Thu, 03 Jan 2013 16:30:47 -0800
8005544: Use 256bit YMM registers in arraycopy stubs on x86
Summary: Use YMM registers in arraycopy and array_fill stubs.
Reviewed-by: roland, twisti
1 /*
2 * Copyright (c) 1997, 2012, 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 */
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 #ifndef SERIALGC
40 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
41 #include "gc_implementation/g1/g1SATBCardTableModRefBS.hpp"
42 #include "gc_implementation/g1/heapRegion.hpp"
43 #endif
45 #ifdef PRODUCT
46 #define BLOCK_COMMENT(str) /* nothing */
47 #define STOP(error) stop(error)
48 #else
49 #define BLOCK_COMMENT(str) block_comment(str)
50 #define STOP(error) block_comment(error); stop(error)
51 #endif
53 #define BIND(label) bind(label); BLOCK_COMMENT(#label ":")
54 // Implementation of AddressLiteral
56 AddressLiteral::AddressLiteral(address target, relocInfo::relocType rtype) {
57 _is_lval = false;
58 _target = target;
59 switch (rtype) {
60 case relocInfo::oop_type:
61 case relocInfo::metadata_type:
62 // Oops are a special case. Normally they would be their own section
63 // but in cases like icBuffer they are literals in the code stream that
64 // we don't have a section for. We use none so that we get a literal address
65 // which is always patchable.
66 break;
67 case relocInfo::external_word_type:
68 _rspec = external_word_Relocation::spec(target);
69 break;
70 case relocInfo::internal_word_type:
71 _rspec = internal_word_Relocation::spec(target);
72 break;
73 case relocInfo::opt_virtual_call_type:
74 _rspec = opt_virtual_call_Relocation::spec();
75 break;
76 case relocInfo::static_call_type:
77 _rspec = static_call_Relocation::spec();
78 break;
79 case relocInfo::runtime_call_type:
80 _rspec = runtime_call_Relocation::spec();
81 break;
82 case relocInfo::poll_type:
83 case relocInfo::poll_return_type:
84 _rspec = Relocation::spec_simple(rtype);
85 break;
86 case relocInfo::none:
87 break;
88 default:
89 ShouldNotReachHere();
90 break;
91 }
92 }
94 // Implementation of Address
96 #ifdef _LP64
98 Address Address::make_array(ArrayAddress adr) {
99 // Not implementable on 64bit machines
100 // Should have been handled higher up the call chain.
101 ShouldNotReachHere();
102 return Address();
103 }
105 // exceedingly dangerous constructor
106 Address::Address(int disp, address loc, relocInfo::relocType rtype) {
107 _base = noreg;
108 _index = noreg;
109 _scale = no_scale;
110 _disp = disp;
111 switch (rtype) {
112 case relocInfo::external_word_type:
113 _rspec = external_word_Relocation::spec(loc);
114 break;
115 case relocInfo::internal_word_type:
116 _rspec = internal_word_Relocation::spec(loc);
117 break;
118 case relocInfo::runtime_call_type:
119 // HMM
120 _rspec = runtime_call_Relocation::spec();
121 break;
122 case relocInfo::poll_type:
123 case relocInfo::poll_return_type:
124 _rspec = Relocation::spec_simple(rtype);
125 break;
126 case relocInfo::none:
127 break;
128 default:
129 ShouldNotReachHere();
130 }
131 }
132 #else // LP64
134 Address Address::make_array(ArrayAddress adr) {
135 AddressLiteral base = adr.base();
136 Address index = adr.index();
137 assert(index._disp == 0, "must not have disp"); // maybe it can?
138 Address array(index._base, index._index, index._scale, (intptr_t) base.target());
139 array._rspec = base._rspec;
140 return array;
141 }
143 // exceedingly dangerous constructor
144 Address::Address(address loc, RelocationHolder spec) {
145 _base = noreg;
146 _index = noreg;
147 _scale = no_scale;
148 _disp = (intptr_t) loc;
149 _rspec = spec;
150 }
152 #endif // _LP64
156 // Convert the raw encoding form into the form expected by the constructor for
157 // Address. An index of 4 (rsp) corresponds to having no index, so convert
158 // that to noreg for the Address constructor.
159 Address Address::make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc) {
160 RelocationHolder rspec;
161 if (disp_reloc != relocInfo::none) {
162 rspec = Relocation::spec_simple(disp_reloc);
163 }
164 bool valid_index = index != rsp->encoding();
165 if (valid_index) {
166 Address madr(as_Register(base), as_Register(index), (Address::ScaleFactor)scale, in_ByteSize(disp));
167 madr._rspec = rspec;
168 return madr;
169 } else {
170 Address madr(as_Register(base), noreg, Address::no_scale, in_ByteSize(disp));
171 madr._rspec = rspec;
172 return madr;
173 }
174 }
176 // Implementation of Assembler
178 int AbstractAssembler::code_fill_byte() {
179 return (u_char)'\xF4'; // hlt
180 }
182 // make this go away someday
183 void Assembler::emit_data(jint data, relocInfo::relocType rtype, int format) {
184 if (rtype == relocInfo::none)
185 emit_long(data);
186 else emit_data(data, Relocation::spec_simple(rtype), format);
187 }
189 void Assembler::emit_data(jint data, RelocationHolder const& rspec, int format) {
190 assert(imm_operand == 0, "default format must be immediate in this file");
191 assert(inst_mark() != NULL, "must be inside InstructionMark");
192 if (rspec.type() != relocInfo::none) {
193 #ifdef ASSERT
194 check_relocation(rspec, format);
195 #endif
196 // Do not use AbstractAssembler::relocate, which is not intended for
197 // embedded words. Instead, relocate to the enclosing instruction.
199 // hack. call32 is too wide for mask so use disp32
200 if (format == call32_operand)
201 code_section()->relocate(inst_mark(), rspec, disp32_operand);
202 else
203 code_section()->relocate(inst_mark(), rspec, format);
204 }
205 emit_long(data);
206 }
208 static int encode(Register r) {
209 int enc = r->encoding();
210 if (enc >= 8) {
211 enc -= 8;
212 }
213 return enc;
214 }
216 static int encode(XMMRegister r) {
217 int enc = r->encoding();
218 if (enc >= 8) {
219 enc -= 8;
220 }
221 return enc;
222 }
224 void Assembler::emit_arith_b(int op1, int op2, Register dst, int imm8) {
225 assert(dst->has_byte_register(), "must have byte register");
226 assert(isByte(op1) && isByte(op2), "wrong opcode");
227 assert(isByte(imm8), "not a byte");
228 assert((op1 & 0x01) == 0, "should be 8bit operation");
229 emit_int8(op1);
230 emit_int8(op2 | encode(dst));
231 emit_int8(imm8);
232 }
235 void Assembler::emit_arith(int op1, int op2, Register dst, int32_t imm32) {
236 assert(isByte(op1) && isByte(op2), "wrong opcode");
237 assert((op1 & 0x01) == 1, "should be 32bit operation");
238 assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
239 if (is8bit(imm32)) {
240 emit_int8(op1 | 0x02); // set sign bit
241 emit_int8(op2 | encode(dst));
242 emit_int8(imm32 & 0xFF);
243 } else {
244 emit_int8(op1);
245 emit_int8(op2 | encode(dst));
246 emit_long(imm32);
247 }
248 }
250 // Force generation of a 4 byte immediate value even if it fits into 8bit
251 void Assembler::emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32) {
252 assert(isByte(op1) && isByte(op2), "wrong opcode");
253 assert((op1 & 0x01) == 1, "should be 32bit operation");
254 assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
255 emit_int8(op1);
256 emit_int8(op2 | encode(dst));
257 emit_long(imm32);
258 }
260 // immediate-to-memory forms
261 void Assembler::emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32) {
262 assert((op1 & 0x01) == 1, "should be 32bit operation");
263 assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
264 if (is8bit(imm32)) {
265 emit_int8(op1 | 0x02); // set sign bit
266 emit_operand(rm, adr, 1);
267 emit_int8(imm32 & 0xFF);
268 } else {
269 emit_int8(op1);
270 emit_operand(rm, adr, 4);
271 emit_long(imm32);
272 }
273 }
276 void Assembler::emit_arith(int op1, int op2, Register dst, Register src) {
277 assert(isByte(op1) && isByte(op2), "wrong opcode");
278 emit_int8(op1);
279 emit_int8(op2 | encode(dst) << 3 | encode(src));
280 }
283 void Assembler::emit_operand(Register reg, Register base, Register index,
284 Address::ScaleFactor scale, int disp,
285 RelocationHolder const& rspec,
286 int rip_relative_correction) {
287 relocInfo::relocType rtype = (relocInfo::relocType) rspec.type();
289 // Encode the registers as needed in the fields they are used in
291 int regenc = encode(reg) << 3;
292 int indexenc = index->is_valid() ? encode(index) << 3 : 0;
293 int baseenc = base->is_valid() ? encode(base) : 0;
295 if (base->is_valid()) {
296 if (index->is_valid()) {
297 assert(scale != Address::no_scale, "inconsistent address");
298 // [base + index*scale + disp]
299 if (disp == 0 && rtype == relocInfo::none &&
300 base != rbp LP64_ONLY(&& base != r13)) {
301 // [base + index*scale]
302 // [00 reg 100][ss index base]
303 assert(index != rsp, "illegal addressing mode");
304 emit_int8(0x04 | regenc);
305 emit_int8(scale << 6 | indexenc | baseenc);
306 } else if (is8bit(disp) && rtype == relocInfo::none) {
307 // [base + index*scale + imm8]
308 // [01 reg 100][ss index base] imm8
309 assert(index != rsp, "illegal addressing mode");
310 emit_int8(0x44 | regenc);
311 emit_int8(scale << 6 | indexenc | baseenc);
312 emit_int8(disp & 0xFF);
313 } else {
314 // [base + index*scale + disp32]
315 // [10 reg 100][ss index base] disp32
316 assert(index != rsp, "illegal addressing mode");
317 emit_int8(0x84 | regenc);
318 emit_int8(scale << 6 | indexenc | baseenc);
319 emit_data(disp, rspec, disp32_operand);
320 }
321 } else if (base == rsp LP64_ONLY(|| base == r12)) {
322 // [rsp + disp]
323 if (disp == 0 && rtype == relocInfo::none) {
324 // [rsp]
325 // [00 reg 100][00 100 100]
326 emit_int8(0x04 | regenc);
327 emit_int8(0x24);
328 } else if (is8bit(disp) && rtype == relocInfo::none) {
329 // [rsp + imm8]
330 // [01 reg 100][00 100 100] disp8
331 emit_int8(0x44 | regenc);
332 emit_int8(0x24);
333 emit_int8(disp & 0xFF);
334 } else {
335 // [rsp + imm32]
336 // [10 reg 100][00 100 100] disp32
337 emit_int8(0x84 | regenc);
338 emit_int8(0x24);
339 emit_data(disp, rspec, disp32_operand);
340 }
341 } else {
342 // [base + disp]
343 assert(base != rsp LP64_ONLY(&& base != r12), "illegal addressing mode");
344 if (disp == 0 && rtype == relocInfo::none &&
345 base != rbp LP64_ONLY(&& base != r13)) {
346 // [base]
347 // [00 reg base]
348 emit_int8(0x00 | regenc | baseenc);
349 } else if (is8bit(disp) && rtype == relocInfo::none) {
350 // [base + disp8]
351 // [01 reg base] disp8
352 emit_int8(0x40 | regenc | baseenc);
353 emit_int8(disp & 0xFF);
354 } else {
355 // [base + disp32]
356 // [10 reg base] disp32
357 emit_int8(0x80 | regenc | baseenc);
358 emit_data(disp, rspec, disp32_operand);
359 }
360 }
361 } else {
362 if (index->is_valid()) {
363 assert(scale != Address::no_scale, "inconsistent address");
364 // [index*scale + disp]
365 // [00 reg 100][ss index 101] disp32
366 assert(index != rsp, "illegal addressing mode");
367 emit_int8(0x04 | regenc);
368 emit_int8(scale << 6 | indexenc | 0x05);
369 emit_data(disp, rspec, disp32_operand);
370 } else if (rtype != relocInfo::none ) {
371 // [disp] (64bit) RIP-RELATIVE (32bit) abs
372 // [00 000 101] disp32
374 emit_int8(0x05 | regenc);
375 // Note that the RIP-rel. correction applies to the generated
376 // disp field, but _not_ to the target address in the rspec.
378 // disp was created by converting the target address minus the pc
379 // at the start of the instruction. That needs more correction here.
380 // intptr_t disp = target - next_ip;
381 assert(inst_mark() != NULL, "must be inside InstructionMark");
382 address next_ip = pc() + sizeof(int32_t) + rip_relative_correction;
383 int64_t adjusted = disp;
384 // Do rip-rel adjustment for 64bit
385 LP64_ONLY(adjusted -= (next_ip - inst_mark()));
386 assert(is_simm32(adjusted),
387 "must be 32bit offset (RIP relative address)");
388 emit_data((int32_t) adjusted, rspec, disp32_operand);
390 } else {
391 // 32bit never did this, did everything as the rip-rel/disp code above
392 // [disp] ABSOLUTE
393 // [00 reg 100][00 100 101] disp32
394 emit_int8(0x04 | regenc);
395 emit_int8(0x25);
396 emit_data(disp, rspec, disp32_operand);
397 }
398 }
399 }
401 void Assembler::emit_operand(XMMRegister reg, Register base, Register index,
402 Address::ScaleFactor scale, int disp,
403 RelocationHolder const& rspec) {
404 emit_operand((Register)reg, base, index, scale, disp, rspec);
405 }
407 // Secret local extension to Assembler::WhichOperand:
408 #define end_pc_operand (_WhichOperand_limit)
410 address Assembler::locate_operand(address inst, WhichOperand which) {
411 // Decode the given instruction, and return the address of
412 // an embedded 32-bit operand word.
414 // If "which" is disp32_operand, selects the displacement portion
415 // of an effective address specifier.
416 // If "which" is imm64_operand, selects the trailing immediate constant.
417 // If "which" is call32_operand, selects the displacement of a call or jump.
418 // Caller is responsible for ensuring that there is such an operand,
419 // and that it is 32/64 bits wide.
421 // If "which" is end_pc_operand, find the end of the instruction.
423 address ip = inst;
424 bool is_64bit = false;
426 debug_only(bool has_disp32 = false);
427 int tail_size = 0; // other random bytes (#32, #16, etc.) at end of insn
429 again_after_prefix:
430 switch (0xFF & *ip++) {
432 // These convenience macros generate groups of "case" labels for the switch.
433 #define REP4(x) (x)+0: case (x)+1: case (x)+2: case (x)+3
434 #define REP8(x) (x)+0: case (x)+1: case (x)+2: case (x)+3: \
435 case (x)+4: case (x)+5: case (x)+6: case (x)+7
436 #define REP16(x) REP8((x)+0): \
437 case REP8((x)+8)
439 case CS_segment:
440 case SS_segment:
441 case DS_segment:
442 case ES_segment:
443 case FS_segment:
444 case GS_segment:
445 // Seems dubious
446 LP64_ONLY(assert(false, "shouldn't have that prefix"));
447 assert(ip == inst+1, "only one prefix allowed");
448 goto again_after_prefix;
450 case 0x67:
451 case REX:
452 case REX_B:
453 case REX_X:
454 case REX_XB:
455 case REX_R:
456 case REX_RB:
457 case REX_RX:
458 case REX_RXB:
459 NOT_LP64(assert(false, "64bit prefixes"));
460 goto again_after_prefix;
462 case REX_W:
463 case REX_WB:
464 case REX_WX:
465 case REX_WXB:
466 case REX_WR:
467 case REX_WRB:
468 case REX_WRX:
469 case REX_WRXB:
470 NOT_LP64(assert(false, "64bit prefixes"));
471 is_64bit = true;
472 goto again_after_prefix;
474 case 0xFF: // pushq a; decl a; incl a; call a; jmp a
475 case 0x88: // movb a, r
476 case 0x89: // movl a, r
477 case 0x8A: // movb r, a
478 case 0x8B: // movl r, a
479 case 0x8F: // popl a
480 debug_only(has_disp32 = true);
481 break;
483 case 0x68: // pushq #32
484 if (which == end_pc_operand) {
485 return ip + 4;
486 }
487 assert(which == imm_operand && !is_64bit, "pushl has no disp32 or 64bit immediate");
488 return ip; // not produced by emit_operand
490 case 0x66: // movw ... (size prefix)
491 again_after_size_prefix2:
492 switch (0xFF & *ip++) {
493 case REX:
494 case REX_B:
495 case REX_X:
496 case REX_XB:
497 case REX_R:
498 case REX_RB:
499 case REX_RX:
500 case REX_RXB:
501 case REX_W:
502 case REX_WB:
503 case REX_WX:
504 case REX_WXB:
505 case REX_WR:
506 case REX_WRB:
507 case REX_WRX:
508 case REX_WRXB:
509 NOT_LP64(assert(false, "64bit prefix found"));
510 goto again_after_size_prefix2;
511 case 0x8B: // movw r, a
512 case 0x89: // movw a, r
513 debug_only(has_disp32 = true);
514 break;
515 case 0xC7: // movw a, #16
516 debug_only(has_disp32 = true);
517 tail_size = 2; // the imm16
518 break;
519 case 0x0F: // several SSE/SSE2 variants
520 ip--; // reparse the 0x0F
521 goto again_after_prefix;
522 default:
523 ShouldNotReachHere();
524 }
525 break;
527 case REP8(0xB8): // movl/q r, #32/#64(oop?)
528 if (which == end_pc_operand) return ip + (is_64bit ? 8 : 4);
529 // these asserts are somewhat nonsensical
530 #ifndef _LP64
531 assert(which == imm_operand || which == disp32_operand,
532 err_msg("which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, ip));
533 #else
534 assert((which == call32_operand || which == imm_operand) && is_64bit ||
535 which == narrow_oop_operand && !is_64bit,
536 err_msg("which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, ip));
537 #endif // _LP64
538 return ip;
540 case 0x69: // imul r, a, #32
541 case 0xC7: // movl a, #32(oop?)
542 tail_size = 4;
543 debug_only(has_disp32 = true); // has both kinds of operands!
544 break;
546 case 0x0F: // movx..., etc.
547 switch (0xFF & *ip++) {
548 case 0x3A: // pcmpestri
549 tail_size = 1;
550 case 0x38: // ptest, pmovzxbw
551 ip++; // skip opcode
552 debug_only(has_disp32 = true); // has both kinds of operands!
553 break;
555 case 0x70: // pshufd r, r/a, #8
556 debug_only(has_disp32 = true); // has both kinds of operands!
557 case 0x73: // psrldq r, #8
558 tail_size = 1;
559 break;
561 case 0x12: // movlps
562 case 0x28: // movaps
563 case 0x2E: // ucomiss
564 case 0x2F: // comiss
565 case 0x54: // andps
566 case 0x55: // andnps
567 case 0x56: // orps
568 case 0x57: // xorps
569 case 0x6E: // movd
570 case 0x7E: // movd
571 case 0xAE: // ldmxcsr, stmxcsr, fxrstor, fxsave, clflush
572 debug_only(has_disp32 = true);
573 break;
575 case 0xAD: // shrd r, a, %cl
576 case 0xAF: // imul r, a
577 case 0xBE: // movsbl r, a (movsxb)
578 case 0xBF: // movswl r, a (movsxw)
579 case 0xB6: // movzbl r, a (movzxb)
580 case 0xB7: // movzwl r, a (movzxw)
581 case REP16(0x40): // cmovl cc, r, a
582 case 0xB0: // cmpxchgb
583 case 0xB1: // cmpxchg
584 case 0xC1: // xaddl
585 case 0xC7: // cmpxchg8
586 case REP16(0x90): // setcc a
587 debug_only(has_disp32 = true);
588 // fall out of the switch to decode the address
589 break;
591 case 0xC4: // pinsrw r, a, #8
592 debug_only(has_disp32 = true);
593 case 0xC5: // pextrw r, r, #8
594 tail_size = 1; // the imm8
595 break;
597 case 0xAC: // shrd r, a, #8
598 debug_only(has_disp32 = true);
599 tail_size = 1; // the imm8
600 break;
602 case REP16(0x80): // jcc rdisp32
603 if (which == end_pc_operand) return ip + 4;
604 assert(which == call32_operand, "jcc has no disp32 or imm");
605 return ip;
606 default:
607 ShouldNotReachHere();
608 }
609 break;
611 case 0x81: // addl a, #32; addl r, #32
612 // also: orl, adcl, sbbl, andl, subl, xorl, cmpl
613 // on 32bit in the case of cmpl, the imm might be an oop
614 tail_size = 4;
615 debug_only(has_disp32 = true); // has both kinds of operands!
616 break;
618 case 0x83: // addl a, #8; addl r, #8
619 // also: orl, adcl, sbbl, andl, subl, xorl, cmpl
620 debug_only(has_disp32 = true); // has both kinds of operands!
621 tail_size = 1;
622 break;
624 case 0x9B:
625 switch (0xFF & *ip++) {
626 case 0xD9: // fnstcw a
627 debug_only(has_disp32 = true);
628 break;
629 default:
630 ShouldNotReachHere();
631 }
632 break;
634 case REP4(0x00): // addb a, r; addl a, r; addb r, a; addl r, a
635 case REP4(0x10): // adc...
636 case REP4(0x20): // and...
637 case REP4(0x30): // xor...
638 case REP4(0x08): // or...
639 case REP4(0x18): // sbb...
640 case REP4(0x28): // sub...
641 case 0xF7: // mull a
642 case 0x8D: // lea r, a
643 case 0x87: // xchg r, a
644 case REP4(0x38): // cmp...
645 case 0x85: // test r, a
646 debug_only(has_disp32 = true); // has both kinds of operands!
647 break;
649 case 0xC1: // sal a, #8; sar a, #8; shl a, #8; shr a, #8
650 case 0xC6: // movb a, #8
651 case 0x80: // cmpb a, #8
652 case 0x6B: // imul r, a, #8
653 debug_only(has_disp32 = true); // has both kinds of operands!
654 tail_size = 1; // the imm8
655 break;
657 case 0xC4: // VEX_3bytes
658 case 0xC5: // VEX_2bytes
659 assert((UseAVX > 0), "shouldn't have VEX prefix");
660 assert(ip == inst+1, "no prefixes allowed");
661 // C4 and C5 are also used as opcodes for PINSRW and PEXTRW instructions
662 // but they have prefix 0x0F and processed when 0x0F processed above.
663 //
664 // In 32-bit mode the VEX first byte C4 and C5 alias onto LDS and LES
665 // instructions (these instructions are not supported in 64-bit mode).
666 // To distinguish them bits [7:6] are set in the VEX second byte since
667 // ModRM byte can not be of the form 11xxxxxx in 32-bit mode. To set
668 // those VEX bits REX and vvvv bits are inverted.
669 //
670 // Fortunately C2 doesn't generate these instructions so we don't need
671 // to check for them in product version.
673 // Check second byte
674 NOT_LP64(assert((0xC0 & *ip) == 0xC0, "shouldn't have LDS and LES instructions"));
676 // First byte
677 if ((0xFF & *inst) == VEX_3bytes) {
678 ip++; // third byte
679 is_64bit = ((VEX_W & *ip) == VEX_W);
680 }
681 ip++; // opcode
682 // To find the end of instruction (which == end_pc_operand).
683 switch (0xFF & *ip) {
684 case 0x61: // pcmpestri r, r/a, #8
685 case 0x70: // pshufd r, r/a, #8
686 case 0x73: // psrldq r, #8
687 tail_size = 1; // the imm8
688 break;
689 default:
690 break;
691 }
692 ip++; // skip opcode
693 debug_only(has_disp32 = true); // has both kinds of operands!
694 break;
696 case 0xD1: // sal a, 1; sar a, 1; shl a, 1; shr a, 1
697 case 0xD3: // sal a, %cl; sar a, %cl; shl a, %cl; shr a, %cl
698 case 0xD9: // fld_s a; fst_s a; fstp_s a; fldcw a
699 case 0xDD: // fld_d a; fst_d a; fstp_d a
700 case 0xDB: // fild_s a; fistp_s a; fld_x a; fstp_x a
701 case 0xDF: // fild_d a; fistp_d a
702 case 0xD8: // fadd_s a; fsubr_s a; fmul_s a; fdivr_s a; fcomp_s a
703 case 0xDC: // fadd_d a; fsubr_d a; fmul_d a; fdivr_d a; fcomp_d a
704 case 0xDE: // faddp_d a; fsubrp_d a; fmulp_d a; fdivrp_d a; fcompp_d a
705 debug_only(has_disp32 = true);
706 break;
708 case 0xE8: // call rdisp32
709 case 0xE9: // jmp rdisp32
710 if (which == end_pc_operand) return ip + 4;
711 assert(which == call32_operand, "call has no disp32 or imm");
712 return ip;
714 case 0xF0: // Lock
715 assert(os::is_MP(), "only on MP");
716 goto again_after_prefix;
718 case 0xF3: // For SSE
719 case 0xF2: // For SSE2
720 switch (0xFF & *ip++) {
721 case REX:
722 case REX_B:
723 case REX_X:
724 case REX_XB:
725 case REX_R:
726 case REX_RB:
727 case REX_RX:
728 case REX_RXB:
729 case REX_W:
730 case REX_WB:
731 case REX_WX:
732 case REX_WXB:
733 case REX_WR:
734 case REX_WRB:
735 case REX_WRX:
736 case REX_WRXB:
737 NOT_LP64(assert(false, "found 64bit prefix"));
738 ip++;
739 default:
740 ip++;
741 }
742 debug_only(has_disp32 = true); // has both kinds of operands!
743 break;
745 default:
746 ShouldNotReachHere();
748 #undef REP8
749 #undef REP16
750 }
752 assert(which != call32_operand, "instruction is not a call, jmp, or jcc");
753 #ifdef _LP64
754 assert(which != imm_operand, "instruction is not a movq reg, imm64");
755 #else
756 // assert(which != imm_operand || has_imm32, "instruction has no imm32 field");
757 assert(which != imm_operand || has_disp32, "instruction has no imm32 field");
758 #endif // LP64
759 assert(which != disp32_operand || has_disp32, "instruction has no disp32 field");
761 // parse the output of emit_operand
762 int op2 = 0xFF & *ip++;
763 int base = op2 & 0x07;
764 int op3 = -1;
765 const int b100 = 4;
766 const int b101 = 5;
767 if (base == b100 && (op2 >> 6) != 3) {
768 op3 = 0xFF & *ip++;
769 base = op3 & 0x07; // refetch the base
770 }
771 // now ip points at the disp (if any)
773 switch (op2 >> 6) {
774 case 0:
775 // [00 reg 100][ss index base]
776 // [00 reg 100][00 100 esp]
777 // [00 reg base]
778 // [00 reg 100][ss index 101][disp32]
779 // [00 reg 101] [disp32]
781 if (base == b101) {
782 if (which == disp32_operand)
783 return ip; // caller wants the disp32
784 ip += 4; // skip the disp32
785 }
786 break;
788 case 1:
789 // [01 reg 100][ss index base][disp8]
790 // [01 reg 100][00 100 esp][disp8]
791 // [01 reg base] [disp8]
792 ip += 1; // skip the disp8
793 break;
795 case 2:
796 // [10 reg 100][ss index base][disp32]
797 // [10 reg 100][00 100 esp][disp32]
798 // [10 reg base] [disp32]
799 if (which == disp32_operand)
800 return ip; // caller wants the disp32
801 ip += 4; // skip the disp32
802 break;
804 case 3:
805 // [11 reg base] (not a memory addressing mode)
806 break;
807 }
809 if (which == end_pc_operand) {
810 return ip + tail_size;
811 }
813 #ifdef _LP64
814 assert(which == narrow_oop_operand && !is_64bit, "instruction is not a movl adr, imm32");
815 #else
816 assert(which == imm_operand, "instruction has only an imm field");
817 #endif // LP64
818 return ip;
819 }
821 address Assembler::locate_next_instruction(address inst) {
822 // Secretly share code with locate_operand:
823 return locate_operand(inst, end_pc_operand);
824 }
827 #ifdef ASSERT
828 void Assembler::check_relocation(RelocationHolder const& rspec, int format) {
829 address inst = inst_mark();
830 assert(inst != NULL && inst < pc(), "must point to beginning of instruction");
831 address opnd;
833 Relocation* r = rspec.reloc();
834 if (r->type() == relocInfo::none) {
835 return;
836 } else if (r->is_call() || format == call32_operand) {
837 // assert(format == imm32_operand, "cannot specify a nonzero format");
838 opnd = locate_operand(inst, call32_operand);
839 } else if (r->is_data()) {
840 assert(format == imm_operand || format == disp32_operand
841 LP64_ONLY(|| format == narrow_oop_operand), "format ok");
842 opnd = locate_operand(inst, (WhichOperand)format);
843 } else {
844 assert(format == imm_operand, "cannot specify a format");
845 return;
846 }
847 assert(opnd == pc(), "must put operand where relocs can find it");
848 }
849 #endif // ASSERT
851 void Assembler::emit_operand32(Register reg, Address adr) {
852 assert(reg->encoding() < 8, "no extended registers");
853 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers");
854 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp,
855 adr._rspec);
856 }
858 void Assembler::emit_operand(Register reg, Address adr,
859 int rip_relative_correction) {
860 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp,
861 adr._rspec,
862 rip_relative_correction);
863 }
865 void Assembler::emit_operand(XMMRegister reg, Address adr) {
866 emit_operand(reg, adr._base, adr._index, adr._scale, adr._disp,
867 adr._rspec);
868 }
870 // MMX operations
871 void Assembler::emit_operand(MMXRegister reg, Address adr) {
872 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers");
873 emit_operand((Register)reg, adr._base, adr._index, adr._scale, adr._disp, adr._rspec);
874 }
876 // work around gcc (3.2.1-7a) bug
877 void Assembler::emit_operand(Address adr, MMXRegister reg) {
878 assert(!adr.base_needs_rex() && !adr.index_needs_rex(), "no extended registers");
879 emit_operand((Register)reg, adr._base, adr._index, adr._scale, adr._disp, adr._rspec);
880 }
883 void Assembler::emit_farith(int b1, int b2, int i) {
884 assert(isByte(b1) && isByte(b2), "wrong opcode");
885 assert(0 <= i && i < 8, "illegal stack offset");
886 emit_int8(b1);
887 emit_int8(b2 + i);
888 }
891 // Now the Assembler instructions (identical for 32/64 bits)
893 void Assembler::adcl(Address dst, int32_t imm32) {
894 InstructionMark im(this);
895 prefix(dst);
896 emit_arith_operand(0x81, rdx, dst, imm32);
897 }
899 void Assembler::adcl(Address dst, Register src) {
900 InstructionMark im(this);
901 prefix(dst, src);
902 emit_int8(0x11);
903 emit_operand(src, dst);
904 }
906 void Assembler::adcl(Register dst, int32_t imm32) {
907 prefix(dst);
908 emit_arith(0x81, 0xD0, dst, imm32);
909 }
911 void Assembler::adcl(Register dst, Address src) {
912 InstructionMark im(this);
913 prefix(src, dst);
914 emit_int8(0x13);
915 emit_operand(dst, src);
916 }
918 void Assembler::adcl(Register dst, Register src) {
919 (void) prefix_and_encode(dst->encoding(), src->encoding());
920 emit_arith(0x13, 0xC0, dst, src);
921 }
923 void Assembler::addl(Address dst, int32_t imm32) {
924 InstructionMark im(this);
925 prefix(dst);
926 emit_arith_operand(0x81, rax, dst, imm32);
927 }
929 void Assembler::addl(Address dst, Register src) {
930 InstructionMark im(this);
931 prefix(dst, src);
932 emit_int8(0x01);
933 emit_operand(src, dst);
934 }
936 void Assembler::addl(Register dst, int32_t imm32) {
937 prefix(dst);
938 emit_arith(0x81, 0xC0, dst, imm32);
939 }
941 void Assembler::addl(Register dst, Address src) {
942 InstructionMark im(this);
943 prefix(src, dst);
944 emit_int8(0x03);
945 emit_operand(dst, src);
946 }
948 void Assembler::addl(Register dst, Register src) {
949 (void) prefix_and_encode(dst->encoding(), src->encoding());
950 emit_arith(0x03, 0xC0, dst, src);
951 }
953 void Assembler::addr_nop_4() {
954 assert(UseAddressNop, "no CPU support");
955 // 4 bytes: NOP DWORD PTR [EAX+0]
956 emit_int8(0x0F);
957 emit_int8(0x1F);
958 emit_int8(0x40); // emit_rm(cbuf, 0x1, EAX_enc, EAX_enc);
959 emit_int8(0); // 8-bits offset (1 byte)
960 }
962 void Assembler::addr_nop_5() {
963 assert(UseAddressNop, "no CPU support");
964 // 5 bytes: NOP DWORD PTR [EAX+EAX*0+0] 8-bits offset
965 emit_int8(0x0F);
966 emit_int8(0x1F);
967 emit_int8(0x44); // emit_rm(cbuf, 0x1, EAX_enc, 0x4);
968 emit_int8(0x00); // emit_rm(cbuf, 0x0, EAX_enc, EAX_enc);
969 emit_int8(0); // 8-bits offset (1 byte)
970 }
972 void Assembler::addr_nop_7() {
973 assert(UseAddressNop, "no CPU support");
974 // 7 bytes: NOP DWORD PTR [EAX+0] 32-bits offset
975 emit_int8(0x0F);
976 emit_int8(0x1F);
977 emit_int8((unsigned char)0x80);
978 // emit_rm(cbuf, 0x2, EAX_enc, EAX_enc);
979 emit_long(0); // 32-bits offset (4 bytes)
980 }
982 void Assembler::addr_nop_8() {
983 assert(UseAddressNop, "no CPU support");
984 // 8 bytes: NOP DWORD PTR [EAX+EAX*0+0] 32-bits offset
985 emit_int8(0x0F);
986 emit_int8(0x1F);
987 emit_int8((unsigned char)0x84);
988 // emit_rm(cbuf, 0x2, EAX_enc, 0x4);
989 emit_int8(0x00); // emit_rm(cbuf, 0x0, EAX_enc, EAX_enc);
990 emit_long(0); // 32-bits offset (4 bytes)
991 }
993 void Assembler::addsd(XMMRegister dst, XMMRegister src) {
994 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
995 emit_simd_arith(0x58, dst, src, VEX_SIMD_F2);
996 }
998 void Assembler::addsd(XMMRegister dst, Address src) {
999 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1000 emit_simd_arith(0x58, dst, src, VEX_SIMD_F2);
1001 }
1003 void Assembler::addss(XMMRegister dst, XMMRegister src) {
1004 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1005 emit_simd_arith(0x58, dst, src, VEX_SIMD_F3);
1006 }
1008 void Assembler::addss(XMMRegister dst, Address src) {
1009 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1010 emit_simd_arith(0x58, dst, src, VEX_SIMD_F3);
1011 }
1013 void Assembler::aesdec(XMMRegister dst, Address src) {
1014 assert(VM_Version::supports_aes(), "");
1015 InstructionMark im(this);
1016 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1017 emit_int8((unsigned char)0xDE);
1018 emit_operand(dst, src);
1019 }
1021 void Assembler::aesdec(XMMRegister dst, XMMRegister src) {
1022 assert(VM_Version::supports_aes(), "");
1023 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1024 emit_int8((unsigned char)0xDE);
1025 emit_int8(0xC0 | encode);
1026 }
1028 void Assembler::aesdeclast(XMMRegister dst, Address src) {
1029 assert(VM_Version::supports_aes(), "");
1030 InstructionMark im(this);
1031 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1032 emit_int8((unsigned char)0xDF);
1033 emit_operand(dst, src);
1034 }
1036 void Assembler::aesdeclast(XMMRegister dst, XMMRegister src) {
1037 assert(VM_Version::supports_aes(), "");
1038 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1039 emit_int8((unsigned char)0xDF);
1040 emit_int8((unsigned char)(0xC0 | encode));
1041 }
1043 void Assembler::aesenc(XMMRegister dst, Address src) {
1044 assert(VM_Version::supports_aes(), "");
1045 InstructionMark im(this);
1046 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1047 emit_int8((unsigned char)0xDC);
1048 emit_operand(dst, src);
1049 }
1051 void Assembler::aesenc(XMMRegister dst, XMMRegister src) {
1052 assert(VM_Version::supports_aes(), "");
1053 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1054 emit_int8((unsigned char)0xDC);
1055 emit_int8(0xC0 | encode);
1056 }
1058 void Assembler::aesenclast(XMMRegister dst, Address src) {
1059 assert(VM_Version::supports_aes(), "");
1060 InstructionMark im(this);
1061 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1062 emit_int8((unsigned char)0xDD);
1063 emit_operand(dst, src);
1064 }
1066 void Assembler::aesenclast(XMMRegister dst, XMMRegister src) {
1067 assert(VM_Version::supports_aes(), "");
1068 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
1069 emit_int8((unsigned char)0xDD);
1070 emit_int8((unsigned char)(0xC0 | encode));
1071 }
1074 void Assembler::andl(Address dst, int32_t imm32) {
1075 InstructionMark im(this);
1076 prefix(dst);
1077 emit_int8((unsigned char)0x81);
1078 emit_operand(rsp, dst, 4);
1079 emit_long(imm32);
1080 }
1082 void Assembler::andl(Register dst, int32_t imm32) {
1083 prefix(dst);
1084 emit_arith(0x81, 0xE0, dst, imm32);
1085 }
1087 void Assembler::andl(Register dst, Address src) {
1088 InstructionMark im(this);
1089 prefix(src, dst);
1090 emit_int8(0x23);
1091 emit_operand(dst, src);
1092 }
1094 void Assembler::andl(Register dst, Register src) {
1095 (void) prefix_and_encode(dst->encoding(), src->encoding());
1096 emit_arith(0x23, 0xC0, dst, src);
1097 }
1099 void Assembler::bsfl(Register dst, Register src) {
1100 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1101 emit_int8(0x0F);
1102 emit_int8((unsigned char)0xBC);
1103 emit_int8((unsigned char)(0xC0 | encode));
1104 }
1106 void Assembler::bsrl(Register dst, Register src) {
1107 assert(!VM_Version::supports_lzcnt(), "encoding is treated as LZCNT");
1108 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1109 emit_int8(0x0F);
1110 emit_int8((unsigned char)0xBD);
1111 emit_int8((unsigned char)(0xC0 | encode));
1112 }
1114 void Assembler::bswapl(Register reg) { // bswap
1115 int encode = prefix_and_encode(reg->encoding());
1116 emit_int8(0x0F);
1117 emit_int8((unsigned char)(0xC8 | encode));
1118 }
1120 void Assembler::call(Label& L, relocInfo::relocType rtype) {
1121 // suspect disp32 is always good
1122 int operand = LP64_ONLY(disp32_operand) NOT_LP64(imm_operand);
1124 if (L.is_bound()) {
1125 const int long_size = 5;
1126 int offs = (int)( target(L) - pc() );
1127 assert(offs <= 0, "assembler error");
1128 InstructionMark im(this);
1129 // 1110 1000 #32-bit disp
1130 emit_int8((unsigned char)0xE8);
1131 emit_data(offs - long_size, rtype, operand);
1132 } else {
1133 InstructionMark im(this);
1134 // 1110 1000 #32-bit disp
1135 L.add_patch_at(code(), locator());
1137 emit_int8((unsigned char)0xE8);
1138 emit_data(int(0), rtype, operand);
1139 }
1140 }
1142 void Assembler::call(Register dst) {
1143 int encode = prefix_and_encode(dst->encoding());
1144 emit_int8((unsigned char)0xFF);
1145 emit_int8((unsigned char)(0xD0 | encode));
1146 }
1149 void Assembler::call(Address adr) {
1150 InstructionMark im(this);
1151 prefix(adr);
1152 emit_int8((unsigned char)0xFF);
1153 emit_operand(rdx, adr);
1154 }
1156 void Assembler::call_literal(address entry, RelocationHolder const& rspec) {
1157 assert(entry != NULL, "call most probably wrong");
1158 InstructionMark im(this);
1159 emit_int8((unsigned char)0xE8);
1160 intptr_t disp = entry - (pc() + sizeof(int32_t));
1161 assert(is_simm32(disp), "must be 32bit offset (call2)");
1162 // Technically, should use call32_operand, but this format is
1163 // implied by the fact that we're emitting a call instruction.
1165 int operand = LP64_ONLY(disp32_operand) NOT_LP64(call32_operand);
1166 emit_data((int) disp, rspec, operand);
1167 }
1169 void Assembler::cdql() {
1170 emit_int8((unsigned char)0x99);
1171 }
1173 void Assembler::cld() {
1174 emit_int8((unsigned char)0xFC);
1175 }
1177 void Assembler::cmovl(Condition cc, Register dst, Register src) {
1178 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction"));
1179 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1180 emit_int8(0x0F);
1181 emit_int8(0x40 | cc);
1182 emit_int8((unsigned char)(0xC0 | encode));
1183 }
1186 void Assembler::cmovl(Condition cc, Register dst, Address src) {
1187 NOT_LP64(guarantee(VM_Version::supports_cmov(), "illegal instruction"));
1188 prefix(src, dst);
1189 emit_int8(0x0F);
1190 emit_int8(0x40 | cc);
1191 emit_operand(dst, src);
1192 }
1194 void Assembler::cmpb(Address dst, int imm8) {
1195 InstructionMark im(this);
1196 prefix(dst);
1197 emit_int8((unsigned char)0x80);
1198 emit_operand(rdi, dst, 1);
1199 emit_int8(imm8);
1200 }
1202 void Assembler::cmpl(Address dst, int32_t imm32) {
1203 InstructionMark im(this);
1204 prefix(dst);
1205 emit_int8((unsigned char)0x81);
1206 emit_operand(rdi, dst, 4);
1207 emit_long(imm32);
1208 }
1210 void Assembler::cmpl(Register dst, int32_t imm32) {
1211 prefix(dst);
1212 emit_arith(0x81, 0xF8, dst, imm32);
1213 }
1215 void Assembler::cmpl(Register dst, Register src) {
1216 (void) prefix_and_encode(dst->encoding(), src->encoding());
1217 emit_arith(0x3B, 0xC0, dst, src);
1218 }
1221 void Assembler::cmpl(Register dst, Address src) {
1222 InstructionMark im(this);
1223 prefix(src, dst);
1224 emit_int8((unsigned char)0x3B);
1225 emit_operand(dst, src);
1226 }
1228 void Assembler::cmpw(Address dst, int imm16) {
1229 InstructionMark im(this);
1230 assert(!dst.base_needs_rex() && !dst.index_needs_rex(), "no extended registers");
1231 emit_int8(0x66);
1232 emit_int8((unsigned char)0x81);
1233 emit_operand(rdi, dst, 2);
1234 emit_int16(imm16);
1235 }
1237 // The 32-bit cmpxchg compares the value at adr with the contents of rax,
1238 // and stores reg into adr if so; otherwise, the value at adr is loaded into rax,.
1239 // The ZF is set if the compared values were equal, and cleared otherwise.
1240 void Assembler::cmpxchgl(Register reg, Address adr) { // cmpxchg
1241 InstructionMark im(this);
1242 prefix(adr, reg);
1243 emit_int8(0x0F);
1244 emit_int8((unsigned char)0xB1);
1245 emit_operand(reg, adr);
1246 }
1248 void Assembler::comisd(XMMRegister dst, Address src) {
1249 // NOTE: dbx seems to decode this as comiss even though the
1250 // 0x66 is there. Strangly ucomisd comes out correct
1251 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1252 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66);
1253 }
1255 void Assembler::comisd(XMMRegister dst, XMMRegister src) {
1256 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1257 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_66);
1258 }
1260 void Assembler::comiss(XMMRegister dst, Address src) {
1261 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1262 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE);
1263 }
1265 void Assembler::comiss(XMMRegister dst, XMMRegister src) {
1266 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1267 emit_simd_arith_nonds(0x2F, dst, src, VEX_SIMD_NONE);
1268 }
1270 void Assembler::cpuid() {
1271 emit_int8(0x0F);
1272 emit_int8((unsigned char)0xA2);
1273 }
1275 void Assembler::cvtdq2pd(XMMRegister dst, XMMRegister src) {
1276 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1277 emit_simd_arith_nonds(0xE6, dst, src, VEX_SIMD_F3);
1278 }
1280 void Assembler::cvtdq2ps(XMMRegister dst, XMMRegister src) {
1281 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1282 emit_simd_arith_nonds(0x5B, dst, src, VEX_SIMD_NONE);
1283 }
1285 void Assembler::cvtsd2ss(XMMRegister dst, XMMRegister src) {
1286 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1287 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2);
1288 }
1290 void Assembler::cvtsd2ss(XMMRegister dst, Address src) {
1291 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1292 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F2);
1293 }
1295 void Assembler::cvtsi2sdl(XMMRegister dst, Register src) {
1296 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1297 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F2);
1298 emit_int8(0x2A);
1299 emit_int8((unsigned char)(0xC0 | encode));
1300 }
1302 void Assembler::cvtsi2sdl(XMMRegister dst, Address src) {
1303 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1304 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F2);
1305 }
1307 void Assembler::cvtsi2ssl(XMMRegister dst, Register src) {
1308 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1309 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_F3);
1310 emit_int8(0x2A);
1311 emit_int8((unsigned char)(0xC0 | encode));
1312 }
1314 void Assembler::cvtsi2ssl(XMMRegister dst, Address src) {
1315 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1316 emit_simd_arith(0x2A, dst, src, VEX_SIMD_F3);
1317 }
1319 void Assembler::cvtss2sd(XMMRegister dst, XMMRegister src) {
1320 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1321 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3);
1322 }
1324 void Assembler::cvtss2sd(XMMRegister dst, Address src) {
1325 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1326 emit_simd_arith(0x5A, dst, src, VEX_SIMD_F3);
1327 }
1330 void Assembler::cvttsd2sil(Register dst, XMMRegister src) {
1331 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1332 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F2);
1333 emit_int8(0x2C);
1334 emit_int8((unsigned char)(0xC0 | encode));
1335 }
1337 void Assembler::cvttss2sil(Register dst, XMMRegister src) {
1338 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1339 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_F3);
1340 emit_int8(0x2C);
1341 emit_int8((unsigned char)(0xC0 | encode));
1342 }
1344 void Assembler::decl(Address dst) {
1345 // Don't use it directly. Use MacroAssembler::decrement() instead.
1346 InstructionMark im(this);
1347 prefix(dst);
1348 emit_int8((unsigned char)0xFF);
1349 emit_operand(rcx, dst);
1350 }
1352 void Assembler::divsd(XMMRegister dst, Address src) {
1353 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1354 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2);
1355 }
1357 void Assembler::divsd(XMMRegister dst, XMMRegister src) {
1358 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1359 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F2);
1360 }
1362 void Assembler::divss(XMMRegister dst, Address src) {
1363 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1364 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3);
1365 }
1367 void Assembler::divss(XMMRegister dst, XMMRegister src) {
1368 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1369 emit_simd_arith(0x5E, dst, src, VEX_SIMD_F3);
1370 }
1372 void Assembler::emms() {
1373 NOT_LP64(assert(VM_Version::supports_mmx(), ""));
1374 emit_int8(0x0F);
1375 emit_int8(0x77);
1376 }
1378 void Assembler::hlt() {
1379 emit_int8((unsigned char)0xF4);
1380 }
1382 void Assembler::idivl(Register src) {
1383 int encode = prefix_and_encode(src->encoding());
1384 emit_int8((unsigned char)0xF7);
1385 emit_int8((unsigned char)(0xF8 | encode));
1386 }
1388 void Assembler::divl(Register src) { // Unsigned
1389 int encode = prefix_and_encode(src->encoding());
1390 emit_int8((unsigned char)0xF7);
1391 emit_int8((unsigned char)(0xF0 | encode));
1392 }
1394 void Assembler::imull(Register dst, Register src) {
1395 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1396 emit_int8(0x0F);
1397 emit_int8((unsigned char)0xAF);
1398 emit_int8((unsigned char)(0xC0 | encode));
1399 }
1402 void Assembler::imull(Register dst, Register src, int value) {
1403 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1404 if (is8bit(value)) {
1405 emit_int8(0x6B);
1406 emit_int8((unsigned char)(0xC0 | encode));
1407 emit_int8(value & 0xFF);
1408 } else {
1409 emit_int8(0x69);
1410 emit_int8((unsigned char)(0xC0 | encode));
1411 emit_long(value);
1412 }
1413 }
1415 void Assembler::incl(Address dst) {
1416 // Don't use it directly. Use MacroAssembler::increment() instead.
1417 InstructionMark im(this);
1418 prefix(dst);
1419 emit_int8((unsigned char)0xFF);
1420 emit_operand(rax, dst);
1421 }
1423 void Assembler::jcc(Condition cc, Label& L, bool maybe_short) {
1424 InstructionMark im(this);
1425 assert((0 <= cc) && (cc < 16), "illegal cc");
1426 if (L.is_bound()) {
1427 address dst = target(L);
1428 assert(dst != NULL, "jcc most probably wrong");
1430 const int short_size = 2;
1431 const int long_size = 6;
1432 intptr_t offs = (intptr_t)dst - (intptr_t)pc();
1433 if (maybe_short && is8bit(offs - short_size)) {
1434 // 0111 tttn #8-bit disp
1435 emit_int8(0x70 | cc);
1436 emit_int8((offs - short_size) & 0xFF);
1437 } else {
1438 // 0000 1111 1000 tttn #32-bit disp
1439 assert(is_simm32(offs - long_size),
1440 "must be 32bit offset (call4)");
1441 emit_int8(0x0F);
1442 emit_int8((unsigned char)(0x80 | cc));
1443 emit_long(offs - long_size);
1444 }
1445 } else {
1446 // Note: could eliminate cond. jumps to this jump if condition
1447 // is the same however, seems to be rather unlikely case.
1448 // Note: use jccb() if label to be bound is very close to get
1449 // an 8-bit displacement
1450 L.add_patch_at(code(), locator());
1451 emit_int8(0x0F);
1452 emit_int8((unsigned char)(0x80 | cc));
1453 emit_long(0);
1454 }
1455 }
1457 void Assembler::jccb(Condition cc, Label& L) {
1458 if (L.is_bound()) {
1459 const int short_size = 2;
1460 address entry = target(L);
1461 #ifdef ASSERT
1462 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size);
1463 intptr_t delta = short_branch_delta();
1464 if (delta != 0) {
1465 dist += (dist < 0 ? (-delta) :delta);
1466 }
1467 assert(is8bit(dist), "Dispacement too large for a short jmp");
1468 #endif
1469 intptr_t offs = (intptr_t)entry - (intptr_t)pc();
1470 // 0111 tttn #8-bit disp
1471 emit_int8(0x70 | cc);
1472 emit_int8((offs - short_size) & 0xFF);
1473 } else {
1474 InstructionMark im(this);
1475 L.add_patch_at(code(), locator());
1476 emit_int8(0x70 | cc);
1477 emit_int8(0);
1478 }
1479 }
1481 void Assembler::jmp(Address adr) {
1482 InstructionMark im(this);
1483 prefix(adr);
1484 emit_int8((unsigned char)0xFF);
1485 emit_operand(rsp, adr);
1486 }
1488 void Assembler::jmp(Label& L, bool maybe_short) {
1489 if (L.is_bound()) {
1490 address entry = target(L);
1491 assert(entry != NULL, "jmp most probably wrong");
1492 InstructionMark im(this);
1493 const int short_size = 2;
1494 const int long_size = 5;
1495 intptr_t offs = entry - pc();
1496 if (maybe_short && is8bit(offs - short_size)) {
1497 emit_int8((unsigned char)0xEB);
1498 emit_int8((offs - short_size) & 0xFF);
1499 } else {
1500 emit_int8((unsigned char)0xE9);
1501 emit_long(offs - long_size);
1502 }
1503 } else {
1504 // By default, forward jumps are always 32-bit displacements, since
1505 // we can't yet know where the label will be bound. If you're sure that
1506 // the forward jump will not run beyond 256 bytes, use jmpb to
1507 // force an 8-bit displacement.
1508 InstructionMark im(this);
1509 L.add_patch_at(code(), locator());
1510 emit_int8((unsigned char)0xE9);
1511 emit_long(0);
1512 }
1513 }
1515 void Assembler::jmp(Register entry) {
1516 int encode = prefix_and_encode(entry->encoding());
1517 emit_int8((unsigned char)0xFF);
1518 emit_int8((unsigned char)(0xE0 | encode));
1519 }
1521 void Assembler::jmp_literal(address dest, RelocationHolder const& rspec) {
1522 InstructionMark im(this);
1523 emit_int8((unsigned char)0xE9);
1524 assert(dest != NULL, "must have a target");
1525 intptr_t disp = dest - (pc() + sizeof(int32_t));
1526 assert(is_simm32(disp), "must be 32bit offset (jmp)");
1527 emit_data(disp, rspec.reloc(), call32_operand);
1528 }
1530 void Assembler::jmpb(Label& L) {
1531 if (L.is_bound()) {
1532 const int short_size = 2;
1533 address entry = target(L);
1534 assert(entry != NULL, "jmp most probably wrong");
1535 #ifdef ASSERT
1536 intptr_t dist = (intptr_t)entry - ((intptr_t)pc() + short_size);
1537 intptr_t delta = short_branch_delta();
1538 if (delta != 0) {
1539 dist += (dist < 0 ? (-delta) :delta);
1540 }
1541 assert(is8bit(dist), "Dispacement too large for a short jmp");
1542 #endif
1543 intptr_t offs = entry - pc();
1544 emit_int8((unsigned char)0xEB);
1545 emit_int8((offs - short_size) & 0xFF);
1546 } else {
1547 InstructionMark im(this);
1548 L.add_patch_at(code(), locator());
1549 emit_int8((unsigned char)0xEB);
1550 emit_int8(0);
1551 }
1552 }
1554 void Assembler::ldmxcsr( Address src) {
1555 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1556 InstructionMark im(this);
1557 prefix(src);
1558 emit_int8(0x0F);
1559 emit_int8((unsigned char)0xAE);
1560 emit_operand(as_Register(2), src);
1561 }
1563 void Assembler::leal(Register dst, Address src) {
1564 InstructionMark im(this);
1565 #ifdef _LP64
1566 emit_int8(0x67); // addr32
1567 prefix(src, dst);
1568 #endif // LP64
1569 emit_int8((unsigned char)0x8D);
1570 emit_operand(dst, src);
1571 }
1573 void Assembler::lfence() {
1574 emit_int8(0x0F);
1575 emit_int8((unsigned char)0xAE);
1576 emit_int8((unsigned char)0xE8);
1577 }
1579 void Assembler::lock() {
1580 emit_int8((unsigned char)0xF0);
1581 }
1583 void Assembler::lzcntl(Register dst, Register src) {
1584 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR");
1585 emit_int8((unsigned char)0xF3);
1586 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1587 emit_int8(0x0F);
1588 emit_int8((unsigned char)0xBD);
1589 emit_int8((unsigned char)(0xC0 | encode));
1590 }
1592 // Emit mfence instruction
1593 void Assembler::mfence() {
1594 NOT_LP64(assert(VM_Version::supports_sse2(), "unsupported");)
1595 emit_int8(0x0F);
1596 emit_int8((unsigned char)0xAE);
1597 emit_int8((unsigned char)0xF0);
1598 }
1600 void Assembler::mov(Register dst, Register src) {
1601 LP64_ONLY(movq(dst, src)) NOT_LP64(movl(dst, src));
1602 }
1604 void Assembler::movapd(XMMRegister dst, XMMRegister src) {
1605 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1606 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_66);
1607 }
1609 void Assembler::movaps(XMMRegister dst, XMMRegister src) {
1610 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1611 emit_simd_arith_nonds(0x28, dst, src, VEX_SIMD_NONE);
1612 }
1614 void Assembler::movlhps(XMMRegister dst, XMMRegister src) {
1615 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1616 int encode = simd_prefix_and_encode(dst, src, src, VEX_SIMD_NONE);
1617 emit_int8(0x16);
1618 emit_int8((unsigned char)(0xC0 | encode));
1619 }
1621 void Assembler::movb(Register dst, Address src) {
1622 NOT_LP64(assert(dst->has_byte_register(), "must have byte register"));
1623 InstructionMark im(this);
1624 prefix(src, dst, true);
1625 emit_int8((unsigned char)0x8A);
1626 emit_operand(dst, src);
1627 }
1630 void Assembler::movb(Address dst, int imm8) {
1631 InstructionMark im(this);
1632 prefix(dst);
1633 emit_int8((unsigned char)0xC6);
1634 emit_operand(rax, dst, 1);
1635 emit_int8(imm8);
1636 }
1639 void Assembler::movb(Address dst, Register src) {
1640 assert(src->has_byte_register(), "must have byte register");
1641 InstructionMark im(this);
1642 prefix(dst, src, true);
1643 emit_int8((unsigned char)0x88);
1644 emit_operand(src, dst);
1645 }
1647 void Assembler::movdl(XMMRegister dst, Register src) {
1648 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1649 int encode = simd_prefix_and_encode(dst, src, VEX_SIMD_66);
1650 emit_int8(0x6E);
1651 emit_int8((unsigned char)(0xC0 | encode));
1652 }
1654 void Assembler::movdl(Register dst, XMMRegister src) {
1655 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1656 // swap src/dst to get correct prefix
1657 int encode = simd_prefix_and_encode(src, dst, VEX_SIMD_66);
1658 emit_int8(0x7E);
1659 emit_int8((unsigned char)(0xC0 | encode));
1660 }
1662 void Assembler::movdl(XMMRegister dst, Address src) {
1663 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1664 InstructionMark im(this);
1665 simd_prefix(dst, src, VEX_SIMD_66);
1666 emit_int8(0x6E);
1667 emit_operand(dst, src);
1668 }
1670 void Assembler::movdl(Address dst, XMMRegister src) {
1671 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1672 InstructionMark im(this);
1673 simd_prefix(dst, src, VEX_SIMD_66);
1674 emit_int8(0x7E);
1675 emit_operand(src, dst);
1676 }
1678 void Assembler::movdqa(XMMRegister dst, XMMRegister src) {
1679 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1680 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_66);
1681 }
1683 void Assembler::movdqu(XMMRegister dst, Address src) {
1684 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1685 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3);
1686 }
1688 void Assembler::movdqu(XMMRegister dst, XMMRegister src) {
1689 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1690 emit_simd_arith_nonds(0x6F, dst, src, VEX_SIMD_F3);
1691 }
1693 void Assembler::movdqu(Address dst, XMMRegister src) {
1694 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1695 InstructionMark im(this);
1696 simd_prefix(dst, src, VEX_SIMD_F3);
1697 emit_int8(0x7F);
1698 emit_operand(src, dst);
1699 }
1701 // Move Unaligned 256bit Vector
1702 void Assembler::vmovdqu(XMMRegister dst, XMMRegister src) {
1703 assert(UseAVX, "");
1704 bool vector256 = true;
1705 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_F3, vector256);
1706 emit_int8(0x6F);
1707 emit_int8((unsigned char)(0xC0 | encode));
1708 }
1710 void Assembler::vmovdqu(XMMRegister dst, Address src) {
1711 assert(UseAVX, "");
1712 InstructionMark im(this);
1713 bool vector256 = true;
1714 vex_prefix(dst, xnoreg, src, VEX_SIMD_F3, vector256);
1715 emit_int8(0x6F);
1716 emit_operand(dst, src);
1717 }
1719 void Assembler::vmovdqu(Address dst, XMMRegister src) {
1720 assert(UseAVX, "");
1721 InstructionMark im(this);
1722 bool vector256 = true;
1723 // swap src<->dst for encoding
1724 assert(src != xnoreg, "sanity");
1725 vex_prefix(src, xnoreg, dst, VEX_SIMD_F3, vector256);
1726 emit_int8(0x7F);
1727 emit_operand(src, dst);
1728 }
1730 // Uses zero extension on 64bit
1732 void Assembler::movl(Register dst, int32_t imm32) {
1733 int encode = prefix_and_encode(dst->encoding());
1734 emit_int8((unsigned char)(0xB8 | encode));
1735 emit_long(imm32);
1736 }
1738 void Assembler::movl(Register dst, Register src) {
1739 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1740 emit_int8((unsigned char)0x8B);
1741 emit_int8((unsigned char)(0xC0 | encode));
1742 }
1744 void Assembler::movl(Register dst, Address src) {
1745 InstructionMark im(this);
1746 prefix(src, dst);
1747 emit_int8((unsigned char)0x8B);
1748 emit_operand(dst, src);
1749 }
1751 void Assembler::movl(Address dst, int32_t imm32) {
1752 InstructionMark im(this);
1753 prefix(dst);
1754 emit_int8((unsigned char)0xC7);
1755 emit_operand(rax, dst, 4);
1756 emit_long(imm32);
1757 }
1759 void Assembler::movl(Address dst, Register src) {
1760 InstructionMark im(this);
1761 prefix(dst, src);
1762 emit_int8((unsigned char)0x89);
1763 emit_operand(src, dst);
1764 }
1766 // New cpus require to use movsd and movss to avoid partial register stall
1767 // when loading from memory. But for old Opteron use movlpd instead of movsd.
1768 // The selection is done in MacroAssembler::movdbl() and movflt().
1769 void Assembler::movlpd(XMMRegister dst, Address src) {
1770 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1771 emit_simd_arith(0x12, dst, src, VEX_SIMD_66);
1772 }
1774 void Assembler::movq( MMXRegister dst, Address src ) {
1775 assert( VM_Version::supports_mmx(), "" );
1776 emit_int8(0x0F);
1777 emit_int8(0x6F);
1778 emit_operand(dst, src);
1779 }
1781 void Assembler::movq( Address dst, MMXRegister src ) {
1782 assert( VM_Version::supports_mmx(), "" );
1783 emit_int8(0x0F);
1784 emit_int8(0x7F);
1785 // workaround gcc (3.2.1-7a) bug
1786 // In that version of gcc with only an emit_operand(MMX, Address)
1787 // gcc will tail jump and try and reverse the parameters completely
1788 // obliterating dst in the process. By having a version available
1789 // that doesn't need to swap the args at the tail jump the bug is
1790 // avoided.
1791 emit_operand(dst, src);
1792 }
1794 void Assembler::movq(XMMRegister dst, Address src) {
1795 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1796 InstructionMark im(this);
1797 simd_prefix(dst, src, VEX_SIMD_F3);
1798 emit_int8(0x7E);
1799 emit_operand(dst, src);
1800 }
1802 void Assembler::movq(Address dst, XMMRegister src) {
1803 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1804 InstructionMark im(this);
1805 simd_prefix(dst, src, VEX_SIMD_66);
1806 emit_int8((unsigned char)0xD6);
1807 emit_operand(src, dst);
1808 }
1810 void Assembler::movsbl(Register dst, Address src) { // movsxb
1811 InstructionMark im(this);
1812 prefix(src, dst);
1813 emit_int8(0x0F);
1814 emit_int8((unsigned char)0xBE);
1815 emit_operand(dst, src);
1816 }
1818 void Assembler::movsbl(Register dst, Register src) { // movsxb
1819 NOT_LP64(assert(src->has_byte_register(), "must have byte register"));
1820 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true);
1821 emit_int8(0x0F);
1822 emit_int8((unsigned char)0xBE);
1823 emit_int8((unsigned char)(0xC0 | encode));
1824 }
1826 void Assembler::movsd(XMMRegister dst, XMMRegister src) {
1827 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1828 emit_simd_arith(0x10, dst, src, VEX_SIMD_F2);
1829 }
1831 void Assembler::movsd(XMMRegister dst, Address src) {
1832 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1833 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F2);
1834 }
1836 void Assembler::movsd(Address dst, XMMRegister src) {
1837 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1838 InstructionMark im(this);
1839 simd_prefix(dst, src, VEX_SIMD_F2);
1840 emit_int8(0x11);
1841 emit_operand(src, dst);
1842 }
1844 void Assembler::movss(XMMRegister dst, XMMRegister src) {
1845 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1846 emit_simd_arith(0x10, dst, src, VEX_SIMD_F3);
1847 }
1849 void Assembler::movss(XMMRegister dst, Address src) {
1850 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1851 emit_simd_arith_nonds(0x10, dst, src, VEX_SIMD_F3);
1852 }
1854 void Assembler::movss(Address dst, XMMRegister src) {
1855 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1856 InstructionMark im(this);
1857 simd_prefix(dst, src, VEX_SIMD_F3);
1858 emit_int8(0x11);
1859 emit_operand(src, dst);
1860 }
1862 void Assembler::movswl(Register dst, Address src) { // movsxw
1863 InstructionMark im(this);
1864 prefix(src, dst);
1865 emit_int8(0x0F);
1866 emit_int8((unsigned char)0xBF);
1867 emit_operand(dst, src);
1868 }
1870 void Assembler::movswl(Register dst, Register src) { // movsxw
1871 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1872 emit_int8(0x0F);
1873 emit_int8((unsigned char)0xBF);
1874 emit_int8((unsigned char)(0xC0 | encode));
1875 }
1877 void Assembler::movw(Address dst, int imm16) {
1878 InstructionMark im(this);
1880 emit_int8(0x66); // switch to 16-bit mode
1881 prefix(dst);
1882 emit_int8((unsigned char)0xC7);
1883 emit_operand(rax, dst, 2);
1884 emit_int16(imm16);
1885 }
1887 void Assembler::movw(Register dst, Address src) {
1888 InstructionMark im(this);
1889 emit_int8(0x66);
1890 prefix(src, dst);
1891 emit_int8((unsigned char)0x8B);
1892 emit_operand(dst, src);
1893 }
1895 void Assembler::movw(Address dst, Register src) {
1896 InstructionMark im(this);
1897 emit_int8(0x66);
1898 prefix(dst, src);
1899 emit_int8((unsigned char)0x89);
1900 emit_operand(src, dst);
1901 }
1903 void Assembler::movzbl(Register dst, Address src) { // movzxb
1904 InstructionMark im(this);
1905 prefix(src, dst);
1906 emit_int8(0x0F);
1907 emit_int8((unsigned char)0xB6);
1908 emit_operand(dst, src);
1909 }
1911 void Assembler::movzbl(Register dst, Register src) { // movzxb
1912 NOT_LP64(assert(src->has_byte_register(), "must have byte register"));
1913 int encode = prefix_and_encode(dst->encoding(), src->encoding(), true);
1914 emit_int8(0x0F);
1915 emit_int8((unsigned char)0xB6);
1916 emit_int8(0xC0 | encode);
1917 }
1919 void Assembler::movzwl(Register dst, Address src) { // movzxw
1920 InstructionMark im(this);
1921 prefix(src, dst);
1922 emit_int8(0x0F);
1923 emit_int8((unsigned char)0xB7);
1924 emit_operand(dst, src);
1925 }
1927 void Assembler::movzwl(Register dst, Register src) { // movzxw
1928 int encode = prefix_and_encode(dst->encoding(), src->encoding());
1929 emit_int8(0x0F);
1930 emit_int8((unsigned char)0xB7);
1931 emit_int8(0xC0 | encode);
1932 }
1934 void Assembler::mull(Address src) {
1935 InstructionMark im(this);
1936 prefix(src);
1937 emit_int8((unsigned char)0xF7);
1938 emit_operand(rsp, src);
1939 }
1941 void Assembler::mull(Register src) {
1942 int encode = prefix_and_encode(src->encoding());
1943 emit_int8((unsigned char)0xF7);
1944 emit_int8((unsigned char)(0xE0 | encode));
1945 }
1947 void Assembler::mulsd(XMMRegister dst, Address src) {
1948 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1949 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2);
1950 }
1952 void Assembler::mulsd(XMMRegister dst, XMMRegister src) {
1953 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
1954 emit_simd_arith(0x59, dst, src, VEX_SIMD_F2);
1955 }
1957 void Assembler::mulss(XMMRegister dst, Address src) {
1958 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1959 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3);
1960 }
1962 void Assembler::mulss(XMMRegister dst, XMMRegister src) {
1963 NOT_LP64(assert(VM_Version::supports_sse(), ""));
1964 emit_simd_arith(0x59, dst, src, VEX_SIMD_F3);
1965 }
1967 void Assembler::negl(Register dst) {
1968 int encode = prefix_and_encode(dst->encoding());
1969 emit_int8((unsigned char)0xF7);
1970 emit_int8((unsigned char)(0xD8 | encode));
1971 }
1973 void Assembler::nop(int i) {
1974 #ifdef ASSERT
1975 assert(i > 0, " ");
1976 // The fancy nops aren't currently recognized by debuggers making it a
1977 // pain to disassemble code while debugging. If asserts are on clearly
1978 // speed is not an issue so simply use the single byte traditional nop
1979 // to do alignment.
1981 for (; i > 0 ; i--) emit_int8((unsigned char)0x90);
1982 return;
1984 #endif // ASSERT
1986 if (UseAddressNop && VM_Version::is_intel()) {
1987 //
1988 // Using multi-bytes nops "0x0F 0x1F [address]" for Intel
1989 // 1: 0x90
1990 // 2: 0x66 0x90
1991 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding)
1992 // 4: 0x0F 0x1F 0x40 0x00
1993 // 5: 0x0F 0x1F 0x44 0x00 0x00
1994 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00
1995 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00
1996 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
1997 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
1998 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
1999 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2001 // The rest coding is Intel specific - don't use consecutive address nops
2003 // 12: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90
2004 // 13: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90
2005 // 14: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90
2006 // 15: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x66 0x66 0x66 0x90
2008 while(i >= 15) {
2009 // For Intel don't generate consecutive addess nops (mix with regular nops)
2010 i -= 15;
2011 emit_int8(0x66); // size prefix
2012 emit_int8(0x66); // size prefix
2013 emit_int8(0x66); // size prefix
2014 addr_nop_8();
2015 emit_int8(0x66); // size prefix
2016 emit_int8(0x66); // size prefix
2017 emit_int8(0x66); // size prefix
2018 emit_int8((unsigned char)0x90);
2019 // nop
2020 }
2021 switch (i) {
2022 case 14:
2023 emit_int8(0x66); // size prefix
2024 case 13:
2025 emit_int8(0x66); // size prefix
2026 case 12:
2027 addr_nop_8();
2028 emit_int8(0x66); // size prefix
2029 emit_int8(0x66); // size prefix
2030 emit_int8(0x66); // size prefix
2031 emit_int8((unsigned char)0x90);
2032 // nop
2033 break;
2034 case 11:
2035 emit_int8(0x66); // size prefix
2036 case 10:
2037 emit_int8(0x66); // size prefix
2038 case 9:
2039 emit_int8(0x66); // size prefix
2040 case 8:
2041 addr_nop_8();
2042 break;
2043 case 7:
2044 addr_nop_7();
2045 break;
2046 case 6:
2047 emit_int8(0x66); // size prefix
2048 case 5:
2049 addr_nop_5();
2050 break;
2051 case 4:
2052 addr_nop_4();
2053 break;
2054 case 3:
2055 // Don't use "0x0F 0x1F 0x00" - need patching safe padding
2056 emit_int8(0x66); // size prefix
2057 case 2:
2058 emit_int8(0x66); // size prefix
2059 case 1:
2060 emit_int8((unsigned char)0x90);
2061 // nop
2062 break;
2063 default:
2064 assert(i == 0, " ");
2065 }
2066 return;
2067 }
2068 if (UseAddressNop && VM_Version::is_amd()) {
2069 //
2070 // Using multi-bytes nops "0x0F 0x1F [address]" for AMD.
2071 // 1: 0x90
2072 // 2: 0x66 0x90
2073 // 3: 0x66 0x66 0x90 (don't use "0x0F 0x1F 0x00" - need patching safe padding)
2074 // 4: 0x0F 0x1F 0x40 0x00
2075 // 5: 0x0F 0x1F 0x44 0x00 0x00
2076 // 6: 0x66 0x0F 0x1F 0x44 0x00 0x00
2077 // 7: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00
2078 // 8: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2079 // 9: 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2080 // 10: 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2081 // 11: 0x66 0x66 0x66 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2083 // The rest coding is AMD specific - use consecutive address nops
2085 // 12: 0x66 0x0F 0x1F 0x44 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00
2086 // 13: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x66 0x0F 0x1F 0x44 0x00 0x00
2087 // 14: 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00
2088 // 15: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x80 0x00 0x00 0x00 0x00
2089 // 16: 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00 0x0F 0x1F 0x84 0x00 0x00 0x00 0x00 0x00
2090 // Size prefixes (0x66) are added for larger sizes
2092 while(i >= 22) {
2093 i -= 11;
2094 emit_int8(0x66); // size prefix
2095 emit_int8(0x66); // size prefix
2096 emit_int8(0x66); // size prefix
2097 addr_nop_8();
2098 }
2099 // Generate first nop for size between 21-12
2100 switch (i) {
2101 case 21:
2102 i -= 1;
2103 emit_int8(0x66); // size prefix
2104 case 20:
2105 case 19:
2106 i -= 1;
2107 emit_int8(0x66); // size prefix
2108 case 18:
2109 case 17:
2110 i -= 1;
2111 emit_int8(0x66); // size prefix
2112 case 16:
2113 case 15:
2114 i -= 8;
2115 addr_nop_8();
2116 break;
2117 case 14:
2118 case 13:
2119 i -= 7;
2120 addr_nop_7();
2121 break;
2122 case 12:
2123 i -= 6;
2124 emit_int8(0x66); // size prefix
2125 addr_nop_5();
2126 break;
2127 default:
2128 assert(i < 12, " ");
2129 }
2131 // Generate second nop for size between 11-1
2132 switch (i) {
2133 case 11:
2134 emit_int8(0x66); // size prefix
2135 case 10:
2136 emit_int8(0x66); // size prefix
2137 case 9:
2138 emit_int8(0x66); // size prefix
2139 case 8:
2140 addr_nop_8();
2141 break;
2142 case 7:
2143 addr_nop_7();
2144 break;
2145 case 6:
2146 emit_int8(0x66); // size prefix
2147 case 5:
2148 addr_nop_5();
2149 break;
2150 case 4:
2151 addr_nop_4();
2152 break;
2153 case 3:
2154 // Don't use "0x0F 0x1F 0x00" - need patching safe padding
2155 emit_int8(0x66); // size prefix
2156 case 2:
2157 emit_int8(0x66); // size prefix
2158 case 1:
2159 emit_int8((unsigned char)0x90);
2160 // nop
2161 break;
2162 default:
2163 assert(i == 0, " ");
2164 }
2165 return;
2166 }
2168 // Using nops with size prefixes "0x66 0x90".
2169 // From AMD Optimization Guide:
2170 // 1: 0x90
2171 // 2: 0x66 0x90
2172 // 3: 0x66 0x66 0x90
2173 // 4: 0x66 0x66 0x66 0x90
2174 // 5: 0x66 0x66 0x90 0x66 0x90
2175 // 6: 0x66 0x66 0x90 0x66 0x66 0x90
2176 // 7: 0x66 0x66 0x66 0x90 0x66 0x66 0x90
2177 // 8: 0x66 0x66 0x66 0x90 0x66 0x66 0x66 0x90
2178 // 9: 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90
2179 // 10: 0x66 0x66 0x66 0x90 0x66 0x66 0x90 0x66 0x66 0x90
2180 //
2181 while(i > 12) {
2182 i -= 4;
2183 emit_int8(0x66); // size prefix
2184 emit_int8(0x66);
2185 emit_int8(0x66);
2186 emit_int8((unsigned char)0x90);
2187 // nop
2188 }
2189 // 1 - 12 nops
2190 if(i > 8) {
2191 if(i > 9) {
2192 i -= 1;
2193 emit_int8(0x66);
2194 }
2195 i -= 3;
2196 emit_int8(0x66);
2197 emit_int8(0x66);
2198 emit_int8((unsigned char)0x90);
2199 }
2200 // 1 - 8 nops
2201 if(i > 4) {
2202 if(i > 6) {
2203 i -= 1;
2204 emit_int8(0x66);
2205 }
2206 i -= 3;
2207 emit_int8(0x66);
2208 emit_int8(0x66);
2209 emit_int8((unsigned char)0x90);
2210 }
2211 switch (i) {
2212 case 4:
2213 emit_int8(0x66);
2214 case 3:
2215 emit_int8(0x66);
2216 case 2:
2217 emit_int8(0x66);
2218 case 1:
2219 emit_int8((unsigned char)0x90);
2220 break;
2221 default:
2222 assert(i == 0, " ");
2223 }
2224 }
2226 void Assembler::notl(Register dst) {
2227 int encode = prefix_and_encode(dst->encoding());
2228 emit_int8((unsigned char)0xF7);
2229 emit_int8((unsigned char)(0xD0 | encode));
2230 }
2232 void Assembler::orl(Address dst, int32_t imm32) {
2233 InstructionMark im(this);
2234 prefix(dst);
2235 emit_arith_operand(0x81, rcx, dst, imm32);
2236 }
2238 void Assembler::orl(Register dst, int32_t imm32) {
2239 prefix(dst);
2240 emit_arith(0x81, 0xC8, dst, imm32);
2241 }
2243 void Assembler::orl(Register dst, Address src) {
2244 InstructionMark im(this);
2245 prefix(src, dst);
2246 emit_int8(0x0B);
2247 emit_operand(dst, src);
2248 }
2250 void Assembler::orl(Register dst, Register src) {
2251 (void) prefix_and_encode(dst->encoding(), src->encoding());
2252 emit_arith(0x0B, 0xC0, dst, src);
2253 }
2255 void Assembler::packuswb(XMMRegister dst, Address src) {
2256 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2257 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2258 emit_simd_arith(0x67, dst, src, VEX_SIMD_66);
2259 }
2261 void Assembler::packuswb(XMMRegister dst, XMMRegister src) {
2262 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2263 emit_simd_arith(0x67, dst, src, VEX_SIMD_66);
2264 }
2266 void Assembler::pcmpestri(XMMRegister dst, Address src, int imm8) {
2267 assert(VM_Version::supports_sse4_2(), "");
2268 InstructionMark im(this);
2269 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_3A);
2270 emit_int8(0x61);
2271 emit_operand(dst, src);
2272 emit_int8(imm8);
2273 }
2275 void Assembler::pcmpestri(XMMRegister dst, XMMRegister src, int imm8) {
2276 assert(VM_Version::supports_sse4_2(), "");
2277 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_3A);
2278 emit_int8(0x61);
2279 emit_int8((unsigned char)(0xC0 | encode));
2280 emit_int8(imm8);
2281 }
2283 void Assembler::pmovzxbw(XMMRegister dst, Address src) {
2284 assert(VM_Version::supports_sse4_1(), "");
2285 InstructionMark im(this);
2286 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2287 emit_int8(0x30);
2288 emit_operand(dst, src);
2289 }
2291 void Assembler::pmovzxbw(XMMRegister dst, XMMRegister src) {
2292 assert(VM_Version::supports_sse4_1(), "");
2293 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2294 emit_int8(0x30);
2295 emit_int8((unsigned char)(0xC0 | encode));
2296 }
2298 // generic
2299 void Assembler::pop(Register dst) {
2300 int encode = prefix_and_encode(dst->encoding());
2301 emit_int8(0x58 | encode);
2302 }
2304 void Assembler::popcntl(Register dst, Address src) {
2305 assert(VM_Version::supports_popcnt(), "must support");
2306 InstructionMark im(this);
2307 emit_int8((unsigned char)0xF3);
2308 prefix(src, dst);
2309 emit_int8(0x0F);
2310 emit_int8((unsigned char)0xB8);
2311 emit_operand(dst, src);
2312 }
2314 void Assembler::popcntl(Register dst, Register src) {
2315 assert(VM_Version::supports_popcnt(), "must support");
2316 emit_int8((unsigned char)0xF3);
2317 int encode = prefix_and_encode(dst->encoding(), src->encoding());
2318 emit_int8(0x0F);
2319 emit_int8((unsigned char)0xB8);
2320 emit_int8((unsigned char)(0xC0 | encode));
2321 }
2323 void Assembler::popf() {
2324 emit_int8((unsigned char)0x9D);
2325 }
2327 #ifndef _LP64 // no 32bit push/pop on amd64
2328 void Assembler::popl(Address dst) {
2329 // NOTE: this will adjust stack by 8byte on 64bits
2330 InstructionMark im(this);
2331 prefix(dst);
2332 emit_int8((unsigned char)0x8F);
2333 emit_operand(rax, dst);
2334 }
2335 #endif
2337 void Assembler::prefetch_prefix(Address src) {
2338 prefix(src);
2339 emit_int8(0x0F);
2340 }
2342 void Assembler::prefetchnta(Address src) {
2343 NOT_LP64(assert(VM_Version::supports_sse(), "must support"));
2344 InstructionMark im(this);
2345 prefetch_prefix(src);
2346 emit_int8(0x18);
2347 emit_operand(rax, src); // 0, src
2348 }
2350 void Assembler::prefetchr(Address src) {
2351 assert(VM_Version::supports_3dnow_prefetch(), "must support");
2352 InstructionMark im(this);
2353 prefetch_prefix(src);
2354 emit_int8(0x0D);
2355 emit_operand(rax, src); // 0, src
2356 }
2358 void Assembler::prefetcht0(Address src) {
2359 NOT_LP64(assert(VM_Version::supports_sse(), "must support"));
2360 InstructionMark im(this);
2361 prefetch_prefix(src);
2362 emit_int8(0x18);
2363 emit_operand(rcx, src); // 1, src
2364 }
2366 void Assembler::prefetcht1(Address src) {
2367 NOT_LP64(assert(VM_Version::supports_sse(), "must support"));
2368 InstructionMark im(this);
2369 prefetch_prefix(src);
2370 emit_int8(0x18);
2371 emit_operand(rdx, src); // 2, src
2372 }
2374 void Assembler::prefetcht2(Address src) {
2375 NOT_LP64(assert(VM_Version::supports_sse(), "must support"));
2376 InstructionMark im(this);
2377 prefetch_prefix(src);
2378 emit_int8(0x18);
2379 emit_operand(rbx, src); // 3, src
2380 }
2382 void Assembler::prefetchw(Address src) {
2383 assert(VM_Version::supports_3dnow_prefetch(), "must support");
2384 InstructionMark im(this);
2385 prefetch_prefix(src);
2386 emit_int8(0x0D);
2387 emit_operand(rcx, src); // 1, src
2388 }
2390 void Assembler::prefix(Prefix p) {
2391 emit_int8(p);
2392 }
2394 void Assembler::pshufb(XMMRegister dst, XMMRegister src) {
2395 assert(VM_Version::supports_ssse3(), "");
2396 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2397 emit_int8(0x00);
2398 emit_int8((unsigned char)(0xC0 | encode));
2399 }
2401 void Assembler::pshufb(XMMRegister dst, Address src) {
2402 assert(VM_Version::supports_ssse3(), "");
2403 InstructionMark im(this);
2404 simd_prefix(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2405 emit_int8(0x00);
2406 emit_operand(dst, src);
2407 }
2409 void Assembler::pshufd(XMMRegister dst, XMMRegister src, int mode) {
2410 assert(isByte(mode), "invalid value");
2411 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2412 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_66);
2413 emit_int8(mode & 0xFF);
2415 }
2417 void Assembler::pshufd(XMMRegister dst, Address src, int mode) {
2418 assert(isByte(mode), "invalid value");
2419 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2420 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2421 InstructionMark im(this);
2422 simd_prefix(dst, src, VEX_SIMD_66);
2423 emit_int8(0x70);
2424 emit_operand(dst, src);
2425 emit_int8(mode & 0xFF);
2426 }
2428 void Assembler::pshuflw(XMMRegister dst, XMMRegister src, int mode) {
2429 assert(isByte(mode), "invalid value");
2430 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2431 emit_simd_arith_nonds(0x70, dst, src, VEX_SIMD_F2);
2432 emit_int8(mode & 0xFF);
2433 }
2435 void Assembler::pshuflw(XMMRegister dst, Address src, int mode) {
2436 assert(isByte(mode), "invalid value");
2437 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2438 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2439 InstructionMark im(this);
2440 simd_prefix(dst, src, VEX_SIMD_F2);
2441 emit_int8(0x70);
2442 emit_operand(dst, src);
2443 emit_int8(mode & 0xFF);
2444 }
2446 void Assembler::psrldq(XMMRegister dst, int shift) {
2447 // Shift 128 bit value in xmm register by number of bytes.
2448 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2449 int encode = simd_prefix_and_encode(xmm3, dst, dst, VEX_SIMD_66);
2450 emit_int8(0x73);
2451 emit_int8((unsigned char)(0xC0 | encode));
2452 emit_int8(shift);
2453 }
2455 void Assembler::ptest(XMMRegister dst, Address src) {
2456 assert(VM_Version::supports_sse4_1(), "");
2457 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2458 InstructionMark im(this);
2459 simd_prefix(dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2460 emit_int8(0x17);
2461 emit_operand(dst, src);
2462 }
2464 void Assembler::ptest(XMMRegister dst, XMMRegister src) {
2465 assert(VM_Version::supports_sse4_1(), "");
2466 int encode = simd_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
2467 emit_int8(0x17);
2468 emit_int8((unsigned char)(0xC0 | encode));
2469 }
2471 void Assembler::punpcklbw(XMMRegister dst, Address src) {
2472 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2473 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2474 emit_simd_arith(0x60, dst, src, VEX_SIMD_66);
2475 }
2477 void Assembler::punpcklbw(XMMRegister dst, XMMRegister src) {
2478 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2479 emit_simd_arith(0x60, dst, src, VEX_SIMD_66);
2480 }
2482 void Assembler::punpckldq(XMMRegister dst, Address src) {
2483 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2484 assert((UseAVX > 0), "SSE mode requires address alignment 16 bytes");
2485 emit_simd_arith(0x62, dst, src, VEX_SIMD_66);
2486 }
2488 void Assembler::punpckldq(XMMRegister dst, XMMRegister src) {
2489 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2490 emit_simd_arith(0x62, dst, src, VEX_SIMD_66);
2491 }
2493 void Assembler::punpcklqdq(XMMRegister dst, XMMRegister src) {
2494 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2495 emit_simd_arith(0x6C, dst, src, VEX_SIMD_66);
2496 }
2498 void Assembler::push(int32_t imm32) {
2499 // in 64bits we push 64bits onto the stack but only
2500 // take a 32bit immediate
2501 emit_int8(0x68);
2502 emit_long(imm32);
2503 }
2505 void Assembler::push(Register src) {
2506 int encode = prefix_and_encode(src->encoding());
2508 emit_int8(0x50 | encode);
2509 }
2511 void Assembler::pushf() {
2512 emit_int8((unsigned char)0x9C);
2513 }
2515 #ifndef _LP64 // no 32bit push/pop on amd64
2516 void Assembler::pushl(Address src) {
2517 // Note this will push 64bit on 64bit
2518 InstructionMark im(this);
2519 prefix(src);
2520 emit_int8((unsigned char)0xFF);
2521 emit_operand(rsi, src);
2522 }
2523 #endif
2525 void Assembler::rcll(Register dst, int imm8) {
2526 assert(isShiftCount(imm8), "illegal shift count");
2527 int encode = prefix_and_encode(dst->encoding());
2528 if (imm8 == 1) {
2529 emit_int8((unsigned char)0xD1);
2530 emit_int8((unsigned char)(0xD0 | encode));
2531 } else {
2532 emit_int8((unsigned char)0xC1);
2533 emit_int8((unsigned char)0xD0 | encode);
2534 emit_int8(imm8);
2535 }
2536 }
2538 // copies data from [esi] to [edi] using rcx pointer sized words
2539 // generic
2540 void Assembler::rep_mov() {
2541 emit_int8((unsigned char)0xF3);
2542 // MOVSQ
2543 LP64_ONLY(prefix(REX_W));
2544 emit_int8((unsigned char)0xA5);
2545 }
2547 // sets rcx bytes with rax, value at [edi]
2548 void Assembler::rep_stosb() {
2549 emit_int8((unsigned char)0xF3); // REP
2550 LP64_ONLY(prefix(REX_W));
2551 emit_int8((unsigned char)0xAA); // STOSB
2552 }
2554 // sets rcx pointer sized words with rax, value at [edi]
2555 // generic
2556 void Assembler::rep_stos() {
2557 emit_int8((unsigned char)0xF3); // REP
2558 LP64_ONLY(prefix(REX_W)); // LP64:STOSQ, LP32:STOSD
2559 emit_int8((unsigned char)0xAB);
2560 }
2562 // scans rcx pointer sized words at [edi] for occurance of rax,
2563 // generic
2564 void Assembler::repne_scan() { // repne_scan
2565 emit_int8((unsigned char)0xF2);
2566 // SCASQ
2567 LP64_ONLY(prefix(REX_W));
2568 emit_int8((unsigned char)0xAF);
2569 }
2571 #ifdef _LP64
2572 // scans rcx 4 byte words at [edi] for occurance of rax,
2573 // generic
2574 void Assembler::repne_scanl() { // repne_scan
2575 emit_int8((unsigned char)0xF2);
2576 // SCASL
2577 emit_int8((unsigned char)0xAF);
2578 }
2579 #endif
2581 void Assembler::ret(int imm16) {
2582 if (imm16 == 0) {
2583 emit_int8((unsigned char)0xC3);
2584 } else {
2585 emit_int8((unsigned char)0xC2);
2586 emit_int16(imm16);
2587 }
2588 }
2590 void Assembler::sahf() {
2591 #ifdef _LP64
2592 // Not supported in 64bit mode
2593 ShouldNotReachHere();
2594 #endif
2595 emit_int8((unsigned char)0x9E);
2596 }
2598 void Assembler::sarl(Register dst, int imm8) {
2599 int encode = prefix_and_encode(dst->encoding());
2600 assert(isShiftCount(imm8), "illegal shift count");
2601 if (imm8 == 1) {
2602 emit_int8((unsigned char)0xD1);
2603 emit_int8((unsigned char)(0xF8 | encode));
2604 } else {
2605 emit_int8((unsigned char)0xC1);
2606 emit_int8((unsigned char)(0xF8 | encode));
2607 emit_int8(imm8);
2608 }
2609 }
2611 void Assembler::sarl(Register dst) {
2612 int encode = prefix_and_encode(dst->encoding());
2613 emit_int8((unsigned char)0xD3);
2614 emit_int8((unsigned char)(0xF8 | encode));
2615 }
2617 void Assembler::sbbl(Address dst, int32_t imm32) {
2618 InstructionMark im(this);
2619 prefix(dst);
2620 emit_arith_operand(0x81, rbx, dst, imm32);
2621 }
2623 void Assembler::sbbl(Register dst, int32_t imm32) {
2624 prefix(dst);
2625 emit_arith(0x81, 0xD8, dst, imm32);
2626 }
2629 void Assembler::sbbl(Register dst, Address src) {
2630 InstructionMark im(this);
2631 prefix(src, dst);
2632 emit_int8(0x1B);
2633 emit_operand(dst, src);
2634 }
2636 void Assembler::sbbl(Register dst, Register src) {
2637 (void) prefix_and_encode(dst->encoding(), src->encoding());
2638 emit_arith(0x1B, 0xC0, dst, src);
2639 }
2641 void Assembler::setb(Condition cc, Register dst) {
2642 assert(0 <= cc && cc < 16, "illegal cc");
2643 int encode = prefix_and_encode(dst->encoding(), true);
2644 emit_int8(0x0F);
2645 emit_int8((unsigned char)0x90 | cc);
2646 emit_int8((unsigned char)(0xC0 | encode));
2647 }
2649 void Assembler::shll(Register dst, int imm8) {
2650 assert(isShiftCount(imm8), "illegal shift count");
2651 int encode = prefix_and_encode(dst->encoding());
2652 if (imm8 == 1 ) {
2653 emit_int8((unsigned char)0xD1);
2654 emit_int8((unsigned char)(0xE0 | encode));
2655 } else {
2656 emit_int8((unsigned char)0xC1);
2657 emit_int8((unsigned char)(0xE0 | encode));
2658 emit_int8(imm8);
2659 }
2660 }
2662 void Assembler::shll(Register dst) {
2663 int encode = prefix_and_encode(dst->encoding());
2664 emit_int8((unsigned char)0xD3);
2665 emit_int8((unsigned char)(0xE0 | encode));
2666 }
2668 void Assembler::shrl(Register dst, int imm8) {
2669 assert(isShiftCount(imm8), "illegal shift count");
2670 int encode = prefix_and_encode(dst->encoding());
2671 emit_int8((unsigned char)0xC1);
2672 emit_int8((unsigned char)(0xE8 | encode));
2673 emit_int8(imm8);
2674 }
2676 void Assembler::shrl(Register dst) {
2677 int encode = prefix_and_encode(dst->encoding());
2678 emit_int8((unsigned char)0xD3);
2679 emit_int8((unsigned char)(0xE8 | encode));
2680 }
2682 // copies a single word from [esi] to [edi]
2683 void Assembler::smovl() {
2684 emit_int8((unsigned char)0xA5);
2685 }
2687 void Assembler::sqrtsd(XMMRegister dst, XMMRegister src) {
2688 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2689 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2);
2690 }
2692 void Assembler::sqrtsd(XMMRegister dst, Address src) {
2693 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2694 emit_simd_arith(0x51, dst, src, VEX_SIMD_F2);
2695 }
2697 void Assembler::sqrtss(XMMRegister dst, XMMRegister src) {
2698 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2699 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3);
2700 }
2702 void Assembler::std() {
2703 emit_int8((unsigned char)0xFD);
2704 }
2706 void Assembler::sqrtss(XMMRegister dst, Address src) {
2707 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2708 emit_simd_arith(0x51, dst, src, VEX_SIMD_F3);
2709 }
2711 void Assembler::stmxcsr( Address dst) {
2712 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2713 InstructionMark im(this);
2714 prefix(dst);
2715 emit_int8(0x0F);
2716 emit_int8((unsigned char)0xAE);
2717 emit_operand(as_Register(3), dst);
2718 }
2720 void Assembler::subl(Address dst, int32_t imm32) {
2721 InstructionMark im(this);
2722 prefix(dst);
2723 emit_arith_operand(0x81, rbp, dst, imm32);
2724 }
2726 void Assembler::subl(Address dst, Register src) {
2727 InstructionMark im(this);
2728 prefix(dst, src);
2729 emit_int8(0x29);
2730 emit_operand(src, dst);
2731 }
2733 void Assembler::subl(Register dst, int32_t imm32) {
2734 prefix(dst);
2735 emit_arith(0x81, 0xE8, dst, imm32);
2736 }
2738 // Force generation of a 4 byte immediate value even if it fits into 8bit
2739 void Assembler::subl_imm32(Register dst, int32_t imm32) {
2740 prefix(dst);
2741 emit_arith_imm32(0x81, 0xE8, dst, imm32);
2742 }
2744 void Assembler::subl(Register dst, Address src) {
2745 InstructionMark im(this);
2746 prefix(src, dst);
2747 emit_int8(0x2B);
2748 emit_operand(dst, src);
2749 }
2751 void Assembler::subl(Register dst, Register src) {
2752 (void) prefix_and_encode(dst->encoding(), src->encoding());
2753 emit_arith(0x2B, 0xC0, dst, src);
2754 }
2756 void Assembler::subsd(XMMRegister dst, XMMRegister src) {
2757 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2758 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2);
2759 }
2761 void Assembler::subsd(XMMRegister dst, Address src) {
2762 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2763 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F2);
2764 }
2766 void Assembler::subss(XMMRegister dst, XMMRegister src) {
2767 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2768 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3);
2769 }
2771 void Assembler::subss(XMMRegister dst, Address src) {
2772 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2773 emit_simd_arith(0x5C, dst, src, VEX_SIMD_F3);
2774 }
2776 void Assembler::testb(Register dst, int imm8) {
2777 NOT_LP64(assert(dst->has_byte_register(), "must have byte register"));
2778 (void) prefix_and_encode(dst->encoding(), true);
2779 emit_arith_b(0xF6, 0xC0, dst, imm8);
2780 }
2782 void Assembler::testl(Register dst, int32_t imm32) {
2783 // not using emit_arith because test
2784 // doesn't support sign-extension of
2785 // 8bit operands
2786 int encode = dst->encoding();
2787 if (encode == 0) {
2788 emit_int8((unsigned char)0xA9);
2789 } else {
2790 encode = prefix_and_encode(encode);
2791 emit_int8((unsigned char)0xF7);
2792 emit_int8((unsigned char)(0xC0 | encode));
2793 }
2794 emit_long(imm32);
2795 }
2797 void Assembler::testl(Register dst, Register src) {
2798 (void) prefix_and_encode(dst->encoding(), src->encoding());
2799 emit_arith(0x85, 0xC0, dst, src);
2800 }
2802 void Assembler::testl(Register dst, Address src) {
2803 InstructionMark im(this);
2804 prefix(src, dst);
2805 emit_int8((unsigned char)0x85);
2806 emit_operand(dst, src);
2807 }
2809 void Assembler::ucomisd(XMMRegister dst, Address src) {
2810 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2811 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66);
2812 }
2814 void Assembler::ucomisd(XMMRegister dst, XMMRegister src) {
2815 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2816 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_66);
2817 }
2819 void Assembler::ucomiss(XMMRegister dst, Address src) {
2820 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2821 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE);
2822 }
2824 void Assembler::ucomiss(XMMRegister dst, XMMRegister src) {
2825 NOT_LP64(assert(VM_Version::supports_sse(), ""));
2826 emit_simd_arith_nonds(0x2E, dst, src, VEX_SIMD_NONE);
2827 }
2830 void Assembler::xaddl(Address dst, Register src) {
2831 InstructionMark im(this);
2832 prefix(dst, src);
2833 emit_int8(0x0F);
2834 emit_int8((unsigned char)0xC1);
2835 emit_operand(src, dst);
2836 }
2838 void Assembler::xchgl(Register dst, Address src) { // xchg
2839 InstructionMark im(this);
2840 prefix(src, dst);
2841 emit_int8((unsigned char)0x87);
2842 emit_operand(dst, src);
2843 }
2845 void Assembler::xchgl(Register dst, Register src) {
2846 int encode = prefix_and_encode(dst->encoding(), src->encoding());
2847 emit_int8((unsigned char)0x87);
2848 emit_int8((unsigned char)(0xC0 | encode));
2849 }
2851 void Assembler::xgetbv() {
2852 emit_int8(0x0F);
2853 emit_int8(0x01);
2854 emit_int8((unsigned char)0xD0);
2855 }
2857 void Assembler::xorl(Register dst, int32_t imm32) {
2858 prefix(dst);
2859 emit_arith(0x81, 0xF0, dst, imm32);
2860 }
2862 void Assembler::xorl(Register dst, Address src) {
2863 InstructionMark im(this);
2864 prefix(src, dst);
2865 emit_int8(0x33);
2866 emit_operand(dst, src);
2867 }
2869 void Assembler::xorl(Register dst, Register src) {
2870 (void) prefix_and_encode(dst->encoding(), src->encoding());
2871 emit_arith(0x33, 0xC0, dst, src);
2872 }
2875 // AVX 3-operands scalar float-point arithmetic instructions
2877 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, Address src) {
2878 assert(VM_Version::supports_avx(), "");
2879 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2880 }
2882 void Assembler::vaddsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2883 assert(VM_Version::supports_avx(), "");
2884 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2885 }
2887 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, Address src) {
2888 assert(VM_Version::supports_avx(), "");
2889 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2890 }
2892 void Assembler::vaddss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2893 assert(VM_Version::supports_avx(), "");
2894 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2895 }
2897 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, Address src) {
2898 assert(VM_Version::supports_avx(), "");
2899 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2900 }
2902 void Assembler::vdivsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2903 assert(VM_Version::supports_avx(), "");
2904 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2905 }
2907 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, Address src) {
2908 assert(VM_Version::supports_avx(), "");
2909 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2910 }
2912 void Assembler::vdivss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2913 assert(VM_Version::supports_avx(), "");
2914 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2915 }
2917 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, Address src) {
2918 assert(VM_Version::supports_avx(), "");
2919 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2920 }
2922 void Assembler::vmulsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2923 assert(VM_Version::supports_avx(), "");
2924 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2925 }
2927 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, Address src) {
2928 assert(VM_Version::supports_avx(), "");
2929 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2930 }
2932 void Assembler::vmulss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2933 assert(VM_Version::supports_avx(), "");
2934 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2935 }
2937 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, Address src) {
2938 assert(VM_Version::supports_avx(), "");
2939 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2940 }
2942 void Assembler::vsubsd(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2943 assert(VM_Version::supports_avx(), "");
2944 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F2, /* vector256 */ false);
2945 }
2947 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, Address src) {
2948 assert(VM_Version::supports_avx(), "");
2949 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2950 }
2952 void Assembler::vsubss(XMMRegister dst, XMMRegister nds, XMMRegister src) {
2953 assert(VM_Version::supports_avx(), "");
2954 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_F3, /* vector256 */ false);
2955 }
2957 //====================VECTOR ARITHMETIC=====================================
2959 // Float-point vector arithmetic
2961 void Assembler::addpd(XMMRegister dst, XMMRegister src) {
2962 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2963 emit_simd_arith(0x58, dst, src, VEX_SIMD_66);
2964 }
2966 void Assembler::addps(XMMRegister dst, XMMRegister src) {
2967 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2968 emit_simd_arith(0x58, dst, src, VEX_SIMD_NONE);
2969 }
2971 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
2972 assert(VM_Version::supports_avx(), "");
2973 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256);
2974 }
2976 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
2977 assert(VM_Version::supports_avx(), "");
2978 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256);
2979 }
2981 void Assembler::vaddpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
2982 assert(VM_Version::supports_avx(), "");
2983 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_66, vector256);
2984 }
2986 void Assembler::vaddps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
2987 assert(VM_Version::supports_avx(), "");
2988 emit_vex_arith(0x58, dst, nds, src, VEX_SIMD_NONE, vector256);
2989 }
2991 void Assembler::subpd(XMMRegister dst, XMMRegister src) {
2992 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2993 emit_simd_arith(0x5C, dst, src, VEX_SIMD_66);
2994 }
2996 void Assembler::subps(XMMRegister dst, XMMRegister src) {
2997 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
2998 emit_simd_arith(0x5C, dst, src, VEX_SIMD_NONE);
2999 }
3001 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3002 assert(VM_Version::supports_avx(), "");
3003 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256);
3004 }
3006 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3007 assert(VM_Version::supports_avx(), "");
3008 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256);
3009 }
3011 void Assembler::vsubpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3012 assert(VM_Version::supports_avx(), "");
3013 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_66, vector256);
3014 }
3016 void Assembler::vsubps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3017 assert(VM_Version::supports_avx(), "");
3018 emit_vex_arith(0x5C, dst, nds, src, VEX_SIMD_NONE, vector256);
3019 }
3021 void Assembler::mulpd(XMMRegister dst, XMMRegister src) {
3022 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3023 emit_simd_arith(0x59, dst, src, VEX_SIMD_66);
3024 }
3026 void Assembler::mulps(XMMRegister dst, XMMRegister src) {
3027 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3028 emit_simd_arith(0x59, dst, src, VEX_SIMD_NONE);
3029 }
3031 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3032 assert(VM_Version::supports_avx(), "");
3033 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256);
3034 }
3036 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3037 assert(VM_Version::supports_avx(), "");
3038 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256);
3039 }
3041 void Assembler::vmulpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3042 assert(VM_Version::supports_avx(), "");
3043 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_66, vector256);
3044 }
3046 void Assembler::vmulps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3047 assert(VM_Version::supports_avx(), "");
3048 emit_vex_arith(0x59, dst, nds, src, VEX_SIMD_NONE, vector256);
3049 }
3051 void Assembler::divpd(XMMRegister dst, XMMRegister src) {
3052 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3053 emit_simd_arith(0x5E, dst, src, VEX_SIMD_66);
3054 }
3056 void Assembler::divps(XMMRegister dst, XMMRegister src) {
3057 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3058 emit_simd_arith(0x5E, dst, src, VEX_SIMD_NONE);
3059 }
3061 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3062 assert(VM_Version::supports_avx(), "");
3063 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256);
3064 }
3066 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3067 assert(VM_Version::supports_avx(), "");
3068 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256);
3069 }
3071 void Assembler::vdivpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3072 assert(VM_Version::supports_avx(), "");
3073 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_66, vector256);
3074 }
3076 void Assembler::vdivps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3077 assert(VM_Version::supports_avx(), "");
3078 emit_vex_arith(0x5E, dst, nds, src, VEX_SIMD_NONE, vector256);
3079 }
3081 void Assembler::andpd(XMMRegister dst, XMMRegister src) {
3082 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3083 emit_simd_arith(0x54, dst, src, VEX_SIMD_66);
3084 }
3086 void Assembler::andps(XMMRegister dst, XMMRegister src) {
3087 NOT_LP64(assert(VM_Version::supports_sse(), ""));
3088 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE);
3089 }
3091 void Assembler::andps(XMMRegister dst, Address src) {
3092 NOT_LP64(assert(VM_Version::supports_sse(), ""));
3093 emit_simd_arith(0x54, dst, src, VEX_SIMD_NONE);
3094 }
3096 void Assembler::andpd(XMMRegister dst, Address src) {
3097 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3098 emit_simd_arith(0x54, dst, src, VEX_SIMD_66);
3099 }
3101 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3102 assert(VM_Version::supports_avx(), "");
3103 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256);
3104 }
3106 void Assembler::vandps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3107 assert(VM_Version::supports_avx(), "");
3108 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256);
3109 }
3111 void Assembler::vandpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3112 assert(VM_Version::supports_avx(), "");
3113 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_66, vector256);
3114 }
3116 void Assembler::vandps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3117 assert(VM_Version::supports_avx(), "");
3118 emit_vex_arith(0x54, dst, nds, src, VEX_SIMD_NONE, vector256);
3119 }
3121 void Assembler::xorpd(XMMRegister dst, XMMRegister src) {
3122 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3123 emit_simd_arith(0x57, dst, src, VEX_SIMD_66);
3124 }
3126 void Assembler::xorps(XMMRegister dst, XMMRegister src) {
3127 NOT_LP64(assert(VM_Version::supports_sse(), ""));
3128 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE);
3129 }
3131 void Assembler::xorpd(XMMRegister dst, Address src) {
3132 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3133 emit_simd_arith(0x57, dst, src, VEX_SIMD_66);
3134 }
3136 void Assembler::xorps(XMMRegister dst, Address src) {
3137 NOT_LP64(assert(VM_Version::supports_sse(), ""));
3138 emit_simd_arith(0x57, dst, src, VEX_SIMD_NONE);
3139 }
3141 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3142 assert(VM_Version::supports_avx(), "");
3143 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256);
3144 }
3146 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3147 assert(VM_Version::supports_avx(), "");
3148 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256);
3149 }
3151 void Assembler::vxorpd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3152 assert(VM_Version::supports_avx(), "");
3153 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_66, vector256);
3154 }
3156 void Assembler::vxorps(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3157 assert(VM_Version::supports_avx(), "");
3158 emit_vex_arith(0x57, dst, nds, src, VEX_SIMD_NONE, vector256);
3159 }
3162 // Integer vector arithmetic
3163 void Assembler::paddb(XMMRegister dst, XMMRegister src) {
3164 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3165 emit_simd_arith(0xFC, dst, src, VEX_SIMD_66);
3166 }
3168 void Assembler::paddw(XMMRegister dst, XMMRegister src) {
3169 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3170 emit_simd_arith(0xFD, dst, src, VEX_SIMD_66);
3171 }
3173 void Assembler::paddd(XMMRegister dst, XMMRegister src) {
3174 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3175 emit_simd_arith(0xFE, dst, src, VEX_SIMD_66);
3176 }
3178 void Assembler::paddq(XMMRegister dst, XMMRegister src) {
3179 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3180 emit_simd_arith(0xD4, dst, src, VEX_SIMD_66);
3181 }
3183 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3184 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3185 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256);
3186 }
3188 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3189 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3190 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256);
3191 }
3193 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3194 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3195 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256);
3196 }
3198 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3199 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3200 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256);
3201 }
3203 void Assembler::vpaddb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3204 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3205 emit_vex_arith(0xFC, dst, nds, src, VEX_SIMD_66, vector256);
3206 }
3208 void Assembler::vpaddw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3209 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3210 emit_vex_arith(0xFD, dst, nds, src, VEX_SIMD_66, vector256);
3211 }
3213 void Assembler::vpaddd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3214 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3215 emit_vex_arith(0xFE, dst, nds, src, VEX_SIMD_66, vector256);
3216 }
3218 void Assembler::vpaddq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3219 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3220 emit_vex_arith(0xD4, dst, nds, src, VEX_SIMD_66, vector256);
3221 }
3223 void Assembler::psubb(XMMRegister dst, XMMRegister src) {
3224 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3225 emit_simd_arith(0xF8, dst, src, VEX_SIMD_66);
3226 }
3228 void Assembler::psubw(XMMRegister dst, XMMRegister src) {
3229 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3230 emit_simd_arith(0xF9, dst, src, VEX_SIMD_66);
3231 }
3233 void Assembler::psubd(XMMRegister dst, XMMRegister src) {
3234 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3235 emit_simd_arith(0xFA, dst, src, VEX_SIMD_66);
3236 }
3238 void Assembler::psubq(XMMRegister dst, XMMRegister src) {
3239 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3240 emit_simd_arith(0xFB, dst, src, VEX_SIMD_66);
3241 }
3243 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3244 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3245 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256);
3246 }
3248 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3249 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3250 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256);
3251 }
3253 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3254 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3255 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256);
3256 }
3258 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3259 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3260 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256);
3261 }
3263 void Assembler::vpsubb(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3264 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3265 emit_vex_arith(0xF8, dst, nds, src, VEX_SIMD_66, vector256);
3266 }
3268 void Assembler::vpsubw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3269 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3270 emit_vex_arith(0xF9, dst, nds, src, VEX_SIMD_66, vector256);
3271 }
3273 void Assembler::vpsubd(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3274 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3275 emit_vex_arith(0xFA, dst, nds, src, VEX_SIMD_66, vector256);
3276 }
3278 void Assembler::vpsubq(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3279 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3280 emit_vex_arith(0xFB, dst, nds, src, VEX_SIMD_66, vector256);
3281 }
3283 void Assembler::pmullw(XMMRegister dst, XMMRegister src) {
3284 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3285 emit_simd_arith(0xD5, dst, src, VEX_SIMD_66);
3286 }
3288 void Assembler::pmulld(XMMRegister dst, XMMRegister src) {
3289 assert(VM_Version::supports_sse4_1(), "");
3290 int encode = simd_prefix_and_encode(dst, dst, src, VEX_SIMD_66, VEX_OPCODE_0F_38);
3291 emit_int8(0x40);
3292 emit_int8((unsigned char)(0xC0 | encode));
3293 }
3295 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3296 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3297 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256);
3298 }
3300 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3301 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3302 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38);
3303 emit_int8(0x40);
3304 emit_int8((unsigned char)(0xC0 | encode));
3305 }
3307 void Assembler::vpmullw(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3308 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3309 emit_vex_arith(0xD5, dst, nds, src, VEX_SIMD_66, vector256);
3310 }
3312 void Assembler::vpmulld(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3313 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3314 InstructionMark im(this);
3315 int dst_enc = dst->encoding();
3316 int nds_enc = nds->is_valid() ? nds->encoding() : 0;
3317 vex_prefix(src, nds_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_38, false, vector256);
3318 emit_int8(0x40);
3319 emit_operand(dst, src);
3320 }
3322 // Shift packed integers left by specified number of bits.
3323 void Assembler::psllw(XMMRegister dst, int shift) {
3324 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3325 // XMM6 is for /6 encoding: 66 0F 71 /6 ib
3326 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66);
3327 emit_int8(0x71);
3328 emit_int8((unsigned char)(0xC0 | encode));
3329 emit_int8(shift & 0xFF);
3330 }
3332 void Assembler::pslld(XMMRegister dst, int shift) {
3333 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3334 // XMM6 is for /6 encoding: 66 0F 72 /6 ib
3335 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66);
3336 emit_int8(0x72);
3337 emit_int8((unsigned char)(0xC0 | encode));
3338 emit_int8(shift & 0xFF);
3339 }
3341 void Assembler::psllq(XMMRegister dst, int shift) {
3342 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3343 // XMM6 is for /6 encoding: 66 0F 73 /6 ib
3344 int encode = simd_prefix_and_encode(xmm6, dst, dst, VEX_SIMD_66);
3345 emit_int8(0x73);
3346 emit_int8((unsigned char)(0xC0 | encode));
3347 emit_int8(shift & 0xFF);
3348 }
3350 void Assembler::psllw(XMMRegister dst, XMMRegister shift) {
3351 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3352 emit_simd_arith(0xF1, dst, shift, VEX_SIMD_66);
3353 }
3355 void Assembler::pslld(XMMRegister dst, XMMRegister shift) {
3356 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3357 emit_simd_arith(0xF2, dst, shift, VEX_SIMD_66);
3358 }
3360 void Assembler::psllq(XMMRegister dst, XMMRegister shift) {
3361 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3362 emit_simd_arith(0xF3, dst, shift, VEX_SIMD_66);
3363 }
3365 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3366 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3367 // XMM6 is for /6 encoding: 66 0F 71 /6 ib
3368 emit_vex_arith(0x71, xmm6, dst, src, VEX_SIMD_66, vector256);
3369 emit_int8(shift & 0xFF);
3370 }
3372 void Assembler::vpslld(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3373 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3374 // XMM6 is for /6 encoding: 66 0F 72 /6 ib
3375 emit_vex_arith(0x72, xmm6, dst, src, VEX_SIMD_66, vector256);
3376 emit_int8(shift & 0xFF);
3377 }
3379 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3380 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3381 // XMM6 is for /6 encoding: 66 0F 73 /6 ib
3382 emit_vex_arith(0x73, xmm6, dst, src, VEX_SIMD_66, vector256);
3383 emit_int8(shift & 0xFF);
3384 }
3386 void Assembler::vpsllw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3387 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3388 emit_vex_arith(0xF1, dst, src, shift, VEX_SIMD_66, vector256);
3389 }
3391 void Assembler::vpslld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3392 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3393 emit_vex_arith(0xF2, dst, src, shift, VEX_SIMD_66, vector256);
3394 }
3396 void Assembler::vpsllq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3397 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3398 emit_vex_arith(0xF3, dst, src, shift, VEX_SIMD_66, vector256);
3399 }
3401 // Shift packed integers logically right by specified number of bits.
3402 void Assembler::psrlw(XMMRegister dst, int shift) {
3403 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3404 // XMM2 is for /2 encoding: 66 0F 71 /2 ib
3405 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66);
3406 emit_int8(0x71);
3407 emit_int8((unsigned char)(0xC0 | encode));
3408 emit_int8(shift & 0xFF);
3409 }
3411 void Assembler::psrld(XMMRegister dst, int shift) {
3412 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3413 // XMM2 is for /2 encoding: 66 0F 72 /2 ib
3414 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66);
3415 emit_int8(0x72);
3416 emit_int8((unsigned char)(0xC0 | encode));
3417 emit_int8(shift & 0xFF);
3418 }
3420 void Assembler::psrlq(XMMRegister dst, int shift) {
3421 // Do not confuse it with psrldq SSE2 instruction which
3422 // shifts 128 bit value in xmm register by number of bytes.
3423 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3424 // XMM2 is for /2 encoding: 66 0F 73 /2 ib
3425 int encode = simd_prefix_and_encode(xmm2, dst, dst, VEX_SIMD_66);
3426 emit_int8(0x73);
3427 emit_int8((unsigned char)(0xC0 | encode));
3428 emit_int8(shift & 0xFF);
3429 }
3431 void Assembler::psrlw(XMMRegister dst, XMMRegister shift) {
3432 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3433 emit_simd_arith(0xD1, dst, shift, VEX_SIMD_66);
3434 }
3436 void Assembler::psrld(XMMRegister dst, XMMRegister shift) {
3437 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3438 emit_simd_arith(0xD2, dst, shift, VEX_SIMD_66);
3439 }
3441 void Assembler::psrlq(XMMRegister dst, XMMRegister shift) {
3442 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3443 emit_simd_arith(0xD3, dst, shift, VEX_SIMD_66);
3444 }
3446 void Assembler::vpsrlw(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3447 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3448 // XMM2 is for /2 encoding: 66 0F 73 /2 ib
3449 emit_vex_arith(0x71, xmm2, dst, src, VEX_SIMD_66, vector256);
3450 emit_int8(shift & 0xFF);
3451 }
3453 void Assembler::vpsrld(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3454 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3455 // XMM2 is for /2 encoding: 66 0F 73 /2 ib
3456 emit_vex_arith(0x72, xmm2, dst, src, VEX_SIMD_66, vector256);
3457 emit_int8(shift & 0xFF);
3458 }
3460 void Assembler::vpsrlq(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3461 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3462 // XMM2 is for /2 encoding: 66 0F 73 /2 ib
3463 emit_vex_arith(0x73, xmm2, dst, src, VEX_SIMD_66, vector256);
3464 emit_int8(shift & 0xFF);
3465 }
3467 void Assembler::vpsrlw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3468 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3469 emit_vex_arith(0xD1, dst, src, shift, VEX_SIMD_66, vector256);
3470 }
3472 void Assembler::vpsrld(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3473 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3474 emit_vex_arith(0xD2, dst, src, shift, VEX_SIMD_66, vector256);
3475 }
3477 void Assembler::vpsrlq(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3478 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3479 emit_vex_arith(0xD3, dst, src, shift, VEX_SIMD_66, vector256);
3480 }
3482 // Shift packed integers arithmetically right by specified number of bits.
3483 void Assembler::psraw(XMMRegister dst, int shift) {
3484 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3485 // XMM4 is for /4 encoding: 66 0F 71 /4 ib
3486 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66);
3487 emit_int8(0x71);
3488 emit_int8((unsigned char)(0xC0 | encode));
3489 emit_int8(shift & 0xFF);
3490 }
3492 void Assembler::psrad(XMMRegister dst, int shift) {
3493 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3494 // XMM4 is for /4 encoding: 66 0F 72 /4 ib
3495 int encode = simd_prefix_and_encode(xmm4, dst, dst, VEX_SIMD_66);
3496 emit_int8(0x72);
3497 emit_int8((unsigned char)(0xC0 | encode));
3498 emit_int8(shift & 0xFF);
3499 }
3501 void Assembler::psraw(XMMRegister dst, XMMRegister shift) {
3502 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3503 emit_simd_arith(0xE1, dst, shift, VEX_SIMD_66);
3504 }
3506 void Assembler::psrad(XMMRegister dst, XMMRegister shift) {
3507 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3508 emit_simd_arith(0xE2, dst, shift, VEX_SIMD_66);
3509 }
3511 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3512 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3513 // XMM4 is for /4 encoding: 66 0F 71 /4 ib
3514 emit_vex_arith(0x71, xmm4, dst, src, VEX_SIMD_66, vector256);
3515 emit_int8(shift & 0xFF);
3516 }
3518 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, int shift, bool vector256) {
3519 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3520 // XMM4 is for /4 encoding: 66 0F 71 /4 ib
3521 emit_vex_arith(0x72, xmm4, dst, src, VEX_SIMD_66, vector256);
3522 emit_int8(shift & 0xFF);
3523 }
3525 void Assembler::vpsraw(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3526 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3527 emit_vex_arith(0xE1, dst, src, shift, VEX_SIMD_66, vector256);
3528 }
3530 void Assembler::vpsrad(XMMRegister dst, XMMRegister src, XMMRegister shift, bool vector256) {
3531 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3532 emit_vex_arith(0xE2, dst, src, shift, VEX_SIMD_66, vector256);
3533 }
3536 // AND packed integers
3537 void Assembler::pand(XMMRegister dst, XMMRegister src) {
3538 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3539 emit_simd_arith(0xDB, dst, src, VEX_SIMD_66);
3540 }
3542 void Assembler::vpand(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3543 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3544 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256);
3545 }
3547 void Assembler::vpand(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3548 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3549 emit_vex_arith(0xDB, dst, nds, src, VEX_SIMD_66, vector256);
3550 }
3552 void Assembler::por(XMMRegister dst, XMMRegister src) {
3553 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3554 emit_simd_arith(0xEB, dst, src, VEX_SIMD_66);
3555 }
3557 void Assembler::vpor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3558 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3559 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256);
3560 }
3562 void Assembler::vpor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3563 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3564 emit_vex_arith(0xEB, dst, nds, src, VEX_SIMD_66, vector256);
3565 }
3567 void Assembler::pxor(XMMRegister dst, XMMRegister src) {
3568 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
3569 emit_simd_arith(0xEF, dst, src, VEX_SIMD_66);
3570 }
3572 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, XMMRegister src, bool vector256) {
3573 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3574 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256);
3575 }
3577 void Assembler::vpxor(XMMRegister dst, XMMRegister nds, Address src, bool vector256) {
3578 assert(VM_Version::supports_avx() && !vector256 || VM_Version::supports_avx2(), "256 bit integer vectors requires AVX2");
3579 emit_vex_arith(0xEF, dst, nds, src, VEX_SIMD_66, vector256);
3580 }
3583 void Assembler::vinsertf128h(XMMRegister dst, XMMRegister nds, XMMRegister src) {
3584 assert(VM_Version::supports_avx(), "");
3585 bool vector256 = true;
3586 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A);
3587 emit_int8(0x18);
3588 emit_int8((unsigned char)(0xC0 | encode));
3589 // 0x00 - insert into lower 128 bits
3590 // 0x01 - insert into upper 128 bits
3591 emit_int8(0x01);
3592 }
3594 void Assembler::vinsertf128h(XMMRegister dst, Address src) {
3595 assert(VM_Version::supports_avx(), "");
3596 InstructionMark im(this);
3597 bool vector256 = true;
3598 assert(dst != xnoreg, "sanity");
3599 int dst_enc = dst->encoding();
3600 // swap src<->dst for encoding
3601 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256);
3602 emit_int8(0x18);
3603 emit_operand(dst, src);
3604 // 0x01 - insert into upper 128 bits
3605 emit_int8(0x01);
3606 }
3608 void Assembler::vextractf128h(Address dst, XMMRegister src) {
3609 assert(VM_Version::supports_avx(), "");
3610 InstructionMark im(this);
3611 bool vector256 = true;
3612 assert(src != xnoreg, "sanity");
3613 int src_enc = src->encoding();
3614 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256);
3615 emit_int8(0x19);
3616 emit_operand(src, dst);
3617 // 0x01 - extract from upper 128 bits
3618 emit_int8(0x01);
3619 }
3621 void Assembler::vinserti128h(XMMRegister dst, XMMRegister nds, XMMRegister src) {
3622 assert(VM_Version::supports_avx2(), "");
3623 bool vector256 = true;
3624 int encode = vex_prefix_and_encode(dst, nds, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_3A);
3625 emit_int8(0x38);
3626 emit_int8((unsigned char)(0xC0 | encode));
3627 // 0x00 - insert into lower 128 bits
3628 // 0x01 - insert into upper 128 bits
3629 emit_int8(0x01);
3630 }
3632 void Assembler::vinserti128h(XMMRegister dst, Address src) {
3633 assert(VM_Version::supports_avx2(), "");
3634 InstructionMark im(this);
3635 bool vector256 = true;
3636 assert(dst != xnoreg, "sanity");
3637 int dst_enc = dst->encoding();
3638 // swap src<->dst for encoding
3639 vex_prefix(src, dst_enc, dst_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256);
3640 emit_int8(0x38);
3641 emit_operand(dst, src);
3642 // 0x01 - insert into upper 128 bits
3643 emit_int8(0x01);
3644 }
3646 void Assembler::vextracti128h(Address dst, XMMRegister src) {
3647 assert(VM_Version::supports_avx2(), "");
3648 InstructionMark im(this);
3649 bool vector256 = true;
3650 assert(src != xnoreg, "sanity");
3651 int src_enc = src->encoding();
3652 vex_prefix(dst, 0, src_enc, VEX_SIMD_66, VEX_OPCODE_0F_3A, false, vector256);
3653 emit_int8(0x39);
3654 emit_operand(src, dst);
3655 // 0x01 - extract from upper 128 bits
3656 emit_int8(0x01);
3657 }
3659 // duplicate 4-bytes integer data from src into 8 locations in dest
3660 void Assembler::vpbroadcastd(XMMRegister dst, XMMRegister src) {
3661 assert(VM_Version::supports_avx2(), "");
3662 bool vector256 = true;
3663 int encode = vex_prefix_and_encode(dst, xnoreg, src, VEX_SIMD_66, vector256, VEX_OPCODE_0F_38);
3664 emit_int8(0x58);
3665 emit_int8((unsigned char)(0xC0 | encode));
3666 }
3668 void Assembler::vzeroupper() {
3669 assert(VM_Version::supports_avx(), "");
3670 (void)vex_prefix_and_encode(xmm0, xmm0, xmm0, VEX_SIMD_NONE);
3671 emit_int8(0x77);
3672 }
3675 #ifndef _LP64
3676 // 32bit only pieces of the assembler
3678 void Assembler::cmp_literal32(Register src1, int32_t imm32, RelocationHolder const& rspec) {
3679 // NO PREFIX AS NEVER 64BIT
3680 InstructionMark im(this);
3681 emit_int8((unsigned char)0x81);
3682 emit_int8((unsigned char)(0xF8 | src1->encoding()));
3683 emit_data(imm32, rspec, 0);
3684 }
3686 void Assembler::cmp_literal32(Address src1, int32_t imm32, RelocationHolder const& rspec) {
3687 // NO PREFIX AS NEVER 64BIT (not even 32bit versions of 64bit regs
3688 InstructionMark im(this);
3689 emit_int8((unsigned char)0x81);
3690 emit_operand(rdi, src1);
3691 emit_data(imm32, rspec, 0);
3692 }
3694 // The 64-bit (32bit platform) cmpxchg compares the value at adr with the contents of rdx:rax,
3695 // and stores rcx:rbx into adr if so; otherwise, the value at adr is loaded
3696 // into rdx:rax. The ZF is set if the compared values were equal, and cleared otherwise.
3697 void Assembler::cmpxchg8(Address adr) {
3698 InstructionMark im(this);
3699 emit_int8(0x0F);
3700 emit_int8((unsigned char)0xC7);
3701 emit_operand(rcx, adr);
3702 }
3704 void Assembler::decl(Register dst) {
3705 // Don't use it directly. Use MacroAssembler::decrementl() instead.
3706 emit_int8(0x48 | dst->encoding());
3707 }
3709 #endif // _LP64
3711 // 64bit typically doesn't use the x87 but needs to for the trig funcs
3713 void Assembler::fabs() {
3714 emit_int8((unsigned char)0xD9);
3715 emit_int8((unsigned char)0xE1);
3716 }
3718 void Assembler::fadd(int i) {
3719 emit_farith(0xD8, 0xC0, i);
3720 }
3722 void Assembler::fadd_d(Address src) {
3723 InstructionMark im(this);
3724 emit_int8((unsigned char)0xDC);
3725 emit_operand32(rax, src);
3726 }
3728 void Assembler::fadd_s(Address src) {
3729 InstructionMark im(this);
3730 emit_int8((unsigned char)0xD8);
3731 emit_operand32(rax, src);
3732 }
3734 void Assembler::fadda(int i) {
3735 emit_farith(0xDC, 0xC0, i);
3736 }
3738 void Assembler::faddp(int i) {
3739 emit_farith(0xDE, 0xC0, i);
3740 }
3742 void Assembler::fchs() {
3743 emit_int8((unsigned char)0xD9);
3744 emit_int8((unsigned char)0xE0);
3745 }
3747 void Assembler::fcom(int i) {
3748 emit_farith(0xD8, 0xD0, i);
3749 }
3751 void Assembler::fcomp(int i) {
3752 emit_farith(0xD8, 0xD8, i);
3753 }
3755 void Assembler::fcomp_d(Address src) {
3756 InstructionMark im(this);
3757 emit_int8((unsigned char)0xDC);
3758 emit_operand32(rbx, src);
3759 }
3761 void Assembler::fcomp_s(Address src) {
3762 InstructionMark im(this);
3763 emit_int8((unsigned char)0xD8);
3764 emit_operand32(rbx, src);
3765 }
3767 void Assembler::fcompp() {
3768 emit_int8((unsigned char)0xDE);
3769 emit_int8((unsigned char)0xD9);
3770 }
3772 void Assembler::fcos() {
3773 emit_int8((unsigned char)0xD9);
3774 emit_int8((unsigned char)0xFF);
3775 }
3777 void Assembler::fdecstp() {
3778 emit_int8((unsigned char)0xD9);
3779 emit_int8((unsigned char)0xF6);
3780 }
3782 void Assembler::fdiv(int i) {
3783 emit_farith(0xD8, 0xF0, i);
3784 }
3786 void Assembler::fdiv_d(Address src) {
3787 InstructionMark im(this);
3788 emit_int8((unsigned char)0xDC);
3789 emit_operand32(rsi, src);
3790 }
3792 void Assembler::fdiv_s(Address src) {
3793 InstructionMark im(this);
3794 emit_int8((unsigned char)0xD8);
3795 emit_operand32(rsi, src);
3796 }
3798 void Assembler::fdiva(int i) {
3799 emit_farith(0xDC, 0xF8, i);
3800 }
3802 // Note: The Intel manual (Pentium Processor User's Manual, Vol.3, 1994)
3803 // is erroneous for some of the floating-point instructions below.
3805 void Assembler::fdivp(int i) {
3806 emit_farith(0xDE, 0xF8, i); // ST(0) <- ST(0) / ST(1) and pop (Intel manual wrong)
3807 }
3809 void Assembler::fdivr(int i) {
3810 emit_farith(0xD8, 0xF8, i);
3811 }
3813 void Assembler::fdivr_d(Address src) {
3814 InstructionMark im(this);
3815 emit_int8((unsigned char)0xDC);
3816 emit_operand32(rdi, src);
3817 }
3819 void Assembler::fdivr_s(Address src) {
3820 InstructionMark im(this);
3821 emit_int8((unsigned char)0xD8);
3822 emit_operand32(rdi, src);
3823 }
3825 void Assembler::fdivra(int i) {
3826 emit_farith(0xDC, 0xF0, i);
3827 }
3829 void Assembler::fdivrp(int i) {
3830 emit_farith(0xDE, 0xF0, i); // ST(0) <- ST(1) / ST(0) and pop (Intel manual wrong)
3831 }
3833 void Assembler::ffree(int i) {
3834 emit_farith(0xDD, 0xC0, i);
3835 }
3837 void Assembler::fild_d(Address adr) {
3838 InstructionMark im(this);
3839 emit_int8((unsigned char)0xDF);
3840 emit_operand32(rbp, adr);
3841 }
3843 void Assembler::fild_s(Address adr) {
3844 InstructionMark im(this);
3845 emit_int8((unsigned char)0xDB);
3846 emit_operand32(rax, adr);
3847 }
3849 void Assembler::fincstp() {
3850 emit_int8((unsigned char)0xD9);
3851 emit_int8((unsigned char)0xF7);
3852 }
3854 void Assembler::finit() {
3855 emit_int8((unsigned char)0x9B);
3856 emit_int8((unsigned char)0xDB);
3857 emit_int8((unsigned char)0xE3);
3858 }
3860 void Assembler::fist_s(Address adr) {
3861 InstructionMark im(this);
3862 emit_int8((unsigned char)0xDB);
3863 emit_operand32(rdx, adr);
3864 }
3866 void Assembler::fistp_d(Address adr) {
3867 InstructionMark im(this);
3868 emit_int8((unsigned char)0xDF);
3869 emit_operand32(rdi, adr);
3870 }
3872 void Assembler::fistp_s(Address adr) {
3873 InstructionMark im(this);
3874 emit_int8((unsigned char)0xDB);
3875 emit_operand32(rbx, adr);
3876 }
3878 void Assembler::fld1() {
3879 emit_int8((unsigned char)0xD9);
3880 emit_int8((unsigned char)0xE8);
3881 }
3883 void Assembler::fld_d(Address adr) {
3884 InstructionMark im(this);
3885 emit_int8((unsigned char)0xDD);
3886 emit_operand32(rax, adr);
3887 }
3889 void Assembler::fld_s(Address adr) {
3890 InstructionMark im(this);
3891 emit_int8((unsigned char)0xD9);
3892 emit_operand32(rax, adr);
3893 }
3896 void Assembler::fld_s(int index) {
3897 emit_farith(0xD9, 0xC0, index);
3898 }
3900 void Assembler::fld_x(Address adr) {
3901 InstructionMark im(this);
3902 emit_int8((unsigned char)0xDB);
3903 emit_operand32(rbp, adr);
3904 }
3906 void Assembler::fldcw(Address src) {
3907 InstructionMark im(this);
3908 emit_int8((unsigned char)0xD9);
3909 emit_operand32(rbp, src);
3910 }
3912 void Assembler::fldenv(Address src) {
3913 InstructionMark im(this);
3914 emit_int8((unsigned char)0xD9);
3915 emit_operand32(rsp, src);
3916 }
3918 void Assembler::fldlg2() {
3919 emit_int8((unsigned char)0xD9);
3920 emit_int8((unsigned char)0xEC);
3921 }
3923 void Assembler::fldln2() {
3924 emit_int8((unsigned char)0xD9);
3925 emit_int8((unsigned char)0xED);
3926 }
3928 void Assembler::fldz() {
3929 emit_int8((unsigned char)0xD9);
3930 emit_int8((unsigned char)0xEE);
3931 }
3933 void Assembler::flog() {
3934 fldln2();
3935 fxch();
3936 fyl2x();
3937 }
3939 void Assembler::flog10() {
3940 fldlg2();
3941 fxch();
3942 fyl2x();
3943 }
3945 void Assembler::fmul(int i) {
3946 emit_farith(0xD8, 0xC8, i);
3947 }
3949 void Assembler::fmul_d(Address src) {
3950 InstructionMark im(this);
3951 emit_int8((unsigned char)0xDC);
3952 emit_operand32(rcx, src);
3953 }
3955 void Assembler::fmul_s(Address src) {
3956 InstructionMark im(this);
3957 emit_int8((unsigned char)0xD8);
3958 emit_operand32(rcx, src);
3959 }
3961 void Assembler::fmula(int i) {
3962 emit_farith(0xDC, 0xC8, i);
3963 }
3965 void Assembler::fmulp(int i) {
3966 emit_farith(0xDE, 0xC8, i);
3967 }
3969 void Assembler::fnsave(Address dst) {
3970 InstructionMark im(this);
3971 emit_int8((unsigned char)0xDD);
3972 emit_operand32(rsi, dst);
3973 }
3975 void Assembler::fnstcw(Address src) {
3976 InstructionMark im(this);
3977 emit_int8((unsigned char)0x9B);
3978 emit_int8((unsigned char)0xD9);
3979 emit_operand32(rdi, src);
3980 }
3982 void Assembler::fnstsw_ax() {
3983 emit_int8((unsigned char)0xDF);
3984 emit_int8((unsigned char)0xE0);
3985 }
3987 void Assembler::fprem() {
3988 emit_int8((unsigned char)0xD9);
3989 emit_int8((unsigned char)0xF8);
3990 }
3992 void Assembler::fprem1() {
3993 emit_int8((unsigned char)0xD9);
3994 emit_int8((unsigned char)0xF5);
3995 }
3997 void Assembler::frstor(Address src) {
3998 InstructionMark im(this);
3999 emit_int8((unsigned char)0xDD);
4000 emit_operand32(rsp, src);
4001 }
4003 void Assembler::fsin() {
4004 emit_int8((unsigned char)0xD9);
4005 emit_int8((unsigned char)0xFE);
4006 }
4008 void Assembler::fsqrt() {
4009 emit_int8((unsigned char)0xD9);
4010 emit_int8((unsigned char)0xFA);
4011 }
4013 void Assembler::fst_d(Address adr) {
4014 InstructionMark im(this);
4015 emit_int8((unsigned char)0xDD);
4016 emit_operand32(rdx, adr);
4017 }
4019 void Assembler::fst_s(Address adr) {
4020 InstructionMark im(this);
4021 emit_int8((unsigned char)0xD9);
4022 emit_operand32(rdx, adr);
4023 }
4025 void Assembler::fstp_d(Address adr) {
4026 InstructionMark im(this);
4027 emit_int8((unsigned char)0xDD);
4028 emit_operand32(rbx, adr);
4029 }
4031 void Assembler::fstp_d(int index) {
4032 emit_farith(0xDD, 0xD8, index);
4033 }
4035 void Assembler::fstp_s(Address adr) {
4036 InstructionMark im(this);
4037 emit_int8((unsigned char)0xD9);
4038 emit_operand32(rbx, adr);
4039 }
4041 void Assembler::fstp_x(Address adr) {
4042 InstructionMark im(this);
4043 emit_int8((unsigned char)0xDB);
4044 emit_operand32(rdi, adr);
4045 }
4047 void Assembler::fsub(int i) {
4048 emit_farith(0xD8, 0xE0, i);
4049 }
4051 void Assembler::fsub_d(Address src) {
4052 InstructionMark im(this);
4053 emit_int8((unsigned char)0xDC);
4054 emit_operand32(rsp, src);
4055 }
4057 void Assembler::fsub_s(Address src) {
4058 InstructionMark im(this);
4059 emit_int8((unsigned char)0xD8);
4060 emit_operand32(rsp, src);
4061 }
4063 void Assembler::fsuba(int i) {
4064 emit_farith(0xDC, 0xE8, i);
4065 }
4067 void Assembler::fsubp(int i) {
4068 emit_farith(0xDE, 0xE8, i); // ST(0) <- ST(0) - ST(1) and pop (Intel manual wrong)
4069 }
4071 void Assembler::fsubr(int i) {
4072 emit_farith(0xD8, 0xE8, i);
4073 }
4075 void Assembler::fsubr_d(Address src) {
4076 InstructionMark im(this);
4077 emit_int8((unsigned char)0xDC);
4078 emit_operand32(rbp, src);
4079 }
4081 void Assembler::fsubr_s(Address src) {
4082 InstructionMark im(this);
4083 emit_int8((unsigned char)0xD8);
4084 emit_operand32(rbp, src);
4085 }
4087 void Assembler::fsubra(int i) {
4088 emit_farith(0xDC, 0xE0, i);
4089 }
4091 void Assembler::fsubrp(int i) {
4092 emit_farith(0xDE, 0xE0, i); // ST(0) <- ST(1) - ST(0) and pop (Intel manual wrong)
4093 }
4095 void Assembler::ftan() {
4096 emit_int8((unsigned char)0xD9);
4097 emit_int8((unsigned char)0xF2);
4098 emit_int8((unsigned char)0xDD);
4099 emit_int8((unsigned char)0xD8);
4100 }
4102 void Assembler::ftst() {
4103 emit_int8((unsigned char)0xD9);
4104 emit_int8((unsigned char)0xE4);
4105 }
4107 void Assembler::fucomi(int i) {
4108 // make sure the instruction is supported (introduced for P6, together with cmov)
4109 guarantee(VM_Version::supports_cmov(), "illegal instruction");
4110 emit_farith(0xDB, 0xE8, i);
4111 }
4113 void Assembler::fucomip(int i) {
4114 // make sure the instruction is supported (introduced for P6, together with cmov)
4115 guarantee(VM_Version::supports_cmov(), "illegal instruction");
4116 emit_farith(0xDF, 0xE8, i);
4117 }
4119 void Assembler::fwait() {
4120 emit_int8((unsigned char)0x9B);
4121 }
4123 void Assembler::fxch(int i) {
4124 emit_farith(0xD9, 0xC8, i);
4125 }
4127 void Assembler::fyl2x() {
4128 emit_int8((unsigned char)0xD9);
4129 emit_int8((unsigned char)0xF1);
4130 }
4132 void Assembler::frndint() {
4133 emit_int8((unsigned char)0xD9);
4134 emit_int8((unsigned char)0xFC);
4135 }
4137 void Assembler::f2xm1() {
4138 emit_int8((unsigned char)0xD9);
4139 emit_int8((unsigned char)0xF0);
4140 }
4142 void Assembler::fldl2e() {
4143 emit_int8((unsigned char)0xD9);
4144 emit_int8((unsigned char)0xEA);
4145 }
4147 // SSE SIMD prefix byte values corresponding to VexSimdPrefix encoding.
4148 static int simd_pre[4] = { 0, 0x66, 0xF3, 0xF2 };
4149 // SSE opcode second byte values (first is 0x0F) corresponding to VexOpcode encoding.
4150 static int simd_opc[4] = { 0, 0, 0x38, 0x3A };
4152 // Generate SSE legacy REX prefix and SIMD opcode based on VEX encoding.
4153 void Assembler::rex_prefix(Address adr, XMMRegister xreg, VexSimdPrefix pre, VexOpcode opc, bool rex_w) {
4154 if (pre > 0) {
4155 emit_int8(simd_pre[pre]);
4156 }
4157 if (rex_w) {
4158 prefixq(adr, xreg);
4159 } else {
4160 prefix(adr, xreg);
4161 }
4162 if (opc > 0) {
4163 emit_int8(0x0F);
4164 int opc2 = simd_opc[opc];
4165 if (opc2 > 0) {
4166 emit_int8(opc2);
4167 }
4168 }
4169 }
4171 int Assembler::rex_prefix_and_encode(int dst_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool rex_w) {
4172 if (pre > 0) {
4173 emit_int8(simd_pre[pre]);
4174 }
4175 int encode = (rex_w) ? prefixq_and_encode(dst_enc, src_enc) :
4176 prefix_and_encode(dst_enc, src_enc);
4177 if (opc > 0) {
4178 emit_int8(0x0F);
4179 int opc2 = simd_opc[opc];
4180 if (opc2 > 0) {
4181 emit_int8(opc2);
4182 }
4183 }
4184 return encode;
4185 }
4188 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) {
4189 if (vex_b || vex_x || vex_w || (opc == VEX_OPCODE_0F_38) || (opc == VEX_OPCODE_0F_3A)) {
4190 prefix(VEX_3bytes);
4192 int byte1 = (vex_r ? VEX_R : 0) | (vex_x ? VEX_X : 0) | (vex_b ? VEX_B : 0);
4193 byte1 = (~byte1) & 0xE0;
4194 byte1 |= opc;
4195 emit_int8(byte1);
4197 int byte2 = ((~nds_enc) & 0xf) << 3;
4198 byte2 |= (vex_w ? VEX_W : 0) | (vector256 ? 4 : 0) | pre;
4199 emit_int8(byte2);
4200 } else {
4201 prefix(VEX_2bytes);
4203 int byte1 = vex_r ? VEX_R : 0;
4204 byte1 = (~byte1) & 0x80;
4205 byte1 |= ((~nds_enc) & 0xf) << 3;
4206 byte1 |= (vector256 ? 4 : 0) | pre;
4207 emit_int8(byte1);
4208 }
4209 }
4211 void Assembler::vex_prefix(Address adr, int nds_enc, int xreg_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256){
4212 bool vex_r = (xreg_enc >= 8);
4213 bool vex_b = adr.base_needs_rex();
4214 bool vex_x = adr.index_needs_rex();
4215 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256);
4216 }
4218 int Assembler::vex_prefix_and_encode(int dst_enc, int nds_enc, int src_enc, VexSimdPrefix pre, VexOpcode opc, bool vex_w, bool vector256) {
4219 bool vex_r = (dst_enc >= 8);
4220 bool vex_b = (src_enc >= 8);
4221 bool vex_x = false;
4222 vex_prefix(vex_r, vex_b, vex_x, vex_w, nds_enc, pre, opc, vector256);
4223 return (((dst_enc & 7) << 3) | (src_enc & 7));
4224 }
4227 void Assembler::simd_prefix(XMMRegister xreg, XMMRegister nds, Address adr, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) {
4228 if (UseAVX > 0) {
4229 int xreg_enc = xreg->encoding();
4230 int nds_enc = nds->is_valid() ? nds->encoding() : 0;
4231 vex_prefix(adr, nds_enc, xreg_enc, pre, opc, rex_w, vector256);
4232 } else {
4233 assert((nds == xreg) || (nds == xnoreg), "wrong sse encoding");
4234 rex_prefix(adr, xreg, pre, opc, rex_w);
4235 }
4236 }
4238 int Assembler::simd_prefix_and_encode(XMMRegister dst, XMMRegister nds, XMMRegister src, VexSimdPrefix pre, VexOpcode opc, bool rex_w, bool vector256) {
4239 int dst_enc = dst->encoding();
4240 int src_enc = src->encoding();
4241 if (UseAVX > 0) {
4242 int nds_enc = nds->is_valid() ? nds->encoding() : 0;
4243 return vex_prefix_and_encode(dst_enc, nds_enc, src_enc, pre, opc, rex_w, vector256);
4244 } else {
4245 assert((nds == dst) || (nds == src) || (nds == xnoreg), "wrong sse encoding");
4246 return rex_prefix_and_encode(dst_enc, src_enc, pre, opc, rex_w);
4247 }
4248 }
4250 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) {
4251 InstructionMark im(this);
4252 simd_prefix(dst, dst, src, pre);
4253 emit_int8(opcode);
4254 emit_operand(dst, src);
4255 }
4257 void Assembler::emit_simd_arith(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) {
4258 int encode = simd_prefix_and_encode(dst, dst, src, pre);
4259 emit_int8(opcode);
4260 emit_int8((unsigned char)(0xC0 | encode));
4261 }
4263 // Versions with no second source register (non-destructive source).
4264 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, Address src, VexSimdPrefix pre) {
4265 InstructionMark im(this);
4266 simd_prefix(dst, xnoreg, src, pre);
4267 emit_int8(opcode);
4268 emit_operand(dst, src);
4269 }
4271 void Assembler::emit_simd_arith_nonds(int opcode, XMMRegister dst, XMMRegister src, VexSimdPrefix pre) {
4272 int encode = simd_prefix_and_encode(dst, xnoreg, src, pre);
4273 emit_int8(opcode);
4274 emit_int8((unsigned char)(0xC0 | encode));
4275 }
4277 // 3-operands AVX instructions
4278 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
4279 Address src, VexSimdPrefix pre, bool vector256) {
4280 InstructionMark im(this);
4281 vex_prefix(dst, nds, src, pre, vector256);
4282 emit_int8(opcode);
4283 emit_operand(dst, src);
4284 }
4286 void Assembler::emit_vex_arith(int opcode, XMMRegister dst, XMMRegister nds,
4287 XMMRegister src, VexSimdPrefix pre, bool vector256) {
4288 int encode = vex_prefix_and_encode(dst, nds, src, pre, vector256);
4289 emit_int8(opcode);
4290 emit_int8((unsigned char)(0xC0 | encode));
4291 }
4293 #ifndef _LP64
4295 void Assembler::incl(Register dst) {
4296 // Don't use it directly. Use MacroAssembler::incrementl() instead.
4297 emit_int8(0x40 | dst->encoding());
4298 }
4300 void Assembler::lea(Register dst, Address src) {
4301 leal(dst, src);
4302 }
4304 void Assembler::mov_literal32(Address dst, int32_t imm32, RelocationHolder const& rspec) {
4305 InstructionMark im(this);
4306 emit_int8((unsigned char)0xC7);
4307 emit_operand(rax, dst);
4308 emit_data((int)imm32, rspec, 0);
4309 }
4311 void Assembler::mov_literal32(Register dst, int32_t imm32, RelocationHolder const& rspec) {
4312 InstructionMark im(this);
4313 int encode = prefix_and_encode(dst->encoding());
4314 emit_int8((unsigned char)(0xB8 | encode));
4315 emit_data((int)imm32, rspec, 0);
4316 }
4318 void Assembler::popa() { // 32bit
4319 emit_int8(0x61);
4320 }
4322 void Assembler::push_literal32(int32_t imm32, RelocationHolder const& rspec) {
4323 InstructionMark im(this);
4324 emit_int8(0x68);
4325 emit_data(imm32, rspec, 0);
4326 }
4328 void Assembler::pusha() { // 32bit
4329 emit_int8(0x60);
4330 }
4332 void Assembler::set_byte_if_not_zero(Register dst) {
4333 emit_int8(0x0F);
4334 emit_int8((unsigned char)0x95);
4335 emit_int8((unsigned char)(0xE0 | dst->encoding()));
4336 }
4338 void Assembler::shldl(Register dst, Register src) {
4339 emit_int8(0x0F);
4340 emit_int8((unsigned char)0xA5);
4341 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding()));
4342 }
4344 void Assembler::shrdl(Register dst, Register src) {
4345 emit_int8(0x0F);
4346 emit_int8((unsigned char)0xAD);
4347 emit_int8((unsigned char)(0xC0 | src->encoding() << 3 | dst->encoding()));
4348 }
4350 #else // LP64
4352 void Assembler::set_byte_if_not_zero(Register dst) {
4353 int enc = prefix_and_encode(dst->encoding(), true);
4354 emit_int8(0x0F);
4355 emit_int8((unsigned char)0x95);
4356 emit_int8((unsigned char)(0xE0 | enc));
4357 }
4359 // 64bit only pieces of the assembler
4360 // This should only be used by 64bit instructions that can use rip-relative
4361 // it cannot be used by instructions that want an immediate value.
4363 bool Assembler::reachable(AddressLiteral adr) {
4364 int64_t disp;
4365 // None will force a 64bit literal to the code stream. Likely a placeholder
4366 // for something that will be patched later and we need to certain it will
4367 // always be reachable.
4368 if (adr.reloc() == relocInfo::none) {
4369 return false;
4370 }
4371 if (adr.reloc() == relocInfo::internal_word_type) {
4372 // This should be rip relative and easily reachable.
4373 return true;
4374 }
4375 if (adr.reloc() == relocInfo::virtual_call_type ||
4376 adr.reloc() == relocInfo::opt_virtual_call_type ||
4377 adr.reloc() == relocInfo::static_call_type ||
4378 adr.reloc() == relocInfo::static_stub_type ) {
4379 // This should be rip relative within the code cache and easily
4380 // reachable until we get huge code caches. (At which point
4381 // ic code is going to have issues).
4382 return true;
4383 }
4384 if (adr.reloc() != relocInfo::external_word_type &&
4385 adr.reloc() != relocInfo::poll_return_type && // these are really external_word but need special
4386 adr.reloc() != relocInfo::poll_type && // relocs to identify them
4387 adr.reloc() != relocInfo::runtime_call_type ) {
4388 return false;
4389 }
4391 // Stress the correction code
4392 if (ForceUnreachable) {
4393 // Must be runtimecall reloc, see if it is in the codecache
4394 // Flipping stuff in the codecache to be unreachable causes issues
4395 // with things like inline caches where the additional instructions
4396 // are not handled.
4397 if (CodeCache::find_blob(adr._target) == NULL) {
4398 return false;
4399 }
4400 }
4401 // For external_word_type/runtime_call_type if it is reachable from where we
4402 // are now (possibly a temp buffer) and where we might end up
4403 // anywhere in the codeCache then we are always reachable.
4404 // This would have to change if we ever save/restore shared code
4405 // to be more pessimistic.
4406 disp = (int64_t)adr._target - ((int64_t)CodeCache::low_bound() + sizeof(int));
4407 if (!is_simm32(disp)) return false;
4408 disp = (int64_t)adr._target - ((int64_t)CodeCache::high_bound() + sizeof(int));
4409 if (!is_simm32(disp)) return false;
4411 disp = (int64_t)adr._target - ((int64_t)pc() + sizeof(int));
4413 // Because rip relative is a disp + address_of_next_instruction and we
4414 // don't know the value of address_of_next_instruction we apply a fudge factor
4415 // to make sure we will be ok no matter the size of the instruction we get placed into.
4416 // We don't have to fudge the checks above here because they are already worst case.
4418 // 12 == override/rex byte, opcode byte, rm byte, sib byte, a 4-byte disp , 4-byte literal
4419 // + 4 because better safe than sorry.
4420 const int fudge = 12 + 4;
4421 if (disp < 0) {
4422 disp -= fudge;
4423 } else {
4424 disp += fudge;
4425 }
4426 return is_simm32(disp);
4427 }
4429 // Check if the polling page is not reachable from the code cache using rip-relative
4430 // addressing.
4431 bool Assembler::is_polling_page_far() {
4432 intptr_t addr = (intptr_t)os::get_polling_page();
4433 return ForceUnreachable ||
4434 !is_simm32(addr - (intptr_t)CodeCache::low_bound()) ||
4435 !is_simm32(addr - (intptr_t)CodeCache::high_bound());
4436 }
4438 void Assembler::emit_data64(jlong data,
4439 relocInfo::relocType rtype,
4440 int format) {
4441 if (rtype == relocInfo::none) {
4442 emit_int64(data);
4443 } else {
4444 emit_data64(data, Relocation::spec_simple(rtype), format);
4445 }
4446 }
4448 void Assembler::emit_data64(jlong data,
4449 RelocationHolder const& rspec,
4450 int format) {
4451 assert(imm_operand == 0, "default format must be immediate in this file");
4452 assert(imm_operand == format, "must be immediate");
4453 assert(inst_mark() != NULL, "must be inside InstructionMark");
4454 // Do not use AbstractAssembler::relocate, which is not intended for
4455 // embedded words. Instead, relocate to the enclosing instruction.
4456 code_section()->relocate(inst_mark(), rspec, format);
4457 #ifdef ASSERT
4458 check_relocation(rspec, format);
4459 #endif
4460 emit_int64(data);
4461 }
4463 int Assembler::prefix_and_encode(int reg_enc, bool byteinst) {
4464 if (reg_enc >= 8) {
4465 prefix(REX_B);
4466 reg_enc -= 8;
4467 } else if (byteinst && reg_enc >= 4) {
4468 prefix(REX);
4469 }
4470 return reg_enc;
4471 }
4473 int Assembler::prefixq_and_encode(int reg_enc) {
4474 if (reg_enc < 8) {
4475 prefix(REX_W);
4476 } else {
4477 prefix(REX_WB);
4478 reg_enc -= 8;
4479 }
4480 return reg_enc;
4481 }
4483 int Assembler::prefix_and_encode(int dst_enc, int src_enc, bool byteinst) {
4484 if (dst_enc < 8) {
4485 if (src_enc >= 8) {
4486 prefix(REX_B);
4487 src_enc -= 8;
4488 } else if (byteinst && src_enc >= 4) {
4489 prefix(REX);
4490 }
4491 } else {
4492 if (src_enc < 8) {
4493 prefix(REX_R);
4494 } else {
4495 prefix(REX_RB);
4496 src_enc -= 8;
4497 }
4498 dst_enc -= 8;
4499 }
4500 return dst_enc << 3 | src_enc;
4501 }
4503 int Assembler::prefixq_and_encode(int dst_enc, int src_enc) {
4504 if (dst_enc < 8) {
4505 if (src_enc < 8) {
4506 prefix(REX_W);
4507 } else {
4508 prefix(REX_WB);
4509 src_enc -= 8;
4510 }
4511 } else {
4512 if (src_enc < 8) {
4513 prefix(REX_WR);
4514 } else {
4515 prefix(REX_WRB);
4516 src_enc -= 8;
4517 }
4518 dst_enc -= 8;
4519 }
4520 return dst_enc << 3 | src_enc;
4521 }
4523 void Assembler::prefix(Register reg) {
4524 if (reg->encoding() >= 8) {
4525 prefix(REX_B);
4526 }
4527 }
4529 void Assembler::prefix(Address adr) {
4530 if (adr.base_needs_rex()) {
4531 if (adr.index_needs_rex()) {
4532 prefix(REX_XB);
4533 } else {
4534 prefix(REX_B);
4535 }
4536 } else {
4537 if (adr.index_needs_rex()) {
4538 prefix(REX_X);
4539 }
4540 }
4541 }
4543 void Assembler::prefixq(Address adr) {
4544 if (adr.base_needs_rex()) {
4545 if (adr.index_needs_rex()) {
4546 prefix(REX_WXB);
4547 } else {
4548 prefix(REX_WB);
4549 }
4550 } else {
4551 if (adr.index_needs_rex()) {
4552 prefix(REX_WX);
4553 } else {
4554 prefix(REX_W);
4555 }
4556 }
4557 }
4560 void Assembler::prefix(Address adr, Register reg, bool byteinst) {
4561 if (reg->encoding() < 8) {
4562 if (adr.base_needs_rex()) {
4563 if (adr.index_needs_rex()) {
4564 prefix(REX_XB);
4565 } else {
4566 prefix(REX_B);
4567 }
4568 } else {
4569 if (adr.index_needs_rex()) {
4570 prefix(REX_X);
4571 } else if (byteinst && reg->encoding() >= 4 ) {
4572 prefix(REX);
4573 }
4574 }
4575 } else {
4576 if (adr.base_needs_rex()) {
4577 if (adr.index_needs_rex()) {
4578 prefix(REX_RXB);
4579 } else {
4580 prefix(REX_RB);
4581 }
4582 } else {
4583 if (adr.index_needs_rex()) {
4584 prefix(REX_RX);
4585 } else {
4586 prefix(REX_R);
4587 }
4588 }
4589 }
4590 }
4592 void Assembler::prefixq(Address adr, Register src) {
4593 if (src->encoding() < 8) {
4594 if (adr.base_needs_rex()) {
4595 if (adr.index_needs_rex()) {
4596 prefix(REX_WXB);
4597 } else {
4598 prefix(REX_WB);
4599 }
4600 } else {
4601 if (adr.index_needs_rex()) {
4602 prefix(REX_WX);
4603 } else {
4604 prefix(REX_W);
4605 }
4606 }
4607 } else {
4608 if (adr.base_needs_rex()) {
4609 if (adr.index_needs_rex()) {
4610 prefix(REX_WRXB);
4611 } else {
4612 prefix(REX_WRB);
4613 }
4614 } else {
4615 if (adr.index_needs_rex()) {
4616 prefix(REX_WRX);
4617 } else {
4618 prefix(REX_WR);
4619 }
4620 }
4621 }
4622 }
4624 void Assembler::prefix(Address adr, XMMRegister reg) {
4625 if (reg->encoding() < 8) {
4626 if (adr.base_needs_rex()) {
4627 if (adr.index_needs_rex()) {
4628 prefix(REX_XB);
4629 } else {
4630 prefix(REX_B);
4631 }
4632 } else {
4633 if (adr.index_needs_rex()) {
4634 prefix(REX_X);
4635 }
4636 }
4637 } else {
4638 if (adr.base_needs_rex()) {
4639 if (adr.index_needs_rex()) {
4640 prefix(REX_RXB);
4641 } else {
4642 prefix(REX_RB);
4643 }
4644 } else {
4645 if (adr.index_needs_rex()) {
4646 prefix(REX_RX);
4647 } else {
4648 prefix(REX_R);
4649 }
4650 }
4651 }
4652 }
4654 void Assembler::prefixq(Address adr, XMMRegister src) {
4655 if (src->encoding() < 8) {
4656 if (adr.base_needs_rex()) {
4657 if (adr.index_needs_rex()) {
4658 prefix(REX_WXB);
4659 } else {
4660 prefix(REX_WB);
4661 }
4662 } else {
4663 if (adr.index_needs_rex()) {
4664 prefix(REX_WX);
4665 } else {
4666 prefix(REX_W);
4667 }
4668 }
4669 } else {
4670 if (adr.base_needs_rex()) {
4671 if (adr.index_needs_rex()) {
4672 prefix(REX_WRXB);
4673 } else {
4674 prefix(REX_WRB);
4675 }
4676 } else {
4677 if (adr.index_needs_rex()) {
4678 prefix(REX_WRX);
4679 } else {
4680 prefix(REX_WR);
4681 }
4682 }
4683 }
4684 }
4686 void Assembler::adcq(Register dst, int32_t imm32) {
4687 (void) prefixq_and_encode(dst->encoding());
4688 emit_arith(0x81, 0xD0, dst, imm32);
4689 }
4691 void Assembler::adcq(Register dst, Address src) {
4692 InstructionMark im(this);
4693 prefixq(src, dst);
4694 emit_int8(0x13);
4695 emit_operand(dst, src);
4696 }
4698 void Assembler::adcq(Register dst, Register src) {
4699 (int) prefixq_and_encode(dst->encoding(), src->encoding());
4700 emit_arith(0x13, 0xC0, dst, src);
4701 }
4703 void Assembler::addq(Address dst, int32_t imm32) {
4704 InstructionMark im(this);
4705 prefixq(dst);
4706 emit_arith_operand(0x81, rax, dst,imm32);
4707 }
4709 void Assembler::addq(Address dst, Register src) {
4710 InstructionMark im(this);
4711 prefixq(dst, src);
4712 emit_int8(0x01);
4713 emit_operand(src, dst);
4714 }
4716 void Assembler::addq(Register dst, int32_t imm32) {
4717 (void) prefixq_and_encode(dst->encoding());
4718 emit_arith(0x81, 0xC0, dst, imm32);
4719 }
4721 void Assembler::addq(Register dst, Address src) {
4722 InstructionMark im(this);
4723 prefixq(src, dst);
4724 emit_int8(0x03);
4725 emit_operand(dst, src);
4726 }
4728 void Assembler::addq(Register dst, Register src) {
4729 (void) prefixq_and_encode(dst->encoding(), src->encoding());
4730 emit_arith(0x03, 0xC0, dst, src);
4731 }
4733 void Assembler::andq(Address dst, int32_t imm32) {
4734 InstructionMark im(this);
4735 prefixq(dst);
4736 emit_int8((unsigned char)0x81);
4737 emit_operand(rsp, dst, 4);
4738 emit_long(imm32);
4739 }
4741 void Assembler::andq(Register dst, int32_t imm32) {
4742 (void) prefixq_and_encode(dst->encoding());
4743 emit_arith(0x81, 0xE0, dst, imm32);
4744 }
4746 void Assembler::andq(Register dst, Address src) {
4747 InstructionMark im(this);
4748 prefixq(src, dst);
4749 emit_int8(0x23);
4750 emit_operand(dst, src);
4751 }
4753 void Assembler::andq(Register dst, Register src) {
4754 (int) prefixq_and_encode(dst->encoding(), src->encoding());
4755 emit_arith(0x23, 0xC0, dst, src);
4756 }
4758 void Assembler::bsfq(Register dst, Register src) {
4759 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
4760 emit_int8(0x0F);
4761 emit_int8((unsigned char)0xBC);
4762 emit_int8((unsigned char)(0xC0 | encode));
4763 }
4765 void Assembler::bsrq(Register dst, Register src) {
4766 assert(!VM_Version::supports_lzcnt(), "encoding is treated as LZCNT");
4767 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
4768 emit_int8(0x0F);
4769 emit_int8((unsigned char)0xBD);
4770 emit_int8((unsigned char)(0xC0 | encode));
4771 }
4773 void Assembler::bswapq(Register reg) {
4774 int encode = prefixq_and_encode(reg->encoding());
4775 emit_int8(0x0F);
4776 emit_int8((unsigned char)(0xC8 | encode));
4777 }
4779 void Assembler::cdqq() {
4780 prefix(REX_W);
4781 emit_int8((unsigned char)0x99);
4782 }
4784 void Assembler::clflush(Address adr) {
4785 prefix(adr);
4786 emit_int8(0x0F);
4787 emit_int8((unsigned char)0xAE);
4788 emit_operand(rdi, adr);
4789 }
4791 void Assembler::cmovq(Condition cc, Register dst, Register src) {
4792 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
4793 emit_int8(0x0F);
4794 emit_int8(0x40 | cc);
4795 emit_int8((unsigned char)(0xC0 | encode));
4796 }
4798 void Assembler::cmovq(Condition cc, Register dst, Address src) {
4799 InstructionMark im(this);
4800 prefixq(src, dst);
4801 emit_int8(0x0F);
4802 emit_int8(0x40 | cc);
4803 emit_operand(dst, src);
4804 }
4806 void Assembler::cmpq(Address dst, int32_t imm32) {
4807 InstructionMark im(this);
4808 prefixq(dst);
4809 emit_int8((unsigned char)0x81);
4810 emit_operand(rdi, dst, 4);
4811 emit_long(imm32);
4812 }
4814 void Assembler::cmpq(Register dst, int32_t imm32) {
4815 (void) prefixq_and_encode(dst->encoding());
4816 emit_arith(0x81, 0xF8, dst, imm32);
4817 }
4819 void Assembler::cmpq(Address dst, Register src) {
4820 InstructionMark im(this);
4821 prefixq(dst, src);
4822 emit_int8(0x3B);
4823 emit_operand(src, dst);
4824 }
4826 void Assembler::cmpq(Register dst, Register src) {
4827 (void) prefixq_and_encode(dst->encoding(), src->encoding());
4828 emit_arith(0x3B, 0xC0, dst, src);
4829 }
4831 void Assembler::cmpq(Register dst, Address src) {
4832 InstructionMark im(this);
4833 prefixq(src, dst);
4834 emit_int8(0x3B);
4835 emit_operand(dst, src);
4836 }
4838 void Assembler::cmpxchgq(Register reg, Address adr) {
4839 InstructionMark im(this);
4840 prefixq(adr, reg);
4841 emit_int8(0x0F);
4842 emit_int8((unsigned char)0xB1);
4843 emit_operand(reg, adr);
4844 }
4846 void Assembler::cvtsi2sdq(XMMRegister dst, Register src) {
4847 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
4848 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F2);
4849 emit_int8(0x2A);
4850 emit_int8((unsigned char)(0xC0 | encode));
4851 }
4853 void Assembler::cvtsi2sdq(XMMRegister dst, Address src) {
4854 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
4855 InstructionMark im(this);
4856 simd_prefix_q(dst, dst, src, VEX_SIMD_F2);
4857 emit_int8(0x2A);
4858 emit_operand(dst, src);
4859 }
4861 void Assembler::cvtsi2ssq(XMMRegister dst, Register src) {
4862 NOT_LP64(assert(VM_Version::supports_sse(), ""));
4863 int encode = simd_prefix_and_encode_q(dst, dst, src, VEX_SIMD_F3);
4864 emit_int8(0x2A);
4865 emit_int8((unsigned char)(0xC0 | encode));
4866 }
4868 void Assembler::cvtsi2ssq(XMMRegister dst, Address src) {
4869 NOT_LP64(assert(VM_Version::supports_sse(), ""));
4870 InstructionMark im(this);
4871 simd_prefix_q(dst, dst, src, VEX_SIMD_F3);
4872 emit_int8(0x2A);
4873 emit_operand(dst, src);
4874 }
4876 void Assembler::cvttsd2siq(Register dst, XMMRegister src) {
4877 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
4878 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F2);
4879 emit_int8(0x2C);
4880 emit_int8((unsigned char)(0xC0 | encode));
4881 }
4883 void Assembler::cvttss2siq(Register dst, XMMRegister src) {
4884 NOT_LP64(assert(VM_Version::supports_sse(), ""));
4885 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_F3);
4886 emit_int8(0x2C);
4887 emit_int8((unsigned char)(0xC0 | encode));
4888 }
4890 void Assembler::decl(Register dst) {
4891 // Don't use it directly. Use MacroAssembler::decrementl() instead.
4892 // Use two-byte form (one-byte form is a REX prefix in 64-bit mode)
4893 int encode = prefix_and_encode(dst->encoding());
4894 emit_int8((unsigned char)0xFF);
4895 emit_int8((unsigned char)(0xC8 | encode));
4896 }
4898 void Assembler::decq(Register dst) {
4899 // Don't use it directly. Use MacroAssembler::decrementq() instead.
4900 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode)
4901 int encode = prefixq_and_encode(dst->encoding());
4902 emit_int8((unsigned char)0xFF);
4903 emit_int8(0xC8 | encode);
4904 }
4906 void Assembler::decq(Address dst) {
4907 // Don't use it directly. Use MacroAssembler::decrementq() instead.
4908 InstructionMark im(this);
4909 prefixq(dst);
4910 emit_int8((unsigned char)0xFF);
4911 emit_operand(rcx, dst);
4912 }
4914 void Assembler::fxrstor(Address src) {
4915 prefixq(src);
4916 emit_int8(0x0F);
4917 emit_int8((unsigned char)0xAE);
4918 emit_operand(as_Register(1), src);
4919 }
4921 void Assembler::fxsave(Address dst) {
4922 prefixq(dst);
4923 emit_int8(0x0F);
4924 emit_int8((unsigned char)0xAE);
4925 emit_operand(as_Register(0), dst);
4926 }
4928 void Assembler::idivq(Register src) {
4929 int encode = prefixq_and_encode(src->encoding());
4930 emit_int8((unsigned char)0xF7);
4931 emit_int8((unsigned char)(0xF8 | encode));
4932 }
4934 void Assembler::imulq(Register dst, Register src) {
4935 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
4936 emit_int8(0x0F);
4937 emit_int8((unsigned char)0xAF);
4938 emit_int8((unsigned char)(0xC0 | encode));
4939 }
4941 void Assembler::imulq(Register dst, Register src, int value) {
4942 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
4943 if (is8bit(value)) {
4944 emit_int8(0x6B);
4945 emit_int8((unsigned char)(0xC0 | encode));
4946 emit_int8(value & 0xFF);
4947 } else {
4948 emit_int8(0x69);
4949 emit_int8((unsigned char)(0xC0 | encode));
4950 emit_long(value);
4951 }
4952 }
4954 void Assembler::incl(Register dst) {
4955 // Don't use it directly. Use MacroAssembler::incrementl() instead.
4956 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode)
4957 int encode = prefix_and_encode(dst->encoding());
4958 emit_int8((unsigned char)0xFF);
4959 emit_int8((unsigned char)(0xC0 | encode));
4960 }
4962 void Assembler::incq(Register dst) {
4963 // Don't use it directly. Use MacroAssembler::incrementq() instead.
4964 // Use two-byte form (one-byte from is a REX prefix in 64-bit mode)
4965 int encode = prefixq_and_encode(dst->encoding());
4966 emit_int8((unsigned char)0xFF);
4967 emit_int8((unsigned char)(0xC0 | encode));
4968 }
4970 void Assembler::incq(Address dst) {
4971 // Don't use it directly. Use MacroAssembler::incrementq() instead.
4972 InstructionMark im(this);
4973 prefixq(dst);
4974 emit_int8((unsigned char)0xFF);
4975 emit_operand(rax, dst);
4976 }
4978 void Assembler::lea(Register dst, Address src) {
4979 leaq(dst, src);
4980 }
4982 void Assembler::leaq(Register dst, Address src) {
4983 InstructionMark im(this);
4984 prefixq(src, dst);
4985 emit_int8((unsigned char)0x8D);
4986 emit_operand(dst, src);
4987 }
4989 void Assembler::mov64(Register dst, int64_t imm64) {
4990 InstructionMark im(this);
4991 int encode = prefixq_and_encode(dst->encoding());
4992 emit_int8((unsigned char)(0xB8 | encode));
4993 emit_int64(imm64);
4994 }
4996 void Assembler::mov_literal64(Register dst, intptr_t imm64, RelocationHolder const& rspec) {
4997 InstructionMark im(this);
4998 int encode = prefixq_and_encode(dst->encoding());
4999 emit_int8(0xB8 | encode);
5000 emit_data64(imm64, rspec);
5001 }
5003 void Assembler::mov_narrow_oop(Register dst, int32_t imm32, RelocationHolder const& rspec) {
5004 InstructionMark im(this);
5005 int encode = prefix_and_encode(dst->encoding());
5006 emit_int8((unsigned char)(0xB8 | encode));
5007 emit_data((int)imm32, rspec, narrow_oop_operand);
5008 }
5010 void Assembler::mov_narrow_oop(Address dst, int32_t imm32, RelocationHolder const& rspec) {
5011 InstructionMark im(this);
5012 prefix(dst);
5013 emit_int8((unsigned char)0xC7);
5014 emit_operand(rax, dst, 4);
5015 emit_data((int)imm32, rspec, narrow_oop_operand);
5016 }
5018 void Assembler::cmp_narrow_oop(Register src1, int32_t imm32, RelocationHolder const& rspec) {
5019 InstructionMark im(this);
5020 int encode = prefix_and_encode(src1->encoding());
5021 emit_int8((unsigned char)0x81);
5022 emit_int8((unsigned char)(0xF8 | encode));
5023 emit_data((int)imm32, rspec, narrow_oop_operand);
5024 }
5026 void Assembler::cmp_narrow_oop(Address src1, int32_t imm32, RelocationHolder const& rspec) {
5027 InstructionMark im(this);
5028 prefix(src1);
5029 emit_int8((unsigned char)0x81);
5030 emit_operand(rax, src1, 4);
5031 emit_data((int)imm32, rspec, narrow_oop_operand);
5032 }
5034 void Assembler::lzcntq(Register dst, Register src) {
5035 assert(VM_Version::supports_lzcnt(), "encoding is treated as BSR");
5036 emit_int8((unsigned char)0xF3);
5037 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5038 emit_int8(0x0F);
5039 emit_int8((unsigned char)0xBD);
5040 emit_int8((unsigned char)(0xC0 | encode));
5041 }
5043 void Assembler::movdq(XMMRegister dst, Register src) {
5044 // table D-1 says MMX/SSE2
5045 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
5046 int encode = simd_prefix_and_encode_q(dst, src, VEX_SIMD_66);
5047 emit_int8(0x6E);
5048 emit_int8((unsigned char)(0xC0 | encode));
5049 }
5051 void Assembler::movdq(Register dst, XMMRegister src) {
5052 // table D-1 says MMX/SSE2
5053 NOT_LP64(assert(VM_Version::supports_sse2(), ""));
5054 // swap src/dst to get correct prefix
5055 int encode = simd_prefix_and_encode_q(src, dst, VEX_SIMD_66);
5056 emit_int8(0x7E);
5057 emit_int8((unsigned char)(0xC0 | encode));
5058 }
5060 void Assembler::movq(Register dst, Register src) {
5061 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5062 emit_int8((unsigned char)0x8B);
5063 emit_int8((unsigned char)(0xC0 | encode));
5064 }
5066 void Assembler::movq(Register dst, Address src) {
5067 InstructionMark im(this);
5068 prefixq(src, dst);
5069 emit_int8((unsigned char)0x8B);
5070 emit_operand(dst, src);
5071 }
5073 void Assembler::movq(Address dst, Register src) {
5074 InstructionMark im(this);
5075 prefixq(dst, src);
5076 emit_int8((unsigned char)0x89);
5077 emit_operand(src, dst);
5078 }
5080 void Assembler::movsbq(Register dst, Address src) {
5081 InstructionMark im(this);
5082 prefixq(src, dst);
5083 emit_int8(0x0F);
5084 emit_int8((unsigned char)0xBE);
5085 emit_operand(dst, src);
5086 }
5088 void Assembler::movsbq(Register dst, Register src) {
5089 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5090 emit_int8(0x0F);
5091 emit_int8((unsigned char)0xBE);
5092 emit_int8((unsigned char)(0xC0 | encode));
5093 }
5095 void Assembler::movslq(Register dst, int32_t imm32) {
5096 // dbx shows movslq(rcx, 3) as movq $0x0000000049000000,(%rbx)
5097 // and movslq(r8, 3); as movl $0x0000000048000000,(%rbx)
5098 // as a result we shouldn't use until tested at runtime...
5099 ShouldNotReachHere();
5100 InstructionMark im(this);
5101 int encode = prefixq_and_encode(dst->encoding());
5102 emit_int8((unsigned char)(0xC7 | encode));
5103 emit_long(imm32);
5104 }
5106 void Assembler::movslq(Address dst, int32_t imm32) {
5107 assert(is_simm32(imm32), "lost bits");
5108 InstructionMark im(this);
5109 prefixq(dst);
5110 emit_int8((unsigned char)0xC7);
5111 emit_operand(rax, dst, 4);
5112 emit_long(imm32);
5113 }
5115 void Assembler::movslq(Register dst, Address src) {
5116 InstructionMark im(this);
5117 prefixq(src, dst);
5118 emit_int8(0x63);
5119 emit_operand(dst, src);
5120 }
5122 void Assembler::movslq(Register dst, Register src) {
5123 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5124 emit_int8(0x63);
5125 emit_int8((unsigned char)(0xC0 | encode));
5126 }
5128 void Assembler::movswq(Register dst, Address src) {
5129 InstructionMark im(this);
5130 prefixq(src, dst);
5131 emit_int8(0x0F);
5132 emit_int8((unsigned char)0xBF);
5133 emit_operand(dst, src);
5134 }
5136 void Assembler::movswq(Register dst, Register src) {
5137 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5138 emit_int8((unsigned char)0x0F);
5139 emit_int8((unsigned char)0xBF);
5140 emit_int8((unsigned char)(0xC0 | encode));
5141 }
5143 void Assembler::movzbq(Register dst, Address src) {
5144 InstructionMark im(this);
5145 prefixq(src, dst);
5146 emit_int8((unsigned char)0x0F);
5147 emit_int8((unsigned char)0xB6);
5148 emit_operand(dst, src);
5149 }
5151 void Assembler::movzbq(Register dst, Register src) {
5152 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5153 emit_int8(0x0F);
5154 emit_int8((unsigned char)0xB6);
5155 emit_int8(0xC0 | encode);
5156 }
5158 void Assembler::movzwq(Register dst, Address src) {
5159 InstructionMark im(this);
5160 prefixq(src, dst);
5161 emit_int8((unsigned char)0x0F);
5162 emit_int8((unsigned char)0xB7);
5163 emit_operand(dst, src);
5164 }
5166 void Assembler::movzwq(Register dst, Register src) {
5167 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5168 emit_int8((unsigned char)0x0F);
5169 emit_int8((unsigned char)0xB7);
5170 emit_int8((unsigned char)(0xC0 | encode));
5171 }
5173 void Assembler::negq(Register dst) {
5174 int encode = prefixq_and_encode(dst->encoding());
5175 emit_int8((unsigned char)0xF7);
5176 emit_int8((unsigned char)(0xD8 | encode));
5177 }
5179 void Assembler::notq(Register dst) {
5180 int encode = prefixq_and_encode(dst->encoding());
5181 emit_int8((unsigned char)0xF7);
5182 emit_int8((unsigned char)(0xD0 | encode));
5183 }
5185 void Assembler::orq(Address dst, int32_t imm32) {
5186 InstructionMark im(this);
5187 prefixq(dst);
5188 emit_int8((unsigned char)0x81);
5189 emit_operand(rcx, dst, 4);
5190 emit_long(imm32);
5191 }
5193 void Assembler::orq(Register dst, int32_t imm32) {
5194 (void) prefixq_and_encode(dst->encoding());
5195 emit_arith(0x81, 0xC8, dst, imm32);
5196 }
5198 void Assembler::orq(Register dst, Address src) {
5199 InstructionMark im(this);
5200 prefixq(src, dst);
5201 emit_int8(0x0B);
5202 emit_operand(dst, src);
5203 }
5205 void Assembler::orq(Register dst, Register src) {
5206 (void) prefixq_and_encode(dst->encoding(), src->encoding());
5207 emit_arith(0x0B, 0xC0, dst, src);
5208 }
5210 void Assembler::popa() { // 64bit
5211 movq(r15, Address(rsp, 0));
5212 movq(r14, Address(rsp, wordSize));
5213 movq(r13, Address(rsp, 2 * wordSize));
5214 movq(r12, Address(rsp, 3 * wordSize));
5215 movq(r11, Address(rsp, 4 * wordSize));
5216 movq(r10, Address(rsp, 5 * wordSize));
5217 movq(r9, Address(rsp, 6 * wordSize));
5218 movq(r8, Address(rsp, 7 * wordSize));
5219 movq(rdi, Address(rsp, 8 * wordSize));
5220 movq(rsi, Address(rsp, 9 * wordSize));
5221 movq(rbp, Address(rsp, 10 * wordSize));
5222 // skip rsp
5223 movq(rbx, Address(rsp, 12 * wordSize));
5224 movq(rdx, Address(rsp, 13 * wordSize));
5225 movq(rcx, Address(rsp, 14 * wordSize));
5226 movq(rax, Address(rsp, 15 * wordSize));
5228 addq(rsp, 16 * wordSize);
5229 }
5231 void Assembler::popcntq(Register dst, Address src) {
5232 assert(VM_Version::supports_popcnt(), "must support");
5233 InstructionMark im(this);
5234 emit_int8((unsigned char)0xF3);
5235 prefixq(src, dst);
5236 emit_int8((unsigned char)0x0F);
5237 emit_int8((unsigned char)0xB8);
5238 emit_operand(dst, src);
5239 }
5241 void Assembler::popcntq(Register dst, Register src) {
5242 assert(VM_Version::supports_popcnt(), "must support");
5243 emit_int8((unsigned char)0xF3);
5244 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5245 emit_int8((unsigned char)0x0F);
5246 emit_int8((unsigned char)0xB8);
5247 emit_int8((unsigned char)(0xC0 | encode));
5248 }
5250 void Assembler::popq(Address dst) {
5251 InstructionMark im(this);
5252 prefixq(dst);
5253 emit_int8((unsigned char)0x8F);
5254 emit_operand(rax, dst);
5255 }
5257 void Assembler::pusha() { // 64bit
5258 // we have to store original rsp. ABI says that 128 bytes
5259 // below rsp are local scratch.
5260 movq(Address(rsp, -5 * wordSize), rsp);
5262 subq(rsp, 16 * wordSize);
5264 movq(Address(rsp, 15 * wordSize), rax);
5265 movq(Address(rsp, 14 * wordSize), rcx);
5266 movq(Address(rsp, 13 * wordSize), rdx);
5267 movq(Address(rsp, 12 * wordSize), rbx);
5268 // skip rsp
5269 movq(Address(rsp, 10 * wordSize), rbp);
5270 movq(Address(rsp, 9 * wordSize), rsi);
5271 movq(Address(rsp, 8 * wordSize), rdi);
5272 movq(Address(rsp, 7 * wordSize), r8);
5273 movq(Address(rsp, 6 * wordSize), r9);
5274 movq(Address(rsp, 5 * wordSize), r10);
5275 movq(Address(rsp, 4 * wordSize), r11);
5276 movq(Address(rsp, 3 * wordSize), r12);
5277 movq(Address(rsp, 2 * wordSize), r13);
5278 movq(Address(rsp, wordSize), r14);
5279 movq(Address(rsp, 0), r15);
5280 }
5282 void Assembler::pushq(Address src) {
5283 InstructionMark im(this);
5284 prefixq(src);
5285 emit_int8((unsigned char)0xFF);
5286 emit_operand(rsi, src);
5287 }
5289 void Assembler::rclq(Register dst, int imm8) {
5290 assert(isShiftCount(imm8 >> 1), "illegal shift count");
5291 int encode = prefixq_and_encode(dst->encoding());
5292 if (imm8 == 1) {
5293 emit_int8((unsigned char)0xD1);
5294 emit_int8((unsigned char)(0xD0 | encode));
5295 } else {
5296 emit_int8((unsigned char)0xC1);
5297 emit_int8((unsigned char)(0xD0 | encode));
5298 emit_int8(imm8);
5299 }
5300 }
5301 void Assembler::sarq(Register dst, int imm8) {
5302 assert(isShiftCount(imm8 >> 1), "illegal shift count");
5303 int encode = prefixq_and_encode(dst->encoding());
5304 if (imm8 == 1) {
5305 emit_int8((unsigned char)0xD1);
5306 emit_int8((unsigned char)(0xF8 | encode));
5307 } else {
5308 emit_int8((unsigned char)0xC1);
5309 emit_int8((unsigned char)(0xF8 | encode));
5310 emit_int8(imm8);
5311 }
5312 }
5314 void Assembler::sarq(Register dst) {
5315 int encode = prefixq_and_encode(dst->encoding());
5316 emit_int8((unsigned char)0xD3);
5317 emit_int8((unsigned char)(0xF8 | encode));
5318 }
5320 void Assembler::sbbq(Address dst, int32_t imm32) {
5321 InstructionMark im(this);
5322 prefixq(dst);
5323 emit_arith_operand(0x81, rbx, dst, imm32);
5324 }
5326 void Assembler::sbbq(Register dst, int32_t imm32) {
5327 (void) prefixq_and_encode(dst->encoding());
5328 emit_arith(0x81, 0xD8, dst, imm32);
5329 }
5331 void Assembler::sbbq(Register dst, Address src) {
5332 InstructionMark im(this);
5333 prefixq(src, dst);
5334 emit_int8(0x1B);
5335 emit_operand(dst, src);
5336 }
5338 void Assembler::sbbq(Register dst, Register src) {
5339 (void) prefixq_and_encode(dst->encoding(), src->encoding());
5340 emit_arith(0x1B, 0xC0, dst, src);
5341 }
5343 void Assembler::shlq(Register dst, int imm8) {
5344 assert(isShiftCount(imm8 >> 1), "illegal shift count");
5345 int encode = prefixq_and_encode(dst->encoding());
5346 if (imm8 == 1) {
5347 emit_int8((unsigned char)0xD1);
5348 emit_int8((unsigned char)(0xE0 | encode));
5349 } else {
5350 emit_int8((unsigned char)0xC1);
5351 emit_int8((unsigned char)(0xE0 | encode));
5352 emit_int8(imm8);
5353 }
5354 }
5356 void Assembler::shlq(Register dst) {
5357 int encode = prefixq_and_encode(dst->encoding());
5358 emit_int8((unsigned char)0xD3);
5359 emit_int8((unsigned char)(0xE0 | encode));
5360 }
5362 void Assembler::shrq(Register dst, int imm8) {
5363 assert(isShiftCount(imm8 >> 1), "illegal shift count");
5364 int encode = prefixq_and_encode(dst->encoding());
5365 emit_int8((unsigned char)0xC1);
5366 emit_int8((unsigned char)(0xE8 | encode));
5367 emit_int8(imm8);
5368 }
5370 void Assembler::shrq(Register dst) {
5371 int encode = prefixq_and_encode(dst->encoding());
5372 emit_int8((unsigned char)0xD3);
5373 emit_int8(0xE8 | encode);
5374 }
5376 void Assembler::subq(Address dst, int32_t imm32) {
5377 InstructionMark im(this);
5378 prefixq(dst);
5379 emit_arith_operand(0x81, rbp, dst, imm32);
5380 }
5382 void Assembler::subq(Address dst, Register src) {
5383 InstructionMark im(this);
5384 prefixq(dst, src);
5385 emit_int8(0x29);
5386 emit_operand(src, dst);
5387 }
5389 void Assembler::subq(Register dst, int32_t imm32) {
5390 (void) prefixq_and_encode(dst->encoding());
5391 emit_arith(0x81, 0xE8, dst, imm32);
5392 }
5394 // Force generation of a 4 byte immediate value even if it fits into 8bit
5395 void Assembler::subq_imm32(Register dst, int32_t imm32) {
5396 (void) prefixq_and_encode(dst->encoding());
5397 emit_arith_imm32(0x81, 0xE8, dst, imm32);
5398 }
5400 void Assembler::subq(Register dst, Address src) {
5401 InstructionMark im(this);
5402 prefixq(src, dst);
5403 emit_int8(0x2B);
5404 emit_operand(dst, src);
5405 }
5407 void Assembler::subq(Register dst, Register src) {
5408 (void) prefixq_and_encode(dst->encoding(), src->encoding());
5409 emit_arith(0x2B, 0xC0, dst, src);
5410 }
5412 void Assembler::testq(Register dst, int32_t imm32) {
5413 // not using emit_arith because test
5414 // doesn't support sign-extension of
5415 // 8bit operands
5416 int encode = dst->encoding();
5417 if (encode == 0) {
5418 prefix(REX_W);
5419 emit_int8((unsigned char)0xA9);
5420 } else {
5421 encode = prefixq_and_encode(encode);
5422 emit_int8((unsigned char)0xF7);
5423 emit_int8((unsigned char)(0xC0 | encode));
5424 }
5425 emit_long(imm32);
5426 }
5428 void Assembler::testq(Register dst, Register src) {
5429 (void) prefixq_and_encode(dst->encoding(), src->encoding());
5430 emit_arith(0x85, 0xC0, dst, src);
5431 }
5433 void Assembler::xaddq(Address dst, Register src) {
5434 InstructionMark im(this);
5435 prefixq(dst, src);
5436 emit_int8(0x0F);
5437 emit_int8((unsigned char)0xC1);
5438 emit_operand(src, dst);
5439 }
5441 void Assembler::xchgq(Register dst, Address src) {
5442 InstructionMark im(this);
5443 prefixq(src, dst);
5444 emit_int8((unsigned char)0x87);
5445 emit_operand(dst, src);
5446 }
5448 void Assembler::xchgq(Register dst, Register src) {
5449 int encode = prefixq_and_encode(dst->encoding(), src->encoding());
5450 emit_int8((unsigned char)0x87);
5451 emit_int8((unsigned char)(0xc0 | encode));
5452 }
5454 void Assembler::xorq(Register dst, Register src) {
5455 (void) prefixq_and_encode(dst->encoding(), src->encoding());
5456 emit_arith(0x33, 0xC0, dst, src);
5457 }
5459 void Assembler::xorq(Register dst, Address src) {
5460 InstructionMark im(this);
5461 prefixq(src, dst);
5462 emit_int8(0x33);
5463 emit_operand(dst, src);
5464 }
5466 #endif // !LP64