1 /*
   2  * Copyright 1997-2008 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_sparc.cpp.incl"
  27 
  28 #ifndef CC_INTERP
  29 #define __ _masm->
  30 
  31 // Misc helpers
  32 
  33 // Do an oop store like *(base + index + offset) = val
  34 // index can be noreg,
  35 static void do_oop_store(InterpreterMacroAssembler* _masm,
  36                          Register base,
  37                          Register index,
  38                          int offset,
  39                          Register val,
  40                          Register tmp,
  41                          BarrierSet::Name barrier,
  42                          bool precise) {
  43   assert(tmp != val && tmp != base && tmp != index, "register collision");
  44   assert(index == noreg || offset == 0, "only one offset");
  45   switch (barrier) {
  46 #ifndef SERIALGC
  47     case BarrierSet::G1SATBCT:
  48     case BarrierSet::G1SATBCTLogging:
  49       {
  50         __ g1_write_barrier_pre( base, index, offset, tmp, /*preserve_o_regs*/true);
  51         if (index == noreg ) {
  52           assert(Assembler::is_simm13(offset), "fix this code");
  53           __ store_heap_oop(val, base, offset);
  54         } else {
  55           __ store_heap_oop(val, base, index);
  56         }
  57 
  58         // No need for post barrier if storing NULL
  59         if (val != G0) {
  60           if (precise) {
  61             if (index == noreg) {
  62               __ add(base, offset, base);
  63             } else {
  64               __ add(base, index, base);
  65             }
  66           }
  67           __ g1_write_barrier_post(base, val, tmp);
  68         }
  69       }
  70       break;
  71 #endif // SERIALGC
  72     case BarrierSet::CardTableModRef:
  73     case BarrierSet::CardTableExtension:
  74       {
  75         if (index == noreg ) {
  76           assert(Assembler::is_simm13(offset), "fix this code");
  77           __ store_heap_oop(val, base, offset);
  78         } else {
  79           __ store_heap_oop(val, base, index);
  80         }
  81         // No need for post barrier if storing NULL
  82         if (val != G0) {
  83           if (precise) {
  84             if (index == noreg) {
  85               __ add(base, offset, base);
  86             } else {
  87               __ add(base, index, base);
  88             }
  89           }
  90           __ card_write_barrier_post(base, val, tmp);
  91         }
  92       }
  93       break;
  94     case BarrierSet::ModRef:
  95     case BarrierSet::Other:
  96       ShouldNotReachHere();
  97       break;
  98     default      :
  99       ShouldNotReachHere();
 100 
 101   }
 102 }
 103 
 104 
 105 //----------------------------------------------------------------------------------------------------
 106 // Platform-dependent initialization
 107 
 108 void TemplateTable::pd_initialize() {
 109   // (none)
 110 }
 111 
 112 
 113 //----------------------------------------------------------------------------------------------------
 114 // Condition conversion
 115 Assembler::Condition ccNot(TemplateTable::Condition cc) {
 116   switch (cc) {
 117     case TemplateTable::equal        : return Assembler::notEqual;
 118     case TemplateTable::not_equal    : return Assembler::equal;
 119     case TemplateTable::less         : return Assembler::greaterEqual;
 120     case TemplateTable::less_equal   : return Assembler::greater;
 121     case TemplateTable::greater      : return Assembler::lessEqual;
 122     case TemplateTable::greater_equal: return Assembler::less;
 123   }
 124   ShouldNotReachHere();
 125   return Assembler::zero;
 126 }
 127 
 128 //----------------------------------------------------------------------------------------------------
 129 // Miscelaneous helper routines
 130 
 131 
 132 Address TemplateTable::at_bcp(int offset) {
 133   assert(_desc->uses_bcp(), "inconsistent uses_bcp information");
 134   return Address( Lbcp, 0, offset);
 135 }
 136 
 137 
 138 void TemplateTable::patch_bytecode(Bytecodes::Code bc, Register Rbyte_code,
 139                                    Register Rscratch,
 140                                    bool load_bc_into_scratch /*=true*/) {
 141   // With sharing on, may need to test methodOop flag.
 142   if (!RewriteBytecodes) return;
 143   if (load_bc_into_scratch) __ set(bc, Rbyte_code);
 144   Label patch_done;
 145   if (JvmtiExport::can_post_breakpoint()) {
 146     Label fast_patch;
 147     __ ldub(at_bcp(0), Rscratch);
 148     __ cmp(Rscratch, Bytecodes::_breakpoint);
 149     __ br(Assembler::notEqual, false, Assembler::pt, fast_patch);
 150     __ delayed()->nop();  // don't bother to hoist the stb here
 151     // perform the quickening, slowly, in the bowels of the breakpoint table
 152     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::set_original_bytecode_at), Lmethod, Lbcp, Rbyte_code);
 153     __ ba(false, patch_done);
 154     __ delayed()->nop();
 155     __ bind(fast_patch);
 156   }
 157 #ifdef ASSERT
 158   Bytecodes::Code orig_bytecode =  Bytecodes::java_code(bc);
 159   Label okay;
 160   __ ldub(at_bcp(0), Rscratch);
 161   __ cmp(Rscratch, orig_bytecode);
 162   __ br(Assembler::equal, false, Assembler::pt, okay);
 163   __ delayed() ->cmp(Rscratch, Rbyte_code);
 164   __ br(Assembler::equal, false, Assembler::pt, okay);
 165   __ delayed()->nop();
 166   __ stop("Rewriting wrong bytecode location");
 167   __ bind(okay);
 168 #endif
 169   __ stb(Rbyte_code, at_bcp(0));
 170   __ bind(patch_done);
 171 }
 172 
 173 //----------------------------------------------------------------------------------------------------
 174 // Individual instructions
 175 
 176 void TemplateTable::nop() {
 177   transition(vtos, vtos);
 178   // nothing to do
 179 }
 180 
 181 void TemplateTable::shouldnotreachhere() {
 182   transition(vtos, vtos);
 183   __ stop("shouldnotreachhere bytecode");
 184 }
 185 
 186 void TemplateTable::aconst_null() {
 187   transition(vtos, atos);
 188   __ clr(Otos_i);
 189 }
 190 
 191 
 192 void TemplateTable::iconst(int value) {
 193   transition(vtos, itos);
 194   __ set(value, Otos_i);
 195 }
 196 
 197 
 198 void TemplateTable::lconst(int value) {
 199   transition(vtos, ltos);
 200   assert(value >= 0, "check this code");
 201 #ifdef _LP64
 202   __ set(value, Otos_l);
 203 #else
 204   __ set(value, Otos_l2);
 205   __ clr( Otos_l1);
 206 #endif
 207 }
 208 
 209 
 210 void TemplateTable::fconst(int value) {
 211   transition(vtos, ftos);
 212   static float zero = 0.0, one = 1.0, two = 2.0;
 213   float* p;
 214   switch( value ) {
 215    default: ShouldNotReachHere();
 216    case 0:  p = &zero;  break;
 217    case 1:  p = &one;   break;
 218    case 2:  p = &two;   break;
 219   }
 220   Address a(G3_scratch, (address)p);
 221   __ sethi(a);
 222   __ ldf(FloatRegisterImpl::S, a, Ftos_f);
 223 }
 224 
 225 
 226 void TemplateTable::dconst(int value) {
 227   transition(vtos, dtos);
 228   static double zero = 0.0, one = 1.0;
 229   double* p;
 230   switch( value ) {
 231    default: ShouldNotReachHere();
 232    case 0:  p = &zero;  break;
 233    case 1:  p = &one;   break;
 234   }
 235   Address a(G3_scratch, (address)p);
 236   __ sethi(a);
 237   __ ldf(FloatRegisterImpl::D, a, Ftos_d);
 238 }
 239 
 240 
 241 // %%%%% Should factore most snippet templates across platforms
 242 
 243 void TemplateTable::bipush() {
 244   transition(vtos, itos);
 245   __ ldsb( at_bcp(1), Otos_i );
 246 }
 247 
 248 void TemplateTable::sipush() {
 249   transition(vtos, itos);
 250   __ get_2_byte_integer_at_bcp(1, G3_scratch, Otos_i, InterpreterMacroAssembler::Signed);
 251 }
 252 
 253 void TemplateTable::ldc(bool wide) {
 254   transition(vtos, vtos);
 255   Label call_ldc, notInt, notString, notClass, exit;
 256 
 257   if (wide) {
 258     __ get_2_byte_integer_at_bcp(1, G3_scratch, O1, InterpreterMacroAssembler::Unsigned);
 259   } else {
 260     __ ldub(Lbcp, 1, O1);
 261   }
 262   __ get_cpool_and_tags(O0, O2);
 263 
 264   const int base_offset = constantPoolOopDesc::header_size() * wordSize;
 265   const int tags_offset = typeArrayOopDesc::header_size(T_BYTE) * wordSize;
 266 
 267   // get type from tags
 268   __ add(O2, tags_offset, O2);
 269   __ ldub(O2, O1, O2);
 270   __ cmp(O2, JVM_CONSTANT_UnresolvedString);    // unresolved string? If so, must resolve
 271   __ brx(Assembler::equal, true, Assembler::pt, call_ldc);
 272   __ delayed()->nop();
 273 
 274   __ cmp(O2, JVM_CONSTANT_UnresolvedClass);     // unresolved class? If so, must resolve
 275   __ brx(Assembler::equal, true, Assembler::pt, call_ldc);
 276   __ delayed()->nop();
 277 
 278   __ cmp(O2, JVM_CONSTANT_UnresolvedClassInError);     // unresolved class in error state
 279   __ brx(Assembler::equal, true, Assembler::pn, call_ldc);
 280   __ delayed()->nop();
 281 
 282   __ cmp(O2, JVM_CONSTANT_Class);      // need to call vm to get java mirror of the class
 283   __ brx(Assembler::notEqual, true, Assembler::pt, notClass);
 284   __ delayed()->add(O0, base_offset, O0);
 285 
 286   __ bind(call_ldc);
 287   __ set(wide, O1);
 288   call_VM(Otos_i, CAST_FROM_FN_PTR(address, InterpreterRuntime::ldc), O1);
 289   __ push(atos);
 290   __ ba(false, exit);
 291   __ delayed()->nop();
 292 
 293   __ bind(notClass);
 294  // __ add(O0, base_offset, O0);
 295   __ sll(O1, LogBytesPerWord, O1);
 296   __ cmp(O2, JVM_CONSTANT_Integer);
 297   __ brx(Assembler::notEqual, true, Assembler::pt, notInt);
 298   __ delayed()->cmp(O2, JVM_CONSTANT_String);
 299   __ ld(O0, O1, Otos_i);
 300   __ push(itos);
 301   __ ba(false, exit);
 302   __ delayed()->nop();
 303 
 304   __ bind(notInt);
 305  // __ cmp(O2, JVM_CONSTANT_String);
 306   __ brx(Assembler::notEqual, true, Assembler::pt, notString);
 307   __ delayed()->ldf(FloatRegisterImpl::S, O0, O1, Ftos_f);
 308   __ ld_ptr(O0, O1, Otos_i);
 309   __ verify_oop(Otos_i);
 310   __ push(atos);
 311   __ ba(false, exit);
 312   __ delayed()->nop();
 313 
 314   __ bind(notString);
 315  // __ ldf(FloatRegisterImpl::S, O0, O1, Ftos_f);
 316   __ push(ftos);
 317 
 318   __ bind(exit);
 319 }
 320 
 321 void TemplateTable::ldc2_w() {
 322   transition(vtos, vtos);
 323   Label retry, resolved, Long, exit;
 324 
 325   __ bind(retry);
 326   __ get_2_byte_integer_at_bcp(1, G3_scratch, O1, InterpreterMacroAssembler::Unsigned);
 327   __ get_cpool_and_tags(O0, O2);
 328 
 329   const int base_offset = constantPoolOopDesc::header_size() * wordSize;
 330   const int tags_offset = typeArrayOopDesc::header_size(T_BYTE) * wordSize;
 331   // get type from tags
 332   __ add(O2, tags_offset, O2);
 333   __ ldub(O2, O1, O2);
 334 
 335   __ sll(O1, LogBytesPerWord, O1);
 336   __ add(O0, O1, G3_scratch);
 337 
 338   __ cmp(O2, JVM_CONSTANT_Double);
 339   __ brx(Assembler::notEqual, false, Assembler::pt, Long);
 340   __ delayed()->nop();
 341   // A double can be placed at word-aligned locations in the constant pool.
 342   // Check out Conversions.java for an example.
 343   // Also constantPoolOopDesc::header_size() is 20, which makes it very difficult
 344   // to double-align double on the constant pool.  SG, 11/7/97
 345 #ifdef _LP64
 346   __ ldf(FloatRegisterImpl::D, G3_scratch, base_offset, Ftos_d);
 347 #else
 348   FloatRegister f = Ftos_d;
 349   __ ldf(FloatRegisterImpl::S, G3_scratch, base_offset, f);
 350   __ ldf(FloatRegisterImpl::S, G3_scratch, base_offset + sizeof(jdouble)/2,
 351          f->successor());
 352 #endif
 353   __ push(dtos);
 354   __ ba(false, exit);
 355   __ delayed()->nop();
 356 
 357   __ bind(Long);
 358 #ifdef _LP64
 359   __ ldx(G3_scratch, base_offset, Otos_l);
 360 #else
 361   __ ld(G3_scratch, base_offset, Otos_l);
 362   __ ld(G3_scratch, base_offset + sizeof(jlong)/2, Otos_l->successor());
 363 #endif
 364   __ push(ltos);
 365 
 366   __ bind(exit);
 367 }
 368 
 369 
 370 void TemplateTable::locals_index(Register reg, int offset) {
 371   __ ldub( at_bcp(offset), reg );
 372 }
 373 
 374 
 375 void TemplateTable::locals_index_wide(Register reg) {
 376   // offset is 2, not 1, because Lbcp points to wide prefix code
 377   __ get_2_byte_integer_at_bcp(2, G4_scratch, reg, InterpreterMacroAssembler::Unsigned);
 378 }
 379 
 380 void TemplateTable::iload() {
 381   transition(vtos, itos);
 382   // Rewrite iload,iload  pair into fast_iload2
 383   //         iload,caload pair into fast_icaload
 384   if (RewriteFrequentPairs) {
 385     Label rewrite, done;
 386 
 387     // get next byte
 388     __ ldub(at_bcp(Bytecodes::length_for(Bytecodes::_iload)), G3_scratch);
 389 
 390     // if _iload, wait to rewrite to iload2.  We only want to rewrite the
 391     // last two iloads in a pair.  Comparing against fast_iload means that
 392     // the next bytecode is neither an iload or a caload, and therefore
 393     // an iload pair.
 394     __ cmp(G3_scratch, (int)Bytecodes::_iload);
 395     __ br(Assembler::equal, false, Assembler::pn, done);
 396     __ delayed()->nop();
 397 
 398     __ cmp(G3_scratch, (int)Bytecodes::_fast_iload);
 399     __ br(Assembler::equal, false, Assembler::pn, rewrite);
 400     __ delayed()->set(Bytecodes::_fast_iload2, G4_scratch);
 401 
 402     __ cmp(G3_scratch, (int)Bytecodes::_caload);
 403     __ br(Assembler::equal, false, Assembler::pn, rewrite);
 404     __ delayed()->set(Bytecodes::_fast_icaload, G4_scratch);
 405 
 406     __ set(Bytecodes::_fast_iload, G4_scratch);  // don't check again
 407     // rewrite
 408     // G4_scratch: fast bytecode
 409     __ bind(rewrite);
 410     patch_bytecode(Bytecodes::_iload, G4_scratch, G3_scratch, false);
 411     __ bind(done);
 412   }
 413 
 414   // Get the local value into tos
 415   locals_index(G3_scratch);
 416   __ access_local_int( G3_scratch, Otos_i );
 417 }
 418 
 419 void TemplateTable::fast_iload2() {
 420   transition(vtos, itos);
 421   locals_index(G3_scratch);
 422   __ access_local_int( G3_scratch, Otos_i );
 423   __ push_i();
 424   locals_index(G3_scratch, 3);  // get next bytecode's local index.
 425   __ access_local_int( G3_scratch, Otos_i );
 426 }
 427 
 428 void TemplateTable::fast_iload() {
 429   transition(vtos, itos);
 430   locals_index(G3_scratch);
 431   __ access_local_int( G3_scratch, Otos_i );
 432 }
 433 
 434 void TemplateTable::lload() {
 435   transition(vtos, ltos);
 436   locals_index(G3_scratch);
 437   __ access_local_long( G3_scratch, Otos_l );
 438 }
 439 
 440 
 441 void TemplateTable::fload() {
 442   transition(vtos, ftos);
 443   locals_index(G3_scratch);
 444   __ access_local_float( G3_scratch, Ftos_f );
 445 }
 446 
 447 
 448 void TemplateTable::dload() {
 449   transition(vtos, dtos);
 450   locals_index(G3_scratch);
 451   __ access_local_double( G3_scratch, Ftos_d );
 452 }
 453 
 454 
 455 void TemplateTable::aload() {
 456   transition(vtos, atos);
 457   locals_index(G3_scratch);
 458   __ access_local_ptr( G3_scratch, Otos_i);
 459 }
 460 
 461 
 462 void TemplateTable::wide_iload() {
 463   transition(vtos, itos);
 464   locals_index_wide(G3_scratch);
 465   __ access_local_int( G3_scratch, Otos_i );
 466 }
 467 
 468 
 469 void TemplateTable::wide_lload() {
 470   transition(vtos, ltos);
 471   locals_index_wide(G3_scratch);
 472   __ access_local_long( G3_scratch, Otos_l );
 473 }
 474 
 475 
 476 void TemplateTable::wide_fload() {
 477   transition(vtos, ftos);
 478   locals_index_wide(G3_scratch);
 479   __ access_local_float( G3_scratch, Ftos_f );
 480 }
 481 
 482 
 483 void TemplateTable::wide_dload() {
 484   transition(vtos, dtos);
 485   locals_index_wide(G3_scratch);
 486   __ access_local_double( G3_scratch, Ftos_d );
 487 }
 488 
 489 
 490 void TemplateTable::wide_aload() {
 491   transition(vtos, atos);
 492   locals_index_wide(G3_scratch);
 493   __ access_local_ptr( G3_scratch, Otos_i );
 494   __ verify_oop(Otos_i);
 495 }
 496 
 497 
 498 void TemplateTable::iaload() {
 499   transition(itos, itos);
 500   // Otos_i: index
 501   // tos: array
 502   __ index_check(O2, Otos_i, LogBytesPerInt, G3_scratch, O3);
 503   __ ld(O3, arrayOopDesc::base_offset_in_bytes(T_INT), Otos_i);
 504 }
 505 
 506 
 507 void TemplateTable::laload() {
 508   transition(itos, ltos);
 509   // Otos_i: index
 510   // O2: array
 511   __ index_check(O2, Otos_i, LogBytesPerLong, G3_scratch, O3);
 512   __ ld_long(O3, arrayOopDesc::base_offset_in_bytes(T_LONG), Otos_l);
 513 }
 514 
 515 
 516 void TemplateTable::faload() {
 517   transition(itos, ftos);
 518   // Otos_i: index
 519   // O2: array
 520   __ index_check(O2, Otos_i, LogBytesPerInt, G3_scratch, O3);
 521   __ ldf(FloatRegisterImpl::S, O3, arrayOopDesc::base_offset_in_bytes(T_FLOAT), Ftos_f);
 522 }
 523 
 524 
 525 void TemplateTable::daload() {
 526   transition(itos, dtos);
 527   // Otos_i: index
 528   // O2: array
 529   __ index_check(O2, Otos_i, LogBytesPerLong, G3_scratch, O3);
 530   __ ldf(FloatRegisterImpl::D, O3, arrayOopDesc::base_offset_in_bytes(T_DOUBLE), Ftos_d);
 531 }
 532 
 533 
 534 void TemplateTable::aaload() {
 535   transition(itos, atos);
 536   // Otos_i: index
 537   // tos: array
 538   __ index_check(O2, Otos_i, UseCompressedOops ? 2 : LogBytesPerWord, G3_scratch, O3);
 539   __ load_heap_oop(O3, arrayOopDesc::base_offset_in_bytes(T_OBJECT), Otos_i);
 540   __ verify_oop(Otos_i);
 541 }
 542 
 543 
 544 void TemplateTable::baload() {
 545   transition(itos, itos);
 546   // Otos_i: index
 547   // tos: array
 548   __ index_check(O2, Otos_i, 0, G3_scratch, O3);
 549   __ ldsb(O3, arrayOopDesc::base_offset_in_bytes(T_BYTE), Otos_i);
 550 }
 551 
 552 
 553 void TemplateTable::caload() {
 554   transition(itos, itos);
 555   // Otos_i: index
 556   // tos: array
 557   __ index_check(O2, Otos_i, LogBytesPerShort, G3_scratch, O3);
 558   __ lduh(O3, arrayOopDesc::base_offset_in_bytes(T_CHAR), Otos_i);
 559 }
 560 
 561 void TemplateTable::fast_icaload() {
 562   transition(vtos, itos);
 563   // Otos_i: index
 564   // tos: array
 565   locals_index(G3_scratch);
 566   __ access_local_int( G3_scratch, Otos_i );
 567   __ index_check(O2, Otos_i, LogBytesPerShort, G3_scratch, O3);
 568   __ lduh(O3, arrayOopDesc::base_offset_in_bytes(T_CHAR), Otos_i);
 569 }
 570 
 571 
 572 void TemplateTable::saload() {
 573   transition(itos, itos);
 574   // Otos_i: index
 575   // tos: array
 576   __ index_check(O2, Otos_i, LogBytesPerShort, G3_scratch, O3);
 577   __ ldsh(O3, arrayOopDesc::base_offset_in_bytes(T_SHORT), Otos_i);
 578 }
 579 
 580 
 581 void TemplateTable::iload(int n) {
 582   transition(vtos, itos);
 583   debug_only(__ verify_local_tag(frame::TagValue, Llocals, Otos_i, n));
 584   __ ld( Llocals, Interpreter::local_offset_in_bytes(n), Otos_i );
 585 }
 586 
 587 
 588 void TemplateTable::lload(int n) {
 589   transition(vtos, ltos);
 590   assert(n+1 < Argument::n_register_parameters, "would need more code");
 591   debug_only(__ verify_local_tag(frame::TagCategory2, Llocals, Otos_l, n));
 592   __ load_unaligned_long(Llocals, Interpreter::local_offset_in_bytes(n+1), Otos_l);
 593 }
 594 
 595 
 596 void TemplateTable::fload(int n) {
 597   transition(vtos, ftos);
 598   assert(n < Argument::n_register_parameters, "would need more code");
 599   debug_only(__ verify_local_tag(frame::TagValue, Llocals, G3_scratch, n));
 600   __ ldf( FloatRegisterImpl::S, Llocals, Interpreter::local_offset_in_bytes(n),     Ftos_f );
 601 }
 602 
 603 
 604 void TemplateTable::dload(int n) {
 605   transition(vtos, dtos);
 606   FloatRegister dst = Ftos_d;
 607   debug_only(__ verify_local_tag(frame::TagCategory2, Llocals, G3_scratch, n));
 608   __ load_unaligned_double(Llocals, Interpreter::local_offset_in_bytes(n+1), dst);
 609 }
 610 
 611 
 612 void TemplateTable::aload(int n) {
 613   transition(vtos, atos);
 614   debug_only(__ verify_local_tag(frame::TagReference, Llocals, Otos_i, n));
 615   __ ld_ptr( Llocals, Interpreter::local_offset_in_bytes(n), Otos_i );
 616 }
 617 
 618 
 619 void TemplateTable::aload_0() {
 620   transition(vtos, atos);
 621 
 622   // According to bytecode histograms, the pairs:
 623   //
 624   // _aload_0, _fast_igetfield (itos)
 625   // _aload_0, _fast_agetfield (atos)
 626   // _aload_0, _fast_fgetfield (ftos)
 627   //
 628   // occur frequently. If RewriteFrequentPairs is set, the (slow) _aload_0
 629   // bytecode checks the next bytecode and then rewrites the current
 630   // bytecode into a pair bytecode; otherwise it rewrites the current
 631   // bytecode into _fast_aload_0 that doesn't do the pair check anymore.
 632   //
 633   if (RewriteFrequentPairs) {
 634     Label rewrite, done;
 635 
 636     // get next byte
 637     __ ldub(at_bcp(Bytecodes::length_for(Bytecodes::_aload_0)), G3_scratch);
 638 
 639     // do actual aload_0
 640     aload(0);
 641 
 642     // if _getfield then wait with rewrite
 643     __ cmp(G3_scratch, (int)Bytecodes::_getfield);
 644     __ br(Assembler::equal, false, Assembler::pn, done);
 645     __ delayed()->nop();
 646 
 647     // if _igetfield then rewrite to _fast_iaccess_0
 648     assert(Bytecodes::java_code(Bytecodes::_fast_iaccess_0) == Bytecodes::_aload_0, "adjust fast bytecode def");
 649     __ cmp(G3_scratch, (int)Bytecodes::_fast_igetfield);
 650     __ br(Assembler::equal, false, Assembler::pn, rewrite);
 651     __ delayed()->set(Bytecodes::_fast_iaccess_0, G4_scratch);
 652 
 653     // if _agetfield then rewrite to _fast_aaccess_0
 654     assert(Bytecodes::java_code(Bytecodes::_fast_aaccess_0) == Bytecodes::_aload_0, "adjust fast bytecode def");
 655     __ cmp(G3_scratch, (int)Bytecodes::_fast_agetfield);
 656     __ br(Assembler::equal, false, Assembler::pn, rewrite);
 657     __ delayed()->set(Bytecodes::_fast_aaccess_0, G4_scratch);
 658 
 659     // if _fgetfield then rewrite to _fast_faccess_0
 660     assert(Bytecodes::java_code(Bytecodes::_fast_faccess_0) == Bytecodes::_aload_0, "adjust fast bytecode def");
 661     __ cmp(G3_scratch, (int)Bytecodes::_fast_fgetfield);
 662     __ br(Assembler::equal, false, Assembler::pn, rewrite);
 663     __ delayed()->set(Bytecodes::_fast_faccess_0, G4_scratch);
 664 
 665     // else rewrite to _fast_aload0
 666     assert(Bytecodes::java_code(Bytecodes::_fast_aload_0) == Bytecodes::_aload_0, "adjust fast bytecode def");
 667     __ set(Bytecodes::_fast_aload_0, G4_scratch);
 668 
 669     // rewrite
 670     // G4_scratch: fast bytecode
 671     __ bind(rewrite);
 672     patch_bytecode(Bytecodes::_aload_0, G4_scratch, G3_scratch, false);
 673     __ bind(done);
 674   } else {
 675     aload(0);
 676   }
 677 }
 678 
 679 
 680 void TemplateTable::istore() {
 681   transition(itos, vtos);
 682   locals_index(G3_scratch);
 683   __ store_local_int( G3_scratch, Otos_i );
 684 }
 685 
 686 
 687 void TemplateTable::lstore() {
 688   transition(ltos, vtos);
 689   locals_index(G3_scratch);
 690   __ store_local_long( G3_scratch, Otos_l );
 691 }
 692 
 693 
 694 void TemplateTable::fstore() {
 695   transition(ftos, vtos);
 696   locals_index(G3_scratch);
 697   __ store_local_float( G3_scratch, Ftos_f );
 698 }
 699 
 700 
 701 void TemplateTable::dstore() {
 702   transition(dtos, vtos);
 703   locals_index(G3_scratch);
 704   __ store_local_double( G3_scratch, Ftos_d );
 705 }
 706 
 707 
 708 void TemplateTable::astore() {
 709   transition(vtos, vtos);
 710   // astore tos can also be a returnAddress, so load and store the tag too
 711   __ load_ptr_and_tag(0, Otos_i, Otos_l2);
 712   __ inc(Lesp, Interpreter::stackElementSize());
 713   __ verify_oop_or_return_address(Otos_i, G3_scratch);
 714   locals_index(G3_scratch);
 715   __ store_local_ptr( G3_scratch, Otos_i, Otos_l2 );
 716 }
 717 
 718 
 719 void TemplateTable::wide_istore() {
 720   transition(vtos, vtos);
 721   __ pop_i();
 722   locals_index_wide(G3_scratch);
 723   __ store_local_int( G3_scratch, Otos_i );
 724 }
 725 
 726 
 727 void TemplateTable::wide_lstore() {
 728   transition(vtos, vtos);
 729   __ pop_l();
 730   locals_index_wide(G3_scratch);
 731   __ store_local_long( G3_scratch, Otos_l );
 732 }
 733 
 734 
 735 void TemplateTable::wide_fstore() {
 736   transition(vtos, vtos);
 737   __ pop_f();
 738   locals_index_wide(G3_scratch);
 739   __ store_local_float( G3_scratch, Ftos_f );
 740 }
 741 
 742 
 743 void TemplateTable::wide_dstore() {
 744   transition(vtos, vtos);
 745   __ pop_d();
 746   locals_index_wide(G3_scratch);
 747   __ store_local_double( G3_scratch, Ftos_d );
 748 }
 749 
 750 
 751 void TemplateTable::wide_astore() {
 752   transition(vtos, vtos);
 753   // astore tos can also be a returnAddress, so load and store the tag too
 754   __ load_ptr_and_tag(0, Otos_i, Otos_l2);
 755   __ inc(Lesp, Interpreter::stackElementSize());
 756   __ verify_oop_or_return_address(Otos_i, G3_scratch);
 757   locals_index_wide(G3_scratch);
 758   __ store_local_ptr( G3_scratch, Otos_i, Otos_l2 );
 759 }
 760 
 761 
 762 void TemplateTable::iastore() {
 763   transition(itos, vtos);
 764   __ pop_i(O2); // index
 765   // Otos_i: val
 766   // O3: array
 767   __ index_check(O3, O2, LogBytesPerInt, G3_scratch, O2);
 768   __ st(Otos_i, O2, arrayOopDesc::base_offset_in_bytes(T_INT));
 769 }
 770 
 771 
 772 void TemplateTable::lastore() {
 773   transition(ltos, vtos);
 774   __ pop_i(O2); // index
 775   // Otos_l: val
 776   // O3: array
 777   __ index_check(O3, O2, LogBytesPerLong, G3_scratch, O2);
 778   __ st_long(Otos_l, O2, arrayOopDesc::base_offset_in_bytes(T_LONG));
 779 }
 780 
 781 
 782 void TemplateTable::fastore() {
 783   transition(ftos, vtos);
 784   __ pop_i(O2); // index
 785   // Ftos_f: val
 786   // O3: array
 787   __ index_check(O3, O2, LogBytesPerInt, G3_scratch, O2);
 788   __ stf(FloatRegisterImpl::S, Ftos_f, O2, arrayOopDesc::base_offset_in_bytes(T_FLOAT));
 789 }
 790 
 791 
 792 void TemplateTable::dastore() {
 793   transition(dtos, vtos);
 794   __ pop_i(O2); // index
 795   // Fos_d: val
 796   // O3: array
 797   __ index_check(O3, O2, LogBytesPerLong, G3_scratch, O2);
 798   __ stf(FloatRegisterImpl::D, Ftos_d, O2, arrayOopDesc::base_offset_in_bytes(T_DOUBLE));
 799 }
 800 
 801 
 802 void TemplateTable::aastore() {
 803   Label store_ok, is_null, done;
 804   transition(vtos, vtos);
 805   __ ld_ptr(Lesp, Interpreter::expr_offset_in_bytes(0), Otos_i);
 806   __ ld(Lesp, Interpreter::expr_offset_in_bytes(1), O2);         // get index
 807   __ ld_ptr(Lesp, Interpreter::expr_offset_in_bytes(2), O3);     // get array
 808   // Otos_i: val
 809   // O2: index
 810   // O3: array
 811   __ verify_oop(Otos_i);
 812   __ index_check_without_pop(O3, O2, UseCompressedOops ? 2 : LogBytesPerWord, G3_scratch, O1);
 813 
 814   // do array store check - check for NULL value first
 815   __ br_null( Otos_i, false, Assembler::pn, is_null );
 816   __ delayed()->nop();
 817 
 818   __ load_klass(O3, O4); // get array klass
 819   __ load_klass(Otos_i, O5); // get value klass
 820 
 821   // do fast instanceof cache test
 822 
 823   __ ld_ptr(O4,     sizeof(oopDesc) + objArrayKlass::element_klass_offset_in_bytes(),  O4);
 824 
 825   assert(Otos_i == O0, "just checking");
 826 
 827   // Otos_i:    value
 828   // O1:        addr - offset
 829   // O2:        index
 830   // O3:        array
 831   // O4:        array element klass
 832   // O5:        value klass
 833 
 834   // Address element(O1, 0, arrayOopDesc::base_offset_in_bytes(T_OBJECT));
 835 
 836   // Generate a fast subtype check.  Branch to store_ok if no
 837   // failure.  Throw if failure.
 838   __ gen_subtype_check( O5, O4, G3_scratch, G4_scratch, G1_scratch, store_ok );
 839 
 840   // Not a subtype; so must throw exception
 841   __ throw_if_not_x( Assembler::never, Interpreter::_throw_ArrayStoreException_entry, G3_scratch );
 842 
 843   // Store is OK.
 844   __ bind(store_ok);
 845   do_oop_store(_masm, O1, noreg, arrayOopDesc::base_offset_in_bytes(T_OBJECT), Otos_i, G3_scratch, _bs->kind(), true);
 846 
 847   __ ba(false,done);
 848   __ delayed()->inc(Lesp, 3* Interpreter::stackElementSize()); // adj sp (pops array, index and value)
 849 
 850   __ bind(is_null);
 851   do_oop_store(_masm, O1, noreg, arrayOopDesc::base_offset_in_bytes(T_OBJECT), G0, G4_scratch, _bs->kind(), true);
 852 
 853   __ profile_null_seen(G3_scratch);
 854   __ inc(Lesp, 3* Interpreter::stackElementSize());     // adj sp (pops array, index and value)
 855   __ bind(done);
 856 }
 857 
 858 
 859 void TemplateTable::bastore() {
 860   transition(itos, vtos);
 861   __ pop_i(O2); // index
 862   // Otos_i: val
 863   // O3: array
 864   __ index_check(O3, O2, 0, G3_scratch, O2);
 865   __ stb(Otos_i, O2, arrayOopDesc::base_offset_in_bytes(T_BYTE));
 866 }
 867 
 868 
 869 void TemplateTable::castore() {
 870   transition(itos, vtos);
 871   __ pop_i(O2); // index
 872   // Otos_i: val
 873   // O3: array
 874   __ index_check(O3, O2, LogBytesPerShort, G3_scratch, O2);
 875   __ sth(Otos_i, O2, arrayOopDesc::base_offset_in_bytes(T_CHAR));
 876 }
 877 
 878 
 879 void TemplateTable::sastore() {
 880   // %%%%% Factor across platform
 881   castore();
 882 }
 883 
 884 
 885 void TemplateTable::istore(int n) {
 886   transition(itos, vtos);
 887   __ tag_local(frame::TagValue, Llocals, Otos_i, n);
 888   __ st(Otos_i, Llocals, Interpreter::local_offset_in_bytes(n));
 889 }
 890 
 891 
 892 void TemplateTable::lstore(int n) {
 893   transition(ltos, vtos);
 894   assert(n+1 < Argument::n_register_parameters, "only handle register cases");
 895   __ tag_local(frame::TagCategory2, Llocals, Otos_l, n);
 896   __ store_unaligned_long(Otos_l, Llocals, Interpreter::local_offset_in_bytes(n+1));
 897 
 898 }
 899 
 900 
 901 void TemplateTable::fstore(int n) {
 902   transition(ftos, vtos);
 903   assert(n < Argument::n_register_parameters, "only handle register cases");
 904   __ tag_local(frame::TagValue, Llocals, Otos_l, n);
 905   __ stf(FloatRegisterImpl::S, Ftos_f, Llocals, Interpreter::local_offset_in_bytes(n));
 906 }
 907 
 908 
 909 void TemplateTable::dstore(int n) {
 910   transition(dtos, vtos);
 911   FloatRegister src = Ftos_d;
 912   __ tag_local(frame::TagCategory2, Llocals, Otos_l, n);
 913   __ store_unaligned_double(src, Llocals, Interpreter::local_offset_in_bytes(n+1));
 914 }
 915 
 916 
 917 void TemplateTable::astore(int n) {
 918   transition(vtos, vtos);
 919   // astore tos can also be a returnAddress, so load and store the tag too
 920   __ load_ptr_and_tag(0, Otos_i, Otos_l2);
 921   __ inc(Lesp, Interpreter::stackElementSize());
 922   __ verify_oop_or_return_address(Otos_i, G3_scratch);
 923   __ store_local_ptr( n, Otos_i, Otos_l2 );
 924 }
 925 
 926 
 927 void TemplateTable::pop() {
 928   transition(vtos, vtos);
 929   __ inc(Lesp, Interpreter::stackElementSize());
 930 }
 931 
 932 
 933 void TemplateTable::pop2() {
 934   transition(vtos, vtos);
 935   __ inc(Lesp, 2 * Interpreter::stackElementSize());
 936 }
 937 
 938 
 939 void TemplateTable::dup() {
 940   transition(vtos, vtos);
 941   // stack: ..., a
 942   // load a and tag
 943   __ load_ptr_and_tag(0, Otos_i, Otos_l2);
 944   __ push_ptr(Otos_i, Otos_l2);
 945   // stack: ..., a, a
 946 }
 947 
 948 
 949 void TemplateTable::dup_x1() {
 950   transition(vtos, vtos);
 951   // stack: ..., a, b
 952   __ load_ptr_and_tag(1, G3_scratch, G4_scratch);   // get a
 953   __ load_ptr_and_tag(0, Otos_l1, Otos_l2);         // get b
 954   __ store_ptr_and_tag(1, Otos_l1, Otos_l2);        // put b
 955   __ store_ptr_and_tag(0, G3_scratch, G4_scratch);  // put a - like swap
 956   __ push_ptr(Otos_l1, Otos_l2);                    // push b
 957   // stack: ..., b, a, b
 958 }
 959 
 960 
 961 void TemplateTable::dup_x2() {
 962   transition(vtos, vtos);
 963   // stack: ..., a, b, c
 964   // get c and push on stack, reuse registers
 965   __ load_ptr_and_tag(0, G3_scratch, G4_scratch);     // get c
 966   __ push_ptr(G3_scratch, G4_scratch);               // push c with tag
 967   // stack: ..., a, b, c, c  (c in reg)  (Lesp - 4)
 968   // (stack offsets n+1 now)
 969   __ load_ptr_and_tag(3, Otos_l1, Otos_l2);          // get a
 970   __ store_ptr_and_tag(3, G3_scratch, G4_scratch);   // put c at 3
 971   // stack: ..., c, b, c, c  (a in reg)
 972   __ load_ptr_and_tag(2, G3_scratch, G4_scratch);    // get b
 973   __ store_ptr_and_tag(2, Otos_l1, Otos_l2);         // put a at 2
 974   // stack: ..., c, a, c, c  (b in reg)
 975   __ store_ptr_and_tag(1, G3_scratch, G4_scratch);   // put b at 1
 976   // stack: ..., c, a, b, c
 977 }
 978 
 979 
 980 void TemplateTable::dup2() {
 981   transition(vtos, vtos);
 982   __ load_ptr_and_tag(1, G3_scratch, G4_scratch);     // get a
 983   __ load_ptr_and_tag(0, Otos_l1, Otos_l2);           // get b
 984   __ push_ptr(G3_scratch, G4_scratch);                // push a
 985   __ push_ptr(Otos_l1, Otos_l2);                      // push b
 986   // stack: ..., a, b, a, b
 987 }
 988 
 989 
 990 void TemplateTable::dup2_x1() {
 991   transition(vtos, vtos);
 992   // stack: ..., a, b, c
 993   __ load_ptr_and_tag(1, Lscratch, G1_scratch);       // get b
 994   __ load_ptr_and_tag(2, Otos_l1, Otos_l2);           // get a
 995   __ store_ptr_and_tag(2, Lscratch, G1_scratch);      // put b at a
 996   // stack: ..., b, b, c
 997   __ load_ptr_and_tag(0, G3_scratch, G4_scratch);     // get c
 998   __ store_ptr_and_tag(1, G3_scratch, G4_scratch);    // put c at b
 999   // stack: ..., b, c, c
1000   __ store_ptr_and_tag(0, Otos_l1, Otos_l2);          // put a at c
1001   // stack: ..., b, c, a
1002   __ push_ptr(Lscratch, G1_scratch);                  // push b
1003   __ push_ptr(G3_scratch, G4_scratch);                // push c
1004   // stack: ..., b, c, a, b, c
1005 }
1006 
1007 
1008 // The spec says that these types can be a mixture of category 1 (1 word)
1009 // types and/or category 2 types (long and doubles)
1010 void TemplateTable::dup2_x2() {
1011   transition(vtos, vtos);
1012   // stack: ..., a, b, c, d
1013   __ load_ptr_and_tag(1, Lscratch, G1_scratch);       // get c
1014   __ load_ptr_and_tag(3, Otos_l1, Otos_l2);           // get a
1015   __ store_ptr_and_tag(3, Lscratch, G1_scratch);      // put c at 3
1016   __ store_ptr_and_tag(1, Otos_l1, Otos_l2);          // put a at 1
1017   // stack: ..., c, b, a, d
1018   __ load_ptr_and_tag(2, G3_scratch, G4_scratch);     // get b
1019   __ load_ptr_and_tag(0, Otos_l1, Otos_l2);           // get d
1020   __ store_ptr_and_tag(0, G3_scratch, G4_scratch);    // put b at 0
1021   __ store_ptr_and_tag(2, Otos_l1, Otos_l2);          // put d at 2
1022   // stack: ..., c, d, a, b
1023   __ push_ptr(Lscratch, G1_scratch);                  // push c
1024   __ push_ptr(Otos_l1, Otos_l2);                      // push d
1025   // stack: ..., c, d, a, b, c, d
1026 }
1027 
1028 
1029 void TemplateTable::swap() {
1030   transition(vtos, vtos);
1031   // stack: ..., a, b
1032   __ load_ptr_and_tag(1, G3_scratch, G4_scratch);     // get a
1033   __ load_ptr_and_tag(0, Otos_l1, Otos_l2);           // get b
1034   __ store_ptr_and_tag(0, G3_scratch, G4_scratch);    // put b
1035   __ store_ptr_and_tag(1, Otos_l1, Otos_l2);          // put a
1036   // stack: ..., b, a
1037 }
1038 
1039 
1040 void TemplateTable::iop2(Operation op) {
1041   transition(itos, itos);
1042   __ pop_i(O1);
1043   switch (op) {
1044    case  add:  __  add(O1, Otos_i, Otos_i);  break;
1045    case  sub:  __  sub(O1, Otos_i, Otos_i);  break;
1046      // %%%%% Mul may not exist: better to call .mul?
1047    case  mul:  __ smul(O1, Otos_i, Otos_i);  break;
1048    case _and:  __  and3(O1, Otos_i, Otos_i);  break;
1049    case  _or:  __   or3(O1, Otos_i, Otos_i);  break;
1050    case _xor:  __  xor3(O1, Otos_i, Otos_i);  break;
1051    case  shl:  __  sll(O1, Otos_i, Otos_i);  break;
1052    case  shr:  __  sra(O1, Otos_i, Otos_i);  break;
1053    case ushr:  __  srl(O1, Otos_i, Otos_i);  break;
1054    default: ShouldNotReachHere();
1055   }
1056 }
1057 
1058 
1059 void TemplateTable::lop2(Operation op) {
1060   transition(ltos, ltos);
1061   __ pop_l(O2);
1062   switch (op) {
1063 #ifdef _LP64
1064    case  add:  __ add(O2, Otos_l, Otos_l);  break;
1065    case  sub:  __ sub(O2, Otos_l, Otos_l);  break;
1066    case _and:  __ and3( O2, Otos_l, Otos_l);  break;
1067    case  _or:  __  or3( O2, Otos_l, Otos_l);  break;
1068    case _xor:  __ xor3( O2, Otos_l, Otos_l);  break;
1069 #else
1070    case  add:  __ addcc(O3, Otos_l2, Otos_l2);  __ addc(O2, Otos_l1, Otos_l1);  break;
1071    case  sub:  __ subcc(O3, Otos_l2, Otos_l2);  __ subc(O2, Otos_l1, Otos_l1);  break;
1072    case _and:  __ and3(  O3, Otos_l2, Otos_l2);  __ and3( O2, Otos_l1, Otos_l1);  break;
1073    case  _or:  __  or3(  O3, Otos_l2, Otos_l2);  __  or3( O2, Otos_l1, Otos_l1);  break;
1074    case _xor:  __ xor3(  O3, Otos_l2, Otos_l2);  __ xor3( O2, Otos_l1, Otos_l1);  break;
1075 #endif
1076    default: ShouldNotReachHere();
1077   }
1078 }
1079 
1080 
1081 void TemplateTable::idiv() {
1082   // %%%%% Later: ForSPARC/V7 call .sdiv library routine,
1083   // %%%%% Use ldsw...sdivx on pure V9 ABI. 64 bit safe.
1084 
1085   transition(itos, itos);
1086   __ pop_i(O1); // get 1st op
1087 
1088   // Y contains upper 32 bits of result, set it to 0 or all ones
1089   __ wry(G0);
1090   __ mov(~0, G3_scratch);
1091 
1092   __ tst(O1);
1093      Label neg;
1094   __ br(Assembler::negative, true, Assembler::pn, neg);
1095   __ delayed()->wry(G3_scratch);
1096   __ bind(neg);
1097 
1098      Label ok;
1099   __ tst(Otos_i);
1100   __ throw_if_not_icc( Assembler::notZero, Interpreter::_throw_ArithmeticException_entry, G3_scratch );
1101 
1102   const int min_int = 0x80000000;
1103   Label regular;
1104   __ cmp(Otos_i, -1);
1105   __ br(Assembler::notEqual, false, Assembler::pt, regular);
1106 #ifdef _LP64
1107   // Don't put set in delay slot
1108   // Set will turn into multiple instructions in 64 bit mode
1109   __ delayed()->nop();
1110   __ set(min_int, G4_scratch);
1111 #else
1112   __ delayed()->set(min_int, G4_scratch);
1113 #endif
1114   Label done;
1115   __ cmp(O1, G4_scratch);
1116   __ br(Assembler::equal, true, Assembler::pt, done);
1117   __ delayed()->mov(O1, Otos_i);   // (mov only executed if branch taken)
1118 
1119   __ bind(regular);
1120   __ sdiv(O1, Otos_i, Otos_i); // note: irem uses O1 after this instruction!
1121   __ bind(done);
1122 }
1123 
1124 
1125 void TemplateTable::irem() {
1126   transition(itos, itos);
1127   __ mov(Otos_i, O2); // save divisor
1128   idiv();                               // %%%% Hack: exploits fact that idiv leaves dividend in O1
1129   __ smul(Otos_i, O2, Otos_i);
1130   __ sub(O1, Otos_i, Otos_i);
1131 }
1132 
1133 
1134 void TemplateTable::lmul() {
1135   transition(ltos, ltos);
1136   __ pop_l(O2);
1137 #ifdef _LP64
1138   __ mulx(Otos_l, O2, Otos_l);
1139 #else
1140   __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::lmul));
1141 #endif
1142 
1143 }
1144 
1145 
1146 void TemplateTable::ldiv() {
1147   transition(ltos, ltos);
1148 
1149   // check for zero
1150   __ pop_l(O2);
1151 #ifdef _LP64
1152   __ tst(Otos_l);
1153   __ throw_if_not_xcc( Assembler::notZero, Interpreter::_throw_ArithmeticException_entry, G3_scratch);
1154   __ sdivx(O2, Otos_l, Otos_l);
1155 #else
1156   __ orcc(Otos_l1, Otos_l2, G0);
1157   __ throw_if_not_icc( Assembler::notZero, Interpreter::_throw_ArithmeticException_entry, G3_scratch);
1158   __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::ldiv));
1159 #endif
1160 }
1161 
1162 
1163 void TemplateTable::lrem() {
1164   transition(ltos, ltos);
1165 
1166   // check for zero
1167   __ pop_l(O2);
1168 #ifdef _LP64
1169   __ tst(Otos_l);
1170   __ throw_if_not_xcc( Assembler::notZero, Interpreter::_throw_ArithmeticException_entry, G3_scratch);
1171   __ sdivx(O2, Otos_l, Otos_l2);
1172   __ mulx (Otos_l2, Otos_l, Otos_l2);
1173   __ sub  (O2, Otos_l2, Otos_l);
1174 #else
1175   __ orcc(Otos_l1, Otos_l2, G0);
1176   __ throw_if_not_icc(Assembler::notZero, Interpreter::_throw_ArithmeticException_entry, G3_scratch);
1177   __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::lrem));
1178 #endif
1179 }
1180 
1181 
1182 void TemplateTable::lshl() {
1183   transition(itos, ltos); // %%%% could optimize, fill delay slot or opt for ultra
1184 
1185   __ pop_l(O2);                          // shift value in O2, O3
1186 #ifdef _LP64
1187   __ sllx(O2, Otos_i, Otos_l);
1188 #else
1189   __ lshl(O2, O3, Otos_i, Otos_l1, Otos_l2, O4);
1190 #endif
1191 }
1192 
1193 
1194 void TemplateTable::lshr() {
1195   transition(itos, ltos); // %%%% see lshl comment
1196 
1197   __ pop_l(O2);                          // shift value in O2, O3
1198 #ifdef _LP64
1199   __ srax(O2, Otos_i, Otos_l);
1200 #else
1201   __ lshr(O2, O3, Otos_i, Otos_l1, Otos_l2, O4);
1202 #endif
1203 }
1204 
1205 
1206 
1207 void TemplateTable::lushr() {
1208   transition(itos, ltos); // %%%% see lshl comment
1209 
1210   __ pop_l(O2);                          // shift value in O2, O3
1211 #ifdef _LP64
1212   __ srlx(O2, Otos_i, Otos_l);
1213 #else
1214   __ lushr(O2, O3, Otos_i, Otos_l1, Otos_l2, O4);
1215 #endif
1216 }
1217 
1218 
1219 void TemplateTable::fop2(Operation op) {
1220   transition(ftos, ftos);
1221   switch (op) {
1222    case  add:  __  pop_f(F4); __ fadd(FloatRegisterImpl::S, F4, Ftos_f, Ftos_f);  break;
1223    case  sub:  __  pop_f(F4); __ fsub(FloatRegisterImpl::S, F4, Ftos_f, Ftos_f);  break;
1224    case  mul:  __  pop_f(F4); __ fmul(FloatRegisterImpl::S, F4, Ftos_f, Ftos_f);  break;
1225    case  div:  __  pop_f(F4); __ fdiv(FloatRegisterImpl::S, F4, Ftos_f, Ftos_f);  break;
1226    case  rem:
1227      assert(Ftos_f == F0, "just checking");
1228 #ifdef _LP64
1229      // LP64 calling conventions use F1, F3 for passing 2 floats
1230      __ pop_f(F1);
1231      __ fmov(FloatRegisterImpl::S, Ftos_f, F3);
1232 #else
1233      __ pop_i(O0);
1234      __ stf(FloatRegisterImpl::S, Ftos_f, __ d_tmp);
1235      __ ld( __ d_tmp, O1 );
1236 #endif
1237      __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::frem));
1238      assert( Ftos_f == F0, "fix this code" );
1239      break;
1240 
1241    default: ShouldNotReachHere();
1242   }
1243 }
1244 
1245 
1246 void TemplateTable::dop2(Operation op) {
1247   transition(dtos, dtos);
1248   switch (op) {
1249    case  add:  __  pop_d(F4); __ fadd(FloatRegisterImpl::D, F4, Ftos_d, Ftos_d);  break;
1250    case  sub:  __  pop_d(F4); __ fsub(FloatRegisterImpl::D, F4, Ftos_d, Ftos_d);  break;
1251    case  mul:  __  pop_d(F4); __ fmul(FloatRegisterImpl::D, F4, Ftos_d, Ftos_d);  break;
1252    case  div:  __  pop_d(F4); __ fdiv(FloatRegisterImpl::D, F4, Ftos_d, Ftos_d);  break;
1253    case  rem:
1254 #ifdef _LP64
1255      // Pass arguments in D0, D2
1256      __ fmov(FloatRegisterImpl::D, Ftos_f, F2 );
1257      __ pop_d( F0 );
1258 #else
1259      // Pass arguments in O0O1, O2O3
1260      __ stf(FloatRegisterImpl::D, Ftos_f, __ d_tmp);
1261      __ ldd( __ d_tmp, O2 );
1262      __ pop_d(Ftos_f);
1263      __ stf(FloatRegisterImpl::D, Ftos_f, __ d_tmp);
1264      __ ldd( __ d_tmp, O0 );
1265 #endif
1266      __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::drem));
1267      assert( Ftos_d == F0, "fix this code" );
1268      break;
1269 
1270    default: ShouldNotReachHere();
1271   }
1272 }
1273 
1274 
1275 void TemplateTable::ineg() {
1276   transition(itos, itos);
1277   __ neg(Otos_i);
1278 }
1279 
1280 
1281 void TemplateTable::lneg() {
1282   transition(ltos, ltos);
1283 #ifdef _LP64
1284   __ sub(G0, Otos_l, Otos_l);
1285 #else
1286   __ lneg(Otos_l1, Otos_l2);
1287 #endif
1288 }
1289 
1290 
1291 void TemplateTable::fneg() {
1292   transition(ftos, ftos);
1293   __ fneg(FloatRegisterImpl::S, Ftos_f);
1294 }
1295 
1296 
1297 void TemplateTable::dneg() {
1298   transition(dtos, dtos);
1299   // v8 has fnegd if source and dest are the same
1300   __ fneg(FloatRegisterImpl::D, Ftos_f);
1301 }
1302 
1303 
1304 void TemplateTable::iinc() {
1305   transition(vtos, vtos);
1306   locals_index(G3_scratch);
1307   __ ldsb(Lbcp, 2, O2);  // load constant
1308   __ access_local_int(G3_scratch, Otos_i);
1309   __ add(Otos_i, O2, Otos_i);
1310   __ st(Otos_i, G3_scratch, Interpreter::value_offset_in_bytes());    // access_local_int puts E.A. in G3_scratch
1311 }
1312 
1313 
1314 void TemplateTable::wide_iinc() {
1315   transition(vtos, vtos);
1316   locals_index_wide(G3_scratch);
1317   __ get_2_byte_integer_at_bcp( 4,  O2, O3, InterpreterMacroAssembler::Signed);
1318   __ access_local_int(G3_scratch, Otos_i);
1319   __ add(Otos_i, O3, Otos_i);
1320   __ st(Otos_i, G3_scratch, Interpreter::value_offset_in_bytes());    // access_local_int puts E.A. in G3_scratch
1321 }
1322 
1323 
1324 void TemplateTable::convert() {
1325 // %%%%% Factor this first part accross platforms
1326   #ifdef ASSERT
1327     TosState tos_in  = ilgl;
1328     TosState tos_out = ilgl;
1329     switch (bytecode()) {
1330       case Bytecodes::_i2l: // fall through
1331       case Bytecodes::_i2f: // fall through
1332       case Bytecodes::_i2d: // fall through
1333       case Bytecodes::_i2b: // fall through
1334       case Bytecodes::_i2c: // fall through
1335       case Bytecodes::_i2s: tos_in = itos; break;
1336       case Bytecodes::_l2i: // fall through
1337       case Bytecodes::_l2f: // fall through
1338       case Bytecodes::_l2d: tos_in = ltos; break;
1339       case Bytecodes::_f2i: // fall through
1340       case Bytecodes::_f2l: // fall through
1341       case Bytecodes::_f2d: tos_in = ftos; break;
1342       case Bytecodes::_d2i: // fall through
1343       case Bytecodes::_d2l: // fall through
1344       case Bytecodes::_d2f: tos_in = dtos; break;
1345       default             : ShouldNotReachHere();
1346     }
1347     switch (bytecode()) {
1348       case Bytecodes::_l2i: // fall through
1349       case Bytecodes::_f2i: // fall through
1350       case Bytecodes::_d2i: // fall through
1351       case Bytecodes::_i2b: // fall through
1352       case Bytecodes::_i2c: // fall through
1353       case Bytecodes::_i2s: tos_out = itos; break;
1354       case Bytecodes::_i2l: // fall through
1355       case Bytecodes::_f2l: // fall through
1356       case Bytecodes::_d2l: tos_out = ltos; break;
1357       case Bytecodes::_i2f: // fall through
1358       case Bytecodes::_l2f: // fall through
1359       case Bytecodes::_d2f: tos_out = ftos; break;
1360       case Bytecodes::_i2d: // fall through
1361       case Bytecodes::_l2d: // fall through
1362       case Bytecodes::_f2d: tos_out = dtos; break;
1363       default             : ShouldNotReachHere();
1364     }
1365     transition(tos_in, tos_out);
1366   #endif
1367 
1368 
1369   // Conversion
1370   Label done;
1371   switch (bytecode()) {
1372    case Bytecodes::_i2l:
1373 #ifdef _LP64
1374     // Sign extend the 32 bits
1375     __ sra ( Otos_i, 0, Otos_l );
1376 #else
1377     __ addcc(Otos_i, 0, Otos_l2);
1378     __ br(Assembler::greaterEqual, true, Assembler::pt, done);
1379     __ delayed()->clr(Otos_l1);
1380     __ set(~0, Otos_l1);
1381 #endif
1382     break;
1383 
1384    case Bytecodes::_i2f:
1385     __ st(Otos_i, __ d_tmp );
1386     __ ldf(FloatRegisterImpl::S,  __ d_tmp, F0);
1387     __ fitof(FloatRegisterImpl::S, F0, Ftos_f);
1388     break;
1389 
1390    case Bytecodes::_i2d:
1391     __ st(Otos_i, __ d_tmp);
1392     __ ldf(FloatRegisterImpl::S,  __ d_tmp, F0);
1393     __ fitof(FloatRegisterImpl::D, F0, Ftos_f);
1394     break;
1395 
1396    case Bytecodes::_i2b:
1397     __ sll(Otos_i, 24, Otos_i);
1398     __ sra(Otos_i, 24, Otos_i);
1399     break;
1400 
1401    case Bytecodes::_i2c:
1402     __ sll(Otos_i, 16, Otos_i);
1403     __ srl(Otos_i, 16, Otos_i);
1404     break;
1405 
1406    case Bytecodes::_i2s:
1407     __ sll(Otos_i, 16, Otos_i);
1408     __ sra(Otos_i, 16, Otos_i);
1409     break;
1410 
1411    case Bytecodes::_l2i:
1412 #ifndef _LP64
1413     __ mov(Otos_l2, Otos_i);
1414 #else
1415     // Sign-extend into the high 32 bits
1416     __ sra(Otos_l, 0, Otos_i);
1417 #endif
1418     break;
1419 
1420    case Bytecodes::_l2f:
1421    case Bytecodes::_l2d:
1422     __ st_long(Otos_l, __ d_tmp);
1423     __ ldf(FloatRegisterImpl::D, __ d_tmp, Ftos_d);
1424 
1425     if (VM_Version::v9_instructions_work()) {
1426       if (bytecode() == Bytecodes::_l2f) {
1427         __ fxtof(FloatRegisterImpl::S, Ftos_d, Ftos_f);
1428       } else {
1429         __ fxtof(FloatRegisterImpl::D, Ftos_d, Ftos_d);
1430       }
1431     } else {
1432       __ call_VM_leaf(
1433         Lscratch,
1434         bytecode() == Bytecodes::_l2f
1435           ? CAST_FROM_FN_PTR(address, SharedRuntime::l2f)
1436           : CAST_FROM_FN_PTR(address, SharedRuntime::l2d)
1437       );
1438     }
1439     break;
1440 
1441   case Bytecodes::_f2i:  {
1442       Label isNaN;
1443       // result must be 0 if value is NaN; test by comparing value to itself
1444       __ fcmp(FloatRegisterImpl::S, Assembler::fcc0, Ftos_f, Ftos_f);
1445       // According to the v8 manual, you have to have a non-fp instruction
1446       // between fcmp and fb.
1447       if (!VM_Version::v9_instructions_work()) {
1448         __ nop();
1449       }
1450       __ fb(Assembler::f_unordered, true, Assembler::pn, isNaN);
1451       __ delayed()->clr(Otos_i);                                     // NaN
1452       __ ftoi(FloatRegisterImpl::S, Ftos_f, F30);
1453       __ stf(FloatRegisterImpl::S, F30, __ d_tmp);
1454       __ ld(__ d_tmp, Otos_i);
1455       __ bind(isNaN);
1456     }
1457     break;
1458 
1459    case Bytecodes::_f2l:
1460     // must uncache tos
1461     __ push_f();
1462 #ifdef _LP64
1463     __ pop_f(F1);
1464 #else
1465     __ pop_i(O0);
1466 #endif
1467     __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::f2l));
1468     break;
1469 
1470    case Bytecodes::_f2d:
1471     __ ftof( FloatRegisterImpl::S, FloatRegisterImpl::D, Ftos_f, Ftos_f);
1472     break;
1473 
1474    case Bytecodes::_d2i:
1475    case Bytecodes::_d2l:
1476     // must uncache tos
1477     __ push_d();
1478 #ifdef _LP64
1479     // LP64 calling conventions pass first double arg in D0
1480     __ pop_d( Ftos_d );
1481 #else
1482     __ pop_i( O0 );
1483     __ pop_i( O1 );
1484 #endif
1485     __ call_VM_leaf(Lscratch,
1486         bytecode() == Bytecodes::_d2i
1487           ? CAST_FROM_FN_PTR(address, SharedRuntime::d2i)
1488           : CAST_FROM_FN_PTR(address, SharedRuntime::d2l));
1489     break;
1490 
1491     case Bytecodes::_d2f:
1492     if (VM_Version::v9_instructions_work()) {
1493       __ ftof( FloatRegisterImpl::D, FloatRegisterImpl::S, Ftos_d, Ftos_f);
1494     }
1495     else {
1496       // must uncache tos
1497       __ push_d();
1498       __ pop_i(O0);
1499       __ pop_i(O1);
1500       __ call_VM_leaf(Lscratch, CAST_FROM_FN_PTR(address, SharedRuntime::d2f));
1501     }
1502     break;
1503 
1504     default: ShouldNotReachHere();
1505   }
1506   __ bind(done);
1507 }
1508 
1509 
1510 void TemplateTable::lcmp() {
1511   transition(ltos, itos);
1512 
1513 #ifdef _LP64
1514   __ pop_l(O1); // pop off value 1, value 2 is in O0
1515   __ lcmp( O1, Otos_l, Otos_i );
1516 #else
1517   __ pop_l(O2); // cmp O2,3 to O0,1
1518   __ lcmp( O2, O3, Otos_l1, Otos_l2, Otos_i );
1519 #endif
1520 }
1521 
1522 
1523 void TemplateTable::float_cmp(bool is_float, int unordered_result) {
1524 
1525   if (is_float) __ pop_f(F2);
1526   else          __ pop_d(F2);
1527 
1528   assert(Ftos_f == F0  &&  Ftos_d == F0,  "alias checking:");
1529 
1530   __ float_cmp( is_float, unordered_result, F2, F0, Otos_i );
1531 }
1532 
1533 void TemplateTable::branch(bool is_jsr, bool is_wide) {
1534   // Note: on SPARC, we use InterpreterMacroAssembler::if_cmp also.
1535   __ verify_oop(Lmethod);
1536   __ verify_thread();
1537 
1538   const Register O2_bumped_count = O2;
1539   __ profile_taken_branch(G3_scratch, O2_bumped_count);
1540 
1541   // get (wide) offset to O1_disp
1542   const Register O1_disp = O1;
1543   if (is_wide)  __ get_4_byte_integer_at_bcp( 1,  G4_scratch, O1_disp,                                    InterpreterMacroAssembler::set_CC);
1544   else          __ get_2_byte_integer_at_bcp( 1,  G4_scratch, O1_disp, InterpreterMacroAssembler::Signed, InterpreterMacroAssembler::set_CC);
1545 
1546   // Handle all the JSR stuff here, then exit.
1547   // It's much shorter and cleaner than intermingling with the
1548   // non-JSR normal-branch stuff occuring below.
1549   if( is_jsr ) {
1550     // compute return address as bci in Otos_i
1551     __ ld_ptr(Address(Lmethod, 0, in_bytes(methodOopDesc::const_offset())), G3_scratch);
1552     __ sub(Lbcp, G3_scratch, G3_scratch);
1553     __ sub(G3_scratch, in_bytes(constMethodOopDesc::codes_offset()) - (is_wide ? 5 : 3), Otos_i);
1554 
1555     // Bump Lbcp to target of JSR
1556     __ add(Lbcp, O1_disp, Lbcp);
1557     // Push returnAddress for "ret" on stack
1558     __ push_ptr(Otos_i, G0); // push ptr sized thing plus 0 for tag.
1559     // And away we go!
1560     __ dispatch_next(vtos);
1561     return;
1562   }
1563 
1564   // Normal (non-jsr) branch handling
1565 
1566   // Save the current Lbcp
1567   const Register O0_cur_bcp = O0;
1568   __ mov( Lbcp, O0_cur_bcp );
1569 
1570   bool increment_invocation_counter_for_backward_branches = UseCompiler && UseLoopCounter;
1571   if ( increment_invocation_counter_for_backward_branches ) {
1572     Label Lforward;
1573     // check branch direction
1574     __ br( Assembler::positive, false,  Assembler::pn, Lforward );
1575     // Bump bytecode pointer by displacement (take the branch)
1576     __ delayed()->add( O1_disp, Lbcp, Lbcp );     // add to bc addr
1577 
1578     // Update Backedge branch separately from invocations
1579     const Register G4_invoke_ctr = G4;
1580     __ increment_backedge_counter(G4_invoke_ctr, G1_scratch);
1581     if (ProfileInterpreter) {
1582       __ test_invocation_counter_for_mdp(G4_invoke_ctr, Lbcp, G3_scratch, Lforward);
1583       if (UseOnStackReplacement) {
1584         __ test_backedge_count_for_osr(O2_bumped_count, O0_cur_bcp, G3_scratch);
1585       }
1586     } else {
1587       if (UseOnStackReplacement) {
1588         __ test_backedge_count_for_osr(G4_invoke_ctr, O0_cur_bcp, G3_scratch);
1589       }
1590     }
1591 
1592     __ bind(Lforward);
1593   } else
1594     // Bump bytecode pointer by displacement (take the branch)
1595     __ add( O1_disp, Lbcp, Lbcp );// add to bc addr
1596 
1597   // continue with bytecode @ target
1598   // %%%%% Like Intel, could speed things up by moving bytecode fetch to code above,
1599   // %%%%% and changing dispatch_next to dispatch_only
1600   __ dispatch_next(vtos);
1601 }
1602 
1603 
1604 // Note Condition in argument is TemplateTable::Condition
1605 // arg scope is within class scope
1606 
1607 void TemplateTable::if_0cmp(Condition cc) {
1608   // no pointers, integer only!
1609   transition(itos, vtos);
1610   // assume branch is more often taken than not (loops use backward branches)
1611   __ cmp( Otos_i, 0);
1612   __ if_cmp(ccNot(cc), false);
1613 }
1614 
1615 
1616 void TemplateTable::if_icmp(Condition cc) {
1617   transition(itos, vtos);
1618   __ pop_i(O1);
1619   __ cmp(O1, Otos_i);
1620   __ if_cmp(ccNot(cc), false);
1621 }
1622 
1623 
1624 void TemplateTable::if_nullcmp(Condition cc) {
1625   transition(atos, vtos);
1626   __ tst(Otos_i);
1627   __ if_cmp(ccNot(cc), true);
1628 }
1629 
1630 
1631 void TemplateTable::if_acmp(Condition cc) {
1632   transition(atos, vtos);
1633   __ pop_ptr(O1);
1634   __ verify_oop(O1);
1635   __ verify_oop(Otos_i);
1636   __ cmp(O1, Otos_i);
1637   __ if_cmp(ccNot(cc), true);
1638 }
1639 
1640 
1641 
1642 void TemplateTable::ret() {
1643   transition(vtos, vtos);
1644   locals_index(G3_scratch);
1645   __ access_local_returnAddress(G3_scratch, Otos_i);
1646   // Otos_i contains the bci, compute the bcp from that
1647 
1648 #ifdef _LP64
1649 #ifdef ASSERT
1650   // jsr result was labeled as an 'itos' not an 'atos' because we cannot GC
1651   // the result.  The return address (really a BCI) was stored with an
1652   // 'astore' because JVM specs claim it's a pointer-sized thing.  Hence in
1653   // the 64-bit build the 32-bit BCI is actually in the low bits of a 64-bit
1654   // loaded value.
1655   { Label zzz ;
1656      __ set (65536, G3_scratch) ;
1657      __ cmp (Otos_i, G3_scratch) ;
1658      __ bp( Assembler::lessEqualUnsigned, false, Assembler::xcc, Assembler::pn, zzz);
1659      __ delayed()->nop();
1660      __ stop("BCI is in the wrong register half?");
1661      __ bind (zzz) ;
1662   }
1663 #endif
1664 #endif
1665 
1666   __ profile_ret(vtos, Otos_i, G4_scratch);
1667 
1668   __ ld_ptr(Address(Lmethod, 0, in_bytes(methodOopDesc::const_offset())), G3_scratch);
1669   __ add(G3_scratch, Otos_i, G3_scratch);
1670   __ add(G3_scratch, in_bytes(constMethodOopDesc::codes_offset()), Lbcp);
1671   __ dispatch_next(vtos);
1672 }
1673 
1674 
1675 void TemplateTable::wide_ret() {
1676   transition(vtos, vtos);
1677   locals_index_wide(G3_scratch);
1678   __ access_local_returnAddress(G3_scratch, Otos_i);
1679   // Otos_i contains the bci, compute the bcp from that
1680 
1681   __ profile_ret(vtos, Otos_i, G4_scratch);
1682 
1683   __ ld_ptr(Address(Lmethod, 0, in_bytes(methodOopDesc::const_offset())), G3_scratch);
1684   __ add(G3_scratch, Otos_i, G3_scratch);
1685   __ add(G3_scratch, in_bytes(constMethodOopDesc::codes_offset()), Lbcp);
1686   __ dispatch_next(vtos);
1687 }
1688 
1689 
1690 void TemplateTable::tableswitch() {
1691   transition(itos, vtos);
1692   Label default_case, continue_execution;
1693 
1694   // align bcp
1695   __ add(Lbcp, BytesPerInt, O1);
1696   __ and3(O1, -BytesPerInt, O1);
1697   // load lo, hi
1698   __ ld(O1, 1 * BytesPerInt, O2);       // Low Byte
1699   __ ld(O1, 2 * BytesPerInt, O3);       // High Byte
1700 #ifdef _LP64
1701   // Sign extend the 32 bits
1702   __ sra ( Otos_i, 0, Otos_i );
1703 #endif /* _LP64 */
1704 
1705   // check against lo & hi
1706   __ cmp( Otos_i, O2);
1707   __ br( Assembler::less, false, Assembler::pn, default_case);
1708   __ delayed()->cmp( Otos_i, O3 );
1709   __ br( Assembler::greater, false, Assembler::pn, default_case);
1710   // lookup dispatch offset
1711   __ delayed()->sub(Otos_i, O2, O2);
1712   __ profile_switch_case(O2, O3, G3_scratch, G4_scratch);
1713   __ sll(O2, LogBytesPerInt, O2);
1714   __ add(O2, 3 * BytesPerInt, O2);
1715   __ ba(false, continue_execution);
1716   __ delayed()->ld(O1, O2, O2);
1717   // handle default
1718   __ bind(default_case);
1719   __ profile_switch_default(O3);
1720   __ ld(O1, 0, O2); // get default offset
1721   // continue execution
1722   __ bind(continue_execution);
1723   __ add(Lbcp, O2, Lbcp);
1724   __ dispatch_next(vtos);
1725 }
1726 
1727 
1728 void TemplateTable::lookupswitch() {
1729   transition(itos, itos);
1730   __ stop("lookupswitch bytecode should have been rewritten");
1731 }
1732 
1733 void TemplateTable::fast_linearswitch() {
1734   transition(itos, vtos);
1735     Label loop_entry, loop, found, continue_execution;
1736   // align bcp
1737   __ add(Lbcp, BytesPerInt, O1);
1738   __ and3(O1, -BytesPerInt, O1);
1739  // set counter
1740   __ ld(O1, BytesPerInt, O2);
1741   __ sll(O2, LogBytesPerInt + 1, O2); // in word-pairs
1742   __ add(O1, 2 * BytesPerInt, O3); // set first pair addr
1743   __ ba(false, loop_entry);
1744   __ delayed()->add(O3, O2, O2); // counter now points past last pair
1745 
1746   // table search
1747   __ bind(loop);
1748   __ cmp(O4, Otos_i);
1749   __ br(Assembler::equal, true, Assembler::pn, found);
1750   __ delayed()->ld(O3, BytesPerInt, O4); // offset -> O4
1751   __ inc(O3, 2 * BytesPerInt);
1752 
1753   __ bind(loop_entry);
1754   __ cmp(O2, O3);
1755   __ brx(Assembler::greaterUnsigned, true, Assembler::pt, loop);
1756   __ delayed()->ld(O3, 0, O4);
1757 
1758   // default case
1759   __ ld(O1, 0, O4); // get default offset
1760   if (ProfileInterpreter) {
1761     __ profile_switch_default(O3);
1762     __ ba(false, continue_execution);
1763     __ delayed()->nop();
1764   }
1765 
1766   // entry found -> get offset
1767   __ bind(found);
1768   if (ProfileInterpreter) {
1769     __ sub(O3, O1, O3);
1770     __ sub(O3, 2*BytesPerInt, O3);
1771     __ srl(O3, LogBytesPerInt + 1, O3); // in word-pairs
1772     __ profile_switch_case(O3, O1, O2, G3_scratch);
1773 
1774     __ bind(continue_execution);
1775   }
1776   __ add(Lbcp, O4, Lbcp);
1777   __ dispatch_next(vtos);
1778 }
1779 
1780 
1781 void TemplateTable::fast_binaryswitch() {
1782   transition(itos, vtos);
1783   // Implementation using the following core algorithm: (copied from Intel)
1784   //
1785   // int binary_search(int key, LookupswitchPair* array, int n) {
1786   //   // Binary search according to "Methodik des Programmierens" by
1787   //   // Edsger W. Dijkstra and W.H.J. Feijen, Addison Wesley Germany 1985.
1788   //   int i = 0;
1789   //   int j = n;
1790   //   while (i+1 < j) {
1791   //     // invariant P: 0 <= i < j <= n and (a[i] <= key < a[j] or Q)
1792   //     // with      Q: for all i: 0 <= i < n: key < a[i]
1793   //     // where a stands for the array and assuming that the (inexisting)
1794   //     // element a[n] is infinitely big.
1795   //     int h = (i + j) >> 1;
1796   //     // i < h < j
1797   //     if (key < array[h].fast_match()) {
1798   //       j = h;
1799   //     } else {
1800   //       i = h;
1801   //     }
1802   //   }
1803   //   // R: a[i] <= key < a[i+1] or Q
1804   //   // (i.e., if key is within array, i is the correct index)
1805   //   return i;
1806   // }
1807 
1808   // register allocation
1809   assert(Otos_i == O0, "alias checking");
1810   const Register Rkey     = Otos_i;                    // already set (tosca)
1811   const Register Rarray   = O1;
1812   const Register Ri       = O2;
1813   const Register Rj       = O3;
1814   const Register Rh       = O4;
1815   const Register Rscratch = O5;
1816 
1817   const int log_entry_size = 3;
1818   const int entry_size = 1 << log_entry_size;
1819 
1820   Label found;
1821   // Find Array start
1822   __ add(Lbcp, 3 * BytesPerInt, Rarray);
1823   __ and3(Rarray, -BytesPerInt, Rarray);
1824   // initialize i & j (in delay slot)
1825   __ clr( Ri );
1826 
1827   // and start
1828   Label entry;
1829   __ ba(false, entry);
1830   __ delayed()->ld( Rarray, -BytesPerInt, Rj);
1831   // (Rj is already in the native byte-ordering.)
1832 
1833   // binary search loop
1834   { Label loop;
1835     __ bind( loop );
1836     // int h = (i + j) >> 1;
1837     __ sra( Rh, 1, Rh );
1838     // if (key < array[h].fast_match()) {
1839     //   j = h;
1840     // } else {
1841     //   i = h;
1842     // }
1843     __ sll( Rh, log_entry_size, Rscratch );
1844     __ ld( Rarray, Rscratch, Rscratch );
1845     // (Rscratch is already in the native byte-ordering.)
1846     __ cmp( Rkey, Rscratch );
1847     if ( VM_Version::v9_instructions_work() ) {
1848       __ movcc( Assembler::less,         false, Assembler::icc, Rh, Rj );  // j = h if (key <  array[h].fast_match())
1849       __ movcc( Assembler::greaterEqual, false, Assembler::icc, Rh, Ri );  // i = h if (key >= array[h].fast_match())
1850     }
1851     else {
1852       Label end_of_if;
1853       __ br( Assembler::less, true, Assembler::pt, end_of_if );
1854       __ delayed()->mov( Rh, Rj ); // if (<) Rj = Rh
1855       __ mov( Rh, Ri );            // else i = h
1856       __ bind(end_of_if);          // }
1857     }
1858 
1859     // while (i+1 < j)
1860     __ bind( entry );
1861     __ add( Ri, 1, Rscratch );
1862     __ cmp(Rscratch, Rj);
1863     __ br( Assembler::less, true, Assembler::pt, loop );
1864     __ delayed()->add( Ri, Rj, Rh ); // start h = i + j  >> 1;
1865   }
1866 
1867   // end of binary search, result index is i (must check again!)
1868   Label default_case;
1869   Label continue_execution;
1870   if (ProfileInterpreter) {
1871     __ mov( Ri, Rh );              // Save index in i for profiling
1872   }
1873   __ sll( Ri, log_entry_size, Ri );
1874   __ ld( Rarray, Ri, Rscratch );
1875   // (Rscratch is already in the native byte-ordering.)
1876   __ cmp( Rkey, Rscratch );
1877   __ br( Assembler::notEqual, true, Assembler::pn, default_case );
1878   __ delayed()->ld( Rarray, -2 * BytesPerInt, Rj ); // load default offset -> j
1879 
1880   // entry found -> j = offset
1881   __ inc( Ri, BytesPerInt );
1882   __ profile_switch_case(Rh, Rj, Rscratch, Rkey);
1883   __ ld( Rarray, Ri, Rj );
1884   // (Rj is already in the native byte-ordering.)
1885 
1886   if (ProfileInterpreter) {
1887     __ ba(false, continue_execution);
1888     __ delayed()->nop();
1889   }
1890 
1891   __ bind(default_case); // fall through (if not profiling)
1892   __ profile_switch_default(Ri);
1893 
1894   __ bind(continue_execution);
1895   __ add( Lbcp, Rj, Lbcp );
1896   __ dispatch_next( vtos );
1897 }
1898 
1899 
1900 void TemplateTable::_return(TosState state) {
1901   transition(state, state);
1902   assert(_desc->calls_vm(), "inconsistent calls_vm information");
1903 
1904   if (_desc->bytecode() == Bytecodes::_return_register_finalizer) {
1905     assert(state == vtos, "only valid state");
1906     __ mov(G0, G3_scratch);
1907     __ access_local_ptr(G3_scratch, Otos_i);
1908     __ load_klass(Otos_i, O2);
1909     __ set(JVM_ACC_HAS_FINALIZER, G3);
1910     __ ld(O2, Klass::access_flags_offset_in_bytes() + sizeof(oopDesc), O2);
1911     __ andcc(G3, O2, G0);
1912     Label skip_register_finalizer;
1913     __ br(Assembler::zero, false, Assembler::pn, skip_register_finalizer);
1914     __ delayed()->nop();
1915 
1916     // Call out to do finalizer registration
1917     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::register_finalizer), Otos_i);
1918 
1919     __ bind(skip_register_finalizer);
1920   }
1921 
1922   __ remove_activation(state, /* throw_monitor_exception */ true);
1923 
1924   // The caller's SP was adjusted upon method entry to accomodate
1925   // the callee's non-argument locals. Undo that adjustment.
1926   __ ret();                             // return to caller
1927   __ delayed()->restore(I5_savedSP, G0, SP);
1928 }
1929 
1930 
1931 // ----------------------------------------------------------------------------
1932 // Volatile variables demand their effects be made known to all CPU's in
1933 // order.  Store buffers on most chips allow reads & writes to reorder; the
1934 // JMM's ReadAfterWrite.java test fails in -Xint mode without some kind of
1935 // memory barrier (i.e., it's not sufficient that the interpreter does not
1936 // reorder volatile references, the hardware also must not reorder them).
1937 //
1938 // According to the new Java Memory Model (JMM):
1939 // (1) All volatiles are serialized wrt to each other.
1940 // ALSO reads & writes act as aquire & release, so:
1941 // (2) A read cannot let unrelated NON-volatile memory refs that happen after
1942 // the read float up to before the read.  It's OK for non-volatile memory refs
1943 // that happen before the volatile read to float down below it.
1944 // (3) Similar a volatile write cannot let unrelated NON-volatile memory refs
1945 // that happen BEFORE the write float down to after the write.  It's OK for
1946 // non-volatile memory refs that happen after the volatile write to float up
1947 // before it.
1948 //
1949 // We only put in barriers around volatile refs (they are expensive), not
1950 // _between_ memory refs (that would require us to track the flavor of the
1951 // previous memory refs).  Requirements (2) and (3) require some barriers
1952 // before volatile stores and after volatile loads.  These nearly cover
1953 // requirement (1) but miss the volatile-store-volatile-load case.  This final
1954 // case is placed after volatile-stores although it could just as well go
1955 // before volatile-loads.
1956 void TemplateTable::volatile_barrier(Assembler::Membar_mask_bits order_constraint) {
1957   // Helper function to insert a is-volatile test and memory barrier
1958   // All current sparc implementations run in TSO, needing only StoreLoad
1959   if ((order_constraint & Assembler::StoreLoad) == 0) return;
1960   __ membar( order_constraint );
1961 }
1962 
1963 // ----------------------------------------------------------------------------
1964 void TemplateTable::resolve_cache_and_index(int byte_no, Register Rcache, Register index) {
1965   assert(byte_no == 1 || byte_no == 2, "byte_no out of range");
1966   // Depends on cpCacheOop layout!
1967   const int shift_count = (1 + byte_no)*BitsPerByte;
1968   Label resolved;
1969 
1970   __ get_cache_and_index_at_bcp(Rcache, index, 1);
1971   __ ld_ptr(Address(Rcache, 0, in_bytes(constantPoolCacheOopDesc::base_offset() +
1972                                         ConstantPoolCacheEntry::indices_offset())), Lbyte_code);
1973 
1974   __ srl(  Lbyte_code, shift_count, Lbyte_code );
1975   __ and3( Lbyte_code,        0xFF, Lbyte_code );
1976   __ cmp(  Lbyte_code, (int)bytecode());
1977   __ br(   Assembler::equal, false, Assembler::pt, resolved);
1978   __ delayed()->set((int)bytecode(), O1);
1979 
1980   address entry;
1981   switch (bytecode()) {
1982     case Bytecodes::_getstatic      : // fall through
1983     case Bytecodes::_putstatic      : // fall through
1984     case Bytecodes::_getfield       : // fall through
1985     case Bytecodes::_putfield       : entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_get_put); break;
1986     case Bytecodes::_invokevirtual  : // fall through
1987     case Bytecodes::_invokespecial  : // fall through
1988     case Bytecodes::_invokestatic   : // fall through
1989     case Bytecodes::_invokeinterface: entry = CAST_FROM_FN_PTR(address, InterpreterRuntime::resolve_invoke);  break;
1990     default                         : ShouldNotReachHere();                                 break;
1991   }
1992   // first time invocation - must resolve first
1993   __ call_VM(noreg, entry, O1);
1994   // Update registers with resolved info
1995   __ get_cache_and_index_at_bcp(Rcache, index, 1);
1996   __ bind(resolved);
1997 }
1998 
1999 void TemplateTable::load_invoke_cp_cache_entry(int byte_no,
2000                                                Register Rmethod,
2001                                                Register Ritable_index,
2002                                                Register Rflags,
2003                                                bool is_invokevirtual,
2004                                                bool is_invokevfinal) {
2005   // Uses both G3_scratch and G4_scratch
2006   Register Rcache = G3_scratch;
2007   Register Rscratch = G4_scratch;
2008   assert_different_registers(Rcache, Rmethod, Ritable_index);
2009 
2010   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2011 
2012   // determine constant pool cache field offsets
2013   const int method_offset = in_bytes(
2014     cp_base_offset +
2015       (is_invokevirtual
2016        ? ConstantPoolCacheEntry::f2_offset()
2017        : ConstantPoolCacheEntry::f1_offset()
2018       )
2019     );
2020   const int flags_offset = in_bytes(cp_base_offset +
2021                                     ConstantPoolCacheEntry::flags_offset());
2022   // access constant pool cache fields
2023   const int index_offset = in_bytes(cp_base_offset +
2024                                     ConstantPoolCacheEntry::f2_offset());
2025 
2026   if (is_invokevfinal) {
2027     __ get_cache_and_index_at_bcp(Rcache, Rscratch, 1);
2028   } else {
2029     resolve_cache_and_index(byte_no, Rcache, Rscratch);
2030   }
2031 
2032   __ ld_ptr(Address(Rcache, 0, method_offset), Rmethod);
2033   if (Ritable_index != noreg) {
2034     __ ld_ptr(Address(Rcache, 0, index_offset), Ritable_index);
2035   }
2036   __ ld_ptr(Address(Rcache, 0, flags_offset),  Rflags);
2037 }
2038 
2039 // The Rcache register must be set before call
2040 void TemplateTable::load_field_cp_cache_entry(Register Robj,
2041                                               Register Rcache,
2042                                               Register index,
2043                                               Register Roffset,
2044                                               Register Rflags,
2045                                               bool is_static) {
2046   assert_different_registers(Rcache, Rflags, Roffset);
2047 
2048   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2049 
2050   __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset +
2051                              ConstantPoolCacheEntry::flags_offset())), Rflags);
2052   __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset +
2053                              ConstantPoolCacheEntry::f2_offset())), Roffset);
2054   if (is_static) {
2055     __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset +
2056                              ConstantPoolCacheEntry::f1_offset())), Robj);
2057   }
2058 }
2059 
2060 // The registers Rcache and index expected to be set before call.
2061 // Correct values of the Rcache and index registers are preserved.
2062 void TemplateTable::jvmti_post_field_access(Register Rcache,
2063                                             Register index,
2064                                             bool is_static,
2065                                             bool has_tos) {
2066   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2067 
2068   if (JvmtiExport::can_post_field_access()) {
2069     // Check to see if a field access watch has been set before we take
2070     // the time to call into the VM.
2071     Label Label1;
2072     assert_different_registers(Rcache, index, G1_scratch);
2073     Address get_field_access_count_addr(G1_scratch,
2074                                         (address)JvmtiExport::get_field_access_count_addr(),
2075                                         relocInfo::none);
2076     __ load_contents(get_field_access_count_addr, G1_scratch);
2077     __ tst(G1_scratch);
2078     __ br(Assembler::zero, false, Assembler::pt, Label1);
2079     __ delayed()->nop();
2080 
2081     __ add(Rcache, in_bytes(cp_base_offset), Rcache);
2082 
2083     if (is_static) {
2084       __ clr(Otos_i);
2085     } else {
2086       if (has_tos) {
2087       // save object pointer before call_VM() clobbers it
2088         __ mov(Otos_i, Lscratch);
2089       } else {
2090         // Load top of stack (do not pop the value off the stack);
2091         __ ld_ptr(Lesp, Interpreter::expr_offset_in_bytes(0), Otos_i);
2092       }
2093       __ verify_oop(Otos_i);
2094     }
2095     // Otos_i: object pointer or NULL if static
2096     // Rcache: cache entry pointer
2097     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_access),
2098                Otos_i, Rcache);
2099     if (!is_static && has_tos) {
2100       __ mov(Lscratch, Otos_i);  // restore object pointer
2101       __ verify_oop(Otos_i);
2102     }
2103     __ get_cache_and_index_at_bcp(Rcache, index, 1);
2104     __ bind(Label1);
2105   }
2106 }
2107 
2108 void TemplateTable::getfield_or_static(int byte_no, bool is_static) {
2109   transition(vtos, vtos);
2110 
2111   Register Rcache = G3_scratch;
2112   Register index  = G4_scratch;
2113   Register Rclass = Rcache;
2114   Register Roffset= G4_scratch;
2115   Register Rflags = G1_scratch;
2116   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2117 
2118   resolve_cache_and_index(byte_no, Rcache, index);
2119   jvmti_post_field_access(Rcache, index, is_static, false);
2120   load_field_cp_cache_entry(Rclass, Rcache, index, Roffset, Rflags, is_static);
2121 
2122   if (!is_static) {
2123     pop_and_check_object(Rclass);
2124   } else {
2125     __ verify_oop(Rclass);
2126   }
2127 
2128   Label exit;
2129 
2130   Assembler::Membar_mask_bits membar_bits =
2131     Assembler::Membar_mask_bits(Assembler::LoadLoad | Assembler::LoadStore);
2132 
2133   if (__ membar_has_effect(membar_bits)) {
2134     // Get volatile flag
2135     __ set((1 << ConstantPoolCacheEntry::volatileField), Lscratch);
2136     __ and3(Rflags, Lscratch, Lscratch);
2137   }
2138 
2139   Label checkVolatile;
2140 
2141   // compute field type
2142   Label notByte, notInt, notShort, notChar, notLong, notFloat, notObj;
2143   __ srl(Rflags, ConstantPoolCacheEntry::tosBits, Rflags);
2144   // Make sure we don't need to mask Rflags for tosBits after the above shift
2145   ConstantPoolCacheEntry::verify_tosBits();
2146 
2147   // Check atos before itos for getstatic, more likely (in Queens at least)
2148   __ cmp(Rflags, atos);
2149   __ br(Assembler::notEqual, false, Assembler::pt, notObj);
2150   __ delayed() ->cmp(Rflags, itos);
2151 
2152   // atos
2153   __ load_heap_oop(Rclass, Roffset, Otos_i);
2154   __ verify_oop(Otos_i);
2155   __ push(atos);
2156   if (!is_static) {
2157     patch_bytecode(Bytecodes::_fast_agetfield, G3_scratch, G4_scratch);
2158   }
2159   __ ba(false, checkVolatile);
2160   __ delayed()->tst(Lscratch);
2161 
2162   __ bind(notObj);
2163 
2164   // cmp(Rflags, itos);
2165   __ br(Assembler::notEqual, false, Assembler::pt, notInt);
2166   __ delayed() ->cmp(Rflags, ltos);
2167 
2168   // itos
2169   __ ld(Rclass, Roffset, Otos_i);
2170   __ push(itos);
2171   if (!is_static) {
2172     patch_bytecode(Bytecodes::_fast_igetfield, G3_scratch, G4_scratch);
2173   }
2174   __ ba(false, checkVolatile);
2175   __ delayed()->tst(Lscratch);
2176 
2177   __ bind(notInt);
2178 
2179   // cmp(Rflags, ltos);
2180   __ br(Assembler::notEqual, false, Assembler::pt, notLong);
2181   __ delayed() ->cmp(Rflags, btos);
2182 
2183   // ltos
2184   // load must be atomic
2185   __ ld_long(Rclass, Roffset, Otos_l);
2186   __ push(ltos);
2187   if (!is_static) {
2188     patch_bytecode(Bytecodes::_fast_lgetfield, G3_scratch, G4_scratch);
2189   }
2190   __ ba(false, checkVolatile);
2191   __ delayed()->tst(Lscratch);
2192 
2193   __ bind(notLong);
2194 
2195   // cmp(Rflags, btos);
2196   __ br(Assembler::notEqual, false, Assembler::pt, notByte);
2197   __ delayed() ->cmp(Rflags, ctos);
2198 
2199   // btos
2200   __ ldsb(Rclass, Roffset, Otos_i);
2201   __ push(itos);
2202   if (!is_static) {
2203     patch_bytecode(Bytecodes::_fast_bgetfield, G3_scratch, G4_scratch);
2204   }
2205   __ ba(false, checkVolatile);
2206   __ delayed()->tst(Lscratch);
2207 
2208   __ bind(notByte);
2209 
2210   // cmp(Rflags, ctos);
2211   __ br(Assembler::notEqual, false, Assembler::pt, notChar);
2212   __ delayed() ->cmp(Rflags, stos);
2213 
2214   // ctos
2215   __ lduh(Rclass, Roffset, Otos_i);
2216   __ push(itos);
2217   if (!is_static) {
2218     patch_bytecode(Bytecodes::_fast_cgetfield, G3_scratch, G4_scratch);
2219   }
2220   __ ba(false, checkVolatile);
2221   __ delayed()->tst(Lscratch);
2222 
2223   __ bind(notChar);
2224 
2225   // cmp(Rflags, stos);
2226   __ br(Assembler::notEqual, false, Assembler::pt, notShort);
2227   __ delayed() ->cmp(Rflags, ftos);
2228 
2229   // stos
2230   __ ldsh(Rclass, Roffset, Otos_i);
2231   __ push(itos);
2232   if (!is_static) {
2233     patch_bytecode(Bytecodes::_fast_sgetfield, G3_scratch, G4_scratch);
2234   }
2235   __ ba(false, checkVolatile);
2236   __ delayed()->tst(Lscratch);
2237 
2238   __ bind(notShort);
2239 
2240 
2241   // cmp(Rflags, ftos);
2242   __ br(Assembler::notEqual, false, Assembler::pt, notFloat);
2243   __ delayed() ->tst(Lscratch);
2244 
2245   // ftos
2246   __ ldf(FloatRegisterImpl::S, Rclass, Roffset, Ftos_f);
2247   __ push(ftos);
2248   if (!is_static) {
2249     patch_bytecode(Bytecodes::_fast_fgetfield, G3_scratch, G4_scratch);
2250   }
2251   __ ba(false, checkVolatile);
2252   __ delayed()->tst(Lscratch);
2253 
2254   __ bind(notFloat);
2255 
2256 
2257   // dtos
2258   __ ldf(FloatRegisterImpl::D, Rclass, Roffset, Ftos_d);
2259   __ push(dtos);
2260   if (!is_static) {
2261     patch_bytecode(Bytecodes::_fast_dgetfield, G3_scratch, G4_scratch);
2262   }
2263 
2264   __ bind(checkVolatile);
2265   if (__ membar_has_effect(membar_bits)) {
2266     // __ tst(Lscratch); executed in delay slot
2267     __ br(Assembler::zero, false, Assembler::pt, exit);
2268     __ delayed()->nop();
2269     volatile_barrier(membar_bits);
2270   }
2271 
2272   __ bind(exit);
2273 }
2274 
2275 
2276 void TemplateTable::getfield(int byte_no) {
2277   getfield_or_static(byte_no, false);
2278 }
2279 
2280 void TemplateTable::getstatic(int byte_no) {
2281   getfield_or_static(byte_no, true);
2282 }
2283 
2284 
2285 void TemplateTable::fast_accessfield(TosState state) {
2286   transition(atos, state);
2287   Register Rcache  = G3_scratch;
2288   Register index   = G4_scratch;
2289   Register Roffset = G4_scratch;
2290   Register Rflags  = Rcache;
2291   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2292 
2293   __ get_cache_and_index_at_bcp(Rcache, index, 1);
2294   jvmti_post_field_access(Rcache, index, /*is_static*/false, /*has_tos*/true);
2295 
2296   __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset + ConstantPoolCacheEntry::f2_offset())), Roffset);
2297 
2298   __ null_check(Otos_i);
2299   __ verify_oop(Otos_i);
2300 
2301   Label exit;
2302 
2303   Assembler::Membar_mask_bits membar_bits =
2304     Assembler::Membar_mask_bits(Assembler::LoadLoad | Assembler::LoadStore);
2305   if (__ membar_has_effect(membar_bits)) {
2306     // Get volatile flag
2307     __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset + ConstantPoolCacheEntry::f2_offset())), Rflags);
2308     __ set((1 << ConstantPoolCacheEntry::volatileField), Lscratch);
2309   }
2310 
2311   switch (bytecode()) {
2312     case Bytecodes::_fast_bgetfield:
2313       __ ldsb(Otos_i, Roffset, Otos_i);
2314       break;
2315     case Bytecodes::_fast_cgetfield:
2316       __ lduh(Otos_i, Roffset, Otos_i);
2317       break;
2318     case Bytecodes::_fast_sgetfield:
2319       __ ldsh(Otos_i, Roffset, Otos_i);
2320       break;
2321     case Bytecodes::_fast_igetfield:
2322       __ ld(Otos_i, Roffset, Otos_i);
2323       break;
2324     case Bytecodes::_fast_lgetfield:
2325       __ ld_long(Otos_i, Roffset, Otos_l);
2326       break;
2327     case Bytecodes::_fast_fgetfield:
2328       __ ldf(FloatRegisterImpl::S, Otos_i, Roffset, Ftos_f);
2329       break;
2330     case Bytecodes::_fast_dgetfield:
2331       __ ldf(FloatRegisterImpl::D, Otos_i, Roffset, Ftos_d);
2332       break;
2333     case Bytecodes::_fast_agetfield:
2334       __ load_heap_oop(Otos_i, Roffset, Otos_i);
2335       break;
2336     default:
2337       ShouldNotReachHere();
2338   }
2339 
2340   if (__ membar_has_effect(membar_bits)) {
2341     __ btst(Lscratch, Rflags);
2342     __ br(Assembler::zero, false, Assembler::pt, exit);
2343     __ delayed()->nop();
2344     volatile_barrier(membar_bits);
2345     __ bind(exit);
2346   }
2347 
2348   if (state == atos) {
2349     __ verify_oop(Otos_i);    // does not blow flags!
2350   }
2351 }
2352 
2353 void TemplateTable::jvmti_post_fast_field_mod() {
2354   if (JvmtiExport::can_post_field_modification()) {
2355     // Check to see if a field modification watch has been set before we take
2356     // the time to call into the VM.
2357     Label done;
2358     Address get_field_modification_count_addr(G4_scratch, (address)JvmtiExport::get_field_modification_count_addr(), relocInfo::none);
2359     __ load_contents(get_field_modification_count_addr, G4_scratch);
2360     __ tst(G4_scratch);
2361     __ br(Assembler::zero, false, Assembler::pt, done);
2362     __ delayed()->nop();
2363     __ pop_ptr(G4_scratch);     // copy the object pointer from tos
2364     __ verify_oop(G4_scratch);
2365     __ push_ptr(G4_scratch);    // put the object pointer back on tos
2366     __ get_cache_entry_pointer_at_bcp(G1_scratch, G3_scratch, 1);
2367     // Save tos values before call_VM() clobbers them. Since we have
2368     // to do it for every data type, we use the saved values as the
2369     // jvalue object.
2370     switch (bytecode()) {  // save tos values before call_VM() clobbers them
2371     case Bytecodes::_fast_aputfield: __ push_ptr(Otos_i); break;
2372     case Bytecodes::_fast_bputfield: // fall through
2373     case Bytecodes::_fast_sputfield: // fall through
2374     case Bytecodes::_fast_cputfield: // fall through
2375     case Bytecodes::_fast_iputfield: __ push_i(Otos_i); break;
2376     case Bytecodes::_fast_dputfield: __ push_d(Ftos_d); break;
2377     case Bytecodes::_fast_fputfield: __ push_f(Ftos_f); break;
2378     // get words in right order for use as jvalue object
2379     case Bytecodes::_fast_lputfield: __ push_l(Otos_l); break;
2380     }
2381     // setup pointer to jvalue object
2382     __ mov(Lesp, G3_scratch);  __ inc(G3_scratch, wordSize);
2383     // G4_scratch:  object pointer
2384     // G1_scratch: cache entry pointer
2385     // G3_scratch: jvalue object on the stack
2386     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification), G4_scratch, G1_scratch, G3_scratch);
2387     switch (bytecode()) {             // restore tos values
2388     case Bytecodes::_fast_aputfield: __ pop_ptr(Otos_i); break;
2389     case Bytecodes::_fast_bputfield: // fall through
2390     case Bytecodes::_fast_sputfield: // fall through
2391     case Bytecodes::_fast_cputfield: // fall through
2392     case Bytecodes::_fast_iputfield: __ pop_i(Otos_i); break;
2393     case Bytecodes::_fast_dputfield: __ pop_d(Ftos_d); break;
2394     case Bytecodes::_fast_fputfield: __ pop_f(Ftos_f); break;
2395     case Bytecodes::_fast_lputfield: __ pop_l(Otos_l); break;
2396     }
2397     __ bind(done);
2398   }
2399 }
2400 
2401 // The registers Rcache and index expected to be set before call.
2402 // The function may destroy various registers, just not the Rcache and index registers.
2403 void TemplateTable::jvmti_post_field_mod(Register Rcache, Register index, bool is_static) {
2404   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2405 
2406   if (JvmtiExport::can_post_field_modification()) {
2407     // Check to see if a field modification watch has been set before we take
2408     // the time to call into the VM.
2409     Label Label1;
2410     assert_different_registers(Rcache, index, G1_scratch);
2411     Address get_field_modification_count_addr(G1_scratch,
2412                                               (address)JvmtiExport::get_field_modification_count_addr(),
2413                                               relocInfo::none);
2414     __ load_contents(get_field_modification_count_addr, G1_scratch);
2415     __ tst(G1_scratch);
2416     __ br(Assembler::zero, false, Assembler::pt, Label1);
2417     __ delayed()->nop();
2418 
2419     // The Rcache and index registers have been already set.
2420     // This allows to eliminate this call but the Rcache and index
2421     // registers must be correspondingly used after this line.
2422     __ get_cache_and_index_at_bcp(G1_scratch, G4_scratch, 1);
2423 
2424     __ add(G1_scratch, in_bytes(cp_base_offset), G3_scratch);
2425     if (is_static) {
2426       // Life is simple.  Null out the object pointer.
2427       __ clr(G4_scratch);
2428     } else {
2429       Register Rflags = G1_scratch;
2430       // Life is harder. The stack holds the value on top, followed by the
2431       // object.  We don't know the size of the value, though; it could be
2432       // one or two words depending on its type. As a result, we must find
2433       // the type to determine where the object is.
2434 
2435       Label two_word, valsizeknown;
2436       __ ld_ptr(Address(G1_scratch, 0, in_bytes(cp_base_offset + ConstantPoolCacheEntry::flags_offset())), Rflags);
2437       __ mov(Lesp, G4_scratch);
2438       __ srl(Rflags, ConstantPoolCacheEntry::tosBits, Rflags);
2439       // Make sure we don't need to mask Rflags for tosBits after the above shift
2440       ConstantPoolCacheEntry::verify_tosBits();
2441       __ cmp(Rflags, ltos);
2442       __ br(Assembler::equal, false, Assembler::pt, two_word);
2443       __ delayed()->cmp(Rflags, dtos);
2444       __ br(Assembler::equal, false, Assembler::pt, two_word);
2445       __ delayed()->nop();
2446       __ inc(G4_scratch, Interpreter::expr_offset_in_bytes(1));
2447       __ br(Assembler::always, false, Assembler::pt, valsizeknown);
2448       __ delayed()->nop();
2449       __ bind(two_word);
2450 
2451       __ inc(G4_scratch, Interpreter::expr_offset_in_bytes(2));
2452 
2453       __ bind(valsizeknown);
2454       // setup object pointer
2455       __ ld_ptr(G4_scratch, 0, G4_scratch);
2456       __ verify_oop(G4_scratch);
2457     }
2458     // setup pointer to jvalue object
2459     __ mov(Lesp, G1_scratch);  __ inc(G1_scratch, wordSize);
2460     // G4_scratch:  object pointer or NULL if static
2461     // G3_scratch: cache entry pointer
2462     // G1_scratch: jvalue object on the stack
2463     __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::post_field_modification),
2464                G4_scratch, G3_scratch, G1_scratch);
2465     __ get_cache_and_index_at_bcp(Rcache, index, 1);
2466     __ bind(Label1);
2467   }
2468 }
2469 
2470 void TemplateTable::pop_and_check_object(Register r) {
2471   __ pop_ptr(r);
2472   __ null_check(r);  // for field access must check obj.
2473   __ verify_oop(r);
2474 }
2475 
2476 void TemplateTable::putfield_or_static(int byte_no, bool is_static) {
2477   transition(vtos, vtos);
2478   Register Rcache = G3_scratch;
2479   Register index  = G4_scratch;
2480   Register Rclass = Rcache;
2481   Register Roffset= G4_scratch;
2482   Register Rflags = G1_scratch;
2483   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2484 
2485   resolve_cache_and_index(byte_no, Rcache, index);
2486   jvmti_post_field_mod(Rcache, index, is_static);
2487   load_field_cp_cache_entry(Rclass, Rcache, index, Roffset, Rflags, is_static);
2488 
2489   Assembler::Membar_mask_bits read_bits =
2490     Assembler::Membar_mask_bits(Assembler::LoadStore | Assembler::StoreStore);
2491   Assembler::Membar_mask_bits write_bits = Assembler::StoreLoad;
2492 
2493   Label notVolatile, checkVolatile, exit;
2494   if (__ membar_has_effect(read_bits) || __ membar_has_effect(write_bits)) {
2495     __ set((1 << ConstantPoolCacheEntry::volatileField), Lscratch);
2496     __ and3(Rflags, Lscratch, Lscratch);
2497 
2498     if (__ membar_has_effect(read_bits)) {
2499       __ tst(Lscratch);
2500       __ br(Assembler::zero, false, Assembler::pt, notVolatile);
2501       __ delayed()->nop();
2502       volatile_barrier(read_bits);
2503       __ bind(notVolatile);
2504     }
2505   }
2506 
2507   __ srl(Rflags, ConstantPoolCacheEntry::tosBits, Rflags);
2508   // Make sure we don't need to mask Rflags for tosBits after the above shift
2509   ConstantPoolCacheEntry::verify_tosBits();
2510 
2511   // compute field type
2512   Label notInt, notShort, notChar, notObj, notByte, notLong, notFloat;
2513 
2514   if (is_static) {
2515     // putstatic with object type most likely, check that first
2516     __ cmp(Rflags, atos );
2517     __ br(Assembler::notEqual, false, Assembler::pt, notObj);
2518     __ delayed() ->cmp(Rflags, itos );
2519 
2520     // atos
2521     __ pop_ptr();
2522     __ verify_oop(Otos_i);
2523 
2524     do_oop_store(_masm, Rclass, Roffset, 0, Otos_i, G1_scratch, _bs->kind(), false);
2525 
2526     __ ba(false, checkVolatile);
2527     __ delayed()->tst(Lscratch);
2528 
2529     __ bind(notObj);
2530 
2531     // cmp(Rflags, itos );
2532     __ br(Assembler::notEqual, false, Assembler::pt, notInt);
2533     __ delayed() ->cmp(Rflags, btos );
2534 
2535     // itos
2536     __ pop_i();
2537     __ st(Otos_i, Rclass, Roffset);
2538     __ ba(false, checkVolatile);
2539     __ delayed()->tst(Lscratch);
2540 
2541     __ bind(notInt);
2542 
2543   } else {
2544     // putfield with int type most likely, check that first
2545     __ cmp(Rflags, itos );
2546     __ br(Assembler::notEqual, false, Assembler::pt, notInt);
2547     __ delayed() ->cmp(Rflags, atos );
2548 
2549     // itos
2550     __ pop_i();
2551     pop_and_check_object(Rclass);
2552     __ st(Otos_i, Rclass, Roffset);
2553     patch_bytecode(Bytecodes::_fast_iputfield, G3_scratch, G4_scratch);
2554     __ ba(false, checkVolatile);
2555     __ delayed()->tst(Lscratch);
2556 
2557     __ bind(notInt);
2558     // cmp(Rflags, atos );
2559     __ br(Assembler::notEqual, false, Assembler::pt, notObj);
2560     __ delayed() ->cmp(Rflags, btos );
2561 
2562     // atos
2563     __ pop_ptr();
2564     pop_and_check_object(Rclass);
2565     __ verify_oop(Otos_i);
2566 
2567     do_oop_store(_masm, Rclass, Roffset, 0, Otos_i, G1_scratch, _bs->kind(), false);
2568 
2569     patch_bytecode(Bytecodes::_fast_aputfield, G3_scratch, G4_scratch);
2570     __ ba(false, checkVolatile);
2571     __ delayed()->tst(Lscratch);
2572 
2573     __ bind(notObj);
2574   }
2575 
2576   // cmp(Rflags, btos );
2577   __ br(Assembler::notEqual, false, Assembler::pt, notByte);
2578   __ delayed() ->cmp(Rflags, ltos );
2579 
2580   // btos
2581   __ pop_i();
2582   if (!is_static) pop_and_check_object(Rclass);
2583   __ stb(Otos_i, Rclass, Roffset);
2584   if (!is_static) {
2585     patch_bytecode(Bytecodes::_fast_bputfield, G3_scratch, G4_scratch);
2586   }
2587   __ ba(false, checkVolatile);
2588   __ delayed()->tst(Lscratch);
2589 
2590   __ bind(notByte);
2591 
2592   // cmp(Rflags, ltos );
2593   __ br(Assembler::notEqual, false, Assembler::pt, notLong);
2594   __ delayed() ->cmp(Rflags, ctos );
2595 
2596   // ltos
2597   __ pop_l();
2598   if (!is_static) pop_and_check_object(Rclass);
2599   __ st_long(Otos_l, Rclass, Roffset);
2600   if (!is_static) {
2601     patch_bytecode(Bytecodes::_fast_lputfield, G3_scratch, G4_scratch);
2602   }
2603   __ ba(false, checkVolatile);
2604   __ delayed()->tst(Lscratch);
2605 
2606   __ bind(notLong);
2607 
2608   // cmp(Rflags, ctos );
2609   __ br(Assembler::notEqual, false, Assembler::pt, notChar);
2610   __ delayed() ->cmp(Rflags, stos );
2611 
2612   // ctos (char)
2613   __ pop_i();
2614   if (!is_static) pop_and_check_object(Rclass);
2615   __ sth(Otos_i, Rclass, Roffset);
2616   if (!is_static) {
2617     patch_bytecode(Bytecodes::_fast_cputfield, G3_scratch, G4_scratch);
2618   }
2619   __ ba(false, checkVolatile);
2620   __ delayed()->tst(Lscratch);
2621 
2622   __ bind(notChar);
2623   // cmp(Rflags, stos );
2624   __ br(Assembler::notEqual, false, Assembler::pt, notShort);
2625   __ delayed() ->cmp(Rflags, ftos );
2626 
2627   // stos (char)
2628   __ pop_i();
2629   if (!is_static) pop_and_check_object(Rclass);
2630   __ sth(Otos_i, Rclass, Roffset);
2631   if (!is_static) {
2632     patch_bytecode(Bytecodes::_fast_sputfield, G3_scratch, G4_scratch);
2633   }
2634   __ ba(false, checkVolatile);
2635   __ delayed()->tst(Lscratch);
2636 
2637   __ bind(notShort);
2638   // cmp(Rflags, ftos );
2639   __ br(Assembler::notZero, false, Assembler::pt, notFloat);
2640   __ delayed()->nop();
2641 
2642   // ftos
2643   __ pop_f();
2644   if (!is_static) pop_and_check_object(Rclass);
2645   __ stf(FloatRegisterImpl::S, Ftos_f, Rclass, Roffset);
2646   if (!is_static) {
2647     patch_bytecode(Bytecodes::_fast_fputfield, G3_scratch, G4_scratch);
2648   }
2649   __ ba(false, checkVolatile);
2650   __ delayed()->tst(Lscratch);
2651 
2652   __ bind(notFloat);
2653 
2654   // dtos
2655   __ pop_d();
2656   if (!is_static) pop_and_check_object(Rclass);
2657   __ stf(FloatRegisterImpl::D, Ftos_d, Rclass, Roffset);
2658   if (!is_static) {
2659     patch_bytecode(Bytecodes::_fast_dputfield, G3_scratch, G4_scratch);
2660   }
2661 
2662   __ bind(checkVolatile);
2663   __ tst(Lscratch);
2664 
2665   if (__ membar_has_effect(write_bits)) {
2666     // __ tst(Lscratch); in delay slot
2667     __ br(Assembler::zero, false, Assembler::pt, exit);
2668     __ delayed()->nop();
2669     volatile_barrier(Assembler::StoreLoad);
2670     __ bind(exit);
2671   }
2672 }
2673 
2674 void TemplateTable::fast_storefield(TosState state) {
2675   transition(state, vtos);
2676   Register Rcache = G3_scratch;
2677   Register Rclass = Rcache;
2678   Register Roffset= G4_scratch;
2679   Register Rflags = G1_scratch;
2680   ByteSize cp_base_offset = constantPoolCacheOopDesc::base_offset();
2681 
2682   jvmti_post_fast_field_mod();
2683 
2684   __ get_cache_and_index_at_bcp(Rcache, G4_scratch, 1);
2685 
2686   Assembler::Membar_mask_bits read_bits =
2687     Assembler::Membar_mask_bits(Assembler::LoadStore | Assembler::StoreStore);
2688   Assembler::Membar_mask_bits write_bits = Assembler::StoreLoad;
2689 
2690   Label notVolatile, checkVolatile, exit;
2691   if (__ membar_has_effect(read_bits) || __ membar_has_effect(write_bits)) {
2692     __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset +
2693                              ConstantPoolCacheEntry::flags_offset())), Rflags);
2694     __ set((1 << ConstantPoolCacheEntry::volatileField), Lscratch);
2695     __ and3(Rflags, Lscratch, Lscratch);
2696     if (__ membar_has_effect(read_bits)) {
2697       __ tst(Lscratch);
2698       __ br(Assembler::zero, false, Assembler::pt, notVolatile);
2699       __ delayed()->nop();
2700       volatile_barrier(read_bits);
2701       __ bind(notVolatile);
2702     }
2703   }
2704 
2705   __ ld_ptr(Address(Rcache, 0, in_bytes(cp_base_offset +
2706                              ConstantPoolCacheEntry::f2_offset())), Roffset);
2707   pop_and_check_object(Rclass);
2708 
2709   switch (bytecode()) {
2710     case Bytecodes::_fast_bputfield: __ stb(Otos_i, Rclass, Roffset); break;
2711     case Bytecodes::_fast_cputfield: /* fall through */
2712     case Bytecodes::_fast_sputfield: __ sth(Otos_i, Rclass, Roffset); break;
2713     case Bytecodes::_fast_iputfield: __ st(Otos_i, Rclass, Roffset);  break;
2714     case Bytecodes::_fast_lputfield: __ st_long(Otos_l, Rclass, Roffset); break;
2715     case Bytecodes::_fast_fputfield:
2716       __ stf(FloatRegisterImpl::S, Ftos_f, Rclass, Roffset);
2717       break;
2718     case Bytecodes::_fast_dputfield:
2719       __ stf(FloatRegisterImpl::D, Ftos_d, Rclass, Roffset);
2720       break;
2721     case Bytecodes::_fast_aputfield:
2722       do_oop_store(_masm, Rclass, Roffset, 0, Otos_i, G1_scratch, _bs->kind(), false);
2723       break;
2724     default:
2725       ShouldNotReachHere();
2726   }
2727 
2728   if (__ membar_has_effect(write_bits)) {
2729     __ tst(Lscratch);
2730     __ br(Assembler::zero, false, Assembler::pt, exit);
2731     __ delayed()->nop();
2732     volatile_barrier(Assembler::StoreLoad);
2733     __ bind(exit);
2734   }
2735 }
2736 
2737 
2738 void TemplateTable::putfield(int byte_no) {
2739   putfield_or_static(byte_no, false);
2740 }
2741 
2742 void TemplateTable::putstatic(int byte_no) {
2743   putfield_or_static(byte_no, true);
2744 }
2745 
2746 
2747 void TemplateTable::fast_xaccess(TosState state) {
2748   transition(vtos, state);
2749   Register Rcache = G3_scratch;
2750   Register Roffset = G4_scratch;
2751   Register Rflags  = G4_scratch;
2752   Register Rreceiver = Lscratch;
2753 
2754   __ ld_ptr(Llocals, Interpreter::value_offset_in_bytes(), Rreceiver);
2755 
2756   // access constant pool cache  (is resolved)
2757   __ get_cache_and_index_at_bcp(Rcache, G4_scratch, 2);
2758   __ ld_ptr(Address(Rcache, 0, in_bytes(constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::f2_offset())), Roffset);
2759   __ add(Lbcp, 1, Lbcp);       // needed to report exception at the correct bcp
2760 
2761   __ verify_oop(Rreceiver);
2762   __ null_check(Rreceiver);
2763   if (state == atos) {
2764     __ load_heap_oop(Rreceiver, Roffset, Otos_i);
2765   } else if (state == itos) {
2766     __ ld (Rreceiver, Roffset, Otos_i) ;
2767   } else if (state == ftos) {
2768     __ ldf(FloatRegisterImpl::S, Rreceiver, Roffset, Ftos_f);
2769   } else {
2770     ShouldNotReachHere();
2771   }
2772 
2773   Assembler::Membar_mask_bits membar_bits =
2774     Assembler::Membar_mask_bits(Assembler::LoadLoad | Assembler::LoadStore);
2775   if (__ membar_has_effect(membar_bits)) {
2776 
2777     // Get is_volatile value in Rflags and check if membar is needed
2778     __ ld_ptr(Address(Rcache, 0, in_bytes(constantPoolCacheOopDesc::base_offset() + ConstantPoolCacheEntry::flags_offset())), Rflags);
2779 
2780     // Test volatile
2781     Label notVolatile;
2782     __ set((1 << ConstantPoolCacheEntry::volatileField), Lscratch);
2783     __ btst(Rflags, Lscratch);
2784     __ br(Assembler::zero, false, Assembler::pt, notVolatile);
2785     __ delayed()->nop();
2786     volatile_barrier(membar_bits);
2787     __ bind(notVolatile);
2788   }
2789 
2790   __ interp_verify_oop(Otos_i, state, __FILE__, __LINE__);
2791   __ sub(Lbcp, 1, Lbcp);
2792 }
2793 
2794 //----------------------------------------------------------------------------------------------------
2795 // Calls
2796 
2797 void TemplateTable::count_calls(Register method, Register temp) {
2798   // implemented elsewhere
2799   ShouldNotReachHere();
2800 }
2801 
2802 void TemplateTable::generate_vtable_call(Register Rrecv, Register Rindex, Register Rret) {
2803   Register Rtemp = G4_scratch;
2804   Register Rcall = Rindex;
2805   assert_different_registers(Rcall, G5_method, Gargs, Rret);
2806 
2807   // get target methodOop & entry point
2808   const int base = instanceKlass::vtable_start_offset() * wordSize;
2809   if (vtableEntry::size() % 3 == 0) {
2810     // scale the vtable index by 12:
2811     int one_third = vtableEntry::size() / 3;
2812     __ sll(Rindex, exact_log2(one_third * 1 * wordSize), Rtemp);
2813     __ sll(Rindex, exact_log2(one_third * 2 * wordSize), Rindex);
2814     __ add(Rindex, Rtemp, Rindex);
2815   } else {
2816     // scale the vtable index by 8:
2817     __ sll(Rindex, exact_log2(vtableEntry::size() * wordSize), Rindex);
2818   }
2819 
2820   __ add(Rrecv, Rindex, Rrecv);
2821   __ ld_ptr(Rrecv, base + vtableEntry::method_offset_in_bytes(), G5_method);
2822 
2823   __ call_from_interpreter(Rcall, Gargs, Rret);
2824 }
2825 
2826 void TemplateTable::invokevirtual(int byte_no) {
2827   transition(vtos, vtos);
2828 
2829   Register Rscratch = G3_scratch;
2830   Register Rtemp = G4_scratch;
2831   Register Rret = Lscratch;
2832   Register Rrecv = G5_method;
2833   Label notFinal;
2834 
2835   load_invoke_cp_cache_entry(byte_no, G5_method, noreg, Rret, true);
2836   __ mov(SP, O5_savedSP); // record SP that we wanted the callee to restore
2837 
2838   // Check for vfinal
2839   __ set((1 << ConstantPoolCacheEntry::vfinalMethod), G4_scratch);
2840   __ btst(Rret, G4_scratch);
2841   __ br(Assembler::zero, false, Assembler::pt, notFinal);
2842   __ delayed()->and3(Rret, 0xFF, G4_scratch);      // gets number of parameters
2843 
2844   patch_bytecode(Bytecodes::_fast_invokevfinal, Rscratch, Rtemp);
2845 
2846   invokevfinal_helper(Rscratch, Rret);
2847 
2848   __ bind(notFinal);
2849 
2850   __ mov(G5_method, Rscratch);  // better scratch register
2851   __ load_receiver(G4_scratch, O0);  // gets receiverOop
2852   // receiver is in O0
2853   __ verify_oop(O0);
2854 
2855   // get return address
2856   Address table(Rtemp, (address)Interpreter::return_3_addrs_by_index_table());
2857   __ load_address(table);
2858   __ srl(Rret, ConstantPoolCacheEntry::tosBits, Rret);          // get return type
2859   // Make sure we don't need to mask Rret for tosBits after the above shift
2860   ConstantPoolCacheEntry::verify_tosBits();
2861   __ sll(Rret,  LogBytesPerWord, Rret);
2862   __ ld_ptr(Rtemp, Rret, Rret);         // get return address
2863 
2864   // get receiver klass
2865   __ null_check(O0, oopDesc::klass_offset_in_bytes());
2866   __ load_klass(O0, Rrecv);
2867   __ verify_oop(Rrecv);
2868 
2869   __ profile_virtual_call(Rrecv, O4);
2870 
2871   generate_vtable_call(Rrecv, Rscratch, Rret);
2872 }
2873 
2874 void TemplateTable::fast_invokevfinal(int byte_no) {
2875   transition(vtos, vtos);
2876 
2877   load_invoke_cp_cache_entry(byte_no, G5_method, noreg, Lscratch, true,
2878                              /*is_invokevfinal*/true);
2879   __ mov(SP, O5_savedSP); // record SP that we wanted the callee to restore
2880   invokevfinal_helper(G3_scratch, Lscratch);
2881 }
2882 
2883 void TemplateTable::invokevfinal_helper(Register Rscratch, Register Rret) {
2884   Register Rtemp = G4_scratch;
2885 
2886   __ verify_oop(G5_method);
2887 
2888   // Load receiver from stack slot
2889   __ lduh(Address(G5_method, 0, in_bytes(methodOopDesc::size_of_parameters_offset())), G4_scratch);
2890   __ load_receiver(G4_scratch, O0);
2891 
2892   // receiver NULL check
2893   __ null_check(O0);
2894 
2895   __ profile_final_call(O4);
2896 
2897   // get return address
2898   Address table(Rtemp, (address)Interpreter::return_3_addrs_by_index_table());
2899   __ load_address(table);
2900   __ srl(Rret, ConstantPoolCacheEntry::tosBits, Rret);          // get return type
2901   // Make sure we don't need to mask Rret for tosBits after the above shift
2902   ConstantPoolCacheEntry::verify_tosBits();
2903   __ sll(Rret,  LogBytesPerWord, Rret);
2904   __ ld_ptr(Rtemp, Rret, Rret);         // get return address
2905 
2906 
2907   // do the call
2908   __ call_from_interpreter(Rscratch, Gargs, Rret);
2909 }
2910 
2911 void TemplateTable::invokespecial(int byte_no) {
2912   transition(vtos, vtos);
2913 
2914   Register Rscratch = G3_scratch;
2915   Register Rtemp = G4_scratch;
2916   Register Rret = Lscratch;
2917 
2918   load_invoke_cp_cache_entry(byte_no, G5_method, noreg, Rret, false);
2919   __ mov(SP, O5_savedSP); // record SP that we wanted the callee to restore
2920 
2921   __ verify_oop(G5_method);
2922 
2923   __ lduh(Address(G5_method, 0, in_bytes(methodOopDesc::size_of_parameters_offset())), G4_scratch);
2924   __ load_receiver(G4_scratch, O0);
2925 
2926   // receiver NULL check
2927   __ null_check(O0);
2928 
2929   __ profile_call(O4);
2930 
2931   // get return address
2932   Address table(Rtemp, (address)Interpreter::return_3_addrs_by_index_table());
2933   __ load_address(table);
2934   __ srl(Rret, ConstantPoolCacheEntry::tosBits, Rret);          // get return type
2935   // Make sure we don't need to mask Rret for tosBits after the above shift
2936   ConstantPoolCacheEntry::verify_tosBits();
2937   __ sll(Rret,  LogBytesPerWord, Rret);
2938   __ ld_ptr(Rtemp, Rret, Rret);         // get return address
2939 
2940   // do the call
2941   __ call_from_interpreter(Rscratch, Gargs, Rret);
2942 }
2943 
2944 void TemplateTable::invokestatic(int byte_no) {
2945   transition(vtos, vtos);
2946 
2947   Register Rscratch = G3_scratch;
2948   Register Rtemp = G4_scratch;
2949   Register Rret = Lscratch;
2950 
2951   load_invoke_cp_cache_entry(byte_no, G5_method, noreg, Rret, false);
2952   __ mov(SP, O5_savedSP); // record SP that we wanted the callee to restore
2953 
2954   __ verify_oop(G5_method);
2955 
2956   __ profile_call(O4);
2957 
2958   // get return address
2959   Address table(Rtemp, (address)Interpreter::return_3_addrs_by_index_table());
2960   __ load_address(table);
2961   __ srl(Rret, ConstantPoolCacheEntry::tosBits, Rret);          // get return type
2962   // Make sure we don't need to mask Rret for tosBits after the above shift
2963   ConstantPoolCacheEntry::verify_tosBits();
2964   __ sll(Rret,  LogBytesPerWord, Rret);
2965   __ ld_ptr(Rtemp, Rret, Rret);         // get return address
2966 
2967   // do the call
2968   __ call_from_interpreter(Rscratch, Gargs, Rret);
2969 }
2970 
2971 
2972 void TemplateTable::invokeinterface_object_method(Register RklassOop,
2973                                                   Register Rcall,
2974                                                   Register Rret,
2975                                                   Register Rflags) {
2976   Register Rscratch = G4_scratch;
2977   Register Rindex = Lscratch;
2978 
2979   assert_different_registers(Rscratch, Rindex, Rret);
2980 
2981   Label notFinal;
2982 
2983   // Check for vfinal
2984   __ set((1 << ConstantPoolCacheEntry::vfinalMethod), Rscratch);
2985   __ btst(Rflags, Rscratch);
2986   __ br(Assembler::zero, false, Assembler::pt, notFinal);
2987   __ delayed()->nop();
2988 
2989   __ profile_final_call(O4);
2990 
2991   // do the call - the index (f2) contains the methodOop
2992   assert_different_registers(G5_method, Gargs, Rcall);
2993   __ mov(Rindex, G5_method);
2994   __ call_from_interpreter(Rcall, Gargs, Rret);
2995   __ bind(notFinal);
2996 
2997   __ profile_virtual_call(RklassOop, O4);
2998   generate_vtable_call(RklassOop, Rindex, Rret);
2999 }
3000 
3001 
3002 void TemplateTable::invokeinterface(int byte_no) {
3003   transition(vtos, vtos);
3004 
3005   Register Rscratch = G4_scratch;
3006   Register Rret = G3_scratch;
3007   Register Rindex = Lscratch;
3008   Register Rinterface = G1_scratch;
3009   Register RklassOop = G5_method;
3010   Register Rflags = O1;
3011   assert_different_registers(Rscratch, G5_method);
3012 
3013   load_invoke_cp_cache_entry(byte_no, Rinterface, Rindex, Rflags, false);
3014   __ mov(SP, O5_savedSP); // record SP that we wanted the callee to restore
3015 
3016   // get receiver
3017   __ and3(Rflags, 0xFF, Rscratch);       // gets number of parameters
3018   __ load_receiver(Rscratch, O0);
3019   __ verify_oop(O0);
3020 
3021   __ mov(Rflags, Rret);
3022 
3023   // get return address
3024   Address table(Rscratch, (address)Interpreter::return_5_addrs_by_index_table());
3025   __ load_address(table);
3026   __ srl(Rret, ConstantPoolCacheEntry::tosBits, Rret);          // get return type
3027   // Make sure we don't need to mask Rret for tosBits after the above shift
3028   ConstantPoolCacheEntry::verify_tosBits();
3029   __ sll(Rret,  LogBytesPerWord, Rret);
3030   __ ld_ptr(Rscratch, Rret, Rret);      // get return address
3031 
3032   // get receiver klass
3033   __ null_check(O0, oopDesc::klass_offset_in_bytes());
3034   __ load_klass(O0, RklassOop);
3035   __ verify_oop(RklassOop);
3036 
3037   // Special case of invokeinterface called for virtual method of
3038   // java.lang.Object.  See cpCacheOop.cpp for details.
3039   // This code isn't produced by javac, but could be produced by
3040   // another compliant java compiler.
3041   Label notMethod;
3042   __ set((1 << ConstantPoolCacheEntry::methodInterface), Rscratch);
3043   __ btst(Rflags, Rscratch);
3044   __ br(Assembler::zero, false, Assembler::pt, notMethod);
3045   __ delayed()->nop();
3046 
3047   invokeinterface_object_method(RklassOop, Rinterface, Rret, Rflags);
3048 
3049   __ bind(notMethod);
3050 
3051   __ profile_virtual_call(RklassOop, O4);
3052 
3053   //
3054   // find entry point to call
3055   //
3056 
3057   // compute start of first itableOffsetEntry (which is at end of vtable)
3058   const int base = instanceKlass::vtable_start_offset() * wordSize;
3059   Label search;
3060   Register Rtemp = Rflags;
3061 
3062   __ ld(Address(RklassOop, 0, instanceKlass::vtable_length_offset() * wordSize), Rtemp);
3063   if (align_object_offset(1) > 1) {
3064     __ round_to(Rtemp, align_object_offset(1));
3065   }
3066   __ sll(Rtemp, LogBytesPerWord, Rtemp);   // Rscratch *= 4;
3067   if (Assembler::is_simm13(base)) {
3068     __ add(Rtemp, base, Rtemp);
3069   } else {
3070     __ set(base, Rscratch);
3071     __ add(Rscratch, Rtemp, Rtemp);
3072   }
3073   __ add(RklassOop, Rtemp, Rscratch);
3074 
3075   __ bind(search);
3076 
3077   __ ld_ptr(Rscratch, itableOffsetEntry::interface_offset_in_bytes(), Rtemp);
3078   {
3079     Label ok;
3080 
3081     // Check that entry is non-null.  Null entries are probably a bytecode
3082     // problem.  If the interface isn't implemented by the reciever class,
3083     // the VM should throw IncompatibleClassChangeError.  linkResolver checks
3084     // this too but that's only if the entry isn't already resolved, so we
3085     // need to check again.
3086     __ br_notnull( Rtemp, false, Assembler::pt, ok);
3087     __ delayed()->nop();
3088     call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_IncompatibleClassChangeError));
3089     __ should_not_reach_here();
3090     __ bind(ok);
3091     __ verify_oop(Rtemp);
3092   }
3093 
3094   __ verify_oop(Rinterface);
3095 
3096   __ cmp(Rinterface, Rtemp);
3097   __ brx(Assembler::notEqual, true, Assembler::pn, search);
3098   __ delayed()->add(Rscratch, itableOffsetEntry::size() * wordSize, Rscratch);
3099 
3100   // entry found and Rscratch points to it
3101   __ ld(Rscratch, itableOffsetEntry::offset_offset_in_bytes(), Rscratch);
3102 
3103   assert(itableMethodEntry::method_offset_in_bytes() == 0, "adjust instruction below");
3104   __ sll(Rindex, exact_log2(itableMethodEntry::size() * wordSize), Rindex);       // Rindex *= 8;
3105   __ add(Rscratch, Rindex, Rscratch);
3106   __ ld_ptr(RklassOop, Rscratch, G5_method);
3107 
3108   // Check for abstract method error.
3109   {
3110     Label ok;
3111     __ tst(G5_method);
3112     __ brx(Assembler::notZero, false, Assembler::pt, ok);
3113     __ delayed()->nop();
3114     call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));
3115     __ should_not_reach_here();
3116     __ bind(ok);
3117   }
3118 
3119   Register Rcall = Rinterface;
3120   assert_different_registers(Rcall, G5_method, Gargs, Rret);
3121 
3122   __ verify_oop(G5_method);
3123   __ call_from_interpreter(Rcall, Gargs, Rret);
3124 
3125 }
3126 
3127 
3128 //----------------------------------------------------------------------------------------------------
3129 // Allocation
3130 
3131 void TemplateTable::_new() {
3132   transition(vtos, atos);
3133 
3134   Label slow_case;
3135   Label done;
3136   Label initialize_header;
3137   Label initialize_object;  // including clearing the fields
3138 
3139   Register RallocatedObject = Otos_i;
3140   Register RinstanceKlass = O1;
3141   Register Roffset = O3;
3142   Register Rscratch = O4;
3143 
3144   __ get_2_byte_integer_at_bcp(1, Rscratch, Roffset, InterpreterMacroAssembler::Unsigned);
3145   __ get_cpool_and_tags(Rscratch, G3_scratch);
3146   // make sure the class we're about to instantiate has been resolved
3147   __ add(G3_scratch, typeArrayOopDesc::header_size(T_BYTE) * wordSize, G3_scratch);
3148   __ ldub(G3_scratch, Roffset, G3_scratch);
3149   __ cmp(G3_scratch, JVM_CONSTANT_Class);
3150   __ br(Assembler::notEqual, false, Assembler::pn, slow_case);
3151   __ delayed()->sll(Roffset, LogBytesPerWord, Roffset);
3152 
3153   //__ sll(Roffset, LogBytesPerWord, Roffset);        // executed in delay slot
3154   __ add(Roffset, sizeof(constantPoolOopDesc), Roffset);
3155   __ ld_ptr(Rscratch, Roffset, RinstanceKlass);
3156 
3157   // make sure klass is fully initialized:
3158   __ ld(RinstanceKlass, instanceKlass::init_state_offset_in_bytes() + sizeof(oopDesc), G3_scratch);
3159   __ cmp(G3_scratch, instanceKlass::fully_initialized);
3160   __ br(Assembler::notEqual, false, Assembler::pn, slow_case);
3161   __ delayed()->ld(RinstanceKlass, Klass::layout_helper_offset_in_bytes() + sizeof(oopDesc), Roffset);
3162 
3163   // get instance_size in instanceKlass (already aligned)
3164   //__ ld(RinstanceKlass, Klass::layout_helper_offset_in_bytes() + sizeof(oopDesc), Roffset);
3165 
3166   // make sure klass does not have has_finalizer, or is abstract, or interface or java/lang/Class
3167   __ btst(Klass::_lh_instance_slow_path_bit, Roffset);
3168   __ br(Assembler::notZero, false, Assembler::pn, slow_case);
3169   __ delayed()->nop();
3170 
3171   // allocate the instance
3172   // 1) Try to allocate in the TLAB
3173   // 2) if fail, and the TLAB is not full enough to discard, allocate in the shared Eden
3174   // 3) if the above fails (or is not applicable), go to a slow case
3175   // (creates a new TLAB, etc.)
3176 
3177   const bool allow_shared_alloc =
3178     Universe::heap()->supports_inline_contig_alloc() && !CMSIncrementalMode;
3179 
3180   if(UseTLAB) {
3181     Register RoldTopValue = RallocatedObject;
3182     Register RtopAddr = G3_scratch, RtlabWasteLimitValue = G3_scratch;
3183     Register RnewTopValue = G1_scratch;
3184     Register RendValue = Rscratch;
3185     Register RfreeValue = RnewTopValue;
3186 
3187     // check if we can allocate in the TLAB
3188     __ ld_ptr(G2_thread, in_bytes(JavaThread::tlab_top_offset()), RoldTopValue); // sets up RalocatedObject
3189     __ ld_ptr(G2_thread, in_bytes(JavaThread::tlab_end_offset()), RendValue);
3190     __ add(RoldTopValue, Roffset, RnewTopValue);
3191 
3192     // if there is enough space, we do not CAS and do not clear
3193     __ cmp(RnewTopValue, RendValue);
3194     if(ZeroTLAB) {
3195       // the fields have already been cleared
3196       __ brx(Assembler::lessEqualUnsigned, true, Assembler::pt, initialize_header);
3197     } else {
3198       // initialize both the header and fields
3199       __ brx(Assembler::lessEqualUnsigned, true, Assembler::pt, initialize_object);
3200     }
3201     __ delayed()->st_ptr(RnewTopValue, G2_thread, in_bytes(JavaThread::tlab_top_offset()));
3202 
3203     if (allow_shared_alloc) {
3204     // Check if tlab should be discarded (refill_waste_limit >= free)
3205     __ ld_ptr(G2_thread, in_bytes(JavaThread::tlab_refill_waste_limit_offset()), RtlabWasteLimitValue);
3206     __ sub(RendValue, RoldTopValue, RfreeValue);
3207 #ifdef _LP64
3208     __ srlx(RfreeValue, LogHeapWordSize, RfreeValue);
3209 #else
3210     __ srl(RfreeValue, LogHeapWordSize, RfreeValue);
3211 #endif
3212     __ cmp(RtlabWasteLimitValue, RfreeValue);
3213     __ brx(Assembler::greaterEqualUnsigned, false, Assembler::pt, slow_case); // tlab waste is small
3214     __ delayed()->nop();
3215 
3216     // increment waste limit to prevent getting stuck on this slow path
3217     __ add(RtlabWasteLimitValue, ThreadLocalAllocBuffer::refill_waste_limit_increment(), RtlabWasteLimitValue);
3218     __ st_ptr(RtlabWasteLimitValue, G2_thread, in_bytes(JavaThread::tlab_refill_waste_limit_offset()));
3219     } else {
3220       // No allocation in the shared eden.
3221       __ br(Assembler::always, false, Assembler::pt, slow_case);
3222       __ delayed()->nop();
3223     }
3224   }
3225 
3226   // Allocation in the shared Eden
3227   if (allow_shared_alloc) {
3228     Register RoldTopValue = G1_scratch;
3229     Register RtopAddr = G3_scratch;
3230     Register RnewTopValue = RallocatedObject;
3231     Register RendValue = Rscratch;
3232 
3233     __ set((intptr_t)Universe::heap()->top_addr(), RtopAddr);
3234 
3235     Label retry;
3236     __ bind(retry);
3237     __ set((intptr_t)Universe::heap()->end_addr(), RendValue);
3238     __ ld_ptr(RendValue, 0, RendValue);
3239     __ ld_ptr(RtopAddr, 0, RoldTopValue);
3240     __ add(RoldTopValue, Roffset, RnewTopValue);
3241 
3242     // RnewTopValue contains the top address after the new object
3243     // has been allocated.
3244     __ cmp(RnewTopValue, RendValue);
3245     __ brx(Assembler::greaterUnsigned, false, Assembler::pn, slow_case);
3246     __ delayed()->nop();
3247 
3248     __ casx_under_lock(RtopAddr, RoldTopValue, RnewTopValue,
3249       VM_Version::v9_instructions_work() ? NULL :
3250       (address)StubRoutines::Sparc::atomic_memory_operation_lock_addr());
3251 
3252     // if someone beat us on the allocation, try again, otherwise continue
3253     __ cmp(RoldTopValue, RnewTopValue);
3254     __ brx(Assembler::notEqual, false, Assembler::pn, retry);
3255     __ delayed()->nop();
3256   }
3257 
3258   if (UseTLAB || Universe::heap()->supports_inline_contig_alloc()) {
3259     // clear object fields
3260     __ bind(initialize_object);
3261     __ deccc(Roffset, sizeof(oopDesc));
3262     __ br(Assembler::zero, false, Assembler::pt, initialize_header);
3263     __ delayed()->add(RallocatedObject, sizeof(oopDesc), G3_scratch);
3264 
3265     // initialize remaining object fields
3266     { Label loop;
3267       __ subcc(Roffset, wordSize, Roffset);
3268       __ bind(loop);
3269       //__ subcc(Roffset, wordSize, Roffset);      // executed above loop or in delay slot
3270       __ st_ptr(G0, G3_scratch, Roffset);
3271       __ br(Assembler::notEqual, false, Assembler::pt, loop);
3272       __ delayed()->subcc(Roffset, wordSize, Roffset);
3273     }
3274     __ br(Assembler::always, false, Assembler::pt, initialize_header);
3275     __ delayed()->nop();
3276   }
3277 
3278   // slow case
3279   __ bind(slow_case);
3280   __ get_2_byte_integer_at_bcp(1, G3_scratch, O2, InterpreterMacroAssembler::Unsigned);
3281   __ get_constant_pool(O1);
3282 
3283   call_VM(Otos_i, CAST_FROM_FN_PTR(address, InterpreterRuntime::_new), O1, O2);
3284 
3285   __ ba(false, done);
3286   __ delayed()->nop();
3287 
3288   // Initialize the header: mark, klass
3289   __ bind(initialize_header);
3290 
3291   if (UseBiasedLocking) {
3292     __ ld_ptr(RinstanceKlass, Klass::prototype_header_offset_in_bytes() + sizeof(oopDesc), G4_scratch);
3293   } else {
3294     __ set((intptr_t)markOopDesc::prototype(), G4_scratch);
3295   }
3296   __ st_ptr(G4_scratch, RallocatedObject, oopDesc::mark_offset_in_bytes());       // mark
3297   __ store_klass_gap(G0, RallocatedObject);         // klass gap if compressed
3298   __ store_klass(RinstanceKlass, RallocatedObject); // klass (last for cms)
3299 
3300   {
3301     SkipIfEqual skip_if(
3302       _masm, G4_scratch, &DTraceAllocProbes, Assembler::zero);
3303     // Trigger dtrace event
3304     __ push(atos);
3305     __ call_VM_leaf(noreg,
3306        CAST_FROM_FN_PTR(address, SharedRuntime::dtrace_object_alloc), O0);
3307     __ pop(atos);
3308   }
3309 
3310   // continue
3311   __ bind(done);
3312 }
3313 
3314 
3315 
3316 void TemplateTable::newarray() {
3317   transition(itos, atos);
3318   __ ldub(Lbcp, 1, O1);
3319      call_VM(Otos_i, CAST_FROM_FN_PTR(address, InterpreterRuntime::newarray), O1, Otos_i);
3320 }
3321 
3322 
3323 void TemplateTable::anewarray() {
3324   transition(itos, atos);
3325   __ get_constant_pool(O1);
3326   __ get_2_byte_integer_at_bcp(1, G4_scratch, O2, InterpreterMacroAssembler::Unsigned);
3327      call_VM(Otos_i, CAST_FROM_FN_PTR(address, InterpreterRuntime::anewarray), O1, O2, Otos_i);
3328 }
3329 
3330 
3331 void TemplateTable::arraylength() {
3332   transition(atos, itos);
3333   Label ok;
3334   __ verify_oop(Otos_i);
3335   __ tst(Otos_i);
3336   __ throw_if_not_1_x( Assembler::notZero, ok );
3337   __ delayed()->ld(Otos_i, arrayOopDesc::length_offset_in_bytes(), Otos_i);
3338   __ throw_if_not_2( Interpreter::_throw_NullPointerException_entry, G3_scratch, ok);
3339 }
3340 
3341 
3342 void TemplateTable::checkcast() {
3343   transition(atos, atos);
3344   Label done, is_null, quicked, cast_ok, resolved;
3345   Register Roffset = G1_scratch;
3346   Register RobjKlass = O5;
3347   Register RspecifiedKlass = O4;
3348 
3349   // Check for casting a NULL
3350   __ br_null(Otos_i, false, Assembler::pn, is_null);
3351   __ delayed()->nop();
3352 
3353   // Get value klass in RobjKlass
3354   __ load_klass(Otos_i, RobjKlass); // get value klass
3355 
3356   // Get constant pool tag
3357   __ get_2_byte_integer_at_bcp(1, Lscratch, Roffset, InterpreterMacroAssembler::Unsigned);
3358 
3359   // See if the checkcast has been quickened
3360   __ get_cpool_and_tags(Lscratch, G3_scratch);
3361   __ add(G3_scratch, typeArrayOopDesc::header_size(T_BYTE) * wordSize, G3_scratch);
3362   __ ldub(G3_scratch, Roffset, G3_scratch);
3363   __ cmp(G3_scratch, JVM_CONSTANT_Class);
3364   __ br(Assembler::equal, true, Assembler::pt, quicked);
3365   __ delayed()->sll(Roffset, LogBytesPerWord, Roffset);
3366 
3367   __ push_ptr(); // save receiver for result, and for GC
3368   call_VM(RspecifiedKlass, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3369   __ pop_ptr(Otos_i, G3_scratch); // restore receiver
3370 
3371   __ br(Assembler::always, false, Assembler::pt, resolved);
3372   __ delayed()->nop();
3373 
3374   // Extract target class from constant pool
3375   __ bind(quicked);
3376   __ add(Roffset, sizeof(constantPoolOopDesc), Roffset);
3377   __ ld_ptr(Lscratch, Roffset, RspecifiedKlass);
3378   __ bind(resolved);
3379   __ load_klass(Otos_i, RobjKlass); // get value klass
3380 
3381   // Generate a fast subtype check.  Branch to cast_ok if no
3382   // failure.  Throw exception if failure.
3383   __ gen_subtype_check( RobjKlass, RspecifiedKlass, G3_scratch, G4_scratch, G1_scratch, cast_ok );
3384 
3385   // Not a subtype; so must throw exception
3386   __ throw_if_not_x( Assembler::never, Interpreter::_throw_ClassCastException_entry, G3_scratch );
3387 
3388   __ bind(cast_ok);
3389 
3390   if (ProfileInterpreter) {
3391     __ ba(false, done);
3392     __ delayed()->nop();
3393   }
3394   __ bind(is_null);
3395   __ profile_null_seen(G3_scratch);
3396   __ bind(done);
3397 }
3398 
3399 
3400 void TemplateTable::instanceof() {
3401   Label done, is_null, quicked, resolved;
3402   transition(atos, itos);
3403   Register Roffset = G1_scratch;
3404   Register RobjKlass = O5;
3405   Register RspecifiedKlass = O4;
3406 
3407   // Check for casting a NULL
3408   __ br_null(Otos_i, false, Assembler::pt, is_null);
3409   __ delayed()->nop();
3410 
3411   // Get value klass in RobjKlass
3412   __ load_klass(Otos_i, RobjKlass); // get value klass
3413 
3414   // Get constant pool tag
3415   __ get_2_byte_integer_at_bcp(1, Lscratch, Roffset, InterpreterMacroAssembler::Unsigned);
3416 
3417   // See if the checkcast has been quickened
3418   __ get_cpool_and_tags(Lscratch, G3_scratch);
3419   __ add(G3_scratch, typeArrayOopDesc::header_size(T_BYTE) * wordSize, G3_scratch);
3420   __ ldub(G3_scratch, Roffset, G3_scratch);
3421   __ cmp(G3_scratch, JVM_CONSTANT_Class);
3422   __ br(Assembler::equal, true, Assembler::pt, quicked);
3423   __ delayed()->sll(Roffset, LogBytesPerWord, Roffset);
3424 
3425   __ push_ptr(); // save receiver for result, and for GC
3426   call_VM(RspecifiedKlass, CAST_FROM_FN_PTR(address, InterpreterRuntime::quicken_io_cc) );
3427   __ pop_ptr(Otos_i, G3_scratch); // restore receiver
3428 
3429   __ br(Assembler::always, false, Assembler::pt, resolved);
3430   __ delayed()->nop();
3431 
3432 
3433   // Extract target class from constant pool
3434   __ bind(quicked);
3435   __ add(Roffset, sizeof(constantPoolOopDesc), Roffset);
3436   __ get_constant_pool(Lscratch);
3437   __ ld_ptr(Lscratch, Roffset, RspecifiedKlass);
3438   __ bind(resolved);
3439   __ load_klass(Otos_i, RobjKlass); // get value klass
3440 
3441   // Generate a fast subtype check.  Branch to cast_ok if no
3442   // failure.  Return 0 if failure.
3443   __ or3(G0, 1, Otos_i);      // set result assuming quick tests succeed
3444   __ gen_subtype_check( RobjKlass, RspecifiedKlass, G3_scratch, G4_scratch, G1_scratch, done );
3445   // Not a subtype; return 0;
3446   __ clr( Otos_i );
3447 
3448   if (ProfileInterpreter) {
3449     __ ba(false, done);
3450     __ delayed()->nop();
3451   }
3452   __ bind(is_null);
3453   __ profile_null_seen(G3_scratch);
3454   __ bind(done);
3455 }
3456 
3457 void TemplateTable::_breakpoint() {
3458 
3459    // Note: We get here even if we are single stepping..
3460    // jbug inists on setting breakpoints at every bytecode
3461    // even if we are in single step mode.
3462 
3463    transition(vtos, vtos);
3464    // get the unpatched byte code
3465    __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::get_original_bytecode_at), Lmethod, Lbcp);
3466    __ mov(O0, Lbyte_code);
3467 
3468    // post the breakpoint event
3469    __ call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::_breakpoint), Lmethod, Lbcp);
3470 
3471    // complete the execution of original bytecode
3472    __ dispatch_normal(vtos);
3473 }
3474 
3475 
3476 //----------------------------------------------------------------------------------------------------
3477 // Exceptions
3478 
3479 void TemplateTable::athrow() {
3480   transition(atos, vtos);
3481 
3482   // This works because exception is cached in Otos_i which is same as O0,
3483   // which is same as what throw_exception_entry_expects
3484   assert(Otos_i == Oexception, "see explanation above");
3485 
3486   __ verify_oop(Otos_i);
3487   __ null_check(Otos_i);
3488   __ throw_if_not_x(Assembler::never, Interpreter::throw_exception_entry(), G3_scratch);
3489 }
3490 
3491 
3492 //----------------------------------------------------------------------------------------------------
3493 // Synchronization
3494 
3495 
3496 // See frame_sparc.hpp for monitor block layout.
3497 // Monitor elements are dynamically allocated by growing stack as needed.
3498 
3499 void TemplateTable::monitorenter() {
3500   transition(atos, vtos);
3501   __ verify_oop(Otos_i);
3502   // Try to acquire a lock on the object
3503   // Repeat until succeeded (i.e., until
3504   // monitorenter returns true).
3505 
3506   {   Label ok;
3507     __ tst(Otos_i);
3508     __ throw_if_not_1_x( Assembler::notZero,  ok);
3509     __ delayed()->mov(Otos_i, Lscratch); // save obj
3510     __ throw_if_not_2( Interpreter::_throw_NullPointerException_entry, G3_scratch, ok);
3511   }
3512 
3513   assert(O0 == Otos_i, "Be sure where the object to lock is");
3514 
3515   // find a free slot in the monitor block
3516 
3517 
3518   // initialize entry pointer
3519   __ clr(O1); // points to free slot or NULL
3520 
3521   {
3522     Label entry, loop, exit;
3523     __ add( __ top_most_monitor(), O2 ); // last one to check
3524     __ ba( false, entry );
3525     __ delayed()->mov( Lmonitors, O3 ); // first one to check
3526 
3527 
3528     __ bind( loop );
3529 
3530     __ verify_oop(O4);          // verify each monitor's oop
3531     __ tst(O4); // is this entry unused?
3532     if (VM_Version::v9_instructions_work())
3533       __ movcc( Assembler::zero, false, Assembler::ptr_cc, O3, O1);
3534     else {
3535       Label L;
3536       __ br( Assembler::zero, true, Assembler::pn, L );
3537       __ delayed()->mov(O3, O1); // rememeber this one if match
3538       __ bind(L);
3539     }
3540 
3541     __ cmp(O4, O0); // check if current entry is for same object
3542     __ brx( Assembler::equal, false, Assembler::pn, exit );
3543     __ delayed()->inc( O3, frame::interpreter_frame_monitor_size() * wordSize ); // check next one
3544 
3545     __ bind( entry );
3546 
3547     __ cmp( O3, O2 );
3548     __ brx( Assembler::lessEqualUnsigned, true, Assembler::pt, loop );
3549     __ delayed()->ld_ptr(O3, BasicObjectLock::obj_offset_in_bytes(), O4);
3550 
3551     __ bind( exit );
3552   }
3553 
3554   { Label allocated;
3555 
3556     // found free slot?
3557     __ br_notnull(O1, false, Assembler::pn, allocated);
3558     __ delayed()->nop();
3559 
3560     __ add_monitor_to_stack( false, O2, O3 );
3561     __ mov(Lmonitors, O1);
3562 
3563     __ bind(allocated);
3564   }
3565 
3566   // Increment bcp to point to the next bytecode, so exception handling for async. exceptions work correctly.
3567   // The object has already been poped from the stack, so the expression stack looks correct.
3568   __ inc(Lbcp);
3569 
3570   __ st_ptr(O0, O1, BasicObjectLock::obj_offset_in_bytes()); // store object
3571   __ lock_object(O1, O0);
3572 
3573   // check if there's enough space on the stack for the monitors after locking
3574   __ generate_stack_overflow_check(0);
3575 
3576   // The bcp has already been incremented. Just need to dispatch to next instruction.
3577   __ dispatch_next(vtos);
3578 }
3579 
3580 
3581 void TemplateTable::monitorexit() {
3582   transition(atos, vtos);
3583   __ verify_oop(Otos_i);
3584   __ tst(Otos_i);
3585   __ throw_if_not_x( Assembler::notZero, Interpreter::_throw_NullPointerException_entry, G3_scratch );
3586 
3587   assert(O0 == Otos_i, "just checking");
3588 
3589   { Label entry, loop, found;
3590     __ add( __ top_most_monitor(), O2 ); // last one to check
3591     __ ba(false, entry );
3592     // use Lscratch to hold monitor elem to check, start with most recent monitor,
3593     // By using a local it survives the call to the C routine.
3594     __ delayed()->mov( Lmonitors, Lscratch );
3595 
3596     __ bind( loop );
3597 
3598     __ verify_oop(O4);          // verify each monitor's oop
3599     __ cmp(O4, O0); // check if current entry is for desired object
3600     __ brx( Assembler::equal, true, Assembler::pt, found );
3601     __ delayed()->mov(Lscratch, O1); // pass found entry as argument to monitorexit
3602 
3603     __ inc( Lscratch, frame::interpreter_frame_monitor_size() * wordSize ); // advance to next
3604 
3605     __ bind( entry );
3606 
3607     __ cmp( Lscratch, O2 );
3608     __ brx( Assembler::lessEqualUnsigned, true, Assembler::pt, loop );
3609     __ delayed()->ld_ptr(Lscratch, BasicObjectLock::obj_offset_in_bytes(), O4);
3610 
3611     call_VM(noreg, CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_illegal_monitor_state_exception));
3612     __ should_not_reach_here();
3613 
3614     __ bind(found);
3615   }
3616   __ unlock_object(O1);
3617 }
3618 
3619 
3620 //----------------------------------------------------------------------------------------------------
3621 // Wide instructions
3622 
3623 void TemplateTable::wide() {
3624   transition(vtos, vtos);
3625   __ ldub(Lbcp, 1, G3_scratch);// get next bc
3626   __ sll(G3_scratch, LogBytesPerWord, G3_scratch);
3627   Address ep(G4_scratch, (address)Interpreter::_wentry_point);
3628   __ load_address(ep);
3629   __ ld_ptr(ep.base(), G3_scratch, G3_scratch);
3630   __ jmp(G3_scratch, G0);
3631   __ delayed()->nop();
3632   // Note: the Lbcp increment step is part of the individual wide bytecode implementations
3633 }
3634 
3635 
3636 //----------------------------------------------------------------------------------------------------
3637 // Multi arrays
3638 
3639 void TemplateTable::multianewarray() {
3640   transition(vtos, atos);
3641      // put ndims * wordSize into Lscratch
3642   __ ldub( Lbcp,     3,               Lscratch);
3643   __ sll(  Lscratch, Interpreter::logStackElementSize(), Lscratch);
3644      // Lesp points past last_dim, so set to O1 to first_dim address
3645   __ add(  Lesp,     Lscratch,        O1);
3646      call_VM(Otos_i, CAST_FROM_FN_PTR(address, InterpreterRuntime::multianewarray), O1);
3647   __ add(  Lesp,     Lscratch,        Lesp); // pop all dimensions off the stack
3648 }
3649 #endif /* !CC_INTERP */