-
-
Notifications
You must be signed in to change notification settings - Fork 606
/
cod1.d
5659 lines (5216 loc) · 189 KB
/
cod1.d
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
/**
* Compiler implementation of the
* $(LINK2 http://www.dlang.org, D programming language).
*
* Copyright: Copyright (C) 1984-1998 by Symantec
* Copyright (C) 2000-2021 by The D Language Foundation, All Rights Reserved
* Authors: $(LINK2 http://www.digitalmars.com, Walter Bright)
* License: $(LINK2 http://www.boost.org/LICENSE_1_0.txt, Boost License 1.0)
* Source: $(LINK2 https://github.com/dlang/dmd/blob/master/src/dmd/backend/cod1.d, backend/cod1.d)
* Coverage: https://codecov.io/gh/dlang/dmd/src/master/src/dmd/backend/cod1.d
*/
module dmd.backend.cod1;
version (SCPP)
version = COMPILE;
version (MARS)
version = COMPILE;
version (COMPILE)
{
import core.stdc.stdio;
import core.stdc.stdlib;
import core.stdc.string;
import dmd.backend.backend;
import dmd.backend.cc;
import dmd.backend.cdef;
import dmd.backend.code;
import dmd.backend.code_x86;
import dmd.backend.codebuilder;
import dmd.backend.mem;
import dmd.backend.el;
import dmd.backend.exh;
import dmd.backend.global;
import dmd.backend.obj;
import dmd.backend.oper;
import dmd.backend.rtlsym;
import dmd.backend.ty;
import dmd.backend.type;
import dmd.backend.xmm;
extern (C++):
nothrow:
@safe:
int REGSIZE();
extern __gshared CGstate cgstate;
extern __gshared ubyte[FLMAX] segfl;
extern __gshared bool[FLMAX] stackfl;
private extern (D) uint mask(uint m) { return 1 << m; }
private void genorreg(ref CodeBuilder c, uint t, uint f) { genregs(c, 0x09, f, t); }
/* array to convert from index register to r/m field */
/* AX CX DX BX SP BP SI DI */
private __gshared const byte[8] regtorm32 = [ 0, 1, 2, 3,-1, 5, 6, 7 ];
__gshared const byte[8] regtorm = [ -1,-1,-1, 7,-1, 6, 4, 5 ];
//void funccall(ref CodeBuilder cdb,elem *e,uint numpara,uint numalign,
// regm_t *pretregs,regm_t keepmsk, bool usefuncarg);
/*********************************
* Determine if we should leave parameter `s` in the register it
* came in, or allocate a register it using the register
* allocator.
* Params:
* s = parameter Symbol
* Returns:
* `true` if `s` is a register parameter and leave it in the register it came in
*/
@trusted
bool regParamInPreg(Symbol* s)
{
//printf("regPAramInPreg %s\n", s.Sident.ptr);
return (s.Sclass == SCfastpar || s.Sclass == SCshadowreg) &&
(!(config.flags4 & CFG4optimized) || !(s.Sflags & GTregcand));
}
/**************************
* Determine if e is a 32 bit scaled index addressing mode.
* Returns:
* 0 not a scaled index addressing mode
* !=0 the value for ss in the SIB byte
*/
@trusted
int isscaledindex(elem *e)
{
targ_uns ss;
assert(!I16);
while (e.Eoper == OPcomma)
e = e.EV.E2;
if (!(e.Eoper == OPshl && !e.Ecount &&
e.EV.E2.Eoper == OPconst &&
(ss = e.EV.E2.EV.Vuns) <= 3
)
)
ss = 0;
return ss;
}
/*********************************************
* Generate code for which isscaledindex(e) returned a non-zero result.
*/
@trusted
/*private*/ void cdisscaledindex(ref CodeBuilder cdb,elem *e,regm_t *pidxregs,regm_t keepmsk)
{
// Load index register with result of e.EV.E1
while (e.Eoper == OPcomma)
{
regm_t r = 0;
scodelem(cdb, e.EV.E1, &r, keepmsk, true);
freenode(e);
e = e.EV.E2;
}
assert(e.Eoper == OPshl);
scodelem(cdb, e.EV.E1, pidxregs, keepmsk, true);
freenode(e.EV.E2);
freenode(e);
}
/***********************************
* Determine index if we can do two LEA instructions as a multiply.
* Returns:
* 0 can't do it
*/
enum
{
SSFLnobp = 1, /// can't have EBP in relconst
SSFLnobase1 = 2, /// no base register for first LEA
SSFLnobase = 4, /// no base register
SSFLlea = 8, /// can do it in one LEA
}
struct Ssindex
{
targ_uns product;
ubyte ss1;
ubyte ss2;
ubyte ssflags; /// SSFLxxxx
}
private __gshared const Ssindex[21] ssindex_array =
[
{ 0, 0, 0 }, // [0] is a place holder
{ 3, 1, 0, SSFLnobp | SSFLlea },
{ 5, 2, 0, SSFLnobp | SSFLlea },
{ 9, 3, 0, SSFLnobp | SSFLlea },
{ 6, 1, 1, SSFLnobase },
{ 12, 1, 2, SSFLnobase },
{ 24, 1, 3, SSFLnobase },
{ 10, 2, 1, SSFLnobase },
{ 20, 2, 2, SSFLnobase },
{ 40, 2, 3, SSFLnobase },
{ 18, 3, 1, SSFLnobase },
{ 36, 3, 2, SSFLnobase },
{ 72, 3, 3, SSFLnobase },
{ 15, 2, 1, SSFLnobp },
{ 25, 2, 2, SSFLnobp },
{ 27, 3, 1, SSFLnobp },
{ 45, 3, 2, SSFLnobp },
{ 81, 3, 3, SSFLnobp },
{ 16, 3, 1, SSFLnobase1 | SSFLnobase },
{ 32, 3, 2, SSFLnobase1 | SSFLnobase },
{ 64, 3, 3, SSFLnobase1 | SSFLnobase },
];
int ssindex(OPER op,targ_uns product)
{
if (op == OPshl)
product = 1 << product;
for (size_t i = 1; i < ssindex_array.length; i++)
{
if (ssindex_array[i].product == product)
return cast(int)i;
}
return 0;
}
/***************************************
* Build an EA of the form disp[base][index*scale].
* Input:
* c struct to fill in
* base base register (-1 if none)
* index index register (-1 if none)
* scale scale factor - 1,2,4,8
* disp displacement
*/
void buildEA(code *c,int base,int index,int scale,targ_size_t disp)
{
ubyte rm;
ubyte sib;
ubyte rex = 0;
sib = 0;
if (!I16)
{ uint ss;
assert(index != SP);
switch (scale)
{ case 1: ss = 0; break;
case 2: ss = 1; break;
case 4: ss = 2; break;
case 8: ss = 3; break;
default: assert(0);
}
if (base == -1)
{
if (index == -1)
rm = modregrm(0,0,5);
else
{
rm = modregrm(0,0,4);
sib = modregrm(ss,index & 7,5);
if (index & 8)
rex |= REX_X;
}
}
else if (index == -1)
{
if (base == SP)
{
rm = modregrm(2, 0, 4);
sib = modregrm(0, 4, SP);
}
else
{ rm = modregrm(2, 0, base & 7);
if (base & 8)
{ rex |= REX_B;
if (base == R12)
{
rm = modregrm(2, 0, 4);
sib = modregrm(0, 4, 4);
}
}
}
}
else
{
rm = modregrm(2, 0, 4);
sib = modregrm(ss,index & 7,base & 7);
if (index & 8)
rex |= REX_X;
if (base & 8)
rex |= REX_B;
}
}
else
{
// -1 AX CX DX BX SP BP SI DI
static immutable ubyte[9][9] EA16rm =
[
[ 0x06,0x09,0x09,0x09,0x87,0x09,0x86,0x84,0x85, ], // -1
[ 0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09, ], // AX
[ 0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09, ], // CX
[ 0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09, ], // DX
[ 0x87,0x09,0x09,0x09,0x09,0x09,0x09,0x80,0x81, ], // BX
[ 0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09,0x09, ], // SP
[ 0x86,0x09,0x09,0x09,0x09,0x09,0x09,0x82,0x83, ], // BP
[ 0x84,0x09,0x09,0x09,0x80,0x09,0x82,0x09,0x09, ], // SI
[ 0x85,0x09,0x09,0x09,0x81,0x09,0x83,0x09,0x09, ] // DI
];
assert(scale == 1);
rm = EA16rm[base + 1][index + 1];
assert(rm != 9);
}
c.Irm = rm;
c.Isib = sib;
c.Irex = rex;
c.IFL1 = FLconst;
c.IEV1.Vuns = cast(targ_uns)disp;
}
/*********************************************
* Build REX, modregrm and sib bytes
*/
uint buildModregrm(int mod, int reg, int rm)
{
uint m;
if (I16)
m = modregrm(mod, reg, rm);
else
{
if ((rm & 7) == SP && mod != 3)
m = (modregrm(0,4,SP) << 8) | modregrm(mod,reg & 7,4);
else
m = modregrm(mod,reg & 7,rm & 7);
if (reg & 8)
m |= REX_R << 16;
if (rm & 8)
m |= REX_B << 16;
}
return m;
}
/****************************************
* Generate code for eecontext
*/
@trusted
void genEEcode()
{
CodeBuilder cdb;
cdb.ctor();
eecontext.EEin++;
regcon.immed.mval = 0;
regm_t retregs = 0; //regmask(eecontext.EEelem.Ety);
assert(EEStack.offset >= REGSIZE);
cod3_stackadj(cdb, cast(int)(EEStack.offset - REGSIZE));
cdb.gen1(0x50 + SI); // PUSH ESI
cdb.genadjesp(cast(int)EEStack.offset);
gencodelem(cdb, eecontext.EEelem, &retregs, false);
code *c = cdb.finish();
assignaddrc(c);
pinholeopt(c,null);
jmpaddr(c);
eecontext.EEcode = gen1(c, 0xCC); // INT 3
eecontext.EEin--;
}
/********************************************
* Gen a save/restore sequence for mask of registers.
* Params:
* regm = mask of registers to save
* cdbsave = save code appended here
* cdbrestore = restore code appended here
* Returns:
* amount of stack consumed
*/
@trusted
uint gensaverestore(regm_t regm,ref CodeBuilder cdbsave,ref CodeBuilder cdbrestore)
{
//printf("gensaverestore2(%s)\n", regm_str(regm));
regm &= mBP | mES | ALLREGS | XMMREGS | mST0 | mST01;
if (!regm)
return 0;
uint stackused = 0;
code *[regm.sizeof * 8] restore;
reg_t i;
for (i = 0; regm; i++)
{
if (regm & 1)
{
code *cs2;
if (i == ES && I16)
{
stackused += REGSIZE;
cdbsave.gen1(0x06); // PUSH ES
cs2 = gen1(null, 0x07); // POP ES
}
else if (i == ST0 || i == ST01)
{
CodeBuilder cdb;
cdb.ctor();
gensaverestore87(1 << i, cdbsave, cdb);
cs2 = cdb.finish();
}
else if (i >= XMM0 || I64 || cgstate.funcarg.size)
{ uint idx;
regsave.save(cdbsave, i, &idx);
CodeBuilder cdb;
cdb.ctor();
regsave.restore(cdb, i, idx);
cs2 = cdb.finish();
}
else
{
stackused += REGSIZE;
cdbsave.gen1(0x50 + (i & 7)); // PUSH i
cs2 = gen1(null, 0x58 + (i & 7)); // POP i
if (i & 8)
{ code_orrex(cdbsave.last(), REX_B);
code_orrex(cs2, REX_B);
}
}
restore[i] = cs2;
}
else
restore[i] = null;
regm >>= 1;
}
while (i)
{
code *c = restore[--i];
if (c)
{
cdbrestore.append(c);
}
}
return stackused;
}
/****************************************
* Clean parameters off stack.
* Input:
* numpara amount to adjust stack pointer
* keepmsk mask of registers to not destroy
*/
@trusted
void genstackclean(ref CodeBuilder cdb,uint numpara,regm_t keepmsk)
{
//dbg_printf("genstackclean(numpara = %d, stackclean = %d)\n",numpara,cgstate.stackclean);
if (numpara && (cgstate.stackclean || STACKALIGN >= 16))
{
/+
if (0 && // won't work if operand of scodelem
numpara == stackpush && // if this is all those pushed
needframe && // and there will be a BP
!config.windows &&
!(regcon.mvar & fregsaved) // and no registers will be pushed
)
genregs(cdb,0x89,BP,SP); // MOV SP,BP
else
+/
{
regm_t scratchm = 0;
if (numpara == REGSIZE && config.flags4 & CFG4space)
{
scratchm = ALLREGS & ~keepmsk & regcon.used & ~regcon.mvar;
}
if (scratchm)
{
reg_t r;
allocreg(cdb, &scratchm, &r, TYint);
cdb.gen1(0x58 + r); // POP r
}
else
cod3_stackadj(cdb, -numpara);
}
stackpush -= numpara;
cdb.genadjesp(-numpara);
}
}
/*********************************
* Generate code for a logical expression.
* Input:
* e elem
* jcond
* bit 1 if true then goto jump address if e
* if false then goto jump address if !e
* 2 don't call save87()
* fltarg FLcode or FLblock, flavor of target if e evaluates to jcond
* targ either code or block pointer to destination
*/
@trusted
void logexp(ref CodeBuilder cdb, elem *e, int jcond, uint fltarg, code *targ)
{
//printf("logexp(e = %p, jcond = %d)\n", e, jcond); elem_print(e);
if (tybasic(e.Ety) == TYnoreturn)
{
con_t regconsave = regcon;
regm_t retregs = 0;
codelem(cdb,e,&retregs,0);
regconsave.used |= regcon.used;
regcon = regconsave;
return;
}
int no87 = (jcond & 2) == 0;
docommas(cdb, &e); // scan down commas
cgstate.stackclean++;
code* c, ce;
if (!OTleaf(e.Eoper) && !e.Ecount) // if operator and not common sub
{
switch (e.Eoper)
{
case OPoror:
{
con_t regconsave;
if (jcond & 1)
{
logexp(cdb, e.EV.E1, jcond, fltarg, targ);
regconsave = regcon;
logexp(cdb, e.EV.E2, jcond, fltarg, targ);
}
else
{
code *cnop = gennop(null);
logexp(cdb, e.EV.E1, jcond | 1, FLcode, cnop);
regconsave = regcon;
logexp(cdb, e.EV.E2, jcond, fltarg, targ);
cdb.append(cnop);
}
andregcon(®consave);
freenode(e);
cgstate.stackclean--;
return;
}
case OPandand:
{
con_t regconsave;
if (jcond & 1)
{
code *cnop = gennop(null); // a dummy target address
logexp(cdb, e.EV.E1, jcond & ~1, FLcode, cnop);
regconsave = regcon;
logexp(cdb, e.EV.E2, jcond, fltarg, targ);
cdb.append(cnop);
}
else
{
logexp(cdb, e.EV.E1, jcond, fltarg, targ);
regconsave = regcon;
logexp(cdb, e.EV.E2, jcond, fltarg, targ);
}
andregcon(®consave);
freenode(e);
cgstate.stackclean--;
return;
}
case OPnot:
jcond ^= 1;
goto case OPbool;
case OPbool:
case OPs8_16:
case OPu8_16:
case OPs16_32:
case OPu16_32:
case OPs32_64:
case OPu32_64:
case OPu32_d:
case OPd_ld:
logexp(cdb, e.EV.E1, jcond, fltarg, targ);
freenode(e);
cgstate.stackclean--;
return;
case OPcond:
{
code *cnop2 = gennop(null); // addresses of start of leaves
code *cnop = gennop(null);
logexp(cdb, e.EV.E1, false, FLcode, cnop2); // eval condition
con_t regconold = regcon;
logexp(cdb, e.EV.E2.EV.E1, jcond, fltarg, targ);
genjmp(cdb, JMP, FLcode, cast(block *) cnop); // skip second leaf
con_t regconsave = regcon;
regcon = regconold;
cdb.append(cnop2);
logexp(cdb, e.EV.E2.EV.E2, jcond, fltarg, targ);
andregcon(®conold);
andregcon(®consave);
freenode(e.EV.E2);
freenode(e);
cdb.append(cnop);
cgstate.stackclean--;
return;
}
default:
break;
}
}
/* Special code for signed long compare.
* Not necessary for I64 until we do cents.
*/
if (OTrel2(e.Eoper) && // if < <= >= >
!e.Ecount &&
( (I16 && tybasic(e.EV.E1.Ety) == TYlong && tybasic(e.EV.E2.Ety) == TYlong) ||
(I32 && tybasic(e.EV.E1.Ety) == TYllong && tybasic(e.EV.E2.Ety) == TYllong))
)
{
longcmp(cdb, e, jcond != 0, fltarg, targ);
cgstate.stackclean--;
return;
}
regm_t retregs = mPSW; // return result in flags
opcode_t op = jmpopcode(e); // get jump opcode
if (!(jcond & 1))
op ^= 0x101; // toggle jump condition(s)
codelem(cdb, e, &retregs, true); // evaluate elem
if (no87)
cse_flush(cdb,no87); // flush CSE's to memory
genjmp(cdb, op, fltarg, cast(block *) targ); // generate jmp instruction
cgstate.stackclean--;
}
/******************************
* Routine to aid in setting things up for gen().
* Look for common subexpression.
* Can handle indirection operators, but not if they're common subs.
* Input:
* e -> elem where we get some of the data from
* cs -> partially filled code to add
* op = opcode
* reg = reg field of (mod reg r/m)
* offset = data to be added to Voffset field
* keepmsk = mask of registers we must not destroy
* desmsk = mask of registers destroyed by executing the instruction
* Returns:
* pointer to code generated
*/
@trusted
void loadea(ref CodeBuilder cdb,elem *e,code *cs,uint op,uint reg,targ_size_t offset,
regm_t keepmsk,regm_t desmsk)
{
code* c, cg, cd;
debug
if (debugw)
printf("loadea: e=%p cs=%p op=x%x reg=%s offset=%lld keepmsk=%s desmsk=%s\n",
e, cs, op, regstring[reg], cast(ulong)offset, regm_str(keepmsk), regm_str(desmsk));
assert(e);
cs.Iflags = 0;
cs.Irex = 0;
cs.Iop = op;
tym_t tym = e.Ety;
int sz = tysize(tym);
/* Determine if location we want to get is in a register. If so, */
/* substitute the register for the EA. */
/* Note that operators don't go through this. CSE'd operators are */
/* picked up by comsub(). */
if (e.Ecount && /* if cse */
e.Ecount != e.Ecomsub && /* and cse was generated */
op != LEA && op != 0xC4 && /* and not an LEA or LES */
(op != 0xFF || reg != 3) && /* and not CALLF MEM16 */
(op & 0xFFF8) != 0xD8) // and not 8087 opcode
{
assert(OTleaf(e.Eoper)); /* can't handle this */
regm_t rm = regcon.cse.mval & ~regcon.cse.mops & ~regcon.mvar; // possible regs
if (op == 0xFF && reg == 6)
rm &= ~XMMREGS; // can't PUSH an XMM register
if (sz > REGSIZE) // value is in 2 or 4 registers
{
if (I16 && sz == 8) // value is in 4 registers
{
static immutable regm_t[4] rmask = [ mDX,mCX,mBX,mAX ];
rm &= rmask[cast(size_t)(offset >> 1)];
}
else if (offset)
rm &= mMSW; /* only high words */
else
rm &= mLSW; /* only low words */
}
for (uint i = 0; rm; i++)
{
if (mask(i) & rm)
{
if (regcon.cse.value[i] == e && // if register has elem
/* watch out for a CWD destroying DX */
!(i == DX && op == 0xF7 && desmsk & mDX))
{
/* if ES, then it can only be a load */
if (i == ES)
{
if (op != 0x8B)
break; // not a load
cs.Iop = 0x8C; /* MOV reg,ES */
cs.Irm = modregrm(3, 0, reg & 7);
if (reg & 8)
code_orrex(cs, REX_B);
}
else // XXX reg,i
{
cs.Irm = modregrm(3, reg & 7, i & 7);
if (reg & 8)
cs.Irex |= REX_R;
if (i & 8)
cs.Irex |= REX_B;
if (sz == 1 && I64 && (i >= 4 || reg >= 4))
cs.Irex |= REX;
if (I64 && (sz == 8 || sz == 16))
cs.Irex |= REX_W;
}
goto L2;
}
rm &= ~mask(i);
}
}
}
getlvalue(cdb, cs, e, keepmsk);
if (offset == REGSIZE)
getlvalue_msw(cs);
else
cs.IEV1.Voffset += offset;
if (I64)
{
if (reg >= 4 && sz == 1) // if byte register
// Can only address those 8 bit registers if a REX byte is present
cs.Irex |= REX;
if ((op & 0xFFFFFFF8) == 0xD8)
cs.Irex &= ~REX_W; // not needed for x87 ops
if (mask(reg) & XMMREGS &&
(op == LODSD || op == STOSD))
cs.Irex &= ~REX_W; // not needed for xmm ops
}
code_newreg(cs, reg); // OR in reg field
if (!I16)
{
if (reg == 6 && op == 0xFF || /* don't PUSH a word */
op == MOVZXw || op == MOVSXw || /* MOVZX/MOVSX */
(op & 0xFFF8) == 0xD8 || /* 8087 instructions */
op == LEA) /* LEA */
{
cs.Iflags &= ~CFopsize;
if (reg == 6 && op == 0xFF) // if PUSH
cs.Irex &= ~REX_W; // REX is ignored for PUSH anyway
}
}
else if ((op & 0xFFF8) == 0xD8 && ADDFWAIT())
cs.Iflags |= CFwait;
L2:
getregs(cdb, desmsk); // save any regs we destroy
/* KLUDGE! fix up DX for divide instructions */
if (op == 0xF7 && desmsk == (mAX|mDX)) /* if we need to fix DX */
{
if (reg == 7) /* if IDIV */
{
cdb.gen1(0x99); // CWD
if (I64 && sz == 8)
code_orrex(cdb.last(), REX_W);
}
else if (reg == 6) // if DIV
genregs(cdb, 0x33, DX, DX); // XOR DX,DX
}
// Eliminate MOV reg,reg
if ((cs.Iop & ~3) == 0x88 &&
(cs.Irm & 0xC7) == modregrm(3,0,reg & 7))
{
uint r = cs.Irm & 7;
if (cs.Irex & REX_B)
r |= 8;
if (r == reg)
cs.Iop = NOP;
}
// Eliminate MOV xmmreg,xmmreg
if ((cs.Iop & ~(LODSD ^ STOSS)) == LODSD && // detect LODSD, LODSS, STOSD, STOSS
(cs.Irm & 0xC7) == modregrm(3,0,reg & 7))
{
reg_t r = cs.Irm & 7;
if (cs.Irex & REX_B)
r |= 8;
if (r == (reg - XMM0))
cs.Iop = NOP;
}
cdb.gen(cs);
}
/**************************
* Get addressing mode.
*/
@trusted
uint getaddrmode(regm_t idxregs)
{
uint mode;
if (I16)
{
static ubyte error() { assert(0); }
mode = (idxregs & mBX) ? modregrm(2,0,7) : /* [BX] */
(idxregs & mDI) ? modregrm(2,0,5): /* [DI] */
(idxregs & mSI) ? modregrm(2,0,4): /* [SI] */
error();
}
else
{
const reg = findreg(idxregs & (ALLREGS | mBP));
if (reg == R12)
mode = (REX_B << 16) | (modregrm(0,4,4) << 8) | modregrm(2,0,4);
else
mode = modregrmx(2,0,reg);
}
return mode;
}
void setaddrmode(code *c, regm_t idxregs)
{
uint mode = getaddrmode(idxregs);
c.Irm = mode & 0xFF;
c.Isib = (mode >> 8) & 0xFF;
c.Irex &= ~REX_B;
c.Irex |= mode >> 16;
}
/**********************************************
*/
@trusted
void getlvalue_msw(code *c)
{
if (c.IFL1 == FLreg)
{
const regmsw = c.IEV1.Vsym.Sregmsw;
c.Irm = (c.Irm & ~7) | (regmsw & 7);
if (regmsw & 8)
c.Irex |= REX_B;
else
c.Irex &= ~REX_B;
}
else
c.IEV1.Voffset += REGSIZE;
}
/**********************************************
*/
@trusted
void getlvalue_lsw(code *c)
{
if (c.IFL1 == FLreg)
{
const reglsw = c.IEV1.Vsym.Sreglsw;
c.Irm = (c.Irm & ~7) | (reglsw & 7);
if (reglsw & 8)
c.Irex |= REX_B;
else
c.Irex &= ~REX_B;
}
else
c.IEV1.Voffset -= REGSIZE;
}
/******************
* Compute addressing mode.
* Generate & return sequence of code (if any).
* Return in cs the info on it.
* Input:
* pcs -> where to store data about addressing mode
* e -> the lvalue elem
* keepmsk mask of registers we must not destroy or use
* if (keepmsk & RMstore), this will be only a store operation
* into the lvalue
* if (keepmsk & RMload), this will be a read operation only
*/
@trusted
void getlvalue(ref CodeBuilder cdb,code *pcs,elem *e,regm_t keepmsk)
{
uint fl, f, opsave;
elem* e1, e11, e12;
bool e1isadd, e1free;
reg_t reg;
tym_t e1ty;
Symbol* s;
//printf("getlvalue(e = %p, keepmsk = %s)\n", e, regm_str(keepmsk));
//elem_print(e);
assert(e);
elem_debug(e);
if (e.Eoper == OPvar || e.Eoper == OPrelconst)
{
s = e.EV.Vsym;
fl = s.Sfl;
if (tyfloating(s.ty()))
objmod.fltused();
}
else
fl = FLoper;
pcs.IFL1 = cast(ubyte)fl;
pcs.Iflags = CFoff; /* only want offsets */
pcs.Irex = 0;
pcs.IEV1.Voffset = 0;
tym_t ty = e.Ety;
uint sz = tysize(ty);
if (tyfloating(ty))
objmod.fltused();
if (I64 && (sz == 8 || sz == 16) && !tyvector(ty))
pcs.Irex |= REX_W;
if (!I16 && sz == SHORTSIZE)
pcs.Iflags |= CFopsize;
if (ty & mTYvolatile)
pcs.Iflags |= CFvolatile;
switch (fl)
{
case FLoper:
debug
if (debugw) printf("getlvalue(e = %p, keepmsk = %s)\n", e, regm_str(keepmsk));
switch (e.Eoper)
{
case OPadd: // this way when we want to do LEA
e1 = e;
e1free = false;
e1isadd = true;
break;
case OPind:
case OPpostinc: // when doing (*p++ = ...)
case OPpostdec: // when doing (*p-- = ...)
case OPbt:
case OPbtc:
case OPbtr:
case OPbts:
case OPvecfill:
e1 = e.EV.E1;
e1free = true;
e1isadd = e1.Eoper == OPadd;
break;
default:
printf("function: %s\n", funcsym_p.Sident.ptr);
elem_print(e);
assert(0);
}
e1ty = tybasic(e1.Ety);
if (e1isadd)
{
e12 = e1.EV.E2;
e11 = e1.EV.E1;
}
/* First see if we can replace *(e+&v) with
* MOV idxreg,e
* EA = [ES:] &v+idxreg
*/
f = FLconst;
/* Is address of `s` relative to RIP ?
*/
static bool relativeToRIP(Symbol* s)
{
if (!I64)
return false;
if (config.exe == EX_WIN64)
return true;
if (config.flags3 & CFG3pie)
{
if (s.Sfl == FLtlsdata || s.ty() & mTYthread)
{
if (s.Sclass == SCglobal || s.Sclass == SCstatic || s.Sclass == SClocstat)
return false;
}
return true;
}
else
return (config.flags3 & CFG3pic) != 0;
}
if (e1isadd &&
((e12.Eoper == OPrelconst &&
!relativeToRIP(e12.EV.Vsym) &&
(f = el_fl(e12)) != FLfardata
) ||
(e12.Eoper == OPconst && !I16 && !e1.Ecount && (!I64 || el_signx32(e12)))) &&
e1.Ecount == e1.Ecomsub &&
(!e1.Ecount || (~keepmsk & ALLREGS & mMSW) || (e1ty != TYfptr && e1ty != TYhptr)) &&
tysize(e11.Ety) == REGSIZE
)
{
uint t; /* component of r/m field */
int ss;
int ssi;
if (e12.Eoper == OPrelconst)
f = el_fl(e12);
/*assert(datafl[f]);*/ /* what if addr of func? */
if (!I16)
{ /* Any register can be an index register */
regm_t idxregs = allregs & ~keepmsk;
assert(idxregs);