1 /* 2 * Copyright (c) 1997, 2011, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #ifndef CPU_SPARC_VM_ASSEMBLER_SPARC_HPP 26 #define CPU_SPARC_VM_ASSEMBLER_SPARC_HPP 27 28 class BiasedLockingCounters; 29 30 // <sys/trap.h> promises that the system will not use traps 16-31 31 #define ST_RESERVED_FOR_USER_0 0x10 32 33 /* Written: David Ungar 4/19/97 */ 34 35 // Contains all the definitions needed for sparc assembly code generation. 36 37 // Register aliases for parts of the system: 38 39 // 64 bit values can be kept in g1-g5, o1-o5 and o7 and all 64 bits are safe 40 // across context switches in V8+ ABI. Of course, there are no 64 bit regs 41 // in V8 ABI. All 64 bits are preserved in V9 ABI for all registers. 42 43 // g2-g4 are scratch registers called "application globals". Their 44 // meaning is reserved to the "compilation system"--which means us! 45 // They are are not supposed to be touched by ordinary C code, although 46 // highly-optimized C code might steal them for temps. They are safe 47 // across thread switches, and the ABI requires that they be safe 48 // across function calls. 49 // 50 // g1 and g3 are touched by more modules. V8 allows g1 to be clobbered 51 // across func calls, and V8+ also allows g5 to be clobbered across 52 // func calls. Also, g1 and g5 can get touched while doing shared 53 // library loading. 54 // 55 // We must not touch g7 (it is the thread-self register) and g6 is 56 // reserved for certain tools. g0, of course, is always zero. 57 // 58 // (Sources: SunSoft Compilers Group, thread library engineers.) 59 60 // %%%% The interpreter should be revisited to reduce global scratch regs. 61 62 // This global always holds the current JavaThread pointer: 63 64 REGISTER_DECLARATION(Register, G2_thread , G2); 65 REGISTER_DECLARATION(Register, G6_heapbase , G6); 66 67 // The following globals are part of the Java calling convention: 68 69 REGISTER_DECLARATION(Register, G5_method , G5); 70 REGISTER_DECLARATION(Register, G5_megamorphic_method , G5_method); 71 REGISTER_DECLARATION(Register, G5_inline_cache_reg , G5_method); 72 73 // The following globals are used for the new C1 & interpreter calling convention: 74 REGISTER_DECLARATION(Register, Gargs , G4); // pointing to the last argument 75 76 // This local is used to preserve G2_thread in the interpreter and in stubs: 77 REGISTER_DECLARATION(Register, L7_thread_cache , L7); 78 79 // These globals are used as scratch registers in the interpreter: 80 81 REGISTER_DECLARATION(Register, Gframe_size , G1); // SAME REG as G1_scratch 82 REGISTER_DECLARATION(Register, G1_scratch , G1); // also SAME 83 REGISTER_DECLARATION(Register, G3_scratch , G3); 84 REGISTER_DECLARATION(Register, G4_scratch , G4); 85 86 // These globals are used as short-lived scratch registers in the compiler: 87 88 REGISTER_DECLARATION(Register, Gtemp , G5); 89 90 // JSR 292 fixed register usages: 91 REGISTER_DECLARATION(Register, G5_method_type , G5); 92 REGISTER_DECLARATION(Register, G3_method_handle , G3); 93 REGISTER_DECLARATION(Register, L7_mh_SP_save , L7); 94 95 // The compiler requires that G5_megamorphic_method is G5_inline_cache_klass, 96 // because a single patchable "set" instruction (NativeMovConstReg, 97 // or NativeMovConstPatching for compiler1) instruction 98 // serves to set up either quantity, depending on whether the compiled 99 // call site is an inline cache or is megamorphic. See the function 100 // CompiledIC::set_to_megamorphic. 101 // 102 // If a inline cache targets an interpreted method, then the 103 // G5 register will be used twice during the call. First, 104 // the call site will be patched to load a compiledICHolder 105 // into G5. (This is an ordered pair of ic_klass, method.) 106 // The c2i adapter will first check the ic_klass, then load 107 // G5_method with the method part of the pair just before 108 // jumping into the interpreter. 109 // 110 // Note that G5_method is only the method-self for the interpreter, 111 // and is logically unrelated to G5_megamorphic_method. 112 // 113 // Invariants on G2_thread (the JavaThread pointer): 114 // - it should not be used for any other purpose anywhere 115 // - it must be re-initialized by StubRoutines::call_stub() 116 // - it must be preserved around every use of call_VM 117 118 // We can consider using g2/g3/g4 to cache more values than the 119 // JavaThread, such as the card-marking base or perhaps pointers into 120 // Eden. It's something of a waste to use them as scratch temporaries, 121 // since they are not supposed to be volatile. (Of course, if we find 122 // that Java doesn't benefit from application globals, then we can just 123 // use them as ordinary temporaries.) 124 // 125 // Since g1 and g5 (and/or g6) are the volatile (caller-save) registers, 126 // it makes sense to use them routinely for procedure linkage, 127 // whenever the On registers are not applicable. Examples: G5_method, 128 // G5_inline_cache_klass, and a double handful of miscellaneous compiler 129 // stubs. This means that compiler stubs, etc., should be kept to a 130 // maximum of two or three G-register arguments. 131 132 133 // stub frames 134 135 REGISTER_DECLARATION(Register, Lentry_args , L0); // pointer to args passed to callee (interpreter) not stub itself 136 137 // Interpreter frames 138 139 #ifdef CC_INTERP 140 REGISTER_DECLARATION(Register, Lstate , L0); // interpreter state object pointer 141 REGISTER_DECLARATION(Register, L1_scratch , L1); // scratch 142 REGISTER_DECLARATION(Register, Lmirror , L1); // mirror (for native methods only) 143 REGISTER_DECLARATION(Register, L2_scratch , L2); 144 REGISTER_DECLARATION(Register, L3_scratch , L3); 145 REGISTER_DECLARATION(Register, L4_scratch , L4); 146 REGISTER_DECLARATION(Register, Lscratch , L5); // C1 uses 147 REGISTER_DECLARATION(Register, Lscratch2 , L6); // C1 uses 148 REGISTER_DECLARATION(Register, L7_scratch , L7); // constant pool cache 149 REGISTER_DECLARATION(Register, O5_savedSP , O5); 150 REGISTER_DECLARATION(Register, I5_savedSP , I5); // Saved SP before bumping for locals. This is simply 151 // a copy SP, so in 64-bit it's a biased value. The bias 152 // is added and removed as needed in the frame code. 153 // Interface to signature handler 154 REGISTER_DECLARATION(Register, Llocals , L7); // pointer to locals for signature handler 155 REGISTER_DECLARATION(Register, Lmethod , L6); // methodOop when calling signature handler 156 157 #else 158 REGISTER_DECLARATION(Register, Lesp , L0); // expression stack pointer 159 REGISTER_DECLARATION(Register, Lbcp , L1); // pointer to next bytecode 160 REGISTER_DECLARATION(Register, Lmethod , L2); 161 REGISTER_DECLARATION(Register, Llocals , L3); 162 REGISTER_DECLARATION(Register, Largs , L3); // pointer to locals for signature handler 163 // must match Llocals in asm interpreter 164 REGISTER_DECLARATION(Register, Lmonitors , L4); 165 REGISTER_DECLARATION(Register, Lbyte_code , L5); 166 // When calling out from the interpreter we record SP so that we can remove any extra stack 167 // space allocated during adapter transitions. This register is only live from the point 168 // of the call until we return. 169 REGISTER_DECLARATION(Register, Llast_SP , L5); 170 REGISTER_DECLARATION(Register, Lscratch , L5); 171 REGISTER_DECLARATION(Register, Lscratch2 , L6); 172 REGISTER_DECLARATION(Register, LcpoolCache , L6); // constant pool cache 173 174 REGISTER_DECLARATION(Register, O5_savedSP , O5); 175 REGISTER_DECLARATION(Register, I5_savedSP , I5); // Saved SP before bumping for locals. This is simply 176 // a copy SP, so in 64-bit it's a biased value. The bias 177 // is added and removed as needed in the frame code. 178 REGISTER_DECLARATION(Register, IdispatchTables , I4); // Base address of the bytecode dispatch tables 179 REGISTER_DECLARATION(Register, IdispatchAddress , I3); // Register which saves the dispatch address for each bytecode 180 REGISTER_DECLARATION(Register, ImethodDataPtr , I2); // Pointer to the current method data 181 #endif /* CC_INTERP */ 182 183 // NOTE: Lscratch2 and LcpoolCache point to the same registers in 184 // the interpreter code. If Lscratch2 needs to be used for some 185 // purpose than LcpoolCache should be restore after that for 186 // the interpreter to work right 187 // (These assignments must be compatible with L7_thread_cache; see above.) 188 189 // Since Lbcp points into the middle of the method object, 190 // it is temporarily converted into a "bcx" during GC. 191 192 // Exception processing 193 // These registers are passed into exception handlers. 194 // All exception handlers require the exception object being thrown. 195 // In addition, an nmethod's exception handler must be passed 196 // the address of the call site within the nmethod, to allow 197 // proper selection of the applicable catch block. 198 // (Interpreter frames use their own bcp() for this purpose.) 199 // 200 // The Oissuing_pc value is not always needed. When jumping to a 201 // handler that is known to be interpreted, the Oissuing_pc value can be 202 // omitted. An actual catch block in compiled code receives (from its 203 // nmethod's exception handler) the thrown exception in the Oexception, 204 // but it doesn't need the Oissuing_pc. 205 // 206 // If an exception handler (either interpreted or compiled) 207 // discovers there is no applicable catch block, it updates 208 // the Oissuing_pc to the continuation PC of its own caller, 209 // pops back to that caller's stack frame, and executes that 210 // caller's exception handler. Obviously, this process will 211 // iterate until the control stack is popped back to a method 212 // containing an applicable catch block. A key invariant is 213 // that the Oissuing_pc value is always a value local to 214 // the method whose exception handler is currently executing. 215 // 216 // Note: The issuing PC value is __not__ a raw return address (I7 value). 217 // It is a "return pc", the address __following__ the call. 218 // Raw return addresses are converted to issuing PCs by frame::pc(), 219 // or by stubs. Issuing PCs can be used directly with PC range tables. 220 // 221 REGISTER_DECLARATION(Register, Oexception , O0); // exception being thrown 222 REGISTER_DECLARATION(Register, Oissuing_pc , O1); // where the exception is coming from 223 224 225 // These must occur after the declarations above 226 #ifndef DONT_USE_REGISTER_DEFINES 227 228 #define Gthread AS_REGISTER(Register, Gthread) 229 #define Gmethod AS_REGISTER(Register, Gmethod) 230 #define Gmegamorphic_method AS_REGISTER(Register, Gmegamorphic_method) 231 #define Ginline_cache_reg AS_REGISTER(Register, Ginline_cache_reg) 232 #define Gargs AS_REGISTER(Register, Gargs) 233 #define Lthread_cache AS_REGISTER(Register, Lthread_cache) 234 #define Gframe_size AS_REGISTER(Register, Gframe_size) 235 #define Gtemp AS_REGISTER(Register, Gtemp) 236 237 #ifdef CC_INTERP 238 #define Lstate AS_REGISTER(Register, Lstate) 239 #define Lesp AS_REGISTER(Register, Lesp) 240 #define L1_scratch AS_REGISTER(Register, L1_scratch) 241 #define Lmirror AS_REGISTER(Register, Lmirror) 242 #define L2_scratch AS_REGISTER(Register, L2_scratch) 243 #define L3_scratch AS_REGISTER(Register, L3_scratch) 244 #define L4_scratch AS_REGISTER(Register, L4_scratch) 245 #define Lscratch AS_REGISTER(Register, Lscratch) 246 #define Lscratch2 AS_REGISTER(Register, Lscratch2) 247 #define L7_scratch AS_REGISTER(Register, L7_scratch) 248 #define Ostate AS_REGISTER(Register, Ostate) 249 #else 250 #define Lesp AS_REGISTER(Register, Lesp) 251 #define Lbcp AS_REGISTER(Register, Lbcp) 252 #define Lmethod AS_REGISTER(Register, Lmethod) 253 #define Llocals AS_REGISTER(Register, Llocals) 254 #define Lmonitors AS_REGISTER(Register, Lmonitors) 255 #define Lbyte_code AS_REGISTER(Register, Lbyte_code) 256 #define Lscratch AS_REGISTER(Register, Lscratch) 257 #define Lscratch2 AS_REGISTER(Register, Lscratch2) 258 #define LcpoolCache AS_REGISTER(Register, LcpoolCache) 259 #endif /* ! CC_INTERP */ 260 261 #define Lentry_args AS_REGISTER(Register, Lentry_args) 262 #define I5_savedSP AS_REGISTER(Register, I5_savedSP) 263 #define O5_savedSP AS_REGISTER(Register, O5_savedSP) 264 #define IdispatchAddress AS_REGISTER(Register, IdispatchAddress) 265 #define ImethodDataPtr AS_REGISTER(Register, ImethodDataPtr) 266 #define IdispatchTables AS_REGISTER(Register, IdispatchTables) 267 268 #define Oexception AS_REGISTER(Register, Oexception) 269 #define Oissuing_pc AS_REGISTER(Register, Oissuing_pc) 270 271 272 #endif 273 274 // Address is an abstraction used to represent a memory location. 275 // 276 // Note: A register location is represented via a Register, not 277 // via an address for efficiency & simplicity reasons. 278 279 class Address VALUE_OBJ_CLASS_SPEC { 280 private: 281 Register _base; // Base register. 282 RegisterOrConstant _index_or_disp; // Index register or constant displacement. 283 RelocationHolder _rspec; 284 285 public: 286 Address() : _base(noreg), _index_or_disp(noreg) {} 287 288 Address(Register base, RegisterOrConstant index_or_disp) 289 : _base(base), 290 _index_or_disp(index_or_disp) { 291 } 292 293 Address(Register base, Register index) 294 : _base(base), 295 _index_or_disp(index) { 296 } 297 298 Address(Register base, int disp) 299 : _base(base), 300 _index_or_disp(disp) { 301 } 302 303 #ifdef ASSERT 304 // ByteSize is only a class when ASSERT is defined, otherwise it's an int. 305 Address(Register base, ByteSize disp) 306 : _base(base), 307 _index_or_disp(in_bytes(disp)) { 308 } 309 #endif 310 311 // accessors 312 Register base() const { return _base; } 313 Register index() const { return _index_or_disp.as_register(); } 314 int disp() const { return _index_or_disp.as_constant(); } 315 316 bool has_index() const { return _index_or_disp.is_register(); } 317 bool has_disp() const { return _index_or_disp.is_constant(); } 318 319 bool uses(Register reg) const { return base() == reg || (has_index() && index() == reg); } 320 321 const relocInfo::relocType rtype() { return _rspec.type(); } 322 const RelocationHolder& rspec() { return _rspec; } 323 324 RelocationHolder rspec(int offset) const { 325 return offset == 0 ? _rspec : _rspec.plus(offset); 326 } 327 328 inline bool is_simm13(int offset = 0); // check disp+offset for overflow 329 330 Address plus_disp(int plusdisp) const { // bump disp by a small amount 331 assert(_index_or_disp.is_constant(), "must have a displacement"); 332 Address a(base(), disp() + plusdisp); 333 return a; 334 } 335 bool is_same_address(Address a) const { 336 // disregard _rspec 337 return base() == a.base() && (has_index() ? index() == a.index() : disp() == a.disp()); 338 } 339 340 Address after_save() const { 341 Address a = (*this); 342 a._base = a._base->after_save(); 343 return a; 344 } 345 346 Address after_restore() const { 347 Address a = (*this); 348 a._base = a._base->after_restore(); 349 return a; 350 } 351 352 // Convert the raw encoding form into the form expected by the 353 // constructor for Address. 354 static Address make_raw(int base, int index, int scale, int disp, bool disp_is_oop); 355 356 friend class Assembler; 357 }; 358 359 360 class AddressLiteral VALUE_OBJ_CLASS_SPEC { 361 private: 362 address _address; 363 RelocationHolder _rspec; 364 365 RelocationHolder rspec_from_rtype(relocInfo::relocType rtype, address addr) { 366 switch (rtype) { 367 case relocInfo::external_word_type: 368 return external_word_Relocation::spec(addr); 369 case relocInfo::internal_word_type: 370 return internal_word_Relocation::spec(addr); 371 #ifdef _LP64 372 case relocInfo::opt_virtual_call_type: 373 return opt_virtual_call_Relocation::spec(); 374 case relocInfo::static_call_type: 375 return static_call_Relocation::spec(); 376 case relocInfo::runtime_call_type: 377 return runtime_call_Relocation::spec(); 378 #endif 379 case relocInfo::none: 380 return RelocationHolder(); 381 default: 382 ShouldNotReachHere(); 383 return RelocationHolder(); 384 } 385 } 386 387 protected: 388 // creation 389 AddressLiteral() : _address(NULL), _rspec(NULL) {} 390 391 public: 392 AddressLiteral(address addr, RelocationHolder const& rspec) 393 : _address(addr), 394 _rspec(rspec) {} 395 396 // Some constructors to avoid casting at the call site. 397 AddressLiteral(jobject obj, RelocationHolder const& rspec) 398 : _address((address) obj), 399 _rspec(rspec) {} 400 401 AddressLiteral(intptr_t value, RelocationHolder const& rspec) 402 : _address((address) value), 403 _rspec(rspec) {} 404 405 AddressLiteral(address addr, relocInfo::relocType rtype = relocInfo::none) 406 : _address((address) addr), 407 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 408 409 // Some constructors to avoid casting at the call site. 410 AddressLiteral(address* addr, relocInfo::relocType rtype = relocInfo::none) 411 : _address((address) addr), 412 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 413 414 AddressLiteral(bool* addr, relocInfo::relocType rtype = relocInfo::none) 415 : _address((address) addr), 416 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 417 418 AddressLiteral(const bool* addr, relocInfo::relocType rtype = relocInfo::none) 419 : _address((address) addr), 420 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 421 422 AddressLiteral(signed char* addr, relocInfo::relocType rtype = relocInfo::none) 423 : _address((address) addr), 424 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 425 426 AddressLiteral(int* addr, relocInfo::relocType rtype = relocInfo::none) 427 : _address((address) addr), 428 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 429 430 AddressLiteral(intptr_t addr, relocInfo::relocType rtype = relocInfo::none) 431 : _address((address) addr), 432 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 433 434 #ifdef _LP64 435 // 32-bit complains about a multiple declaration for int*. 436 AddressLiteral(intptr_t* addr, relocInfo::relocType rtype = relocInfo::none) 437 : _address((address) addr), 438 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 439 #endif 440 441 AddressLiteral(oop addr, relocInfo::relocType rtype = relocInfo::none) 442 : _address((address) addr), 443 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 444 445 AddressLiteral(oop* addr, relocInfo::relocType rtype = relocInfo::none) 446 : _address((address) addr), 447 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 448 449 AddressLiteral(float* addr, relocInfo::relocType rtype = relocInfo::none) 450 : _address((address) addr), 451 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 452 453 AddressLiteral(double* addr, relocInfo::relocType rtype = relocInfo::none) 454 : _address((address) addr), 455 _rspec(rspec_from_rtype(rtype, (address) addr)) {} 456 457 intptr_t value() const { return (intptr_t) _address; } 458 int low10() const; 459 460 const relocInfo::relocType rtype() const { return _rspec.type(); } 461 const RelocationHolder& rspec() const { return _rspec; } 462 463 RelocationHolder rspec(int offset) const { 464 return offset == 0 ? _rspec : _rspec.plus(offset); 465 } 466 }; 467 468 // Convenience classes 469 class ExternalAddress: public AddressLiteral { 470 private: 471 static relocInfo::relocType reloc_for_target(address target) { 472 // Sometimes ExternalAddress is used for values which aren't 473 // exactly addresses, like the card table base. 474 // external_word_type can't be used for values in the first page 475 // so just skip the reloc in that case. 476 return external_word_Relocation::can_be_relocated(target) ? relocInfo::external_word_type : relocInfo::none; 477 } 478 479 public: 480 ExternalAddress(address target) : AddressLiteral(target, reloc_for_target( target)) {} 481 ExternalAddress(oop* target) : AddressLiteral(target, reloc_for_target((address) target)) {} 482 }; 483 484 inline Address RegisterImpl::address_in_saved_window() const { 485 return (Address(SP, (sp_offset_in_saved_window() * wordSize) + STACK_BIAS)); 486 } 487 488 489 490 // Argument is an abstraction used to represent an outgoing 491 // actual argument or an incoming formal parameter, whether 492 // it resides in memory or in a register, in a manner consistent 493 // with the SPARC Application Binary Interface, or ABI. This is 494 // often referred to as the native or C calling convention. 495 496 class Argument VALUE_OBJ_CLASS_SPEC { 497 private: 498 int _number; 499 bool _is_in; 500 501 public: 502 #ifdef _LP64 503 enum { 504 n_register_parameters = 6, // only 6 registers may contain integer parameters 505 n_float_register_parameters = 16 // Can have up to 16 floating registers 506 }; 507 #else 508 enum { 509 n_register_parameters = 6 // only 6 registers may contain integer parameters 510 }; 511 #endif 512 513 // creation 514 Argument(int number, bool is_in) : _number(number), _is_in(is_in) {} 515 516 int number() const { return _number; } 517 bool is_in() const { return _is_in; } 518 bool is_out() const { return !is_in(); } 519 520 Argument successor() const { return Argument(number() + 1, is_in()); } 521 Argument as_in() const { return Argument(number(), true ); } 522 Argument as_out() const { return Argument(number(), false); } 523 524 // locating register-based arguments: 525 bool is_register() const { return _number < n_register_parameters; } 526 527 #ifdef _LP64 528 // locating Floating Point register-based arguments: 529 bool is_float_register() const { return _number < n_float_register_parameters; } 530 531 FloatRegister as_float_register() const { 532 assert(is_float_register(), "must be a register argument"); 533 return as_FloatRegister(( number() *2 ) + 1); 534 } 535 FloatRegister as_double_register() const { 536 assert(is_float_register(), "must be a register argument"); 537 return as_FloatRegister(( number() *2 )); 538 } 539 #endif 540 541 Register as_register() const { 542 assert(is_register(), "must be a register argument"); 543 return is_in() ? as_iRegister(number()) : as_oRegister(number()); 544 } 545 546 // locating memory-based arguments 547 Address as_address() const { 548 assert(!is_register(), "must be a memory argument"); 549 return address_in_frame(); 550 } 551 552 // When applied to a register-based argument, give the corresponding address 553 // into the 6-word area "into which callee may store register arguments" 554 // (This is a different place than the corresponding register-save area location.) 555 Address address_in_frame() const; 556 557 // debugging 558 const char* name() const; 559 560 friend class Assembler; 561 }; 562 563 564 // The SPARC Assembler: Pure assembler doing NO optimizations on the instruction 565 // level; i.e., what you write 566 // is what you get. The Assembler is generating code into a CodeBuffer. 567 568 class Assembler : public AbstractAssembler { 569 protected: 570 571 static void print_instruction(int inst); 572 static int patched_branch(int dest_pos, int inst, int inst_pos); 573 static int branch_destination(int inst, int pos); 574 575 576 friend class AbstractAssembler; 577 friend class AddressLiteral; 578 579 // code patchers need various routines like inv_wdisp() 580 friend class NativeInstruction; 581 friend class NativeGeneralJump; 582 friend class Relocation; 583 friend class Label; 584 585 public: 586 // op carries format info; see page 62 & 267 587 588 enum ops { 589 call_op = 1, // fmt 1 590 branch_op = 0, // also sethi (fmt2) 591 arith_op = 2, // fmt 3, arith & misc 592 ldst_op = 3 // fmt 3, load/store 593 }; 594 595 enum op2s { 596 bpr_op2 = 3, 597 fb_op2 = 6, 598 fbp_op2 = 5, 599 br_op2 = 2, 600 bp_op2 = 1, 601 cb_op2 = 7, // V8 602 sethi_op2 = 4 603 }; 604 605 enum op3s { 606 // selected op3s 607 add_op3 = 0x00, 608 and_op3 = 0x01, 609 or_op3 = 0x02, 610 xor_op3 = 0x03, 611 sub_op3 = 0x04, 612 andn_op3 = 0x05, 613 orn_op3 = 0x06, 614 xnor_op3 = 0x07, 615 addc_op3 = 0x08, 616 mulx_op3 = 0x09, 617 umul_op3 = 0x0a, 618 smul_op3 = 0x0b, 619 subc_op3 = 0x0c, 620 udivx_op3 = 0x0d, 621 udiv_op3 = 0x0e, 622 sdiv_op3 = 0x0f, 623 624 addcc_op3 = 0x10, 625 andcc_op3 = 0x11, 626 orcc_op3 = 0x12, 627 xorcc_op3 = 0x13, 628 subcc_op3 = 0x14, 629 andncc_op3 = 0x15, 630 orncc_op3 = 0x16, 631 xnorcc_op3 = 0x17, 632 addccc_op3 = 0x18, 633 umulcc_op3 = 0x1a, 634 smulcc_op3 = 0x1b, 635 subccc_op3 = 0x1c, 636 udivcc_op3 = 0x1e, 637 sdivcc_op3 = 0x1f, 638 639 taddcc_op3 = 0x20, 640 tsubcc_op3 = 0x21, 641 taddcctv_op3 = 0x22, 642 tsubcctv_op3 = 0x23, 643 mulscc_op3 = 0x24, 644 sll_op3 = 0x25, 645 sllx_op3 = 0x25, 646 srl_op3 = 0x26, 647 srlx_op3 = 0x26, 648 sra_op3 = 0x27, 649 srax_op3 = 0x27, 650 rdreg_op3 = 0x28, 651 membar_op3 = 0x28, 652 653 flushw_op3 = 0x2b, 654 movcc_op3 = 0x2c, 655 sdivx_op3 = 0x2d, 656 popc_op3 = 0x2e, 657 movr_op3 = 0x2f, 658 659 sir_op3 = 0x30, 660 wrreg_op3 = 0x30, 661 saved_op3 = 0x31, 662 663 fpop1_op3 = 0x34, 664 fpop2_op3 = 0x35, 665 impdep1_op3 = 0x36, 666 impdep2_op3 = 0x37, 667 jmpl_op3 = 0x38, 668 rett_op3 = 0x39, 669 trap_op3 = 0x3a, 670 flush_op3 = 0x3b, 671 save_op3 = 0x3c, 672 restore_op3 = 0x3d, 673 done_op3 = 0x3e, 674 retry_op3 = 0x3e, 675 676 lduw_op3 = 0x00, 677 ldub_op3 = 0x01, 678 lduh_op3 = 0x02, 679 ldd_op3 = 0x03, 680 stw_op3 = 0x04, 681 stb_op3 = 0x05, 682 sth_op3 = 0x06, 683 std_op3 = 0x07, 684 ldsw_op3 = 0x08, 685 ldsb_op3 = 0x09, 686 ldsh_op3 = 0x0a, 687 ldx_op3 = 0x0b, 688 689 ldstub_op3 = 0x0d, 690 stx_op3 = 0x0e, 691 swap_op3 = 0x0f, 692 693 stwa_op3 = 0x14, 694 stxa_op3 = 0x1e, 695 696 ldf_op3 = 0x20, 697 ldfsr_op3 = 0x21, 698 ldqf_op3 = 0x22, 699 lddf_op3 = 0x23, 700 stf_op3 = 0x24, 701 stfsr_op3 = 0x25, 702 stqf_op3 = 0x26, 703 stdf_op3 = 0x27, 704 705 prefetch_op3 = 0x2d, 706 707 708 ldc_op3 = 0x30, 709 ldcsr_op3 = 0x31, 710 lddc_op3 = 0x33, 711 stc_op3 = 0x34, 712 stcsr_op3 = 0x35, 713 stdcq_op3 = 0x36, 714 stdc_op3 = 0x37, 715 716 casa_op3 = 0x3c, 717 casxa_op3 = 0x3e, 718 719 alt_bit_op3 = 0x10, 720 cc_bit_op3 = 0x10 721 }; 722 723 enum opfs { 724 // selected opfs 725 fmovs_opf = 0x01, 726 fmovd_opf = 0x02, 727 728 fnegs_opf = 0x05, 729 fnegd_opf = 0x06, 730 731 fadds_opf = 0x41, 732 faddd_opf = 0x42, 733 fsubs_opf = 0x45, 734 fsubd_opf = 0x46, 735 736 fmuls_opf = 0x49, 737 fmuld_opf = 0x4a, 738 fdivs_opf = 0x4d, 739 fdivd_opf = 0x4e, 740 741 fcmps_opf = 0x51, 742 fcmpd_opf = 0x52, 743 744 fstox_opf = 0x81, 745 fdtox_opf = 0x82, 746 fxtos_opf = 0x84, 747 fxtod_opf = 0x88, 748 fitos_opf = 0xc4, 749 fdtos_opf = 0xc6, 750 fitod_opf = 0xc8, 751 fstod_opf = 0xc9, 752 fstoi_opf = 0xd1, 753 fdtoi_opf = 0xd2 754 }; 755 756 enum RCondition { rc_z = 1, rc_lez = 2, rc_lz = 3, rc_nz = 5, rc_gz = 6, rc_gez = 7 }; 757 758 enum Condition { 759 // for FBfcc & FBPfcc instruction 760 f_never = 0, 761 f_notEqual = 1, 762 f_notZero = 1, 763 f_lessOrGreater = 2, 764 f_unorderedOrLess = 3, 765 f_less = 4, 766 f_unorderedOrGreater = 5, 767 f_greater = 6, 768 f_unordered = 7, 769 f_always = 8, 770 f_equal = 9, 771 f_zero = 9, 772 f_unorderedOrEqual = 10, 773 f_greaterOrEqual = 11, 774 f_unorderedOrGreaterOrEqual = 12, 775 f_lessOrEqual = 13, 776 f_unorderedOrLessOrEqual = 14, 777 f_ordered = 15, 778 779 // V8 coproc, pp 123 v8 manual 780 781 cp_always = 8, 782 cp_never = 0, 783 cp_3 = 7, 784 cp_2 = 6, 785 cp_2or3 = 5, 786 cp_1 = 4, 787 cp_1or3 = 3, 788 cp_1or2 = 2, 789 cp_1or2or3 = 1, 790 cp_0 = 9, 791 cp_0or3 = 10, 792 cp_0or2 = 11, 793 cp_0or2or3 = 12, 794 cp_0or1 = 13, 795 cp_0or1or3 = 14, 796 cp_0or1or2 = 15, 797 798 799 // for integers 800 801 never = 0, 802 equal = 1, 803 zero = 1, 804 lessEqual = 2, 805 less = 3, 806 lessEqualUnsigned = 4, 807 lessUnsigned = 5, 808 carrySet = 5, 809 negative = 6, 810 overflowSet = 7, 811 always = 8, 812 notEqual = 9, 813 notZero = 9, 814 greater = 10, 815 greaterEqual = 11, 816 greaterUnsigned = 12, 817 greaterEqualUnsigned = 13, 818 carryClear = 13, 819 positive = 14, 820 overflowClear = 15 821 }; 822 823 enum CC { 824 icc = 0, xcc = 2, 825 // ptr_cc is the correct condition code for a pointer or intptr_t: 826 ptr_cc = NOT_LP64(icc) LP64_ONLY(xcc), 827 fcc0 = 0, fcc1 = 1, fcc2 = 2, fcc3 = 3 828 }; 829 830 enum PrefetchFcn { 831 severalReads = 0, oneRead = 1, severalWritesAndPossiblyReads = 2, oneWrite = 3, page = 4 832 }; 833 834 public: 835 // Helper functions for groups of instructions 836 837 enum Predict { pt = 1, pn = 0 }; // pt = predict taken 838 839 enum Membar_mask_bits { // page 184, v9 840 StoreStore = 1 << 3, 841 LoadStore = 1 << 2, 842 StoreLoad = 1 << 1, 843 LoadLoad = 1 << 0, 844 845 Sync = 1 << 6, 846 MemIssue = 1 << 5, 847 Lookaside = 1 << 4 848 }; 849 850 // test if x is within signed immediate range for nbits 851 static bool is_simm(intptr_t x, int nbits) { return -( intptr_t(1) << nbits-1 ) <= x && x < ( intptr_t(1) << nbits-1 ); } 852 853 // test if -4096 <= x <= 4095 854 static bool is_simm13(intptr_t x) { return is_simm(x, 13); } 855 856 static bool is_in_wdisp_range(address a, address b, int nbits) { 857 intptr_t d = intptr_t(b) - intptr_t(a); 858 return is_simm(d, nbits + 2); 859 } 860 861 // test if label is in simm16 range in words (wdisp16). 862 bool is_in_wdisp16_range(Label& L) { 863 return is_in_wdisp_range(target(L), pc(), 16); 864 } 865 // test if the distance between two addresses fits in simm30 range in words 866 static bool is_in_wdisp30_range(address a, address b) { 867 return is_in_wdisp_range(a, b, 30); 868 } 869 870 enum ASIs { // page 72, v9 871 ASI_PRIMARY = 0x80, 872 ASI_PRIMARY_LITTLE = 0x88 873 // add more from book as needed 874 }; 875 876 protected: 877 // helpers 878 879 // x is supposed to fit in a field "nbits" wide 880 // and be sign-extended. Check the range. 881 882 static void assert_signed_range(intptr_t x, int nbits) { 883 assert(nbits == 32 || (-(1 << nbits-1) <= x && x < ( 1 << nbits-1)), 884 err_msg("value out of range: x=" INTPTR_FORMAT ", nbits=%d", x, nbits)); 885 } 886 887 static void assert_signed_word_disp_range(intptr_t x, int nbits) { 888 assert( (x & 3) == 0, "not word aligned"); 889 assert_signed_range(x, nbits + 2); 890 } 891 892 static void assert_unsigned_const(int x, int nbits) { 893 assert( juint(x) < juint(1 << nbits), "unsigned constant out of range"); 894 } 895 896 // fields: note bits numbered from LSB = 0, 897 // fields known by inclusive bit range 898 899 static int fmask(juint hi_bit, juint lo_bit) { 900 assert( hi_bit >= lo_bit && 0 <= lo_bit && hi_bit < 32, "bad bits"); 901 return (1 << ( hi_bit-lo_bit + 1 )) - 1; 902 } 903 904 // inverse of u_field 905 906 static int inv_u_field(int x, int hi_bit, int lo_bit) { 907 juint r = juint(x) >> lo_bit; 908 r &= fmask( hi_bit, lo_bit); 909 return int(r); 910 } 911 912 913 // signed version: extract from field and sign-extend 914 915 static int inv_s_field(int x, int hi_bit, int lo_bit) { 916 int sign_shift = 31 - hi_bit; 917 return inv_u_field( ((x << sign_shift) >> sign_shift), hi_bit, lo_bit); 918 } 919 920 // given a field that ranges from hi_bit to lo_bit (inclusive, 921 // LSB = 0), and an unsigned value for the field, 922 // shift it into the field 923 924 #ifdef ASSERT 925 static int u_field(int x, int hi_bit, int lo_bit) { 926 assert( ( x & ~fmask(hi_bit, lo_bit)) == 0, 927 "value out of range"); 928 int r = x << lo_bit; 929 assert( inv_u_field(r, hi_bit, lo_bit) == x, "just checking"); 930 return r; 931 } 932 #else 933 // make sure this is inlined as it will reduce code size significantly 934 #define u_field(x, hi_bit, lo_bit) ((x) << (lo_bit)) 935 #endif 936 937 static int inv_op( int x ) { return inv_u_field(x, 31, 30); } 938 static int inv_op2( int x ) { return inv_u_field(x, 24, 22); } 939 static int inv_op3( int x ) { return inv_u_field(x, 24, 19); } 940 static int inv_cond( int x ){ return inv_u_field(x, 28, 25); } 941 942 static bool inv_immed( int x ) { return (x & Assembler::immed(true)) != 0; } 943 944 static Register inv_rd( int x ) { return as_Register(inv_u_field(x, 29, 25)); } 945 static Register inv_rs1( int x ) { return as_Register(inv_u_field(x, 18, 14)); } 946 static Register inv_rs2( int x ) { return as_Register(inv_u_field(x, 4, 0)); } 947 948 static int op( int x) { return u_field(x, 31, 30); } 949 static int rd( Register r) { return u_field(r->encoding(), 29, 25); } 950 static int fcn( int x) { return u_field(x, 29, 25); } 951 static int op3( int x) { return u_field(x, 24, 19); } 952 static int rs1( Register r) { return u_field(r->encoding(), 18, 14); } 953 static int rs2( Register r) { return u_field(r->encoding(), 4, 0); } 954 static int annul( bool a) { return u_field(a ? 1 : 0, 29, 29); } 955 static int cond( int x) { return u_field(x, 28, 25); } 956 static int cond_mov( int x) { return u_field(x, 17, 14); } 957 static int rcond( RCondition x) { return u_field(x, 12, 10); } 958 static int op2( int x) { return u_field(x, 24, 22); } 959 static int predict( bool p) { return u_field(p ? 1 : 0, 19, 19); } 960 static int branchcc( CC fcca) { return u_field(fcca, 21, 20); } 961 static int cmpcc( CC fcca) { return u_field(fcca, 26, 25); } 962 static int imm_asi( int x) { return u_field(x, 12, 5); } 963 static int immed( bool i) { return u_field(i ? 1 : 0, 13, 13); } 964 static int opf_low6( int w) { return u_field(w, 10, 5); } 965 static int opf_low5( int w) { return u_field(w, 9, 5); } 966 static int trapcc( CC cc) { return u_field(cc, 12, 11); } 967 static int sx( int i) { return u_field(i, 12, 12); } // shift x=1 means 64-bit 968 static int opf( int x) { return u_field(x, 13, 5); } 969 970 static int opf_cc( CC c, bool useFloat ) { return u_field((useFloat ? 0 : 4) + c, 13, 11); } 971 static int mov_cc( CC c, bool useFloat ) { return u_field(useFloat ? 0 : 1, 18, 18) | u_field(c, 12, 11); } 972 973 static int fd( FloatRegister r, FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa), 29, 25); }; 974 static int fs1(FloatRegister r, FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa), 18, 14); }; 975 static int fs2(FloatRegister r, FloatRegisterImpl::Width fwa) { return u_field(r->encoding(fwa), 4, 0); }; 976 977 // some float instructions use this encoding on the op3 field 978 static int alt_op3(int op, FloatRegisterImpl::Width w) { 979 int r; 980 switch(w) { 981 case FloatRegisterImpl::S: r = op + 0; break; 982 case FloatRegisterImpl::D: r = op + 3; break; 983 case FloatRegisterImpl::Q: r = op + 2; break; 984 default: ShouldNotReachHere(); break; 985 } 986 return op3(r); 987 } 988 989 990 // compute inverse of simm 991 static int inv_simm(int x, int nbits) { 992 return (int)(x << (32 - nbits)) >> (32 - nbits); 993 } 994 995 static int inv_simm13( int x ) { return inv_simm(x, 13); } 996 997 // signed immediate, in low bits, nbits long 998 static int simm(int x, int nbits) { 999 assert_signed_range(x, nbits); 1000 return x & (( 1 << nbits ) - 1); 1001 } 1002 1003 // compute inverse of wdisp16 1004 static intptr_t inv_wdisp16(int x, intptr_t pos) { 1005 int lo = x & (( 1 << 14 ) - 1); 1006 int hi = (x >> 20) & 3; 1007 if (hi >= 2) hi |= ~1; 1008 return (((hi << 14) | lo) << 2) + pos; 1009 } 1010 1011 // word offset, 14 bits at LSend, 2 bits at B21, B20 1012 static int wdisp16(intptr_t x, intptr_t off) { 1013 intptr_t xx = x - off; 1014 assert_signed_word_disp_range(xx, 16); 1015 int r = (xx >> 2) & ((1 << 14) - 1) 1016 | ( ( (xx>>(2+14)) & 3 ) << 20 ); 1017 assert( inv_wdisp16(r, off) == x, "inverse is not inverse"); 1018 return r; 1019 } 1020 1021 1022 // word displacement in low-order nbits bits 1023 1024 static intptr_t inv_wdisp( int x, intptr_t pos, int nbits ) { 1025 int pre_sign_extend = x & (( 1 << nbits ) - 1); 1026 int r = pre_sign_extend >= ( 1 << (nbits-1) ) 1027 ? pre_sign_extend | ~(( 1 << nbits ) - 1) 1028 : pre_sign_extend; 1029 return (r << 2) + pos; 1030 } 1031 1032 static int wdisp( intptr_t x, intptr_t off, int nbits ) { 1033 intptr_t xx = x - off; 1034 assert_signed_word_disp_range(xx, nbits); 1035 int r = (xx >> 2) & (( 1 << nbits ) - 1); 1036 assert( inv_wdisp( r, off, nbits ) == x, "inverse not inverse"); 1037 return r; 1038 } 1039 1040 1041 // Extract the top 32 bits in a 64 bit word 1042 static int32_t hi32( int64_t x ) { 1043 int32_t r = int32_t( (uint64_t)x >> 32 ); 1044 return r; 1045 } 1046 1047 // given a sethi instruction, extract the constant, left-justified 1048 static int inv_hi22( int x ) { 1049 return x << 10; 1050 } 1051 1052 // create an imm22 field, given a 32-bit left-justified constant 1053 static int hi22( int x ) { 1054 int r = int( juint(x) >> 10 ); 1055 assert( (r & ~((1 << 22) - 1)) == 0, "just checkin'"); 1056 return r; 1057 } 1058 1059 // create a low10 __value__ (not a field) for a given a 32-bit constant 1060 static int low10( int x ) { 1061 return x & ((1 << 10) - 1); 1062 } 1063 1064 // instruction only in v9 1065 static void v9_only() { assert( VM_Version::v9_instructions_work(), "This instruction only works on SPARC V9"); } 1066 1067 // instruction only in v8 1068 static void v8_only() { assert( VM_Version::v8_instructions_work(), "This instruction only works on SPARC V8"); } 1069 1070 // instruction deprecated in v9 1071 static void v9_dep() { } // do nothing for now 1072 1073 // some float instructions only exist for single prec. on v8 1074 static void v8_s_only(FloatRegisterImpl::Width w) { if (w != FloatRegisterImpl::S) v9_only(); } 1075 1076 // v8 has no CC field 1077 static void v8_no_cc(CC cc) { if (cc) v9_only(); } 1078 1079 protected: 1080 // Simple delay-slot scheme: 1081 // In order to check the programmer, the assembler keeps track of deley slots. 1082 // It forbids CTIs in delay slots (conservative, but should be OK). 1083 // Also, when putting an instruction into a delay slot, you must say 1084 // asm->delayed()->add(...), in order to check that you don't omit 1085 // delay-slot instructions. 1086 // To implement this, we use a simple FSA 1087 1088 #ifdef ASSERT 1089 #define CHECK_DELAY 1090 #endif 1091 #ifdef CHECK_DELAY 1092 enum Delay_state { no_delay, at_delay_slot, filling_delay_slot } delay_state; 1093 #endif 1094 1095 public: 1096 // Tells assembler next instruction must NOT be in delay slot. 1097 // Use at start of multinstruction macros. 1098 void assert_not_delayed() { 1099 // This is a separate overloading to avoid creation of string constants 1100 // in non-asserted code--with some compilers this pollutes the object code. 1101 #ifdef CHECK_DELAY 1102 assert_not_delayed("next instruction should not be a delay slot"); 1103 #endif 1104 } 1105 void assert_not_delayed(const char* msg) { 1106 #ifdef CHECK_DELAY 1107 assert(delay_state == no_delay, msg); 1108 #endif 1109 } 1110 1111 protected: 1112 // Delay slot helpers 1113 // cti is called when emitting control-transfer instruction, 1114 // BEFORE doing the emitting. 1115 // Only effective when assertion-checking is enabled. 1116 void cti() { 1117 #ifdef CHECK_DELAY 1118 assert_not_delayed("cti should not be in delay slot"); 1119 #endif 1120 } 1121 1122 // called when emitting cti with a delay slot, AFTER emitting 1123 void has_delay_slot() { 1124 #ifdef CHECK_DELAY 1125 assert_not_delayed("just checking"); 1126 delay_state = at_delay_slot; 1127 #endif 1128 } 1129 1130 public: 1131 // Tells assembler you know that next instruction is delayed 1132 Assembler* delayed() { 1133 #ifdef CHECK_DELAY 1134 assert ( delay_state == at_delay_slot, "delayed instruction is not in delay slot"); 1135 delay_state = filling_delay_slot; 1136 #endif 1137 return this; 1138 } 1139 1140 void flush() { 1141 #ifdef CHECK_DELAY 1142 assert ( delay_state == no_delay, "ending code with a delay slot"); 1143 #endif 1144 AbstractAssembler::flush(); 1145 } 1146 1147 inline void emit_long(int); // shadows AbstractAssembler::emit_long 1148 inline void emit_data(int x) { emit_long(x); } 1149 inline void emit_data(int, RelocationHolder const&); 1150 inline void emit_data(int, relocInfo::relocType rtype); 1151 // helper for above fcns 1152 inline void check_delay(); 1153 1154 1155 public: 1156 // instructions, refer to page numbers in the SPARC Architecture Manual, V9 1157 1158 // pp 135 (addc was addx in v8) 1159 1160 inline void add(Register s1, Register s2, Register d ); 1161 inline void add(Register s1, int simm13a, Register d, relocInfo::relocType rtype = relocInfo::none); 1162 inline void add(Register s1, int simm13a, Register d, RelocationHolder const& rspec); 1163 inline void add(Register s1, RegisterOrConstant s2, Register d, int offset = 0); 1164 inline void add(const Address& a, Register d, int offset = 0); 1165 1166 void addcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1167 void addcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(add_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1168 void addc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 ) | rs1(s1) | rs2(s2) ); } 1169 void addc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1170 void addccc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1171 void addccc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(addc_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1172 1173 // pp 136 1174 1175 inline void bpr( RCondition c, bool a, Predict p, Register s1, address d, relocInfo::relocType rt = relocInfo::none ); 1176 inline void bpr( RCondition c, bool a, Predict p, Register s1, Label& L); 1177 1178 protected: // use MacroAssembler::br instead 1179 1180 // pp 138 1181 1182 inline void fb( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none ); 1183 inline void fb( Condition c, bool a, Label& L ); 1184 1185 // pp 141 1186 1187 inline void fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1188 inline void fbp( Condition c, bool a, CC cc, Predict p, Label& L ); 1189 1190 public: 1191 1192 // pp 144 1193 1194 inline void br( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none ); 1195 inline void br( Condition c, bool a, Label& L ); 1196 1197 // pp 146 1198 1199 inline void bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1200 inline void bp( Condition c, bool a, CC cc, Predict p, Label& L ); 1201 1202 // pp 121 (V8) 1203 1204 inline void cb( Condition c, bool a, address d, relocInfo::relocType rt = relocInfo::none ); 1205 inline void cb( Condition c, bool a, Label& L ); 1206 1207 // pp 149 1208 1209 inline void call( address d, relocInfo::relocType rt = relocInfo::runtime_call_type ); 1210 inline void call( Label& L, relocInfo::relocType rt = relocInfo::runtime_call_type ); 1211 1212 // pp 150 1213 1214 // These instructions compare the contents of s2 with the contents of 1215 // memory at address in s1. If the values are equal, the contents of memory 1216 // at address s1 is swapped with the data in d. If the values are not equal, 1217 // the the contents of memory at s1 is loaded into d, without the swap. 1218 1219 void casa( Register s1, Register s2, Register d, int ia = -1 ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(casa_op3 ) | rs1(s1) | (ia == -1 ? immed(true) : imm_asi(ia)) | rs2(s2)); } 1220 void casxa( Register s1, Register s2, Register d, int ia = -1 ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(casxa_op3) | rs1(s1) | (ia == -1 ? immed(true) : imm_asi(ia)) | rs2(s2)); } 1221 1222 // pp 152 1223 1224 void udiv( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 ) | rs1(s1) | rs2(s2)); } 1225 void udiv( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1226 void sdiv( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 ) | rs1(s1) | rs2(s2)); } 1227 void sdiv( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1228 void udivcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 | cc_bit_op3) | rs1(s1) | rs2(s2)); } 1229 void udivcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(udiv_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1230 void sdivcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 | cc_bit_op3) | rs1(s1) | rs2(s2)); } 1231 void sdivcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sdiv_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1232 1233 // pp 155 1234 1235 void done() { v9_only(); cti(); emit_long( op(arith_op) | fcn(0) | op3(done_op3) ); } 1236 void retry() { v9_only(); cti(); emit_long( op(arith_op) | fcn(1) | op3(retry_op3) ); } 1237 1238 // pp 156 1239 1240 void fadd( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x40 + w) | fs2(s2, w)); } 1241 void fsub( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x44 + w) | fs2(s2, w)); } 1242 1243 // pp 157 1244 1245 void fcmp( FloatRegisterImpl::Width w, CC cc, FloatRegister s1, FloatRegister s2) { v8_no_cc(cc); emit_long( op(arith_op) | cmpcc(cc) | op3(fpop2_op3) | fs1(s1, w) | opf(0x50 + w) | fs2(s2, w)); } 1246 void fcmpe( FloatRegisterImpl::Width w, CC cc, FloatRegister s1, FloatRegister s2) { v8_no_cc(cc); emit_long( op(arith_op) | cmpcc(cc) | op3(fpop2_op3) | fs1(s1, w) | opf(0x54 + w) | fs2(s2, w)); } 1247 1248 // pp 159 1249 1250 void ftox( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v9_only(); emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x80 + w) | fs2(s, w)); } 1251 void ftoi( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0xd0 + w) | fs2(s, w)); } 1252 1253 // pp 160 1254 1255 void ftof( FloatRegisterImpl::Width sw, FloatRegisterImpl::Width dw, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, dw) | op3(fpop1_op3) | opf(0xc0 + sw + dw*4) | fs2(s, sw)); } 1256 1257 // pp 161 1258 1259 void fxtof( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v9_only(); emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x80 + w*4) | fs2(s, w)); } 1260 void fitof( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0xc0 + w*4) | fs2(s, w)); } 1261 1262 // pp 162 1263 1264 void fmov( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w); emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x00 + w) | fs2(s, w)); } 1265 1266 void fneg( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w); emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x04 + w) | fs2(s, w)); } 1267 1268 // page 144 sparc v8 architecture (double prec works on v8 if the source and destination registers are the same). fnegs is the only instruction available 1269 // on v8 to do negation of single, double and quad precision floats. 1270 1271 void fneg( FloatRegisterImpl::Width w, FloatRegister sd ) { if (VM_Version::v9_instructions_work()) emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x04 + w) | fs2(sd, w)); else emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x05) | fs2(sd, w)); } 1272 1273 void fabs( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { v8_s_only(w); emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x08 + w) | fs2(s, w)); } 1274 1275 // page 144 sparc v8 architecture (double prec works on v8 if the source and destination registers are the same). fabss is the only instruction available 1276 // on v8 to do abs operation on single/double/quad precision floats. 1277 1278 void fabs( FloatRegisterImpl::Width w, FloatRegister sd ) { if (VM_Version::v9_instructions_work()) emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x08 + w) | fs2(sd, w)); else emit_long( op(arith_op) | fd(sd, w) | op3(fpop1_op3) | opf(0x09) | fs2(sd, w)); } 1279 1280 // pp 163 1281 1282 void fmul( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x48 + w) | fs2(s2, w)); } 1283 void fmul( FloatRegisterImpl::Width sw, FloatRegisterImpl::Width dw, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, dw) | op3(fpop1_op3) | fs1(s1, sw) | opf(0x60 + sw + dw*4) | fs2(s2, sw)); } 1284 void fdiv( FloatRegisterImpl::Width w, FloatRegister s1, FloatRegister s2, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | fs1(s1, w) | opf(0x4c + w) | fs2(s2, w)); } 1285 1286 // pp 164 1287 1288 void fsqrt( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d ) { emit_long( op(arith_op) | fd(d, w) | op3(fpop1_op3) | opf(0x28 + w) | fs2(s, w)); } 1289 1290 // pp 165 1291 1292 inline void flush( Register s1, Register s2 ); 1293 inline void flush( Register s1, int simm13a); 1294 1295 // pp 167 1296 1297 void flushw() { v9_only(); emit_long( op(arith_op) | op3(flushw_op3) ); } 1298 1299 // pp 168 1300 1301 void illtrap( int const22a) { if (const22a != 0) v9_only(); emit_long( op(branch_op) | u_field(const22a, 21, 0) ); } 1302 // v8 unimp == illtrap(0) 1303 1304 // pp 169 1305 1306 void impdep1( int id1, int const19a ) { v9_only(); emit_long( op(arith_op) | fcn(id1) | op3(impdep1_op3) | u_field(const19a, 18, 0)); } 1307 void impdep2( int id1, int const19a ) { v9_only(); emit_long( op(arith_op) | fcn(id1) | op3(impdep2_op3) | u_field(const19a, 18, 0)); } 1308 1309 // pp 149 (v8) 1310 1311 void cpop1( int opc, int cr1, int cr2, int crd ) { v8_only(); emit_long( op(arith_op) | fcn(crd) | op3(impdep1_op3) | u_field(cr1, 18, 14) | opf(opc) | u_field(cr2, 4, 0)); } 1312 void cpop2( int opc, int cr1, int cr2, int crd ) { v8_only(); emit_long( op(arith_op) | fcn(crd) | op3(impdep2_op3) | u_field(cr1, 18, 14) | opf(opc) | u_field(cr2, 4, 0)); } 1313 1314 // pp 170 1315 1316 void jmpl( Register s1, Register s2, Register d ); 1317 void jmpl( Register s1, int simm13a, Register d, RelocationHolder const& rspec = RelocationHolder() ); 1318 1319 // 171 1320 1321 inline void ldf(FloatRegisterImpl::Width w, Register s1, RegisterOrConstant s2, FloatRegister d); 1322 inline void ldf(FloatRegisterImpl::Width w, Register s1, Register s2, FloatRegister d); 1323 inline void ldf(FloatRegisterImpl::Width w, Register s1, int simm13a, FloatRegister d, RelocationHolder const& rspec = RelocationHolder()); 1324 1325 inline void ldf(FloatRegisterImpl::Width w, const Address& a, FloatRegister d, int offset = 0); 1326 1327 1328 inline void ldfsr( Register s1, Register s2 ); 1329 inline void ldfsr( Register s1, int simm13a); 1330 inline void ldxfsr( Register s1, Register s2 ); 1331 inline void ldxfsr( Register s1, int simm13a); 1332 1333 // pp 94 (v8) 1334 1335 inline void ldc( Register s1, Register s2, int crd ); 1336 inline void ldc( Register s1, int simm13a, int crd); 1337 inline void lddc( Register s1, Register s2, int crd ); 1338 inline void lddc( Register s1, int simm13a, int crd); 1339 inline void ldcsr( Register s1, Register s2, int crd ); 1340 inline void ldcsr( Register s1, int simm13a, int crd); 1341 1342 1343 // 173 1344 1345 void ldfa( FloatRegisterImpl::Width w, Register s1, Register s2, int ia, FloatRegister d ) { v9_only(); emit_long( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3 | alt_bit_op3, w) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1346 void ldfa( FloatRegisterImpl::Width w, Register s1, int simm13a, FloatRegister d ) { v9_only(); emit_long( op(ldst_op) | fd(d, w) | alt_op3(ldf_op3 | alt_bit_op3, w) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1347 1348 // pp 175, lduw is ld on v8 1349 1350 inline void ldsb( Register s1, Register s2, Register d ); 1351 inline void ldsb( Register s1, int simm13a, Register d); 1352 inline void ldsh( Register s1, Register s2, Register d ); 1353 inline void ldsh( Register s1, int simm13a, Register d); 1354 inline void ldsw( Register s1, Register s2, Register d ); 1355 inline void ldsw( Register s1, int simm13a, Register d); 1356 inline void ldub( Register s1, Register s2, Register d ); 1357 inline void ldub( Register s1, int simm13a, Register d); 1358 inline void lduh( Register s1, Register s2, Register d ); 1359 inline void lduh( Register s1, int simm13a, Register d); 1360 inline void lduw( Register s1, Register s2, Register d ); 1361 inline void lduw( Register s1, int simm13a, Register d); 1362 inline void ldx( Register s1, Register s2, Register d ); 1363 inline void ldx( Register s1, int simm13a, Register d); 1364 inline void ld( Register s1, Register s2, Register d ); 1365 inline void ld( Register s1, int simm13a, Register d); 1366 inline void ldd( Register s1, Register s2, Register d ); 1367 inline void ldd( Register s1, int simm13a, Register d); 1368 1369 #ifdef ASSERT 1370 // ByteSize is only a class when ASSERT is defined, otherwise it's an int. 1371 inline void ld( Register s1, ByteSize simm13a, Register d); 1372 #endif 1373 1374 inline void ldsb(const Address& a, Register d, int offset = 0); 1375 inline void ldsh(const Address& a, Register d, int offset = 0); 1376 inline void ldsw(const Address& a, Register d, int offset = 0); 1377 inline void ldub(const Address& a, Register d, int offset = 0); 1378 inline void lduh(const Address& a, Register d, int offset = 0); 1379 inline void lduw(const Address& a, Register d, int offset = 0); 1380 inline void ldx( const Address& a, Register d, int offset = 0); 1381 inline void ld( const Address& a, Register d, int offset = 0); 1382 inline void ldd( const Address& a, Register d, int offset = 0); 1383 1384 inline void ldub( Register s1, RegisterOrConstant s2, Register d ); 1385 inline void ldsb( Register s1, RegisterOrConstant s2, Register d ); 1386 inline void lduh( Register s1, RegisterOrConstant s2, Register d ); 1387 inline void ldsh( Register s1, RegisterOrConstant s2, Register d ); 1388 inline void lduw( Register s1, RegisterOrConstant s2, Register d ); 1389 inline void ldsw( Register s1, RegisterOrConstant s2, Register d ); 1390 inline void ldx( Register s1, RegisterOrConstant s2, Register d ); 1391 inline void ld( Register s1, RegisterOrConstant s2, Register d ); 1392 inline void ldd( Register s1, RegisterOrConstant s2, Register d ); 1393 1394 // pp 177 1395 1396 void ldsba( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldsb_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1397 void ldsba( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldsb_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1398 void ldsha( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldsh_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1399 void ldsha( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldsh_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1400 void ldswa( Register s1, Register s2, int ia, Register d ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(ldsw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1401 void ldswa( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(ldsw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1402 void lduba( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldub_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1403 void lduba( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldub_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1404 void lduha( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(lduh_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1405 void lduha( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(lduh_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1406 void lduwa( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(lduw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1407 void lduwa( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(lduw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1408 void ldxa( Register s1, Register s2, int ia, Register d ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(ldx_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1409 void ldxa( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(ldx_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1410 void ldda( Register s1, Register s2, int ia, Register d ) { v9_dep(); emit_long( op(ldst_op) | rd(d) | op3(ldd_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1411 void ldda( Register s1, int simm13a, Register d ) { v9_dep(); emit_long( op(ldst_op) | rd(d) | op3(ldd_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1412 1413 // pp 179 1414 1415 inline void ldstub( Register s1, Register s2, Register d ); 1416 inline void ldstub( Register s1, int simm13a, Register d); 1417 1418 // pp 180 1419 1420 void ldstuba( Register s1, Register s2, int ia, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldstub_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1421 void ldstuba( Register s1, int simm13a, Register d ) { emit_long( op(ldst_op) | rd(d) | op3(ldstub_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1422 1423 // pp 181 1424 1425 void and3( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3 ) | rs1(s1) | rs2(s2) ); } 1426 void and3( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1427 void andcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1428 void andcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(and_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1429 void andn( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 ) | rs1(s1) | rs2(s2) ); } 1430 void andn( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1431 void andn( Register s1, RegisterOrConstant s2, Register d); 1432 void andncc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1433 void andncc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(andn_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1434 void or3( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3 ) | rs1(s1) | rs2(s2) ); } 1435 void or3( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1436 void orcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1437 void orcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(or_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1438 void orn( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3) | rs1(s1) | rs2(s2) ); } 1439 void orn( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1440 void orncc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1441 void orncc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(orn_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1442 void xor3( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3 ) | rs1(s1) | rs2(s2) ); } 1443 void xor3( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1444 void xorcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1445 void xorcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xor_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1446 void xnor( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 ) | rs1(s1) | rs2(s2) ); } 1447 void xnor( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1448 void xnorcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1449 void xnorcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(xnor_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1450 1451 // pp 183 1452 1453 void membar( Membar_mask_bits const7a ) { v9_only(); emit_long( op(arith_op) | op3(membar_op3) | rs1(O7) | immed(true) | u_field( int(const7a), 6, 0)); } 1454 1455 // pp 185 1456 1457 void fmov( FloatRegisterImpl::Width w, Condition c, bool floatCC, CC cca, FloatRegister s2, FloatRegister d ) { v9_only(); emit_long( op(arith_op) | fd(d, w) | op3(fpop2_op3) | cond_mov(c) | opf_cc(cca, floatCC) | opf_low6(w) | fs2(s2, w)); } 1458 1459 // pp 189 1460 1461 void fmov( FloatRegisterImpl::Width w, RCondition c, Register s1, FloatRegister s2, FloatRegister d ) { v9_only(); emit_long( op(arith_op) | fd(d, w) | op3(fpop2_op3) | rs1(s1) | rcond(c) | opf_low5(4 + w) | fs2(s2, w)); } 1462 1463 // pp 191 1464 1465 void movcc( Condition c, bool floatCC, CC cca, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(movcc_op3) | mov_cc(cca, floatCC) | cond_mov(c) | rs2(s2) ); } 1466 void movcc( Condition c, bool floatCC, CC cca, int simm11a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(movcc_op3) | mov_cc(cca, floatCC) | cond_mov(c) | immed(true) | simm(simm11a, 11) ); } 1467 1468 // pp 195 1469 1470 void movr( RCondition c, Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(movr_op3) | rs1(s1) | rcond(c) | rs2(s2) ); } 1471 void movr( RCondition c, Register s1, int simm10a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(movr_op3) | rs1(s1) | rcond(c) | immed(true) | simm(simm10a, 10) ); } 1472 1473 // pp 196 1474 1475 void mulx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(mulx_op3 ) | rs1(s1) | rs2(s2) ); } 1476 void mulx( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(mulx_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1477 void sdivx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sdivx_op3) | rs1(s1) | rs2(s2) ); } 1478 void sdivx( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sdivx_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1479 void udivx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(udivx_op3) | rs1(s1) | rs2(s2) ); } 1480 void udivx( Register s1, int simm13a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(udivx_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1481 1482 // pp 197 1483 1484 void umul( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 ) | rs1(s1) | rs2(s2) ); } 1485 void umul( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1486 void smul( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 ) | rs1(s1) | rs2(s2) ); } 1487 void smul( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1488 void umulcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1489 void umulcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(umul_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1490 void smulcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1491 void smulcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(smul_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1492 1493 // pp 199 1494 1495 void mulscc( Register s1, Register s2, Register d ) { v9_dep(); emit_long( op(arith_op) | rd(d) | op3(mulscc_op3) | rs1(s1) | rs2(s2) ); } 1496 void mulscc( Register s1, int simm13a, Register d ) { v9_dep(); emit_long( op(arith_op) | rd(d) | op3(mulscc_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1497 1498 // pp 201 1499 1500 void nop() { emit_long( op(branch_op) | op2(sethi_op2) ); } 1501 1502 1503 // pp 202 1504 1505 void popc( Register s, Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(popc_op3) | rs2(s)); } 1506 void popc( int simm13a, Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(popc_op3) | immed(true) | simm(simm13a, 13)); } 1507 1508 // pp 203 1509 1510 void prefetch( Register s1, Register s2, PrefetchFcn f); 1511 void prefetch( Register s1, int simm13a, PrefetchFcn f); 1512 void prefetcha( Register s1, Register s2, int ia, PrefetchFcn f ) { v9_only(); emit_long( op(ldst_op) | fcn(f) | op3(prefetch_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1513 void prefetcha( Register s1, int simm13a, PrefetchFcn f ) { v9_only(); emit_long( op(ldst_op) | fcn(f) | op3(prefetch_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1514 1515 inline void prefetch(const Address& a, PrefetchFcn F, int offset = 0); 1516 1517 // pp 208 1518 1519 // not implementing read privileged register 1520 1521 inline void rdy( Register d) { v9_dep(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(0, 18, 14)); } 1522 inline void rdccr( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(2, 18, 14)); } 1523 inline void rdasi( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(3, 18, 14)); } 1524 inline void rdtick( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(4, 18, 14)); } // Spoon! 1525 inline void rdpc( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(5, 18, 14)); } 1526 inline void rdfprs( Register d) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(rdreg_op3) | u_field(6, 18, 14)); } 1527 1528 // pp 213 1529 1530 inline void rett( Register s1, Register s2); 1531 inline void rett( Register s1, int simm13a, relocInfo::relocType rt = relocInfo::none); 1532 1533 // pp 214 1534 1535 void save( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(save_op3) | rs1(s1) | rs2(s2) ); } 1536 void save( Register s1, int simm13a, Register d ) { 1537 // make sure frame is at least large enough for the register save area 1538 assert(-simm13a >= 16 * wordSize, "frame too small"); 1539 emit_long( op(arith_op) | rd(d) | op3(save_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); 1540 } 1541 1542 void restore( Register s1 = G0, Register s2 = G0, Register d = G0 ) { emit_long( op(arith_op) | rd(d) | op3(restore_op3) | rs1(s1) | rs2(s2) ); } 1543 void restore( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(restore_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1544 1545 // pp 216 1546 1547 void saved() { v9_only(); emit_long( op(arith_op) | fcn(0) | op3(saved_op3)); } 1548 void restored() { v9_only(); emit_long( op(arith_op) | fcn(1) | op3(saved_op3)); } 1549 1550 // pp 217 1551 1552 inline void sethi( int imm22a, Register d, RelocationHolder const& rspec = RelocationHolder() ); 1553 // pp 218 1554 1555 void sll( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(0) | rs2(s2) ); } 1556 void sll( Register s1, int imm5a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); } 1557 void srl( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(0) | rs2(s2) ); } 1558 void srl( Register s1, int imm5a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); } 1559 void sra( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(0) | rs2(s2) ); } 1560 void sra( Register s1, int imm5a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(0) | immed(true) | u_field(imm5a, 4, 0) ); } 1561 1562 void sllx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(1) | rs2(s2) ); } 1563 void sllx( Register s1, int imm6a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sll_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); } 1564 void srlx( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(1) | rs2(s2) ); } 1565 void srlx( Register s1, int imm6a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(srl_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); } 1566 void srax( Register s1, Register s2, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(1) | rs2(s2) ); } 1567 void srax( Register s1, int imm6a, Register d ) { v9_only(); emit_long( op(arith_op) | rd(d) | op3(sra_op3) | rs1(s1) | sx(1) | immed(true) | u_field(imm6a, 5, 0) ); } 1568 1569 // pp 220 1570 1571 void sir( int simm13a ) { emit_long( op(arith_op) | fcn(15) | op3(sir_op3) | immed(true) | simm(simm13a, 13)); } 1572 1573 // pp 221 1574 1575 void stbar() { emit_long( op(arith_op) | op3(membar_op3) | u_field(15, 18, 14)); } 1576 1577 // pp 222 1578 1579 inline void stf( FloatRegisterImpl::Width w, FloatRegister d, Register s1, RegisterOrConstant s2); 1580 inline void stf( FloatRegisterImpl::Width w, FloatRegister d, Register s1, Register s2); 1581 inline void stf( FloatRegisterImpl::Width w, FloatRegister d, Register s1, int simm13a); 1582 inline void stf( FloatRegisterImpl::Width w, FloatRegister d, const Address& a, int offset = 0); 1583 1584 inline void stfsr( Register s1, Register s2 ); 1585 inline void stfsr( Register s1, int simm13a); 1586 inline void stxfsr( Register s1, Register s2 ); 1587 inline void stxfsr( Register s1, int simm13a); 1588 1589 // pp 224 1590 1591 void stfa( FloatRegisterImpl::Width w, FloatRegister d, Register s1, Register s2, int ia ) { v9_only(); emit_long( op(ldst_op) | fd(d, w) | alt_op3(stf_op3 | alt_bit_op3, w) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1592 void stfa( FloatRegisterImpl::Width w, FloatRegister d, Register s1, int simm13a ) { v9_only(); emit_long( op(ldst_op) | fd(d, w) | alt_op3(stf_op3 | alt_bit_op3, w) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1593 1594 // p 226 1595 1596 inline void stb( Register d, Register s1, Register s2 ); 1597 inline void stb( Register d, Register s1, int simm13a); 1598 inline void sth( Register d, Register s1, Register s2 ); 1599 inline void sth( Register d, Register s1, int simm13a); 1600 inline void stw( Register d, Register s1, Register s2 ); 1601 inline void stw( Register d, Register s1, int simm13a); 1602 inline void st( Register d, Register s1, Register s2 ); 1603 inline void st( Register d, Register s1, int simm13a); 1604 inline void stx( Register d, Register s1, Register s2 ); 1605 inline void stx( Register d, Register s1, int simm13a); 1606 inline void std( Register d, Register s1, Register s2 ); 1607 inline void std( Register d, Register s1, int simm13a); 1608 1609 #ifdef ASSERT 1610 // ByteSize is only a class when ASSERT is defined, otherwise it's an int. 1611 inline void st( Register d, Register s1, ByteSize simm13a); 1612 #endif 1613 1614 inline void stb( Register d, const Address& a, int offset = 0 ); 1615 inline void sth( Register d, const Address& a, int offset = 0 ); 1616 inline void stw( Register d, const Address& a, int offset = 0 ); 1617 inline void stx( Register d, const Address& a, int offset = 0 ); 1618 inline void st( Register d, const Address& a, int offset = 0 ); 1619 inline void std( Register d, const Address& a, int offset = 0 ); 1620 1621 inline void stb( Register d, Register s1, RegisterOrConstant s2 ); 1622 inline void sth( Register d, Register s1, RegisterOrConstant s2 ); 1623 inline void stw( Register d, Register s1, RegisterOrConstant s2 ); 1624 inline void stx( Register d, Register s1, RegisterOrConstant s2 ); 1625 inline void std( Register d, Register s1, RegisterOrConstant s2 ); 1626 inline void st( Register d, Register s1, RegisterOrConstant s2 ); 1627 1628 // pp 177 1629 1630 void stba( Register d, Register s1, Register s2, int ia ) { emit_long( op(ldst_op) | rd(d) | op3(stb_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1631 void stba( Register d, Register s1, int simm13a ) { emit_long( op(ldst_op) | rd(d) | op3(stb_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1632 void stha( Register d, Register s1, Register s2, int ia ) { emit_long( op(ldst_op) | rd(d) | op3(sth_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1633 void stha( Register d, Register s1, int simm13a ) { emit_long( op(ldst_op) | rd(d) | op3(sth_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1634 void stwa( Register d, Register s1, Register s2, int ia ) { emit_long( op(ldst_op) | rd(d) | op3(stw_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1635 void stwa( Register d, Register s1, int simm13a ) { emit_long( op(ldst_op) | rd(d) | op3(stw_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1636 void stxa( Register d, Register s1, Register s2, int ia ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(stx_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1637 void stxa( Register d, Register s1, int simm13a ) { v9_only(); emit_long( op(ldst_op) | rd(d) | op3(stx_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1638 void stda( Register d, Register s1, Register s2, int ia ) { emit_long( op(ldst_op) | rd(d) | op3(std_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1639 void stda( Register d, Register s1, int simm13a ) { emit_long( op(ldst_op) | rd(d) | op3(std_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1640 1641 // pp 97 (v8) 1642 1643 inline void stc( int crd, Register s1, Register s2 ); 1644 inline void stc( int crd, Register s1, int simm13a); 1645 inline void stdc( int crd, Register s1, Register s2 ); 1646 inline void stdc( int crd, Register s1, int simm13a); 1647 inline void stcsr( int crd, Register s1, Register s2 ); 1648 inline void stcsr( int crd, Register s1, int simm13a); 1649 inline void stdcq( int crd, Register s1, Register s2 ); 1650 inline void stdcq( int crd, Register s1, int simm13a); 1651 1652 // pp 230 1653 1654 void sub( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 ) | rs1(s1) | rs2(s2) ); } 1655 void sub( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1656 1657 // Note: offset is added to s2. 1658 inline void sub(Register s1, RegisterOrConstant s2, Register d, int offset = 0); 1659 1660 void subcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 | cc_bit_op3 ) | rs1(s1) | rs2(s2) ); } 1661 void subcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(sub_op3 | cc_bit_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1662 void subc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 ) | rs1(s1) | rs2(s2) ); } 1663 void subc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1664 void subccc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 | cc_bit_op3) | rs1(s1) | rs2(s2) ); } 1665 void subccc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(subc_op3 | cc_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1666 1667 // pp 231 1668 1669 inline void swap( Register s1, Register s2, Register d ); 1670 inline void swap( Register s1, int simm13a, Register d); 1671 inline void swap( Address& a, Register d, int offset = 0 ); 1672 1673 // pp 232 1674 1675 void swapa( Register s1, Register s2, int ia, Register d ) { v9_dep(); emit_long( op(ldst_op) | rd(d) | op3(swap_op3 | alt_bit_op3) | rs1(s1) | imm_asi(ia) | rs2(s2) ); } 1676 void swapa( Register s1, int simm13a, Register d ) { v9_dep(); emit_long( op(ldst_op) | rd(d) | op3(swap_op3 | alt_bit_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1677 1678 // pp 234, note op in book is wrong, see pp 268 1679 1680 void taddcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(taddcc_op3 ) | rs1(s1) | rs2(s2) ); } 1681 void taddcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(taddcc_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1682 void taddcctv( Register s1, Register s2, Register d ) { v9_dep(); emit_long( op(arith_op) | rd(d) | op3(taddcctv_op3) | rs1(s1) | rs2(s2) ); } 1683 void taddcctv( Register s1, int simm13a, Register d ) { v9_dep(); emit_long( op(arith_op) | rd(d) | op3(taddcctv_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1684 1685 // pp 235 1686 1687 void tsubcc( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcc_op3 ) | rs1(s1) | rs2(s2) ); } 1688 void tsubcc( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcc_op3 ) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1689 void tsubcctv( Register s1, Register s2, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcctv_op3) | rs1(s1) | rs2(s2) ); } 1690 void tsubcctv( Register s1, int simm13a, Register d ) { emit_long( op(arith_op) | rd(d) | op3(tsubcctv_op3) | rs1(s1) | immed(true) | simm(simm13a, 13) ); } 1691 1692 // pp 237 1693 1694 void trap( Condition c, CC cc, Register s1, Register s2 ) { v8_no_cc(cc); emit_long( op(arith_op) | cond(c) | op3(trap_op3) | rs1(s1) | trapcc(cc) | rs2(s2)); } 1695 void trap( Condition c, CC cc, Register s1, int trapa ) { v8_no_cc(cc); emit_long( op(arith_op) | cond(c) | op3(trap_op3) | rs1(s1) | trapcc(cc) | immed(true) | u_field(trapa, 6, 0)); } 1696 // simple uncond. trap 1697 void trap( int trapa ) { trap( always, icc, G0, trapa ); } 1698 1699 // pp 239 omit write priv register for now 1700 1701 inline void wry( Register d) { v9_dep(); emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(0, 29, 25)); } 1702 inline void wrccr(Register s) { v9_only(); emit_long( op(arith_op) | rs1(s) | op3(wrreg_op3) | u_field(2, 29, 25)); } 1703 inline void wrccr(Register s, int simm13a) { v9_only(); emit_long( op(arith_op) | 1704 rs1(s) | 1705 op3(wrreg_op3) | 1706 u_field(2, 29, 25) | 1707 u_field(1, 13, 13) | 1708 simm(simm13a, 13)); } 1709 inline void wrasi( Register d) { v9_only(); emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(3, 29, 25)); } 1710 inline void wrfprs( Register d) { v9_only(); emit_long( op(arith_op) | rs1(d) | op3(wrreg_op3) | u_field(6, 29, 25)); } 1711 1712 // For a given register condition, return the appropriate condition code 1713 // Condition (the one you would use to get the same effect after "tst" on 1714 // the target register.) 1715 Assembler::Condition reg_cond_to_cc_cond(RCondition in); 1716 1717 1718 // Creation 1719 Assembler(CodeBuffer* code) : AbstractAssembler(code) { 1720 #ifdef CHECK_DELAY 1721 delay_state = no_delay; 1722 #endif 1723 } 1724 1725 // Testing 1726 #ifndef PRODUCT 1727 void test_v9(); 1728 void test_v8_onlys(); 1729 #endif 1730 }; 1731 1732 1733 class RegistersForDebugging : public StackObj { 1734 public: 1735 intptr_t i[8], l[8], o[8], g[8]; 1736 float f[32]; 1737 double d[32]; 1738 1739 void print(outputStream* s); 1740 1741 static int i_offset(int j) { return offset_of(RegistersForDebugging, i[j]); } 1742 static int l_offset(int j) { return offset_of(RegistersForDebugging, l[j]); } 1743 static int o_offset(int j) { return offset_of(RegistersForDebugging, o[j]); } 1744 static int g_offset(int j) { return offset_of(RegistersForDebugging, g[j]); } 1745 static int f_offset(int j) { return offset_of(RegistersForDebugging, f[j]); } 1746 static int d_offset(int j) { return offset_of(RegistersForDebugging, d[j / 2]); } 1747 1748 // gen asm code to save regs 1749 static void save_registers(MacroAssembler* a); 1750 1751 // restore global registers in case C code disturbed them 1752 static void restore_registers(MacroAssembler* a, Register r); 1753 1754 1755 }; 1756 1757 1758 // MacroAssembler extends Assembler by a few frequently used macros. 1759 // 1760 // Most of the standard SPARC synthetic ops are defined here. 1761 // Instructions for which a 'better' code sequence exists depending 1762 // on arguments should also go in here. 1763 1764 #define JMP2(r1, r2) jmp(r1, r2, __FILE__, __LINE__) 1765 #define JMP(r1, off) jmp(r1, off, __FILE__, __LINE__) 1766 #define JUMP(a, temp, off) jump(a, temp, off, __FILE__, __LINE__) 1767 #define JUMPL(a, temp, d, off) jumpl(a, temp, d, off, __FILE__, __LINE__) 1768 1769 1770 class MacroAssembler: public Assembler { 1771 protected: 1772 // Support for VM calls 1773 // This is the base routine called by the different versions of call_VM_leaf. The interpreter 1774 // may customize this version by overriding it for its purposes (e.g., to save/restore 1775 // additional registers when doing a VM call). 1776 #ifdef CC_INTERP 1777 #define VIRTUAL 1778 #else 1779 #define VIRTUAL virtual 1780 #endif 1781 1782 VIRTUAL void call_VM_leaf_base(Register thread_cache, address entry_point, int number_of_arguments); 1783 1784 // 1785 // It is imperative that all calls into the VM are handled via the call_VM macros. 1786 // They make sure that the stack linkage is setup correctly. call_VM's correspond 1787 // to ENTRY/ENTRY_X entry points while call_VM_leaf's correspond to LEAF entry points. 1788 // 1789 // This is the base routine called by the different versions of call_VM. The interpreter 1790 // may customize this version by overriding it for its purposes (e.g., to save/restore 1791 // additional registers when doing a VM call). 1792 // 1793 // A non-volatile java_thread_cache register should be specified so 1794 // that the G2_thread value can be preserved across the call. 1795 // (If java_thread_cache is noreg, then a slow get_thread call 1796 // will re-initialize the G2_thread.) call_VM_base returns the register that contains the 1797 // thread. 1798 // 1799 // If no last_java_sp is specified (noreg) than SP will be used instead. 1800 1801 virtual void call_VM_base( 1802 Register oop_result, // where an oop-result ends up if any; use noreg otherwise 1803 Register java_thread_cache, // the thread if computed before ; use noreg otherwise 1804 Register last_java_sp, // to set up last_Java_frame in stubs; use noreg otherwise 1805 address entry_point, // the entry point 1806 int number_of_arguments, // the number of arguments (w/o thread) to pop after call 1807 bool check_exception=true // flag which indicates if exception should be checked 1808 ); 1809 1810 // This routine should emit JVMTI PopFrame and ForceEarlyReturn handling code. 1811 // The implementation is only non-empty for the InterpreterMacroAssembler, 1812 // as only the interpreter handles and ForceEarlyReturn PopFrame requests. 1813 virtual void check_and_handle_popframe(Register scratch_reg); 1814 virtual void check_and_handle_earlyret(Register scratch_reg); 1815 1816 public: 1817 MacroAssembler(CodeBuffer* code) : Assembler(code) {} 1818 1819 // Support for NULL-checks 1820 // 1821 // Generates code that causes a NULL OS exception if the content of reg is NULL. 1822 // If the accessed location is M[reg + offset] and the offset is known, provide the 1823 // offset. No explicit code generation is needed if the offset is within a certain 1824 // range (0 <= offset <= page_size). 1825 // 1826 // %%%%%% Currently not done for SPARC 1827 1828 void null_check(Register reg, int offset = -1); 1829 static bool needs_explicit_null_check(intptr_t offset); 1830 1831 // support for delayed instructions 1832 MacroAssembler* delayed() { Assembler::delayed(); return this; } 1833 1834 // branches that use right instruction for v8 vs. v9 1835 inline void br( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1836 inline void br( Condition c, bool a, Predict p, Label& L ); 1837 1838 inline void fb( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1839 inline void fb( Condition c, bool a, Predict p, Label& L ); 1840 1841 // compares register with zero and branches (V9 and V8 instructions) 1842 void br_zero( Condition c, bool a, Predict p, Register s1, Label& L); 1843 // Compares a pointer register with zero and branches on (not)null. 1844 // Does a test & branch on 32-bit systems and a register-branch on 64-bit. 1845 void br_null ( Register s1, bool a, Predict p, Label& L ); 1846 void br_notnull( Register s1, bool a, Predict p, Label& L ); 1847 1848 // These versions will do the most efficient thing on v8 and v9. Perhaps 1849 // this is what the routine above was meant to do, but it didn't (and 1850 // didn't cover both target address kinds.) 1851 void br_on_reg_cond( RCondition c, bool a, Predict p, Register s1, address d, relocInfo::relocType rt = relocInfo::none ); 1852 void br_on_reg_cond( RCondition c, bool a, Predict p, Register s1, Label& L); 1853 1854 inline void bp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1855 inline void bp( Condition c, bool a, CC cc, Predict p, Label& L ); 1856 1857 // Branch that tests xcc in LP64 and icc in !LP64 1858 inline void brx( Condition c, bool a, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1859 inline void brx( Condition c, bool a, Predict p, Label& L ); 1860 1861 // unconditional short branch 1862 inline void ba( bool a, Label& L ); 1863 1864 // Branch that tests fp condition codes 1865 inline void fbp( Condition c, bool a, CC cc, Predict p, address d, relocInfo::relocType rt = relocInfo::none ); 1866 inline void fbp( Condition c, bool a, CC cc, Predict p, Label& L ); 1867 1868 // get PC the best way 1869 inline int get_pc( Register d ); 1870 1871 // Sparc shorthands(pp 85, V8 manual, pp 289 V9 manual) 1872 inline void cmp( Register s1, Register s2 ) { subcc( s1, s2, G0 ); } 1873 inline void cmp( Register s1, int simm13a ) { subcc( s1, simm13a, G0 ); } 1874 1875 inline void jmp( Register s1, Register s2 ); 1876 inline void jmp( Register s1, int simm13a, RelocationHolder const& rspec = RelocationHolder() ); 1877 1878 // Check if the call target is out of wdisp30 range (relative to the code cache) 1879 static inline bool is_far_target(address d); 1880 inline void call( address d, relocInfo::relocType rt = relocInfo::runtime_call_type ); 1881 inline void call( Label& L, relocInfo::relocType rt = relocInfo::runtime_call_type ); 1882 inline void callr( Register s1, Register s2 ); 1883 inline void callr( Register s1, int simm13a, RelocationHolder const& rspec = RelocationHolder() ); 1884 1885 // Emits nothing on V8 1886 inline void iprefetch( address d, relocInfo::relocType rt = relocInfo::none ); 1887 inline void iprefetch( Label& L); 1888 1889 inline void tst( Register s ) { orcc( G0, s, G0 ); } 1890 1891 #ifdef PRODUCT 1892 inline void ret( bool trace = TraceJumps ) { if (trace) { 1893 mov(I7, O7); // traceable register 1894 JMP(O7, 2 * BytesPerInstWord); 1895 } else { 1896 jmpl( I7, 2 * BytesPerInstWord, G0 ); 1897 } 1898 } 1899 1900 inline void retl( bool trace = TraceJumps ) { if (trace) JMP(O7, 2 * BytesPerInstWord); 1901 else jmpl( O7, 2 * BytesPerInstWord, G0 ); } 1902 #else 1903 void ret( bool trace = TraceJumps ); 1904 void retl( bool trace = TraceJumps ); 1905 #endif /* PRODUCT */ 1906 1907 // Required platform-specific helpers for Label::patch_instructions. 1908 // They _shadow_ the declarations in AbstractAssembler, which are undefined. 1909 void pd_patch_instruction(address branch, address target); 1910 #ifndef PRODUCT 1911 static void pd_print_patched_instruction(address branch); 1912 #endif 1913 1914 // sethi Macro handles optimizations and relocations 1915 private: 1916 void internal_sethi(const AddressLiteral& addrlit, Register d, bool ForceRelocatable); 1917 public: 1918 void sethi(const AddressLiteral& addrlit, Register d); 1919 void patchable_sethi(const AddressLiteral& addrlit, Register d); 1920 1921 // compute the number of instructions for a sethi/set 1922 static int insts_for_sethi( address a, bool worst_case = false ); 1923 static int worst_case_insts_for_set(); 1924 1925 // set may be either setsw or setuw (high 32 bits may be zero or sign) 1926 private: 1927 void internal_set(const AddressLiteral& al, Register d, bool ForceRelocatable); 1928 static int insts_for_internal_set(intptr_t value); 1929 public: 1930 void set(const AddressLiteral& addrlit, Register d); 1931 void set(intptr_t value, Register d); 1932 void set(address addr, Register d, RelocationHolder const& rspec); 1933 static int insts_for_set(intptr_t value) { return insts_for_internal_set(value); } 1934 1935 void patchable_set(const AddressLiteral& addrlit, Register d); 1936 void patchable_set(intptr_t value, Register d); 1937 void set64(jlong value, Register d, Register tmp); 1938 static int insts_for_set64(jlong value); 1939 1940 // sign-extend 32 to 64 1941 inline void signx( Register s, Register d ) { sra( s, G0, d); } 1942 inline void signx( Register d ) { sra( d, G0, d); } 1943 1944 inline void not1( Register s, Register d ) { xnor( s, G0, d ); } 1945 inline void not1( Register d ) { xnor( d, G0, d ); } 1946 1947 inline void neg( Register s, Register d ) { sub( G0, s, d ); } 1948 inline void neg( Register d ) { sub( G0, d, d ); } 1949 1950 inline void cas( Register s1, Register s2, Register d) { casa( s1, s2, d, ASI_PRIMARY); } 1951 inline void casx( Register s1, Register s2, Register d) { casxa(s1, s2, d, ASI_PRIMARY); } 1952 // Functions for isolating 64 bit atomic swaps for LP64 1953 // cas_ptr will perform cas for 32 bit VM's and casx for 64 bit VM's 1954 inline void cas_ptr( Register s1, Register s2, Register d) { 1955 #ifdef _LP64 1956 casx( s1, s2, d ); 1957 #else 1958 cas( s1, s2, d ); 1959 #endif 1960 } 1961 1962 // Functions for isolating 64 bit shifts for LP64 1963 inline void sll_ptr( Register s1, Register s2, Register d ); 1964 inline void sll_ptr( Register s1, int imm6a, Register d ); 1965 inline void sll_ptr( Register s1, RegisterOrConstant s2, Register d ); 1966 inline void srl_ptr( Register s1, Register s2, Register d ); 1967 inline void srl_ptr( Register s1, int imm6a, Register d ); 1968 1969 // little-endian 1970 inline void casl( Register s1, Register s2, Register d) { casa( s1, s2, d, ASI_PRIMARY_LITTLE); } 1971 inline void casxl( Register s1, Register s2, Register d) { casxa(s1, s2, d, ASI_PRIMARY_LITTLE); } 1972 1973 inline void inc( Register d, int const13 = 1 ) { add( d, const13, d); } 1974 inline void inccc( Register d, int const13 = 1 ) { addcc( d, const13, d); } 1975 1976 inline void dec( Register d, int const13 = 1 ) { sub( d, const13, d); } 1977 inline void deccc( Register d, int const13 = 1 ) { subcc( d, const13, d); } 1978 1979 inline void btst( Register s1, Register s2 ) { andcc( s1, s2, G0 ); } 1980 inline void btst( int simm13a, Register s ) { andcc( s, simm13a, G0 ); } 1981 1982 inline void bset( Register s1, Register s2 ) { or3( s1, s2, s2 ); } 1983 inline void bset( int simm13a, Register s ) { or3( s, simm13a, s ); } 1984 1985 inline void bclr( Register s1, Register s2 ) { andn( s1, s2, s2 ); } 1986 inline void bclr( int simm13a, Register s ) { andn( s, simm13a, s ); } 1987 1988 inline void btog( Register s1, Register s2 ) { xor3( s1, s2, s2 ); } 1989 inline void btog( int simm13a, Register s ) { xor3( s, simm13a, s ); } 1990 1991 inline void clr( Register d ) { or3( G0, G0, d ); } 1992 1993 inline void clrb( Register s1, Register s2); 1994 inline void clrh( Register s1, Register s2); 1995 inline void clr( Register s1, Register s2); 1996 inline void clrx( Register s1, Register s2); 1997 1998 inline void clrb( Register s1, int simm13a); 1999 inline void clrh( Register s1, int simm13a); 2000 inline void clr( Register s1, int simm13a); 2001 inline void clrx( Register s1, int simm13a); 2002 2003 // copy & clear upper word 2004 inline void clruw( Register s, Register d ) { srl( s, G0, d); } 2005 // clear upper word 2006 inline void clruwu( Register d ) { srl( d, G0, d); } 2007 2008 // membar psuedo instruction. takes into account target memory model. 2009 inline void membar( Assembler::Membar_mask_bits const7a ); 2010 2011 // returns if membar generates anything. 2012 inline bool membar_has_effect( Assembler::Membar_mask_bits const7a ); 2013 2014 // mov pseudo instructions 2015 inline void mov( Register s, Register d) { 2016 if ( s != d ) or3( G0, s, d); 2017 else assert_not_delayed(); // Put something useful in the delay slot! 2018 } 2019 2020 inline void mov_or_nop( Register s, Register d) { 2021 if ( s != d ) or3( G0, s, d); 2022 else nop(); 2023 } 2024 2025 inline void mov( int simm13a, Register d) { or3( G0, simm13a, d); } 2026 2027 // address pseudos: make these names unlike instruction names to avoid confusion 2028 inline intptr_t load_pc_address( Register reg, int bytes_to_skip ); 2029 inline void load_contents(const AddressLiteral& addrlit, Register d, int offset = 0); 2030 inline void load_ptr_contents(const AddressLiteral& addrlit, Register d, int offset = 0); 2031 inline void store_contents(Register s, const AddressLiteral& addrlit, Register temp, int offset = 0); 2032 inline void store_ptr_contents(Register s, const AddressLiteral& addrlit, Register temp, int offset = 0); 2033 inline void jumpl_to(const AddressLiteral& addrlit, Register temp, Register d, int offset = 0); 2034 inline void jump_to(const AddressLiteral& addrlit, Register temp, int offset = 0); 2035 inline void jump_indirect_to(Address& a, Register temp, int ld_offset = 0, int jmp_offset = 0); 2036 2037 // ring buffer traceable jumps 2038 2039 void jmp2( Register r1, Register r2, const char* file, int line ); 2040 void jmp ( Register r1, int offset, const char* file, int line ); 2041 2042 void jumpl(const AddressLiteral& addrlit, Register temp, Register d, int offset, const char* file, int line); 2043 void jump (const AddressLiteral& addrlit, Register temp, int offset, const char* file, int line); 2044 2045 2046 // argument pseudos: 2047 2048 inline void load_argument( Argument& a, Register d ); 2049 inline void store_argument( Register s, Argument& a ); 2050 inline void store_ptr_argument( Register s, Argument& a ); 2051 inline void store_float_argument( FloatRegister s, Argument& a ); 2052 inline void store_double_argument( FloatRegister s, Argument& a ); 2053 inline void store_long_argument( Register s, Argument& a ); 2054 2055 // handy macros: 2056 2057 inline void round_to( Register r, int modulus ) { 2058 assert_not_delayed(); 2059 inc( r, modulus - 1 ); 2060 and3( r, -modulus, r ); 2061 } 2062 2063 // -------------------------------------------------- 2064 2065 // Functions for isolating 64 bit loads for LP64 2066 // ld_ptr will perform ld for 32 bit VM's and ldx for 64 bit VM's 2067 // st_ptr will perform st for 32 bit VM's and stx for 64 bit VM's 2068 inline void ld_ptr(Register s1, Register s2, Register d); 2069 inline void ld_ptr(Register s1, int simm13a, Register d); 2070 inline void ld_ptr(Register s1, RegisterOrConstant s2, Register d); 2071 inline void ld_ptr(const Address& a, Register d, int offset = 0); 2072 inline void st_ptr(Register d, Register s1, Register s2); 2073 inline void st_ptr(Register d, Register s1, int simm13a); 2074 inline void st_ptr(Register d, Register s1, RegisterOrConstant s2); 2075 inline void st_ptr(Register d, const Address& a, int offset = 0); 2076 2077 #ifdef ASSERT 2078 // ByteSize is only a class when ASSERT is defined, otherwise it's an int. 2079 inline void ld_ptr(Register s1, ByteSize simm13a, Register d); 2080 inline void st_ptr(Register d, Register s1, ByteSize simm13a); 2081 #endif 2082 2083 // ld_long will perform ldd for 32 bit VM's and ldx for 64 bit VM's 2084 // st_long will perform std for 32 bit VM's and stx for 64 bit VM's 2085 inline void ld_long(Register s1, Register s2, Register d); 2086 inline void ld_long(Register s1, int simm13a, Register d); 2087 inline void ld_long(Register s1, RegisterOrConstant s2, Register d); 2088 inline void ld_long(const Address& a, Register d, int offset = 0); 2089 inline void st_long(Register d, Register s1, Register s2); 2090 inline void st_long(Register d, Register s1, int simm13a); 2091 inline void st_long(Register d, Register s1, RegisterOrConstant s2); 2092 inline void st_long(Register d, const Address& a, int offset = 0); 2093 2094 // Helpers for address formation. 2095 // - They emit only a move if s2 is a constant zero. 2096 // - If dest is a constant and either s1 or s2 is a register, the temp argument is required and becomes the result. 2097 // - If dest is a register and either s1 or s2 is a non-simm13 constant, the temp argument is required and used to materialize the constant. 2098 RegisterOrConstant regcon_andn_ptr(RegisterOrConstant s1, RegisterOrConstant s2, RegisterOrConstant d, Register temp = noreg); 2099 RegisterOrConstant regcon_inc_ptr( RegisterOrConstant s1, RegisterOrConstant s2, RegisterOrConstant d, Register temp = noreg); 2100 RegisterOrConstant regcon_sll_ptr( RegisterOrConstant s1, RegisterOrConstant s2, RegisterOrConstant d, Register temp = noreg); 2101 2102 RegisterOrConstant ensure_simm13_or_reg(RegisterOrConstant src, Register temp) { 2103 if (is_simm13(src.constant_or_zero())) 2104 return src; // register or short constant 2105 guarantee(temp != noreg, "constant offset overflow"); 2106 set(src.as_constant(), temp); 2107 return temp; 2108 } 2109 2110 // -------------------------------------------------- 2111 2112 public: 2113 // traps as per trap.h (SPARC ABI?) 2114 2115 void breakpoint_trap(); 2116 void breakpoint_trap(Condition c, CC cc = icc); 2117 void flush_windows_trap(); 2118 void clean_windows_trap(); 2119 void get_psr_trap(); 2120 void set_psr_trap(); 2121 2122 // V8/V9 flush_windows 2123 void flush_windows(); 2124 2125 // Support for serializing memory accesses between threads 2126 void serialize_memory(Register thread, Register tmp1, Register tmp2); 2127 2128 // Stack frame creation/removal 2129 void enter(); 2130 void leave(); 2131 2132 // V8/V9 integer multiply 2133 void mult(Register s1, Register s2, Register d); 2134 void mult(Register s1, int simm13a, Register d); 2135 2136 // V8/V9 read and write of condition codes. 2137 void read_ccr(Register d); 2138 void write_ccr(Register s); 2139 2140 // Manipulation of C++ bools 2141 // These are idioms to flag the need for care with accessing bools but on 2142 // this platform we assume byte size 2143 2144 inline void stbool(Register d, const Address& a) { stb(d, a); } 2145 inline void ldbool(const Address& a, Register d) { ldsb(a, d); } 2146 inline void tstbool( Register s ) { tst(s); } 2147 inline void movbool( bool boolconst, Register d) { mov( (int) boolconst, d); } 2148 2149 // klass oop manipulations if compressed 2150 void load_klass(Register src_oop, Register klass); 2151 void store_klass(Register klass, Register dst_oop); 2152 void store_klass_gap(Register s, Register dst_oop); 2153 2154 // oop manipulations 2155 void load_heap_oop(const Address& s, Register d); 2156 void load_heap_oop(Register s1, Register s2, Register d); 2157 void load_heap_oop(Register s1, int simm13a, Register d); 2158 void load_heap_oop(Register s1, RegisterOrConstant s2, Register d); 2159 void store_heap_oop(Register d, Register s1, Register s2); 2160 void store_heap_oop(Register d, Register s1, int simm13a); 2161 void store_heap_oop(Register d, const Address& a, int offset = 0); 2162 2163 void encode_heap_oop(Register src, Register dst); 2164 void encode_heap_oop(Register r) { 2165 encode_heap_oop(r, r); 2166 } 2167 void decode_heap_oop(Register src, Register dst); 2168 void decode_heap_oop(Register r) { 2169 decode_heap_oop(r, r); 2170 } 2171 void encode_heap_oop_not_null(Register r); 2172 void decode_heap_oop_not_null(Register r); 2173 void encode_heap_oop_not_null(Register src, Register dst); 2174 void decode_heap_oop_not_null(Register src, Register dst); 2175 2176 // Support for managing the JavaThread pointer (i.e.; the reference to 2177 // thread-local information). 2178 void get_thread(); // load G2_thread 2179 void verify_thread(); // verify G2_thread contents 2180 void save_thread (const Register threache); // save to cache 2181 void restore_thread(const Register thread_cache); // restore from cache 2182 2183 // Support for last Java frame (but use call_VM instead where possible) 2184 void set_last_Java_frame(Register last_java_sp, Register last_Java_pc); 2185 void reset_last_Java_frame(void); 2186 2187 // Call into the VM. 2188 // Passes the thread pointer (in O0) as a prepended argument. 2189 // Makes sure oop return values are visible to the GC. 2190 void call_VM(Register oop_result, address entry_point, int number_of_arguments = 0, bool check_exceptions = true); 2191 void call_VM(Register oop_result, address entry_point, Register arg_1, bool check_exceptions = true); 2192 void call_VM(Register oop_result, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true); 2193 void call_VM(Register oop_result, address entry_point, Register arg_1, Register arg_2, Register arg_3, bool check_exceptions = true); 2194 2195 // these overloadings are not presently used on SPARC: 2196 void call_VM(Register oop_result, Register last_java_sp, address entry_point, int number_of_arguments = 0, bool check_exceptions = true); 2197 void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, bool check_exceptions = true); 2198 void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, bool check_exceptions = true); 2199 void call_VM(Register oop_result, Register last_java_sp, address entry_point, Register arg_1, Register arg_2, Register arg_3, bool check_exceptions = true); 2200 2201 void call_VM_leaf(Register thread_cache, address entry_point, int number_of_arguments = 0); 2202 void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1); 2203 void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1, Register arg_2); 2204 void call_VM_leaf(Register thread_cache, address entry_point, Register arg_1, Register arg_2, Register arg_3); 2205 2206 void get_vm_result (Register oop_result); 2207 void get_vm_result_2(Register oop_result); 2208 2209 // vm result is currently getting hijacked to for oop preservation 2210 void set_vm_result(Register oop_result); 2211 2212 // if call_VM_base was called with check_exceptions=false, then call 2213 // check_and_forward_exception to handle exceptions when it is safe 2214 void check_and_forward_exception(Register scratch_reg); 2215 2216 private: 2217 // For V8 2218 void read_ccr_trap(Register ccr_save); 2219 void write_ccr_trap(Register ccr_save1, Register scratch1, Register scratch2); 2220 2221 #ifdef ASSERT 2222 // For V8 debugging. Uses V8 instruction sequence and checks 2223 // result with V9 insturctions rdccr and wrccr. 2224 // Uses Gscatch and Gscatch2 2225 void read_ccr_v8_assert(Register ccr_save); 2226 void write_ccr_v8_assert(Register ccr_save); 2227 #endif // ASSERT 2228 2229 public: 2230 2231 // Write to card table for - register is destroyed afterwards. 2232 void card_table_write(jbyte* byte_map_base, Register tmp, Register obj); 2233 2234 void card_write_barrier_post(Register store_addr, Register new_val, Register tmp); 2235 2236 #ifndef SERIALGC 2237 // General G1 pre-barrier generator. 2238 void g1_write_barrier_pre(Register obj, Register index, int offset, Register pre_val, Register tmp, bool preserve_o_regs); 2239 2240 // General G1 post-barrier generator 2241 void g1_write_barrier_post(Register store_addr, Register new_val, Register tmp); 2242 #endif // SERIALGC 2243 2244 // pushes double TOS element of FPU stack on CPU stack; pops from FPU stack 2245 void push_fTOS(); 2246 2247 // pops double TOS element from CPU stack and pushes on FPU stack 2248 void pop_fTOS(); 2249 2250 void empty_FPU_stack(); 2251 2252 void push_IU_state(); 2253 void pop_IU_state(); 2254 2255 void push_FPU_state(); 2256 void pop_FPU_state(); 2257 2258 void push_CPU_state(); 2259 void pop_CPU_state(); 2260 2261 // if heap base register is used - reinit it with the correct value 2262 void reinit_heapbase(); 2263 2264 // Debugging 2265 void _verify_oop(Register reg, const char * msg, const char * file, int line); 2266 void _verify_oop_addr(Address addr, const char * msg, const char * file, int line); 2267 2268 #define verify_oop(reg) _verify_oop(reg, "broken oop " #reg, __FILE__, __LINE__) 2269 #define verify_oop_addr(addr) _verify_oop_addr(addr, "broken oop addr ", __FILE__, __LINE__) 2270 2271 // only if +VerifyOops 2272 void verify_FPU(int stack_depth, const char* s = "illegal FPU state"); 2273 // only if +VerifyFPU 2274 void stop(const char* msg); // prints msg, dumps registers and stops execution 2275 void warn(const char* msg); // prints msg, but don't stop 2276 void untested(const char* what = ""); 2277 void unimplemented(const char* what = "") { char* b = new char[1024]; jio_snprintf(b, 1024, "unimplemented: %s", what); stop(b); } 2278 void should_not_reach_here() { stop("should not reach here"); } 2279 void print_CPU_state(); 2280 2281 // oops in code 2282 AddressLiteral allocate_oop_address(jobject obj); // allocate_index 2283 AddressLiteral constant_oop_address(jobject obj); // find_index 2284 inline void set_oop (jobject obj, Register d); // uses allocate_oop_address 2285 inline void set_oop_constant (jobject obj, Register d); // uses constant_oop_address 2286 inline void set_oop (const AddressLiteral& obj_addr, Register d); // same as load_address 2287 2288 void set_narrow_oop( jobject obj, Register d ); 2289 2290 // nop padding 2291 void align(int modulus); 2292 2293 // declare a safepoint 2294 void safepoint(); 2295 2296 // factor out part of stop into subroutine to save space 2297 void stop_subroutine(); 2298 // factor out part of verify_oop into subroutine to save space 2299 void verify_oop_subroutine(); 2300 2301 // side-door communication with signalHandler in os_solaris.cpp 2302 static address _verify_oop_implicit_branch[3]; 2303 2304 #ifndef PRODUCT 2305 static void test(); 2306 #endif 2307 2308 // convert an incoming arglist to varargs format; put the pointer in d 2309 void set_varargs( Argument a, Register d ); 2310 2311 int total_frame_size_in_bytes(int extraWords); 2312 2313 // used when extraWords known statically 2314 void save_frame(int extraWords = 0); 2315 void save_frame_c1(int size_in_bytes); 2316 // make a frame, and simultaneously pass up one or two register value 2317 // into the new register window 2318 void save_frame_and_mov(int extraWords, Register s1, Register d1, Register s2 = Register(), Register d2 = Register()); 2319 2320 // give no. (outgoing) params, calc # of words will need on frame 2321 void calc_mem_param_words(Register Rparam_words, Register Rresult); 2322 2323 // used to calculate frame size dynamically 2324 // result is in bytes and must be negated for save inst 2325 void calc_frame_size(Register extraWords, Register resultReg); 2326 2327 // calc and also save 2328 void calc_frame_size_and_save(Register extraWords, Register resultReg); 2329 2330 static void debug(char* msg, RegistersForDebugging* outWindow); 2331 2332 // implementations of bytecodes used by both interpreter and compiler 2333 2334 void lcmp( Register Ra_hi, Register Ra_low, 2335 Register Rb_hi, Register Rb_low, 2336 Register Rresult); 2337 2338 void lneg( Register Rhi, Register Rlow ); 2339 2340 void lshl( Register Rin_high, Register Rin_low, Register Rcount, 2341 Register Rout_high, Register Rout_low, Register Rtemp ); 2342 2343 void lshr( Register Rin_high, Register Rin_low, Register Rcount, 2344 Register Rout_high, Register Rout_low, Register Rtemp ); 2345 2346 void lushr( Register Rin_high, Register Rin_low, Register Rcount, 2347 Register Rout_high, Register Rout_low, Register Rtemp ); 2348 2349 #ifdef _LP64 2350 void lcmp( Register Ra, Register Rb, Register Rresult); 2351 #endif 2352 2353 // Load and store values by size and signed-ness 2354 void load_sized_value( Address src, Register dst, size_t size_in_bytes, bool is_signed); 2355 void store_sized_value(Register src, Address dst, size_t size_in_bytes); 2356 2357 void float_cmp( bool is_float, int unordered_result, 2358 FloatRegister Fa, FloatRegister Fb, 2359 Register Rresult); 2360 2361 void fneg( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d); 2362 void fneg( FloatRegisterImpl::Width w, FloatRegister sd ) { Assembler::fneg(w, sd); } 2363 void fmov( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d); 2364 void fabs( FloatRegisterImpl::Width w, FloatRegister s, FloatRegister d); 2365 2366 void save_all_globals_into_locals(); 2367 void restore_globals_from_locals(); 2368 2369 void casx_under_lock(Register top_ptr_reg, Register top_reg, Register ptr_reg, 2370 address lock_addr=0, bool use_call_vm=false); 2371 void cas_under_lock(Register top_ptr_reg, Register top_reg, Register ptr_reg, 2372 address lock_addr=0, bool use_call_vm=false); 2373 void casn (Register addr_reg, Register cmp_reg, Register set_reg) ; 2374 2375 // These set the icc condition code to equal if the lock succeeded 2376 // and notEqual if it failed and requires a slow case 2377 void compiler_lock_object(Register Roop, Register Rmark, Register Rbox, 2378 Register Rscratch, 2379 BiasedLockingCounters* counters = NULL, 2380 bool try_bias = UseBiasedLocking); 2381 void compiler_unlock_object(Register Roop, Register Rmark, Register Rbox, 2382 Register Rscratch, 2383 bool try_bias = UseBiasedLocking); 2384 2385 // Biased locking support 2386 // Upon entry, lock_reg must point to the lock record on the stack, 2387 // obj_reg must contain the target object, and mark_reg must contain 2388 // the target object's header. 2389 // Destroys mark_reg if an attempt is made to bias an anonymously 2390 // biased lock. In this case a failure will go either to the slow 2391 // case or fall through with the notEqual condition code set with 2392 // the expectation that the slow case in the runtime will be called. 2393 // In the fall-through case where the CAS-based lock is done, 2394 // mark_reg is not destroyed. 2395 void biased_locking_enter(Register obj_reg, Register mark_reg, Register temp_reg, 2396 Label& done, Label* slow_case = NULL, 2397 BiasedLockingCounters* counters = NULL); 2398 // Upon entry, the base register of mark_addr must contain the oop. 2399 // Destroys temp_reg. 2400 2401 // If allow_delay_slot_filling is set to true, the next instruction 2402 // emitted after this one will go in an annulled delay slot if the 2403 // biased locking exit case failed. 2404 void biased_locking_exit(Address mark_addr, Register temp_reg, Label& done, bool allow_delay_slot_filling = false); 2405 2406 // allocation 2407 void eden_allocate( 2408 Register obj, // result: pointer to object after successful allocation 2409 Register var_size_in_bytes, // object size in bytes if unknown at compile time; invalid otherwise 2410 int con_size_in_bytes, // object size in bytes if known at compile time 2411 Register t1, // temp register 2412 Register t2, // temp register 2413 Label& slow_case // continuation point if fast allocation fails 2414 ); 2415 void tlab_allocate( 2416 Register obj, // result: pointer to object after successful allocation 2417 Register var_size_in_bytes, // object size in bytes if unknown at compile time; invalid otherwise 2418 int con_size_in_bytes, // object size in bytes if known at compile time 2419 Register t1, // temp register 2420 Label& slow_case // continuation point if fast allocation fails 2421 ); 2422 void tlab_refill(Label& retry_tlab, Label& try_eden, Label& slow_case); 2423 void incr_allocated_bytes(RegisterOrConstant size_in_bytes, 2424 Register t1, Register t2); 2425 2426 // interface method calling 2427 void lookup_interface_method(Register recv_klass, 2428 Register intf_klass, 2429 RegisterOrConstant itable_index, 2430 Register method_result, 2431 Register temp_reg, Register temp2_reg, 2432 Label& no_such_interface); 2433 2434 // Test sub_klass against super_klass, with fast and slow paths. 2435 2436 // The fast path produces a tri-state answer: yes / no / maybe-slow. 2437 // One of the three labels can be NULL, meaning take the fall-through. 2438 // If super_check_offset is -1, the value is loaded up from super_klass. 2439 // No registers are killed, except temp_reg and temp2_reg. 2440 // If super_check_offset is not -1, temp2_reg is not used and can be noreg. 2441 void check_klass_subtype_fast_path(Register sub_klass, 2442 Register super_klass, 2443 Register temp_reg, 2444 Register temp2_reg, 2445 Label* L_success, 2446 Label* L_failure, 2447 Label* L_slow_path, 2448 RegisterOrConstant super_check_offset = RegisterOrConstant(-1), 2449 Register instanceof_hack = noreg); 2450 2451 // The rest of the type check; must be wired to a corresponding fast path. 2452 // It does not repeat the fast path logic, so don't use it standalone. 2453 // The temp_reg can be noreg, if no temps are available. 2454 // It can also be sub_klass or super_klass, meaning it's OK to kill that one. 2455 // Updates the sub's secondary super cache as necessary. 2456 void check_klass_subtype_slow_path(Register sub_klass, 2457 Register super_klass, 2458 Register temp_reg, 2459 Register temp2_reg, 2460 Register temp3_reg, 2461 Register temp4_reg, 2462 Label* L_success, 2463 Label* L_failure); 2464 2465 // Simplified, combined version, good for typical uses. 2466 // Falls through on failure. 2467 void check_klass_subtype(Register sub_klass, 2468 Register super_klass, 2469 Register temp_reg, 2470 Register temp2_reg, 2471 Label& L_success); 2472 2473 // method handles (JSR 292) 2474 void check_method_handle_type(Register mtype_reg, Register mh_reg, 2475 Register temp_reg, 2476 Label& wrong_method_type); 2477 void load_method_handle_vmslots(Register vmslots_reg, Register mh_reg, 2478 Register temp_reg); 2479 void jump_to_method_handle_entry(Register mh_reg, Register temp_reg, bool emit_delayed_nop = true); 2480 // offset relative to Gargs of argument at tos[arg_slot]. 2481 // (arg_slot == 0 means the last argument, not the first). 2482 RegisterOrConstant argument_offset(RegisterOrConstant arg_slot, 2483 Register temp_reg, 2484 int extra_slot_offset = 0); 2485 // Address of Gargs and argument_offset. 2486 Address argument_address(RegisterOrConstant arg_slot, 2487 Register temp_reg, 2488 int extra_slot_offset = 0); 2489 2490 // Stack overflow checking 2491 2492 // Note: this clobbers G3_scratch 2493 void bang_stack_with_offset(int offset) { 2494 // stack grows down, caller passes positive offset 2495 assert(offset > 0, "must bang with negative offset"); 2496 set((-offset)+STACK_BIAS, G3_scratch); 2497 st(G0, SP, G3_scratch); 2498 } 2499 2500 // Writes to stack successive pages until offset reached to check for 2501 // stack overflow + shadow pages. Clobbers tsp and scratch registers. 2502 void bang_stack_size(Register Rsize, Register Rtsp, Register Rscratch); 2503 2504 virtual RegisterOrConstant delayed_value_impl(intptr_t* delayed_value_addr, Register tmp, int offset); 2505 2506 void verify_tlab(); 2507 2508 Condition negate_condition(Condition cond); 2509 2510 // Helper functions for statistics gathering. 2511 // Conditionally (non-atomically) increments passed counter address, preserving condition codes. 2512 void cond_inc(Condition cond, address counter_addr, Register Rtemp1, Register Rtemp2); 2513 // Unconditional increment. 2514 void inc_counter(address counter_addr, Register Rtmp1, Register Rtmp2); 2515 void inc_counter(int* counter_addr, Register Rtmp1, Register Rtmp2); 2516 2517 // Compare char[] arrays aligned to 4 bytes. 2518 void char_arrays_equals(Register ary1, Register ary2, 2519 Register limit, Register result, 2520 Register chr1, Register chr2, Label& Ldone); 2521 2522 #undef VIRTUAL 2523 2524 }; 2525 2526 /** 2527 * class SkipIfEqual: 2528 * 2529 * Instantiating this class will result in assembly code being output that will 2530 * jump around any code emitted between the creation of the instance and it's 2531 * automatic destruction at the end of a scope block, depending on the value of 2532 * the flag passed to the constructor, which will be checked at run-time. 2533 */ 2534 class SkipIfEqual : public StackObj { 2535 private: 2536 MacroAssembler* _masm; 2537 Label _label; 2538 2539 public: 2540 // 'temp' is a temp register that this object can use (and trash) 2541 SkipIfEqual(MacroAssembler*, Register temp, 2542 const bool* flag_addr, Assembler::Condition condition); 2543 ~SkipIfEqual(); 2544 }; 2545 2546 #ifdef ASSERT 2547 // On RISC, there's no benefit to verifying instruction boundaries. 2548 inline bool AbstractAssembler::pd_check_instruction_mark() { return false; } 2549 #endif 2550 2551 #endif // CPU_SPARC_VM_ASSEMBLER_SPARC_HPP