summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/i965/brw_context.c
blob: 6ca20f24b947f54bff88dfc318d62c1391d6abf5 (plain)
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
/*
 Copyright 2003 VMware, Inc.
 Copyright (C) Intel Corp.  2006.  All Rights Reserved.
 Intel funded Tungsten Graphics to
 develop this 3D driver.

 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.

 **********************************************************************/
 /*
  * Authors:
  *   Keith Whitwell <keithw@vmware.com>
  */


#include "compiler/nir/nir.h"
#include "main/api_exec.h"
#include "main/context.h"
#include "main/fbobject.h"
#include "main/extensions.h"
#include "main/glthread.h"
#include "main/macros.h"
#include "main/points.h"
#include "main/version.h"
#include "main/vtxfmt.h"
#include "main/texobj.h"
#include "main/framebuffer.h"
#include "main/stencil.h"
#include "main/state.h"
#include "main/spirv_extensions.h"

#include "vbo/vbo.h"

#include "drivers/common/driverfuncs.h"
#include "drivers/common/meta.h"
#include "utils.h"

#include "brw_context.h"
#include "brw_defines.h"
#include "brw_blorp.h"
#include "brw_draw.h"
#include "brw_state.h"

#include "intel_batchbuffer.h"
#include "intel_buffer_objects.h"
#include "intel_buffers.h"
#include "intel_fbo.h"
#include "intel_mipmap_tree.h"
#include "intel_pixel.h"
#include "intel_image.h"
#include "intel_tex.h"
#include "intel_tex_obj.h"

#include "swrast_setup/swrast_setup.h"
#include "tnl/tnl.h"
#include "tnl/t_pipeline.h"
#include "util/ralloc.h"
#include "util/debug.h"
#include "util/disk_cache.h"
#include "util/u_memory.h"
#include "isl/isl.h"

#include "common/gen_defines.h"

#include "compiler/spirv/nir_spirv.h"
/***************************************
 * Mesa's Driver Functions
 ***************************************/

const char *const brw_vendor_string = "Intel Open Source Technology Center";

static const char *
get_bsw_model(const struct intel_screen *screen)
{
   switch (screen->eu_total) {
   case 16:
      return "405";
   case 12:
      return "400";
   default:
      return "   ";
   }
}

const char *
brw_get_renderer_string(const struct intel_screen *screen)
{
   static char buf[128];
   const char *name = gen_get_device_name(screen->deviceID);

   if (!name)
      name = "Intel Unknown";

   snprintf(buf, sizeof(buf), "Mesa DRI %s", name);

   /* Braswell branding is funny, so we have to fix it up here */
   if (screen->deviceID == 0x22B1) {
      char *needle = strstr(buf, "XXX");
      if (needle)
         memcpy(needle, get_bsw_model(screen), 3);
   }

   return buf;
}

static const GLubyte *
intel_get_string(struct gl_context * ctx, GLenum name)
{
   const struct brw_context *const brw = brw_context(ctx);

   switch (name) {
   case GL_VENDOR:
      return (GLubyte *) brw_vendor_string;

   case GL_RENDERER:
      return
         (GLubyte *) brw_get_renderer_string(brw->screen);

   default:
      return NULL;
   }
}

static void
brw_set_background_context(struct gl_context *ctx,
                           struct util_queue_monitoring *queue_info)
{
   struct brw_context *brw = brw_context(ctx);
   __DRIcontext *driContext = brw->driContext;
   __DRIscreen *driScreen = driContext->driScreenPriv;
   const __DRIbackgroundCallableExtension *backgroundCallable =
      driScreen->dri2.backgroundCallable;

   /* Note: Mesa will only call this function if we've called
    * _mesa_enable_multithreading().  We only do that if the loader exposed
    * the __DRI_BACKGROUND_CALLABLE extension.  So we know that
    * backgroundCallable is not NULL.
    */
   backgroundCallable->setBackgroundContext(driContext->loaderPrivate);
}

static void
intel_viewport(struct gl_context *ctx)
{
   struct brw_context *brw = brw_context(ctx);
   __DRIcontext *driContext = brw->driContext;

   if (_mesa_is_winsys_fbo(ctx->DrawBuffer)) {
      if (driContext->driDrawablePriv)
         dri2InvalidateDrawable(driContext->driDrawablePriv);
      if (driContext->driReadablePriv)
         dri2InvalidateDrawable(driContext->driReadablePriv);
   }
}

static void
intel_update_framebuffer(struct gl_context *ctx,
                         struct gl_framebuffer *fb)
{
   struct brw_context *brw = brw_context(ctx);

   /* Quantize the derived default number of samples
    */
   fb->DefaultGeometry._NumSamples =
      intel_quantize_num_samples(brw->screen,
                                 fb->DefaultGeometry.NumSamples);
}

static void
intel_update_state(struct gl_context * ctx)
{
   GLuint new_state = ctx->NewState;
   struct brw_context *brw = brw_context(ctx);

   if (ctx->swrast_context)
      _swrast_InvalidateState(ctx, new_state);

   brw->NewGLState |= new_state;

   if (new_state & (_NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT))
      _mesa_update_draw_buffer_bounds(ctx, ctx->DrawBuffer);

   if (new_state & (_NEW_STENCIL | _NEW_BUFFERS)) {
      brw->stencil_enabled = _mesa_stencil_is_enabled(ctx);
      brw->stencil_two_sided = _mesa_stencil_is_two_sided(ctx);
      brw->stencil_write_enabled =
         _mesa_stencil_is_write_enabled(ctx, brw->stencil_two_sided);
   }

   if (new_state & _NEW_POLYGON)
      brw->polygon_front_bit = _mesa_polygon_get_front_bit(ctx);

   if (new_state & _NEW_BUFFERS) {
      intel_update_framebuffer(ctx, ctx->DrawBuffer);
      if (ctx->DrawBuffer != ctx->ReadBuffer)
         intel_update_framebuffer(ctx, ctx->ReadBuffer);
   }
}

#define flushFront(screen)      ((screen)->image.loader ? (screen)->image.loader->flushFrontBuffer : (screen)->dri2.loader->flushFrontBuffer)

static void
intel_flush_front(struct gl_context *ctx)
{
   struct brw_context *brw = brw_context(ctx);
   __DRIcontext *driContext = brw->driContext;
   __DRIdrawable *driDrawable = driContext->driDrawablePriv;
   __DRIscreen *const dri_screen = brw->screen->driScrnPriv;

   if (brw->front_buffer_dirty && _mesa_is_winsys_fbo(ctx->DrawBuffer)) {
      if (flushFront(dri_screen) && driDrawable &&
          driDrawable->loaderPrivate) {

         /* Resolve before flushing FAKE_FRONT_LEFT to FRONT_LEFT.
          *
          * This potentially resolves both front and back buffer. It
          * is unnecessary to resolve the back, but harms nothing except
          * performance. And no one cares about front-buffer render
          * performance.
          */
         intel_resolve_for_dri2_flush(brw, driDrawable);
         intel_batchbuffer_flush(brw);

         flushFront(dri_screen)(driDrawable, driDrawable->loaderPrivate);

         /* We set the dirty bit in intel_prepare_render() if we're
          * front buffer rendering once we get there.
          */
         brw->front_buffer_dirty = false;
      }
   }
}

static void
brw_display_shared_buffer(struct brw_context *brw)
{
   __DRIcontext *dri_context = brw->driContext;
   __DRIdrawable *dri_drawable = dri_context->driDrawablePriv;
   __DRIscreen *dri_screen = brw->screen->driScrnPriv;
   int fence_fd = -1;

   if (!brw->is_shared_buffer_bound)
      return;

   if (!brw->is_shared_buffer_dirty)
      return;

   if (brw->screen->has_exec_fence) {
      /* This function is always called during a flush operation, so there is
       * no need to flush again here. But we want to provide a fence_fd to the
       * loader, and a redundant flush is the easiest way to acquire one.
       */
      if (intel_batchbuffer_flush_fence(brw, -1, &fence_fd))
         return;
   }

   dri_screen->mutableRenderBuffer.loader
      ->displaySharedBuffer(dri_drawable, fence_fd,
                            dri_drawable->loaderPrivate);
   brw->is_shared_buffer_dirty = false;
}

static void
intel_glFlush(struct gl_context *ctx)
{
   struct brw_context *brw = brw_context(ctx);

   intel_batchbuffer_flush(brw);
   intel_flush_front(ctx);
   brw_display_shared_buffer(brw);
   brw->need_flush_throttle = true;
}

static void
intel_glEnable(struct gl_context *ctx, GLenum cap, GLboolean state)
{
   struct brw_context *brw = brw_context(ctx);

   switch (cap) {
   case GL_BLACKHOLE_RENDER_INTEL:
      brw->frontend_noop = state;
      intel_batchbuffer_flush(brw);
      intel_batchbuffer_maybe_noop(brw);
      /* Because we started previous batches with a potential
       * MI_BATCH_BUFFER_END if NOOP was enabled, that means that anything
       * that was ever emitted after that never made it to the HW. So when the
       * blackhole state changes from NOOP->!NOOP reupload the entire state.
       */
      if (!brw->frontend_noop) {
         brw->NewGLState = ~0u;
         brw->ctx.NewDriverState = ~0ull;
      }
      break;
   default:
      break;
   }
}

static void
intel_finish(struct gl_context * ctx)
{
   struct brw_context *brw = brw_context(ctx);

   intel_glFlush(ctx);

   if (brw->batch.last_bo)
      brw_bo_wait_rendering(brw->batch.last_bo);
}

static void
brw_init_driver_functions(struct brw_context *brw,
                          struct dd_function_table *functions)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;

   _mesa_init_driver_functions(functions);

   /* GLX uses DRI2 invalidate events to handle window resizing.
    * Unfortunately, EGL does not - libEGL is written in XCB (not Xlib),
    * which doesn't provide a mechanism for snooping the event queues.
    *
    * So EGL still relies on viewport hacks to handle window resizing.
    * This should go away with DRI3000.
    */
   if (!brw->driContext->driScreenPriv->dri2.useInvalidate)
      functions->Viewport = intel_viewport;

   functions->Enable = intel_glEnable;
   functions->Flush = intel_glFlush;
   functions->Finish = intel_finish;
   functions->GetString = intel_get_string;
   functions->UpdateState = intel_update_state;

   brw_init_draw_functions(functions);
   intelInitTextureFuncs(functions);
   intelInitTextureImageFuncs(functions);
   intelInitTextureCopyImageFuncs(functions);
   intelInitCopyImageFuncs(functions);
   intelInitClearFuncs(functions);
   intelInitBufferFuncs(functions);
   intelInitPixelFuncs(functions);
   intelInitBufferObjectFuncs(functions);
   brw_init_syncobj_functions(functions);
   brw_init_object_purgeable_functions(functions);

   brwInitFragProgFuncs( functions );
   brw_init_common_queryobj_functions(functions);
   if (devinfo->gen >= 8 || devinfo->is_haswell)
      hsw_init_queryobj_functions(functions);
   else if (devinfo->gen >= 6)
      gen6_init_queryobj_functions(functions);
   else
      gen4_init_queryobj_functions(functions);
   brw_init_compute_functions(functions);
   brw_init_conditional_render_functions(functions);

   functions->GenerateMipmap = brw_generate_mipmap;

   functions->QueryInternalFormat = brw_query_internal_format;

   functions->NewTransformFeedback = brw_new_transform_feedback;
   functions->DeleteTransformFeedback = brw_delete_transform_feedback;
   if (can_do_mi_math_and_lrr(brw->screen)) {
      functions->BeginTransformFeedback = hsw_begin_transform_feedback;
      functions->EndTransformFeedback = hsw_end_transform_feedback;
      functions->PauseTransformFeedback = hsw_pause_transform_feedback;
      functions->ResumeTransformFeedback = hsw_resume_transform_feedback;
   } else if (devinfo->gen >= 7) {
      functions->BeginTransformFeedback = gen7_begin_transform_feedback;
      functions->EndTransformFeedback = gen7_end_transform_feedback;
      functions->PauseTransformFeedback = gen7_pause_transform_feedback;
      functions->ResumeTransformFeedback = gen7_resume_transform_feedback;
      functions->GetTransformFeedbackVertexCount =
         brw_get_transform_feedback_vertex_count;
   } else {
      functions->BeginTransformFeedback = brw_begin_transform_feedback;
      functions->EndTransformFeedback = brw_end_transform_feedback;
      functions->PauseTransformFeedback = brw_pause_transform_feedback;
      functions->ResumeTransformFeedback = brw_resume_transform_feedback;
      functions->GetTransformFeedbackVertexCount =
         brw_get_transform_feedback_vertex_count;
   }

   if (devinfo->gen >= 6)
      functions->GetSamplePosition = gen6_get_sample_position;

   /* GL_ARB_get_program_binary */
   brw_program_binary_init(brw->screen->deviceID);
   functions->GetProgramBinaryDriverSHA1 = brw_get_program_binary_driver_sha1;
   functions->ProgramBinarySerializeDriverBlob = brw_serialize_program_binary;
   functions->ProgramBinaryDeserializeDriverBlob =
      brw_deserialize_program_binary;

   if (brw->screen->disk_cache) {
      functions->ShaderCacheSerializeDriverBlob = brw_program_serialize_nir;
   }

   functions->SetBackgroundContext = brw_set_background_context;
}

static void
brw_initialize_spirv_supported_capabilities(struct brw_context *brw)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   struct gl_context *ctx = &brw->ctx;

   /* The following SPIR-V capabilities are only supported on gen7+. In theory
    * you should enable the extension only on gen7+, but just in case let's
    * assert it.
    */
   assert(devinfo->gen >= 7);

   ctx->Const.SpirVCapabilities.atomic_storage = devinfo->gen >= 7;
   ctx->Const.SpirVCapabilities.draw_parameters = true;
   ctx->Const.SpirVCapabilities.float64 = devinfo->gen >= 8;
   ctx->Const.SpirVCapabilities.geometry_streams = devinfo->gen >= 7;
   ctx->Const.SpirVCapabilities.image_write_without_format = true;
   ctx->Const.SpirVCapabilities.int64 = devinfo->gen >= 8;
   ctx->Const.SpirVCapabilities.tessellation = true;
   ctx->Const.SpirVCapabilities.transform_feedback = devinfo->gen >= 7;
   ctx->Const.SpirVCapabilities.variable_pointers = true;
   ctx->Const.SpirVCapabilities.integer_functions2 = devinfo->gen >= 8;
}

static void
brw_initialize_context_constants(struct brw_context *brw)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   struct gl_context *ctx = &brw->ctx;
   const struct brw_compiler *compiler = brw->screen->compiler;

   const bool stage_exists[MESA_SHADER_STAGES] = {
      [MESA_SHADER_VERTEX] = true,
      [MESA_SHADER_TESS_CTRL] = devinfo->gen >= 7,
      [MESA_SHADER_TESS_EVAL] = devinfo->gen >= 7,
      [MESA_SHADER_GEOMETRY] = devinfo->gen >= 6,
      [MESA_SHADER_FRAGMENT] = true,
      [MESA_SHADER_COMPUTE] =
         (_mesa_is_desktop_gl(ctx) &&
          ctx->Const.MaxComputeWorkGroupSize[0] >= 1024) ||
         (ctx->API == API_OPENGLES2 &&
          ctx->Const.MaxComputeWorkGroupSize[0] >= 128),
   };

   unsigned num_stages = 0;
   for (int i = 0; i < MESA_SHADER_STAGES; i++) {
      if (stage_exists[i])
         num_stages++;
   }

   unsigned max_samplers =
      devinfo->gen >= 8 || devinfo->is_haswell ? BRW_MAX_TEX_UNIT : 16;

   ctx->Const.MaxDualSourceDrawBuffers = 1;
   ctx->Const.MaxDrawBuffers = BRW_MAX_DRAW_BUFFERS;
   ctx->Const.MaxCombinedShaderOutputResources =
      MAX_IMAGE_UNITS + BRW_MAX_DRAW_BUFFERS;

   /* The timestamp register we can read for glGetTimestamp() is
    * sometimes only 32 bits, before scaling to nanoseconds (depending
    * on kernel).
    *
    * Once scaled to nanoseconds the timestamp would roll over at a
    * non-power-of-two, so an application couldn't use
    * GL_QUERY_COUNTER_BITS to handle rollover correctly.  Instead, we
    * report 36 bits and truncate at that (rolling over 5 times as
    * often as the HW counter), and when the 32-bit counter rolls
    * over, it happens to also be at a rollover in the reported value
    * from near (1<<36) to 0.
    *
    * The low 32 bits rolls over in ~343 seconds.  Our 36-bit result
    * rolls over every ~69 seconds.
    */
   ctx->Const.QueryCounterBits.Timestamp = 36;

   ctx->Const.MaxTextureCoordUnits = 8; /* Mesa limit */
   ctx->Const.MaxImageUnits = MAX_IMAGE_UNITS;
   if (devinfo->gen >= 7) {
      ctx->Const.MaxRenderbufferSize = 16384;
      ctx->Const.MaxTextureSize = 16384;
      ctx->Const.MaxCubeTextureLevels = 15; /* 16384 */
   } else {
      ctx->Const.MaxRenderbufferSize = 8192;
      ctx->Const.MaxTextureSize = 8192;
      ctx->Const.MaxCubeTextureLevels = 14; /* 8192 */
   }
   ctx->Const.Max3DTextureLevels = 12; /* 2048 */
   ctx->Const.MaxArrayTextureLayers = devinfo->gen >= 7 ? 2048 : 512;
   ctx->Const.MaxTextureMbytes = 1536;
   ctx->Const.MaxTextureRectSize = devinfo->gen >= 7 ? 16384 : 8192;
   ctx->Const.MaxTextureMaxAnisotropy = 16.0;
   ctx->Const.MaxTextureLodBias = 15.0;
   ctx->Const.StripTextureBorder = true;
   if (devinfo->gen >= 7) {
      ctx->Const.MaxProgramTextureGatherComponents = 4;
      ctx->Const.MinProgramTextureGatherOffset = -32;
      ctx->Const.MaxProgramTextureGatherOffset = 31;
   } else if (devinfo->gen == 6) {
      ctx->Const.MaxProgramTextureGatherComponents = 1;
      ctx->Const.MinProgramTextureGatherOffset = -8;
      ctx->Const.MaxProgramTextureGatherOffset = 7;
   }

   ctx->Const.MaxUniformBlockSize = 65536;

   for (int i = 0; i < MESA_SHADER_STAGES; i++) {
      struct gl_program_constants *prog = &ctx->Const.Program[i];

      if (!stage_exists[i])
         continue;

      prog->MaxTextureImageUnits = max_samplers;

      prog->MaxUniformBlocks = BRW_MAX_UBO;
      prog->MaxCombinedUniformComponents =
         prog->MaxUniformComponents +
         ctx->Const.MaxUniformBlockSize / 4 * prog->MaxUniformBlocks;

      prog->MaxAtomicCounters = MAX_ATOMIC_COUNTERS;
      prog->MaxAtomicBuffers = BRW_MAX_ABO;
      prog->MaxImageUniforms = compiler->scalar_stage[i] ? BRW_MAX_IMAGES : 0;
      prog->MaxShaderStorageBlocks = BRW_MAX_SSBO;
   }

   ctx->Const.MaxTextureUnits =
      MIN2(ctx->Const.MaxTextureCoordUnits,
           ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits);

   ctx->Const.MaxUniformBufferBindings = num_stages * BRW_MAX_UBO;
   ctx->Const.MaxCombinedUniformBlocks = num_stages * BRW_MAX_UBO;
   ctx->Const.MaxCombinedAtomicBuffers = num_stages * BRW_MAX_ABO;
   ctx->Const.MaxCombinedShaderStorageBlocks = num_stages * BRW_MAX_SSBO;
   ctx->Const.MaxShaderStorageBufferBindings = num_stages * BRW_MAX_SSBO;
   ctx->Const.MaxCombinedTextureImageUnits = num_stages * max_samplers;
   ctx->Const.MaxCombinedImageUniforms = num_stages * BRW_MAX_IMAGES;


   /* Hardware only supports a limited number of transform feedback buffers.
    * So we need to override the Mesa default (which is based only on software
    * limits).
    */
   ctx->Const.MaxTransformFeedbackBuffers = BRW_MAX_SOL_BUFFERS;

   /* On Gen6, in the worst case, we use up one binding table entry per
    * transform feedback component (see comments above the definition of
    * BRW_MAX_SOL_BINDINGS, in brw_context.h), so we need to advertise a value
    * for MAX_TRANSFORM_FEEDBACK_INTERLEAVED_COMPONENTS equal to
    * BRW_MAX_SOL_BINDINGS.
    *
    * In "separate components" mode, we need to divide this value by
    * BRW_MAX_SOL_BUFFERS, so that the total number of binding table entries
    * used up by all buffers will not exceed BRW_MAX_SOL_BINDINGS.
    */
   ctx->Const.MaxTransformFeedbackInterleavedComponents = BRW_MAX_SOL_BINDINGS;
   ctx->Const.MaxTransformFeedbackSeparateComponents =
      BRW_MAX_SOL_BINDINGS / BRW_MAX_SOL_BUFFERS;

   ctx->Const.AlwaysUseGetTransformFeedbackVertexCount =
      !can_do_mi_math_and_lrr(brw->screen);

   int max_samples;
   const int *msaa_modes = intel_supported_msaa_modes(brw->screen);
   const int clamp_max_samples =
      driQueryOptioni(&brw->optionCache, "clamp_max_samples");

   if (clamp_max_samples < 0) {
      max_samples = msaa_modes[0];
   } else {
      /* Select the largest supported MSAA mode that does not exceed
       * clamp_max_samples.
       */
      max_samples = 0;
      for (int i = 0; msaa_modes[i] != 0; ++i) {
         if (msaa_modes[i] <= clamp_max_samples) {
            max_samples = msaa_modes[i];
            break;
         }
      }
   }

   ctx->Const.MaxSamples = max_samples;
   ctx->Const.MaxColorTextureSamples = max_samples;
   ctx->Const.MaxDepthTextureSamples = max_samples;
   ctx->Const.MaxIntegerSamples = max_samples;
   ctx->Const.MaxImageSamples = 0;

   ctx->Const.MinLineWidth = 1.0;
   ctx->Const.MinLineWidthAA = 1.0;
   if (devinfo->gen >= 6) {
      ctx->Const.MaxLineWidth = 7.375;
      ctx->Const.MaxLineWidthAA = 7.375;
      ctx->Const.LineWidthGranularity = 0.125;
   } else {
      ctx->Const.MaxLineWidth = 7.0;
      ctx->Const.MaxLineWidthAA = 7.0;
      ctx->Const.LineWidthGranularity = 0.5;
   }

   /* For non-antialiased lines, we have to round the line width to the
    * nearest whole number. Make sure that we don't advertise a line
    * width that, when rounded, will be beyond the actual hardware
    * maximum.
    */
   assert(roundf(ctx->Const.MaxLineWidth) <= ctx->Const.MaxLineWidth);

   ctx->Const.MinPointSize = 1.0;
   ctx->Const.MinPointSizeAA = 1.0;
   ctx->Const.MaxPointSize = 255.0;
   ctx->Const.MaxPointSizeAA = 255.0;
   ctx->Const.PointSizeGranularity = 1.0;

   if (devinfo->gen >= 5 || devinfo->is_g4x)
      ctx->Const.MaxClipPlanes = 8;

   ctx->Const.GLSLFragCoordIsSysVal = true;
   ctx->Const.GLSLFrontFacingIsSysVal = true;
   ctx->Const.GLSLTessLevelsAsInputs = true;
   ctx->Const.PrimitiveRestartForPatches = true;

   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeInstructions = 16 * 1024;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxAluInstructions = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxTexInstructions = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxTexIndirections = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeAluInstructions = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeTexInstructions = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeTexIndirections = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeAttribs = 16;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeTemps = 256;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeAddressRegs = 1;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeParameters = 1024;
   ctx->Const.Program[MESA_SHADER_VERTEX].MaxEnvParams =
      MIN2(ctx->Const.Program[MESA_SHADER_VERTEX].MaxNativeParameters,
	   ctx->Const.Program[MESA_SHADER_VERTEX].MaxEnvParams);

   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeInstructions = 1024;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeAluInstructions = 1024;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeTexInstructions = 1024;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeTexIndirections = 1024;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeAttribs = 12;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeTemps = 256;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeAddressRegs = 0;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeParameters = 1024;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxEnvParams =
      MIN2(ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxNativeParameters,
	   ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxEnvParams);

   /* Fragment shaders use real, 32-bit twos-complement integers for all
    * integer types.
    */
   ctx->Const.Program[MESA_SHADER_FRAGMENT].LowInt.RangeMin = 31;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].LowInt.RangeMax = 30;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].LowInt.Precision = 0;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].HighInt = ctx->Const.Program[MESA_SHADER_FRAGMENT].LowInt;
   ctx->Const.Program[MESA_SHADER_FRAGMENT].MediumInt = ctx->Const.Program[MESA_SHADER_FRAGMENT].LowInt;

   ctx->Const.Program[MESA_SHADER_VERTEX].LowInt.RangeMin = 31;
   ctx->Const.Program[MESA_SHADER_VERTEX].LowInt.RangeMax = 30;
   ctx->Const.Program[MESA_SHADER_VERTEX].LowInt.Precision = 0;
   ctx->Const.Program[MESA_SHADER_VERTEX].HighInt = ctx->Const.Program[MESA_SHADER_VERTEX].LowInt;
   ctx->Const.Program[MESA_SHADER_VERTEX].MediumInt = ctx->Const.Program[MESA_SHADER_VERTEX].LowInt;

   /* Gen6 converts quads to polygon in beginning of 3D pipeline,
    * but we're not sure how it's actually done for vertex order,
    * that affect provoking vertex decision. Always use last vertex
    * convention for quad primitive which works as expected for now.
    */
   if (devinfo->gen >= 6)
      ctx->Const.QuadsFollowProvokingVertexConvention = false;

   ctx->Const.NativeIntegers = true;

   /* Regarding the CMP instruction, the Ivybridge PRM says:
    *
    *   "For each enabled channel 0b or 1b is assigned to the appropriate flag
    *    bit and 0/all zeros or all ones (e.g, byte 0xFF, word 0xFFFF, DWord
    *    0xFFFFFFFF) is assigned to dst."
    *
    * but PRMs for earlier generations say
    *
    *   "In dword format, one GRF may store up to 8 results. When the register
    *    is used later as a vector of Booleans, as only LSB at each channel
    *    contains meaning [sic] data, software should make sure all higher bits
    *    are masked out (e.g. by 'and-ing' an [sic] 0x01 constant)."
    *
    * We select the representation of a true boolean uniform to be ~0, and fix
    * the results of Gen <= 5 CMP instruction's with -(result & 1).
    */
   ctx->Const.UniformBooleanTrue = ~0;

   /* From the gen4 PRM, volume 4 page 127:
    *
    *     "For SURFTYPE_BUFFER non-rendertarget surfaces, this field specifies
    *      the base address of the first element of the surface, computed in
    *      software by adding the surface base address to the byte offset of
    *      the element in the buffer."
    *
    * However, unaligned accesses are slower, so enforce buffer alignment.
    *
    * In order to push UBO data, 3DSTATE_CONSTANT_XS imposes an additional
    * restriction: the start of the buffer needs to be 32B aligned.
    */
   ctx->Const.UniformBufferOffsetAlignment = 32;

   /* ShaderStorageBufferOffsetAlignment should be a cacheline (64 bytes) so
    * that we can safely have the CPU and GPU writing the same SSBO on
    * non-cachecoherent systems (our Atom CPUs). With UBOs, the GPU never
    * writes, so there's no problem. For an SSBO, the GPU and the CPU can
    * be updating disjoint regions of the buffer simultaneously and that will
    * break if the regions overlap the same cacheline.
    */
   ctx->Const.ShaderStorageBufferOffsetAlignment = 64;
   ctx->Const.TextureBufferOffsetAlignment = 16;
   ctx->Const.MaxTextureBufferSize = 128 * 1024 * 1024;

   if (devinfo->gen >= 6) {
      ctx->Const.MaxVarying = 32;
      ctx->Const.Program[MESA_SHADER_VERTEX].MaxOutputComponents = 128;
      ctx->Const.Program[MESA_SHADER_GEOMETRY].MaxInputComponents =
         compiler->scalar_stage[MESA_SHADER_GEOMETRY] ? 128 : 64;
      ctx->Const.Program[MESA_SHADER_GEOMETRY].MaxOutputComponents = 128;
      ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxInputComponents = 128;
      ctx->Const.Program[MESA_SHADER_TESS_CTRL].MaxInputComponents = 128;
      ctx->Const.Program[MESA_SHADER_TESS_CTRL].MaxOutputComponents = 128;
      ctx->Const.Program[MESA_SHADER_TESS_EVAL].MaxInputComponents = 128;
      ctx->Const.Program[MESA_SHADER_TESS_EVAL].MaxOutputComponents = 128;
   }

   /* We want the GLSL compiler to emit code that uses condition codes */
   for (int i = 0; i < MESA_SHADER_STAGES; i++) {
      ctx->Const.ShaderCompilerOptions[i] =
         brw->screen->compiler->glsl_compiler_options[i];
   }

   if (devinfo->gen >= 7) {
      ctx->Const.MaxViewportWidth = 32768;
      ctx->Const.MaxViewportHeight = 32768;
   }

   /* ARB_viewport_array, OES_viewport_array */
   if (devinfo->gen >= 6) {
      ctx->Const.MaxViewports = GEN6_NUM_VIEWPORTS;
      ctx->Const.ViewportSubpixelBits = 8;

      /* Cast to float before negating because MaxViewportWidth is unsigned.
       */
      ctx->Const.ViewportBounds.Min = -(float)ctx->Const.MaxViewportWidth;
      ctx->Const.ViewportBounds.Max = ctx->Const.MaxViewportWidth;
   }

   /* ARB_gpu_shader5 */
   if (devinfo->gen >= 7)
      ctx->Const.MaxVertexStreams = MIN2(4, MAX_VERTEX_STREAMS);

   /* ARB_framebuffer_no_attachments */
   ctx->Const.MaxFramebufferWidth = 16384;
   ctx->Const.MaxFramebufferHeight = 16384;
   ctx->Const.MaxFramebufferLayers = ctx->Const.MaxArrayTextureLayers;
   ctx->Const.MaxFramebufferSamples = max_samples;

   /* OES_primitive_bounding_box */
   ctx->Const.NoPrimitiveBoundingBoxOutput = true;

   /* TODO: We should be able to use STD430 packing by default on all hardware
    * but some piglit tests [1] currently fail on SNB when this is enabled.
    * The problem is the messages we're using for doing uniform pulls
    * in the vec4 back-end on SNB is the OWORD block load instruction, which
    * takes its offset in units of OWORDS (16 bytes).  On IVB+, we use the
    * sampler which doesn't have these restrictions.
    *
    * In the scalar back-end, we use the sampler for dynamic uniform loads and
    * pull an entire cache line at a time for constant offset loads both of
    * which support almost any alignment.
    *
    * [1] glsl-1.40/uniform_buffer/vs-float-array-variable-index.shader_test
    */
   if (devinfo->gen >= 7)
      ctx->Const.UseSTD430AsDefaultPacking = true;

   if (!(ctx->Const.ContextFlags & GL_CONTEXT_FLAG_DEBUG_BIT))
      ctx->Const.AllowMappedBuffersDuringExecution = true;

   /* GL_ARB_get_program_binary */
   ctx->Const.NumProgramBinaryFormats = 1;
}

static void
brw_initialize_cs_context_constants(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   const struct intel_screen *screen = brw->screen;
   struct gen_device_info *devinfo = &brw->screen->devinfo;

   /* FINISHME: Do this for all platforms that the kernel supports */
   if (devinfo->is_cherryview &&
       screen->subslice_total > 0 && screen->eu_total > 0) {
      /* Logical CS threads = EUs per subslice * 7 threads per EU */
      uint32_t max_cs_threads = screen->eu_total / screen->subslice_total * 7;

      /* Fuse configurations may give more threads than expected, never less. */
      if (max_cs_threads > devinfo->max_cs_threads)
         devinfo->max_cs_threads = max_cs_threads;
   }

   /* Maximum number of scalar compute shader invocations that can be run in
    * parallel in the same subslice assuming SIMD32 dispatch.
    *
    * We don't advertise more than 64 threads, because we are limited to 64 by
    * our usage of thread_width_max in the gpgpu walker command. This only
    * currently impacts Haswell, which otherwise might be able to advertise 70
    * threads. With SIMD32 and 64 threads, Haswell still provides twice the
    * required the number of invocation needed for ARB_compute_shader.
    */
   const unsigned max_threads = MIN2(64, devinfo->max_cs_threads);
   const uint32_t max_invocations = 32 * max_threads;
   ctx->Const.MaxComputeWorkGroupSize[0] = max_invocations;
   ctx->Const.MaxComputeWorkGroupSize[1] = max_invocations;
   ctx->Const.MaxComputeWorkGroupSize[2] = max_invocations;
   ctx->Const.MaxComputeWorkGroupInvocations = max_invocations;
   ctx->Const.MaxComputeSharedMemorySize = 64 * 1024;

   /* Constants used for ARB_compute_variable_group_size. */
   if (devinfo->gen >= 7) {
      assert(max_invocations >= 512);
      ctx->Const.MaxComputeVariableGroupSize[0] = max_invocations;
      ctx->Const.MaxComputeVariableGroupSize[1] = max_invocations;
      ctx->Const.MaxComputeVariableGroupSize[2] = max_invocations;
      ctx->Const.MaxComputeVariableGroupInvocations = max_invocations;
   }
}

/**
 * Process driconf (drirc) options, setting appropriate context flags.
 *
 * intelInitExtensions still pokes at optionCache directly, in order to
 * avoid advertising various extensions.  No flags are set, so it makes
 * sense to continue doing that there.
 */
static void
brw_process_driconf_options(struct brw_context *brw)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   struct gl_context *ctx = &brw->ctx;

   driOptionCache *options = &brw->optionCache;
   driParseConfigFiles(options, &brw->screen->optionCache,
                       brw->driContext->driScreenPriv->myNum,
                       "i965", NULL, NULL, 0);

   if (INTEL_DEBUG & DEBUG_NO_HIZ) {
       brw->has_hiz = false;
       /* On gen6, you can only do separate stencil with HIZ. */
       if (devinfo->gen == 6)
          brw->has_separate_stencil = false;
   }

   if (driQueryOptionb(options, "mesa_no_error"))
      ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_NO_ERROR_BIT_KHR;

   if (driQueryOptionb(options, "always_flush_batch")) {
      fprintf(stderr, "flushing batchbuffer before/after each draw call\n");
      brw->always_flush_batch = true;
   }

   if (driQueryOptionb(options, "always_flush_cache")) {
      fprintf(stderr, "flushing GPU caches before/after each draw call\n");
      brw->always_flush_cache = true;
   }

   if (driQueryOptionb(options, "disable_throttling")) {
      fprintf(stderr, "disabling flush throttling\n");
      brw->disable_throttling = true;
   }

   brw->precompile = driQueryOptionb(&brw->optionCache, "shader_precompile");

   if (driQueryOptionb(&brw->optionCache, "precise_trig"))
      brw->screen->compiler->precise_trig = true;

   ctx->Const.ForceGLSLExtensionsWarn =
      driQueryOptionb(options, "force_glsl_extensions_warn");

   ctx->Const.ForceGLSLVersion =
      driQueryOptioni(options, "force_glsl_version");

   ctx->Const.DisableGLSLLineContinuations =
      driQueryOptionb(options, "disable_glsl_line_continuations");

   ctx->Const.AllowGLSLExtensionDirectiveMidShader =
      driQueryOptionb(options, "allow_glsl_extension_directive_midshader");

   ctx->Const.AllowGLSLBuiltinVariableRedeclaration =
      driQueryOptionb(options, "allow_glsl_builtin_variable_redeclaration");

   ctx->Const.AllowHigherCompatVersion =
      driQueryOptionb(options, "allow_higher_compat_version");

   ctx->Const.ForceGLSLAbsSqrt =
      driQueryOptionb(options, "force_glsl_abs_sqrt");

   ctx->Const.GLSLZeroInit = driQueryOptionb(options, "glsl_zero_init") ? 1 : 0;

   brw->dual_color_blend_by_location =
      driQueryOptionb(options, "dual_color_blend_by_location");

   ctx->Const.AllowGLSLCrossStageInterpolationMismatch =
      driQueryOptionb(options, "allow_glsl_cross_stage_interpolation_mismatch");

   ctx->Const.dri_config_options_sha1 = ralloc_array(brw, unsigned char, 20);
   driComputeOptionsSha1(&brw->screen->optionCache,
                         ctx->Const.dri_config_options_sha1);
}

GLboolean
brwCreateContext(gl_api api,
                 const struct gl_config *mesaVis,
                 __DRIcontext *driContextPriv,
                 const struct __DriverContextConfig *ctx_config,
                 unsigned *dri_ctx_error,
                 void *sharedContextPrivate)
{
   struct gl_context *shareCtx = (struct gl_context *) sharedContextPrivate;
   struct intel_screen *screen = driContextPriv->driScreenPriv->driverPrivate;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct dd_function_table functions;

   /* Only allow the __DRI_CTX_FLAG_ROBUST_BUFFER_ACCESS flag if the kernel
    * provides us with context reset notifications.
    */
   uint32_t allowed_flags = __DRI_CTX_FLAG_DEBUG |
                            __DRI_CTX_FLAG_FORWARD_COMPATIBLE |
                            __DRI_CTX_FLAG_NO_ERROR;

   if (screen->has_context_reset_notification)
      allowed_flags |= __DRI_CTX_FLAG_ROBUST_BUFFER_ACCESS;

   if (ctx_config->flags & ~allowed_flags) {
      *dri_ctx_error = __DRI_CTX_ERROR_UNKNOWN_FLAG;
      return false;
   }

   if (ctx_config->attribute_mask &
       ~(__DRIVER_CONTEXT_ATTRIB_RESET_STRATEGY |
         __DRIVER_CONTEXT_ATTRIB_PRIORITY)) {
      *dri_ctx_error = __DRI_CTX_ERROR_UNKNOWN_ATTRIBUTE;
      return false;
   }

   bool notify_reset =
      ((ctx_config->attribute_mask & __DRIVER_CONTEXT_ATTRIB_RESET_STRATEGY) &&
       ctx_config->reset_strategy != __DRI_CTX_RESET_NO_NOTIFICATION);

   struct brw_context *brw = rzalloc(NULL, struct brw_context);
   if (!brw) {
      fprintf(stderr, "%s: failed to alloc context\n", __func__);
      *dri_ctx_error = __DRI_CTX_ERROR_NO_MEMORY;
      return false;
   }
   brw->perf_ctx = gen_perf_new_context(brw);

   driContextPriv->driverPrivate = brw;
   brw->driContext = driContextPriv;
   brw->screen = screen;
   brw->bufmgr = screen->bufmgr;

   brw->has_hiz = devinfo->has_hiz_and_separate_stencil;
   brw->has_separate_stencil = devinfo->has_hiz_and_separate_stencil;

   brw->has_swizzling = screen->hw_has_swizzling;

   brw->isl_dev = screen->isl_dev;

   brw->vs.base.stage = MESA_SHADER_VERTEX;
   brw->tcs.base.stage = MESA_SHADER_TESS_CTRL;
   brw->tes.base.stage = MESA_SHADER_TESS_EVAL;
   brw->gs.base.stage = MESA_SHADER_GEOMETRY;
   brw->wm.base.stage = MESA_SHADER_FRAGMENT;
   brw->cs.base.stage = MESA_SHADER_COMPUTE;

   brw_init_driver_functions(brw, &functions);

   if (notify_reset)
      functions.GetGraphicsResetStatus = brw_get_graphics_reset_status;

   brw_process_driconf_options(brw);

   if (api == API_OPENGL_CORE &&
       driQueryOptionb(&screen->optionCache, "force_compat_profile")) {
      api = API_OPENGL_COMPAT;
   }

   struct gl_context *ctx = &brw->ctx;

   if (!_mesa_initialize_context(ctx, api, mesaVis, shareCtx, &functions)) {
      *dri_ctx_error = __DRI_CTX_ERROR_NO_MEMORY;
      fprintf(stderr, "%s: failed to init mesa context\n", __func__);
      intelDestroyContext(driContextPriv);
      return false;
   }

   driContextSetFlags(ctx, ctx_config->flags);

   /* Initialize the software rasterizer and helper modules.
    *
    * As of GL 3.1 core, the gen4+ driver doesn't need the swrast context for
    * software fallbacks (which we have to support on legacy GL to do weird
    * glDrawPixels(), glBitmap(), and other functions).
    */
   if (api != API_OPENGL_CORE && api != API_OPENGLES2) {
      _swrast_CreateContext(ctx);
   }

   _vbo_CreateContext(ctx, true);
   if (ctx->swrast_context) {
      _tnl_CreateContext(ctx);
      TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
      _swsetup_CreateContext(ctx);

      /* Configure swrast to match hardware characteristics: */
      _swrast_allow_pixel_fog(ctx, false);
      _swrast_allow_vertex_fog(ctx, true);
   }

   _mesa_meta_init(ctx);

   if (INTEL_DEBUG & DEBUG_PERF)
      brw->perf_debug = true;

   brw_initialize_cs_context_constants(brw);
   brw_initialize_context_constants(brw);

   ctx->Const.ResetStrategy = notify_reset
      ? GL_LOSE_CONTEXT_ON_RESET_ARB : GL_NO_RESET_NOTIFICATION_ARB;

   /* Reinitialize the context point state.  It depends on ctx->Const values. */
   _mesa_init_point(ctx);

   intel_fbo_init(brw);

   intel_batchbuffer_init(brw);

   /* Create a new hardware context.  Using a hardware context means that
    * our GPU state will be saved/restored on context switch, allowing us
    * to assume that the GPU is in the same state we left it in.
    *
    * This is required for transform feedback buffer offsets, query objects,
    * and also allows us to reduce how much state we have to emit.
    */
   brw->hw_ctx = brw_create_hw_context(brw->bufmgr);
   if (!brw->hw_ctx && devinfo->gen >= 6) {
      fprintf(stderr, "Failed to create hardware context.\n");
      intelDestroyContext(driContextPriv);
      return false;
   }

   if (brw->hw_ctx) {
      int hw_priority = GEN_CONTEXT_MEDIUM_PRIORITY;
      if (ctx_config->attribute_mask & __DRIVER_CONTEXT_ATTRIB_PRIORITY) {
         switch (ctx_config->priority) {
         case __DRI_CTX_PRIORITY_LOW:
            hw_priority = GEN_CONTEXT_LOW_PRIORITY;
            break;
         case __DRI_CTX_PRIORITY_HIGH:
            hw_priority = GEN_CONTEXT_HIGH_PRIORITY;
            break;
         }
      }
      if (hw_priority != I915_CONTEXT_DEFAULT_PRIORITY &&
          brw_hw_context_set_priority(brw->bufmgr, brw->hw_ctx, hw_priority)) {
         fprintf(stderr,
		 "Failed to set priority [%d:%d] for hardware context.\n",
                 ctx_config->priority, hw_priority);
         intelDestroyContext(driContextPriv);
         return false;
      }
   }

   if (brw_init_pipe_control(brw, devinfo)) {
      *dri_ctx_error = __DRI_CTX_ERROR_NO_MEMORY;
      intelDestroyContext(driContextPriv);
      return false;
   }

   brw_upload_init(&brw->upload, brw->bufmgr, 65536);

   brw_init_state(brw);

   intelInitExtensions(ctx);

   brw_init_surface_formats(brw);

   brw_blorp_init(brw);

   brw->urb.size = devinfo->urb.size;

   if (devinfo->gen == 6)
      brw->urb.gs_present = false;

   brw->prim_restart.in_progress = false;
   brw->prim_restart.enable_cut_index = false;
   brw->gs.enabled = false;
   brw->clip.viewport_count = 1;

   brw->predicate.state = BRW_PREDICATE_STATE_RENDER;

   brw->max_gtt_map_object_size = screen->max_gtt_map_object_size;

   ctx->VertexProgram._MaintainTnlProgram = true;
   ctx->FragmentProgram._MaintainTexEnvProgram = true;

   brw_draw_init( brw );

   if ((ctx_config->flags & __DRI_CTX_FLAG_DEBUG) != 0) {
      /* Turn on some extra GL_ARB_debug_output generation. */
      brw->perf_debug = true;
   }

   if ((ctx_config->flags & __DRI_CTX_FLAG_ROBUST_BUFFER_ACCESS) != 0) {
      ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_ROBUST_ACCESS_BIT_ARB;
      ctx->Const.RobustAccess = GL_TRUE;
   }

   if (INTEL_DEBUG & DEBUG_SHADER_TIME)
      brw_init_shader_time(brw);

   _mesa_override_extensions(ctx);
   _mesa_compute_version(ctx);

   /* GL_ARB_gl_spirv */
   if (ctx->Extensions.ARB_gl_spirv) {
      brw_initialize_spirv_supported_capabilities(brw);

      if (ctx->Extensions.ARB_spirv_extensions) {
         /* GL_ARB_spirv_extensions */
         ctx->Const.SpirVExtensions = MALLOC_STRUCT(spirv_supported_extensions);
         _mesa_fill_supported_spirv_extensions(ctx->Const.SpirVExtensions,
                                               &ctx->Const.SpirVCapabilities);
      }
   }

   _mesa_initialize_dispatch_tables(ctx);
   _mesa_initialize_vbo_vtxfmt(ctx);

   if (ctx->Extensions.INTEL_performance_query)
      brw_init_performance_queries(brw);

   brw->ctx.Cache = brw->screen->disk_cache;

   if (driContextPriv->driScreenPriv->dri2.backgroundCallable &&
       driQueryOptionb(&screen->optionCache, "mesa_glthread")) {
      /* Loader supports multithreading, and so do we. */
      _mesa_glthread_init(ctx);
   }

   return true;
}

void
intelDestroyContext(__DRIcontext * driContextPriv)
{
   struct brw_context *brw =
      (struct brw_context *) driContextPriv->driverPrivate;
   struct gl_context *ctx = &brw->ctx;

   GET_CURRENT_CONTEXT(curctx);

   if (curctx == NULL) {
      /* No current context, but we need one to release
       * renderbuffer surface when we release framebuffer.
       * So temporarily bind the context.
       */
      _mesa_make_current(ctx, NULL, NULL);
   }

   _mesa_glthread_destroy(&brw->ctx);

   _mesa_meta_free(&brw->ctx);

   if (INTEL_DEBUG & DEBUG_SHADER_TIME) {
      /* Force a report. */
      brw->shader_time.report_time = 0;

      brw_collect_and_report_shader_time(brw);
      brw_destroy_shader_time(brw);
   }

   blorp_finish(&brw->blorp);

   brw_destroy_state(brw);
   brw_draw_destroy(brw);

   brw_bo_unreference(brw->curbe.curbe_bo);

   brw_bo_unreference(brw->vs.base.scratch_bo);
   brw_bo_unreference(brw->tcs.base.scratch_bo);
   brw_bo_unreference(brw->tes.base.scratch_bo);
   brw_bo_unreference(brw->gs.base.scratch_bo);
   brw_bo_unreference(brw->wm.base.scratch_bo);

   brw_bo_unreference(brw->vs.base.push_const_bo);
   brw_bo_unreference(brw->tcs.base.push_const_bo);
   brw_bo_unreference(brw->tes.base.push_const_bo);
   brw_bo_unreference(brw->gs.base.push_const_bo);
   brw_bo_unreference(brw->wm.base.push_const_bo);

   brw_destroy_hw_context(brw->bufmgr, brw->hw_ctx);

   if (ctx->swrast_context) {
      _swsetup_DestroyContext(&brw->ctx);
      _tnl_DestroyContext(&brw->ctx);
   }
   _vbo_DestroyContext(&brw->ctx);

   if (ctx->swrast_context)
      _swrast_DestroyContext(&brw->ctx);

   brw_fini_pipe_control(brw);
   intel_batchbuffer_free(&brw->batch);

   brw_bo_unreference(brw->throttle_batch[1]);
   brw_bo_unreference(brw->throttle_batch[0]);
   brw->throttle_batch[1] = NULL;
   brw->throttle_batch[0] = NULL;

   driDestroyOptionCache(&brw->optionCache);

   /* free the Mesa context */
   _mesa_free_context_data(&brw->ctx);

   ralloc_free(brw);
   driContextPriv->driverPrivate = NULL;
}

GLboolean
intelUnbindContext(__DRIcontext * driContextPriv)
{
   struct gl_context *ctx = driContextPriv->driverPrivate;
   _mesa_glthread_finish(ctx);

   /* Unset current context and dispath table */
   _mesa_make_current(NULL, NULL, NULL);

   return true;
}

/**
 * Fixes up the context for GLES23 with our default-to-sRGB-capable behavior
 * on window system framebuffers.
 *
 * Desktop GL is fairly reasonable in its handling of sRGB: You can ask if
 * your renderbuffer can do sRGB encode, and you can flip a switch that does
 * sRGB encode if the renderbuffer can handle it.  You can ask specifically
 * for a visual where you're guaranteed to be capable, but it turns out that
 * everyone just makes all their ARGB8888 visuals capable and doesn't offer
 * incapable ones, because there's no difference between the two in resources
 * used.  Applications thus get built that accidentally rely on the default
 * visual choice being sRGB, so we make ours sRGB capable.  Everything sounds
 * great...
 *
 * But for GLES2/3, they decided that it was silly to not turn on sRGB encode
 * for sRGB renderbuffers you made with the GL_EXT_texture_sRGB equivalent.
 * So they removed the enable knob and made it "if the renderbuffer is sRGB
 * capable, do sRGB encode".  Then, for your window system renderbuffers, you
 * can ask for sRGB visuals and get sRGB encode, or not ask for sRGB visuals
 * and get no sRGB encode (assuming that both kinds of visual are available).
 * Thus our choice to support sRGB by default on our visuals for desktop would
 * result in broken rendering of GLES apps that aren't expecting sRGB encode.
 *
 * Unfortunately, renderbuffer setup happens before a context is created.  So
 * in intel_screen.c we always set up sRGB, and here, if you're a GLES2/3
 * context (without an sRGB visual), we go turn that back off before anyone
 * finds out.
 */
static void
intel_gles3_srgb_workaround(struct brw_context *brw,
                            struct gl_framebuffer *fb)
{
   struct gl_context *ctx = &brw->ctx;

   if (_mesa_is_desktop_gl(ctx) || !fb->Visual.sRGBCapable)
      return;

   for (int i = 0; i < BUFFER_COUNT; i++) {
      struct gl_renderbuffer *rb = fb->Attachment[i].Renderbuffer;

      /* Check if sRGB was specifically asked for. */
      struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, i);
      if (irb && irb->need_srgb)
         return;

      if (rb)
         rb->Format = _mesa_get_srgb_format_linear(rb->Format);
   }
   /* Disable sRGB from framebuffers that are not compatible. */
   fb->Visual.sRGBCapable = false;
}

GLboolean
intelMakeCurrent(__DRIcontext * driContextPriv,
                 __DRIdrawable * driDrawPriv,
                 __DRIdrawable * driReadPriv)
{
   struct brw_context *brw;

   if (driContextPriv)
      brw = (struct brw_context *) driContextPriv->driverPrivate;
   else
      brw = NULL;

   if (driContextPriv) {
      struct gl_context *ctx = &brw->ctx;
      struct gl_framebuffer *fb, *readFb;

      if (driDrawPriv == NULL) {
         fb = _mesa_get_incomplete_framebuffer();
      } else {
         fb = driDrawPriv->driverPrivate;
         driContextPriv->dri2.draw_stamp = driDrawPriv->dri2.stamp - 1;
      }

      if (driReadPriv == NULL) {
         readFb = _mesa_get_incomplete_framebuffer();
      } else {
         readFb = driReadPriv->driverPrivate;
         driContextPriv->dri2.read_stamp = driReadPriv->dri2.stamp - 1;
      }

      /* The sRGB workaround changes the renderbuffer's format. We must change
       * the format before the renderbuffer's miptree get's allocated, otherwise
       * the formats of the renderbuffer and its miptree will differ.
       */
      intel_gles3_srgb_workaround(brw, fb);
      intel_gles3_srgb_workaround(brw, readFb);

      /* If the context viewport hasn't been initialized, force a call out to
       * the loader to get buffers so we have a drawable size for the initial
       * viewport. */
      if (!brw->ctx.ViewportInitialized)
         intel_prepare_render(brw);

      _mesa_make_current(ctx, fb, readFb);
   } else {
      GET_CURRENT_CONTEXT(ctx);
      _mesa_glthread_finish(ctx);
      _mesa_make_current(NULL, NULL, NULL);
   }

   return true;
}

void
intel_resolve_for_dri2_flush(struct brw_context *brw,
                             __DRIdrawable *drawable)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;

   if (devinfo->gen < 6) {
      /* MSAA and fast color clear are not supported, so don't waste time
       * checking whether a resolve is needed.
       */
      return;
   }

   struct gl_framebuffer *fb = drawable->driverPrivate;
   struct intel_renderbuffer *rb;

   /* Usually, only the back buffer will need to be downsampled. However,
    * the front buffer will also need it if the user has rendered into it.
    */
   static const gl_buffer_index buffers[2] = {
         BUFFER_BACK_LEFT,
         BUFFER_FRONT_LEFT,
   };

   for (int i = 0; i < 2; ++i) {
      rb = intel_get_renderbuffer(fb, buffers[i]);
      if (rb == NULL || rb->mt == NULL)
         continue;
      if (rb->mt->surf.samples == 1) {
         assert(rb->mt_layer == 0 && rb->mt_level == 0 &&
                rb->layer_count == 1);
         intel_miptree_prepare_external(brw, rb->mt);
      } else {
         intel_renderbuffer_downsample(brw, rb);

         /* Call prepare_external on the single-sample miptree to do any
          * needed resolves prior to handing it off to the window system.
          * This is needed in the case that rb->singlesample_mt is Y-tiled
          * with CCS_E enabled but without I915_FORMAT_MOD_Y_TILED_CCS_E.  In
          * this case, the MSAA resolve above will write compressed data into
          * rb->singlesample_mt.
          *
          * TODO: Some day, if we decide to care about the tiny performance
          * hit we're taking by doing the MSAA resolve and then a CCS resolve,
          * we could detect this case and just allocate the single-sampled
          * miptree without aux.  However, that would be a lot of plumbing and
          * this is a rather exotic case so it's not really worth it.
          */
         intel_miptree_prepare_external(brw, rb->singlesample_mt);
      }
   }
}

static unsigned
intel_bits_per_pixel(const struct intel_renderbuffer *rb)
{
   return _mesa_get_format_bytes(intel_rb_format(rb)) * 8;
}

static void
intel_query_dri2_buffers(struct brw_context *brw,
                         __DRIdrawable *drawable,
                         __DRIbuffer **buffers,
                         int *count);

static void
intel_process_dri2_buffer(struct brw_context *brw,
                          __DRIdrawable *drawable,
                          __DRIbuffer *buffer,
                          struct intel_renderbuffer *rb,
                          const char *buffer_name);

static void
intel_update_image_buffers(struct brw_context *brw, __DRIdrawable *drawable);

static void
intel_update_dri2_buffers(struct brw_context *brw, __DRIdrawable *drawable)
{
   struct gl_framebuffer *fb = drawable->driverPrivate;
   struct intel_renderbuffer *rb;
   __DRIbuffer *buffers = NULL;
   int count;
   const char *region_name;

   /* Set this up front, so that in case our buffers get invalidated
    * while we're getting new buffers, we don't clobber the stamp and
    * thus ignore the invalidate. */
   drawable->lastStamp = drawable->dri2.stamp;

   if (unlikely(INTEL_DEBUG & DEBUG_DRI))
      fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);

   intel_query_dri2_buffers(brw, drawable, &buffers, &count);

   if (buffers == NULL)
      return;

   for (int i = 0; i < count; i++) {
       switch (buffers[i].attachment) {
       case __DRI_BUFFER_FRONT_LEFT:
           rb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT);
           region_name = "dri2 front buffer";
           break;

       case __DRI_BUFFER_FAKE_FRONT_LEFT:
           rb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT);
           region_name = "dri2 fake front buffer";
           break;

       case __DRI_BUFFER_BACK_LEFT:
           rb = intel_get_renderbuffer(fb, BUFFER_BACK_LEFT);
           region_name = "dri2 back buffer";
           break;

       case __DRI_BUFFER_DEPTH:
       case __DRI_BUFFER_HIZ:
       case __DRI_BUFFER_DEPTH_STENCIL:
       case __DRI_BUFFER_STENCIL:
       case __DRI_BUFFER_ACCUM:
       default:
           fprintf(stderr,
                   "unhandled buffer attach event, attachment type %d\n",
                   buffers[i].attachment);
           return;
       }

       intel_process_dri2_buffer(brw, drawable, &buffers[i], rb, region_name);
   }

}

void
intel_update_renderbuffers(__DRIcontext *context, __DRIdrawable *drawable)
{
   struct brw_context *brw = context->driverPrivate;
   __DRIscreen *dri_screen = brw->screen->driScrnPriv;

   /* Set this up front, so that in case our buffers get invalidated
    * while we're getting new buffers, we don't clobber the stamp and
    * thus ignore the invalidate. */
   drawable->lastStamp = drawable->dri2.stamp;

   if (unlikely(INTEL_DEBUG & DEBUG_DRI))
      fprintf(stderr, "enter %s, drawable %p\n", __func__, drawable);

   if (dri_screen->image.loader)
      intel_update_image_buffers(brw, drawable);
   else
      intel_update_dri2_buffers(brw, drawable);

   driUpdateFramebufferSize(&brw->ctx, drawable);
}

/**
 * intel_prepare_render should be called anywhere that curent read/drawbuffer
 * state is required.
 */
void
intel_prepare_render(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   __DRIcontext *driContext = brw->driContext;
   __DRIdrawable *drawable;

   drawable = driContext->driDrawablePriv;
   if (drawable && drawable->dri2.stamp != driContext->dri2.draw_stamp) {
      if (drawable->lastStamp != drawable->dri2.stamp)
         intel_update_renderbuffers(driContext, drawable);
      driContext->dri2.draw_stamp = drawable->dri2.stamp;
   }

   drawable = driContext->driReadablePriv;
   if (drawable && drawable->dri2.stamp != driContext->dri2.read_stamp) {
      if (drawable->lastStamp != drawable->dri2.stamp)
         intel_update_renderbuffers(driContext, drawable);
      driContext->dri2.read_stamp = drawable->dri2.stamp;
   }

   /* If we're currently rendering to the front buffer, the rendering
    * that will happen next will probably dirty the front buffer.  So
    * mark it as dirty here.
    */
   if (_mesa_is_front_buffer_drawing(ctx->DrawBuffer) &&
       ctx->DrawBuffer != _mesa_get_incomplete_framebuffer()) {
      brw->front_buffer_dirty = true;
   }

   if (brw->is_shared_buffer_bound) {
      /* Subsequent rendering will probably dirty the shared buffer. */
      brw->is_shared_buffer_dirty = true;
   }
}

/**
 * \brief Query DRI2 to obtain a DRIdrawable's buffers.
 *
 * To determine which DRI buffers to request, examine the renderbuffers
 * attached to the drawable's framebuffer. Then request the buffers with
 * DRI2GetBuffers() or DRI2GetBuffersWithFormat().
 *
 * This is called from intel_update_renderbuffers().
 *
 * \param drawable      Drawable whose buffers are queried.
 * \param buffers       [out] List of buffers returned by DRI2 query.
 * \param buffer_count  [out] Number of buffers returned.
 *
 * \see intel_update_renderbuffers()
 * \see DRI2GetBuffers()
 * \see DRI2GetBuffersWithFormat()
 */
static void
intel_query_dri2_buffers(struct brw_context *brw,
                         __DRIdrawable *drawable,
                         __DRIbuffer **buffers,
                         int *buffer_count)
{
   __DRIscreen *dri_screen = brw->screen->driScrnPriv;
   struct gl_framebuffer *fb = drawable->driverPrivate;
   int i = 0;
   unsigned attachments[8];

   struct intel_renderbuffer *front_rb;
   struct intel_renderbuffer *back_rb;

   front_rb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT);
   back_rb = intel_get_renderbuffer(fb, BUFFER_BACK_LEFT);

   memset(attachments, 0, sizeof(attachments));
   if ((_mesa_is_front_buffer_drawing(fb) ||
        _mesa_is_front_buffer_reading(fb) ||
        !back_rb) && front_rb) {
      /* If a fake front buffer is in use, then querying for
       * __DRI_BUFFER_FRONT_LEFT will cause the server to copy the image from
       * the real front buffer to the fake front buffer.  So before doing the
       * query, we need to make sure all the pending drawing has landed in the
       * real front buffer.
       */
      intel_batchbuffer_flush(brw);
      intel_flush_front(&brw->ctx);

      attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
      attachments[i++] = intel_bits_per_pixel(front_rb);
   } else if (front_rb && brw->front_buffer_dirty) {
      /* We have pending front buffer rendering, but we aren't querying for a
       * front buffer.  If the front buffer we have is a fake front buffer,
       * the X server is going to throw it away when it processes the query.
       * So before doing the query, make sure all the pending drawing has
       * landed in the real front buffer.
       */
      intel_batchbuffer_flush(brw);
      intel_flush_front(&brw->ctx);
   }

   if (back_rb) {
      attachments[i++] = __DRI_BUFFER_BACK_LEFT;
      attachments[i++] = intel_bits_per_pixel(back_rb);
   }

   assert(i <= ARRAY_SIZE(attachments));

   *buffers =
      dri_screen->dri2.loader->getBuffersWithFormat(drawable,
                                                    &drawable->w,
                                                    &drawable->h,
                                                    attachments, i / 2,
                                                    buffer_count,
                                                    drawable->loaderPrivate);
}

/**
 * \brief Assign a DRI buffer's DRM region to a renderbuffer.
 *
 * This is called from intel_update_renderbuffers().
 *
 * \par Note:
 *    DRI buffers whose attachment point is DRI2BufferStencil or
 *    DRI2BufferDepthStencil are handled as special cases.
 *
 * \param buffer_name is a human readable name, such as "dri2 front buffer",
 *        that is passed to brw_bo_gem_create_from_name().
 *
 * \see intel_update_renderbuffers()
 */
static void
intel_process_dri2_buffer(struct brw_context *brw,
                          __DRIdrawable *drawable,
                          __DRIbuffer *buffer,
                          struct intel_renderbuffer *rb,
                          const char *buffer_name)
{
   struct gl_framebuffer *fb = drawable->driverPrivate;
   struct brw_bo *bo;

   if (!rb)
      return;

   unsigned num_samples = rb->Base.Base.NumSamples;

   /* We try to avoid closing and reopening the same BO name, because the first
    * use of a mapping of the buffer involves a bunch of page faulting which is
    * moderately expensive.
    */
   struct intel_mipmap_tree *last_mt;
   if (num_samples == 0)
      last_mt = rb->mt;
   else
      last_mt = rb->singlesample_mt;

   uint32_t old_name = 0;
   if (last_mt) {
       /* The bo already has a name because the miptree was created by a
	* previous call to intel_process_dri2_buffer(). If a bo already has a
	* name, then brw_bo_flink() is a low-cost getter.  It does not
	* create a new name.
	*/
      brw_bo_flink(last_mt->bo, &old_name);
   }

   if (old_name == buffer->name)
      return;

   if (unlikely(INTEL_DEBUG & DEBUG_DRI)) {
      fprintf(stderr,
              "attaching buffer %d, at %d, cpp %d, pitch %d\n",
              buffer->name, buffer->attachment,
              buffer->cpp, buffer->pitch);
   }

   bo = brw_bo_gem_create_from_name(brw->bufmgr, buffer_name,
                                          buffer->name);
   if (!bo) {
      fprintf(stderr,
              "Failed to open BO for returned DRI2 buffer "
              "(%dx%d, %s, named %d).\n"
              "This is likely a bug in the X Server that will lead to a "
              "crash soon.\n",
              drawable->w, drawable->h, buffer_name, buffer->name);
      return;
   }

   uint32_t tiling, swizzle;
   brw_bo_get_tiling(bo, &tiling, &swizzle);

   struct intel_mipmap_tree *mt =
      intel_miptree_create_for_bo(brw,
                                  bo,
                                  intel_rb_format(rb),
                                  0,
                                  drawable->w,
                                  drawable->h,
                                  1,
                                  buffer->pitch,
                                  isl_tiling_from_i915_tiling(tiling),
                                  MIPTREE_CREATE_DEFAULT);
   if (!mt) {
      brw_bo_unreference(bo);
      return;
   }

   /* We got this BO from X11.  We cana't assume that we have coherent texture
    * access because X may suddenly decide to use it for scan-out which would
    * destroy coherency.
    */
   bo->cache_coherent = false;

   if (!intel_update_winsys_renderbuffer_miptree(brw, rb, mt,
                                                 drawable->w, drawable->h,
                                                 buffer->pitch)) {
      brw_bo_unreference(bo);
      intel_miptree_release(&mt);
      return;
   }

   if (_mesa_is_front_buffer_drawing(fb) &&
       (buffer->attachment == __DRI_BUFFER_FRONT_LEFT ||
        buffer->attachment == __DRI_BUFFER_FAKE_FRONT_LEFT) &&
       rb->Base.Base.NumSamples > 1) {
      intel_renderbuffer_upsample(brw, rb);
   }

   assert(rb->mt);

   brw_bo_unreference(bo);
}

/**
 * \brief Query DRI image loader to obtain a DRIdrawable's buffers.
 *
 * To determine which DRI buffers to request, examine the renderbuffers
 * attached to the drawable's framebuffer. Then request the buffers from
 * the image loader
 *
 * This is called from intel_update_renderbuffers().
 *
 * \param drawable      Drawable whose buffers are queried.
 * \param buffers       [out] List of buffers returned by DRI2 query.
 * \param buffer_count  [out] Number of buffers returned.
 *
 * \see intel_update_renderbuffers()
 */

static void
intel_update_image_buffer(struct brw_context *intel,
                          __DRIdrawable *drawable,
                          struct intel_renderbuffer *rb,
                          __DRIimage *buffer,
                          enum __DRIimageBufferMask buffer_type)
{
   struct gl_framebuffer *fb = drawable->driverPrivate;

   if (!rb || !buffer->bo)
      return;

   unsigned num_samples = rb->Base.Base.NumSamples;

   /* Check and see if we're already bound to the right
    * buffer object
    */
   struct intel_mipmap_tree *last_mt;
   if (num_samples == 0)
      last_mt = rb->mt;
   else
      last_mt = rb->singlesample_mt;

   if (last_mt && last_mt->bo == buffer->bo) {
      if (buffer_type == __DRI_IMAGE_BUFFER_SHARED) {
         intel_miptree_make_shareable(intel, last_mt);
      }
      return;
   }

   /* Only allow internal compression if samples == 0.  For multisampled
    * window system buffers, the only thing the single-sampled buffer is used
    * for is as a resolve target.  If we do any compression beyond what is
    * supported by the window system, we will just have to resolve so it's
    * probably better to just not bother.
    */
   const bool allow_internal_aux = (num_samples == 0);

   struct intel_mipmap_tree *mt =
      intel_miptree_create_for_dri_image(intel, buffer, GL_TEXTURE_2D,
                                         intel_rb_format(rb),
                                         allow_internal_aux);
   if (!mt)
      return;

   if (!intel_update_winsys_renderbuffer_miptree(intel, rb, mt,
                                                 buffer->width, buffer->height,
                                                 buffer->pitch)) {
      intel_miptree_release(&mt);
      return;
   }

   if (_mesa_is_front_buffer_drawing(fb) &&
       buffer_type == __DRI_IMAGE_BUFFER_FRONT &&
       rb->Base.Base.NumSamples > 1) {
      intel_renderbuffer_upsample(intel, rb);
   }

   if (buffer_type == __DRI_IMAGE_BUFFER_SHARED) {
      /* The compositor and the application may access this image
       * concurrently. The display hardware may even scanout the image while
       * the GPU is rendering to it.  Aux surfaces cause difficulty with
       * concurrent access, so permanently disable aux for this miptree.
       *
       * Perhaps we could improve overall application performance by
       * re-enabling the aux surface when EGL_RENDER_BUFFER transitions to
       * EGL_BACK_BUFFER, then disabling it again when EGL_RENDER_BUFFER
       * returns to EGL_SINGLE_BUFFER. I expect the wins and losses with this
       * approach to be highly dependent on the application's GL usage.
       *
       * I [chadv] expect clever disabling/reenabling to be counterproductive
       * in the use cases I care about: applications that render nearly
       * realtime handwriting to the surface while possibly undergiong
       * simultaneously scanout as a display plane. The app requires low
       * render latency. Even though the app spends most of its time in
       * shared-buffer mode, it also frequently transitions between
       * shared-buffer (EGL_SINGLE_BUFFER) and double-buffer (EGL_BACK_BUFFER)
       * mode.  Visual sutter during the transitions should be avoided.
       *
       * In this case, I [chadv] believe reducing the GPU workload at
       * shared-buffer/double-buffer transitions would offer a smoother app
       * experience than any savings due to aux compression. But I've
       * collected no data to prove my theory.
       */
      intel_miptree_make_shareable(intel, mt);
   }
}

static void
intel_update_image_buffers(struct brw_context *brw, __DRIdrawable *drawable)
{
   struct gl_framebuffer *fb = drawable->driverPrivate;
   __DRIscreen *dri_screen = brw->screen->driScrnPriv;
   struct intel_renderbuffer *front_rb;
   struct intel_renderbuffer *back_rb;
   struct __DRIimageList images;
   mesa_format format;
   uint32_t buffer_mask = 0;
   int ret;

   front_rb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT);
   back_rb = intel_get_renderbuffer(fb, BUFFER_BACK_LEFT);

   if (back_rb)
      format = intel_rb_format(back_rb);
   else if (front_rb)
      format = intel_rb_format(front_rb);
   else
      return;

   if (front_rb && (_mesa_is_front_buffer_drawing(fb) ||
                    _mesa_is_front_buffer_reading(fb) || !back_rb)) {
      buffer_mask |= __DRI_IMAGE_BUFFER_FRONT;
   }

   if (back_rb)
      buffer_mask |= __DRI_IMAGE_BUFFER_BACK;

   ret = dri_screen->image.loader->getBuffers(drawable,
                                              driGLFormatToImageFormat(format),
                                              &drawable->dri2.stamp,
                                              drawable->loaderPrivate,
                                              buffer_mask,
                                              &images);
   if (!ret)
      return;

   if (images.image_mask & __DRI_IMAGE_BUFFER_FRONT) {
      drawable->w = images.front->width;
      drawable->h = images.front->height;
      intel_update_image_buffer(brw,
                                drawable,
                                front_rb,
                                images.front,
                                __DRI_IMAGE_BUFFER_FRONT);
   }

   if (images.image_mask & __DRI_IMAGE_BUFFER_BACK) {
      drawable->w = images.back->width;
      drawable->h = images.back->height;
      intel_update_image_buffer(brw,
                                drawable,
                                back_rb,
                                images.back,
                                __DRI_IMAGE_BUFFER_BACK);
   }

   if (images.image_mask & __DRI_IMAGE_BUFFER_SHARED) {
      assert(images.image_mask == __DRI_IMAGE_BUFFER_SHARED);
      drawable->w = images.back->width;
      drawable->h = images.back->height;
      intel_update_image_buffer(brw,
                                drawable,
                                back_rb,
                                images.back,
                                __DRI_IMAGE_BUFFER_SHARED);
      brw->is_shared_buffer_bound = true;
   } else {
      brw->is_shared_buffer_bound = false;
      brw->is_shared_buffer_dirty = false;
   }
}