntry::is_forced_virtual_shift));
3169   __ jcc(Assembler::zero, notMethod);
3170 
3171   invokevirtual_helper(rbx, rcx, rdx);
3172   __ bind(notMethod);
3173 
3174   // Get receiver klass into rdx - also a null check
3175   __ restore_locals();  // restore r14
3176   __ null_check(rcx, oopDesc::klass_offset_in_bytes());
3177   __ load_klass(rdx, rcx);
3178 
3179   Label no_such_interface, no_such_method;
3180 
3181   // Receiver subtype check against REFC.
3182   // Superklass in rax. Subklass in rdx. Blows rcx, rdi.
3183   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3184                              rdx, rax, noreg,
3185                              // outputs: scan temp. reg, scan temp. reg
3186                              r13, r14,
3187                              no_such_interface,
3188                              /*return_method=*/false);
3189 
3190   // profile this call
3191   __ restore_bcp(); // rbcp was destroyed by receiver type check
3192   __ profile_virtual_call(rdx, r13, r14);
3193 
3194   // Get declaring interface class from method, and itable index
3195   __ movptr(rax, Address(rbx, Method::const_offset()));
3196   __ movptr(rax, Address(rax, ConstMethod::constants_offset()));
3197   __ movptr(rax, Address(rax, ConstantPool::pool_holder_offset_in_bytes()));
3198   __ movl(rbx, Address(rbx, Method::itable_index_offset()));
3199   __ subl(rbx, Method::itable_index_max);
3200   __ negl(rbx);
3201 
3202   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3203                              rdx, rax, rbx,
3204                              // outputs: method, scan temp. reg
3205                              rbx, r13,
3206                              no_such_interface);
3207 
3208   // rbx: Method* to call
3209   // rcx: receiver
3210   // Check for abstract method error
3211   // Note: This should be done more efficiently via a throw_abstract_method_error
3212   //       interpreter entry point and a conditional jump to it in case of a null
3213   //       method.
3214   __ testptr(rbx, rbx);
3215   __ jcc(Assembler::zero, no_such_method);
3216 
3217   __ profile_arguments_type(rdx, rbx, r13, true);
3218 
3219   // do the call
3220   // rcx: receiver
3221   // rbx,: Method*
3222   __ jump_from_interpreted(rbx, rdx);
3223   __ should_not_reach_here();
3224 
3225   // exception handling code follows...
3226   // note: must restore interpreter registers to canonical
3227   //       state for exception handling to work correctly!
3228 
3229   __ bind(no_such_method);
3230   // throw exception
3231   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3232   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3233   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3234   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3235   // the call_VM checks for exception, so we should never return here.
3236   __ should_not_reach_here();
3237 
3238   __ bind(no_such_interface);
3239   // throw exception
3240   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3241   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3242   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3243   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3244                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3245   // the call_VM checks for exception, so we should never return here.
3246   __ should_not_reach_here();
3247 }
3248 
3249 
3250 void TemplateTable::invokehandle(int byte_no) {
3251   transition(vtos, vtos);
3252   assert(byte_no == f1_byte, "use this argument");
3253   const Register rbx_method = rbx;
3254   const Register rax_mtype  = rax;
3255   const Register rcx_recv   = rcx;
3256   const Register rdx_flags  = rdx;
3257 
3258   if (!EnableInvokeDynamic) {
3259     // rewriter does not generate this bytecode
3260     __ should_not_reach_here();
3261     return;
3262   }
3263 
3264   prepare_invoke(byte_no, rbx_method, rax_mtype, rcx_recv);
3265   __ verify_method_ptr(rbx_method);
3266   __ verify_oop(rcx_recv);
3267   __ null_check(rcx_recv);
3268 
3269   // rax: MethodType object (from cpool->resolved_references[f1], if necessary)
3270   // rbx: MH.invokeExact_MT method (from f2)
3271 
3272   // Note:  rax_mtype is already pushed (if necessary) by prepare_invoke
3273 
3274   // FIXME: profile the LambdaForm also
3275   __ profile_final_call(rax);
3276   __ profile_arguments_type(rdx, rbx_method, r13, true);
3277 
3278   __ jump_from_interpreted(rbx_method, rdx);
3279 }
3280 
3281 
3282 void TemplateTable::invokedynamic(int byte_no) {
3283   transition(vtos, vtos);
3284   assert(byte_no == f1_byte, "use this argument");
3285 
3286   if (!EnableInvokeDynamic) {
3287     // We should not encounter this bytecode if !EnableInvokeDynamic.
3288     // The verifier will stop it.  However, if we get past the verifier,
3289     // this will stop the thread in a reasonable way, without crashing the JVM.
3290     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3291                      InterpreterRuntime::throw_IncompatibleClassChangeError));
3292     // the call_VM checks for exception, so we should never return here.
3293     __ should_not_reach_here();
3294     return;
3295   }
3296 
3297   const Register rbx_method   = rbx;
3298   const Register rax_callsite = rax;
3299 
3300   prepare_invoke(byte_no, rbx_method, rax_callsite);
3301 
3302   // rax: CallSite object (from cpool->resolved_references[f1])
3303   // rbx: MH.linkToCallSite method (from f2)
3304 
3305   // Note:  rax_callsite is already pushed by prepare_invoke
3306 
3307   // %%% should make a type profile for any invokedynamic that takes a ref argument
3308   // profile this call
3309   __ profile_call(r13);
3310   __ profile_arguments_type(rdx, rbx_method, r13, false);
3311 
3312   __ verify_oop(rax_callsite);
3313 
3314   __ jump_from_interpreted(rbx_method, rdx);
3315 }
3316 
3317 
3318 //-----------------------------------------------------------------------------
3319 // Allocation
3320 
3321 void TemplateTable::_new() {
3322   transition(vtos, atos);
3323   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3324   Label slow_case;
3325   Label done;
3326   Label initialize_header;
3327   Label initialize_object; // including clearing the fields
3328   Label allocate_shared;
3329 
3330   __ get_cpool_and_tags(rsi, rax);
3331   // Make sure the class we're about to instantiate has been resolved.
3332   // This is done before loading InstanceKlass to be consistent with the order
3333   // how Constant Pool is updated (see ConstantPool::klass_at_put)
3334   const int tags_offset = Array<u1>::base_offset_in_bytes();
3335   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset),
3336           JVM_CONSTANT_Class);
3337   __ jcc(Assembler::notEqual, slow_case);
3338 
3339   // get InstanceKlass
3340   __ movptr(rsi, Address(rsi, rdx,
3341             Address::times_8, sizeof(ConstantPool)));
3342 
3343   // make sure klass is initialized & doesn't have finalizer
3344   // make sure klass is fully initialized
3345   __ cmpb(Address(rsi,
3346                   InstanceKlass::init_state_offset()),
3347           InstanceKlass::fully_initialized);
3348   __ jcc(Assembler::notEqual, slow_case);
3349 
3350   // get instance_size in InstanceKlass (scaled to a count of bytes)
3351   __ movl(rdx,
3352           Address(rsi,
3353                   Klass::layout_helper_offset()));
3354   // test to see if it has a finalizer or is malformed in some way
3355   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3356   __ jcc(Assembler::notZero, slow_case);
3357 
3358   // Allocate the instance
3359   // 1) Try to allocate in the TLAB
3360   // 2) if fail and the object is large allocate in the shared Eden
3361   // 3) if the above fails (or is not applicable), go to a slow case
3362   // (creates a new TLAB, etc.)
3363 
3364   const bool allow_shared_alloc =
3365     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3366 
3367   if (UseTLAB) {
3368     __ movptr(rax, Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())));
3369     __ lea(rbx, Address(rax, rdx, Address::times_1));
3370     __ cmpptr(rbx, Address(r15_thread, in_bytes(JavaThread::tlab_end_offset())));
3371     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3372     __ movptr(Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3373     if (ZeroTLAB) {
3374       // the fields have been already cleared
3375       __ jmp(initialize_header);
3376     } else {
3377       // initialize both the header and fields
3378       __ jmp(initialize_object);
3379     }
3380   }
3381 
3382   // Allocation in the shared Eden, if allowed.
3383   //
3384   // rdx: instance size in bytes
3385   if (allow_shared_alloc) {
3386     __ bind(allocate_shared);
3387 
3388     ExternalAddress top((address)Universe::heap()->top_addr());
3389     ExternalAddress end((address)Universe::heap()->end_addr());
3390 
3391     const Register RtopAddr = rscratch1;
3392     const Register RendAddr = rscratch2;
3393 
3394     __ lea(RtopAddr, top);
3395     __ lea(RendAddr, end);
3396     __ movptr(rax, Address(RtopAddr, 0));
3397 
3398     // For retries rax gets set by cmpxchgq
3399     Label retry;
3400     __ bind(retry);
3401     __ lea(rbx, Address(rax, rdx, Address::times_1));
3402     __ cmpptr(rbx, Address(RendAddr, 0));
3403     __ jcc(Assembler::above, slow_case);
3404 
3405     // Compare rax with the top addr, and if still equal, store the new
3406     // top addr in rbx at the address of the top addr pointer. Sets ZF if was
3407     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3408     //
3409     // rax: object begin
3410     // rbx: object end
3411     // rdx: instance size in bytes
3412     if (os::is_MP()) {
3413       __ lock();
3414     }
3415     __ cmpxchgptr(rbx, Address(RtopAddr, 0));
3416 
3417     // if someone beat us on the allocation, try again, otherwise continue
3418     __ jcc(Assembler::notEqual, retry);
3419 
3420     __ incr_allocated_bytes(r15_thread, rdx, 0);
3421   }
3422 
3423   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3424     // The object is initialized before the header.  If the object size is
3425     // zero, go directly to the header initialization.
3426     __ bind(initialize_object);
3427     __ decrementl(rdx, sizeof(oopDesc));
3428     __ jcc(Assembler::zero, initialize_header);
3429 
3430     // Initialize object fields
3431     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3432     __ shrl(rdx, LogBytesPerLong);  // divide by oopSize to simplify the loop
3433     {
3434       Label loop;
3435       __ bind(loop);
3436       __ movq(Address(rax, rdx, Address::times_8,
3437                       sizeof(oopDesc) - oopSize),
3438               rcx);
3439       __ decrementl(rdx);
3440       __ jcc(Assembler::notZero, loop);
3441     }
3442 
3443     // initialize object header only.
3444     __ bind(initialize_header);
3445     if (UseBiasedLocking) {
3446       __ movptr(rscratch1, Address(rsi, Klass::prototype_header_offset()));
3447       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()), rscratch1);
3448     } else {
3449       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()),
3450                (intptr_t) markOopDesc::prototype()); // header (address 0x1)
3451     }
3452     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3453     __ store_klass_gap(rax, rcx);  // zero klass gap for compressed oops
3454     __ store_klass(rax, rsi);      // store klass last
3455 
3456     {
3457       SkipIfEqual skip(_masm, &DTraceAllocProbes, false);
3458       // Trigger dtrace event for fastpath
3459       __ push(atos); // save the return value
3460       __ call_VM_leaf(
3461            CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3462       __ pop(atos); // restore the return value
3463 
3464     }
3465     __ jmp(done);
3466   }
3467 
3468 
3469   // slow case
3470   __ bind(slow_case);
3471   __ get_constant_pool(c_rarg1);
3472   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3473   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), c_rarg1, c_rarg2);
3474   __ verify_oop(rax);
3475 
3476   // continue
3477   __ bind(done);
3478 }
3479 
3480 void TemplateTable::newarray() {
3481   transition(itos, atos);
3482   __ load_unsigned_byte(c_rarg1, at_bcp(1));
3483   __ movl(c_rarg2, rax);
3484   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray),
3485           c_rarg1, c_rarg2);
3486 }
3487 
3488 void TemplateTable::anewarray() {
3489   transition(itos, atos);
3490   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3491   __ get_constant_pool(c_rarg1);
3492   __ movl(c_rarg3, rax);
3493   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray),
3494           c_rarg1, c_rarg2, c_rarg3);
3495 }
3496 
3497 void TemplateTable::arraylength() {
3498   transition(atos, itos);
3499   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3500   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3501 }
3502 
3503 void TemplateTable::checkcast() {
3504   transition(atos, atos);
3505   Label done, is_null, ok_is_subtype, quicked, resolved;
3506   __ testptr(rax, rax); // object is in rax
3507   __ jcc(Assembler::zero, is_null);
3508 
3509   // Get cpool & tags index
3510   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3511   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3512   // See if bytecode has already been quicked
3513   __ cmpb(Address(rdx, rbx,
3514                   Address::times_1,
3515                   Array<u1>::base_offset_in_bytes()),
3516           JVM_CONSTANT_Class);
3517   __ jcc(Assembler::equal, quicked);
3518   __ push(atos); // save receiver for result, and for GC
3519   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3520   // vm_result_2 has metadata result
3521   __ get_vm_result_2(rax, r15_thread);
3522   __ pop_ptr(rdx); // restore receiver
3523   __ jmpb(resolved);
3524 
3525   // Get superklass in rax and subklass in rbx
3526   __ bind(quicked);
3527   __ mov(rdx, rax); // Save object in rdx; rax needed for subtype check
3528   __ movptr(rax, Address(rcx, rbx,
3529                        Address::times_8, sizeof(ConstantPool)));
3530 
3531   __ bind(resolved);
3532   __ load_klass(rbx, rdx);
3533 
3534   // Generate subtype check.  Blows rcx, rdi.  Object in rdx.
3535   // Superklass in rax.  Subklass in rbx.
3536   __ gen_subtype_check(rbx, ok_is_subtype);
3537 
3538   // Come here on failure
3539   __ push_ptr(rdx);
3540   // object is at TOS
3541   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3542 
3543   // Come here on success
3544   __ bind(ok_is_subtype);
3545   __ mov(rax, rdx); // Restore object in rdx
3546 
3547   // Collect counts on whether this check-cast sees NULLs a lot or not.
3548   if (ProfileInterpreter) {
3549     __ jmp(done);
3550     __ bind(is_null);
3551     __ profile_null_seen(rcx);
3552   } else {
3553     __ bind(is_null);   // same as 'done'
3554   }
3555   __ bind(done);
3556 }
3557 
3558 void TemplateTable::instanceof() {
3559   transition(atos, itos);
3560   Label done, is_null, ok_is_subtype, quicked, resolved;
3561   __ testptr(rax, rax);
3562   __ jcc(Assembler::zero, is_null);
3563 
3564   // Get cpool & tags index
3565   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3566   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3567   // See if bytecode has already been quicked
3568   __ cmpb(Address(rdx, rbx,
3569                   Address::times_1,
3570                   Array<u1>::base_offset_in_bytes()),
3571           JVM_CONSTANT_Class);
3572   __ jcc(Assembler::equal, quicked);
3573 
3574   __ push(atos); // save receiver for result, and for GC
3575   call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3576   // vm_result_2 has metadata result
3577   __ get_vm_result_2(rax, r15_thread);
3578   __ pop_ptr(rdx); // restore receiver
3579   __ verify_oop(rdx);
3580   __ load_klass(rdx, rdx);
3581   __ jmpb(resolved);
3582 
3583   // Get superklass in rax and subklass in rdx
3584   __ bind(quicked);
3585   __ load_klass(rdx, rax);
3586   __ movptr(rax, Address(rcx, rbx,
3587                          Address::times_8, sizeof(ConstantPool)));
3588 
3589   __ bind(resolved);
3590 
3591   // Generate subtype check.  Blows rcx, rdi
3592   // Superklass in rax.  Subklass in rdx.
3593   __ gen_subtype_check(rdx, ok_is_subtype);
3594 
3595   // Come here on failure
3596   __ xorl(rax, rax);
3597   __ jmpb(done);
3598   // Come here on success
3599   __ bind(ok_is_subtype);
3600   __ movl(rax, 1);
3601 
3602   // Collect counts on whether this test sees NULLs a lot or not.
3603   if (ProfileInterpreter) {
3604     __ jmp(done);
3605     __ bind(is_null);
3606     __ profile_null_seen(rcx);
3607   } else {
3608     __ bind(is_null);   // same as 'done'
3609   }
3610   __ bind(done);
3611   // rax = 0: obj == NULL or  obj is not an instanceof the specified klass
3612   // rax = 1: obj != NULL and obj is     an instanceof the specified klass
3613 }
3614 
3615 //-----------------------------------------------------------------------------
3616 // Breakpoints
3617 void TemplateTable::_breakpoint() {
3618   // Note: We get here even if we are single stepping..
3619   // jbug inists on setting breakpoints at every bytecode
3620   // even if we are in single step mode.
3621 
3622   transition(vtos, vtos);
3623 
3624   // get the unpatched byte code
3625   __ get_method(c_rarg1);
3626   __ call_VM(noreg,
3627              CAST_FROM_FN_PTR(address,
3628                               InterpreterRuntime::get_original_bytecode_at),
3629              c_rarg1, r13);
3630   __ mov(rbx, rax);
3631 
3632   // post the breakpoint event
3633   __ get_method(c_rarg1);
3634   __ call_VM(noreg,
3635              CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint),
3636              c_rarg1, r13);
3637 
3638   // complete the execution of original bytecode
3639   __ dispatch_only_normal(vtos);
3640 }
3641 
3642 //-----------------------------------------------------------------------------
3643 // Exceptions
3644 
3645 void TemplateTable::athrow() {
3646   transition(atos, vtos);
3647   __ null_check(rax);
3648   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3649 }
3650 
3651 //-----------------------------------------------------------------------------
3652 // Synchronization
3653 //
3654 // Note: monitorenter & exit are symmetric routines; which is reflected
3655 //       in the assembly code structure as well
3656 //
3657 // Stack layout:
3658 //
3659 // [expressions  ] <--- rsp               = expression stack top
3660 // ..
3661 // [expressions  ]
3662 // [monitor entry] <--- monitor block top = expression stack bot
3663 // ..
3664 // [monitor entry]
3665 // [frame data   ] <--- monitor block bot
3666 // ...
3667 // [saved rbp    ] <--- rbp
3668 void TemplateTable::monitorenter() {
3669   transition(atos, vtos);
3670 
3671   // check for NULL object
3672   __ null_check(rax);
3673 
3674   const Address monitor_block_top(
3675         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3676   const Address monitor_block_bot(
3677         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3678   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3679 
3680   Label allocated;
3681 
3682   // initialize entry pointer
3683   __ xorl(c_rarg1, c_rarg1); // points to free slot or NULL
3684 
3685   // find a free slot in the monitor block (result in c_rarg1)
3686   {
3687     Label entry, loop, exit;
3688     __ movptr(c_rarg3, monitor_block_top); // points to current entry,
3689                                      // starting with top-most entry
3690     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3691                                      // of monitor block
3692     __ jmpb(entry);
3693 
3694     __ bind(loop);
3695     // check if current entry is used
3696     __ cmpptr(Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()), (int32_t) NULL_WORD);
3697     // if not used then remember entry in c_rarg1
3698     __ cmov(Assembler::equal, c_rarg1, c_rarg3);
3699     // check if current entry is for same object
3700     __ cmpptr(rax, Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()));
3701     // if same object then stop searching
3702     __ jccb(Assembler::equal, exit);
3703     // otherwise advance to next entry
3704     __ addptr(c_rarg3, entry_size);
3705     __ bind(entry);
3706     // check if bottom reached
3707     __ cmpptr(c_rarg3, c_rarg2);
3708     // if not at bottom then check this entry
3709     __ jcc(Assembler::notEqual, loop);
3710     __ bind(exit);
3711   }
3712 
3713   __ testptr(c_rarg1, c_rarg1); // check if a slot has been found
3714   __ jcc(Assembler::notZero, allocated); // if found, continue with that one
3715 
3716   // allocate one if there's no free slot
3717   {
3718     Label entry, loop;
3719     // 1. compute new pointers             // rsp: old expression stack top
3720     __ movptr(c_rarg1, monitor_block_bot); // c_rarg1: old expression stack bottom
3721     __ subptr(rsp, entry_size);            // move expression stack top
3722     __ subptr(c_rarg1, entry_size);        // move expression stack bottom
3723     __ mov(c_rarg3, rsp);                  // set start value for copy loop
3724     __ movptr(monitor_block_bot, c_rarg1); // set new monitor block bottom
3725     __ jmp(entry);
3726     // 2. move expression stack contents
3727     __ bind(loop);
3728     __ movptr(c_rarg2, Address(c_rarg3, entry_size)); // load expression stack
3729                                                       // word from old location
3730     __ movptr(Address(c_rarg3, 0), c_rarg2);          // and store it at new location
3731     __ addptr(c_rarg3, wordSize);                     // advance to next word
3732     __ bind(entry);
3733     __ cmpptr(c_rarg3, c_rarg1);            // check if bottom reached
3734     __ jcc(Assembler::notEqual, loop);      // if not at bottom then
3735                                             // copy next word
3736   }
3737 
3738   // call run-time routine
3739   // c_rarg1: points to monitor entry
3740   __ bind(allocated);
3741 
3742   // Increment bcp to point to the next bytecode, so exception
3743   // handling for async. exceptions work correctly.
3744   // The object has already been poped from the stack, so the
3745   // expression stack looks correct.
3746   __ increment(r13);
3747 
3748   // store object
3749   __ movptr(Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()), rax);
3750   __ lock_object(c_rarg1);
3751 
3752   // check to make sure this monitor doesn't cause stack overflow after locking
3753   __ save_bcp();  // in case of exception
3754   __ generate_stack_overflow_check(0);
3755 
3756   // The bcp has already been incremented. Just need to dispatch to
3757   // next instruction.
3758   __ dispatch_next(vtos);
3759 }
3760 
3761 
3762 void TemplateTable::monitorexit() {
3763   transition(atos, vtos);
3764 
3765   // check for NULL object
3766   __ null_check(rax);
3767 
3768   const Address monitor_block_top(
3769         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3770   const Address monitor_block_bot(
3771         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3772   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3773 
3774   Label found;
3775 
3776   // find matching slot
3777   {
3778     Label entry, loop;
3779     __ movptr(c_rarg1, monitor_block_top); // points to current entry,
3780                                      // starting with top-most entry
3781     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3782                                      // of monitor block
3783     __ jmpb(entry);
3784 
3785     __ bind(loop);
3786     // check if current entry is for same object
3787     __ cmpptr(rax, Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()));
3788     // if same object then stop searching
3789     __ jcc(Assembler::equal, found);
3790     // otherwise advance to next entry
3791     __ addptr(c_rarg1, entry_size);
3792     __ bind(entry);
3793     // check if bottom reached
3794     __ cmpptr(c_rarg1, c_rarg2);
3795     // if not at bottom then check this entry
3796     __ jcc(Assembler::notEqual, loop);
3797   }
3798 
3799   // error handling. Unlocking was not block-structured
3800   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3801                    InterpreterRuntime::throw_illegal_monitor_state_exception));
3802   __ should_not_reach_here();
3803 
3804   // call run-time routine
3805   // rsi: points to monitor entry
3806   __ bind(found);
3807   __ push_ptr(rax); // make sure object is on stack (contract with oopMaps)
3808   __ unlock_object(c_rarg1);
3809   __ pop_ptr(rax); // discard object
3810 }
3811 
3812 
3813 // Wide instructions
3814 void TemplateTable::wide() {
3815   transition(vtos, vtos);
3816   __ load_unsigned_byte(rbx, at_bcp(1));
3817   __ lea(rscratch1, ExternalAddress((address)Interpreter::_wentry_point));
3818   __ jmp(Address(rscratch1, rbx, Address::times_8));
3819   // Note: the r13 increment step is part of the individual wide
3820   // bytecode implementations
3821 }
3822 
3823 
3824 // Multi arrays
3825 void TemplateTable::multianewarray() {
3826   transition(vtos, atos);
3827   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3828   // last dim is on top of stack; we want address of first one:
3829   // first_addr = last_addr + (ndims - 1) * wordSize
3830   __ lea(c_rarg1, Address(rsp, rax, Address::times_8, -wordSize));
3831   call_VM(rax,
3832           CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray),
3833           c_rarg1);
3834   __ load_unsigned_byte(rbx, at_bcp(3));
3835   __ lea(rsp, Address(rsp, rbx, Address::times_8));
3836 }
3837 #endif // !CC_INTERP