-
Notifications
You must be signed in to change notification settings - Fork 224
/
Copy pathassembler_x86.cpp
11119 lines (9786 loc) · 470 KB
/
assembler_x86.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* Copyright (c) 1997, 2021, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*
*/
#include "precompiled.hpp"
#include "asm/assembler.hpp"
#include "asm/assembler.inline.hpp"
#include "gc/shared/cardTableBarrierSet.hpp"
#include "gc/shared/collectedHeap.inline.hpp"
#include "interpreter/interpreter.hpp"
#include "memory/resourceArea.hpp"
#include "prims/methodHandles.hpp"
#include "runtime/biasedLocking.hpp"
#include "runtime/objectMonitor.hpp"
#include "runtime/os.hpp"
#include "runtime/sharedRuntime.hpp"
#include "runtime/stubRoutines.hpp"
#include "runtime/vm_version.hpp"
#include "utilities/macros.hpp"
#ifdef PRODUCT
#define BLOCK_COMMENT(str) /* nothing */
#define STOP(error) stop(error)
#else
#define BLOCK_COMMENT(str) block_comment(str)
#define STOP(error) block_comment(error); stop(error)
#endif
#define BIND(label) bind(label); BLOCK_COMMENT(#label ":")
// Implementation of AddressLiteral
// A 2-D table for managing compressed displacement(disp8) on EVEX enabled platforms.
unsigned char tuple_table[Assembler::EVEX_ETUP + 1][Assembler::AVX_512bit + 1] = {
// -----------------Table 4.5 -------------------- //
16, 32, 64, // EVEX_FV(0)
4, 4, 4, // EVEX_FV(1) - with Evex.b
16, 32, 64, // EVEX_FV(2) - with Evex.w
8, 8, 8, // EVEX_FV(3) - with Evex.w and Evex.b
8, 16, 32, // EVEX_HV(0)
4, 4, 4, // EVEX_HV(1) - with Evex.b
// -----------------Table 4.6 -------------------- //
16, 32, 64, // EVEX_FVM(0)
1, 1, 1, // EVEX_T1S(0)
2, 2, 2, // EVEX_T1S(1)
4, 4, 4, // EVEX_T1S(2)
8, 8, 8, // EVEX_T1S(3)
4, 4, 4, // EVEX_T1F(0)
8, 8, 8, // EVEX_T1F(1)
8, 8, 8, // EVEX_T2(0)
0, 16, 16, // EVEX_T2(1)
0, 16, 16, // EVEX_T4(0)
0, 0, 32, // EVEX_T4(1)
0, 0, 32, // EVEX_T8(0)
8, 16, 32, // EVEX_HVM(0)
4, 8, 16, // EVEX_QVM(0)
2, 4, 8, // EVEX_OVM(0)
16, 16, 16, // EVEX_M128(0)
8, 32, 64, // EVEX_DUP(0)
0, 0, 0 // EVEX_NTUP
};
AddressLiteral::AddressLiteral(address target, relocInfo::relocType rtype) {
_is_lval = false;
_target = target;
switch (rtype) {
case relocInfo::oop_type:
case relocInfo::metadata_type:
// Oops are a special case. Normally they would be their own section
// but in cases like icBuffer they are literals in the code stream that
// we don't have a section for. We use none so that we get a literal address
// which is always patchable.
break;
case relocInfo::external_word_type:
_rspec = external_word_Relocation::spec(target);
break;
case relocInfo::internal_word_type:
_rspec = internal_word_Relocation::spec(target);
break;
case relocInfo::opt_virtual_call_type:
_rspec = opt_virtual_call_Relocation::spec();
break;
case relocInfo::static_call_type:
_rspec = static_call_Relocation::spec();
break;
case relocInfo::runtime_call_type:
_rspec = runtime_call_Relocation::spec();
break;
case relocInfo::poll_type:
case relocInfo::poll_return_type:
_rspec = Relocation::spec_simple(rtype);
break;
case relocInfo::none:
break;
default:
ShouldNotReachHere();
break;
}
}
// Implementation of Address
#ifdef _LP64
Address Address::make_array(ArrayAddress adr) {
// Not implementable on 64bit machines
// Should have been handled higher up the call chain.
ShouldNotReachHere();
return Address();
}
// exceedingly dangerous constructor
Address::Address(int disp, address loc, relocInfo::relocType rtype) {
_base = noreg;
_index = noreg;
_scale = no_scale;
_disp = disp;
_xmmindex = xnoreg;
_isxmmindex = false;
switch (rtype) {
case relocInfo::external_word_type:
_rspec = external_word_Relocation::spec(loc);
break;
case relocInfo::internal_word_type:
_rspec = internal_word_Relocation::spec(loc);
break;
case relocInfo::runtime_call_type:
// HMM
_rspec = runtime_call_Relocation::spec();
break;
case relocInfo::poll_type:
case relocInfo::poll_return_type:
_rspec = Relocation::spec_simple(rtype);
break;
case relocInfo::none:
break;
default:
ShouldNotReachHere();
}
}
#else // LP64
Address Address::make_array(ArrayAddress adr) {
AddressLiteral base = adr.base();
Address index = adr.index();
assert(index._disp == 0, "must not have disp"); // maybe it can?
Address array(index._base, index._index, index._scale, (intptr_t) base.target());
array._rspec = base._rspec;
return array;
}
// exceedingly dangerous constructor
Address::Address(address loc, RelocationHolder spec) {
_base = noreg;
_index = noreg;
_scale = no_scale;
_disp = (intptr_t) loc;
_rspec = spec;
_xmmindex = xnoreg;
_isxmmindex = false;
}
#endif // _LP64
// Convert the raw encoding form into the form expected by the constructor for
// Address. An index of 4 (rsp) corresponds to having no index, so convert
// that to noreg for the Address constructor.
Address Address::make_raw(int base, int index, int scale, int disp, relocInfo::relocType disp_reloc) {
RelocationHolder rspec = RelocationHolder::none;
if (disp_reloc != relocInfo::none) {
rspec = Relocation::spec_simple(disp_reloc);
}
bool valid_index = index != rsp->encoding();
if (valid_index) {
Address madr(as_Register(base), as_Register(index), (Address::ScaleFactor)scale, in_ByteSize(disp));
madr._rspec = rspec;
return madr;
} else {
Address madr(as_Register(base), noreg, Address::no_scale, in_ByteSize(disp));
madr._rspec = rspec;
return madr;
}
}
// Implementation of Assembler
int AbstractAssembler::code_fill_byte() {
return (u_char)'\xF4'; // hlt
}
void Assembler::init_attributes(void) {
_legacy_mode_bw = (VM_Version::supports_avx512bw() == false);
_legacy_mode_dq = (VM_Version::supports_avx512dq() == false);
_legacy_mode_vl = (VM_Version::supports_avx512vl() == false);
_legacy_mode_vlbw = (VM_Version::supports_avx512vlbw() == false);
NOT_LP64(_is_managed = false;)
_attributes = NULL;
}
void Assembler::set_attributes(InstructionAttr* attributes) {
// Record the assembler in the attributes, so the attributes destructor can
// clear the assembler's attributes, cleaning up the otherwise dangling
// pointer. gcc13 has a false positive warning, because it doesn't tie that
// cleanup to the assignment of _attributes here.
attributes->set_current_assembler(this);
PRAGMA_DIAG_PUSH
PRAGMA_DANGLING_POINTER_IGNORED
_attributes = attributes;
PRAGMA_DIAG_POP
}
void Assembler::membar(Membar_mask_bits order_constraint) {
// We only have to handle StoreLoad
if (order_constraint & StoreLoad) {
// All usable chips support "locked" instructions which suffice
// as barriers, and are much faster than the alternative of
// using cpuid instruction. We use here a locked add [esp-C],0.
// This is conveniently otherwise a no-op except for blowing
// flags, and introducing a false dependency on target memory
// location. We can't do anything with flags, but we can avoid
// memory dependencies in the current method by locked-adding
// somewhere else on the stack. Doing [esp+C] will collide with
// something on stack in current method, hence we go for [esp-C].
// It is convenient since it is almost always in data cache, for
// any small C. We need to step back from SP to avoid data
// dependencies with other things on below SP (callee-saves, for
// example). Without a clear way to figure out the minimal safe
// distance from SP, it makes sense to step back the complete
// cache line, as this will also avoid possible second-order effects
// with locked ops against the cache line. Our choice of offset
// is bounded by x86 operand encoding, which should stay within
// [-128; +127] to have the 8-byte displacement encoding.
//
// Any change to this code may need to revisit other places in
// the code where this idiom is used, in particular the
// orderAccess code.
int offset = -VM_Version::L1_line_size();
if (offset < -128) {
offset = -128;
}
lock();
addl(Address(rsp, offset), 0);// Assert the lock# signal here
}
}
// make this go away someday
void Assembler::emit_data(jint data, relocInfo::relocType rtype, int format) {
if (rtype == relocInfo::none)
emit_int32(data);
else
emit_data(data, Relocation::spec_simple(rtype), format);
}
void Assembler::emit_data(jint data, RelocationHolder const& rspec, int format) {
assert(imm_operand == 0, "default format must be immediate in this file");
assert(inst_mark() != NULL, "must be inside InstructionMark");
if (rspec.type() != relocInfo::none) {
#ifdef ASSERT
check_relocation(rspec, format);
#endif
// Do not use AbstractAssembler::relocate, which is not intended for
// embedded words. Instead, relocate to the enclosing instruction.
// hack. call32 is too wide for mask so use disp32
if (format == call32_operand)
code_section()->relocate(inst_mark(), rspec, disp32_operand);
else
code_section()->relocate(inst_mark(), rspec, format);
}
emit_int32(data);
}
static int encode(Register r) {
int enc = r->encoding();
if (enc >= 8) {
enc -= 8;
}
return enc;
}
void Assembler::emit_arith_b(int op1, int op2, Register dst, int imm8) {
assert(dst->has_byte_register(), "must have byte register");
assert(isByte(op1) && isByte(op2), "wrong opcode");
assert(isByte(imm8), "not a byte");
assert((op1 & 0x01) == 0, "should be 8bit operation");
emit_int24(op1, (op2 | encode(dst)), imm8);
}
void Assembler::emit_arith(int op1, int op2, Register dst, int32_t imm32) {
assert(isByte(op1) && isByte(op2), "wrong opcode");
assert((op1 & 0x01) == 1, "should be 32bit operation");
assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
if (is8bit(imm32)) {
emit_int24(op1 | 0x02, // set sign bit
op2 | encode(dst),
imm32 & 0xFF);
} else {
emit_int16(op1, (op2 | encode(dst)));
emit_int32(imm32);
}
}
// Force generation of a 4 byte immediate value even if it fits into 8bit
void Assembler::emit_arith_imm32(int op1, int op2, Register dst, int32_t imm32) {
assert(isByte(op1) && isByte(op2), "wrong opcode");
assert((op1 & 0x01) == 1, "should be 32bit operation");
assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
emit_int16(op1, (op2 | encode(dst)));
emit_int32(imm32);
}
// immediate-to-memory forms
void Assembler::emit_arith_operand(int op1, Register rm, Address adr, int32_t imm32) {
assert((op1 & 0x01) == 1, "should be 32bit operation");
assert((op1 & 0x02) == 0, "sign-extension bit should not be set");
if (is8bit(imm32)) {
emit_int8(op1 | 0x02); // set sign bit
emit_operand(rm, adr, 1);
emit_int8(imm32 & 0xFF);
} else {
emit_int8(op1);
emit_operand(rm, adr, 4);
emit_int32(imm32);
}
}
void Assembler::emit_arith(int op1, int op2, Register dst, Register src) {
assert(isByte(op1) && isByte(op2), "wrong opcode");
emit_int16(op1, (op2 | encode(dst) << 3 | encode(src)));
}
bool Assembler::query_compressed_disp_byte(int disp, bool is_evex_inst, int vector_len,
int cur_tuple_type, int in_size_in_bits, int cur_encoding) {
int mod_idx = 0;
// We will test if the displacement fits the compressed format and if so
// apply the compression to the displacment iff the result is8bit.
if (VM_Version::supports_evex() && is_evex_inst) {
switch (cur_tuple_type) {
case EVEX_FV:
if ((cur_encoding & VEX_W) == VEX_W) {
mod_idx = ((cur_encoding & EVEX_Rb) == EVEX_Rb) ? 3 : 2;
} else {
mod_idx = ((cur_encoding & EVEX_Rb) == EVEX_Rb) ? 1 : 0;
}
break;
case EVEX_HV:
mod_idx = ((cur_encoding & EVEX_Rb) == EVEX_Rb) ? 1 : 0;
break;
case EVEX_FVM:
break;
case EVEX_T1S:
switch (in_size_in_bits) {
case EVEX_8bit:
break;
case EVEX_16bit:
mod_idx = 1;
break;
case EVEX_32bit:
mod_idx = 2;
break;
case EVEX_64bit:
mod_idx = 3;
break;
}
break;
case EVEX_T1F:
case EVEX_T2:
case EVEX_T4:
mod_idx = (in_size_in_bits == EVEX_64bit) ? 1 : 0;
break;
case EVEX_T8:
break;
case EVEX_HVM:
break;
case EVEX_QVM:
break;
case EVEX_OVM:
break;
case EVEX_M128:
break;
case EVEX_DUP:
break;
default:
assert(0, "no valid evex tuple_table entry");
break;
}
if (vector_len >= AVX_128bit && vector_len <= AVX_512bit) {
int disp_factor = tuple_table[cur_tuple_type + mod_idx][vector_len];
if ((disp % disp_factor) == 0) {
int new_disp = disp / disp_factor;
if ((-0x80 <= new_disp && new_disp < 0x80)) {
disp = new_disp;
}
} else {
return false;
}
}
}
return (-0x80 <= disp && disp < 0x80);
}
bool Assembler::emit_compressed_disp_byte(int &disp) {
int mod_idx = 0;
// We will test if the displacement fits the compressed format and if so
// apply the compression to the displacment iff the result is8bit.
if (VM_Version::supports_evex() && _attributes && _attributes->is_evex_instruction()) {
int evex_encoding = _attributes->get_evex_encoding();
int tuple_type = _attributes->get_tuple_type();
switch (tuple_type) {
case EVEX_FV:
if ((evex_encoding & VEX_W) == VEX_W) {
mod_idx = ((evex_encoding & EVEX_Rb) == EVEX_Rb) ? 3 : 2;
} else {
mod_idx = ((evex_encoding & EVEX_Rb) == EVEX_Rb) ? 1 : 0;
}
break;
case EVEX_HV:
mod_idx = ((evex_encoding & EVEX_Rb) == EVEX_Rb) ? 1 : 0;
break;
case EVEX_FVM:
break;
case EVEX_T1S:
switch (_attributes->get_input_size()) {
case EVEX_8bit:
break;
case EVEX_16bit:
mod_idx = 1;
break;
case EVEX_32bit:
mod_idx = 2;
break;
case EVEX_64bit:
mod_idx = 3;
break;
}
break;
case EVEX_T1F:
case EVEX_T2:
case EVEX_T4:
mod_idx = (_attributes->get_input_size() == EVEX_64bit) ? 1 : 0;
break;
case EVEX_T8:
break;
case EVEX_HVM:
break;
case EVEX_QVM:
break;
case EVEX_OVM:
break;
case EVEX_M128:
break;
case EVEX_DUP:
break;
default:
assert(0, "no valid evex tuple_table entry");
break;
}
int vector_len = _attributes->get_vector_len();
if (vector_len >= AVX_128bit && vector_len <= AVX_512bit) {
int disp_factor = tuple_table[tuple_type + mod_idx][vector_len];
if ((disp % disp_factor) == 0) {
int new_disp = disp / disp_factor;
if (is8bit(new_disp)) {
disp = new_disp;
}
} else {
return false;
}
}
}
return is8bit(disp);
}
static bool is_valid_encoding(int reg_enc) {
return reg_enc >= 0;
}
static int raw_encode(Register reg) {
assert(reg == noreg || reg->is_valid(), "sanity");
int reg_enc = (intptr_t)reg;
assert(reg_enc == -1 || is_valid_encoding(reg_enc), "sanity");
return reg_enc;
}
static int raw_encode(XMMRegister xmmreg) {
assert(xmmreg == xnoreg || xmmreg->is_valid(), "sanity");
int xmmreg_enc = (intptr_t)xmmreg;
assert(xmmreg_enc == -1 || is_valid_encoding(xmmreg_enc), "sanity");
return xmmreg_enc;
}
static int modrm_encoding(int mod, int dst_enc, int src_enc) {
return (mod & 3) << 6 | (dst_enc & 7) << 3 | (src_enc & 7);
}
static int sib_encoding(Address::ScaleFactor scale, int index_enc, int base_enc) {
return (scale & 3) << 6 | (index_enc & 7) << 3 | (base_enc & 7);
}
inline void Assembler::emit_modrm(int mod, int dst_enc, int src_enc) {
assert((mod & 3) != 0b11, "forbidden");
int modrm = modrm_encoding(mod, dst_enc, src_enc);
emit_int8(modrm);
}
inline void Assembler::emit_modrm_disp8(int mod, int dst_enc, int src_enc,
int disp) {
int modrm = modrm_encoding(mod, dst_enc, src_enc);
emit_int16(modrm, disp & 0xFF);
}
inline void Assembler::emit_modrm_sib(int mod, int dst_enc, int src_enc,
Address::ScaleFactor scale, int index_enc, int base_enc) {
int modrm = modrm_encoding(mod, dst_enc, src_enc);
int sib = sib_encoding(scale, index_enc, base_enc);
emit_int16(modrm, sib);
}
inline void Assembler::emit_modrm_sib_disp8(int mod, int dst_enc, int src_enc,
Address::ScaleFactor scale, int index_enc, int base_enc,
int disp) {
int modrm = modrm_encoding(mod, dst_enc, src_enc);
int sib = sib_encoding(scale, index_enc, base_enc);
emit_int24(modrm, sib, disp & 0xFF);
}
void Assembler::emit_operand_helper(int reg_enc, int base_enc, int index_enc,
Address::ScaleFactor scale, int disp,
RelocationHolder const& rspec,
int rip_relative_correction) {
bool no_relocation = (rspec.type() == relocInfo::none);
if (is_valid_encoding(base_enc)) {
if (is_valid_encoding(index_enc)) {
assert(scale != Address::no_scale, "inconsistent address");
// [base + index*scale + disp]
if (disp == 0 && no_relocation &&
base_enc != rbp->encoding() LP64_ONLY(&& base_enc != r13->encoding())) {
// [base + index*scale]
// [00 reg 100][ss index base]
emit_modrm_sib(0b00, reg_enc, 0b100,
scale, index_enc, base_enc);
} else if (emit_compressed_disp_byte(disp) && no_relocation) {
// [base + index*scale + imm8]
// [01 reg 100][ss index base] imm8
emit_modrm_sib_disp8(0b01, reg_enc, 0b100,
scale, index_enc, base_enc,
disp);
} else {
// [base + index*scale + disp32]
// [10 reg 100][ss index base] disp32
emit_modrm_sib(0b10, reg_enc, 0b100,
scale, index_enc, base_enc);
emit_data(disp, rspec, disp32_operand);
}
} else if (base_enc == rsp->encoding() LP64_ONLY(|| base_enc == r12->encoding())) {
// [rsp + disp]
if (disp == 0 && no_relocation) {
// [rsp]
// [00 reg 100][00 100 100]
emit_modrm_sib(0b00, reg_enc, 0b100,
Address::times_1, 0b100, 0b100);
} else if (emit_compressed_disp_byte(disp) && no_relocation) {
// [rsp + imm8]
// [01 reg 100][00 100 100] disp8
emit_modrm_sib_disp8(0b01, reg_enc, 0b100,
Address::times_1, 0b100, 0b100,
disp);
} else {
// [rsp + imm32]
// [10 reg 100][00 100 100] disp32
emit_modrm_sib(0b10, reg_enc, 0b100,
Address::times_1, 0b100, 0b100);
emit_data(disp, rspec, disp32_operand);
}
} else {
// [base + disp]
assert(base_enc != rsp->encoding() LP64_ONLY(&& base_enc != r12->encoding()), "illegal addressing mode");
if (disp == 0 && no_relocation &&
base_enc != rbp->encoding() LP64_ONLY(&& base_enc != r13->encoding())) {
// [base]
// [00 reg base]
emit_modrm(0, reg_enc, base_enc);
} else if (emit_compressed_disp_byte(disp) && no_relocation) {
// [base + disp8]
// [01 reg base] disp8
emit_modrm_disp8(0b01, reg_enc, base_enc,
disp);
} else {
// [base + disp32]
// [10 reg base] disp32
emit_modrm(0b10, reg_enc, base_enc);
emit_data(disp, rspec, disp32_operand);
}
}
} else {
if (is_valid_encoding(index_enc)) {
assert(scale != Address::no_scale, "inconsistent address");
// base == noreg
// [index*scale + disp]
// [00 reg 100][ss index 101] disp32
emit_modrm_sib(0b00, reg_enc, 0b100,
scale, index_enc, 0b101 /* no base */);
emit_data(disp, rspec, disp32_operand);
} else if (!no_relocation) {
// base == noreg, index == noreg
// [disp] (64bit) RIP-RELATIVE (32bit) abs
// [00 reg 101] disp32
emit_modrm(0b00, reg_enc, 0b101 /* no base */);
// Note that the RIP-rel. correction applies to the generated
// disp field, but _not_ to the target address in the rspec.
// disp was created by converting the target address minus the pc
// at the start of the instruction. That needs more correction here.
// intptr_t disp = target - next_ip;
assert(inst_mark() != NULL, "must be inside InstructionMark");
address next_ip = pc() + sizeof(int32_t) + rip_relative_correction;
int64_t adjusted = disp;
// Do rip-rel adjustment for 64bit
LP64_ONLY(adjusted -= (next_ip - inst_mark()));
assert(is_simm32(adjusted),
"must be 32bit offset (RIP relative address)");
emit_data((int32_t) adjusted, rspec, disp32_operand);
} else {
// base == noreg, index == noreg, no_relocation == true
// 32bit never did this, did everything as the rip-rel/disp code above
// [disp] ABSOLUTE
// [00 reg 100][00 100 101] disp32
emit_modrm_sib(0b00, reg_enc, 0b100 /* no base */,
Address::times_1, 0b100, 0b101);
emit_data(disp, rspec, disp32_operand);
}
}
}
void Assembler::emit_operand(Register reg, Register base, Register index,
Address::ScaleFactor scale, int disp,
RelocationHolder const& rspec,
int rip_relative_correction) {
assert(!index->is_valid() || index != rsp, "illegal addressing mode");
emit_operand_helper(raw_encode(reg), raw_encode(base), raw_encode(index),
scale, disp, rspec, rip_relative_correction);
}
void Assembler::emit_operand(XMMRegister xmmreg, Register base, Register index,
Address::ScaleFactor scale, int disp,
RelocationHolder const& rspec) {
assert(!index->is_valid() || index != rsp, "illegal addressing mode");
assert(xmmreg->encoding() < 16 || UseAVX > 2, "not supported");
emit_operand_helper(raw_encode(xmmreg), raw_encode(base), raw_encode(index),
scale, disp, rspec);
}
void Assembler::emit_operand(XMMRegister xmmreg, Register base, XMMRegister xmmindex,
Address::ScaleFactor scale, int disp,
RelocationHolder const& rspec) {
assert(xmmreg->encoding() < 16 || UseAVX > 2, "not supported");
assert(xmmindex->encoding() < 16 || UseAVX > 2, "not supported");
emit_operand_helper(raw_encode(xmmreg), raw_encode(base), raw_encode(xmmindex),
scale, disp, rspec, /* rip_relative_correction */ 0);
}
// Secret local extension to Assembler::WhichOperand:
#define end_pc_operand (_WhichOperand_limit)
address Assembler::locate_operand(address inst, WhichOperand which) {
// Decode the given instruction, and return the address of
// an embedded 32-bit operand word.
// If "which" is disp32_operand, selects the displacement portion
// of an effective address specifier.
// If "which" is imm64_operand, selects the trailing immediate constant.
// If "which" is call32_operand, selects the displacement of a call or jump.
// Caller is responsible for ensuring that there is such an operand,
// and that it is 32/64 bits wide.
// If "which" is end_pc_operand, find the end of the instruction.
address ip = inst;
bool is_64bit = false;
debug_only(bool has_disp32 = false);
int tail_size = 0; // other random bytes (#32, #16, etc.) at end of insn
again_after_prefix:
switch (0xFF & *ip++) {
// These convenience macros generate groups of "case" labels for the switch.
#define REP4(x) (x)+0: case (x)+1: case (x)+2: case (x)+3
#define REP8(x) (x)+0: case (x)+1: case (x)+2: case (x)+3: \
case (x)+4: case (x)+5: case (x)+6: case (x)+7
#define REP16(x) REP8((x)+0): \
case REP8((x)+8)
case CS_segment:
case SS_segment:
case DS_segment:
case ES_segment:
case FS_segment:
case GS_segment:
// Seems dubious
LP64_ONLY(assert(false, "shouldn't have that prefix"));
assert(ip == inst+1, "only one prefix allowed");
goto again_after_prefix;
case 0x67:
case REX:
case REX_B:
case REX_X:
case REX_XB:
case REX_R:
case REX_RB:
case REX_RX:
case REX_RXB:
NOT_LP64(assert(false, "64bit prefixes"));
goto again_after_prefix;
case REX_W:
case REX_WB:
case REX_WX:
case REX_WXB:
case REX_WR:
case REX_WRB:
case REX_WRX:
case REX_WRXB:
NOT_LP64(assert(false, "64bit prefixes"));
is_64bit = true;
goto again_after_prefix;
case 0xFF: // pushq a; decl a; incl a; call a; jmp a
case 0x88: // movb a, r
case 0x89: // movl a, r
case 0x8A: // movb r, a
case 0x8B: // movl r, a
case 0x8F: // popl a
debug_only(has_disp32 = true);
break;
case 0x68: // pushq #32
if (which == end_pc_operand) {
return ip + 4;
}
assert(which == imm_operand && !is_64bit, "pushl has no disp32 or 64bit immediate");
return ip; // not produced by emit_operand
case 0x66: // movw ... (size prefix)
again_after_size_prefix2:
switch (0xFF & *ip++) {
case REX:
case REX_B:
case REX_X:
case REX_XB:
case REX_R:
case REX_RB:
case REX_RX:
case REX_RXB:
case REX_W:
case REX_WB:
case REX_WX:
case REX_WXB:
case REX_WR:
case REX_WRB:
case REX_WRX:
case REX_WRXB:
NOT_LP64(assert(false, "64bit prefix found"));
goto again_after_size_prefix2;
case 0x8B: // movw r, a
case 0x89: // movw a, r
debug_only(has_disp32 = true);
break;
case 0xC7: // movw a, #16
debug_only(has_disp32 = true);
tail_size = 2; // the imm16
break;
case 0x0F: // several SSE/SSE2 variants
ip--; // reparse the 0x0F
goto again_after_prefix;
default:
ShouldNotReachHere();
}
break;
case REP8(0xB8): // movl/q r, #32/#64(oop?)
if (which == end_pc_operand) return ip + (is_64bit ? 8 : 4);
// these asserts are somewhat nonsensical
#ifndef _LP64
assert(which == imm_operand || which == disp32_operand,
"which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, p2i(ip));
#else
assert((which == call32_operand || which == imm_operand) && is_64bit ||
which == narrow_oop_operand && !is_64bit,
"which %d is_64_bit %d ip " INTPTR_FORMAT, which, is_64bit, p2i(ip));
#endif // _LP64
return ip;
case 0x69: // imul r, a, #32
case 0xC7: // movl a, #32(oop?)
tail_size = 4;
debug_only(has_disp32 = true); // has both kinds of operands!
break;
case 0x0F: // movx..., etc.
switch (0xFF & *ip++) {
case 0x3A: // pcmpestri
tail_size = 1;
case 0x38: // ptest, pmovzxbw
ip++; // skip opcode
debug_only(has_disp32 = true); // has both kinds of operands!
break;
case 0x70: // pshufd r, r/a, #8
debug_only(has_disp32 = true); // has both kinds of operands!
case 0x73: // psrldq r, #8
tail_size = 1;
break;
case 0x12: // movlps
case 0x28: // movaps
case 0x2E: // ucomiss
case 0x2F: // comiss
case 0x54: // andps
case 0x55: // andnps
case 0x56: // orps
case 0x57: // xorps
case 0x58: // addpd
case 0x59: // mulpd
case 0x6E: // movd
case 0x7E: // movd
case 0x6F: // movdq
case 0x7F: // movdq
case 0xAE: // ldmxcsr, stmxcsr, fxrstor, fxsave, clflush
case 0xFE: // paddd
debug_only(has_disp32 = true);
break;
case 0xAD: // shrd r, a, %cl
case 0xAF: // imul r, a
case 0xBE: // movsbl r, a (movsxb)
case 0xBF: // movswl r, a (movsxw)
case 0xB6: // movzbl r, a (movzxb)
case 0xB7: // movzwl r, a (movzxw)
case REP16(0x40): // cmovl cc, r, a
case 0xB0: // cmpxchgb
case 0xB1: // cmpxchg
case 0xC1: // xaddl
case 0xC7: // cmpxchg8
case REP16(0x90): // setcc a
debug_only(has_disp32 = true);
// fall out of the switch to decode the address
break;
case 0xC4: // pinsrw r, a, #8
debug_only(has_disp32 = true);
case 0xC5: // pextrw r, r, #8
tail_size = 1; // the imm8
break;
case 0xAC: // shrd r, a, #8
debug_only(has_disp32 = true);
tail_size = 1; // the imm8
break;
case REP16(0x80): // jcc rdisp32
if (which == end_pc_operand) return ip + 4;
assert(which == call32_operand, "jcc has no disp32 or imm");
return ip;
default:
ShouldNotReachHere();
}
break;
case 0x81: // addl a, #32; addl r, #32
// also: orl, adcl, sbbl, andl, subl, xorl, cmpl
// on 32bit in the case of cmpl, the imm might be an oop
tail_size = 4;
debug_only(has_disp32 = true); // has both kinds of operands!
break;
case 0x83: // addl a, #8; addl r, #8
// also: orl, adcl, sbbl, andl, subl, xorl, cmpl
debug_only(has_disp32 = true); // has both kinds of operands!
tail_size = 1;
break;
case 0x9B:
switch (0xFF & *ip++) {
case 0xD9: // fnstcw a
debug_only(has_disp32 = true);
break;
default:
ShouldNotReachHere();
}
break;
case REP4(0x00): // addb a, r; addl a, r; addb r, a; addl r, a
case REP4(0x10): // adc...
case REP4(0x20): // and...
case REP4(0x30): // xor...
case REP4(0x08): // or...
case REP4(0x18): // sbb...
case REP4(0x28): // sub...
case 0xF7: // mull a
case 0x8D: // lea r, a
case 0x87: // xchg r, a
case REP4(0x38): // cmp...
case 0x85: // test r, a
debug_only(has_disp32 = true); // has both kinds of operands!
break;
case 0xC1: // sal a, #8; sar a, #8; shl a, #8; shr a, #8
case 0xC6: // movb a, #8
case 0x80: // cmpb a, #8
case 0x6B: // imul r, a, #8
debug_only(has_disp32 = true); // has both kinds of operands!
tail_size = 1; // the imm8
break;
case 0xC4: // VEX_3bytes
case 0xC5: // VEX_2bytes
assert((UseAVX > 0), "shouldn't have VEX prefix");
assert(ip == inst+1, "no prefixes allowed");
// C4 and C5 are also used as opcodes for PINSRW and PEXTRW instructions
// but they have prefix 0x0F and processed when 0x0F processed above.
//
// In 32-bit mode the VEX first byte C4 and C5 alias onto LDS and LES
// instructions (these instructions are not supported in 64-bit mode).
// To distinguish them bits [7:6] are set in the VEX second byte since
// ModRM byte can not be of the form 11xxxxxx in 32-bit mode. To set
// those VEX bits REX and vvvv bits are inverted.
//
// Fortunately C2 doesn't generate these instructions so we don't need
// to check for them in product version.
// Check second byte
NOT_LP64(assert((0xC0 & *ip) == 0xC0, "shouldn't have LDS and LES instructions"));
int vex_opcode;
// First byte
if ((0xFF & *inst) == VEX_3bytes) {
vex_opcode = VEX_OPCODE_MASK & *ip;
ip++; // third byte