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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
|
/*
* Copyright © 2010 Intel Corporation
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice (including the next
* paragraph) shall be included in all copies or substantial portions of the
* Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
* IN THE SOFTWARE.
*/
/** @file brw_fs_generator.cpp
*
* This file supports generating code from the FS LIR to the actual
* native instructions.
*/
extern "C" {
#include "main/macros.h"
#include "brw_context.h"
#include "brw_eu.h"
} /* extern "C" */
#include "brw_fs.h"
#include "brw_cfg.h"
fs_generator::fs_generator(struct brw_context *brw,
struct brw_wm_compile *c,
struct gl_shader_program *prog,
struct gl_fragment_program *fp,
bool dual_source_output)
: brw(brw), c(c), prog(prog), fp(fp), dual_source_output(dual_source_output)
{
ctx = &brw->ctx;
shader = prog ? prog->_LinkedShaders[MESA_SHADER_FRAGMENT] : NULL;
mem_ctx = c;
p = rzalloc(mem_ctx, struct brw_compile);
brw_init_compile(brw, p, mem_ctx);
}
fs_generator::~fs_generator()
{
}
void
fs_generator::mark_surface_used(unsigned surf_index)
{
assert(surf_index < BRW_MAX_SURFACES);
c->prog_data.base.binding_table.size_bytes =
MAX2(c->prog_data.base.binding_table.size_bytes, (surf_index + 1) * 4);
}
void
fs_generator::patch_discard_jumps_to_fb_writes()
{
if (brw->gen < 6 || this->discard_halt_patches.is_empty())
return;
/* There is a somewhat strange undocumented requirement of using
* HALT, according to the simulator. If some channel has HALTed to
* a particular UIP, then by the end of the program, every channel
* must have HALTed to that UIP. Furthermore, the tracking is a
* stack, so you can't do the final halt of a UIP after starting
* halting to a new UIP.
*
* Symptoms of not emitting this instruction on actual hardware
* included GPU hangs and sparkly rendering on the piglit discard
* tests.
*/
struct brw_instruction *last_halt = gen6_HALT(p);
last_halt->bits3.break_cont.uip = 2;
last_halt->bits3.break_cont.jip = 2;
int ip = p->nr_insn;
foreach_list(node, &this->discard_halt_patches) {
ip_record *patch_ip = (ip_record *)node;
struct brw_instruction *patch = &p->store[patch_ip->ip];
assert(patch->header.opcode == BRW_OPCODE_HALT);
/* HALT takes a half-instruction distance from the pre-incremented IP. */
patch->bits3.break_cont.uip = (ip - patch_ip->ip) * 2;
}
this->discard_halt_patches.make_empty();
}
void
fs_generator::generate_fb_write(fs_inst *inst)
{
bool eot = inst->eot;
struct brw_reg implied_header;
uint32_t msg_control;
/* Header is 2 regs, g0 and g1 are the contents. g0 will be implied
* move, here's g1.
*/
brw_push_insn_state(p);
brw_set_mask_control(p, BRW_MASK_DISABLE);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
if (fp->UsesKill) {
struct brw_reg pixel_mask;
if (brw->gen >= 6)
pixel_mask = retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_UW);
else
pixel_mask = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
brw_MOV(p, pixel_mask, brw_flag_reg(0, 1));
}
if (inst->header_present) {
if (brw->gen >= 6) {
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
brw_MOV(p,
retype(brw_message_reg(inst->base_mrf), BRW_REGISTER_TYPE_UD),
retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UD));
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
if (inst->target > 0 && c->key.replicate_alpha) {
/* Set "Source0 Alpha Present to RenderTarget" bit in message
* header.
*/
brw_OR(p,
vec1(retype(brw_message_reg(inst->base_mrf), BRW_REGISTER_TYPE_UD)),
vec1(retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UD)),
brw_imm_ud(0x1 << 11));
}
if (inst->target > 0) {
/* Set the render target index for choosing BLEND_STATE. */
brw_MOV(p, retype(brw_vec1_reg(BRW_MESSAGE_REGISTER_FILE,
inst->base_mrf, 2),
BRW_REGISTER_TYPE_UD),
brw_imm_ud(inst->target));
}
implied_header = brw_null_reg();
} else {
implied_header = retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW);
brw_MOV(p,
brw_message_reg(inst->base_mrf + 1),
brw_vec8_grf(1, 0));
}
} else {
implied_header = brw_null_reg();
}
if (this->dual_source_output)
msg_control = BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01;
else if (dispatch_width == 16)
msg_control = BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE;
else
msg_control = BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01;
brw_pop_insn_state(p);
uint32_t surf_index =
c->prog_data.binding_table.render_target_start + inst->target;
brw_fb_WRITE(p,
dispatch_width,
inst->base_mrf,
implied_header,
msg_control,
surf_index,
inst->mlen,
0,
eot,
inst->header_present);
mark_surface_used(surf_index);
}
/* Computes the integer pixel x,y values from the origin.
*
* This is the basis of gl_FragCoord computation, but is also used
* pre-gen6 for computing the deltas from v0 for computing
* interpolation.
*/
void
fs_generator::generate_pixel_xy(struct brw_reg dst, bool is_x)
{
struct brw_reg g1_uw = retype(brw_vec1_grf(1, 0), BRW_REGISTER_TYPE_UW);
struct brw_reg src;
struct brw_reg deltas;
if (is_x) {
src = stride(suboffset(g1_uw, 4), 2, 4, 0);
deltas = brw_imm_v(0x10101010);
} else {
src = stride(suboffset(g1_uw, 5), 2, 4, 0);
deltas = brw_imm_v(0x11001100);
}
if (dispatch_width == 16) {
dst = vec16(dst);
}
/* We do this 8 or 16-wide, but since the destination is UW we
* don't do compression in the 16-wide case.
*/
brw_push_insn_state(p);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_ADD(p, dst, src, deltas);
brw_pop_insn_state(p);
}
void
fs_generator::generate_linterp(fs_inst *inst,
struct brw_reg dst, struct brw_reg *src)
{
struct brw_reg delta_x = src[0];
struct brw_reg delta_y = src[1];
struct brw_reg interp = src[2];
if (brw->has_pln &&
delta_y.nr == delta_x.nr + 1 &&
(brw->gen >= 6 || (delta_x.nr & 1) == 0)) {
brw_PLN(p, dst, interp, delta_x);
} else {
brw_LINE(p, brw_null_reg(), interp, delta_x);
brw_MAC(p, dst, suboffset(interp, 1), delta_y);
}
}
void
fs_generator::generate_math1_gen7(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src0)
{
assert(inst->mlen == 0);
brw_math(p, dst,
brw_math_function(inst->opcode),
0, src0,
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
}
void
fs_generator::generate_math2_gen7(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src0,
struct brw_reg src1)
{
assert(inst->mlen == 0);
brw_math2(p, dst, brw_math_function(inst->opcode), src0, src1);
}
void
fs_generator::generate_math1_gen6(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src0)
{
int op = brw_math_function(inst->opcode);
assert(inst->mlen == 0);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_math(p, dst,
op,
0, src0,
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_math(p, sechalf(dst),
op,
0, sechalf(src0),
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
}
}
void
fs_generator::generate_math2_gen6(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src0,
struct brw_reg src1)
{
int op = brw_math_function(inst->opcode);
assert(inst->mlen == 0);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_math2(p, dst, op, src0, src1);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_math2(p, sechalf(dst), op, sechalf(src0), sechalf(src1));
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
}
}
void
fs_generator::generate_math_gen4(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src)
{
int op = brw_math_function(inst->opcode);
assert(inst->mlen >= 1);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_math(p, dst,
op,
inst->base_mrf, src,
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_math(p, sechalf(dst),
op,
inst->base_mrf + 1, sechalf(src),
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
}
}
void
fs_generator::generate_math_g45(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src)
{
if (inst->opcode == SHADER_OPCODE_POW ||
inst->opcode == SHADER_OPCODE_INT_QUOTIENT ||
inst->opcode == SHADER_OPCODE_INT_REMAINDER) {
generate_math_gen4(inst, dst, src);
return;
}
int op = brw_math_function(inst->opcode);
assert(inst->mlen >= 1);
brw_math(p, dst,
op,
inst->base_mrf, src,
BRW_MATH_DATA_VECTOR,
BRW_MATH_PRECISION_FULL);
}
void
fs_generator::generate_tex(fs_inst *inst, struct brw_reg dst, struct brw_reg src)
{
int msg_type = -1;
int rlen = 4;
uint32_t simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD8;
uint32_t return_format;
switch (dst.type) {
case BRW_REGISTER_TYPE_D:
return_format = BRW_SAMPLER_RETURN_FORMAT_SINT32;
break;
case BRW_REGISTER_TYPE_UD:
return_format = BRW_SAMPLER_RETURN_FORMAT_UINT32;
break;
default:
return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32;
break;
}
if (dispatch_width == 16 &&
!inst->force_uncompressed && !inst->force_sechalf)
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
if (brw->gen >= 5) {
switch (inst->opcode) {
case SHADER_OPCODE_TEX:
if (inst->shadow_compare) {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_COMPARE;
} else {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE;
}
break;
case FS_OPCODE_TXB:
if (inst->shadow_compare) {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_BIAS_COMPARE;
} else {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_BIAS;
}
break;
case SHADER_OPCODE_TXL:
if (inst->shadow_compare) {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_LOD_COMPARE;
} else {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_LOD;
}
break;
case SHADER_OPCODE_TXS:
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_RESINFO;
break;
case SHADER_OPCODE_TXD:
if (inst->shadow_compare) {
/* Gen7.5+. Otherwise, lowered by brw_lower_texture_gradients(). */
assert(brw->is_haswell);
msg_type = HSW_SAMPLER_MESSAGE_SAMPLE_DERIV_COMPARE;
} else {
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_DERIVS;
}
break;
case SHADER_OPCODE_TXF:
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_LD;
break;
case SHADER_OPCODE_TXF_MS:
if (brw->gen >= 7)
msg_type = GEN7_SAMPLER_MESSAGE_SAMPLE_LD2DMS;
else
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_LD;
break;
case SHADER_OPCODE_LOD:
msg_type = GEN5_SAMPLER_MESSAGE_LOD;
break;
case SHADER_OPCODE_TG4:
assert(brw->gen >= 6);
msg_type = GEN7_SAMPLER_MESSAGE_SAMPLE_GATHER4;
break;
default:
assert(!"not reached");
break;
}
} else {
switch (inst->opcode) {
case SHADER_OPCODE_TEX:
/* Note that G45 and older determines shadow compare and dispatch width
* from message length for most messages.
*/
assert(dispatch_width == 8);
msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE;
if (inst->shadow_compare) {
assert(inst->mlen == 6);
} else {
assert(inst->mlen <= 4);
}
break;
case FS_OPCODE_TXB:
if (inst->shadow_compare) {
assert(inst->mlen == 6);
msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_BIAS_COMPARE;
} else {
assert(inst->mlen == 9);
msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
}
break;
case SHADER_OPCODE_TXL:
if (inst->shadow_compare) {
assert(inst->mlen == 6);
msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_LOD_COMPARE;
} else {
assert(inst->mlen == 9);
msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
}
break;
case SHADER_OPCODE_TXD:
/* There is no sample_d_c message; comparisons are done manually */
assert(inst->mlen == 7 || inst->mlen == 10);
msg_type = BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS;
break;
case SHADER_OPCODE_TXF:
assert(inst->mlen == 9);
msg_type = BRW_SAMPLER_MESSAGE_SIMD16_LD;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
break;
case SHADER_OPCODE_TXS:
assert(inst->mlen == 3);
msg_type = BRW_SAMPLER_MESSAGE_SIMD16_RESINFO;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
break;
default:
assert(!"not reached");
break;
}
}
assert(msg_type != -1);
if (simd_mode == BRW_SAMPLER_SIMD_MODE_SIMD16) {
rlen = 8;
dst = vec16(dst);
}
if (brw->gen >= 7 && inst->header_present && dispatch_width == 16) {
/* The send-from-GRF for 16-wide texturing with a header has an extra
* hardware register allocated to it, which we need to skip over (since
* our coordinates in the payload are in the even-numbered registers,
* and the header comes right before the first one).
*/
assert(src.file == BRW_GENERAL_REGISTER_FILE);
src.nr++;
}
/* Load the message header if present. If there's a texture offset,
* we need to set it up explicitly and load the offset bitfield.
* Otherwise, we can use an implied move from g0 to the first message reg.
*/
if (inst->texture_offset) {
struct brw_reg header_reg;
if (brw->gen >= 7) {
header_reg = src;
} else {
assert(inst->base_mrf != -1);
header_reg = retype(brw_message_reg(inst->base_mrf),
BRW_REGISTER_TYPE_UD);
}
brw_push_insn_state(p);
brw_set_mask_control(p, BRW_MASK_DISABLE);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
/* Explicitly set up the message header by copying g0 to the MRF. */
brw_MOV(p, header_reg, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UD));
/* Then set the offset bits in DWord 2. */
brw_MOV(p, retype(brw_vec1_reg(header_reg.file,
header_reg.nr, 2), BRW_REGISTER_TYPE_UD),
brw_imm_ud(inst->texture_offset));
brw_pop_insn_state(p);
} else if (inst->header_present) {
assert(brw->gen < 7);
/* Set up an implied move from g0 to the MRF. */
src = retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW);
}
uint32_t surface_index = (inst->opcode == SHADER_OPCODE_TG4
? c->prog_data.base.binding_table.gather_texture_start
: c->prog_data.base.binding_table.texture_start) + inst->sampler;
brw_SAMPLE(p,
retype(dst, BRW_REGISTER_TYPE_UW),
inst->base_mrf,
src,
surface_index,
inst->sampler,
msg_type,
rlen,
inst->mlen,
inst->header_present,
simd_mode,
return_format);
mark_surface_used(surface_index);
}
/* For OPCODE_DDX and OPCODE_DDY, per channel of output we've got input
* looking like:
*
* arg0: ss0.tl ss0.tr ss0.bl ss0.br ss1.tl ss1.tr ss1.bl ss1.br
*
* Ideally, we want to produce:
*
* DDX DDY
* dst: (ss0.tr - ss0.tl) (ss0.tl - ss0.bl)
* (ss0.tr - ss0.tl) (ss0.tr - ss0.br)
* (ss0.br - ss0.bl) (ss0.tl - ss0.bl)
* (ss0.br - ss0.bl) (ss0.tr - ss0.br)
* (ss1.tr - ss1.tl) (ss1.tl - ss1.bl)
* (ss1.tr - ss1.tl) (ss1.tr - ss1.br)
* (ss1.br - ss1.bl) (ss1.tl - ss1.bl)
* (ss1.br - ss1.bl) (ss1.tr - ss1.br)
*
* and add another set of two more subspans if in 16-pixel dispatch mode.
*
* For DDX, it ends up being easy: width = 2, horiz=0 gets us the same result
* for each pair, and vertstride = 2 jumps us 2 elements after processing a
* pair. But the ideal approximation may impose a huge performance cost on
* sample_d. On at least Haswell, sample_d instruction does some
* optimizations if the same LOD is used for all pixels in the subspan.
*
* For DDY, we need to use ALIGN16 mode since it's capable of doing the
* appropriate swizzling.
*/
void
fs_generator::generate_ddx(fs_inst *inst, struct brw_reg dst, struct brw_reg src)
{
unsigned vstride, width;
if (c->key.high_quality_derivatives) {
/* produce accurate derivatives */
vstride = BRW_VERTICAL_STRIDE_2;
width = BRW_WIDTH_2;
}
else {
/* replicate the derivative at the top-left pixel to other pixels */
vstride = BRW_VERTICAL_STRIDE_4;
width = BRW_WIDTH_4;
}
struct brw_reg src0 = brw_reg(src.file, src.nr, 1,
BRW_REGISTER_TYPE_F,
vstride,
width,
BRW_HORIZONTAL_STRIDE_0,
BRW_SWIZZLE_XYZW, WRITEMASK_XYZW);
struct brw_reg src1 = brw_reg(src.file, src.nr, 0,
BRW_REGISTER_TYPE_F,
vstride,
width,
BRW_HORIZONTAL_STRIDE_0,
BRW_SWIZZLE_XYZW, WRITEMASK_XYZW);
brw_ADD(p, dst, src0, negate(src1));
}
/* The negate_value boolean is used to negate the derivative computation for
* FBOs, since they place the origin at the upper left instead of the lower
* left.
*/
void
fs_generator::generate_ddy(fs_inst *inst, struct brw_reg dst, struct brw_reg src,
bool negate_value)
{
if (c->key.high_quality_derivatives) {
/* produce accurate derivatives */
struct brw_reg src0 = brw_reg(src.file, src.nr, 0,
BRW_REGISTER_TYPE_F,
BRW_VERTICAL_STRIDE_4,
BRW_WIDTH_4,
BRW_HORIZONTAL_STRIDE_1,
BRW_SWIZZLE_XYXY, WRITEMASK_XYZW);
struct brw_reg src1 = brw_reg(src.file, src.nr, 0,
BRW_REGISTER_TYPE_F,
BRW_VERTICAL_STRIDE_4,
BRW_WIDTH_4,
BRW_HORIZONTAL_STRIDE_1,
BRW_SWIZZLE_ZWZW, WRITEMASK_XYZW);
brw_push_insn_state(p);
brw_set_access_mode(p, BRW_ALIGN_16);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
if (negate_value)
brw_ADD(p, dst, src1, negate(src0));
else
brw_ADD(p, dst, src0, negate(src1));
if (dispatch_width == 16) {
/* From page 340 of the i965 PRM:
*
* "A compressed instruction must be in Align1 access
* mode. Align16 mode instructions cannot be compressed."
*
* Therefore, when doing a 16-wide dispatch, we need to manually
* unroll to two instructions.
*/
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
src0 = sechalf(src0);
src1 = sechalf(src1);
dst = sechalf(dst);
if (negate_value)
brw_ADD(p, dst, src1, negate(src0));
else
brw_ADD(p, dst, src0, negate(src1));
}
brw_pop_insn_state(p);
} else {
/* replicate the derivative at the top-left pixel to other pixels */
struct brw_reg src0 = brw_reg(src.file, src.nr, 0,
BRW_REGISTER_TYPE_F,
BRW_VERTICAL_STRIDE_4,
BRW_WIDTH_4,
BRW_HORIZONTAL_STRIDE_0,
BRW_SWIZZLE_XYZW, WRITEMASK_XYZW);
struct brw_reg src1 = brw_reg(src.file, src.nr, 2,
BRW_REGISTER_TYPE_F,
BRW_VERTICAL_STRIDE_4,
BRW_WIDTH_4,
BRW_HORIZONTAL_STRIDE_0,
BRW_SWIZZLE_XYZW, WRITEMASK_XYZW);
if (negate_value)
brw_ADD(p, dst, src1, negate(src0));
else
brw_ADD(p, dst, src0, negate(src1));
}
}
void
fs_generator::generate_discard_jump(fs_inst *inst)
{
assert(brw->gen >= 6);
/* This HALT will be patched up at FB write time to point UIP at the end of
* the program, and at brw_uip_jip() JIP will be set to the end of the
* current block (or the program).
*/
this->discard_halt_patches.push_tail(new(mem_ctx) ip_record(p->nr_insn));
brw_push_insn_state(p);
brw_set_mask_control(p, BRW_MASK_DISABLE);
gen6_HALT(p);
brw_pop_insn_state(p);
}
void
fs_generator::generate_spill(fs_inst *inst, struct brw_reg src)
{
assert(inst->mlen != 0);
brw_MOV(p,
retype(brw_message_reg(inst->base_mrf + 1), BRW_REGISTER_TYPE_UD),
retype(src, BRW_REGISTER_TYPE_UD));
brw_oword_block_write_scratch(p, brw_message_reg(inst->base_mrf), 1,
inst->offset);
}
void
fs_generator::generate_unspill(fs_inst *inst, struct brw_reg dst)
{
assert(inst->mlen != 0);
brw_oword_block_read_scratch(p, dst, brw_message_reg(inst->base_mrf), 1,
inst->offset);
}
void
fs_generator::generate_uniform_pull_constant_load(fs_inst *inst,
struct brw_reg dst,
struct brw_reg index,
struct brw_reg offset)
{
assert(inst->mlen != 0);
assert(index.file == BRW_IMMEDIATE_VALUE &&
index.type == BRW_REGISTER_TYPE_UD);
uint32_t surf_index = index.dw1.ud;
assert(offset.file == BRW_IMMEDIATE_VALUE &&
offset.type == BRW_REGISTER_TYPE_UD);
uint32_t read_offset = offset.dw1.ud;
brw_oword_block_read(p, dst, brw_message_reg(inst->base_mrf),
read_offset, surf_index);
mark_surface_used(surf_index);
}
void
fs_generator::generate_uniform_pull_constant_load_gen7(fs_inst *inst,
struct brw_reg dst,
struct brw_reg index,
struct brw_reg offset)
{
assert(inst->mlen == 0);
assert(index.file == BRW_IMMEDIATE_VALUE &&
index.type == BRW_REGISTER_TYPE_UD);
uint32_t surf_index = index.dw1.ud;
assert(offset.file == BRW_GENERAL_REGISTER_FILE);
/* Reference just the dword we need, to avoid angering validate_reg(). */
offset = brw_vec1_grf(offset.nr, 0);
brw_push_insn_state(p);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_set_mask_control(p, BRW_MASK_DISABLE);
struct brw_instruction *send = brw_next_insn(p, BRW_OPCODE_SEND);
brw_pop_insn_state(p);
/* We use the SIMD4x2 mode because we want to end up with 4 components in
* the destination loaded consecutively from the same offset (which appears
* in the first component, and the rest are ignored).
*/
dst.width = BRW_WIDTH_4;
brw_set_dest(p, send, dst);
brw_set_src0(p, send, offset);
brw_set_sampler_message(p, send,
surf_index,
0, /* LD message ignores sampler unit */
GEN5_SAMPLER_MESSAGE_SAMPLE_LD,
1, /* rlen */
1, /* mlen */
false, /* no header */
BRW_SAMPLER_SIMD_MODE_SIMD4X2,
0);
mark_surface_used(surf_index);
}
void
fs_generator::generate_varying_pull_constant_load(fs_inst *inst,
struct brw_reg dst,
struct brw_reg index,
struct brw_reg offset)
{
assert(brw->gen < 7); /* Should use the gen7 variant. */
assert(inst->header_present);
assert(inst->mlen);
assert(index.file == BRW_IMMEDIATE_VALUE &&
index.type == BRW_REGISTER_TYPE_UD);
uint32_t surf_index = index.dw1.ud;
uint32_t simd_mode, rlen, msg_type;
if (dispatch_width == 16) {
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
rlen = 8;
} else {
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD8;
rlen = 4;
}
if (brw->gen >= 5)
msg_type = GEN5_SAMPLER_MESSAGE_SAMPLE_LD;
else {
/* We always use the SIMD16 message so that we only have to load U, and
* not V or R.
*/
msg_type = BRW_SAMPLER_MESSAGE_SIMD16_LD;
assert(inst->mlen == 3);
assert(inst->regs_written == 8);
rlen = 8;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
}
struct brw_reg offset_mrf = retype(brw_message_reg(inst->base_mrf + 1),
BRW_REGISTER_TYPE_D);
brw_MOV(p, offset_mrf, offset);
struct brw_reg header = brw_vec8_grf(0, 0);
gen6_resolve_implied_move(p, &header, inst->base_mrf);
struct brw_instruction *send = brw_next_insn(p, BRW_OPCODE_SEND);
send->header.compression_control = BRW_COMPRESSION_NONE;
brw_set_dest(p, send, dst);
brw_set_src0(p, send, header);
if (brw->gen < 6)
send->header.destreg__conditionalmod = inst->base_mrf;
/* Our surface is set up as floats, regardless of what actual data is
* stored in it.
*/
uint32_t return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32;
brw_set_sampler_message(p, send,
surf_index,
0, /* sampler (unused) */
msg_type,
rlen,
inst->mlen,
inst->header_present,
simd_mode,
return_format);
mark_surface_used(surf_index);
}
void
fs_generator::generate_varying_pull_constant_load_gen7(fs_inst *inst,
struct brw_reg dst,
struct brw_reg index,
struct brw_reg offset)
{
assert(brw->gen >= 7);
/* Varying-offset pull constant loads are treated as a normal expression on
* gen7, so the fact that it's a send message is hidden at the IR level.
*/
assert(!inst->header_present);
assert(!inst->mlen);
assert(index.file == BRW_IMMEDIATE_VALUE &&
index.type == BRW_REGISTER_TYPE_UD);
uint32_t surf_index = index.dw1.ud;
uint32_t simd_mode, rlen, mlen;
if (dispatch_width == 16) {
mlen = 2;
rlen = 8;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD16;
} else {
mlen = 1;
rlen = 4;
simd_mode = BRW_SAMPLER_SIMD_MODE_SIMD8;
}
struct brw_instruction *send = brw_next_insn(p, BRW_OPCODE_SEND);
brw_set_dest(p, send, dst);
brw_set_src0(p, send, offset);
brw_set_sampler_message(p, send,
surf_index,
0, /* LD message ignores sampler unit */
GEN5_SAMPLER_MESSAGE_SAMPLE_LD,
rlen,
mlen,
false, /* no header */
simd_mode,
0);
mark_surface_used(surf_index);
}
/**
* Cause the current pixel/sample mask (from R1.7 bits 15:0) to be transferred
* into the flags register (f0.0).
*
* Used only on Gen6 and above.
*/
void
fs_generator::generate_mov_dispatch_to_flags(fs_inst *inst)
{
struct brw_reg flags = brw_flag_reg(0, inst->flag_subreg);
struct brw_reg dispatch_mask;
if (brw->gen >= 6)
dispatch_mask = retype(brw_vec1_grf(1, 7), BRW_REGISTER_TYPE_UW);
else
dispatch_mask = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW);
brw_push_insn_state(p);
brw_set_mask_control(p, BRW_MASK_DISABLE);
brw_MOV(p, flags, dispatch_mask);
brw_pop_insn_state(p);
}
static uint32_t brw_file_from_reg(fs_reg *reg)
{
switch (reg->file) {
case GRF:
return BRW_GENERAL_REGISTER_FILE;
case MRF:
return BRW_MESSAGE_REGISTER_FILE;
case IMM:
return BRW_IMMEDIATE_VALUE;
default:
assert(!"not reached");
return BRW_GENERAL_REGISTER_FILE;
}
}
static struct brw_reg
brw_reg_from_fs_reg(fs_reg *reg)
{
struct brw_reg brw_reg;
switch (reg->file) {
case GRF:
case MRF:
if (reg->smear == -1) {
brw_reg = brw_vec8_reg(brw_file_from_reg(reg), reg->reg, 0);
} else {
brw_reg = brw_vec1_reg(brw_file_from_reg(reg), reg->reg, reg->smear);
}
brw_reg = retype(brw_reg, reg->type);
if (reg->sechalf)
brw_reg = sechalf(brw_reg);
break;
case IMM:
switch (reg->type) {
case BRW_REGISTER_TYPE_F:
brw_reg = brw_imm_f(reg->imm.f);
break;
case BRW_REGISTER_TYPE_D:
brw_reg = brw_imm_d(reg->imm.i);
break;
case BRW_REGISTER_TYPE_UD:
brw_reg = brw_imm_ud(reg->imm.u);
break;
default:
assert(!"not reached");
brw_reg = brw_null_reg();
break;
}
break;
case HW_REG:
brw_reg = reg->fixed_hw_reg;
break;
case BAD_FILE:
/* Probably unused. */
brw_reg = brw_null_reg();
break;
case UNIFORM:
assert(!"not reached");
brw_reg = brw_null_reg();
break;
default:
assert(!"not reached");
brw_reg = brw_null_reg();
break;
}
if (reg->abs)
brw_reg = brw_abs(brw_reg);
if (reg->negate)
brw_reg = negate(brw_reg);
return brw_reg;
}
/**
* Sets the first word of a vgrf for gen7+ simd4x2 uniform pull constant
* sampler LD messages.
*
* We don't want to bake it into the send message's code generation because
* that means we don't get a chance to schedule the instructions.
*/
void
fs_generator::generate_set_simd4x2_offset(fs_inst *inst,
struct brw_reg dst,
struct brw_reg value)
{
assert(value.file == BRW_IMMEDIATE_VALUE);
brw_push_insn_state(p);
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_set_mask_control(p, BRW_MASK_DISABLE);
brw_MOV(p, retype(brw_vec1_reg(dst.file, dst.nr, 0), value.type), value);
brw_pop_insn_state(p);
}
/**
* Change the register's data type from UD to W, doubling the strides in order
* to compensate for halving the data type width.
*/
static struct brw_reg
ud_reg_to_w(struct brw_reg r)
{
assert(r.type == BRW_REGISTER_TYPE_UD);
r.type = BRW_REGISTER_TYPE_W;
/* The BRW_*_STRIDE enums are defined so that incrementing the field
* doubles the real stride.
*/
if (r.hstride != 0)
++r.hstride;
if (r.vstride != 0)
++r.vstride;
return r;
}
void
fs_generator::generate_pack_half_2x16_split(fs_inst *inst,
struct brw_reg dst,
struct brw_reg x,
struct brw_reg y)
{
assert(brw->gen >= 7);
assert(dst.type == BRW_REGISTER_TYPE_UD);
assert(x.type == BRW_REGISTER_TYPE_F);
assert(y.type == BRW_REGISTER_TYPE_F);
/* From the Ivybridge PRM, Vol4, Part3, Section 6.27 f32to16:
*
* Because this instruction does not have a 16-bit floating-point type,
* the destination data type must be Word (W).
*
* The destination must be DWord-aligned and specify a horizontal stride
* (HorzStride) of 2. The 16-bit result is stored in the lower word of
* each destination channel and the upper word is not modified.
*/
struct brw_reg dst_w = ud_reg_to_w(dst);
/* Give each 32-bit channel of dst the form below , where "." means
* unchanged.
* 0x....hhhh
*/
brw_F32TO16(p, dst_w, y);
/* Now the form:
* 0xhhhh0000
*/
brw_SHL(p, dst, dst, brw_imm_ud(16u));
/* And, finally the form of packHalf2x16's output:
* 0xhhhhllll
*/
brw_F32TO16(p, dst_w, x);
}
void
fs_generator::generate_unpack_half_2x16_split(fs_inst *inst,
struct brw_reg dst,
struct brw_reg src)
{
assert(brw->gen >= 7);
assert(dst.type == BRW_REGISTER_TYPE_F);
assert(src.type == BRW_REGISTER_TYPE_UD);
/* From the Ivybridge PRM, Vol4, Part3, Section 6.26 f16to32:
*
* Because this instruction does not have a 16-bit floating-point type,
* the source data type must be Word (W). The destination type must be
* F (Float).
*/
struct brw_reg src_w = ud_reg_to_w(src);
/* Each channel of src has the form of unpackHalf2x16's input: 0xhhhhllll.
* For the Y case, we wish to access only the upper word; therefore
* a 16-bit subregister offset is needed.
*/
assert(inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X ||
inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y);
if (inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y)
src_w.subnr += 2;
brw_F16TO32(p, dst, src_w);
}
void
fs_generator::generate_shader_time_add(fs_inst *inst,
struct brw_reg payload,
struct brw_reg offset,
struct brw_reg value)
{
assert(brw->gen >= 7);
brw_push_insn_state(p);
brw_set_mask_control(p, true);
assert(payload.file == BRW_GENERAL_REGISTER_FILE);
struct brw_reg payload_offset = retype(brw_vec1_grf(payload.nr, 0),
offset.type);
struct brw_reg payload_value = retype(brw_vec1_grf(payload.nr + 1, 0),
value.type);
assert(offset.file == BRW_IMMEDIATE_VALUE);
if (value.file == BRW_GENERAL_REGISTER_FILE) {
value.width = BRW_WIDTH_1;
value.hstride = BRW_HORIZONTAL_STRIDE_0;
value.vstride = BRW_VERTICAL_STRIDE_0;
} else {
assert(value.file == BRW_IMMEDIATE_VALUE);
}
/* Trying to deal with setup of the params from the IR is crazy in the FS8
* case, and we don't really care about squeezing every bit of performance
* out of this path, so we just emit the MOVs from here.
*/
brw_MOV(p, payload_offset, offset);
brw_MOV(p, payload_value, value);
brw_shader_time_add(p, payload,
c->prog_data.base.binding_table.shader_time_start);
brw_pop_insn_state(p);
mark_surface_used(c->prog_data.base.binding_table.shader_time_start);
}
void
fs_generator::generate_code(exec_list *instructions)
{
int last_native_insn_offset = p->next_insn_offset;
const char *last_annotation_string = NULL;
const void *last_annotation_ir = NULL;
if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
if (shader) {
printf("Native code for fragment shader %d (%d-wide dispatch):\n",
prog->Name, dispatch_width);
} else {
printf("Native code for fragment program %d (%d-wide dispatch):\n",
fp->Base.Id, dispatch_width);
}
}
cfg_t *cfg = NULL;
if (unlikely(INTEL_DEBUG & DEBUG_WM))
cfg = new(mem_ctx) cfg_t(mem_ctx, instructions);
foreach_list(node, instructions) {
fs_inst *inst = (fs_inst *)node;
struct brw_reg src[3], dst;
if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
foreach_list(node, &cfg->block_list) {
bblock_link *link = (bblock_link *)node;
bblock_t *block = link->block;
if (block->start == inst) {
printf(" START B%d", block->block_num);
foreach_list(predecessor_node, &block->parents) {
bblock_link *predecessor_link =
(bblock_link *)predecessor_node;
bblock_t *predecessor_block = predecessor_link->block;
printf(" <-B%d", predecessor_block->block_num);
}
printf("\n");
}
}
if (last_annotation_ir != inst->ir) {
last_annotation_ir = inst->ir;
if (last_annotation_ir) {
printf(" ");
if (shader)
((ir_instruction *)inst->ir)->print();
else {
const prog_instruction *fpi;
fpi = (const prog_instruction *)inst->ir;
printf("%d: ", (int)(fpi - fp->Base.Instructions));
_mesa_fprint_instruction_opt(stdout,
fpi,
0, PROG_PRINT_DEBUG, NULL);
}
printf("\n");
}
}
if (last_annotation_string != inst->annotation) {
last_annotation_string = inst->annotation;
if (last_annotation_string)
printf(" %s\n", last_annotation_string);
}
}
for (unsigned int i = 0; i < 3; i++) {
src[i] = brw_reg_from_fs_reg(&inst->src[i]);
/* The accumulator result appears to get used for the
* conditional modifier generation. When negating a UD
* value, there is a 33rd bit generated for the sign in the
* accumulator value, so now you can't check, for example,
* equality with a 32-bit value. See piglit fs-op-neg-uvec4.
*/
assert(!inst->conditional_mod ||
inst->src[i].type != BRW_REGISTER_TYPE_UD ||
!inst->src[i].negate);
}
dst = brw_reg_from_fs_reg(&inst->dst);
brw_set_conditionalmod(p, inst->conditional_mod);
brw_set_predicate_control(p, inst->predicate);
brw_set_predicate_inverse(p, inst->predicate_inverse);
brw_set_flag_reg(p, 0, inst->flag_subreg);
brw_set_saturate(p, inst->saturate);
brw_set_mask_control(p, inst->force_writemask_all);
if (inst->force_uncompressed || dispatch_width == 8) {
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
} else if (inst->force_sechalf) {
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
} else {
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
}
switch (inst->opcode) {
case BRW_OPCODE_MOV:
brw_MOV(p, dst, src[0]);
break;
case BRW_OPCODE_ADD:
brw_ADD(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_MUL:
brw_MUL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_MACH:
brw_set_acc_write_control(p, 1);
brw_MACH(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_MAD:
assert(brw->gen >= 6);
brw_set_access_mode(p, BRW_ALIGN_16);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_MAD(p, dst, src[0], src[1], src[2]);
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_MAD(p, sechalf(dst), sechalf(src[0]), sechalf(src[1]), sechalf(src[2]));
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
} else {
brw_MAD(p, dst, src[0], src[1], src[2]);
}
brw_set_access_mode(p, BRW_ALIGN_1);
break;
case BRW_OPCODE_LRP:
assert(brw->gen >= 6);
brw_set_access_mode(p, BRW_ALIGN_16);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_LRP(p, dst, src[0], src[1], src[2]);
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_LRP(p, sechalf(dst), sechalf(src[0]), sechalf(src[1]), sechalf(src[2]));
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
} else {
brw_LRP(p, dst, src[0], src[1], src[2]);
}
brw_set_access_mode(p, BRW_ALIGN_1);
break;
case BRW_OPCODE_FRC:
brw_FRC(p, dst, src[0]);
break;
case BRW_OPCODE_RNDD:
brw_RNDD(p, dst, src[0]);
break;
case BRW_OPCODE_RNDE:
brw_RNDE(p, dst, src[0]);
break;
case BRW_OPCODE_RNDZ:
brw_RNDZ(p, dst, src[0]);
break;
case BRW_OPCODE_AND:
brw_AND(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_OR:
brw_OR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_XOR:
brw_XOR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_NOT:
brw_NOT(p, dst, src[0]);
break;
case BRW_OPCODE_ASR:
brw_ASR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_SHR:
brw_SHR(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_SHL:
brw_SHL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_F32TO16:
assert(brw->gen >= 7);
brw_F32TO16(p, dst, src[0]);
break;
case BRW_OPCODE_F16TO32:
assert(brw->gen >= 7);
brw_F16TO32(p, dst, src[0]);
break;
case BRW_OPCODE_CMP:
brw_CMP(p, dst, inst->conditional_mod, src[0], src[1]);
break;
case BRW_OPCODE_SEL:
brw_SEL(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_BFREV:
assert(brw->gen >= 7);
/* BFREV only supports UD type for src and dst. */
brw_BFREV(p, retype(dst, BRW_REGISTER_TYPE_UD),
retype(src[0], BRW_REGISTER_TYPE_UD));
break;
case BRW_OPCODE_FBH:
assert(brw->gen >= 7);
/* FBH only supports UD type for dst. */
brw_FBH(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_FBL:
assert(brw->gen >= 7);
/* FBL only supports UD type for dst. */
brw_FBL(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_CBIT:
assert(brw->gen >= 7);
/* CBIT only supports UD type for dst. */
brw_CBIT(p, retype(dst, BRW_REGISTER_TYPE_UD), src[0]);
break;
case BRW_OPCODE_ADDC:
assert(brw->gen >= 7);
brw_set_acc_write_control(p, 1);
brw_ADDC(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_SUBB:
assert(brw->gen >= 7);
brw_set_acc_write_control(p, 1);
brw_SUBB(p, dst, src[0], src[1]);
brw_set_acc_write_control(p, 0);
break;
case BRW_OPCODE_BFE:
assert(brw->gen >= 7);
brw_set_access_mode(p, BRW_ALIGN_16);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_BFE(p, dst, src[0], src[1], src[2]);
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_BFE(p, sechalf(dst), sechalf(src[0]), sechalf(src[1]), sechalf(src[2]));
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
} else {
brw_BFE(p, dst, src[0], src[1], src[2]);
}
brw_set_access_mode(p, BRW_ALIGN_1);
break;
case BRW_OPCODE_BFI1:
assert(brw->gen >= 7);
brw_BFI1(p, dst, src[0], src[1]);
break;
case BRW_OPCODE_BFI2:
assert(brw->gen >= 7);
brw_set_access_mode(p, BRW_ALIGN_16);
if (dispatch_width == 16) {
brw_set_compression_control(p, BRW_COMPRESSION_NONE);
brw_BFI2(p, dst, src[0], src[1], src[2]);
brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF);
brw_BFI2(p, sechalf(dst), sechalf(src[0]), sechalf(src[1]), sechalf(src[2]));
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
} else {
brw_BFI2(p, dst, src[0], src[1], src[2]);
}
brw_set_access_mode(p, BRW_ALIGN_1);
break;
case BRW_OPCODE_IF:
if (inst->src[0].file != BAD_FILE) {
/* The instruction has an embedded compare (only allowed on gen6) */
assert(brw->gen == 6);
gen6_IF(p, inst->conditional_mod, src[0], src[1]);
} else {
brw_IF(p, dispatch_width == 16 ? BRW_EXECUTE_16 : BRW_EXECUTE_8);
}
break;
case BRW_OPCODE_ELSE:
brw_ELSE(p);
break;
case BRW_OPCODE_ENDIF:
brw_ENDIF(p);
break;
case BRW_OPCODE_DO:
brw_DO(p, BRW_EXECUTE_8);
break;
case BRW_OPCODE_BREAK:
brw_BREAK(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
break;
case BRW_OPCODE_CONTINUE:
/* FINISHME: We need to write the loop instruction support still. */
if (brw->gen >= 6)
gen6_CONT(p);
else
brw_CONT(p);
brw_set_predicate_control(p, BRW_PREDICATE_NONE);
break;
case BRW_OPCODE_WHILE:
brw_WHILE(p);
break;
case SHADER_OPCODE_RCP:
case SHADER_OPCODE_RSQ:
case SHADER_OPCODE_SQRT:
case SHADER_OPCODE_EXP2:
case SHADER_OPCODE_LOG2:
case SHADER_OPCODE_SIN:
case SHADER_OPCODE_COS:
if (brw->gen >= 7) {
generate_math1_gen7(inst, dst, src[0]);
} else if (brw->gen == 6) {
generate_math1_gen6(inst, dst, src[0]);
} else if (brw->gen == 5 || brw->is_g4x) {
generate_math_g45(inst, dst, src[0]);
} else {
generate_math_gen4(inst, dst, src[0]);
}
break;
case SHADER_OPCODE_INT_QUOTIENT:
case SHADER_OPCODE_INT_REMAINDER:
case SHADER_OPCODE_POW:
if (brw->gen >= 7) {
generate_math2_gen7(inst, dst, src[0], src[1]);
} else if (brw->gen == 6) {
generate_math2_gen6(inst, dst, src[0], src[1]);
} else {
generate_math_gen4(inst, dst, src[0]);
}
break;
case FS_OPCODE_PIXEL_X:
generate_pixel_xy(dst, true);
break;
case FS_OPCODE_PIXEL_Y:
generate_pixel_xy(dst, false);
break;
case FS_OPCODE_CINTERP:
brw_MOV(p, dst, src[0]);
break;
case FS_OPCODE_LINTERP:
generate_linterp(inst, dst, src);
break;
case SHADER_OPCODE_TEX:
case FS_OPCODE_TXB:
case SHADER_OPCODE_TXD:
case SHADER_OPCODE_TXF:
case SHADER_OPCODE_TXF_MS:
case SHADER_OPCODE_TXL:
case SHADER_OPCODE_TXS:
case SHADER_OPCODE_LOD:
case SHADER_OPCODE_TG4:
generate_tex(inst, dst, src[0]);
break;
case FS_OPCODE_DDX:
generate_ddx(inst, dst, src[0]);
break;
case FS_OPCODE_DDY:
/* Make sure fp->UsesDFdy flag got set (otherwise there's no
* guarantee that c->key.render_to_fbo is set).
*/
assert(fp->UsesDFdy);
generate_ddy(inst, dst, src[0], c->key.render_to_fbo);
break;
case FS_OPCODE_SPILL:
generate_spill(inst, src[0]);
break;
case FS_OPCODE_UNSPILL:
generate_unspill(inst, dst);
break;
case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD:
generate_uniform_pull_constant_load(inst, dst, src[0], src[1]);
break;
case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD_GEN7:
generate_uniform_pull_constant_load_gen7(inst, dst, src[0], src[1]);
break;
case FS_OPCODE_VARYING_PULL_CONSTANT_LOAD:
generate_varying_pull_constant_load(inst, dst, src[0], src[1]);
break;
case FS_OPCODE_VARYING_PULL_CONSTANT_LOAD_GEN7:
generate_varying_pull_constant_load_gen7(inst, dst, src[0], src[1]);
break;
case FS_OPCODE_FB_WRITE:
generate_fb_write(inst);
break;
case FS_OPCODE_MOV_DISPATCH_TO_FLAGS:
generate_mov_dispatch_to_flags(inst);
break;
case FS_OPCODE_DISCARD_JUMP:
generate_discard_jump(inst);
break;
case SHADER_OPCODE_SHADER_TIME_ADD:
generate_shader_time_add(inst, src[0], src[1], src[2]);
break;
case FS_OPCODE_SET_SIMD4X2_OFFSET:
generate_set_simd4x2_offset(inst, dst, src[0]);
break;
case FS_OPCODE_PACK_HALF_2x16_SPLIT:
generate_pack_half_2x16_split(inst, dst, src[0], src[1]);
break;
case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X:
case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y:
generate_unpack_half_2x16_split(inst, dst, src[0]);
break;
case FS_OPCODE_PLACEHOLDER_HALT:
/* This is the place where the final HALT needs to be inserted if
* we've emitted any discards. If not, this will emit no code.
*/
patch_discard_jumps_to_fb_writes();
break;
default:
if (inst->opcode < (int) ARRAY_SIZE(opcode_descs)) {
_mesa_problem(ctx, "Unsupported opcode `%s' in FS",
opcode_descs[inst->opcode].name);
} else {
_mesa_problem(ctx, "Unsupported opcode %d in FS", inst->opcode);
}
abort();
}
if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
brw_dump_compile(p, stdout,
last_native_insn_offset, p->next_insn_offset);
foreach_list(node, &cfg->block_list) {
bblock_link *link = (bblock_link *)node;
bblock_t *block = link->block;
if (block->end == inst) {
printf(" END B%d", block->block_num);
foreach_list(successor_node, &block->children) {
bblock_link *successor_link =
(bblock_link *)successor_node;
bblock_t *successor_block = successor_link->block;
printf(" ->B%d", successor_block->block_num);
}
printf("\n");
}
}
}
last_native_insn_offset = p->next_insn_offset;
}
if (unlikely(INTEL_DEBUG & DEBUG_WM)) {
printf("\n");
}
brw_set_uip_jip(p);
/* OK, while the INTEL_DEBUG=wm above is very nice for debugging FS
* emit issues, it doesn't get the jump distances into the output,
* which is often something we want to debug. So this is here in
* case you're doing that.
*/
if (0) {
brw_dump_compile(p, stdout, 0, p->next_insn_offset);
}
}
const unsigned *
fs_generator::generate_assembly(exec_list *simd8_instructions,
exec_list *simd16_instructions,
unsigned *assembly_size)
{
dispatch_width = 8;
generate_code(simd8_instructions);
if (simd16_instructions) {
/* We have to do a compaction pass now, or the one at the end of
* execution will squash down where our prog_offset start needs
* to be.
*/
brw_compact_instructions(p);
/* align to 64 byte boundary. */
while ((p->nr_insn * sizeof(struct brw_instruction)) % 64) {
brw_NOP(p);
}
/* Save off the start of this 16-wide program */
c->prog_data.prog_offset_16 = p->nr_insn * sizeof(struct brw_instruction);
brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED);
dispatch_width = 16;
generate_code(simd16_instructions);
}
return brw_get_program(p, assembly_size);
}
|