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
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
|
/*
* Copyright (C) 2019 Connor Abbott <cwabbott0@gmail.com>
* Copyright (C) 2019 Lyude Paul <thatslyude@gmail.com>
* Copyright (C) 2019 Ryan Houdek <Sonicadvance1@gmail.com>
*
* 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.
*/
#include <stdbool.h>
#include <stdio.h>
#include <stdint.h>
#include <assert.h>
#include <inttypes.h>
#include <string.h>
#include "bifrost.h"
#include "disassemble.h"
#include "bi_print.h"
#include "util/macros.h"
// return bits (high, lo]
static uint64_t bits(uint32_t word, unsigned lo, unsigned high)
{
if (high == 32)
return word >> lo;
return (word & ((1 << high) - 1)) >> lo;
}
// each of these structs represents an instruction that's dispatched in one
// cycle. Note that these instructions are packed in funny ways within the
// clause, hence the need for a separate struct.
struct bifrost_alu_inst {
uint32_t fma_bits;
uint32_t add_bits;
uint64_t reg_bits;
};
static unsigned get_reg0(struct bifrost_regs regs)
{
if (regs.ctrl == 0)
return regs.reg0 | ((regs.reg1 & 0x1) << 5);
return regs.reg0 <= regs.reg1 ? regs.reg0 : 63 - regs.reg0;
}
static unsigned get_reg1(struct bifrost_regs regs)
{
return regs.reg0 <= regs.reg1 ? regs.reg1 : 63 - regs.reg1;
}
// this represents the decoded version of the ctrl register field.
struct bifrost_reg_ctrl {
bool read_reg0;
bool read_reg1;
bool read_reg3;
enum bifrost_reg_write_unit fma_write_unit;
enum bifrost_reg_write_unit add_write_unit;
bool clause_start;
};
enum fma_src_type {
FMA_ONE_SRC,
FMA_TWO_SRC,
FMA_FADD,
FMA_FMINMAX,
FMA_FADD16,
FMA_FMINMAX16,
FMA_FCMP,
FMA_FCMP16,
FMA_THREE_SRC,
FMA_SHIFT,
FMA_FMA,
FMA_FMA16,
FMA_CSEL4,
FMA_FMA_MSCALE,
FMA_SHIFT_ADD64,
};
struct fma_op_info {
bool extended;
unsigned op;
char name[30];
enum fma_src_type src_type;
};
enum add_src_type {
ADD_ONE_SRC,
ADD_TWO_SRC,
ADD_FADD,
ADD_FMINMAX,
ADD_FADD16,
ADD_FMINMAX16,
ADD_THREE_SRC,
ADD_SHIFT,
ADD_FADDMscale,
ADD_FCMP,
ADD_FCMP16,
ADD_TEX_COMPACT, // texture instruction with embedded sampler
ADD_TEX, // texture instruction with sampler/etc. in uniform port
ADD_VARYING_INTERP,
ADD_BLENDING,
ADD_LOAD_ATTR,
ADD_VARYING_ADDRESS,
ADD_BRANCH,
};
struct add_op_info {
unsigned op;
char name[30];
enum add_src_type src_type;
bool has_data_reg;
};
void dump_header(FILE *fp, struct bifrost_header header, bool verbose);
void dump_instr(FILE *fp, const struct bifrost_alu_inst *instr,
struct bifrost_regs next_regs, uint64_t *consts,
unsigned data_reg, unsigned offset, bool verbose);
bool dump_clause(FILE *fp, uint32_t *words, unsigned *size, unsigned offset, bool verbose);
void dump_header(FILE *fp, struct bifrost_header header, bool verbose)
{
fprintf(fp, "id(%du) ", header.scoreboard_index);
if (header.clause_type != 0) {
const char *name = bi_clause_type_name(header.clause_type);
if (name[0] == '?')
fprintf(fp, "unk%u ", header.clause_type);
else
fprintf(fp, "%s ", name);
}
if (header.scoreboard_deps != 0) {
fprintf(fp, "next-wait(");
bool first = true;
for (unsigned i = 0; i < 8; i++) {
if (header.scoreboard_deps & (1 << i)) {
if (!first) {
fprintf(fp, ", ");
}
fprintf(fp, "%d", i);
first = false;
}
}
fprintf(fp, ") ");
}
if (header.datareg_writebarrier)
fprintf(fp, "data-reg-barrier ");
if (!header.no_end_of_shader)
fprintf(fp, "eos ");
if (!header.back_to_back) {
fprintf(fp, "nbb ");
if (header.branch_cond)
fprintf(fp, "branch-cond ");
else
fprintf(fp, "branch-uncond ");
}
if (header.elide_writes)
fprintf(fp, "we ");
if (header.suppress_inf)
fprintf(fp, "suppress-inf ");
if (header.suppress_nan)
fprintf(fp, "suppress-nan ");
if (header.unk0)
fprintf(fp, "unk0 ");
if (header.unk1)
fprintf(fp, "unk1 ");
if (header.unk2)
fprintf(fp, "unk2 ");
if (header.unk3)
fprintf(fp, "unk3 ");
if (header.unk4)
fprintf(fp, "unk4 ");
fprintf(fp, "\n");
if (verbose) {
fprintf(fp, "# clause type %d, next clause type %d\n",
header.clause_type, header.next_clause_type);
}
}
static struct bifrost_reg_ctrl DecodeRegCtrl(FILE *fp, struct bifrost_regs regs)
{
struct bifrost_reg_ctrl decoded = {};
unsigned ctrl;
if (regs.ctrl == 0) {
ctrl = regs.reg1 >> 2;
decoded.read_reg0 = !(regs.reg1 & 0x2);
decoded.read_reg1 = false;
} else {
ctrl = regs.ctrl;
decoded.read_reg0 = decoded.read_reg1 = true;
}
switch (ctrl) {
case 1:
decoded.fma_write_unit = REG_WRITE_TWO;
break;
case 2:
case 3:
decoded.fma_write_unit = REG_WRITE_TWO;
decoded.read_reg3 = true;
break;
case 4:
decoded.read_reg3 = true;
break;
case 5:
decoded.add_write_unit = REG_WRITE_TWO;
break;
case 6:
decoded.add_write_unit = REG_WRITE_TWO;
decoded.read_reg3 = true;
break;
case 8:
decoded.clause_start = true;
break;
case 9:
decoded.fma_write_unit = REG_WRITE_TWO;
decoded.clause_start = true;
break;
case 11:
break;
case 12:
decoded.read_reg3 = true;
decoded.clause_start = true;
break;
case 13:
decoded.add_write_unit = REG_WRITE_TWO;
decoded.clause_start = true;
break;
case 7:
case 15:
decoded.fma_write_unit = REG_WRITE_THREE;
decoded.add_write_unit = REG_WRITE_TWO;
break;
default:
fprintf(fp, "# unknown reg ctrl %d\n", ctrl);
}
return decoded;
}
// Pass in the add_write_unit or fma_write_unit, and this returns which register
// the ADD/FMA units are writing to
static unsigned GetRegToWrite(enum bifrost_reg_write_unit unit, struct bifrost_regs regs)
{
switch (unit) {
case REG_WRITE_TWO:
return regs.reg2;
case REG_WRITE_THREE:
return regs.reg3;
default: /* REG_WRITE_NONE */
assert(0);
return 0;
}
}
static void dump_regs(FILE *fp, struct bifrost_regs srcs)
{
struct bifrost_reg_ctrl ctrl = DecodeRegCtrl(fp, srcs);
fprintf(fp, "# ");
if (ctrl.read_reg0)
fprintf(fp, "port 0: R%d ", get_reg0(srcs));
if (ctrl.read_reg1)
fprintf(fp, "port 1: R%d ", get_reg1(srcs));
if (ctrl.fma_write_unit == REG_WRITE_TWO)
fprintf(fp, "port 2: R%d (write FMA) ", srcs.reg2);
else if (ctrl.add_write_unit == REG_WRITE_TWO)
fprintf(fp, "port 2: R%d (write ADD) ", srcs.reg2);
if (ctrl.fma_write_unit == REG_WRITE_THREE)
fprintf(fp, "port 3: R%d (write FMA) ", srcs.reg3);
else if (ctrl.add_write_unit == REG_WRITE_THREE)
fprintf(fp, "port 3: R%d (write ADD) ", srcs.reg3);
else if (ctrl.read_reg3)
fprintf(fp, "port 3: R%d (read) ", srcs.reg3);
if (srcs.uniform_const) {
if (srcs.uniform_const & 0x80) {
fprintf(fp, "uniform: U%d", (srcs.uniform_const & 0x7f) * 2);
}
}
fprintf(fp, "\n");
}
static void dump_const_imm(FILE *fp, uint32_t imm)
{
union {
float f;
uint32_t i;
} fi;
fi.i = imm;
fprintf(fp, "0x%08x /* %f */", imm, fi.f);
}
static uint64_t get_const(uint64_t *consts, struct bifrost_regs srcs)
{
unsigned low_bits = srcs.uniform_const & 0xf;
uint64_t imm;
switch (srcs.uniform_const >> 4) {
case 4:
imm = consts[0];
break;
case 5:
imm = consts[1];
break;
case 6:
imm = consts[2];
break;
case 7:
imm = consts[3];
break;
case 2:
imm = consts[4];
break;
case 3:
imm = consts[5];
break;
default:
assert(0);
break;
}
return imm | low_bits;
}
static void dump_uniform_const_src(FILE *fp, struct bifrost_regs srcs, uint64_t *consts, bool high32)
{
if (srcs.uniform_const & 0x80) {
unsigned uniform = (srcs.uniform_const & 0x7f) * 2;
fprintf(fp, "U%d", uniform + (high32 ? 1 : 0));
} else if (srcs.uniform_const >= 0x20) {
uint64_t imm = get_const(consts, srcs);
if (high32)
dump_const_imm(fp, imm >> 32);
else
dump_const_imm(fp, imm);
} else {
switch (srcs.uniform_const) {
case 0:
fprintf(fp, "0");
break;
case 5:
fprintf(fp, "atest-data");
break;
case 6:
fprintf(fp, "sample-ptr");
break;
case 8:
case 9:
case 10:
case 11:
case 12:
case 13:
case 14:
case 15:
fprintf(fp, "blend-descriptor%u", (unsigned) srcs.uniform_const - 8);
break;
default:
fprintf(fp, "unkConst%u", (unsigned) srcs.uniform_const);
break;
}
if (high32)
fprintf(fp, ".y");
else
fprintf(fp, ".x");
}
}
static void dump_src(FILE *fp, unsigned src, struct bifrost_regs srcs, uint64_t *consts, bool isFMA)
{
switch (src) {
case 0:
fprintf(fp, "R%d", get_reg0(srcs));
break;
case 1:
fprintf(fp, "R%d", get_reg1(srcs));
break;
case 2:
fprintf(fp, "R%d", srcs.reg3);
break;
case 3:
if (isFMA)
fprintf(fp, "0");
else
fprintf(fp, "T"); // i.e. the output of FMA this cycle
break;
case 4:
dump_uniform_const_src(fp, srcs, consts, false);
break;
case 5:
dump_uniform_const_src(fp, srcs, consts, true);
break;
case 6:
fprintf(fp, "T0");
break;
case 7:
fprintf(fp, "T1");
break;
}
}
static const struct fma_op_info FMAOpInfos[] = {
{ false, 0x00000, "FMA.f32", FMA_FMA },
{ false, 0x40000, "MAX.f32", FMA_FMINMAX },
{ false, 0x44000, "MIN.f32", FMA_FMINMAX },
{ false, 0x48000, "FCMP.GL", FMA_FCMP },
{ false, 0x4c000, "FCMP.D3D", FMA_FCMP },
{ false, 0x4ff98, "ADD.i32", FMA_TWO_SRC },
{ false, 0x4ffd8, "SUB.i32", FMA_TWO_SRC },
{ false, 0x4fff0, "SUBB.i32", FMA_TWO_SRC },
{ false, 0x50000, "FMA_MSCALE", FMA_FMA_MSCALE },
{ false, 0x58000, "ADD.f32", FMA_FADD },
{ false, 0x5c000, "CSEL4", FMA_CSEL4 },
{ false, 0x5d8d0, "ICMP.D3D.GT.v2i16", FMA_TWO_SRC },
{ false, 0x5d9d0, "UCMP.D3D.GT.v2i16", FMA_TWO_SRC },
{ false, 0x5dad0, "ICMP.D3D.GE.v2i16", FMA_TWO_SRC },
{ false, 0x5dbd0, "UCMP.D3D.GE.v2i16", FMA_TWO_SRC },
{ false, 0x5dcd0, "ICMP.D3D.EQ.v2i16", FMA_TWO_SRC },
{ false, 0x5de40, "ICMP.GL.GT.i32", FMA_TWO_SRC }, // src0 > src1 ? 1 : 0
{ false, 0x5de48, "ICMP.GL.GE.i32", FMA_TWO_SRC },
{ false, 0x5de50, "UCMP.GL.GT.i32", FMA_TWO_SRC },
{ false, 0x5de58, "UCMP.GL.GE.i32", FMA_TWO_SRC },
{ false, 0x5de60, "ICMP.GL.EQ.i32", FMA_TWO_SRC },
{ false, 0x5dec0, "ICMP.D3D.GT.i32", FMA_TWO_SRC }, // src0 > src1 ? ~0 : 0
{ false, 0x5dec8, "ICMP.D3D.GE.i32", FMA_TWO_SRC },
{ false, 0x5ded0, "UCMP.D3D.GT.i32", FMA_TWO_SRC },
{ false, 0x5ded8, "UCMP.D3D.GE.i32", FMA_TWO_SRC },
{ false, 0x5dee0, "ICMP.D3D.EQ.i32", FMA_TWO_SRC },
{ false, 0x60000, "RSHIFT_NAND", FMA_SHIFT },
{ false, 0x61000, "RSHIFT_AND", FMA_SHIFT },
{ false, 0x62000, "LSHIFT_NAND", FMA_SHIFT },
{ false, 0x63000, "LSHIFT_AND", FMA_SHIFT }, // (src0 << src2) & src1
{ false, 0x64000, "RSHIFT_XOR", FMA_SHIFT },
{ false, 0x65200, "LSHIFT_ADD.i32", FMA_THREE_SRC },
{ false, 0x65600, "LSHIFT_SUB.i32", FMA_THREE_SRC }, // (src0 << src2) - src1
{ false, 0x65a00, "LSHIFT_RSUB.i32", FMA_THREE_SRC }, // src1 - (src0 << src2)
{ false, 0x65e00, "RSHIFT_ADD.i32", FMA_THREE_SRC },
{ false, 0x66200, "RSHIFT_SUB.i32", FMA_THREE_SRC },
{ false, 0x66600, "RSHIFT_RSUB.i32", FMA_THREE_SRC },
{ false, 0x66a00, "ARSHIFT_ADD.i32", FMA_THREE_SRC },
{ false, 0x66e00, "ARSHIFT_SUB.i32", FMA_THREE_SRC },
{ false, 0x67200, "ARSHIFT_RSUB.i32", FMA_THREE_SRC },
{ false, 0x80000, "FMA.v2f16", FMA_FMA16 },
{ false, 0xc0000, "MAX.v2f16", FMA_FMINMAX16 },
{ false, 0xc4000, "MIN.v2f16", FMA_FMINMAX16 },
{ false, 0xc8000, "FCMP.GL", FMA_FCMP16 },
{ false, 0xcc000, "FCMP.D3D", FMA_FCMP16 },
{ false, 0xcf900, "ADD.v2i16", FMA_TWO_SRC },
{ false, 0xcfc10, "ADDC.i32", FMA_TWO_SRC },
{ false, 0xcfd80, "ADD.i32.i16.X", FMA_TWO_SRC },
{ false, 0xcfd90, "ADD.i32.u16.X", FMA_TWO_SRC },
{ false, 0xcfdc0, "ADD.i32.i16.Y", FMA_TWO_SRC },
{ false, 0xcfdd0, "ADD.i32.u16.Y", FMA_TWO_SRC },
{ false, 0xd8000, "ADD.v2f16", FMA_FADD16 },
{ false, 0xdc000, "CSEL4.v16", FMA_CSEL4 },
{ false, 0xdd000, "F32_TO_F16", FMA_TWO_SRC },
/* TODO: Combine to bifrost_fma_f2i_i2f16 */
{ true, 0x00046, "F16_TO_I16.XX", FMA_ONE_SRC },
{ true, 0x00047, "F16_TO_U16.XX", FMA_ONE_SRC },
{ true, 0x0004e, "F16_TO_I16.YX", FMA_ONE_SRC },
{ true, 0x0004f, "F16_TO_U16.YX", FMA_ONE_SRC },
{ true, 0x00056, "F16_TO_I16.XY", FMA_ONE_SRC },
{ true, 0x00057, "F16_TO_U16.XY", FMA_ONE_SRC },
{ true, 0x0005e, "F16_TO_I16.YY", FMA_ONE_SRC },
{ true, 0x0005f, "F16_TO_U16.YY", FMA_ONE_SRC },
{ true, 0x000c0, "I16_TO_F16.XX", FMA_ONE_SRC },
{ true, 0x000c1, "U16_TO_F16.XX", FMA_ONE_SRC },
{ true, 0x000c8, "I16_TO_F16.YX", FMA_ONE_SRC },
{ true, 0x000c9, "U16_TO_F16.YX", FMA_ONE_SRC },
{ true, 0x000d0, "I16_TO_F16.XY", FMA_ONE_SRC },
{ true, 0x000d1, "U16_TO_F16.XY", FMA_ONE_SRC },
{ true, 0x000d8, "I16_TO_F16.YY", FMA_ONE_SRC },
{ true, 0x000d9, "U16_TO_F16.YY", FMA_ONE_SRC },
{ true, 0x00136, "F32_TO_I32", FMA_ONE_SRC },
{ true, 0x00137, "F32_TO_U32", FMA_ONE_SRC },
{ true, 0x00178, "I32_TO_F32", FMA_ONE_SRC },
{ true, 0x00179, "U32_TO_F32", FMA_ONE_SRC },
/* TODO: cleanup to use bifrost_fma_int16_to_32 */
{ true, 0x00198, "I16_TO_I32.X", FMA_ONE_SRC },
{ true, 0x00199, "U16_TO_U32.X", FMA_ONE_SRC },
{ true, 0x0019a, "I16_TO_I32.Y", FMA_ONE_SRC },
{ true, 0x0019b, "U16_TO_U32.Y", FMA_ONE_SRC },
{ true, 0x0019c, "I16_TO_F32.X", FMA_ONE_SRC },
{ true, 0x0019d, "U16_TO_F32.X", FMA_ONE_SRC },
{ true, 0x0019e, "I16_TO_F32.Y", FMA_ONE_SRC },
{ true, 0x0019f, "U16_TO_F32.Y", FMA_ONE_SRC },
{ true, 0x001a2, "F16_TO_F32.X", FMA_ONE_SRC },
{ true, 0x001a3, "F16_TO_F32.Y", FMA_ONE_SRC },
{ true, 0x0032c, "NOP", FMA_ONE_SRC },
{ true, 0x0032d, "MOV", FMA_ONE_SRC },
{ true, 0x0032f, "SWZ.YY.v2i16", FMA_ONE_SRC },
{ true, 0x00345, "LOG_FREXPM", FMA_ONE_SRC },
{ true, 0x00365, "FRCP_FREXPM", FMA_ONE_SRC },
{ true, 0x00375, "FSQRT_FREXPM", FMA_ONE_SRC },
{ true, 0x0038d, "FRCP_FREXPE", FMA_ONE_SRC },
{ true, 0x003a5, "FSQRT_FREXPE", FMA_ONE_SRC },
{ true, 0x003ad, "FRSQ_FREXPE", FMA_ONE_SRC },
{ true, 0x003c5, "LOG_FREXPE", FMA_ONE_SRC },
{ true, 0x003fa, "CLZ", FMA_ONE_SRC },
{ true, 0x00b80, "IMAX3", FMA_THREE_SRC },
{ true, 0x00bc0, "UMAX3", FMA_THREE_SRC },
{ true, 0x00c00, "IMIN3", FMA_THREE_SRC },
{ true, 0x00c40, "UMIN3", FMA_THREE_SRC },
{ true, 0x00ec2, "ROUND.v2f16", FMA_ONE_SRC },
{ true, 0x00ec5, "ROUND.f32", FMA_ONE_SRC },
{ true, 0x00f40, "CSEL", FMA_THREE_SRC }, // src2 != 0 ? src1 : src0
{ true, 0x00fc0, "MUX.i32", FMA_THREE_SRC }, // see ADD comment
{ true, 0x01802, "ROUNDEVEN.v2f16", FMA_ONE_SRC },
{ true, 0x01805, "ROUNDEVEN.f32", FMA_ONE_SRC },
{ true, 0x01842, "CEIL.v2f16", FMA_ONE_SRC },
{ true, 0x01845, "CEIL.f32", FMA_ONE_SRC },
{ true, 0x01882, "FLOOR.v2f16", FMA_ONE_SRC },
{ true, 0x01885, "FLOOR.f32", FMA_ONE_SRC },
{ true, 0x018c2, "TRUNC.v2f16", FMA_ONE_SRC },
{ true, 0x018c5, "TRUNC.f32", FMA_ONE_SRC },
{ true, 0x019b0, "ATAN_LDEXP.Y.f32", FMA_TWO_SRC },
{ true, 0x019b8, "ATAN_LDEXP.X.f32", FMA_TWO_SRC },
{ true, 0x01c80, "LSHIFT_ADD_LOW32.u32", FMA_SHIFT_ADD64 },
{ true, 0x01cc0, "LSHIFT_ADD_LOW32.i64", FMA_SHIFT_ADD64 },
{ true, 0x01d80, "LSHIFT_ADD_LOW32.i32", FMA_SHIFT_ADD64 },
{ true, 0x01e00, "SEL.XX.i16", FMA_TWO_SRC },
{ true, 0x01e08, "SEL.YX.i16", FMA_TWO_SRC },
{ true, 0x01e10, "SEL.XY.i16", FMA_TWO_SRC },
{ true, 0x01e18, "SEL.YY.i16", FMA_TWO_SRC },
{ true, 0x01e80, "ADD_FREXPM.f32", FMA_TWO_SRC },
{ true, 0x02000, "SWZ.XXXX.v4i8", FMA_ONE_SRC },
{ true, 0x03e00, "SWZ.ZZZZ.v4i8", FMA_ONE_SRC },
{ true, 0x00800, "IMAD", FMA_THREE_SRC },
{ true, 0x078db, "POPCNT", FMA_ONE_SRC },
};
static struct fma_op_info find_fma_op_info(unsigned op, bool extended)
{
for (unsigned i = 0; i < ARRAY_SIZE(FMAOpInfos); i++) {
unsigned opCmp = ~0;
if (FMAOpInfos[i].extended != extended)
continue;
if (extended)
op &= ~0xe0000;
switch (FMAOpInfos[i].src_type) {
case FMA_ONE_SRC:
opCmp = op;
break;
case FMA_TWO_SRC:
opCmp = op & ~0x7;
break;
case FMA_FCMP:
case FMA_FCMP16:
opCmp = op & ~0x1fff;
break;
case FMA_THREE_SRC:
case FMA_SHIFT_ADD64:
opCmp = op & ~0x3f;
break;
case FMA_FADD:
case FMA_FMINMAX:
case FMA_FADD16:
case FMA_FMINMAX16:
opCmp = op & ~0x3fff;
break;
case FMA_FMA:
case FMA_FMA16:
opCmp = op & ~0x3ffff;
break;
case FMA_CSEL4:
case FMA_SHIFT:
opCmp = op & ~0xfff;
break;
case FMA_FMA_MSCALE:
opCmp = op & ~0x7fff;
break;
default:
opCmp = ~0;
break;
}
if (FMAOpInfos[i].op == opCmp)
return FMAOpInfos[i];
}
struct fma_op_info info;
snprintf(info.name, sizeof(info.name), "op%04x", op);
info.op = op;
info.src_type = FMA_THREE_SRC;
return info;
}
static void dump_fcmp(FILE *fp, unsigned op)
{
switch (op) {
case 0:
fprintf(fp, ".OEQ");
break;
case 1:
fprintf(fp, ".OGT");
break;
case 2:
fprintf(fp, ".OGE");
break;
case 3:
fprintf(fp, ".UNE");
break;
case 4:
fprintf(fp, ".OLT");
break;
case 5:
fprintf(fp, ".OLE");
break;
default:
fprintf(fp, ".unk%d", op);
break;
}
}
static void dump_16swizzle(FILE *fp, unsigned swiz)
{
if (swiz == 2)
return;
fprintf(fp, ".%c%c", "xy"[swiz & 1], "xy"[(swiz >> 1) & 1]);
}
static void dump_fma_expand_src0(FILE *fp, unsigned ctrl)
{
switch (ctrl) {
case 3:
case 4:
case 6:
fprintf(fp, ".x");
break;
case 5:
case 7:
fprintf(fp, ".y");
break;
case 0:
case 1:
case 2:
break;
default:
fprintf(fp, ".unk");
break;
}
}
static void dump_fma_expand_src1(FILE *fp, unsigned ctrl)
{
switch (ctrl) {
case 1:
case 3:
fprintf(fp, ".x");
break;
case 2:
case 4:
case 5:
fprintf(fp, ".y");
break;
case 0:
case 6:
case 7:
break;
default:
fprintf(fp, ".unk");
break;
}
}
static void dump_fma(FILE *fp, uint64_t word, struct bifrost_regs regs, struct bifrost_regs next_regs, uint64_t *consts, bool verbose)
{
if (verbose) {
fprintf(fp, "# FMA: %016" PRIx64 "\n", word);
}
struct bifrost_fma_inst FMA;
memcpy((char *) &FMA, (char *) &word, sizeof(struct bifrost_fma_inst));
struct fma_op_info info = find_fma_op_info(FMA.op, (FMA.op & 0xe0000) == 0xe0000);
fprintf(fp, "%s", info.name);
if (info.src_type == FMA_FADD ||
info.src_type == FMA_FMINMAX ||
info.src_type == FMA_FMA ||
info.src_type == FMA_FADD16 ||
info.src_type == FMA_FMINMAX16 ||
info.src_type == FMA_FMA16) {
fprintf(fp, "%s", bi_output_mod_name(bits(FMA.op, 12, 14)));
switch (info.src_type) {
case FMA_FADD:
case FMA_FMA:
case FMA_FADD16:
case FMA_FMA16:
fprintf(fp, "%s", bi_round_mode_name(bits(FMA.op, 10, 12)));
break;
case FMA_FMINMAX:
case FMA_FMINMAX16:
fprintf(fp, "%s", bi_minmax_mode_name(bits(FMA.op, 10, 12)));
break;
default:
assert(0);
}
} else if (info.src_type == FMA_FCMP || info.src_type == FMA_FCMP16) {
dump_fcmp(fp, bits(FMA.op, 10, 13));
if (info.src_type == FMA_FCMP)
fprintf(fp, ".f32");
else
fprintf(fp, ".v2f16");
} else if (info.src_type == FMA_FMA_MSCALE) {
if (FMA.op & (1 << 11)) {
switch ((FMA.op >> 9) & 0x3) {
case 0:
/* This mode seems to do a few things:
* - Makes 0 * infinity (and incidentally 0 * nan) return 0,
* since generating a nan would poison the result of
* 1/infinity and 1/0.
* - Fiddles with which nan is returned in nan * nan,
* presumably to make sure that the same exact nan is
* returned for 1/nan.
*/
fprintf(fp, ".rcp_mode");
break;
case 3:
/* Similar to the above, but src0 always wins when multiplying
* 0 by infinity.
*/
fprintf(fp, ".sqrt_mode");
break;
default:
fprintf(fp, ".unk%d_mode", (int) (FMA.op >> 9) & 0x3);
}
} else {
fprintf(fp, "%s", bi_output_mod_name(bits(FMA.op, 9, 11)));
}
} else if (info.src_type == FMA_SHIFT) {
struct bifrost_shift_fma shift;
memcpy(&shift, &FMA, sizeof(shift));
if (shift.half == 0x7)
fprintf(fp, ".v2i16");
else if (shift.half == 0)
fprintf(fp, ".i32");
else if (shift.half == 0x4)
fprintf(fp, ".v4i8");
else
fprintf(fp, ".unk%u", shift.half);
if (!shift.unk)
fprintf(fp, ".no_unk");
if (shift.invert_1)
fprintf(fp, ".invert_1");
if (shift.invert_2)
fprintf(fp, ".invert_2");
}
fprintf(fp, " ");
struct bifrost_reg_ctrl next_ctrl = DecodeRegCtrl(fp, next_regs);
if (next_ctrl.fma_write_unit != REG_WRITE_NONE) {
fprintf(fp, "{R%d, T0}, ", GetRegToWrite(next_ctrl.fma_write_unit, next_regs));
} else {
fprintf(fp, "T0, ");
}
switch (info.src_type) {
case FMA_ONE_SRC:
dump_src(fp, FMA.src0, regs, consts, true);
break;
case FMA_TWO_SRC:
dump_src(fp, FMA.src0, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
break;
case FMA_FADD:
case FMA_FMINMAX:
if (FMA.op & 0x10)
fprintf(fp, "-");
if (FMA.op & 0x200)
fprintf(fp, "abs(");
dump_src(fp, FMA.src0, regs, consts, true);
dump_fma_expand_src0(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & 0x200)
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & 0x20)
fprintf(fp, "-");
if (FMA.op & 0x8)
fprintf(fp, "abs(");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_fma_expand_src1(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & 0x8)
fprintf(fp, ")");
break;
case FMA_FADD16:
case FMA_FMINMAX16: {
bool abs1 = FMA.op & 0x8;
bool abs2 = (FMA.op & 0x7) < FMA.src0;
if (FMA.op & 0x10)
fprintf(fp, "-");
if (abs1 || abs2)
fprintf(fp, "abs(");
dump_src(fp, FMA.src0, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 6) & 0x3);
if (abs1 || abs2)
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & 0x20)
fprintf(fp, "-");
if (abs1 && abs2)
fprintf(fp, "abs(");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 8) & 0x3);
if (abs1 && abs2)
fprintf(fp, ")");
break;
}
case FMA_FCMP:
if (FMA.op & 0x200)
fprintf(fp, "abs(");
dump_src(fp, FMA.src0, regs, consts, true);
dump_fma_expand_src0(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & 0x200)
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & 0x20)
fprintf(fp, "-");
if (FMA.op & 0x8)
fprintf(fp, "abs(");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_fma_expand_src1(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & 0x8)
fprintf(fp, ")");
break;
case FMA_FCMP16:
dump_src(fp, FMA.src0, regs, consts, true);
// Note: this is kinda a guess, I haven't seen the blob set this to
// anything other than the identity, but it matches FMA_TWO_SRCFmod16
dump_16swizzle(fp, (FMA.op >> 6) & 0x3);
fprintf(fp, ", ");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 8) & 0x3);
break;
case FMA_SHIFT_ADD64:
dump_src(fp, FMA.src0, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
fprintf(fp, ", ");
fprintf(fp, "shift:%u", (FMA.op >> 3) & 0x7);
break;
case FMA_THREE_SRC:
dump_src(fp, FMA.src0, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, (FMA.op >> 3) & 0x7, regs, consts, true);
break;
case FMA_SHIFT: {
struct bifrost_shift_fma shift;
memcpy(&shift, &FMA, sizeof(shift));
dump_src(fp, shift.src0, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, shift.src1, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, shift.src2, regs, consts, true);
break;
}
case FMA_FMA:
if (FMA.op & (1 << 14))
fprintf(fp, "-");
if (FMA.op & (1 << 9))
fprintf(fp, "abs(");
dump_src(fp, FMA.src0, regs, consts, true);
dump_fma_expand_src0(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & (1 << 9))
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & (1 << 16))
fprintf(fp, "abs(");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_fma_expand_src1(fp, (FMA.op >> 6) & 0x7);
if (FMA.op & (1 << 16))
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & (1 << 15))
fprintf(fp, "-");
if (FMA.op & (1 << 17))
fprintf(fp, "abs(");
dump_src(fp, (FMA.op >> 3) & 0x7, regs, consts, true);
if (FMA.op & (1 << 17))
fprintf(fp, ")");
break;
case FMA_FMA16:
if (FMA.op & (1 << 14))
fprintf(fp, "-");
dump_src(fp, FMA.src0, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 6) & 0x3);
fprintf(fp, ", ");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 8) & 0x3);
fprintf(fp, ", ");
if (FMA.op & (1 << 15))
fprintf(fp, "-");
dump_src(fp, (FMA.op >> 3) & 0x7, regs, consts, true);
dump_16swizzle(fp, (FMA.op >> 16) & 0x3);
break;
case FMA_CSEL4: {
struct bifrost_csel4 csel;
memcpy(&csel, &FMA, sizeof(csel));
fprintf(fp, ".%s ", bi_csel_cond_name(csel.cond));
dump_src(fp, csel.src0, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, csel.src1, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, csel.src2, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, csel.src3, regs, consts, true);
break;
}
case FMA_FMA_MSCALE:
if (FMA.op & (1 << 12))
fprintf(fp, "abs(");
dump_src(fp, FMA.src0, regs, consts, true);
if (FMA.op & (1 << 12))
fprintf(fp, ")");
fprintf(fp, ", ");
if (FMA.op & (1 << 13))
fprintf(fp, "-");
dump_src(fp, FMA.op & 0x7, regs, consts, true);
fprintf(fp, ", ");
if (FMA.op & (1 << 14))
fprintf(fp, "-");
dump_src(fp, (FMA.op >> 3) & 0x7, regs, consts, true);
fprintf(fp, ", ");
dump_src(fp, (FMA.op >> 6) & 0x7, regs, consts, true);
break;
}
fprintf(fp, "\n");
}
static const struct add_op_info add_op_infos[] = {
{ 0x00000, "MAX.f32", ADD_FMINMAX },
{ 0x02000, "MIN.f32", ADD_FMINMAX },
{ 0x04000, "ADD.f32", ADD_FADD },
{ 0x06000, "FCMP.GL", ADD_FCMP },
{ 0x07000, "FCMP.D3D", ADD_FCMP },
{ 0x07856, "F16_TO_I16", ADD_ONE_SRC },
{ 0x07857, "F16_TO_U16", ADD_ONE_SRC },
{ 0x078c0, "I16_TO_F16.XX", ADD_ONE_SRC },
{ 0x078c1, "U16_TO_F16.XX", ADD_ONE_SRC },
{ 0x078c8, "I16_TO_F16.YX", ADD_ONE_SRC },
{ 0x078c9, "U16_TO_F16.YX", ADD_ONE_SRC },
{ 0x078d0, "I16_TO_F16.XY", ADD_ONE_SRC },
{ 0x078d1, "U16_TO_F16.XY", ADD_ONE_SRC },
{ 0x078d8, "I16_TO_F16.YY", ADD_ONE_SRC },
{ 0x078d9, "U16_TO_F16.YY", ADD_ONE_SRC },
{ 0x07909, "B1_TO_F16", ADD_ONE_SRC },
{ 0x07936, "F32_TO_I32", ADD_ONE_SRC },
{ 0x07937, "F32_TO_U32", ADD_ONE_SRC },
{ 0x07971, "B1_TO_F32", ADD_ONE_SRC },
{ 0x07978, "I32_TO_F32", ADD_ONE_SRC },
{ 0x07979, "U32_TO_F32", ADD_ONE_SRC },
{ 0x07998, "I16_TO_I32.X", ADD_ONE_SRC },
{ 0x07999, "U16_TO_U32.X", ADD_ONE_SRC },
{ 0x0799a, "I16_TO_I32.Y", ADD_ONE_SRC },
{ 0x0799b, "U16_TO_U32.Y", ADD_ONE_SRC },
{ 0x0799c, "I16_TO_F32.X", ADD_ONE_SRC },
{ 0x0799d, "U16_TO_F32.X", ADD_ONE_SRC },
{ 0x0799e, "I16_TO_F32.Y", ADD_ONE_SRC },
{ 0x0799f, "U16_TO_F32.Y", ADD_ONE_SRC },
{ 0x079a2, "F16_TO_F32.X", ADD_ONE_SRC },
{ 0x079a3, "F16_TO_F32.Y", ADD_ONE_SRC },
{ 0x07b2b, "SWZ.YX.v2i16", ADD_ONE_SRC },
{ 0x07b2c, "NOP", ADD_ONE_SRC },
{ 0x07b29, "SWZ.XX.v2i16", ADD_ONE_SRC },
{ 0x07b2d, "MOV", ADD_ONE_SRC },
{ 0x07b2f, "SWZ.YY.v2i16", ADD_ONE_SRC },
{ 0x07b65, "FRCP_FREXPM", ADD_ONE_SRC },
{ 0x07b75, "FSQRT_FREXPM", ADD_ONE_SRC },
{ 0x07b8d, "FRCP_FREXPE", ADD_ONE_SRC },
{ 0x07ba5, "FSQRT_FREXPE", ADD_ONE_SRC },
{ 0x07bad, "FRSQ_FREXPE", ADD_ONE_SRC },
{ 0x07bc5, "FLOG_FREXPE", ADD_ONE_SRC },
{ 0x07d42, "CEIL.v2f16", ADD_ONE_SRC },
{ 0x07d45, "CEIL.f32", ADD_ONE_SRC },
{ 0x07d82, "FLOOR.v2f16", ADD_ONE_SRC },
{ 0x07d85, "FLOOR.f32", ADD_ONE_SRC },
{ 0x07dc2, "TRUNC.v2f16", ADD_ONE_SRC },
{ 0x07dc5, "TRUNC.f32", ADD_ONE_SRC },
{ 0x07f18, "LSHIFT_ADD_HIGH32.i32", ADD_TWO_SRC },
{ 0x08000, "LD_ATTR", ADD_LOAD_ATTR, true },
{ 0x0a000, "LD_VAR.32", ADD_VARYING_INTERP, true },
{ 0x0b000, "TEX", ADD_TEX_COMPACT, true },
{ 0x0c188, "LOAD.i32", ADD_TWO_SRC, true },
{ 0x0c1a0, "LD_UBO.i32", ADD_TWO_SRC, true },
{ 0x0c1b8, "LD_SCRATCH.v2i32", ADD_TWO_SRC, true },
{ 0x0c1c8, "LOAD.v2i32", ADD_TWO_SRC, true },
{ 0x0c1e0, "LD_UBO.v2i32", ADD_TWO_SRC, true },
{ 0x0c1f8, "LD_SCRATCH.v2i32", ADD_TWO_SRC, true },
{ 0x0c208, "LOAD.v4i32", ADD_TWO_SRC, true },
{ 0x0c220, "LD_UBO.v4i32", ADD_TWO_SRC, true },
{ 0x0c238, "LD_SCRATCH.v4i32", ADD_TWO_SRC, true },
{ 0x0c248, "STORE.v4i32", ADD_TWO_SRC, true },
{ 0x0c278, "ST_SCRATCH.v4i32", ADD_TWO_SRC, true },
{ 0x0c588, "STORE.i32", ADD_TWO_SRC, true },
{ 0x0c5b8, "ST_SCRATCH.i32", ADD_TWO_SRC, true },
{ 0x0c5c8, "STORE.v2i32", ADD_TWO_SRC, true },
{ 0x0c5f8, "ST_SCRATCH.v2i32", ADD_TWO_SRC, true },
{ 0x0c648, "LOAD.u16", ADD_TWO_SRC, true }, // zero-extends
{ 0x0ca88, "LOAD.v3i32", ADD_TWO_SRC, true },
{ 0x0caa0, "LD_UBO.v3i32", ADD_TWO_SRC, true },
{ 0x0cab8, "LD_SCRATCH.v3i32", ADD_TWO_SRC, true },
{ 0x0cb88, "STORE.v3i32", ADD_TWO_SRC, true },
{ 0x0cbb8, "ST_SCRATCH.v3i32", ADD_TWO_SRC, true },
{ 0x0cc00, "FRCP_FAST.f32", ADD_ONE_SRC },
{ 0x0cc20, "FRSQ_FAST.f32", ADD_ONE_SRC },
{ 0x0cc68, "FLOG2_U.f32", ADD_ONE_SRC },
{ 0x0cd58, "FEXP2_FAST.f32", ADD_ONE_SRC },
{ 0x0ce00, "FRCP_TABLE", ADD_ONE_SRC },
{ 0x0ce10, "FRCP_FAST.f16.X", ADD_ONE_SRC },
{ 0x0ce20, "FRSQ_TABLE", ADD_ONE_SRC },
{ 0x0ce30, "FRCP_FAST.f16.Y", ADD_ONE_SRC },
{ 0x0ce50, "FRSQ_FAST.f16.X", ADD_ONE_SRC },
{ 0x0ce60, "FRCP_APPROX", ADD_ONE_SRC },
{ 0x0ce70, "FRSQ_FAST.f16.Y", ADD_ONE_SRC },
{ 0x0cf40, "ATAN_ASSIST", ADD_TWO_SRC },
{ 0x0cf48, "ATAN_TABLE", ADD_TWO_SRC },
{ 0x0cf50, "SIN_TABLE", ADD_ONE_SRC },
{ 0x0cf51, "COS_TABLE", ADD_ONE_SRC },
{ 0x0cf58, "EXP_TABLE", ADD_ONE_SRC },
{ 0x0cf60, "FLOG2_TABLE", ADD_ONE_SRC },
{ 0x0cf64, "FLOGE_TABLE", ADD_ONE_SRC },
{ 0x0d000, "BRANCH", ADD_BRANCH },
{ 0x0e8c0, "MUX", ADD_THREE_SRC },
{ 0x0e9b0, "ATAN_LDEXP.Y.f32", ADD_TWO_SRC },
{ 0x0e9b8, "ATAN_LDEXP.X.f32", ADD_TWO_SRC },
{ 0x0ea60, "SEL.XX.i16", ADD_TWO_SRC },
{ 0x0ea70, "SEL.XY.i16", ADD_TWO_SRC },
{ 0x0ea68, "SEL.YX.i16", ADD_TWO_SRC },
{ 0x0ea78, "SEL.YY.i16", ADD_TWO_SRC },
{ 0x0ec00, "F32_TO_F16", ADD_TWO_SRC },
{ 0x0e840, "CSEL.64", ADD_THREE_SRC }, // u2u32(src2) ? src0 : src1
{ 0x0e940, "CSEL.8", ADD_THREE_SRC }, // (src2 != 0) ? src0 : src1
{ 0x0f640, "ICMP.GL.GT", ADD_TWO_SRC }, // src0 > src1 ? 1 : 0
{ 0x0f648, "ICMP.GL.GE", ADD_TWO_SRC },
{ 0x0f650, "UCMP.GL.GT", ADD_TWO_SRC },
{ 0x0f658, "UCMP.GL.GE", ADD_TWO_SRC },
{ 0x0f660, "ICMP.GL.EQ", ADD_TWO_SRC },
{ 0x0f669, "ICMP.GL.NEQ", ADD_TWO_SRC },
{ 0x0f690, "UCMP.8.GT", ADD_TWO_SRC },
{ 0x0f698, "UCMP.8.GE", ADD_TWO_SRC },
{ 0x0f6a8, "ICMP.8.NE", ADD_TWO_SRC },
{ 0x0f6c0, "ICMP.D3D.GT", ADD_TWO_SRC }, // src0 > src1 ? ~0 : 0
{ 0x0f6c8, "ICMP.D3D.GE", ADD_TWO_SRC },
{ 0x0f6d0, "UCMP.D3D.GT", ADD_TWO_SRC },
{ 0x0f6d8, "UCMP.D3D.GE", ADD_TWO_SRC },
{ 0x0f6e0, "ICMP.D3D.EQ", ADD_TWO_SRC },
{ 0x0f700, "ICMP.64.GT.PT1", ADD_TWO_SRC },
{ 0x0f708, "ICMP.64.GE.PT1", ADD_TWO_SRC },
{ 0x0f710, "UCMP.64.GT.PT1", ADD_TWO_SRC },
{ 0x0f718, "UCMP.64.GE.PT1", ADD_TWO_SRC },
{ 0x0f720, "ICMP.64.EQ.PT1", ADD_TWO_SRC },
{ 0x0f728, "ICMP.64.NE.PT1", ADD_TWO_SRC },
{ 0x0f7c0, "ICMP.64.PT2", ADD_THREE_SRC }, // src3 = result of PT1
{ 0x10000, "MAX.v2f16", ADD_FMINMAX16 },
{ 0x11000, "ADD_MSCALE.f32", ADD_FADDMscale },
{ 0x12000, "MIN.v2f16", ADD_FMINMAX16 },
{ 0x14000, "ADD.v2f16", ADD_FADD16 },
{ 0x16000, "FCMP.GL", ADD_FCMP16 },
{ 0x17000, "FCMP.D3D", ADD_FCMP16 },
{ 0x17880, "ADD.v4i8", ADD_TWO_SRC },
{ 0x178c0, "ADD.i32", ADD_TWO_SRC },
{ 0x17900, "ADD.v2i16", ADD_TWO_SRC },
{ 0x17ac0, "SUB.i32", ADD_TWO_SRC },
{ 0x17c10, "ADDC.i32", ADD_TWO_SRC }, // adds src0 to the bottom bit of src1
{ 0x17d80, "ADD.i32.i16.X", ADD_TWO_SRC },
{ 0x17d90, "ADD.i32.u16.X", ADD_TWO_SRC },
{ 0x17dc0, "ADD.i32.i16.Y", ADD_TWO_SRC },
{ 0x17dd0, "ADD.i32.u16.Y", ADD_TWO_SRC },
{ 0x18000, "LD_VAR_ADDR", ADD_VARYING_ADDRESS, true },
{ 0x19181, "DISCARD.FEQ.f32", ADD_TWO_SRC, true },
{ 0x19189, "DISCARD.FNE.f32", ADD_TWO_SRC, true },
{ 0x1918C, "DISCARD.GL.f32", ADD_TWO_SRC, true }, /* Consumes ICMP.GL/etc with fixed 0 argument */
{ 0x19190, "DISCARD.FLE.f32", ADD_TWO_SRC, true },
{ 0x19198, "DISCARD.FLT.f32", ADD_TWO_SRC, true },
{ 0x191e8, "ATEST.f32", ADD_TWO_SRC, true },
{ 0x191f0, "ATEST.X.f16", ADD_TWO_SRC, true },
{ 0x191f8, "ATEST.Y.f16", ADD_TWO_SRC, true },
{ 0x19300, "ST_VAR.v1", ADD_THREE_SRC, true },
{ 0x19340, "ST_VAR.v2", ADD_THREE_SRC, true },
{ 0x19380, "ST_VAR.v3", ADD_THREE_SRC, true },
{ 0x193c0, "ST_VAR.v4", ADD_THREE_SRC, true },
{ 0x1952c, "BLEND", ADD_BLENDING, true },
{ 0x1a000, "LD_VAR.16", ADD_VARYING_INTERP, true },
{ 0x1ae60, "TEX", ADD_TEX, true },
{ 0x1b000, "TEX.f16", ADD_TEX_COMPACT, true },
{ 0x1c000, "RSHIFT_NAND.i32", ADD_SHIFT },
{ 0x1c400, "RSHIFT_AND.i32", ADD_SHIFT },
{ 0x1c800, "LSHIFT_NAND.i32", ADD_SHIFT },
{ 0x1cc00, "LSHIFT_AND.i32", ADD_SHIFT },
{ 0x1d000, "RSHIFT_XOR.i32", ADD_SHIFT },
{ 0x1d400, "LSHIFT_ADD.i32", ADD_SHIFT },
{ 0x1d800, "RSHIFT_SUB.i32", ADD_SHIFT },
{ 0x1dd18, "OR.i32", ADD_TWO_SRC },
{ 0x1dd20, "AND.i32", ADD_TWO_SRC },
{ 0x1dd60, "LSHIFT.i32", ADD_TWO_SRC },
{ 0x1dd50, "XOR.i32", ADD_TWO_SRC },
{ 0x1dd80, "RSHIFT.i32", ADD_TWO_SRC },
{ 0x1dda0, "ARSHIFT.i32", ADD_TWO_SRC },
};
static struct add_op_info find_add_op_info(unsigned op)
{
for (unsigned i = 0; i < ARRAY_SIZE(add_op_infos); i++) {
unsigned opCmp = ~0;
switch (add_op_infos[i].src_type) {
case ADD_ONE_SRC:
case ADD_BLENDING:
opCmp = op;
break;
case ADD_TWO_SRC:
opCmp = op & ~0x7;
break;
case ADD_THREE_SRC:
opCmp = op & ~0x3f;
break;
case ADD_SHIFT:
opCmp = op & ~0x3ff;
break;
case ADD_TEX:
opCmp = op & ~0xf;
break;
case ADD_FADD:
case ADD_FMINMAX:
case ADD_FADD16:
opCmp = op & ~0x1fff;
break;
case ADD_FMINMAX16:
case ADD_FADDMscale:
opCmp = op & ~0xfff;
break;
case ADD_FCMP:
case ADD_FCMP16:
opCmp = op & ~0x7ff;
break;
case ADD_TEX_COMPACT:
opCmp = op & ~0x3ff;
break;
case ADD_VARYING_INTERP:
opCmp = op & ~0x7ff;
break;
case ADD_VARYING_ADDRESS:
opCmp = op & ~0xfff;
break;
case ADD_LOAD_ATTR:
case ADD_BRANCH:
opCmp = op & ~0xfff;
break;
default:
opCmp = ~0;
break;
}
if (add_op_infos[i].op == opCmp)
return add_op_infos[i];
}
struct add_op_info info;
snprintf(info.name, sizeof(info.name), "op%04x", op);
info.op = op;
info.src_type = ADD_TWO_SRC;
info.has_data_reg = true;
return info;
}
static void dump_add(FILE *fp, uint64_t word, struct bifrost_regs regs,
struct bifrost_regs next_regs, uint64_t *consts,
unsigned data_reg, unsigned offset, bool verbose)
{
if (verbose) {
fprintf(fp, "# ADD: %016" PRIx64 "\n", word);
}
struct bifrost_add_inst ADD;
memcpy((char *) &ADD, (char *) &word, sizeof(ADD));
struct add_op_info info = find_add_op_info(ADD.op);
fprintf(fp, "%s", info.name);
// float16 seems like it doesn't support output modifiers
if (info.src_type == ADD_FADD || info.src_type == ADD_FMINMAX) {
// output modifiers
fprintf(fp, "%s", bi_output_mod_name(bits(ADD.op, 8, 10)));
if (info.src_type == ADD_FADD)
fprintf(fp, "%s", bi_round_mode_name(bits(ADD.op, 10, 12)));
else
fprintf(fp, "%s", bi_minmax_mode_name(bits(ADD.op, 10, 12)));
} else if (info.src_type == ADD_FCMP || info.src_type == ADD_FCMP16) {
dump_fcmp(fp, bits(ADD.op, 3, 6));
if (info.src_type == ADD_FCMP)
fprintf(fp, ".f32");
else
fprintf(fp, ".v2f16");
} else if (info.src_type == ADD_FADDMscale) {
switch ((ADD.op >> 6) & 0x7) {
case 0:
break;
// causes GPU hangs on G71
case 1:
fprintf(fp, ".invalid");
break;
// Same as usual outmod value.
case 2:
fprintf(fp, ".clamp_0_1");
break;
// If src0 is infinite or NaN, flush it to zero so that the other
// source is passed through unmodified.
case 3:
fprintf(fp, ".flush_src0_inf_nan");
break;
// Vice versa.
case 4:
fprintf(fp, ".flush_src1_inf_nan");
break;
// Every other case seems to behave the same as the above?
default:
fprintf(fp, ".unk%d", (ADD.op >> 6) & 0x7);
break;
}
} else if (info.src_type == ADD_VARYING_INTERP) {
if (ADD.op & 0x200)
fprintf(fp, ".reuse");
if (ADD.op & 0x400)
fprintf(fp, ".flat");
fprintf(fp, "%s", bi_interp_mode_name((ADD.op >> 7) & 0x3));
fprintf(fp, ".v%d", ((ADD.op >> 5) & 0x3) + 1);
} else if (info.src_type == ADD_BRANCH) {
enum bifrost_branch_code branchCode = (enum bifrost_branch_code) ((ADD.op >> 6) & 0x3f);
if (branchCode == BR_ALWAYS) {
// unconditional branch
} else {
enum bifrost_branch_cond cond = (enum bifrost_branch_cond) ((ADD.op >> 6) & 0x7);
enum branch_bit_size size = (enum branch_bit_size) ((ADD.op >> 9) & 0x7);
bool portSwapped = (ADD.op & 0x7) < ADD.src0;
// See the comment in branch_bit_size
if (size == BR_SIZE_16YX0)
portSwapped = true;
if (size == BR_SIZE_16YX1)
portSwapped = false;
// These sizes are only for floating point comparisons, so the
// non-floating-point comparisons are reused to encode the flipped
// versions.
if (size == BR_SIZE_32_AND_16X || size == BR_SIZE_32_AND_16Y)
portSwapped = false;
// There's only one argument, so we reuse the extra argument to
// encode this.
if (size == BR_SIZE_ZERO)
portSwapped = !(ADD.op & 1);
switch (cond) {
case BR_COND_LT:
if (portSwapped)
fprintf(fp, ".LT.u");
else
fprintf(fp, ".LT.i");
break;
case BR_COND_LE:
if (size == BR_SIZE_32_AND_16X || size == BR_SIZE_32_AND_16Y) {
fprintf(fp, ".UNE.f");
} else {
if (portSwapped)
fprintf(fp, ".LE.u");
else
fprintf(fp, ".LE.i");
}
break;
case BR_COND_GT:
if (portSwapped)
fprintf(fp, ".GT.u");
else
fprintf(fp, ".GT.i");
break;
case BR_COND_GE:
if (portSwapped)
fprintf(fp, ".GE.u");
else
fprintf(fp, ".GE.i");
break;
case BR_COND_EQ:
if (portSwapped)
fprintf(fp, ".NE.i");
else
fprintf(fp, ".EQ.i");
break;
case BR_COND_OEQ:
if (portSwapped)
fprintf(fp, ".UNE.f");
else
fprintf(fp, ".OEQ.f");
break;
case BR_COND_OGT:
if (portSwapped)
fprintf(fp, ".OGT.unk.f");
else
fprintf(fp, ".OGT.f");
break;
case BR_COND_OLT:
if (portSwapped)
fprintf(fp, ".OLT.unk.f");
else
fprintf(fp, ".OLT.f");
break;
}
switch (size) {
case BR_SIZE_32:
case BR_SIZE_32_AND_16X:
case BR_SIZE_32_AND_16Y:
fprintf(fp, "32");
break;
case BR_SIZE_16XX:
case BR_SIZE_16YY:
case BR_SIZE_16YX0:
case BR_SIZE_16YX1:
fprintf(fp, "16");
break;
case BR_SIZE_ZERO: {
unsigned ctrl = (ADD.op >> 1) & 0x3;
if (ctrl == 0)
fprintf(fp, "32.Z");
else
fprintf(fp, "16.Z");
break;
}
}
}
} else if (info.src_type == ADD_SHIFT) {
struct bifrost_shift_add shift;
memcpy(&shift, &ADD, sizeof(ADD));
if (shift.invert_1)
fprintf(fp, ".invert_1");
if (shift.invert_2)
fprintf(fp, ".invert_2");
if (shift.zero)
fprintf(fp, ".unk%u", shift.zero);
} else if (info.src_type == ADD_VARYING_ADDRESS) {
struct bifrost_ld_var_addr ld;
memcpy(&ld, &ADD, sizeof(ADD));
fprintf(fp, ".%s", bi_ldst_type_name(ld.type));
} else if (info.src_type == ADD_LOAD_ATTR) {
struct bifrost_ld_attr ld;
memcpy(&ld, &ADD, sizeof(ADD));
if (ld.channels)
fprintf(fp, ".v%d%s", ld.channels + 1, bi_ldst_type_name(ld.type));
else
fprintf(fp, ".%s", bi_ldst_type_name(ld.type));
}
fprintf(fp, " ");
struct bifrost_reg_ctrl next_ctrl = DecodeRegCtrl(fp, next_regs);
if (next_ctrl.add_write_unit != REG_WRITE_NONE) {
fprintf(fp, "{R%d, T1}, ", GetRegToWrite(next_ctrl.add_write_unit, next_regs));
} else {
fprintf(fp, "T1, ");
}
switch (info.src_type) {
case ADD_BLENDING:
// Note: in this case, regs.uniform_const == location | 0x8
// This probably means we can't load uniforms or immediates in the
// same instruction. This re-uses the encoding that normally means
// "disabled", where the low 4 bits are ignored. Perhaps the extra
// 0x8 or'd in indicates this is happening.
fprintf(fp, "location:%d, ", regs.uniform_const & 0x7);
// fallthrough
case ADD_ONE_SRC:
dump_src(fp, ADD.src0, regs, consts, false);
break;
case ADD_TEX:
case ADD_TEX_COMPACT: {
int tex_index;
int sampler_index;
bool dualTex = false;
fprintf(fp, "coords <");
dump_src(fp, ADD.src0, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
fprintf(fp, ">, ");
if (info.src_type == ADD_TEX_COMPACT) {
tex_index = (ADD.op >> 3) & 0x7;
sampler_index = (ADD.op >> 7) & 0x7;
bool unknown = (ADD.op & 0x40);
// TODO: figure out if the unknown bit is ever 0
if (!unknown)
fprintf(fp, "unknown ");
} else {
uint64_t constVal = get_const(consts, regs);
uint32_t controlBits = (ADD.op & 0x8) ? (constVal >> 32) : constVal;
struct bifrost_tex_ctrl ctrl;
memcpy((char *) &ctrl, (char *) &controlBits, sizeof(ctrl));
/* Dual-tex triggered for adjacent texturing
* instructions with the same coordinates to different
* textures/samplers. Observed for the compact
* (2D/normal) case. */
if ((ctrl.result_type & 7) == 1) {
bool f32 = ctrl.result_type & 8;
struct bifrost_dual_tex_ctrl dualCtrl;
memcpy((char *) &dualCtrl, (char *) &controlBits, sizeof(ctrl));
fprintf(fp, "(dualtex) tex0:%d samp0:%d tex1:%d samp1:%d %s",
dualCtrl.tex_index0, dualCtrl.sampler_index0,
dualCtrl.tex_index1, dualCtrl.sampler_index1,
f32 ? "f32" : "f16");
if (dualCtrl.unk0 != 3)
fprintf(fp, "unk:%d ", dualCtrl.unk0);
dualTex = true;
} else {
if (ctrl.no_merge_index) {
tex_index = ctrl.tex_index;
sampler_index = ctrl.sampler_index;
} else {
tex_index = sampler_index = ctrl.tex_index;
unsigned unk = ctrl.sampler_index >> 2;
if (unk != 3)
fprintf(fp, "unk:%d ", unk);
if (ctrl.sampler_index & 1)
tex_index = -1;
if (ctrl.sampler_index & 2)
sampler_index = -1;
}
if (ctrl.unk0 != 3)
fprintf(fp, "unk0:%d ", ctrl.unk0);
if (ctrl.unk1)
fprintf(fp, "unk1 ");
if (ctrl.unk2 != 0xf)
fprintf(fp, "unk2:%x ", ctrl.unk2);
switch (ctrl.result_type) {
case 0x4:
fprintf(fp, "f32 ");
break;
case 0xe:
fprintf(fp, "i32 ");
break;
case 0xf:
fprintf(fp, "u32 ");
break;
default:
fprintf(fp, "unktype(%x) ", ctrl.result_type);
}
switch (ctrl.tex_type) {
case 0:
fprintf(fp, "cube ");
break;
case 1:
fprintf(fp, "buffer ");
break;
case 2:
fprintf(fp, "2D ");
break;
case 3:
fprintf(fp, "3D ");
break;
}
if (ctrl.is_shadow)
fprintf(fp, "shadow ");
if (ctrl.is_array)
fprintf(fp, "array ");
if (!ctrl.filter) {
if (ctrl.calc_gradients) {
int comp = (controlBits >> 20) & 0x3;
fprintf(fp, "txg comp:%d ", comp);
} else {
fprintf(fp, "txf ");
}
} else {
if (!ctrl.not_supply_lod) {
if (ctrl.compute_lod)
fprintf(fp, "lod_bias ");
else
fprintf(fp, "lod ");
}
if (!ctrl.calc_gradients)
fprintf(fp, "grad ");
}
if (ctrl.texel_offset)
fprintf(fp, "offset ");
}
}
if (!dualTex) {
if (tex_index == -1)
fprintf(fp, "tex:indirect ");
else
fprintf(fp, "tex:%d ", tex_index);
if (sampler_index == -1)
fprintf(fp, "samp:indirect ");
else
fprintf(fp, "samp:%d ", sampler_index);
}
break;
}
case ADD_VARYING_INTERP: {
unsigned addr = ADD.op & 0x1f;
if (addr < 0b10100) {
// direct addr
fprintf(fp, "%d", addr);
} else if (addr < 0b11000) {
if (addr == 22)
fprintf(fp, "fragw");
else if (addr == 23)
fprintf(fp, "fragz");
else
fprintf(fp, "unk%d", addr);
} else {
dump_src(fp, ADD.op & 0x7, regs, consts, false);
}
fprintf(fp, ", ");
dump_src(fp, ADD.src0, regs, consts, false);
break;
}
case ADD_VARYING_ADDRESS: {
dump_src(fp, ADD.src0, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
fprintf(fp, ", ");
unsigned location = (ADD.op >> 3) & 0x1f;
if (location < 16) {
fprintf(fp, "location:%d", location);
} else if (location == 20) {
fprintf(fp, "location:%u", (uint32_t) get_const(consts, regs));
} else if (location == 21) {
fprintf(fp, "location:%u", (uint32_t) (get_const(consts, regs) >> 32));
} else {
fprintf(fp, "location:%d(unk)", location);
}
break;
}
case ADD_LOAD_ATTR:
fprintf(fp, "location:%d, ", (ADD.op >> 3) & 0x1f);
case ADD_TWO_SRC:
dump_src(fp, ADD.src0, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
break;
case ADD_THREE_SRC:
dump_src(fp, ADD.src0, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, (ADD.op >> 3) & 0x7, regs, consts, false);
break;
case ADD_SHIFT: {
struct bifrost_shift_add shift;
memcpy(&shift, &ADD, sizeof(ADD));
dump_src(fp, shift.src0, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, shift.src1, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, shift.src2, regs, consts, false);
break;
}
case ADD_FADD:
case ADD_FMINMAX:
if (ADD.op & 0x10)
fprintf(fp, "-");
if (ADD.op & 0x1000)
fprintf(fp, "abs(");
dump_src(fp, ADD.src0, regs, consts, false);
switch ((ADD.op >> 6) & 0x3) {
case 3:
fprintf(fp, ".x");
break;
default:
break;
}
if (ADD.op & 0x1000)
fprintf(fp, ")");
fprintf(fp, ", ");
if (ADD.op & 0x20)
fprintf(fp, "-");
if (ADD.op & 0x8)
fprintf(fp, "abs(");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
switch ((ADD.op >> 6) & 0x3) {
case 1:
case 3:
fprintf(fp, ".x");
break;
case 2:
fprintf(fp, ".y");
break;
case 0:
break;
default:
fprintf(fp, ".unk");
break;
}
if (ADD.op & 0x8)
fprintf(fp, ")");
break;
case ADD_FADD16:
if (ADD.op & 0x10)
fprintf(fp, "-");
if (ADD.op & 0x1000)
fprintf(fp, "abs(");
dump_src(fp, ADD.src0, regs, consts, false);
if (ADD.op & 0x1000)
fprintf(fp, ")");
dump_16swizzle(fp, (ADD.op >> 6) & 0x3);
fprintf(fp, ", ");
if (ADD.op & 0x20)
fprintf(fp, "-");
if (ADD.op & 0x8)
fprintf(fp, "abs(");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
dump_16swizzle(fp, (ADD.op >> 8) & 0x3);
if (ADD.op & 0x8)
fprintf(fp, ")");
break;
case ADD_FMINMAX16: {
bool abs1 = ADD.op & 0x8;
bool abs2 = (ADD.op & 0x7) < ADD.src0;
if (ADD.op & 0x10)
fprintf(fp, "-");
if (abs1 || abs2)
fprintf(fp, "abs(");
dump_src(fp, ADD.src0, regs, consts, false);
dump_16swizzle(fp, (ADD.op >> 6) & 0x3);
if (abs1 || abs2)
fprintf(fp, ")");
fprintf(fp, ", ");
if (ADD.op & 0x20)
fprintf(fp, "-");
if (abs1 && abs2)
fprintf(fp, "abs(");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
dump_16swizzle(fp, (ADD.op >> 8) & 0x3);
if (abs1 && abs2)
fprintf(fp, ")");
fprintf(fp, "/* %X */\n", (ADD.op >> 10) & 0x3); /* mode */
break;
}
case ADD_FADDMscale: {
if (ADD.op & 0x400)
fprintf(fp, "-");
if (ADD.op & 0x200)
fprintf(fp, "abs(");
dump_src(fp, ADD.src0, regs, consts, false);
if (ADD.op & 0x200)
fprintf(fp, ")");
fprintf(fp, ", ");
if (ADD.op & 0x800)
fprintf(fp, "-");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
fprintf(fp, ", ");
dump_src(fp, (ADD.op >> 3) & 0x7, regs, consts, false);
break;
}
case ADD_FCMP:
if (ADD.op & 0x400) {
fprintf(fp, "-");
}
if (ADD.op & 0x100) {
fprintf(fp, "abs(");
}
dump_src(fp, ADD.src0, regs, consts, false);
switch ((ADD.op >> 6) & 0x3) {
case 3:
fprintf(fp, ".x");
break;
default:
break;
}
if (ADD.op & 0x100) {
fprintf(fp, ")");
}
fprintf(fp, ", ");
if (ADD.op & 0x200) {
fprintf(fp, "abs(");
}
dump_src(fp, ADD.op & 0x7, regs, consts, false);
switch ((ADD.op >> 6) & 0x3) {
case 1:
case 3:
fprintf(fp, ".x");
break;
case 2:
fprintf(fp, ".y");
break;
case 0:
break;
default:
fprintf(fp, ".unk");
break;
}
if (ADD.op & 0x200) {
fprintf(fp, ")");
}
break;
case ADD_FCMP16:
dump_src(fp, ADD.src0, regs, consts, false);
dump_16swizzle(fp, (ADD.op >> 6) & 0x3);
fprintf(fp, ", ");
dump_src(fp, ADD.op & 0x7, regs, consts, false);
dump_16swizzle(fp, (ADD.op >> 8) & 0x3);
break;
case ADD_BRANCH: {
enum bifrost_branch_code code = (enum bifrost_branch_code) ((ADD.op >> 6) & 0x3f);
enum branch_bit_size size = (enum branch_bit_size) ((ADD.op >> 9) & 0x7);
if (code != BR_ALWAYS) {
dump_src(fp, ADD.src0, regs, consts, false);
switch (size) {
case BR_SIZE_16XX:
fprintf(fp, ".x");
break;
case BR_SIZE_16YY:
case BR_SIZE_16YX0:
case BR_SIZE_16YX1:
fprintf(fp, ".y");
break;
case BR_SIZE_ZERO: {
unsigned ctrl = (ADD.op >> 1) & 0x3;
switch (ctrl) {
case 1:
fprintf(fp, ".y");
break;
case 2:
fprintf(fp, ".x");
break;
default:
break;
}
}
default:
break;
}
fprintf(fp, ", ");
}
if (code != BR_ALWAYS && size != BR_SIZE_ZERO) {
dump_src(fp, ADD.op & 0x7, regs, consts, false);
switch (size) {
case BR_SIZE_16XX:
case BR_SIZE_16YX0:
case BR_SIZE_16YX1:
case BR_SIZE_32_AND_16X:
fprintf(fp, ".x");
break;
case BR_SIZE_16YY:
case BR_SIZE_32_AND_16Y:
fprintf(fp, ".y");
break;
default:
break;
}
fprintf(fp, ", ");
}
// I haven't had the chance to test if this actually specifies the
// branch offset, since I couldn't get it to produce values other
// than 5 (uniform/const high), but these three bits are always
// consistent across branch instructions, so it makes sense...
int offsetSrc = (ADD.op >> 3) & 0x7;
if (offsetSrc == 4 || offsetSrc == 5) {
// If the offset is known/constant, we can decode it
uint32_t raw_offset;
if (offsetSrc == 4)
raw_offset = get_const(consts, regs);
else
raw_offset = get_const(consts, regs) >> 32;
// The high 4 bits are flags, while the rest is the
// twos-complement offset in bytes (here we convert to
// clauses).
int32_t branch_offset = ((int32_t) raw_offset << 4) >> 8;
// If high4 is the high 4 bits of the last 64-bit constant,
// this is calculated as (high4 + 4) & 0xf, or 0 if the branch
// offset itself is the last constant. Not sure if this is
// actually used, or just garbage in unused bits, but in any
// case, we can just ignore it here since it's redundant. Note
// that if there is any padding, this will be 4 since the
// padding counts as the last constant.
unsigned flags = raw_offset >> 28;
(void) flags;
// Note: the offset is in bytes, relative to the beginning of the
// current clause, so a zero offset would be a loop back to the
// same clause (annoyingly different from Midgard).
fprintf(fp, "clause_%d", offset + branch_offset);
} else {
dump_src(fp, offsetSrc, regs, consts, false);
}
}
}
if (info.has_data_reg) {
fprintf(fp, ", R%d", data_reg);
}
fprintf(fp, "\n");
}
void dump_instr(FILE *fp, const struct bifrost_alu_inst *instr,
struct bifrost_regs next_regs, uint64_t *consts,
unsigned data_reg, unsigned offset, bool verbose)
{
struct bifrost_regs regs;
memcpy((char *) ®s, (char *) &instr->reg_bits, sizeof(regs));
if (verbose) {
fprintf(fp, "# regs: %016" PRIx64 "\n", instr->reg_bits);
dump_regs(fp, regs);
}
dump_fma(fp, instr->fma_bits, regs, next_regs, consts, verbose);
dump_add(fp, instr->add_bits, regs, next_regs, consts, data_reg, offset, verbose);
}
bool dump_clause(FILE *fp, uint32_t *words, unsigned *size, unsigned offset, bool verbose)
{
// State for a decoded clause
struct bifrost_alu_inst instrs[8] = {};
uint64_t consts[6] = {};
unsigned num_instrs = 0;
unsigned num_consts = 0;
uint64_t header_bits = 0;
bool stopbit = false;
unsigned i;
for (i = 0; ; i++, words += 4) {
if (verbose) {
fprintf(fp, "# ");
for (int j = 0; j < 4; j++)
fprintf(fp, "%08x ", words[3 - j]); // low bit on the right
fprintf(fp, "\n");
}
unsigned tag = bits(words[0], 0, 8);
// speculatively decode some things that are common between many formats, so we can share some code
struct bifrost_alu_inst main_instr = {};
// 20 bits
main_instr.add_bits = bits(words[2], 2, 32 - 13);
// 23 bits
main_instr.fma_bits = bits(words[1], 11, 32) | bits(words[2], 0, 2) << (32 - 11);
// 35 bits
main_instr.reg_bits = ((uint64_t) bits(words[1], 0, 11)) << 24 | (uint64_t) bits(words[0], 8, 32);
uint64_t const0 = bits(words[0], 8, 32) << 4 | (uint64_t) words[1] << 28 | bits(words[2], 0, 4) << 60;
uint64_t const1 = bits(words[2], 4, 32) << 4 | (uint64_t) words[3] << 32;
bool stop = tag & 0x40;
if (verbose) {
fprintf(fp, "# tag: 0x%02x\n", tag);
}
if (tag & 0x80) {
unsigned idx = stop ? 5 : 2;
main_instr.add_bits |= ((tag >> 3) & 0x7) << 17;
instrs[idx + 1] = main_instr;
instrs[idx].add_bits = bits(words[3], 0, 17) | ((tag & 0x7) << 17);
instrs[idx].fma_bits |= bits(words[2], 19, 32) << 10;
consts[0] = bits(words[3], 17, 32) << 4;
} else {
bool done = false;
switch ((tag >> 3) & 0x7) {
case 0x0:
switch (tag & 0x7) {
case 0x3:
main_instr.add_bits |= bits(words[3], 29, 32) << 17;
instrs[1] = main_instr;
num_instrs = 2;
done = stop;
break;
case 0x4:
instrs[2].add_bits = bits(words[3], 0, 17) | bits(words[3], 29, 32) << 17;
instrs[2].fma_bits |= bits(words[2], 19, 32) << 10;
consts[0] = const0;
num_instrs = 3;
num_consts = 1;
done = stop;
break;
case 0x1:
case 0x5:
instrs[2].add_bits = bits(words[3], 0, 17) | bits(words[3], 29, 32) << 17;
instrs[2].fma_bits |= bits(words[2], 19, 32) << 10;
main_instr.add_bits |= bits(words[3], 26, 29) << 17;
instrs[3] = main_instr;
if ((tag & 0x7) == 0x5) {
num_instrs = 4;
done = stop;
}
break;
case 0x6:
instrs[5].add_bits = bits(words[3], 0, 17) | bits(words[3], 29, 32) << 17;
instrs[5].fma_bits |= bits(words[2], 19, 32) << 10;
consts[0] = const0;
num_instrs = 6;
num_consts = 1;
done = stop;
break;
case 0x7:
instrs[5].add_bits = bits(words[3], 0, 17) | bits(words[3], 29, 32) << 17;
instrs[5].fma_bits |= bits(words[2], 19, 32) << 10;
main_instr.add_bits |= bits(words[3], 26, 29) << 17;
instrs[6] = main_instr;
num_instrs = 7;
done = stop;
break;
default:
fprintf(fp, "unknown tag bits 0x%02x\n", tag);
}
break;
case 0x2:
case 0x3: {
unsigned idx = ((tag >> 3) & 0x7) == 2 ? 4 : 7;
main_instr.add_bits |= (tag & 0x7) << 17;
instrs[idx] = main_instr;
consts[0] |= (bits(words[2], 19, 32) | ((uint64_t) words[3] << 13)) << 19;
num_consts = 1;
num_instrs = idx + 1;
done = stop;
break;
}
case 0x4: {
unsigned idx = stop ? 4 : 1;
main_instr.add_bits |= (tag & 0x7) << 17;
instrs[idx] = main_instr;
instrs[idx + 1].fma_bits |= bits(words[3], 22, 32);
instrs[idx + 1].reg_bits = bits(words[2], 19, 32) | (bits(words[3], 0, 22) << (32 - 19));
break;
}
case 0x1:
// only constants can come after this
num_instrs = 1;
done = stop;
case 0x5:
header_bits = bits(words[2], 19, 32) | ((uint64_t) words[3] << (32 - 19));
main_instr.add_bits |= (tag & 0x7) << 17;
instrs[0] = main_instr;
break;
case 0x6:
case 0x7: {
unsigned pos = tag & 0xf;
// note that `pos' encodes both the total number of
// instructions and the position in the constant stream,
// presumably because decoded constants and instructions
// share a buffer in the decoder, but we only care about
// the position in the constant stream; the total number of
// instructions is redundant.
unsigned const_idx = 0;
switch (pos) {
case 0:
case 1:
case 2:
case 6:
const_idx = 0;
break;
case 3:
case 4:
case 7:
case 9:
const_idx = 1;
break;
case 5:
case 0xa:
const_idx = 2;
break;
case 8:
case 0xb:
case 0xc:
const_idx = 3;
break;
case 0xd:
const_idx = 4;
break;
default:
fprintf(fp, "# unknown pos 0x%x\n", pos);
break;
}
if (num_consts < const_idx + 2)
num_consts = const_idx + 2;
consts[const_idx] = const0;
consts[const_idx + 1] = const1;
done = stop;
break;
}
default:
break;
}
if (done)
break;
}
}
*size = i + 1;
if (verbose) {
fprintf(fp, "# header: %012" PRIx64 "\n", header_bits);
}
struct bifrost_header header;
memcpy((char *) &header, (char *) &header_bits, sizeof(struct bifrost_header));
dump_header(fp, header, verbose);
if (!header.no_end_of_shader)
stopbit = true;
fprintf(fp, "{\n");
for (i = 0; i < num_instrs; i++) {
struct bifrost_regs next_regs;
if (i + 1 == num_instrs) {
memcpy((char *) &next_regs, (char *) &instrs[0].reg_bits,
sizeof(next_regs));
} else {
memcpy((char *) &next_regs, (char *) &instrs[i + 1].reg_bits,
sizeof(next_regs));
}
dump_instr(fp, &instrs[i], next_regs, consts, header.datareg, offset, verbose);
}
fprintf(fp, "}\n");
if (verbose) {
for (unsigned i = 0; i < num_consts; i++) {
fprintf(fp, "# const%d: %08" PRIx64 "\n", 2 * i, consts[i] & 0xffffffff);
fprintf(fp, "# const%d: %08" PRIx64 "\n", 2 * i + 1, consts[i] >> 32);
}
}
return stopbit;
}
void disassemble_bifrost(FILE *fp, uint8_t *code, size_t size, bool verbose)
{
uint32_t *words = (uint32_t *) code;
uint32_t *words_end = words + (size / 4);
// used for displaying branch targets
unsigned offset = 0;
while (words != words_end) {
// we don't know what the program-end bit is quite yet, so for now just
// assume that an all-0 quadword is padding
uint32_t zero[4] = {};
if (memcmp(words, zero, 4 * sizeof(uint32_t)) == 0)
break;
fprintf(fp, "clause_%d:\n", offset);
unsigned size;
if (dump_clause(fp, words, &size, offset, verbose) == true) {
break;
}
words += size * 4;
offset += size;
}
}
|