1 /* 2 * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #include "precompiled.hpp" 26 #include "asm/assembler.hpp" 27 #include "asm/assembler.inline.hpp" 28 #include "gc_interface/collectedHeap.inline.hpp" 29 #include "interpreter/interpreter.hpp" 30 #include "memory/cardTableModRefBS.hpp" 31 #include "memory/resourceArea.hpp" 32 #include "prims/methodHandles.hpp" 33 #include "runtime/biasedLocking.hpp" 34 #include "runtime/interfaceSupport.hpp" 35 #include "runtime/objectMonitor.hpp" 36 #include "runtime/os.hpp" 37 #include "runtime/sharedRuntime.hpp" 38 #include "runtime/stubRoutines.hpp" 39 #include "utilities/macros.hpp" 40 #if INCLUDE_ALL_GCS 41 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp" 42 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp" 43 #include "gc_implementation/g1/heapRegion.hpp" 44 #endif // INCLUDE_ALL_GCS 45 46 #ifdef PRODUCT 47 #define BLOCK_COMMENT(str) /* nothing */ 48 #define STOP(error) stop(error) 49 #else 50 #define BLOCK_COMMENT(str) block_comment(str) 51 #define STOP(error) block_comment(error); stop(error) 52 #endif 53 54 #define BIND(label) bind(label); BLOCK_COMMENT(#label ":") 55 // Implementation of AddressLiteral 56 57 AddressLiteral::AddressLiteral(address target, relocInfo::relocType rtype) { 58 _is_lval = false; 59 _target = target; 60 switch (rtype) { 61 case relocInfo::oop_type: 62 case relocInfo::metadata_type: 63 // Oops are a special case. Normally they would be their own section 64 // but in cases like icBuffer they are literals in the code stream that 65 // we don't have a section for. We use none so that we get a literal address 66 // which is always patchable. 67 break; 68 case relocInfo::external_word_type: 69 _rspec = external_word_Relocation::spec(target); 70 break; 71 case relocInfo::internal_word_type: 72 _rspec = internal_word_Relocation::spec(target); 73 break; 74 case relocInfo::opt_virtual_call_type: 75 _rspec = opt_virtual_call_Relocation::spec(); 76 break; 77 case relocInfo::static_call_type: 78 _rspec = static_call_Relocation::spec(); 79 break; 80 case relocInfo::runtime_call_type: 81 _rspec = runtime_call_Relocation::spec(); 82 break; 83 case relocInfo::poll_type: 84 case relocInfo::poll_return_type: 85 _rspec = Relocation::spec_simple(rtype); 86 break; 87 case relocInfo::none: 88 break; 89 default: 90 ShouldNotReachHere(); 91 break; 92 } 93 } 94 95 // Implementation of Address 96 97 #ifdef _LP64 98 99 Address Address::make_array(ArrayAddress adr) { 100 // Not implementable on 64bit machines 101 // Should have been handled higher up the call chain. 102 ShouldNotReachHere(); 103 return Address(); 104 } 105 106 // exceedingly dangerous constructor 107 Address::Address(int disp, address loc, relocInfo::relocType rtype) { 108 _base = noreg; 109 _index = noreg; 110 _scale = no_scale; 111 _disp = disp; 112 switch (rtype) { 113 case relocInfo::external_word_type: 114 _rspec = external_word_Relocation::spec(loc); 115 break; 116 case relocInfo::internal_word_type: 117 _rspec = internal_word_Relocation::spec(loc); 118 break; 119 case relocInfo::runtime_call_type: 120 // HMM 121 _rspec = runtime_call_Relocation::spec(); 122 break; 123 case relocInfo::poll_type: 124 case relocInfo::poll_return_type: 125 _rspec = Relocation::spec_simple(rtype); 126 break; 127 case relocInfo::none: 128 break; 129 default: 130 ShouldNotReachHere(); 131 } 132 } 133 #else // LP64 134 135 Address Address::make_array(ArrayAddress adr) { 136 AddressLiteral base = adr.base(); 137 Address index = adr.index(); 138 assert(index._disp == 0, "must not have disp"); // maybe it can? 139 Address array(index._base, index._index, index._scale, (intptr_t) base.target()); 140 array._rspec = base._rspec; 141 return array; 142 } 143 144 // exceedingly dangerous constructor 145 Address::Address(address loc, RelocationHolder spec) { 146 _base = noreg; 147 _index = noreg; 148 _scale = no_scale; 149 _disp = (intptr_t) loc; 150 _rspec = spec; 151 } 152 153 #endif // _LP64 154 155 156 157 // Convert the raw encoding form into the form expected by the constructor for 158 // Address. An index of 4 (rsp) corresponds to having no index, so convert 159 // that to noreg for the Address constructor. 160 Address Address::make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc) { 161 RelocationHolder rspec; 162 if (disp_reloc != relocInfo::none) { 163 rspec = Relocation::spec_simple(disp_reloc); 164 } 165 bool valid_index = index != rsp->encoding(); 166 if (valid_index) { 167 Address madr(as_Register(base), as_Register(index), (Address::ScaleFactor)scale, in_ByteSize(disp)); 168 madr._rspec = rspec; 169 return madr; 170 } else { 171 Address madr(as_Register(base), noreg, Address::no_scale, in_ByteSize(disp)); 172 madr._rspec = rspec; 173 return madr; 174 } 175 } 176 177 // Implementation of Assembler 178 179 int AbstractAssembler::code_fill_byte() { 180 return (u_char)'\xF4'; // hlt 181 } 182 183 // make this go away someday 184 void Assembler::emit_data(jint data, relocInfo::relocType rtype, int format) { 185 if (rtype == relocInfo::none) 186 emit_int32(data); 187 else emit_data(data, Relocation::spec_simple(rtype), format); 188 } 189 190 void Assembler::emit_data(jint data, RelocationHolder const& rspec, int format) { 191 assert(imm_operand == 0, "default format must be immediate in this file"); 192 assert(inst_mark() != NULL, "must be inside InstructionMark"); 193 if (rspec.type() != relocInfo::none) { 194 #ifdef ASSERT 195 check_relocation(rspec, format); 196 #endif 197 // Do not use AbstractAssembler::relocate, which is not intended for 198 // embedded words. Instead, relocate to the enclosing instruction. 199 200 // hack. call32 is too wide for mask so use disp32 201 if (format == call32_operand) 202 code_section()->relocate(inst_mark(), rspec, disp32_operand); 203 else 204 code_section()->relocate(inst_mark(), rspec, format); 205 } 206 emit_int32(data); 207 } 208 209 static int encode(Register r) { 210 int enc = r->encoding(); 211 if (enc >= 8) { 212 enc -= 8; 213 } 214 return enc; 215 } 216 217 void Assembler::emit_arith_b(int op1, int op2, Register dst, int imm8) { 218 assert(dst->has_byte_register(), "must have byte register"); 219 assert(isByte(op1) && isByte(op2), "wrong opcode"); 220 assert(isByte(imm8), "not a byte"); 221 assert((op1 & 0x01) == 0, "should be 8bit operation"); 222 emit_int8(op1); 223 emit_int8(op2 | encode(dst)); 224 emit_int8(imm8); 225 } 226 227 228 void Assembler::emit_arith(int op1, int op2, Register dst, int32_t imm32) { 229 assert(isByte(op1) && isByte(op2), "wrong opcode"); 230 assert((op1 & 0x01) == 1, "should be 32bit operation"); 231 assert((op1 & 0x02) == 0, "sign-extension bit should not be set"); 232 if (is8bit(imm32)) { 233 emit_int8(op1 | 0x02); // set sign bit 234 emit_int8(op2 | encode(dst)); 235 emit_int8(imm32 & 0xFF); 236 } else { 237 emit_int8(op1); 238 emit_int8(op2 | encode(dst)); 239 emit_int32(imm32); 240 } 241 } 242 243 // Force generation of a 4 byte immediate value even if it fits into 8bit 244 void Assembler::emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32) { 245 assert(isByte(op1) && isByte(op2), "wrong opcode"); 246 assert((op1 & 0x01) == 1, "should be 32bit operation"); 247 assert((op1 & 0x02) == 0, "sign-extension bit should not be set"); 248 emit_int8(op1); 249 emit_int8(op2 | encode(dst)); 250 emit_int32(imm32); 251 } 252 253 // immediate-to-memory forms 254 void Assembler::emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32) { 255 assert((op1 & 0x01) == 1, "should be 32bit operation"); 256 assert((op1 & 0x02) == 0, "sign-extension bit should not be set"); 257 if (is8bit(imm32)) { 258 emit_int8(op1 | 0x02); // set sign bit 259 emit_operand(rm, adr, 1); 260 emit_int8(imm32 & 0xFF); 261 } else { 262 emit_int8(op1); 263 emit_operand(rm, adr, 4); 264 emit_int32(imm32); 265 } 266 } 267 268 269 void Assembler::emit_arith(int op1, int op2, Register dst, Register src) { 270 assert(isByte(op1) && isByte(op2), "wrong opcode"); 271 emit_int8(op1); 272 emit_int8(op2 | encode(dst) << 3 | encode(src)); 273 } 274 275 276 void Assembler::emit_operand(Register reg, Register base, Register index, 277 Address::ScaleFactor scale, int disp, 278 RelocationHolder const& rspec, 279 int rip_relative_correction) { 280 relocInfo::relocType rtype = (relocInfo::relocType) rspec.type(); 281 282 // Encode the registers as needed in the fields they are used in 283 284 int regenc = encode(reg) << 3; 285 int indexenc = index->is_valid() ? encode(index) << 3 : 0; 286 int baseenc = base->is_valid() ? encode(base) : 0; 287 288 if (base->is_valid()) { 289 if (index->is_valid()) { 290 assert(scale != Address::no_scale, "inconsistent address"); 291 // [base + index*scale + disp] 292 if (disp == 0 && rtype == relocInfo::none && 293 base != rbp LP64_ONLY(&& base != r13)) { 294 // [base + index*scale] 295 // [00 reg 100][ss index base] 296 assert(index != rsp, "illegal addressing mode"); 297 emit_int8(0x04 | regenc); 298 emit_int8(scale << 6 | indexenc | baseenc); 299 } else if (is8bit(disp) && rtype == relocInfo::none) { 300 // [base + index*scale + imm8] 301 // [01 reg 100][ss index base] imm8 302 assert(index != rsp, "illegal addressing mode"); 303 emit_int8(0x44 | regenc); 304 emit_int8(scale << 6 | indexenc | baseenc); 305 emit_int8(disp & 0xFF); 306 } else { 307 // [base + index*scale + disp32] 308 // [10 reg 100][ss index base] disp32 309 assert(index != rsp, "illegal addressing mode"); 310 emit_int8(0x84 | regenc); 311 emit_int8(scale << 6 | indexenc | baseenc); 312 emit_data(disp, rspec, disp32_operand); 313 } 314 } else if (base == rsp LP64_ONLY(|| base == r12)) { 315 // [rsp + disp] 316 if (disp == 0 && rtype == relocInfo::none) { 317 // [rsp] 318 // [00 reg 100][00 100 100] 319 emit_int8(0x04 | regenc); 320 emit_int8(0x24); 321 } else if (is8bit(disp) && rtype == relocInfo::none) { 322 // [rsp + imm8] 323 // [01 reg 100][00 100 100] disp8 324 emit_int8(0x44 | regenc); 325 emit_int8(0x24); 326 emit_int8(disp & 0xFF); 327 } else { 328 // [rsp + imm32] 329 // [10 reg 100][00 100 100] disp32 330 emit_int8(0x84 | regenc); 331 emit_int8(0x24); 332 emit_data(disp, rspec, disp32_operand); 333 } 334 } else { 335 // [base + disp] 336 assert(base != rsp LP64_ONLY(&& base != r12), "illegal addressing mode"); 337 if (disp == 0 && rtype == relocInfo::none && 338 base != rbp LP64_ONLY(&& base != r13)) { 339 // [base] 340 // [00 reg base] 341 emit_int8(0x00 | regenc | baseenc); 342 } else if (is8bit(disp) && rtype == relocInfo::none) { 343 // [base + disp8] 344 // [01 reg base] disp8 345 emit_int8(0x40 | regenc | baseenc); 346 emit_int8(disp & 0xFF); 347 } else { 348 // [base + disp32] 349 // [10 reg base] disp32 350 emit_int8(0x80 | regenc | baseenc); 351 emit_data(disp, rspec, disp32_operand); 352 } 353 } 354 } else { 355 if (index->is_valid()) { 356 assert(scale != Address::no_scale, "inconsistent address"); 357 // [index*scale + disp] 358 // [00 reg 100][ss index 101] disp32 359 assert(index != rsp, "illegal addressing mode"); 360 emit_int8(0x04 | regenc); 361 emit_int8(scale << 6 | indexenc | 0x05); 362 emit_data(disp, rspec, disp32_operand); 363 } else if (rtype != relocInfo::none ) { 364 // [disp] (64bit) RIP-RELATIVE (32bit) abs 365 // [00 000 101] disp32 366 367 emit_int8(0x05 | regenc); 368 // Note that the RIP-rel. correction applies to the generated 369 // disp field, but _not_ to the target address in the rspec. 370 371 // disp was created by converting the target address minus the pc 372 // at the start of the instruction. That needs more correction here. 373 // intptr_t disp = target - next_ip; 374 assert(inst_mark() != NULL, "must be inside InstructionMark"); 375 address next_ip = pc() + sizeof(int32_t) + rip_relative_correction; 376 int64_t adjusted = disp; 377 // Do rip-rel adjustment for 64bit 378 LP64_ONLY(adjusted -= (next_ip - inst_mark())); 379 assert(is_simm32(adjusted), 380 "must be 32bit offset (RIP relative address)"); 381 emit_data((int32_t) adjusted, rspec, disp32_operand); 382 383 } else { 384 // 32bit never did this, did everything as the rip-rel/disp code above 385 // [disp] ABSOLUTE 386 // [00 reg 100][00 100 101] disp32 387 emit_int8(0x04 | regenc); 388 emit_int8(0x25); 389 emit_data(disp, rspec, disp32_operand); 390 } 391 } 392 } 393 394 void Assembler::emit_operand(XMMRegister reg, Register base, Register index, 395 Address::ScaleFactor scale, int disp, 396 RelocationHolder const& rspec) { 397 emit_operand((Register)reg, base, index, scale, disp, rspec); 398 } 399 400 // Secret local extension to Assembler::WhichOperand: 401 #define end_pc_operand (_WhichOperand_limit) 402 403 address Assembler::locate_operand(address inst, WhichOperand which) { 404 // Decode the given instruction, and return the address of 405 // an embedded 32-bit operand word. 406 407 // If "which" is disp32_operand, selects the displacement portion 408 // of an effective address specifier. 409 // If "which" is imm64_operand, selects the trailing immediate constant. 410 // If "which" is call32_operand, selects the displacement of a call or jump. 411 // Caller is responsible for ensuring that there is such an operand, 412 // and that it is 32/64 bits wide. 413 414 // If "which" is end_pc_operand, find the end of the instruction. 415 416 address ip = inst; 417 bool is_64bit = false; 418 419 debug_only(bool has_disp32 = false); 420 int tail_size = 0; // other random bytes (#32, #16, etc.) at end of insn 421 422 again_after_prefix: 423 switch (0xFF & *ip++) { 424 425 // These convenience macros generate groups of "case" labels for the switch. 426 #define REP4(x) (x)+0: case (x)+1: case (x)+2: case (x)+3 427 #define REP8(x) (x)+0: case (x)+1: case (x)+2: case (x)+3: \ 428 case (x)+4: case (x)+5: case (x)+6: case (x)+7 429 #define REP16(x) REP8((x)+0): \ 430 case REP8((x)+8) 431 432 case CS_segment: 433 case SS_segment: 434 case DS_segment: 435 case ES_segment: 436 case FS_segment: 437 case GS_segment: 438 // Seems dubious 439 LP64_ONLY(assert(false, "shouldn't have that prefix")); 440 assert(ip == inst+1, "only one prefix allowed"); 441 goto again_after_prefix; 442 443 case 0x67: 444 case REX: 445 case REX_B: 446 case REX_X: 447 case REX_XB: 448 case REX_R: 449 case REX_RB: 450 case REX_RX: 451 case REX_RXB: 452 NOT_LP64(assert(false, "64bit prefixes")); 453 goto again_after_prefix; 454 455 case REX_W: 456 case REX_WB: 457 case REX_WX: 458 case REX_WXB: 459 case REX_WR: 460 case REX_WRB: 461 case REX_WRX: 462 case REX_WRXB: 463 NOT_LP64(assert(false, "64bit prefixes")); 464 is_64bit = true; 465 goto again_after_prefix; 466 467 case 0xFF: // pushq a; decl a; incl a; call a; jmp a 468 case 0x88: // movb a, r 469 case 0x89: // movl a, r 470 case 0x8A: // movb r, a 471 case 0x8B: // movl r, a 472 case 0x8F: // popl a 473 debug_only(has_disp32 = true); 474 break; 475 476 case 0x68: // pushq #32 477 if (which == end_pc_operand) { 478 return ip + 4; 479 } 480 assert(which == imm_operand && !is_64bit, "pushl has no disp32 or 64bit immediate"); 481 return ip; // not produced by emit_operand 482 483 case 0x66: // movw ... (size prefix) 484 again_after_size_prefix2: 485 switch (0xFF & *ip++) { 486 case REX: 487 case REX_B: 488 case REX_X: 489 case REX_XB: 490 case REX_R: 491 case REX_RB: 492 case REX_RX: 493 case REX_RXB: 494 case REX_W: 495 case REX_WB: 496 case REX_WX: 497 case REX_WXB: 498 case REX_WR: 499 case REX_WRB: 500 case REX_WRX: 501 case REX_WRXB: 502 NOT_LP64(assert(false, "64bit prefix found")); 503 goto again_after_size_prefix2; 504 case 0x8B: // movw r, a 505 case 0x89: // movw a, r 506 debug_only(has_disp32 = true); 507 break; 508 case 0xC7: // movw a, #16 509 debug_only(has_disp32 = true); 510 tail_size = 2; // the imm16 511 break; 512 case 0x0F: // several SSE/SSE2 variants 513 ip--; // reparse the 0x0F 514 goto again_after_prefix; 515 default: 516 ShouldNotReachHere(); 517 } 518 break; 519 520 case REP8(0xB8): // movl/q r, #32/#64(oop?) 521 if (which == end_pc_operand) return ip + (is_64bit ? 8 : 4); 522 // these asserts are somewhat nonsensical 523 #ifndef _LP64 524 assert(which == imm_operand || which == disp32_operand, 525 err_msg("which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, ip)); 526 #else 527 assert((which == call32_operand || which == imm_operand) && is_64bit || 528 which == narrow_oop_operand && !is_64bit, 529 err_msg("which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, ip)); 530 #endif // _LP64 531 return ip; 532 533 case 0x69: // imul r, a, #32 534 case 0xC7: // movl a, #32(oop?) 535 tail_size = 4; 536 debug_only(has_disp32 = true); // has both kinds of operands! 537 break; 538 539 case 0x0F: // movx..., etc. 540 switch (0xFF & *ip++) { 541 case 0x3A: // pcmpestri 542 tail_size = 1; 543 case 0x38: // ptest, pmovzxbw 544 ip++; // skip opcode 545 debug_only(has_disp32 = true); // has both kinds of operands! 546 break; 547 548 case 0x70: // pshufd r, r/a, #8 549 debug_only(has_disp32 = true); // has both kinds of operands! 550 case 0x73: // psrldq r, #8 551 tail_size = 1; 552 break; 553 554 case 0x12: // movlps 555 case 0x28: // movaps 556 case 0x2E: // ucomiss 557 case 0x2F: // comiss 558 case 0x54: // andps 559 case 0x55: // andnps 560 case 0x56: // orps 561 case 0x57: // xorps 562 case 0x6E: // movd 563 case 0x7E: // movd 564 case 0xAE: // ldmxcsr, stmxcsr, fxrstor, fxsave, clflush 565 debug_only(has_disp32 = true); 566 break; 567 568 case 0xAD: // shrd r, a, %cl 569 case 0xAF: // imul r, a 570 case 0xBE: // movsbl r, a (movsxb) 571 case 0xBF: // movswl r, a (movsxw) 572 case 0xB6: // movzbl r, a (movzxb) 573 case 0xB7: // movzwl r, a (movzxw) 574 case REP16(0x40): // cmovl cc, r, a 575 case 0xB0: // cmpxchgb 576 case 0xB1: // cmpxchg 577 case 0xC1: // xaddl 578 case 0xC7: // cmpxchg8 579 case REP16(0x90): // setcc a 580 debug_only(has_disp32 = true); 581 // fall out of the switch to decode the address 582 break; 583 584 case 0xC4: // pinsrw r, a, #8 585 debug_only(has_disp32 = true); 586 case 0xC5: // pextrw r, r, #8 587 tail_size = 1; // the imm8 588 break; 589 590 case 0xAC: // shrd r, a, #8 591 debug_only(has_disp32 = true); 592 tail_size = 1; // the imm8 593 break; 594 595 case REP16(0x80): // jcc rdisp32 596 if (which == end_pc_operand) return ip + 4; 597 assert(which == call32_operand, "jcc has no disp32 or imm"); 598 return ip; 599 default: 600 ShouldNotReachHere(); 601 } 602 break; 603 604 case 0x81: // addl a, #32; addl r, #32 605 // also: orl, adcl, sbbl, andl, subl, xorl, cmpl 606 // on 32bit in the case of cmpl, the imm might be an oop 607 tail_size = 4; 608 debug_only(has_disp32 = true); // has both kinds of operands! 609 break; 610 611 case 0x83: // addl a, #8; addl r, #8 612 // also: orl, adcl, sbbl, andl, subl, xorl, cmpl 613 debug_only(has_disp32 = true); // has both kinds of operands! 614 tail_size = 1; 615 break; 616 617 case 0x9B: 618 switch (0xFF & *ip++) { 619 case 0xD9: // fnstcw a 620 debug_only(has_disp32 = true); 621 break; 622 default: 623 ShouldNotReachHere(); 624 } 625 break; 626 627 case REP4(0x00): // addb a, r; addl a, r; addb r, a; addl r, a 628 case REP4(0x10): // adc... 629 case REP4(0x20): // and... 630 case REP4(0x30): // xor... 631 case REP4(0x08): // or... 632 case REP4(0x18): // sbb... 633 case REP4(0x28): // sub... 634 case 0xF7: // mull a 635 case 0x8D: // lea r, a 636 case 0x87: // xchg r, a 637 case REP4(0x38): // cmp... 638 case 0x85: // test r, a 639 debug_only(has_disp32 = true); // has both kinds of operands! 640 break; 641 642 case 0xC1: // sal a, #8; sar a, #8; shl a, #8; shr a, #8 643 case 0xC6: // movb a, #8 644 case 0x80: // cmpb a, #8 645 case 0x6B: // imul r, a, #8 646 debug_only(has_disp32 = true); // has both kinds of operands! 647 tail_size = 1; // the imm8 648 break; 649 650 case 0xC4: // VEX_3bytes 651 case 0xC5: // VEX_2bytes 652 assert((UseAVX > 0), "shouldn't have VEX prefix"); 653 assert(ip == inst+1, "no prefixes allowed"); 654 // C4 and C5 are also used as opcodes for PINSRW and PEXTRW instructions 655 // but they have prefix 0x0F and processed when 0x0F processed above. 656 // 657 // In 32-bit mode the VEX first byte C4 and C5 alias onto LDS and LES 658 // instructions (these instructions are not supported in 64-bit mode). 659 // To distinguish them bits [7:6] are set in the VEX second byte since 660 // ModRM byte can not be of the form 11xxxxxx in 32-bit mode. To set 661 // those VEX bits REX and vvvv bits are inverted. 662 // 663 // Fortunately C2 doesn't generate these instructions so we don't need 664 // to check for them in product version. 665 666 // Check second byte 667 NOT_LP64(assert((0xC0 & *ip) == 0xC0, "shouldn't have LDS and LES instructions")); 668 669 // First byte 670 if ((0xFF & *inst) == VEX_3bytes) { 671 ip++; // third byte 672 is_64bit = ((VEX_W & *ip) == VEX_W); 673 } 674 ip++; // opcode 675 // To find the end of instruction (which == end_pc_operand). 676 switch (0xFF & *ip) { 677 case 0x61: // pcmpestri r, r/a, #8 678 case 0x70: // pshufd r, r/a, #8 679 case 0x73: // psrldq r, #8 680 tail_size = 1; // the imm8 681 break; 682 default: 683 break; 684 } 685 ip++; // skip opcode 686 debug_only(has_disp32 = true); // has both kinds of operands! 687 break; 688 689 case 0xD1: // sal a, 1; sar a, 1; shl a, 1; shr a, 1 690 case 0xD3: // sal a, %cl; sar a, %cl; shl a, %cl; shr a, %cl 691 case 0xD9: // fld_s a; fst_s a; fstp_s a; fldcw a 692 case 0xDD: // fld_d a; fst_d a; fstp_d a 693 case 0xDB: // fild_s a; fistp_s a; fld_x a; fstp_x a 694 case 0xDF: // fild_d a; fistp_d a 695 case 0xD8: // fadd_s a; fsubr_s a; fmul_s a; fdivr_s a; fcomp_s a 696 case 0xDC: // fadd_d a; fsubr_d a; fmul_d a; fdivr_d a; fcomp_d a 697 case 0xDE: // faddp_d a; fsubrp_d a; fmulp_d a; fdivrp_d a; fcompp_d a 698 debug_only(has_disp32 = true); 699 break; 700 701 case 0xE8: // call rdisp32 702 case 0xE9: // jmp rdisp32 703 if (which == end_pc_operand) return ip + 4; 704 assert(which == call32_operand, "call has no disp32 or imm"); 705 return ip; 706 707 case 0xF0: // Lock 708 assert(os::is_MP(), "only on MP"); 709 goto again_after_prefix; 710 711 case 0xF3: // For SSE 712 case 0xF2: // For SSE2 713 switch (0xFF & *ip++) { 714 case REX: 715 case REX_B: 716 case REX_X: 717 case REX_XB: 718 case REX_R: 719 case REX_RB: 720 case REX_RX: 721 case REX_RXB: 722 case REX_W: 723 case REX_WB: 724 case REX_WX: 725 case REX_WXB: 726 case REX_WR: 727 case REX_WRB: 728 case REX_WRX: 729 case REX_WRXB: 730 NOT_LP64(assert(false, "found 64bit prefix")); 731 ip++; 732 default: 733 ip++; 734 } 735 debug_only(has_disp32 = true); // has both kinds of operands! 736 break; 737 738 default: 739 ShouldNotReachHere(); 740 741 #undef REP8 742 #undef REP16 743 } 744 745 assert(which != call32_operand, "instruction is not a call, jmp, or jcc"); 746 #ifdef _LP64 747 assert(which != imm_operand, "instruction is not a movq reg, imm64"); 748 #else 749 // assert(which != imm_operand || has_imm32, "instruction has no imm32 field"); 750 assert(which != imm_operand || has_disp32, "instruction has no imm32 field"); 751 #endif // LP64 752 assert(which != disp32_operand || has_disp32, "instruction has no disp32 field"); 753 754 // parse the output of emit_operand 755 int op2 = 0xFF & *ip++; 756 int base = op2 & 0x07; 757 int op3 = -1; 758 const int b100 = 4; 759 const int b101 = 5; 760 if (base == b100 && (op2 >> 6) != 3) { 761 op3 = 0xFF & *ip++; 762 base = op3 & 0x07; // refetch the base 763 } 764 // now ip points at the disp (if any) 765 766 switch (op2 >> 6) { 767 case 0: 768 // [00 reg 100][ss index base] 769 // [00 reg 100][00 100 esp] 770 // [00 reg base] 771 // [00 reg 100][ss index 101][disp32] 772 // [00 reg 101] [disp32] 773 774 if (base == b101) { 775 if (which == disp32_operand) 776 return ip; // caller wants the disp32 777 ip += 4; // skip the disp32 778 } 779 break; 780 781 case 1: 782 // [01 reg 100][ss index base][disp8] 783 // [01 reg 100][00 100 esp][disp8] 784 // [01 reg base] [disp8] 785 ip += 1; // skip the disp8 786 break; 787 788 case 2: 789 // [10 reg 100][ss index base][disp32] 790 // [10 reg 100][00 100 esp][disp32] 791 // [10 reg base] [disp32] 792 if (which == disp32_operand) 793 return ip; // caller wants the disp32 794 ip += 4; // skip the disp32 795 break; 796 797 case 3: 798 // [11 reg base] (not a memory addressing mode) 799 break; 800 } 801 802 if (which == end_pc_operand) { 803 return ip + tail_size; 804 } 805 806 #ifdef _LP64 807 assert(which == narrow_oop_operand && !is_64bit, "instruction is not a movl adr, imm32"); 808 #else 809 assert(which == imm_operand, "instruction has only an imm field"); 810 #endif // LP64 811 return ip; 812 } 813 814 address Assembler::locate_next_instruction(address inst) { 815 // Secretly share code with locate_operand: 816 return locate_operand(inst, end_pc_operand); 817 } 818 819 820 #ifdef ASSERT 821 void Assembler::check_relocation(RelocationHolder const& rspec, int format) { 822 address inst = inst_mark(); 823 assert(inst != NULL && inst < pc(), "must point to beginning of instruction"); 824 address opnd; 825 826 Relocation* r = rspec.reloc(); 827 if (r->type() == relocInfo::none) { 828 return; 829 } else if (r->is_call() || format == call32_operand) { 830 // assert(format == imm32_operand, "cannot specify a nonzero format"); 831 opnd = locate_operand(inst, call32_operand); 832 } else if (r->is_data()) { 833 assert(format == imm_operand || format == disp32_operand 834 LP64_ONLY(|| format == narrow_oop_operand), "format ok"); 835 opnd = locate_operand(inst, (WhichOperand)format); 836 } else { 837 assert(format == imm_operand, "cannot specify a format"); 838 return; 839 } 840 assert(opnd == pc(), "must put operand where relocs can find it"); 841 } 842 #endif // ASSERT 843 844 void Assembler::emit_operand32(Register reg, Address adr) { 845 assert(reg->encoding() < 8, "no extended registers"); 846 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers"); 847 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp, 848 adr._rspec); 849 } 850 851 void Assembler::emit_operand(Register reg, Address adr, 852 int rip_relative_correction) { 853 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp, 854 adr._rspec, 855 rip_relative_correction); 856 } 857 858 void Assembler::emit_operand(XMMRegister reg, Address adr) { 859 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp, 860 adr._rspec); 861 } 862 863 // MMX operations 864 void Assembler::emit_operand(MMXRegister reg, Address adr) { 865 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers"); 866 emit_operand((Register)reg, adr._base, adr._index, adr._scale, adr._disp, adr._rspec); 867 } 868 869 // work around gcc (3.2.1-7a) bug 870 void Assembler::emit_operand(Address adr, MMXRegister reg) { 871 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers"); 872 emit_operand((Register)reg, adr._base, adr._index, adr._scale, adr._disp, adr._rspec); 873 } 874 875 876 void Assembler::emit_farith(int b1, int b2, int i) { 877 assert(isByte(b1) && isByte(b2), "wrong opcode"); 878 assert(0 <= i && i < 8, "illegal stack offset"); 879 emit_int8(b1); 880 emit_int8(b2 + i); 881 } 882 883 884 // Now the Assembler instructions (identical for 32/64 bits) 885 886 void Assembler::adcl(Address dst, int32_t imm32) { 887 InstructionMark im(this); 888 prefix(dst); 889 emit_arith_operand(0x81, rdx, dst, imm32); 890 } 891 892 void Assembler::adcl(Address dst, Register src) { 893 InstructionMark im(this); 894 prefix(dst, src); 895 emit_int8(0x11); 896 emit_operand(src, dst); 897 } 898 899 void Assembler::adcl(Register dst, int32_t imm32) { 900 prefix(dst); 901 emit_arith(0x81, 0xD0, dst, imm32); 902 } 903 904 void Assembler::adcl(Register dst, Address src) { 905 InstructionMark im(this); 906 prefix(src, dst); 907 emit_int8(0x13); 908 emit_operand(dst, src); 909 } 910 911 void Assembler::adcl(Register dst, Register src) { 912 (void) prefix_and_encode(dst->encoding(), src->encoding()); 913 emit_arith(0x13, 0xC0, dst, src); 914 } 915 916 void Assembler::addl(Address dst, int32_t imm32) { 917 InstructionMark im(this); 918 prefix(dst); 919 emit_arith_operand(0x81, rax, dst, imm32); 920 } 921 922 void Assembler::addl(Address dst, Register src) { 923 InstructionMark im(this); 924 prefix(dst, src); 925 emit_int8(0x01); 926 emit_operand(src, dst); 927 } 928 929 void Assembler::addl(Register dst, int32_t imm32) { 930 prefix(dst); 931 emit_arith(0x81, 0xC0, dst, imm32); 932 } 933 934 void Assembler::addl(Register dst, Address src) { 935 InstructionMark im(this); 936 prefix(src, dst); 937 emit_int8(0x03); 938 emit_operand(dst, src); 939 } 940 941 void Assembler::addl(Register dst, Register src) { 942 (void) prefix_and_encode(dst->encoding(), src->encoding()); 943 emit_arith(0x03, 0xC0, dst, src); 944 } 945 946 void Assembler::addr_nop_4() { 947 assert(UseAddressNop, "no CPU support"); 948 // 4 bytes: NOP DWORD PTR [EAX+0] 949 emit_int8(0x0F); 950 emit_int8(0x1F); 951 emit_int8(0x40); // emit_rm(cbuf, 0x1, EAX_enc, EAX_enc); 952 emit_int8(0); // 8-bits offset (1 byte) 953 } 954 955 void Assembler::addr_nop_5() { 956 assert(UseAddressNop, "no CPU support"); 957 // 5 bytes: NOP DWORD PTR [EAX+EAX*0+0] 8-bits offset 958 emit_int8(0x0F); 959 emit_int8(0x1F); 960 emit_int8(0x44); // emit_rm(cbuf, 0x1, EAX_enc, 0x4); 961 emit_int8(0x00); // emit_rm(cbuf, 0x0, EAX_enc, EAX_enc); 962 emit_int8(0); // 8-bits offset (1 byte) 963 } 964 965 void Assembler::addr_nop_7() { 966 assert(UseAddressNop, "no CPU support"); 967 // 7 bytes: NOP DWORD PTR [EAX+0] 32-bits offset 968 emit_int8(0x0F); 969 emit_int8(0x1F); 970 emit_int8((unsigned char)0x80); 971 // emit_rm(cbuf, 0x2, EAX_enc, EAX_enc); 972 emit_int32(0); // 32-bits offset (4 bytes) 973 } 974 975 void Assembler::addr_nop_8() { 976 assert(UseAddressNop, "no CPU support"); 977 // 8 bytes: NOP DWORD PTR [EAX+EAX*0+0] 32-bits offset 978 emit_int8(0x0F); 979 emit_int8(0x1F); 980 emit_int8((unsigned char)0x84); 981 // emit_rm(cbuf, 0x2, EAX_enc, 0x4); 982 emit_int8(0x00); // emit_rm(cbuf, 0x0, EAX_enc, EAX_enc); 983 emit_int32(0); // 32-bits offset (4 bytes) 984 } 985 986 void Assembler::addsd(XMMRegister dst, XMMRegister src) { 987 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 988 emit_simd_arith(0x58, dst, src, VEX_SIMD_F2); 989 } 990 991 void Assembler::addsd(XMMRegister dst, Address src) { 992 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 993 emit_simd_arith(0x58, dst, src, VEX_SIMD_F2); 994 } 995 996 void Assembler::addss(XMMRegister dst, XMMRegister src) { 997 NOT_LP64(assert(VM_Version::supports_sse(), "")); 998 emit_simd_arith(0x58, dst, src, VEX_SIMD_F3); 999 } 1000 1001 void Assembler::addss(XMMRegister dst, Address src) { 1002 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1003 emit_simd_arith(0x58, dst, src, VEX_SIMD_F3); 1004 } 1005 1006 void Assembler::aesdec(XMMRegister dst, Address src) { 1007 assert(VM_Version::supports_aes(), ""); 1008 InstructionMark im(this); 1009 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1010 emit_int8((unsigned char)0xDE); 1011 emit_operand(dst, src); 1012 } 1013 1014 void Assembler::aesdec(XMMRegister dst, XMMRegister src) { 1015 assert(VM_Version::supports_aes(), ""); 1016 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1017 emit_int8((unsigned char)0xDE); 1018 emit_int8(0xC0 | encode); 1019 } 1020 1021 void Assembler::aesdeclast(XMMRegister dst, Address src) { 1022 assert(VM_Version::supports_aes(), ""); 1023 InstructionMark im(this); 1024 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1025 emit_int8((unsigned char)0xDF); 1026 emit_operand(dst, src); 1027 } 1028 1029 void Assembler::aesdeclast(XMMRegister dst, XMMRegister src) { 1030 assert(VM_Version::supports_aes(), ""); 1031 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1032 emit_int8((unsigned char)0xDF); 1033 emit_int8((unsigned char)(0xC0 | encode)); 1034 } 1035 1036 void Assembler::aesenc(XMMRegister dst, Address src) { 1037 assert(VM_Version::supports_aes(), ""); 1038 InstructionMark im(this); 1039 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1040 emit_int8((unsigned char)0xDC); 1041 emit_operand(dst, src); 1042 } 1043 1044 void Assembler::aesenc(XMMRegister dst, XMMRegister src) { 1045 assert(VM_Version::supports_aes(), ""); 1046 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1047 emit_int8((unsigned char)0xDC); 1048 emit_int8(0xC0 | encode); 1049 } 1050 1051 void Assembler::aesenclast(XMMRegister dst, Address src) { 1052 assert(VM_Version::supports_aes(), ""); 1053 InstructionMark im(this); 1054 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1055 emit_int8((unsigned char)0xDD); 1056 emit_operand(dst, src); 1057 } 1058 1059 void Assembler::aesenclast(XMMRegister dst, XMMRegister src) { 1060 assert(VM_Version::supports_aes(), ""); 1061 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 1062 emit_int8((unsigned char)0xDD); 1063 emit_int8((unsigned char)(0xC0 | encode)); 1064 } 1065 1066 1067 void Assembler::andl(Address dst, int32_t imm32) { 1068 InstructionMark im(this); 1069 prefix(dst); 1070 emit_int8((unsigned char)0x81); 1071 emit_operand(rsp, dst, 4); 1072 emit_int32(imm32); 1073 } 1074 1075 void Assembler::andl(Register dst, int32_t imm32) { 1076 prefix(dst); 1077 emit_arith(0x81, 0xE0, dst, imm32); 1078 } 1079 1080 void Assembler::andl(Register dst, Address src) { 1081 InstructionMark im(this); 1082 prefix(src, dst); 1083 emit_int8(0x23); 1084 emit_operand(dst, src); 1085 } 1086 1087 void Assembler::andl(Register dst, Register src) { 1088 (void) prefix_and_encode(dst->encoding(), src->encoding()); 1089 emit_arith(0x23, 0xC0, dst, src); 1090 } 1091 1092 void Assembler::andnl(Register dst, Register src1, Register src2) { 1093 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1094 int encode = vex_prefix_0F38_and_encode(dst, src1, src2); 1095 emit_int8((unsigned char)0xF2); 1096 emit_int8((unsigned char)(0xC0 | encode)); 1097 } 1098 1099 void Assembler::andnl(Register dst, Register src1, Address src2) { 1100 InstructionMark im(this); 1101 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1102 vex_prefix_0F38(dst, src1, src2); 1103 emit_int8((unsigned char)0xF2); 1104 emit_operand(dst, src2); 1105 } 1106 1107 void Assembler::bsfl(Register dst, Register src) { 1108 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1109 emit_int8(0x0F); 1110 emit_int8((unsigned char)0xBC); 1111 emit_int8((unsigned char)(0xC0 | encode)); 1112 } 1113 1114 void Assembler::bsrl(Register dst, Register src) { 1115 assert(!VM_Version::supports_lzcnt(), "encoding is treated as LZCNT"); 1116 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1117 emit_int8(0x0F); 1118 emit_int8((unsigned char)0xBD); 1119 emit_int8((unsigned char)(0xC0 | encode)); 1120 } 1121 1122 void Assembler::bswapl(Register reg) { // bswap 1123 int encode = prefix_and_encode(reg->encoding()); 1124 emit_int8(0x0F); 1125 emit_int8((unsigned char)(0xC8 | encode)); 1126 } 1127 1128 void Assembler::blsil(Register dst, Register src) { 1129 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1130 int encode = vex_prefix_0F38_and_encode(rbx, dst, src); 1131 emit_int8((unsigned char)0xF3); 1132 emit_int8((unsigned char)(0xC0 | encode)); 1133 } 1134 1135 void Assembler::blsil(Register dst, Address src) { 1136 InstructionMark im(this); 1137 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1138 vex_prefix_0F38(rbx, dst, src); 1139 emit_int8((unsigned char)0xF3); 1140 emit_operand(rbx, src); 1141 } 1142 1143 void Assembler::blsmskl(Register dst, Register src) { 1144 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1145 int encode = vex_prefix_0F38_and_encode(rdx, dst, src); 1146 emit_int8((unsigned char)0xF3); 1147 emit_int8((unsigned char)(0xC0 | encode)); 1148 } 1149 1150 void Assembler::blsmskl(Register dst, Address src) { 1151 InstructionMark im(this); 1152 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1153 vex_prefix_0F38(rdx, dst, src); 1154 emit_int8((unsigned char)0xF3); 1155 emit_operand(rdx, src); 1156 } 1157 1158 void Assembler::blsrl(Register dst, Register src) { 1159 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1160 int encode = vex_prefix_0F38_and_encode(rcx, dst, src); 1161 emit_int8((unsigned char)0xF3); 1162 emit_int8((unsigned char)(0xC0 | encode)); 1163 } 1164 1165 void Assembler::blsrl(Register dst, Address src) { 1166 InstructionMark im(this); 1167 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1168 vex_prefix_0F38(rcx, dst, src); 1169 emit_int8((unsigned char)0xF3); 1170 emit_operand(rcx, src); 1171 } 1172 1173 void Assembler::call(Label& L, relocInfo::relocType rtype) { 1174 // suspect disp32 is always good 1175 int operand = LP64_ONLY(disp32_operand) NOT_LP64(imm_operand); 1176 1177 if (L.is_bound()) { 1178 const int long_size = 5; 1179 int offs = (int)( target(L) - pc() ); 1180 assert(offs <= 0, "assembler error"); 1181 InstructionMark im(this); 1182 // 1110 1000 #32-bit disp 1183 emit_int8((unsigned char)0xE8); 1184 emit_data(offs - long_size, rtype, operand); 1185 } else { 1186 InstructionMark im(this); 1187 // 1110 1000 #32-bit disp 1188 L.add_patch_at(code(), locator()); 1189 1190 emit_int8((unsigned char)0xE8); 1191 emit_data(int(0), rtype, operand); 1192 } 1193 } 1194 1195 void Assembler::call(Register dst) { 1196 int encode = prefix_and_encode(dst->encoding()); 1197 emit_int8((unsigned char)0xFF); 1198 emit_int8((unsigned char)(0xD0 | encode)); 1199 } 1200 1201 1202 void Assembler::call(Address adr) { 1203 InstructionMark im(this); 1204 prefix(adr); 1205 emit_int8((unsigned char)0xFF); 1206 emit_operand(rdx, adr); 1207 } 1208 1209 void Assembler::call_literal(address entry, RelocationHolder const& rspec) { 1210 assert(entry != NULL, "call most probably wrong"); 1211 InstructionMark im(this); 1212 emit_int8((unsigned char)0xE8); 1213 intptr_t disp = entry - (pc() + sizeof(int32_t)); 1214 assert(is_simm32(disp), "must be 32bit offset (call2)"); 1215 // Technically, should use call32_operand, but this format is 1216 // implied by the fact that we're emitting a call instruction. 1217 1218 int operand = LP64_ONLY(disp32_operand) NOT_LP64(call32_operand); 1219 emit_data((int) disp, rspec, operand); 1220 } 1221 1222 void Assembler::cdql() { 1223 emit_int8((unsigned char)0x99); 1224 } 1225 1226 void Assembler::cld() { 1227 emit_int8((unsigned char)0xFC); 1228 } 1229 1230 void Assembler::cmovl(Condition cc, Register dst, Register src) { 1231 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction")); 1232 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1233 emit_int8(0x0F); 1234 emit_int8(0x40 | cc); 1235 emit_int8((unsigned char)(0xC0 | encode)); 1236 } 1237 1238 1239 void Assembler::cmovl(Condition cc, Register dst, Address src) { 1240 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction")); 1241 prefix(src, dst); 1242 emit_int8(0x0F); 1243 emit_int8(0x40 | cc); 1244 emit_operand(dst, src); 1245 } 1246 1247 void Assembler::cmpb(Address dst, int imm8) { 1248 InstructionMark im(this); 1249 prefix(dst); 1250 emit_int8((unsigned char)0x80); 1251 emit_operand(rdi, dst, 1); 1252 emit_int8(imm8); 1253 } 1254 1255 void Assembler::cmpl(Address dst, int32_t imm32) { 1256 InstructionMark im(this); 1257 prefix(dst); 1258 emit_int8((unsigned char)0x81); 1259 emit_operand(rdi, dst, 4); 1260 emit_int32(imm32); 1261 } 1262 1263 void Assembler::cmpl(Register dst, int32_t imm32) { 1264 prefix(dst); 1265 emit_arith(0x81, 0xF8, dst, imm32); 1266 } 1267 1268 void Assembler::cmpl(Register dst, Register src) { 1269 (void) prefix_and_encode(dst->encoding(), src->encoding()); 1270 emit_arith(0x3B, 0xC0, dst, src); 1271 } 1272 1273 1274 void Assembler::cmpl(Register dst, Address src) { 1275 InstructionMark im(this); 1276 prefix(src, dst); 1277 emit_int8((unsigned char)0x3B); 1278 emit_operand(dst, src); 1279 } 1280 1281 void Assembler::cmpw(Address dst, int imm16) { 1282 InstructionMark im(this); 1283 assert(!dst.base_needs_rex() && !dst.index_needs_rex(), "no extended registers"); 1284 emit_int8(0x66); 1285 emit_int8((unsigned char)0x81); 1286 emit_operand(rdi, dst, 2); 1287 emit_int16(imm16); 1288 } 1289 1290 // The 32-bit cmpxchg compares the value at adr with the contents of rax, 1291 // and stores reg into adr if so; otherwise, the value at adr is loaded into rax,. 1292 // The ZF is set if the compared values were equal, and cleared otherwise. 1293 void Assembler::cmpxchgl(Register reg, Address adr) { // cmpxchg 1294 InstructionMark im(this); 1295 prefix(adr, reg); 1296 emit_int8(0x0F); 1297 emit_int8((unsigned char)0xB1); 1298 emit_operand(reg, adr); 1299 } 1300 1301 void Assembler::comisd(XMMRegister dst, Address src) { 1302 // NOTE: dbx seems to decode this as comiss even though the 1303 // 0x66 is there. Strangly ucomisd comes out correct 1304 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1305 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66); 1306 } 1307 1308 void Assembler::comisd(XMMRegister dst, XMMRegister src) { 1309 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1310 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66); 1311 } 1312 1313 void Assembler::comiss(XMMRegister dst, Address src) { 1314 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1315 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE); 1316 } 1317 1318 void Assembler::comiss(XMMRegister dst, XMMRegister src) { 1319 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1320 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE); 1321 } 1322 1323 void Assembler::cpuid() { 1324 emit_int8(0x0F); 1325 emit_int8((unsigned char)0xA2); 1326 } 1327 1328 void Assembler::cvtdq2pd(XMMRegister dst, XMMRegister src) { 1329 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1330 emit_simd_arith_nonds(0xE6, dst, src, VEX_SIMD_F3); 1331 } 1332 1333 void Assembler::cvtdq2ps(XMMRegister dst, XMMRegister src) { 1334 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1335 emit_simd_arith_nonds(0x5B, dst, src, VEX_SIMD_NONE); 1336 } 1337 1338 void Assembler::cvtsd2ss(XMMRegister dst, XMMRegister src) { 1339 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1340 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2); 1341 } 1342 1343 void Assembler::cvtsd2ss(XMMRegister dst, Address src) { 1344 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1345 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2); 1346 } 1347 1348 void Assembler::cvtsi2sdl(XMMRegister dst, Register src) { 1349 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1350 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2); 1351 emit_int8(0x2A); 1352 emit_int8((unsigned char)(0xC0 | encode)); 1353 } 1354 1355 void Assembler::cvtsi2sdl(XMMRegister dst, Address src) { 1356 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1357 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F2); 1358 } 1359 1360 void Assembler::cvtsi2ssl(XMMRegister dst, Register src) { 1361 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1362 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3); 1363 emit_int8(0x2A); 1364 emit_int8((unsigned char)(0xC0 | encode)); 1365 } 1366 1367 void Assembler::cvtsi2ssl(XMMRegister dst, Address src) { 1368 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1369 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F3); 1370 } 1371 1372 void Assembler::cvtss2sd(XMMRegister dst, XMMRegister src) { 1373 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1374 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3); 1375 } 1376 1377 void Assembler::cvtss2sd(XMMRegister dst, Address src) { 1378 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1379 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3); 1380 } 1381 1382 1383 void Assembler::cvttsd2sil(Register dst, XMMRegister src) { 1384 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1385 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F2); 1386 emit_int8(0x2C); 1387 emit_int8((unsigned char)(0xC0 | encode)); 1388 } 1389 1390 void Assembler::cvttss2sil(Register dst, XMMRegister src) { 1391 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1392 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F3); 1393 emit_int8(0x2C); 1394 emit_int8((unsigned char)(0xC0 | encode)); 1395 } 1396 1397 void Assembler::decl(Address dst) { 1398 // Don't use it directly. Use MacroAssembler::decrement() instead. 1399 InstructionMark im(this); 1400 prefix(dst); 1401 emit_int8((unsigned char)0xFF); 1402 emit_operand(rcx, dst); 1403 } 1404 1405 void Assembler::divsd(XMMRegister dst, Address src) { 1406 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1407 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2); 1408 } 1409 1410 void Assembler::divsd(XMMRegister dst, XMMRegister src) { 1411 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1412 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2); 1413 } 1414 1415 void Assembler::divss(XMMRegister dst, Address src) { 1416 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1417 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3); 1418 } 1419 1420 void Assembler::divss(XMMRegister dst, XMMRegister src) { 1421 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1422 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3); 1423 } 1424 1425 void Assembler::emms() { 1426 NOT_LP64(assert(VM_Version::supports_mmx(), "")); 1427 emit_int8(0x0F); 1428 emit_int8(0x77); 1429 } 1430 1431 void Assembler::hlt() { 1432 emit_int8((unsigned char)0xF4); 1433 } 1434 1435 void Assembler::idivl(Register src) { 1436 int encode = prefix_and_encode(src->encoding()); 1437 emit_int8((unsigned char)0xF7); 1438 emit_int8((unsigned char)(0xF8 | encode)); 1439 } 1440 1441 void Assembler::divl(Register src) { // Unsigned 1442 int encode = prefix_and_encode(src->encoding()); 1443 emit_int8((unsigned char)0xF7); 1444 emit_int8((unsigned char)(0xF0 | encode)); 1445 } 1446 1447 void Assembler::imull(Register dst, Register src) { 1448 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1449 emit_int8(0x0F); 1450 emit_int8((unsigned char)0xAF); 1451 emit_int8((unsigned char)(0xC0 | encode)); 1452 } 1453 1454 1455 void Assembler::imull(Register dst, Register src, int value) { 1456 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1457 if (is8bit(value)) { 1458 emit_int8(0x6B); 1459 emit_int8((unsigned char)(0xC0 | encode)); 1460 emit_int8(value & 0xFF); 1461 } else { 1462 emit_int8(0x69); 1463 emit_int8((unsigned char)(0xC0 | encode)); 1464 emit_int32(value); 1465 } 1466 } 1467 1468 void Assembler::imull(Register dst, Address src) { 1469 InstructionMark im(this); 1470 prefix(src, dst); 1471 emit_int8(0x0F); 1472 emit_int8((unsigned char) 0xAF); 1473 emit_operand(dst, src); 1474 } 1475 1476 1477 void Assembler::incl(Address dst) { 1478 // Don't use it directly. Use MacroAssembler::increment() instead. 1479 InstructionMark im(this); 1480 prefix(dst); 1481 emit_int8((unsigned char)0xFF); 1482 emit_operand(rax, dst); 1483 } 1484 1485 void Assembler::jcc(Condition cc, Label& L, bool maybe_short) { 1486 InstructionMark im(this); 1487 assert((0 <= cc) && (cc < 16), "illegal cc"); 1488 if (L.is_bound()) { 1489 address dst = target(L); 1490 assert(dst != NULL, "jcc most probably wrong"); 1491 1492 const int short_size = 2; 1493 const int long_size = 6; 1494 intptr_t offs = (intptr_t)dst - (intptr_t)pc(); 1495 if (maybe_short && is8bit(offs - short_size)) { 1496 // 0111 tttn #8-bit disp 1497 emit_int8(0x70 | cc); 1498 emit_int8((offs - short_size) & 0xFF); 1499 } else { 1500 // 0000 1111 1000 tttn #32-bit disp 1501 assert(is_simm32(offs - long_size), 1502 "must be 32bit offset (call4)"); 1503 emit_int8(0x0F); 1504 emit_int8((unsigned char)(0x80 | cc)); 1505 emit_int32(offs - long_size); 1506 } 1507 } else { 1508 // Note: could eliminate cond. jumps to this jump if condition 1509 // is the same however, seems to be rather unlikely case. 1510 // Note: use jccb() if label to be bound is very close to get 1511 // an 8-bit displacement 1512 L.add_patch_at(code(), locator()); 1513 emit_int8(0x0F); 1514 emit_int8((unsigned char)(0x80 | cc)); 1515 emit_int32(0); 1516 } 1517 } 1518 1519 void Assembler::jccb(Condition cc, Label& L) { 1520 if (L.is_bound()) { 1521 const int short_size = 2; 1522 address entry = target(L); 1523 #ifdef ASSERT 1524 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size); 1525 intptr_t delta = short_branch_delta(); 1526 if (delta != 0) { 1527 dist += (dist < 0 ? (-delta) :delta); 1528 } 1529 assert(is8bit(dist), "Dispacement too large for a short jmp"); 1530 #endif 1531 intptr_t offs = (intptr_t)entry - (intptr_t)pc(); 1532 // 0111 tttn #8-bit disp 1533 emit_int8(0x70 | cc); 1534 emit_int8((offs - short_size) & 0xFF); 1535 } else { 1536 InstructionMark im(this); 1537 L.add_patch_at(code(), locator()); 1538 emit_int8(0x70 | cc); 1539 emit_int8(0); 1540 } 1541 } 1542 1543 void Assembler::jmp(Address adr) { 1544 InstructionMark im(this); 1545 prefix(adr); 1546 emit_int8((unsigned char)0xFF); 1547 emit_operand(rsp, adr); 1548 } 1549 1550 void Assembler::jmp(Label& L, bool maybe_short) { 1551 if (L.is_bound()) { 1552 address entry = target(L); 1553 assert(entry != NULL, "jmp most probably wrong"); 1554 InstructionMark im(this); 1555 const int short_size = 2; 1556 const int long_size = 5; 1557 intptr_t offs = entry - pc(); 1558 if (maybe_short && is8bit(offs - short_size)) { 1559 emit_int8((unsigned char)0xEB); 1560 emit_int8((offs - short_size) & 0xFF); 1561 } else { 1562 emit_int8((unsigned char)0xE9); 1563 emit_int32(offs - long_size); 1564 } 1565 } else { 1566 // By default, forward jumps are always 32-bit displacements, since 1567 // we can't yet know where the label will be bound. If you're sure that 1568 // the forward jump will not run beyond 256 bytes, use jmpb to 1569 // force an 8-bit displacement. 1570 InstructionMark im(this); 1571 L.add_patch_at(code(), locator()); 1572 emit_int8((unsigned char)0xE9); 1573 emit_int32(0); 1574 } 1575 } 1576 1577 void Assembler::jmp(Register entry) { 1578 int encode = prefix_and_encode(entry->encoding()); 1579 emit_int8((unsigned char)0xFF); 1580 emit_int8((unsigned char)(0xE0 | encode)); 1581 } 1582 1583 void Assembler::jmp_literal(address dest, RelocationHolder const& rspec) { 1584 InstructionMark im(this); 1585 emit_int8((unsigned char)0xE9); 1586 assert(dest != NULL, "must have a target"); 1587 intptr_t disp = dest - (pc() + sizeof(int32_t)); 1588 assert(is_simm32(disp), "must be 32bit offset (jmp)"); 1589 emit_data(disp, rspec.reloc(), call32_operand); 1590 } 1591 1592 void Assembler::jmpb(Label& L) { 1593 if (L.is_bound()) { 1594 const int short_size = 2; 1595 address entry = target(L); 1596 assert(entry != NULL, "jmp most probably wrong"); 1597 #ifdef ASSERT 1598 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size); 1599 intptr_t delta = short_branch_delta(); 1600 if (delta != 0) { 1601 dist += (dist < 0 ? (-delta) :delta); 1602 } 1603 assert(is8bit(dist), "Dispacement too large for a short jmp"); 1604 #endif 1605 intptr_t offs = entry - pc(); 1606 emit_int8((unsigned char)0xEB); 1607 emit_int8((offs - short_size) & 0xFF); 1608 } else { 1609 InstructionMark im(this); 1610 L.add_patch_at(code(), locator()); 1611 emit_int8((unsigned char)0xEB); 1612 emit_int8(0); 1613 } 1614 } 1615 1616 void Assembler::ldmxcsr( Address src) { 1617 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1618 InstructionMark im(this); 1619 prefix(src); 1620 emit_int8(0x0F); 1621 emit_int8((unsigned char)0xAE); 1622 emit_operand(as_Register(2), src); 1623 } 1624 1625 void Assembler::leal(Register dst, Address src) { 1626 InstructionMark im(this); 1627 #ifdef _LP64 1628 emit_int8(0x67); // addr32 1629 prefix(src, dst); 1630 #endif // LP64 1631 emit_int8((unsigned char)0x8D); 1632 emit_operand(dst, src); 1633 } 1634 1635 void Assembler::lfence() { 1636 emit_int8(0x0F); 1637 emit_int8((unsigned char)0xAE); 1638 emit_int8((unsigned char)0xE8); 1639 } 1640 1641 void Assembler::lock() { 1642 emit_int8((unsigned char)0xF0); 1643 } 1644 1645 void Assembler::lzcntl(Register dst, Register src) { 1646 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR"); 1647 emit_int8((unsigned char)0xF3); 1648 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1649 emit_int8(0x0F); 1650 emit_int8((unsigned char)0xBD); 1651 emit_int8((unsigned char)(0xC0 | encode)); 1652 } 1653 1654 // Emit mfence instruction 1655 void Assembler::mfence() { 1656 NOT_LP64(assert(VM_Version::supports_sse2(), "unsupported");) 1657 emit_int8(0x0F); 1658 emit_int8((unsigned char)0xAE); 1659 emit_int8((unsigned char)0xF0); 1660 } 1661 1662 void Assembler::mov(Register dst, Register src) { 1663 LP64_ONLY(movq(dst, src)) NOT_LP64(movl(dst, src)); 1664 } 1665 1666 void Assembler::movapd(XMMRegister dst, XMMRegister src) { 1667 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1668 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_66); 1669 } 1670 1671 void Assembler::movaps(XMMRegister dst, XMMRegister src) { 1672 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1673 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_NONE); 1674 } 1675 1676 void Assembler::movlhps(XMMRegister dst, XMMRegister src) { 1677 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1678 int encode = simd_prefix_and_encode(dst, src, src, VEX_SIMD_NONE); 1679 emit_int8(0x16); 1680 emit_int8((unsigned char)(0xC0 | encode)); 1681 } 1682 1683 void Assembler::movb(Register dst, Address src) { 1684 NOT_LP64(assert(dst->has_byte_register(), "must have byte register")); 1685 InstructionMark im(this); 1686 prefix(src, dst, true); 1687 emit_int8((unsigned char)0x8A); 1688 emit_operand(dst, src); 1689 } 1690 1691 1692 void Assembler::movb(Address dst, int imm8) { 1693 InstructionMark im(this); 1694 prefix(dst); 1695 emit_int8((unsigned char)0xC6); 1696 emit_operand(rax, dst, 1); 1697 emit_int8(imm8); 1698 } 1699 1700 1701 void Assembler::movb(Address dst, Register src) { 1702 assert(src->has_byte_register(), "must have byte register"); 1703 InstructionMark im(this); 1704 prefix(dst, src, true); 1705 emit_int8((unsigned char)0x88); 1706 emit_operand(src, dst); 1707 } 1708 1709 void Assembler::movdl(XMMRegister dst, Register src) { 1710 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1711 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_66); 1712 emit_int8(0x6E); 1713 emit_int8((unsigned char)(0xC0 | encode)); 1714 } 1715 1716 void Assembler::movdl(Register dst, XMMRegister src) { 1717 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1718 // swap src/dst to get correct prefix 1719 int encode = simd_prefix_and_encode(src, dst, VEX_SIMD_66); 1720 emit_int8(0x7E); 1721 emit_int8((unsigned char)(0xC0 | encode)); 1722 } 1723 1724 void Assembler::movdl(XMMRegister dst, Address src) { 1725 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1726 InstructionMark im(this); 1727 simd_prefix(dst, src, VEX_SIMD_66); 1728 emit_int8(0x6E); 1729 emit_operand(dst, src); 1730 } 1731 1732 void Assembler::movdl(Address dst, XMMRegister src) { 1733 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1734 InstructionMark im(this); 1735 simd_prefix(dst, src, VEX_SIMD_66); 1736 emit_int8(0x7E); 1737 emit_operand(src, dst); 1738 } 1739 1740 void Assembler::movdqa(XMMRegister dst, XMMRegister src) { 1741 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1742 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_66); 1743 } 1744 1745 void Assembler::movdqa(XMMRegister dst, Address src) { 1746 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1747 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_66); 1748 } 1749 1750 void Assembler::movdqu(XMMRegister dst, Address src) { 1751 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1752 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3); 1753 } 1754 1755 void Assembler::movdqu(XMMRegister dst, XMMRegister src) { 1756 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1757 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3); 1758 } 1759 1760 void Assembler::movdqu(Address dst, XMMRegister src) { 1761 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1762 InstructionMark im(this); 1763 simd_prefix(dst, src, VEX_SIMD_F3); 1764 emit_int8(0x7F); 1765 emit_operand(src, dst); 1766 } 1767 1768 // Move Unaligned 256bit Vector 1769 void Assembler::vmovdqu(XMMRegister dst, XMMRegister src) { 1770 assert(UseAVX, ""); 1771 bool vector256 = true; 1772 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_F3, vector256); 1773 emit_int8(0x6F); 1774 emit_int8((unsigned char)(0xC0 | encode)); 1775 } 1776 1777 void Assembler::vmovdqu(XMMRegister dst, Address src) { 1778 assert(UseAVX, ""); 1779 InstructionMark im(this); 1780 bool vector256 = true; 1781 vex_prefix(dst, xnoreg, src, VEX_SIMD_F3, vector256); 1782 emit_int8(0x6F); 1783 emit_operand(dst, src); 1784 } 1785 1786 void Assembler::vmovdqu(Address dst, XMMRegister src) { 1787 assert(UseAVX, ""); 1788 InstructionMark im(this); 1789 bool vector256 = true; 1790 // swap src<->dst for encoding 1791 assert(src != xnoreg, "sanity"); 1792 vex_prefix(src, xnoreg, dst, VEX_SIMD_F3, vector256); 1793 emit_int8(0x7F); 1794 emit_operand(src, dst); 1795 } 1796 1797 // Uses zero extension on 64bit 1798 1799 void Assembler::movl(Register dst, int32_t imm32) { 1800 int encode = prefix_and_encode(dst->encoding()); 1801 emit_int8((unsigned char)(0xB8 | encode)); 1802 emit_int32(imm32); 1803 } 1804 1805 void Assembler::movl(Register dst, Register src) { 1806 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1807 emit_int8((unsigned char)0x8B); 1808 emit_int8((unsigned char)(0xC0 | encode)); 1809 } 1810 1811 void Assembler::movl(Register dst, Address src) { 1812 InstructionMark im(this); 1813 prefix(src, dst); 1814 emit_int8((unsigned char)0x8B); 1815 emit_operand(dst, src); 1816 } 1817 1818 void Assembler::movl(Address dst, int32_t imm32) { 1819 InstructionMark im(this); 1820 prefix(dst); 1821 emit_int8((unsigned char)0xC7); 1822 emit_operand(rax, dst, 4); 1823 emit_int32(imm32); 1824 } 1825 1826 void Assembler::movl(Address dst, Register src) { 1827 InstructionMark im(this); 1828 prefix(dst, src); 1829 emit_int8((unsigned char)0x89); 1830 emit_operand(src, dst); 1831 } 1832 1833 // New cpus require to use movsd and movss to avoid partial register stall 1834 // when loading from memory. But for old Opteron use movlpd instead of movsd. 1835 // The selection is done in MacroAssembler::movdbl() and movflt(). 1836 void Assembler::movlpd(XMMRegister dst, Address src) { 1837 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1838 emit_simd_arith(0x12, dst, src, VEX_SIMD_66); 1839 } 1840 1841 void Assembler::movq( MMXRegister dst, Address src ) { 1842 assert( VM_Version::supports_mmx(), "" ); 1843 emit_int8(0x0F); 1844 emit_int8(0x6F); 1845 emit_operand(dst, src); 1846 } 1847 1848 void Assembler::movq( Address dst, MMXRegister src ) { 1849 assert( VM_Version::supports_mmx(), "" ); 1850 emit_int8(0x0F); 1851 emit_int8(0x7F); 1852 // workaround gcc (3.2.1-7a) bug 1853 // In that version of gcc with only an emit_operand(MMX, Address) 1854 // gcc will tail jump and try and reverse the parameters completely 1855 // obliterating dst in the process. By having a version available 1856 // that doesn't need to swap the args at the tail jump the bug is 1857 // avoided. 1858 emit_operand(dst, src); 1859 } 1860 1861 void Assembler::movq(XMMRegister dst, Address src) { 1862 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1863 InstructionMark im(this); 1864 simd_prefix(dst, src, VEX_SIMD_F3); 1865 emit_int8(0x7E); 1866 emit_operand(dst, src); 1867 } 1868 1869 void Assembler::movq(Address dst, XMMRegister src) { 1870 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1871 InstructionMark im(this); 1872 simd_prefix(dst, src, VEX_SIMD_66); 1873 emit_int8((unsigned char)0xD6); 1874 emit_operand(src, dst); 1875 } 1876 1877 void Assembler::movsbl(Register dst, Address src) { // movsxb 1878 InstructionMark im(this); 1879 prefix(src, dst); 1880 emit_int8(0x0F); 1881 emit_int8((unsigned char)0xBE); 1882 emit_operand(dst, src); 1883 } 1884 1885 void Assembler::movsbl(Register dst, Register src) { // movsxb 1886 NOT_LP64(assert(src->has_byte_register(), "must have byte register")); 1887 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true); 1888 emit_int8(0x0F); 1889 emit_int8((unsigned char)0xBE); 1890 emit_int8((unsigned char)(0xC0 | encode)); 1891 } 1892 1893 void Assembler::movsd(XMMRegister dst, XMMRegister src) { 1894 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1895 emit_simd_arith(0x10, dst, src, VEX_SIMD_F2); 1896 } 1897 1898 void Assembler::movsd(XMMRegister dst, Address src) { 1899 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1900 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F2); 1901 } 1902 1903 void Assembler::movsd(Address dst, XMMRegister src) { 1904 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1905 InstructionMark im(this); 1906 simd_prefix(dst, src, VEX_SIMD_F2); 1907 emit_int8(0x11); 1908 emit_operand(src, dst); 1909 } 1910 1911 void Assembler::movss(XMMRegister dst, XMMRegister src) { 1912 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1913 emit_simd_arith(0x10, dst, src, VEX_SIMD_F3); 1914 } 1915 1916 void Assembler::movss(XMMRegister dst, Address src) { 1917 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1918 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F3); 1919 } 1920 1921 void Assembler::movss(Address dst, XMMRegister src) { 1922 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1923 InstructionMark im(this); 1924 simd_prefix(dst, src, VEX_SIMD_F3); 1925 emit_int8(0x11); 1926 emit_operand(src, dst); 1927 } 1928 1929 void Assembler::movswl(Register dst, Address src) { // movsxw 1930 InstructionMark im(this); 1931 prefix(src, dst); 1932 emit_int8(0x0F); 1933 emit_int8((unsigned char)0xBF); 1934 emit_operand(dst, src); 1935 } 1936 1937 void Assembler::movswl(Register dst, Register src) { // movsxw 1938 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1939 emit_int8(0x0F); 1940 emit_int8((unsigned char)0xBF); 1941 emit_int8((unsigned char)(0xC0 | encode)); 1942 } 1943 1944 void Assembler::movw(Address dst, int imm16) { 1945 InstructionMark im(this); 1946 1947 emit_int8(0x66); // switch to 16-bit mode 1948 prefix(dst); 1949 emit_int8((unsigned char)0xC7); 1950 emit_operand(rax, dst, 2); 1951 emit_int16(imm16); 1952 } 1953 1954 void Assembler::movw(Register dst, Address src) { 1955 InstructionMark im(this); 1956 emit_int8(0x66); 1957 prefix(src, dst); 1958 emit_int8((unsigned char)0x8B); 1959 emit_operand(dst, src); 1960 } 1961 1962 void Assembler::movw(Address dst, Register src) { 1963 InstructionMark im(this); 1964 emit_int8(0x66); 1965 prefix(dst, src); 1966 emit_int8((unsigned char)0x89); 1967 emit_operand(src, dst); 1968 } 1969 1970 void Assembler::movzbl(Register dst, Address src) { // movzxb 1971 InstructionMark im(this); 1972 prefix(src, dst); 1973 emit_int8(0x0F); 1974 emit_int8((unsigned char)0xB6); 1975 emit_operand(dst, src); 1976 } 1977 1978 void Assembler::movzbl(Register dst, Register src) { // movzxb 1979 NOT_LP64(assert(src->has_byte_register(), "must have byte register")); 1980 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true); 1981 emit_int8(0x0F); 1982 emit_int8((unsigned char)0xB6); 1983 emit_int8(0xC0 | encode); 1984 } 1985 1986 void Assembler::movzwl(Register dst, Address src) { // movzxw 1987 InstructionMark im(this); 1988 prefix(src, dst); 1989 emit_int8(0x0F); 1990 emit_int8((unsigned char)0xB7); 1991 emit_operand(dst, src); 1992 } 1993 1994 void Assembler::movzwl(Register dst, Register src) { // movzxw 1995 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1996 emit_int8(0x0F); 1997 emit_int8((unsigned char)0xB7); 1998 emit_int8(0xC0 | encode); 1999 } 2000 2001 void Assembler::mull(Address src) { 2002 InstructionMark im(this); 2003 prefix(src); 2004 emit_int8((unsigned char)0xF7); 2005 emit_operand(rsp, src); 2006 } 2007 2008 void Assembler::mull(Register src) { 2009 int encode = prefix_and_encode(src->encoding()); 2010 emit_int8((unsigned char)0xF7); 2011 emit_int8((unsigned char)(0xE0 | encode)); 2012 } 2013 2014 void Assembler::mulsd(XMMRegister dst, Address src) { 2015 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2016 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2); 2017 } 2018 2019 void Assembler::mulsd(XMMRegister dst, XMMRegister src) { 2020 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2021 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2); 2022 } 2023 2024 void Assembler::mulss(XMMRegister dst, Address src) { 2025 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2026 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3); 2027 } 2028 2029 void Assembler::mulss(XMMRegister dst, XMMRegister src) { 2030 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2031 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3); 2032 } 2033 2034 void Assembler::negl(Register dst) { 2035 int encode = prefix_and_encode(dst->encoding()); 2036 emit_int8((unsigned char)0xF7); 2037 emit_int8((unsigned char)(0xD8 | encode)); 2038 } 2039 2040 void Assembler::nop(int i) { 2041 #ifdef ASSERT 2042 assert(i > 0, " "); 2043 // The fancy nops aren't currently recognized by debuggers making it a 2044 // pain to disassemble code while debugging. If asserts are on clearly 2045 // speed is not an issue so simply use the single byte traditional nop 2046 // to do alignment. 2047 2048 for (; i > 0 ; i--) emit_int8((unsigned char)0x90); 2049 return; 2050 2051 #endif // ASSERT 2052 2053 if (UseAddressNop && VM_Version::is_intel()) { 2054 // 2055 // Using multi-bytes nops "0x0F 0x1F [address]" for Intel 2056 // 1: 0x90 2057 // 2: 0x66 0x90 2058 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding) 2059 // 4: 0x0F 0x1F 0x40 0x00 2060 // 5: 0x0F 0x1F 0x44 0x00 0x00 2061 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00 2062 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2063 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2064 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2065 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2066 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2067 2068 // The rest coding is Intel specific - don't use consecutive address nops 2069 2070 // 12: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2071 // 13: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2072 // 14: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2073 // 15: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2074 2075 while(i >= 15) { 2076 // For Intel don't generate consecutive addess nops (mix with regular nops) 2077 i -= 15; 2078 emit_int8(0x66); // size prefix 2079 emit_int8(0x66); // size prefix 2080 emit_int8(0x66); // size prefix 2081 addr_nop_8(); 2082 emit_int8(0x66); // size prefix 2083 emit_int8(0x66); // size prefix 2084 emit_int8(0x66); // size prefix 2085 emit_int8((unsigned char)0x90); 2086 // nop 2087 } 2088 switch (i) { 2089 case 14: 2090 emit_int8(0x66); // size prefix 2091 case 13: 2092 emit_int8(0x66); // size prefix 2093 case 12: 2094 addr_nop_8(); 2095 emit_int8(0x66); // size prefix 2096 emit_int8(0x66); // size prefix 2097 emit_int8(0x66); // size prefix 2098 emit_int8((unsigned char)0x90); 2099 // nop 2100 break; 2101 case 11: 2102 emit_int8(0x66); // size prefix 2103 case 10: 2104 emit_int8(0x66); // size prefix 2105 case 9: 2106 emit_int8(0x66); // size prefix 2107 case 8: 2108 addr_nop_8(); 2109 break; 2110 case 7: 2111 addr_nop_7(); 2112 break; 2113 case 6: 2114 emit_int8(0x66); // size prefix 2115 case 5: 2116 addr_nop_5(); 2117 break; 2118 case 4: 2119 addr_nop_4(); 2120 break; 2121 case 3: 2122 // Don't use "0x0F 0x1F 0x00" - need patching safe padding 2123 emit_int8(0x66); // size prefix 2124 case 2: 2125 emit_int8(0x66); // size prefix 2126 case 1: 2127 emit_int8((unsigned char)0x90); 2128 // nop 2129 break; 2130 default: 2131 assert(i == 0, " "); 2132 } 2133 return; 2134 } 2135 if (UseAddressNop && VM_Version::is_amd()) { 2136 // 2137 // Using multi-bytes nops "0x0F 0x1F [address]" for AMD. 2138 // 1: 0x90 2139 // 2: 0x66 0x90 2140 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding) 2141 // 4: 0x0F 0x1F 0x40 0x00 2142 // 5: 0x0F 0x1F 0x44 0x00 0x00 2143 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00 2144 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2145 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2146 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2147 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2148 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2149 2150 // The rest coding is AMD specific - use consecutive address nops 2151 2152 // 12: 0x66 0x0F 0x1F 0x44 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00 2153 // 13: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00 2154 // 14: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2155 // 15: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2156 // 16: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2157 // Size prefixes (0x66) are added for larger sizes 2158 2159 while(i >= 22) { 2160 i -= 11; 2161 emit_int8(0x66); // size prefix 2162 emit_int8(0x66); // size prefix 2163 emit_int8(0x66); // size prefix 2164 addr_nop_8(); 2165 } 2166 // Generate first nop for size between 21-12 2167 switch (i) { 2168 case 21: 2169 i -= 1; 2170 emit_int8(0x66); // size prefix 2171 case 20: 2172 case 19: 2173 i -= 1; 2174 emit_int8(0x66); // size prefix 2175 case 18: 2176 case 17: 2177 i -= 1; 2178 emit_int8(0x66); // size prefix 2179 case 16: 2180 case 15: 2181 i -= 8; 2182 addr_nop_8(); 2183 break; 2184 case 14: 2185 case 13: 2186 i -= 7; 2187 addr_nop_7(); 2188 break; 2189 case 12: 2190 i -= 6; 2191 emit_int8(0x66); // size prefix 2192 addr_nop_5(); 2193 break; 2194 default: 2195 assert(i < 12, " "); 2196 } 2197 2198 // Generate second nop for size between 11-1 2199 switch (i) { 2200 case 11: 2201 emit_int8(0x66); // size prefix 2202 case 10: 2203 emit_int8(0x66); // size prefix 2204 case 9: 2205 emit_int8(0x66); // size prefix 2206 case 8: 2207 addr_nop_8(); 2208 break; 2209 case 7: 2210 addr_nop_7(); 2211 break; 2212 case 6: 2213 emit_int8(0x66); // size prefix 2214 case 5: 2215 addr_nop_5(); 2216 break; 2217 case 4: 2218 addr_nop_4(); 2219 break; 2220 case 3: 2221 // Don't use "0x0F 0x1F 0x00" - need patching safe padding 2222 emit_int8(0x66); // size prefix 2223 case 2: 2224 emit_int8(0x66); // size prefix 2225 case 1: 2226 emit_int8((unsigned char)0x90); 2227 // nop 2228 break; 2229 default: 2230 assert(i == 0, " "); 2231 } 2232 return; 2233 } 2234 2235 // Using nops with size prefixes "0x66 0x90". 2236 // From AMD Optimization Guide: 2237 // 1: 0x90 2238 // 2: 0x66 0x90 2239 // 3: 0x66 0x66 0x90 2240 // 4: 0x66 0x66 0x66 0x90 2241 // 5: 0x66 0x66 0x90 0x66 0x90 2242 // 6: 0x66 0x66 0x90 0x66 0x66 0x90 2243 // 7: 0x66 0x66 0x66 0x90 0x66 0x66 0x90 2244 // 8: 0x66 0x66 0x66 0x90 0x66 0x66 0x66 0x90 2245 // 9: 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90 2246 // 10: 0x66 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90 2247 // 2248 while(i > 12) { 2249 i -= 4; 2250 emit_int8(0x66); // size prefix 2251 emit_int8(0x66); 2252 emit_int8(0x66); 2253 emit_int8((unsigned char)0x90); 2254 // nop 2255 } 2256 // 1 - 12 nops 2257 if(i > 8) { 2258 if(i > 9) { 2259 i -= 1; 2260 emit_int8(0x66); 2261 } 2262 i -= 3; 2263 emit_int8(0x66); 2264 emit_int8(0x66); 2265 emit_int8((unsigned char)0x90); 2266 } 2267 // 1 - 8 nops 2268 if(i > 4) { 2269 if(i > 6) { 2270 i -= 1; 2271 emit_int8(0x66); 2272 } 2273 i -= 3; 2274 emit_int8(0x66); 2275 emit_int8(0x66); 2276 emit_int8((unsigned char)0x90); 2277 } 2278 switch (i) { 2279 case 4: 2280 emit_int8(0x66); 2281 case 3: 2282 emit_int8(0x66); 2283 case 2: 2284 emit_int8(0x66); 2285 case 1: 2286 emit_int8((unsigned char)0x90); 2287 break; 2288 default: 2289 assert(i == 0, " "); 2290 } 2291 } 2292 2293 void Assembler::notl(Register dst) { 2294 int encode = prefix_and_encode(dst->encoding()); 2295 emit_int8((unsigned char)0xF7); 2296 emit_int8((unsigned char)(0xD0 | encode)); 2297 } 2298 2299 void Assembler::orl(Address dst, int32_t imm32) { 2300 InstructionMark im(this); 2301 prefix(dst); 2302 emit_arith_operand(0x81, rcx, dst, imm32); 2303 } 2304 2305 void Assembler::orl(Register dst, int32_t imm32) { 2306 prefix(dst); 2307 emit_arith(0x81, 0xC8, dst, imm32); 2308 } 2309 2310 void Assembler::orl(Register dst, Address src) { 2311 InstructionMark im(this); 2312 prefix(src, dst); 2313 emit_int8(0x0B); 2314 emit_operand(dst, src); 2315 } 2316 2317 void Assembler::orl(Register dst, Register src) { 2318 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2319 emit_arith(0x0B, 0xC0, dst, src); 2320 } 2321 2322 void Assembler::packuswb(XMMRegister dst, Address src) { 2323 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2324 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2325 emit_simd_arith(0x67, dst, src, VEX_SIMD_66); 2326 } 2327 2328 void Assembler::packuswb(XMMRegister dst, XMMRegister src) { 2329 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2330 emit_simd_arith(0x67, dst, src, VEX_SIMD_66); 2331 } 2332 2333 void Assembler::vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 2334 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 2335 emit_vex_arith(0x67, dst, nds, src, VEX_SIMD_66, vector256); 2336 } 2337 2338 void Assembler::vpermq(XMMRegister dst, XMMRegister src, int imm8, bool vector256) { 2339 assert(VM_Version::supports_avx2(), ""); 2340 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, true, vector256); 2341 emit_int8(0x00); 2342 emit_int8(0xC0 | encode); 2343 emit_int8(imm8); 2344 } 2345 2346 void Assembler::pcmpestri(XMMRegister dst, Address src, int imm8) { 2347 assert(VM_Version::supports_sse4_2(), ""); 2348 InstructionMark im(this); 2349 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_3A); 2350 emit_int8(0x61); 2351 emit_operand(dst, src); 2352 emit_int8(imm8); 2353 } 2354 2355 void Assembler::pcmpestri(XMMRegister dst, XMMRegister src, int imm8) { 2356 assert(VM_Version::supports_sse4_2(), ""); 2357 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A); 2358 emit_int8(0x61); 2359 emit_int8((unsigned char)(0xC0 | encode)); 2360 emit_int8(imm8); 2361 } 2362 2363 void Assembler::pextrd(Register dst, XMMRegister src, int imm8) { 2364 assert(VM_Version::supports_sse4_1(), ""); 2365 int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, false); 2366 emit_int8(0x16); 2367 emit_int8((unsigned char)(0xC0 | encode)); 2368 emit_int8(imm8); 2369 } 2370 2371 void Assembler::pextrq(Register dst, XMMRegister src, int imm8) { 2372 assert(VM_Version::supports_sse4_1(), ""); 2373 int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, true); 2374 emit_int8(0x16); 2375 emit_int8((unsigned char)(0xC0 | encode)); 2376 emit_int8(imm8); 2377 } 2378 2379 void Assembler::pinsrd(XMMRegister dst, Register src, int imm8) { 2380 assert(VM_Version::supports_sse4_1(), ""); 2381 int encode = simd_prefix_and_encode(dst, dst, as_XMMRegister(src->encoding()), VEX_SIMD_66, VEX_OPCODE_0F_3A, false); 2382 emit_int8(0x22); 2383 emit_int8((unsigned char)(0xC0 | encode)); 2384 emit_int8(imm8); 2385 } 2386 2387 void Assembler::pinsrq(XMMRegister dst, Register src, int imm8) { 2388 assert(VM_Version::supports_sse4_1(), ""); 2389 int encode = simd_prefix_and_encode(dst, dst, as_XMMRegister(src->encoding()), VEX_SIMD_66, VEX_OPCODE_0F_3A, true); 2390 emit_int8(0x22); 2391 emit_int8((unsigned char)(0xC0 | encode)); 2392 emit_int8(imm8); 2393 } 2394 2395 void Assembler::pmovzxbw(XMMRegister dst, Address src) { 2396 assert(VM_Version::supports_sse4_1(), ""); 2397 InstructionMark im(this); 2398 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2399 emit_int8(0x30); 2400 emit_operand(dst, src); 2401 } 2402 2403 void Assembler::pmovzxbw(XMMRegister dst, XMMRegister src) { 2404 assert(VM_Version::supports_sse4_1(), ""); 2405 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2406 emit_int8(0x30); 2407 emit_int8((unsigned char)(0xC0 | encode)); 2408 } 2409 2410 // generic 2411 void Assembler::pop(Register dst) { 2412 int encode = prefix_and_encode(dst->encoding()); 2413 emit_int8(0x58 | encode); 2414 } 2415 2416 void Assembler::popcntl(Register dst, Address src) { 2417 assert(VM_Version::supports_popcnt(), "must support"); 2418 InstructionMark im(this); 2419 emit_int8((unsigned char)0xF3); 2420 prefix(src, dst); 2421 emit_int8(0x0F); 2422 emit_int8((unsigned char)0xB8); 2423 emit_operand(dst, src); 2424 } 2425 2426 void Assembler::popcntl(Register dst, Register src) { 2427 assert(VM_Version::supports_popcnt(), "must support"); 2428 emit_int8((unsigned char)0xF3); 2429 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2430 emit_int8(0x0F); 2431 emit_int8((unsigned char)0xB8); 2432 emit_int8((unsigned char)(0xC0 | encode)); 2433 } 2434 2435 void Assembler::popf() { 2436 emit_int8((unsigned char)0x9D); 2437 } 2438 2439 #ifndef _LP64 // no 32bit push/pop on amd64 2440 void Assembler::popl(Address dst) { 2441 // NOTE: this will adjust stack by 8byte on 64bits 2442 InstructionMark im(this); 2443 prefix(dst); 2444 emit_int8((unsigned char)0x8F); 2445 emit_operand(rax, dst); 2446 } 2447 #endif 2448 2449 void Assembler::prefetch_prefix(Address src) { 2450 prefix(src); 2451 emit_int8(0x0F); 2452 } 2453 2454 void Assembler::prefetchnta(Address src) { 2455 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2456 InstructionMark im(this); 2457 prefetch_prefix(src); 2458 emit_int8(0x18); 2459 emit_operand(rax, src); // 0, src 2460 } 2461 2462 void Assembler::prefetchr(Address src) { 2463 assert(VM_Version::supports_3dnow_prefetch(), "must support"); 2464 InstructionMark im(this); 2465 prefetch_prefix(src); 2466 emit_int8(0x0D); 2467 emit_operand(rax, src); // 0, src 2468 } 2469 2470 void Assembler::prefetcht0(Address src) { 2471 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2472 InstructionMark im(this); 2473 prefetch_prefix(src); 2474 emit_int8(0x18); 2475 emit_operand(rcx, src); // 1, src 2476 } 2477 2478 void Assembler::prefetcht1(Address src) { 2479 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2480 InstructionMark im(this); 2481 prefetch_prefix(src); 2482 emit_int8(0x18); 2483 emit_operand(rdx, src); // 2, src 2484 } 2485 2486 void Assembler::prefetcht2(Address src) { 2487 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2488 InstructionMark im(this); 2489 prefetch_prefix(src); 2490 emit_int8(0x18); 2491 emit_operand(rbx, src); // 3, src 2492 } 2493 2494 void Assembler::prefetchw(Address src) { 2495 assert(VM_Version::supports_3dnow_prefetch(), "must support"); 2496 InstructionMark im(this); 2497 prefetch_prefix(src); 2498 emit_int8(0x0D); 2499 emit_operand(rcx, src); // 1, src 2500 } 2501 2502 void Assembler::prefix(Prefix p) { 2503 emit_int8(p); 2504 } 2505 2506 void Assembler::pshufb(XMMRegister dst, XMMRegister src) { 2507 assert(VM_Version::supports_ssse3(), ""); 2508 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2509 emit_int8(0x00); 2510 emit_int8((unsigned char)(0xC0 | encode)); 2511 } 2512 2513 void Assembler::pshufb(XMMRegister dst, Address src) { 2514 assert(VM_Version::supports_ssse3(), ""); 2515 InstructionMark im(this); 2516 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2517 emit_int8(0x00); 2518 emit_operand(dst, src); 2519 } 2520 2521 void Assembler::pshufd(XMMRegister dst, XMMRegister src, int mode) { 2522 assert(isByte(mode), "invalid value"); 2523 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2524 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_66); 2525 emit_int8(mode & 0xFF); 2526 2527 } 2528 2529 void Assembler::pshufd(XMMRegister dst, Address src, int mode) { 2530 assert(isByte(mode), "invalid value"); 2531 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2532 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2533 InstructionMark im(this); 2534 simd_prefix(dst, src, VEX_SIMD_66); 2535 emit_int8(0x70); 2536 emit_operand(dst, src); 2537 emit_int8(mode & 0xFF); 2538 } 2539 2540 void Assembler::pshuflw(XMMRegister dst, XMMRegister src, int mode) { 2541 assert(isByte(mode), "invalid value"); 2542 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2543 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_F2); 2544 emit_int8(mode & 0xFF); 2545 } 2546 2547 void Assembler::pshuflw(XMMRegister dst, Address src, int mode) { 2548 assert(isByte(mode), "invalid value"); 2549 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2550 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2551 InstructionMark im(this); 2552 simd_prefix(dst, src, VEX_SIMD_F2); 2553 emit_int8(0x70); 2554 emit_operand(dst, src); 2555 emit_int8(mode & 0xFF); 2556 } 2557 2558 void Assembler::psrldq(XMMRegister dst, int shift) { 2559 // Shift 128 bit value in xmm register by number of bytes. 2560 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2561 int encode = simd_prefix_and_encode(xmm3, dst, dst, VEX_SIMD_66); 2562 emit_int8(0x73); 2563 emit_int8((unsigned char)(0xC0 | encode)); 2564 emit_int8(shift); 2565 } 2566 2567 void Assembler::ptest(XMMRegister dst, Address src) { 2568 assert(VM_Version::supports_sse4_1(), ""); 2569 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2570 InstructionMark im(this); 2571 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2572 emit_int8(0x17); 2573 emit_operand(dst, src); 2574 } 2575 2576 void Assembler::ptest(XMMRegister dst, XMMRegister src) { 2577 assert(VM_Version::supports_sse4_1(), ""); 2578 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2579 emit_int8(0x17); 2580 emit_int8((unsigned char)(0xC0 | encode)); 2581 } 2582 2583 void Assembler::vptest(XMMRegister dst, Address src) { 2584 assert(VM_Version::supports_avx(), ""); 2585 InstructionMark im(this); 2586 bool vector256 = true; 2587 assert(dst != xnoreg, "sanity"); 2588 int dst_enc = dst->encoding(); 2589 // swap src<->dst for encoding 2590 vex_prefix(src, 0, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256); 2591 emit_int8(0x17); 2592 emit_operand(dst, src); 2593 } 2594 2595 void Assembler::vptest(XMMRegister dst, XMMRegister src) { 2596 assert(VM_Version::supports_avx(), ""); 2597 bool vector256 = true; 2598 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 2599 emit_int8(0x17); 2600 emit_int8((unsigned char)(0xC0 | encode)); 2601 } 2602 2603 void Assembler::punpcklbw(XMMRegister dst, Address src) { 2604 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2605 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2606 emit_simd_arith(0x60, dst, src, VEX_SIMD_66); 2607 } 2608 2609 void Assembler::punpcklbw(XMMRegister dst, XMMRegister src) { 2610 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2611 emit_simd_arith(0x60, dst, src, VEX_SIMD_66); 2612 } 2613 2614 void Assembler::punpckldq(XMMRegister dst, Address src) { 2615 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2616 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2617 emit_simd_arith(0x62, dst, src, VEX_SIMD_66); 2618 } 2619 2620 void Assembler::punpckldq(XMMRegister dst, XMMRegister src) { 2621 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2622 emit_simd_arith(0x62, dst, src, VEX_SIMD_66); 2623 } 2624 2625 void Assembler::punpcklqdq(XMMRegister dst, XMMRegister src) { 2626 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2627 emit_simd_arith(0x6C, dst, src, VEX_SIMD_66); 2628 } 2629 2630 void Assembler::push(int32_t imm32) { 2631 // in 64bits we push 64bits onto the stack but only 2632 // take a 32bit immediate 2633 emit_int8(0x68); 2634 emit_int32(imm32); 2635 } 2636 2637 void Assembler::push(Register src) { 2638 int encode = prefix_and_encode(src->encoding()); 2639 2640 emit_int8(0x50 | encode); 2641 } 2642 2643 void Assembler::pushf() { 2644 emit_int8((unsigned char)0x9C); 2645 } 2646 2647 #ifndef _LP64 // no 32bit push/pop on amd64 2648 void Assembler::pushl(Address src) { 2649 // Note this will push 64bit on 64bit 2650 InstructionMark im(this); 2651 prefix(src); 2652 emit_int8((unsigned char)0xFF); 2653 emit_operand(rsi, src); 2654 } 2655 #endif 2656 2657 void Assembler::rcll(Register dst, int imm8) { 2658 assert(isShiftCount(imm8), "illegal shift count"); 2659 int encode = prefix_and_encode(dst->encoding()); 2660 if (imm8 == 1) { 2661 emit_int8((unsigned char)0xD1); 2662 emit_int8((unsigned char)(0xD0 | encode)); 2663 } else { 2664 emit_int8((unsigned char)0xC1); 2665 emit_int8((unsigned char)0xD0 | encode); 2666 emit_int8(imm8); 2667 } 2668 } 2669 2670 // copies data from [esi] to [edi] using rcx pointer sized words 2671 // generic 2672 void Assembler::rep_mov() { 2673 emit_int8((unsigned char)0xF3); 2674 // MOVSQ 2675 LP64_ONLY(prefix(REX_W)); 2676 emit_int8((unsigned char)0xA5); 2677 } 2678 2679 // sets rcx bytes with rax, value at [edi] 2680 void Assembler::rep_stosb() { 2681 emit_int8((unsigned char)0xF3); // REP 2682 LP64_ONLY(prefix(REX_W)); 2683 emit_int8((unsigned char)0xAA); // STOSB 2684 } 2685 2686 // sets rcx pointer sized words with rax, value at [edi] 2687 // generic 2688 void Assembler::rep_stos() { 2689 emit_int8((unsigned char)0xF3); // REP 2690 LP64_ONLY(prefix(REX_W)); // LP64:STOSQ, LP32:STOSD 2691 emit_int8((unsigned char)0xAB); 2692 } 2693 2694 // scans rcx pointer sized words at [edi] for occurance of rax, 2695 // generic 2696 void Assembler::repne_scan() { // repne_scan 2697 emit_int8((unsigned char)0xF2); 2698 // SCASQ 2699 LP64_ONLY(prefix(REX_W)); 2700 emit_int8((unsigned char)0xAF); 2701 } 2702 2703 #ifdef _LP64 2704 // scans rcx 4 byte words at [edi] for occurance of rax, 2705 // generic 2706 void Assembler::repne_scanl() { // repne_scan 2707 emit_int8((unsigned char)0xF2); 2708 // SCASL 2709 emit_int8((unsigned char)0xAF); 2710 } 2711 #endif 2712 2713 void Assembler::ret(int imm16) { 2714 if (imm16 == 0) { 2715 emit_int8((unsigned char)0xC3); 2716 } else { 2717 emit_int8((unsigned char)0xC2); 2718 emit_int16(imm16); 2719 } 2720 } 2721 2722 void Assembler::sahf() { 2723 #ifdef _LP64 2724 // Not supported in 64bit mode 2725 ShouldNotReachHere(); 2726 #endif 2727 emit_int8((unsigned char)0x9E); 2728 } 2729 2730 void Assembler::sarl(Register dst, int imm8) { 2731 int encode = prefix_and_encode(dst->encoding()); 2732 assert(isShiftCount(imm8), "illegal shift count"); 2733 if (imm8 == 1) { 2734 emit_int8((unsigned char)0xD1); 2735 emit_int8((unsigned char)(0xF8 | encode)); 2736 } else { 2737 emit_int8((unsigned char)0xC1); 2738 emit_int8((unsigned char)(0xF8 | encode)); 2739 emit_int8(imm8); 2740 } 2741 } 2742 2743 void Assembler::sarl(Register dst) { 2744 int encode = prefix_and_encode(dst->encoding()); 2745 emit_int8((unsigned char)0xD3); 2746 emit_int8((unsigned char)(0xF8 | encode)); 2747 } 2748 2749 void Assembler::sbbl(Address dst, int32_t imm32) { 2750 InstructionMark im(this); 2751 prefix(dst); 2752 emit_arith_operand(0x81, rbx, dst, imm32); 2753 } 2754 2755 void Assembler::sbbl(Register dst, int32_t imm32) { 2756 prefix(dst); 2757 emit_arith(0x81, 0xD8, dst, imm32); 2758 } 2759 2760 2761 void Assembler::sbbl(Register dst, Address src) { 2762 InstructionMark im(this); 2763 prefix(src, dst); 2764 emit_int8(0x1B); 2765 emit_operand(dst, src); 2766 } 2767 2768 void Assembler::sbbl(Register dst, Register src) { 2769 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2770 emit_arith(0x1B, 0xC0, dst, src); 2771 } 2772 2773 void Assembler::setb(Condition cc, Register dst) { 2774 assert(0 <= cc && cc < 16, "illegal cc"); 2775 int encode = prefix_and_encode(dst->encoding(), true); 2776 emit_int8(0x0F); 2777 emit_int8((unsigned char)0x90 | cc); 2778 emit_int8((unsigned char)(0xC0 | encode)); 2779 } 2780 2781 void Assembler::shll(Register dst, int imm8) { 2782 assert(isShiftCount(imm8), "illegal shift count"); 2783 int encode = prefix_and_encode(dst->encoding()); 2784 if (imm8 == 1 ) { 2785 emit_int8((unsigned char)0xD1); 2786 emit_int8((unsigned char)(0xE0 | encode)); 2787 } else { 2788 emit_int8((unsigned char)0xC1); 2789 emit_int8((unsigned char)(0xE0 | encode)); 2790 emit_int8(imm8); 2791 } 2792 } 2793 2794 void Assembler::shll(Register dst) { 2795 int encode = prefix_and_encode(dst->encoding()); 2796 emit_int8((unsigned char)0xD3); 2797 emit_int8((unsigned char)(0xE0 | encode)); 2798 } 2799 2800 void Assembler::shrl(Register dst, int imm8) { 2801 assert(isShiftCount(imm8), "illegal shift count"); 2802 int encode = prefix_and_encode(dst->encoding()); 2803 emit_int8((unsigned char)0xC1); 2804 emit_int8((unsigned char)(0xE8 | encode)); 2805 emit_int8(imm8); 2806 } 2807 2808 void Assembler::shrl(Register dst) { 2809 int encode = prefix_and_encode(dst->encoding()); 2810 emit_int8((unsigned char)0xD3); 2811 emit_int8((unsigned char)(0xE8 | encode)); 2812 } 2813 2814 // copies a single word from [esi] to [edi] 2815 void Assembler::smovl() { 2816 emit_int8((unsigned char)0xA5); 2817 } 2818 2819 void Assembler::sqrtsd(XMMRegister dst, XMMRegister src) { 2820 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2821 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2); 2822 } 2823 2824 void Assembler::sqrtsd(XMMRegister dst, Address src) { 2825 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2826 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2); 2827 } 2828 2829 void Assembler::sqrtss(XMMRegister dst, XMMRegister src) { 2830 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2831 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3); 2832 } 2833 2834 void Assembler::std() { 2835 emit_int8((unsigned char)0xFD); 2836 } 2837 2838 void Assembler::sqrtss(XMMRegister dst, Address src) { 2839 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2840 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3); 2841 } 2842 2843 void Assembler::stmxcsr( Address dst) { 2844 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2845 InstructionMark im(this); 2846 prefix(dst); 2847 emit_int8(0x0F); 2848 emit_int8((unsigned char)0xAE); 2849 emit_operand(as_Register(3), dst); 2850 } 2851 2852 void Assembler::subl(Address dst, int32_t imm32) { 2853 InstructionMark im(this); 2854 prefix(dst); 2855 emit_arith_operand(0x81, rbp, dst, imm32); 2856 } 2857 2858 void Assembler::subl(Address dst, Register src) { 2859 InstructionMark im(this); 2860 prefix(dst, src); 2861 emit_int8(0x29); 2862 emit_operand(src, dst); 2863 } 2864 2865 void Assembler::subl(Register dst, int32_t imm32) { 2866 prefix(dst); 2867 emit_arith(0x81, 0xE8, dst, imm32); 2868 } 2869 2870 // Force generation of a 4 byte immediate value even if it fits into 8bit 2871 void Assembler::subl_imm32(Register dst, int32_t imm32) { 2872 prefix(dst); 2873 emit_arith_imm32(0x81, 0xE8, dst, imm32); 2874 } 2875 2876 void Assembler::subl(Register dst, Address src) { 2877 InstructionMark im(this); 2878 prefix(src, dst); 2879 emit_int8(0x2B); 2880 emit_operand(dst, src); 2881 } 2882 2883 void Assembler::subl(Register dst, Register src) { 2884 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2885 emit_arith(0x2B, 0xC0, dst, src); 2886 } 2887 2888 void Assembler::subsd(XMMRegister dst, XMMRegister src) { 2889 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2890 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2); 2891 } 2892 2893 void Assembler::subsd(XMMRegister dst, Address src) { 2894 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2895 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2); 2896 } 2897 2898 void Assembler::subss(XMMRegister dst, XMMRegister src) { 2899 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2900 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3); 2901 } 2902 2903 void Assembler::subss(XMMRegister dst, Address src) { 2904 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2905 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3); 2906 } 2907 2908 void Assembler::testb(Register dst, int imm8) { 2909 NOT_LP64(assert(dst->has_byte_register(), "must have byte register")); 2910 (void) prefix_and_encode(dst->encoding(), true); 2911 emit_arith_b(0xF6, 0xC0, dst, imm8); 2912 } 2913 2914 void Assembler::testl(Register dst, int32_t imm32) { 2915 // not using emit_arith because test 2916 // doesn't support sign-extension of 2917 // 8bit operands 2918 int encode = dst->encoding(); 2919 if (encode == 0) { 2920 emit_int8((unsigned char)0xA9); 2921 } else { 2922 encode = prefix_and_encode(encode); 2923 emit_int8((unsigned char)0xF7); 2924 emit_int8((unsigned char)(0xC0 | encode)); 2925 } 2926 emit_int32(imm32); 2927 } 2928 2929 void Assembler::testl(Register dst, Register src) { 2930 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2931 emit_arith(0x85, 0xC0, dst, src); 2932 } 2933 2934 void Assembler::testl(Register dst, Address src) { 2935 InstructionMark im(this); 2936 prefix(src, dst); 2937 emit_int8((unsigned char)0x85); 2938 emit_operand(dst, src); 2939 } 2940 2941 void Assembler::tzcntl(Register dst, Register src) { 2942 assert(VM_Version::supports_bmi1(), "tzcnt instruction not supported"); 2943 emit_int8((unsigned char)0xF3); 2944 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2945 emit_int8(0x0F); 2946 emit_int8((unsigned char)0xBC); 2947 emit_int8((unsigned char)0xC0 | encode); 2948 } 2949 2950 void Assembler::tzcntq(Register dst, Register src) { 2951 assert(VM_Version::supports_bmi1(), "tzcnt instruction not supported"); 2952 emit_int8((unsigned char)0xF3); 2953 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 2954 emit_int8(0x0F); 2955 emit_int8((unsigned char)0xBC); 2956 emit_int8((unsigned char)(0xC0 | encode)); 2957 } 2958 2959 void Assembler::ucomisd(XMMRegister dst, Address src) { 2960 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2961 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66); 2962 } 2963 2964 void Assembler::ucomisd(XMMRegister dst, XMMRegister src) { 2965 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2966 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66); 2967 } 2968 2969 void Assembler::ucomiss(XMMRegister dst, Address src) { 2970 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2971 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE); 2972 } 2973 2974 void Assembler::ucomiss(XMMRegister dst, XMMRegister src) { 2975 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2976 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE); 2977 } 2978 2979 2980 void Assembler::xaddl(Address dst, Register src) { 2981 InstructionMark im(this); 2982 prefix(dst, src); 2983 emit_int8(0x0F); 2984 emit_int8((unsigned char)0xC1); 2985 emit_operand(src, dst); 2986 } 2987 2988 void Assembler::xchgl(Register dst, Address src) { // xchg 2989 InstructionMark im(this); 2990 prefix(src, dst); 2991 emit_int8((unsigned char)0x87); 2992 emit_operand(dst, src); 2993 } 2994 2995 void Assembler::xchgl(Register dst, Register src) { 2996 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2997 emit_int8((unsigned char)0x87); 2998 emit_int8((unsigned char)(0xC0 | encode)); 2999 } 3000 3001 void Assembler::xgetbv() { 3002 emit_int8(0x0F); 3003 emit_int8(0x01); 3004 emit_int8((unsigned char)0xD0); 3005 } 3006 3007 void Assembler::xorl(Register dst, int32_t imm32) { 3008 prefix(dst); 3009 emit_arith(0x81, 0xF0, dst, imm32); 3010 } 3011 3012 void Assembler::xorl(Register dst, Address src) { 3013 InstructionMark im(this); 3014 prefix(src, dst); 3015 emit_int8(0x33); 3016 emit_operand(dst, src); 3017 } 3018 3019 void Assembler::xorl(Register dst, Register src) { 3020 (void) prefix_and_encode(dst->encoding(), src->encoding()); 3021 emit_arith(0x33, 0xC0, dst, src); 3022 } 3023 3024 3025 // AVX 3-operands scalar float-point arithmetic instructions 3026 3027 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, Address src) { 3028 assert(VM_Version::supports_avx(), ""); 3029 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3030 } 3031 3032 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3033 assert(VM_Version::supports_avx(), ""); 3034 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3035 } 3036 3037 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, Address src) { 3038 assert(VM_Version::supports_avx(), ""); 3039 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3040 } 3041 3042 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3043 assert(VM_Version::supports_avx(), ""); 3044 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3045 } 3046 3047 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, Address src) { 3048 assert(VM_Version::supports_avx(), ""); 3049 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3050 } 3051 3052 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3053 assert(VM_Version::supports_avx(), ""); 3054 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3055 } 3056 3057 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, Address src) { 3058 assert(VM_Version::supports_avx(), ""); 3059 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3060 } 3061 3062 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3063 assert(VM_Version::supports_avx(), ""); 3064 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3065 } 3066 3067 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, Address src) { 3068 assert(VM_Version::supports_avx(), ""); 3069 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3070 } 3071 3072 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3073 assert(VM_Version::supports_avx(), ""); 3074 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3075 } 3076 3077 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, Address src) { 3078 assert(VM_Version::supports_avx(), ""); 3079 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3080 } 3081 3082 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3083 assert(VM_Version::supports_avx(), ""); 3084 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3085 } 3086 3087 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, Address src) { 3088 assert(VM_Version::supports_avx(), ""); 3089 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3090 } 3091 3092 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3093 assert(VM_Version::supports_avx(), ""); 3094 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3095 } 3096 3097 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, Address src) { 3098 assert(VM_Version::supports_avx(), ""); 3099 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3100 } 3101 3102 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3103 assert(VM_Version::supports_avx(), ""); 3104 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3105 } 3106 3107 //====================VECTOR ARITHMETIC===================================== 3108 3109 // Float-point vector arithmetic 3110 3111 void Assembler::addpd(XMMRegister dst, XMMRegister src) { 3112 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3113 emit_simd_arith(0x58, dst, src, VEX_SIMD_66); 3114 } 3115 3116 void Assembler::addps(XMMRegister dst, XMMRegister src) { 3117 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3118 emit_simd_arith(0x58, dst, src, VEX_SIMD_NONE); 3119 } 3120 3121 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3122 assert(VM_Version::supports_avx(), ""); 3123 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256); 3124 } 3125 3126 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3127 assert(VM_Version::supports_avx(), ""); 3128 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256); 3129 } 3130 3131 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3132 assert(VM_Version::supports_avx(), ""); 3133 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256); 3134 } 3135 3136 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3137 assert(VM_Version::supports_avx(), ""); 3138 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256); 3139 } 3140 3141 void Assembler::subpd(XMMRegister dst, XMMRegister src) { 3142 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3143 emit_simd_arith(0x5C, dst, src, VEX_SIMD_66); 3144 } 3145 3146 void Assembler::subps(XMMRegister dst, XMMRegister src) { 3147 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3148 emit_simd_arith(0x5C, dst, src, VEX_SIMD_NONE); 3149 } 3150 3151 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3152 assert(VM_Version::supports_avx(), ""); 3153 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256); 3154 } 3155 3156 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3157 assert(VM_Version::supports_avx(), ""); 3158 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256); 3159 } 3160 3161 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3162 assert(VM_Version::supports_avx(), ""); 3163 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256); 3164 } 3165 3166 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3167 assert(VM_Version::supports_avx(), ""); 3168 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256); 3169 } 3170 3171 void Assembler::mulpd(XMMRegister dst, XMMRegister src) { 3172 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3173 emit_simd_arith(0x59, dst, src, VEX_SIMD_66); 3174 } 3175 3176 void Assembler::mulps(XMMRegister dst, XMMRegister src) { 3177 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3178 emit_simd_arith(0x59, dst, src, VEX_SIMD_NONE); 3179 } 3180 3181 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3182 assert(VM_Version::supports_avx(), ""); 3183 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256); 3184 } 3185 3186 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3187 assert(VM_Version::supports_avx(), ""); 3188 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256); 3189 } 3190 3191 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3192 assert(VM_Version::supports_avx(), ""); 3193 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256); 3194 } 3195 3196 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3197 assert(VM_Version::supports_avx(), ""); 3198 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256); 3199 } 3200 3201 void Assembler::divpd(XMMRegister dst, XMMRegister src) { 3202 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3203 emit_simd_arith(0x5E, dst, src, VEX_SIMD_66); 3204 } 3205 3206 void Assembler::divps(XMMRegister dst, XMMRegister src) { 3207 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3208 emit_simd_arith(0x5E, dst, src, VEX_SIMD_NONE); 3209 } 3210 3211 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3212 assert(VM_Version::supports_avx(), ""); 3213 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256); 3214 } 3215 3216 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3217 assert(VM_Version::supports_avx(), ""); 3218 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256); 3219 } 3220 3221 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3222 assert(VM_Version::supports_avx(), ""); 3223 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256); 3224 } 3225 3226 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3227 assert(VM_Version::supports_avx(), ""); 3228 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256); 3229 } 3230 3231 void Assembler::andpd(XMMRegister dst, XMMRegister src) { 3232 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3233 emit_simd_arith(0x54, dst, src, VEX_SIMD_66); 3234 } 3235 3236 void Assembler::andps(XMMRegister dst, XMMRegister src) { 3237 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3238 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE); 3239 } 3240 3241 void Assembler::andps(XMMRegister dst, Address src) { 3242 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3243 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE); 3244 } 3245 3246 void Assembler::andpd(XMMRegister dst, Address src) { 3247 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3248 emit_simd_arith(0x54, dst, src, VEX_SIMD_66); 3249 } 3250 3251 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3252 assert(VM_Version::supports_avx(), ""); 3253 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256); 3254 } 3255 3256 void Assembler::vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3257 assert(VM_Version::supports_avx(), ""); 3258 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256); 3259 } 3260 3261 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3262 assert(VM_Version::supports_avx(), ""); 3263 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256); 3264 } 3265 3266 void Assembler::vandps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3267 assert(VM_Version::supports_avx(), ""); 3268 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256); 3269 } 3270 3271 void Assembler::xorpd(XMMRegister dst, XMMRegister src) { 3272 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3273 emit_simd_arith(0x57, dst, src, VEX_SIMD_66); 3274 } 3275 3276 void Assembler::xorps(XMMRegister dst, XMMRegister src) { 3277 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3278 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE); 3279 } 3280 3281 void Assembler::xorpd(XMMRegister dst, Address src) { 3282 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3283 emit_simd_arith(0x57, dst, src, VEX_SIMD_66); 3284 } 3285 3286 void Assembler::xorps(XMMRegister dst, Address src) { 3287 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3288 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE); 3289 } 3290 3291 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3292 assert(VM_Version::supports_avx(), ""); 3293 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256); 3294 } 3295 3296 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3297 assert(VM_Version::supports_avx(), ""); 3298 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256); 3299 } 3300 3301 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3302 assert(VM_Version::supports_avx(), ""); 3303 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256); 3304 } 3305 3306 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3307 assert(VM_Version::supports_avx(), ""); 3308 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256); 3309 } 3310 3311 3312 // Integer vector arithmetic 3313 void Assembler::paddb(XMMRegister dst, XMMRegister src) { 3314 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3315 emit_simd_arith(0xFC, dst, src, VEX_SIMD_66); 3316 } 3317 3318 void Assembler::paddw(XMMRegister dst, XMMRegister src) { 3319 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3320 emit_simd_arith(0xFD, dst, src, VEX_SIMD_66); 3321 } 3322 3323 void Assembler::paddd(XMMRegister dst, XMMRegister src) { 3324 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3325 emit_simd_arith(0xFE, dst, src, VEX_SIMD_66); 3326 } 3327 3328 void Assembler::paddq(XMMRegister dst, XMMRegister src) { 3329 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3330 emit_simd_arith(0xD4, dst, src, VEX_SIMD_66); 3331 } 3332 3333 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3334 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3335 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256); 3336 } 3337 3338 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3339 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3340 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256); 3341 } 3342 3343 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3344 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3345 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256); 3346 } 3347 3348 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3349 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3350 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256); 3351 } 3352 3353 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3354 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3355 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256); 3356 } 3357 3358 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3359 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3360 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256); 3361 } 3362 3363 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3364 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3365 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256); 3366 } 3367 3368 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3369 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3370 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256); 3371 } 3372 3373 void Assembler::psubb(XMMRegister dst, XMMRegister src) { 3374 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3375 emit_simd_arith(0xF8, dst, src, VEX_SIMD_66); 3376 } 3377 3378 void Assembler::psubw(XMMRegister dst, XMMRegister src) { 3379 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3380 emit_simd_arith(0xF9, dst, src, VEX_SIMD_66); 3381 } 3382 3383 void Assembler::psubd(XMMRegister dst, XMMRegister src) { 3384 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3385 emit_simd_arith(0xFA, dst, src, VEX_SIMD_66); 3386 } 3387 3388 void Assembler::psubq(XMMRegister dst, XMMRegister src) { 3389 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3390 emit_simd_arith(0xFB, dst, src, VEX_SIMD_66); 3391 } 3392 3393 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3394 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3395 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256); 3396 } 3397 3398 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3399 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3400 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256); 3401 } 3402 3403 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3404 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3405 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256); 3406 } 3407 3408 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3409 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3410 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256); 3411 } 3412 3413 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3414 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3415 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256); 3416 } 3417 3418 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3419 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3420 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256); 3421 } 3422 3423 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3424 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3425 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256); 3426 } 3427 3428 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3429 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3430 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256); 3431 } 3432 3433 void Assembler::pmullw(XMMRegister dst, XMMRegister src) { 3434 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3435 emit_simd_arith(0xD5, dst, src, VEX_SIMD_66); 3436 } 3437 3438 void Assembler::pmulld(XMMRegister dst, XMMRegister src) { 3439 assert(VM_Version::supports_sse4_1(), ""); 3440 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 3441 emit_int8(0x40); 3442 emit_int8((unsigned char)(0xC0 | encode)); 3443 } 3444 3445 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3446 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3447 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256); 3448 } 3449 3450 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3451 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3452 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 3453 emit_int8(0x40); 3454 emit_int8((unsigned char)(0xC0 | encode)); 3455 } 3456 3457 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3458 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3459 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256); 3460 } 3461 3462 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3463 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3464 InstructionMark im(this); 3465 int dst_enc = dst->encoding(); 3466 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 3467 vex_prefix(src, nds_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256); 3468 emit_int8(0x40); 3469 emit_operand(dst, src); 3470 } 3471 3472 // Shift packed integers left by specified number of bits. 3473 void Assembler::psllw(XMMRegister dst, int shift) { 3474 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3475 // XMM6 is for /6 encoding: 66 0F 71 /6 ib 3476 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3477 emit_int8(0x71); 3478 emit_int8((unsigned char)(0xC0 | encode)); 3479 emit_int8(shift & 0xFF); 3480 } 3481 3482 void Assembler::pslld(XMMRegister dst, int shift) { 3483 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3484 // XMM6 is for /6 encoding: 66 0F 72 /6 ib 3485 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3486 emit_int8(0x72); 3487 emit_int8((unsigned char)(0xC0 | encode)); 3488 emit_int8(shift & 0xFF); 3489 } 3490 3491 void Assembler::psllq(XMMRegister dst, int shift) { 3492 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3493 // XMM6 is for /6 encoding: 66 0F 73 /6 ib 3494 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3495 emit_int8(0x73); 3496 emit_int8((unsigned char)(0xC0 | encode)); 3497 emit_int8(shift & 0xFF); 3498 } 3499 3500 void Assembler::psllw(XMMRegister dst, XMMRegister shift) { 3501 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3502 emit_simd_arith(0xF1, dst, shift, VEX_SIMD_66); 3503 } 3504 3505 void Assembler::pslld(XMMRegister dst, XMMRegister shift) { 3506 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3507 emit_simd_arith(0xF2, dst, shift, VEX_SIMD_66); 3508 } 3509 3510 void Assembler::psllq(XMMRegister dst, XMMRegister shift) { 3511 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3512 emit_simd_arith(0xF3, dst, shift, VEX_SIMD_66); 3513 } 3514 3515 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3516 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3517 // XMM6 is for /6 encoding: 66 0F 71 /6 ib 3518 emit_vex_arith(0x71, xmm6, dst, src, VEX_SIMD_66, vector256); 3519 emit_int8(shift & 0xFF); 3520 } 3521 3522 void Assembler::vpslld(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3523 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3524 // XMM6 is for /6 encoding: 66 0F 72 /6 ib 3525 emit_vex_arith(0x72, xmm6, dst, src, VEX_SIMD_66, vector256); 3526 emit_int8(shift & 0xFF); 3527 } 3528 3529 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3530 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3531 // XMM6 is for /6 encoding: 66 0F 73 /6 ib 3532 emit_vex_arith(0x73, xmm6, dst, src, VEX_SIMD_66, vector256); 3533 emit_int8(shift & 0xFF); 3534 } 3535 3536 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3537 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3538 emit_vex_arith(0xF1, dst, src, shift, VEX_SIMD_66, vector256); 3539 } 3540 3541 void Assembler::vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3542 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3543 emit_vex_arith(0xF2, dst, src, shift, VEX_SIMD_66, vector256); 3544 } 3545 3546 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3547 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3548 emit_vex_arith(0xF3, dst, src, shift, VEX_SIMD_66, vector256); 3549 } 3550 3551 // Shift packed integers logically right by specified number of bits. 3552 void Assembler::psrlw(XMMRegister dst, int shift) { 3553 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3554 // XMM2 is for /2 encoding: 66 0F 71 /2 ib 3555 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3556 emit_int8(0x71); 3557 emit_int8((unsigned char)(0xC0 | encode)); 3558 emit_int8(shift & 0xFF); 3559 } 3560 3561 void Assembler::psrld(XMMRegister dst, int shift) { 3562 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3563 // XMM2 is for /2 encoding: 66 0F 72 /2 ib 3564 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3565 emit_int8(0x72); 3566 emit_int8((unsigned char)(0xC0 | encode)); 3567 emit_int8(shift & 0xFF); 3568 } 3569 3570 void Assembler::psrlq(XMMRegister dst, int shift) { 3571 // Do not confuse it with psrldq SSE2 instruction which 3572 // shifts 128 bit value in xmm register by number of bytes. 3573 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3574 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3575 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3576 emit_int8(0x73); 3577 emit_int8((unsigned char)(0xC0 | encode)); 3578 emit_int8(shift & 0xFF); 3579 } 3580 3581 void Assembler::psrlw(XMMRegister dst, XMMRegister shift) { 3582 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3583 emit_simd_arith(0xD1, dst, shift, VEX_SIMD_66); 3584 } 3585 3586 void Assembler::psrld(XMMRegister dst, XMMRegister shift) { 3587 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3588 emit_simd_arith(0xD2, dst, shift, VEX_SIMD_66); 3589 } 3590 3591 void Assembler::psrlq(XMMRegister dst, XMMRegister shift) { 3592 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3593 emit_simd_arith(0xD3, dst, shift, VEX_SIMD_66); 3594 } 3595 3596 void Assembler::vpsrlw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3597 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3598 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3599 emit_vex_arith(0x71, xmm2, dst, src, VEX_SIMD_66, vector256); 3600 emit_int8(shift & 0xFF); 3601 } 3602 3603 void Assembler::vpsrld(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3604 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3605 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3606 emit_vex_arith(0x72, xmm2, dst, src, VEX_SIMD_66, vector256); 3607 emit_int8(shift & 0xFF); 3608 } 3609 3610 void Assembler::vpsrlq(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3611 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3612 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3613 emit_vex_arith(0x73, xmm2, dst, src, VEX_SIMD_66, vector256); 3614 emit_int8(shift & 0xFF); 3615 } 3616 3617 void Assembler::vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3618 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3619 emit_vex_arith(0xD1, dst, src, shift, VEX_SIMD_66, vector256); 3620 } 3621 3622 void Assembler::vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3623 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3624 emit_vex_arith(0xD2, dst, src, shift, VEX_SIMD_66, vector256); 3625 } 3626 3627 void Assembler::vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3628 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3629 emit_vex_arith(0xD3, dst, src, shift, VEX_SIMD_66, vector256); 3630 } 3631 3632 // Shift packed integers arithmetically right by specified number of bits. 3633 void Assembler::psraw(XMMRegister dst, int shift) { 3634 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3635 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3636 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66); 3637 emit_int8(0x71); 3638 emit_int8((unsigned char)(0xC0 | encode)); 3639 emit_int8(shift & 0xFF); 3640 } 3641 3642 void Assembler::psrad(XMMRegister dst, int shift) { 3643 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3644 // XMM4 is for /4 encoding: 66 0F 72 /4 ib 3645 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66); 3646 emit_int8(0x72); 3647 emit_int8((unsigned char)(0xC0 | encode)); 3648 emit_int8(shift & 0xFF); 3649 } 3650 3651 void Assembler::psraw(XMMRegister dst, XMMRegister shift) { 3652 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3653 emit_simd_arith(0xE1, dst, shift, VEX_SIMD_66); 3654 } 3655 3656 void Assembler::psrad(XMMRegister dst, XMMRegister shift) { 3657 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3658 emit_simd_arith(0xE2, dst, shift, VEX_SIMD_66); 3659 } 3660 3661 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3662 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3663 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3664 emit_vex_arith(0x71, xmm4, dst, src, VEX_SIMD_66, vector256); 3665 emit_int8(shift & 0xFF); 3666 } 3667 3668 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3669 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3670 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3671 emit_vex_arith(0x72, xmm4, dst, src, VEX_SIMD_66, vector256); 3672 emit_int8(shift & 0xFF); 3673 } 3674 3675 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3676 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3677 emit_vex_arith(0xE1, dst, src, shift, VEX_SIMD_66, vector256); 3678 } 3679 3680 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3681 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3682 emit_vex_arith(0xE2, dst, src, shift, VEX_SIMD_66, vector256); 3683 } 3684 3685 3686 // AND packed integers 3687 void Assembler::pand(XMMRegister dst, XMMRegister src) { 3688 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3689 emit_simd_arith(0xDB, dst, src, VEX_SIMD_66); 3690 } 3691 3692 void Assembler::vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3693 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3694 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256); 3695 } 3696 3697 void Assembler::vpand(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3698 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3699 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256); 3700 } 3701 3702 void Assembler::por(XMMRegister dst, XMMRegister src) { 3703 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3704 emit_simd_arith(0xEB, dst, src, VEX_SIMD_66); 3705 } 3706 3707 void Assembler::vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3708 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3709 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256); 3710 } 3711 3712 void Assembler::vpor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3713 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3714 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256); 3715 } 3716 3717 void Assembler::pxor(XMMRegister dst, XMMRegister src) { 3718 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3719 emit_simd_arith(0xEF, dst, src, VEX_SIMD_66); 3720 } 3721 3722 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3723 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3724 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256); 3725 } 3726 3727 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3728 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3729 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256); 3730 } 3731 3732 3733 void Assembler::vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3734 assert(VM_Version::supports_avx(), ""); 3735 bool vector256 = true; 3736 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3737 emit_int8(0x18); 3738 emit_int8((unsigned char)(0xC0 | encode)); 3739 // 0x00 - insert into lower 128 bits 3740 // 0x01 - insert into upper 128 bits 3741 emit_int8(0x01); 3742 } 3743 3744 void Assembler::vinsertf128h(XMMRegister dst, Address src) { 3745 assert(VM_Version::supports_avx(), ""); 3746 InstructionMark im(this); 3747 bool vector256 = true; 3748 assert(dst != xnoreg, "sanity"); 3749 int dst_enc = dst->encoding(); 3750 // swap src<->dst for encoding 3751 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3752 emit_int8(0x18); 3753 emit_operand(dst, src); 3754 // 0x01 - insert into upper 128 bits 3755 emit_int8(0x01); 3756 } 3757 3758 void Assembler::vextractf128h(Address dst, XMMRegister src) { 3759 assert(VM_Version::supports_avx(), ""); 3760 InstructionMark im(this); 3761 bool vector256 = true; 3762 assert(src != xnoreg, "sanity"); 3763 int src_enc = src->encoding(); 3764 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3765 emit_int8(0x19); 3766 emit_operand(src, dst); 3767 // 0x01 - extract from upper 128 bits 3768 emit_int8(0x01); 3769 } 3770 3771 void Assembler::vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3772 assert(VM_Version::supports_avx2(), ""); 3773 bool vector256 = true; 3774 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3775 emit_int8(0x38); 3776 emit_int8((unsigned char)(0xC0 | encode)); 3777 // 0x00 - insert into lower 128 bits 3778 // 0x01 - insert into upper 128 bits 3779 emit_int8(0x01); 3780 } 3781 3782 void Assembler::vinserti128h(XMMRegister dst, Address src) { 3783 assert(VM_Version::supports_avx2(), ""); 3784 InstructionMark im(this); 3785 bool vector256 = true; 3786 assert(dst != xnoreg, "sanity"); 3787 int dst_enc = dst->encoding(); 3788 // swap src<->dst for encoding 3789 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3790 emit_int8(0x38); 3791 emit_operand(dst, src); 3792 // 0x01 - insert into upper 128 bits 3793 emit_int8(0x01); 3794 } 3795 3796 void Assembler::vextracti128h(Address dst, XMMRegister src) { 3797 assert(VM_Version::supports_avx2(), ""); 3798 InstructionMark im(this); 3799 bool vector256 = true; 3800 assert(src != xnoreg, "sanity"); 3801 int src_enc = src->encoding(); 3802 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3803 emit_int8(0x39); 3804 emit_operand(src, dst); 3805 // 0x01 - extract from upper 128 bits 3806 emit_int8(0x01); 3807 } 3808 3809 // duplicate 4-bytes integer data from src into 8 locations in dest 3810 void Assembler::vpbroadcastd(XMMRegister dst, XMMRegister src) { 3811 assert(VM_Version::supports_avx2(), ""); 3812 bool vector256 = true; 3813 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 3814 emit_int8(0x58); 3815 emit_int8((unsigned char)(0xC0 | encode)); 3816 } 3817 3818 // Carry-Less Multiplication Quadword 3819 void Assembler::vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask) { 3820 assert(VM_Version::supports_avx() && VM_Version::supports_clmul(), ""); 3821 bool vector256 = false; 3822 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3823 emit_int8(0x44); 3824 emit_int8((unsigned char)(0xC0 | encode)); 3825 emit_int8((unsigned char)mask); 3826 } 3827 3828 void Assembler::vzeroupper() { 3829 assert(VM_Version::supports_avx(), ""); 3830 (void)vex_prefix_and_encode(xmm0, xmm0, xmm0, VEX_SIMD_NONE); 3831 emit_int8(0x77); 3832 } 3833 3834 3835 #ifndef _LP64 3836 // 32bit only pieces of the assembler 3837 3838 void Assembler::cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec) { 3839 // NO PREFIX AS NEVER 64BIT 3840 InstructionMark im(this); 3841 emit_int8((unsigned char)0x81); 3842 emit_int8((unsigned char)(0xF8 | src1->encoding())); 3843 emit_data(imm32, rspec, 0); 3844 } 3845 3846 void Assembler::cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec) { 3847 // NO PREFIX AS NEVER 64BIT (not even 32bit versions of 64bit regs 3848 InstructionMark im(this); 3849 emit_int8((unsigned char)0x81); 3850 emit_operand(rdi, src1); 3851 emit_data(imm32, rspec, 0); 3852 } 3853 3854 // The 64-bit (32bit platform) cmpxchg compares the value at adr with the contents of rdx:rax, 3855 // and stores rcx:rbx into adr if so; otherwise, the value at adr is loaded 3856 // into rdx:rax. The ZF is set if the compared values were equal, and cleared otherwise. 3857 void Assembler::cmpxchg8(Address adr) { 3858 InstructionMark im(this); 3859 emit_int8(0x0F); 3860 emit_int8((unsigned char)0xC7); 3861 emit_operand(rcx, adr); 3862 } 3863 3864 void Assembler::decl(Register dst) { 3865 // Don't use it directly. Use MacroAssembler::decrementl() instead. 3866 emit_int8(0x48 | dst->encoding()); 3867 } 3868 3869 #endif // _LP64 3870 3871 // 64bit typically doesn't use the x87 but needs to for the trig funcs 3872 3873 void Assembler::fabs() { 3874 emit_int8((unsigned char)0xD9); 3875 emit_int8((unsigned char)0xE1); 3876 } 3877 3878 void Assembler::fadd(int i) { 3879 emit_farith(0xD8, 0xC0, i); 3880 } 3881 3882 void Assembler::fadd_d(Address src) { 3883 InstructionMark im(this); 3884 emit_int8((unsigned char)0xDC); 3885 emit_operand32(rax, src); 3886 } 3887 3888 void Assembler::fadd_s(Address src) { 3889 InstructionMark im(this); 3890 emit_int8((unsigned char)0xD8); 3891 emit_operand32(rax, src); 3892 } 3893 3894 void Assembler::fadda(int i) { 3895 emit_farith(0xDC, 0xC0, i); 3896 } 3897 3898 void Assembler::faddp(int i) { 3899 emit_farith(0xDE, 0xC0, i); 3900 } 3901 3902 void Assembler::fchs() { 3903 emit_int8((unsigned char)0xD9); 3904 emit_int8((unsigned char)0xE0); 3905 } 3906 3907 void Assembler::fcom(int i) { 3908 emit_farith(0xD8, 0xD0, i); 3909 } 3910 3911 void Assembler::fcomp(int i) { 3912 emit_farith(0xD8, 0xD8, i); 3913 } 3914 3915 void Assembler::fcomp_d(Address src) { 3916 InstructionMark im(this); 3917 emit_int8((unsigned char)0xDC); 3918 emit_operand32(rbx, src); 3919 } 3920 3921 void Assembler::fcomp_s(Address src) { 3922 InstructionMark im(this); 3923 emit_int8((unsigned char)0xD8); 3924 emit_operand32(rbx, src); 3925 } 3926 3927 void Assembler::fcompp() { 3928 emit_int8((unsigned char)0xDE); 3929 emit_int8((unsigned char)0xD9); 3930 } 3931 3932 void Assembler::fcos() { 3933 emit_int8((unsigned char)0xD9); 3934 emit_int8((unsigned char)0xFF); 3935 } 3936 3937 void Assembler::fdecstp() { 3938 emit_int8((unsigned char)0xD9); 3939 emit_int8((unsigned char)0xF6); 3940 } 3941 3942 void Assembler::fdiv(int i) { 3943 emit_farith(0xD8, 0xF0, i); 3944 } 3945 3946 void Assembler::fdiv_d(Address src) { 3947 InstructionMark im(this); 3948 emit_int8((unsigned char)0xDC); 3949 emit_operand32(rsi, src); 3950 } 3951 3952 void Assembler::fdiv_s(Address src) { 3953 InstructionMark im(this); 3954 emit_int8((unsigned char)0xD8); 3955 emit_operand32(rsi, src); 3956 } 3957 3958 void Assembler::fdiva(int i) { 3959 emit_farith(0xDC, 0xF8, i); 3960 } 3961 3962 // Note: The Intel manual (Pentium Processor User's Manual, Vol.3, 1994) 3963 // is erroneous for some of the floating-point instructions below. 3964 3965 void Assembler::fdivp(int i) { 3966 emit_farith(0xDE, 0xF8, i); // ST(0) <- ST(0) / ST(1) and pop (Intel manual wrong) 3967 } 3968 3969 void Assembler::fdivr(int i) { 3970 emit_farith(0xD8, 0xF8, i); 3971 } 3972 3973 void Assembler::fdivr_d(Address src) { 3974 InstructionMark im(this); 3975 emit_int8((unsigned char)0xDC); 3976 emit_operand32(rdi, src); 3977 } 3978 3979 void Assembler::fdivr_s(Address src) { 3980 InstructionMark im(this); 3981 emit_int8((unsigned char)0xD8); 3982 emit_operand32(rdi, src); 3983 } 3984 3985 void Assembler::fdivra(int i) { 3986 emit_farith(0xDC, 0xF0, i); 3987 } 3988 3989 void Assembler::fdivrp(int i) { 3990 emit_farith(0xDE, 0xF0, i); // ST(0) <- ST(1) / ST(0) and pop (Intel manual wrong) 3991 } 3992 3993 void Assembler::ffree(int i) { 3994 emit_farith(0xDD, 0xC0, i); 3995 } 3996 3997 void Assembler::fild_d(Address adr) { 3998 InstructionMark im(this); 3999 emit_int8((unsigned char)0xDF); 4000 emit_operand32(rbp, adr); 4001 } 4002 4003 void Assembler::fild_s(Address adr) { 4004 InstructionMark im(this); 4005 emit_int8((unsigned char)0xDB); 4006 emit_operand32(rax, adr); 4007 } 4008 4009 void Assembler::fincstp() { 4010 emit_int8((unsigned char)0xD9); 4011 emit_int8((unsigned char)0xF7); 4012 } 4013 4014 void Assembler::finit() { 4015 emit_int8((unsigned char)0x9B); 4016 emit_int8((unsigned char)0xDB); 4017 emit_int8((unsigned char)0xE3); 4018 } 4019 4020 void Assembler::fist_s(Address adr) { 4021 InstructionMark im(this); 4022 emit_int8((unsigned char)0xDB); 4023 emit_operand32(rdx, adr); 4024 } 4025 4026 void Assembler::fistp_d(Address adr) { 4027 InstructionMark im(this); 4028 emit_int8((unsigned char)0xDF); 4029 emit_operand32(rdi, adr); 4030 } 4031 4032 void Assembler::fistp_s(Address adr) { 4033 InstructionMark im(this); 4034 emit_int8((unsigned char)0xDB); 4035 emit_operand32(rbx, adr); 4036 } 4037 4038 void Assembler::fld1() { 4039 emit_int8((unsigned char)0xD9); 4040 emit_int8((unsigned char)0xE8); 4041 } 4042 4043 void Assembler::fld_d(Address adr) { 4044 InstructionMark im(this); 4045 emit_int8((unsigned char)0xDD); 4046 emit_operand32(rax, adr); 4047 } 4048 4049 void Assembler::fld_s(Address adr) { 4050 InstructionMark im(this); 4051 emit_int8((unsigned char)0xD9); 4052 emit_operand32(rax, adr); 4053 } 4054 4055 4056 void Assembler::fld_s(int index) { 4057 emit_farith(0xD9, 0xC0, index); 4058 } 4059 4060 void Assembler::fld_x(Address adr) { 4061 InstructionMark im(this); 4062 emit_int8((unsigned char)0xDB); 4063 emit_operand32(rbp, adr); 4064 } 4065 4066 void Assembler::fldcw(Address src) { 4067 InstructionMark im(this); 4068 emit_int8((unsigned char)0xD9); 4069 emit_operand32(rbp, src); 4070 } 4071 4072 void Assembler::fldenv(Address src) { 4073 InstructionMark im(this); 4074 emit_int8((unsigned char)0xD9); 4075 emit_operand32(rsp, src); 4076 } 4077 4078 void Assembler::fldlg2() { 4079 emit_int8((unsigned char)0xD9); 4080 emit_int8((unsigned char)0xEC); 4081 } 4082 4083 void Assembler::fldln2() { 4084 emit_int8((unsigned char)0xD9); 4085 emit_int8((unsigned char)0xED); 4086 } 4087 4088 void Assembler::fldz() { 4089 emit_int8((unsigned char)0xD9); 4090 emit_int8((unsigned char)0xEE); 4091 } 4092 4093 void Assembler::flog() { 4094 fldln2(); 4095 fxch(); 4096 fyl2x(); 4097 } 4098 4099 void Assembler::flog10() { 4100 fldlg2(); 4101 fxch(); 4102 fyl2x(); 4103 } 4104 4105 void Assembler::fmul(int i) { 4106 emit_farith(0xD8, 0xC8, i); 4107 } 4108 4109 void Assembler::fmul_d(Address src) { 4110 InstructionMark im(this); 4111 emit_int8((unsigned char)0xDC); 4112 emit_operand32(rcx, src); 4113 } 4114 4115 void Assembler::fmul_s(Address src) { 4116 InstructionMark im(this); 4117 emit_int8((unsigned char)0xD8); 4118 emit_operand32(rcx, src); 4119 } 4120 4121 void Assembler::fmula(int i) { 4122 emit_farith(0xDC, 0xC8, i); 4123 } 4124 4125 void Assembler::fmulp(int i) { 4126 emit_farith(0xDE, 0xC8, i); 4127 } 4128 4129 void Assembler::fnsave(Address dst) { 4130 InstructionMark im(this); 4131 emit_int8((unsigned char)0xDD); 4132 emit_operand32(rsi, dst); 4133 } 4134 4135 void Assembler::fnstcw(Address src) { 4136 InstructionMark im(this); 4137 emit_int8((unsigned char)0x9B); 4138 emit_int8((unsigned char)0xD9); 4139 emit_operand32(rdi, src); 4140 } 4141 4142 void Assembler::fnstsw_ax() { 4143 emit_int8((unsigned char)0xDF); 4144 emit_int8((unsigned char)0xE0); 4145 } 4146 4147 void Assembler::fprem() { 4148 emit_int8((unsigned char)0xD9); 4149 emit_int8((unsigned char)0xF8); 4150 } 4151 4152 void Assembler::fprem1() { 4153 emit_int8((unsigned char)0xD9); 4154 emit_int8((unsigned char)0xF5); 4155 } 4156 4157 void Assembler::frstor(Address src) { 4158 InstructionMark im(this); 4159 emit_int8((unsigned char)0xDD); 4160 emit_operand32(rsp, src); 4161 } 4162 4163 void Assembler::fsin() { 4164 emit_int8((unsigned char)0xD9); 4165 emit_int8((unsigned char)0xFE); 4166 } 4167 4168 void Assembler::fsqrt() { 4169 emit_int8((unsigned char)0xD9); 4170 emit_int8((unsigned char)0xFA); 4171 } 4172 4173 void Assembler::fst_d(Address adr) { 4174 InstructionMark im(this); 4175 emit_int8((unsigned char)0xDD); 4176 emit_operand32(rdx, adr); 4177 } 4178 4179 void Assembler::fst_s(Address adr) { 4180 InstructionMark im(this); 4181 emit_int8((unsigned char)0xD9); 4182 emit_operand32(rdx, adr); 4183 } 4184 4185 void Assembler::fstp_d(Address adr) { 4186 InstructionMark im(this); 4187 emit_int8((unsigned char)0xDD); 4188 emit_operand32(rbx, adr); 4189 } 4190 4191 void Assembler::fstp_d(int index) { 4192 emit_farith(0xDD, 0xD8, index); 4193 } 4194 4195 void Assembler::fstp_s(Address adr) { 4196 InstructionMark im(this); 4197 emit_int8((unsigned char)0xD9); 4198 emit_operand32(rbx, adr); 4199 } 4200 4201 void Assembler::fstp_x(Address adr) { 4202 InstructionMark im(this); 4203 emit_int8((unsigned char)0xDB); 4204 emit_operand32(rdi, adr); 4205 } 4206 4207 void Assembler::fsub(int i) { 4208 emit_farith(0xD8, 0xE0, i); 4209 } 4210 4211 void Assembler::fsub_d(Address src) { 4212 InstructionMark im(this); 4213 emit_int8((unsigned char)0xDC); 4214 emit_operand32(rsp, src); 4215 } 4216 4217 void Assembler::fsub_s(Address src) { 4218 InstructionMark im(this); 4219 emit_int8((unsigned char)0xD8); 4220 emit_operand32(rsp, src); 4221 } 4222 4223 void Assembler::fsuba(int i) { 4224 emit_farith(0xDC, 0xE8, i); 4225 } 4226 4227 void Assembler::fsubp(int i) { 4228 emit_farith(0xDE, 0xE8, i); // ST(0) <- ST(0) - ST(1) and pop (Intel manual wrong) 4229 } 4230 4231 void Assembler::fsubr(int i) { 4232 emit_farith(0xD8, 0xE8, i); 4233 } 4234 4235 void Assembler::fsubr_d(Address src) { 4236 InstructionMark im(this); 4237 emit_int8((unsigned char)0xDC); 4238 emit_operand32(rbp, src); 4239 } 4240 4241 void Assembler::fsubr_s(Address src) { 4242 InstructionMark im(this); 4243 emit_int8((unsigned char)0xD8); 4244 emit_operand32(rbp, src); 4245 } 4246 4247 void Assembler::fsubra(int i) { 4248 emit_farith(0xDC, 0xE0, i); 4249 } 4250 4251 void Assembler::fsubrp(int i) { 4252 emit_farith(0xDE, 0xE0, i); // ST(0) <- ST(1) - ST(0) and pop (Intel manual wrong) 4253 } 4254 4255 void Assembler::ftan() { 4256 emit_int8((unsigned char)0xD9); 4257 emit_int8((unsigned char)0xF2); 4258 emit_int8((unsigned char)0xDD); 4259 emit_int8((unsigned char)0xD8); 4260 } 4261 4262 void Assembler::ftst() { 4263 emit_int8((unsigned char)0xD9); 4264 emit_int8((unsigned char)0xE4); 4265 } 4266 4267 void Assembler::fucomi(int i) { 4268 // make sure the instruction is supported (introduced for P6, together with cmov) 4269 guarantee(VM_Version::supports_cmov(), "illegal instruction"); 4270 emit_farith(0xDB, 0xE8, i); 4271 } 4272 4273 void Assembler::fucomip(int i) { 4274 // make sure the instruction is supported (introduced for P6, together with cmov) 4275 guarantee(VM_Version::supports_cmov(), "illegal instruction"); 4276 emit_farith(0xDF, 0xE8, i); 4277 } 4278 4279 void Assembler::fwait() { 4280 emit_int8((unsigned char)0x9B); 4281 } 4282 4283 void Assembler::fxch(int i) { 4284 emit_farith(0xD9, 0xC8, i); 4285 } 4286 4287 void Assembler::fyl2x() { 4288 emit_int8((unsigned char)0xD9); 4289 emit_int8((unsigned char)0xF1); 4290 } 4291 4292 void Assembler::frndint() { 4293 emit_int8((unsigned char)0xD9); 4294 emit_int8((unsigned char)0xFC); 4295 } 4296 4297 void Assembler::f2xm1() { 4298 emit_int8((unsigned char)0xD9); 4299 emit_int8((unsigned char)0xF0); 4300 } 4301 4302 void Assembler::fldl2e() { 4303 emit_int8((unsigned char)0xD9); 4304 emit_int8((unsigned char)0xEA); 4305 } 4306 4307 // SSE SIMD prefix byte values corresponding to VexSimdPrefix encoding. 4308 static int simd_pre[4] = { 0, 0x66, 0xF3, 0xF2 }; 4309 // SSE opcode second byte values (first is 0x0F) corresponding to VexOpcode encoding. 4310 static int simd_opc[4] = { 0, 0, 0x38, 0x3A }; 4311 4312 // Generate SSE legacy REX prefix and SIMD opcode based on VEX encoding. 4313 void Assembler::rex_prefix(Address adr, XMMRegister xreg, VexSimdPrefix pre, VexOpcode opc, bool rex_w) { 4314 if (pre > 0) { 4315 emit_int8(simd_pre[pre]); 4316 } 4317 if (rex_w) { 4318 prefixq(adr, xreg); 4319 } else { 4320 prefix(adr, xreg); 4321 } 4322 if (opc > 0) { 4323 emit_int8(0x0F); 4324 int opc2 = simd_opc[opc]; 4325 if (opc2 > 0) { 4326 emit_int8(opc2); 4327 } 4328 } 4329 } 4330 4331 int Assembler::rex_prefix_and_encode(int dst_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool rex_w) { 4332 if (pre > 0) { 4333 emit_int8(simd_pre[pre]); 4334 } 4335 int encode = (rex_w) ? prefixq_and_encode(dst_enc, src_enc) : 4336 prefix_and_encode(dst_enc, src_enc); 4337 if (opc > 0) { 4338 emit_int8(0x0F); 4339 int opc2 = simd_opc[opc]; 4340 if (opc2 > 0) { 4341 emit_int8(opc2); 4342 } 4343 } 4344 return encode; 4345 } 4346 4347 4348 void Assembler::vex_prefix(bool vex_r, bool vex_b, bool vex_x, bool vex_w, int nds_enc, VexSimdPrefix pre, VexOpcode opc, bool vector256) { 4349 if (vex_b || vex_x || vex_w || (opc == VEX_OPCODE_0F_38) || (opc == VEX_OPCODE_0F_3A)) { 4350 prefix(VEX_3bytes); 4351 4352 int byte1 = (vex_r ? VEX_R : 0) | (vex_x ? VEX_X : 0) | (vex_b ? VEX_B : 0); 4353 byte1 = (~byte1) & 0xE0; 4354 byte1 |= opc; 4355 emit_int8(byte1); 4356 4357 int byte2 = ((~nds_enc) & 0xf) << 3; 4358 byte2 |= (vex_w ? VEX_W : 0) | (vector256 ? 4 : 0) | pre; 4359 emit_int8(byte2); 4360 } else { 4361 prefix(VEX_2bytes); 4362 4363 int byte1 = vex_r ? VEX_R : 0; 4364 byte1 = (~byte1) & 0x80; 4365 byte1 |= ((~nds_enc) & 0xf) << 3; 4366 byte1 |= (vector256 ? 4 : 0) | pre; 4367 emit_int8(byte1); 4368 } 4369 } 4370 4371 void Assembler::vex_prefix(Address adr, int nds_enc, int xreg_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256){ 4372 bool vex_r = (xreg_enc >= 8); 4373 bool vex_b = adr.base_needs_rex(); 4374 bool vex_x = adr.index_needs_rex(); 4375 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256); 4376 } 4377 4378 int Assembler::vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256) { 4379 bool vex_r = (dst_enc >= 8); 4380 bool vex_b = (src_enc >= 8); 4381 bool vex_x = false; 4382 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256); 4383 return (((dst_enc & 7) << 3) | (src_enc & 7)); 4384 } 4385 4386 4387 void Assembler::simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) { 4388 if (UseAVX > 0) { 4389 int xreg_enc = xreg->encoding(); 4390 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 4391 vex_prefix(adr, nds_enc, xreg_enc, pre, opc, rex_w, vector256); 4392 } else { 4393 assert((nds == xreg) || (nds == xnoreg), "wrong sse encoding"); 4394 rex_prefix(adr, xreg, pre, opc, rex_w); 4395 } 4396 } 4397 4398 int Assembler::simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) { 4399 int dst_enc = dst->encoding(); 4400 int src_enc = src->encoding(); 4401 if (UseAVX > 0) { 4402 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 4403 return vex_prefix_and_encode(dst_enc, nds_enc, src_enc, pre, opc, rex_w, vector256); 4404 } else { 4405 assert((nds == dst) || (nds == src) || (nds == xnoreg), "wrong sse encoding"); 4406 return rex_prefix_and_encode(dst_enc, src_enc, pre, opc, rex_w); 4407 } 4408 } 4409 4410 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) { 4411 InstructionMark im(this); 4412 simd_prefix(dst, dst, src, pre); 4413 emit_int8(opcode); 4414 emit_operand(dst, src); 4415 } 4416 4417 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) { 4418 int encode = simd_prefix_and_encode(dst, dst, src, pre); 4419 emit_int8(opcode); 4420 emit_int8((unsigned char)(0xC0 | encode)); 4421 } 4422 4423 // Versions with no second source register (non-destructive source). 4424 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) { 4425 InstructionMark im(this); 4426 simd_prefix(dst, xnoreg, src, pre); 4427 emit_int8(opcode); 4428 emit_operand(dst, src); 4429 } 4430 4431 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) { 4432 int encode = simd_prefix_and_encode(dst, xnoreg, src, pre); 4433 emit_int8(opcode); 4434 emit_int8((unsigned char)(0xC0 | encode)); 4435 } 4436 4437 // 3-operands AVX instructions 4438 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds, 4439 Address src, VexSimdPrefix pre, bool vector256) { 4440 InstructionMark im(this); 4441 vex_prefix(dst, nds, src, pre, vector256); 4442 emit_int8(opcode); 4443 emit_operand(dst, src); 4444 } 4445 4446 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds, 4447 XMMRegister src, VexSimdPrefix pre, bool vector256) { 4448 int encode = vex_prefix_and_encode(dst, nds, src, pre, vector256); 4449 emit_int8(opcode); 4450 emit_int8((unsigned char)(0xC0 | encode)); 4451 } 4452 4453 #ifndef _LP64 4454 4455 void Assembler::incl(Register dst) { 4456 // Don't use it directly. Use MacroAssembler::incrementl() instead. 4457 emit_int8(0x40 | dst->encoding()); 4458 } 4459 4460 void Assembler::lea(Register dst, Address src) { 4461 leal(dst, src); 4462 } 4463 4464 void Assembler::mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec) { 4465 InstructionMark im(this); 4466 emit_int8((unsigned char)0xC7); 4467 emit_operand(rax, dst); 4468 emit_data((int)imm32, rspec, 0); 4469 } 4470 4471 void Assembler::mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec) { 4472 InstructionMark im(this); 4473 int encode = prefix_and_encode(dst->encoding()); 4474 emit_int8((unsigned char)(0xB8 | encode)); 4475 emit_data((int)imm32, rspec, 0); 4476 } 4477 4478 void Assembler::popa() { // 32bit 4479 emit_int8(0x61); 4480 } 4481 4482 void Assembler::push_literal32(int32_t imm32, RelocationHolder const& rspec) { 4483 InstructionMark im(this); 4484 emit_int8(0x68); 4485 emit_data(imm32, rspec, 0); 4486 } 4487 4488 void Assembler::pusha() { // 32bit 4489 emit_int8(0x60); 4490 } 4491 4492 void Assembler::set_byte_if_not_zero(Register dst) { 4493 emit_int8(0x0F); 4494 emit_int8((unsigned char)0x95); 4495 emit_int8((unsigned char)(0xE0 | dst->encoding())); 4496 } 4497 4498 void Assembler::shldl(Register dst, Register src) { 4499 emit_int8(0x0F); 4500 emit_int8((unsigned char)0xA5); 4501 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding())); 4502 } 4503 4504 void Assembler::shrdl(Register dst, Register src) { 4505 emit_int8(0x0F); 4506 emit_int8((unsigned char)0xAD); 4507 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding())); 4508 } 4509 4510 #else // LP64 4511 4512 void Assembler::set_byte_if_not_zero(Register dst) { 4513 int enc = prefix_and_encode(dst->encoding(), true); 4514 emit_int8(0x0F); 4515 emit_int8((unsigned char)0x95); 4516 emit_int8((unsigned char)(0xE0 | enc)); 4517 } 4518 4519 // 64bit only pieces of the assembler 4520 // This should only be used by 64bit instructions that can use rip-relative 4521 // it cannot be used by instructions that want an immediate value. 4522 4523 bool Assembler::reachable(AddressLiteral adr) { 4524 int64_t disp; 4525 // None will force a 64bit literal to the code stream. Likely a placeholder 4526 // for something that will be patched later and we need to certain it will 4527 // always be reachable. 4528 if (adr.reloc() == relocInfo::none) { 4529 return false; 4530 } 4531 if (adr.reloc() == relocInfo::internal_word_type) { 4532 // This should be rip relative and easily reachable. 4533 return true; 4534 } 4535 if (adr.reloc() == relocInfo::virtual_call_type || 4536 adr.reloc() == relocInfo::opt_virtual_call_type || 4537 adr.reloc() == relocInfo::static_call_type || 4538 adr.reloc() == relocInfo::static_stub_type ) { 4539 // This should be rip relative within the code cache and easily 4540 // reachable until we get huge code caches. (At which point 4541 // ic code is going to have issues). 4542 return true; 4543 } 4544 if (adr.reloc() != relocInfo::external_word_type && 4545 adr.reloc() != relocInfo::poll_return_type && // these are really external_word but need special 4546 adr.reloc() != relocInfo::poll_type && // relocs to identify them 4547 adr.reloc() != relocInfo::runtime_call_type ) { 4548 return false; 4549 } 4550 4551 // Stress the correction code 4552 if (ForceUnreachable) { 4553 // Must be runtimecall reloc, see if it is in the codecache 4554 // Flipping stuff in the codecache to be unreachable causes issues 4555 // with things like inline caches where the additional instructions 4556 // are not handled. 4557 if (CodeCache::find_blob(adr._target) == NULL) { 4558 return false; 4559 } 4560 } 4561 // For external_word_type/runtime_call_type if it is reachable from where we 4562 // are now (possibly a temp buffer) and where we might end up 4563 // anywhere in the codeCache then we are always reachable. 4564 // This would have to change if we ever save/restore shared code 4565 // to be more pessimistic. 4566 disp = (int64_t)adr._target - ((int64_t)CodeCache::low_bound() + sizeof(int)); 4567 if (!is_simm32(disp)) return false; 4568 disp = (int64_t)adr._target - ((int64_t)CodeCache::high_bound() + sizeof(int)); 4569 if (!is_simm32(disp)) return false; 4570 4571 disp = (int64_t)adr._target - ((int64_t)pc() + sizeof(int)); 4572 4573 // Because rip relative is a disp + address_of_next_instruction and we 4574 // don't know the value of address_of_next_instruction we apply a fudge factor 4575 // to make sure we will be ok no matter the size of the instruction we get placed into. 4576 // We don't have to fudge the checks above here because they are already worst case. 4577 4578 // 12 == override/rex byte, opcode byte, rm byte, sib byte, a 4-byte disp , 4-byte literal 4579 // + 4 because better safe than sorry. 4580 const int fudge = 12 + 4; 4581 if (disp < 0) { 4582 disp -= fudge; 4583 } else { 4584 disp += fudge; 4585 } 4586 return is_simm32(disp); 4587 } 4588 4589 // Check if the polling page is not reachable from the code cache using rip-relative 4590 // addressing. 4591 bool Assembler::is_polling_page_far() { 4592 intptr_t addr = (intptr_t)os::get_polling_page(); 4593 return ForceUnreachable || 4594 !is_simm32(addr - (intptr_t)CodeCache::low_bound()) || 4595 !is_simm32(addr - (intptr_t)CodeCache::high_bound()); 4596 } 4597 4598 void Assembler::emit_data64(jlong data, 4599 relocInfo::relocType rtype, 4600 int format) { 4601 if (rtype == relocInfo::none) { 4602 emit_int64(data); 4603 } else { 4604 emit_data64(data, Relocation::spec_simple(rtype), format); 4605 } 4606 } 4607 4608 void Assembler::emit_data64(jlong data, 4609 RelocationHolder const& rspec, 4610 int format) { 4611 assert(imm_operand == 0, "default format must be immediate in this file"); 4612 assert(imm_operand == format, "must be immediate"); 4613 assert(inst_mark() != NULL, "must be inside InstructionMark"); 4614 // Do not use AbstractAssembler::relocate, which is not intended for 4615 // embedded words. Instead, relocate to the enclosing instruction. 4616 code_section()->relocate(inst_mark(), rspec, format); 4617 #ifdef ASSERT 4618 check_relocation(rspec, format); 4619 #endif 4620 emit_int64(data); 4621 } 4622 4623 int Assembler::prefix_and_encode(int reg_enc, bool byteinst) { 4624 if (reg_enc >= 8) { 4625 prefix(REX_B); 4626 reg_enc -= 8; 4627 } else if (byteinst && reg_enc >= 4) { 4628 prefix(REX); 4629 } 4630 return reg_enc; 4631 } 4632 4633 int Assembler::prefixq_and_encode(int reg_enc) { 4634 if (reg_enc < 8) { 4635 prefix(REX_W); 4636 } else { 4637 prefix(REX_WB); 4638 reg_enc -= 8; 4639 } 4640 return reg_enc; 4641 } 4642 4643 int Assembler::prefix_and_encode(int dst_enc, int src_enc, bool byteinst) { 4644 if (dst_enc < 8) { 4645 if (src_enc >= 8) { 4646 prefix(REX_B); 4647 src_enc -= 8; 4648 } else if (byteinst && src_enc >= 4) { 4649 prefix(REX); 4650 } 4651 } else { 4652 if (src_enc < 8) { 4653 prefix(REX_R); 4654 } else { 4655 prefix(REX_RB); 4656 src_enc -= 8; 4657 } 4658 dst_enc -= 8; 4659 } 4660 return dst_enc << 3 | src_enc; 4661 } 4662 4663 int Assembler::prefixq_and_encode(int dst_enc, int src_enc) { 4664 if (dst_enc < 8) { 4665 if (src_enc < 8) { 4666 prefix(REX_W); 4667 } else { 4668 prefix(REX_WB); 4669 src_enc -= 8; 4670 } 4671 } else { 4672 if (src_enc < 8) { 4673 prefix(REX_WR); 4674 } else { 4675 prefix(REX_WRB); 4676 src_enc -= 8; 4677 } 4678 dst_enc -= 8; 4679 } 4680 return dst_enc << 3 | src_enc; 4681 } 4682 4683 void Assembler::prefix(Register reg) { 4684 if (reg->encoding() >= 8) { 4685 prefix(REX_B); 4686 } 4687 } 4688 4689 void Assembler::prefix(Address adr) { 4690 if (adr.base_needs_rex()) { 4691 if (adr.index_needs_rex()) { 4692 prefix(REX_XB); 4693 } else { 4694 prefix(REX_B); 4695 } 4696 } else { 4697 if (adr.index_needs_rex()) { 4698 prefix(REX_X); 4699 } 4700 } 4701 } 4702 4703 void Assembler::prefixq(Address adr) { 4704 if (adr.base_needs_rex()) { 4705 if (adr.index_needs_rex()) { 4706 prefix(REX_WXB); 4707 } else { 4708 prefix(REX_WB); 4709 } 4710 } else { 4711 if (adr.index_needs_rex()) { 4712 prefix(REX_WX); 4713 } else { 4714 prefix(REX_W); 4715 } 4716 } 4717 } 4718 4719 4720 void Assembler::prefix(Address adr, Register reg, bool byteinst) { 4721 if (reg->encoding() < 8) { 4722 if (adr.base_needs_rex()) { 4723 if (adr.index_needs_rex()) { 4724 prefix(REX_XB); 4725 } else { 4726 prefix(REX_B); 4727 } 4728 } else { 4729 if (adr.index_needs_rex()) { 4730 prefix(REX_X); 4731 } else if (byteinst && reg->encoding() >= 4 ) { 4732 prefix(REX); 4733 } 4734 } 4735 } else { 4736 if (adr.base_needs_rex()) { 4737 if (adr.index_needs_rex()) { 4738 prefix(REX_RXB); 4739 } else { 4740 prefix(REX_RB); 4741 } 4742 } else { 4743 if (adr.index_needs_rex()) { 4744 prefix(REX_RX); 4745 } else { 4746 prefix(REX_R); 4747 } 4748 } 4749 } 4750 } 4751 4752 void Assembler::prefixq(Address adr, Register src) { 4753 if (src->encoding() < 8) { 4754 if (adr.base_needs_rex()) { 4755 if (adr.index_needs_rex()) { 4756 prefix(REX_WXB); 4757 } else { 4758 prefix(REX_WB); 4759 } 4760 } else { 4761 if (adr.index_needs_rex()) { 4762 prefix(REX_WX); 4763 } else { 4764 prefix(REX_W); 4765 } 4766 } 4767 } else { 4768 if (adr.base_needs_rex()) { 4769 if (adr.index_needs_rex()) { 4770 prefix(REX_WRXB); 4771 } else { 4772 prefix(REX_WRB); 4773 } 4774 } else { 4775 if (adr.index_needs_rex()) { 4776 prefix(REX_WRX); 4777 } else { 4778 prefix(REX_WR); 4779 } 4780 } 4781 } 4782 } 4783 4784 void Assembler::prefix(Address adr, XMMRegister reg) { 4785 if (reg->encoding() < 8) { 4786 if (adr.base_needs_rex()) { 4787 if (adr.index_needs_rex()) { 4788 prefix(REX_XB); 4789 } else { 4790 prefix(REX_B); 4791 } 4792 } else { 4793 if (adr.index_needs_rex()) { 4794 prefix(REX_X); 4795 } 4796 } 4797 } else { 4798 if (adr.base_needs_rex()) { 4799 if (adr.index_needs_rex()) { 4800 prefix(REX_RXB); 4801 } else { 4802 prefix(REX_RB); 4803 } 4804 } else { 4805 if (adr.index_needs_rex()) { 4806 prefix(REX_RX); 4807 } else { 4808 prefix(REX_R); 4809 } 4810 } 4811 } 4812 } 4813 4814 void Assembler::prefixq(Address adr, XMMRegister src) { 4815 if (src->encoding() < 8) { 4816 if (adr.base_needs_rex()) { 4817 if (adr.index_needs_rex()) { 4818 prefix(REX_WXB); 4819 } else { 4820 prefix(REX_WB); 4821 } 4822 } else { 4823 if (adr.index_needs_rex()) { 4824 prefix(REX_WX); 4825 } else { 4826 prefix(REX_W); 4827 } 4828 } 4829 } else { 4830 if (adr.base_needs_rex()) { 4831 if (adr.index_needs_rex()) { 4832 prefix(REX_WRXB); 4833 } else { 4834 prefix(REX_WRB); 4835 } 4836 } else { 4837 if (adr.index_needs_rex()) { 4838 prefix(REX_WRX); 4839 } else { 4840 prefix(REX_WR); 4841 } 4842 } 4843 } 4844 } 4845 4846 void Assembler::adcq(Register dst, int32_t imm32) { 4847 (void) prefixq_and_encode(dst->encoding()); 4848 emit_arith(0x81, 0xD0, dst, imm32); 4849 } 4850 4851 void Assembler::adcq(Register dst, Address src) { 4852 InstructionMark im(this); 4853 prefixq(src, dst); 4854 emit_int8(0x13); 4855 emit_operand(dst, src); 4856 } 4857 4858 void Assembler::adcq(Register dst, Register src) { 4859 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 4860 emit_arith(0x13, 0xC0, dst, src); 4861 } 4862 4863 void Assembler::addq(Address dst, int32_t imm32) { 4864 InstructionMark im(this); 4865 prefixq(dst); 4866 emit_arith_operand(0x81, rax, dst,imm32); 4867 } 4868 4869 void Assembler::addq(Address dst, Register src) { 4870 InstructionMark im(this); 4871 prefixq(dst, src); 4872 emit_int8(0x01); 4873 emit_operand(src, dst); 4874 } 4875 4876 void Assembler::addq(Register dst, int32_t imm32) { 4877 (void) prefixq_and_encode(dst->encoding()); 4878 emit_arith(0x81, 0xC0, dst, imm32); 4879 } 4880 4881 void Assembler::addq(Register dst, Address src) { 4882 InstructionMark im(this); 4883 prefixq(src, dst); 4884 emit_int8(0x03); 4885 emit_operand(dst, src); 4886 } 4887 4888 void Assembler::addq(Register dst, Register src) { 4889 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 4890 emit_arith(0x03, 0xC0, dst, src); 4891 } 4892 4893 void Assembler::andq(Address dst, int32_t imm32) { 4894 InstructionMark im(this); 4895 prefixq(dst); 4896 emit_int8((unsigned char)0x81); 4897 emit_operand(rsp, dst, 4); 4898 emit_int32(imm32); 4899 } 4900 4901 void Assembler::andq(Register dst, int32_t imm32) { 4902 (void) prefixq_and_encode(dst->encoding()); 4903 emit_arith(0x81, 0xE0, dst, imm32); 4904 } 4905 4906 void Assembler::andq(Register dst, Address src) { 4907 InstructionMark im(this); 4908 prefixq(src, dst); 4909 emit_int8(0x23); 4910 emit_operand(dst, src); 4911 } 4912 4913 void Assembler::andq(Register dst, Register src) { 4914 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 4915 emit_arith(0x23, 0xC0, dst, src); 4916 } 4917 4918 void Assembler::andnq(Register dst, Register src1, Register src2) { 4919 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4920 int encode = vex_prefix_0F38_and_encode_q(dst, src1, src2); 4921 emit_int8((unsigned char)0xF2); 4922 emit_int8((unsigned char)(0xC0 | encode)); 4923 } 4924 4925 void Assembler::andnq(Register dst, Register src1, Address src2) { 4926 InstructionMark im(this); 4927 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4928 vex_prefix_0F38_q(dst, src1, src2); 4929 emit_int8((unsigned char)0xF2); 4930 emit_operand(dst, src2); 4931 } 4932 4933 void Assembler::bsfq(Register dst, Register src) { 4934 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 4935 emit_int8(0x0F); 4936 emit_int8((unsigned char)0xBC); 4937 emit_int8((unsigned char)(0xC0 | encode)); 4938 } 4939 4940 void Assembler::bsrq(Register dst, Register src) { 4941 assert(!VM_Version::supports_lzcnt(), "encoding is treated as LZCNT"); 4942 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 4943 emit_int8(0x0F); 4944 emit_int8((unsigned char)0xBD); 4945 emit_int8((unsigned char)(0xC0 | encode)); 4946 } 4947 4948 void Assembler::bswapq(Register reg) { 4949 int encode = prefixq_and_encode(reg->encoding()); 4950 emit_int8(0x0F); 4951 emit_int8((unsigned char)(0xC8 | encode)); 4952 } 4953 4954 void Assembler::blsiq(Register dst, Register src) { 4955 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4956 int encode = vex_prefix_0F38_and_encode_q(rbx, dst, src); 4957 emit_int8((unsigned char)0xF3); 4958 emit_int8((unsigned char)(0xC0 | encode)); 4959 } 4960 4961 void Assembler::blsiq(Register dst, Address src) { 4962 InstructionMark im(this); 4963 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4964 vex_prefix_0F38_q(rbx, dst, src); 4965 emit_int8((unsigned char)0xF3); 4966 emit_operand(rbx, src); 4967 } 4968 4969 void Assembler::blsmskq(Register dst, Register src) { 4970 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4971 int encode = vex_prefix_0F38_and_encode_q(rdx, dst, src); 4972 emit_int8((unsigned char)0xF3); 4973 emit_int8((unsigned char)(0xC0 | encode)); 4974 } 4975 4976 void Assembler::blsmskq(Register dst, Address src) { 4977 InstructionMark im(this); 4978 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4979 vex_prefix_0F38_q(rdx, dst, src); 4980 emit_int8((unsigned char)0xF3); 4981 emit_operand(rdx, src); 4982 } 4983 4984 void Assembler::blsrq(Register dst, Register src) { 4985 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4986 int encode = vex_prefix_0F38_and_encode_q(rcx, dst, src); 4987 emit_int8((unsigned char)0xF3); 4988 emit_int8((unsigned char)(0xC0 | encode)); 4989 } 4990 4991 void Assembler::blsrq(Register dst, Address src) { 4992 InstructionMark im(this); 4993 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 4994 vex_prefix_0F38_q(rcx, dst, src); 4995 emit_int8((unsigned char)0xF3); 4996 emit_operand(rcx, src); 4997 } 4998 4999 void Assembler::cdqq() { 5000 prefix(REX_W); 5001 emit_int8((unsigned char)0x99); 5002 } 5003 5004 void Assembler::clflush(Address adr) { 5005 prefix(adr); 5006 emit_int8(0x0F); 5007 emit_int8((unsigned char)0xAE); 5008 emit_operand(rdi, adr); 5009 } 5010 5011 void Assembler::cmovq(Condition cc, Register dst, Register src) { 5012 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5013 emit_int8(0x0F); 5014 emit_int8(0x40 | cc); 5015 emit_int8((unsigned char)(0xC0 | encode)); 5016 } 5017 5018 void Assembler::cmovq(Condition cc, Register dst, Address src) { 5019 InstructionMark im(this); 5020 prefixq(src, dst); 5021 emit_int8(0x0F); 5022 emit_int8(0x40 | cc); 5023 emit_operand(dst, src); 5024 } 5025 5026 void Assembler::cmpq(Address dst, int32_t imm32) { 5027 InstructionMark im(this); 5028 prefixq(dst); 5029 emit_int8((unsigned char)0x81); 5030 emit_operand(rdi, dst, 4); 5031 emit_int32(imm32); 5032 } 5033 5034 void Assembler::cmpq(Register dst, int32_t imm32) { 5035 (void) prefixq_and_encode(dst->encoding()); 5036 emit_arith(0x81, 0xF8, dst, imm32); 5037 } 5038 5039 void Assembler::cmpq(Address dst, Register src) { 5040 InstructionMark im(this); 5041 prefixq(dst, src); 5042 emit_int8(0x3B); 5043 emit_operand(src, dst); 5044 } 5045 5046 void Assembler::cmpq(Register dst, Register src) { 5047 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5048 emit_arith(0x3B, 0xC0, dst, src); 5049 } 5050 5051 void Assembler::cmpq(Register dst, Address src) { 5052 InstructionMark im(this); 5053 prefixq(src, dst); 5054 emit_int8(0x3B); 5055 emit_operand(dst, src); 5056 } 5057 5058 void Assembler::cmpxchgq(Register reg, Address adr) { 5059 InstructionMark im(this); 5060 prefixq(adr, reg); 5061 emit_int8(0x0F); 5062 emit_int8((unsigned char)0xB1); 5063 emit_operand(reg, adr); 5064 } 5065 5066 void Assembler::cvtsi2sdq(XMMRegister dst, Register src) { 5067 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5068 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F2); 5069 emit_int8(0x2A); 5070 emit_int8((unsigned char)(0xC0 | encode)); 5071 } 5072 5073 void Assembler::cvtsi2sdq(XMMRegister dst, Address src) { 5074 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5075 InstructionMark im(this); 5076 simd_prefix_q(dst, dst, src, VEX_SIMD_F2); 5077 emit_int8(0x2A); 5078 emit_operand(dst, src); 5079 } 5080 5081 void Assembler::cvtsi2ssq(XMMRegister dst, Register src) { 5082 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5083 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F3); 5084 emit_int8(0x2A); 5085 emit_int8((unsigned char)(0xC0 | encode)); 5086 } 5087 5088 void Assembler::cvtsi2ssq(XMMRegister dst, Address src) { 5089 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5090 InstructionMark im(this); 5091 simd_prefix_q(dst, dst, src, VEX_SIMD_F3); 5092 emit_int8(0x2A); 5093 emit_operand(dst, src); 5094 } 5095 5096 void Assembler::cvttsd2siq(Register dst, XMMRegister src) { 5097 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5098 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F2); 5099 emit_int8(0x2C); 5100 emit_int8((unsigned char)(0xC0 | encode)); 5101 } 5102 5103 void Assembler::cvttss2siq(Register dst, XMMRegister src) { 5104 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5105 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F3); 5106 emit_int8(0x2C); 5107 emit_int8((unsigned char)(0xC0 | encode)); 5108 } 5109 5110 void Assembler::decl(Register dst) { 5111 // Don't use it directly. Use MacroAssembler::decrementl() instead. 5112 // Use two-byte form (one-byte form is a REX prefix in 64-bit mode) 5113 int encode = prefix_and_encode(dst->encoding()); 5114 emit_int8((unsigned char)0xFF); 5115 emit_int8((unsigned char)(0xC8 | encode)); 5116 } 5117 5118 void Assembler::decq(Register dst) { 5119 // Don't use it directly. Use MacroAssembler::decrementq() instead. 5120 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5121 int encode = prefixq_and_encode(dst->encoding()); 5122 emit_int8((unsigned char)0xFF); 5123 emit_int8(0xC8 | encode); 5124 } 5125 5126 void Assembler::decq(Address dst) { 5127 // Don't use it directly. Use MacroAssembler::decrementq() instead. 5128 InstructionMark im(this); 5129 prefixq(dst); 5130 emit_int8((unsigned char)0xFF); 5131 emit_operand(rcx, dst); 5132 } 5133 5134 void Assembler::fxrstor(Address src) { 5135 prefixq(src); 5136 emit_int8(0x0F); 5137 emit_int8((unsigned char)0xAE); 5138 emit_operand(as_Register(1), src); 5139 } 5140 5141 void Assembler::fxsave(Address dst) { 5142 prefixq(dst); 5143 emit_int8(0x0F); 5144 emit_int8((unsigned char)0xAE); 5145 emit_operand(as_Register(0), dst); 5146 } 5147 5148 void Assembler::idivq(Register src) { 5149 int encode = prefixq_and_encode(src->encoding()); 5150 emit_int8((unsigned char)0xF7); 5151 emit_int8((unsigned char)(0xF8 | encode)); 5152 } 5153 5154 void Assembler::imulq(Register dst, Register src) { 5155 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5156 emit_int8(0x0F); 5157 emit_int8((unsigned char)0xAF); 5158 emit_int8((unsigned char)(0xC0 | encode)); 5159 } 5160 5161 void Assembler::imulq(Register dst, Register src, int value) { 5162 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5163 if (is8bit(value)) { 5164 emit_int8(0x6B); 5165 emit_int8((unsigned char)(0xC0 | encode)); 5166 emit_int8(value & 0xFF); 5167 } else { 5168 emit_int8(0x69); 5169 emit_int8((unsigned char)(0xC0 | encode)); 5170 emit_int32(value); 5171 } 5172 } 5173 5174 void Assembler::imulq(Register dst, Address src) { 5175 InstructionMark im(this); 5176 prefixq(src, dst); 5177 emit_int8(0x0F); 5178 emit_int8((unsigned char) 0xAF); 5179 emit_operand(dst, src); 5180 } 5181 5182 void Assembler::incl(Register dst) { 5183 // Don't use it directly. Use MacroAssembler::incrementl() instead. 5184 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5185 int encode = prefix_and_encode(dst->encoding()); 5186 emit_int8((unsigned char)0xFF); 5187 emit_int8((unsigned char)(0xC0 | encode)); 5188 } 5189 5190 void Assembler::incq(Register dst) { 5191 // Don't use it directly. Use MacroAssembler::incrementq() instead. 5192 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5193 int encode = prefixq_and_encode(dst->encoding()); 5194 emit_int8((unsigned char)0xFF); 5195 emit_int8((unsigned char)(0xC0 | encode)); 5196 } 5197 5198 void Assembler::incq(Address dst) { 5199 // Don't use it directly. Use MacroAssembler::incrementq() instead. 5200 InstructionMark im(this); 5201 prefixq(dst); 5202 emit_int8((unsigned char)0xFF); 5203 emit_operand(rax, dst); 5204 } 5205 5206 void Assembler::lea(Register dst, Address src) { 5207 leaq(dst, src); 5208 } 5209 5210 void Assembler::leaq(Register dst, Address src) { 5211 InstructionMark im(this); 5212 prefixq(src, dst); 5213 emit_int8((unsigned char)0x8D); 5214 emit_operand(dst, src); 5215 } 5216 5217 void Assembler::mov64(Register dst, int64_t imm64) { 5218 InstructionMark im(this); 5219 int encode = prefixq_and_encode(dst->encoding()); 5220 emit_int8((unsigned char)(0xB8 | encode)); 5221 emit_int64(imm64); 5222 } 5223 5224 void Assembler::mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec) { 5225 InstructionMark im(this); 5226 int encode = prefixq_and_encode(dst->encoding()); 5227 emit_int8(0xB8 | encode); 5228 emit_data64(imm64, rspec); 5229 } 5230 5231 void Assembler::mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec) { 5232 InstructionMark im(this); 5233 int encode = prefix_and_encode(dst->encoding()); 5234 emit_int8((unsigned char)(0xB8 | encode)); 5235 emit_data((int)imm32, rspec, narrow_oop_operand); 5236 } 5237 5238 void Assembler::mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec) { 5239 InstructionMark im(this); 5240 prefix(dst); 5241 emit_int8((unsigned char)0xC7); 5242 emit_operand(rax, dst, 4); 5243 emit_data((int)imm32, rspec, narrow_oop_operand); 5244 } 5245 5246 void Assembler::cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec) { 5247 InstructionMark im(this); 5248 int encode = prefix_and_encode(src1->encoding()); 5249 emit_int8((unsigned char)0x81); 5250 emit_int8((unsigned char)(0xF8 | encode)); 5251 emit_data((int)imm32, rspec, narrow_oop_operand); 5252 } 5253 5254 void Assembler::cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec) { 5255 InstructionMark im(this); 5256 prefix(src1); 5257 emit_int8((unsigned char)0x81); 5258 emit_operand(rax, src1, 4); 5259 emit_data((int)imm32, rspec, narrow_oop_operand); 5260 } 5261 5262 void Assembler::lzcntq(Register dst, Register src) { 5263 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR"); 5264 emit_int8((unsigned char)0xF3); 5265 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5266 emit_int8(0x0F); 5267 emit_int8((unsigned char)0xBD); 5268 emit_int8((unsigned char)(0xC0 | encode)); 5269 } 5270 5271 void Assembler::movdq(XMMRegister dst, Register src) { 5272 // table D-1 says MMX/SSE2 5273 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5274 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_66); 5275 emit_int8(0x6E); 5276 emit_int8((unsigned char)(0xC0 | encode)); 5277 } 5278 5279 void Assembler::movdq(Register dst, XMMRegister src) { 5280 // table D-1 says MMX/SSE2 5281 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5282 // swap src/dst to get correct prefix 5283 int encode = simd_prefix_and_encode_q(src, dst, VEX_SIMD_66); 5284 emit_int8(0x7E); 5285 emit_int8((unsigned char)(0xC0 | encode)); 5286 } 5287 5288 void Assembler::movq(Register dst, Register src) { 5289 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5290 emit_int8((unsigned char)0x8B); 5291 emit_int8((unsigned char)(0xC0 | encode)); 5292 } 5293 5294 void Assembler::movq(Register dst, Address src) { 5295 InstructionMark im(this); 5296 prefixq(src, dst); 5297 emit_int8((unsigned char)0x8B); 5298 emit_operand(dst, src); 5299 } 5300 5301 void Assembler::movq(Address dst, Register src) { 5302 InstructionMark im(this); 5303 prefixq(dst, src); 5304 emit_int8((unsigned char)0x89); 5305 emit_operand(src, dst); 5306 } 5307 5308 void Assembler::movsbq(Register dst, Address src) { 5309 InstructionMark im(this); 5310 prefixq(src, dst); 5311 emit_int8(0x0F); 5312 emit_int8((unsigned char)0xBE); 5313 emit_operand(dst, src); 5314 } 5315 5316 void Assembler::movsbq(Register dst, Register src) { 5317 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5318 emit_int8(0x0F); 5319 emit_int8((unsigned char)0xBE); 5320 emit_int8((unsigned char)(0xC0 | encode)); 5321 } 5322 5323 void Assembler::movslq(Register dst, int32_t imm32) { 5324 // dbx shows movslq(rcx, 3) as movq $0x0000000049000000,(%rbx) 5325 // and movslq(r8, 3); as movl $0x0000000048000000,(%rbx) 5326 // as a result we shouldn't use until tested at runtime... 5327 ShouldNotReachHere(); 5328 InstructionMark im(this); 5329 int encode = prefixq_and_encode(dst->encoding()); 5330 emit_int8((unsigned char)(0xC7 | encode)); 5331 emit_int32(imm32); 5332 } 5333 5334 void Assembler::movslq(Address dst, int32_t imm32) { 5335 assert(is_simm32(imm32), "lost bits"); 5336 InstructionMark im(this); 5337 prefixq(dst); 5338 emit_int8((unsigned char)0xC7); 5339 emit_operand(rax, dst, 4); 5340 emit_int32(imm32); 5341 } 5342 5343 void Assembler::movslq(Register dst, Address src) { 5344 InstructionMark im(this); 5345 prefixq(src, dst); 5346 emit_int8(0x63); 5347 emit_operand(dst, src); 5348 } 5349 5350 void Assembler::movslq(Register dst, Register src) { 5351 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5352 emit_int8(0x63); 5353 emit_int8((unsigned char)(0xC0 | encode)); 5354 } 5355 5356 void Assembler::movswq(Register dst, Address src) { 5357 InstructionMark im(this); 5358 prefixq(src, dst); 5359 emit_int8(0x0F); 5360 emit_int8((unsigned char)0xBF); 5361 emit_operand(dst, src); 5362 } 5363 5364 void Assembler::movswq(Register dst, Register src) { 5365 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5366 emit_int8((unsigned char)0x0F); 5367 emit_int8((unsigned char)0xBF); 5368 emit_int8((unsigned char)(0xC0 | encode)); 5369 } 5370 5371 void Assembler::movzbq(Register dst, Address src) { 5372 InstructionMark im(this); 5373 prefixq(src, dst); 5374 emit_int8((unsigned char)0x0F); 5375 emit_int8((unsigned char)0xB6); 5376 emit_operand(dst, src); 5377 } 5378 5379 void Assembler::movzbq(Register dst, Register src) { 5380 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5381 emit_int8(0x0F); 5382 emit_int8((unsigned char)0xB6); 5383 emit_int8(0xC0 | encode); 5384 } 5385 5386 void Assembler::movzwq(Register dst, Address src) { 5387 InstructionMark im(this); 5388 prefixq(src, dst); 5389 emit_int8((unsigned char)0x0F); 5390 emit_int8((unsigned char)0xB7); 5391 emit_operand(dst, src); 5392 } 5393 5394 void Assembler::movzwq(Register dst, Register src) { 5395 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5396 emit_int8((unsigned char)0x0F); 5397 emit_int8((unsigned char)0xB7); 5398 emit_int8((unsigned char)(0xC0 | encode)); 5399 } 5400 5401 void Assembler::negq(Register dst) { 5402 int encode = prefixq_and_encode(dst->encoding()); 5403 emit_int8((unsigned char)0xF7); 5404 emit_int8((unsigned char)(0xD8 | encode)); 5405 } 5406 5407 void Assembler::notq(Register dst) { 5408 int encode = prefixq_and_encode(dst->encoding()); 5409 emit_int8((unsigned char)0xF7); 5410 emit_int8((unsigned char)(0xD0 | encode)); 5411 } 5412 5413 void Assembler::orq(Address dst, int32_t imm32) { 5414 InstructionMark im(this); 5415 prefixq(dst); 5416 emit_int8((unsigned char)0x81); 5417 emit_operand(rcx, dst, 4); 5418 emit_int32(imm32); 5419 } 5420 5421 void Assembler::orq(Register dst, int32_t imm32) { 5422 (void) prefixq_and_encode(dst->encoding()); 5423 emit_arith(0x81, 0xC8, dst, imm32); 5424 } 5425 5426 void Assembler::orq(Register dst, Address src) { 5427 InstructionMark im(this); 5428 prefixq(src, dst); 5429 emit_int8(0x0B); 5430 emit_operand(dst, src); 5431 } 5432 5433 void Assembler::orq(Register dst, Register src) { 5434 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5435 emit_arith(0x0B, 0xC0, dst, src); 5436 } 5437 5438 void Assembler::popa() { // 64bit 5439 movq(r15, Address(rsp, 0)); 5440 movq(r14, Address(rsp, wordSize)); 5441 movq(r13, Address(rsp, 2 * wordSize)); 5442 movq(r12, Address(rsp, 3 * wordSize)); 5443 movq(r11, Address(rsp, 4 * wordSize)); 5444 movq(r10, Address(rsp, 5 * wordSize)); 5445 movq(r9, Address(rsp, 6 * wordSize)); 5446 movq(r8, Address(rsp, 7 * wordSize)); 5447 movq(rdi, Address(rsp, 8 * wordSize)); 5448 movq(rsi, Address(rsp, 9 * wordSize)); 5449 movq(rbp, Address(rsp, 10 * wordSize)); 5450 // skip rsp 5451 movq(rbx, Address(rsp, 12 * wordSize)); 5452 movq(rdx, Address(rsp, 13 * wordSize)); 5453 movq(rcx, Address(rsp, 14 * wordSize)); 5454 movq(rax, Address(rsp, 15 * wordSize)); 5455 5456 addq(rsp, 16 * wordSize); 5457 } 5458 5459 void Assembler::popcntq(Register dst, Address src) { 5460 assert(VM_Version::supports_popcnt(), "must support"); 5461 InstructionMark im(this); 5462 emit_int8((unsigned char)0xF3); 5463 prefixq(src, dst); 5464 emit_int8((unsigned char)0x0F); 5465 emit_int8((unsigned char)0xB8); 5466 emit_operand(dst, src); 5467 } 5468 5469 void Assembler::popcntq(Register dst, Register src) { 5470 assert(VM_Version::supports_popcnt(), "must support"); 5471 emit_int8((unsigned char)0xF3); 5472 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5473 emit_int8((unsigned char)0x0F); 5474 emit_int8((unsigned char)0xB8); 5475 emit_int8((unsigned char)(0xC0 | encode)); 5476 } 5477 5478 void Assembler::popq(Address dst) { 5479 InstructionMark im(this); 5480 prefixq(dst); 5481 emit_int8((unsigned char)0x8F); 5482 emit_operand(rax, dst); 5483 } 5484 5485 void Assembler::pusha() { // 64bit 5486 // we have to store original rsp. ABI says that 128 bytes 5487 // below rsp are local scratch. 5488 movq(Address(rsp, -5 * wordSize), rsp); 5489 5490 subq(rsp, 16 * wordSize); 5491 5492 movq(Address(rsp, 15 * wordSize), rax); 5493 movq(Address(rsp, 14 * wordSize), rcx); 5494 movq(Address(rsp, 13 * wordSize), rdx); 5495 movq(Address(rsp, 12 * wordSize), rbx); 5496 // skip rsp 5497 movq(Address(rsp, 10 * wordSize), rbp); 5498 movq(Address(rsp, 9 * wordSize), rsi); 5499 movq(Address(rsp, 8 * wordSize), rdi); 5500 movq(Address(rsp, 7 * wordSize), r8); 5501 movq(Address(rsp, 6 * wordSize), r9); 5502 movq(Address(rsp, 5 * wordSize), r10); 5503 movq(Address(rsp, 4 * wordSize), r11); 5504 movq(Address(rsp, 3 * wordSize), r12); 5505 movq(Address(rsp, 2 * wordSize), r13); 5506 movq(Address(rsp, wordSize), r14); 5507 movq(Address(rsp, 0), r15); 5508 } 5509 5510 void Assembler::pushq(Address src) { 5511 InstructionMark im(this); 5512 prefixq(src); 5513 emit_int8((unsigned char)0xFF); 5514 emit_operand(rsi, src); 5515 } 5516 5517 void Assembler::rclq(Register dst, int imm8) { 5518 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5519 int encode = prefixq_and_encode(dst->encoding()); 5520 if (imm8 == 1) { 5521 emit_int8((unsigned char)0xD1); 5522 emit_int8((unsigned char)(0xD0 | encode)); 5523 } else { 5524 emit_int8((unsigned char)0xC1); 5525 emit_int8((unsigned char)(0xD0 | encode)); 5526 emit_int8(imm8); 5527 } 5528 } 5529 void Assembler::sarq(Register dst, int imm8) { 5530 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5531 int encode = prefixq_and_encode(dst->encoding()); 5532 if (imm8 == 1) { 5533 emit_int8((unsigned char)0xD1); 5534 emit_int8((unsigned char)(0xF8 | encode)); 5535 } else { 5536 emit_int8((unsigned char)0xC1); 5537 emit_int8((unsigned char)(0xF8 | encode)); 5538 emit_int8(imm8); 5539 } 5540 } 5541 5542 void Assembler::sarq(Register dst) { 5543 int encode = prefixq_and_encode(dst->encoding()); 5544 emit_int8((unsigned char)0xD3); 5545 emit_int8((unsigned char)(0xF8 | encode)); 5546 } 5547 5548 void Assembler::sbbq(Address dst, int32_t imm32) { 5549 InstructionMark im(this); 5550 prefixq(dst); 5551 emit_arith_operand(0x81, rbx, dst, imm32); 5552 } 5553 5554 void Assembler::sbbq(Register dst, int32_t imm32) { 5555 (void) prefixq_and_encode(dst->encoding()); 5556 emit_arith(0x81, 0xD8, dst, imm32); 5557 } 5558 5559 void Assembler::sbbq(Register dst, Address src) { 5560 InstructionMark im(this); 5561 prefixq(src, dst); 5562 emit_int8(0x1B); 5563 emit_operand(dst, src); 5564 } 5565 5566 void Assembler::sbbq(Register dst, Register src) { 5567 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5568 emit_arith(0x1B, 0xC0, dst, src); 5569 } 5570 5571 void Assembler::shlq(Register dst, int imm8) { 5572 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5573 int encode = prefixq_and_encode(dst->encoding()); 5574 if (imm8 == 1) { 5575 emit_int8((unsigned char)0xD1); 5576 emit_int8((unsigned char)(0xE0 | encode)); 5577 } else { 5578 emit_int8((unsigned char)0xC1); 5579 emit_int8((unsigned char)(0xE0 | encode)); 5580 emit_int8(imm8); 5581 } 5582 } 5583 5584 void Assembler::shlq(Register dst) { 5585 int encode = prefixq_and_encode(dst->encoding()); 5586 emit_int8((unsigned char)0xD3); 5587 emit_int8((unsigned char)(0xE0 | encode)); 5588 } 5589 5590 void Assembler::shrq(Register dst, int imm8) { 5591 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5592 int encode = prefixq_and_encode(dst->encoding()); 5593 emit_int8((unsigned char)0xC1); 5594 emit_int8((unsigned char)(0xE8 | encode)); 5595 emit_int8(imm8); 5596 } 5597 5598 void Assembler::shrq(Register dst) { 5599 int encode = prefixq_and_encode(dst->encoding()); 5600 emit_int8((unsigned char)0xD3); 5601 emit_int8(0xE8 | encode); 5602 } 5603 5604 void Assembler::subq(Address dst, int32_t imm32) { 5605 InstructionMark im(this); 5606 prefixq(dst); 5607 emit_arith_operand(0x81, rbp, dst, imm32); 5608 } 5609 5610 void Assembler::subq(Address dst, Register src) { 5611 InstructionMark im(this); 5612 prefixq(dst, src); 5613 emit_int8(0x29); 5614 emit_operand(src, dst); 5615 } 5616 5617 void Assembler::subq(Register dst, int32_t imm32) { 5618 (void) prefixq_and_encode(dst->encoding()); 5619 emit_arith(0x81, 0xE8, dst, imm32); 5620 } 5621 5622 // Force generation of a 4 byte immediate value even if it fits into 8bit 5623 void Assembler::subq_imm32(Register dst, int32_t imm32) { 5624 (void) prefixq_and_encode(dst->encoding()); 5625 emit_arith_imm32(0x81, 0xE8, dst, imm32); 5626 } 5627 5628 void Assembler::subq(Register dst, Address src) { 5629 InstructionMark im(this); 5630 prefixq(src, dst); 5631 emit_int8(0x2B); 5632 emit_operand(dst, src); 5633 } 5634 5635 void Assembler::subq(Register dst, Register src) { 5636 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5637 emit_arith(0x2B, 0xC0, dst, src); 5638 } 5639 5640 void Assembler::testq(Register dst, int32_t imm32) { 5641 // not using emit_arith because test 5642 // doesn't support sign-extension of 5643 // 8bit operands 5644 int encode = dst->encoding(); 5645 if (encode == 0) { 5646 prefix(REX_W); 5647 emit_int8((unsigned char)0xA9); 5648 } else { 5649 encode = prefixq_and_encode(encode); 5650 emit_int8((unsigned char)0xF7); 5651 emit_int8((unsigned char)(0xC0 | encode)); 5652 } 5653 emit_int32(imm32); 5654 } 5655 5656 void Assembler::testq(Register dst, Register src) { 5657 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5658 emit_arith(0x85, 0xC0, dst, src); 5659 } 5660 5661 void Assembler::xaddq(Address dst, Register src) { 5662 InstructionMark im(this); 5663 prefixq(dst, src); 5664 emit_int8(0x0F); 5665 emit_int8((unsigned char)0xC1); 5666 emit_operand(src, dst); 5667 } 5668 5669 void Assembler::xchgq(Register dst, Address src) { 5670 InstructionMark im(this); 5671 prefixq(src, dst); 5672 emit_int8((unsigned char)0x87); 5673 emit_operand(dst, src); 5674 } 5675 5676 void Assembler::xchgq(Register dst, Register src) { 5677 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5678 emit_int8((unsigned char)0x87); 5679 emit_int8((unsigned char)(0xc0 | encode)); 5680 } 5681 5682 void Assembler::xorq(Register dst, Register src) { 5683 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5684 emit_arith(0x33, 0xC0, dst, src); 5685 } 5686 5687 void Assembler::xorq(Register dst, Address src) { 5688 InstructionMark im(this); 5689 prefixq(src, dst); 5690 emit_int8(0x33); 5691 emit_operand(dst, src); 5692 } 5693 5694 #endif // !LP64