1 /* 2 * Copyright (c) 1997, 2015, 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, p2i(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, p2i(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 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1116 emit_int8(0x0F); 1117 emit_int8((unsigned char)0xBD); 1118 emit_int8((unsigned char)(0xC0 | encode)); 1119 } 1120 1121 void Assembler::bswapl(Register reg) { // bswap 1122 int encode = prefix_and_encode(reg->encoding()); 1123 emit_int8(0x0F); 1124 emit_int8((unsigned char)(0xC8 | encode)); 1125 } 1126 1127 void Assembler::blsil(Register dst, Register src) { 1128 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1129 int encode = vex_prefix_0F38_and_encode(rbx, dst, src); 1130 emit_int8((unsigned char)0xF3); 1131 emit_int8((unsigned char)(0xC0 | encode)); 1132 } 1133 1134 void Assembler::blsil(Register dst, Address src) { 1135 InstructionMark im(this); 1136 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1137 vex_prefix_0F38(rbx, dst, src); 1138 emit_int8((unsigned char)0xF3); 1139 emit_operand(rbx, src); 1140 } 1141 1142 void Assembler::blsmskl(Register dst, Register src) { 1143 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1144 int encode = vex_prefix_0F38_and_encode(rdx, dst, src); 1145 emit_int8((unsigned char)0xF3); 1146 emit_int8((unsigned char)(0xC0 | encode)); 1147 } 1148 1149 void Assembler::blsmskl(Register dst, Address src) { 1150 InstructionMark im(this); 1151 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1152 vex_prefix_0F38(rdx, dst, src); 1153 emit_int8((unsigned char)0xF3); 1154 emit_operand(rdx, src); 1155 } 1156 1157 void Assembler::blsrl(Register dst, Register src) { 1158 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1159 int encode = vex_prefix_0F38_and_encode(rcx, dst, src); 1160 emit_int8((unsigned char)0xF3); 1161 emit_int8((unsigned char)(0xC0 | encode)); 1162 } 1163 1164 void Assembler::blsrl(Register dst, Address src) { 1165 InstructionMark im(this); 1166 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 1167 vex_prefix_0F38(rcx, dst, src); 1168 emit_int8((unsigned char)0xF3); 1169 emit_operand(rcx, src); 1170 } 1171 1172 void Assembler::call(Label& L, relocInfo::relocType rtype) { 1173 // suspect disp32 is always good 1174 int operand = LP64_ONLY(disp32_operand) NOT_LP64(imm_operand); 1175 1176 if (L.is_bound()) { 1177 const int long_size = 5; 1178 int offs = (int)( target(L) - pc() ); 1179 assert(offs <= 0, "assembler error"); 1180 InstructionMark im(this); 1181 // 1110 1000 #32-bit disp 1182 emit_int8((unsigned char)0xE8); 1183 emit_data(offs - long_size, rtype, operand); 1184 } else { 1185 InstructionMark im(this); 1186 // 1110 1000 #32-bit disp 1187 L.add_patch_at(code(), locator()); 1188 1189 emit_int8((unsigned char)0xE8); 1190 emit_data(int(0), rtype, operand); 1191 } 1192 } 1193 1194 void Assembler::call(Register dst) { 1195 int encode = prefix_and_encode(dst->encoding()); 1196 emit_int8((unsigned char)0xFF); 1197 emit_int8((unsigned char)(0xD0 | encode)); 1198 } 1199 1200 1201 void Assembler::call(Address adr) { 1202 InstructionMark im(this); 1203 prefix(adr); 1204 emit_int8((unsigned char)0xFF); 1205 emit_operand(rdx, adr); 1206 } 1207 1208 void Assembler::call_literal(address entry, RelocationHolder const& rspec) { 1209 assert(entry != NULL, "call most probably wrong"); 1210 InstructionMark im(this); 1211 emit_int8((unsigned char)0xE8); 1212 intptr_t disp = entry - (pc() + sizeof(int32_t)); 1213 assert(is_simm32(disp), "must be 32bit offset (call2)"); 1214 // Technically, should use call32_operand, but this format is 1215 // implied by the fact that we're emitting a call instruction. 1216 1217 int operand = LP64_ONLY(disp32_operand) NOT_LP64(call32_operand); 1218 emit_data((int) disp, rspec, operand); 1219 } 1220 1221 void Assembler::cdql() { 1222 emit_int8((unsigned char)0x99); 1223 } 1224 1225 void Assembler::cld() { 1226 emit_int8((unsigned char)0xFC); 1227 } 1228 1229 void Assembler::cmovl(Condition cc, Register dst, Register src) { 1230 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction")); 1231 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1232 emit_int8(0x0F); 1233 emit_int8(0x40 | cc); 1234 emit_int8((unsigned char)(0xC0 | encode)); 1235 } 1236 1237 1238 void Assembler::cmovl(Condition cc, Register dst, Address src) { 1239 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction")); 1240 prefix(src, dst); 1241 emit_int8(0x0F); 1242 emit_int8(0x40 | cc); 1243 emit_operand(dst, src); 1244 } 1245 1246 void Assembler::cmpb(Address dst, int imm8) { 1247 InstructionMark im(this); 1248 prefix(dst); 1249 emit_int8((unsigned char)0x80); 1250 emit_operand(rdi, dst, 1); 1251 emit_int8(imm8); 1252 } 1253 1254 void Assembler::cmpl(Address dst, int32_t imm32) { 1255 InstructionMark im(this); 1256 prefix(dst); 1257 emit_int8((unsigned char)0x81); 1258 emit_operand(rdi, dst, 4); 1259 emit_int32(imm32); 1260 } 1261 1262 void Assembler::cmpl(Register dst, int32_t imm32) { 1263 prefix(dst); 1264 emit_arith(0x81, 0xF8, dst, imm32); 1265 } 1266 1267 void Assembler::cmpl(Register dst, Register src) { 1268 (void) prefix_and_encode(dst->encoding(), src->encoding()); 1269 emit_arith(0x3B, 0xC0, dst, src); 1270 } 1271 1272 1273 void Assembler::cmpl(Register dst, Address src) { 1274 InstructionMark im(this); 1275 prefix(src, dst); 1276 emit_int8((unsigned char)0x3B); 1277 emit_operand(dst, src); 1278 } 1279 1280 void Assembler::cmpw(Address dst, int imm16) { 1281 InstructionMark im(this); 1282 assert(!dst.base_needs_rex() && !dst.index_needs_rex(), "no extended registers"); 1283 emit_int8(0x66); 1284 emit_int8((unsigned char)0x81); 1285 emit_operand(rdi, dst, 2); 1286 emit_int16(imm16); 1287 } 1288 1289 // The 32-bit cmpxchg compares the value at adr with the contents of rax, 1290 // and stores reg into adr if so; otherwise, the value at adr is loaded into rax,. 1291 // The ZF is set if the compared values were equal, and cleared otherwise. 1292 void Assembler::cmpxchgl(Register reg, Address adr) { // cmpxchg 1293 InstructionMark im(this); 1294 prefix(adr, reg); 1295 emit_int8(0x0F); 1296 emit_int8((unsigned char)0xB1); 1297 emit_operand(reg, adr); 1298 } 1299 1300 // The 8-bit cmpxchg compares the value at adr with the contents of rax, 1301 // and stores reg into adr if so; otherwise, the value at adr is loaded into rax,. 1302 // The ZF is set if the compared values were equal, and cleared otherwise. 1303 void Assembler::cmpxchgb(Register reg, Address adr) { // cmpxchg 1304 InstructionMark im(this); 1305 prefix(adr, reg, true); 1306 emit_int8(0x0F); 1307 emit_int8((unsigned char)0xB0); 1308 emit_operand(reg, adr); 1309 } 1310 1311 void Assembler::comisd(XMMRegister dst, Address src) { 1312 // NOTE: dbx seems to decode this as comiss even though the 1313 // 0x66 is there. Strangly ucomisd comes out correct 1314 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1315 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66); 1316 } 1317 1318 void Assembler::comisd(XMMRegister dst, XMMRegister src) { 1319 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1320 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66); 1321 } 1322 1323 void Assembler::comiss(XMMRegister dst, Address src) { 1324 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1325 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE); 1326 } 1327 1328 void Assembler::comiss(XMMRegister dst, XMMRegister src) { 1329 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1330 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE); 1331 } 1332 1333 void Assembler::cpuid() { 1334 emit_int8(0x0F); 1335 emit_int8((unsigned char)0xA2); 1336 } 1337 1338 void Assembler::cvtdq2pd(XMMRegister dst, XMMRegister src) { 1339 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1340 emit_simd_arith_nonds(0xE6, dst, src, VEX_SIMD_F3); 1341 } 1342 1343 void Assembler::cvtdq2ps(XMMRegister dst, XMMRegister src) { 1344 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1345 emit_simd_arith_nonds(0x5B, dst, src, VEX_SIMD_NONE); 1346 } 1347 1348 void Assembler::cvtsd2ss(XMMRegister dst, XMMRegister src) { 1349 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1350 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2); 1351 } 1352 1353 void Assembler::cvtsd2ss(XMMRegister dst, Address src) { 1354 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1355 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2); 1356 } 1357 1358 void Assembler::cvtsi2sdl(XMMRegister dst, Register src) { 1359 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1360 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2); 1361 emit_int8(0x2A); 1362 emit_int8((unsigned char)(0xC0 | encode)); 1363 } 1364 1365 void Assembler::cvtsi2sdl(XMMRegister dst, Address src) { 1366 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1367 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F2); 1368 } 1369 1370 void Assembler::cvtsi2ssl(XMMRegister dst, Register src) { 1371 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1372 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3); 1373 emit_int8(0x2A); 1374 emit_int8((unsigned char)(0xC0 | encode)); 1375 } 1376 1377 void Assembler::cvtsi2ssl(XMMRegister dst, Address src) { 1378 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1379 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F3); 1380 } 1381 1382 void Assembler::cvtss2sd(XMMRegister dst, XMMRegister src) { 1383 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1384 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3); 1385 } 1386 1387 void Assembler::cvtss2sd(XMMRegister dst, Address src) { 1388 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1389 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3); 1390 } 1391 1392 1393 void Assembler::cvttsd2sil(Register dst, XMMRegister src) { 1394 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1395 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F2); 1396 emit_int8(0x2C); 1397 emit_int8((unsigned char)(0xC0 | encode)); 1398 } 1399 1400 void Assembler::cvttss2sil(Register dst, XMMRegister src) { 1401 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1402 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F3); 1403 emit_int8(0x2C); 1404 emit_int8((unsigned char)(0xC0 | encode)); 1405 } 1406 1407 void Assembler::decl(Address dst) { 1408 // Don't use it directly. Use MacroAssembler::decrement() instead. 1409 InstructionMark im(this); 1410 prefix(dst); 1411 emit_int8((unsigned char)0xFF); 1412 emit_operand(rcx, dst); 1413 } 1414 1415 void Assembler::divsd(XMMRegister dst, Address src) { 1416 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1417 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2); 1418 } 1419 1420 void Assembler::divsd(XMMRegister dst, XMMRegister src) { 1421 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1422 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2); 1423 } 1424 1425 void Assembler::divss(XMMRegister dst, Address src) { 1426 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1427 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3); 1428 } 1429 1430 void Assembler::divss(XMMRegister dst, XMMRegister src) { 1431 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1432 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3); 1433 } 1434 1435 void Assembler::emms() { 1436 NOT_LP64(assert(VM_Version::supports_mmx(), "")); 1437 emit_int8(0x0F); 1438 emit_int8(0x77); 1439 } 1440 1441 void Assembler::hlt() { 1442 emit_int8((unsigned char)0xF4); 1443 } 1444 1445 void Assembler::idivl(Register src) { 1446 int encode = prefix_and_encode(src->encoding()); 1447 emit_int8((unsigned char)0xF7); 1448 emit_int8((unsigned char)(0xF8 | encode)); 1449 } 1450 1451 void Assembler::divl(Register src) { // Unsigned 1452 int encode = prefix_and_encode(src->encoding()); 1453 emit_int8((unsigned char)0xF7); 1454 emit_int8((unsigned char)(0xF0 | encode)); 1455 } 1456 1457 void Assembler::imull(Register dst, Register src) { 1458 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1459 emit_int8(0x0F); 1460 emit_int8((unsigned char)0xAF); 1461 emit_int8((unsigned char)(0xC0 | encode)); 1462 } 1463 1464 1465 void Assembler::imull(Register dst, Register src, int value) { 1466 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1467 if (is8bit(value)) { 1468 emit_int8(0x6B); 1469 emit_int8((unsigned char)(0xC0 | encode)); 1470 emit_int8(value & 0xFF); 1471 } else { 1472 emit_int8(0x69); 1473 emit_int8((unsigned char)(0xC0 | encode)); 1474 emit_int32(value); 1475 } 1476 } 1477 1478 void Assembler::imull(Register dst, Address src) { 1479 InstructionMark im(this); 1480 prefix(src, dst); 1481 emit_int8(0x0F); 1482 emit_int8((unsigned char) 0xAF); 1483 emit_operand(dst, src); 1484 } 1485 1486 1487 void Assembler::incl(Address dst) { 1488 // Don't use it directly. Use MacroAssembler::increment() instead. 1489 InstructionMark im(this); 1490 prefix(dst); 1491 emit_int8((unsigned char)0xFF); 1492 emit_operand(rax, dst); 1493 } 1494 1495 void Assembler::jcc(Condition cc, Label& L, bool maybe_short) { 1496 InstructionMark im(this); 1497 assert((0 <= cc) && (cc < 16), "illegal cc"); 1498 if (L.is_bound()) { 1499 address dst = target(L); 1500 assert(dst != NULL, "jcc most probably wrong"); 1501 1502 const int short_size = 2; 1503 const int long_size = 6; 1504 intptr_t offs = (intptr_t)dst - (intptr_t)pc(); 1505 if (maybe_short && is8bit(offs - short_size)) { 1506 // 0111 tttn #8-bit disp 1507 emit_int8(0x70 | cc); 1508 emit_int8((offs - short_size) & 0xFF); 1509 } else { 1510 // 0000 1111 1000 tttn #32-bit disp 1511 assert(is_simm32(offs - long_size), 1512 "must be 32bit offset (call4)"); 1513 emit_int8(0x0F); 1514 emit_int8((unsigned char)(0x80 | cc)); 1515 emit_int32(offs - long_size); 1516 } 1517 } else { 1518 // Note: could eliminate cond. jumps to this jump if condition 1519 // is the same however, seems to be rather unlikely case. 1520 // Note: use jccb() if label to be bound is very close to get 1521 // an 8-bit displacement 1522 L.add_patch_at(code(), locator()); 1523 emit_int8(0x0F); 1524 emit_int8((unsigned char)(0x80 | cc)); 1525 emit_int32(0); 1526 } 1527 } 1528 1529 void Assembler::jccb(Condition cc, Label& L) { 1530 if (L.is_bound()) { 1531 const int short_size = 2; 1532 address entry = target(L); 1533 #ifdef ASSERT 1534 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size); 1535 intptr_t delta = short_branch_delta(); 1536 if (delta != 0) { 1537 dist += (dist < 0 ? (-delta) :delta); 1538 } 1539 assert(is8bit(dist), "Dispacement too large for a short jmp"); 1540 #endif 1541 intptr_t offs = (intptr_t)entry - (intptr_t)pc(); 1542 // 0111 tttn #8-bit disp 1543 emit_int8(0x70 | cc); 1544 emit_int8((offs - short_size) & 0xFF); 1545 } else { 1546 InstructionMark im(this); 1547 L.add_patch_at(code(), locator()); 1548 emit_int8(0x70 | cc); 1549 emit_int8(0); 1550 } 1551 } 1552 1553 void Assembler::jmp(Address adr) { 1554 InstructionMark im(this); 1555 prefix(adr); 1556 emit_int8((unsigned char)0xFF); 1557 emit_operand(rsp, adr); 1558 } 1559 1560 void Assembler::jmp(Label& L, bool maybe_short) { 1561 if (L.is_bound()) { 1562 address entry = target(L); 1563 assert(entry != NULL, "jmp most probably wrong"); 1564 InstructionMark im(this); 1565 const int short_size = 2; 1566 const int long_size = 5; 1567 intptr_t offs = entry - pc(); 1568 if (maybe_short && is8bit(offs - short_size)) { 1569 emit_int8((unsigned char)0xEB); 1570 emit_int8((offs - short_size) & 0xFF); 1571 } else { 1572 emit_int8((unsigned char)0xE9); 1573 emit_int32(offs - long_size); 1574 } 1575 } else { 1576 // By default, forward jumps are always 32-bit displacements, since 1577 // we can't yet know where the label will be bound. If you're sure that 1578 // the forward jump will not run beyond 256 bytes, use jmpb to 1579 // force an 8-bit displacement. 1580 InstructionMark im(this); 1581 L.add_patch_at(code(), locator()); 1582 emit_int8((unsigned char)0xE9); 1583 emit_int32(0); 1584 } 1585 } 1586 1587 void Assembler::jmp(Register entry) { 1588 int encode = prefix_and_encode(entry->encoding()); 1589 emit_int8((unsigned char)0xFF); 1590 emit_int8((unsigned char)(0xE0 | encode)); 1591 } 1592 1593 void Assembler::jmp_literal(address dest, RelocationHolder const& rspec) { 1594 InstructionMark im(this); 1595 emit_int8((unsigned char)0xE9); 1596 assert(dest != NULL, "must have a target"); 1597 intptr_t disp = dest - (pc() + sizeof(int32_t)); 1598 assert(is_simm32(disp), "must be 32bit offset (jmp)"); 1599 emit_data(disp, rspec.reloc(), call32_operand); 1600 } 1601 1602 void Assembler::jmpb(Label& L) { 1603 if (L.is_bound()) { 1604 const int short_size = 2; 1605 address entry = target(L); 1606 assert(entry != NULL, "jmp most probably wrong"); 1607 #ifdef ASSERT 1608 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size); 1609 intptr_t delta = short_branch_delta(); 1610 if (delta != 0) { 1611 dist += (dist < 0 ? (-delta) :delta); 1612 } 1613 assert(is8bit(dist), "Dispacement too large for a short jmp"); 1614 #endif 1615 intptr_t offs = entry - pc(); 1616 emit_int8((unsigned char)0xEB); 1617 emit_int8((offs - short_size) & 0xFF); 1618 } else { 1619 InstructionMark im(this); 1620 L.add_patch_at(code(), locator()); 1621 emit_int8((unsigned char)0xEB); 1622 emit_int8(0); 1623 } 1624 } 1625 1626 void Assembler::ldmxcsr( Address src) { 1627 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1628 InstructionMark im(this); 1629 prefix(src); 1630 emit_int8(0x0F); 1631 emit_int8((unsigned char)0xAE); 1632 emit_operand(as_Register(2), src); 1633 } 1634 1635 void Assembler::leal(Register dst, Address src) { 1636 InstructionMark im(this); 1637 #ifdef _LP64 1638 emit_int8(0x67); // addr32 1639 prefix(src, dst); 1640 #endif // LP64 1641 emit_int8((unsigned char)0x8D); 1642 emit_operand(dst, src); 1643 } 1644 1645 void Assembler::lfence() { 1646 emit_int8(0x0F); 1647 emit_int8((unsigned char)0xAE); 1648 emit_int8((unsigned char)0xE8); 1649 } 1650 1651 void Assembler::lock() { 1652 emit_int8((unsigned char)0xF0); 1653 } 1654 1655 void Assembler::lzcntl(Register dst, Register src) { 1656 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR"); 1657 emit_int8((unsigned char)0xF3); 1658 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1659 emit_int8(0x0F); 1660 emit_int8((unsigned char)0xBD); 1661 emit_int8((unsigned char)(0xC0 | encode)); 1662 } 1663 1664 // Emit mfence instruction 1665 void Assembler::mfence() { 1666 NOT_LP64(assert(VM_Version::supports_sse2(), "unsupported");) 1667 emit_int8(0x0F); 1668 emit_int8((unsigned char)0xAE); 1669 emit_int8((unsigned char)0xF0); 1670 } 1671 1672 void Assembler::mov(Register dst, Register src) { 1673 LP64_ONLY(movq(dst, src)) NOT_LP64(movl(dst, src)); 1674 } 1675 1676 void Assembler::movapd(XMMRegister dst, XMMRegister src) { 1677 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1678 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_66); 1679 } 1680 1681 void Assembler::movaps(XMMRegister dst, XMMRegister src) { 1682 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1683 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_NONE); 1684 } 1685 1686 void Assembler::movlhps(XMMRegister dst, XMMRegister src) { 1687 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1688 int encode = simd_prefix_and_encode(dst, src, src, VEX_SIMD_NONE); 1689 emit_int8(0x16); 1690 emit_int8((unsigned char)(0xC0 | encode)); 1691 } 1692 1693 void Assembler::movb(Register dst, Address src) { 1694 NOT_LP64(assert(dst->has_byte_register(), "must have byte register")); 1695 InstructionMark im(this); 1696 prefix(src, dst, true); 1697 emit_int8((unsigned char)0x8A); 1698 emit_operand(dst, src); 1699 } 1700 1701 1702 void Assembler::movb(Address dst, int imm8) { 1703 InstructionMark im(this); 1704 prefix(dst); 1705 emit_int8((unsigned char)0xC6); 1706 emit_operand(rax, dst, 1); 1707 emit_int8(imm8); 1708 } 1709 1710 1711 void Assembler::movb(Address dst, Register src) { 1712 assert(src->has_byte_register(), "must have byte register"); 1713 InstructionMark im(this); 1714 prefix(dst, src, true); 1715 emit_int8((unsigned char)0x88); 1716 emit_operand(src, dst); 1717 } 1718 1719 void Assembler::movdl(XMMRegister dst, Register src) { 1720 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1721 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_66); 1722 emit_int8(0x6E); 1723 emit_int8((unsigned char)(0xC0 | encode)); 1724 } 1725 1726 void Assembler::movdl(Register dst, XMMRegister src) { 1727 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1728 // swap src/dst to get correct prefix 1729 int encode = simd_prefix_and_encode(src, dst, VEX_SIMD_66); 1730 emit_int8(0x7E); 1731 emit_int8((unsigned char)(0xC0 | encode)); 1732 } 1733 1734 void Assembler::movdl(XMMRegister dst, Address src) { 1735 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1736 InstructionMark im(this); 1737 simd_prefix(dst, src, VEX_SIMD_66); 1738 emit_int8(0x6E); 1739 emit_operand(dst, src); 1740 } 1741 1742 void Assembler::movdl(Address dst, XMMRegister src) { 1743 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1744 InstructionMark im(this); 1745 simd_prefix(dst, src, VEX_SIMD_66); 1746 emit_int8(0x7E); 1747 emit_operand(src, dst); 1748 } 1749 1750 void Assembler::movdqa(XMMRegister dst, XMMRegister src) { 1751 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1752 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_66); 1753 } 1754 1755 void Assembler::movdqa(XMMRegister dst, Address src) { 1756 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1757 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_66); 1758 } 1759 1760 void Assembler::movdqu(XMMRegister dst, Address src) { 1761 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1762 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3); 1763 } 1764 1765 void Assembler::movdqu(XMMRegister dst, XMMRegister src) { 1766 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1767 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3); 1768 } 1769 1770 void Assembler::movdqu(Address dst, XMMRegister src) { 1771 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1772 InstructionMark im(this); 1773 simd_prefix(dst, src, VEX_SIMD_F3); 1774 emit_int8(0x7F); 1775 emit_operand(src, dst); 1776 } 1777 1778 // Move Unaligned 256bit Vector 1779 void Assembler::vmovdqu(XMMRegister dst, XMMRegister src) { 1780 assert(UseAVX > 0, ""); 1781 bool vector256 = true; 1782 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_F3, vector256); 1783 emit_int8(0x6F); 1784 emit_int8((unsigned char)(0xC0 | encode)); 1785 } 1786 1787 void Assembler::vmovdqu(XMMRegister dst, Address src) { 1788 assert(UseAVX > 0, ""); 1789 InstructionMark im(this); 1790 bool vector256 = true; 1791 vex_prefix(dst, xnoreg, src, VEX_SIMD_F3, vector256); 1792 emit_int8(0x6F); 1793 emit_operand(dst, src); 1794 } 1795 1796 void Assembler::vmovdqu(Address dst, XMMRegister src) { 1797 assert(UseAVX > 0, ""); 1798 InstructionMark im(this); 1799 bool vector256 = true; 1800 // swap src<->dst for encoding 1801 assert(src != xnoreg, "sanity"); 1802 vex_prefix(src, xnoreg, dst, VEX_SIMD_F3, vector256); 1803 emit_int8(0x7F); 1804 emit_operand(src, dst); 1805 } 1806 1807 // Uses zero extension on 64bit 1808 1809 void Assembler::movl(Register dst, int32_t imm32) { 1810 int encode = prefix_and_encode(dst->encoding()); 1811 emit_int8((unsigned char)(0xB8 | encode)); 1812 emit_int32(imm32); 1813 } 1814 1815 void Assembler::movl(Register dst, Register src) { 1816 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1817 emit_int8((unsigned char)0x8B); 1818 emit_int8((unsigned char)(0xC0 | encode)); 1819 } 1820 1821 void Assembler::movl(Register dst, Address src) { 1822 InstructionMark im(this); 1823 prefix(src, dst); 1824 emit_int8((unsigned char)0x8B); 1825 emit_operand(dst, src); 1826 } 1827 1828 void Assembler::movl(Address dst, int32_t imm32) { 1829 InstructionMark im(this); 1830 prefix(dst); 1831 emit_int8((unsigned char)0xC7); 1832 emit_operand(rax, dst, 4); 1833 emit_int32(imm32); 1834 } 1835 1836 void Assembler::movl(Address dst, Register src) { 1837 InstructionMark im(this); 1838 prefix(dst, src); 1839 emit_int8((unsigned char)0x89); 1840 emit_operand(src, dst); 1841 } 1842 1843 // New cpus require to use movsd and movss to avoid partial register stall 1844 // when loading from memory. But for old Opteron use movlpd instead of movsd. 1845 // The selection is done in MacroAssembler::movdbl() and movflt(). 1846 void Assembler::movlpd(XMMRegister dst, Address src) { 1847 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1848 emit_simd_arith(0x12, dst, src, VEX_SIMD_66); 1849 } 1850 1851 void Assembler::movq( MMXRegister dst, Address src ) { 1852 assert( VM_Version::supports_mmx(), "" ); 1853 emit_int8(0x0F); 1854 emit_int8(0x6F); 1855 emit_operand(dst, src); 1856 } 1857 1858 void Assembler::movq( Address dst, MMXRegister src ) { 1859 assert( VM_Version::supports_mmx(), "" ); 1860 emit_int8(0x0F); 1861 emit_int8(0x7F); 1862 // workaround gcc (3.2.1-7a) bug 1863 // In that version of gcc with only an emit_operand(MMX, Address) 1864 // gcc will tail jump and try and reverse the parameters completely 1865 // obliterating dst in the process. By having a version available 1866 // that doesn't need to swap the args at the tail jump the bug is 1867 // avoided. 1868 emit_operand(dst, src); 1869 } 1870 1871 void Assembler::movq(XMMRegister dst, Address src) { 1872 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1873 InstructionMark im(this); 1874 simd_prefix(dst, src, VEX_SIMD_F3); 1875 emit_int8(0x7E); 1876 emit_operand(dst, src); 1877 } 1878 1879 void Assembler::movq(Address dst, XMMRegister src) { 1880 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1881 InstructionMark im(this); 1882 simd_prefix(dst, src, VEX_SIMD_66); 1883 emit_int8((unsigned char)0xD6); 1884 emit_operand(src, dst); 1885 } 1886 1887 void Assembler::movsbl(Register dst, Address src) { // movsxb 1888 InstructionMark im(this); 1889 prefix(src, dst); 1890 emit_int8(0x0F); 1891 emit_int8((unsigned char)0xBE); 1892 emit_operand(dst, src); 1893 } 1894 1895 void Assembler::movsbl(Register dst, Register src) { // movsxb 1896 NOT_LP64(assert(src->has_byte_register(), "must have byte register")); 1897 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true); 1898 emit_int8(0x0F); 1899 emit_int8((unsigned char)0xBE); 1900 emit_int8((unsigned char)(0xC0 | encode)); 1901 } 1902 1903 void Assembler::movsd(XMMRegister dst, XMMRegister src) { 1904 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1905 emit_simd_arith(0x10, dst, src, VEX_SIMD_F2); 1906 } 1907 1908 void Assembler::movsd(XMMRegister dst, Address src) { 1909 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1910 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F2); 1911 } 1912 1913 void Assembler::movsd(Address dst, XMMRegister src) { 1914 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 1915 InstructionMark im(this); 1916 simd_prefix(dst, src, VEX_SIMD_F2); 1917 emit_int8(0x11); 1918 emit_operand(src, dst); 1919 } 1920 1921 void Assembler::movss(XMMRegister dst, XMMRegister src) { 1922 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1923 emit_simd_arith(0x10, dst, src, VEX_SIMD_F3); 1924 } 1925 1926 void Assembler::movss(XMMRegister dst, Address src) { 1927 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1928 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F3); 1929 } 1930 1931 void Assembler::movss(Address dst, XMMRegister src) { 1932 NOT_LP64(assert(VM_Version::supports_sse(), "")); 1933 InstructionMark im(this); 1934 simd_prefix(dst, src, VEX_SIMD_F3); 1935 emit_int8(0x11); 1936 emit_operand(src, dst); 1937 } 1938 1939 void Assembler::movswl(Register dst, Address src) { // movsxw 1940 InstructionMark im(this); 1941 prefix(src, dst); 1942 emit_int8(0x0F); 1943 emit_int8((unsigned char)0xBF); 1944 emit_operand(dst, src); 1945 } 1946 1947 void Assembler::movswl(Register dst, Register src) { // movsxw 1948 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 1949 emit_int8(0x0F); 1950 emit_int8((unsigned char)0xBF); 1951 emit_int8((unsigned char)(0xC0 | encode)); 1952 } 1953 1954 void Assembler::movw(Address dst, int imm16) { 1955 InstructionMark im(this); 1956 1957 emit_int8(0x66); // switch to 16-bit mode 1958 prefix(dst); 1959 emit_int8((unsigned char)0xC7); 1960 emit_operand(rax, dst, 2); 1961 emit_int16(imm16); 1962 } 1963 1964 void Assembler::movw(Register dst, Address src) { 1965 InstructionMark im(this); 1966 emit_int8(0x66); 1967 prefix(src, dst); 1968 emit_int8((unsigned char)0x8B); 1969 emit_operand(dst, src); 1970 } 1971 1972 void Assembler::movw(Address dst, Register src) { 1973 InstructionMark im(this); 1974 emit_int8(0x66); 1975 prefix(dst, src); 1976 emit_int8((unsigned char)0x89); 1977 emit_operand(src, dst); 1978 } 1979 1980 void Assembler::movzbl(Register dst, Address src) { // movzxb 1981 InstructionMark im(this); 1982 prefix(src, dst); 1983 emit_int8(0x0F); 1984 emit_int8((unsigned char)0xB6); 1985 emit_operand(dst, src); 1986 } 1987 1988 void Assembler::movzbl(Register dst, Register src) { // movzxb 1989 NOT_LP64(assert(src->has_byte_register(), "must have byte register")); 1990 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true); 1991 emit_int8(0x0F); 1992 emit_int8((unsigned char)0xB6); 1993 emit_int8(0xC0 | encode); 1994 } 1995 1996 void Assembler::movzwl(Register dst, Address src) { // movzxw 1997 InstructionMark im(this); 1998 prefix(src, dst); 1999 emit_int8(0x0F); 2000 emit_int8((unsigned char)0xB7); 2001 emit_operand(dst, src); 2002 } 2003 2004 void Assembler::movzwl(Register dst, Register src) { // movzxw 2005 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2006 emit_int8(0x0F); 2007 emit_int8((unsigned char)0xB7); 2008 emit_int8(0xC0 | encode); 2009 } 2010 2011 void Assembler::mull(Address src) { 2012 InstructionMark im(this); 2013 prefix(src); 2014 emit_int8((unsigned char)0xF7); 2015 emit_operand(rsp, src); 2016 } 2017 2018 void Assembler::mull(Register src) { 2019 int encode = prefix_and_encode(src->encoding()); 2020 emit_int8((unsigned char)0xF7); 2021 emit_int8((unsigned char)(0xE0 | encode)); 2022 } 2023 2024 void Assembler::mulsd(XMMRegister dst, Address src) { 2025 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2026 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2); 2027 } 2028 2029 void Assembler::mulsd(XMMRegister dst, XMMRegister src) { 2030 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2031 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2); 2032 } 2033 2034 void Assembler::mulss(XMMRegister dst, Address src) { 2035 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2036 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3); 2037 } 2038 2039 void Assembler::mulss(XMMRegister dst, XMMRegister src) { 2040 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2041 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3); 2042 } 2043 2044 void Assembler::negl(Register dst) { 2045 int encode = prefix_and_encode(dst->encoding()); 2046 emit_int8((unsigned char)0xF7); 2047 emit_int8((unsigned char)(0xD8 | encode)); 2048 } 2049 2050 void Assembler::nop(int i) { 2051 #ifdef ASSERT 2052 assert(i > 0, " "); 2053 // The fancy nops aren't currently recognized by debuggers making it a 2054 // pain to disassemble code while debugging. If asserts are on clearly 2055 // speed is not an issue so simply use the single byte traditional nop 2056 // to do alignment. 2057 2058 for (; i > 0 ; i--) emit_int8((unsigned char)0x90); 2059 return; 2060 2061 #endif // ASSERT 2062 2063 if (UseAddressNop && VM_Version::is_intel()) { 2064 // 2065 // Using multi-bytes nops "0x0F 0x1F [address]" for Intel 2066 // 1: 0x90 2067 // 2: 0x66 0x90 2068 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding) 2069 // 4: 0x0F 0x1F 0x40 0x00 2070 // 5: 0x0F 0x1F 0x44 0x00 0x00 2071 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00 2072 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2073 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2074 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2075 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2076 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2077 2078 // The rest coding is Intel specific - don't use consecutive address nops 2079 2080 // 12: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2081 // 13: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2082 // 14: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2083 // 15: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90 2084 2085 while(i >= 15) { 2086 // For Intel don't generate consecutive addess nops (mix with regular nops) 2087 i -= 15; 2088 emit_int8(0x66); // size prefix 2089 emit_int8(0x66); // size prefix 2090 emit_int8(0x66); // size prefix 2091 addr_nop_8(); 2092 emit_int8(0x66); // size prefix 2093 emit_int8(0x66); // size prefix 2094 emit_int8(0x66); // size prefix 2095 emit_int8((unsigned char)0x90); 2096 // nop 2097 } 2098 switch (i) { 2099 case 14: 2100 emit_int8(0x66); // size prefix 2101 case 13: 2102 emit_int8(0x66); // size prefix 2103 case 12: 2104 addr_nop_8(); 2105 emit_int8(0x66); // size prefix 2106 emit_int8(0x66); // size prefix 2107 emit_int8(0x66); // size prefix 2108 emit_int8((unsigned char)0x90); 2109 // nop 2110 break; 2111 case 11: 2112 emit_int8(0x66); // size prefix 2113 case 10: 2114 emit_int8(0x66); // size prefix 2115 case 9: 2116 emit_int8(0x66); // size prefix 2117 case 8: 2118 addr_nop_8(); 2119 break; 2120 case 7: 2121 addr_nop_7(); 2122 break; 2123 case 6: 2124 emit_int8(0x66); // size prefix 2125 case 5: 2126 addr_nop_5(); 2127 break; 2128 case 4: 2129 addr_nop_4(); 2130 break; 2131 case 3: 2132 // Don't use "0x0F 0x1F 0x00" - need patching safe padding 2133 emit_int8(0x66); // size prefix 2134 case 2: 2135 emit_int8(0x66); // size prefix 2136 case 1: 2137 emit_int8((unsigned char)0x90); 2138 // nop 2139 break; 2140 default: 2141 assert(i == 0, " "); 2142 } 2143 return; 2144 } 2145 if (UseAddressNop && VM_Version::is_amd()) { 2146 // 2147 // Using multi-bytes nops "0x0F 0x1F [address]" for AMD. 2148 // 1: 0x90 2149 // 2: 0x66 0x90 2150 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding) 2151 // 4: 0x0F 0x1F 0x40 0x00 2152 // 5: 0x0F 0x1F 0x44 0x00 0x00 2153 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00 2154 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2155 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2156 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2157 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2158 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2159 2160 // The rest coding is AMD specific - use consecutive address nops 2161 2162 // 12: 0x66 0x0F 0x1F 0x44 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00 2163 // 13: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00 2164 // 14: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2165 // 15: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 2166 // 16: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 2167 // Size prefixes (0x66) are added for larger sizes 2168 2169 while(i >= 22) { 2170 i -= 11; 2171 emit_int8(0x66); // size prefix 2172 emit_int8(0x66); // size prefix 2173 emit_int8(0x66); // size prefix 2174 addr_nop_8(); 2175 } 2176 // Generate first nop for size between 21-12 2177 switch (i) { 2178 case 21: 2179 i -= 1; 2180 emit_int8(0x66); // size prefix 2181 case 20: 2182 case 19: 2183 i -= 1; 2184 emit_int8(0x66); // size prefix 2185 case 18: 2186 case 17: 2187 i -= 1; 2188 emit_int8(0x66); // size prefix 2189 case 16: 2190 case 15: 2191 i -= 8; 2192 addr_nop_8(); 2193 break; 2194 case 14: 2195 case 13: 2196 i -= 7; 2197 addr_nop_7(); 2198 break; 2199 case 12: 2200 i -= 6; 2201 emit_int8(0x66); // size prefix 2202 addr_nop_5(); 2203 break; 2204 default: 2205 assert(i < 12, " "); 2206 } 2207 2208 // Generate second nop for size between 11-1 2209 switch (i) { 2210 case 11: 2211 emit_int8(0x66); // size prefix 2212 case 10: 2213 emit_int8(0x66); // size prefix 2214 case 9: 2215 emit_int8(0x66); // size prefix 2216 case 8: 2217 addr_nop_8(); 2218 break; 2219 case 7: 2220 addr_nop_7(); 2221 break; 2222 case 6: 2223 emit_int8(0x66); // size prefix 2224 case 5: 2225 addr_nop_5(); 2226 break; 2227 case 4: 2228 addr_nop_4(); 2229 break; 2230 case 3: 2231 // Don't use "0x0F 0x1F 0x00" - need patching safe padding 2232 emit_int8(0x66); // size prefix 2233 case 2: 2234 emit_int8(0x66); // size prefix 2235 case 1: 2236 emit_int8((unsigned char)0x90); 2237 // nop 2238 break; 2239 default: 2240 assert(i == 0, " "); 2241 } 2242 return; 2243 } 2244 2245 // Using nops with size prefixes "0x66 0x90". 2246 // From AMD Optimization Guide: 2247 // 1: 0x90 2248 // 2: 0x66 0x90 2249 // 3: 0x66 0x66 0x90 2250 // 4: 0x66 0x66 0x66 0x90 2251 // 5: 0x66 0x66 0x90 0x66 0x90 2252 // 6: 0x66 0x66 0x90 0x66 0x66 0x90 2253 // 7: 0x66 0x66 0x66 0x90 0x66 0x66 0x90 2254 // 8: 0x66 0x66 0x66 0x90 0x66 0x66 0x66 0x90 2255 // 9: 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90 2256 // 10: 0x66 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90 2257 // 2258 while(i > 12) { 2259 i -= 4; 2260 emit_int8(0x66); // size prefix 2261 emit_int8(0x66); 2262 emit_int8(0x66); 2263 emit_int8((unsigned char)0x90); 2264 // nop 2265 } 2266 // 1 - 12 nops 2267 if(i > 8) { 2268 if(i > 9) { 2269 i -= 1; 2270 emit_int8(0x66); 2271 } 2272 i -= 3; 2273 emit_int8(0x66); 2274 emit_int8(0x66); 2275 emit_int8((unsigned char)0x90); 2276 } 2277 // 1 - 8 nops 2278 if(i > 4) { 2279 if(i > 6) { 2280 i -= 1; 2281 emit_int8(0x66); 2282 } 2283 i -= 3; 2284 emit_int8(0x66); 2285 emit_int8(0x66); 2286 emit_int8((unsigned char)0x90); 2287 } 2288 switch (i) { 2289 case 4: 2290 emit_int8(0x66); 2291 case 3: 2292 emit_int8(0x66); 2293 case 2: 2294 emit_int8(0x66); 2295 case 1: 2296 emit_int8((unsigned char)0x90); 2297 break; 2298 default: 2299 assert(i == 0, " "); 2300 } 2301 } 2302 2303 void Assembler::notl(Register dst) { 2304 int encode = prefix_and_encode(dst->encoding()); 2305 emit_int8((unsigned char)0xF7); 2306 emit_int8((unsigned char)(0xD0 | encode)); 2307 } 2308 2309 void Assembler::orl(Address dst, int32_t imm32) { 2310 InstructionMark im(this); 2311 prefix(dst); 2312 emit_arith_operand(0x81, rcx, dst, imm32); 2313 } 2314 2315 void Assembler::orl(Register dst, int32_t imm32) { 2316 prefix(dst); 2317 emit_arith(0x81, 0xC8, dst, imm32); 2318 } 2319 2320 void Assembler::orl(Register dst, Address src) { 2321 InstructionMark im(this); 2322 prefix(src, dst); 2323 emit_int8(0x0B); 2324 emit_operand(dst, src); 2325 } 2326 2327 void Assembler::orl(Register dst, Register src) { 2328 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2329 emit_arith(0x0B, 0xC0, dst, src); 2330 } 2331 2332 void Assembler::packuswb(XMMRegister dst, Address src) { 2333 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2334 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2335 emit_simd_arith(0x67, dst, src, VEX_SIMD_66); 2336 } 2337 2338 void Assembler::packuswb(XMMRegister dst, XMMRegister src) { 2339 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2340 emit_simd_arith(0x67, dst, src, VEX_SIMD_66); 2341 } 2342 2343 void Assembler::vpackuswb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 2344 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 2345 emit_vex_arith(0x67, dst, nds, src, VEX_SIMD_66, vector256); 2346 } 2347 2348 void Assembler::vpermq(XMMRegister dst, XMMRegister src, int imm8, bool vector256) { 2349 assert(VM_Version::supports_avx2(), ""); 2350 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, true, vector256); 2351 emit_int8(0x00); 2352 emit_int8(0xC0 | encode); 2353 emit_int8(imm8); 2354 } 2355 2356 void Assembler::pause() { 2357 emit_int8((unsigned char)0xF3); 2358 emit_int8((unsigned char)0x90); 2359 } 2360 2361 void Assembler::pcmpestri(XMMRegister dst, Address src, int imm8) { 2362 assert(VM_Version::supports_sse4_2(), ""); 2363 InstructionMark im(this); 2364 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_3A); 2365 emit_int8(0x61); 2366 emit_operand(dst, src); 2367 emit_int8(imm8); 2368 } 2369 2370 void Assembler::pcmpestri(XMMRegister dst, XMMRegister src, int imm8) { 2371 assert(VM_Version::supports_sse4_2(), ""); 2372 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A); 2373 emit_int8(0x61); 2374 emit_int8((unsigned char)(0xC0 | encode)); 2375 emit_int8(imm8); 2376 } 2377 2378 void Assembler::pextrd(Register dst, XMMRegister src, int imm8) { 2379 assert(VM_Version::supports_sse4_1(), ""); 2380 int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, false); 2381 emit_int8(0x16); 2382 emit_int8((unsigned char)(0xC0 | encode)); 2383 emit_int8(imm8); 2384 } 2385 2386 void Assembler::pextrq(Register dst, XMMRegister src, int imm8) { 2387 assert(VM_Version::supports_sse4_1(), ""); 2388 int encode = simd_prefix_and_encode(as_XMMRegister(dst->encoding()), xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A, true); 2389 emit_int8(0x16); 2390 emit_int8((unsigned char)(0xC0 | encode)); 2391 emit_int8(imm8); 2392 } 2393 2394 void Assembler::pinsrd(XMMRegister dst, Register src, int imm8) { 2395 assert(VM_Version::supports_sse4_1(), ""); 2396 int encode = simd_prefix_and_encode(dst, dst, as_XMMRegister(src->encoding()), VEX_SIMD_66, VEX_OPCODE_0F_3A, false); 2397 emit_int8(0x22); 2398 emit_int8((unsigned char)(0xC0 | encode)); 2399 emit_int8(imm8); 2400 } 2401 2402 void Assembler::pinsrq(XMMRegister dst, Register src, int imm8) { 2403 assert(VM_Version::supports_sse4_1(), ""); 2404 int encode = simd_prefix_and_encode(dst, dst, as_XMMRegister(src->encoding()), VEX_SIMD_66, VEX_OPCODE_0F_3A, true); 2405 emit_int8(0x22); 2406 emit_int8((unsigned char)(0xC0 | encode)); 2407 emit_int8(imm8); 2408 } 2409 2410 void Assembler::pmovzxbw(XMMRegister dst, Address src) { 2411 assert(VM_Version::supports_sse4_1(), ""); 2412 InstructionMark im(this); 2413 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2414 emit_int8(0x30); 2415 emit_operand(dst, src); 2416 } 2417 2418 void Assembler::pmovzxbw(XMMRegister dst, XMMRegister src) { 2419 assert(VM_Version::supports_sse4_1(), ""); 2420 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2421 emit_int8(0x30); 2422 emit_int8((unsigned char)(0xC0 | encode)); 2423 } 2424 2425 // generic 2426 void Assembler::pop(Register dst) { 2427 int encode = prefix_and_encode(dst->encoding()); 2428 emit_int8(0x58 | encode); 2429 } 2430 2431 void Assembler::popcntl(Register dst, Address src) { 2432 assert(VM_Version::supports_popcnt(), "must support"); 2433 InstructionMark im(this); 2434 emit_int8((unsigned char)0xF3); 2435 prefix(src, dst); 2436 emit_int8(0x0F); 2437 emit_int8((unsigned char)0xB8); 2438 emit_operand(dst, src); 2439 } 2440 2441 void Assembler::popcntl(Register dst, Register src) { 2442 assert(VM_Version::supports_popcnt(), "must support"); 2443 emit_int8((unsigned char)0xF3); 2444 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2445 emit_int8(0x0F); 2446 emit_int8((unsigned char)0xB8); 2447 emit_int8((unsigned char)(0xC0 | encode)); 2448 } 2449 2450 void Assembler::popf() { 2451 emit_int8((unsigned char)0x9D); 2452 } 2453 2454 #ifndef _LP64 // no 32bit push/pop on amd64 2455 void Assembler::popl(Address dst) { 2456 // NOTE: this will adjust stack by 8byte on 64bits 2457 InstructionMark im(this); 2458 prefix(dst); 2459 emit_int8((unsigned char)0x8F); 2460 emit_operand(rax, dst); 2461 } 2462 #endif 2463 2464 void Assembler::prefetch_prefix(Address src) { 2465 prefix(src); 2466 emit_int8(0x0F); 2467 } 2468 2469 void Assembler::prefetchnta(Address src) { 2470 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2471 InstructionMark im(this); 2472 prefetch_prefix(src); 2473 emit_int8(0x18); 2474 emit_operand(rax, src); // 0, src 2475 } 2476 2477 void Assembler::prefetchr(Address src) { 2478 assert(VM_Version::supports_3dnow_prefetch(), "must support"); 2479 InstructionMark im(this); 2480 prefetch_prefix(src); 2481 emit_int8(0x0D); 2482 emit_operand(rax, src); // 0, src 2483 } 2484 2485 void Assembler::prefetcht0(Address src) { 2486 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2487 InstructionMark im(this); 2488 prefetch_prefix(src); 2489 emit_int8(0x18); 2490 emit_operand(rcx, src); // 1, src 2491 } 2492 2493 void Assembler::prefetcht1(Address src) { 2494 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2495 InstructionMark im(this); 2496 prefetch_prefix(src); 2497 emit_int8(0x18); 2498 emit_operand(rdx, src); // 2, src 2499 } 2500 2501 void Assembler::prefetcht2(Address src) { 2502 NOT_LP64(assert(VM_Version::supports_sse(), "must support")); 2503 InstructionMark im(this); 2504 prefetch_prefix(src); 2505 emit_int8(0x18); 2506 emit_operand(rbx, src); // 3, src 2507 } 2508 2509 void Assembler::prefetchw(Address src) { 2510 assert(VM_Version::supports_3dnow_prefetch(), "must support"); 2511 InstructionMark im(this); 2512 prefetch_prefix(src); 2513 emit_int8(0x0D); 2514 emit_operand(rcx, src); // 1, src 2515 } 2516 2517 void Assembler::prefix(Prefix p) { 2518 emit_int8(p); 2519 } 2520 2521 void Assembler::pshufb(XMMRegister dst, XMMRegister src) { 2522 assert(VM_Version::supports_ssse3(), ""); 2523 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2524 emit_int8(0x00); 2525 emit_int8((unsigned char)(0xC0 | encode)); 2526 } 2527 2528 void Assembler::pshufb(XMMRegister dst, Address src) { 2529 assert(VM_Version::supports_ssse3(), ""); 2530 InstructionMark im(this); 2531 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2532 emit_int8(0x00); 2533 emit_operand(dst, src); 2534 } 2535 2536 void Assembler::pshufd(XMMRegister dst, XMMRegister src, int mode) { 2537 assert(isByte(mode), "invalid value"); 2538 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2539 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_66); 2540 emit_int8(mode & 0xFF); 2541 2542 } 2543 2544 void Assembler::pshufd(XMMRegister dst, Address src, int mode) { 2545 assert(isByte(mode), "invalid value"); 2546 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2547 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2548 InstructionMark im(this); 2549 simd_prefix(dst, src, VEX_SIMD_66); 2550 emit_int8(0x70); 2551 emit_operand(dst, src); 2552 emit_int8(mode & 0xFF); 2553 } 2554 2555 void Assembler::pshuflw(XMMRegister dst, XMMRegister src, int mode) { 2556 assert(isByte(mode), "invalid value"); 2557 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2558 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_F2); 2559 emit_int8(mode & 0xFF); 2560 } 2561 2562 void Assembler::pshuflw(XMMRegister dst, Address src, int mode) { 2563 assert(isByte(mode), "invalid value"); 2564 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2565 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2566 InstructionMark im(this); 2567 simd_prefix(dst, src, VEX_SIMD_F2); 2568 emit_int8(0x70); 2569 emit_operand(dst, src); 2570 emit_int8(mode & 0xFF); 2571 } 2572 2573 void Assembler::psrldq(XMMRegister dst, int shift) { 2574 // Shift 128 bit value in xmm register by number of bytes. 2575 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2576 int encode = simd_prefix_and_encode(xmm3, dst, dst, VEX_SIMD_66); 2577 emit_int8(0x73); 2578 emit_int8((unsigned char)(0xC0 | encode)); 2579 emit_int8(shift); 2580 } 2581 2582 void Assembler::pslldq(XMMRegister dst, int shift) { 2583 // Shift left 128 bit value in xmm register by number of bytes. 2584 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2585 int encode = simd_prefix_and_encode(xmm7, dst, dst, VEX_SIMD_66); 2586 emit_int8(0x73); 2587 emit_int8((unsigned char)(0xC0 | encode)); 2588 emit_int8(shift); 2589 } 2590 2591 void Assembler::ptest(XMMRegister dst, Address src) { 2592 assert(VM_Version::supports_sse4_1(), ""); 2593 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2594 InstructionMark im(this); 2595 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2596 emit_int8(0x17); 2597 emit_operand(dst, src); 2598 } 2599 2600 void Assembler::ptest(XMMRegister dst, XMMRegister src) { 2601 assert(VM_Version::supports_sse4_1(), ""); 2602 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 2603 emit_int8(0x17); 2604 emit_int8((unsigned char)(0xC0 | encode)); 2605 } 2606 2607 void Assembler::vptest(XMMRegister dst, Address src) { 2608 assert(VM_Version::supports_avx(), ""); 2609 InstructionMark im(this); 2610 bool vector256 = true; 2611 assert(dst != xnoreg, "sanity"); 2612 int dst_enc = dst->encoding(); 2613 // swap src<->dst for encoding 2614 vex_prefix(src, 0, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256); 2615 emit_int8(0x17); 2616 emit_operand(dst, src); 2617 } 2618 2619 void Assembler::vptest(XMMRegister dst, XMMRegister src) { 2620 assert(VM_Version::supports_avx(), ""); 2621 bool vector256 = true; 2622 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 2623 emit_int8(0x17); 2624 emit_int8((unsigned char)(0xC0 | encode)); 2625 } 2626 2627 void Assembler::punpcklbw(XMMRegister dst, Address src) { 2628 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2629 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2630 emit_simd_arith(0x60, dst, src, VEX_SIMD_66); 2631 } 2632 2633 void Assembler::punpcklbw(XMMRegister dst, XMMRegister src) { 2634 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2635 emit_simd_arith(0x60, dst, src, VEX_SIMD_66); 2636 } 2637 2638 void Assembler::punpckldq(XMMRegister dst, Address src) { 2639 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2640 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes"); 2641 emit_simd_arith(0x62, dst, src, VEX_SIMD_66); 2642 } 2643 2644 void Assembler::punpckldq(XMMRegister dst, XMMRegister src) { 2645 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2646 emit_simd_arith(0x62, dst, src, VEX_SIMD_66); 2647 } 2648 2649 void Assembler::punpcklqdq(XMMRegister dst, XMMRegister src) { 2650 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2651 emit_simd_arith(0x6C, dst, src, VEX_SIMD_66); 2652 } 2653 2654 void Assembler::push(int32_t imm32) { 2655 // in 64bits we push 64bits onto the stack but only 2656 // take a 32bit immediate 2657 emit_int8(0x68); 2658 emit_int32(imm32); 2659 } 2660 2661 void Assembler::push(Register src) { 2662 int encode = prefix_and_encode(src->encoding()); 2663 2664 emit_int8(0x50 | encode); 2665 } 2666 2667 void Assembler::pushf() { 2668 emit_int8((unsigned char)0x9C); 2669 } 2670 2671 #ifndef _LP64 // no 32bit push/pop on amd64 2672 void Assembler::pushl(Address src) { 2673 // Note this will push 64bit on 64bit 2674 InstructionMark im(this); 2675 prefix(src); 2676 emit_int8((unsigned char)0xFF); 2677 emit_operand(rsi, src); 2678 } 2679 #endif 2680 2681 void Assembler::rcll(Register dst, int imm8) { 2682 assert(isShiftCount(imm8), "illegal shift count"); 2683 int encode = prefix_and_encode(dst->encoding()); 2684 if (imm8 == 1) { 2685 emit_int8((unsigned char)0xD1); 2686 emit_int8((unsigned char)(0xD0 | encode)); 2687 } else { 2688 emit_int8((unsigned char)0xC1); 2689 emit_int8((unsigned char)0xD0 | encode); 2690 emit_int8(imm8); 2691 } 2692 } 2693 2694 void Assembler::rdtsc() { 2695 emit_int8((unsigned char)0x0F); 2696 emit_int8((unsigned char)0x31); 2697 } 2698 2699 // copies data from [esi] to [edi] using rcx pointer sized words 2700 // generic 2701 void Assembler::rep_mov() { 2702 emit_int8((unsigned char)0xF3); 2703 // MOVSQ 2704 LP64_ONLY(prefix(REX_W)); 2705 emit_int8((unsigned char)0xA5); 2706 } 2707 2708 // sets rcx bytes with rax, value at [edi] 2709 void Assembler::rep_stosb() { 2710 emit_int8((unsigned char)0xF3); // REP 2711 LP64_ONLY(prefix(REX_W)); 2712 emit_int8((unsigned char)0xAA); // STOSB 2713 } 2714 2715 // sets rcx pointer sized words with rax, value at [edi] 2716 // generic 2717 void Assembler::rep_stos() { 2718 emit_int8((unsigned char)0xF3); // REP 2719 LP64_ONLY(prefix(REX_W)); // LP64:STOSQ, LP32:STOSD 2720 emit_int8((unsigned char)0xAB); 2721 } 2722 2723 // scans rcx pointer sized words at [edi] for occurance of rax, 2724 // generic 2725 void Assembler::repne_scan() { // repne_scan 2726 emit_int8((unsigned char)0xF2); 2727 // SCASQ 2728 LP64_ONLY(prefix(REX_W)); 2729 emit_int8((unsigned char)0xAF); 2730 } 2731 2732 #ifdef _LP64 2733 // scans rcx 4 byte words at [edi] for occurance of rax, 2734 // generic 2735 void Assembler::repne_scanl() { // repne_scan 2736 emit_int8((unsigned char)0xF2); 2737 // SCASL 2738 emit_int8((unsigned char)0xAF); 2739 } 2740 #endif 2741 2742 void Assembler::ret(int imm16) { 2743 if (imm16 == 0) { 2744 emit_int8((unsigned char)0xC3); 2745 } else { 2746 emit_int8((unsigned char)0xC2); 2747 emit_int16(imm16); 2748 } 2749 } 2750 2751 void Assembler::sahf() { 2752 #ifdef _LP64 2753 // Not supported in 64bit mode 2754 ShouldNotReachHere(); 2755 #endif 2756 emit_int8((unsigned char)0x9E); 2757 } 2758 2759 void Assembler::sarl(Register dst, int imm8) { 2760 int encode = prefix_and_encode(dst->encoding()); 2761 assert(isShiftCount(imm8), "illegal shift count"); 2762 if (imm8 == 1) { 2763 emit_int8((unsigned char)0xD1); 2764 emit_int8((unsigned char)(0xF8 | encode)); 2765 } else { 2766 emit_int8((unsigned char)0xC1); 2767 emit_int8((unsigned char)(0xF8 | encode)); 2768 emit_int8(imm8); 2769 } 2770 } 2771 2772 void Assembler::sarl(Register dst) { 2773 int encode = prefix_and_encode(dst->encoding()); 2774 emit_int8((unsigned char)0xD3); 2775 emit_int8((unsigned char)(0xF8 | encode)); 2776 } 2777 2778 void Assembler::sbbl(Address dst, int32_t imm32) { 2779 InstructionMark im(this); 2780 prefix(dst); 2781 emit_arith_operand(0x81, rbx, dst, imm32); 2782 } 2783 2784 void Assembler::sbbl(Register dst, int32_t imm32) { 2785 prefix(dst); 2786 emit_arith(0x81, 0xD8, dst, imm32); 2787 } 2788 2789 2790 void Assembler::sbbl(Register dst, Address src) { 2791 InstructionMark im(this); 2792 prefix(src, dst); 2793 emit_int8(0x1B); 2794 emit_operand(dst, src); 2795 } 2796 2797 void Assembler::sbbl(Register dst, Register src) { 2798 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2799 emit_arith(0x1B, 0xC0, dst, src); 2800 } 2801 2802 void Assembler::setb(Condition cc, Register dst) { 2803 assert(0 <= cc && cc < 16, "illegal cc"); 2804 int encode = prefix_and_encode(dst->encoding(), true); 2805 emit_int8(0x0F); 2806 emit_int8((unsigned char)0x90 | cc); 2807 emit_int8((unsigned char)(0xC0 | encode)); 2808 } 2809 2810 void Assembler::shll(Register dst, int imm8) { 2811 assert(isShiftCount(imm8), "illegal shift count"); 2812 int encode = prefix_and_encode(dst->encoding()); 2813 if (imm8 == 1 ) { 2814 emit_int8((unsigned char)0xD1); 2815 emit_int8((unsigned char)(0xE0 | encode)); 2816 } else { 2817 emit_int8((unsigned char)0xC1); 2818 emit_int8((unsigned char)(0xE0 | encode)); 2819 emit_int8(imm8); 2820 } 2821 } 2822 2823 void Assembler::shll(Register dst) { 2824 int encode = prefix_and_encode(dst->encoding()); 2825 emit_int8((unsigned char)0xD3); 2826 emit_int8((unsigned char)(0xE0 | encode)); 2827 } 2828 2829 void Assembler::shrl(Register dst, int imm8) { 2830 assert(isShiftCount(imm8), "illegal shift count"); 2831 int encode = prefix_and_encode(dst->encoding()); 2832 emit_int8((unsigned char)0xC1); 2833 emit_int8((unsigned char)(0xE8 | encode)); 2834 emit_int8(imm8); 2835 } 2836 2837 void Assembler::shrl(Register dst) { 2838 int encode = prefix_and_encode(dst->encoding()); 2839 emit_int8((unsigned char)0xD3); 2840 emit_int8((unsigned char)(0xE8 | encode)); 2841 } 2842 2843 // copies a single word from [esi] to [edi] 2844 void Assembler::smovl() { 2845 emit_int8((unsigned char)0xA5); 2846 } 2847 2848 void Assembler::sqrtsd(XMMRegister dst, XMMRegister src) { 2849 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2850 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2); 2851 } 2852 2853 void Assembler::sqrtsd(XMMRegister dst, Address src) { 2854 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2855 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2); 2856 } 2857 2858 void Assembler::sqrtss(XMMRegister dst, XMMRegister src) { 2859 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2860 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3); 2861 } 2862 2863 void Assembler::std() { 2864 emit_int8((unsigned char)0xFD); 2865 } 2866 2867 void Assembler::sqrtss(XMMRegister dst, Address src) { 2868 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2869 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3); 2870 } 2871 2872 void Assembler::stmxcsr( Address dst) { 2873 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2874 InstructionMark im(this); 2875 prefix(dst); 2876 emit_int8(0x0F); 2877 emit_int8((unsigned char)0xAE); 2878 emit_operand(as_Register(3), dst); 2879 } 2880 2881 void Assembler::subl(Address dst, int32_t imm32) { 2882 InstructionMark im(this); 2883 prefix(dst); 2884 emit_arith_operand(0x81, rbp, dst, imm32); 2885 } 2886 2887 void Assembler::subl(Address dst, Register src) { 2888 InstructionMark im(this); 2889 prefix(dst, src); 2890 emit_int8(0x29); 2891 emit_operand(src, dst); 2892 } 2893 2894 void Assembler::subl(Register dst, int32_t imm32) { 2895 prefix(dst); 2896 emit_arith(0x81, 0xE8, dst, imm32); 2897 } 2898 2899 // Force generation of a 4 byte immediate value even if it fits into 8bit 2900 void Assembler::subl_imm32(Register dst, int32_t imm32) { 2901 prefix(dst); 2902 emit_arith_imm32(0x81, 0xE8, dst, imm32); 2903 } 2904 2905 void Assembler::subl(Register dst, Address src) { 2906 InstructionMark im(this); 2907 prefix(src, dst); 2908 emit_int8(0x2B); 2909 emit_operand(dst, src); 2910 } 2911 2912 void Assembler::subl(Register dst, Register src) { 2913 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2914 emit_arith(0x2B, 0xC0, dst, src); 2915 } 2916 2917 void Assembler::subsd(XMMRegister dst, XMMRegister src) { 2918 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2919 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2); 2920 } 2921 2922 void Assembler::subsd(XMMRegister dst, Address src) { 2923 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2924 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2); 2925 } 2926 2927 void Assembler::subss(XMMRegister dst, XMMRegister src) { 2928 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2929 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3); 2930 } 2931 2932 void Assembler::subss(XMMRegister dst, Address src) { 2933 NOT_LP64(assert(VM_Version::supports_sse(), "")); 2934 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3); 2935 } 2936 2937 void Assembler::testb(Register dst, int imm8) { 2938 NOT_LP64(assert(dst->has_byte_register(), "must have byte register")); 2939 (void) prefix_and_encode(dst->encoding(), true); 2940 emit_arith_b(0xF6, 0xC0, dst, imm8); 2941 } 2942 2943 void Assembler::testl(Register dst, int32_t imm32) { 2944 // not using emit_arith because test 2945 // doesn't support sign-extension of 2946 // 8bit operands 2947 int encode = dst->encoding(); 2948 if (encode == 0) { 2949 emit_int8((unsigned char)0xA9); 2950 } else { 2951 encode = prefix_and_encode(encode); 2952 emit_int8((unsigned char)0xF7); 2953 emit_int8((unsigned char)(0xC0 | encode)); 2954 } 2955 emit_int32(imm32); 2956 } 2957 2958 void Assembler::testl(Register dst, Register src) { 2959 (void) prefix_and_encode(dst->encoding(), src->encoding()); 2960 emit_arith(0x85, 0xC0, dst, src); 2961 } 2962 2963 void Assembler::testl(Register dst, Address src) { 2964 InstructionMark im(this); 2965 prefix(src, dst); 2966 emit_int8((unsigned char)0x85); 2967 emit_operand(dst, src); 2968 } 2969 2970 void Assembler::tzcntl(Register dst, Register src) { 2971 assert(VM_Version::supports_bmi1(), "tzcnt instruction not supported"); 2972 emit_int8((unsigned char)0xF3); 2973 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 2974 emit_int8(0x0F); 2975 emit_int8((unsigned char)0xBC); 2976 emit_int8((unsigned char)0xC0 | encode); 2977 } 2978 2979 void Assembler::tzcntq(Register dst, Register src) { 2980 assert(VM_Version::supports_bmi1(), "tzcnt instruction not supported"); 2981 emit_int8((unsigned char)0xF3); 2982 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 2983 emit_int8(0x0F); 2984 emit_int8((unsigned char)0xBC); 2985 emit_int8((unsigned char)(0xC0 | encode)); 2986 } 2987 2988 void Assembler::ucomisd(XMMRegister dst, Address src) { 2989 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2990 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66); 2991 } 2992 2993 void Assembler::ucomisd(XMMRegister dst, XMMRegister src) { 2994 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 2995 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66); 2996 } 2997 2998 void Assembler::ucomiss(XMMRegister dst, Address src) { 2999 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3000 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE); 3001 } 3002 3003 void Assembler::ucomiss(XMMRegister dst, XMMRegister src) { 3004 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3005 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE); 3006 } 3007 3008 void Assembler::xabort(int8_t imm8) { 3009 emit_int8((unsigned char)0xC6); 3010 emit_int8((unsigned char)0xF8); 3011 emit_int8((unsigned char)(imm8 & 0xFF)); 3012 } 3013 3014 void Assembler::xaddl(Address dst, Register src) { 3015 InstructionMark im(this); 3016 prefix(dst, src); 3017 emit_int8(0x0F); 3018 emit_int8((unsigned char)0xC1); 3019 emit_operand(src, dst); 3020 } 3021 3022 void Assembler::xbegin(Label& abort, relocInfo::relocType rtype) { 3023 InstructionMark im(this); 3024 relocate(rtype); 3025 if (abort.is_bound()) { 3026 address entry = target(abort); 3027 assert(entry != NULL, "abort entry NULL"); 3028 intptr_t offset = entry - pc(); 3029 emit_int8((unsigned char)0xC7); 3030 emit_int8((unsigned char)0xF8); 3031 emit_int32(offset - 6); // 2 opcode + 4 address 3032 } else { 3033 abort.add_patch_at(code(), locator()); 3034 emit_int8((unsigned char)0xC7); 3035 emit_int8((unsigned char)0xF8); 3036 emit_int32(0); 3037 } 3038 } 3039 3040 void Assembler::xchgl(Register dst, Address src) { // xchg 3041 InstructionMark im(this); 3042 prefix(src, dst); 3043 emit_int8((unsigned char)0x87); 3044 emit_operand(dst, src); 3045 } 3046 3047 void Assembler::xchgl(Register dst, Register src) { 3048 int encode = prefix_and_encode(dst->encoding(), src->encoding()); 3049 emit_int8((unsigned char)0x87); 3050 emit_int8((unsigned char)(0xC0 | encode)); 3051 } 3052 3053 void Assembler::xend() { 3054 emit_int8((unsigned char)0x0F); 3055 emit_int8((unsigned char)0x01); 3056 emit_int8((unsigned char)0xD5); 3057 } 3058 3059 void Assembler::xgetbv() { 3060 emit_int8(0x0F); 3061 emit_int8(0x01); 3062 emit_int8((unsigned char)0xD0); 3063 } 3064 3065 void Assembler::xorl(Register dst, int32_t imm32) { 3066 prefix(dst); 3067 emit_arith(0x81, 0xF0, dst, imm32); 3068 } 3069 3070 void Assembler::xorl(Register dst, Address src) { 3071 InstructionMark im(this); 3072 prefix(src, dst); 3073 emit_int8(0x33); 3074 emit_operand(dst, src); 3075 } 3076 3077 void Assembler::xorl(Register dst, Register src) { 3078 (void) prefix_and_encode(dst->encoding(), src->encoding()); 3079 emit_arith(0x33, 0xC0, dst, src); 3080 } 3081 3082 3083 // AVX 3-operands scalar float-point arithmetic instructions 3084 3085 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, Address src) { 3086 assert(VM_Version::supports_avx(), ""); 3087 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3088 } 3089 3090 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3091 assert(VM_Version::supports_avx(), ""); 3092 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3093 } 3094 3095 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, Address src) { 3096 assert(VM_Version::supports_avx(), ""); 3097 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3098 } 3099 3100 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3101 assert(VM_Version::supports_avx(), ""); 3102 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3103 } 3104 3105 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, Address src) { 3106 assert(VM_Version::supports_avx(), ""); 3107 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3108 } 3109 3110 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3111 assert(VM_Version::supports_avx(), ""); 3112 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3113 } 3114 3115 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, Address src) { 3116 assert(VM_Version::supports_avx(), ""); 3117 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3118 } 3119 3120 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3121 assert(VM_Version::supports_avx(), ""); 3122 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3123 } 3124 3125 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, Address src) { 3126 assert(VM_Version::supports_avx(), ""); 3127 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3128 } 3129 3130 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3131 assert(VM_Version::supports_avx(), ""); 3132 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3133 } 3134 3135 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, Address src) { 3136 assert(VM_Version::supports_avx(), ""); 3137 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3138 } 3139 3140 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3141 assert(VM_Version::supports_avx(), ""); 3142 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3143 } 3144 3145 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, Address src) { 3146 assert(VM_Version::supports_avx(), ""); 3147 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3148 } 3149 3150 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3151 assert(VM_Version::supports_avx(), ""); 3152 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false); 3153 } 3154 3155 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, Address src) { 3156 assert(VM_Version::supports_avx(), ""); 3157 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3158 } 3159 3160 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3161 assert(VM_Version::supports_avx(), ""); 3162 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false); 3163 } 3164 3165 //====================VECTOR ARITHMETIC===================================== 3166 3167 // Float-point vector arithmetic 3168 3169 void Assembler::addpd(XMMRegister dst, XMMRegister src) { 3170 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3171 emit_simd_arith(0x58, dst, src, VEX_SIMD_66); 3172 } 3173 3174 void Assembler::addps(XMMRegister dst, XMMRegister src) { 3175 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3176 emit_simd_arith(0x58, dst, src, VEX_SIMD_NONE); 3177 } 3178 3179 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3180 assert(VM_Version::supports_avx(), ""); 3181 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256); 3182 } 3183 3184 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3185 assert(VM_Version::supports_avx(), ""); 3186 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256); 3187 } 3188 3189 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3190 assert(VM_Version::supports_avx(), ""); 3191 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256); 3192 } 3193 3194 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3195 assert(VM_Version::supports_avx(), ""); 3196 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256); 3197 } 3198 3199 void Assembler::subpd(XMMRegister dst, XMMRegister src) { 3200 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3201 emit_simd_arith(0x5C, dst, src, VEX_SIMD_66); 3202 } 3203 3204 void Assembler::subps(XMMRegister dst, XMMRegister src) { 3205 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3206 emit_simd_arith(0x5C, dst, src, VEX_SIMD_NONE); 3207 } 3208 3209 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3210 assert(VM_Version::supports_avx(), ""); 3211 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256); 3212 } 3213 3214 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3215 assert(VM_Version::supports_avx(), ""); 3216 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256); 3217 } 3218 3219 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3220 assert(VM_Version::supports_avx(), ""); 3221 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256); 3222 } 3223 3224 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3225 assert(VM_Version::supports_avx(), ""); 3226 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256); 3227 } 3228 3229 void Assembler::mulpd(XMMRegister dst, XMMRegister src) { 3230 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3231 emit_simd_arith(0x59, dst, src, VEX_SIMD_66); 3232 } 3233 3234 void Assembler::mulps(XMMRegister dst, XMMRegister src) { 3235 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3236 emit_simd_arith(0x59, dst, src, VEX_SIMD_NONE); 3237 } 3238 3239 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3240 assert(VM_Version::supports_avx(), ""); 3241 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256); 3242 } 3243 3244 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3245 assert(VM_Version::supports_avx(), ""); 3246 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256); 3247 } 3248 3249 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3250 assert(VM_Version::supports_avx(), ""); 3251 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256); 3252 } 3253 3254 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3255 assert(VM_Version::supports_avx(), ""); 3256 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256); 3257 } 3258 3259 void Assembler::divpd(XMMRegister dst, XMMRegister src) { 3260 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3261 emit_simd_arith(0x5E, dst, src, VEX_SIMD_66); 3262 } 3263 3264 void Assembler::divps(XMMRegister dst, XMMRegister src) { 3265 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3266 emit_simd_arith(0x5E, dst, src, VEX_SIMD_NONE); 3267 } 3268 3269 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3270 assert(VM_Version::supports_avx(), ""); 3271 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256); 3272 } 3273 3274 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3275 assert(VM_Version::supports_avx(), ""); 3276 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256); 3277 } 3278 3279 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3280 assert(VM_Version::supports_avx(), ""); 3281 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256); 3282 } 3283 3284 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3285 assert(VM_Version::supports_avx(), ""); 3286 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256); 3287 } 3288 3289 void Assembler::andpd(XMMRegister dst, XMMRegister src) { 3290 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3291 emit_simd_arith(0x54, dst, src, VEX_SIMD_66); 3292 } 3293 3294 void Assembler::andps(XMMRegister dst, XMMRegister src) { 3295 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3296 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE); 3297 } 3298 3299 void Assembler::andps(XMMRegister dst, Address src) { 3300 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3301 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE); 3302 } 3303 3304 void Assembler::andpd(XMMRegister dst, Address src) { 3305 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3306 emit_simd_arith(0x54, dst, src, VEX_SIMD_66); 3307 } 3308 3309 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3310 assert(VM_Version::supports_avx(), ""); 3311 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256); 3312 } 3313 3314 void Assembler::vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3315 assert(VM_Version::supports_avx(), ""); 3316 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256); 3317 } 3318 3319 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3320 assert(VM_Version::supports_avx(), ""); 3321 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256); 3322 } 3323 3324 void Assembler::vandps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3325 assert(VM_Version::supports_avx(), ""); 3326 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256); 3327 } 3328 3329 void Assembler::xorpd(XMMRegister dst, XMMRegister src) { 3330 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3331 emit_simd_arith(0x57, dst, src, VEX_SIMD_66); 3332 } 3333 3334 void Assembler::xorps(XMMRegister dst, XMMRegister src) { 3335 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3336 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE); 3337 } 3338 3339 void Assembler::xorpd(XMMRegister dst, Address src) { 3340 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3341 emit_simd_arith(0x57, dst, src, VEX_SIMD_66); 3342 } 3343 3344 void Assembler::xorps(XMMRegister dst, Address src) { 3345 NOT_LP64(assert(VM_Version::supports_sse(), "")); 3346 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE); 3347 } 3348 3349 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3350 assert(VM_Version::supports_avx(), ""); 3351 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256); 3352 } 3353 3354 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3355 assert(VM_Version::supports_avx(), ""); 3356 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256); 3357 } 3358 3359 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3360 assert(VM_Version::supports_avx(), ""); 3361 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256); 3362 } 3363 3364 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3365 assert(VM_Version::supports_avx(), ""); 3366 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256); 3367 } 3368 3369 3370 // Integer vector arithmetic 3371 void Assembler::paddb(XMMRegister dst, XMMRegister src) { 3372 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3373 emit_simd_arith(0xFC, dst, src, VEX_SIMD_66); 3374 } 3375 3376 void Assembler::paddw(XMMRegister dst, XMMRegister src) { 3377 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3378 emit_simd_arith(0xFD, dst, src, VEX_SIMD_66); 3379 } 3380 3381 void Assembler::paddd(XMMRegister dst, XMMRegister src) { 3382 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3383 emit_simd_arith(0xFE, dst, src, VEX_SIMD_66); 3384 } 3385 3386 void Assembler::paddq(XMMRegister dst, XMMRegister src) { 3387 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3388 emit_simd_arith(0xD4, dst, src, VEX_SIMD_66); 3389 } 3390 3391 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3392 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3393 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256); 3394 } 3395 3396 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3397 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3398 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256); 3399 } 3400 3401 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3402 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3403 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256); 3404 } 3405 3406 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3407 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3408 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256); 3409 } 3410 3411 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3412 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3413 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256); 3414 } 3415 3416 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3417 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3418 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256); 3419 } 3420 3421 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3422 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3423 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256); 3424 } 3425 3426 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3427 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3428 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256); 3429 } 3430 3431 void Assembler::psubb(XMMRegister dst, XMMRegister src) { 3432 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3433 emit_simd_arith(0xF8, dst, src, VEX_SIMD_66); 3434 } 3435 3436 void Assembler::psubw(XMMRegister dst, XMMRegister src) { 3437 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3438 emit_simd_arith(0xF9, dst, src, VEX_SIMD_66); 3439 } 3440 3441 void Assembler::psubd(XMMRegister dst, XMMRegister src) { 3442 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3443 emit_simd_arith(0xFA, dst, src, VEX_SIMD_66); 3444 } 3445 3446 void Assembler::psubq(XMMRegister dst, XMMRegister src) { 3447 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3448 emit_simd_arith(0xFB, dst, src, VEX_SIMD_66); 3449 } 3450 3451 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3452 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3453 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256); 3454 } 3455 3456 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3457 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3458 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256); 3459 } 3460 3461 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3462 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3463 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256); 3464 } 3465 3466 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3467 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3468 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256); 3469 } 3470 3471 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3472 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3473 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256); 3474 } 3475 3476 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3477 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3478 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256); 3479 } 3480 3481 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3482 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3483 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256); 3484 } 3485 3486 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3487 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3488 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256); 3489 } 3490 3491 void Assembler::pmullw(XMMRegister dst, XMMRegister src) { 3492 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3493 emit_simd_arith(0xD5, dst, src, VEX_SIMD_66); 3494 } 3495 3496 void Assembler::pmulld(XMMRegister dst, XMMRegister src) { 3497 assert(VM_Version::supports_sse4_1(), ""); 3498 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38); 3499 emit_int8(0x40); 3500 emit_int8((unsigned char)(0xC0 | encode)); 3501 } 3502 3503 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3504 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3505 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256); 3506 } 3507 3508 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3509 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3510 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 3511 emit_int8(0x40); 3512 emit_int8((unsigned char)(0xC0 | encode)); 3513 } 3514 3515 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3516 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3517 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256); 3518 } 3519 3520 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3521 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3522 InstructionMark im(this); 3523 int dst_enc = dst->encoding(); 3524 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 3525 vex_prefix(src, nds_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256); 3526 emit_int8(0x40); 3527 emit_operand(dst, src); 3528 } 3529 3530 // Shift packed integers left by specified number of bits. 3531 void Assembler::psllw(XMMRegister dst, int shift) { 3532 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3533 // XMM6 is for /6 encoding: 66 0F 71 /6 ib 3534 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3535 emit_int8(0x71); 3536 emit_int8((unsigned char)(0xC0 | encode)); 3537 emit_int8(shift & 0xFF); 3538 } 3539 3540 void Assembler::pslld(XMMRegister dst, int shift) { 3541 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3542 // XMM6 is for /6 encoding: 66 0F 72 /6 ib 3543 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3544 emit_int8(0x72); 3545 emit_int8((unsigned char)(0xC0 | encode)); 3546 emit_int8(shift & 0xFF); 3547 } 3548 3549 void Assembler::psllq(XMMRegister dst, int shift) { 3550 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3551 // XMM6 is for /6 encoding: 66 0F 73 /6 ib 3552 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66); 3553 emit_int8(0x73); 3554 emit_int8((unsigned char)(0xC0 | encode)); 3555 emit_int8(shift & 0xFF); 3556 } 3557 3558 void Assembler::psllw(XMMRegister dst, XMMRegister shift) { 3559 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3560 emit_simd_arith(0xF1, dst, shift, VEX_SIMD_66); 3561 } 3562 3563 void Assembler::pslld(XMMRegister dst, XMMRegister shift) { 3564 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3565 emit_simd_arith(0xF2, dst, shift, VEX_SIMD_66); 3566 } 3567 3568 void Assembler::psllq(XMMRegister dst, XMMRegister shift) { 3569 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3570 emit_simd_arith(0xF3, dst, shift, VEX_SIMD_66); 3571 } 3572 3573 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3574 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3575 // XMM6 is for /6 encoding: 66 0F 71 /6 ib 3576 emit_vex_arith(0x71, xmm6, dst, src, VEX_SIMD_66, vector256); 3577 emit_int8(shift & 0xFF); 3578 } 3579 3580 void Assembler::vpslld(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3581 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3582 // XMM6 is for /6 encoding: 66 0F 72 /6 ib 3583 emit_vex_arith(0x72, xmm6, dst, src, VEX_SIMD_66, vector256); 3584 emit_int8(shift & 0xFF); 3585 } 3586 3587 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3588 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3589 // XMM6 is for /6 encoding: 66 0F 73 /6 ib 3590 emit_vex_arith(0x73, xmm6, dst, src, VEX_SIMD_66, vector256); 3591 emit_int8(shift & 0xFF); 3592 } 3593 3594 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3595 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3596 emit_vex_arith(0xF1, dst, src, shift, VEX_SIMD_66, vector256); 3597 } 3598 3599 void Assembler::vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3600 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3601 emit_vex_arith(0xF2, dst, src, shift, VEX_SIMD_66, vector256); 3602 } 3603 3604 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3605 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3606 emit_vex_arith(0xF3, dst, src, shift, VEX_SIMD_66, vector256); 3607 } 3608 3609 // Shift packed integers logically right by specified number of bits. 3610 void Assembler::psrlw(XMMRegister dst, int shift) { 3611 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3612 // XMM2 is for /2 encoding: 66 0F 71 /2 ib 3613 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3614 emit_int8(0x71); 3615 emit_int8((unsigned char)(0xC0 | encode)); 3616 emit_int8(shift & 0xFF); 3617 } 3618 3619 void Assembler::psrld(XMMRegister dst, int shift) { 3620 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3621 // XMM2 is for /2 encoding: 66 0F 72 /2 ib 3622 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3623 emit_int8(0x72); 3624 emit_int8((unsigned char)(0xC0 | encode)); 3625 emit_int8(shift & 0xFF); 3626 } 3627 3628 void Assembler::psrlq(XMMRegister dst, int shift) { 3629 // Do not confuse it with psrldq SSE2 instruction which 3630 // shifts 128 bit value in xmm register by number of bytes. 3631 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3632 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3633 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66); 3634 emit_int8(0x73); 3635 emit_int8((unsigned char)(0xC0 | encode)); 3636 emit_int8(shift & 0xFF); 3637 } 3638 3639 void Assembler::psrlw(XMMRegister dst, XMMRegister shift) { 3640 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3641 emit_simd_arith(0xD1, dst, shift, VEX_SIMD_66); 3642 } 3643 3644 void Assembler::psrld(XMMRegister dst, XMMRegister shift) { 3645 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3646 emit_simd_arith(0xD2, dst, shift, VEX_SIMD_66); 3647 } 3648 3649 void Assembler::psrlq(XMMRegister dst, XMMRegister shift) { 3650 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3651 emit_simd_arith(0xD3, dst, shift, VEX_SIMD_66); 3652 } 3653 3654 void Assembler::vpsrlw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3655 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3656 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3657 emit_vex_arith(0x71, xmm2, dst, src, VEX_SIMD_66, vector256); 3658 emit_int8(shift & 0xFF); 3659 } 3660 3661 void Assembler::vpsrld(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 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3664 emit_vex_arith(0x72, xmm2, dst, src, VEX_SIMD_66, vector256); 3665 emit_int8(shift & 0xFF); 3666 } 3667 3668 void Assembler::vpsrlq(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 // XMM2 is for /2 encoding: 66 0F 73 /2 ib 3671 emit_vex_arith(0x73, xmm2, dst, src, VEX_SIMD_66, vector256); 3672 emit_int8(shift & 0xFF); 3673 } 3674 3675 void Assembler::vpsrlw(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(0xD1, dst, src, shift, VEX_SIMD_66, vector256); 3678 } 3679 3680 void Assembler::vpsrld(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(0xD2, dst, src, shift, VEX_SIMD_66, vector256); 3683 } 3684 3685 void Assembler::vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3686 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3687 emit_vex_arith(0xD3, dst, src, shift, VEX_SIMD_66, vector256); 3688 } 3689 3690 // Shift packed integers arithmetically right by specified number of bits. 3691 void Assembler::psraw(XMMRegister dst, int shift) { 3692 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3693 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3694 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66); 3695 emit_int8(0x71); 3696 emit_int8((unsigned char)(0xC0 | encode)); 3697 emit_int8(shift & 0xFF); 3698 } 3699 3700 void Assembler::psrad(XMMRegister dst, int shift) { 3701 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3702 // XMM4 is for /4 encoding: 66 0F 72 /4 ib 3703 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66); 3704 emit_int8(0x72); 3705 emit_int8((unsigned char)(0xC0 | encode)); 3706 emit_int8(shift & 0xFF); 3707 } 3708 3709 void Assembler::psraw(XMMRegister dst, XMMRegister shift) { 3710 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3711 emit_simd_arith(0xE1, dst, shift, VEX_SIMD_66); 3712 } 3713 3714 void Assembler::psrad(XMMRegister dst, XMMRegister shift) { 3715 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3716 emit_simd_arith(0xE2, dst, shift, VEX_SIMD_66); 3717 } 3718 3719 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3720 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3721 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3722 emit_vex_arith(0x71, xmm4, dst, src, VEX_SIMD_66, vector256); 3723 emit_int8(shift & 0xFF); 3724 } 3725 3726 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, int shift, bool vector256) { 3727 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3728 // XMM4 is for /4 encoding: 66 0F 71 /4 ib 3729 emit_vex_arith(0x72, xmm4, dst, src, VEX_SIMD_66, vector256); 3730 emit_int8(shift & 0xFF); 3731 } 3732 3733 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3734 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3735 emit_vex_arith(0xE1, dst, src, shift, VEX_SIMD_66, vector256); 3736 } 3737 3738 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) { 3739 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3740 emit_vex_arith(0xE2, dst, src, shift, VEX_SIMD_66, vector256); 3741 } 3742 3743 3744 // AND packed integers 3745 void Assembler::pand(XMMRegister dst, XMMRegister src) { 3746 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3747 emit_simd_arith(0xDB, dst, src, VEX_SIMD_66); 3748 } 3749 3750 void Assembler::vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3751 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3752 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256); 3753 } 3754 3755 void Assembler::vpand(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3756 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3757 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256); 3758 } 3759 3760 void Assembler::por(XMMRegister dst, XMMRegister src) { 3761 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3762 emit_simd_arith(0xEB, dst, src, VEX_SIMD_66); 3763 } 3764 3765 void Assembler::vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3766 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3767 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256); 3768 } 3769 3770 void Assembler::vpor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3771 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3772 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256); 3773 } 3774 3775 void Assembler::pxor(XMMRegister dst, XMMRegister src) { 3776 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 3777 emit_simd_arith(0xEF, dst, src, VEX_SIMD_66); 3778 } 3779 3780 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) { 3781 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3782 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256); 3783 } 3784 3785 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) { 3786 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2"); 3787 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256); 3788 } 3789 3790 3791 void Assembler::vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3792 assert(VM_Version::supports_avx(), ""); 3793 bool vector256 = true; 3794 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3795 emit_int8(0x18); 3796 emit_int8((unsigned char)(0xC0 | encode)); 3797 // 0x00 - insert into lower 128 bits 3798 // 0x01 - insert into upper 128 bits 3799 emit_int8(0x01); 3800 } 3801 3802 void Assembler::vinsertf128h(XMMRegister dst, Address src) { 3803 assert(VM_Version::supports_avx(), ""); 3804 InstructionMark im(this); 3805 bool vector256 = true; 3806 assert(dst != xnoreg, "sanity"); 3807 int dst_enc = dst->encoding(); 3808 // swap src<->dst for encoding 3809 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3810 emit_int8(0x18); 3811 emit_operand(dst, src); 3812 // 0x01 - insert into upper 128 bits 3813 emit_int8(0x01); 3814 } 3815 3816 void Assembler::vextractf128h(Address dst, XMMRegister src) { 3817 assert(VM_Version::supports_avx(), ""); 3818 InstructionMark im(this); 3819 bool vector256 = true; 3820 assert(src != xnoreg, "sanity"); 3821 int src_enc = src->encoding(); 3822 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3823 emit_int8(0x19); 3824 emit_operand(src, dst); 3825 // 0x01 - extract from upper 128 bits 3826 emit_int8(0x01); 3827 } 3828 3829 void Assembler::vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src) { 3830 assert(VM_Version::supports_avx2(), ""); 3831 bool vector256 = true; 3832 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3833 emit_int8(0x38); 3834 emit_int8((unsigned char)(0xC0 | encode)); 3835 // 0x00 - insert into lower 128 bits 3836 // 0x01 - insert into upper 128 bits 3837 emit_int8(0x01); 3838 } 3839 3840 void Assembler::vinserti128h(XMMRegister dst, Address src) { 3841 assert(VM_Version::supports_avx2(), ""); 3842 InstructionMark im(this); 3843 bool vector256 = true; 3844 assert(dst != xnoreg, "sanity"); 3845 int dst_enc = dst->encoding(); 3846 // swap src<->dst for encoding 3847 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3848 emit_int8(0x38); 3849 emit_operand(dst, src); 3850 // 0x01 - insert into upper 128 bits 3851 emit_int8(0x01); 3852 } 3853 3854 void Assembler::vextracti128h(Address dst, XMMRegister src) { 3855 assert(VM_Version::supports_avx2(), ""); 3856 InstructionMark im(this); 3857 bool vector256 = true; 3858 assert(src != xnoreg, "sanity"); 3859 int src_enc = src->encoding(); 3860 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256); 3861 emit_int8(0x39); 3862 emit_operand(src, dst); 3863 // 0x01 - extract from upper 128 bits 3864 emit_int8(0x01); 3865 } 3866 3867 // duplicate 4-bytes integer data from src into 8 locations in dest 3868 void Assembler::vpbroadcastd(XMMRegister dst, XMMRegister src) { 3869 assert(VM_Version::supports_avx2(), ""); 3870 bool vector256 = true; 3871 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38); 3872 emit_int8(0x58); 3873 emit_int8((unsigned char)(0xC0 | encode)); 3874 } 3875 3876 // Carry-Less Multiplication Quadword 3877 void Assembler::pclmulqdq(XMMRegister dst, XMMRegister src, int mask) { 3878 assert(VM_Version::supports_clmul(), ""); 3879 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_3A); 3880 emit_int8(0x44); 3881 emit_int8((unsigned char)(0xC0 | encode)); 3882 emit_int8((unsigned char)mask); 3883 } 3884 3885 // Carry-Less Multiplication Quadword 3886 void Assembler::vpclmulqdq(XMMRegister dst, XMMRegister nds, XMMRegister src, int mask) { 3887 assert(VM_Version::supports_avx() && VM_Version::supports_clmul(), ""); 3888 bool vector256 = false; 3889 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A); 3890 emit_int8(0x44); 3891 emit_int8((unsigned char)(0xC0 | encode)); 3892 emit_int8((unsigned char)mask); 3893 } 3894 3895 void Assembler::vzeroupper() { 3896 assert(VM_Version::supports_avx(), ""); 3897 (void)vex_prefix_and_encode(xmm0, xmm0, xmm0, VEX_SIMD_NONE); 3898 emit_int8(0x77); 3899 } 3900 3901 3902 #ifndef _LP64 3903 // 32bit only pieces of the assembler 3904 3905 void Assembler::cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec) { 3906 // NO PREFIX AS NEVER 64BIT 3907 InstructionMark im(this); 3908 emit_int8((unsigned char)0x81); 3909 emit_int8((unsigned char)(0xF8 | src1->encoding())); 3910 emit_data(imm32, rspec, 0); 3911 } 3912 3913 void Assembler::cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec) { 3914 // NO PREFIX AS NEVER 64BIT (not even 32bit versions of 64bit regs 3915 InstructionMark im(this); 3916 emit_int8((unsigned char)0x81); 3917 emit_operand(rdi, src1); 3918 emit_data(imm32, rspec, 0); 3919 } 3920 3921 // The 64-bit (32bit platform) cmpxchg compares the value at adr with the contents of rdx:rax, 3922 // and stores rcx:rbx into adr if so; otherwise, the value at adr is loaded 3923 // into rdx:rax. The ZF is set if the compared values were equal, and cleared otherwise. 3924 void Assembler::cmpxchg8(Address adr) { 3925 InstructionMark im(this); 3926 emit_int8(0x0F); 3927 emit_int8((unsigned char)0xC7); 3928 emit_operand(rcx, adr); 3929 } 3930 3931 void Assembler::decl(Register dst) { 3932 // Don't use it directly. Use MacroAssembler::decrementl() instead. 3933 emit_int8(0x48 | dst->encoding()); 3934 } 3935 3936 #endif // _LP64 3937 3938 // 64bit typically doesn't use the x87 but needs to for the trig funcs 3939 3940 void Assembler::fabs() { 3941 emit_int8((unsigned char)0xD9); 3942 emit_int8((unsigned char)0xE1); 3943 } 3944 3945 void Assembler::fadd(int i) { 3946 emit_farith(0xD8, 0xC0, i); 3947 } 3948 3949 void Assembler::fadd_d(Address src) { 3950 InstructionMark im(this); 3951 emit_int8((unsigned char)0xDC); 3952 emit_operand32(rax, src); 3953 } 3954 3955 void Assembler::fadd_s(Address src) { 3956 InstructionMark im(this); 3957 emit_int8((unsigned char)0xD8); 3958 emit_operand32(rax, src); 3959 } 3960 3961 void Assembler::fadda(int i) { 3962 emit_farith(0xDC, 0xC0, i); 3963 } 3964 3965 void Assembler::faddp(int i) { 3966 emit_farith(0xDE, 0xC0, i); 3967 } 3968 3969 void Assembler::fchs() { 3970 emit_int8((unsigned char)0xD9); 3971 emit_int8((unsigned char)0xE0); 3972 } 3973 3974 void Assembler::fcom(int i) { 3975 emit_farith(0xD8, 0xD0, i); 3976 } 3977 3978 void Assembler::fcomp(int i) { 3979 emit_farith(0xD8, 0xD8, i); 3980 } 3981 3982 void Assembler::fcomp_d(Address src) { 3983 InstructionMark im(this); 3984 emit_int8((unsigned char)0xDC); 3985 emit_operand32(rbx, src); 3986 } 3987 3988 void Assembler::fcomp_s(Address src) { 3989 InstructionMark im(this); 3990 emit_int8((unsigned char)0xD8); 3991 emit_operand32(rbx, src); 3992 } 3993 3994 void Assembler::fcompp() { 3995 emit_int8((unsigned char)0xDE); 3996 emit_int8((unsigned char)0xD9); 3997 } 3998 3999 void Assembler::fcos() { 4000 emit_int8((unsigned char)0xD9); 4001 emit_int8((unsigned char)0xFF); 4002 } 4003 4004 void Assembler::fdecstp() { 4005 emit_int8((unsigned char)0xD9); 4006 emit_int8((unsigned char)0xF6); 4007 } 4008 4009 void Assembler::fdiv(int i) { 4010 emit_farith(0xD8, 0xF0, i); 4011 } 4012 4013 void Assembler::fdiv_d(Address src) { 4014 InstructionMark im(this); 4015 emit_int8((unsigned char)0xDC); 4016 emit_operand32(rsi, src); 4017 } 4018 4019 void Assembler::fdiv_s(Address src) { 4020 InstructionMark im(this); 4021 emit_int8((unsigned char)0xD8); 4022 emit_operand32(rsi, src); 4023 } 4024 4025 void Assembler::fdiva(int i) { 4026 emit_farith(0xDC, 0xF8, i); 4027 } 4028 4029 // Note: The Intel manual (Pentium Processor User's Manual, Vol.3, 1994) 4030 // is erroneous for some of the floating-point instructions below. 4031 4032 void Assembler::fdivp(int i) { 4033 emit_farith(0xDE, 0xF8, i); // ST(0) <- ST(0) / ST(1) and pop (Intel manual wrong) 4034 } 4035 4036 void Assembler::fdivr(int i) { 4037 emit_farith(0xD8, 0xF8, i); 4038 } 4039 4040 void Assembler::fdivr_d(Address src) { 4041 InstructionMark im(this); 4042 emit_int8((unsigned char)0xDC); 4043 emit_operand32(rdi, src); 4044 } 4045 4046 void Assembler::fdivr_s(Address src) { 4047 InstructionMark im(this); 4048 emit_int8((unsigned char)0xD8); 4049 emit_operand32(rdi, src); 4050 } 4051 4052 void Assembler::fdivra(int i) { 4053 emit_farith(0xDC, 0xF0, i); 4054 } 4055 4056 void Assembler::fdivrp(int i) { 4057 emit_farith(0xDE, 0xF0, i); // ST(0) <- ST(1) / ST(0) and pop (Intel manual wrong) 4058 } 4059 4060 void Assembler::ffree(int i) { 4061 emit_farith(0xDD, 0xC0, i); 4062 } 4063 4064 void Assembler::fild_d(Address adr) { 4065 InstructionMark im(this); 4066 emit_int8((unsigned char)0xDF); 4067 emit_operand32(rbp, adr); 4068 } 4069 4070 void Assembler::fild_s(Address adr) { 4071 InstructionMark im(this); 4072 emit_int8((unsigned char)0xDB); 4073 emit_operand32(rax, adr); 4074 } 4075 4076 void Assembler::fincstp() { 4077 emit_int8((unsigned char)0xD9); 4078 emit_int8((unsigned char)0xF7); 4079 } 4080 4081 void Assembler::finit() { 4082 emit_int8((unsigned char)0x9B); 4083 emit_int8((unsigned char)0xDB); 4084 emit_int8((unsigned char)0xE3); 4085 } 4086 4087 void Assembler::fist_s(Address adr) { 4088 InstructionMark im(this); 4089 emit_int8((unsigned char)0xDB); 4090 emit_operand32(rdx, adr); 4091 } 4092 4093 void Assembler::fistp_d(Address adr) { 4094 InstructionMark im(this); 4095 emit_int8((unsigned char)0xDF); 4096 emit_operand32(rdi, adr); 4097 } 4098 4099 void Assembler::fistp_s(Address adr) { 4100 InstructionMark im(this); 4101 emit_int8((unsigned char)0xDB); 4102 emit_operand32(rbx, adr); 4103 } 4104 4105 void Assembler::fld1() { 4106 emit_int8((unsigned char)0xD9); 4107 emit_int8((unsigned char)0xE8); 4108 } 4109 4110 void Assembler::fld_d(Address adr) { 4111 InstructionMark im(this); 4112 emit_int8((unsigned char)0xDD); 4113 emit_operand32(rax, adr); 4114 } 4115 4116 void Assembler::fld_s(Address adr) { 4117 InstructionMark im(this); 4118 emit_int8((unsigned char)0xD9); 4119 emit_operand32(rax, adr); 4120 } 4121 4122 4123 void Assembler::fld_s(int index) { 4124 emit_farith(0xD9, 0xC0, index); 4125 } 4126 4127 void Assembler::fld_x(Address adr) { 4128 InstructionMark im(this); 4129 emit_int8((unsigned char)0xDB); 4130 emit_operand32(rbp, adr); 4131 } 4132 4133 void Assembler::fldcw(Address src) { 4134 InstructionMark im(this); 4135 emit_int8((unsigned char)0xD9); 4136 emit_operand32(rbp, src); 4137 } 4138 4139 void Assembler::fldenv(Address src) { 4140 InstructionMark im(this); 4141 emit_int8((unsigned char)0xD9); 4142 emit_operand32(rsp, src); 4143 } 4144 4145 void Assembler::fldlg2() { 4146 emit_int8((unsigned char)0xD9); 4147 emit_int8((unsigned char)0xEC); 4148 } 4149 4150 void Assembler::fldln2() { 4151 emit_int8((unsigned char)0xD9); 4152 emit_int8((unsigned char)0xED); 4153 } 4154 4155 void Assembler::fldz() { 4156 emit_int8((unsigned char)0xD9); 4157 emit_int8((unsigned char)0xEE); 4158 } 4159 4160 void Assembler::flog() { 4161 fldln2(); 4162 fxch(); 4163 fyl2x(); 4164 } 4165 4166 void Assembler::flog10() { 4167 fldlg2(); 4168 fxch(); 4169 fyl2x(); 4170 } 4171 4172 void Assembler::fmul(int i) { 4173 emit_farith(0xD8, 0xC8, i); 4174 } 4175 4176 void Assembler::fmul_d(Address src) { 4177 InstructionMark im(this); 4178 emit_int8((unsigned char)0xDC); 4179 emit_operand32(rcx, src); 4180 } 4181 4182 void Assembler::fmul_s(Address src) { 4183 InstructionMark im(this); 4184 emit_int8((unsigned char)0xD8); 4185 emit_operand32(rcx, src); 4186 } 4187 4188 void Assembler::fmula(int i) { 4189 emit_farith(0xDC, 0xC8, i); 4190 } 4191 4192 void Assembler::fmulp(int i) { 4193 emit_farith(0xDE, 0xC8, i); 4194 } 4195 4196 void Assembler::fnsave(Address dst) { 4197 InstructionMark im(this); 4198 emit_int8((unsigned char)0xDD); 4199 emit_operand32(rsi, dst); 4200 } 4201 4202 void Assembler::fnstcw(Address src) { 4203 InstructionMark im(this); 4204 emit_int8((unsigned char)0x9B); 4205 emit_int8((unsigned char)0xD9); 4206 emit_operand32(rdi, src); 4207 } 4208 4209 void Assembler::fnstsw_ax() { 4210 emit_int8((unsigned char)0xDF); 4211 emit_int8((unsigned char)0xE0); 4212 } 4213 4214 void Assembler::fprem() { 4215 emit_int8((unsigned char)0xD9); 4216 emit_int8((unsigned char)0xF8); 4217 } 4218 4219 void Assembler::fprem1() { 4220 emit_int8((unsigned char)0xD9); 4221 emit_int8((unsigned char)0xF5); 4222 } 4223 4224 void Assembler::frstor(Address src) { 4225 InstructionMark im(this); 4226 emit_int8((unsigned char)0xDD); 4227 emit_operand32(rsp, src); 4228 } 4229 4230 void Assembler::fsin() { 4231 emit_int8((unsigned char)0xD9); 4232 emit_int8((unsigned char)0xFE); 4233 } 4234 4235 void Assembler::fsqrt() { 4236 emit_int8((unsigned char)0xD9); 4237 emit_int8((unsigned char)0xFA); 4238 } 4239 4240 void Assembler::fst_d(Address adr) { 4241 InstructionMark im(this); 4242 emit_int8((unsigned char)0xDD); 4243 emit_operand32(rdx, adr); 4244 } 4245 4246 void Assembler::fst_s(Address adr) { 4247 InstructionMark im(this); 4248 emit_int8((unsigned char)0xD9); 4249 emit_operand32(rdx, adr); 4250 } 4251 4252 void Assembler::fstp_d(Address adr) { 4253 InstructionMark im(this); 4254 emit_int8((unsigned char)0xDD); 4255 emit_operand32(rbx, adr); 4256 } 4257 4258 void Assembler::fstp_d(int index) { 4259 emit_farith(0xDD, 0xD8, index); 4260 } 4261 4262 void Assembler::fstp_s(Address adr) { 4263 InstructionMark im(this); 4264 emit_int8((unsigned char)0xD9); 4265 emit_operand32(rbx, adr); 4266 } 4267 4268 void Assembler::fstp_x(Address adr) { 4269 InstructionMark im(this); 4270 emit_int8((unsigned char)0xDB); 4271 emit_operand32(rdi, adr); 4272 } 4273 4274 void Assembler::fsub(int i) { 4275 emit_farith(0xD8, 0xE0, i); 4276 } 4277 4278 void Assembler::fsub_d(Address src) { 4279 InstructionMark im(this); 4280 emit_int8((unsigned char)0xDC); 4281 emit_operand32(rsp, src); 4282 } 4283 4284 void Assembler::fsub_s(Address src) { 4285 InstructionMark im(this); 4286 emit_int8((unsigned char)0xD8); 4287 emit_operand32(rsp, src); 4288 } 4289 4290 void Assembler::fsuba(int i) { 4291 emit_farith(0xDC, 0xE8, i); 4292 } 4293 4294 void Assembler::fsubp(int i) { 4295 emit_farith(0xDE, 0xE8, i); // ST(0) <- ST(0) - ST(1) and pop (Intel manual wrong) 4296 } 4297 4298 void Assembler::fsubr(int i) { 4299 emit_farith(0xD8, 0xE8, i); 4300 } 4301 4302 void Assembler::fsubr_d(Address src) { 4303 InstructionMark im(this); 4304 emit_int8((unsigned char)0xDC); 4305 emit_operand32(rbp, src); 4306 } 4307 4308 void Assembler::fsubr_s(Address src) { 4309 InstructionMark im(this); 4310 emit_int8((unsigned char)0xD8); 4311 emit_operand32(rbp, src); 4312 } 4313 4314 void Assembler::fsubra(int i) { 4315 emit_farith(0xDC, 0xE0, i); 4316 } 4317 4318 void Assembler::fsubrp(int i) { 4319 emit_farith(0xDE, 0xE0, i); // ST(0) <- ST(1) - ST(0) and pop (Intel manual wrong) 4320 } 4321 4322 void Assembler::ftan() { 4323 emit_int8((unsigned char)0xD9); 4324 emit_int8((unsigned char)0xF2); 4325 emit_int8((unsigned char)0xDD); 4326 emit_int8((unsigned char)0xD8); 4327 } 4328 4329 void Assembler::ftst() { 4330 emit_int8((unsigned char)0xD9); 4331 emit_int8((unsigned char)0xE4); 4332 } 4333 4334 void Assembler::fucomi(int i) { 4335 // make sure the instruction is supported (introduced for P6, together with cmov) 4336 guarantee(VM_Version::supports_cmov(), "illegal instruction"); 4337 emit_farith(0xDB, 0xE8, i); 4338 } 4339 4340 void Assembler::fucomip(int i) { 4341 // make sure the instruction is supported (introduced for P6, together with cmov) 4342 guarantee(VM_Version::supports_cmov(), "illegal instruction"); 4343 emit_farith(0xDF, 0xE8, i); 4344 } 4345 4346 void Assembler::fwait() { 4347 emit_int8((unsigned char)0x9B); 4348 } 4349 4350 void Assembler::fxch(int i) { 4351 emit_farith(0xD9, 0xC8, i); 4352 } 4353 4354 void Assembler::fyl2x() { 4355 emit_int8((unsigned char)0xD9); 4356 emit_int8((unsigned char)0xF1); 4357 } 4358 4359 void Assembler::frndint() { 4360 emit_int8((unsigned char)0xD9); 4361 emit_int8((unsigned char)0xFC); 4362 } 4363 4364 void Assembler::f2xm1() { 4365 emit_int8((unsigned char)0xD9); 4366 emit_int8((unsigned char)0xF0); 4367 } 4368 4369 void Assembler::fldl2e() { 4370 emit_int8((unsigned char)0xD9); 4371 emit_int8((unsigned char)0xEA); 4372 } 4373 4374 // SSE SIMD prefix byte values corresponding to VexSimdPrefix encoding. 4375 static int simd_pre[4] = { 0, 0x66, 0xF3, 0xF2 }; 4376 // SSE opcode second byte values (first is 0x0F) corresponding to VexOpcode encoding. 4377 static int simd_opc[4] = { 0, 0, 0x38, 0x3A }; 4378 4379 // Generate SSE legacy REX prefix and SIMD opcode based on VEX encoding. 4380 void Assembler::rex_prefix(Address adr, XMMRegister xreg, VexSimdPrefix pre, VexOpcode opc, bool rex_w) { 4381 if (pre > 0) { 4382 emit_int8(simd_pre[pre]); 4383 } 4384 if (rex_w) { 4385 prefixq(adr, xreg); 4386 } else { 4387 prefix(adr, xreg); 4388 } 4389 if (opc > 0) { 4390 emit_int8(0x0F); 4391 int opc2 = simd_opc[opc]; 4392 if (opc2 > 0) { 4393 emit_int8(opc2); 4394 } 4395 } 4396 } 4397 4398 int Assembler::rex_prefix_and_encode(int dst_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool rex_w) { 4399 if (pre > 0) { 4400 emit_int8(simd_pre[pre]); 4401 } 4402 int encode = (rex_w) ? prefixq_and_encode(dst_enc, src_enc) : 4403 prefix_and_encode(dst_enc, src_enc); 4404 if (opc > 0) { 4405 emit_int8(0x0F); 4406 int opc2 = simd_opc[opc]; 4407 if (opc2 > 0) { 4408 emit_int8(opc2); 4409 } 4410 } 4411 return encode; 4412 } 4413 4414 4415 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) { 4416 if (vex_b || vex_x || vex_w || (opc == VEX_OPCODE_0F_38) || (opc == VEX_OPCODE_0F_3A)) { 4417 prefix(VEX_3bytes); 4418 4419 int byte1 = (vex_r ? VEX_R : 0) | (vex_x ? VEX_X : 0) | (vex_b ? VEX_B : 0); 4420 byte1 = (~byte1) & 0xE0; 4421 byte1 |= opc; 4422 emit_int8(byte1); 4423 4424 int byte2 = ((~nds_enc) & 0xf) << 3; 4425 byte2 |= (vex_w ? VEX_W : 0) | (vector256 ? 4 : 0) | pre; 4426 emit_int8(byte2); 4427 } else { 4428 prefix(VEX_2bytes); 4429 4430 int byte1 = vex_r ? VEX_R : 0; 4431 byte1 = (~byte1) & 0x80; 4432 byte1 |= ((~nds_enc) & 0xf) << 3; 4433 byte1 |= (vector256 ? 4 : 0) | pre; 4434 emit_int8(byte1); 4435 } 4436 } 4437 4438 void Assembler::vex_prefix(Address adr, int nds_enc, int xreg_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256){ 4439 bool vex_r = (xreg_enc >= 8); 4440 bool vex_b = adr.base_needs_rex(); 4441 bool vex_x = adr.index_needs_rex(); 4442 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256); 4443 } 4444 4445 int Assembler::vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256) { 4446 bool vex_r = (dst_enc >= 8); 4447 bool vex_b = (src_enc >= 8); 4448 bool vex_x = false; 4449 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256); 4450 return (((dst_enc & 7) << 3) | (src_enc & 7)); 4451 } 4452 4453 4454 void Assembler::simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) { 4455 if (UseAVX > 0) { 4456 int xreg_enc = xreg->encoding(); 4457 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 4458 vex_prefix(adr, nds_enc, xreg_enc, pre, opc, rex_w, vector256); 4459 } else { 4460 assert((nds == xreg) || (nds == xnoreg), "wrong sse encoding"); 4461 rex_prefix(adr, xreg, pre, opc, rex_w); 4462 } 4463 } 4464 4465 int Assembler::simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) { 4466 int dst_enc = dst->encoding(); 4467 int src_enc = src->encoding(); 4468 if (UseAVX > 0) { 4469 int nds_enc = nds->is_valid() ? nds->encoding() : 0; 4470 return vex_prefix_and_encode(dst_enc, nds_enc, src_enc, pre, opc, rex_w, vector256); 4471 } else { 4472 assert((nds == dst) || (nds == src) || (nds == xnoreg), "wrong sse encoding"); 4473 return rex_prefix_and_encode(dst_enc, src_enc, pre, opc, rex_w); 4474 } 4475 } 4476 4477 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) { 4478 InstructionMark im(this); 4479 simd_prefix(dst, dst, src, pre); 4480 emit_int8(opcode); 4481 emit_operand(dst, src); 4482 } 4483 4484 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) { 4485 int encode = simd_prefix_and_encode(dst, dst, src, pre); 4486 emit_int8(opcode); 4487 emit_int8((unsigned char)(0xC0 | encode)); 4488 } 4489 4490 // Versions with no second source register (non-destructive source). 4491 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) { 4492 InstructionMark im(this); 4493 simd_prefix(dst, xnoreg, src, pre); 4494 emit_int8(opcode); 4495 emit_operand(dst, src); 4496 } 4497 4498 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) { 4499 int encode = simd_prefix_and_encode(dst, xnoreg, src, pre); 4500 emit_int8(opcode); 4501 emit_int8((unsigned char)(0xC0 | encode)); 4502 } 4503 4504 // 3-operands AVX instructions 4505 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds, 4506 Address src, VexSimdPrefix pre, bool vector256) { 4507 InstructionMark im(this); 4508 vex_prefix(dst, nds, src, pre, vector256); 4509 emit_int8(opcode); 4510 emit_operand(dst, src); 4511 } 4512 4513 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds, 4514 XMMRegister src, VexSimdPrefix pre, bool vector256) { 4515 int encode = vex_prefix_and_encode(dst, nds, src, pre, vector256); 4516 emit_int8(opcode); 4517 emit_int8((unsigned char)(0xC0 | encode)); 4518 } 4519 4520 #ifndef _LP64 4521 4522 void Assembler::incl(Register dst) { 4523 // Don't use it directly. Use MacroAssembler::incrementl() instead. 4524 emit_int8(0x40 | dst->encoding()); 4525 } 4526 4527 void Assembler::lea(Register dst, Address src) { 4528 leal(dst, src); 4529 } 4530 4531 void Assembler::mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec) { 4532 InstructionMark im(this); 4533 emit_int8((unsigned char)0xC7); 4534 emit_operand(rax, dst); 4535 emit_data((int)imm32, rspec, 0); 4536 } 4537 4538 void Assembler::mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec) { 4539 InstructionMark im(this); 4540 int encode = prefix_and_encode(dst->encoding()); 4541 emit_int8((unsigned char)(0xB8 | encode)); 4542 emit_data((int)imm32, rspec, 0); 4543 } 4544 4545 void Assembler::popa() { // 32bit 4546 emit_int8(0x61); 4547 } 4548 4549 void Assembler::push_literal32(int32_t imm32, RelocationHolder const& rspec) { 4550 InstructionMark im(this); 4551 emit_int8(0x68); 4552 emit_data(imm32, rspec, 0); 4553 } 4554 4555 void Assembler::pusha() { // 32bit 4556 emit_int8(0x60); 4557 } 4558 4559 void Assembler::set_byte_if_not_zero(Register dst) { 4560 emit_int8(0x0F); 4561 emit_int8((unsigned char)0x95); 4562 emit_int8((unsigned char)(0xE0 | dst->encoding())); 4563 } 4564 4565 void Assembler::shldl(Register dst, Register src) { 4566 emit_int8(0x0F); 4567 emit_int8((unsigned char)0xA5); 4568 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding())); 4569 } 4570 4571 void Assembler::shrdl(Register dst, Register src) { 4572 emit_int8(0x0F); 4573 emit_int8((unsigned char)0xAD); 4574 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding())); 4575 } 4576 4577 #else // LP64 4578 4579 void Assembler::set_byte_if_not_zero(Register dst) { 4580 int enc = prefix_and_encode(dst->encoding(), true); 4581 emit_int8(0x0F); 4582 emit_int8((unsigned char)0x95); 4583 emit_int8((unsigned char)(0xE0 | enc)); 4584 } 4585 4586 // 64bit only pieces of the assembler 4587 // This should only be used by 64bit instructions that can use rip-relative 4588 // it cannot be used by instructions that want an immediate value. 4589 4590 bool Assembler::reachable(AddressLiteral adr) { 4591 int64_t disp; 4592 // None will force a 64bit literal to the code stream. Likely a placeholder 4593 // for something that will be patched later and we need to certain it will 4594 // always be reachable. 4595 if (adr.reloc() == relocInfo::none) { 4596 return false; 4597 } 4598 if (adr.reloc() == relocInfo::internal_word_type) { 4599 // This should be rip relative and easily reachable. 4600 return true; 4601 } 4602 if (adr.reloc() == relocInfo::virtual_call_type || 4603 adr.reloc() == relocInfo::opt_virtual_call_type || 4604 adr.reloc() == relocInfo::static_call_type || 4605 adr.reloc() == relocInfo::static_stub_type ) { 4606 // This should be rip relative within the code cache and easily 4607 // reachable until we get huge code caches. (At which point 4608 // ic code is going to have issues). 4609 return true; 4610 } 4611 if (adr.reloc() != relocInfo::external_word_type && 4612 adr.reloc() != relocInfo::poll_return_type && // these are really external_word but need special 4613 adr.reloc() != relocInfo::poll_type && // relocs to identify them 4614 adr.reloc() != relocInfo::runtime_call_type ) { 4615 return false; 4616 } 4617 4618 // Stress the correction code 4619 if (ForceUnreachable) { 4620 // Must be runtimecall reloc, see if it is in the codecache 4621 // Flipping stuff in the codecache to be unreachable causes issues 4622 // with things like inline caches where the additional instructions 4623 // are not handled. 4624 if (CodeCache::find_blob(adr._target) == NULL) { 4625 return false; 4626 } 4627 } 4628 // For external_word_type/runtime_call_type if it is reachable from where we 4629 // are now (possibly a temp buffer) and where we might end up 4630 // anywhere in the codeCache then we are always reachable. 4631 // This would have to change if we ever save/restore shared code 4632 // to be more pessimistic. 4633 disp = (int64_t)adr._target - ((int64_t)CodeCache::low_bound() + sizeof(int)); 4634 if (!is_simm32(disp)) return false; 4635 disp = (int64_t)adr._target - ((int64_t)CodeCache::high_bound() + sizeof(int)); 4636 if (!is_simm32(disp)) return false; 4637 4638 disp = (int64_t)adr._target - ((int64_t)pc() + sizeof(int)); 4639 4640 // Because rip relative is a disp + address_of_next_instruction and we 4641 // don't know the value of address_of_next_instruction we apply a fudge factor 4642 // to make sure we will be ok no matter the size of the instruction we get placed into. 4643 // We don't have to fudge the checks above here because they are already worst case. 4644 4645 // 12 == override/rex byte, opcode byte, rm byte, sib byte, a 4-byte disp , 4-byte literal 4646 // + 4 because better safe than sorry. 4647 const int fudge = 12 + 4; 4648 if (disp < 0) { 4649 disp -= fudge; 4650 } else { 4651 disp += fudge; 4652 } 4653 return is_simm32(disp); 4654 } 4655 4656 // Check if the polling page is not reachable from the code cache using rip-relative 4657 // addressing. 4658 bool Assembler::is_polling_page_far() { 4659 intptr_t addr = (intptr_t)os::get_polling_page(); 4660 return ForceUnreachable || 4661 !is_simm32(addr - (intptr_t)CodeCache::low_bound()) || 4662 !is_simm32(addr - (intptr_t)CodeCache::high_bound()); 4663 } 4664 4665 void Assembler::emit_data64(jlong data, 4666 relocInfo::relocType rtype, 4667 int format) { 4668 if (rtype == relocInfo::none) { 4669 emit_int64(data); 4670 } else { 4671 emit_data64(data, Relocation::spec_simple(rtype), format); 4672 } 4673 } 4674 4675 void Assembler::emit_data64(jlong data, 4676 RelocationHolder const& rspec, 4677 int format) { 4678 assert(imm_operand == 0, "default format must be immediate in this file"); 4679 assert(imm_operand == format, "must be immediate"); 4680 assert(inst_mark() != NULL, "must be inside InstructionMark"); 4681 // Do not use AbstractAssembler::relocate, which is not intended for 4682 // embedded words. Instead, relocate to the enclosing instruction. 4683 code_section()->relocate(inst_mark(), rspec, format); 4684 #ifdef ASSERT 4685 check_relocation(rspec, format); 4686 #endif 4687 emit_int64(data); 4688 } 4689 4690 int Assembler::prefix_and_encode(int reg_enc, bool byteinst) { 4691 if (reg_enc >= 8) { 4692 prefix(REX_B); 4693 reg_enc -= 8; 4694 } else if (byteinst && reg_enc >= 4) { 4695 prefix(REX); 4696 } 4697 return reg_enc; 4698 } 4699 4700 int Assembler::prefixq_and_encode(int reg_enc) { 4701 if (reg_enc < 8) { 4702 prefix(REX_W); 4703 } else { 4704 prefix(REX_WB); 4705 reg_enc -= 8; 4706 } 4707 return reg_enc; 4708 } 4709 4710 int Assembler::prefix_and_encode(int dst_enc, int src_enc, bool byteinst) { 4711 if (dst_enc < 8) { 4712 if (src_enc >= 8) { 4713 prefix(REX_B); 4714 src_enc -= 8; 4715 } else if (byteinst && src_enc >= 4) { 4716 prefix(REX); 4717 } 4718 } else { 4719 if (src_enc < 8) { 4720 prefix(REX_R); 4721 } else { 4722 prefix(REX_RB); 4723 src_enc -= 8; 4724 } 4725 dst_enc -= 8; 4726 } 4727 return dst_enc << 3 | src_enc; 4728 } 4729 4730 int Assembler::prefixq_and_encode(int dst_enc, int src_enc) { 4731 if (dst_enc < 8) { 4732 if (src_enc < 8) { 4733 prefix(REX_W); 4734 } else { 4735 prefix(REX_WB); 4736 src_enc -= 8; 4737 } 4738 } else { 4739 if (src_enc < 8) { 4740 prefix(REX_WR); 4741 } else { 4742 prefix(REX_WRB); 4743 src_enc -= 8; 4744 } 4745 dst_enc -= 8; 4746 } 4747 return dst_enc << 3 | src_enc; 4748 } 4749 4750 void Assembler::prefix(Register reg) { 4751 if (reg->encoding() >= 8) { 4752 prefix(REX_B); 4753 } 4754 } 4755 4756 void Assembler::prefix(Address adr) { 4757 if (adr.base_needs_rex()) { 4758 if (adr.index_needs_rex()) { 4759 prefix(REX_XB); 4760 } else { 4761 prefix(REX_B); 4762 } 4763 } else { 4764 if (adr.index_needs_rex()) { 4765 prefix(REX_X); 4766 } 4767 } 4768 } 4769 4770 void Assembler::prefixq(Address adr) { 4771 if (adr.base_needs_rex()) { 4772 if (adr.index_needs_rex()) { 4773 prefix(REX_WXB); 4774 } else { 4775 prefix(REX_WB); 4776 } 4777 } else { 4778 if (adr.index_needs_rex()) { 4779 prefix(REX_WX); 4780 } else { 4781 prefix(REX_W); 4782 } 4783 } 4784 } 4785 4786 4787 void Assembler::prefix(Address adr, Register reg, bool byteinst) { 4788 if (reg->encoding() < 8) { 4789 if (adr.base_needs_rex()) { 4790 if (adr.index_needs_rex()) { 4791 prefix(REX_XB); 4792 } else { 4793 prefix(REX_B); 4794 } 4795 } else { 4796 if (adr.index_needs_rex()) { 4797 prefix(REX_X); 4798 } else if (byteinst && reg->encoding() >= 4 ) { 4799 prefix(REX); 4800 } 4801 } 4802 } else { 4803 if (adr.base_needs_rex()) { 4804 if (adr.index_needs_rex()) { 4805 prefix(REX_RXB); 4806 } else { 4807 prefix(REX_RB); 4808 } 4809 } else { 4810 if (adr.index_needs_rex()) { 4811 prefix(REX_RX); 4812 } else { 4813 prefix(REX_R); 4814 } 4815 } 4816 } 4817 } 4818 4819 void Assembler::prefixq(Address adr, Register src) { 4820 if (src->encoding() < 8) { 4821 if (adr.base_needs_rex()) { 4822 if (adr.index_needs_rex()) { 4823 prefix(REX_WXB); 4824 } else { 4825 prefix(REX_WB); 4826 } 4827 } else { 4828 if (adr.index_needs_rex()) { 4829 prefix(REX_WX); 4830 } else { 4831 prefix(REX_W); 4832 } 4833 } 4834 } else { 4835 if (adr.base_needs_rex()) { 4836 if (adr.index_needs_rex()) { 4837 prefix(REX_WRXB); 4838 } else { 4839 prefix(REX_WRB); 4840 } 4841 } else { 4842 if (adr.index_needs_rex()) { 4843 prefix(REX_WRX); 4844 } else { 4845 prefix(REX_WR); 4846 } 4847 } 4848 } 4849 } 4850 4851 void Assembler::prefix(Address adr, XMMRegister reg) { 4852 if (reg->encoding() < 8) { 4853 if (adr.base_needs_rex()) { 4854 if (adr.index_needs_rex()) { 4855 prefix(REX_XB); 4856 } else { 4857 prefix(REX_B); 4858 } 4859 } else { 4860 if (adr.index_needs_rex()) { 4861 prefix(REX_X); 4862 } 4863 } 4864 } else { 4865 if (adr.base_needs_rex()) { 4866 if (adr.index_needs_rex()) { 4867 prefix(REX_RXB); 4868 } else { 4869 prefix(REX_RB); 4870 } 4871 } else { 4872 if (adr.index_needs_rex()) { 4873 prefix(REX_RX); 4874 } else { 4875 prefix(REX_R); 4876 } 4877 } 4878 } 4879 } 4880 4881 void Assembler::prefixq(Address adr, XMMRegister src) { 4882 if (src->encoding() < 8) { 4883 if (adr.base_needs_rex()) { 4884 if (adr.index_needs_rex()) { 4885 prefix(REX_WXB); 4886 } else { 4887 prefix(REX_WB); 4888 } 4889 } else { 4890 if (adr.index_needs_rex()) { 4891 prefix(REX_WX); 4892 } else { 4893 prefix(REX_W); 4894 } 4895 } 4896 } else { 4897 if (adr.base_needs_rex()) { 4898 if (adr.index_needs_rex()) { 4899 prefix(REX_WRXB); 4900 } else { 4901 prefix(REX_WRB); 4902 } 4903 } else { 4904 if (adr.index_needs_rex()) { 4905 prefix(REX_WRX); 4906 } else { 4907 prefix(REX_WR); 4908 } 4909 } 4910 } 4911 } 4912 4913 void Assembler::adcq(Register dst, int32_t imm32) { 4914 (void) prefixq_and_encode(dst->encoding()); 4915 emit_arith(0x81, 0xD0, dst, imm32); 4916 } 4917 4918 void Assembler::adcq(Register dst, Address src) { 4919 InstructionMark im(this); 4920 prefixq(src, dst); 4921 emit_int8(0x13); 4922 emit_operand(dst, src); 4923 } 4924 4925 void Assembler::adcq(Register dst, Register src) { 4926 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 4927 emit_arith(0x13, 0xC0, dst, src); 4928 } 4929 4930 void Assembler::addq(Address dst, int32_t imm32) { 4931 InstructionMark im(this); 4932 prefixq(dst); 4933 emit_arith_operand(0x81, rax, dst,imm32); 4934 } 4935 4936 void Assembler::addq(Address dst, Register src) { 4937 InstructionMark im(this); 4938 prefixq(dst, src); 4939 emit_int8(0x01); 4940 emit_operand(src, dst); 4941 } 4942 4943 void Assembler::addq(Register dst, int32_t imm32) { 4944 (void) prefixq_and_encode(dst->encoding()); 4945 emit_arith(0x81, 0xC0, dst, imm32); 4946 } 4947 4948 void Assembler::addq(Register dst, Address src) { 4949 InstructionMark im(this); 4950 prefixq(src, dst); 4951 emit_int8(0x03); 4952 emit_operand(dst, src); 4953 } 4954 4955 void Assembler::addq(Register dst, Register src) { 4956 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 4957 emit_arith(0x03, 0xC0, dst, src); 4958 } 4959 4960 void Assembler::adcxq(Register dst, Register src) { 4961 //assert(VM_Version::supports_adx(), "adx instructions not supported"); 4962 emit_int8((unsigned char)0x66); 4963 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 4964 emit_int8(0x0F); 4965 emit_int8(0x38); 4966 emit_int8((unsigned char)0xF6); 4967 emit_int8((unsigned char)(0xC0 | encode)); 4968 } 4969 4970 void Assembler::adoxq(Register dst, Register src) { 4971 //assert(VM_Version::supports_adx(), "adx instructions not supported"); 4972 emit_int8((unsigned char)0xF3); 4973 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 4974 emit_int8(0x0F); 4975 emit_int8(0x38); 4976 emit_int8((unsigned char)0xF6); 4977 emit_int8((unsigned char)(0xC0 | encode)); 4978 } 4979 4980 void Assembler::andq(Address dst, int32_t imm32) { 4981 InstructionMark im(this); 4982 prefixq(dst); 4983 emit_int8((unsigned char)0x81); 4984 emit_operand(rsp, dst, 4); 4985 emit_int32(imm32); 4986 } 4987 4988 void Assembler::andq(Register dst, int32_t imm32) { 4989 (void) prefixq_and_encode(dst->encoding()); 4990 emit_arith(0x81, 0xE0, dst, imm32); 4991 } 4992 4993 void Assembler::andq(Register dst, Address src) { 4994 InstructionMark im(this); 4995 prefixq(src, dst); 4996 emit_int8(0x23); 4997 emit_operand(dst, src); 4998 } 4999 5000 void Assembler::andq(Register dst, Register src) { 5001 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5002 emit_arith(0x23, 0xC0, dst, src); 5003 } 5004 5005 void Assembler::andnq(Register dst, Register src1, Register src2) { 5006 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5007 int encode = vex_prefix_0F38_and_encode_q(dst, src1, src2); 5008 emit_int8((unsigned char)0xF2); 5009 emit_int8((unsigned char)(0xC0 | encode)); 5010 } 5011 5012 void Assembler::andnq(Register dst, Register src1, Address src2) { 5013 InstructionMark im(this); 5014 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5015 vex_prefix_0F38_q(dst, src1, src2); 5016 emit_int8((unsigned char)0xF2); 5017 emit_operand(dst, src2); 5018 } 5019 5020 void Assembler::bsfq(Register dst, Register src) { 5021 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5022 emit_int8(0x0F); 5023 emit_int8((unsigned char)0xBC); 5024 emit_int8((unsigned char)(0xC0 | encode)); 5025 } 5026 5027 void Assembler::bsrq(Register dst, Register src) { 5028 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5029 emit_int8(0x0F); 5030 emit_int8((unsigned char)0xBD); 5031 emit_int8((unsigned char)(0xC0 | encode)); 5032 } 5033 5034 void Assembler::bswapq(Register reg) { 5035 int encode = prefixq_and_encode(reg->encoding()); 5036 emit_int8(0x0F); 5037 emit_int8((unsigned char)(0xC8 | encode)); 5038 } 5039 5040 void Assembler::blsiq(Register dst, Register src) { 5041 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5042 int encode = vex_prefix_0F38_and_encode_q(rbx, dst, src); 5043 emit_int8((unsigned char)0xF3); 5044 emit_int8((unsigned char)(0xC0 | encode)); 5045 } 5046 5047 void Assembler::blsiq(Register dst, Address src) { 5048 InstructionMark im(this); 5049 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5050 vex_prefix_0F38_q(rbx, dst, src); 5051 emit_int8((unsigned char)0xF3); 5052 emit_operand(rbx, src); 5053 } 5054 5055 void Assembler::blsmskq(Register dst, Register src) { 5056 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5057 int encode = vex_prefix_0F38_and_encode_q(rdx, dst, src); 5058 emit_int8((unsigned char)0xF3); 5059 emit_int8((unsigned char)(0xC0 | encode)); 5060 } 5061 5062 void Assembler::blsmskq(Register dst, Address src) { 5063 InstructionMark im(this); 5064 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5065 vex_prefix_0F38_q(rdx, dst, src); 5066 emit_int8((unsigned char)0xF3); 5067 emit_operand(rdx, src); 5068 } 5069 5070 void Assembler::blsrq(Register dst, Register src) { 5071 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5072 int encode = vex_prefix_0F38_and_encode_q(rcx, dst, src); 5073 emit_int8((unsigned char)0xF3); 5074 emit_int8((unsigned char)(0xC0 | encode)); 5075 } 5076 5077 void Assembler::blsrq(Register dst, Address src) { 5078 InstructionMark im(this); 5079 assert(VM_Version::supports_bmi1(), "bit manipulation instructions not supported"); 5080 vex_prefix_0F38_q(rcx, dst, src); 5081 emit_int8((unsigned char)0xF3); 5082 emit_operand(rcx, src); 5083 } 5084 5085 void Assembler::cdqq() { 5086 prefix(REX_W); 5087 emit_int8((unsigned char)0x99); 5088 } 5089 5090 void Assembler::clflush(Address adr) { 5091 prefix(adr); 5092 emit_int8(0x0F); 5093 emit_int8((unsigned char)0xAE); 5094 emit_operand(rdi, adr); 5095 } 5096 5097 void Assembler::cmovq(Condition cc, Register dst, Register src) { 5098 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5099 emit_int8(0x0F); 5100 emit_int8(0x40 | cc); 5101 emit_int8((unsigned char)(0xC0 | encode)); 5102 } 5103 5104 void Assembler::cmovq(Condition cc, Register dst, Address src) { 5105 InstructionMark im(this); 5106 prefixq(src, dst); 5107 emit_int8(0x0F); 5108 emit_int8(0x40 | cc); 5109 emit_operand(dst, src); 5110 } 5111 5112 void Assembler::cmpq(Address dst, int32_t imm32) { 5113 InstructionMark im(this); 5114 prefixq(dst); 5115 emit_int8((unsigned char)0x81); 5116 emit_operand(rdi, dst, 4); 5117 emit_int32(imm32); 5118 } 5119 5120 void Assembler::cmpq(Register dst, int32_t imm32) { 5121 (void) prefixq_and_encode(dst->encoding()); 5122 emit_arith(0x81, 0xF8, dst, imm32); 5123 } 5124 5125 void Assembler::cmpq(Address dst, Register src) { 5126 InstructionMark im(this); 5127 prefixq(dst, src); 5128 emit_int8(0x3B); 5129 emit_operand(src, dst); 5130 } 5131 5132 void Assembler::cmpq(Register dst, Register src) { 5133 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5134 emit_arith(0x3B, 0xC0, dst, src); 5135 } 5136 5137 void Assembler::cmpq(Register dst, Address src) { 5138 InstructionMark im(this); 5139 prefixq(src, dst); 5140 emit_int8(0x3B); 5141 emit_operand(dst, src); 5142 } 5143 5144 void Assembler::cmpxchgq(Register reg, Address adr) { 5145 InstructionMark im(this); 5146 prefixq(adr, reg); 5147 emit_int8(0x0F); 5148 emit_int8((unsigned char)0xB1); 5149 emit_operand(reg, adr); 5150 } 5151 5152 void Assembler::cvtsi2sdq(XMMRegister dst, Register src) { 5153 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5154 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F2); 5155 emit_int8(0x2A); 5156 emit_int8((unsigned char)(0xC0 | encode)); 5157 } 5158 5159 void Assembler::cvtsi2sdq(XMMRegister dst, Address src) { 5160 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5161 InstructionMark im(this); 5162 simd_prefix_q(dst, dst, src, VEX_SIMD_F2); 5163 emit_int8(0x2A); 5164 emit_operand(dst, src); 5165 } 5166 5167 void Assembler::cvtsi2ssq(XMMRegister dst, Register src) { 5168 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5169 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F3); 5170 emit_int8(0x2A); 5171 emit_int8((unsigned char)(0xC0 | encode)); 5172 } 5173 5174 void Assembler::cvtsi2ssq(XMMRegister dst, Address src) { 5175 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5176 InstructionMark im(this); 5177 simd_prefix_q(dst, dst, src, VEX_SIMD_F3); 5178 emit_int8(0x2A); 5179 emit_operand(dst, src); 5180 } 5181 5182 void Assembler::cvttsd2siq(Register dst, XMMRegister src) { 5183 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5184 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F2); 5185 emit_int8(0x2C); 5186 emit_int8((unsigned char)(0xC0 | encode)); 5187 } 5188 5189 void Assembler::cvttss2siq(Register dst, XMMRegister src) { 5190 NOT_LP64(assert(VM_Version::supports_sse(), "")); 5191 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F3); 5192 emit_int8(0x2C); 5193 emit_int8((unsigned char)(0xC0 | encode)); 5194 } 5195 5196 void Assembler::decl(Register dst) { 5197 // Don't use it directly. Use MacroAssembler::decrementl() instead. 5198 // Use two-byte form (one-byte form is a REX prefix in 64-bit mode) 5199 int encode = prefix_and_encode(dst->encoding()); 5200 emit_int8((unsigned char)0xFF); 5201 emit_int8((unsigned char)(0xC8 | encode)); 5202 } 5203 5204 void Assembler::decq(Register dst) { 5205 // Don't use it directly. Use MacroAssembler::decrementq() instead. 5206 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5207 int encode = prefixq_and_encode(dst->encoding()); 5208 emit_int8((unsigned char)0xFF); 5209 emit_int8(0xC8 | encode); 5210 } 5211 5212 void Assembler::decq(Address dst) { 5213 // Don't use it directly. Use MacroAssembler::decrementq() instead. 5214 InstructionMark im(this); 5215 prefixq(dst); 5216 emit_int8((unsigned char)0xFF); 5217 emit_operand(rcx, dst); 5218 } 5219 5220 void Assembler::fxrstor(Address src) { 5221 prefixq(src); 5222 emit_int8(0x0F); 5223 emit_int8((unsigned char)0xAE); 5224 emit_operand(as_Register(1), src); 5225 } 5226 5227 void Assembler::fxsave(Address dst) { 5228 prefixq(dst); 5229 emit_int8(0x0F); 5230 emit_int8((unsigned char)0xAE); 5231 emit_operand(as_Register(0), dst); 5232 } 5233 5234 void Assembler::idivq(Register src) { 5235 int encode = prefixq_and_encode(src->encoding()); 5236 emit_int8((unsigned char)0xF7); 5237 emit_int8((unsigned char)(0xF8 | encode)); 5238 } 5239 5240 void Assembler::imulq(Register dst, Register src) { 5241 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5242 emit_int8(0x0F); 5243 emit_int8((unsigned char)0xAF); 5244 emit_int8((unsigned char)(0xC0 | encode)); 5245 } 5246 5247 void Assembler::imulq(Register dst, Register src, int value) { 5248 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5249 if (is8bit(value)) { 5250 emit_int8(0x6B); 5251 emit_int8((unsigned char)(0xC0 | encode)); 5252 emit_int8(value & 0xFF); 5253 } else { 5254 emit_int8(0x69); 5255 emit_int8((unsigned char)(0xC0 | encode)); 5256 emit_int32(value); 5257 } 5258 } 5259 5260 void Assembler::imulq(Register dst, Address src) { 5261 InstructionMark im(this); 5262 prefixq(src, dst); 5263 emit_int8(0x0F); 5264 emit_int8((unsigned char) 0xAF); 5265 emit_operand(dst, src); 5266 } 5267 5268 void Assembler::incl(Register dst) { 5269 // Don't use it directly. Use MacroAssembler::incrementl() instead. 5270 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5271 int encode = prefix_and_encode(dst->encoding()); 5272 emit_int8((unsigned char)0xFF); 5273 emit_int8((unsigned char)(0xC0 | encode)); 5274 } 5275 5276 void Assembler::incq(Register dst) { 5277 // Don't use it directly. Use MacroAssembler::incrementq() instead. 5278 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode) 5279 int encode = prefixq_and_encode(dst->encoding()); 5280 emit_int8((unsigned char)0xFF); 5281 emit_int8((unsigned char)(0xC0 | encode)); 5282 } 5283 5284 void Assembler::incq(Address dst) { 5285 // Don't use it directly. Use MacroAssembler::incrementq() instead. 5286 InstructionMark im(this); 5287 prefixq(dst); 5288 emit_int8((unsigned char)0xFF); 5289 emit_operand(rax, dst); 5290 } 5291 5292 void Assembler::lea(Register dst, Address src) { 5293 leaq(dst, src); 5294 } 5295 5296 void Assembler::leaq(Register dst, Address src) { 5297 InstructionMark im(this); 5298 prefixq(src, dst); 5299 emit_int8((unsigned char)0x8D); 5300 emit_operand(dst, src); 5301 } 5302 5303 void Assembler::mov64(Register dst, int64_t imm64) { 5304 InstructionMark im(this); 5305 int encode = prefixq_and_encode(dst->encoding()); 5306 emit_int8((unsigned char)(0xB8 | encode)); 5307 emit_int64(imm64); 5308 } 5309 5310 void Assembler::mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec) { 5311 InstructionMark im(this); 5312 int encode = prefixq_and_encode(dst->encoding()); 5313 emit_int8(0xB8 | encode); 5314 emit_data64(imm64, rspec); 5315 } 5316 5317 void Assembler::mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec) { 5318 InstructionMark im(this); 5319 int encode = prefix_and_encode(dst->encoding()); 5320 emit_int8((unsigned char)(0xB8 | encode)); 5321 emit_data((int)imm32, rspec, narrow_oop_operand); 5322 } 5323 5324 void Assembler::mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec) { 5325 InstructionMark im(this); 5326 prefix(dst); 5327 emit_int8((unsigned char)0xC7); 5328 emit_operand(rax, dst, 4); 5329 emit_data((int)imm32, rspec, narrow_oop_operand); 5330 } 5331 5332 void Assembler::cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec) { 5333 InstructionMark im(this); 5334 int encode = prefix_and_encode(src1->encoding()); 5335 emit_int8((unsigned char)0x81); 5336 emit_int8((unsigned char)(0xF8 | encode)); 5337 emit_data((int)imm32, rspec, narrow_oop_operand); 5338 } 5339 5340 void Assembler::cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec) { 5341 InstructionMark im(this); 5342 prefix(src1); 5343 emit_int8((unsigned char)0x81); 5344 emit_operand(rax, src1, 4); 5345 emit_data((int)imm32, rspec, narrow_oop_operand); 5346 } 5347 5348 void Assembler::lzcntq(Register dst, Register src) { 5349 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR"); 5350 emit_int8((unsigned char)0xF3); 5351 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5352 emit_int8(0x0F); 5353 emit_int8((unsigned char)0xBD); 5354 emit_int8((unsigned char)(0xC0 | encode)); 5355 } 5356 5357 void Assembler::movdq(XMMRegister dst, Register src) { 5358 // table D-1 says MMX/SSE2 5359 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5360 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_66); 5361 emit_int8(0x6E); 5362 emit_int8((unsigned char)(0xC0 | encode)); 5363 } 5364 5365 void Assembler::movdq(Register dst, XMMRegister src) { 5366 // table D-1 says MMX/SSE2 5367 NOT_LP64(assert(VM_Version::supports_sse2(), "")); 5368 // swap src/dst to get correct prefix 5369 int encode = simd_prefix_and_encode_q(src, dst, VEX_SIMD_66); 5370 emit_int8(0x7E); 5371 emit_int8((unsigned char)(0xC0 | encode)); 5372 } 5373 5374 void Assembler::movq(Register dst, Register src) { 5375 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5376 emit_int8((unsigned char)0x8B); 5377 emit_int8((unsigned char)(0xC0 | encode)); 5378 } 5379 5380 void Assembler::movq(Register dst, Address src) { 5381 InstructionMark im(this); 5382 prefixq(src, dst); 5383 emit_int8((unsigned char)0x8B); 5384 emit_operand(dst, src); 5385 } 5386 5387 void Assembler::movq(Address dst, Register src) { 5388 InstructionMark im(this); 5389 prefixq(dst, src); 5390 emit_int8((unsigned char)0x89); 5391 emit_operand(src, dst); 5392 } 5393 5394 void Assembler::movsbq(Register dst, Address src) { 5395 InstructionMark im(this); 5396 prefixq(src, dst); 5397 emit_int8(0x0F); 5398 emit_int8((unsigned char)0xBE); 5399 emit_operand(dst, src); 5400 } 5401 5402 void Assembler::movsbq(Register dst, Register src) { 5403 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5404 emit_int8(0x0F); 5405 emit_int8((unsigned char)0xBE); 5406 emit_int8((unsigned char)(0xC0 | encode)); 5407 } 5408 5409 void Assembler::movslq(Register dst, int32_t imm32) { 5410 // dbx shows movslq(rcx, 3) as movq $0x0000000049000000,(%rbx) 5411 // and movslq(r8, 3); as movl $0x0000000048000000,(%rbx) 5412 // as a result we shouldn't use until tested at runtime... 5413 ShouldNotReachHere(); 5414 InstructionMark im(this); 5415 int encode = prefixq_and_encode(dst->encoding()); 5416 emit_int8((unsigned char)(0xC7 | encode)); 5417 emit_int32(imm32); 5418 } 5419 5420 void Assembler::movslq(Address dst, int32_t imm32) { 5421 assert(is_simm32(imm32), "lost bits"); 5422 InstructionMark im(this); 5423 prefixq(dst); 5424 emit_int8((unsigned char)0xC7); 5425 emit_operand(rax, dst, 4); 5426 emit_int32(imm32); 5427 } 5428 5429 void Assembler::movslq(Register dst, Address src) { 5430 InstructionMark im(this); 5431 prefixq(src, dst); 5432 emit_int8(0x63); 5433 emit_operand(dst, src); 5434 } 5435 5436 void Assembler::movslq(Register dst, Register src) { 5437 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5438 emit_int8(0x63); 5439 emit_int8((unsigned char)(0xC0 | encode)); 5440 } 5441 5442 void Assembler::movswq(Register dst, Address src) { 5443 InstructionMark im(this); 5444 prefixq(src, dst); 5445 emit_int8(0x0F); 5446 emit_int8((unsigned char)0xBF); 5447 emit_operand(dst, src); 5448 } 5449 5450 void Assembler::movswq(Register dst, Register src) { 5451 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5452 emit_int8((unsigned char)0x0F); 5453 emit_int8((unsigned char)0xBF); 5454 emit_int8((unsigned char)(0xC0 | encode)); 5455 } 5456 5457 void Assembler::movzbq(Register dst, Address src) { 5458 InstructionMark im(this); 5459 prefixq(src, dst); 5460 emit_int8((unsigned char)0x0F); 5461 emit_int8((unsigned char)0xB6); 5462 emit_operand(dst, src); 5463 } 5464 5465 void Assembler::movzbq(Register dst, Register src) { 5466 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5467 emit_int8(0x0F); 5468 emit_int8((unsigned char)0xB6); 5469 emit_int8(0xC0 | encode); 5470 } 5471 5472 void Assembler::movzwq(Register dst, Address src) { 5473 InstructionMark im(this); 5474 prefixq(src, dst); 5475 emit_int8((unsigned char)0x0F); 5476 emit_int8((unsigned char)0xB7); 5477 emit_operand(dst, src); 5478 } 5479 5480 void Assembler::movzwq(Register dst, Register src) { 5481 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5482 emit_int8((unsigned char)0x0F); 5483 emit_int8((unsigned char)0xB7); 5484 emit_int8((unsigned char)(0xC0 | encode)); 5485 } 5486 5487 void Assembler::mulq(Address src) { 5488 InstructionMark im(this); 5489 prefixq(src); 5490 emit_int8((unsigned char)0xF7); 5491 emit_operand(rsp, src); 5492 } 5493 5494 void Assembler::mulq(Register src) { 5495 int encode = prefixq_and_encode(src->encoding()); 5496 emit_int8((unsigned char)0xF7); 5497 emit_int8((unsigned char)(0xE0 | encode)); 5498 } 5499 5500 void Assembler::mulxq(Register dst1, Register dst2, Register src) { 5501 assert(VM_Version::supports_bmi2(), "bit manipulation instructions not supported"); 5502 int encode = vex_prefix_and_encode(dst1->encoding(), dst2->encoding(), src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F_38, true, false); 5503 emit_int8((unsigned char)0xF6); 5504 emit_int8((unsigned char)(0xC0 | encode)); 5505 } 5506 5507 void Assembler::negq(Register dst) { 5508 int encode = prefixq_and_encode(dst->encoding()); 5509 emit_int8((unsigned char)0xF7); 5510 emit_int8((unsigned char)(0xD8 | encode)); 5511 } 5512 5513 void Assembler::notq(Register dst) { 5514 int encode = prefixq_and_encode(dst->encoding()); 5515 emit_int8((unsigned char)0xF7); 5516 emit_int8((unsigned char)(0xD0 | encode)); 5517 } 5518 5519 void Assembler::orq(Address dst, int32_t imm32) { 5520 InstructionMark im(this); 5521 prefixq(dst); 5522 emit_int8((unsigned char)0x81); 5523 emit_operand(rcx, dst, 4); 5524 emit_int32(imm32); 5525 } 5526 5527 void Assembler::orq(Register dst, int32_t imm32) { 5528 (void) prefixq_and_encode(dst->encoding()); 5529 emit_arith(0x81, 0xC8, dst, imm32); 5530 } 5531 5532 void Assembler::orq(Register dst, Address src) { 5533 InstructionMark im(this); 5534 prefixq(src, dst); 5535 emit_int8(0x0B); 5536 emit_operand(dst, src); 5537 } 5538 5539 void Assembler::orq(Register dst, Register src) { 5540 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5541 emit_arith(0x0B, 0xC0, dst, src); 5542 } 5543 5544 void Assembler::popa() { // 64bit 5545 movq(r15, Address(rsp, 0)); 5546 movq(r14, Address(rsp, wordSize)); 5547 movq(r13, Address(rsp, 2 * wordSize)); 5548 movq(r12, Address(rsp, 3 * wordSize)); 5549 movq(r11, Address(rsp, 4 * wordSize)); 5550 movq(r10, Address(rsp, 5 * wordSize)); 5551 movq(r9, Address(rsp, 6 * wordSize)); 5552 movq(r8, Address(rsp, 7 * wordSize)); 5553 movq(rdi, Address(rsp, 8 * wordSize)); 5554 movq(rsi, Address(rsp, 9 * wordSize)); 5555 movq(rbp, Address(rsp, 10 * wordSize)); 5556 // skip rsp 5557 movq(rbx, Address(rsp, 12 * wordSize)); 5558 movq(rdx, Address(rsp, 13 * wordSize)); 5559 movq(rcx, Address(rsp, 14 * wordSize)); 5560 movq(rax, Address(rsp, 15 * wordSize)); 5561 5562 addq(rsp, 16 * wordSize); 5563 } 5564 5565 void Assembler::popcntq(Register dst, Address src) { 5566 assert(VM_Version::supports_popcnt(), "must support"); 5567 InstructionMark im(this); 5568 emit_int8((unsigned char)0xF3); 5569 prefixq(src, dst); 5570 emit_int8((unsigned char)0x0F); 5571 emit_int8((unsigned char)0xB8); 5572 emit_operand(dst, src); 5573 } 5574 5575 void Assembler::popcntq(Register dst, Register src) { 5576 assert(VM_Version::supports_popcnt(), "must support"); 5577 emit_int8((unsigned char)0xF3); 5578 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5579 emit_int8((unsigned char)0x0F); 5580 emit_int8((unsigned char)0xB8); 5581 emit_int8((unsigned char)(0xC0 | encode)); 5582 } 5583 5584 void Assembler::popq(Address dst) { 5585 InstructionMark im(this); 5586 prefixq(dst); 5587 emit_int8((unsigned char)0x8F); 5588 emit_operand(rax, dst); 5589 } 5590 5591 void Assembler::pusha() { // 64bit 5592 // we have to store original rsp. ABI says that 128 bytes 5593 // below rsp are local scratch. 5594 movq(Address(rsp, -5 * wordSize), rsp); 5595 5596 subq(rsp, 16 * wordSize); 5597 5598 movq(Address(rsp, 15 * wordSize), rax); 5599 movq(Address(rsp, 14 * wordSize), rcx); 5600 movq(Address(rsp, 13 * wordSize), rdx); 5601 movq(Address(rsp, 12 * wordSize), rbx); 5602 // skip rsp 5603 movq(Address(rsp, 10 * wordSize), rbp); 5604 movq(Address(rsp, 9 * wordSize), rsi); 5605 movq(Address(rsp, 8 * wordSize), rdi); 5606 movq(Address(rsp, 7 * wordSize), r8); 5607 movq(Address(rsp, 6 * wordSize), r9); 5608 movq(Address(rsp, 5 * wordSize), r10); 5609 movq(Address(rsp, 4 * wordSize), r11); 5610 movq(Address(rsp, 3 * wordSize), r12); 5611 movq(Address(rsp, 2 * wordSize), r13); 5612 movq(Address(rsp, wordSize), r14); 5613 movq(Address(rsp, 0), r15); 5614 } 5615 5616 void Assembler::pushq(Address src) { 5617 InstructionMark im(this); 5618 prefixq(src); 5619 emit_int8((unsigned char)0xFF); 5620 emit_operand(rsi, src); 5621 } 5622 5623 void Assembler::rclq(Register dst, int imm8) { 5624 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5625 int encode = prefixq_and_encode(dst->encoding()); 5626 if (imm8 == 1) { 5627 emit_int8((unsigned char)0xD1); 5628 emit_int8((unsigned char)(0xD0 | encode)); 5629 } else { 5630 emit_int8((unsigned char)0xC1); 5631 emit_int8((unsigned char)(0xD0 | encode)); 5632 emit_int8(imm8); 5633 } 5634 } 5635 5636 void Assembler::rorq(Register dst, int imm8) { 5637 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5638 int encode = prefixq_and_encode(dst->encoding()); 5639 if (imm8 == 1) { 5640 emit_int8((unsigned char)0xD1); 5641 emit_int8((unsigned char)(0xC8 | encode)); 5642 } else { 5643 emit_int8((unsigned char)0xC1); 5644 emit_int8((unsigned char)(0xc8 | encode)); 5645 emit_int8(imm8); 5646 } 5647 } 5648 5649 void Assembler::rorxq(Register dst, Register src, int imm8) { 5650 assert(VM_Version::supports_bmi2(), "bit manipulation instructions not supported"); 5651 int encode = vex_prefix_and_encode(dst->encoding(), 0, src->encoding(), VEX_SIMD_F2, VEX_OPCODE_0F_3A, true, false); 5652 emit_int8((unsigned char)0xF0); 5653 emit_int8((unsigned char)(0xC0 | encode)); 5654 emit_int8(imm8); 5655 } 5656 5657 void Assembler::sarq(Register dst, int imm8) { 5658 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5659 int encode = prefixq_and_encode(dst->encoding()); 5660 if (imm8 == 1) { 5661 emit_int8((unsigned char)0xD1); 5662 emit_int8((unsigned char)(0xF8 | encode)); 5663 } else { 5664 emit_int8((unsigned char)0xC1); 5665 emit_int8((unsigned char)(0xF8 | encode)); 5666 emit_int8(imm8); 5667 } 5668 } 5669 5670 void Assembler::sarq(Register dst) { 5671 int encode = prefixq_and_encode(dst->encoding()); 5672 emit_int8((unsigned char)0xD3); 5673 emit_int8((unsigned char)(0xF8 | encode)); 5674 } 5675 5676 void Assembler::sbbq(Address dst, int32_t imm32) { 5677 InstructionMark im(this); 5678 prefixq(dst); 5679 emit_arith_operand(0x81, rbx, dst, imm32); 5680 } 5681 5682 void Assembler::sbbq(Register dst, int32_t imm32) { 5683 (void) prefixq_and_encode(dst->encoding()); 5684 emit_arith(0x81, 0xD8, dst, imm32); 5685 } 5686 5687 void Assembler::sbbq(Register dst, Address src) { 5688 InstructionMark im(this); 5689 prefixq(src, dst); 5690 emit_int8(0x1B); 5691 emit_operand(dst, src); 5692 } 5693 5694 void Assembler::sbbq(Register dst, Register src) { 5695 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5696 emit_arith(0x1B, 0xC0, dst, src); 5697 } 5698 5699 void Assembler::shlq(Register dst, int imm8) { 5700 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5701 int encode = prefixq_and_encode(dst->encoding()); 5702 if (imm8 == 1) { 5703 emit_int8((unsigned char)0xD1); 5704 emit_int8((unsigned char)(0xE0 | encode)); 5705 } else { 5706 emit_int8((unsigned char)0xC1); 5707 emit_int8((unsigned char)(0xE0 | encode)); 5708 emit_int8(imm8); 5709 } 5710 } 5711 5712 void Assembler::shlq(Register dst) { 5713 int encode = prefixq_and_encode(dst->encoding()); 5714 emit_int8((unsigned char)0xD3); 5715 emit_int8((unsigned char)(0xE0 | encode)); 5716 } 5717 5718 void Assembler::shrq(Register dst, int imm8) { 5719 assert(isShiftCount(imm8 >> 1), "illegal shift count"); 5720 int encode = prefixq_and_encode(dst->encoding()); 5721 emit_int8((unsigned char)0xC1); 5722 emit_int8((unsigned char)(0xE8 | encode)); 5723 emit_int8(imm8); 5724 } 5725 5726 void Assembler::shrq(Register dst) { 5727 int encode = prefixq_and_encode(dst->encoding()); 5728 emit_int8((unsigned char)0xD3); 5729 emit_int8(0xE8 | encode); 5730 } 5731 5732 void Assembler::subq(Address dst, int32_t imm32) { 5733 InstructionMark im(this); 5734 prefixq(dst); 5735 emit_arith_operand(0x81, rbp, dst, imm32); 5736 } 5737 5738 void Assembler::subq(Address dst, Register src) { 5739 InstructionMark im(this); 5740 prefixq(dst, src); 5741 emit_int8(0x29); 5742 emit_operand(src, dst); 5743 } 5744 5745 void Assembler::subq(Register dst, int32_t imm32) { 5746 (void) prefixq_and_encode(dst->encoding()); 5747 emit_arith(0x81, 0xE8, dst, imm32); 5748 } 5749 5750 // Force generation of a 4 byte immediate value even if it fits into 8bit 5751 void Assembler::subq_imm32(Register dst, int32_t imm32) { 5752 (void) prefixq_and_encode(dst->encoding()); 5753 emit_arith_imm32(0x81, 0xE8, dst, imm32); 5754 } 5755 5756 void Assembler::subq(Register dst, Address src) { 5757 InstructionMark im(this); 5758 prefixq(src, dst); 5759 emit_int8(0x2B); 5760 emit_operand(dst, src); 5761 } 5762 5763 void Assembler::subq(Register dst, Register src) { 5764 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5765 emit_arith(0x2B, 0xC0, dst, src); 5766 } 5767 5768 void Assembler::testq(Register dst, int32_t imm32) { 5769 // not using emit_arith because test 5770 // doesn't support sign-extension of 5771 // 8bit operands 5772 int encode = dst->encoding(); 5773 if (encode == 0) { 5774 prefix(REX_W); 5775 emit_int8((unsigned char)0xA9); 5776 } else { 5777 encode = prefixq_and_encode(encode); 5778 emit_int8((unsigned char)0xF7); 5779 emit_int8((unsigned char)(0xC0 | encode)); 5780 } 5781 emit_int32(imm32); 5782 } 5783 5784 void Assembler::testq(Register dst, Register src) { 5785 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5786 emit_arith(0x85, 0xC0, dst, src); 5787 } 5788 5789 void Assembler::xaddq(Address dst, Register src) { 5790 InstructionMark im(this); 5791 prefixq(dst, src); 5792 emit_int8(0x0F); 5793 emit_int8((unsigned char)0xC1); 5794 emit_operand(src, dst); 5795 } 5796 5797 void Assembler::xchgq(Register dst, Address src) { 5798 InstructionMark im(this); 5799 prefixq(src, dst); 5800 emit_int8((unsigned char)0x87); 5801 emit_operand(dst, src); 5802 } 5803 5804 void Assembler::xchgq(Register dst, Register src) { 5805 int encode = prefixq_and_encode(dst->encoding(), src->encoding()); 5806 emit_int8((unsigned char)0x87); 5807 emit_int8((unsigned char)(0xc0 | encode)); 5808 } 5809 5810 void Assembler::xorq(Register dst, Register src) { 5811 (void) prefixq_and_encode(dst->encoding(), src->encoding()); 5812 emit_arith(0x33, 0xC0, dst, src); 5813 } 5814 5815 void Assembler::xorq(Register dst, Address src) { 5816 InstructionMark im(this); 5817 prefixq(src, dst); 5818 emit_int8(0x33); 5819 emit_operand(dst, src); 5820 } 5821 5822 #endif // !LP64