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
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
|
/*
* Copyright © 2017 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 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 iris_resource.c
*
* Resources are images, buffers, and other objects used by the GPU.
*
* XXX: explain resources
*/
#include <stdio.h>
#include <errno.h>
#include "pipe/p_defines.h"
#include "pipe/p_state.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "util/os_memory.h"
#include "util/u_cpu_detect.h"
#include "util/u_inlines.h"
#include "util/format/u_format.h"
#include "util/u_threaded_context.h"
#include "util/u_transfer.h"
#include "util/u_transfer_helper.h"
#include "util/u_upload_mgr.h"
#include "util/ralloc.h"
#include "iris_batch.h"
#include "iris_context.h"
#include "iris_resource.h"
#include "iris_screen.h"
#include "intel/common/gen_aux_map.h"
#include "intel/dev/gen_debug.h"
#include "isl/isl.h"
#include "drm-uapi/drm_fourcc.h"
#include "drm-uapi/i915_drm.h"
enum modifier_priority {
MODIFIER_PRIORITY_INVALID = 0,
MODIFIER_PRIORITY_LINEAR,
MODIFIER_PRIORITY_X,
MODIFIER_PRIORITY_Y,
MODIFIER_PRIORITY_Y_CCS,
};
static const uint64_t priority_to_modifier[] = {
[MODIFIER_PRIORITY_INVALID] = DRM_FORMAT_MOD_INVALID,
[MODIFIER_PRIORITY_LINEAR] = DRM_FORMAT_MOD_LINEAR,
[MODIFIER_PRIORITY_X] = I915_FORMAT_MOD_X_TILED,
[MODIFIER_PRIORITY_Y] = I915_FORMAT_MOD_Y_TILED,
[MODIFIER_PRIORITY_Y_CCS] = I915_FORMAT_MOD_Y_TILED_CCS,
};
static bool
modifier_is_supported(const struct gen_device_info *devinfo,
enum pipe_format pfmt, uint64_t modifier)
{
/* XXX: do something real */
switch (modifier) {
case I915_FORMAT_MOD_Y_TILED_CCS: {
if (unlikely(INTEL_DEBUG & DEBUG_NO_RBC))
return false;
enum isl_format rt_format =
iris_format_for_usage(devinfo, pfmt,
ISL_SURF_USAGE_RENDER_TARGET_BIT).fmt;
enum isl_format linear_format = isl_format_srgb_to_linear(rt_format);
if (linear_format == ISL_FORMAT_UNSUPPORTED ||
!isl_format_supports_ccs_e(devinfo, linear_format))
return false;
return devinfo->gen >= 9 && devinfo->gen <= 11;
}
case I915_FORMAT_MOD_Y_TILED:
case I915_FORMAT_MOD_X_TILED:
case DRM_FORMAT_MOD_LINEAR:
return true;
case DRM_FORMAT_MOD_INVALID:
default:
return false;
}
}
static uint64_t
select_best_modifier(struct gen_device_info *devinfo, enum pipe_format pfmt,
const uint64_t *modifiers,
int count)
{
enum modifier_priority prio = MODIFIER_PRIORITY_INVALID;
for (int i = 0; i < count; i++) {
if (!modifier_is_supported(devinfo, pfmt, modifiers[i]))
continue;
switch (modifiers[i]) {
case I915_FORMAT_MOD_Y_TILED_CCS:
prio = MAX2(prio, MODIFIER_PRIORITY_Y_CCS);
break;
case I915_FORMAT_MOD_Y_TILED:
prio = MAX2(prio, MODIFIER_PRIORITY_Y);
break;
case I915_FORMAT_MOD_X_TILED:
prio = MAX2(prio, MODIFIER_PRIORITY_X);
break;
case DRM_FORMAT_MOD_LINEAR:
prio = MAX2(prio, MODIFIER_PRIORITY_LINEAR);
break;
case DRM_FORMAT_MOD_INVALID:
default:
break;
}
}
return priority_to_modifier[prio];
}
enum isl_surf_dim
target_to_isl_surf_dim(enum pipe_texture_target target)
{
switch (target) {
case PIPE_BUFFER:
case PIPE_TEXTURE_1D:
case PIPE_TEXTURE_1D_ARRAY:
return ISL_SURF_DIM_1D;
case PIPE_TEXTURE_2D:
case PIPE_TEXTURE_CUBE:
case PIPE_TEXTURE_RECT:
case PIPE_TEXTURE_2D_ARRAY:
case PIPE_TEXTURE_CUBE_ARRAY:
return ISL_SURF_DIM_2D;
case PIPE_TEXTURE_3D:
return ISL_SURF_DIM_3D;
case PIPE_MAX_TEXTURE_TYPES:
break;
}
unreachable("invalid texture type");
}
static void
iris_query_dmabuf_modifiers(struct pipe_screen *pscreen,
enum pipe_format pfmt,
int max,
uint64_t *modifiers,
unsigned int *external_only,
int *count)
{
struct iris_screen *screen = (void *) pscreen;
const struct gen_device_info *devinfo = &screen->devinfo;
uint64_t all_modifiers[] = {
DRM_FORMAT_MOD_LINEAR,
I915_FORMAT_MOD_X_TILED,
I915_FORMAT_MOD_Y_TILED,
I915_FORMAT_MOD_Y_TILED_CCS,
};
int supported_mods = 0;
for (int i = 0; i < ARRAY_SIZE(all_modifiers); i++) {
if (!modifier_is_supported(devinfo, pfmt, all_modifiers[i]))
continue;
if (supported_mods < max) {
if (modifiers)
modifiers[supported_mods] = all_modifiers[i];
if (external_only)
external_only[supported_mods] = util_format_is_yuv(pfmt);
}
supported_mods++;
}
*count = supported_mods;
}
static isl_surf_usage_flags_t
pipe_bind_to_isl_usage(unsigned bindings)
{
isl_surf_usage_flags_t usage = 0;
if (bindings & PIPE_BIND_RENDER_TARGET)
usage |= ISL_SURF_USAGE_RENDER_TARGET_BIT;
if (bindings & PIPE_BIND_SAMPLER_VIEW)
usage |= ISL_SURF_USAGE_TEXTURE_BIT;
if (bindings & (PIPE_BIND_SHADER_IMAGE | PIPE_BIND_SHADER_BUFFER))
usage |= ISL_SURF_USAGE_STORAGE_BIT;
if (bindings & PIPE_BIND_DISPLAY_TARGET)
usage |= ISL_SURF_USAGE_DISPLAY_BIT;
return usage;
}
enum isl_format
iris_image_view_get_format(struct iris_context *ice,
const struct pipe_image_view *img)
{
struct iris_screen *screen = (struct iris_screen *)ice->ctx.screen;
const struct gen_device_info *devinfo = &screen->devinfo;
isl_surf_usage_flags_t usage = ISL_SURF_USAGE_STORAGE_BIT;
enum isl_format isl_fmt =
iris_format_for_usage(devinfo, img->format, usage).fmt;
if (img->shader_access & PIPE_IMAGE_ACCESS_READ) {
/* On Gen8, try to use typed surfaces reads (which support a
* limited number of formats), and if not possible, fall back
* to untyped reads.
*/
if (devinfo->gen == 8 &&
!isl_has_matching_typed_storage_image_format(devinfo, isl_fmt))
return ISL_FORMAT_RAW;
else
return isl_lower_storage_image_format(devinfo, isl_fmt);
}
return isl_fmt;
}
struct pipe_resource *
iris_resource_get_separate_stencil(struct pipe_resource *p_res)
{
/* For packed depth-stencil, we treat depth as the primary resource
* and store S8 as the "second plane" resource.
*/
if (p_res->next && p_res->next->format == PIPE_FORMAT_S8_UINT)
return p_res->next;
return NULL;
}
static void
iris_resource_set_separate_stencil(struct pipe_resource *p_res,
struct pipe_resource *stencil)
{
assert(util_format_has_depth(util_format_description(p_res->format)));
pipe_resource_reference(&p_res->next, stencil);
}
void
iris_get_depth_stencil_resources(struct pipe_resource *res,
struct iris_resource **out_z,
struct iris_resource **out_s)
{
if (!res) {
*out_z = NULL;
*out_s = NULL;
return;
}
if (res->format != PIPE_FORMAT_S8_UINT) {
*out_z = (void *) res;
*out_s = (void *) iris_resource_get_separate_stencil(res);
} else {
*out_z = NULL;
*out_s = (void *) res;
}
}
enum isl_dim_layout
iris_get_isl_dim_layout(const struct gen_device_info *devinfo,
enum isl_tiling tiling,
enum pipe_texture_target target)
{
switch (target) {
case PIPE_TEXTURE_1D:
case PIPE_TEXTURE_1D_ARRAY:
return (devinfo->gen >= 9 && tiling == ISL_TILING_LINEAR ?
ISL_DIM_LAYOUT_GEN9_1D : ISL_DIM_LAYOUT_GEN4_2D);
case PIPE_TEXTURE_2D:
case PIPE_TEXTURE_2D_ARRAY:
case PIPE_TEXTURE_RECT:
case PIPE_TEXTURE_CUBE:
case PIPE_TEXTURE_CUBE_ARRAY:
return ISL_DIM_LAYOUT_GEN4_2D;
case PIPE_TEXTURE_3D:
return (devinfo->gen >= 9 ?
ISL_DIM_LAYOUT_GEN4_2D : ISL_DIM_LAYOUT_GEN4_3D);
case PIPE_MAX_TEXTURE_TYPES:
case PIPE_BUFFER:
break;
}
unreachable("invalid texture type");
}
void
iris_resource_disable_aux(struct iris_resource *res)
{
iris_bo_unreference(res->aux.bo);
iris_bo_unreference(res->aux.clear_color_bo);
free(res->aux.state);
res->aux.usage = ISL_AUX_USAGE_NONE;
res->aux.possible_usages = 1 << ISL_AUX_USAGE_NONE;
res->aux.sampler_usages = 1 << ISL_AUX_USAGE_NONE;
res->aux.has_hiz = 0;
res->aux.surf.size_B = 0;
res->aux.bo = NULL;
res->aux.extra_aux.surf.size_B = 0;
res->aux.clear_color_bo = NULL;
res->aux.state = NULL;
}
static void
iris_resource_destroy(struct pipe_screen *screen,
struct pipe_resource *resource)
{
struct iris_resource *res = (struct iris_resource *)resource;
if (resource->target == PIPE_BUFFER)
util_range_destroy(&res->valid_buffer_range);
iris_resource_disable_aux(res);
iris_bo_unreference(res->bo);
iris_pscreen_unref(res->base.screen);
free(res);
}
static struct iris_resource *
iris_alloc_resource(struct pipe_screen *pscreen,
const struct pipe_resource *templ)
{
struct iris_resource *res = calloc(1, sizeof(struct iris_resource));
if (!res)
return NULL;
res->base = *templ;
res->base.screen = iris_pscreen_ref(pscreen);
pipe_reference_init(&res->base.reference, 1);
res->aux.possible_usages = 1 << ISL_AUX_USAGE_NONE;
res->aux.sampler_usages = 1 << ISL_AUX_USAGE_NONE;
if (templ->target == PIPE_BUFFER)
util_range_init(&res->valid_buffer_range);
return res;
}
unsigned
iris_get_num_logical_layers(const struct iris_resource *res, unsigned level)
{
if (res->surf.dim == ISL_SURF_DIM_3D)
return minify(res->surf.logical_level0_px.depth, level);
else
return res->surf.logical_level0_px.array_len;
}
static enum isl_aux_state **
create_aux_state_map(struct iris_resource *res, enum isl_aux_state initial)
{
assert(res->aux.state == NULL);
uint32_t total_slices = 0;
for (uint32_t level = 0; level < res->surf.levels; level++)
total_slices += iris_get_num_logical_layers(res, level);
const size_t per_level_array_size =
res->surf.levels * sizeof(enum isl_aux_state *);
/* We're going to allocate a single chunk of data for both the per-level
* reference array and the arrays of aux_state. This makes cleanup
* significantly easier.
*/
const size_t total_size =
per_level_array_size + total_slices * sizeof(enum isl_aux_state);
void *data = malloc(total_size);
if (!data)
return NULL;
enum isl_aux_state **per_level_arr = data;
enum isl_aux_state *s = data + per_level_array_size;
for (uint32_t level = 0; level < res->surf.levels; level++) {
per_level_arr[level] = s;
const unsigned level_layers = iris_get_num_logical_layers(res, level);
for (uint32_t a = 0; a < level_layers; a++)
*(s++) = initial;
}
assert((void *)s == data + total_size);
return per_level_arr;
}
static unsigned
iris_get_aux_clear_color_state_size(struct iris_screen *screen)
{
const struct gen_device_info *devinfo = &screen->devinfo;
return devinfo->gen >= 10 ? screen->isl_dev.ss.clear_color_state_size : 0;
}
static void
map_aux_addresses(struct iris_screen *screen, struct iris_resource *res)
{
const struct gen_device_info *devinfo = &screen->devinfo;
if (devinfo->gen >= 12 && isl_aux_usage_has_ccs(res->aux.usage)) {
void *aux_map_ctx = iris_bufmgr_get_aux_map_context(screen->bufmgr);
assert(aux_map_ctx);
const unsigned aux_offset = res->aux.extra_aux.surf.size_B > 0 ?
res->aux.extra_aux.offset : res->aux.offset;
gen_aux_map_add_image(aux_map_ctx, &res->surf, res->bo->gtt_offset,
res->aux.bo->gtt_offset + aux_offset);
res->bo->aux_map_address = res->aux.bo->gtt_offset;
}
}
static bool
want_ccs_e_for_format(const struct gen_device_info *devinfo,
enum isl_format format)
{
if (!isl_format_supports_ccs_e(devinfo, format))
return false;
const struct isl_format_layout *fmtl = isl_format_get_layout(format);
/* CCS_E seems to significantly hurt performance with 32-bit floating
* point formats. For example, Paraview's "Wavelet Volume" case uses
* both R32_FLOAT and R32G32B32A32_FLOAT, and enabling CCS_E for those
* formats causes a 62% FPS drop.
*
* However, many benchmarks seem to use 16-bit float with no issues.
*/
if (fmtl->channels.r.bits == 32 && fmtl->channels.r.type == ISL_SFLOAT)
return false;
return true;
}
/**
* Configure aux for the resource, but don't allocate it. For images which
* might be shared with modifiers, we must allocate the image and aux data in
* a single bo.
*
* Returns false on unexpected error (e.g. allocation failed, or invalid
* configuration result).
*/
static bool
iris_resource_configure_aux(struct iris_screen *screen,
struct iris_resource *res, bool imported,
uint64_t *aux_size_B,
uint32_t *alloc_flags)
{
const struct gen_device_info *devinfo = &screen->devinfo;
/* Try to create the auxiliary surfaces allowed by the modifier or by
* the user if no modifier is specified.
*/
assert(!res->mod_info || res->mod_info->aux_usage == ISL_AUX_USAGE_NONE ||
res->mod_info->aux_usage == ISL_AUX_USAGE_CCS_E);
const bool has_mcs = !res->mod_info &&
isl_surf_get_mcs_surf(&screen->isl_dev, &res->surf, &res->aux.surf);
const bool has_hiz = !res->mod_info && !(INTEL_DEBUG & DEBUG_NO_HIZ) &&
isl_surf_get_hiz_surf(&screen->isl_dev, &res->surf, &res->aux.surf);
const bool has_ccs =
((!res->mod_info && !(INTEL_DEBUG & DEBUG_NO_RBC)) ||
(res->mod_info && res->mod_info->aux_usage != ISL_AUX_USAGE_NONE)) &&
isl_surf_get_ccs_surf(&screen->isl_dev, &res->surf, &res->aux.surf,
&res->aux.extra_aux.surf, 0);
/* Having both HIZ and MCS is impossible. */
assert(!has_mcs || !has_hiz);
/* Ensure aux surface creation for MCS_CCS and HIZ_CCS is correct. */
if (has_ccs && (has_mcs || has_hiz)) {
assert(res->aux.extra_aux.surf.size_B > 0 &&
res->aux.extra_aux.surf.usage & ISL_SURF_USAGE_CCS_BIT);
assert(res->aux.surf.size_B > 0 &&
res->aux.surf.usage &
(ISL_SURF_USAGE_HIZ_BIT | ISL_SURF_USAGE_MCS_BIT));
}
if (res->mod_info && has_ccs) {
/* Only allow a CCS modifier if the aux was created successfully. */
res->aux.possible_usages |= 1 << res->mod_info->aux_usage;
} else if (has_mcs) {
res->aux.possible_usages |=
1 << (has_ccs ? ISL_AUX_USAGE_MCS_CCS : ISL_AUX_USAGE_MCS);
} else if (has_hiz) {
if (!has_ccs) {
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_HIZ;
} else if (res->surf.samples == 1 &&
(res->surf.usage & ISL_SURF_USAGE_TEXTURE_BIT)) {
/* If this resource is single-sampled and will be used as a texture,
* put the HiZ surface in write-through mode so that we can sample
* from it.
*/
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_HIZ_CCS_WT;
} else {
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_HIZ_CCS;
}
} else if (has_ccs && isl_surf_usage_is_stencil(res->surf.usage)) {
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_STC_CCS;
} else if (has_ccs) {
if (want_ccs_e_for_format(devinfo, res->surf.format))
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_CCS_E;
if (isl_format_supports_ccs_d(devinfo, res->surf.format))
res->aux.possible_usages |= 1 << ISL_AUX_USAGE_CCS_D;
}
res->aux.usage = util_last_bit(res->aux.possible_usages) - 1;
res->aux.sampler_usages = res->aux.possible_usages;
/* We don't always support sampling with hiz. But when we do, it must be
* single sampled.
*/
if (!devinfo->has_sample_with_hiz || res->surf.samples > 1)
res->aux.sampler_usages &= ~(1 << ISL_AUX_USAGE_HIZ);
/* ISL_AUX_USAGE_HIZ_CCS doesn't support sampling at all */
res->aux.sampler_usages &= ~(1 << ISL_AUX_USAGE_HIZ_CCS);
enum isl_aux_state initial_state;
*aux_size_B = 0;
*alloc_flags = 0;
assert(!res->aux.bo);
switch (res->aux.usage) {
case ISL_AUX_USAGE_NONE:
/* Having no aux buffer is only okay if there's no modifier with aux. */
return !res->mod_info || res->mod_info->aux_usage == ISL_AUX_USAGE_NONE;
case ISL_AUX_USAGE_HIZ:
case ISL_AUX_USAGE_HIZ_CCS:
case ISL_AUX_USAGE_HIZ_CCS_WT:
initial_state = ISL_AUX_STATE_AUX_INVALID;
break;
case ISL_AUX_USAGE_MCS:
case ISL_AUX_USAGE_MCS_CCS:
/* The Ivybridge PRM, Vol 2 Part 1 p326 says:
*
* "When MCS buffer is enabled and bound to MSRT, it is required
* that it is cleared prior to any rendering."
*
* Since we only use the MCS buffer for rendering, we just clear it
* immediately on allocation. The clear value for MCS buffers is all
* 1's, so we simply memset it to 0xff.
*/
initial_state = ISL_AUX_STATE_CLEAR;
break;
case ISL_AUX_USAGE_CCS_D:
case ISL_AUX_USAGE_CCS_E:
case ISL_AUX_USAGE_STC_CCS:
/* When CCS_E is used, we need to ensure that the CCS starts off in
* a valid state. From the Sky Lake PRM, "MCS Buffer for Render
* Target(s)":
*
* "If Software wants to enable Color Compression without Fast
* clear, Software needs to initialize MCS with zeros."
*
* A CCS value of 0 indicates that the corresponding block is in the
* pass-through state which is what we want.
*
* For CCS_D, do the same thing. On Gen9+, this avoids having any
* undefined bits in the aux buffer.
*/
if (imported) {
assert(res->aux.usage != ISL_AUX_USAGE_STC_CCS);
initial_state =
isl_drm_modifier_get_default_aux_state(res->mod_info->modifier);
} else {
initial_state = ISL_AUX_STATE_PASS_THROUGH;
}
*alloc_flags |= BO_ALLOC_ZEROED;
break;
case ISL_AUX_USAGE_MC:
unreachable("Unsupported aux mode");
}
/* Create the aux_state for the auxiliary buffer. */
res->aux.state = create_aux_state_map(res, initial_state);
if (!res->aux.state)
return false;
/* Increase the aux offset if the main and aux surfaces will share a BO. */
res->aux.offset =
!res->mod_info || res->mod_info->aux_usage == res->aux.usage ?
ALIGN(res->surf.size_B, res->aux.surf.alignment_B) : 0;
uint64_t size = res->aux.surf.size_B;
/* Allocate space in the buffer for storing the CCS. */
if (res->aux.extra_aux.surf.size_B > 0) {
const uint64_t padded_aux_size =
ALIGN(size, res->aux.extra_aux.surf.alignment_B);
res->aux.extra_aux.offset = res->aux.offset + padded_aux_size;
size = padded_aux_size + res->aux.extra_aux.surf.size_B;
}
/* Allocate space in the buffer for storing the clear color. On modern
* platforms (gen > 9), we can read it directly from such buffer.
*
* On gen <= 9, we are going to store the clear color on the buffer
* anyways, and copy it back to the surface state during state emission.
*
* Also add some padding to make sure the fast clear color state buffer
* starts at a 4K alignment. We believe that 256B might be enough, but due
* to lack of testing we will leave this as 4K for now.
*/
size = ALIGN(size, 4096);
res->aux.clear_color_offset = res->aux.offset + size;
size += iris_get_aux_clear_color_state_size(screen);
*aux_size_B = size;
if (isl_aux_usage_has_hiz(res->aux.usage)) {
for (unsigned level = 0; level < res->surf.levels; ++level) {
uint32_t width = u_minify(res->surf.phys_level0_sa.width, level);
uint32_t height = u_minify(res->surf.phys_level0_sa.height, level);
/* Disable HiZ for LOD > 0 unless the width/height are 8x4 aligned.
* For LOD == 0, we can grow the dimensions to make it work.
*/
if (level == 0 || ((width & 7) == 0 && (height & 3) == 0))
res->aux.has_hiz |= 1 << level;
}
}
return true;
}
/**
* Initialize the aux buffer contents.
*
* Returns false on unexpected error (e.g. mapping a BO failed).
*/
static bool
iris_resource_init_aux_buf(struct iris_resource *res, uint32_t alloc_flags,
unsigned clear_color_state_size)
{
if (!(alloc_flags & BO_ALLOC_ZEROED)) {
void *map = iris_bo_map(NULL, res->aux.bo, MAP_WRITE | MAP_RAW);
if (!map)
return false;
if (iris_resource_get_aux_state(res, 0, 0) != ISL_AUX_STATE_AUX_INVALID) {
uint8_t memset_value = isl_aux_usage_has_mcs(res->aux.usage) ? 0xFF : 0;
memset((char*)map + res->aux.offset, memset_value,
res->aux.surf.size_B);
}
memset((char*)map + res->aux.extra_aux.offset,
0, res->aux.extra_aux.surf.size_B);
/* Zero the indirect clear color to match ::fast_clear_color. */
memset((char *)map + res->aux.clear_color_offset, 0,
clear_color_state_size);
iris_bo_unmap(res->aux.bo);
}
if (clear_color_state_size > 0) {
res->aux.clear_color_bo = res->aux.bo;
iris_bo_reference(res->aux.clear_color_bo);
}
return true;
}
/**
* Allocate the initial aux surface for a resource based on aux.usage
*
* Returns false on unexpected error (e.g. allocation failed, or invalid
* configuration result).
*/
static bool
iris_resource_alloc_separate_aux(struct iris_screen *screen,
struct iris_resource *res)
{
uint32_t alloc_flags;
uint64_t size;
if (!iris_resource_configure_aux(screen, res, false, &size, &alloc_flags))
return false;
if (size == 0)
return true;
/* Allocate the auxiliary buffer. ISL has stricter set of alignment rules
* the drm allocator. Therefore, one can pass the ISL dimensions in terms
* of bytes instead of trying to recalculate based on different format
* block sizes.
*/
res->aux.bo = iris_bo_alloc_tiled(screen->bufmgr, "aux buffer", size, 4096,
IRIS_MEMZONE_OTHER,
isl_tiling_to_i915_tiling(res->aux.surf.tiling),
res->aux.surf.row_pitch_B, alloc_flags);
if (!res->aux.bo) {
return false;
}
if (!iris_resource_init_aux_buf(res, alloc_flags,
iris_get_aux_clear_color_state_size(screen)))
return false;
map_aux_addresses(screen, res);
return true;
}
void
iris_resource_finish_aux_import(struct pipe_screen *pscreen,
struct iris_resource *res)
{
struct iris_screen *screen = (struct iris_screen *)pscreen;
assert(iris_resource_unfinished_aux_import(res));
assert(!res->mod_info->supports_clear_color);
struct iris_resource *aux_res = (void *) res->base.next;
assert(aux_res->aux.surf.row_pitch_B && aux_res->aux.offset &&
aux_res->aux.bo);
assert(res->bo == aux_res->aux.bo);
iris_bo_reference(aux_res->aux.bo);
res->aux.bo = aux_res->aux.bo;
res->aux.offset = aux_res->aux.offset;
assert(res->bo->size >= (res->aux.offset + res->aux.surf.size_B));
assert(res->aux.clear_color_bo == NULL);
res->aux.clear_color_offset = 0;
assert(aux_res->aux.surf.row_pitch_B == res->aux.surf.row_pitch_B);
unsigned clear_color_state_size =
iris_get_aux_clear_color_state_size(screen);
if (clear_color_state_size > 0) {
res->aux.clear_color_bo =
iris_bo_alloc(screen->bufmgr, "clear color buffer",
clear_color_state_size, IRIS_MEMZONE_OTHER);
res->aux.clear_color_offset = 0;
}
iris_resource_destroy(&screen->base, res->base.next);
res->base.next = NULL;
}
static struct pipe_resource *
iris_resource_create_for_buffer(struct pipe_screen *pscreen,
const struct pipe_resource *templ)
{
struct iris_screen *screen = (struct iris_screen *)pscreen;
struct iris_resource *res = iris_alloc_resource(pscreen, templ);
assert(templ->target == PIPE_BUFFER);
assert(templ->height0 <= 1);
assert(templ->depth0 <= 1);
assert(templ->format == PIPE_FORMAT_NONE ||
util_format_get_blocksize(templ->format) == 1);
res->internal_format = templ->format;
res->surf.tiling = ISL_TILING_LINEAR;
enum iris_memory_zone memzone = IRIS_MEMZONE_OTHER;
const char *name = templ->target == PIPE_BUFFER ? "buffer" : "miptree";
if (templ->flags & IRIS_RESOURCE_FLAG_SHADER_MEMZONE) {
memzone = IRIS_MEMZONE_SHADER;
name = "shader kernels";
} else if (templ->flags & IRIS_RESOURCE_FLAG_SURFACE_MEMZONE) {
memzone = IRIS_MEMZONE_SURFACE;
name = "surface state";
} else if (templ->flags & IRIS_RESOURCE_FLAG_DYNAMIC_MEMZONE) {
memzone = IRIS_MEMZONE_DYNAMIC;
name = "dynamic state";
}
res->bo = iris_bo_alloc(screen->bufmgr, name, templ->width0, memzone);
if (!res->bo) {
iris_resource_destroy(pscreen, &res->base);
return NULL;
}
if (templ->bind & PIPE_BIND_SHARED)
iris_bo_make_external(res->bo);
return &res->base;
}
static struct pipe_resource *
iris_resource_create_with_modifiers(struct pipe_screen *pscreen,
const struct pipe_resource *templ,
const uint64_t *modifiers,
int modifiers_count)
{
struct iris_screen *screen = (struct iris_screen *)pscreen;
struct gen_device_info *devinfo = &screen->devinfo;
struct iris_resource *res = iris_alloc_resource(pscreen, templ);
if (!res)
return NULL;
const struct util_format_description *format_desc =
util_format_description(templ->format);
const bool has_depth = util_format_has_depth(format_desc);
uint64_t modifier =
select_best_modifier(devinfo, templ->format, modifiers, modifiers_count);
isl_tiling_flags_t tiling_flags = ISL_TILING_ANY_MASK;
if (modifier != DRM_FORMAT_MOD_INVALID) {
res->mod_info = isl_drm_modifier_get_info(modifier);
tiling_flags = 1 << res->mod_info->tiling;
} else {
if (modifiers_count > 0) {
fprintf(stderr, "Unsupported modifier, resource creation failed.\n");
goto fail;
}
/* Use linear for staging buffers */
if (templ->usage == PIPE_USAGE_STAGING ||
templ->bind & (PIPE_BIND_LINEAR | PIPE_BIND_CURSOR) )
tiling_flags = ISL_TILING_LINEAR_BIT;
else if (templ->bind & PIPE_BIND_SCANOUT)
tiling_flags = ISL_TILING_X_BIT;
}
isl_surf_usage_flags_t usage = pipe_bind_to_isl_usage(templ->bind);
if (templ->target == PIPE_TEXTURE_CUBE ||
templ->target == PIPE_TEXTURE_CUBE_ARRAY)
usage |= ISL_SURF_USAGE_CUBE_BIT;
if (templ->usage != PIPE_USAGE_STAGING) {
if (templ->format == PIPE_FORMAT_S8_UINT)
usage |= ISL_SURF_USAGE_STENCIL_BIT;
else if (has_depth)
usage |= ISL_SURF_USAGE_DEPTH_BIT;
}
enum pipe_format pfmt = templ->format;
res->internal_format = pfmt;
/* Should be handled by u_transfer_helper */
assert(!util_format_is_depth_and_stencil(pfmt));
struct iris_format_info fmt = iris_format_for_usage(devinfo, pfmt, usage);
assert(fmt.fmt != ISL_FORMAT_UNSUPPORTED);
UNUSED const bool isl_surf_created_successfully =
isl_surf_init(&screen->isl_dev, &res->surf,
.dim = target_to_isl_surf_dim(templ->target),
.format = fmt.fmt,
.width = templ->width0,
.height = templ->height0,
.depth = templ->depth0,
.levels = templ->last_level + 1,
.array_len = templ->array_size,
.samples = MAX2(templ->nr_samples, 1),
.min_alignment_B = 0,
.row_pitch_B = 0,
.usage = usage,
.tiling_flags = tiling_flags);
assert(isl_surf_created_successfully);
const char *name = "miptree";
enum iris_memory_zone memzone = IRIS_MEMZONE_OTHER;
unsigned int flags = 0;
if (templ->usage == PIPE_USAGE_STAGING)
flags |= BO_ALLOC_COHERENT;
/* These are for u_upload_mgr buffers only */
assert(!(templ->flags & (IRIS_RESOURCE_FLAG_SHADER_MEMZONE |
IRIS_RESOURCE_FLAG_SURFACE_MEMZONE |
IRIS_RESOURCE_FLAG_DYNAMIC_MEMZONE)));
uint32_t aux_preferred_alloc_flags;
uint64_t aux_size = 0;
if (!iris_resource_configure_aux(screen, res, false, &aux_size,
&aux_preferred_alloc_flags)) {
goto fail;
}
/* Modifiers require the aux data to be in the same buffer as the main
* surface, but we combine them even when a modifiers is not being used.
*/
const uint64_t bo_size =
MAX2(res->surf.size_B, res->aux.offset + aux_size);
uint32_t alignment = MAX2(4096, res->surf.alignment_B);
res->bo = iris_bo_alloc_tiled(screen->bufmgr, name, bo_size, alignment,
memzone,
isl_tiling_to_i915_tiling(res->surf.tiling),
res->surf.row_pitch_B, flags);
if (!res->bo)
goto fail;
if (aux_size > 0) {
res->aux.bo = res->bo;
iris_bo_reference(res->aux.bo);
unsigned clear_color_state_size =
iris_get_aux_clear_color_state_size(screen);
if (!iris_resource_init_aux_buf(res, flags, clear_color_state_size))
goto fail;
map_aux_addresses(screen, res);
}
if (templ->bind & PIPE_BIND_SHARED)
iris_bo_make_external(res->bo);
return &res->base;
fail:
fprintf(stderr, "XXX: resource creation failed\n");
iris_resource_destroy(pscreen, &res->base);
return NULL;
}
static struct pipe_resource *
iris_resource_create(struct pipe_screen *pscreen,
const struct pipe_resource *templ)
{
if (templ->target == PIPE_BUFFER)
return iris_resource_create_for_buffer(pscreen, templ);
else
return iris_resource_create_with_modifiers(pscreen, templ, NULL, 0);
}
static uint64_t
tiling_to_modifier(uint32_t tiling)
{
static const uint64_t map[] = {
[I915_TILING_NONE] = DRM_FORMAT_MOD_LINEAR,
[I915_TILING_X] = I915_FORMAT_MOD_X_TILED,
[I915_TILING_Y] = I915_FORMAT_MOD_Y_TILED,
};
assert(tiling < ARRAY_SIZE(map));
return map[tiling];
}
static struct pipe_resource *
iris_resource_from_user_memory(struct pipe_screen *pscreen,
const struct pipe_resource *templ,
void *user_memory)
{
struct iris_screen *screen = (struct iris_screen *)pscreen;
struct iris_bufmgr *bufmgr = screen->bufmgr;
struct iris_resource *res = iris_alloc_resource(pscreen, templ);
if (!res)
return NULL;
assert(templ->target == PIPE_BUFFER);
res->internal_format = templ->format;
res->bo = iris_bo_create_userptr(bufmgr, "user",
user_memory, templ->width0,
IRIS_MEMZONE_OTHER);
if (!res->bo) {
iris_resource_destroy(pscreen, &res->base);
return NULL;
}
util_range_add(&res->base, &res->valid_buffer_range, 0, templ->width0);
return &res->base;
}
static struct pipe_resource *
iris_resource_from_handle(struct pipe_screen *pscreen,
const struct pipe_resource *templ,
struct winsys_handle *whandle,
unsigned usage)
{
struct iris_screen *screen = (struct iris_screen *)pscreen;
struct gen_device_info *devinfo = &screen->devinfo;
struct iris_bufmgr *bufmgr = screen->bufmgr;
struct iris_resource *res = iris_alloc_resource(pscreen, templ);
const struct isl_drm_modifier_info *mod_inf =
isl_drm_modifier_get_info(whandle->modifier);
uint32_t tiling;
if (!res)
return NULL;
switch (whandle->type) {
case WINSYS_HANDLE_TYPE_FD:
if (mod_inf)
tiling = isl_tiling_to_i915_tiling(mod_inf->tiling);
else
tiling = I915_TILING_LAST + 1;
res->bo = iris_bo_import_dmabuf(bufmgr, whandle->handle,
tiling, whandle->stride);
break;
case WINSYS_HANDLE_TYPE_SHARED:
res->bo = iris_bo_gem_create_from_name(bufmgr, "winsys image",
whandle->handle);
break;
default:
unreachable("invalid winsys handle type");
}
if (!res->bo)
goto fail;
res->offset = whandle->offset;
if (mod_inf == NULL) {
mod_inf =
isl_drm_modifier_get_info(tiling_to_modifier(res->bo->tiling_mode));
}
assert(mod_inf);
res->external_format = whandle->format;
res->mod_info = mod_inf;
isl_surf_usage_flags_t isl_usage = pipe_bind_to_isl_usage(templ->bind);
const struct iris_format_info fmt =
iris_format_for_usage(devinfo, templ->format, isl_usage);
res->internal_format = templ->format;
if (templ->target == PIPE_BUFFER) {
res->surf.tiling = ISL_TILING_LINEAR;
} else {
/* Create a surface for each plane specified by the external format. */
if (whandle->plane < util_format_get_num_planes(whandle->format)) {
UNUSED const bool isl_surf_created_successfully =
isl_surf_init(&screen->isl_dev, &res->surf,
.dim = target_to_isl_surf_dim(templ->target),
.format = fmt.fmt,
.width = templ->width0,
.height = templ->height0,
.depth = templ->depth0,
.levels = templ->last_level + 1,
.array_len = templ->array_size,
.samples = MAX2(templ->nr_samples, 1),
.min_alignment_B = 0,
.row_pitch_B = whandle->stride,
.usage = isl_usage,
.tiling_flags = 1 << res->mod_info->tiling);
assert(isl_surf_created_successfully);
assert(res->bo->tiling_mode ==
isl_tiling_to_i915_tiling(res->surf.tiling));
// XXX: create_ccs_buf_for_image?
if (whandle->modifier == DRM_FORMAT_MOD_INVALID) {
if (!iris_resource_alloc_separate_aux(screen, res))
goto fail;
} else {
if (res->mod_info->aux_usage != ISL_AUX_USAGE_NONE) {
uint32_t alloc_flags;
uint64_t size;
bool ok = iris_resource_configure_aux(screen, res, true, &size,
&alloc_flags);
assert(ok);
/* The gallium dri layer will create a separate plane resource
* for the aux image. iris_resource_finish_aux_import will
* merge the separate aux parameters back into a single
* iris_resource.
*/
}
}
} else {
/* Save modifier import information to reconstruct later. After
* import, this will be available under a second image accessible
* from the main image with res->base.next. See
* iris_resource_finish_aux_import.
*/
res->aux.surf.row_pitch_B = whandle->stride;
res->aux.offset = whandle->offset;
res->aux.bo = res->bo;
res->bo = NULL;
}
}
return &res->base;
fail:
iris_resource_destroy(pscreen, &res->base);
return NULL;
}
static void
iris_flush_resource(struct pipe_context *ctx, struct pipe_resource *resource)
{
struct iris_context *ice = (struct iris_context *)ctx;
struct iris_batch *render_batch = &ice->batches[IRIS_BATCH_RENDER];
struct iris_resource *res = (void *) resource;
const struct isl_drm_modifier_info *mod = res->mod_info;
iris_resource_prepare_access(ice, render_batch, res,
0, INTEL_REMAINING_LEVELS,
0, INTEL_REMAINING_LAYERS,
mod ? mod->aux_usage : ISL_AUX_USAGE_NONE,
mod ? mod->supports_clear_color : false);
}
static void
iris_resource_disable_aux_on_first_query(struct pipe_resource *resource,
unsigned usage)
{
struct iris_resource *res = (struct iris_resource *)resource;
bool mod_with_aux =
res->mod_info && res->mod_info->aux_usage != ISL_AUX_USAGE_NONE;
/* Disable aux usage if explicit flush not set and this is the first time
* we are dealing with this resource and the resource was not created with
* a modifier with aux.
*/
if (!mod_with_aux &&
(!(usage & PIPE_HANDLE_USAGE_EXPLICIT_FLUSH) && res->aux.usage != 0) &&
p_atomic_read(&resource->reference.count) == 1) {
iris_resource_disable_aux(res);
}
}
static bool
iris_resource_get_param(struct pipe_screen *screen,
struct pipe_context *context,
struct pipe_resource *resource,
unsigned plane,
unsigned layer,
enum pipe_resource_param param,
unsigned handle_usage,
uint64_t *value)
{
struct iris_resource *res = (struct iris_resource *)resource;
bool mod_with_aux =
res->mod_info && res->mod_info->aux_usage != ISL_AUX_USAGE_NONE;
bool wants_aux = mod_with_aux && plane > 0;
bool result;
unsigned handle;
if (iris_resource_unfinished_aux_import(res))
iris_resource_finish_aux_import(screen, res);
struct iris_bo *bo = wants_aux ? res->aux.bo : res->bo;
iris_resource_disable_aux_on_first_query(resource, handle_usage);
switch (param) {
case PIPE_RESOURCE_PARAM_NPLANES:
if (mod_with_aux) {
*value = 2;
} else {
unsigned count = 0;
for (struct pipe_resource *cur = resource; cur; cur = cur->next)
count++;
*value = count;
}
return true;
case PIPE_RESOURCE_PARAM_STRIDE:
*value = wants_aux ? res->aux.surf.row_pitch_B : res->surf.row_pitch_B;
return true;
case PIPE_RESOURCE_PARAM_OFFSET:
*value = wants_aux ? res->aux.offset : 0;
return true;
case PIPE_RESOURCE_PARAM_MODIFIER:
*value = res->mod_info ? res->mod_info->modifier :
tiling_to_modifier(res->bo->tiling_mode);
return true;
case PIPE_RESOURCE_PARAM_HANDLE_TYPE_SHARED:
result = iris_bo_flink(bo, &handle) == 0;
if (result)
*value = handle;
return result;
case PIPE_RESOURCE_PARAM_HANDLE_TYPE_KMS:
*value = iris_bo_export_gem_handle(bo);
return true;
case PIPE_RESOURCE_PARAM_HANDLE_TYPE_FD:
result = iris_bo_export_dmabuf(bo, (int *) &handle) == 0;
if (result)
*value = handle;
return result;
default:
return false;
}
}
static bool
iris_resource_get_handle(struct pipe_screen *pscreen,
struct pipe_context *ctx,
struct pipe_resource *resource,
struct winsys_handle *whandle,
unsigned usage)
{
struct iris_resource *res = (struct iris_resource *)resource;
bool mod_with_aux =
res->mod_info && res->mod_info->aux_usage != ISL_AUX_USAGE_NONE;
iris_resource_disable_aux_on_first_query(resource, usage);
struct iris_bo *bo;
if (mod_with_aux && whandle->plane > 0) {
assert(res->aux.bo);
bo = res->aux.bo;
whandle->stride = res->aux.surf.row_pitch_B;
whandle->offset = res->aux.offset;
} else {
/* If this is a buffer, stride should be 0 - no need to special case */
whandle->stride = res->surf.row_pitch_B;
bo = res->bo;
}
whandle->format = res->external_format;
whandle->modifier =
res->mod_info ? res->mod_info->modifier
: tiling_to_modifier(res->bo->tiling_mode);
#ifndef NDEBUG
enum isl_aux_usage allowed_usage =
res->mod_info ? res->mod_info->aux_usage : ISL_AUX_USAGE_NONE;
if (res->aux.usage != allowed_usage) {
enum isl_aux_state aux_state = iris_resource_get_aux_state(res, 0, 0);
assert(aux_state == ISL_AUX_STATE_RESOLVED ||
aux_state == ISL_AUX_STATE_PASS_THROUGH);
}
#endif
switch (whandle->type) {
case WINSYS_HANDLE_TYPE_SHARED:
return iris_bo_flink(bo, &whandle->handle) == 0;
case WINSYS_HANDLE_TYPE_KMS:
whandle->handle = iris_bo_export_gem_handle(bo);
return true;
case WINSYS_HANDLE_TYPE_FD:
return iris_bo_export_dmabuf(bo, (int *) &whandle->handle) == 0;
}
return false;
}
static bool
resource_is_busy(struct iris_context *ice,
struct iris_resource *res)
{
bool busy = iris_bo_busy(res->bo);
for (int i = 0; i < IRIS_BATCH_COUNT; i++)
busy |= iris_batch_references(&ice->batches[i], res->bo);
return busy;
}
static void
iris_invalidate_resource(struct pipe_context *ctx,
struct pipe_resource *resource)
{
struct iris_screen *screen = (void *) ctx->screen;
struct iris_context *ice = (void *) ctx;
struct iris_resource *res = (void *) resource;
if (resource->target != PIPE_BUFFER)
return;
/* If it's already invalidated, don't bother doing anything. */
if (res->valid_buffer_range.start > res->valid_buffer_range.end)
return;
if (!resource_is_busy(ice, res)) {
/* The resource is idle, so just mark that it contains no data and
* keep using the same underlying buffer object.
*/
util_range_set_empty(&res->valid_buffer_range);
return;
}
/* Otherwise, try and replace the backing storage with a new BO. */
/* We can't reallocate memory we didn't allocate in the first place. */
if (res->bo->userptr)
return;
// XXX: We should support this.
if (res->bind_history & PIPE_BIND_STREAM_OUTPUT)
return;
struct iris_bo *old_bo = res->bo;
struct iris_bo *new_bo =
iris_bo_alloc(screen->bufmgr, res->bo->name, resource->width0,
iris_memzone_for_address(old_bo->gtt_offset));
if (!new_bo)
return;
/* Swap out the backing storage */
res->bo = new_bo;
/* Rebind the buffer, replacing any state referring to the old BO's
* address, and marking state dirty so it's reemitted.
*/
screen->vtbl.rebind_buffer(ice, res);
util_range_set_empty(&res->valid_buffer_range);
iris_bo_unreference(old_bo);
}
static void
iris_flush_staging_region(struct pipe_transfer *xfer,
const struct pipe_box *flush_box)
{
if (!(xfer->usage & PIPE_TRANSFER_WRITE))
return;
struct iris_transfer *map = (void *) xfer;
struct pipe_box src_box = *flush_box;
/* Account for extra alignment padding in staging buffer */
if (xfer->resource->target == PIPE_BUFFER)
src_box.x += xfer->box.x % IRIS_MAP_BUFFER_ALIGNMENT;
struct pipe_box dst_box = (struct pipe_box) {
.x = xfer->box.x + flush_box->x,
.y = xfer->box.y + flush_box->y,
.z = xfer->box.z + flush_box->z,
.width = flush_box->width,
.height = flush_box->height,
.depth = flush_box->depth,
};
iris_copy_region(map->blorp, map->batch, xfer->resource, xfer->level,
dst_box.x, dst_box.y, dst_box.z, map->staging, 0,
&src_box);
}
static void
iris_unmap_copy_region(struct iris_transfer *map)
{
iris_resource_destroy(map->staging->screen, map->staging);
map->ptr = NULL;
}
static void
iris_map_copy_region(struct iris_transfer *map)
{
struct pipe_screen *pscreen = &map->batch->screen->base;
struct pipe_transfer *xfer = &map->base;
struct pipe_box *box = &xfer->box;
struct iris_resource *res = (void *) xfer->resource;
unsigned extra = xfer->resource->target == PIPE_BUFFER ?
box->x % IRIS_MAP_BUFFER_ALIGNMENT : 0;
struct pipe_resource templ = (struct pipe_resource) {
.usage = PIPE_USAGE_STAGING,
.width0 = box->width + extra,
.height0 = box->height,
.depth0 = 1,
.nr_samples = xfer->resource->nr_samples,
.nr_storage_samples = xfer->resource->nr_storage_samples,
.array_size = box->depth,
.format = res->internal_format,
};
if (xfer->resource->target == PIPE_BUFFER)
templ.target = PIPE_BUFFER;
else if (templ.array_size > 1)
templ.target = PIPE_TEXTURE_2D_ARRAY;
else
templ.target = PIPE_TEXTURE_2D;
map->staging = iris_resource_create(pscreen, &templ);
assert(map->staging);
if (templ.target != PIPE_BUFFER) {
struct isl_surf *surf = &((struct iris_resource *) map->staging)->surf;
xfer->stride = isl_surf_get_row_pitch_B(surf);
xfer->layer_stride = isl_surf_get_array_pitch(surf);
}
if (!(xfer->usage & PIPE_TRANSFER_DISCARD_RANGE)) {
iris_copy_region(map->blorp, map->batch, map->staging, 0, extra, 0, 0,
xfer->resource, xfer->level, box);
/* Ensure writes to the staging BO land before we map it below. */
iris_emit_pipe_control_flush(map->batch,
"transfer read: flush before mapping",
PIPE_CONTROL_RENDER_TARGET_FLUSH |
PIPE_CONTROL_CS_STALL);
}
struct iris_bo *staging_bo = iris_resource_bo(map->staging);
if (iris_batch_references(map->batch, staging_bo))
iris_batch_flush(map->batch);
map->ptr =
iris_bo_map(map->dbg, staging_bo, xfer->usage & MAP_FLAGS) + extra;
map->unmap = iris_unmap_copy_region;
}
static void
get_image_offset_el(const struct isl_surf *surf, unsigned level, unsigned z,
unsigned *out_x0_el, unsigned *out_y0_el)
{
if (surf->dim == ISL_SURF_DIM_3D) {
isl_surf_get_image_offset_el(surf, level, 0, z, out_x0_el, out_y0_el);
} else {
isl_surf_get_image_offset_el(surf, level, z, 0, out_x0_el, out_y0_el);
}
}
/**
* This function computes the tile_w (in bytes) and tile_h (in rows) of
* different tiling patterns.
*/
static void
iris_resource_get_tile_dims(enum isl_tiling tiling, uint32_t cpp,
uint32_t *tile_w, uint32_t *tile_h)
{
switch (tiling) {
case ISL_TILING_X:
*tile_w = 512;
*tile_h = 8;
break;
case ISL_TILING_Y0:
*tile_w = 128;
*tile_h = 32;
break;
case ISL_TILING_LINEAR:
*tile_w = cpp;
*tile_h = 1;
break;
default:
unreachable("not reached");
}
}
/**
* This function computes masks that may be used to select the bits of the X
* and Y coordinates that indicate the offset within a tile. If the BO is
* untiled, the masks are set to 0.
*/
static void
iris_resource_get_tile_masks(enum isl_tiling tiling, uint32_t cpp,
uint32_t *mask_x, uint32_t *mask_y)
{
uint32_t tile_w_bytes, tile_h;
iris_resource_get_tile_dims(tiling, cpp, &tile_w_bytes, &tile_h);
*mask_x = tile_w_bytes / cpp - 1;
*mask_y = tile_h - 1;
}
/**
* Compute the offset (in bytes) from the start of the BO to the given x
* and y coordinate. For tiled BOs, caller must ensure that x and y are
* multiples of the tile size.
*/
static uint32_t
iris_resource_get_aligned_offset(const struct iris_resource *res,
uint32_t x, uint32_t y)
{
const struct isl_format_layout *fmtl = isl_format_get_layout(res->surf.format);
unsigned cpp = fmtl->bpb / 8;
uint32_t pitch = res->surf.row_pitch_B;
switch (res->surf.tiling) {
default:
unreachable("not reached");
case ISL_TILING_LINEAR:
return y * pitch + x * cpp;
case ISL_TILING_X:
assert((x % (512 / cpp)) == 0);
assert((y % 8) == 0);
return y * pitch + x / (512 / cpp) * 4096;
case ISL_TILING_Y0:
assert((x % (128 / cpp)) == 0);
assert((y % 32) == 0);
return y * pitch + x / (128 / cpp) * 4096;
}
}
/**
* Rendering with tiled buffers requires that the base address of the buffer
* be aligned to a page boundary. For renderbuffers, and sometimes with
* textures, we may want the surface to point at a texture image level that
* isn't at a page boundary.
*
* This function returns an appropriately-aligned base offset
* according to the tiling restrictions, plus any required x/y offset
* from there.
*/
uint32_t
iris_resource_get_tile_offsets(const struct iris_resource *res,
uint32_t level, uint32_t z,
uint32_t *tile_x, uint32_t *tile_y)
{
uint32_t x, y;
uint32_t mask_x, mask_y;
const struct isl_format_layout *fmtl = isl_format_get_layout(res->surf.format);
const unsigned cpp = fmtl->bpb / 8;
iris_resource_get_tile_masks(res->surf.tiling, cpp, &mask_x, &mask_y);
get_image_offset_el(&res->surf, level, z, &x, &y);
*tile_x = x & mask_x;
*tile_y = y & mask_y;
return iris_resource_get_aligned_offset(res, x & ~mask_x, y & ~mask_y);
}
/**
* Get pointer offset into stencil buffer.
*
* The stencil buffer is W tiled. Since the GTT is incapable of W fencing, we
* must decode the tile's layout in software.
*
* See
* - PRM, 2011 Sandy Bridge, Volume 1, Part 2, Section 4.5.2.1 W-Major Tile
* Format.
* - PRM, 2011 Sandy Bridge, Volume 1, Part 2, Section 4.5.3 Tiling Algorithm
*
* Even though the returned offset is always positive, the return type is
* signed due to
* commit e8b1c6d6f55f5be3bef25084fdd8b6127517e137
* mesa: Fix return type of _mesa_get_format_bytes() (#37351)
*/
static intptr_t
s8_offset(uint32_t stride, uint32_t x, uint32_t y)
{
uint32_t tile_size = 4096;
uint32_t tile_width = 64;
uint32_t tile_height = 64;
uint32_t row_size = 64 * stride / 2; /* Two rows are interleaved. */
uint32_t tile_x = x / tile_width;
uint32_t tile_y = y / tile_height;
/* The byte's address relative to the tile's base addres. */
uint32_t byte_x = x % tile_width;
uint32_t byte_y = y % tile_height;
uintptr_t u = tile_y * row_size
+ tile_x * tile_size
+ 512 * (byte_x / 8)
+ 64 * (byte_y / 8)
+ 32 * ((byte_y / 4) % 2)
+ 16 * ((byte_x / 4) % 2)
+ 8 * ((byte_y / 2) % 2)
+ 4 * ((byte_x / 2) % 2)
+ 2 * (byte_y % 2)
+ 1 * (byte_x % 2);
return u;
}
static void
iris_unmap_s8(struct iris_transfer *map)
{
struct pipe_transfer *xfer = &map->base;
const struct pipe_box *box = &xfer->box;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
struct isl_surf *surf = &res->surf;
if (xfer->usage & PIPE_TRANSFER_WRITE) {
uint8_t *untiled_s8_map = map->ptr;
uint8_t *tiled_s8_map =
iris_bo_map(map->dbg, res->bo, (xfer->usage | MAP_RAW) & MAP_FLAGS);
for (int s = 0; s < box->depth; s++) {
unsigned x0_el, y0_el;
get_image_offset_el(surf, xfer->level, box->z + s, &x0_el, &y0_el);
for (uint32_t y = 0; y < box->height; y++) {
for (uint32_t x = 0; x < box->width; x++) {
ptrdiff_t offset = s8_offset(surf->row_pitch_B,
x0_el + box->x + x,
y0_el + box->y + y);
tiled_s8_map[offset] =
untiled_s8_map[s * xfer->layer_stride + y * xfer->stride + x];
}
}
}
}
free(map->buffer);
}
static void
iris_map_s8(struct iris_transfer *map)
{
struct pipe_transfer *xfer = &map->base;
const struct pipe_box *box = &xfer->box;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
struct isl_surf *surf = &res->surf;
xfer->stride = surf->row_pitch_B;
xfer->layer_stride = xfer->stride * box->height;
/* The tiling and detiling functions require that the linear buffer has
* a 16-byte alignment (that is, its `x0` is 16-byte aligned). Here we
* over-allocate the linear buffer to get the proper alignment.
*/
map->buffer = map->ptr = malloc(xfer->layer_stride * box->depth);
assert(map->buffer);
/* One of either READ_BIT or WRITE_BIT or both is set. READ_BIT implies no
* INVALIDATE_RANGE_BIT. WRITE_BIT needs the original values read in unless
* invalidate is set, since we'll be writing the whole rectangle from our
* temporary buffer back out.
*/
if (!(xfer->usage & PIPE_TRANSFER_DISCARD_RANGE)) {
uint8_t *untiled_s8_map = map->ptr;
uint8_t *tiled_s8_map =
iris_bo_map(map->dbg, res->bo, (xfer->usage | MAP_RAW) & MAP_FLAGS);
for (int s = 0; s < box->depth; s++) {
unsigned x0_el, y0_el;
get_image_offset_el(surf, xfer->level, box->z + s, &x0_el, &y0_el);
for (uint32_t y = 0; y < box->height; y++) {
for (uint32_t x = 0; x < box->width; x++) {
ptrdiff_t offset = s8_offset(surf->row_pitch_B,
x0_el + box->x + x,
y0_el + box->y + y);
untiled_s8_map[s * xfer->layer_stride + y * xfer->stride + x] =
tiled_s8_map[offset];
}
}
}
}
map->unmap = iris_unmap_s8;
}
/* Compute extent parameters for use with tiled_memcpy functions.
* xs are in units of bytes and ys are in units of strides.
*/
static inline void
tile_extents(const struct isl_surf *surf,
const struct pipe_box *box,
unsigned level, int z,
unsigned *x1_B, unsigned *x2_B,
unsigned *y1_el, unsigned *y2_el)
{
const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
const unsigned cpp = fmtl->bpb / 8;
assert(box->x % fmtl->bw == 0);
assert(box->y % fmtl->bh == 0);
unsigned x0_el, y0_el;
get_image_offset_el(surf, level, box->z + z, &x0_el, &y0_el);
*x1_B = (box->x / fmtl->bw + x0_el) * cpp;
*y1_el = box->y / fmtl->bh + y0_el;
*x2_B = (DIV_ROUND_UP(box->x + box->width, fmtl->bw) + x0_el) * cpp;
*y2_el = DIV_ROUND_UP(box->y + box->height, fmtl->bh) + y0_el;
}
static void
iris_unmap_tiled_memcpy(struct iris_transfer *map)
{
struct pipe_transfer *xfer = &map->base;
const struct pipe_box *box = &xfer->box;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
struct isl_surf *surf = &res->surf;
const bool has_swizzling = false;
if (xfer->usage & PIPE_TRANSFER_WRITE) {
char *dst =
iris_bo_map(map->dbg, res->bo, (xfer->usage | MAP_RAW) & MAP_FLAGS);
for (int s = 0; s < box->depth; s++) {
unsigned x1, x2, y1, y2;
tile_extents(surf, box, xfer->level, s, &x1, &x2, &y1, &y2);
void *ptr = map->ptr + s * xfer->layer_stride;
isl_memcpy_linear_to_tiled(x1, x2, y1, y2, dst, ptr,
surf->row_pitch_B, xfer->stride,
has_swizzling, surf->tiling, ISL_MEMCPY);
}
}
os_free_aligned(map->buffer);
map->buffer = map->ptr = NULL;
}
static void
iris_map_tiled_memcpy(struct iris_transfer *map)
{
struct pipe_transfer *xfer = &map->base;
const struct pipe_box *box = &xfer->box;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
struct isl_surf *surf = &res->surf;
xfer->stride = ALIGN(surf->row_pitch_B, 16);
xfer->layer_stride = xfer->stride * box->height;
unsigned x1, x2, y1, y2;
tile_extents(surf, box, xfer->level, 0, &x1, &x2, &y1, &y2);
/* The tiling and detiling functions require that the linear buffer has
* a 16-byte alignment (that is, its `x0` is 16-byte aligned). Here we
* over-allocate the linear buffer to get the proper alignment.
*/
map->buffer =
os_malloc_aligned(xfer->layer_stride * box->depth, 16);
assert(map->buffer);
map->ptr = (char *)map->buffer + (x1 & 0xf);
const bool has_swizzling = false;
if (!(xfer->usage & PIPE_TRANSFER_DISCARD_RANGE)) {
char *src =
iris_bo_map(map->dbg, res->bo, (xfer->usage | MAP_RAW) & MAP_FLAGS);
for (int s = 0; s < box->depth; s++) {
unsigned x1, x2, y1, y2;
tile_extents(surf, box, xfer->level, s, &x1, &x2, &y1, &y2);
/* Use 's' rather than 'box->z' to rebase the first slice to 0. */
void *ptr = map->ptr + s * xfer->layer_stride;
isl_memcpy_tiled_to_linear(x1, x2, y1, y2, ptr, src, xfer->stride,
surf->row_pitch_B, has_swizzling,
surf->tiling, ISL_MEMCPY_STREAMING_LOAD);
}
}
map->unmap = iris_unmap_tiled_memcpy;
}
static void
iris_map_direct(struct iris_transfer *map)
{
struct pipe_transfer *xfer = &map->base;
struct pipe_box *box = &xfer->box;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
void *ptr = iris_bo_map(map->dbg, res->bo, xfer->usage & MAP_FLAGS);
if (res->base.target == PIPE_BUFFER) {
xfer->stride = 0;
xfer->layer_stride = 0;
map->ptr = ptr + box->x;
} else {
struct isl_surf *surf = &res->surf;
const struct isl_format_layout *fmtl =
isl_format_get_layout(surf->format);
const unsigned cpp = fmtl->bpb / 8;
unsigned x0_el, y0_el;
get_image_offset_el(surf, xfer->level, box->z, &x0_el, &y0_el);
xfer->stride = isl_surf_get_row_pitch_B(surf);
xfer->layer_stride = isl_surf_get_array_pitch(surf);
map->ptr = ptr + (y0_el + box->y) * xfer->stride + (x0_el + box->x) * cpp;
}
}
static bool
can_promote_to_async(const struct iris_resource *res,
const struct pipe_box *box,
enum pipe_transfer_usage usage)
{
/* If we're writing to a section of the buffer that hasn't even been
* initialized with useful data, then we can safely promote this write
* to be unsynchronized. This helps the common pattern of appending data.
*/
return res->base.target == PIPE_BUFFER && (usage & PIPE_TRANSFER_WRITE) &&
!(usage & TC_TRANSFER_MAP_NO_INFER_UNSYNCHRONIZED) &&
!util_ranges_intersect(&res->valid_buffer_range, box->x,
box->x + box->width);
}
static void *
iris_transfer_map(struct pipe_context *ctx,
struct pipe_resource *resource,
unsigned level,
enum pipe_transfer_usage usage,
const struct pipe_box *box,
struct pipe_transfer **ptransfer)
{
struct iris_context *ice = (struct iris_context *)ctx;
struct iris_resource *res = (struct iris_resource *)resource;
struct isl_surf *surf = &res->surf;
if (usage & PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE) {
/* Replace the backing storage with a fresh buffer for non-async maps */
if (!(usage & (PIPE_TRANSFER_UNSYNCHRONIZED |
TC_TRANSFER_MAP_NO_INVALIDATE)))
iris_invalidate_resource(ctx, resource);
/* If we can discard the whole resource, we can discard the range. */
usage |= PIPE_TRANSFER_DISCARD_RANGE;
}
if (!(usage & PIPE_TRANSFER_UNSYNCHRONIZED) &&
can_promote_to_async(res, box, usage)) {
usage |= PIPE_TRANSFER_UNSYNCHRONIZED;
}
bool need_resolve = false;
bool need_color_resolve = false;
if (resource->target != PIPE_BUFFER) {
bool need_hiz_resolve = iris_resource_level_has_hiz(res, level);
bool need_stencil_resolve = res->aux.usage == ISL_AUX_USAGE_STC_CCS;
need_color_resolve =
(res->aux.usage == ISL_AUX_USAGE_CCS_D ||
res->aux.usage == ISL_AUX_USAGE_CCS_E) &&
iris_has_color_unresolved(res, level, 1, box->z, box->depth);
need_resolve = need_color_resolve ||
need_hiz_resolve ||
need_stencil_resolve;
}
bool map_would_stall = false;
if (!(usage & PIPE_TRANSFER_UNSYNCHRONIZED)) {
map_would_stall = need_resolve || resource_is_busy(ice, res);
if (map_would_stall && (usage & PIPE_TRANSFER_DONTBLOCK) &&
(usage & PIPE_TRANSFER_MAP_DIRECTLY))
return NULL;
}
if (surf->tiling != ISL_TILING_LINEAR &&
(usage & PIPE_TRANSFER_MAP_DIRECTLY))
return NULL;
struct iris_transfer *map = slab_alloc(&ice->transfer_pool);
struct pipe_transfer *xfer = &map->base;
if (!map)
return NULL;
memset(map, 0, sizeof(*map));
map->dbg = &ice->dbg;
pipe_resource_reference(&xfer->resource, resource);
xfer->level = level;
xfer->usage = usage;
xfer->box = *box;
*ptransfer = xfer;
map->dest_had_defined_contents =
util_ranges_intersect(&res->valid_buffer_range, box->x,
box->x + box->width);
if (usage & PIPE_TRANSFER_WRITE)
util_range_add(&res->base, &res->valid_buffer_range, box->x, box->x + box->width);
/* Avoid using GPU copies for persistent/coherent buffers, as the idea
* there is to access them simultaneously on the CPU & GPU. This also
* avoids trying to use GPU copies for our u_upload_mgr buffers which
* contain state we're constructing for a GPU draw call, which would
* kill us with infinite stack recursion.
*/
bool no_gpu = usage & (PIPE_TRANSFER_PERSISTENT |
PIPE_TRANSFER_COHERENT |
PIPE_TRANSFER_MAP_DIRECTLY);
/* GPU copies are not useful for buffer reads. Instead of stalling to
* read from the original buffer, we'd simply copy it to a temporary...
* then stall (a bit longer) to read from that buffer.
*
* Images are less clear-cut. Color resolves are destructive, removing
* the underlying compression, so we'd rather blit the data to a linear
* temporary and map that, to avoid the resolve. (It might be better to
* a tiled temporary and use the tiled_memcpy paths...)
*/
if (!(usage & PIPE_TRANSFER_DISCARD_RANGE) && !need_color_resolve)
no_gpu = true;
const struct isl_format_layout *fmtl = isl_format_get_layout(surf->format);
if (fmtl->txc == ISL_TXC_ASTC)
no_gpu = true;
if ((map_would_stall || res->aux.usage == ISL_AUX_USAGE_CCS_E) && !no_gpu) {
/* If we need a synchronous mapping and the resource is busy, or needs
* resolving, we copy to/from a linear temporary buffer using the GPU.
*/
map->batch = &ice->batches[IRIS_BATCH_RENDER];
map->blorp = &ice->blorp;
iris_map_copy_region(map);
} else {
/* Otherwise we're free to map on the CPU. */
if (need_resolve) {
iris_resource_access_raw(ice, &ice->batches[IRIS_BATCH_RENDER], res,
level, box->z, box->depth,
usage & PIPE_TRANSFER_WRITE);
}
if (!(usage & PIPE_TRANSFER_UNSYNCHRONIZED)) {
for (int i = 0; i < IRIS_BATCH_COUNT; i++) {
if (iris_batch_references(&ice->batches[i], res->bo))
iris_batch_flush(&ice->batches[i]);
}
}
if (surf->tiling == ISL_TILING_W) {
/* TODO: Teach iris_map_tiled_memcpy about W-tiling... */
iris_map_s8(map);
} else if (surf->tiling != ISL_TILING_LINEAR) {
iris_map_tiled_memcpy(map);
} else {
iris_map_direct(map);
}
}
return map->ptr;
}
static void
iris_transfer_flush_region(struct pipe_context *ctx,
struct pipe_transfer *xfer,
const struct pipe_box *box)
{
struct iris_context *ice = (struct iris_context *)ctx;
struct iris_resource *res = (struct iris_resource *) xfer->resource;
struct iris_transfer *map = (void *) xfer;
if (map->staging)
iris_flush_staging_region(xfer, box);
uint32_t history_flush = 0;
if (res->base.target == PIPE_BUFFER) {
if (map->staging)
history_flush |= PIPE_CONTROL_RENDER_TARGET_FLUSH;
if (map->dest_had_defined_contents)
history_flush |= iris_flush_bits_for_history(res);
util_range_add(&res->base, &res->valid_buffer_range, box->x, box->x + box->width);
}
if (history_flush & ~PIPE_CONTROL_CS_STALL) {
for (int i = 0; i < IRIS_BATCH_COUNT; i++) {
struct iris_batch *batch = &ice->batches[i];
if (batch->contains_draw || batch->cache.render->entries) {
iris_batch_maybe_flush(batch, 24);
iris_emit_pipe_control_flush(batch,
"cache history: transfer flush",
history_flush);
}
}
}
/* Make sure we flag constants dirty even if there's no need to emit
* any PIPE_CONTROLs to a batch.
*/
iris_dirty_for_history(ice, res);
}
static void
iris_transfer_unmap(struct pipe_context *ctx, struct pipe_transfer *xfer)
{
struct iris_context *ice = (struct iris_context *)ctx;
struct iris_transfer *map = (void *) xfer;
if (!(xfer->usage & (PIPE_TRANSFER_FLUSH_EXPLICIT |
PIPE_TRANSFER_COHERENT))) {
struct pipe_box flush_box = {
.x = 0, .y = 0, .z = 0,
.width = xfer->box.width,
.height = xfer->box.height,
.depth = xfer->box.depth,
};
iris_transfer_flush_region(ctx, xfer, &flush_box);
}
if (map->unmap)
map->unmap(map);
pipe_resource_reference(&xfer->resource, NULL);
slab_free(&ice->transfer_pool, map);
}
/**
* Mark state dirty that needs to be re-emitted when a resource is written.
*/
void
iris_dirty_for_history(struct iris_context *ice,
struct iris_resource *res)
{
uint64_t dirty = 0ull;
if (res->bind_history & PIPE_BIND_CONSTANT_BUFFER) {
dirty |= ((uint64_t)res->bind_stages) << IRIS_SHIFT_FOR_DIRTY_CONSTANTS;
}
ice->state.dirty |= dirty;
}
/**
* Produce a set of PIPE_CONTROL bits which ensure data written to a
* resource becomes visible, and any stale read cache data is invalidated.
*/
uint32_t
iris_flush_bits_for_history(struct iris_resource *res)
{
uint32_t flush = PIPE_CONTROL_CS_STALL;
if (res->bind_history & PIPE_BIND_CONSTANT_BUFFER) {
flush |= PIPE_CONTROL_CONST_CACHE_INVALIDATE |
PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
}
if (res->bind_history & PIPE_BIND_SAMPLER_VIEW)
flush |= PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
if (res->bind_history & (PIPE_BIND_VERTEX_BUFFER | PIPE_BIND_INDEX_BUFFER))
flush |= PIPE_CONTROL_VF_CACHE_INVALIDATE;
if (res->bind_history & (PIPE_BIND_SHADER_BUFFER | PIPE_BIND_SHADER_IMAGE))
flush |= PIPE_CONTROL_DATA_CACHE_FLUSH;
return flush;
}
void
iris_flush_and_dirty_for_history(struct iris_context *ice,
struct iris_batch *batch,
struct iris_resource *res,
uint32_t extra_flags,
const char *reason)
{
if (res->base.target != PIPE_BUFFER)
return;
uint32_t flush = iris_flush_bits_for_history(res) | extra_flags;
iris_emit_pipe_control_flush(batch, reason, flush);
iris_dirty_for_history(ice, res);
}
bool
iris_resource_set_clear_color(struct iris_context *ice,
struct iris_resource *res,
union isl_color_value color)
{
if (memcmp(&res->aux.clear_color, &color, sizeof(color)) != 0) {
res->aux.clear_color = color;
return true;
}
return false;
}
union isl_color_value
iris_resource_get_clear_color(const struct iris_resource *res,
struct iris_bo **clear_color_bo,
uint64_t *clear_color_offset)
{
assert(res->aux.bo);
if (clear_color_bo)
*clear_color_bo = res->aux.clear_color_bo;
if (clear_color_offset)
*clear_color_offset = res->aux.clear_color_offset;
return res->aux.clear_color;
}
static enum pipe_format
iris_resource_get_internal_format(struct pipe_resource *p_res)
{
struct iris_resource *res = (void *) p_res;
return res->internal_format;
}
static const struct u_transfer_vtbl transfer_vtbl = {
.resource_create = iris_resource_create,
.resource_destroy = iris_resource_destroy,
.transfer_map = iris_transfer_map,
.transfer_unmap = iris_transfer_unmap,
.transfer_flush_region = iris_transfer_flush_region,
.get_internal_format = iris_resource_get_internal_format,
.set_stencil = iris_resource_set_separate_stencil,
.get_stencil = iris_resource_get_separate_stencil,
};
void
iris_init_screen_resource_functions(struct pipe_screen *pscreen)
{
pscreen->query_dmabuf_modifiers = iris_query_dmabuf_modifiers;
pscreen->resource_create_with_modifiers =
iris_resource_create_with_modifiers;
pscreen->resource_create = u_transfer_helper_resource_create;
pscreen->resource_from_user_memory = iris_resource_from_user_memory;
pscreen->resource_from_handle = iris_resource_from_handle;
pscreen->resource_get_handle = iris_resource_get_handle;
pscreen->resource_get_param = iris_resource_get_param;
pscreen->resource_destroy = u_transfer_helper_resource_destroy;
pscreen->transfer_helper =
u_transfer_helper_create(&transfer_vtbl, true, true, false, true);
}
void
iris_init_resource_functions(struct pipe_context *ctx)
{
ctx->flush_resource = iris_flush_resource;
ctx->invalidate_resource = iris_invalidate_resource;
ctx->transfer_map = u_transfer_helper_transfer_map;
ctx->transfer_flush_region = u_transfer_helper_transfer_flush_region;
ctx->transfer_unmap = u_transfer_helper_transfer_unmap;
ctx->buffer_subdata = u_default_buffer_subdata;
ctx->texture_subdata = u_default_texture_subdata;
}
|