summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/panfrost/pan_context.c
blob: af8f94354e9b275c6b3f3afa17b8d550c306d00f (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
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
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
/*
 * © Copyright 2018 Alyssa Rosenzweig
 * Copyright © 2014-2017 Broadcom
 * Copyright (C) 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 (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 *
 */

#include <sys/poll.h>
#include <errno.h>

#include "pan_bo.h"
#include "pan_context.h"
#include "pan_format.h"
#include "panfrost-quirks.h"

#include "util/macros.h"
#include "util/format/u_format.h"
#include "util/u_inlines.h"
#include "util/u_upload_mgr.h"
#include "util/u_memory.h"
#include "util/u_vbuf.h"
#include "util/half_float.h"
#include "util/u_helpers.h"
#include "util/format/u_format.h"
#include "util/u_prim.h"
#include "util/u_prim_restart.h"
#include "indices/u_primconvert.h"
#include "tgsi/tgsi_parse.h"
#include "tgsi/tgsi_from_mesa.h"
#include "util/u_math.h"

#include "pan_screen.h"
#include "pan_blending.h"
#include "pan_blend_shaders.h"
#include "pan_util.h"

/* Framebuffer descriptor */

static struct midgard_tiler_descriptor
panfrost_emit_midg_tiler(struct panfrost_batch *batch, unsigned vertex_count)
{
        struct panfrost_screen *screen = pan_screen(batch->ctx->base.screen);
        struct midgard_tiler_descriptor t = {0};
        unsigned height = batch->key.height;
        unsigned width = batch->key.width;

        t.hierarchy_mask =
                panfrost_choose_hierarchy_mask(width, height, vertex_count);

        /* Compute the polygon header size and use that to offset the body */

        unsigned header_size = panfrost_tiler_header_size(
                                       width, height, t.hierarchy_mask);

        t.polygon_list_size = panfrost_tiler_full_size(
                                     width, height, t.hierarchy_mask);

        /* Sanity check */

        if (t.hierarchy_mask) {
                struct panfrost_bo *tiler_heap;

                tiler_heap = panfrost_batch_get_tiler_heap(batch);
                t.polygon_list = panfrost_batch_get_polygon_list(batch,
                                                                 header_size +
                                                                 t.polygon_list_size);


                /* Allow the entire tiler heap */
                t.heap_start = tiler_heap->gpu;
                t.heap_end = tiler_heap->gpu + tiler_heap->size;
        } else {
                struct panfrost_bo *tiler_dummy;

                tiler_dummy = panfrost_batch_get_tiler_dummy(batch);

                /* The tiler is disabled, so don't allow the tiler heap */
                t.heap_start = tiler_dummy->gpu;
                t.heap_end = t.heap_start;

                /* Use a dummy polygon list */
                t.polygon_list = tiler_dummy->gpu;

                /* Disable the tiler */
                t.hierarchy_mask |= MALI_TILER_DISABLED;

                if (screen->quirks & MIDGARD_SFBD) {
                        t.hierarchy_mask = 0xFFF; /* TODO: What's this? */
                        t.polygon_list_size = 0x200;

                        /* We don't have a SET_VALUE job, so write the polygon list manually */
                        uint32_t *polygon_list_body = (uint32_t *) (tiler_dummy->cpu + header_size);
                        polygon_list_body[0] = 0xa0000000; /* TODO: Just that? */
                }
        }

        t.polygon_list_body =
                t.polygon_list + header_size;

        return t;
}

struct mali_single_framebuffer
panfrost_emit_sfbd(struct panfrost_batch *batch, unsigned vertex_count)
{
        unsigned width = batch->key.width;
        unsigned height = batch->key.height;

        struct mali_single_framebuffer framebuffer = {
                .width = MALI_POSITIVE(width),
                .height = MALI_POSITIVE(height),
                .unknown2 = 0x1f,
                .format = {
                        .unk3 = 0x3,
                },
                .clear_flags = 0x1000,
                .unknown_address_0 = panfrost_batch_get_scratchpad(batch)->gpu,
                .tiler = panfrost_emit_midg_tiler(batch, vertex_count),
        };

        return framebuffer;
}

struct bifrost_framebuffer
panfrost_emit_mfbd(struct panfrost_batch *batch, unsigned vertex_count)
{
        unsigned width = batch->key.width;
        unsigned height = batch->key.height;

        struct bifrost_framebuffer framebuffer = {
                .unk0 = 0x1e5, /* 1e4 if no spill */
                .width1 = MALI_POSITIVE(width),
                .height1 = MALI_POSITIVE(height),
                .width2 = MALI_POSITIVE(width),
                .height2 = MALI_POSITIVE(height),

                .unk1 = 0x1080,

                .rt_count_1 = MALI_POSITIVE(batch->key.nr_cbufs),
                .rt_count_2 = 4,

                .unknown2 = 0x1f,

                .scratchpad = panfrost_batch_get_scratchpad(batch)->gpu,
                .tiler = panfrost_emit_midg_tiler(batch, vertex_count)
        };

        return framebuffer;
}

static void
panfrost_clear(
        struct pipe_context *pipe,
        unsigned buffers,
        const union pipe_color_union *color,
        double depth, unsigned stencil)
{
        struct panfrost_context *ctx = pan_context(pipe);

        /* TODO: panfrost_get_fresh_batch_for_fbo() instantiates a new batch if
         * the existing batch targeting this FBO has draws. We could probably
         * avoid that by replacing plain clears by quad-draws with a specific
         * color/depth/stencil value, thus avoiding the generation of extra
         * fragment/set_value jobs.
         */
        struct panfrost_batch *batch = panfrost_get_fresh_batch_for_fbo(ctx);

        panfrost_batch_add_fbo_bos(batch);
        panfrost_batch_clear(batch, buffers, color, depth, stencil);
}

static mali_ptr
panfrost_attach_vt_mfbd(struct panfrost_batch *batch)
{
        struct bifrost_framebuffer mfbd = panfrost_emit_mfbd(batch, ~0);

        return panfrost_upload_transient(batch, &mfbd, sizeof(mfbd)) | MALI_MFBD;
}

static mali_ptr
panfrost_attach_vt_sfbd(struct panfrost_batch *batch)
{
        struct mali_single_framebuffer sfbd = panfrost_emit_sfbd(batch, ~0);

        return panfrost_upload_transient(batch, &sfbd, sizeof(sfbd)) | MALI_SFBD;
}

static void
panfrost_attach_vt_framebuffer(struct panfrost_context *ctx)
{
        /* Skip the attach if we can */

        if (ctx->payloads[PIPE_SHADER_VERTEX].postfix.framebuffer) {
                assert(ctx->payloads[PIPE_SHADER_FRAGMENT].postfix.framebuffer);
                return;
        }

        struct panfrost_screen *screen = pan_screen(ctx->base.screen);
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        if (!batch->framebuffer)
                batch->framebuffer = (screen->quirks & MIDGARD_SFBD) ?
                                     panfrost_attach_vt_sfbd(batch) :
                                     panfrost_attach_vt_mfbd(batch);

        for (unsigned i = 0; i < PIPE_SHADER_TYPES; ++i)
                ctx->payloads[i].postfix.framebuffer = batch->framebuffer;
}

/* Reset per-frame context, called on context initialisation as well as after
 * flushing a frame */

void
panfrost_invalidate_frame(struct panfrost_context *ctx)
{
        for (unsigned i = 0; i < PIPE_SHADER_TYPES; ++i)
                ctx->payloads[i].postfix.framebuffer = 0;

        if (ctx->rasterizer)
                ctx->dirty |= PAN_DIRTY_RASTERIZER;

        /* XXX */
        ctx->dirty |= PAN_DIRTY_SAMPLERS | PAN_DIRTY_TEXTURES;

        /* TODO: When does this need to be handled? */
        ctx->active_queries = true;
}

/* In practice, every field of these payloads should be configurable
 * arbitrarily, which means these functions are basically catch-all's for
 * as-of-yet unwavering unknowns */

static void
panfrost_emit_vertex_payload(struct panfrost_context *ctx)
{
        /* 0x2 bit clear on 32-bit T6XX */

        struct midgard_payload_vertex_tiler payload = {
                .gl_enables = 0x4 | 0x2,
        };

        /* Vertex and compute are closely coupled, so share a payload */

        memcpy(&ctx->payloads[PIPE_SHADER_VERTEX], &payload, sizeof(payload));
        memcpy(&ctx->payloads[PIPE_SHADER_COMPUTE], &payload, sizeof(payload));
}

static void
panfrost_emit_tiler_payload(struct panfrost_context *ctx)
{
        struct midgard_payload_vertex_tiler payload = {
                .prefix = {
                        .zero1 = 0xffff, /* Why is this only seen on test-quad-textured? */
                },
        };

        memcpy(&ctx->payloads[PIPE_SHADER_FRAGMENT], &payload, sizeof(payload));
}

static unsigned
translate_tex_wrap(enum pipe_tex_wrap w)
{
        switch (w) {
        case PIPE_TEX_WRAP_REPEAT:
                return MALI_WRAP_REPEAT;

                /* TODO: lower GL_CLAMP? */
        case PIPE_TEX_WRAP_CLAMP:
        case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
                return MALI_WRAP_CLAMP_TO_EDGE;

        case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
                return MALI_WRAP_CLAMP_TO_BORDER;

        case PIPE_TEX_WRAP_MIRROR_REPEAT:
                return MALI_WRAP_MIRRORED_REPEAT;

        default:
                unreachable("Invalid wrap");
        }
}

static unsigned
panfrost_translate_compare_func(enum pipe_compare_func in)
{
        switch (in) {
        case PIPE_FUNC_NEVER:
                return MALI_FUNC_NEVER;

        case PIPE_FUNC_LESS:
                return MALI_FUNC_LESS;

        case PIPE_FUNC_EQUAL:
                return MALI_FUNC_EQUAL;

        case PIPE_FUNC_LEQUAL:
                return MALI_FUNC_LEQUAL;

        case PIPE_FUNC_GREATER:
                return MALI_FUNC_GREATER;

        case PIPE_FUNC_NOTEQUAL:
                return MALI_FUNC_NOTEQUAL;

        case PIPE_FUNC_GEQUAL:
                return MALI_FUNC_GEQUAL;

        case PIPE_FUNC_ALWAYS:
                return MALI_FUNC_ALWAYS;

        default:
                unreachable("Invalid func");
        }
}

static unsigned
panfrost_translate_alt_compare_func(enum pipe_compare_func in)
{
        switch (in) {
        case PIPE_FUNC_NEVER:
                return MALI_ALT_FUNC_NEVER;

        case PIPE_FUNC_LESS:
                return MALI_ALT_FUNC_LESS;

        case PIPE_FUNC_EQUAL:
                return MALI_ALT_FUNC_EQUAL;

        case PIPE_FUNC_LEQUAL:
                return MALI_ALT_FUNC_LEQUAL;

        case PIPE_FUNC_GREATER:
                return MALI_ALT_FUNC_GREATER;

        case PIPE_FUNC_NOTEQUAL:
                return MALI_ALT_FUNC_NOTEQUAL;

        case PIPE_FUNC_GEQUAL:
                return MALI_ALT_FUNC_GEQUAL;

        case PIPE_FUNC_ALWAYS:
                return MALI_ALT_FUNC_ALWAYS;

        default:
                unreachable("Invalid alt func");
        }
}

static unsigned
panfrost_translate_stencil_op(enum pipe_stencil_op in)
{
        switch (in) {
        case PIPE_STENCIL_OP_KEEP:
                return MALI_STENCIL_KEEP;

        case PIPE_STENCIL_OP_ZERO:
                return MALI_STENCIL_ZERO;

        case PIPE_STENCIL_OP_REPLACE:
                return MALI_STENCIL_REPLACE;

        case PIPE_STENCIL_OP_INCR:
                return MALI_STENCIL_INCR;

        case PIPE_STENCIL_OP_DECR:
                return MALI_STENCIL_DECR;

        case PIPE_STENCIL_OP_INCR_WRAP:
                return MALI_STENCIL_INCR_WRAP;

        case PIPE_STENCIL_OP_DECR_WRAP:
                return MALI_STENCIL_DECR_WRAP;

        case PIPE_STENCIL_OP_INVERT:
                return MALI_STENCIL_INVERT;

        default:
                unreachable("Invalid stencil op");
        }
}

static void
panfrost_make_stencil_state(const struct pipe_stencil_state *in, struct mali_stencil_test *out)
{
        out->ref = 0; /* Gallium gets it from elsewhere */

        out->mask = in->valuemask;
        out->func = panfrost_translate_compare_func(in->func);
        out->sfail = panfrost_translate_stencil_op(in->fail_op);
        out->dpfail = panfrost_translate_stencil_op(in->zfail_op);
        out->dppass = panfrost_translate_stencil_op(in->zpass_op);
}

static void
panfrost_default_shader_backend(struct panfrost_context *ctx)
{
        struct panfrost_screen *screen = pan_screen(ctx->base.screen);
        struct mali_shader_meta shader = {
                .alpha_coverage = ~MALI_ALPHA_COVERAGE(0.000000),

                .unknown2_3 = MALI_DEPTH_FUNC(MALI_FUNC_ALWAYS) | 0x3010,
                .unknown2_4 = MALI_NO_MSAA | 0x4e0,
        };

        /* unknown2_4 has 0x10 bit set on T6XX and T720. We don't know why this is
         * required (independent of 32-bit/64-bit descriptors), or why it's not
         * used on later GPU revisions. Otherwise, all shader jobs fault on
         * these earlier chips (perhaps this is a chicken bit of some kind).
         * More investigation is needed. */

	if (screen->quirks & MIDGARD_SFBD)
		shader.unknown2_4 |= 0x10;

        struct pipe_stencil_state default_stencil = {
                .enabled = 0,
                .func = PIPE_FUNC_ALWAYS,
                .fail_op = MALI_STENCIL_KEEP,
                .zfail_op = MALI_STENCIL_KEEP,
                .zpass_op = MALI_STENCIL_KEEP,
                .writemask = 0xFF,
                .valuemask = 0xFF
        };

        panfrost_make_stencil_state(&default_stencil, &shader.stencil_front);
        shader.stencil_mask_front = default_stencil.writemask;

        panfrost_make_stencil_state(&default_stencil, &shader.stencil_back);
        shader.stencil_mask_back = default_stencil.writemask;

        if (default_stencil.enabled)
                shader.unknown2_4 |= MALI_STENCIL_TEST;

        memcpy(&ctx->fragment_shader_core, &shader, sizeof(shader));
}

/* Generates a vertex/tiler job. This is, in some sense, the heart of the
 * graphics command stream. It should be called once per draw, accordding to
 * presentations. Set is_tiler for "tiler" jobs (fragment shader jobs, but in
 * Mali parlance, "fragment" refers to framebuffer writeout). Clear it for
 * vertex jobs. */

struct panfrost_transfer
panfrost_vertex_tiler_job(struct panfrost_context *ctx, bool is_tiler)
{
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        struct mali_job_descriptor_header job = {
                .job_type = is_tiler ? JOB_TYPE_TILER : JOB_TYPE_VERTEX,
                .job_descriptor_size = 1,
        };

        struct midgard_payload_vertex_tiler *payload = is_tiler ? &ctx->payloads[PIPE_SHADER_FRAGMENT] : &ctx->payloads[PIPE_SHADER_VERTEX];

        struct panfrost_transfer transfer = panfrost_allocate_transient(batch, sizeof(job) + sizeof(*payload));
        memcpy(transfer.cpu, &job, sizeof(job));
        memcpy(transfer.cpu + sizeof(job), payload, sizeof(*payload));
        return transfer;
}

mali_ptr
panfrost_vertex_buffer_address(struct panfrost_context *ctx, unsigned i)
{
        struct pipe_vertex_buffer *buf = &ctx->vertex_buffers[i];
        struct panfrost_resource *rsrc = (struct panfrost_resource *) (buf->buffer.resource);

        return rsrc->bo->gpu + buf->buffer_offset;
}

static bool
panfrost_writes_point_size(struct panfrost_context *ctx)
{
        assert(ctx->shader[PIPE_SHADER_VERTEX]);
        struct panfrost_shader_state *vs = &ctx->shader[PIPE_SHADER_VERTEX]->variants[ctx->shader[PIPE_SHADER_VERTEX]->active_variant];

        return vs->writes_point_size && ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.draw_mode == MALI_POINTS;
}

/* Stage the attribute descriptors so we can adjust src_offset
 * to let BOs align nicely */

static void
panfrost_stage_attributes(struct panfrost_context *ctx)
{
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        struct panfrost_vertex_state *so = ctx->vertex;

        size_t sz = sizeof(struct mali_attr_meta) * so->num_elements;
        struct panfrost_transfer transfer = panfrost_allocate_transient(batch, sz);
        struct mali_attr_meta *target = (struct mali_attr_meta *) transfer.cpu;

        /* Copy as-is for the first pass */
        memcpy(target, so->hw, sz);

        /* Fixup offsets for the second pass. Recall that the hardware
         * calculates attribute addresses as:
         *
         *      addr = base + (stride * vtx) + src_offset;
         *
         * However, on Mali, base must be aligned to 64-bytes, so we
         * instead let:
         *
         *      base' = base & ~63 = base - (base & 63)
         *
         * To compensate when using base' (see emit_vertex_data), we have
         * to adjust src_offset by the masked off piece:
         *
         *      addr' = base' + (stride * vtx) + (src_offset + (base & 63))
         *            = base - (base & 63) + (stride * vtx) + src_offset + (base & 63)
         *            = base + (stride * vtx) + src_offset
         *            = addr;
         *
         * QED.
         */

        unsigned start = ctx->payloads[PIPE_SHADER_VERTEX].offset_start;

        for (unsigned i = 0; i < so->num_elements; ++i) {
                unsigned vbi = so->pipe[i].vertex_buffer_index;
                struct pipe_vertex_buffer *buf = &ctx->vertex_buffers[vbi];
                mali_ptr addr = panfrost_vertex_buffer_address(ctx, vbi);

                /* Adjust by the masked off bits of the offset */
                target[i].src_offset += (addr & 63);

                /* Also, somewhat obscurely per-instance data needs to be
                 * offset in response to a delayed start in an indexed draw */

                if (so->pipe[i].instance_divisor && ctx->instance_count > 1 && start) {
                        target[i].src_offset -= buf->stride * start;
                }


        }

        ctx->payloads[PIPE_SHADER_VERTEX].postfix.attribute_meta = transfer.gpu;
}

static void
panfrost_upload_sampler_descriptors(struct panfrost_context *ctx)
{
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        size_t desc_size = sizeof(struct mali_sampler_descriptor);

        for (int t = 0; t <= PIPE_SHADER_FRAGMENT; ++t) {
                mali_ptr upload = 0;

                if (ctx->sampler_count[t] && ctx->sampler_view_count[t]) {
                        size_t transfer_size = desc_size * ctx->sampler_count[t];

                        struct panfrost_transfer transfer =
                                panfrost_allocate_transient(batch, transfer_size);

                        struct mali_sampler_descriptor *desc =
                                (struct mali_sampler_descriptor *) transfer.cpu;

                        for (int i = 0; i < ctx->sampler_count[t]; ++i)
                                desc[i] = ctx->samplers[t][i]->hw;

                        upload = transfer.gpu;
                }

                ctx->payloads[t].postfix.sampler_descriptor = upload;
        }
}

static enum mali_texture_layout
panfrost_layout_for_texture(struct panfrost_resource *rsrc)
{
        /* TODO: other linear depth textures */
        bool is_depth = rsrc->base.format == PIPE_FORMAT_Z32_UNORM;

        switch (rsrc->layout) {
        case PAN_AFBC:
                return MALI_TEXTURE_AFBC;
        case PAN_TILED:
                assert(!is_depth);
                return MALI_TEXTURE_TILED;
        case PAN_LINEAR:
                return is_depth ? MALI_TEXTURE_TILED : MALI_TEXTURE_LINEAR;
        default:
                unreachable("Invalid texture layout");
        }
}

static mali_ptr
panfrost_upload_tex(
        struct panfrost_context *ctx,
        enum pipe_shader_type st,
        struct panfrost_sampler_view *view)
{
        if (!view)
                return (mali_ptr) 0;

        struct pipe_sampler_view *pview = &view->base;
        struct panfrost_resource *rsrc = pan_resource(pview->texture);

        /* Do we interleave an explicit stride with every element? */

        bool has_manual_stride = view->manual_stride;

        /* For easy access */

        bool is_buffer = pview->target == PIPE_BUFFER;
        unsigned first_level = is_buffer ? 0 : pview->u.tex.first_level;
        unsigned last_level  = is_buffer ? 0 : pview->u.tex.last_level;
        unsigned first_layer = is_buffer ? 0 : pview->u.tex.first_layer;
        unsigned last_layer  = is_buffer ? 0 : pview->u.tex.last_layer;

        /* Lower-bit is set when sampling from colour AFBC */
        bool is_afbc = rsrc->layout == PAN_AFBC;
        bool is_zs = rsrc->base.bind & PIPE_BIND_DEPTH_STENCIL;
        unsigned afbc_bit = (is_afbc && !is_zs) ? 1 : 0;

        /* Add the BO to the job so it's retained until the job is done. */
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        panfrost_batch_add_bo(batch, rsrc->bo,
                              PAN_BO_ACCESS_SHARED | PAN_BO_ACCESS_READ |
                              panfrost_bo_access_for_stage(st));

        /* Add the usage flags in, since they can change across the CSO
         * lifetime due to layout switches */

        view->hw.format.layout = panfrost_layout_for_texture(rsrc);
        view->hw.format.manual_stride = has_manual_stride;

        /* Inject the addresses in, interleaving mip levels, cube faces, and
         * strides in that order */

        unsigned idx = 0;

        for (unsigned l = first_level; l <= last_level; ++l) {
                for (unsigned f = first_layer; f <= last_layer; ++f) {

                        view->hw.payload[idx++] =
                                panfrost_get_texture_address(rsrc, l, f) + afbc_bit;

                        if (has_manual_stride) {
                                view->hw.payload[idx++] =
                                        rsrc->slices[l].stride;
                        }
                }
        }

        return panfrost_upload_transient(batch, &view->hw,
                                         sizeof(struct mali_texture_descriptor));
}

static void
panfrost_upload_texture_descriptors(struct panfrost_context *ctx)
{
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        for (int t = 0; t <= PIPE_SHADER_FRAGMENT; ++t) {
                mali_ptr trampoline = 0;

                if (ctx->sampler_view_count[t]) {
                        uint64_t trampolines[PIPE_MAX_SHADER_SAMPLER_VIEWS];

                        for (int i = 0; i < ctx->sampler_view_count[t]; ++i)
                                trampolines[i] =
                                        panfrost_upload_tex(ctx, t, ctx->sampler_views[t][i]);

                        trampoline = panfrost_upload_transient(batch, trampolines, sizeof(uint64_t) * ctx->sampler_view_count[t]);
                }

                ctx->payloads[t].postfix.texture_trampoline = trampoline;
        }
}

struct sysval_uniform {
        union {
                float f[4];
                int32_t i[4];
                uint32_t u[4];
                uint64_t du[2];
        };
};

static void panfrost_upload_viewport_scale_sysval(struct panfrost_context *ctx,
                struct sysval_uniform *uniform)
{
        const struct pipe_viewport_state *vp = &ctx->pipe_viewport;

        uniform->f[0] = vp->scale[0];
        uniform->f[1] = vp->scale[1];
        uniform->f[2] = vp->scale[2];
}

static void panfrost_upload_viewport_offset_sysval(struct panfrost_context *ctx,
                struct sysval_uniform *uniform)
{
        const struct pipe_viewport_state *vp = &ctx->pipe_viewport;

        uniform->f[0] = vp->translate[0];
        uniform->f[1] = vp->translate[1];
        uniform->f[2] = vp->translate[2];
}

static void panfrost_upload_txs_sysval(struct panfrost_context *ctx,
                                       enum pipe_shader_type st,
                                       unsigned int sysvalid,
                                       struct sysval_uniform *uniform)
{
        unsigned texidx = PAN_SYSVAL_ID_TO_TXS_TEX_IDX(sysvalid);
        unsigned dim = PAN_SYSVAL_ID_TO_TXS_DIM(sysvalid);
        bool is_array = PAN_SYSVAL_ID_TO_TXS_IS_ARRAY(sysvalid);
        struct pipe_sampler_view *tex = &ctx->sampler_views[st][texidx]->base;

        assert(dim);
        uniform->i[0] = u_minify(tex->texture->width0, tex->u.tex.first_level);

        if (dim > 1)
                uniform->i[1] = u_minify(tex->texture->height0,
                                         tex->u.tex.first_level);

        if (dim > 2)
                uniform->i[2] = u_minify(tex->texture->depth0,
                                         tex->u.tex.first_level);

        if (is_array)
                uniform->i[dim] = tex->texture->array_size;
}

static void panfrost_upload_ssbo_sysval(
        struct panfrost_context *ctx,
        enum pipe_shader_type st,
        unsigned ssbo_id,
        struct sysval_uniform *uniform)
{
        assert(ctx->ssbo_mask[st] & (1 << ssbo_id));
        struct pipe_shader_buffer sb = ctx->ssbo[st][ssbo_id];

        /* Compute address */
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        struct panfrost_bo *bo = pan_resource(sb.buffer)->bo;

        panfrost_batch_add_bo(batch, bo,
                              PAN_BO_ACCESS_SHARED | PAN_BO_ACCESS_RW |
                              panfrost_bo_access_for_stage(st));

        /* Upload address and size as sysval */
        uniform->du[0] = bo->gpu + sb.buffer_offset;
        uniform->u[2] = sb.buffer_size;
}

static void
panfrost_upload_sampler_sysval(
                struct panfrost_context *ctx,
                enum pipe_shader_type st,
                unsigned sampler_index,
                struct sysval_uniform *uniform)
{
        struct pipe_sampler_state *sampl =
                &ctx->samplers[st][sampler_index]->base;

        uniform->f[0] = sampl->min_lod;
        uniform->f[1] = sampl->max_lod;
        uniform->f[2] = sampl->lod_bias;

        /* Even without any errata, Midgard represents "no mipmapping" as
         * fixing the LOD with the clamps; keep behaviour consistent. c.f.
         * panfrost_create_sampler_state which also explains our choice of
         * epsilon value (again to keep behaviour consistent) */

        if (sampl->min_mip_filter == PIPE_TEX_MIPFILTER_NONE)
                uniform->f[1] = uniform->f[0] + (1.0/256.0);
}

static void panfrost_upload_num_work_groups_sysval(struct panfrost_context *ctx,
                struct sysval_uniform *uniform)
{
        uniform->u[0] = ctx->compute_grid->grid[0];
        uniform->u[1] = ctx->compute_grid->grid[1];
        uniform->u[2] = ctx->compute_grid->grid[2];
}

static void panfrost_upload_sysvals(struct panfrost_context *ctx, void *buf,
                                    struct panfrost_shader_state *ss,
                                    enum pipe_shader_type st)
{
        struct sysval_uniform *uniforms = (void *)buf;

        for (unsigned i = 0; i < ss->sysval_count; ++i) {
                int sysval = ss->sysval[i];

                switch (PAN_SYSVAL_TYPE(sysval)) {
                case PAN_SYSVAL_VIEWPORT_SCALE:
                        panfrost_upload_viewport_scale_sysval(ctx, &uniforms[i]);
                        break;
                case PAN_SYSVAL_VIEWPORT_OFFSET:
                        panfrost_upload_viewport_offset_sysval(ctx, &uniforms[i]);
                        break;
                case PAN_SYSVAL_TEXTURE_SIZE:
                        panfrost_upload_txs_sysval(ctx, st, PAN_SYSVAL_ID(sysval),
                                                   &uniforms[i]);
                        break;
                case PAN_SYSVAL_SSBO:
                        panfrost_upload_ssbo_sysval(ctx, st, PAN_SYSVAL_ID(sysval),
                                                    &uniforms[i]);
                        break;
                case PAN_SYSVAL_NUM_WORK_GROUPS:
                        panfrost_upload_num_work_groups_sysval(ctx, &uniforms[i]);
                        break;
                case PAN_SYSVAL_SAMPLER:
                        panfrost_upload_sampler_sysval(ctx, st, PAN_SYSVAL_ID(sysval),
                                                    &uniforms[i]);
                        break;
                default:
                        assert(0);
                }
        }
}

static const void *
panfrost_map_constant_buffer_cpu(struct panfrost_constant_buffer *buf, unsigned index)
{
        struct pipe_constant_buffer *cb = &buf->cb[index];
        struct panfrost_resource *rsrc = pan_resource(cb->buffer);

        if (rsrc)
                return rsrc->bo->cpu;
        else if (cb->user_buffer)
                return cb->user_buffer;
        else
                unreachable("No constant buffer");
}

static mali_ptr
panfrost_map_constant_buffer_gpu(
        struct panfrost_context *ctx,
        enum pipe_shader_type st,
        struct panfrost_constant_buffer *buf,
        unsigned index)
{
        struct pipe_constant_buffer *cb = &buf->cb[index];
        struct panfrost_resource *rsrc = pan_resource(cb->buffer);
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        if (rsrc) {
                panfrost_batch_add_bo(batch, rsrc->bo,
                                      PAN_BO_ACCESS_SHARED |
                                      PAN_BO_ACCESS_READ |
                                      panfrost_bo_access_for_stage(st));
                return rsrc->bo->gpu;
	} else if (cb->user_buffer) {
                return panfrost_upload_transient(batch, cb->user_buffer, cb->buffer_size);
	} else {
                unreachable("No constant buffer");
        }
}

/* Compute number of UBOs active (more specifically, compute the highest UBO
 * number addressable -- if there are gaps, include them in the count anyway).
 * We always include UBO #0 in the count, since we *need* uniforms enabled for
 * sysvals. */

static unsigned
panfrost_ubo_count(struct panfrost_context *ctx, enum pipe_shader_type stage)
{
        unsigned mask = ctx->constant_buffer[stage].enabled_mask | 1;
        return 32 - __builtin_clz(mask);
}

/* Fixes up a shader state with current state, returning a GPU address to the
 * patched shader */

static mali_ptr
panfrost_patch_shader_state(
        struct panfrost_context *ctx,
        struct panfrost_shader_state *ss,
        enum pipe_shader_type stage,
        bool should_upload)
{
        ss->tripipe->texture_count = ctx->sampler_view_count[stage];
        ss->tripipe->sampler_count = ctx->sampler_count[stage];

        ss->tripipe->midgard1.flags = 0x220;

        unsigned ubo_count = panfrost_ubo_count(ctx, stage);
        ss->tripipe->midgard1.uniform_buffer_count = ubo_count;

        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        /* Add the shader BO to the batch. */
        panfrost_batch_add_bo(batch, ss->bo,
                              PAN_BO_ACCESS_PRIVATE |
                              PAN_BO_ACCESS_READ |
			      panfrost_bo_access_for_stage(stage));

        /* We can't reuse over frames; that's not safe. The descriptor must be
         * transient uploaded */

        if (should_upload) {
                return panfrost_upload_transient(batch, ss->tripipe,
                                                 sizeof(struct mali_shader_meta));
        }

        /* If we don't need an upload, don't bother */
        return 0;

}

static void
panfrost_patch_shader_state_compute(
        struct panfrost_context *ctx,
        enum pipe_shader_type stage,
        bool should_upload)
{
        struct panfrost_shader_variants *all = ctx->shader[stage];

        if (!all) {
                ctx->payloads[stage].postfix.shader = 0;
                return;
        }

        struct panfrost_shader_state *s = &all->variants[all->active_variant];

        ctx->payloads[stage].postfix.shader =
                panfrost_patch_shader_state(ctx, s, stage, should_upload);
}

/* Go through dirty flags and actualise them in the cmdstream. */

void
panfrost_emit_for_draw(struct panfrost_context *ctx, bool with_vertex_data)
{
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
        struct panfrost_screen *screen = pan_screen(ctx->base.screen);

        panfrost_batch_add_fbo_bos(batch);
        panfrost_attach_vt_framebuffer(ctx);

        if (with_vertex_data) {
                panfrost_emit_vertex_data(batch);

                /* Varyings emitted for -all- geometry */
                unsigned total_count = ctx->padded_count * ctx->instance_count;
                panfrost_emit_varying_descriptor(ctx, total_count);
        }

        bool msaa = ctx->rasterizer->base.multisample;

        if (ctx->dirty & PAN_DIRTY_RASTERIZER) {
                ctx->payloads[PIPE_SHADER_FRAGMENT].gl_enables = ctx->rasterizer->tiler_gl_enables;

                /* TODO: Sample size */
                SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_HAS_MSAA, msaa);
                SET_BIT(ctx->fragment_shader_core.unknown2_4, MALI_NO_MSAA, !msaa);
        }

        panfrost_batch_set_requirements(batch);

        if (ctx->occlusion_query) {
                ctx->payloads[PIPE_SHADER_FRAGMENT].gl_enables |= MALI_OCCLUSION_QUERY;
                ctx->payloads[PIPE_SHADER_FRAGMENT].postfix.occlusion_counter = ctx->occlusion_query->bo->gpu;
        }

        panfrost_patch_shader_state_compute(ctx, PIPE_SHADER_VERTEX, true);
        panfrost_patch_shader_state_compute(ctx, PIPE_SHADER_COMPUTE, true);

        if (ctx->dirty & (PAN_DIRTY_RASTERIZER | PAN_DIRTY_VS)) {
                /* Check if we need to link the gl_PointSize varying */
                if (!panfrost_writes_point_size(ctx)) {
                        /* If the size is constant, write it out. Otherwise,
                         * don't touch primitive_size (since we would clobber
                         * the pointer there) */

                        ctx->payloads[PIPE_SHADER_FRAGMENT].primitive_size.constant = ctx->rasterizer->base.line_width;
                }
        }

        /* TODO: Maybe dirty track FS, maybe not. For now, it's transient. */
        if (ctx->shader[PIPE_SHADER_FRAGMENT])
                ctx->dirty |= PAN_DIRTY_FS;

        if (ctx->dirty & PAN_DIRTY_FS) {
                assert(ctx->shader[PIPE_SHADER_FRAGMENT]);
                struct panfrost_shader_state *variant = &ctx->shader[PIPE_SHADER_FRAGMENT]->variants[ctx->shader[PIPE_SHADER_FRAGMENT]->active_variant];

                panfrost_patch_shader_state(ctx, variant, PIPE_SHADER_FRAGMENT, false);

#define COPY(name) ctx->fragment_shader_core.name = variant->tripipe->name

                COPY(shader);
                COPY(attribute_count);
                COPY(varying_count);
                COPY(texture_count);
                COPY(sampler_count);
                COPY(midgard1.uniform_count);
                COPY(midgard1.uniform_buffer_count);
                COPY(midgard1.work_count);
                COPY(midgard1.flags);
                COPY(midgard1.unknown2);

#undef COPY

                /* Get blending setup */
                unsigned rt_count = MAX2(ctx->pipe_framebuffer.nr_cbufs, 1);

                struct panfrost_blend_final blend[PIPE_MAX_COLOR_BUFS];

                for (unsigned c = 0; c < rt_count; ++c)
                        blend[c] = panfrost_get_blend_for_context(ctx, c);

                /* If there is a blend shader, work registers are shared. XXX: opt */

                for (unsigned c = 0; c < rt_count; ++c) {
                        if (blend[c].is_shader)
                                ctx->fragment_shader_core.midgard1.work_count = 16;
                }

                /* Depending on whether it's legal to in the given shader, we
                 * try to enable early-z testing (or forward-pixel kill?) */

                SET_BIT(ctx->fragment_shader_core.midgard1.flags, MALI_EARLY_Z, !variant->can_discard);

                /* Any time texturing is used, derivatives are implicitly
                 * calculated, so we need to enable helper invocations */

                SET_BIT(ctx->fragment_shader_core.midgard1.flags, MALI_HELPER_INVOCATIONS, variant->helper_invocations);

                /* Assign the stencil refs late */

                unsigned front_ref = ctx->stencil_ref.ref_value[0];
                unsigned back_ref = ctx->stencil_ref.ref_value[1];
                bool back_enab = ctx->depth_stencil->stencil[1].enabled;

                ctx->fragment_shader_core.stencil_front.ref = front_ref;
                ctx->fragment_shader_core.stencil_back.ref = back_enab ? back_ref : front_ref;

                /* CAN_DISCARD should be set if the fragment shader possibly
                 * contains a 'discard' instruction. It is likely this is
                 * related to optimizations related to forward-pixel kill, as
                 * per "Mali Performance 3: Is EGL_BUFFER_PRESERVED a good
                 * thing?" by Peter Harris
                 */

                SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_CAN_DISCARD, variant->can_discard);
                SET_BIT(ctx->fragment_shader_core.midgard1.flags, 0x400, variant->can_discard);

                /* Even on MFBD, the shader descriptor gets blend shaders. It's
                 * *also* copied to the blend_meta appended (by convention),
                 * but this is the field actually read by the hardware. (Or
                 * maybe both are read...?) */

                if (blend[0].is_shader) {
                        ctx->fragment_shader_core.blend.shader =
                                blend[0].shader.bo->gpu | blend[0].shader.first_tag;
                } else {
                        ctx->fragment_shader_core.blend.shader = 0;
                }

                if (screen->quirks & MIDGARD_SFBD) {
                        /* When only a single render target platform is used, the blend
                         * information is inside the shader meta itself. We
                         * additionally need to signal CAN_DISCARD for nontrivial blend
                         * modes (so we're able to read back the destination buffer) */

                        SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_HAS_BLEND_SHADER, blend[0].is_shader);

                        if (!blend[0].is_shader) {
                                ctx->fragment_shader_core.blend.equation =
                                        *blend[0].equation.equation;
                                ctx->fragment_shader_core.blend.constant =
                                        blend[0].equation.constant;
                        }

                        SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_CAN_DISCARD, !blend[0].no_blending);
                }

                size_t size = sizeof(struct mali_shader_meta) + (sizeof(struct midgard_blend_rt) * rt_count);
                struct panfrost_transfer transfer = panfrost_allocate_transient(batch, size);
                memcpy(transfer.cpu, &ctx->fragment_shader_core, sizeof(struct mali_shader_meta));

                ctx->payloads[PIPE_SHADER_FRAGMENT].postfix.shader = transfer.gpu;

                if (!(screen->quirks & MIDGARD_SFBD)) {
                        /* Additional blend descriptor tacked on for jobs using MFBD */

                        struct midgard_blend_rt rts[4];

                        for (unsigned i = 0; i < rt_count; ++i) {
                                rts[i].flags = 0x200;

                                bool is_srgb =
                                        (ctx->pipe_framebuffer.nr_cbufs > i) &&
                                        (ctx->pipe_framebuffer.cbufs[i]) &&
                                        util_format_is_srgb(ctx->pipe_framebuffer.cbufs[i]->format);

                                SET_BIT(rts[i].flags, MALI_BLEND_MRT_SHADER, blend[i].is_shader);
                                SET_BIT(rts[i].flags, MALI_BLEND_LOAD_TIB, !blend[i].no_blending);
                                SET_BIT(rts[i].flags, MALI_BLEND_SRGB, is_srgb);
                                SET_BIT(rts[i].flags, MALI_BLEND_NO_DITHER, !ctx->blend->base.dither);

                                /* TODO: sRGB in blend shaders is currently
                                 * unimplemented. Contact me (Alyssa) if you're
                                 * interested in working on this. We have
                                 * native Midgard ops for helping here, but
                                 * they're not well-understood yet. */

                                assert(!(is_srgb && blend[i].is_shader));

                                if (blend[i].is_shader) {
                                        rts[i].blend.shader = blend[i].shader.bo->gpu | blend[i].shader.first_tag;
                                } else {
                                        rts[i].blend.equation = *blend[i].equation.equation;
                                        rts[i].blend.constant = blend[i].equation.constant;
                                }
                        }

                        memcpy(transfer.cpu + sizeof(struct mali_shader_meta), rts, sizeof(rts[0]) * rt_count);
                }
        }

        /* We stage to transient, so always dirty.. */
        if (ctx->vertex)
                panfrost_stage_attributes(ctx);

        if (ctx->dirty & PAN_DIRTY_SAMPLERS)
                panfrost_upload_sampler_descriptors(ctx);

        if (ctx->dirty & PAN_DIRTY_TEXTURES)
                panfrost_upload_texture_descriptors(ctx);

        const struct pipe_viewport_state *vp = &ctx->pipe_viewport;

        for (int i = 0; i < PIPE_SHADER_TYPES; ++i) {
                struct panfrost_shader_variants *all = ctx->shader[i];

                if (!all)
                        continue;

                struct panfrost_constant_buffer *buf = &ctx->constant_buffer[i];

                struct panfrost_shader_state *ss = &all->variants[all->active_variant];

                /* Uniforms are implicitly UBO #0 */
                bool has_uniforms = buf->enabled_mask & (1 << 0);

                /* Allocate room for the sysval and the uniforms */
                size_t sys_size = sizeof(float) * 4 * ss->sysval_count;
                size_t uniform_size = has_uniforms ? (buf->cb[0].buffer_size) : 0;
                size_t size = sys_size + uniform_size;
                struct panfrost_transfer transfer = panfrost_allocate_transient(batch, size);

                /* Upload sysvals requested by the shader */
                panfrost_upload_sysvals(ctx, transfer.cpu, ss, i);

                /* Upload uniforms */
                if (has_uniforms) {
                        const void *cpu = panfrost_map_constant_buffer_cpu(buf, 0);
                        memcpy(transfer.cpu + sys_size, cpu, uniform_size);
                }

                int uniform_count =
                        ctx->shader[i]->variants[ctx->shader[i]->active_variant].uniform_count;

                struct mali_vertex_tiler_postfix *postfix =
                        &ctx->payloads[i].postfix;

                /* Next up, attach UBOs. UBO #0 is the uniforms we just
                 * uploaded */

                unsigned ubo_count = panfrost_ubo_count(ctx, i);
                assert(ubo_count >= 1);

                size_t sz = sizeof(struct mali_uniform_buffer_meta) * ubo_count;
                struct mali_uniform_buffer_meta ubos[PAN_MAX_CONST_BUFFERS];

                /* Upload uniforms as a UBO */
                ubos[0].size = MALI_POSITIVE((2 + uniform_count));
                ubos[0].ptr = transfer.gpu >> 2;

                /* The rest are honest-to-goodness UBOs */

                for (unsigned ubo = 1; ubo < ubo_count; ++ubo) {
                        size_t usz = buf->cb[ubo].buffer_size;

                        bool enabled = buf->enabled_mask & (1 << ubo);
                        bool empty = usz == 0;

                        if (!enabled || empty) {
                                /* Stub out disabled UBOs to catch accesses */

                                ubos[ubo].size = 0;
                                ubos[ubo].ptr = 0xDEAD0000;
                                continue;
                        }

                        mali_ptr gpu = panfrost_map_constant_buffer_gpu(ctx, i, buf, ubo);

                        unsigned bytes_per_field = 16;
                        unsigned aligned = ALIGN_POT(usz, bytes_per_field);
                        unsigned fields = aligned / bytes_per_field;

                        ubos[ubo].size = MALI_POSITIVE(fields);
                        ubos[ubo].ptr = gpu >> 2;
                }

                mali_ptr ubufs = panfrost_upload_transient(batch, ubos, sz);
                postfix->uniforms = transfer.gpu;
                postfix->uniform_buffers = ubufs;

                buf->dirty_mask = 0;
        }

        /* TODO: Upload the viewport somewhere more appropriate */

        /* Clip bounds are encoded as floats. The viewport itself is encoded as
         * (somewhat) asymmetric ints. */
        const struct pipe_scissor_state *ss = &ctx->scissor;

        struct mali_viewport view = {
                /* By default, do no viewport clipping, i.e. clip to (-inf,
                 * inf) in each direction. Clipping to the viewport in theory
                 * should work, but in practice causes issues when we're not
                 * explicitly trying to scissor */

                .clip_minx = -INFINITY,
                .clip_miny = -INFINITY,
                .clip_maxx = INFINITY,
                .clip_maxy = INFINITY,
        };

        /* Always scissor to the viewport by default. */
        float vp_minx = (int) (vp->translate[0] - fabsf(vp->scale[0]));
        float vp_maxx = (int) (vp->translate[0] + fabsf(vp->scale[0]));

        float vp_miny = (int) (vp->translate[1] - fabsf(vp->scale[1]));
        float vp_maxy = (int) (vp->translate[1] + fabsf(vp->scale[1]));

        float minz = (vp->translate[2] - fabsf(vp->scale[2]));
        float maxz = (vp->translate[2] + fabsf(vp->scale[2]));

        /* Apply the scissor test */

        unsigned minx, miny, maxx, maxy;

        if (ss && ctx->rasterizer && ctx->rasterizer->base.scissor) {
                minx = MAX2(ss->minx, vp_minx);
                miny = MAX2(ss->miny, vp_miny);
                maxx = MIN2(ss->maxx, vp_maxx);
                maxy = MIN2(ss->maxy, vp_maxy);
        } else {
                minx = vp_minx;
                miny = vp_miny;
                maxx = vp_maxx;
                maxy = vp_maxy;
        }

        /* Hardware needs the min/max to be strictly ordered, so flip if we
         * need to. The viewport transformation in the vertex shader will
         * handle the negatives if we don't */

        if (miny > maxy) {
                unsigned temp = miny;
                miny = maxy;
                maxy = temp;
        }

        if (minx > maxx) {
                unsigned temp = minx;
                minx = maxx;
                maxx = temp;
        }

        if (minz > maxz) {
                float temp = minz;
                minz = maxz;
                maxz = temp;
        }

        /* Clamp to the framebuffer size as a last check */

        minx = MIN2(ctx->pipe_framebuffer.width, minx);
        maxx = MIN2(ctx->pipe_framebuffer.width, maxx);

        miny = MIN2(ctx->pipe_framebuffer.height, miny);
        maxy = MIN2(ctx->pipe_framebuffer.height, maxy);

        /* Update the job, unless we're doing wallpapering (whose lack of
         * scissor we can ignore, since if we "miss" a tile of wallpaper, it'll
         * just... be faster :) */

        if (!ctx->wallpaper_batch)
                panfrost_batch_union_scissor(batch, minx, miny, maxx, maxy);

        /* Upload */

        view.viewport0[0] = minx;
        view.viewport1[0] = MALI_POSITIVE(maxx);

        view.viewport0[1] = miny;
        view.viewport1[1] = MALI_POSITIVE(maxy);

        view.clip_minz = minz;
        view.clip_maxz = maxz;

        ctx->payloads[PIPE_SHADER_FRAGMENT].postfix.viewport =
                panfrost_upload_transient(batch,
                                          &view,
                                          sizeof(struct mali_viewport));

        ctx->dirty = 0;
}

/* Corresponds to exactly one draw, but does not submit anything */

static void
panfrost_queue_draw(struct panfrost_context *ctx)
{
        /* Handle dirty flags now */
        panfrost_emit_for_draw(ctx, true);

        /* If rasterizer discard is enable, only submit the vertex */

        bool rasterizer_discard = ctx->rasterizer
                                  && ctx->rasterizer->base.rasterizer_discard;

        struct panfrost_transfer vertex = panfrost_vertex_tiler_job(ctx, false);
        struct panfrost_transfer tiler;

        if (!rasterizer_discard)
                tiler = panfrost_vertex_tiler_job(ctx, true);

        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        if (rasterizer_discard)
                panfrost_scoreboard_queue_vertex_job(batch, vertex, FALSE);
        else if (ctx->wallpaper_batch && batch->first_tiler.gpu)
                panfrost_scoreboard_queue_fused_job_prepend(batch, vertex, tiler);
        else
                panfrost_scoreboard_queue_fused_job(batch, vertex, tiler);
}

/* The entire frame is in memory -- send it off to the kernel! */

void
panfrost_flush(
        struct pipe_context *pipe,
        struct pipe_fence_handle **fence,
        unsigned flags)
{
        struct panfrost_context *ctx = pan_context(pipe);
        struct util_dynarray fences;

        /* We must collect the fences before the flush is done, otherwise we'll
         * lose track of them.
         */
        if (fence) {
                util_dynarray_init(&fences, NULL);
                hash_table_foreach(ctx->batches, hentry) {
                        struct panfrost_batch *batch = hentry->data;

                        panfrost_batch_fence_reference(batch->out_sync);
                        util_dynarray_append(&fences,
                                             struct panfrost_batch_fence *,
                                             batch->out_sync);
                }
        }

        /* Submit all pending jobs */
        panfrost_flush_all_batches(ctx, false);

        if (fence) {
                struct panfrost_fence *f = panfrost_fence_create(ctx, &fences);
                pipe->screen->fence_reference(pipe->screen, fence, NULL);
                *fence = (struct pipe_fence_handle *)f;

                util_dynarray_foreach(&fences, struct panfrost_batch_fence *, fence)
                        panfrost_batch_fence_unreference(*fence);

                util_dynarray_fini(&fences);
        }
}

#define DEFINE_CASE(c) case PIPE_PRIM_##c: return MALI_##c;

static int
g2m_draw_mode(enum pipe_prim_type mode)
{
        switch (mode) {
                DEFINE_CASE(POINTS);
                DEFINE_CASE(LINES);
                DEFINE_CASE(LINE_LOOP);
                DEFINE_CASE(LINE_STRIP);
                DEFINE_CASE(TRIANGLES);
                DEFINE_CASE(TRIANGLE_STRIP);
                DEFINE_CASE(TRIANGLE_FAN);
                DEFINE_CASE(QUADS);
                DEFINE_CASE(QUAD_STRIP);
                DEFINE_CASE(POLYGON);

        default:
                unreachable("Invalid draw mode");
        }
}

#undef DEFINE_CASE

static unsigned
panfrost_translate_index_size(unsigned size)
{
        switch (size) {
        case 1:
                return MALI_DRAW_INDEXED_UINT8;

        case 2:
                return MALI_DRAW_INDEXED_UINT16;

        case 4:
                return MALI_DRAW_INDEXED_UINT32;

        default:
                unreachable("Invalid index size");
        }
}

/* Gets a GPU address for the associated index buffer. Only gauranteed to be
 * good for the duration of the draw (transient), could last longer */

static mali_ptr
panfrost_get_index_buffer_mapped(struct panfrost_context *ctx, const struct pipe_draw_info *info)
{
        struct panfrost_resource *rsrc = (struct panfrost_resource *) (info->index.resource);

        off_t offset = info->start * info->index_size;
        struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);

        if (!info->has_user_indices) {
                /* Only resources can be directly mapped */
                panfrost_batch_add_bo(batch, rsrc->bo,
                                      PAN_BO_ACCESS_SHARED |
                                      PAN_BO_ACCESS_READ |
                                      PAN_BO_ACCESS_VERTEX_TILER);
                return rsrc->bo->gpu + offset;
        } else {
                /* Otherwise, we need to upload to transient memory */
                const uint8_t *ibuf8 = (const uint8_t *) info->index.user;
                return panfrost_upload_transient(batch, ibuf8 + offset, info->count * info->index_size);
        }
}

static bool
panfrost_scissor_culls_everything(struct panfrost_context *ctx)
{
        const struct pipe_scissor_state *ss = &ctx->scissor;

        /* Check if we're scissoring at all */

        if (!(ctx->rasterizer && ctx->rasterizer->base.scissor))
                return false;

        return (ss->minx == ss->maxx) || (ss->miny == ss->maxy);
}

/* Count generated primitives (when there is no geom/tess shaders) for
 * transform feedback */

static void
panfrost_statistics_record(
                struct panfrost_context *ctx,
                const struct pipe_draw_info *info)
{
        if (!ctx->active_queries)
                return;

        uint32_t prims = u_prims_for_vertices(info->mode, info->count);
        ctx->prims_generated += prims;

        if (!ctx->streamout.num_targets)
                return;

        ctx->tf_prims_generated += prims;
}

static void
panfrost_draw_vbo(
        struct pipe_context *pipe,
        const struct pipe_draw_info *info)
{
        struct panfrost_context *ctx = pan_context(pipe);

        /* First of all, check the scissor to see if anything is drawn at all.
         * If it's not, we drop the draw (mostly a conformance issue;
         * well-behaved apps shouldn't hit this) */

        if (panfrost_scissor_culls_everything(ctx))
                return;

        int mode = info->mode;

        /* Fallback unsupported restart index */
        unsigned primitive_index = (1 << (info->index_size * 8)) - 1;

        if (info->primitive_restart && info->index_size
            && info->restart_index != primitive_index) {
                util_draw_vbo_without_prim_restart(pipe, info);
                return;
        }

        /* Fallback for unsupported modes */

        assert(ctx->rasterizer != NULL);

        if (!(ctx->draw_modes & (1 << mode))) {
                if (mode == PIPE_PRIM_QUADS && info->count == 4 && !ctx->rasterizer->base.flatshade) {
                        mode = PIPE_PRIM_TRIANGLE_FAN;
                } else {
                        if (info->count < 4) {
                                /* Degenerate case? */
                                return;
                        }

                        util_primconvert_save_rasterizer_state(ctx->primconvert, &ctx->rasterizer->base);
                        util_primconvert_draw_vbo(ctx->primconvert, info);
                        return;
                }
        }

        ctx->payloads[PIPE_SHADER_VERTEX].offset_start = info->start;
        ctx->payloads[PIPE_SHADER_FRAGMENT].offset_start = info->start;

        /* Now that we have a guaranteed terminating path, find the job.
         * Assignment commented out to prevent unused warning */

        /* struct panfrost_batch *batch = */ panfrost_get_batch_for_fbo(ctx);

        ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.draw_mode = g2m_draw_mode(mode);

        /* Take into account a negative bias */
        ctx->vertex_count = info->count + abs(info->index_bias);
        ctx->instance_count = info->instance_count;
        ctx->active_prim = info->mode;

        /* For non-indexed draws, they're the same */
        unsigned vertex_count = ctx->vertex_count;

        unsigned draw_flags = 0;

        /* The draw flags interpret how primitive size is interpreted */

        if (panfrost_writes_point_size(ctx))
                draw_flags |= MALI_DRAW_VARYING_SIZE;

        if (info->primitive_restart)
                draw_flags |= MALI_DRAW_PRIMITIVE_RESTART_FIXED_INDEX;

        /* For higher amounts of vertices (greater than what fits in a 16-bit
         * short), the other value is needed, otherwise there will be bizarre
         * rendering artefacts. It's not clear what these values mean yet. This
         * change is also needed for instancing and sometimes points (perhaps
         * related to dynamically setting gl_PointSize) */

        bool is_points = mode == PIPE_PRIM_POINTS;
        bool many_verts = ctx->vertex_count > 0xFFFF;
        bool instanced = ctx->instance_count > 1;

        draw_flags |= (is_points || many_verts || instanced) ? 0x3000 : 0x18000;

        /* This doesn't make much sense */
        if (mode == PIPE_PRIM_LINE_STRIP) {
                draw_flags |= 0x800;
        }

        panfrost_statistics_record(ctx, info);

        if (info->index_size) {
                /* Calculate the min/max index used so we can figure out how
                 * many times to invoke the vertex shader */

                /* Fetch / calculate index bounds */
                unsigned min_index = 0, max_index = 0;

                if (info->max_index == ~0u) {
                        u_vbuf_get_minmax_index(pipe, info, &min_index, &max_index);
                } else {
                        min_index = info->min_index;
                        max_index = info->max_index;
                }

                /* Use the corresponding values */
                vertex_count = max_index - min_index + 1;
                ctx->payloads[PIPE_SHADER_VERTEX].offset_start = min_index + info->index_bias;
                ctx->payloads[PIPE_SHADER_FRAGMENT].offset_start = min_index + info->index_bias;

                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.offset_bias_correction = -min_index;
                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.index_count = MALI_POSITIVE(info->count);

                //assert(!info->restart_index); /* TODO: Research */

                draw_flags |= panfrost_translate_index_size(info->index_size);
                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.indices = panfrost_get_index_buffer_mapped(ctx, info);
        } else {
                /* Index count == vertex count, if no indexing is applied, as
                 * if it is internally indexed in the expected order */

                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.offset_bias_correction = 0;
                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.index_count = MALI_POSITIVE(ctx->vertex_count);

                /* Reverse index state */
                ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.indices = (u64) NULL;
        }

        /* Dispatch "compute jobs" for the vertex/tiler pair as (1,
         * vertex_count, 1) */

        panfrost_pack_work_groups_fused(
                &ctx->payloads[PIPE_SHADER_VERTEX].prefix,
                &ctx->payloads[PIPE_SHADER_FRAGMENT].prefix,
                1, vertex_count, info->instance_count,
                1, 1, 1);

        ctx->payloads[PIPE_SHADER_FRAGMENT].prefix.unknown_draw = draw_flags;

        /* Encode the padded vertex count */

        if (info->instance_count > 1) {
                /* Triangles have non-even vertex counts so they change how
                 * padding works internally */

                bool is_triangle =
                        mode == PIPE_PRIM_TRIANGLES ||
                        mode == PIPE_PRIM_TRIANGLE_STRIP ||
                        mode == PIPE_PRIM_TRIANGLE_FAN;

                struct pan_shift_odd so =
                        panfrost_padded_vertex_count(vertex_count, !is_triangle);

                ctx->payloads[PIPE_SHADER_VERTEX].instance_shift = so.shift;
                ctx->payloads[PIPE_SHADER_FRAGMENT].instance_shift = so.shift;

                ctx->payloads[PIPE_SHADER_VERTEX].instance_odd = so.odd;
                ctx->payloads[PIPE_SHADER_FRAGMENT].instance_odd = so.odd;

                ctx->padded_count = pan_expand_shift_odd(so);
        } else {
                ctx->padded_count = vertex_count;

                /* Reset instancing state */
                ctx->payloads[PIPE_SHADER_VERTEX].instance_shift = 0;
                ctx->payloads[PIPE_SHADER_VERTEX].instance_odd = 0;
                ctx->payloads[PIPE_SHADER_FRAGMENT].instance_shift = 0;
                ctx->payloads[PIPE_SHADER_FRAGMENT].instance_odd = 0;
        }

        /* Fire off the draw itself */
        panfrost_queue_draw(ctx);

        /* Increment transform feedback offsets */

        for (unsigned i = 0; i < ctx->streamout.num_targets; ++i) {
                unsigned output_count = u_stream_outputs_for_vertices(
                                ctx->active_prim, ctx->vertex_count);

                ctx->streamout.offsets[i] += output_count;
        }
}

/* CSO state */

static void
panfrost_generic_cso_delete(struct pipe_context *pctx, void *hwcso)
{
        free(hwcso);
}

static void *
panfrost_create_rasterizer_state(
        struct pipe_context *pctx,
        const struct pipe_rasterizer_state *cso)
{
        struct panfrost_rasterizer *so = CALLOC_STRUCT(panfrost_rasterizer);

        so->base = *cso;

        /* Bitmask, unknown meaning of the start value. 0x105 on 32-bit T6XX */
        so->tiler_gl_enables = 0x7;

        if (cso->front_ccw)
                so->tiler_gl_enables |= MALI_FRONT_CCW_TOP;

        if (cso->cull_face & PIPE_FACE_FRONT)
                so->tiler_gl_enables |= MALI_CULL_FACE_FRONT;

        if (cso->cull_face & PIPE_FACE_BACK)
                so->tiler_gl_enables |= MALI_CULL_FACE_BACK;

        return so;
}

static void
panfrost_bind_rasterizer_state(
        struct pipe_context *pctx,
        void *hwcso)
{
        struct panfrost_context *ctx = pan_context(pctx);

        /* TODO: Why can't rasterizer be NULL ever? Other drivers are fine.. */
        if (!hwcso)
                return;

        ctx->rasterizer = hwcso;
        ctx->dirty |= PAN_DIRTY_RASTERIZER;

        ctx->fragment_shader_core.depth_units = ctx->rasterizer->base.offset_units * 2.0f;
        ctx->fragment_shader_core.depth_factor = ctx->rasterizer->base.offset_scale;

        /* Gauranteed with the core GL call, so don't expose ARB_polygon_offset */
        assert(ctx->rasterizer->base.offset_clamp == 0.0);

        /* XXX: Which bit is which? Does this maybe allow offseting not-tri? */

        SET_BIT(ctx->fragment_shader_core.unknown2_4, MALI_DEPTH_RANGE_A, ctx->rasterizer->base.offset_tri);
        SET_BIT(ctx->fragment_shader_core.unknown2_4, MALI_DEPTH_RANGE_B, ctx->rasterizer->base.offset_tri);

        /* Point sprites are emulated */

        struct panfrost_shader_state *variant =
                        ctx->shader[PIPE_SHADER_FRAGMENT] ? &ctx->shader[PIPE_SHADER_FRAGMENT]->variants[ctx->shader[PIPE_SHADER_FRAGMENT]->active_variant] : NULL;

        if (ctx->rasterizer->base.sprite_coord_enable || (variant && variant->point_sprite_mask))
                ctx->base.bind_fs_state(&ctx->base, ctx->shader[PIPE_SHADER_FRAGMENT]);
}

static void *
panfrost_create_vertex_elements_state(
        struct pipe_context *pctx,
        unsigned num_elements,
        const struct pipe_vertex_element *elements)
{
        struct panfrost_vertex_state *so = CALLOC_STRUCT(panfrost_vertex_state);

        so->num_elements = num_elements;
        memcpy(so->pipe, elements, sizeof(*elements) * num_elements);

        for (int i = 0; i < num_elements; ++i) {
                so->hw[i].index = i;

                enum pipe_format fmt = elements[i].src_format;
                const struct util_format_description *desc = util_format_description(fmt);
                so->hw[i].unknown1 = 0x2;
                so->hw[i].swizzle = panfrost_get_default_swizzle(desc->nr_channels);

                so->hw[i].format = panfrost_find_format(desc);

                /* The field itself should probably be shifted over */
                so->hw[i].src_offset = elements[i].src_offset;
        }

        return so;
}

static void
panfrost_bind_vertex_elements_state(
        struct pipe_context *pctx,
        void *hwcso)
{
        struct panfrost_context *ctx = pan_context(pctx);

        ctx->vertex = hwcso;
        ctx->dirty |= PAN_DIRTY_VERTEX;
}

static void *
panfrost_create_shader_state(
        struct pipe_context *pctx,
        const struct pipe_shader_state *cso)
{
        struct panfrost_shader_variants *so = CALLOC_STRUCT(panfrost_shader_variants);
        so->base = *cso;

        /* Token deep copy to prevent memory corruption */

        if (cso->type == PIPE_SHADER_IR_TGSI)
                so->base.tokens = tgsi_dup_tokens(so->base.tokens);

        return so;
}

static void
panfrost_delete_shader_state(
        struct pipe_context *pctx,
        void *so)
{
        struct panfrost_shader_variants *cso = (struct panfrost_shader_variants *) so;

        if (cso->base.type == PIPE_SHADER_IR_TGSI) {
                DBG("Deleting TGSI shader leaks duplicated tokens\n");
        }

        for (unsigned i = 0; i < cso->variant_count; ++i) {
                struct panfrost_shader_state *shader_state = &cso->variants[i];
                panfrost_bo_unreference(shader_state->bo);
                shader_state->bo = NULL;
        }

        free(so);
}

static void *
panfrost_create_sampler_state(
        struct pipe_context *pctx,
        const struct pipe_sampler_state *cso)
{
        struct panfrost_sampler_state *so = CALLOC_STRUCT(panfrost_sampler_state);
        so->base = *cso;

        /* sampler_state corresponds to mali_sampler_descriptor, which we can generate entirely here */

        bool min_nearest = cso->min_img_filter == PIPE_TEX_FILTER_NEAREST;
        bool mag_nearest = cso->mag_img_filter == PIPE_TEX_FILTER_NEAREST;
        bool mip_linear  = cso->min_mip_filter == PIPE_TEX_MIPFILTER_LINEAR;

        unsigned min_filter = min_nearest ? MALI_SAMP_MIN_NEAREST : 0;
        unsigned mag_filter = mag_nearest ? MALI_SAMP_MAG_NEAREST : 0;
        unsigned mip_filter = mip_linear  ?
                (MALI_SAMP_MIP_LINEAR_1 | MALI_SAMP_MIP_LINEAR_2) : 0;
        unsigned normalized = cso->normalized_coords ? MALI_SAMP_NORM_COORDS : 0;

        struct mali_sampler_descriptor sampler_descriptor = {
                .filter_mode = min_filter | mag_filter | mip_filter | normalized,
                .wrap_s = translate_tex_wrap(cso->wrap_s),
                .wrap_t = translate_tex_wrap(cso->wrap_t),
                .wrap_r = translate_tex_wrap(cso->wrap_r),
                .compare_func = panfrost_translate_alt_compare_func(cso->compare_func),
                .border_color = {
                        cso->border_color.f[0],
                        cso->border_color.f[1],
                        cso->border_color.f[2],
                        cso->border_color.f[3]
                },
                .min_lod = FIXED_16(cso->min_lod),
                .max_lod = FIXED_16(cso->max_lod),
                .lod_bias = FIXED_16(cso->lod_bias),
                .seamless_cube_map = cso->seamless_cube_map,
        };

        /* If necessary, we disable mipmapping in the sampler descriptor by
         * clamping the LOD as tight as possible (from 0 to epsilon,
         * essentially -- remember these are fixed point numbers, so
         * epsilon=1/256) */

        if (cso->min_mip_filter == PIPE_TEX_MIPFILTER_NONE)
                sampler_descriptor.max_lod = sampler_descriptor.min_lod;

        /* Enforce that there is something in the middle by adding epsilon*/

        if (sampler_descriptor.min_lod == sampler_descriptor.max_lod)
                sampler_descriptor.max_lod++;

        /* Sanity check */
        assert(sampler_descriptor.max_lod > sampler_descriptor.min_lod);

        so->hw = sampler_descriptor;

        return so;
}

static void
panfrost_bind_sampler_states(
        struct pipe_context *pctx,
        enum pipe_shader_type shader,
        unsigned start_slot, unsigned num_sampler,
        void **sampler)
{
        assert(start_slot == 0);

        struct panfrost_context *ctx = pan_context(pctx);

        /* XXX: Should upload, not just copy? */
        ctx->sampler_count[shader] = num_sampler;
        memcpy(ctx->samplers[shader], sampler, num_sampler * sizeof (void *));

        ctx->dirty |= PAN_DIRTY_SAMPLERS;
}

static bool
panfrost_variant_matches(
        struct panfrost_context *ctx,
        struct panfrost_shader_state *variant,
        enum pipe_shader_type type)
{
        struct pipe_rasterizer_state *rasterizer = &ctx->rasterizer->base;
        struct pipe_alpha_state *alpha = &ctx->depth_stencil->alpha;

        bool is_fragment = (type == PIPE_SHADER_FRAGMENT);

        if (is_fragment && (alpha->enabled || variant->alpha_state.enabled)) {
                /* Make sure enable state is at least the same */
                if (alpha->enabled != variant->alpha_state.enabled) {
                        return false;
                }

                /* Check that the contents of the test are the same */
                bool same_func = alpha->func == variant->alpha_state.func;
                bool same_ref = alpha->ref_value == variant->alpha_state.ref_value;

                if (!(same_func && same_ref)) {
                        return false;
                }
        }

        if (is_fragment && rasterizer && (rasterizer->sprite_coord_enable |
                                          variant->point_sprite_mask)) {
                /* Ensure the same varyings are turned to point sprites */
                if (rasterizer->sprite_coord_enable != variant->point_sprite_mask)
                        return false;

                /* Ensure the orientation is correct */
                bool upper_left =
                        rasterizer->sprite_coord_mode ==
                        PIPE_SPRITE_COORD_UPPER_LEFT;

                if (variant->point_sprite_upper_left != upper_left)
                        return false;
        }

        /* Otherwise, we're good to go */
        return true;
}

/**
 * Fix an uncompiled shader's stream output info, and produce a bitmask
 * of which VARYING_SLOT_* are captured for stream output.
 *
 * Core Gallium stores output->register_index as a "slot" number, where
 * slots are assigned consecutively to all outputs in info->outputs_written.
 * This naive packing of outputs doesn't work for us - we too have slots,
 * but the layout is defined by the VUE map, which we won't have until we
 * compile a specific shader variant.  So, we remap these and simply store
 * VARYING_SLOT_* in our copy's output->register_index fields.
 *
 * We then produce a bitmask of outputs which are used for SO.
 *
 * Implementation from iris.
 */

static uint64_t
update_so_info(struct pipe_stream_output_info *so_info,
               uint64_t outputs_written)
{
	uint64_t so_outputs = 0;
	uint8_t reverse_map[64] = {0};
	unsigned slot = 0;

	while (outputs_written)
		reverse_map[slot++] = u_bit_scan64(&outputs_written);

	for (unsigned i = 0; i < so_info->num_outputs; i++) {
		struct pipe_stream_output *output = &so_info->output[i];

		/* Map Gallium's condensed "slots" back to real VARYING_SLOT_* enums */
		output->register_index = reverse_map[output->register_index];

		so_outputs |= 1ull << output->register_index;
	}

	return so_outputs;
}

static void
panfrost_bind_shader_state(
        struct pipe_context *pctx,
        void *hwcso,
        enum pipe_shader_type type)
{
        struct panfrost_context *ctx = pan_context(pctx);

        ctx->shader[type] = hwcso;

        if (type == PIPE_SHADER_FRAGMENT)
                ctx->dirty |= PAN_DIRTY_FS;
        else
                ctx->dirty |= PAN_DIRTY_VS;

        if (!hwcso) return;

        /* Match the appropriate variant */

        signed variant = -1;
        struct panfrost_shader_variants *variants = (struct panfrost_shader_variants *) hwcso;

        for (unsigned i = 0; i < variants->variant_count; ++i) {
                if (panfrost_variant_matches(ctx, &variants->variants[i], type)) {
                        variant = i;
                        break;
                }
        }

        if (variant == -1) {
                /* No variant matched, so create a new one */
                variant = variants->variant_count++;
                assert(variants->variant_count < MAX_SHADER_VARIANTS);

                struct panfrost_shader_state *v =
                                &variants->variants[variant];

                if (type == PIPE_SHADER_FRAGMENT) {
                        v->alpha_state = ctx->depth_stencil->alpha;

                        if (ctx->rasterizer) {
                                v->point_sprite_mask = ctx->rasterizer->base.sprite_coord_enable;
                                v->point_sprite_upper_left =
                                        ctx->rasterizer->base.sprite_coord_mode ==
                                        PIPE_SPRITE_COORD_UPPER_LEFT;
                        }
                }

                variants->variants[variant].tripipe = calloc(1, sizeof(struct mali_shader_meta));

        }

        /* Select this variant */
        variants->active_variant = variant;

        struct panfrost_shader_state *shader_state = &variants->variants[variant];
        assert(panfrost_variant_matches(ctx, shader_state, type));

        /* We finally have a variant, so compile it */

        if (!shader_state->compiled) {
                uint64_t outputs_written = 0;

                panfrost_shader_compile(ctx, shader_state->tripipe,
                              variants->base.type,
                              variants->base.type == PIPE_SHADER_IR_NIR ?
                                      variants->base.ir.nir :
                                      variants->base.tokens,
                                        tgsi_processor_to_shader_stage(type), shader_state,
                                        &outputs_written);

                shader_state->compiled = true;

                /* Fixup the stream out information, since what Gallium returns
                 * normally is mildly insane */

                shader_state->stream_output = variants->base.stream_output;
                shader_state->so_mask =
                        update_so_info(&shader_state->stream_output, outputs_written);
        }
}

static void
panfrost_bind_vs_state(struct pipe_context *pctx, void *hwcso)
{
        panfrost_bind_shader_state(pctx, hwcso, PIPE_SHADER_VERTEX);
}

static void
panfrost_bind_fs_state(struct pipe_context *pctx, void *hwcso)
{
        panfrost_bind_shader_state(pctx, hwcso, PIPE_SHADER_FRAGMENT);
}

static void
panfrost_set_vertex_buffers(
        struct pipe_context *pctx,
        unsigned start_slot,
        unsigned num_buffers,
        const struct pipe_vertex_buffer *buffers)
{
        struct panfrost_context *ctx = pan_context(pctx);

        util_set_vertex_buffers_mask(ctx->vertex_buffers, &ctx->vb_mask, buffers, start_slot, num_buffers);
}

static void
panfrost_set_constant_buffer(
        struct pipe_context *pctx,
        enum pipe_shader_type shader, uint index,
        const struct pipe_constant_buffer *buf)
{
        struct panfrost_context *ctx = pan_context(pctx);
        struct panfrost_constant_buffer *pbuf = &ctx->constant_buffer[shader];

        util_copy_constant_buffer(&pbuf->cb[index], buf);

        unsigned mask = (1 << index);

        if (unlikely(!buf)) {
                pbuf->enabled_mask &= ~mask;
                pbuf->dirty_mask &= ~mask;
                return;
        }

        pbuf->enabled_mask |= mask;
        pbuf->dirty_mask |= mask;
}

static void
panfrost_set_stencil_ref(
        struct pipe_context *pctx,
        const struct pipe_stencil_ref *ref)
{
        struct panfrost_context *ctx = pan_context(pctx);
        ctx->stencil_ref = *ref;

        /* Shader core dirty */
        ctx->dirty |= PAN_DIRTY_FS;
}

static enum mali_texture_type
panfrost_translate_texture_type(enum pipe_texture_target t) {
        switch (t)
        {
        case PIPE_BUFFER:
        case PIPE_TEXTURE_1D:
        case PIPE_TEXTURE_1D_ARRAY:
                return MALI_TEX_1D;

        case PIPE_TEXTURE_2D:
        case PIPE_TEXTURE_2D_ARRAY:
        case PIPE_TEXTURE_RECT:
                return MALI_TEX_2D;

        case PIPE_TEXTURE_3D:
                return MALI_TEX_3D;

        case PIPE_TEXTURE_CUBE:
        case PIPE_TEXTURE_CUBE_ARRAY:
                return MALI_TEX_CUBE;

        default:
                unreachable("Unknown target");
        }
}

static struct pipe_sampler_view *
panfrost_create_sampler_view(
        struct pipe_context *pctx,
        struct pipe_resource *texture,
        const struct pipe_sampler_view *template)
{
        struct panfrost_sampler_view *so = rzalloc(pctx, struct panfrost_sampler_view);
        int bytes_per_pixel = util_format_get_blocksize(texture->format);

        pipe_reference(NULL, &texture->reference);

        struct panfrost_resource *prsrc = (struct panfrost_resource *) texture;
        assert(prsrc->bo);

        so->base = *template;
        so->base.texture = texture;
        so->base.reference.count = 1;
        so->base.context = pctx;

        /* sampler_views correspond to texture descriptors, minus the texture
         * (data) itself. So, we serialise the descriptor here and cache it for
         * later. */

        const struct util_format_description *desc = util_format_description(prsrc->base.format);

        unsigned char user_swizzle[4] = {
                template->swizzle_r,
                template->swizzle_g,
                template->swizzle_b,
                template->swizzle_a
        };

        enum mali_format format = panfrost_find_format(desc);

        /* Check if we need to set a custom stride by computing the "expected"
         * stride and comparing it to what the BO actually wants. Only applies
         * to linear textures, since tiled/compressed textures have strict
         * alignment requirements for their strides as it is */

        unsigned first_level = template->u.tex.first_level;
        unsigned last_level = template->u.tex.last_level;

        if (prsrc->layout == PAN_LINEAR) {
                for (unsigned l = first_level; l <= last_level; ++l) {
                        unsigned actual_stride = prsrc->slices[l].stride;
                        unsigned width = u_minify(texture->width0, l);
                        unsigned comp_stride = width * bytes_per_pixel;

                        if (comp_stride != actual_stride) {
                                so->manual_stride = true;
                                break;
                        }
                }
        }

        /* In the hardware, array_size refers specifically to array textures,
         * whereas in Gallium, it also covers cubemaps */

        unsigned array_size = texture->array_size;

        if (template->target == PIPE_TEXTURE_CUBE) {
                /* TODO: Cubemap arrays */
                assert(array_size == 6);
                array_size /= 6;
        }

        struct mali_texture_descriptor texture_descriptor = {
                .width = MALI_POSITIVE(u_minify(texture->width0, first_level)),
                .height = MALI_POSITIVE(u_minify(texture->height0, first_level)),
                .depth = MALI_POSITIVE(u_minify(texture->depth0, first_level)),
                .array_size = MALI_POSITIVE(array_size),

                .format = {
                        .swizzle = panfrost_translate_swizzle_4(desc->swizzle),
                        .format = format,
                        .srgb = desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB,
                        .type = panfrost_translate_texture_type(template->target),
                        .unknown2 = 0x1,
                },

                .swizzle = panfrost_translate_swizzle_4(user_swizzle)
        };

        texture_descriptor.levels = last_level - first_level;

        so->hw = texture_descriptor;

        return (struct pipe_sampler_view *) so;
}

static void
panfrost_set_sampler_views(
        struct pipe_context *pctx,
        enum pipe_shader_type shader,
        unsigned start_slot, unsigned num_views,
        struct pipe_sampler_view **views)
{
        struct panfrost_context *ctx = pan_context(pctx);

        assert(start_slot == 0);

        unsigned new_nr = 0;
        for (unsigned i = 0; i < num_views; ++i) {
                if (views[i])
                        new_nr = i + 1;
        }

        ctx->sampler_view_count[shader] = new_nr;
        memcpy(ctx->sampler_views[shader], views, num_views * sizeof (void *));

        ctx->dirty |= PAN_DIRTY_TEXTURES;
}

static void
panfrost_sampler_view_destroy(
        struct pipe_context *pctx,
        struct pipe_sampler_view *view)
{
        pipe_resource_reference(&view->texture, NULL);
        ralloc_free(view);
}

static void
panfrost_set_shader_buffers(
        struct pipe_context *pctx,
        enum pipe_shader_type shader,
        unsigned start, unsigned count,
        const struct pipe_shader_buffer *buffers,
        unsigned writable_bitmask)
{
        struct panfrost_context *ctx = pan_context(pctx);

        util_set_shader_buffers_mask(ctx->ssbo[shader], &ctx->ssbo_mask[shader],
                        buffers, start, count);
}

/* Hints that a framebuffer should use AFBC where possible */

static void
panfrost_hint_afbc(
                struct panfrost_screen *screen,
                const struct pipe_framebuffer_state *fb)
{
        /* AFBC implemenation incomplete; hide it */
        if (!(pan_debug & PAN_DBG_AFBC)) return;

        /* Hint AFBC to the resources bound to each color buffer */

        for (unsigned i = 0; i < fb->nr_cbufs; ++i) {
                struct pipe_surface *surf = fb->cbufs[i];
                struct panfrost_resource *rsrc = pan_resource(surf->texture);
                panfrost_resource_hint_layout(screen, rsrc, PAN_AFBC, 1);
        }

        /* Also hint it to the depth buffer */

        if (fb->zsbuf) {
                struct panfrost_resource *rsrc = pan_resource(fb->zsbuf->texture);
                panfrost_resource_hint_layout(screen, rsrc, PAN_AFBC, 1);
        }
}

static void
panfrost_set_framebuffer_state(struct pipe_context *pctx,
                               const struct pipe_framebuffer_state *fb)
{
        struct panfrost_context *ctx = pan_context(pctx);

        panfrost_hint_afbc(pan_screen(pctx->screen), fb);
        util_copy_framebuffer_state(&ctx->pipe_framebuffer, fb);
        ctx->batch = NULL;
        panfrost_invalidate_frame(ctx);
}

static void *
panfrost_create_depth_stencil_state(struct pipe_context *pipe,
                                    const struct pipe_depth_stencil_alpha_state *depth_stencil)
{
        return mem_dup(depth_stencil, sizeof(*depth_stencil));
}

static void
panfrost_bind_depth_stencil_state(struct pipe_context *pipe,
                                  void *cso)
{
        struct panfrost_context *ctx = pan_context(pipe);
        struct pipe_depth_stencil_alpha_state *depth_stencil = cso;
        ctx->depth_stencil = depth_stencil;

        if (!depth_stencil)
                return;

        /* Alpha does not exist in the hardware (it's not in ES3), so it's
         * emulated in the fragment shader */

        if (depth_stencil->alpha.enabled) {
                /* We need to trigger a new shader (maybe) */
                ctx->base.bind_fs_state(&ctx->base, ctx->shader[PIPE_SHADER_FRAGMENT]);
        }

        /* Stencil state */
        SET_BIT(ctx->fragment_shader_core.unknown2_4, MALI_STENCIL_TEST, depth_stencil->stencil[0].enabled);

        panfrost_make_stencil_state(&depth_stencil->stencil[0], &ctx->fragment_shader_core.stencil_front);
        ctx->fragment_shader_core.stencil_mask_front = depth_stencil->stencil[0].writemask;

        /* If back-stencil is not enabled, use the front values */
        bool back_enab = ctx->depth_stencil->stencil[1].enabled;
        unsigned back_index = back_enab ? 1 : 0;

        panfrost_make_stencil_state(&depth_stencil->stencil[back_index], &ctx->fragment_shader_core.stencil_back);
        ctx->fragment_shader_core.stencil_mask_back = depth_stencil->stencil[back_index].writemask;

        /* Depth state (TODO: Refactor) */
        SET_BIT(ctx->fragment_shader_core.unknown2_3, MALI_DEPTH_WRITEMASK,
                depth_stencil->depth.writemask);

        int func = depth_stencil->depth.enabled ? depth_stencil->depth.func : PIPE_FUNC_ALWAYS;

        ctx->fragment_shader_core.unknown2_3 &= ~MALI_DEPTH_FUNC_MASK;
        ctx->fragment_shader_core.unknown2_3 |= MALI_DEPTH_FUNC(panfrost_translate_compare_func(func));

        /* Bounds test not implemented */
        assert(!depth_stencil->depth.bounds_test);

        ctx->dirty |= PAN_DIRTY_FS;
}

static void
panfrost_delete_depth_stencil_state(struct pipe_context *pipe, void *depth)
{
        free( depth );
}

static void
panfrost_set_sample_mask(struct pipe_context *pipe,
                         unsigned sample_mask)
{
}

static void
panfrost_set_clip_state(struct pipe_context *pipe,
                        const struct pipe_clip_state *clip)
{
        //struct panfrost_context *panfrost = pan_context(pipe);
}

static void
panfrost_set_viewport_states(struct pipe_context *pipe,
                             unsigned start_slot,
                             unsigned num_viewports,
                             const struct pipe_viewport_state *viewports)
{
        struct panfrost_context *ctx = pan_context(pipe);

        assert(start_slot == 0);
        assert(num_viewports == 1);

        ctx->pipe_viewport = *viewports;
}

static void
panfrost_set_scissor_states(struct pipe_context *pipe,
                            unsigned start_slot,
                            unsigned num_scissors,
                            const struct pipe_scissor_state *scissors)
{
        struct panfrost_context *ctx = pan_context(pipe);

        assert(start_slot == 0);
        assert(num_scissors == 1);

        ctx->scissor = *scissors;
}

static void
panfrost_set_polygon_stipple(struct pipe_context *pipe,
                             const struct pipe_poly_stipple *stipple)
{
        //struct panfrost_context *panfrost = pan_context(pipe);
}

static void
panfrost_set_active_query_state(struct pipe_context *pipe,
                                bool enable)
{
        struct panfrost_context *ctx = pan_context(pipe);
        ctx->active_queries = enable;
}

static void
panfrost_destroy(struct pipe_context *pipe)
{
        struct panfrost_context *panfrost = pan_context(pipe);

        if (panfrost->blitter)
                util_blitter_destroy(panfrost->blitter);

        if (panfrost->blitter_wallpaper)
                util_blitter_destroy(panfrost->blitter_wallpaper);

        util_unreference_framebuffer_state(&panfrost->pipe_framebuffer);
        u_upload_destroy(pipe->stream_uploader);

        ralloc_free(pipe);
}

static struct pipe_query *
panfrost_create_query(struct pipe_context *pipe,
                      unsigned type,
                      unsigned index)
{
        struct panfrost_query *q = rzalloc(pipe, struct panfrost_query);

        q->type = type;
        q->index = index;

        return (struct pipe_query *) q;
}

static void
panfrost_destroy_query(struct pipe_context *pipe, struct pipe_query *q)
{
        struct panfrost_query *query = (struct panfrost_query *) q;

        if (query->bo) {
                panfrost_bo_unreference(query->bo);
                query->bo = NULL;
        }

        ralloc_free(q);
}

static bool
panfrost_begin_query(struct pipe_context *pipe, struct pipe_query *q)
{
        struct panfrost_context *ctx = pan_context(pipe);
        struct panfrost_query *query = (struct panfrost_query *) q;

        switch (query->type) {
        case PIPE_QUERY_OCCLUSION_COUNTER:
        case PIPE_QUERY_OCCLUSION_PREDICATE:
        case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
                /* Allocate a bo for the query results to be stored */
                if (!query->bo) {
                        query->bo = panfrost_bo_create(
                                        pan_screen(ctx->base.screen),
                                        sizeof(unsigned), 0);
                }

                unsigned *result = (unsigned *)query->bo->cpu;
                *result = 0; /* Default to 0 if nothing at all drawn. */
                ctx->occlusion_query = query;
                break;

        /* Geometry statistics are computed in the driver. XXX: geom/tess
         * shaders.. */

        case PIPE_QUERY_PRIMITIVES_GENERATED:
                query->start = ctx->prims_generated;
                break;
        case PIPE_QUERY_PRIMITIVES_EMITTED:
                query->start = ctx->tf_prims_generated;
                break;

        default:
                fprintf(stderr, "Skipping query %u\n", query->type);
                break;
        }

        return true;
}

static bool
panfrost_end_query(struct pipe_context *pipe, struct pipe_query *q)
{
        struct panfrost_context *ctx = pan_context(pipe);
        struct panfrost_query *query = (struct panfrost_query *) q;

        switch (query->type) {
        case PIPE_QUERY_OCCLUSION_COUNTER:
        case PIPE_QUERY_OCCLUSION_PREDICATE:
        case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
                ctx->occlusion_query = NULL;
                break;
        case PIPE_QUERY_PRIMITIVES_GENERATED:
                query->end = ctx->prims_generated;
                break;
        case PIPE_QUERY_PRIMITIVES_EMITTED:
                query->end = ctx->tf_prims_generated;
                break;
        }

        return true;
}

static bool
panfrost_get_query_result(struct pipe_context *pipe,
                          struct pipe_query *q,
                          bool wait,
                          union pipe_query_result *vresult)
{
        struct panfrost_query *query = (struct panfrost_query *) q;
        struct panfrost_context *ctx = pan_context(pipe);


        switch (query->type) {
        case PIPE_QUERY_OCCLUSION_COUNTER:
        case PIPE_QUERY_OCCLUSION_PREDICATE:
        case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
                /* Flush first */
                panfrost_flush_all_batches(ctx, true);

                /* Read back the query results */
                unsigned *result = (unsigned *) query->bo->cpu;
                unsigned passed = *result;

                if (query->type == PIPE_QUERY_OCCLUSION_COUNTER) {
                        vresult->u64 = passed;
                } else {
                        vresult->b = !!passed;
                }

                break;

        case PIPE_QUERY_PRIMITIVES_GENERATED:
        case PIPE_QUERY_PRIMITIVES_EMITTED:
                panfrost_flush_all_batches(ctx, true);
                vresult->u64 = query->end - query->start;
                break;

        default:
                DBG("Skipped query get %u\n", query->type);
                break;
        }

        return true;
}

static struct pipe_stream_output_target *
panfrost_create_stream_output_target(struct pipe_context *pctx,
                                     struct pipe_resource *prsc,
                                     unsigned buffer_offset,
                                     unsigned buffer_size)
{
        struct pipe_stream_output_target *target;

        target = rzalloc(pctx, struct pipe_stream_output_target);

        if (!target)
                return NULL;

        pipe_reference_init(&target->reference, 1);
        pipe_resource_reference(&target->buffer, prsc);

        target->context = pctx;
        target->buffer_offset = buffer_offset;
        target->buffer_size = buffer_size;

        return target;
}

static void
panfrost_stream_output_target_destroy(struct pipe_context *pctx,
                                      struct pipe_stream_output_target *target)
{
        pipe_resource_reference(&target->buffer, NULL);
        ralloc_free(target);
}

static void
panfrost_set_stream_output_targets(struct pipe_context *pctx,
                                   unsigned num_targets,
                                   struct pipe_stream_output_target **targets,
                                   const unsigned *offsets)
{
        struct panfrost_context *ctx = pan_context(pctx);
        struct panfrost_streamout *so = &ctx->streamout;

        assert(num_targets <= ARRAY_SIZE(so->targets));

        for (unsigned i = 0; i < num_targets; i++) {
                if (offsets[i] != -1)
                        so->offsets[i] = offsets[i];

                pipe_so_target_reference(&so->targets[i], targets[i]);
        }

        for (unsigned i = 0; i < so->num_targets; i++)
                pipe_so_target_reference(&so->targets[i], NULL);

        so->num_targets = num_targets;
}

struct pipe_context *
panfrost_create_context(struct pipe_screen *screen, void *priv, unsigned flags)
{
        struct panfrost_context *ctx = rzalloc(screen, struct panfrost_context);
        struct panfrost_screen *pscreen = pan_screen(screen);
        struct pipe_context *gallium = (struct pipe_context *) ctx;

        gallium->screen = screen;

        gallium->destroy = panfrost_destroy;

        gallium->set_framebuffer_state = panfrost_set_framebuffer_state;

        gallium->flush = panfrost_flush;
        gallium->clear = panfrost_clear;
        gallium->draw_vbo = panfrost_draw_vbo;

        gallium->set_vertex_buffers = panfrost_set_vertex_buffers;
        gallium->set_constant_buffer = panfrost_set_constant_buffer;
        gallium->set_shader_buffers = panfrost_set_shader_buffers;

        gallium->set_stencil_ref = panfrost_set_stencil_ref;

        gallium->create_sampler_view = panfrost_create_sampler_view;
        gallium->set_sampler_views = panfrost_set_sampler_views;
        gallium->sampler_view_destroy = panfrost_sampler_view_destroy;

        gallium->create_rasterizer_state = panfrost_create_rasterizer_state;
        gallium->bind_rasterizer_state = panfrost_bind_rasterizer_state;
        gallium->delete_rasterizer_state = panfrost_generic_cso_delete;

        gallium->create_vertex_elements_state = panfrost_create_vertex_elements_state;
        gallium->bind_vertex_elements_state = panfrost_bind_vertex_elements_state;
        gallium->delete_vertex_elements_state = panfrost_generic_cso_delete;

        gallium->create_fs_state = panfrost_create_shader_state;
        gallium->delete_fs_state = panfrost_delete_shader_state;
        gallium->bind_fs_state = panfrost_bind_fs_state;

        gallium->create_vs_state = panfrost_create_shader_state;
        gallium->delete_vs_state = panfrost_delete_shader_state;
        gallium->bind_vs_state = panfrost_bind_vs_state;

        gallium->create_sampler_state = panfrost_create_sampler_state;
        gallium->delete_sampler_state = panfrost_generic_cso_delete;
        gallium->bind_sampler_states = panfrost_bind_sampler_states;

        gallium->create_depth_stencil_alpha_state = panfrost_create_depth_stencil_state;
        gallium->bind_depth_stencil_alpha_state   = panfrost_bind_depth_stencil_state;
        gallium->delete_depth_stencil_alpha_state = panfrost_delete_depth_stencil_state;

        gallium->set_sample_mask = panfrost_set_sample_mask;

        gallium->set_clip_state = panfrost_set_clip_state;
        gallium->set_viewport_states = panfrost_set_viewport_states;
        gallium->set_scissor_states = panfrost_set_scissor_states;
        gallium->set_polygon_stipple = panfrost_set_polygon_stipple;
        gallium->set_active_query_state = panfrost_set_active_query_state;

        gallium->create_query = panfrost_create_query;
        gallium->destroy_query = panfrost_destroy_query;
        gallium->begin_query = panfrost_begin_query;
        gallium->end_query = panfrost_end_query;
        gallium->get_query_result = panfrost_get_query_result;

        gallium->create_stream_output_target = panfrost_create_stream_output_target;
        gallium->stream_output_target_destroy = panfrost_stream_output_target_destroy;
        gallium->set_stream_output_targets = panfrost_set_stream_output_targets;

        panfrost_resource_context_init(gallium);
        panfrost_blend_context_init(gallium);
        panfrost_compute_context_init(gallium);

        /* XXX: leaks */
        gallium->stream_uploader = u_upload_create_default(gallium);
        gallium->const_uploader = gallium->stream_uploader;
        assert(gallium->stream_uploader);

        /* Midgard supports ES modes, plus QUADS/QUAD_STRIPS/POLYGON */
        ctx->draw_modes = (1 << (PIPE_PRIM_POLYGON + 1)) - 1;

        ctx->primconvert = util_primconvert_create(gallium, ctx->draw_modes);

        ctx->blitter = util_blitter_create(gallium);
        ctx->blitter_wallpaper = util_blitter_create(gallium);

        assert(ctx->blitter);
        assert(ctx->blitter_wallpaper);

        /* Prepare for render! */

        panfrost_batch_init(ctx);
        panfrost_emit_vertex_payload(ctx);
        panfrost_emit_tiler_payload(ctx);
        panfrost_invalidate_frame(ctx);
        panfrost_default_shader_backend(ctx);

        return gallium;
}