1 /*
   2  * Copyright 2003-2009 Sun Microsystems, Inc.  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 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
  20  * CA 95054 USA or visit www.sun.com if you need additional information or
  21  * have any questions.
  22  *
  23  */
  24 
  25 #include "incls/_precompiled.incl"
  26 #include "incls/_templateTable_x86_64.cpp.incl"
  27 
  28 #ifndef CC_INTERP
  29 
  30 #define __ _masm->
  31 
  32 // Platform-dependent initialization
  33 
  34 void TemplateTable::pd_initialize() {
  35   // No amd64 specific initialization
  36 }
  37 
  38 // Address computation: local variables
  39 
  40 static inline Address iaddress(int n) {
  41   return Address(r14, Interpreter::local_offset_in_bytes(n));
  42 }
  43 
  44 static inline Address laddress(int n) {
  45   return iaddress(n + 1);
  46 }
  47 
  48 static inline Address faddress(int n) {
  49   return iaddress(n);
  50 }
  51 
  52 static inline Address daddress(int n) {
  53   return laddress(n);
  54 }
  55 
  56 static inline Address aaddress(int n) {
  57   return iaddress(n);
  58 }
  59 
  60 static inline Address iaddress(Register r) {
  61   return Address(r14, r, Address::times_8, Interpreter::value_offset_in_bytes());
  62 }
  63 
  64 static inline Address laddress(Register r) {
  65   return Address(r14, r, Address::times_8, Interpreter::local_offset_in_bytes(1));
  66 }
  67 
  68 static inline Address faddress(Register r) {
  69   return iaddress(r);
  70 }
  71 
  72 static inline Address daddress(Register r) {
  73   return laddress(r);
  74 }
  75 
  76 static inline Address aaddress(Register r) {
  77   return iaddress(r);
  78 }
  79 
  80 static inline Address at_rsp() {
  81   return Address(rsp, 0);
  82 }
  83 
  84 // At top of Java expression stack which may be different than esp().  It
  85 // isn't for category 1 objects.
  86 static inline Address at_tos   () {
  87   return Address(rsp,  Interpreter::expr_offset_in_bytes(0));
  88 }
  89 
  90 static inline Address at_tos_p1() {
  91   return Address(rsp,  Interpreter::expr_offset_in_bytes(1));
  92 }
  93 
  94 static inline Address at_tos_p2() {
  95   return Address(rsp,  Interpreter::expr_offset_in_bytes(2));
  96 }
  97 
  98 static inline Address at_tos_p3() {
  99   return Address(rsp,  Interpreter::expr_offset_in_bytes(3));
 100 }
 101 
 102 // Condition conversion
 103 static Assembler::Condition j_not(TemplateTable::Condition cc) {
 104   switch (cc) {
 105   case TemplateTable::equal        : return Assembler::notEqual;
 106   case TemplateTable::not_equal    : return Assembler::equal;
 107   case TemplateTable::less         : return Assembler::greaterEqual;
 108   case TemplateTable::less_equal   : return Assembler::greater;
 109   case TemplateTable::greater      : return Assembler::lessEqual;
 110   case TemplateTable::greater_equal: return Assembler::less;
 111   }
 112   ShouldNotReachHere();
 113   return Assembler::zero;
 114 }
 115 
 116 
 117 // Miscelaneous helper routines
 118 // Store an oop (or NULL) at the address described by obj.
 119 // If val == noreg this means store a NULL
 120 
 121 static void do_oop_store(InterpreterMacroAssembler* _masm,
 122                          Address obj,
 123                          Register val,
 124                          BarrierSet::Name barrier,
 125                          bool precise) {
 126   assert(val == noreg || val == rax, "parameter is just for looks");
 127   switch (barrier) {
 128 #ifndef SERIALGC
 129     case BarrierSet::G1SATBCT:
 130     case BarrierSet::G1SATBCTLogging:
 131       {
 132         // flatten object address if needed
 133         if (obj.index() == noreg && obj.disp() == 0) {
 134           if (obj.base() != rdx) {
 135             __ movq(rdx, obj.base());
 136           }
 137         } else {
 138           __ leaq(rdx, obj);
 139         }
 140         __ g1_write_barrier_pre(rdx, r8, rbx, val != noreg);
 141         if (val == noreg) {
 142           __ store_heap_oop_null(Address(rdx, 0));
 143         } else {
 144           __ store_heap_oop(Address(rdx, 0), val);
 145           __ g1_write_barrier_post(rdx, val, r8, rbx);
 146         }
 147 
 148       }
 149       break;
 150 #endif // SERIALGC
 151     case BarrierSet::CardTableModRef:
 152     case BarrierSet::CardTableExtension:
 153       {
 154         if (val == noreg) {
 155           __ store_heap_oop_null(obj);
 156         } else {
 157           __ store_heap_oop(obj, val);
 158           // flatten object address if needed
 159           if (!precise || (obj.index() == noreg && obj.disp() == 0)) {
 160             __ store_check(obj.base());
 161           } else {
 162             __ leaq(rdx, obj);
 163             __ store_check(rdx);
 164           }
 165         }
 166       }
 167       break;
 168     case BarrierSet::ModRef:
 169     case BarrierSet::Other:
 170       if (val == noreg) {
 171         __ store_heap_oop_null(obj);
 172       } else {
 173         __ store_heap_oop(obj, val);
 174       }
 175       break;
 176     default      :
 177       ShouldNotReachHere();
 178 
 179   }
 180 }
 181 
 182 Address TemplateTable::at_bcp(int offset) {
 183   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 184   return Address(r13, offset);
 185 }
 186 
 187 void TemplateTable::patch_bytecode(Bytecodes::Code bytecode, Register bc,
 188                                    Register scratch,
 189                                    bool load_bc_into_scratch/*=true*/) {
 190   if (!RewriteBytecodes) {
 191     return;
 192   }
 193   // the pair bytecodes have already done the load.
 194   if (load_bc_into_scratch) {
 195     __ movl(bc, bytecode);
 196   }
 197   Label patch_done;
 198   if (JvmtiExport::can_post_breakpoint()) {
 199     Label fast_patch;
 200     // if a breakpoint is present we can't rewrite the stream directly
 201     __ movzbl(scratch, at_bcp(0));
 202     __ cmpl(scratch, Bytecodes::_breakpoint);
 203     __ jcc(Assembler::notEqual, fast_patch);
 204     __ get_method(scratch);
 205     // Let breakpoint table handling rewrite to quicker bytecode
 206     __ call_VM(noreg,
 207                CAST_FROM_FN_PTR(address,
 208                                 InterpreterRuntime::set_original_bytecode_at),
 209                scratch, r13, bc);
 210 #ifndef ASSERT
 211     __ jmpb(patch_done);
 212     __ bind(fast_patch);
 213   }
 214 #else
 215     __ jmp(patch_done);
 216     __ bind(fast_patch);
 217   }
 218   Label okay;
 219   __ load_unsigned_byte(scratch, at_bcp(0));
 220   __ cmpl(scratch, (int) Bytecodes::java_code(bytecode));
 221   __ jcc(Assembler::equal, okay);
 222   __ cmpl(scratch, bc);
 223   __ jcc(Assembler::equal, okay);
 224   __ stop("patching the wrong bytecode");
 225   __ bind(okay);
 226 #endif
 227   // patch bytecode
 228   __ movb(at_bcp(0), bc);
 229   __ bind(patch_done);
 230 }
 231 
 232 
 233 // Individual instructions
 234 
 235 void TemplateTable::nop() {
 236   transition(vtos, vtos);
 237   // nothing to do
 238 }
 239 
 240 void TemplateTable::shouldnotreachhere() {
 241   transition(vtos, vtos);
 242   __ stop("shouldnotreachhere bytecode");
 243 }
 244 
 245 void TemplateTable::aconst_null() {
 246   transition(vtos, atos);
 247   __ xorl(rax, rax);
 248 }
 249 
 250 void TemplateTable::iconst(int value) {
 251   transition(vtos, itos);
 252   if (value == 0) {
 253     __ xorl(rax, rax);
 254   } else {
 255     __ movl(rax, value);
 256   }
 257 }
 258 
 259 void TemplateTable::lconst(int value) {
 260   transition(vtos, ltos);
 261   if (value == 0) {
 262     __ xorl(rax, rax);
 263   } else {
 264     __ movl(rax, value);
 265   }
 266 }
 267 
 268 void TemplateTable::fconst(int value) {
 269   transition(vtos, ftos);
 270   static float one = 1.0f, two = 2.0f;
 271   switch (value) {
 272   case 0:
 273     __ xorps(xmm0, xmm0);
 274     break;
 275   case 1:
 276     __ movflt(xmm0, ExternalAddress((address) &one));
 277     break;
 278   case 2:
 279     __ movflt(xmm0, ExternalAddress((address) &two));
 280     break;
 281   default:
 282     ShouldNotReachHere();
 283     break;
 284   }
 285 }
 286 
 287 void TemplateTable::dconst(int value) {
 288   transition(vtos, dtos);
 289   static double one = 1.0;
 290   switch (value) {
 291   case 0:
 292     __ xorpd(xmm0, xmm0);
 293     break;
 294   case 1:
 295     __ movdbl(xmm0, ExternalAddress((address) &one));
 296     break;
 297   default:
 298     ShouldNotReachHere();
 299     break;
 300   }
 301 }
 302 
 303 void TemplateTable::bipush() {
 304   transition(vtos, itos);
 305   __ load_signed_byte(rax, at_bcp(1));
 306 }
 307 
 308 void TemplateTable::sipush() {
 309   transition(vtos, itos);
 310   __ load_unsigned_short(rax, at_bcp(1));
 311   __ bswapl(rax);
 312   __ sarl(rax, 16);
 313 }
 314 
 315 void TemplateTable::ldc(bool wide) {
 316   transition(vtos, vtos);
 317   Label call_ldc, notFloat, notClass, Done;
 318 
 319   if (wide) {
 320     __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 321   } else {
 322     __ load_unsigned_byte(rbx, at_bcp(1));
 323   }
 324 
 325   __ get_cpool_and_tags(rcx, rax);
 326   const int base_offset = constantPoolOopDesc::header_size() * wordSize;
 327   const int tags_offset = typeArrayOopDesc::header_size(T_BYTE) * wordSize;
 328 
 329   // get type
 330   __ movzbl(rdx, Address(rax, rbx, Address::times_1, tags_offset));
 331 
 332   // unresolved string - get the resolved string
 333   __ cmpl(rdx, JVM_CONSTANT_UnresolvedString);
 334   __ jccb(Assembler::equal, call_ldc);
 335 
 336   // unresolved class - get the resolved class
 337   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClass);
 338   __ jccb(Assembler::equal, call_ldc);
 339 
 340   // unresolved class in error state - call into runtime to throw the error
 341   // from the first resolution attempt
 342   __ cmpl(rdx, JVM_CONSTANT_UnresolvedClassInError);
 343   __ jccb(Assembler::equal, call_ldc);
 344 
 345   // resolved class - need to call vm to get java mirror of the class
 346   __ cmpl(rdx, JVM_CONSTANT_Class);
 347   __ jcc(Assembler::notEqual, notClass);
 348 
 349   __ bind(call_ldc);
 350   __ movl(c_rarg1, wide);
 351   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), c_rarg1);
 352   __ push_ptr(rax);
 353   __ verify_oop(rax);
 354   __ jmp(Done);
 355 
 356   __ bind(notClass);
 357   __ cmpl(rdx, JVM_CONSTANT_Float);
 358   __ jccb(Assembler::notEqual, notFloat);
 359   // ftos
 360   __ movflt(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 361   __ push_f();
 362   __ jmp(Done);
 363 
 364   __ bind(notFloat);
 365 #ifdef ASSERT
 366   {
 367     Label L;
 368     __ cmpl(rdx, JVM_CONSTANT_Integer);
 369     __ jcc(Assembler::equal, L);
 370     __ cmpl(rdx, JVM_CONSTANT_String);
 371     __ jcc(Assembler::equal, L);
 372     __ stop("unexpected tag type in ldc");
 373     __ bind(L);
 374   }
 375 #endif
 376   // atos and itos
 377   Label isOop;
 378   __ cmpl(rdx, JVM_CONSTANT_Integer);
 379   __ jcc(Assembler::notEqual, isOop);
 380   __ movl(rax, Address(rcx, rbx, Address::times_8, base_offset));
 381   __ push_i(rax);
 382   __ jmp(Done);
 383 
 384   __ bind(isOop);
 385   __ movptr(rax, Address(rcx, rbx, Address::times_8, base_offset));
 386   __ push_ptr(rax);
 387 
 388   if (VerifyOops) {
 389     __ verify_oop(rax);
 390   }
 391 
 392   __ bind(Done);
 393 }
 394 
 395 void TemplateTable::ldc2_w() {
 396   transition(vtos, vtos);
 397   Label Long, Done;
 398   __ get_unsigned_2_byte_index_at_bcp(rbx, 1);
 399 
 400   __ get_cpool_and_tags(rcx, rax);
 401   const int base_offset = constantPoolOopDesc::header_size() * wordSize;
 402   const int tags_offset = typeArrayOopDesc::header_size(T_BYTE) * wordSize;
 403 
 404   // get type
 405   __ cmpb(Address(rax, rbx, Address::times_1, tags_offset),
 406           JVM_CONSTANT_Double);
 407   __ jccb(Assembler::notEqual, Long);
 408   // dtos
 409   __ movdbl(xmm0, Address(rcx, rbx, Address::times_8, base_offset));
 410   __ push_d();
 411   __ jmpb(Done);
 412 
 413   __ bind(Long);
 414   // ltos
 415   __ movq(rax, Address(rcx, rbx, Address::times_8, base_offset));
 416   __ push_l();
 417 
 418   __ bind(Done);
 419 }
 420 
 421 void TemplateTable::locals_index(Register reg, int offset) {
 422   __ load_unsigned_byte(reg, at_bcp(offset));
 423   __ negptr(reg);
 424   if (TaggedStackInterpreter) __ shlptr(reg, 1);  // index = index*2
 425 }
 426 
 427 void TemplateTable::iload() {
 428   transition(vtos, itos);
 429   if (RewriteFrequentPairs) {
 430     Label rewrite, done;
 431     const Register bc = c_rarg3;
 432     assert(rbx != bc, "register damaged");
 433 
 434     // get next byte
 435     __ load_unsigned_byte(rbx,
 436                           at_bcp(Bytecodes::length_for(Bytecodes::_iload)));
 437     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 438     // last two iloads in a pair.  Comparing against fast_iload means that
 439     // the next bytecode is neither an iload or a caload, and therefore
 440     // an iload pair.
 441     __ cmpl(rbx, Bytecodes::_iload);
 442     __ jcc(Assembler::equal, done);
 443 
 444     __ cmpl(rbx, Bytecodes::_fast_iload);
 445     __ movl(bc, Bytecodes::_fast_iload2);
 446     __ jccb(Assembler::equal, rewrite);
 447 
 448     // if _caload, rewrite to fast_icaload
 449     __ cmpl(rbx, Bytecodes::_caload);
 450     __ movl(bc, Bytecodes::_fast_icaload);
 451     __ jccb(Assembler::equal, rewrite);
 452 
 453     // rewrite so iload doesn't check again.
 454     __ movl(bc, Bytecodes::_fast_iload);
 455 
 456     // rewrite
 457     // bc: fast bytecode
 458     __ bind(rewrite);
 459     patch_bytecode(Bytecodes::_iload, bc, rbx, false);
 460     __ bind(done);
 461   }
 462 
 463   // Get the local value into tos
 464   locals_index(rbx);
 465   __ movl(rax, iaddress(rbx));
 466   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 467 }
 468 
 469 void TemplateTable::fast_iload2() {
 470   transition(vtos, itos);
 471   locals_index(rbx);
 472   __ movl(rax, iaddress(rbx));
 473   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 474   __ push(itos);
 475   locals_index(rbx, 3);
 476   __ movl(rax, iaddress(rbx));
 477   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 478 }
 479 
 480 void TemplateTable::fast_iload() {
 481   transition(vtos, itos);
 482   locals_index(rbx);
 483   __ movl(rax, iaddress(rbx));
 484   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 485 }
 486 
 487 void TemplateTable::lload() {
 488   transition(vtos, ltos);
 489   locals_index(rbx);
 490   __ movq(rax, laddress(rbx));
 491   debug_only(__ verify_local_tag(frame::TagCategory2, rbx));
 492 }
 493 
 494 void TemplateTable::fload() {
 495   transition(vtos, ftos);
 496   locals_index(rbx);
 497   __ movflt(xmm0, faddress(rbx));
 498   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 499 }
 500 
 501 void TemplateTable::dload() {
 502   transition(vtos, dtos);
 503   locals_index(rbx);
 504   __ movdbl(xmm0, daddress(rbx));
 505   debug_only(__ verify_local_tag(frame::TagCategory2, rbx));
 506 }
 507 
 508 void TemplateTable::aload() {
 509   transition(vtos, atos);
 510   locals_index(rbx);
 511   __ movptr(rax, aaddress(rbx));
 512   debug_only(__ verify_local_tag(frame::TagReference, rbx));
 513 }
 514 
 515 void TemplateTable::locals_index_wide(Register reg) {
 516   __ movl(reg, at_bcp(2));
 517   __ bswapl(reg);
 518   __ shrl(reg, 16);
 519   __ negptr(reg);
 520   if (TaggedStackInterpreter) __ shlptr(reg, 1);  // index = index*2
 521 }
 522 
 523 void TemplateTable::wide_iload() {
 524   transition(vtos, itos);
 525   locals_index_wide(rbx);
 526   __ movl(rax, iaddress(rbx));
 527   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 528 }
 529 
 530 void TemplateTable::wide_lload() {
 531   transition(vtos, ltos);
 532   locals_index_wide(rbx);
 533   __ movq(rax, laddress(rbx));
 534   debug_only(__ verify_local_tag(frame::TagCategory2, rbx));
 535 }
 536 
 537 void TemplateTable::wide_fload() {
 538   transition(vtos, ftos);
 539   locals_index_wide(rbx);
 540   __ movflt(xmm0, faddress(rbx));
 541   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 542 }
 543 
 544 void TemplateTable::wide_dload() {
 545   transition(vtos, dtos);
 546   locals_index_wide(rbx);
 547   __ movdbl(xmm0, daddress(rbx));
 548   debug_only(__ verify_local_tag(frame::TagCategory2, rbx));
 549 }
 550 
 551 void TemplateTable::wide_aload() {
 552   transition(vtos, atos);
 553   locals_index_wide(rbx);
 554   __ movptr(rax, aaddress(rbx));
 555   debug_only(__ verify_local_tag(frame::TagReference, rbx));
 556 }
 557 
 558 void TemplateTable::index_check(Register array, Register index) {
 559   // destroys rbx
 560   // check array
 561   __ null_check(array, arrayOopDesc::length_offset_in_bytes());
 562   // sign extend index for use by indexed load
 563   __ movl2ptr(index, index);
 564   // check index
 565   __ cmpl(index, Address(array, arrayOopDesc::length_offset_in_bytes()));
 566   if (index != rbx) {
 567     // ??? convention: move aberrant index into ebx for exception message
 568     assert(rbx != array, "different registers");
 569     __ movl(rbx, index);
 570   }
 571   __ jump_cc(Assembler::aboveEqual,
 572              ExternalAddress(Interpreter::_throw_ArrayIndexOutOfBoundsException_entry));
 573 }
 574 
 575 void TemplateTable::iaload() {
 576   transition(itos, itos);
 577   __ pop_ptr(rdx);
 578   // eax: index
 579   // rdx: array
 580   index_check(rdx, rax); // kills rbx
 581   __ movl(rax, Address(rdx, rax,
 582                        Address::times_4,
 583                        arrayOopDesc::base_offset_in_bytes(T_INT)));
 584 }
 585 
 586 void TemplateTable::laload() {
 587   transition(itos, ltos);
 588   __ pop_ptr(rdx);
 589   // eax: index
 590   // rdx: array
 591   index_check(rdx, rax); // kills rbx
 592   __ movq(rax, Address(rdx, rbx,
 593                        Address::times_8,
 594                        arrayOopDesc::base_offset_in_bytes(T_LONG)));
 595 }
 596 
 597 void TemplateTable::faload() {
 598   transition(itos, ftos);
 599   __ pop_ptr(rdx);
 600   // eax: index
 601   // rdx: array
 602   index_check(rdx, rax); // kills rbx
 603   __ movflt(xmm0, Address(rdx, rax,
 604                          Address::times_4,
 605                          arrayOopDesc::base_offset_in_bytes(T_FLOAT)));
 606 }
 607 
 608 void TemplateTable::daload() {
 609   transition(itos, dtos);
 610   __ pop_ptr(rdx);
 611   // eax: index
 612   // rdx: array
 613   index_check(rdx, rax); // kills rbx
 614   __ movdbl(xmm0, Address(rdx, rax,
 615                           Address::times_8,
 616                           arrayOopDesc::base_offset_in_bytes(T_DOUBLE)));
 617 }
 618 
 619 void TemplateTable::aaload() {
 620   transition(itos, atos);
 621   __ pop_ptr(rdx);
 622   // eax: index
 623   // rdx: array
 624   index_check(rdx, rax); // kills rbx
 625   __ load_heap_oop(rax, Address(rdx, rax,
 626                                 UseCompressedOops ? Address::times_4 : Address::times_8,
 627                                 arrayOopDesc::base_offset_in_bytes(T_OBJECT)));
 628 }
 629 
 630 void TemplateTable::baload() {
 631   transition(itos, itos);
 632   __ pop_ptr(rdx);
 633   // eax: index
 634   // rdx: array
 635   index_check(rdx, rax); // kills rbx
 636   __ load_signed_byte(rax,
 637                       Address(rdx, rax,
 638                               Address::times_1,
 639                               arrayOopDesc::base_offset_in_bytes(T_BYTE)));
 640 }
 641 
 642 void TemplateTable::caload() {
 643   transition(itos, itos);
 644   __ pop_ptr(rdx);
 645   // eax: index
 646   // rdx: array
 647   index_check(rdx, rax); // kills rbx
 648   __ load_unsigned_short(rax,
 649                          Address(rdx, rax,
 650                                  Address::times_2,
 651                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 652 }
 653 
 654 // iload followed by caload frequent pair
 655 void TemplateTable::fast_icaload() {
 656   transition(vtos, itos);
 657   // load index out of locals
 658   locals_index(rbx);
 659   __ movl(rax, iaddress(rbx));
 660   debug_only(__ verify_local_tag(frame::TagValue, rbx));
 661 
 662   // eax: index
 663   // rdx: array
 664   __ pop_ptr(rdx);
 665   index_check(rdx, rax); // kills rbx
 666   __ load_unsigned_short(rax,
 667                          Address(rdx, rax,
 668                                  Address::times_2,
 669                                  arrayOopDesc::base_offset_in_bytes(T_CHAR)));
 670 }
 671 
 672 void TemplateTable::saload() {
 673   transition(itos, itos);
 674   __ pop_ptr(rdx);
 675   // eax: index
 676   // rdx: array
 677   index_check(rdx, rax); // kills rbx
 678   __ load_signed_short(rax,
 679                        Address(rdx, rax,
 680                                Address::times_2,
 681                                arrayOopDesc::base_offset_in_bytes(T_SHORT)));
 682 }
 683 
 684 void TemplateTable::iload(int n) {
 685   transition(vtos, itos);
 686   __ movl(rax, iaddress(n));
 687   debug_only(__ verify_local_tag(frame::TagValue, n));
 688 }
 689 
 690 void TemplateTable::lload(int n) {
 691   transition(vtos, ltos);
 692   __ movq(rax, laddress(n));
 693   debug_only(__ verify_local_tag(frame::TagCategory2, n));
 694 }
 695 
 696 void TemplateTable::fload(int n) {
 697   transition(vtos, ftos);
 698   __ movflt(xmm0, faddress(n));
 699   debug_only(__ verify_local_tag(frame::TagValue, n));
 700 }
 701 
 702 void TemplateTable::dload(int n) {
 703   transition(vtos, dtos);
 704   __ movdbl(xmm0, daddress(n));
 705   debug_only(__ verify_local_tag(frame::TagCategory2, n));
 706 }
 707 
 708 void TemplateTable::aload(int n) {
 709   transition(vtos, atos);
 710   __ movptr(rax, aaddress(n));
 711   debug_only(__ verify_local_tag(frame::TagReference, n));
 712 }
 713 
 714 void TemplateTable::aload_0() {
 715   transition(vtos, atos);
 716   // According to bytecode histograms, the pairs:
 717   //
 718   // _aload_0, _fast_igetfield
 719   // _aload_0, _fast_agetfield
 720   // _aload_0, _fast_fgetfield
 721   //
 722   // occur frequently. If RewriteFrequentPairs is set, the (slow)
 723   // _aload_0 bytecode checks if the next bytecode is either
 724   // _fast_igetfield, _fast_agetfield or _fast_fgetfield and then
 725   // rewrites the current bytecode into a pair bytecode; otherwise it
 726   // rewrites the current bytecode into _fast_aload_0 that doesn't do
 727   // the pair check anymore.
 728   //
 729   // Note: If the next bytecode is _getfield, the rewrite must be
 730   //       delayed, otherwise we may miss an opportunity for a pair.
 731   //
 732   // Also rewrite frequent pairs
 733   //   aload_0, aload_1
 734   //   aload_0, iload_1
 735   // These bytecodes with a small amount of code are most profitable
 736   // to rewrite
 737   if (RewriteFrequentPairs) {
 738     Label rewrite, done;
 739     const Register bc = c_rarg3;
 740     assert(rbx != bc, "register damaged");
 741     // get next byte
 742     __ load_unsigned_byte(rbx,
 743                           at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)));
 744 
 745     // do actual aload_0
 746     aload(0);
 747 
 748     // if _getfield then wait with rewrite
 749     __ cmpl(rbx, Bytecodes::_getfield);
 750     __ jcc(Assembler::equal, done);
 751 
 752     // if _igetfield then reqrite to _fast_iaccess_0
 753     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) ==
 754            Bytecodes::_aload_0,
 755            "fix bytecode definition");
 756     __ cmpl(rbx, Bytecodes::_fast_igetfield);
 757     __ movl(bc, Bytecodes::_fast_iaccess_0);
 758     __ jccb(Assembler::equal, rewrite);
 759 
 760     // if _agetfield then reqrite to _fast_aaccess_0
 761     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) ==
 762            Bytecodes::_aload_0,
 763            "fix bytecode definition");
 764     __ cmpl(rbx, Bytecodes::_fast_agetfield);
 765     __ movl(bc, Bytecodes::_fast_aaccess_0);
 766     __ jccb(Assembler::equal, rewrite);
 767 
 768     // if _fgetfield then reqrite to _fast_faccess_0
 769     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) ==
 770            Bytecodes::_aload_0,
 771            "fix bytecode definition");
 772     __ cmpl(rbx, Bytecodes::_fast_fgetfield);
 773     __ movl(bc, Bytecodes::_fast_faccess_0);
 774     __ jccb(Assembler::equal, rewrite);
 775 
 776     // else rewrite to _fast_aload0
 777     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) ==
 778            Bytecodes::_aload_0,
 779            "fix bytecode definition");
 780     __ movl(bc, Bytecodes::_fast_aload_0);
 781 
 782     // rewrite
 783     // bc: fast bytecode
 784     __ bind(rewrite);
 785     patch_bytecode(Bytecodes::_aload_0, bc, rbx, false);
 786 
 787     __ bind(done);
 788   } else {
 789     aload(0);
 790   }
 791 }
 792 
 793 void TemplateTable::istore() {
 794   transition(itos, vtos);
 795   locals_index(rbx);
 796   __ movl(iaddress(rbx), rax);
 797   __ tag_local(frame::TagValue, rbx);
 798 }
 799 
 800 void TemplateTable::lstore() {
 801   transition(ltos, vtos);
 802   locals_index(rbx);
 803   __ movq(laddress(rbx), rax);
 804   __ tag_local(frame::TagCategory2, rbx);
 805 }
 806 
 807 void TemplateTable::fstore() {
 808   transition(ftos, vtos);
 809   locals_index(rbx);
 810   __ movflt(faddress(rbx), xmm0);
 811   __ tag_local(frame::TagValue, rbx);
 812 }
 813 
 814 void TemplateTable::dstore() {
 815   transition(dtos, vtos);
 816   locals_index(rbx);
 817   __ movdbl(daddress(rbx), xmm0);
 818   __ tag_local(frame::TagCategory2, rbx);
 819 }
 820 
 821 void TemplateTable::astore() {
 822   transition(vtos, vtos);
 823   __ pop_ptr(rax, rdx);    // will need to pop tag too
 824   locals_index(rbx);
 825   __ movptr(aaddress(rbx), rax);
 826   __ tag_local(rdx, rbx);  // store tag from stack, might be returnAddr
 827 }
 828 
 829 void TemplateTable::wide_istore() {
 830   transition(vtos, vtos);
 831   __ pop_i();
 832   locals_index_wide(rbx);
 833   __ movl(iaddress(rbx), rax);
 834   __ tag_local(frame::TagValue, rbx);
 835 }
 836 
 837 void TemplateTable::wide_lstore() {
 838   transition(vtos, vtos);
 839   __ pop_l();
 840   locals_index_wide(rbx);
 841   __ movq(laddress(rbx), rax);
 842   __ tag_local(frame::TagCategory2, rbx);
 843 }
 844 
 845 void TemplateTable::wide_fstore() {
 846   transition(vtos, vtos);
 847   __ pop_f();
 848   locals_index_wide(rbx);
 849   __ movflt(faddress(rbx), xmm0);
 850   __ tag_local(frame::TagValue, rbx);
 851 }
 852 
 853 void TemplateTable::wide_dstore() {
 854   transition(vtos, vtos);
 855   __ pop_d();
 856   locals_index_wide(rbx);
 857   __ movdbl(daddress(rbx), xmm0);
 858   __ tag_local(frame::TagCategory2, rbx);
 859 }
 860 
 861 void TemplateTable::wide_astore() {
 862   transition(vtos, vtos);
 863   __ pop_ptr(rax, rdx);    // will need to pop tag too
 864   locals_index_wide(rbx);
 865   __ movptr(aaddress(rbx), rax);
 866   __ tag_local(rdx, rbx);  // store tag from stack, might be returnAddr
 867 }
 868 
 869 void TemplateTable::iastore() {
 870   transition(itos, vtos);
 871   __ pop_i(rbx);
 872   __ pop_ptr(rdx);
 873   // eax: value
 874   // ebx: index
 875   // rdx: array
 876   index_check(rdx, rbx); // prefer index in ebx
 877   __ movl(Address(rdx, rbx,
 878                   Address::times_4,
 879                   arrayOopDesc::base_offset_in_bytes(T_INT)),
 880           rax);
 881 }
 882 
 883 void TemplateTable::lastore() {
 884   transition(ltos, vtos);
 885   __ pop_i(rbx);
 886   __ pop_ptr(rdx);
 887   // rax: value
 888   // ebx: index
 889   // rdx: array
 890   index_check(rdx, rbx); // prefer index in ebx
 891   __ movq(Address(rdx, rbx,
 892                   Address::times_8,
 893                   arrayOopDesc::base_offset_in_bytes(T_LONG)),
 894           rax);
 895 }
 896 
 897 void TemplateTable::fastore() {
 898   transition(ftos, vtos);
 899   __ pop_i(rbx);
 900   __ pop_ptr(rdx);
 901   // xmm0: value
 902   // ebx:  index
 903   // rdx:  array
 904   index_check(rdx, rbx); // prefer index in ebx
 905   __ movflt(Address(rdx, rbx,
 906                    Address::times_4,
 907                    arrayOopDesc::base_offset_in_bytes(T_FLOAT)),
 908            xmm0);
 909 }
 910 
 911 void TemplateTable::dastore() {
 912   transition(dtos, vtos);
 913   __ pop_i(rbx);
 914   __ pop_ptr(rdx);
 915   // xmm0: value
 916   // ebx:  index
 917   // rdx:  array
 918   index_check(rdx, rbx); // prefer index in ebx
 919   __ movdbl(Address(rdx, rbx,
 920                    Address::times_8,
 921                    arrayOopDesc::base_offset_in_bytes(T_DOUBLE)),
 922            xmm0);
 923 }
 924 
 925 void TemplateTable::aastore() {
 926   Label is_null, ok_is_subtype, done;
 927   transition(vtos, vtos);
 928   // stack: ..., array, index, value
 929   __ movptr(rax, at_tos());    // value
 930   __ movl(rcx, at_tos_p1()); // index
 931   __ movptr(rdx, at_tos_p2()); // array
 932 
 933   Address element_address(rdx, rcx,
 934                           UseCompressedOops? Address::times_4 : Address::times_8,
 935                           arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 936 
 937   index_check(rdx, rcx);     // kills rbx
 938   // do array store check - check for NULL value first
 939   __ testptr(rax, rax);
 940   __ jcc(Assembler::zero, is_null);
 941 
 942   // Move subklass into rbx
 943   __ load_klass(rbx, rax);
 944   // Move superklass into rax
 945   __ load_klass(rax, rdx);
 946   __ movptr(rax, Address(rax,
 947                          sizeof(oopDesc) +
 948                          objArrayKlass::element_klass_offset_in_bytes()));
 949   // Compress array + index*oopSize + 12 into a single register.  Frees rcx.
 950   __ lea(rdx, element_address);
 951 
 952   // Generate subtype check.  Blows rcx, rdi
 953   // Superklass in rax.  Subklass in rbx.
 954   __ gen_subtype_check(rbx, ok_is_subtype);
 955 
 956   // Come here on failure
 957   // object is at TOS
 958   __ jump(ExternalAddress(Interpreter::_throw_ArrayStoreException_entry));
 959 
 960   // Come here on success
 961   __ bind(ok_is_subtype);
 962 
 963   // Get the value we will store
 964   __ movptr(rax, at_tos());
 965   // Now store using the appropriate barrier
 966   do_oop_store(_masm, Address(rdx, 0), rax, _bs->kind(), true);
 967   __ jmp(done);
 968 
 969   // Have a NULL in rax, rdx=array, ecx=index.  Store NULL at ary[idx]
 970   __ bind(is_null);
 971   __ profile_null_seen(rbx);
 972 
 973   // Store a NULL
 974   do_oop_store(_masm, element_address, noreg, _bs->kind(), true);
 975 
 976   // Pop stack arguments
 977   __ bind(done);
 978   __ addptr(rsp, 3 * Interpreter::stackElementSize());
 979 }
 980 
 981 void TemplateTable::bastore() {
 982   transition(itos, vtos);
 983   __ pop_i(rbx);
 984   __ pop_ptr(rdx);
 985   // eax: value
 986   // ebx: index
 987   // rdx: array
 988   index_check(rdx, rbx); // prefer index in ebx
 989   __ movb(Address(rdx, rbx,
 990                   Address::times_1,
 991                   arrayOopDesc::base_offset_in_bytes(T_BYTE)),
 992           rax);
 993 }
 994 
 995 void TemplateTable::castore() {
 996   transition(itos, vtos);
 997   __ pop_i(rbx);
 998   __ pop_ptr(rdx);
 999   // eax: value
1000   // ebx: index
1001   // rdx: array
1002   index_check(rdx, rbx);  // prefer index in ebx
1003   __ movw(Address(rdx, rbx,
1004                   Address::times_2,
1005                   arrayOopDesc::base_offset_in_bytes(T_CHAR)),
1006           rax);
1007 }
1008 
1009 void TemplateTable::sastore() {
1010   castore();
1011 }
1012 
1013 void TemplateTable::istore(int n) {
1014   transition(itos, vtos);
1015   __ movl(iaddress(n), rax);
1016   __ tag_local(frame::TagValue, n);
1017 }
1018 
1019 void TemplateTable::lstore(int n) {
1020   transition(ltos, vtos);
1021   __ movq(laddress(n), rax);
1022   __ tag_local(frame::TagCategory2, n);
1023 }
1024 
1025 void TemplateTable::fstore(int n) {
1026   transition(ftos, vtos);
1027   __ movflt(faddress(n), xmm0);
1028   __ tag_local(frame::TagValue, n);
1029 }
1030 
1031 void TemplateTable::dstore(int n) {
1032   transition(dtos, vtos);
1033   __ movdbl(daddress(n), xmm0);
1034   __ tag_local(frame::TagCategory2, n);
1035 }
1036 
1037 void TemplateTable::astore(int n) {
1038   transition(vtos, vtos);
1039   __ pop_ptr(rax, rdx);
1040   __ movptr(aaddress(n), rax);
1041   __ tag_local(rdx, n);
1042 }
1043 
1044 void TemplateTable::pop() {
1045   transition(vtos, vtos);
1046   __ addptr(rsp, Interpreter::stackElementSize());
1047 }
1048 
1049 void TemplateTable::pop2() {
1050   transition(vtos, vtos);
1051   __ addptr(rsp, 2 * Interpreter::stackElementSize());
1052 }
1053 
1054 void TemplateTable::dup() {
1055   transition(vtos, vtos);
1056   __ load_ptr_and_tag(0, rax, rdx);
1057   __ push_ptr(rax, rdx);
1058   // stack: ..., a, a
1059 }
1060 
1061 void TemplateTable::dup_x1() {
1062   transition(vtos, vtos);
1063   // stack: ..., a, b
1064   __ load_ptr_and_tag(0, rax, rdx);  // load b
1065   __ load_ptr_and_tag(1, rcx, rbx);  // load a
1066   __ store_ptr_and_tag(1, rax, rdx); // store b
1067   __ store_ptr_and_tag(0, rcx, rbx); // store a
1068   __ push_ptr(rax, rdx);             // push b
1069   // stack: ..., b, a, b
1070 }
1071 
1072 void TemplateTable::dup_x2() {
1073   transition(vtos, vtos);
1074   // stack: ..., a, b, c
1075   __ load_ptr_and_tag(0, rax, rdx);  // load c
1076   __ load_ptr_and_tag(2, rcx, rbx);  // load a
1077   __ store_ptr_and_tag(2, rax, rdx); // store c in a
1078   __ push_ptr(rax, rdx);             // push c
1079   // stack: ..., c, b, c, c
1080   __ load_ptr_and_tag(2, rax, rdx);  // load b
1081   __ store_ptr_and_tag(2, rcx, rbx); // store a in b
1082   // stack: ..., c, a, c, c
1083   __ store_ptr_and_tag(1, rax, rdx); // store b in c
1084   // stack: ..., c, a, b, c
1085 }
1086 
1087 void TemplateTable::dup2() {
1088   transition(vtos, vtos);
1089   // stack: ..., a, b
1090   __ load_ptr_and_tag(1, rax, rdx);  // load a
1091   __ push_ptr(rax, rdx);             // push a
1092   __ load_ptr_and_tag(1, rax, rdx);  // load b
1093   __ push_ptr(rax, rdx);             // push b
1094   // stack: ..., a, b, a, b
1095 }
1096 
1097 void TemplateTable::dup2_x1() {
1098   transition(vtos, vtos);
1099   // stack: ..., a, b, c
1100   __ load_ptr_and_tag(0, rcx, rbx);  // load c
1101   __ load_ptr_and_tag(1, rax, rdx);  // load b
1102   __ push_ptr(rax, rdx);             // push b
1103   __ push_ptr(rcx, rbx);             // push c
1104   // stack: ..., a, b, c, b, c
1105   __ store_ptr_and_tag(3, rcx, rbx); // store c in b
1106   // stack: ..., a, c, c, b, c
1107   __ load_ptr_and_tag(4, rcx, rbx);  // load a
1108   __ store_ptr_and_tag(2, rcx, rbx); // store a in 2nd c
1109   // stack: ..., a, c, a, b, c
1110   __ store_ptr_and_tag(4, rax, rdx); // store b in a
1111   // stack: ..., b, c, a, b, c
1112 }
1113 
1114 void TemplateTable::dup2_x2() {
1115   transition(vtos, vtos);
1116   // stack: ..., a, b, c, d
1117   __ load_ptr_and_tag(0, rcx, rbx);  // load d
1118   __ load_ptr_and_tag(1, rax, rdx);  // load c
1119   __ push_ptr(rax, rdx);             // push c
1120   __ push_ptr(rcx, rbx);             // push d
1121   // stack: ..., a, b, c, d, c, d
1122   __ load_ptr_and_tag(4, rax, rdx);  // load b
1123   __ store_ptr_and_tag(2, rax, rdx); // store b in d
1124   __ store_ptr_and_tag(4, rcx, rbx); // store d in b
1125   // stack: ..., a, d, c, b, c, d
1126   __ load_ptr_and_tag(5, rcx, rbx);  // load a
1127   __ load_ptr_and_tag(3, rax, rdx);  // load c
1128   __ store_ptr_and_tag(3, rcx, rbx); // store a in c
1129   __ store_ptr_and_tag(5, rax, rdx); // store c in a
1130   // stack: ..., c, d, a, b, c, d
1131 }
1132 
1133 void TemplateTable::swap() {
1134   transition(vtos, vtos);
1135   // stack: ..., a, b
1136   __ load_ptr_and_tag(1, rcx, rbx);  // load a
1137   __ load_ptr_and_tag(0, rax, rdx);  // load b
1138   __ store_ptr_and_tag(0, rcx, rbx); // store a in b
1139   __ store_ptr_and_tag(1, rax, rdx); // store b in a
1140   // stack: ..., b, a
1141 }
1142 
1143 void TemplateTable::iop2(Operation op) {
1144   transition(itos, itos);
1145   switch (op) {
1146   case add  :                    __ pop_i(rdx); __ addl (rax, rdx); break;
1147   case sub  : __ movl(rdx, rax); __ pop_i(rax); __ subl (rax, rdx); break;
1148   case mul  :                    __ pop_i(rdx); __ imull(rax, rdx); break;
1149   case _and :                    __ pop_i(rdx); __ andl (rax, rdx); break;
1150   case _or  :                    __ pop_i(rdx); __ orl  (rax, rdx); break;
1151   case _xor :                    __ pop_i(rdx); __ xorl (rax, rdx); break;
1152   case shl  : __ movl(rcx, rax); __ pop_i(rax); __ shll (rax);      break;
1153   case shr  : __ movl(rcx, rax); __ pop_i(rax); __ sarl (rax);      break;
1154   case ushr : __ movl(rcx, rax); __ pop_i(rax); __ shrl (rax);      break;
1155   default   : ShouldNotReachHere();
1156   }
1157 }
1158 
1159 void TemplateTable::lop2(Operation op) {
1160   transition(ltos, ltos);
1161   switch (op) {
1162   case add  :                    __ pop_l(rdx); __ addptr (rax, rdx); break;
1163   case sub  : __ mov(rdx, rax);  __ pop_l(rax); __ subptr (rax, rdx); break;
1164   case _and :                    __ pop_l(rdx); __ andptr (rax, rdx); break;
1165   case _or  :                    __ pop_l(rdx); __ orptr  (rax, rdx); break;
1166   case _xor :                    __ pop_l(rdx); __ xorptr (rax, rdx); break;
1167   default : ShouldNotReachHere();
1168   }
1169 }
1170 
1171 void TemplateTable::idiv() {
1172   transition(itos, itos);
1173   __ movl(rcx, rax);
1174   __ pop_i(rax);
1175   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1176   //       they are not equal, one could do a normal division (no correction
1177   //       needed), which may speed up this implementation for the common case.
1178   //       (see also JVM spec., p.243 & p.271)
1179   __ corrected_idivl(rcx);
1180 }
1181 
1182 void TemplateTable::irem() {
1183   transition(itos, itos);
1184   __ movl(rcx, rax);
1185   __ pop_i(rax);
1186   // Note: could xor eax and ecx and compare with (-1 ^ min_int). If
1187   //       they are not equal, one could do a normal division (no correction
1188   //       needed), which may speed up this implementation for the common case.
1189   //       (see also JVM spec., p.243 & p.271)
1190   __ corrected_idivl(rcx);
1191   __ movl(rax, rdx);
1192 }
1193 
1194 void TemplateTable::lmul() {
1195   transition(ltos, ltos);
1196   __ pop_l(rdx);
1197   __ imulq(rax, rdx);
1198 }
1199 
1200 void TemplateTable::ldiv() {
1201   transition(ltos, ltos);
1202   __ mov(rcx, rax);
1203   __ pop_l(rax);
1204   // generate explicit div0 check
1205   __ testq(rcx, rcx);
1206   __ jump_cc(Assembler::zero,
1207              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1208   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1209   //       they are not equal, one could do a normal division (no correction
1210   //       needed), which may speed up this implementation for the common case.
1211   //       (see also JVM spec., p.243 & p.271)
1212   __ corrected_idivq(rcx); // kills rbx
1213 }
1214 
1215 void TemplateTable::lrem() {
1216   transition(ltos, ltos);
1217   __ mov(rcx, rax);
1218   __ pop_l(rax);
1219   __ testq(rcx, rcx);
1220   __ jump_cc(Assembler::zero,
1221              ExternalAddress(Interpreter::_throw_ArithmeticException_entry));
1222   // Note: could xor rax and rcx and compare with (-1 ^ min_int). If
1223   //       they are not equal, one could do a normal division (no correction
1224   //       needed), which may speed up this implementation for the common case.
1225   //       (see also JVM spec., p.243 & p.271)
1226   __ corrected_idivq(rcx); // kills rbx
1227   __ mov(rax, rdx);
1228 }
1229 
1230 void TemplateTable::lshl() {
1231   transition(itos, ltos);
1232   __ movl(rcx, rax);                             // get shift count
1233   __ pop_l(rax);                                 // get shift value
1234   __ shlq(rax);
1235 }
1236 
1237 void TemplateTable::lshr() {
1238   transition(itos, ltos);
1239   __ movl(rcx, rax);                             // get shift count
1240   __ pop_l(rax);                                 // get shift value
1241   __ sarq(rax);
1242 }
1243 
1244 void TemplateTable::lushr() {
1245   transition(itos, ltos);
1246   __ movl(rcx, rax);                             // get shift count
1247   __ pop_l(rax);                                 // get shift value
1248   __ shrq(rax);
1249 }
1250 
1251 void TemplateTable::fop2(Operation op) {
1252   transition(ftos, ftos);
1253   switch (op) {
1254   case add:
1255     __ addss(xmm0, at_rsp());
1256     __ addptr(rsp, Interpreter::stackElementSize());
1257     break;
1258   case sub:
1259     __ movflt(xmm1, xmm0);
1260     __ pop_f(xmm0);
1261     __ subss(xmm0, xmm1);
1262     break;
1263   case mul:
1264     __ mulss(xmm0, at_rsp());
1265     __ addptr(rsp, Interpreter::stackElementSize());
1266     break;
1267   case div:
1268     __ movflt(xmm1, xmm0);
1269     __ pop_f(xmm0);
1270     __ divss(xmm0, xmm1);
1271     break;
1272   case rem:
1273     __ movflt(xmm1, xmm0);
1274     __ pop_f(xmm0);
1275     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::frem), 2);
1276     break;
1277   default:
1278     ShouldNotReachHere();
1279     break;
1280   }
1281 }
1282 
1283 void TemplateTable::dop2(Operation op) {
1284   transition(dtos, dtos);
1285   switch (op) {
1286   case add:
1287     __ addsd(xmm0, at_rsp());
1288     __ addptr(rsp, 2 * Interpreter::stackElementSize());
1289     break;
1290   case sub:
1291     __ movdbl(xmm1, xmm0);
1292     __ pop_d(xmm0);
1293     __ subsd(xmm0, xmm1);
1294     break;
1295   case mul:
1296     __ mulsd(xmm0, at_rsp());
1297     __ addptr(rsp, 2 * Interpreter::stackElementSize());
1298     break;
1299   case div:
1300     __ movdbl(xmm1, xmm0);
1301     __ pop_d(xmm0);
1302     __ divsd(xmm0, xmm1);
1303     break;
1304   case rem:
1305     __ movdbl(xmm1, xmm0);
1306     __ pop_d(xmm0);
1307     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::drem), 2);
1308     break;
1309   default:
1310     ShouldNotReachHere();
1311     break;
1312   }
1313 }
1314 
1315 void TemplateTable::ineg() {
1316   transition(itos, itos);
1317   __ negl(rax);
1318 }
1319 
1320 void TemplateTable::lneg() {
1321   transition(ltos, ltos);
1322   __ negq(rax);
1323 }
1324 
1325 // Note: 'double' and 'long long' have 32-bits alignment on x86.
1326 static jlong* double_quadword(jlong *adr, jlong lo, jlong hi) {
1327   // Use the expression (adr)&(~0xF) to provide 128-bits aligned address
1328   // of 128-bits operands for SSE instructions.
1329   jlong *operand = (jlong*)(((intptr_t)adr)&((intptr_t)(~0xF)));
1330   // Store the value to a 128-bits operand.
1331   operand[0] = lo;
1332   operand[1] = hi;
1333   return operand;
1334 }
1335 
1336 // Buffer for 128-bits masks used by SSE instructions.
1337 static jlong float_signflip_pool[2*2];
1338 static jlong double_signflip_pool[2*2];
1339 
1340 void TemplateTable::fneg() {
1341   transition(ftos, ftos);
1342   static jlong *float_signflip  = double_quadword(&float_signflip_pool[1], 0x8000000080000000, 0x8000000080000000);
1343   __ xorps(xmm0, ExternalAddress((address) float_signflip));
1344 }
1345 
1346 void TemplateTable::dneg() {
1347   transition(dtos, dtos);
1348   static jlong *double_signflip  = double_quadword(&double_signflip_pool[1], 0x8000000000000000, 0x8000000000000000);
1349   __ xorpd(xmm0, ExternalAddress((address) double_signflip));
1350 }
1351 
1352 void TemplateTable::iinc() {
1353   transition(vtos, vtos);
1354   __ load_signed_byte(rdx, at_bcp(2)); // get constant
1355   locals_index(rbx);
1356   __ addl(iaddress(rbx), rdx);
1357 }
1358 
1359 void TemplateTable::wide_iinc() {
1360   transition(vtos, vtos);
1361   __ movl(rdx, at_bcp(4)); // get constant
1362   locals_index_wide(rbx);
1363   __ bswapl(rdx); // swap bytes & sign-extend constant
1364   __ sarl(rdx, 16);
1365   __ addl(iaddress(rbx), rdx);
1366   // Note: should probably use only one movl to get both
1367   //       the index and the constant -> fix this
1368 }
1369 
1370 void TemplateTable::convert() {
1371   // Checking
1372 #ifdef ASSERT
1373   {
1374     TosState tos_in  = ilgl;
1375     TosState tos_out = ilgl;
1376     switch (bytecode()) {
1377     case Bytecodes::_i2l: // fall through
1378     case Bytecodes::_i2f: // fall through
1379     case Bytecodes::_i2d: // fall through
1380     case Bytecodes::_i2b: // fall through
1381     case Bytecodes::_i2c: // fall through
1382     case Bytecodes::_i2s: tos_in = itos; break;
1383     case Bytecodes::_l2i: // fall through
1384     case Bytecodes::_l2f: // fall through
1385     case Bytecodes::_l2d: tos_in = ltos; break;
1386     case Bytecodes::_f2i: // fall through
1387     case Bytecodes::_f2l: // fall through
1388     case Bytecodes::_f2d: tos_in = ftos; break;
1389     case Bytecodes::_d2i: // fall through
1390     case Bytecodes::_d2l: // fall through
1391     case Bytecodes::_d2f: tos_in = dtos; break;
1392     default             : ShouldNotReachHere();
1393     }
1394     switch (bytecode()) {
1395     case Bytecodes::_l2i: // fall through
1396     case Bytecodes::_f2i: // fall through
1397     case Bytecodes::_d2i: // fall through
1398     case Bytecodes::_i2b: // fall through
1399     case Bytecodes::_i2c: // fall through
1400     case Bytecodes::_i2s: tos_out = itos; break;
1401     case Bytecodes::_i2l: // fall through
1402     case Bytecodes::_f2l: // fall through
1403     case Bytecodes::_d2l: tos_out = ltos; break;
1404     case Bytecodes::_i2f: // fall through
1405     case Bytecodes::_l2f: // fall through
1406     case Bytecodes::_d2f: tos_out = ftos; break;
1407     case Bytecodes::_i2d: // fall through
1408     case Bytecodes::_l2d: // fall through
1409     case Bytecodes::_f2d: tos_out = dtos; break;
1410     default             : ShouldNotReachHere();
1411     }
1412     transition(tos_in, tos_out);
1413   }
1414 #endif // ASSERT
1415 
1416   static const int64_t is_nan = 0x8000000000000000L;
1417 
1418   // Conversion
1419   switch (bytecode()) {
1420   case Bytecodes::_i2l:
1421     __ movslq(rax, rax);
1422     break;
1423   case Bytecodes::_i2f:
1424     __ cvtsi2ssl(xmm0, rax);
1425     break;
1426   case Bytecodes::_i2d:
1427     __ cvtsi2sdl(xmm0, rax);
1428     break;
1429   case Bytecodes::_i2b:
1430     __ movsbl(rax, rax);
1431     break;
1432   case Bytecodes::_i2c:
1433     __ movzwl(rax, rax);
1434     break;
1435   case Bytecodes::_i2s:
1436     __ movswl(rax, rax);
1437     break;
1438   case Bytecodes::_l2i:
1439     __ movl(rax, rax);
1440     break;
1441   case Bytecodes::_l2f:
1442     __ cvtsi2ssq(xmm0, rax);
1443     break;
1444   case Bytecodes::_l2d:
1445     __ cvtsi2sdq(xmm0, rax);
1446     break;
1447   case Bytecodes::_f2i:
1448   {
1449     Label L;
1450     __ cvttss2sil(rax, xmm0);
1451     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1452     __ jcc(Assembler::notEqual, L);
1453     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2i), 1);
1454     __ bind(L);
1455   }
1456     break;
1457   case Bytecodes::_f2l:
1458   {
1459     Label L;
1460     __ cvttss2siq(rax, xmm0);
1461     // NaN or overflow/underflow?
1462     __ cmp64(rax, ExternalAddress((address) &is_nan));
1463     __ jcc(Assembler::notEqual, L);
1464     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::f2l), 1);
1465     __ bind(L);
1466   }
1467     break;
1468   case Bytecodes::_f2d:
1469     __ cvtss2sd(xmm0, xmm0);
1470     break;
1471   case Bytecodes::_d2i:
1472   {
1473     Label L;
1474     __ cvttsd2sil(rax, xmm0);
1475     __ cmpl(rax, 0x80000000); // NaN or overflow/underflow?
1476     __ jcc(Assembler::notEqual, L);
1477     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2i), 1);
1478     __ bind(L);
1479   }
1480     break;
1481   case Bytecodes::_d2l:
1482   {
1483     Label L;
1484     __ cvttsd2siq(rax, xmm0);
1485     // NaN or overflow/underflow?
1486     __ cmp64(rax, ExternalAddress((address) &is_nan));
1487     __ jcc(Assembler::notEqual, L);
1488     __ call_VM_leaf(CAST_FROM_FN_PTR(address, SharedRuntime::d2l), 1);
1489     __ bind(L);
1490   }
1491     break;
1492   case Bytecodes::_d2f:
1493     __ cvtsd2ss(xmm0, xmm0);
1494     break;
1495   default:
1496     ShouldNotReachHere();
1497   }
1498 }
1499 
1500 void TemplateTable::lcmp() {
1501   transition(ltos, itos);
1502   Label done;
1503   __ pop_l(rdx);
1504   __ cmpq(rdx, rax);
1505   __ movl(rax, -1);
1506   __ jccb(Assembler::less, done);
1507   __ setb(Assembler::notEqual, rax);
1508   __ movzbl(rax, rax);
1509   __ bind(done);
1510 }
1511 
1512 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1513   Label done;
1514   if (is_float) {
1515     // XXX get rid of pop here, use ... reg, mem32
1516     __ pop_f(xmm1);
1517     __ ucomiss(xmm1, xmm0);
1518   } else {
1519     // XXX get rid of pop here, use ... reg, mem64
1520     __ pop_d(xmm1);
1521     __ ucomisd(xmm1, xmm0);
1522   }
1523   if (unordered_result < 0) {
1524     __ movl(rax, -1);
1525     __ jccb(Assembler::parity, done);
1526     __ jccb(Assembler::below, done);
1527     __ setb(Assembler::notEqual, rdx);
1528     __ movzbl(rax, rdx);
1529   } else {
1530     __ movl(rax, 1);
1531     __ jccb(Assembler::parity, done);
1532     __ jccb(Assembler::above, done);
1533     __ movl(rax, 0);
1534     __ jccb(Assembler::equal, done);
1535     __ decrementl(rax);
1536   }
1537   __ bind(done);
1538 }
1539 
1540 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1541   __ get_method(rcx); // rcx holds method
1542   __ profile_taken_branch(rax, rbx); // rax holds updated MDP, rbx
1543                                      // holds bumped taken count
1544 
1545   const ByteSize be_offset = methodOopDesc::backedge_counter_offset() +
1546                              InvocationCounter::counter_offset();
1547   const ByteSize inv_offset = methodOopDesc::invocation_counter_offset() +
1548                               InvocationCounter::counter_offset();
1549   const int method_offset = frame::interpreter_frame_method_offset * wordSize;
1550 
1551   // Load up edx with the branch displacement
1552   __ movl(rdx, at_bcp(1));
1553   __ bswapl(rdx);
1554 
1555   if (!is_wide) {
1556     __ sarl(rdx, 16);
1557   }
1558   __ movl2ptr(rdx, rdx);
1559 
1560   // Handle all the JSR stuff here, then exit.
1561   // It's much shorter and cleaner than intermingling with the non-JSR
1562   // normal-branch stuff occurring below.
1563   if (is_jsr) {
1564     // Pre-load the next target bytecode into rbx
1565     __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1, 0));
1566 
1567     // compute return address as bci in rax
1568     __ lea(rax, at_bcp((is_wide ? 5 : 3) -
1569                         in_bytes(constMethodOopDesc::codes_offset())));
1570     __ subptr(rax, Address(rcx, methodOopDesc::const_offset()));
1571     // Adjust the bcp in r13 by the displacement in rdx
1572     __ addptr(r13, rdx);
1573     // jsr returns atos that is not an oop
1574     __ push_i(rax);
1575     __ dispatch_only(vtos);
1576     return;
1577   }
1578 
1579   // Normal (non-jsr) branch handling
1580 
1581   // Adjust the bcp in r13 by the displacement in rdx
1582   __ addptr(r13, rdx);
1583 
1584   assert(UseLoopCounter || !UseOnStackReplacement,
1585          "on-stack-replacement requires loop counters");
1586   Label backedge_counter_overflow;
1587   Label profile_method;
1588   Label dispatch;
1589   if (UseLoopCounter) {
1590     // increment backedge counter for backward branches
1591     // rax: MDO
1592     // ebx: MDO bumped taken-count
1593     // rcx: method
1594     // rdx: target offset
1595     // r13: target bcp
1596     // r14: locals pointer
1597     __ testl(rdx, rdx);             // check if forward or backward branch
1598     __ jcc(Assembler::positive, dispatch); // count only if backward branch
1599 
1600     // increment counter
1601     __ movl(rax, Address(rcx, be_offset));        // load backedge counter
1602     __ incrementl(rax, InvocationCounter::count_increment); // increment
1603                                                             // counter
1604     __ movl(Address(rcx, be_offset), rax);        // store counter
1605 
1606     __ movl(rax, Address(rcx, inv_offset));    // load invocation counter
1607     __ andl(rax, InvocationCounter::count_mask_value); // and the status bits
1608     __ addl(rax, Address(rcx, be_offset));        // add both counters
1609 
1610     if (ProfileInterpreter) {
1611       // Test to see if we should create a method data oop
1612       __ cmp32(rax,
1613                ExternalAddress((address) &InvocationCounter::InterpreterProfileLimit));
1614       __ jcc(Assembler::less, dispatch);
1615 
1616       // if no method data exists, go to profile method
1617       __ test_method_data_pointer(rax, profile_method);
1618 
1619       if (UseOnStackReplacement) {
1620         // check for overflow against ebx which is the MDO taken count
1621         __ cmp32(rbx,
1622                  ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1623         __ jcc(Assembler::below, dispatch);
1624 
1625         // When ProfileInterpreter is on, the backedge_count comes
1626         // from the methodDataOop, which value does not get reset on
1627         // the call to frequency_counter_overflow().  To avoid
1628         // excessive calls to the overflow routine while the method is
1629         // being compiled, add a second test to make sure the overflow
1630         // function is called only once every overflow_frequency.
1631         const int overflow_frequency = 1024;
1632         __ andl(rbx, overflow_frequency - 1);
1633         __ jcc(Assembler::zero, backedge_counter_overflow);
1634 
1635       }
1636     } else {
1637       if (UseOnStackReplacement) {
1638         // check for overflow against eax, which is the sum of the
1639         // counters
1640         __ cmp32(rax,
1641                  ExternalAddress((address) &InvocationCounter::InterpreterBackwardBranchLimit));
1642         __ jcc(Assembler::aboveEqual, backedge_counter_overflow);
1643 
1644       }
1645     }
1646     __ bind(dispatch);
1647   }
1648 
1649   // Pre-load the next target bytecode into rbx
1650   __ load_unsigned_byte(rbx, Address(r13, 0));
1651 
1652   // continue with the bytecode @ target
1653   // eax: return bci for jsr's, unused otherwise
1654   // ebx: target bytecode
1655   // r13: target bcp
1656   __ dispatch_only(vtos);
1657 
1658   if (UseLoopCounter) {
1659     if (ProfileInterpreter) {
1660       // Out-of-line code to allocate method data oop.
1661       __ bind(profile_method);
1662       __ call_VM(noreg,
1663                  CAST_FROM_FN_PTR(address,
1664                                   InterpreterRuntime::profile_method), r13);
1665       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1666       __ movptr(rcx, Address(rbp, method_offset));
1667       __ movptr(rcx, Address(rcx,
1668                              in_bytes(methodOopDesc::method_data_offset())));
1669       __ movptr(Address(rbp, frame::interpreter_frame_mdx_offset * wordSize),
1670                 rcx);
1671       __ test_method_data_pointer(rcx, dispatch);
1672       // offset non-null mdp by MDO::data_offset() + IR::profile_method()
1673       __ addptr(rcx, in_bytes(methodDataOopDesc::data_offset()));
1674       __ addptr(rcx, rax);
1675       __ movptr(Address(rbp, frame::interpreter_frame_mdx_offset * wordSize),
1676                 rcx);
1677       __ jmp(dispatch);
1678     }
1679 
1680     if (UseOnStackReplacement) {
1681       // invocation counter overflow
1682       __ bind(backedge_counter_overflow);
1683       __ negptr(rdx);
1684       __ addptr(rdx, r13); // branch bcp
1685       // IcoResult frequency_counter_overflow([JavaThread*], address branch_bcp)
1686       __ call_VM(noreg,
1687                  CAST_FROM_FN_PTR(address,
1688                                   InterpreterRuntime::frequency_counter_overflow),
1689                  rdx);
1690       __ load_unsigned_byte(rbx, Address(r13, 0));  // restore target bytecode
1691 
1692       // rax: osr nmethod (osr ok) or NULL (osr not possible)
1693       // ebx: target bytecode
1694       // rdx: scratch
1695       // r14: locals pointer
1696       // r13: bcp
1697       __ testptr(rax, rax);                        // test result
1698       __ jcc(Assembler::zero, dispatch);         // no osr if null
1699       // nmethod may have been invalidated (VM may block upon call_VM return)
1700       __ movl(rcx, Address(rax, nmethod::entry_bci_offset()));
1701       __ cmpl(rcx, InvalidOSREntryBci);
1702       __ jcc(Assembler::equal, dispatch);
1703 
1704       // We have the address of an on stack replacement routine in eax
1705       // We need to prepare to execute the OSR method. First we must
1706       // migrate the locals and monitors off of the stack.
1707 
1708       __ mov(r13, rax);                             // save the nmethod
1709 
1710       call_VM(noreg, CAST_FROM_FN_PTR(address, SharedRuntime::OSR_migration_begin));
1711 
1712       // eax is OSR buffer, move it to expected parameter location
1713       __ mov(j_rarg0, rax);
1714 
1715       // We use j_rarg definitions here so that registers don't conflict as parameter
1716       // registers change across platforms as we are in the midst of a calling
1717       // sequence to the OSR nmethod and we don't want collision. These are NOT parameters.
1718 
1719       const Register retaddr = j_rarg2;
1720       const Register sender_sp = j_rarg1;
1721 
1722       // pop the interpreter frame
1723       __ movptr(sender_sp, Address(rbp, frame::interpreter_frame_sender_sp_offset * wordSize)); // get sender sp
1724       __ leave();                                // remove frame anchor
1725       __ pop(retaddr);                           // get return address
1726       __ mov(rsp, sender_sp);                   // set sp to sender sp
1727       // Ensure compiled code always sees stack at proper alignment
1728       __ andptr(rsp, -(StackAlignmentInBytes));
1729 
1730       // unlike x86 we need no specialized return from compiled code
1731       // to the interpreter or the call stub.
1732 
1733       // push the return address
1734       __ push(retaddr);
1735 
1736       // and begin the OSR nmethod
1737       __ jmp(Address(r13, nmethod::osr_entry_point_offset()));
1738     }
1739   }
1740 }
1741 
1742 
1743 void TemplateTable::if_0cmp(Condition cc) {
1744   transition(itos, vtos);
1745   // assume branch is more often taken than not (loops use backward branches)
1746   Label not_taken;
1747   __ testl(rax, rax);
1748   __ jcc(j_not(cc), not_taken);
1749   branch(false, false);
1750   __ bind(not_taken);
1751   __ profile_not_taken_branch(rax);
1752 }
1753 
1754 void TemplateTable::if_icmp(Condition cc) {
1755   transition(itos, vtos);
1756   // assume branch is more often taken than not (loops use backward branches)
1757   Label not_taken;
1758   __ pop_i(rdx);
1759   __ cmpl(rdx, rax);
1760   __ jcc(j_not(cc), not_taken);
1761   branch(false, false);
1762   __ bind(not_taken);
1763   __ profile_not_taken_branch(rax);
1764 }
1765 
1766 void TemplateTable::if_nullcmp(Condition cc) {
1767   transition(atos, vtos);
1768   // assume branch is more often taken than not (loops use backward branches)
1769   Label not_taken;
1770   __ testptr(rax, rax);
1771   __ jcc(j_not(cc), not_taken);
1772   branch(false, false);
1773   __ bind(not_taken);
1774   __ profile_not_taken_branch(rax);
1775 }
1776 
1777 void TemplateTable::if_acmp(Condition cc) {
1778   transition(atos, vtos);
1779   // assume branch is more often taken than not (loops use backward branches)
1780   Label not_taken;
1781   __ pop_ptr(rdx);
1782   __ cmpptr(rdx, rax);
1783   __ jcc(j_not(cc), not_taken);
1784   branch(false, false);
1785   __ bind(not_taken);
1786   __ profile_not_taken_branch(rax);
1787 }
1788 
1789 void TemplateTable::ret() {
1790   transition(vtos, vtos);
1791   locals_index(rbx);
1792   __ movslq(rbx, iaddress(rbx)); // get return bci, compute return bcp
1793   __ profile_ret(rbx, rcx);
1794   __ get_method(rax);
1795   __ movptr(r13, Address(rax, methodOopDesc::const_offset()));
1796   __ lea(r13, Address(r13, rbx, Address::times_1,
1797                       constMethodOopDesc::codes_offset()));
1798   __ dispatch_next(vtos);
1799 }
1800 
1801 void TemplateTable::wide_ret() {
1802   transition(vtos, vtos);
1803   locals_index_wide(rbx);
1804   __ movptr(rbx, aaddress(rbx)); // get return bci, compute return bcp
1805   __ profile_ret(rbx, rcx);
1806   __ get_method(rax);
1807   __ movptr(r13, Address(rax, methodOopDesc::const_offset()));
1808   __ lea(r13, Address(r13, rbx, Address::times_1, constMethodOopDesc::codes_offset()));
1809   __ dispatch_next(vtos);
1810 }
1811 
1812 void TemplateTable::tableswitch() {
1813   Label default_case, continue_execution;
1814   transition(itos, vtos);
1815   // align r13
1816   __ lea(rbx, at_bcp(BytesPerInt));
1817   __ andptr(rbx, -BytesPerInt);
1818   // load lo & hi
1819   __ movl(rcx, Address(rbx, BytesPerInt));
1820   __ movl(rdx, Address(rbx, 2 * BytesPerInt));
1821   __ bswapl(rcx);
1822   __ bswapl(rdx);
1823   // check against lo & hi
1824   __ cmpl(rax, rcx);
1825   __ jcc(Assembler::less, default_case);
1826   __ cmpl(rax, rdx);
1827   __ jcc(Assembler::greater, default_case);
1828   // lookup dispatch offset
1829   __ subl(rax, rcx);
1830   __ movl(rdx, Address(rbx, rax, Address::times_4, 3 * BytesPerInt));
1831   __ profile_switch_case(rax, rbx, rcx);
1832   // continue execution
1833   __ bind(continue_execution);
1834   __ bswapl(rdx);
1835   __ movl2ptr(rdx, rdx);
1836   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1837   __ addptr(r13, rdx);
1838   __ dispatch_only(vtos);
1839   // handle default
1840   __ bind(default_case);
1841   __ profile_switch_default(rax);
1842   __ movl(rdx, Address(rbx, 0));
1843   __ jmp(continue_execution);
1844 }
1845 
1846 void TemplateTable::lookupswitch() {
1847   transition(itos, itos);
1848   __ stop("lookupswitch bytecode should have been rewritten");
1849 }
1850 
1851 void TemplateTable::fast_linearswitch() {
1852   transition(itos, vtos);
1853   Label loop_entry, loop, found, continue_execution;
1854   // bswap rax so we can avoid bswapping the table entries
1855   __ bswapl(rax);
1856   // align r13
1857   __ lea(rbx, at_bcp(BytesPerInt)); // btw: should be able to get rid of
1858                                     // this instruction (change offsets
1859                                     // below)
1860   __ andptr(rbx, -BytesPerInt);
1861   // set counter
1862   __ movl(rcx, Address(rbx, BytesPerInt));
1863   __ bswapl(rcx);
1864   __ jmpb(loop_entry);
1865   // table search
1866   __ bind(loop);
1867   __ cmpl(rax, Address(rbx, rcx, Address::times_8, 2 * BytesPerInt));
1868   __ jcc(Assembler::equal, found);
1869   __ bind(loop_entry);
1870   __ decrementl(rcx);
1871   __ jcc(Assembler::greaterEqual, loop);
1872   // default case
1873   __ profile_switch_default(rax);
1874   __ movl(rdx, Address(rbx, 0));
1875   __ jmp(continue_execution);
1876   // entry found -> get offset
1877   __ bind(found);
1878   __ movl(rdx, Address(rbx, rcx, Address::times_8, 3 * BytesPerInt));
1879   __ profile_switch_case(rcx, rax, rbx);
1880   // continue execution
1881   __ bind(continue_execution);
1882   __ bswapl(rdx);
1883   __ movl2ptr(rdx, rdx);
1884   __ load_unsigned_byte(rbx, Address(r13, rdx, Address::times_1));
1885   __ addptr(r13, rdx);
1886   __ dispatch_only(vtos);
1887 }
1888 
1889 void TemplateTable::fast_binaryswitch() {
1890   transition(itos, vtos);
1891   // Implementation using the following core algorithm:
1892   //
1893   // int binary_search(int key, LookupswitchPair* array, int n) {
1894   //   // Binary search according to "Methodik des Programmierens" by
1895   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1896   //   int i = 0;
1897   //   int j = n;
1898   //   while (i+1 < j) {
1899   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1900   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1901   //     // where a stands for the array and assuming that the (inexisting)
1902   //     // element a[n] is infinitely big.
1903   //     int h = (i + j) >> 1;
1904   //     // i < h < j
1905   //     if (key < array[h].fast_match()) {
1906   //       j = h;
1907   //     } else {
1908   //       i = h;
1909   //     }
1910   //   }
1911   //   // R: a[i] <= key < a[i+1] or Q
1912   //   // (i.e., if key is within array, i is the correct index)
1913   //   return i;
1914   // }
1915 
1916   // Register allocation
1917   const Register key   = rax; // already set (tosca)
1918   const Register array = rbx;
1919   const Register i     = rcx;
1920   const Register j     = rdx;
1921   const Register h     = rdi;
1922   const Register temp  = rsi;
1923 
1924   // Find array start
1925   __ lea(array, at_bcp(3 * BytesPerInt)); // btw: should be able to
1926                                           // get rid of this
1927                                           // instruction (change
1928                                           // offsets below)
1929   __ andptr(array, -BytesPerInt);
1930 
1931   // Initialize i & j
1932   __ xorl(i, i);                            // i = 0;
1933   __ movl(j, Address(array, -BytesPerInt)); // j = length(array);
1934 
1935   // Convert j into native byteordering
1936   __ bswapl(j);
1937 
1938   // And start
1939   Label entry;
1940   __ jmp(entry);
1941 
1942   // binary search loop
1943   {
1944     Label loop;
1945     __ bind(loop);
1946     // int h = (i + j) >> 1;
1947     __ leal(h, Address(i, j, Address::times_1)); // h = i + j;
1948     __ sarl(h, 1);                               // h = (i + j) >> 1;
1949     // if (key < array[h].fast_match()) {
1950     //   j = h;
1951     // } else {
1952     //   i = h;
1953     // }
1954     // Convert array[h].match to native byte-ordering before compare
1955     __ movl(temp, Address(array, h, Address::times_8));
1956     __ bswapl(temp);
1957     __ cmpl(key, temp);
1958     // j = h if (key <  array[h].fast_match())
1959     __ cmovl(Assembler::less, j, h);
1960     // i = h if (key >= array[h].fast_match())
1961     __ cmovl(Assembler::greaterEqual, i, h);
1962     // while (i+1 < j)
1963     __ bind(entry);
1964     __ leal(h, Address(i, 1)); // i+1
1965     __ cmpl(h, j);             // i+1 < j
1966     __ jcc(Assembler::less, loop);
1967   }
1968 
1969   // end of binary search, result index is i (must check again!)
1970   Label default_case;
1971   // Convert array[i].match to native byte-ordering before compare
1972   __ movl(temp, Address(array, i, Address::times_8));
1973   __ bswapl(temp);
1974   __ cmpl(key, temp);
1975   __ jcc(Assembler::notEqual, default_case);
1976 
1977   // entry found -> j = offset
1978   __ movl(j , Address(array, i, Address::times_8, BytesPerInt));
1979   __ profile_switch_case(i, key, array);
1980   __ bswapl(j);
1981   __ movl2ptr(j, j);
1982   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
1983   __ addptr(r13, j);
1984   __ dispatch_only(vtos);
1985 
1986   // default case -> j = default offset
1987   __ bind(default_case);
1988   __ profile_switch_default(i);
1989   __ movl(j, Address(array, -2 * BytesPerInt));
1990   __ bswapl(j);
1991   __ movl2ptr(j, j);
1992   __ load_unsigned_byte(rbx, Address(r13, j, Address::times_1));
1993   __ addptr(r13, j);
1994   __ dispatch_only(vtos);
1995 }
1996 
1997 
1998 void TemplateTable::_return(TosState state) {
1999   transition(state, state);
2000   assert(_desc->calls_vm(),
2001          "inconsistent calls_vm information"); // call in remove_activation
2002 
2003   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
2004     assert(state == vtos, "only valid state");
2005     __ movptr(c_rarg1, aaddress(0));
2006     __ load_klass(rdi, c_rarg1);
2007     __ movl(rdi, Address(rdi, Klass::access_flags_offset_in_bytes() + sizeof(oopDesc)));
2008     __ testl(rdi, JVM_ACC_HAS_FINALIZER);
2009     Label skip_register_finalizer;
2010     __ jcc(Assembler::zero, skip_register_finalizer);
2011 
2012     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), c_rarg1);
2013 
2014     __ bind(skip_register_finalizer);
2015   }
2016 
2017   __ remove_activation(state, r13);
2018   __ jmp(r13);
2019 }
2020 
2021 // ----------------------------------------------------------------------------
2022 // Volatile variables demand their effects be made known to all CPU's
2023 // in order.  Store buffers on most chips allow reads & writes to
2024 // reorder; the JMM's ReadAfterWrite.java test fails in -Xint mode
2025 // without some kind of memory barrier (i.e., it's not sufficient that
2026 // the interpreter does not reorder volatile references, the hardware
2027 // also must not reorder them).
2028 //
2029 // According to the new Java Memory Model (JMM):
2030 // (1) All volatiles are serialized wrt to each other.  ALSO reads &
2031 //     writes act as aquire & release, so:
2032 // (2) A read cannot let unrelated NON-volatile memory refs that
2033 //     happen after the read float up to before the read.  It's OK for
2034 //     non-volatile memory refs that happen before the volatile read to
2035 //     float down below it.
2036 // (3) Similar a volatile write cannot let unrelated NON-volatile
2037 //     memory refs that happen BEFORE the write float down to after the
2038 //     write.  It's OK for non-volatile memory refs that happen after the
2039 //     volatile write to float up before it.
2040 //
2041 // We only put in barriers around volatile refs (they are expensive),
2042 // not _between_ memory refs (that would require us to track the
2043 // flavor of the previous memory refs).  Requirements (2) and (3)
2044 // require some barriers before volatile stores and after volatile
2045 // loads.  These nearly cover requirement (1) but miss the
2046 // volatile-store-volatile-load case.  This final case is placed after
2047 // volatile-stores although it could just as well go before
2048 // volatile-loads.
2049 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits
2050                                      order_constraint) {
2051   // Helper function to insert a is-volatile test and memory barrier
2052   if (os::is_MP()) { // Not needed on single CPU
2053     __ membar(order_constraint);
2054   }
2055 }
2056 
2057 void TemplateTable::resolve_cache_and_index(int byte_no,
2058                                             Register Rcache,
2059                                             Register index) {
2060   assert(byte_no == 1 || byte_no == 2, "byte_no out of range");
2061 
2062   const Register temp = rbx;
2063   assert_different_registers(Rcache, index, temp);
2064 
2065   const int shift_count = (1 + byte_no) * BitsPerByte;
2066   Label resolved;
2067   __ get_cache_and_index_at_bcp(Rcache, index, 1);
2068   __ movl(temp, Address(Rcache,
2069                         index, Address::times_8,
2070                         constantPoolCacheOopDesc::base_offset() +
2071                         ConstantPoolCacheEntry::indices_offset()));
2072   __ shrl(temp, shift_count);
2073   // have we resolved this bytecode?
2074   __ andl(temp, 0xFF);
2075   __ cmpl(temp, (int) bytecode());
2076   __ jcc(Assembler::equal, resolved);
2077 
2078   // resolve first time through
2079   address entry;
2080   switch (bytecode()) {
2081   case Bytecodes::_getstatic:
2082   case Bytecodes::_putstatic:
2083   case Bytecodes::_getfield:
2084   case Bytecodes::_putfield:
2085     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put);
2086     break;
2087   case Bytecodes::_invokevirtual:
2088   case Bytecodes::_invokespecial:
2089   case Bytecodes::_invokestatic:
2090   case Bytecodes::_invokeinterface:
2091     entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);
2092     break;
2093   default:
2094     ShouldNotReachHere();
2095     break;
2096   }
2097   __ movl(temp, (int) bytecode());
2098   __ call_VM(noreg, entry, temp);
2099 
2100   // Update registers with resolved info
2101   __ get_cache_and_index_at_bcp(Rcache, index, 1);
2102   __ bind(resolved);
2103 }
2104 
2105 // The Rcache and index registers must be set before call
2106 void TemplateTable::load_field_cp_cache_entry(Register obj,
2107                                               Register cache,
2108                                               Register index,
2109                                               Register off,
2110                                               Register flags,
2111                                               bool is_static = false) {
2112   assert_different_registers(cache, index, flags, off);
2113 
2114   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2115   // Field offset
2116   __ movptr(off, Address(cache, index, Address::times_8,
2117                          in_bytes(cp_base_offset +
2118                                   ConstantPoolCacheEntry::f2_offset())));
2119   // Flags
2120   __ movl(flags, Address(cache, index, Address::times_8,
2121                          in_bytes(cp_base_offset +
2122                                   ConstantPoolCacheEntry::flags_offset())));
2123 
2124   // klass overwrite register
2125   if (is_static) {
2126     __ movptr(obj, Address(cache, index, Address::times_8,
2127                            in_bytes(cp_base_offset +
2128                                     ConstantPoolCacheEntry::f1_offset())));
2129   }
2130 }
2131 
2132 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2133                                                Register method,
2134                                                Register itable_index,
2135                                                Register flags,
2136                                                bool is_invokevirtual,
2137                                                bool is_invokevfinal /*unused*/) {
2138   // setup registers
2139   const Register cache = rcx;
2140   const Register index = rdx;
2141   assert_different_registers(method, flags);
2142   assert_different_registers(method, cache, index);
2143   assert_different_registers(itable_index, flags);
2144   assert_different_registers(itable_index, cache, index);
2145   // determine constant pool cache field offsets
2146   const int method_offset = in_bytes(
2147     constantPoolCacheOopDesc::base_offset() +
2148       (is_invokevirtual
2149        ? ConstantPoolCacheEntry::f2_offset()
2150        : ConstantPoolCacheEntry::f1_offset()));
2151   const int flags_offset = in_bytes(constantPoolCacheOopDesc::base_offset() +
2152                                     ConstantPoolCacheEntry::flags_offset());
2153   // access constant pool cache fields
2154   const int index_offset = in_bytes(constantPoolCacheOopDesc::base_offset() +
2155                                     ConstantPoolCacheEntry::f2_offset());
2156 
2157   resolve_cache_and_index(byte_no, cache, index);
2158 
2159   assert(wordSize == 8, "adjust code below");
2160   __ movptr(method, Address(cache, index, Address::times_8, method_offset));
2161   if (itable_index != noreg) {
2162     __ movptr(itable_index,
2163             Address(cache, index, Address::times_8, index_offset));
2164   }
2165   __ movl(flags , Address(cache, index, Address::times_8, flags_offset));
2166 }
2167 
2168 
2169 // The registers cache and index expected to be set before call.
2170 // Correct values of the cache and index registers are preserved.
2171 void TemplateTable::jvmti_post_field_access(Register cache, Register index,
2172                                             bool is_static, bool has_tos) {
2173   // do the JVMTI work here to avoid disturbing the register state below
2174   // We use c_rarg registers here because we want to use the register used in
2175   // the call to the VM
2176   if (JvmtiExport::can_post_field_access()) {
2177     // Check to see if a field access watch has been set before we
2178     // take the time to call into the VM.
2179     Label L1;
2180     assert_different_registers(cache, index, rax);
2181     __ mov32(rax, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2182     __ testl(rax, rax);
2183     __ jcc(Assembler::zero, L1);
2184 
2185     __ get_cache_and_index_at_bcp(c_rarg2, c_rarg3, 1);
2186 
2187     // cache entry pointer
2188     __ addptr(c_rarg2, in_bytes(constantPoolCacheOopDesc::base_offset()));
2189     __ shll(c_rarg3, LogBytesPerWord);
2190     __ addptr(c_rarg2, c_rarg3);
2191     if (is_static) {
2192       __ xorl(c_rarg1, c_rarg1); // NULL object reference
2193     } else {
2194       __ movptr(c_rarg1, at_tos()); // get object pointer without popping it
2195       __ verify_oop(c_rarg1);
2196     }
2197     // c_rarg1: object pointer or NULL
2198     // c_rarg2: cache entry pointer
2199     // c_rarg3: jvalue object on the stack
2200     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
2201                                        InterpreterRuntime::post_field_access),
2202                c_rarg1, c_rarg2, c_rarg3);
2203     __ get_cache_and_index_at_bcp(cache, index, 1);
2204     __ bind(L1);
2205   }
2206 }
2207 
2208 void TemplateTable::pop_and_check_object(Register r) {
2209   __ pop_ptr(r);
2210   __ null_check(r);  // for field access must check obj.
2211   __ verify_oop(r);
2212 }
2213 
2214 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2215   transition(vtos, vtos);
2216 
2217   const Register cache = rcx;
2218   const Register index = rdx;
2219   const Register obj   = c_rarg3;
2220   const Register off   = rbx;
2221   const Register flags = rax;
2222   const Register bc = c_rarg3; // uses same reg as obj, so don't mix them
2223 
2224   resolve_cache_and_index(byte_no, cache, index);
2225   jvmti_post_field_access(cache, index, is_static, false);
2226   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2227 
2228   if (!is_static) {
2229     // obj is on the stack
2230     pop_and_check_object(obj);
2231   }
2232 
2233   const Address field(obj, off, Address::times_1);
2234 
2235   Label Done, notByte, notInt, notShort, notChar,
2236               notLong, notFloat, notObj, notDouble;
2237 
2238   __ shrl(flags, ConstantPoolCacheEntry::tosBits);
2239   assert(btos == 0, "change code, btos != 0");
2240 
2241   __ andl(flags, 0x0F);
2242   __ jcc(Assembler::notZero, notByte);
2243   // btos
2244   __ load_signed_byte(rax, field);
2245   __ push(btos);
2246   // Rewrite bytecode to be faster
2247   if (!is_static) {
2248     patch_bytecode(Bytecodes::_fast_bgetfield, bc, rbx);
2249   }
2250   __ jmp(Done);
2251 
2252   __ bind(notByte);
2253   __ cmpl(flags, atos);
2254   __ jcc(Assembler::notEqual, notObj);
2255   // atos
2256   __ load_heap_oop(rax, field);
2257   __ push(atos);
2258   if (!is_static) {
2259     patch_bytecode(Bytecodes::_fast_agetfield, bc, rbx);
2260   }
2261   __ jmp(Done);
2262 
2263   __ bind(notObj);
2264   __ cmpl(flags, itos);
2265   __ jcc(Assembler::notEqual, notInt);
2266   // itos
2267   __ movl(rax, field);
2268   __ push(itos);
2269   // Rewrite bytecode to be faster
2270   if (!is_static) {
2271     patch_bytecode(Bytecodes::_fast_igetfield, bc, rbx);
2272   }
2273   __ jmp(Done);
2274 
2275   __ bind(notInt);
2276   __ cmpl(flags, ctos);
2277   __ jcc(Assembler::notEqual, notChar);
2278   // ctos
2279   __ load_unsigned_short(rax, field);
2280   __ push(ctos);
2281   // Rewrite bytecode to be faster
2282   if (!is_static) {
2283     patch_bytecode(Bytecodes::_fast_cgetfield, bc, rbx);
2284   }
2285   __ jmp(Done);
2286 
2287   __ bind(notChar);
2288   __ cmpl(flags, stos);
2289   __ jcc(Assembler::notEqual, notShort);
2290   // stos
2291   __ load_signed_short(rax, field);
2292   __ push(stos);
2293   // Rewrite bytecode to be faster
2294   if (!is_static) {
2295     patch_bytecode(Bytecodes::_fast_sgetfield, bc, rbx);
2296   }
2297   __ jmp(Done);
2298 
2299   __ bind(notShort);
2300   __ cmpl(flags, ltos);
2301   __ jcc(Assembler::notEqual, notLong);
2302   // ltos
2303   __ movq(rax, field);
2304   __ push(ltos);
2305   // Rewrite bytecode to be faster
2306   if (!is_static) {
2307     patch_bytecode(Bytecodes::_fast_lgetfield, bc, rbx);
2308   }
2309   __ jmp(Done);
2310 
2311   __ bind(notLong);
2312   __ cmpl(flags, ftos);
2313   __ jcc(Assembler::notEqual, notFloat);
2314   // ftos
2315   __ movflt(xmm0, field);
2316   __ push(ftos);
2317   // Rewrite bytecode to be faster
2318   if (!is_static) {
2319     patch_bytecode(Bytecodes::_fast_fgetfield, bc, rbx);
2320   }
2321   __ jmp(Done);
2322 
2323   __ bind(notFloat);
2324 #ifdef ASSERT
2325   __ cmpl(flags, dtos);
2326   __ jcc(Assembler::notEqual, notDouble);
2327 #endif
2328   // dtos
2329   __ movdbl(xmm0, field);
2330   __ push(dtos);
2331   // Rewrite bytecode to be faster
2332   if (!is_static) {
2333     patch_bytecode(Bytecodes::_fast_dgetfield, bc, rbx);
2334   }
2335 #ifdef ASSERT
2336   __ jmp(Done);
2337 
2338   __ bind(notDouble);
2339   __ stop("Bad state");
2340 #endif
2341 
2342   __ bind(Done);
2343   // [jk] not needed currently
2344   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadLoad |
2345   //                                              Assembler::LoadStore));
2346 }
2347 
2348 
2349 void TemplateTable::getfield(int byte_no) {
2350   getfield_or_static(byte_no, false);
2351 }
2352 
2353 void TemplateTable::getstatic(int byte_no) {
2354   getfield_or_static(byte_no, true);
2355 }
2356 
2357 // The registers cache and index expected to be set before call.
2358 // The function may destroy various registers, just not the cache and index registers.
2359 void TemplateTable::jvmti_post_field_mod(Register cache, Register index, bool is_static) {
2360   transition(vtos, vtos);
2361 
2362   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2363 
2364   if (JvmtiExport::can_post_field_modification()) {
2365     // Check to see if a field modification watch has been set before
2366     // we take the time to call into the VM.
2367     Label L1;
2368     assert_different_registers(cache, index, rax);
2369     __ mov32(rax, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2370     __ testl(rax, rax);
2371     __ jcc(Assembler::zero, L1);
2372 
2373     __ get_cache_and_index_at_bcp(c_rarg2, rscratch1, 1);
2374 
2375     if (is_static) {
2376       // Life is simple.  Null out the object pointer.
2377       __ xorl(c_rarg1, c_rarg1);
2378     } else {
2379       // Life is harder. The stack holds the value on top, followed by
2380       // the object.  We don't know the size of the value, though; it
2381       // could be one or two words depending on its type. As a result,
2382       // we must find the type to determine where the object is.
2383       __ movl(c_rarg3, Address(c_rarg2, rscratch1,
2384                            Address::times_8,
2385                            in_bytes(cp_base_offset +
2386                                      ConstantPoolCacheEntry::flags_offset())));
2387       __ shrl(c_rarg3, ConstantPoolCacheEntry::tosBits);
2388       // Make sure we don't need to mask rcx for tosBits after the
2389       // above shift
2390       ConstantPoolCacheEntry::verify_tosBits();
2391       __ movptr(c_rarg1, at_tos_p1());  // initially assume a one word jvalue
2392       __ cmpl(c_rarg3, ltos);
2393       __ cmovptr(Assembler::equal,
2394                  c_rarg1, at_tos_p2()); // ltos (two word jvalue)
2395       __ cmpl(c_rarg3, dtos);
2396       __ cmovptr(Assembler::equal,
2397                  c_rarg1, at_tos_p2()); // dtos (two word jvalue)
2398     }
2399     // cache entry pointer
2400     __ addptr(c_rarg2, in_bytes(cp_base_offset));
2401     __ shll(rscratch1, LogBytesPerWord);
2402     __ addptr(c_rarg2, rscratch1);
2403     // object (tos)
2404     __ mov(c_rarg3, rsp);
2405     // c_rarg1: object pointer set up above (NULL if static)
2406     // c_rarg2: cache entry pointer
2407     // c_rarg3: jvalue object on the stack
2408     __ call_VM(noreg,
2409                CAST_FROM_FN_PTR(address,
2410                                 InterpreterRuntime::post_field_modification),
2411                c_rarg1, c_rarg2, c_rarg3);
2412     __ get_cache_and_index_at_bcp(cache, index, 1);
2413     __ bind(L1);
2414   }
2415 }
2416 
2417 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2418   transition(vtos, vtos);
2419 
2420   const Register cache = rcx;
2421   const Register index = rdx;
2422   const Register obj   = rcx;
2423   const Register off   = rbx;
2424   const Register flags = rax;
2425   const Register bc    = c_rarg3;
2426 
2427   resolve_cache_and_index(byte_no, cache, index);
2428   jvmti_post_field_mod(cache, index, is_static);
2429   load_field_cp_cache_entry(obj, cache, index, off, flags, is_static);
2430 
2431   // [jk] not needed currently
2432   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2433   //                                              Assembler::StoreStore));
2434 
2435   Label notVolatile, Done;
2436   __ movl(rdx, flags);
2437   __ shrl(rdx, ConstantPoolCacheEntry::volatileField);
2438   __ andl(rdx, 0x1);
2439 
2440   // field address
2441   const Address field(obj, off, Address::times_1);
2442 
2443   Label notByte, notInt, notShort, notChar,
2444         notLong, notFloat, notObj, notDouble;
2445 
2446   __ shrl(flags, ConstantPoolCacheEntry::tosBits);
2447 
2448   assert(btos == 0, "change code, btos != 0");
2449   __ andl(flags, 0x0f);
2450   __ jcc(Assembler::notZero, notByte);
2451   // btos
2452   __ pop(btos);
2453   if (!is_static) pop_and_check_object(obj);
2454   __ movb(field, rax);
2455   if (!is_static) {
2456     patch_bytecode(Bytecodes::_fast_bputfield, bc, rbx);
2457   }
2458   __ jmp(Done);
2459 
2460   __ bind(notByte);
2461   __ cmpl(flags, atos);
2462   __ jcc(Assembler::notEqual, notObj);
2463   // atos
2464   __ pop(atos);
2465   if (!is_static) pop_and_check_object(obj);
2466 
2467   // Store into the field
2468   do_oop_store(_masm, field, rax, _bs->kind(), false);
2469 
2470   if (!is_static) {
2471     patch_bytecode(Bytecodes::_fast_aputfield, bc, rbx);
2472   }
2473   __ jmp(Done);
2474 
2475   __ bind(notObj);
2476   __ cmpl(flags, itos);
2477   __ jcc(Assembler::notEqual, notInt);
2478   // itos
2479   __ pop(itos);
2480   if (!is_static) pop_and_check_object(obj);
2481   __ movl(field, rax);
2482   if (!is_static) {
2483     patch_bytecode(Bytecodes::_fast_iputfield, bc, rbx);
2484   }
2485   __ jmp(Done);
2486 
2487   __ bind(notInt);
2488   __ cmpl(flags, ctos);
2489   __ jcc(Assembler::notEqual, notChar);
2490   // ctos
2491   __ pop(ctos);
2492   if (!is_static) pop_and_check_object(obj);
2493   __ movw(field, rax);
2494   if (!is_static) {
2495     patch_bytecode(Bytecodes::_fast_cputfield, bc, rbx);
2496   }
2497   __ jmp(Done);
2498 
2499   __ bind(notChar);
2500   __ cmpl(flags, stos);
2501   __ jcc(Assembler::notEqual, notShort);
2502   // stos
2503   __ pop(stos);
2504   if (!is_static) pop_and_check_object(obj);
2505   __ movw(field, rax);
2506   if (!is_static) {
2507     patch_bytecode(Bytecodes::_fast_sputfield, bc, rbx);
2508   }
2509   __ jmp(Done);
2510 
2511   __ bind(notShort);
2512   __ cmpl(flags, ltos);
2513   __ jcc(Assembler::notEqual, notLong);
2514   // ltos
2515   __ pop(ltos);
2516   if (!is_static) pop_and_check_object(obj);
2517   __ movq(field, rax);
2518   if (!is_static) {
2519     patch_bytecode(Bytecodes::_fast_lputfield, bc, rbx);
2520   }
2521   __ jmp(Done);
2522 
2523   __ bind(notLong);
2524   __ cmpl(flags, ftos);
2525   __ jcc(Assembler::notEqual, notFloat);
2526   // ftos
2527   __ pop(ftos);
2528   if (!is_static) pop_and_check_object(obj);
2529   __ movflt(field, xmm0);
2530   if (!is_static) {
2531     patch_bytecode(Bytecodes::_fast_fputfield, bc, rbx);
2532   }
2533   __ jmp(Done);
2534 
2535   __ bind(notFloat);
2536 #ifdef ASSERT
2537   __ cmpl(flags, dtos);
2538   __ jcc(Assembler::notEqual, notDouble);
2539 #endif
2540   // dtos
2541   __ pop(dtos);
2542   if (!is_static) pop_and_check_object(obj);
2543   __ movdbl(field, xmm0);
2544   if (!is_static) {
2545     patch_bytecode(Bytecodes::_fast_dputfield, bc, rbx);
2546   }
2547 
2548 #ifdef ASSERT
2549   __ jmp(Done);
2550 
2551   __ bind(notDouble);
2552   __ stop("Bad state");
2553 #endif
2554 
2555   __ bind(Done);
2556   // Check for volatile store
2557   __ testl(rdx, rdx);
2558   __ jcc(Assembler::zero, notVolatile);
2559   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2560                                                Assembler::StoreStore));
2561 
2562   __ bind(notVolatile);
2563 }
2564 
2565 void TemplateTable::putfield(int byte_no) {
2566   putfield_or_static(byte_no, false);
2567 }
2568 
2569 void TemplateTable::putstatic(int byte_no) {
2570   putfield_or_static(byte_no, true);
2571 }
2572 
2573 void TemplateTable::jvmti_post_fast_field_mod() {
2574   if (JvmtiExport::can_post_field_modification()) {
2575     // Check to see if a field modification watch has been set before
2576     // we take the time to call into the VM.
2577     Label L2;
2578     __ mov32(c_rarg3, ExternalAddress((address)JvmtiExport::get_field_modification_count_addr()));
2579     __ testl(c_rarg3, c_rarg3);
2580     __ jcc(Assembler::zero, L2);
2581     __ pop_ptr(rbx);                  // copy the object pointer from tos
2582     __ verify_oop(rbx);
2583     __ push_ptr(rbx);                 // put the object pointer back on tos
2584     __ subptr(rsp, sizeof(jvalue));  // add space for a jvalue object
2585     __ mov(c_rarg3, rsp);
2586     const Address field(c_rarg3, 0);
2587 
2588     switch (bytecode()) {          // load values into the jvalue object
2589     case Bytecodes::_fast_aputfield: __ movq(field, rax); break;
2590     case Bytecodes::_fast_lputfield: __ movq(field, rax); break;
2591     case Bytecodes::_fast_iputfield: __ movl(field, rax); break;
2592     case Bytecodes::_fast_bputfield: __ movb(field, rax); break;
2593     case Bytecodes::_fast_sputfield: // fall through
2594     case Bytecodes::_fast_cputfield: __ movw(field, rax); break;
2595     case Bytecodes::_fast_fputfield: __ movflt(field, xmm0); break;
2596     case Bytecodes::_fast_dputfield: __ movdbl(field, xmm0); break;
2597     default:
2598       ShouldNotReachHere();
2599     }
2600 
2601     // Save rax because call_VM() will clobber it, then use it for
2602     // JVMTI purposes
2603     __ push(rax);
2604     // access constant pool cache entry
2605     __ get_cache_entry_pointer_at_bcp(c_rarg2, rax, 1);
2606     __ verify_oop(rbx);
2607     // rbx: object pointer copied above
2608     // c_rarg2: cache entry pointer
2609     // c_rarg3: jvalue object on the stack
2610     __ call_VM(noreg,
2611                CAST_FROM_FN_PTR(address,
2612                                 InterpreterRuntime::post_field_modification),
2613                rbx, c_rarg2, c_rarg3);
2614     __ pop(rax);     // restore lower value
2615     __ addptr(rsp, sizeof(jvalue));  // release jvalue object space
2616     __ bind(L2);
2617   }
2618 }
2619 
2620 void TemplateTable::fast_storefield(TosState state) {
2621   transition(state, vtos);
2622 
2623   ByteSize base = constantPoolCacheOopDesc::base_offset();
2624 
2625   jvmti_post_fast_field_mod();
2626 
2627   // access constant pool cache
2628   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2629 
2630   // test for volatile with rdx
2631   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2632                        in_bytes(base +
2633                                 ConstantPoolCacheEntry::flags_offset())));
2634 
2635   // replace index with field offset from cache entry
2636   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2637                          in_bytes(base + ConstantPoolCacheEntry::f2_offset())));
2638 
2639   // [jk] not needed currently
2640   // volatile_barrier(Assembler::Membar_mask_bits(Assembler::LoadStore |
2641   //                                              Assembler::StoreStore));
2642 
2643   Label notVolatile;
2644   __ shrl(rdx, ConstantPoolCacheEntry::volatileField);
2645   __ andl(rdx, 0x1);
2646 
2647   // Get object from stack
2648   pop_and_check_object(rcx);
2649 
2650   // field address
2651   const Address field(rcx, rbx, Address::times_1);
2652 
2653   // access field
2654   switch (bytecode()) {
2655   case Bytecodes::_fast_aputfield:
2656     do_oop_store(_masm, field, rax, _bs->kind(), false);
2657     break;
2658   case Bytecodes::_fast_lputfield:
2659     __ movq(field, rax);
2660     break;
2661   case Bytecodes::_fast_iputfield:
2662     __ movl(field, rax);
2663     break;
2664   case Bytecodes::_fast_bputfield:
2665     __ movb(field, rax);
2666     break;
2667   case Bytecodes::_fast_sputfield:
2668     // fall through
2669   case Bytecodes::_fast_cputfield:
2670     __ movw(field, rax);
2671     break;
2672   case Bytecodes::_fast_fputfield:
2673     __ movflt(field, xmm0);
2674     break;
2675   case Bytecodes::_fast_dputfield:
2676     __ movdbl(field, xmm0);
2677     break;
2678   default:
2679     ShouldNotReachHere();
2680   }
2681 
2682   // Check for volatile store
2683   __ testl(rdx, rdx);
2684   __ jcc(Assembler::zero, notVolatile);
2685   volatile_barrier(Assembler::Membar_mask_bits(Assembler::StoreLoad |
2686                                                Assembler::StoreStore));
2687   __ bind(notVolatile);
2688 }
2689 
2690 
2691 void TemplateTable::fast_accessfield(TosState state) {
2692   transition(atos, state);
2693 
2694   // Do the JVMTI work here to avoid disturbing the register state below
2695   if (JvmtiExport::can_post_field_access()) {
2696     // Check to see if a field access watch has been set before we
2697     // take the time to call into the VM.
2698     Label L1;
2699     __ mov32(rcx, ExternalAddress((address) JvmtiExport::get_field_access_count_addr()));
2700     __ testl(rcx, rcx);
2701     __ jcc(Assembler::zero, L1);
2702     // access constant pool cache entry
2703     __ get_cache_entry_pointer_at_bcp(c_rarg2, rcx, 1);
2704     __ verify_oop(rax);
2705     __ mov(r12, rax);  // save object pointer before call_VM() clobbers it
2706     __ mov(c_rarg1, rax);
2707     // c_rarg1: object pointer copied above
2708     // c_rarg2: cache entry pointer
2709     __ call_VM(noreg,
2710                CAST_FROM_FN_PTR(address,
2711                                 InterpreterRuntime::post_field_access),
2712                c_rarg1, c_rarg2);
2713     __ mov(rax, r12); // restore object pointer
2714     __ reinit_heapbase();
2715     __ bind(L1);
2716   }
2717 
2718   // access constant pool cache
2719   __ get_cache_and_index_at_bcp(rcx, rbx, 1);
2720   // replace index with field offset from cache entry
2721   // [jk] not needed currently
2722   // if (os::is_MP()) {
2723   //   __ movl(rdx, Address(rcx, rbx, Address::times_8,
2724   //                        in_bytes(constantPoolCacheOopDesc::base_offset() +
2725   //                                 ConstantPoolCacheEntry::flags_offset())));
2726   //   __ shrl(rdx, ConstantPoolCacheEntry::volatileField);
2727   //   __ andl(rdx, 0x1);
2728   // }
2729   __ movptr(rbx, Address(rcx, rbx, Address::times_8,
2730                          in_bytes(constantPoolCacheOopDesc::base_offset() +
2731                                   ConstantPoolCacheEntry::f2_offset())));
2732 
2733   // rax: object
2734   __ verify_oop(rax);
2735   __ null_check(rax);
2736   Address field(rax, rbx, Address::times_1);
2737 
2738   // access field
2739   switch (bytecode()) {
2740   case Bytecodes::_fast_agetfield:
2741     __ load_heap_oop(rax, field);
2742     __ verify_oop(rax);
2743     break;
2744   case Bytecodes::_fast_lgetfield:
2745     __ movq(rax, field);
2746     break;
2747   case Bytecodes::_fast_igetfield:
2748     __ movl(rax, field);
2749     break;
2750   case Bytecodes::_fast_bgetfield:
2751     __ movsbl(rax, field);
2752     break;
2753   case Bytecodes::_fast_sgetfield:
2754     __ load_signed_short(rax, field);
2755     break;
2756   case Bytecodes::_fast_cgetfield:
2757     __ load_unsigned_short(rax, field);
2758     break;
2759   case Bytecodes::_fast_fgetfield:
2760     __ movflt(xmm0, field);
2761     break;
2762   case Bytecodes::_fast_dgetfield:
2763     __ movdbl(xmm0, field);
2764     break;
2765   default:
2766     ShouldNotReachHere();
2767   }
2768   // [jk] not needed currently
2769   // if (os::is_MP()) {
2770   //   Label notVolatile;
2771   //   __ testl(rdx, rdx);
2772   //   __ jcc(Assembler::zero, notVolatile);
2773   //   __ membar(Assembler::LoadLoad);
2774   //   __ bind(notVolatile);
2775   //};
2776 }
2777 
2778 void TemplateTable::fast_xaccess(TosState state) {
2779   transition(vtos, state);
2780 
2781   // get receiver
2782   __ movptr(rax, aaddress(0));
2783   debug_only(__ verify_local_tag(frame::TagReference, 0));
2784   // access constant pool cache
2785   __ get_cache_and_index_at_bcp(rcx, rdx, 2);
2786   __ movptr(rbx,
2787             Address(rcx, rdx, Address::times_8,
2788                     in_bytes(constantPoolCacheOopDesc::base_offset() +
2789                              ConstantPoolCacheEntry::f2_offset())));
2790   // make sure exception is reported in correct bcp range (getfield is
2791   // next instruction)
2792   __ increment(r13);
2793   __ null_check(rax);
2794   switch (state) {
2795   case itos:
2796     __ movl(rax, Address(rax, rbx, Address::times_1));
2797     break;
2798   case atos:
2799     __ load_heap_oop(rax, Address(rax, rbx, Address::times_1));
2800     __ verify_oop(rax);
2801     break;
2802   case ftos:
2803     __ movflt(xmm0, Address(rax, rbx, Address::times_1));
2804     break;
2805   default:
2806     ShouldNotReachHere();
2807   }
2808 
2809   // [jk] not needed currently
2810   // if (os::is_MP()) {
2811   //   Label notVolatile;
2812   //   __ movl(rdx, Address(rcx, rdx, Address::times_8,
2813   //                        in_bytes(constantPoolCacheOopDesc::base_offset() +
2814   //                                 ConstantPoolCacheEntry::flags_offset())));
2815   //   __ shrl(rdx, ConstantPoolCacheEntry::volatileField);
2816   //   __ testl(rdx, 0x1);
2817   //   __ jcc(Assembler::zero, notVolatile);
2818   //   __ membar(Assembler::LoadLoad);
2819   //   __ bind(notVolatile);
2820   // }
2821 
2822   __ decrement(r13);
2823 }
2824 
2825 
2826 
2827 //-----------------------------------------------------------------------------
2828 // Calls
2829 
2830 void TemplateTable::count_calls(Register method, Register temp) {
2831   // implemented elsewhere
2832   ShouldNotReachHere();
2833 }
2834 
2835 void TemplateTable::prepare_invoke(Register method,
2836                                    Register index,
2837                                    int byte_no,
2838                                    Bytecodes::Code code) {
2839   // determine flags
2840   const bool is_invokeinterface  = code == Bytecodes::_invokeinterface;
2841   const bool is_invokevirtual    = code == Bytecodes::_invokevirtual;
2842   const bool is_invokespecial    = code == Bytecodes::_invokespecial;
2843   const bool load_receiver       = code != Bytecodes::_invokestatic;
2844   const bool receiver_null_check = is_invokespecial;
2845   const bool save_flags = is_invokeinterface || is_invokevirtual;
2846   // setup registers & access constant pool cache
2847   const Register recv   = rcx;
2848   const Register flags  = rdx;
2849   assert_different_registers(method, index, recv, flags);
2850 
2851   // save 'interpreter return address'
2852   __ save_bcp();
2853 
2854   load_invoke_cp_cache_entry(byte_no, method, index, flags, is_invokevirtual);
2855 
2856   // load receiver if needed (note: no return address pushed yet)
2857   if (load_receiver) {
2858     __ movl(recv, flags);
2859     __ andl(recv, 0xFF);
2860     if (TaggedStackInterpreter) __ shll(recv, 1);  // index*2
2861     __ movptr(recv, Address(rsp, recv, Address::times_8,
2862                                  -Interpreter::expr_offset_in_bytes(1)));
2863     __ verify_oop(recv);
2864   }
2865 
2866   // do null check if needed
2867   if (receiver_null_check) {
2868     __ null_check(recv);
2869   }
2870 
2871   if (save_flags) {
2872     __ movl(r13, flags);
2873   }
2874 
2875   // compute return type
2876   __ shrl(flags, ConstantPoolCacheEntry::tosBits);
2877   // Make sure we don't need to mask flags for tosBits after the above shift
2878   ConstantPoolCacheEntry::verify_tosBits();
2879   // load return address
2880   {
2881     ExternalAddress return_5((address)Interpreter::return_5_addrs_by_index_table());
2882     ExternalAddress return_3((address)Interpreter::return_3_addrs_by_index_table());
2883     __ lea(rscratch1, (is_invokeinterface ? return_5 : return_3));
2884     __ movptr(flags, Address(rscratch1, flags, Address::times_8));
2885   }
2886 
2887   // push return address
2888   __ push(flags);
2889 
2890   // Restore flag field from the constant pool cache, and restore esi
2891   // for later null checks.  r13 is the bytecode pointer
2892   if (save_flags) {
2893     __ movl(flags, r13);
2894     __ restore_bcp();
2895   }
2896 }
2897 
2898 
2899 void TemplateTable::invokevirtual_helper(Register index,
2900                                          Register recv,
2901                                          Register flags) {
2902   // Uses temporary registers rax, rdx  assert_different_registers(index, recv, rax, rdx);
2903 
2904   // Test for an invoke of a final method
2905   Label notFinal;
2906   __ movl(rax, flags);
2907   __ andl(rax, (1 << ConstantPoolCacheEntry::vfinalMethod));
2908   __ jcc(Assembler::zero, notFinal);
2909 
2910   const Register method = index;  // method must be rbx
2911   assert(method == rbx,
2912          "methodOop must be rbx for interpreter calling convention");
2913 
2914   // do the call - the index is actually the method to call
2915   __ verify_oop(method);
2916 
2917   // It's final, need a null check here!
2918   __ null_check(recv);
2919 
2920   // profile this call
2921   __ profile_final_call(rax);
2922 
2923   __ jump_from_interpreted(method, rax);
2924 
2925   __ bind(notFinal);
2926 
2927   // get receiver klass
2928   __ null_check(recv, oopDesc::klass_offset_in_bytes());
2929   __ load_klass(rax, recv);
2930 
2931   __ verify_oop(rax);
2932 
2933   // profile this call
2934   __ profile_virtual_call(rax, r14, rdx);
2935 
2936   // get target methodOop & entry point
2937   const int base = instanceKlass::vtable_start_offset() * wordSize;
2938   assert(vtableEntry::size() * wordSize == 8,
2939          "adjust the scaling in the code below");
2940   __ movptr(method, Address(rax, index,
2941                                  Address::times_8,
2942                                  base + vtableEntry::method_offset_in_bytes()));
2943   __ movptr(rdx, Address(method, methodOopDesc::interpreter_entry_offset()));
2944   __ jump_from_interpreted(method, rdx);
2945 }
2946 
2947 
2948 void TemplateTable::invokevirtual(int byte_no) {
2949   transition(vtos, vtos);
2950   prepare_invoke(rbx, noreg, byte_no, bytecode());
2951 
2952   // rbx: index
2953   // rcx: receiver
2954   // rdx: flags
2955 
2956   invokevirtual_helper(rbx, rcx, rdx);
2957 }
2958 
2959 
2960 void TemplateTable::invokespecial(int byte_no) {
2961   transition(vtos, vtos);
2962   prepare_invoke(rbx, noreg, byte_no, bytecode());
2963   // do the call
2964   __ verify_oop(rbx);
2965   __ profile_call(rax);
2966   __ jump_from_interpreted(rbx, rax);
2967 }
2968 
2969 
2970 void TemplateTable::invokestatic(int byte_no) {
2971   transition(vtos, vtos);
2972   prepare_invoke(rbx, noreg, byte_no, bytecode());
2973   // do the call
2974   __ verify_oop(rbx);
2975   __ profile_call(rax);
2976   __ jump_from_interpreted(rbx, rax);
2977 }
2978 
2979 void TemplateTable::fast_invokevfinal(int byte_no) {
2980   transition(vtos, vtos);
2981   __ stop("fast_invokevfinal not used on amd64");
2982 }
2983 
2984 void TemplateTable::invokeinterface(int byte_no) {
2985   transition(vtos, vtos);
2986   prepare_invoke(rax, rbx, byte_no, bytecode());
2987 
2988   // rax: Interface
2989   // rbx: index
2990   // rcx: receiver
2991   // rdx: flags
2992 
2993   // Special case of invokeinterface called for virtual method of
2994   // java.lang.Object.  See cpCacheOop.cpp for details.
2995   // This code isn't produced by javac, but could be produced by
2996   // another compliant java compiler.
2997   Label notMethod;
2998   __ movl(r14, rdx);
2999   __ andl(r14, (1 << ConstantPoolCacheEntry::methodInterface));
3000   __ jcc(Assembler::zero, notMethod);
3001 
3002   invokevirtual_helper(rbx, rcx, rdx);
3003   __ bind(notMethod);
3004 
3005   // Get receiver klass into rdx - also a null check
3006   __ restore_locals(); // restore r14
3007   __ load_klass(rdx, rcx);
3008   __ verify_oop(rdx);
3009 
3010   // profile this call
3011   __ profile_virtual_call(rdx, r13, r14);
3012 
3013   Label no_such_interface, no_such_method;
3014 
3015   __ lookup_interface_method(// inputs: rec. class, interface, itable index
3016                              rdx, rax, rbx,
3017                              // outputs: method, scan temp. reg
3018                              rbx, r13,
3019                              no_such_interface);
3020 
3021   // rbx,: methodOop to call
3022   // rcx: receiver
3023   // Check for abstract method error
3024   // Note: This should be done more efficiently via a throw_abstract_method_error
3025   //       interpreter entry point and a conditional jump to it in case of a null
3026   //       method.
3027   __ testptr(rbx, rbx);
3028   __ jcc(Assembler::zero, no_such_method);
3029 
3030   // do the call
3031   // rcx: receiver
3032   // rbx,: methodOop
3033   __ jump_from_interpreted(rbx, rdx);
3034   __ should_not_reach_here();
3035 
3036   // exception handling code follows...
3037   // note: must restore interpreter registers to canonical
3038   //       state for exception handling to work correctly!
3039 
3040   __ bind(no_such_method);
3041   // throw exception
3042   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3043   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3044   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3045   __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3046   // the call_VM checks for exception, so we should never return here.
3047   __ should_not_reach_here();
3048 
3049   __ bind(no_such_interface);
3050   // throw exception
3051   __ pop(rbx);           // pop return address (pushed by prepare_invoke)
3052   __ restore_bcp();      // r13 must be correct for exception handler   (was destroyed)
3053   __ restore_locals();   // make sure locals pointer is correct as well (was destroyed)
3054   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3055                    InterpreterRuntime::throw_IncompatibleClassChangeError));
3056   // the call_VM checks for exception, so we should never return here.
3057   __ should_not_reach_here();
3058   return;
3059 }
3060 
3061 void TemplateTable::invokedynamic(int byte_no) {
3062   transition(vtos, vtos);
3063 
3064   if (!EnableInvokeDynamic) {
3065     // We should not encounter this bytecode if !EnableInvokeDynamic.
3066     // The verifier will stop it.  However, if we get past the verifier,
3067     // this will stop the thread in a reasonable way, without crashing the JVM.
3068     __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3069                      InterpreterRuntime::throw_IncompatibleClassChangeError));
3070     // the call_VM checks for exception, so we should never return here.
3071     __ should_not_reach_here();
3072     return;
3073   }
3074 
3075   __ stop("invokedynamic NYI");//6815692//
3076 }
3077 
3078 
3079 //-----------------------------------------------------------------------------
3080 // Allocation
3081 
3082 void TemplateTable::_new() {
3083   transition(vtos, atos);
3084   __ get_unsigned_2_byte_index_at_bcp(rdx, 1);
3085   Label slow_case;
3086   Label done;
3087   Label initialize_header;
3088   Label initialize_object; // including clearing the fields
3089   Label allocate_shared;
3090 
3091   __ get_cpool_and_tags(rsi, rax);
3092   // get instanceKlass
3093   __ movptr(rsi, Address(rsi, rdx,
3094                          Address::times_8, sizeof(constantPoolOopDesc)));
3095 
3096   // make sure the class we're about to instantiate has been
3097   // resolved. Note: slow_case does a pop of stack, which is why we
3098   // loaded class/pushed above
3099   const int tags_offset = typeArrayOopDesc::header_size(T_BYTE) * wordSize;
3100   __ cmpb(Address(rax, rdx, Address::times_1, tags_offset),
3101           JVM_CONSTANT_Class);
3102   __ jcc(Assembler::notEqual, slow_case);
3103 
3104   // make sure klass is initialized & doesn't have finalizer
3105   // make sure klass is fully initialized
3106   __ cmpl(Address(rsi,
3107                   instanceKlass::init_state_offset_in_bytes() +
3108                   sizeof(oopDesc)),
3109           instanceKlass::fully_initialized);
3110   __ jcc(Assembler::notEqual, slow_case);
3111 
3112   // get instance_size in instanceKlass (scaled to a count of bytes)
3113   __ movl(rdx,
3114           Address(rsi,
3115                   Klass::layout_helper_offset_in_bytes() + sizeof(oopDesc)));
3116   // test to see if it has a finalizer or is malformed in some way
3117   __ testl(rdx, Klass::_lh_instance_slow_path_bit);
3118   __ jcc(Assembler::notZero, slow_case);
3119 
3120   // Allocate the instance
3121   // 1) Try to allocate in the TLAB
3122   // 2) if fail and the object is large allocate in the shared Eden
3123   // 3) if the above fails (or is not applicable), go to a slow case
3124   // (creates a new TLAB, etc.)
3125 
3126   const bool allow_shared_alloc =
3127     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3128 
3129   if (UseTLAB) {
3130     __ movptr(rax, Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())));
3131     __ lea(rbx, Address(rax, rdx, Address::times_1));
3132     __ cmpptr(rbx, Address(r15_thread, in_bytes(JavaThread::tlab_end_offset())));
3133     __ jcc(Assembler::above, allow_shared_alloc ? allocate_shared : slow_case);
3134     __ movptr(Address(r15_thread, in_bytes(JavaThread::tlab_top_offset())), rbx);
3135     if (ZeroTLAB) {
3136       // the fields have been already cleared
3137       __ jmp(initialize_header);
3138     } else {
3139       // initialize both the header and fields
3140       __ jmp(initialize_object);
3141     }
3142   }
3143 
3144   // Allocation in the shared Eden, if allowed.
3145   //
3146   // rdx: instance size in bytes
3147   if (allow_shared_alloc) {
3148     __ bind(allocate_shared);
3149 
3150     ExternalAddress top((address)Universe::heap()->top_addr());
3151     ExternalAddress end((address)Universe::heap()->end_addr());
3152 
3153     const Register RtopAddr = rscratch1;
3154     const Register RendAddr = rscratch2;
3155 
3156     __ lea(RtopAddr, top);
3157     __ lea(RendAddr, end);
3158     __ movptr(rax, Address(RtopAddr, 0));
3159 
3160     // For retries rax gets set by cmpxchgq
3161     Label retry;
3162     __ bind(retry);
3163     __ lea(rbx, Address(rax, rdx, Address::times_1));
3164     __ cmpptr(rbx, Address(RendAddr, 0));
3165     __ jcc(Assembler::above, slow_case);
3166 
3167     // Compare rax with the top addr, and if still equal, store the new
3168     // top addr in rbx at the address of the top addr pointer. Sets ZF if was
3169     // equal, and clears it otherwise. Use lock prefix for atomicity on MPs.
3170     //
3171     // rax: object begin
3172     // rbx: object end
3173     // rdx: instance size in bytes
3174     if (os::is_MP()) {
3175       __ lock();
3176     }
3177     __ cmpxchgptr(rbx, Address(RtopAddr, 0));
3178 
3179     // if someone beat us on the allocation, try again, otherwise continue
3180     __ jcc(Assembler::notEqual, retry);
3181   }
3182 
3183   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3184     // The object is initialized before the header.  If the object size is
3185     // zero, go directly to the header initialization.
3186     __ bind(initialize_object);
3187     __ decrementl(rdx, sizeof(oopDesc));
3188     __ jcc(Assembler::zero, initialize_header);
3189 
3190     // Initialize object fields
3191     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3192     __ shrl(rdx, LogBytesPerLong);  // divide by oopSize to simplify the loop
3193     {
3194       Label loop;
3195       __ bind(loop);
3196       __ movq(Address(rax, rdx, Address::times_8,
3197                       sizeof(oopDesc) - oopSize),
3198               rcx);
3199       __ decrementl(rdx);
3200       __ jcc(Assembler::notZero, loop);
3201     }
3202 
3203     // initialize object header only.
3204     __ bind(initialize_header);
3205     if (UseBiasedLocking) {
3206       __ movptr(rscratch1, Address(rsi, Klass::prototype_header_offset_in_bytes() + klassOopDesc::klass_part_offset_in_bytes()));
3207       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()), rscratch1);
3208     } else {
3209       __ movptr(Address(rax, oopDesc::mark_offset_in_bytes()),
3210                (intptr_t) markOopDesc::prototype()); // header (address 0x1)
3211     }
3212     __ xorl(rcx, rcx); // use zero reg to clear memory (shorter code)
3213     __ store_klass_gap(rax, rcx);  // zero klass gap for compressed oops
3214     __ store_klass(rax, rsi);      // store klass last
3215     __ jmp(done);
3216   }
3217 
3218   {
3219     SkipIfEqual skip(_masm, &DTraceAllocProbes, false);
3220     // Trigger dtrace event for fastpath
3221     __ push(atos); // save the return value
3222     __ call_VM_leaf(
3223          CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), rax);
3224     __ pop(atos); // restore the return value
3225   }
3226 
3227   // slow case
3228   __ bind(slow_case);
3229   __ get_constant_pool(c_rarg1);
3230   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3231   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), c_rarg1, c_rarg2);
3232   __ verify_oop(rax);
3233 
3234   // continue
3235   __ bind(done);
3236 }
3237 
3238 void TemplateTable::newarray() {
3239   transition(itos, atos);
3240   __ load_unsigned_byte(c_rarg1, at_bcp(1));
3241   __ movl(c_rarg2, rax);
3242   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray),
3243           c_rarg1, c_rarg2);
3244 }
3245 
3246 void TemplateTable::anewarray() {
3247   transition(itos, atos);
3248   __ get_unsigned_2_byte_index_at_bcp(c_rarg2, 1);
3249   __ get_constant_pool(c_rarg1);
3250   __ movl(c_rarg3, rax);
3251   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray),
3252           c_rarg1, c_rarg2, c_rarg3);
3253 }
3254 
3255 void TemplateTable::arraylength() {
3256   transition(atos, itos);
3257   __ null_check(rax, arrayOopDesc::length_offset_in_bytes());
3258   __ movl(rax, Address(rax, arrayOopDesc::length_offset_in_bytes()));
3259 }
3260 
3261 void TemplateTable::checkcast() {
3262   transition(atos, atos);
3263   Label done, is_null, ok_is_subtype, quicked, resolved;
3264   __ testptr(rax, rax); // object is in rax
3265   __ jcc(Assembler::zero, is_null);
3266 
3267   // Get cpool & tags index
3268   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3269   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3270   // See if bytecode has already been quicked
3271   __ cmpb(Address(rdx, rbx,
3272                   Address::times_1,
3273                   typeArrayOopDesc::header_size(T_BYTE) * wordSize),
3274           JVM_CONSTANT_Class);
3275   __ jcc(Assembler::equal, quicked);
3276   __ push(atos); // save receiver for result, and for GC
3277   __ mov(r12, rcx); // save rcx XXX
3278   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3279   __ movq(rcx, r12); // restore rcx XXX
3280   __ reinit_heapbase();
3281   __ pop_ptr(rdx); // restore receiver
3282   __ jmpb(resolved);
3283 
3284   // Get superklass in rax and subklass in rbx
3285   __ bind(quicked);
3286   __ mov(rdx, rax); // Save object in rdx; rax needed for subtype check
3287   __ movptr(rax, Address(rcx, rbx,
3288                        Address::times_8, sizeof(constantPoolOopDesc)));
3289 
3290   __ bind(resolved);
3291   __ load_klass(rbx, rdx);
3292 
3293   // Generate subtype check.  Blows rcx, rdi.  Object in rdx.
3294   // Superklass in rax.  Subklass in rbx.
3295   __ gen_subtype_check(rbx, ok_is_subtype);
3296 
3297   // Come here on failure
3298   __ push_ptr(rdx);
3299   // object is at TOS
3300   __ jump(ExternalAddress(Interpreter::_throw_ClassCastException_entry));
3301 
3302   // Come here on success
3303   __ bind(ok_is_subtype);
3304   __ mov(rax, rdx); // Restore object in rdx
3305 
3306   // Collect counts on whether this check-cast sees NULLs a lot or not.
3307   if (ProfileInterpreter) {
3308     __ jmp(done);
3309     __ bind(is_null);
3310     __ profile_null_seen(rcx);
3311   } else {
3312     __ bind(is_null);   // same as 'done'
3313   }
3314   __ bind(done);
3315 }
3316 
3317 void TemplateTable::instanceof() {
3318   transition(atos, itos);
3319   Label done, is_null, ok_is_subtype, quicked, resolved;
3320   __ testptr(rax, rax);
3321   __ jcc(Assembler::zero, is_null);
3322 
3323   // Get cpool & tags index
3324   __ get_cpool_and_tags(rcx, rdx); // rcx=cpool, rdx=tags array
3325   __ get_unsigned_2_byte_index_at_bcp(rbx, 1); // rbx=index
3326   // See if bytecode has already been quicked
3327   __ cmpb(Address(rdx, rbx,
3328                   Address::times_1,
3329                   typeArrayOopDesc::header_size(T_BYTE) * wordSize),
3330           JVM_CONSTANT_Class);
3331   __ jcc(Assembler::equal, quicked);
3332 
3333   __ push(atos); // save receiver for result, and for GC
3334   __ mov(r12, rcx); // save rcx
3335   call_VM(rax, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc));
3336   __ movq(rcx, r12); // restore rcx
3337   __ reinit_heapbase();
3338   __ pop_ptr(rdx); // restore receiver
3339   __ load_klass(rdx, rdx);
3340   __ jmpb(resolved);
3341 
3342   // Get superklass in rax and subklass in rdx
3343   __ bind(quicked);
3344   __ load_klass(rdx, rax);
3345   __ movptr(rax, Address(rcx, rbx,
3346                          Address::times_8, sizeof(constantPoolOopDesc)));
3347 
3348   __ bind(resolved);
3349 
3350   // Generate subtype check.  Blows rcx, rdi
3351   // Superklass in rax.  Subklass in rdx.
3352   __ gen_subtype_check(rdx, ok_is_subtype);
3353 
3354   // Come here on failure
3355   __ xorl(rax, rax);
3356   __ jmpb(done);
3357   // Come here on success
3358   __ bind(ok_is_subtype);
3359   __ movl(rax, 1);
3360 
3361   // Collect counts on whether this test sees NULLs a lot or not.
3362   if (ProfileInterpreter) {
3363     __ jmp(done);
3364     __ bind(is_null);
3365     __ profile_null_seen(rcx);
3366   } else {
3367     __ bind(is_null);   // same as 'done'
3368   }
3369   __ bind(done);
3370   // rax = 0: obj == NULL or  obj is not an instanceof the specified klass
3371   // rax = 1: obj != NULL and obj is     an instanceof the specified klass
3372 }
3373 
3374 //-----------------------------------------------------------------------------
3375 // Breakpoints
3376 void TemplateTable::_breakpoint() {
3377   // Note: We get here even if we are single stepping..
3378   // jbug inists on setting breakpoints at every bytecode
3379   // even if we are in single step mode.
3380 
3381   transition(vtos, vtos);
3382 
3383   // get the unpatched byte code
3384   __ get_method(c_rarg1);
3385   __ call_VM(noreg,
3386              CAST_FROM_FN_PTR(address,
3387                               InterpreterRuntime::get_original_bytecode_at),
3388              c_rarg1, r13);
3389   __ mov(rbx, rax);
3390 
3391   // post the breakpoint event
3392   __ get_method(c_rarg1);
3393   __ call_VM(noreg,
3394              CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint),
3395              c_rarg1, r13);
3396 
3397   // complete the execution of original bytecode
3398   __ dispatch_only_normal(vtos);
3399 }
3400 
3401 //-----------------------------------------------------------------------------
3402 // Exceptions
3403 
3404 void TemplateTable::athrow() {
3405   transition(atos, vtos);
3406   __ null_check(rax);
3407   __ jump(ExternalAddress(Interpreter::throw_exception_entry()));
3408 }
3409 
3410 //-----------------------------------------------------------------------------
3411 // Synchronization
3412 //
3413 // Note: monitorenter & exit are symmetric routines; which is reflected
3414 //       in the assembly code structure as well
3415 //
3416 // Stack layout:
3417 //
3418 // [expressions  ] <--- rsp               = expression stack top
3419 // ..
3420 // [expressions  ]
3421 // [monitor entry] <--- monitor block top = expression stack bot
3422 // ..
3423 // [monitor entry]
3424 // [frame data   ] <--- monitor block bot
3425 // ...
3426 // [saved rbp    ] <--- rbp
3427 void TemplateTable::monitorenter() {
3428   transition(atos, vtos);
3429 
3430   // check for NULL object
3431   __ null_check(rax);
3432 
3433   const Address monitor_block_top(
3434         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3435   const Address monitor_block_bot(
3436         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3437   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3438 
3439   Label allocated;
3440 
3441   // initialize entry pointer
3442   __ xorl(c_rarg1, c_rarg1); // points to free slot or NULL
3443 
3444   // find a free slot in the monitor block (result in c_rarg1)
3445   {
3446     Label entry, loop, exit;
3447     __ movptr(c_rarg3, monitor_block_top); // points to current entry,
3448                                      // starting with top-most entry
3449     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3450                                      // of monitor block
3451     __ jmpb(entry);
3452 
3453     __ bind(loop);
3454     // check if current entry is used
3455     __ cmpptr(Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()), (int32_t) NULL_WORD);
3456     // if not used then remember entry in c_rarg1
3457     __ cmov(Assembler::equal, c_rarg1, c_rarg3);
3458     // check if current entry is for same object
3459     __ cmpptr(rax, Address(c_rarg3, BasicObjectLock::obj_offset_in_bytes()));
3460     // if same object then stop searching
3461     __ jccb(Assembler::equal, exit);
3462     // otherwise advance to next entry
3463     __ addptr(c_rarg3, entry_size);
3464     __ bind(entry);
3465     // check if bottom reached
3466     __ cmpptr(c_rarg3, c_rarg2);
3467     // if not at bottom then check this entry
3468     __ jcc(Assembler::notEqual, loop);
3469     __ bind(exit);
3470   }
3471 
3472   __ testptr(c_rarg1, c_rarg1); // check if a slot has been found
3473   __ jcc(Assembler::notZero, allocated); // if found, continue with that one
3474 
3475   // allocate one if there's no free slot
3476   {
3477     Label entry, loop;
3478     // 1. compute new pointers             // rsp: old expression stack top
3479     __ movptr(c_rarg1, monitor_block_bot); // c_rarg1: old expression stack bottom
3480     __ subptr(rsp, entry_size);            // move expression stack top
3481     __ subptr(c_rarg1, entry_size);        // move expression stack bottom
3482     __ mov(c_rarg3, rsp);                  // set start value for copy loop
3483     __ movptr(monitor_block_bot, c_rarg1); // set new monitor block bottom
3484     __ jmp(entry);
3485     // 2. move expression stack contents
3486     __ bind(loop);
3487     __ movptr(c_rarg2, Address(c_rarg3, entry_size)); // load expression stack
3488                                                       // word from old location
3489     __ movptr(Address(c_rarg3, 0), c_rarg2);          // and store it at new location
3490     __ addptr(c_rarg3, wordSize);                     // advance to next word
3491     __ bind(entry);
3492     __ cmpptr(c_rarg3, c_rarg1);            // check if bottom reached
3493     __ jcc(Assembler::notEqual, loop);      // if not at bottom then
3494                                             // copy next word
3495   }
3496 
3497   // call run-time routine
3498   // c_rarg1: points to monitor entry
3499   __ bind(allocated);
3500 
3501   // Increment bcp to point to the next bytecode, so exception
3502   // handling for async. exceptions work correctly.
3503   // The object has already been poped from the stack, so the
3504   // expression stack looks correct.
3505   __ increment(r13);
3506 
3507   // store object
3508   __ movptr(Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()), rax);
3509   __ lock_object(c_rarg1);
3510 
3511   // check to make sure this monitor doesn't cause stack overflow after locking
3512   __ save_bcp();  // in case of exception
3513   __ generate_stack_overflow_check(0);
3514 
3515   // The bcp has already been incremented. Just need to dispatch to
3516   // next instruction.
3517   __ dispatch_next(vtos);
3518 }
3519 
3520 
3521 void TemplateTable::monitorexit() {
3522   transition(atos, vtos);
3523 
3524   // check for NULL object
3525   __ null_check(rax);
3526 
3527   const Address monitor_block_top(
3528         rbp, frame::interpreter_frame_monitor_block_top_offset * wordSize);
3529   const Address monitor_block_bot(
3530         rbp, frame::interpreter_frame_initial_sp_offset * wordSize);
3531   const int entry_size = frame::interpreter_frame_monitor_size() * wordSize;
3532 
3533   Label found;
3534 
3535   // find matching slot
3536   {
3537     Label entry, loop;
3538     __ movptr(c_rarg1, monitor_block_top); // points to current entry,
3539                                      // starting with top-most entry
3540     __ lea(c_rarg2, monitor_block_bot); // points to word before bottom
3541                                      // of monitor block
3542     __ jmpb(entry);
3543 
3544     __ bind(loop);
3545     // check if current entry is for same object
3546     __ cmpptr(rax, Address(c_rarg1, BasicObjectLock::obj_offset_in_bytes()));
3547     // if same object then stop searching
3548     __ jcc(Assembler::equal, found);
3549     // otherwise advance to next entry
3550     __ addptr(c_rarg1, entry_size);
3551     __ bind(entry);
3552     // check if bottom reached
3553     __ cmpptr(c_rarg1, c_rarg2);
3554     // if not at bottom then check this entry
3555     __ jcc(Assembler::notEqual, loop);
3556   }
3557 
3558   // error handling. Unlocking was not block-structured
3559   __ call_VM(noreg, CAST_FROM_FN_PTR(address,
3560                    InterpreterRuntime::throw_illegal_monitor_state_exception));
3561   __ should_not_reach_here();
3562 
3563   // call run-time routine
3564   // rsi: points to monitor entry
3565   __ bind(found);
3566   __ push_ptr(rax); // make sure object is on stack (contract with oopMaps)
3567   __ unlock_object(c_rarg1);
3568   __ pop_ptr(rax); // discard object
3569 }
3570 
3571 
3572 // Wide instructions
3573 void TemplateTable::wide() {
3574   transition(vtos, vtos);
3575   __ load_unsigned_byte(rbx, at_bcp(1));
3576   __ lea(rscratch1, ExternalAddress((address)Interpreter::_wentry_point));
3577   __ jmp(Address(rscratch1, rbx, Address::times_8));
3578   // Note: the r13 increment step is part of the individual wide
3579   // bytecode implementations
3580 }
3581 
3582 
3583 // Multi arrays
3584 void TemplateTable::multianewarray() {
3585   transition(vtos, atos);
3586   __ load_unsigned_byte(rax, at_bcp(3)); // get number of dimensions
3587   // last dim is on top of stack; we want address of first one:
3588   // first_addr = last_addr + (ndims - 1) * wordSize
3589   if (TaggedStackInterpreter) __ shll(rax, 1);  // index*2
3590   __ lea(c_rarg1, Address(rsp, rax, Address::times_8, -wordSize));
3591   call_VM(rax,
3592           CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray),
3593           c_rarg1);
3594   __ load_unsigned_byte(rbx, at_bcp(3));
3595   if (TaggedStackInterpreter) __ shll(rbx, 1);  // index*2
3596   __ lea(rsp, Address(rsp, rbx, Address::times_8));
3597 }
3598 #endif // !CC_INTERP