summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/iris/iris_state.c
blob: dec914b550b1f111db33d2976fbd0cbd93c70be6 (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
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
4990
4991
4992
4993
4994
4995
4996
4997
4998
4999
5000
5001
5002
5003
5004
5005
5006
5007
5008
5009
5010
5011
5012
5013
5014
5015
5016
5017
5018
5019
5020
5021
5022
5023
5024
5025
5026
5027
5028
5029
5030
5031
5032
5033
5034
5035
5036
5037
5038
5039
5040
5041
5042
5043
5044
5045
5046
5047
5048
5049
5050
5051
5052
5053
5054
5055
5056
5057
5058
5059
5060
5061
5062
5063
5064
5065
5066
5067
5068
5069
5070
5071
5072
5073
5074
5075
5076
5077
5078
5079
5080
5081
5082
5083
5084
5085
5086
5087
5088
5089
5090
5091
5092
5093
5094
5095
5096
5097
5098
5099
5100
5101
5102
5103
5104
5105
5106
5107
5108
5109
5110
5111
5112
5113
5114
5115
5116
5117
5118
5119
5120
5121
5122
5123
5124
5125
5126
5127
5128
5129
5130
5131
5132
5133
5134
5135
5136
5137
5138
5139
5140
5141
5142
5143
5144
5145
5146
5147
5148
5149
5150
5151
5152
5153
5154
5155
5156
5157
5158
5159
5160
5161
5162
5163
5164
5165
5166
5167
5168
5169
5170
5171
5172
5173
5174
5175
5176
5177
5178
5179
5180
5181
5182
5183
5184
5185
5186
5187
5188
5189
5190
5191
5192
5193
5194
5195
5196
5197
5198
5199
5200
5201
5202
5203
5204
5205
5206
5207
/*
 * Copyright © 2017 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included
 * in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
 * DEALINGS IN THE SOFTWARE.
 */

/**
 * @file iris_state.c
 *
 * ============================= GENXML CODE =============================
 *              [This file is compiled once per generation.]
 * =======================================================================
 *
 * This is the main state upload code.
 *
 * Gallium uses Constant State Objects, or CSOs, for most state.  Large,
 * complex, or highly reusable state can be created once, and bound and
 * rebound multiple times.  This is modeled with the pipe->create_*_state()
 * and pipe->bind_*_state() hooks.  Highly dynamic or inexpensive state is
 * streamed out on the fly, via pipe->set_*_state() hooks.
 *
 * OpenGL involves frequently mutating context state, which is mirrored in
 * core Mesa by highly mutable data structures.  However, most applications
 * typically draw the same things over and over - from frame to frame, most
 * of the same objects are still visible and need to be redrawn.  So, rather
 * than inventing new state all the time, applications usually mutate to swap
 * between known states that we've seen before.
 *
 * Gallium isolates us from this mutation by tracking API state, and
 * distilling it into a set of Constant State Objects, or CSOs.  Large,
 * complex, or typically reusable state can be created once, then reused
 * multiple times.  Drivers can create and store their own associated data.
 * This create/bind model corresponds to the pipe->create_*_state() and
 * pipe->bind_*_state() driver hooks.
 *
 * Some state is cheap to create, or expected to be highly dynamic.  Rather
 * than creating and caching piles of CSOs for these, Gallium simply streams
 * them out, via the pipe->set_*_state() driver hooks.
 *
 * To reduce draw time overhead, we try to compute as much state at create
 * time as possible.  Wherever possible, we translate the Gallium pipe state
 * to 3DSTATE commands, and store those commands in the CSO.  At draw time,
 * we can simply memcpy them into a batch buffer.
 *
 * No hardware matches the abstraction perfectly, so some commands require
 * information from multiple CSOs.  In this case, we can store two copies
 * of the packet (one in each CSO), and simply | together their DWords at
 * draw time.  Sometimes the second set is trivial (one or two fields), so
 * we simply pack it at draw time.
 *
 * There are two main components in the file below.  First, the CSO hooks
 * create/bind/track state.  The second are the draw-time upload functions,
 * iris_upload_render_state() and iris_upload_compute_state(), which read
 * the context state and emit the commands into the actual batch.
 */

#include <stdio.h>
#include <errno.h>

#if HAVE_VALGRIND
#include <valgrind.h>
#include <memcheck.h>
#define VG(x) x
#ifndef NDEBUG
#define __gen_validate_value(x) VALGRIND_CHECK_MEM_IS_DEFINED(&(x), sizeof(x))
#endif
#else
#define VG(x)
#endif

#include "pipe/p_defines.h"
#include "pipe/p_state.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "util/u_inlines.h"
#include "util/u_format.h"
#include "util/u_framebuffer.h"
#include "util/u_transfer.h"
#include "util/u_upload_mgr.h"
#include "util/u_viewport.h"
#include "i915_drm.h"
#include "nir.h"
#include "intel/compiler/brw_compiler.h"
#include "intel/common/gen_l3_config.h"
#include "intel/common/gen_sample_positions.h"
#include "iris_batch.h"
#include "iris_context.h"
#include "iris_pipe.h"
#include "iris_resource.h"

#define __gen_address_type struct iris_address
#define __gen_user_data struct iris_batch

#define ARRAY_BYTES(x) (sizeof(uint32_t) * ARRAY_SIZE(x))

static uint64_t
__gen_combine_address(struct iris_batch *batch, void *location,
                      struct iris_address addr, uint32_t delta)
{
   uint64_t result = addr.offset + delta;

   if (addr.bo) {
      iris_use_pinned_bo(batch, addr.bo, addr.write);
      /* Assume this is a general address, not relative to a base. */
      result += addr.bo->gtt_offset;
   }

   return result;
}

#define __genxml_cmd_length(cmd) cmd ## _length
#define __genxml_cmd_length_bias(cmd) cmd ## _length_bias
#define __genxml_cmd_header(cmd) cmd ## _header
#define __genxml_cmd_pack(cmd) cmd ## _pack

#define _iris_pack_command(batch, cmd, dst, name)                 \
   for (struct cmd name = { __genxml_cmd_header(cmd) },           \
        *_dst = (void *)(dst); __builtin_expect(_dst != NULL, 1); \
        ({ __genxml_cmd_pack(cmd)(batch, (void *)_dst, &name);    \
           _dst = NULL;                                           \
           }))

#define iris_pack_command(cmd, dst, name) \
   _iris_pack_command(NULL, cmd, dst, name)

#define iris_pack_state(cmd, dst, name)                           \
   for (struct cmd name = {},                                     \
        *_dst = (void *)(dst); __builtin_expect(_dst != NULL, 1); \
        __genxml_cmd_pack(cmd)(NULL, (void *)_dst, &name),        \
        _dst = NULL)

#define iris_emit_cmd(batch, cmd, name) \
   _iris_pack_command(batch, cmd, iris_get_command_space(batch, 4 * __genxml_cmd_length(cmd)), name)

#define iris_emit_merge(batch, dwords0, dwords1, num_dwords)   \
   do {                                                        \
      uint32_t *dw = iris_get_command_space(batch, 4 * num_dwords); \
      for (uint32_t i = 0; i < num_dwords; i++)                \
         dw[i] = (dwords0)[i] | (dwords1)[i];                  \
      VG(VALGRIND_CHECK_MEM_IS_DEFINED(dw, num_dwords));       \
   } while (0)

#include "genxml/genX_pack.h"
#include "genxml/gen_macros.h"
#include "genxml/genX_bits.h"

#define MOCS_WB (2 << 1)

/**
 * Statically assert that PIPE_* enums match the hardware packets.
 * (As long as they match, we don't need to translate them.)
 */
UNUSED static void pipe_asserts()
{
#define PIPE_ASSERT(x) STATIC_ASSERT((int)x)

   /* pipe_logicop happens to match the hardware. */
   PIPE_ASSERT(PIPE_LOGICOP_CLEAR == LOGICOP_CLEAR);
   PIPE_ASSERT(PIPE_LOGICOP_NOR == LOGICOP_NOR);
   PIPE_ASSERT(PIPE_LOGICOP_AND_INVERTED == LOGICOP_AND_INVERTED);
   PIPE_ASSERT(PIPE_LOGICOP_COPY_INVERTED == LOGICOP_COPY_INVERTED);
   PIPE_ASSERT(PIPE_LOGICOP_AND_REVERSE == LOGICOP_AND_REVERSE);
   PIPE_ASSERT(PIPE_LOGICOP_INVERT == LOGICOP_INVERT);
   PIPE_ASSERT(PIPE_LOGICOP_XOR == LOGICOP_XOR);
   PIPE_ASSERT(PIPE_LOGICOP_NAND == LOGICOP_NAND);
   PIPE_ASSERT(PIPE_LOGICOP_AND == LOGICOP_AND);
   PIPE_ASSERT(PIPE_LOGICOP_EQUIV == LOGICOP_EQUIV);
   PIPE_ASSERT(PIPE_LOGICOP_NOOP == LOGICOP_NOOP);
   PIPE_ASSERT(PIPE_LOGICOP_OR_INVERTED == LOGICOP_OR_INVERTED);
   PIPE_ASSERT(PIPE_LOGICOP_COPY == LOGICOP_COPY);
   PIPE_ASSERT(PIPE_LOGICOP_OR_REVERSE == LOGICOP_OR_REVERSE);
   PIPE_ASSERT(PIPE_LOGICOP_OR == LOGICOP_OR);
   PIPE_ASSERT(PIPE_LOGICOP_SET == LOGICOP_SET);

   /* pipe_blend_func happens to match the hardware. */
   PIPE_ASSERT(PIPE_BLENDFACTOR_ONE == BLENDFACTOR_ONE);
   PIPE_ASSERT(PIPE_BLENDFACTOR_SRC_COLOR == BLENDFACTOR_SRC_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_SRC_ALPHA == BLENDFACTOR_SRC_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_DST_ALPHA == BLENDFACTOR_DST_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_DST_COLOR == BLENDFACTOR_DST_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE == BLENDFACTOR_SRC_ALPHA_SATURATE);
   PIPE_ASSERT(PIPE_BLENDFACTOR_CONST_COLOR == BLENDFACTOR_CONST_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_CONST_ALPHA == BLENDFACTOR_CONST_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_SRC1_COLOR == BLENDFACTOR_SRC1_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_SRC1_ALPHA == BLENDFACTOR_SRC1_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_ZERO == BLENDFACTOR_ZERO);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_SRC_COLOR == BLENDFACTOR_INV_SRC_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_SRC_ALPHA == BLENDFACTOR_INV_SRC_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_DST_ALPHA == BLENDFACTOR_INV_DST_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_DST_COLOR == BLENDFACTOR_INV_DST_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_CONST_COLOR == BLENDFACTOR_INV_CONST_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_CONST_ALPHA == BLENDFACTOR_INV_CONST_ALPHA);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_SRC1_COLOR == BLENDFACTOR_INV_SRC1_COLOR);
   PIPE_ASSERT(PIPE_BLENDFACTOR_INV_SRC1_ALPHA == BLENDFACTOR_INV_SRC1_ALPHA);

   /* pipe_blend_func happens to match the hardware. */
   PIPE_ASSERT(PIPE_BLEND_ADD == BLENDFUNCTION_ADD);
   PIPE_ASSERT(PIPE_BLEND_SUBTRACT == BLENDFUNCTION_SUBTRACT);
   PIPE_ASSERT(PIPE_BLEND_REVERSE_SUBTRACT == BLENDFUNCTION_REVERSE_SUBTRACT);
   PIPE_ASSERT(PIPE_BLEND_MIN == BLENDFUNCTION_MIN);
   PIPE_ASSERT(PIPE_BLEND_MAX == BLENDFUNCTION_MAX);

   /* pipe_stencil_op happens to match the hardware. */
   PIPE_ASSERT(PIPE_STENCIL_OP_KEEP == STENCILOP_KEEP);
   PIPE_ASSERT(PIPE_STENCIL_OP_ZERO == STENCILOP_ZERO);
   PIPE_ASSERT(PIPE_STENCIL_OP_REPLACE == STENCILOP_REPLACE);
   PIPE_ASSERT(PIPE_STENCIL_OP_INCR == STENCILOP_INCRSAT);
   PIPE_ASSERT(PIPE_STENCIL_OP_DECR == STENCILOP_DECRSAT);
   PIPE_ASSERT(PIPE_STENCIL_OP_INCR_WRAP == STENCILOP_INCR);
   PIPE_ASSERT(PIPE_STENCIL_OP_DECR_WRAP == STENCILOP_DECR);
   PIPE_ASSERT(PIPE_STENCIL_OP_INVERT == STENCILOP_INVERT);

   /* pipe_sprite_coord_mode happens to match 3DSTATE_SBE */
   PIPE_ASSERT(PIPE_SPRITE_COORD_UPPER_LEFT == UPPERLEFT);
   PIPE_ASSERT(PIPE_SPRITE_COORD_LOWER_LEFT == LOWERLEFT);
#undef PIPE_ASSERT
}

static unsigned
translate_prim_type(enum pipe_prim_type prim, uint8_t verts_per_patch)
{
   static const unsigned map[] = {
      [PIPE_PRIM_POINTS]                   = _3DPRIM_POINTLIST,
      [PIPE_PRIM_LINES]                    = _3DPRIM_LINELIST,
      [PIPE_PRIM_LINE_LOOP]                = _3DPRIM_LINELOOP,
      [PIPE_PRIM_LINE_STRIP]               = _3DPRIM_LINESTRIP,
      [PIPE_PRIM_TRIANGLES]                = _3DPRIM_TRILIST,
      [PIPE_PRIM_TRIANGLE_STRIP]           = _3DPRIM_TRISTRIP,
      [PIPE_PRIM_TRIANGLE_FAN]             = _3DPRIM_TRIFAN,
      [PIPE_PRIM_QUADS]                    = _3DPRIM_QUADLIST,
      [PIPE_PRIM_QUAD_STRIP]               = _3DPRIM_QUADSTRIP,
      [PIPE_PRIM_POLYGON]                  = _3DPRIM_POLYGON,
      [PIPE_PRIM_LINES_ADJACENCY]          = _3DPRIM_LINELIST_ADJ,
      [PIPE_PRIM_LINE_STRIP_ADJACENCY]     = _3DPRIM_LINESTRIP_ADJ,
      [PIPE_PRIM_TRIANGLES_ADJACENCY]      = _3DPRIM_TRILIST_ADJ,
      [PIPE_PRIM_TRIANGLE_STRIP_ADJACENCY] = _3DPRIM_TRISTRIP_ADJ,
      [PIPE_PRIM_PATCHES]                  = _3DPRIM_PATCHLIST_1 - 1,
   };

   return map[prim] + (prim == PIPE_PRIM_PATCHES ? verts_per_patch : 0);
}

static unsigned
translate_compare_func(enum pipe_compare_func pipe_func)
{
   static const unsigned map[] = {
      [PIPE_FUNC_NEVER]    = COMPAREFUNCTION_NEVER,
      [PIPE_FUNC_LESS]     = COMPAREFUNCTION_LESS,
      [PIPE_FUNC_EQUAL]    = COMPAREFUNCTION_EQUAL,
      [PIPE_FUNC_LEQUAL]   = COMPAREFUNCTION_LEQUAL,
      [PIPE_FUNC_GREATER]  = COMPAREFUNCTION_GREATER,
      [PIPE_FUNC_NOTEQUAL] = COMPAREFUNCTION_NOTEQUAL,
      [PIPE_FUNC_GEQUAL]   = COMPAREFUNCTION_GEQUAL,
      [PIPE_FUNC_ALWAYS]   = COMPAREFUNCTION_ALWAYS,
   };
   return map[pipe_func];
}

static unsigned
translate_shadow_func(enum pipe_compare_func pipe_func)
{
   /* Gallium specifies the result of shadow comparisons as:
    *
    *    1 if ref <op> texel,
    *    0 otherwise.
    *
    * The hardware does:
    *
    *    0 if texel <op> ref,
    *    1 otherwise.
    *
    * So we need to flip the operator and also negate.
    */
   static const unsigned map[] = {
      [PIPE_FUNC_NEVER]    = PREFILTEROPALWAYS,
      [PIPE_FUNC_LESS]     = PREFILTEROPLEQUAL,
      [PIPE_FUNC_EQUAL]    = PREFILTEROPNOTEQUAL,
      [PIPE_FUNC_LEQUAL]   = PREFILTEROPLESS,
      [PIPE_FUNC_GREATER]  = PREFILTEROPGEQUAL,
      [PIPE_FUNC_NOTEQUAL] = PREFILTEROPEQUAL,
      [PIPE_FUNC_GEQUAL]   = PREFILTEROPGREATER,
      [PIPE_FUNC_ALWAYS]   = PREFILTEROPNEVER,
   };
   return map[pipe_func];
}

static unsigned
translate_cull_mode(unsigned pipe_face)
{
   static const unsigned map[4] = {
      [PIPE_FACE_NONE]           = CULLMODE_NONE,
      [PIPE_FACE_FRONT]          = CULLMODE_FRONT,
      [PIPE_FACE_BACK]           = CULLMODE_BACK,
      [PIPE_FACE_FRONT_AND_BACK] = CULLMODE_BOTH,
   };
   return map[pipe_face];
}

static unsigned
translate_fill_mode(unsigned pipe_polymode)
{
   static const unsigned map[4] = {
      [PIPE_POLYGON_MODE_FILL]           = FILL_MODE_SOLID,
      [PIPE_POLYGON_MODE_LINE]           = FILL_MODE_WIREFRAME,
      [PIPE_POLYGON_MODE_POINT]          = FILL_MODE_POINT,
      [PIPE_POLYGON_MODE_FILL_RECTANGLE] = FILL_MODE_SOLID,
   };
   return map[pipe_polymode];
}

static unsigned
translate_mip_filter(enum pipe_tex_mipfilter pipe_mip)
{
   static const unsigned map[] = {
      [PIPE_TEX_MIPFILTER_NEAREST] = MIPFILTER_NEAREST,
      [PIPE_TEX_MIPFILTER_LINEAR]  = MIPFILTER_LINEAR,
      [PIPE_TEX_MIPFILTER_NONE]    = MIPFILTER_NONE,
   };
   return map[pipe_mip];
}

static uint32_t
translate_wrap(unsigned pipe_wrap)
{
   static const unsigned map[] = {
      [PIPE_TEX_WRAP_REPEAT]                 = TCM_WRAP,
      [PIPE_TEX_WRAP_CLAMP]                  = TCM_HALF_BORDER,
      [PIPE_TEX_WRAP_CLAMP_TO_EDGE]          = TCM_CLAMP,
      [PIPE_TEX_WRAP_CLAMP_TO_BORDER]        = TCM_CLAMP_BORDER,
      [PIPE_TEX_WRAP_MIRROR_REPEAT]          = TCM_MIRROR,
      [PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE]   = TCM_MIRROR_ONCE,

      /* These are unsupported. */
      [PIPE_TEX_WRAP_MIRROR_CLAMP]           = -1,
      [PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER] = -1,
   };
   return map[pipe_wrap];
}

static struct iris_address
ro_bo(struct iris_bo *bo, uint64_t offset)
{
   /* CSOs must pass NULL for bo!  Otherwise it will add the BO to the
    * validation list at CSO creation time, instead of draw time.
    */
   return (struct iris_address) { .bo = bo, .offset = offset };
}

static struct iris_address
rw_bo(struct iris_bo *bo, uint64_t offset)
{
   /* CSOs must pass NULL for bo!  Otherwise it will add the BO to the
    * validation list at CSO creation time, instead of draw time.
    */
   return (struct iris_address) { .bo = bo, .offset = offset, .write = true };
}

/**
 * Allocate space for some indirect state.
 *
 * Return a pointer to the map (to fill it out) and a state ref (for
 * referring to the state in GPU commands).
 */
static void *
upload_state(struct u_upload_mgr *uploader,
             struct iris_state_ref *ref,
             unsigned size,
             unsigned alignment)
{
   void *p = NULL;
   u_upload_alloc(uploader, 0, size, alignment, &ref->offset, &ref->res, &p);
   return p;
}

/**
 * Stream out temporary/short-lived state.
 *
 * This allocates space, pins the BO, and includes the BO address in the
 * returned offset (which works because all state lives in 32-bit memory
 * zones).
 */
static uint32_t *
stream_state(struct iris_batch *batch,
             struct u_upload_mgr *uploader,
             struct pipe_resource **out_res,
             unsigned size,
             unsigned alignment,
             uint32_t *out_offset)
{
   void *ptr = NULL;

   u_upload_alloc(uploader, 0, size, alignment, out_offset, out_res, &ptr);

   struct iris_bo *bo = iris_resource_bo(*out_res);
   iris_use_pinned_bo(batch, bo, false);

   *out_offset += iris_bo_offset_from_base_address(bo);

   return ptr;
}

/**
 * stream_state() + memcpy.
 */
static uint32_t
emit_state(struct iris_batch *batch,
           struct u_upload_mgr *uploader,
           struct pipe_resource **out_res,
           const void *data,
           unsigned size,
           unsigned alignment)
{
   unsigned offset = 0;
   uint32_t *map =
      stream_state(batch, uploader, out_res, size, alignment, &offset);

   if (map)
      memcpy(map, data, size);

   return offset;
}

/**
 * Did field 'x' change between 'old_cso' and 'new_cso'?
 *
 * (If so, we may want to set some dirty flags.)
 */
#define cso_changed(x) (!old_cso || (old_cso->x != new_cso->x))
#define cso_changed_memcmp(x) \
   (!old_cso || memcmp(old_cso->x, new_cso->x, sizeof(old_cso->x)) != 0)

static void
flush_for_state_base_change(struct iris_batch *batch)
{
   /* Flush before emitting STATE_BASE_ADDRESS.
    *
    * This isn't documented anywhere in the PRM.  However, it seems to be
    * necessary prior to changing the surface state base adress.  We've
    * seen issues in Vulkan where we get GPU hangs when using multi-level
    * command buffers which clear depth, reset state base address, and then
    * go render stuff.
    *
    * Normally, in GL, we would trust the kernel to do sufficient stalls
    * and flushes prior to executing our batch.  However, it doesn't seem
    * as if the kernel's flushing is always sufficient and we don't want to
    * rely on it.
    *
    * We make this an end-of-pipe sync instead of a normal flush because we
    * do not know the current status of the GPU.  On Haswell at least,
    * having a fast-clear operation in flight at the same time as a normal
    * rendering operation can cause hangs.  Since the kernel's flushing is
    * insufficient, we need to ensure that any rendering operations from
    * other processes are definitely complete before we try to do our own
    * rendering.  It's a bit of a big hammer but it appears to work.
    */
   iris_emit_end_of_pipe_sync(batch,
                              PIPE_CONTROL_RENDER_TARGET_FLUSH |
                              PIPE_CONTROL_DEPTH_CACHE_FLUSH |
                              PIPE_CONTROL_DATA_CACHE_FLUSH);
}

static void
_iris_emit_lri(struct iris_batch *batch, uint32_t reg, uint32_t val)
{
   iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
      lri.RegisterOffset = reg;
      lri.DataDWord      = val;
   }
}
#define iris_emit_lri(b, r, v) _iris_emit_lri(b, GENX(r##_num), v)

static void
emit_pipeline_select(struct iris_batch *batch, uint32_t pipeline)
{
#if GEN_GEN >= 8 && GEN_GEN < 10
   /* From the Broadwell PRM, Volume 2a: Instructions, PIPELINE_SELECT:
    *
    *   Software must clear the COLOR_CALC_STATE Valid field in
    *   3DSTATE_CC_STATE_POINTERS command prior to send a PIPELINE_SELECT
    *   with Pipeline Select set to GPGPU.
    *
    * The internal hardware docs recommend the same workaround for Gen9
    * hardware too.
    */
   if (pipeline == GPGPU)
      iris_emit_cmd(batch, GENX(3DSTATE_CC_STATE_POINTERS), t);
#endif


   /* From "BXML » GT » MI » vol1a GPU Overview » [Instruction]
    * PIPELINE_SELECT [DevBWR+]":
    *
    *    "Project: DEVSNB+
    *
    *     Software must ensure all the write caches are flushed through a
    *     stalling PIPE_CONTROL command followed by another PIPE_CONTROL
    *     command to invalidate read only caches prior to programming
    *     MI_PIPELINE_SELECT command to change the Pipeline Select Mode."
    */
    iris_emit_pipe_control_flush(batch,
                                 PIPE_CONTROL_RENDER_TARGET_FLUSH |
                                 PIPE_CONTROL_DEPTH_CACHE_FLUSH |
                                 PIPE_CONTROL_DATA_CACHE_FLUSH |
                                 PIPE_CONTROL_CS_STALL);

    iris_emit_pipe_control_flush(batch,
                                 PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE |
                                 PIPE_CONTROL_CONST_CACHE_INVALIDATE |
                                 PIPE_CONTROL_STATE_CACHE_INVALIDATE |
                                 PIPE_CONTROL_INSTRUCTION_INVALIDATE);

   iris_emit_cmd(batch, GENX(PIPELINE_SELECT), sel) {
#if GEN_GEN >= 9
      sel.MaskBits = 3;
#endif
      sel.PipelineSelection = pipeline;
   }
}

UNUSED static void
init_glk_barrier_mode(struct iris_batch *batch, uint32_t value)
{
#if GEN_GEN == 9
   /* Project: DevGLK
    *
    *    "This chicken bit works around a hardware issue with barrier
    *     logic encountered when switching between GPGPU and 3D pipelines.
    *     To workaround the issue, this mode bit should be set after a
    *     pipeline is selected."
    */
   uint32_t reg_val;
   iris_pack_state(GENX(SLICE_COMMON_ECO_CHICKEN1), &reg_val, reg) {
      reg.GLKBarrierMode = value;
      reg.GLKBarrierModeMask = 1;
   }
   iris_emit_lri(batch, SLICE_COMMON_ECO_CHICKEN1, reg_val);
#endif
}

static void
init_state_base_address(struct iris_batch *batch)
{
   flush_for_state_base_change(batch);

   /* We program most base addresses once at context initialization time.
    * Each base address points at a 4GB memory zone, and never needs to
    * change.  See iris_bufmgr.h for a description of the memory zones.
    *
    * The one exception is Surface State Base Address, which needs to be
    * updated occasionally.  See iris_binder.c for the details there.
    */
   iris_emit_cmd(batch, GENX(STATE_BASE_ADDRESS), sba) {
   #if 0
   // XXX: MOCS is stupid for this.
      sba.GeneralStateMemoryObjectControlState            = MOCS_WB;
      sba.StatelessDataPortAccessMemoryObjectControlState = MOCS_WB;
      sba.DynamicStateMemoryObjectControlState            = MOCS_WB;
      sba.IndirectObjectMemoryObjectControlState          = MOCS_WB;
      sba.InstructionMemoryObjectControlState             = MOCS_WB;
      sba.BindlessSurfaceStateMemoryObjectControlState    = MOCS_WB;
   #endif

      sba.GeneralStateBaseAddressModifyEnable   = true;
      sba.DynamicStateBaseAddressModifyEnable   = true;
      sba.IndirectObjectBaseAddressModifyEnable = true;
      sba.InstructionBaseAddressModifyEnable    = true;
      sba.GeneralStateBufferSizeModifyEnable    = true;
      sba.DynamicStateBufferSizeModifyEnable    = true;
      sba.BindlessSurfaceStateBaseAddressModifyEnable = true;
      sba.IndirectObjectBufferSizeModifyEnable  = true;
      sba.InstructionBuffersizeModifyEnable     = true;

      sba.InstructionBaseAddress  = ro_bo(NULL, IRIS_MEMZONE_SHADER_START);
      sba.DynamicStateBaseAddress = ro_bo(NULL, IRIS_MEMZONE_DYNAMIC_START);

      sba.GeneralStateBufferSize   = 0xfffff;
      sba.IndirectObjectBufferSize = 0xfffff;
      sba.InstructionBufferSize    = 0xfffff;
      sba.DynamicStateBufferSize   = 0xfffff;
   }
}

/**
 * Upload the initial GPU state for a render context.
 *
 * This sets some invariant state that needs to be programmed a particular
 * way, but we never actually change.
 */
static void
iris_init_render_context(struct iris_screen *screen,
                         struct iris_batch *batch,
                         struct iris_vtable *vtbl,
                         struct pipe_debug_callback *dbg)
{
   UNUSED const struct gen_device_info *devinfo = &screen->devinfo;
   uint32_t reg_val;

   emit_pipeline_select(batch, _3D);

   init_state_base_address(batch);

   // XXX: INSTPM on Gen8
   iris_pack_state(GENX(CS_DEBUG_MODE2), &reg_val, reg) {
      reg.CONSTANT_BUFFERAddressOffsetDisable = true;
      reg.CONSTANT_BUFFERAddressOffsetDisableMask = true;
   }
   iris_emit_lri(batch, CS_DEBUG_MODE2, reg_val);

#if GEN_GEN == 9
   iris_pack_state(GENX(CACHE_MODE_1), &reg_val, reg) {
      reg.FloatBlendOptimizationEnable = true;
      reg.FloatBlendOptimizationEnableMask = true;
      reg.PartialResolveDisableInVC = true;
      reg.PartialResolveDisableInVCMask = true;
   }
   iris_emit_lri(batch, CACHE_MODE_1, reg_val);

   if (devinfo->is_geminilake)
      init_glk_barrier_mode(batch, GLK_BARRIER_MODE_3D_HULL);
#endif

#if GEN_GEN == 11
      iris_pack_state(GENX(SAMPLER_MODE), &reg_val, reg) {
         reg.HeaderlessMessageforPreemptableContexts = 1;
         reg.HeaderlessMessageforPreemptableContextsMask = 1;
      }
      iris_emit_lri(batch, SAMPLER_MODE, reg_val);

      // XXX: 3D_MODE?
#endif

   /* 3DSTATE_DRAWING_RECTANGLE is non-pipelined, so we want to avoid
    * changing it dynamically.  We set it to the maximum size here, and
    * instead include the render target dimensions in the viewport, so
    * viewport extents clipping takes care of pruning stray geometry.
    */
   iris_emit_cmd(batch, GENX(3DSTATE_DRAWING_RECTANGLE), rect) {
      rect.ClippedDrawingRectangleXMax = UINT16_MAX;
      rect.ClippedDrawingRectangleYMax = UINT16_MAX;
   }

   /* Set the initial MSAA sample positions. */
   iris_emit_cmd(batch, GENX(3DSTATE_SAMPLE_PATTERN), pat) {
      GEN_SAMPLE_POS_1X(pat._1xSample);
      GEN_SAMPLE_POS_2X(pat._2xSample);
      GEN_SAMPLE_POS_4X(pat._4xSample);
      GEN_SAMPLE_POS_8X(pat._8xSample);
      GEN_SAMPLE_POS_16X(pat._16xSample);
   }

   /* Use the legacy AA line coverage computation. */
   iris_emit_cmd(batch, GENX(3DSTATE_AA_LINE_PARAMETERS), foo);

   /* Disable chromakeying (it's for media) */
   iris_emit_cmd(batch, GENX(3DSTATE_WM_CHROMAKEY), foo);

   /* We want regular rendering, not special HiZ operations. */
   iris_emit_cmd(batch, GENX(3DSTATE_WM_HZ_OP), foo);

   /* No polygon stippling offsets are necessary. */
   // XXX: may need to set an offset for origin-UL framebuffers
   iris_emit_cmd(batch, GENX(3DSTATE_POLY_STIPPLE_OFFSET), foo);

   /* Set a static partitioning of the push constant area. */
   // XXX: this may be a bad idea...could starve the push ringbuffers...
   for (int i = 0; i <= MESA_SHADER_FRAGMENT; i++) {
      iris_emit_cmd(batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_VS), alloc) {
         alloc._3DCommandSubOpcode = 18 + i;
         alloc.ConstantBufferOffset = 6 * i;
         alloc.ConstantBufferSize = i == MESA_SHADER_FRAGMENT ? 8 : 6;
      }
   }
}

static void
iris_init_compute_context(struct iris_screen *screen,
                          struct iris_batch *batch,
                          struct iris_vtable *vtbl,
                          struct pipe_debug_callback *dbg)
{
   UNUSED const struct gen_device_info *devinfo = &screen->devinfo;

   emit_pipeline_select(batch, GPGPU);

   init_state_base_address(batch);

#if GEN_GEN == 9
   if (devinfo->is_geminilake)
      init_glk_barrier_mode(batch, GLK_BARRIER_MODE_GPGPU);
#endif
}

struct iris_vertex_buffer_state {
   /** The 3DSTATE_VERTEX_BUFFERS hardware packet. */
   uint32_t vertex_buffers[1 + 33 * GENX(VERTEX_BUFFER_STATE_length)];

   /** The resource to source vertex data from. */
   struct pipe_resource *resources[33];

   /** The number of bound vertex buffers. */
   unsigned num_buffers;
};

struct iris_depth_buffer_state {
   /* Depth/HiZ/Stencil related hardware packets. */
   uint32_t packets[GENX(3DSTATE_DEPTH_BUFFER_length) +
                    GENX(3DSTATE_STENCIL_BUFFER_length) +
                    GENX(3DSTATE_HIER_DEPTH_BUFFER_length) +
                    GENX(3DSTATE_CLEAR_PARAMS_length)];
};

/**
 * Generation-specific context state (ice->state.genx->...).
 *
 * Most state can go in iris_context directly, but these encode hardware
 * packets which vary by generation.
 */
struct iris_genx_state {
   /** SF_CLIP_VIEWPORT */
   uint32_t sf_cl_vp[GENX(SF_CLIP_VIEWPORT_length) * IRIS_MAX_VIEWPORTS];

   struct iris_vertex_buffer_state vertex_buffers;
   struct iris_depth_buffer_state depth_buffer;

   uint32_t so_buffers[4 * GENX(3DSTATE_SO_BUFFER_length)];
   uint32_t streamout[4 * GENX(3DSTATE_STREAMOUT_length)];
};

/**
 * The pipe->set_blend_color() driver hook.
 *
 * This corresponds to our COLOR_CALC_STATE.
 */
static void
iris_set_blend_color(struct pipe_context *ctx,
                     const struct pipe_blend_color *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;

   /* Our COLOR_CALC_STATE is exactly pipe_blend_color, so just memcpy */
   memcpy(&ice->state.blend_color, state, sizeof(struct pipe_blend_color));
   ice->state.dirty |= IRIS_DIRTY_COLOR_CALC_STATE;
}

/**
 * Gallium CSO for blend state (see pipe_blend_state).
 */
struct iris_blend_state {
   /** Partial 3DSTATE_PS_BLEND */
   uint32_t ps_blend[GENX(3DSTATE_PS_BLEND_length)];

   /** Partial BLEND_STATE */
   uint32_t blend_state[GENX(BLEND_STATE_length) +
                        BRW_MAX_DRAW_BUFFERS * GENX(BLEND_STATE_ENTRY_length)];

   bool alpha_to_coverage; /* for shader key */
};

/**
 * The pipe->create_blend_state() driver hook.
 *
 * Translates a pipe_blend_state into iris_blend_state.
 */
static void *
iris_create_blend_state(struct pipe_context *ctx,
                        const struct pipe_blend_state *state)
{
   struct iris_blend_state *cso = malloc(sizeof(struct iris_blend_state));
   uint32_t *blend_state = cso->blend_state;

   cso->alpha_to_coverage = state->alpha_to_coverage;

   iris_pack_command(GENX(3DSTATE_PS_BLEND), cso->ps_blend, pb) {
      /* pb.HasWriteableRT is filled in at draw time. */
      /* pb.AlphaTestEnable is filled in at draw time. */
      pb.AlphaToCoverageEnable = state->alpha_to_coverage;
      pb.IndependentAlphaBlendEnable = state->independent_blend_enable;

      pb.ColorBufferBlendEnable = state->rt[0].blend_enable;

      pb.SourceBlendFactor           = state->rt[0].rgb_src_factor;
      pb.SourceAlphaBlendFactor      = state->rt[0].alpha_func;
      pb.DestinationBlendFactor      = state->rt[0].rgb_dst_factor;
      pb.DestinationAlphaBlendFactor = state->rt[0].alpha_dst_factor;
   }

   iris_pack_state(GENX(BLEND_STATE), blend_state, bs) {
      bs.AlphaToCoverageEnable = state->alpha_to_coverage;
      bs.IndependentAlphaBlendEnable = state->independent_blend_enable;
      bs.AlphaToOneEnable = state->alpha_to_one;
      bs.AlphaToCoverageDitherEnable = state->alpha_to_coverage;
      bs.ColorDitherEnable = state->dither;
      /* bl.AlphaTestEnable and bs.AlphaTestFunction are filled in later. */
   }

   blend_state += GENX(BLEND_STATE_length);

   for (int i = 0; i < BRW_MAX_DRAW_BUFFERS; i++) {
      const struct pipe_rt_blend_state *rt =
         &state->rt[state->independent_blend_enable ? i : 0];
      iris_pack_state(GENX(BLEND_STATE_ENTRY), blend_state, be) {
         be.LogicOpEnable = state->logicop_enable;
         be.LogicOpFunction = state->logicop_func;

         be.PreBlendSourceOnlyClampEnable = false;
         be.ColorClampRange = COLORCLAMP_RTFORMAT;
         be.PreBlendColorClampEnable = true;
         be.PostBlendColorClampEnable = true;

         be.ColorBufferBlendEnable = rt->blend_enable;

         be.ColorBlendFunction          = rt->rgb_func;
         be.AlphaBlendFunction          = rt->alpha_func;
         be.SourceBlendFactor           = rt->rgb_src_factor;
         be.SourceAlphaBlendFactor      = rt->alpha_func;
         be.DestinationBlendFactor      = rt->rgb_dst_factor;
         be.DestinationAlphaBlendFactor = rt->alpha_dst_factor;

         be.WriteDisableRed   = !(rt->colormask & PIPE_MASK_R);
         be.WriteDisableGreen = !(rt->colormask & PIPE_MASK_G);
         be.WriteDisableBlue  = !(rt->colormask & PIPE_MASK_B);
         be.WriteDisableAlpha = !(rt->colormask & PIPE_MASK_A);
      }
      blend_state += GENX(BLEND_STATE_ENTRY_length);
   }

   return cso;
}

/**
 * The pipe->bind_blend_state() driver hook.
 *
 * Bind a blending CSO and flag related dirty bits.
 */
static void
iris_bind_blend_state(struct pipe_context *ctx, void *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   ice->state.cso_blend = state;
   ice->state.dirty |= IRIS_DIRTY_PS_BLEND;
   ice->state.dirty |= IRIS_DIRTY_BLEND_STATE;
   ice->state.dirty |= ice->state.dirty_for_nos[IRIS_NOS_BLEND];
}

/**
 * Gallium CSO for depth, stencil, and alpha testing state.
 */
struct iris_depth_stencil_alpha_state {
   /** Partial 3DSTATE_WM_DEPTH_STENCIL. */
   uint32_t wmds[GENX(3DSTATE_WM_DEPTH_STENCIL_length)];

   /** Outbound to BLEND_STATE, 3DSTATE_PS_BLEND, COLOR_CALC_STATE. */
   struct pipe_alpha_state alpha;

   /** Outbound to resolve and cache set tracking. */
   bool depth_writes_enabled;
   bool stencil_writes_enabled;
};

/**
 * The pipe->create_depth_stencil_alpha_state() driver hook.
 *
 * We encode most of 3DSTATE_WM_DEPTH_STENCIL, and just save off the alpha
 * testing state since we need pieces of it in a variety of places.
 */
static void *
iris_create_zsa_state(struct pipe_context *ctx,
                      const struct pipe_depth_stencil_alpha_state *state)
{
   struct iris_depth_stencil_alpha_state *cso =
      malloc(sizeof(struct iris_depth_stencil_alpha_state));

   bool two_sided_stencil = state->stencil[1].enabled;

   cso->alpha = state->alpha;
   cso->depth_writes_enabled = state->depth.writemask;
   cso->stencil_writes_enabled =
      state->stencil[0].writemask != 0 ||
      (two_sided_stencil && state->stencil[1].writemask != 1);

   /* The state tracker needs to optimize away EQUAL writes for us. */
   assert(!(state->depth.func == PIPE_FUNC_EQUAL && state->depth.writemask));

   iris_pack_command(GENX(3DSTATE_WM_DEPTH_STENCIL), cso->wmds, wmds) {
      wmds.StencilFailOp = state->stencil[0].fail_op;
      wmds.StencilPassDepthFailOp = state->stencil[0].zfail_op;
      wmds.StencilPassDepthPassOp = state->stencil[0].zpass_op;
      wmds.StencilTestFunction =
         translate_compare_func(state->stencil[0].func);
      wmds.BackfaceStencilFailOp = state->stencil[1].fail_op;
      wmds.BackfaceStencilPassDepthFailOp = state->stencil[1].zfail_op;
      wmds.BackfaceStencilPassDepthPassOp = state->stencil[1].zpass_op;
      wmds.BackfaceStencilTestFunction =
         translate_compare_func(state->stencil[1].func);
      wmds.DepthTestFunction = translate_compare_func(state->depth.func);
      wmds.DoubleSidedStencilEnable = two_sided_stencil;
      wmds.StencilTestEnable = state->stencil[0].enabled;
      wmds.StencilBufferWriteEnable =
         state->stencil[0].writemask != 0 ||
         (two_sided_stencil && state->stencil[1].writemask != 0);
      wmds.DepthTestEnable = state->depth.enabled;
      wmds.DepthBufferWriteEnable = state->depth.writemask;
      wmds.StencilTestMask = state->stencil[0].valuemask;
      wmds.StencilWriteMask = state->stencil[0].writemask;
      wmds.BackfaceStencilTestMask = state->stencil[1].valuemask;
      wmds.BackfaceStencilWriteMask = state->stencil[1].writemask;
      /* wmds.[Backface]StencilReferenceValue are merged later */
   }

   return cso;
}

/**
 * The pipe->bind_depth_stencil_alpha_state() driver hook.
 *
 * Bind a depth/stencil/alpha CSO and flag related dirty bits.
 */
static void
iris_bind_zsa_state(struct pipe_context *ctx, void *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_depth_stencil_alpha_state *old_cso = ice->state.cso_zsa;
   struct iris_depth_stencil_alpha_state *new_cso = state;

   if (new_cso) {
      if (cso_changed(alpha.ref_value))
         ice->state.dirty |= IRIS_DIRTY_COLOR_CALC_STATE;

      if (cso_changed(alpha.enabled))
         ice->state.dirty |= IRIS_DIRTY_PS_BLEND | IRIS_DIRTY_BLEND_STATE;

      if (cso_changed(alpha.func))
         ice->state.dirty |= IRIS_DIRTY_BLEND_STATE;

      ice->state.depth_writes_enabled = new_cso->depth_writes_enabled;
      ice->state.stencil_writes_enabled = new_cso->stencil_writes_enabled;
   }

   ice->state.cso_zsa = new_cso;
   ice->state.dirty |= IRIS_DIRTY_CC_VIEWPORT;
   ice->state.dirty |= IRIS_DIRTY_WM_DEPTH_STENCIL;
   ice->state.dirty |= ice->state.dirty_for_nos[IRIS_NOS_DEPTH_STENCIL_ALPHA];
}

/**
 * Gallium CSO for rasterizer state.
 */
struct iris_rasterizer_state {
   uint32_t sf[GENX(3DSTATE_SF_length)];
   uint32_t clip[GENX(3DSTATE_CLIP_length)];
   uint32_t raster[GENX(3DSTATE_RASTER_length)];
   uint32_t wm[GENX(3DSTATE_WM_length)];
   uint32_t line_stipple[GENX(3DSTATE_LINE_STIPPLE_length)];

   bool clip_halfz; /* for CC_VIEWPORT */
   bool depth_clip_near; /* for CC_VIEWPORT */
   bool depth_clip_far; /* for CC_VIEWPORT */
   bool flatshade; /* for shader state */
   bool flatshade_first; /* for stream output */
   bool clamp_fragment_color; /* for shader state */
   bool light_twoside; /* for shader state */
   bool rasterizer_discard; /* for 3DSTATE_STREAMOUT */
   bool half_pixel_center; /* for 3DSTATE_MULTISAMPLE */
   bool line_stipple_enable;
   bool poly_stipple_enable;
   bool multisample;
   bool force_persample_interp;
   enum pipe_sprite_coord_mode sprite_coord_mode; /* PIPE_SPRITE_* */
   uint16_t sprite_coord_enable;
};

static float
get_line_width(const struct pipe_rasterizer_state *state)
{
   float line_width = state->line_width;

   /* From the OpenGL 4.4 spec:
    *
    * "The actual width of non-antialiased lines is determined by rounding
    *  the supplied width to the nearest integer, then clamping it to the
    *  implementation-dependent maximum non-antialiased line width."
    */
   if (!state->multisample && !state->line_smooth)
      line_width = roundf(state->line_width);

   if (!state->multisample && state->line_smooth && line_width < 1.5f) {
      /* For 1 pixel line thickness or less, the general anti-aliasing
       * algorithm gives up, and a garbage line is generated.  Setting a
       * Line Width of 0.0 specifies the rasterization of the "thinnest"
       * (one-pixel-wide), non-antialiased lines.
       *
       * Lines rendered with zero Line Width are rasterized using the
       * "Grid Intersection Quantization" rules as specified by the
       * "Zero-Width (Cosmetic) Line Rasterization" section of the docs.
       */
      line_width = 0.0f;
   }

   return line_width;
}

/**
 * The pipe->create_rasterizer_state() driver hook.
 */
static void *
iris_create_rasterizer_state(struct pipe_context *ctx,
                             const struct pipe_rasterizer_state *state)
{
   struct iris_rasterizer_state *cso =
      malloc(sizeof(struct iris_rasterizer_state));

#if 0
   point_quad_rasterization -> SBE?

   not necessary?
   {
      poly_smooth
      force_persample_interp - ?
      bottom_edge_rule

      offset_units_unscaled - cap not exposed
   }
   #endif

   // XXX: it may make more sense just to store the pipe_rasterizer_state,
   // we're copying a lot of booleans here.  But we don't need all of them...

   cso->multisample = state->multisample;
   cso->force_persample_interp = state->force_persample_interp;
   cso->clip_halfz = state->clip_halfz;
   cso->depth_clip_near = state->depth_clip_near;
   cso->depth_clip_far = state->depth_clip_far;
   cso->flatshade = state->flatshade;
   cso->flatshade_first = state->flatshade_first;
   cso->clamp_fragment_color = state->clamp_fragment_color;
   cso->light_twoside = state->light_twoside;
   cso->rasterizer_discard = state->rasterizer_discard;
   cso->half_pixel_center = state->half_pixel_center;
   cso->sprite_coord_mode = state->sprite_coord_mode;
   cso->sprite_coord_enable = state->sprite_coord_enable;
   cso->line_stipple_enable = state->line_stipple_enable;
   cso->poly_stipple_enable = state->poly_stipple_enable;

   float line_width = get_line_width(state);

   iris_pack_command(GENX(3DSTATE_SF), cso->sf, sf) {
      sf.StatisticsEnable = true;
      sf.ViewportTransformEnable = true;
      sf.AALineDistanceMode = AALINEDISTANCE_TRUE;
      sf.LineEndCapAntialiasingRegionWidth =
         state->line_smooth ? _10pixels : _05pixels;
      sf.LastPixelEnable = state->line_last_pixel;
      sf.LineWidth = line_width;
      sf.SmoothPointEnable = state->point_smooth || state->multisample;
      sf.PointWidthSource = state->point_size_per_vertex ? Vertex : State;
      sf.PointWidth = state->point_size;

      if (state->flatshade_first) {
         sf.TriangleFanProvokingVertexSelect = 1;
      } else {
         sf.TriangleStripListProvokingVertexSelect = 2;
         sf.TriangleFanProvokingVertexSelect = 2;
         sf.LineStripListProvokingVertexSelect = 1;
      }
   }

   iris_pack_command(GENX(3DSTATE_RASTER), cso->raster, rr) {
      rr.FrontWinding = state->front_ccw ? CounterClockwise : Clockwise;
      rr.CullMode = translate_cull_mode(state->cull_face);
      rr.FrontFaceFillMode = translate_fill_mode(state->fill_front);
      rr.BackFaceFillMode = translate_fill_mode(state->fill_back);
      rr.DXMultisampleRasterizationEnable = state->multisample;
      rr.GlobalDepthOffsetEnableSolid = state->offset_tri;
      rr.GlobalDepthOffsetEnableWireframe = state->offset_line;
      rr.GlobalDepthOffsetEnablePoint = state->offset_point;
      rr.GlobalDepthOffsetConstant = state->offset_units * 2;
      rr.GlobalDepthOffsetScale = state->offset_scale;
      rr.GlobalDepthOffsetClamp = state->offset_clamp;
      rr.SmoothPointEnable = state->point_smooth || state->multisample;
      rr.AntialiasingEnable = state->line_smooth;
      rr.ScissorRectangleEnable = state->scissor;
      rr.ViewportZNearClipTestEnable = state->depth_clip_near;
      rr.ViewportZFarClipTestEnable = state->depth_clip_far;
      //rr.ConservativeRasterizationEnable = not yet supported by Gallium...
   }

   iris_pack_command(GENX(3DSTATE_CLIP), cso->clip, cl) {
      /* cl.NonPerspectiveBarycentricEnable is filled in at draw time from
       * the FS program; cl.ForceZeroRTAIndexEnable is filled in from the FB.
       */
      cl.StatisticsEnable = true;
      cl.EarlyCullEnable = true;
      cl.UserClipDistanceClipTestEnableBitmask = state->clip_plane_enable;
      cl.ForceUserClipDistanceClipTestEnableBitmask = true;
      cl.APIMode = state->clip_halfz ? APIMODE_D3D : APIMODE_OGL;
      cl.GuardbandClipTestEnable = true;
      cl.ClipMode = CLIPMODE_NORMAL;
      cl.ClipEnable = true;
      cl.ViewportXYClipTestEnable = state->point_tri_clip;
      cl.MinimumPointWidth = 0.125;
      cl.MaximumPointWidth = 255.875;

      if (state->flatshade_first) {
         cl.TriangleFanProvokingVertexSelect = 1;
      } else {
         cl.TriangleStripListProvokingVertexSelect = 2;
         cl.TriangleFanProvokingVertexSelect = 2;
         cl.LineStripListProvokingVertexSelect = 1;
      }
   }

   iris_pack_command(GENX(3DSTATE_WM), cso->wm, wm) {
      /* wm.BarycentricInterpolationMode and wm.EarlyDepthStencilControl are
       * filled in at draw time from the FS program.
       */
      wm.LineAntialiasingRegionWidth = _10pixels;
      wm.LineEndCapAntialiasingRegionWidth = _05pixels;
      wm.PointRasterizationRule = RASTRULE_UPPER_RIGHT;
      wm.LineStippleEnable = state->line_stipple_enable;
      wm.PolygonStippleEnable = state->poly_stipple_enable;
   }

   /* Remap from 0..255 back to 1..256 */
   const unsigned line_stipple_factor = state->line_stipple_factor + 1;

   iris_pack_command(GENX(3DSTATE_LINE_STIPPLE), cso->line_stipple, line) {
      line.LineStipplePattern = state->line_stipple_pattern;
      line.LineStippleInverseRepeatCount = 1.0f / line_stipple_factor;
      line.LineStippleRepeatCount = line_stipple_factor;
   }

   return cso;
}

/**
 * The pipe->bind_rasterizer_state() driver hook.
 *
 * Bind a rasterizer CSO and flag related dirty bits.
 */
static void
iris_bind_rasterizer_state(struct pipe_context *ctx, void *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_rasterizer_state *old_cso = ice->state.cso_rast;
   struct iris_rasterizer_state *new_cso = state;

   if (new_cso) {
      /* Try to avoid re-emitting 3DSTATE_LINE_STIPPLE, it's non-pipelined */
      if (cso_changed_memcmp(line_stipple))
         ice->state.dirty |= IRIS_DIRTY_LINE_STIPPLE;

      if (cso_changed(half_pixel_center))
         ice->state.dirty |= IRIS_DIRTY_MULTISAMPLE;

      if (cso_changed(line_stipple_enable) || cso_changed(poly_stipple_enable))
         ice->state.dirty |= IRIS_DIRTY_WM;

      if (cso_changed(rasterizer_discard) || cso_changed(flatshade_first))
         ice->state.dirty |= IRIS_DIRTY_STREAMOUT;

      if (cso_changed(depth_clip_near) || cso_changed(depth_clip_far) ||
          cso_changed(clip_halfz))
         ice->state.dirty |= IRIS_DIRTY_CC_VIEWPORT;

      if (cso_changed(sprite_coord_enable) || cso_changed(light_twoside))
         ice->state.dirty |= IRIS_DIRTY_SBE;
   }

   ice->state.cso_rast = new_cso;
   ice->state.dirty |= IRIS_DIRTY_RASTER;
   ice->state.dirty |= IRIS_DIRTY_CLIP;
   ice->state.dirty |= ice->state.dirty_for_nos[IRIS_NOS_RASTERIZER];
}

/**
 * Return true if the given wrap mode requires the border color to exist.
 *
 * (We can skip uploading it if the sampler isn't going to use it.)
 */
static bool
wrap_mode_needs_border_color(unsigned wrap_mode)
{
   return wrap_mode == TCM_CLAMP_BORDER || wrap_mode == TCM_HALF_BORDER;
}

/**
 * Gallium CSO for sampler state.
 */
struct iris_sampler_state {
   union pipe_color_union border_color;
   bool needs_border_color;

   uint32_t sampler_state[GENX(SAMPLER_STATE_length)];
};

/**
 * The pipe->create_sampler_state() driver hook.
 *
 * We fill out SAMPLER_STATE (except for the border color pointer), and
 * store that on the CPU.  It doesn't make sense to upload it to a GPU
 * buffer object yet, because 3DSTATE_SAMPLER_STATE_POINTERS requires
 * all bound sampler states to be in contiguous memor.
 */
static void *
iris_create_sampler_state(struct pipe_context *ctx,
                          const struct pipe_sampler_state *state)
{
   struct iris_sampler_state *cso = CALLOC_STRUCT(iris_sampler_state);

   if (!cso)
      return NULL;

   STATIC_ASSERT(PIPE_TEX_FILTER_NEAREST == MAPFILTER_NEAREST);
   STATIC_ASSERT(PIPE_TEX_FILTER_LINEAR == MAPFILTER_LINEAR);

   unsigned wrap_s = translate_wrap(state->wrap_s);
   unsigned wrap_t = translate_wrap(state->wrap_t);
   unsigned wrap_r = translate_wrap(state->wrap_r);

   memcpy(&cso->border_color, &state->border_color, sizeof(cso->border_color));

   cso->needs_border_color = wrap_mode_needs_border_color(wrap_s) ||
                             wrap_mode_needs_border_color(wrap_t) ||
                             wrap_mode_needs_border_color(wrap_r);

   float min_lod = state->min_lod;
   unsigned mag_img_filter = state->mag_img_filter;

   // XXX: explain this code ported from ilo...I don't get it at all...
   if (state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE &&
       state->min_lod > 0.0f) {
      min_lod = 0.0f;
      mag_img_filter = state->min_img_filter;
   }

   iris_pack_state(GENX(SAMPLER_STATE), cso->sampler_state, samp) {
      samp.TCXAddressControlMode = wrap_s;
      samp.TCYAddressControlMode = wrap_t;
      samp.TCZAddressControlMode = wrap_r;
      samp.CubeSurfaceControlMode = state->seamless_cube_map;
      samp.NonnormalizedCoordinateEnable = !state->normalized_coords;
      samp.MinModeFilter = state->min_img_filter;
      samp.MagModeFilter = mag_img_filter;
      samp.MipModeFilter = translate_mip_filter(state->min_mip_filter);
      samp.MaximumAnisotropy = RATIO21;

      if (state->max_anisotropy >= 2) {
         if (state->min_img_filter == PIPE_TEX_FILTER_LINEAR) {
            samp.MinModeFilter = MAPFILTER_ANISOTROPIC;
            samp.AnisotropicAlgorithm = EWAApproximation;
         }

         if (state->mag_img_filter == PIPE_TEX_FILTER_LINEAR)
            samp.MagModeFilter = MAPFILTER_ANISOTROPIC;

         samp.MaximumAnisotropy =
            MIN2((state->max_anisotropy - 2) / 2, RATIO161);
      }

      /* Set address rounding bits if not using nearest filtering. */
      if (state->min_img_filter != PIPE_TEX_FILTER_NEAREST) {
         samp.UAddressMinFilterRoundingEnable = true;
         samp.VAddressMinFilterRoundingEnable = true;
         samp.RAddressMinFilterRoundingEnable = true;
      }

      if (state->mag_img_filter != PIPE_TEX_FILTER_NEAREST) {
         samp.UAddressMagFilterRoundingEnable = true;
         samp.VAddressMagFilterRoundingEnable = true;
         samp.RAddressMagFilterRoundingEnable = true;
      }

      if (state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE)
         samp.ShadowFunction = translate_shadow_func(state->compare_func);

      const float hw_max_lod = GEN_GEN >= 7 ? 14 : 13;

      samp.LODPreClampMode = CLAMP_MODE_OGL;
      samp.MinLOD = CLAMP(min_lod, 0, hw_max_lod);
      samp.MaxLOD = CLAMP(state->max_lod, 0, hw_max_lod);
      samp.TextureLODBias = CLAMP(state->lod_bias, -16, 15);

      /* .BorderColorPointer is filled in by iris_bind_sampler_states. */
   }

   return cso;
}

/**
 * The pipe->bind_sampler_states() driver hook.
 *
 * Now that we know all the sampler states, we upload them all into a
 * contiguous area of GPU memory, for 3DSTATE_SAMPLER_STATE_POINTERS_*.
 * We also fill out the border color state pointers at this point.
 *
 * We could defer this work to draw time, but we assume that binding
 * will be less frequent than drawing.
 */
// XXX: this may be a bad idea, need to make sure that st/mesa calls us
// XXX: with the complete set of shaders.  If it makes multiple calls to
// XXX: things one at a time, we could waste a lot of time assembling things.
// XXX: it doesn't even BUY us anything to do it here, because we only flag
// XXX: IRIS_DIRTY_SAMPLER_STATE when this is called...
static void
iris_bind_sampler_states(struct pipe_context *ctx,
                         enum pipe_shader_type p_stage,
                         unsigned start, unsigned count,
                         void **states)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   gl_shader_stage stage = stage_from_pipe(p_stage);
   struct iris_shader_state *shs = &ice->state.shaders[stage];

   assert(start + count <= IRIS_MAX_TEXTURE_SAMPLERS);
   shs->num_samplers = MAX2(shs->num_samplers, start + count);

   for (int i = 0; i < count; i++) {
      shs->samplers[start + i] = states[i];
   }

   /* Assemble the SAMPLER_STATEs into a contiguous table that lives
    * in the dynamic state memory zone, so we can point to it via the
    * 3DSTATE_SAMPLER_STATE_POINTERS_* commands.
    */
   uint32_t *map =
      upload_state(ice->state.dynamic_uploader, &shs->sampler_table,
                   count * 4 * GENX(SAMPLER_STATE_length), 32);
   if (unlikely(!map))
      return;

   struct pipe_resource *res = shs->sampler_table.res;
   shs->sampler_table.offset +=
      iris_bo_offset_from_base_address(iris_resource_bo(res));

   /* Make sure all land in the same BO */
   iris_border_color_pool_reserve(ice, IRIS_MAX_TEXTURE_SAMPLERS);

   for (int i = 0; i < count; i++) {
      struct iris_sampler_state *state = shs->samplers[i];

      if (!state) {
         memset(map, 0, 4 * GENX(SAMPLER_STATE_length));
      } else if (!state->needs_border_color) {
         memcpy(map, state->sampler_state, 4 * GENX(SAMPLER_STATE_length));
      } else {
         ice->state.need_border_colors = true;

         /* Stream out the border color and merge the pointer. */
         uint32_t offset =
            iris_upload_border_color(ice, &state->border_color);

         uint32_t dynamic[GENX(SAMPLER_STATE_length)];
         iris_pack_state(GENX(SAMPLER_STATE), dynamic, dyns) {
            dyns.BorderColorPointer = offset;
         }

         for (uint32_t j = 0; j < GENX(SAMPLER_STATE_length); j++)
            map[j] = state->sampler_state[j] | dynamic[j];
      }

      map += GENX(SAMPLER_STATE_length);
   }

   ice->state.dirty |= IRIS_DIRTY_SAMPLER_STATES_VS << stage;
}

static enum isl_channel_select
fmt_swizzle(const struct iris_format_info *fmt, enum pipe_swizzle swz)
{
   switch (swz) {
   case PIPE_SWIZZLE_X: return fmt->swizzle.r;
   case PIPE_SWIZZLE_Y: return fmt->swizzle.g;
   case PIPE_SWIZZLE_Z: return fmt->swizzle.b;
   case PIPE_SWIZZLE_W: return fmt->swizzle.a;
   case PIPE_SWIZZLE_1: return SCS_ONE;
   case PIPE_SWIZZLE_0: return SCS_ZERO;
   default: unreachable("invalid swizzle");
   }
}

static void
fill_buffer_surface_state(struct isl_device *isl_dev,
                          struct iris_bo *bo,
                          void *map,
                          enum isl_format format,
                          unsigned offset,
                          unsigned size)
{
   const struct isl_format_layout *fmtl = isl_format_get_layout(format);
   const unsigned cpp = fmtl->bpb / 8;

   /* The ARB_texture_buffer_specification says:
    *
    *    "The number of texels in the buffer texture's texel array is given by
    *
    *       floor(<buffer_size> / (<components> * sizeof(<base_type>)),
    *
    *     where <buffer_size> is the size of the buffer object, in basic
    *     machine units and <components> and <base_type> are the element count
    *     and base data type for elements, as specified in Table X.1.  The
    *     number of texels in the texel array is then clamped to the
    *     implementation-dependent limit MAX_TEXTURE_BUFFER_SIZE_ARB."
    *
    * We need to clamp the size in bytes to MAX_TEXTURE_BUFFER_SIZE * stride,
    * so that when ISL divides by stride to obtain the number of texels, that
    * texel count is clamped to MAX_TEXTURE_BUFFER_SIZE.
    */
   unsigned final_size =
      MIN3(size, bo->size - offset, IRIS_MAX_TEXTURE_BUFFER_SIZE * cpp);

   isl_buffer_fill_state(isl_dev, map,
                         .address = bo->gtt_offset + offset,
                         .size_B = final_size,
                         .format = format,
                         .stride_B = cpp,
                         .mocs = MOCS_WB);
}

/**
 * The pipe->create_sampler_view() driver hook.
 */
static struct pipe_sampler_view *
iris_create_sampler_view(struct pipe_context *ctx,
                         struct pipe_resource *tex,
                         const struct pipe_sampler_view *tmpl)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct iris_sampler_view *isv = calloc(1, sizeof(struct iris_sampler_view));

   if (!isv)
      return NULL;

   /* initialize base object */
   isv->base = *tmpl;
   isv->base.context = ctx;
   isv->base.texture = NULL;
   pipe_reference_init(&isv->base.reference, 1);
   pipe_resource_reference(&isv->base.texture, tex);

   void *map = upload_state(ice->state.surface_uploader, &isv->surface_state,
                            4 * GENX(RENDER_SURFACE_STATE_length), 64);
   if (!unlikely(map))
      return NULL;

   struct iris_bo *state_bo = iris_resource_bo(isv->surface_state.res);
   isv->surface_state.offset += iris_bo_offset_from_base_address(state_bo);

   if (util_format_is_depth_or_stencil(tmpl->format)) {
      struct iris_resource *zres, *sres;
      const struct util_format_description *desc =
         util_format_description(tmpl->format);

      iris_get_depth_stencil_resources(tex, &zres, &sres);

      tex = util_format_has_depth(desc) ? &zres->base : &sres->base;
   }

   isv->res = (struct iris_resource *) tex;

   isl_surf_usage_flags_t usage =
      ISL_SURF_USAGE_TEXTURE_BIT |
      (isv->res->surf.usage & ISL_SURF_USAGE_CUBE_BIT);

   const struct iris_format_info fmt =
      iris_format_for_usage(devinfo, tmpl->format, usage);

   isv->view = (struct isl_view) {
      .format = fmt.fmt,
      .swizzle = (struct isl_swizzle) {
         .r = fmt_swizzle(&fmt, tmpl->swizzle_r),
         .g = fmt_swizzle(&fmt, tmpl->swizzle_g),
         .b = fmt_swizzle(&fmt, tmpl->swizzle_b),
         .a = fmt_swizzle(&fmt, tmpl->swizzle_a),
      },
      .usage = usage,
   };

   /* Fill out SURFACE_STATE for this view. */
   if (tmpl->target != PIPE_BUFFER) {
      isv->view.base_level = tmpl->u.tex.first_level;
      isv->view.levels = tmpl->u.tex.last_level - tmpl->u.tex.first_level + 1;
      isv->view.base_array_layer = tmpl->u.tex.first_layer;
      isv->view.array_len =
         tmpl->u.tex.last_layer - tmpl->u.tex.first_layer + 1;

      isl_surf_fill_state(&screen->isl_dev, map,
                          .surf = &isv->res->surf, .view = &isv->view,
                          .mocs = MOCS_WB,
                          .address = isv->res->bo->gtt_offset);
                          // .aux_surf =
                          // .clear_color = clear_color,
   } else {
      fill_buffer_surface_state(&screen->isl_dev, isv->res->bo, map,
                                isv->view.format, tmpl->u.buf.offset,
                                tmpl->u.buf.size);
   }

   return &isv->base;
}

static void
iris_sampler_view_destroy(struct pipe_context *ctx,
                          struct pipe_sampler_view *state)
{
   struct iris_sampler_view *isv = (void *) state;
   pipe_resource_reference(&state->texture, NULL);
   pipe_resource_reference(&isv->surface_state.res, NULL);
   free(isv);
}

/**
 * The pipe->create_surface() driver hook.
 *
 * In Gallium nomenclature, "surfaces" are a view of a resource that
 * can be bound as a render target or depth/stencil buffer.
 */
static struct pipe_surface *
iris_create_surface(struct pipe_context *ctx,
                    struct pipe_resource *tex,
                    const struct pipe_surface *tmpl)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct iris_surface *surf = calloc(1, sizeof(struct iris_surface));
   struct pipe_surface *psurf = &surf->base;
   struct iris_resource *res = (struct iris_resource *) tex;

   if (!surf)
      return NULL;

   pipe_reference_init(&psurf->reference, 1);
   pipe_resource_reference(&psurf->texture, tex);
   psurf->context = ctx;
   psurf->format = tmpl->format;
   psurf->width = tex->width0;
   psurf->height = tex->height0;
   psurf->texture = tex;
   psurf->u.tex.first_layer = tmpl->u.tex.first_layer;
   psurf->u.tex.last_layer = tmpl->u.tex.last_layer;
   psurf->u.tex.level = tmpl->u.tex.level;

   isl_surf_usage_flags_t usage = 0;
   if (tmpl->writable)
      usage = ISL_SURF_USAGE_STORAGE_BIT;
   else if (util_format_is_depth_or_stencil(tmpl->format))
      usage = ISL_SURF_USAGE_DEPTH_BIT;
   else
      usage = ISL_SURF_USAGE_RENDER_TARGET_BIT;

   const struct iris_format_info fmt =
      iris_format_for_usage(devinfo, psurf->format, usage);

   if ((usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) &&
       !isl_format_supports_rendering(devinfo, fmt.fmt)) {
      /* Framebuffer validation will reject this invalid case, but it
       * hasn't had the opportunity yet.  In the meantime, we need to
       * avoid hitting ISL asserts about unsupported formats below.
       */
      free(surf);
      return NULL;
   }

   surf->view = (struct isl_view) {
      .format = fmt.fmt,
      .base_level = tmpl->u.tex.level,
      .levels = 1,
      .base_array_layer = tmpl->u.tex.first_layer,
      .array_len = tmpl->u.tex.last_layer - tmpl->u.tex.first_layer + 1,
      .swizzle = ISL_SWIZZLE_IDENTITY,
      .usage = usage,
   };

   /* Bail early for depth/stencil - we don't want SURFACE_STATE for them. */
   if (res->surf.usage & (ISL_SURF_USAGE_DEPTH_BIT |
                          ISL_SURF_USAGE_STENCIL_BIT))
      return psurf;


   void *map = upload_state(ice->state.surface_uploader, &surf->surface_state,
                            4 * GENX(RENDER_SURFACE_STATE_length), 64);
   if (!unlikely(map))
      return NULL;

   struct iris_bo *state_bo = iris_resource_bo(surf->surface_state.res);
   surf->surface_state.offset += iris_bo_offset_from_base_address(state_bo);

   isl_surf_fill_state(&screen->isl_dev, map,
                       .surf = &res->surf, .view = &surf->view,
                       .mocs = MOCS_WB,
                       .address = res->bo->gtt_offset);
                       // .aux_surf =
                       // .clear_color = clear_color,

   return psurf;
}

/**
 * The pipe->set_shader_images() driver hook.
 */
static void
iris_set_shader_images(struct pipe_context *ctx,
                       enum pipe_shader_type p_stage,
                       unsigned start_slot, unsigned count,
                       const struct pipe_image_view *p_images)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   gl_shader_stage stage = stage_from_pipe(p_stage);
   struct iris_shader_state *shs = &ice->state.shaders[stage];

   for (unsigned i = 0; i < count; i++) {
      if (p_images && p_images[i].resource) {
         const struct pipe_image_view *img = &p_images[i];
         struct iris_resource *res = (void *) img->resource;
         pipe_resource_reference(&shs->image[start_slot + i].res, &res->base);

         // XXX: these are not retained forever, use a separate uploader?
         void *map =
            upload_state(ice->state.surface_uploader,
                         &shs->image[start_slot + i].surface_state,
                         4 * GENX(RENDER_SURFACE_STATE_length), 64);
         if (!unlikely(map)) {
            pipe_resource_reference(&shs->image[start_slot + i].res, NULL);
            return;
         }

         struct iris_bo *surf_state_bo =
            iris_resource_bo(shs->image[start_slot + i].surface_state.res);
         shs->image[start_slot + i].surface_state.offset +=
            iris_bo_offset_from_base_address(surf_state_bo);

         isl_surf_usage_flags_t usage = ISL_SURF_USAGE_STORAGE_BIT;
         enum isl_format isl_format =
            iris_format_for_usage(devinfo, img->format, usage).fmt;

         if (img->shader_access & PIPE_IMAGE_ACCESS_READ)
            isl_format = isl_lower_storage_image_format(devinfo, isl_format);

         shs->image[start_slot + i].access = img->shader_access;

         if (res->base.target != PIPE_BUFFER) {
            struct isl_view view = {
               .format = isl_format,
               .base_level = img->u.tex.level,
               .levels = 1,
               .base_array_layer = img->u.tex.first_layer,
               .array_len = img->u.tex.last_layer - img->u.tex.first_layer + 1,
               .swizzle = ISL_SWIZZLE_IDENTITY,
               .usage = usage,
            };

            isl_surf_fill_state(&screen->isl_dev, map,
                                .surf = &res->surf, .view = &view,
                                .mocs = MOCS_WB,
                                .address = res->bo->gtt_offset);
                                // .aux_surf =
                                // .clear_color = clear_color,
         } else {
            fill_buffer_surface_state(&screen->isl_dev, res->bo, map,
                                      isl_format, img->u.buf.offset,
                                      img->u.buf.size);
         }
      } else {
         pipe_resource_reference(&shs->image[start_slot + i].res, NULL);
         pipe_resource_reference(&shs->image[start_slot + i].surface_state.res,
                                 NULL);
      }
   }

   ice->state.dirty |= IRIS_DIRTY_BINDINGS_VS << stage;
}


/**
 * The pipe->set_sampler_views() driver hook.
 */
static void
iris_set_sampler_views(struct pipe_context *ctx,
                       enum pipe_shader_type p_stage,
                       unsigned start, unsigned count,
                       struct pipe_sampler_view **views)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   gl_shader_stage stage = stage_from_pipe(p_stage);
   struct iris_shader_state *shs = &ice->state.shaders[stage];

   unsigned i;
   for (i = 0; i < count; i++) {
      pipe_sampler_view_reference((struct pipe_sampler_view **)
                                  &shs->textures[i], views[i]);
   }
   for (; i < shs->num_textures; i++) {
      pipe_sampler_view_reference((struct pipe_sampler_view **)
                                  &shs->textures[i], NULL);
   }

   shs->num_textures = count;

   ice->state.dirty |= (IRIS_DIRTY_BINDINGS_VS << stage);
}

/**
 * The pipe->set_tess_state() driver hook.
 */
static void
iris_set_tess_state(struct pipe_context *ctx,
                    const float default_outer_level[4],
                    const float default_inner_level[2])
{
   struct iris_context *ice = (struct iris_context *) ctx;

   memcpy(&ice->state.default_outer_level[0], &default_outer_level[0], 4 * sizeof(float));
   memcpy(&ice->state.default_inner_level[0], &default_inner_level[0], 2 * sizeof(float));

   ice->state.dirty |= IRIS_DIRTY_CONSTANTS_TCS;
}

static void
iris_surface_destroy(struct pipe_context *ctx, struct pipe_surface *p_surf)
{
   struct iris_surface *surf = (void *) p_surf;
   pipe_resource_reference(&p_surf->texture, NULL);
   pipe_resource_reference(&surf->surface_state.res, NULL);
   free(surf);
}

// XXX: actually implement user clip planes
static void
iris_set_clip_state(struct pipe_context *ctx,
                    const struct pipe_clip_state *state)
{
}

/**
 * The pipe->set_polygon_stipple() driver hook.
 */
static void
iris_set_polygon_stipple(struct pipe_context *ctx,
                         const struct pipe_poly_stipple *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   memcpy(&ice->state.poly_stipple, state, sizeof(*state));
   ice->state.dirty |= IRIS_DIRTY_POLYGON_STIPPLE;
}

/**
 * The pipe->set_sample_mask() driver hook.
 */
static void
iris_set_sample_mask(struct pipe_context *ctx, unsigned sample_mask)
{
   struct iris_context *ice = (struct iris_context *) ctx;

   /* We only support 16x MSAA, so we have 16 bits of sample maks.
    * st/mesa may pass us 0xffffffff though, meaning "enable all samples".
    */
   ice->state.sample_mask = sample_mask & 0xffff;
   ice->state.dirty |= IRIS_DIRTY_SAMPLE_MASK;
}

/**
 * The pipe->set_scissor_states() driver hook.
 *
 * This corresponds to our SCISSOR_RECT state structures.  It's an
 * exact match, so we just store them, and memcpy them out later.
 */
static void
iris_set_scissor_states(struct pipe_context *ctx,
                        unsigned start_slot,
                        unsigned num_scissors,
                        const struct pipe_scissor_state *rects)
{
   struct iris_context *ice = (struct iris_context *) ctx;

   for (unsigned i = 0; i < num_scissors; i++) {
      if (rects[i].minx == rects[i].maxx || rects[i].miny == rects[i].maxy) {
         /* If the scissor was out of bounds and got clamped to 0 width/height
          * at the bounds, the subtraction of 1 from maximums could produce a
          * negative number and thus not clip anything.  Instead, just provide
          * a min > max scissor inside the bounds, which produces the expected
          * no rendering.
          */
         ice->state.scissors[start_slot + i] = (struct pipe_scissor_state) {
            .minx = 1, .maxx = 0, .miny = 1, .maxy = 0,
         };
      } else {
         ice->state.scissors[start_slot + i] = (struct pipe_scissor_state) {
            .minx = rects[i].minx,     .miny = rects[i].miny,
            .maxx = rects[i].maxx - 1, .maxy = rects[i].maxy - 1,
         };
      }
   }

   ice->state.dirty |= IRIS_DIRTY_SCISSOR_RECT;
}

/**
 * The pipe->set_stencil_ref() driver hook.
 *
 * This is added to 3DSTATE_WM_DEPTH_STENCIL dynamically at draw time.
 */
static void
iris_set_stencil_ref(struct pipe_context *ctx,
                     const struct pipe_stencil_ref *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   memcpy(&ice->state.stencil_ref, state, sizeof(*state));
   ice->state.dirty |= IRIS_DIRTY_WM_DEPTH_STENCIL;
}

static float
viewport_extent(const struct pipe_viewport_state *state, int axis, float sign)
{
   return copysignf(state->scale[axis], sign) + state->translate[axis];
}

#if 0
static void
calculate_guardband_size(uint32_t fb_width, uint32_t fb_height,
                         float m00, float m11, float m30, float m31,
                         float *xmin, float *xmax,
                         float *ymin, float *ymax)
{
   /* According to the "Vertex X,Y Clamping and Quantization" section of the
    * Strips and Fans documentation:
    *
    * "The vertex X and Y screen-space coordinates are also /clamped/ to the
    *  fixed-point "guardband" range supported by the rasterization hardware"
    *
    * and
    *
    * "In almost all circumstances, if an object’s vertices are actually
    *  modified by this clamping (i.e., had X or Y coordinates outside of
    *  the guardband extent the rendered object will not match the intended
    *  result.  Therefore software should take steps to ensure that this does
    *  not happen - e.g., by clipping objects such that they do not exceed
    *  these limits after the Drawing Rectangle is applied."
    *
    * I believe the fundamental restriction is that the rasterizer (in
    * the SF/WM stages) have a limit on the number of pixels that can be
    * rasterized.  We need to ensure any coordinates beyond the rasterizer
    * limit are handled by the clipper.  So effectively that limit becomes
    * the clipper's guardband size.
    *
    * It goes on to say:
    *
    * "In addition, in order to be correctly rendered, objects must have a
    *  screenspace bounding box not exceeding 8K in the X or Y direction.
    *  This additional restriction must also be comprehended by software,
    *  i.e., enforced by use of clipping."
    *
    * This makes no sense.  Gen7+ hardware supports 16K render targets,
    * and you definitely need to be able to draw polygons that fill the
    * surface.  Our assumption is that the rasterizer was limited to 8K
    * on Sandybridge, which only supports 8K surfaces, and it was actually
    * increased to 16K on Ivybridge and later.
    *
    * So, limit the guardband to 16K on Gen7+ and 8K on Sandybridge.
    */
   const float gb_size = GEN_GEN >= 7 ? 16384.0f : 8192.0f;

   if (m00 != 0 && m11 != 0) {
      /* First, we compute the screen-space render area */
      const float ss_ra_xmin = MIN3(        0, m30 + m00, m30 - m00);
      const float ss_ra_xmax = MAX3( fb_width, m30 + m00, m30 - m00);
      const float ss_ra_ymin = MIN3(        0, m31 + m11, m31 - m11);
      const float ss_ra_ymax = MAX3(fb_height, m31 + m11, m31 - m11);

      /* We want the guardband to be centered on that */
      const float ss_gb_xmin = (ss_ra_xmin + ss_ra_xmax) / 2 - gb_size;
      const float ss_gb_xmax = (ss_ra_xmin + ss_ra_xmax) / 2 + gb_size;
      const float ss_gb_ymin = (ss_ra_ymin + ss_ra_ymax) / 2 - gb_size;
      const float ss_gb_ymax = (ss_ra_ymin + ss_ra_ymax) / 2 + gb_size;

      /* Now we need it in native device coordinates */
      const float ndc_gb_xmin = (ss_gb_xmin - m30) / m00;
      const float ndc_gb_xmax = (ss_gb_xmax - m30) / m00;
      const float ndc_gb_ymin = (ss_gb_ymin - m31) / m11;
      const float ndc_gb_ymax = (ss_gb_ymax - m31) / m11;

      /* Thanks to Y-flipping and ORIGIN_UPPER_LEFT, the Y coordinates may be
       * flipped upside-down.  X should be fine though.
       */
      assert(ndc_gb_xmin <= ndc_gb_xmax);
      *xmin = ndc_gb_xmin;
      *xmax = ndc_gb_xmax;
      *ymin = MIN2(ndc_gb_ymin, ndc_gb_ymax);
      *ymax = MAX2(ndc_gb_ymin, ndc_gb_ymax);
   } else {
      /* The viewport scales to 0, so nothing will be rendered. */
      *xmin = 0.0f;
      *xmax = 0.0f;
      *ymin = 0.0f;
      *ymax = 0.0f;
   }
}
#endif

/**
 * The pipe->set_viewport_states() driver hook.
 *
 * This corresponds to our SF_CLIP_VIEWPORT states.  We can't calculate
 * the guardband yet, as we need the framebuffer dimensions, but we can
 * at least fill out the rest.
 */
static void
iris_set_viewport_states(struct pipe_context *ctx,
                         unsigned start_slot,
                         unsigned count,
                         const struct pipe_viewport_state *states)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_genx_state *genx = ice->state.genx;
   uint32_t *vp_map =
      &genx->sf_cl_vp[start_slot * GENX(SF_CLIP_VIEWPORT_length)];

   for (unsigned i = 0; i < count; i++) {
      const struct pipe_viewport_state *state = &states[i];

      memcpy(&ice->state.viewports[start_slot + i], state, sizeof(*state));

      iris_pack_state(GENX(SF_CLIP_VIEWPORT), vp_map, vp) {
         vp.ViewportMatrixElementm00 = state->scale[0];
         vp.ViewportMatrixElementm11 = state->scale[1];
         vp.ViewportMatrixElementm22 = state->scale[2];
         vp.ViewportMatrixElementm30 = state->translate[0];
         vp.ViewportMatrixElementm31 = state->translate[1];
         vp.ViewportMatrixElementm32 = state->translate[2];
         /* XXX: in i965 this is computed based on the drawbuffer size,
          * but we don't have that here...
          */
         vp.XMinClipGuardband = -1.0;
         vp.XMaxClipGuardband = 1.0;
         vp.YMinClipGuardband = -1.0;
         vp.YMaxClipGuardband = 1.0;
         vp.XMinViewPort = viewport_extent(state, 0, -1.0f);
         vp.XMaxViewPort = viewport_extent(state, 0,  1.0f) - 1;
         vp.YMinViewPort = viewport_extent(state, 1, -1.0f);
         vp.YMaxViewPort = viewport_extent(state, 1,  1.0f) - 1;
      }

      vp_map += GENX(SF_CLIP_VIEWPORT_length);
   }

   ice->state.dirty |= IRIS_DIRTY_SF_CL_VIEWPORT;

   if (ice->state.cso_rast && (!ice->state.cso_rast->depth_clip_near ||
                               !ice->state.cso_rast->depth_clip_far))
      ice->state.dirty |= IRIS_DIRTY_CC_VIEWPORT;
}

/**
 * The pipe->set_framebuffer_state() driver hook.
 *
 * Sets the current draw FBO, including color render targets, depth,
 * and stencil buffers.
 */
static void
iris_set_framebuffer_state(struct pipe_context *ctx,
                           const struct pipe_framebuffer_state *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   struct isl_device *isl_dev = &screen->isl_dev;
   struct pipe_framebuffer_state *cso = &ice->state.framebuffer;
   struct iris_resource *zres;
   struct iris_resource *stencil_res;

   unsigned samples = util_framebuffer_get_num_samples(state);

   if (cso->samples != samples) {
      ice->state.dirty |= IRIS_DIRTY_MULTISAMPLE;
   }

   if (cso->nr_cbufs != state->nr_cbufs) {
      ice->state.dirty |= IRIS_DIRTY_BLEND_STATE;
   }

   if ((cso->layers == 0) != (state->layers == 0)) {
      ice->state.dirty |= IRIS_DIRTY_CLIP;
   }

   util_copy_framebuffer_state(cso, state);
   cso->samples = samples;

   struct iris_depth_buffer_state *cso_z = &ice->state.genx->depth_buffer;

   struct isl_view view = {
      .base_level = 0,
      .levels = 1,
      .base_array_layer = 0,
      .array_len = 1,
      .swizzle = ISL_SWIZZLE_IDENTITY,
   };

   struct isl_depth_stencil_hiz_emit_info info = {
      .view = &view,
      .mocs = MOCS_WB,
   };

   if (cso->zsbuf) {
      iris_get_depth_stencil_resources(cso->zsbuf->texture, &zres,
                                       &stencil_res);

      view.base_level = cso->zsbuf->u.tex.level;
      view.base_array_layer = cso->zsbuf->u.tex.first_layer;
      view.array_len =
         cso->zsbuf->u.tex.last_layer - cso->zsbuf->u.tex.first_layer + 1;

      if (zres) {
         view.usage |= ISL_SURF_USAGE_DEPTH_BIT;

         info.depth_surf = &zres->surf;
         info.depth_address = zres->bo->gtt_offset;
         info.hiz_usage = ISL_AUX_USAGE_NONE;

         view.format = zres->surf.format;
      }

      if (stencil_res) {
         view.usage |= ISL_SURF_USAGE_STENCIL_BIT;
         info.stencil_surf = &stencil_res->surf;
         info.stencil_address = stencil_res->bo->gtt_offset;
         if (!zres)
            view.format = stencil_res->surf.format;
      }
   }

   isl_emit_depth_stencil_hiz_s(isl_dev, cso_z->packets, &info);

   /* Make a null surface for unbound buffers */
   void *null_surf_map =
      upload_state(ice->state.surface_uploader, &ice->state.null_fb,
                   4 * GENX(RENDER_SURFACE_STATE_length), 64);
   isl_null_fill_state(&screen->isl_dev, null_surf_map,
                       isl_extent3d(MAX2(cso->width, 1),
                                    MAX2(cso->height, 1),
                                    cso->layers ? cso->layers : 1));
   ice->state.null_fb.offset +=
      iris_bo_offset_from_base_address(iris_resource_bo(ice->state.null_fb.res));

   ice->state.dirty |= IRIS_DIRTY_DEPTH_BUFFER;

   /* Render target change */
   ice->state.dirty |= IRIS_DIRTY_BINDINGS_FS;

   ice->state.dirty |= ice->state.dirty_for_nos[IRIS_NOS_FRAMEBUFFER];

#if GEN_GEN == 11
   // XXX: we may want to flag IRIS_DIRTY_MULTISAMPLE (or SAMPLE_MASK?)
   // XXX: see commit 979fc1bc9bcc64027ff2cfafd285676f31b930a6

   /* The PIPE_CONTROL command description says:
    *
    *   "Whenever a Binding Table Index (BTI) used by a Render Target Message
    *    points to a different RENDER_SURFACE_STATE, SW must issue a Render
    *    Target Cache Flush by enabling this bit. When render target flush
    *    is set due to new association of BTI, PS Scoreboard Stall bit must
    *    be set in this packet."
    */
   // XXX: does this need to happen at 3DSTATE_BTP_PS time?
   iris_emit_pipe_control_flush(&ice->render_batch,
                                PIPE_CONTROL_RENDER_TARGET_FLUSH |
                                PIPE_CONTROL_STALL_AT_SCOREBOARD);
#endif
}

/**
 * The pipe->set_constant_buffer() driver hook.
 *
 * This uploads any constant data in user buffers, and references
 * any UBO resources containing constant data.
 */
static void
iris_set_constant_buffer(struct pipe_context *ctx,
                         enum pipe_shader_type p_stage, unsigned index,
                         const struct pipe_constant_buffer *input)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   gl_shader_stage stage = stage_from_pipe(p_stage);
   struct iris_shader_state *shs = &ice->state.shaders[stage];
   struct iris_const_buffer *cbuf = &shs->constbuf[index];

   if (input && (input->buffer || input->user_buffer)) {
      if (input->user_buffer) {
         u_upload_data(ctx->const_uploader, 0, input->buffer_size, 32,
                       input->user_buffer, &cbuf->data.offset,
                       &cbuf->data.res);
      } else {
         pipe_resource_reference(&cbuf->data.res, input->buffer);
         cbuf->data.offset = input->buffer_offset;
      }

      // XXX: these are not retained forever, use a separate uploader?
      void *map =
         upload_state(ice->state.surface_uploader, &cbuf->surface_state,
                      4 * GENX(RENDER_SURFACE_STATE_length), 64);
      if (!unlikely(map)) {
         pipe_resource_reference(&cbuf->data.res, NULL);
         return;
      }

      struct iris_resource *res = (void *) cbuf->data.res;
      struct iris_bo *surf_bo = iris_resource_bo(cbuf->surface_state.res);
      cbuf->surface_state.offset += iris_bo_offset_from_base_address(surf_bo);

      isl_buffer_fill_state(&screen->isl_dev, map,
                            .address = res->bo->gtt_offset + cbuf->data.offset,
                            .size_B = MIN2(input->buffer_size,
                                           res->bo->size - cbuf->data.offset),
                            .format = ISL_FORMAT_R32G32B32A32_FLOAT,
                            .stride_B = 1,
                            .mocs = MOCS_WB)
   } else {
      pipe_resource_reference(&cbuf->data.res, NULL);
      pipe_resource_reference(&cbuf->surface_state.res, NULL);
   }

   ice->state.dirty |= IRIS_DIRTY_CONSTANTS_VS << stage;
   // XXX: maybe not necessary all the time...?
   // XXX: we need 3DS_BTP to commit these changes, and if we fell back to
   // XXX: pull model we may need actual new bindings...
   ice->state.dirty |= IRIS_DIRTY_BINDINGS_VS << stage;
}

/**
 * The pipe->set_shader_buffers() driver hook.
 *
 * This binds SSBOs and ABOs.  Unfortunately, we need to stream out
 * SURFACE_STATE here, as the buffer offset may change each time.
 */
static void
iris_set_shader_buffers(struct pipe_context *ctx,
                        enum pipe_shader_type p_stage,
                        unsigned start_slot, unsigned count,
                        const struct pipe_shader_buffer *buffers)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   gl_shader_stage stage = stage_from_pipe(p_stage);
   struct iris_shader_state *shs = &ice->state.shaders[stage];

   for (unsigned i = 0; i < count; i++) {
      if (buffers && buffers[i].buffer) {
         const struct pipe_shader_buffer *buffer = &buffers[i];
         struct iris_resource *res = (void *) buffer->buffer;
         pipe_resource_reference(&shs->ssbo[start_slot + i], &res->base);

         // XXX: these are not retained forever, use a separate uploader?
         void *map =
            upload_state(ice->state.surface_uploader,
                         &shs->ssbo_surface_state[start_slot + i],
                         4 * GENX(RENDER_SURFACE_STATE_length), 64);
         if (!unlikely(map)) {
            pipe_resource_reference(&shs->ssbo[start_slot + i], NULL);
            return;
         }

         struct iris_bo *surf_state_bo =
            iris_resource_bo(shs->ssbo_surface_state[start_slot + i].res);
         shs->ssbo_surface_state[start_slot + i].offset +=
            iris_bo_offset_from_base_address(surf_state_bo);

         isl_buffer_fill_state(&screen->isl_dev, map,
                               .address =
                                  res->bo->gtt_offset + buffer->buffer_offset,
                               .size_B =
                                  MIN2(buffer->buffer_size,
                                       res->bo->size - buffer->buffer_offset),
                               .format = ISL_FORMAT_RAW,
                               .stride_B = 1,
                               .mocs = MOCS_WB);
      } else {
         pipe_resource_reference(&shs->ssbo[start_slot + i], NULL);
         pipe_resource_reference(&shs->ssbo_surface_state[start_slot + i].res,
                                 NULL);
      }
   }

   ice->state.dirty |= IRIS_DIRTY_BINDINGS_VS << stage;
}

static void
iris_delete_state(struct pipe_context *ctx, void *state)
{
   free(state);
}

static void
iris_free_vertex_buffers(struct iris_vertex_buffer_state *cso)
{
   for (unsigned i = 0; i < cso->num_buffers; i++)
      pipe_resource_reference(&cso->resources[i], NULL);
}

/**
 * The pipe->set_vertex_buffers() driver hook.
 *
 * This translates pipe_vertex_buffer to our 3DSTATE_VERTEX_BUFFERS packet.
 */
static void
iris_set_vertex_buffers(struct pipe_context *ctx,
                        unsigned start_slot, unsigned count,
                        const struct pipe_vertex_buffer *buffers)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_vertex_buffer_state *cso = &ice->state.genx->vertex_buffers;

   iris_free_vertex_buffers(&ice->state.genx->vertex_buffers);

   if (!buffers)
      count = 0;

   cso->num_buffers = count;

   iris_pack_command(GENX(3DSTATE_VERTEX_BUFFERS), cso->vertex_buffers, vb) {
      vb.DWordLength = 4 * MAX2(cso->num_buffers, 1) - 1;
   }

   uint32_t *vb_pack_dest = &cso->vertex_buffers[1];

   if (count == 0) {
      iris_pack_state(GENX(VERTEX_BUFFER_STATE), vb_pack_dest, vb) {
         vb.VertexBufferIndex = start_slot;
         vb.NullVertexBuffer = true;
         vb.AddressModifyEnable = true;
      }
   }

   for (unsigned i = 0; i < count; i++) {
      assert(!buffers[i].is_user_buffer);

      pipe_resource_reference(&cso->resources[i], buffers[i].buffer.resource);
      struct iris_resource *res = (void *) cso->resources[i];

      iris_pack_state(GENX(VERTEX_BUFFER_STATE), vb_pack_dest, vb) {
         vb.VertexBufferIndex = start_slot + i;
         vb.MOCS = MOCS_WB;
         vb.AddressModifyEnable = true;
         vb.BufferPitch = buffers[i].stride;
         if (res) {
            vb.BufferSize = res->bo->size;
            vb.BufferStartingAddress =
               ro_bo(NULL, res->bo->gtt_offset + buffers[i].buffer_offset);
         } else {
            vb.NullVertexBuffer = true;
         }
      }

      vb_pack_dest += GENX(VERTEX_BUFFER_STATE_length);
   }

   ice->state.dirty |= IRIS_DIRTY_VERTEX_BUFFERS;
}

/**
 * Gallium CSO for vertex elements.
 */
struct iris_vertex_element_state {
   uint32_t vertex_elements[1 + 33 * GENX(VERTEX_ELEMENT_STATE_length)];
   uint32_t vf_instancing[33 * GENX(3DSTATE_VF_INSTANCING_length)];
   unsigned count;
};

/**
 * The pipe->create_vertex_elements() driver hook.
 *
 * This translates pipe_vertex_element to our 3DSTATE_VERTEX_ELEMENTS
 * and 3DSTATE_VF_INSTANCING commands.  SGVs are handled at draw time.
 */
static void *
iris_create_vertex_elements(struct pipe_context *ctx,
                            unsigned count,
                            const struct pipe_vertex_element *state)
{
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct iris_vertex_element_state *cso =
      malloc(sizeof(struct iris_vertex_element_state));

   cso->count = count;

   /* TODO:
    *  - create edge flag one
    *  - create SGV ones
    *  - if those are necessary, use count + 1/2/3... OR in the length
    */
   iris_pack_command(GENX(3DSTATE_VERTEX_ELEMENTS), cso->vertex_elements, ve) {
      ve.DWordLength =
         1 + GENX(VERTEX_ELEMENT_STATE_length) * MAX2(count, 1) - 2;
   }

   uint32_t *ve_pack_dest = &cso->vertex_elements[1];
   uint32_t *vfi_pack_dest = cso->vf_instancing;

   if (count == 0) {
      iris_pack_state(GENX(VERTEX_ELEMENT_STATE), ve_pack_dest, ve) {
         ve.Valid = true;
         ve.SourceElementFormat = ISL_FORMAT_R32G32B32A32_FLOAT;
         ve.Component0Control = VFCOMP_STORE_0;
         ve.Component1Control = VFCOMP_STORE_0;
         ve.Component2Control = VFCOMP_STORE_0;
         ve.Component3Control = VFCOMP_STORE_1_FP;
      }

      iris_pack_command(GENX(3DSTATE_VF_INSTANCING), vfi_pack_dest, vi) {
      }
   }

   for (int i = 0; i < count; i++) {
      const struct iris_format_info fmt =
         iris_format_for_usage(devinfo, state[i].src_format, 0);
      unsigned comp[4] = { VFCOMP_STORE_SRC, VFCOMP_STORE_SRC,
                           VFCOMP_STORE_SRC, VFCOMP_STORE_SRC };

      switch (isl_format_get_num_channels(fmt.fmt)) {
      case 0: comp[0] = VFCOMP_STORE_0;
      case 1: comp[1] = VFCOMP_STORE_0;
      case 2: comp[2] = VFCOMP_STORE_0;
      case 3:
         comp[3] = isl_format_has_int_channel(fmt.fmt) ? VFCOMP_STORE_1_INT
                                                       : VFCOMP_STORE_1_FP;
         break;
      }
      iris_pack_state(GENX(VERTEX_ELEMENT_STATE), ve_pack_dest, ve) {
         ve.VertexBufferIndex = state[i].vertex_buffer_index;
         ve.Valid = true;
         ve.SourceElementOffset = state[i].src_offset;
         ve.SourceElementFormat = fmt.fmt;
         ve.Component0Control = comp[0];
         ve.Component1Control = comp[1];
         ve.Component2Control = comp[2];
         ve.Component3Control = comp[3];
      }

      iris_pack_command(GENX(3DSTATE_VF_INSTANCING), vfi_pack_dest, vi) {
         vi.VertexElementIndex = i;
         vi.InstancingEnable = state[i].instance_divisor > 0;
         vi.InstanceDataStepRate = state[i].instance_divisor;
      }

      ve_pack_dest += GENX(VERTEX_ELEMENT_STATE_length);
      vfi_pack_dest += GENX(3DSTATE_VF_INSTANCING_length);
   }

   return cso;
}

/**
 * The pipe->bind_vertex_elements_state() driver hook.
 */
static void
iris_bind_vertex_elements_state(struct pipe_context *ctx, void *state)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_vertex_element_state *old_cso = ice->state.cso_vertex_elements;
   struct iris_vertex_element_state *new_cso = state;

   /* 3DSTATE_VF_SGVs overrides the last VE, so if the count is changing,
    * we need to re-emit it to ensure we're overriding the right one.
    */
   if (new_cso && cso_changed(count))
      ice->state.dirty |= IRIS_DIRTY_VF_SGVS;

   ice->state.cso_vertex_elements = state;
   ice->state.dirty |= IRIS_DIRTY_VERTEX_ELEMENTS;
}

/**
 * Gallium CSO for stream output (transform feedback) targets.
 */
struct iris_stream_output_target {
   struct pipe_stream_output_target base;

   uint32_t so_buffer[GENX(3DSTATE_SO_BUFFER_length)];

   /** Storage holding the offset where we're writing in the buffer */
   struct iris_state_ref offset;
};

/**
 * The pipe->create_stream_output_target() driver hook.
 *
 * "Target" here refers to a destination buffer.  We translate this into
 * a 3DSTATE_SO_BUFFER packet.  We can handle most fields, but don't yet
 * know which buffer this represents, or whether we ought to zero the
 * write-offsets, or append.  Those are handled in the set() hook.
 */
static struct pipe_stream_output_target *
iris_create_stream_output_target(struct pipe_context *ctx,
                                 struct pipe_resource *res,
                                 unsigned buffer_offset,
                                 unsigned buffer_size)
{
   struct iris_stream_output_target *cso = calloc(1, sizeof(*cso));
   if (!cso)
      return NULL;

   pipe_reference_init(&cso->base.reference, 1);
   pipe_resource_reference(&cso->base.buffer, res);
   cso->base.buffer_offset = buffer_offset;
   cso->base.buffer_size = buffer_size;
   cso->base.context = ctx;

   upload_state(ctx->stream_uploader, &cso->offset, 4 * sizeof(uint32_t), 4);

   iris_pack_command(GENX(3DSTATE_SO_BUFFER), cso->so_buffer, sob) {
      sob.SurfaceBaseAddress =
         rw_bo(NULL, iris_resource_bo(res)->gtt_offset + buffer_offset);
      sob.SOBufferEnable = true;
      sob.StreamOffsetWriteEnable = true;
      sob.StreamOutputBufferOffsetAddressEnable = true;
      sob.MOCS = MOCS_WB; // XXX: MOCS

      sob.SurfaceSize = MAX2(buffer_size / 4, 1) - 1;

      /* .SOBufferIndex, .StreamOffset, and .StreamOutputBufferOffsetAddress
       * are filled in later when we have stream IDs.
       */
   }

   return &cso->base;
}

static void
iris_stream_output_target_destroy(struct pipe_context *ctx,
                                  struct pipe_stream_output_target *state)
{
   struct iris_stream_output_target *cso = (void *) state;

   pipe_resource_reference(&cso->base.buffer, NULL);
   pipe_resource_reference(&cso->offset.res, NULL);

   free(cso);
}

/**
 * The pipe->set_stream_output_targets() driver hook.
 *
 * At this point, we know which targets are bound to a particular index,
 * and also whether we want to append or start over.  We can finish the
 * 3DSTATE_SO_BUFFER packets we started earlier.
 */
static void
iris_set_stream_output_targets(struct pipe_context *ctx,
                               unsigned num_targets,
                               struct pipe_stream_output_target **targets,
                               const unsigned *offsets)
{
   struct iris_context *ice = (struct iris_context *) ctx;
   struct iris_genx_state *genx = ice->state.genx;
   uint32_t *so_buffers = genx->so_buffers;

   const bool active = num_targets > 0;
   if (ice->state.streamout_active != active) {
      ice->state.streamout_active = active;
      ice->state.dirty |= IRIS_DIRTY_STREAMOUT;

      /* We only emit 3DSTATE_SO_DECL_LIST when streamout is active, because
       * it's a non-pipelined command.  If we're switching streamout on, we
       * may have missed emitting it earlier, so do so now.  (We're already
       * taking a stall to update 3DSTATE_SO_BUFFERS anyway...)
       */
      if (active)
         ice->state.dirty |= IRIS_DIRTY_SO_DECL_LIST;
   }

   for (int i = 0; i < 4; i++) {
      pipe_so_target_reference(&ice->state.so_target[i],
                               i < num_targets ? targets[i] : NULL);
   }

   /* No need to update 3DSTATE_SO_BUFFER unless SOL is active. */
   if (!active)
      return;

   for (unsigned i = 0; i < 4; i++,
        so_buffers += GENX(3DSTATE_SO_BUFFER_length)) {

      if (i >= num_targets || !targets[i]) {
         iris_pack_command(GENX(3DSTATE_SO_BUFFER), so_buffers, sob)
            sob.SOBufferIndex = i;
         continue;
      }

      struct iris_stream_output_target *tgt = (void *) targets[i];

      /* Note that offsets[i] will either be 0, causing us to zero
       * the value in the buffer, or 0xFFFFFFFF, which happens to mean
       * "continue appending at the existing offset."
       */
      assert(offsets[i] == 0 || offsets[i] == 0xFFFFFFFF);

      uint32_t dynamic[GENX(3DSTATE_SO_BUFFER_length)];
      iris_pack_state(GENX(3DSTATE_SO_BUFFER), dynamic, dyns) {
         dyns.SOBufferIndex = i;
         dyns.StreamOffset = offsets[i];
         dyns.StreamOutputBufferOffsetAddress =
            rw_bo(NULL, iris_resource_bo(tgt->offset.res)->gtt_offset + tgt->offset.offset + i * sizeof(uint32_t));
      }

      for (uint32_t j = 0; j < GENX(3DSTATE_SO_BUFFER_length); j++) {
         so_buffers[j] = tgt->so_buffer[j] | dynamic[j];
      }
   }

   ice->state.dirty |= IRIS_DIRTY_SO_BUFFERS;
}

/**
 * An iris-vtable helper for encoding the 3DSTATE_SO_DECL_LIST and
 * 3DSTATE_STREAMOUT packets.
 *
 * 3DSTATE_SO_DECL_LIST is a list of shader outputs we want the streamout
 * hardware to record.  We can create it entirely based on the shader, with
 * no dynamic state dependencies.
 *
 * 3DSTATE_STREAMOUT is an annoying mix of shader-based information and
 * state-based settings.  We capture the shader-related ones here, and merge
 * the rest in at draw time.
 */
static uint32_t *
iris_create_so_decl_list(const struct pipe_stream_output_info *info,
                         const struct brw_vue_map *vue_map)
{
   struct GENX(SO_DECL) so_decl[MAX_VERTEX_STREAMS][128];
   int buffer_mask[MAX_VERTEX_STREAMS] = {0, 0, 0, 0};
   int next_offset[MAX_VERTEX_STREAMS] = {0, 0, 0, 0};
   int decls[MAX_VERTEX_STREAMS] = {0, 0, 0, 0};
   int max_decls = 0;
   STATIC_ASSERT(ARRAY_SIZE(so_decl[0]) >= MAX_PROGRAM_OUTPUTS);

   memset(so_decl, 0, sizeof(so_decl));

   /* Construct the list of SO_DECLs to be emitted.  The formatting of the
    * command feels strange -- each dword pair contains a SO_DECL per stream.
    */
   for (unsigned i = 0; i < info->num_outputs; i++) {
      const struct pipe_stream_output *output = &info->output[i];
      const int buffer = output->output_buffer;
      const int varying = output->register_index;
      const unsigned stream_id = output->stream;
      assert(stream_id < MAX_VERTEX_STREAMS);

      buffer_mask[stream_id] |= 1 << buffer;

      assert(vue_map->varying_to_slot[varying] >= 0);

      /* Mesa doesn't store entries for gl_SkipComponents in the Outputs[]
       * array.  Instead, it simply increments DstOffset for the following
       * input by the number of components that should be skipped.
       *
       * Our hardware is unusual in that it requires us to program SO_DECLs
       * for fake "hole" components, rather than simply taking the offset
       * for each real varying.  Each hole can have size 1, 2, 3, or 4; we
       * program as many size = 4 holes as we can, then a final hole to
       * accommodate the final 1, 2, or 3 remaining.
       */
      int skip_components = output->dst_offset - next_offset[buffer];

      while (skip_components > 0) {
         so_decl[stream_id][decls[stream_id]++] = (struct GENX(SO_DECL)) {
            .HoleFlag = 1,
            .OutputBufferSlot = output->output_buffer,
            .ComponentMask = (1 << MIN2(skip_components, 4)) - 1,
         };
         skip_components -= 4;
      }

      next_offset[buffer] = output->dst_offset + output->num_components;

      so_decl[stream_id][decls[stream_id]++] = (struct GENX(SO_DECL)) {
         .OutputBufferSlot = output->output_buffer,
         .RegisterIndex = vue_map->varying_to_slot[varying],
         .ComponentMask =
            ((1 << output->num_components) - 1) << output->start_component,
      };

      if (decls[stream_id] > max_decls)
         max_decls = decls[stream_id];
   }

   unsigned dwords = GENX(3DSTATE_STREAMOUT_length) + (3 + 2 * max_decls);
   uint32_t *map = ralloc_size(NULL, sizeof(uint32_t) * dwords);
   uint32_t *so_decl_map = map + GENX(3DSTATE_STREAMOUT_length);

   iris_pack_command(GENX(3DSTATE_STREAMOUT), map, sol) {
      int urb_entry_read_offset = 0;
      int urb_entry_read_length = (vue_map->num_slots + 1) / 2 -
         urb_entry_read_offset;

      /* We always read the whole vertex.  This could be reduced at some
       * point by reading less and offsetting the register index in the
       * SO_DECLs.
       */
      sol.Stream0VertexReadOffset = urb_entry_read_offset;
      sol.Stream0VertexReadLength = urb_entry_read_length - 1;
      sol.Stream1VertexReadOffset = urb_entry_read_offset;
      sol.Stream1VertexReadLength = urb_entry_read_length - 1;
      sol.Stream2VertexReadOffset = urb_entry_read_offset;
      sol.Stream2VertexReadLength = urb_entry_read_length - 1;
      sol.Stream3VertexReadOffset = urb_entry_read_offset;
      sol.Stream3VertexReadLength = urb_entry_read_length - 1;

      /* Set buffer pitches; 0 means unbound. */
      sol.Buffer0SurfacePitch = 4 * info->stride[0];
      sol.Buffer1SurfacePitch = 4 * info->stride[1];
      sol.Buffer2SurfacePitch = 4 * info->stride[2];
      sol.Buffer3SurfacePitch = 4 * info->stride[3];
   }

   iris_pack_command(GENX(3DSTATE_SO_DECL_LIST), so_decl_map, list) {
      list.DWordLength = 3 + 2 * max_decls - 2;
      list.StreamtoBufferSelects0 = buffer_mask[0];
      list.StreamtoBufferSelects1 = buffer_mask[1];
      list.StreamtoBufferSelects2 = buffer_mask[2];
      list.StreamtoBufferSelects3 = buffer_mask[3];
      list.NumEntries0 = decls[0];
      list.NumEntries1 = decls[1];
      list.NumEntries2 = decls[2];
      list.NumEntries3 = decls[3];
   }

   for (int i = 0; i < max_decls; i++) {
      iris_pack_state(GENX(SO_DECL_ENTRY), so_decl_map + 3 + i * 2, entry) {
         entry.Stream0Decl = so_decl[0][i];
         entry.Stream1Decl = so_decl[1][i];
         entry.Stream2Decl = so_decl[2][i];
         entry.Stream3Decl = so_decl[3][i];
      }
   }

   return map;
}

static void
iris_compute_sbe_urb_read_interval(uint64_t fs_input_slots,
                                   const struct brw_vue_map *last_vue_map,
                                   bool two_sided_color,
                                   unsigned *out_offset,
                                   unsigned *out_length)
{
   /* The compiler computes the first URB slot without considering COL/BFC
    * swizzling (because it doesn't know whether it's enabled), so we need
    * to do that here too.  This may result in a smaller offset, which
    * should be safe.
    */
   const unsigned first_slot =
      brw_compute_first_urb_slot_required(fs_input_slots, last_vue_map);

   /* This becomes the URB read offset (counted in pairs of slots). */
   assert(first_slot % 2 == 0);
   *out_offset = first_slot / 2;

   /* We need to adjust the inputs read to account for front/back color
    * swizzling, as it can make the URB length longer.
    */
   for (int c = 0; c <= 1; c++) {
      if (fs_input_slots & (VARYING_BIT_COL0 << c)) {
         /* If two sided color is enabled, the fragment shader's gl_Color
          * (COL0) input comes from either the gl_FrontColor (COL0) or
          * gl_BackColor (BFC0) input varyings.  Mark BFC as used, too.
          */
         if (two_sided_color)
            fs_input_slots |= (VARYING_BIT_BFC0 << c);

         /* If front color isn't written, we opt to give them back color
          * instead of an undefined value.  Switch from COL to BFC.
          */
         if (last_vue_map->varying_to_slot[VARYING_SLOT_COL0 + c] == -1) {
            fs_input_slots &= ~(VARYING_BIT_COL0 << c);
            fs_input_slots |= (VARYING_BIT_BFC0 << c);
         }
      }
   }

   /* Compute the minimum URB Read Length necessary for the FS inputs.
    *
    * From the Sandy Bridge PRM, Volume 2, Part 1, documentation for
    * 3DSTATE_SF DWord 1 bits 15:11, "Vertex URB Entry Read Length":
    *
    * "This field should be set to the minimum length required to read the
    *  maximum source attribute.  The maximum source attribute is indicated
    *  by the maximum value of the enabled Attribute # Source Attribute if
    *  Attribute Swizzle Enable is set, Number of Output Attributes-1 if
    *  enable is not set.
    *  read_length = ceiling((max_source_attr + 1) / 2)
    *
    *  [errata] Corruption/Hang possible if length programmed larger than
    *  recommended"
    *
    * Similar text exists for Ivy Bridge.
    *
    * We find the last URB slot that's actually read by the FS.
    */
   unsigned last_read_slot = last_vue_map->num_slots - 1;
   while (last_read_slot > first_slot && !(fs_input_slots &
          (1ull << last_vue_map->slot_to_varying[last_read_slot])))
      --last_read_slot;

   /* The URB read length is the difference of the two, counted in pairs. */
   *out_length = DIV_ROUND_UP(last_read_slot - first_slot + 1, 2);
}

static void
iris_emit_sbe_swiz(struct iris_batch *batch,
                   const struct iris_context *ice,
                   unsigned urb_read_offset,
                   unsigned sprite_coord_enables)
{
   struct GENX(SF_OUTPUT_ATTRIBUTE_DETAIL) attr_overrides[16] = {};
   const struct brw_wm_prog_data *wm_prog_data = (void *)
      ice->shaders.prog[MESA_SHADER_FRAGMENT]->prog_data;
   const struct brw_vue_map *vue_map = ice->shaders.last_vue_map;
   const struct iris_rasterizer_state *cso_rast = ice->state.cso_rast;

   /* XXX: this should be generated when putting programs in place */

   // XXX: raster->sprite_coord_enable

   for (int fs_attr = 0; fs_attr < VARYING_SLOT_MAX; fs_attr++) {
      const int input_index = wm_prog_data->urb_setup[fs_attr];
      if (input_index < 0 || input_index >= 16)
         continue;

      struct GENX(SF_OUTPUT_ATTRIBUTE_DETAIL) *attr =
         &attr_overrides[input_index];
      int slot = vue_map->varying_to_slot[fs_attr];

      /* Viewport and Layer are stored in the VUE header.  We need to override
       * them to zero if earlier stages didn't write them, as GL requires that
       * they read back as zero when not explicitly set.
       */
      switch (fs_attr) {
      case VARYING_SLOT_VIEWPORT:
      case VARYING_SLOT_LAYER:
         attr->ComponentOverrideX = true;
         attr->ComponentOverrideW = true;
         attr->ConstantSource = CONST_0000;

         if (!(vue_map->slots_valid & VARYING_BIT_LAYER))
            attr->ComponentOverrideY = true;
         if (!(vue_map->slots_valid & VARYING_BIT_VIEWPORT))
            attr->ComponentOverrideZ = true;
         continue;

      case VARYING_SLOT_PRIMITIVE_ID:
         /* Override if the previous shader stage didn't write gl_PrimitiveID. */
         if (slot == -1) {
            attr->ComponentOverrideX = true;
            attr->ComponentOverrideY = true;
            attr->ComponentOverrideZ = true;
            attr->ComponentOverrideW = true;
            attr->ConstantSource = PRIM_ID;
            continue;
         }

      default:
         break;
      }

      if (sprite_coord_enables & (1 << input_index))
         continue;

      /* If there was only a back color written but not front, use back
       * as the color instead of undefined.
       */
      if (slot == -1 && fs_attr == VARYING_SLOT_COL0)
         slot = vue_map->varying_to_slot[VARYING_SLOT_BFC0];
      if (slot == -1 && fs_attr == VARYING_SLOT_COL1)
         slot = vue_map->varying_to_slot[VARYING_SLOT_BFC1];

      /* Not written by the previous stage - undefined. */
      if (slot == -1) {
         attr->ComponentOverrideX = true;
         attr->ComponentOverrideY = true;
         attr->ComponentOverrideZ = true;
         attr->ComponentOverrideW = true;
         attr->ConstantSource = CONST_0001_FLOAT;
         continue;
      }

      /* Compute the location of the attribute relative to the read offset,
       * which is counted in 256-bit increments (two 128-bit VUE slots).
       */
      const int source_attr = slot - 2 * urb_read_offset;
      assert(source_attr >= 0 && source_attr <= 32);
      attr->SourceAttribute = source_attr;

      /* If we are doing two-sided color, and the VUE slot following this one
       * represents a back-facing color, then we need to instruct the SF unit
       * to do back-facing swizzling.
       */
      if (cso_rast->light_twoside &&
          ((vue_map->slot_to_varying[slot] == VARYING_SLOT_COL0 &&
            vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC0) ||
           (vue_map->slot_to_varying[slot] == VARYING_SLOT_COL1 &&
            vue_map->slot_to_varying[slot+1] == VARYING_SLOT_BFC1)))
         attr->SwizzleSelect = INPUTATTR_FACING;
   }

   iris_emit_cmd(batch, GENX(3DSTATE_SBE_SWIZ), sbes) {
      for (int i = 0; i < 16; i++)
         sbes.Attribute[i] = attr_overrides[i];
   }
}

static unsigned
iris_calculate_point_sprite_overrides(const struct brw_wm_prog_data *prog_data,
                                      const struct iris_rasterizer_state *cso)
{
   unsigned overrides = 0;

   if (prog_data->urb_setup[VARYING_SLOT_PNTC] != -1)
      overrides |= 1 << prog_data->urb_setup[VARYING_SLOT_PNTC];

   for (int i = 0; i < 8; i++) {
      if ((cso->sprite_coord_enable & (1 << i)) &&
          prog_data->urb_setup[VARYING_SLOT_TEX0 + i] != -1)
         overrides |= 1 << prog_data->urb_setup[VARYING_SLOT_TEX0 + i];
   }

   return overrides;
}

static void
iris_emit_sbe(struct iris_batch *batch, const struct iris_context *ice)
{
   const struct iris_rasterizer_state *cso_rast = ice->state.cso_rast;
   const struct brw_wm_prog_data *wm_prog_data = (void *)
      ice->shaders.prog[MESA_SHADER_FRAGMENT]->prog_data;
   const struct shader_info *fs_info =
      iris_get_shader_info(ice, MESA_SHADER_FRAGMENT);

   unsigned urb_read_offset, urb_read_length;
   iris_compute_sbe_urb_read_interval(fs_info->inputs_read,
                                      ice->shaders.last_vue_map,
                                      cso_rast->light_twoside,
                                      &urb_read_offset, &urb_read_length);

   unsigned sprite_coord_overrides =
      iris_calculate_point_sprite_overrides(wm_prog_data, cso_rast);

   iris_emit_cmd(batch, GENX(3DSTATE_SBE), sbe) {
      sbe.AttributeSwizzleEnable = true;
      sbe.NumberofSFOutputAttributes = wm_prog_data->num_varying_inputs;
      sbe.PointSpriteTextureCoordinateOrigin = cso_rast->sprite_coord_mode;
      sbe.VertexURBEntryReadOffset = urb_read_offset;
      sbe.VertexURBEntryReadLength = urb_read_length;
      sbe.ForceVertexURBEntryReadOffset = true;
      sbe.ForceVertexURBEntryReadLength = true;
      sbe.ConstantInterpolationEnable = wm_prog_data->flat_inputs;
      sbe.PointSpriteTextureCoordinateEnable = sprite_coord_overrides;

      for (int i = 0; i < 32; i++) {
         sbe.AttributeActiveComponentFormat[i] = ACTIVE_COMPONENT_XYZW;
      }
   }

   iris_emit_sbe_swiz(batch, ice, urb_read_offset, sprite_coord_overrides);
}

/* ------------------------------------------------------------------- */

/**
 * Set sampler-related program key fields based on the current state.
 */
static void
iris_populate_sampler_key(const struct iris_context *ice,
                          struct brw_sampler_prog_key_data *key)
{
   for (int i = 0; i < MAX_SAMPLERS; i++) {
      key->swizzles[i] = 0x688; /* XYZW */
   }
}

/**
 * Populate VS program key fields based on the current state.
 */
static void
iris_populate_vs_key(const struct iris_context *ice,
                     struct brw_vs_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);
}

/**
 * Populate TCS program key fields based on the current state.
 */
static void
iris_populate_tcs_key(const struct iris_context *ice,
                      struct brw_tcs_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);
}

/**
 * Populate TES program key fields based on the current state.
 */
static void
iris_populate_tes_key(const struct iris_context *ice,
                      struct brw_tes_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);
}

/**
 * Populate GS program key fields based on the current state.
 */
static void
iris_populate_gs_key(const struct iris_context *ice,
                     struct brw_gs_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);
}

/**
 * Populate FS program key fields based on the current state.
 */
static void
iris_populate_fs_key(const struct iris_context *ice,
                     struct brw_wm_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);

   /* XXX: dirty flags? */
   const struct pipe_framebuffer_state *fb = &ice->state.framebuffer;
   const struct iris_depth_stencil_alpha_state *zsa = ice->state.cso_zsa;
   const struct iris_rasterizer_state *rast = ice->state.cso_rast;
   const struct iris_blend_state *blend = ice->state.cso_blend;

   key->nr_color_regions = fb->nr_cbufs;

   key->clamp_fragment_color = rast->clamp_fragment_color;

   key->replicate_alpha = fb->nr_cbufs > 1 &&
      (zsa->alpha.enabled || blend->alpha_to_coverage);

   /* XXX: only bother if COL0/1 are read */
   key->flat_shade = rast->flatshade;

   key->persample_interp = rast->force_persample_interp;
   key->multisample_fbo = rast->multisample && fb->samples > 1;

   key->coherent_fb_fetch = true;

   // XXX: uint64_t input_slots_valid; - for >16 inputs

   // XXX: key->force_dual_color_blend for unigine
   // XXX: respect hint for high_quality_derivatives:1;
}

static void
iris_populate_cs_key(const struct iris_context *ice,
                     struct brw_cs_prog_key *key)
{
   iris_populate_sampler_key(ice, &key->tex);
}

#if 0
   // XXX: these need to go in INIT_THREAD_DISPATCH_FIELDS
   pkt.SamplerCount =                                                     \
      DIV_ROUND_UP(CLAMP(stage_state->sampler_count, 0, 16), 4);          \

#endif

static uint64_t
KSP(const struct iris_compiled_shader *shader)
{
   struct iris_resource *res = (void *) shader->assembly.res;
   return iris_bo_offset_from_base_address(res->bo) + shader->assembly.offset;
}

// Gen11 workaround table #2056 WABTPPrefetchDisable suggests to disable
// prefetching of binding tables in A0 and B0 steppings.  XXX: Revisit
// this WA on C0 stepping.

#define INIT_THREAD_DISPATCH_FIELDS(pkt, prefix, stage)                   \
   pkt.KernelStartPointer = KSP(shader);                                  \
   pkt.BindingTableEntryCount = GEN_GEN == 11 ? 0 :                       \
      prog_data->binding_table.size_bytes / 4;                            \
   pkt.FloatingPointMode = prog_data->use_alt_mode;                       \
                                                                          \
   pkt.DispatchGRFStartRegisterForURBData =                               \
      prog_data->dispatch_grf_start_reg;                                  \
   pkt.prefix##URBEntryReadLength = vue_prog_data->urb_read_length;       \
   pkt.prefix##URBEntryReadOffset = 0;                                    \
                                                                          \
   pkt.StatisticsEnable = true;                                           \
   pkt.Enable           = true;                                           \
                                                                          \
   if (prog_data->total_scratch) {                                        \
      uint32_t scratch_addr =                                             \
         iris_get_scratch_space(ice, prog_data->total_scratch, stage);    \
      pkt.PerThreadScratchSpace = ffs(prog_data->total_scratch) - 11;     \
      pkt.ScratchSpaceBasePointer = rw_bo(NULL, scratch_addr);            \
   }

/**
 * Encode most of 3DSTATE_VS based on the compiled shader.
 */
static void
iris_store_vs_state(struct iris_context *ice,
                    const struct gen_device_info *devinfo,
                    struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_vue_prog_data *vue_prog_data = (void *) prog_data;

   iris_pack_command(GENX(3DSTATE_VS), shader->derived_data, vs) {
      INIT_THREAD_DISPATCH_FIELDS(vs, Vertex, MESA_SHADER_VERTEX);
      vs.MaximumNumberofThreads = devinfo->max_vs_threads - 1;
      vs.SIMD8DispatchEnable = true;
      vs.UserClipDistanceCullTestEnableBitmask =
         vue_prog_data->cull_distance_mask;
   }
}

/**
 * Encode most of 3DSTATE_HS based on the compiled shader.
 */
static void
iris_store_tcs_state(struct iris_context *ice,
                     const struct gen_device_info *devinfo,
                     struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_vue_prog_data *vue_prog_data = (void *) prog_data;
   struct brw_tcs_prog_data *tcs_prog_data = (void *) prog_data;

   iris_pack_command(GENX(3DSTATE_HS), shader->derived_data, hs) {
      INIT_THREAD_DISPATCH_FIELDS(hs, Vertex, MESA_SHADER_TESS_CTRL);

      hs.InstanceCount = tcs_prog_data->instances - 1;
      hs.MaximumNumberofThreads = devinfo->max_tcs_threads - 1;
      hs.IncludeVertexHandles = true;
   }
}

/**
 * Encode 3DSTATE_TE and most of 3DSTATE_DS based on the compiled shader.
 */
static void
iris_store_tes_state(struct iris_context *ice,
                     const struct gen_device_info *devinfo,
                     struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_vue_prog_data *vue_prog_data = (void *) prog_data;
   struct brw_tes_prog_data *tes_prog_data = (void *) prog_data;

   uint32_t *te_state = (void *) shader->derived_data;
   uint32_t *ds_state = te_state + GENX(3DSTATE_TE_length);

   iris_pack_command(GENX(3DSTATE_TE), te_state, te) {
      te.Partitioning = tes_prog_data->partitioning;
      te.OutputTopology = tes_prog_data->output_topology;
      te.TEDomain = tes_prog_data->domain;
      te.TEEnable = true;
      te.MaximumTessellationFactorOdd = 63.0;
      te.MaximumTessellationFactorNotOdd = 64.0;
   }

   iris_pack_command(GENX(3DSTATE_DS), ds_state, ds) {
      INIT_THREAD_DISPATCH_FIELDS(ds, Patch, MESA_SHADER_TESS_EVAL);

      ds.DispatchMode = DISPATCH_MODE_SIMD8_SINGLE_PATCH;
      ds.MaximumNumberofThreads = devinfo->max_tes_threads - 1;
      ds.ComputeWCoordinateEnable =
         tes_prog_data->domain == BRW_TESS_DOMAIN_TRI;

      ds.UserClipDistanceCullTestEnableBitmask =
         vue_prog_data->cull_distance_mask;
   }

}

/**
 * Encode most of 3DSTATE_GS based on the compiled shader.
 */
static void
iris_store_gs_state(struct iris_context *ice,
                    const struct gen_device_info *devinfo,
                    struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_vue_prog_data *vue_prog_data = (void *) prog_data;
   struct brw_gs_prog_data *gs_prog_data = (void *) prog_data;

   iris_pack_command(GENX(3DSTATE_GS), shader->derived_data, gs) {
      INIT_THREAD_DISPATCH_FIELDS(gs, Vertex, MESA_SHADER_GEOMETRY);

      gs.OutputVertexSize = gs_prog_data->output_vertex_size_hwords * 2 - 1;
      gs.OutputTopology = gs_prog_data->output_topology;
      gs.ControlDataHeaderSize =
         gs_prog_data->control_data_header_size_hwords;
      gs.InstanceControl = gs_prog_data->invocations - 1;
      gs.DispatchMode = DISPATCH_MODE_SIMD8;
      gs.IncludePrimitiveID = gs_prog_data->include_primitive_id;
      gs.ControlDataFormat = gs_prog_data->control_data_format;
      gs.ReorderMode = TRAILING;
      gs.ExpectedVertexCount = gs_prog_data->vertices_in;
      gs.MaximumNumberofThreads =
         GEN_GEN == 8 ? (devinfo->max_gs_threads / 2 - 1)
                      : (devinfo->max_gs_threads - 1);

      if (gs_prog_data->static_vertex_count != -1) {
         gs.StaticOutput = true;
         gs.StaticOutputVertexCount = gs_prog_data->static_vertex_count;
      }
      gs.IncludeVertexHandles = vue_prog_data->include_vue_handles;

      gs.UserClipDistanceCullTestEnableBitmask =
         vue_prog_data->cull_distance_mask;

      const int urb_entry_write_offset = 1;
      const uint32_t urb_entry_output_length =
         DIV_ROUND_UP(vue_prog_data->vue_map.num_slots, 2) -
         urb_entry_write_offset;

      gs.VertexURBEntryOutputReadOffset = urb_entry_write_offset;
      gs.VertexURBEntryOutputLength = MAX2(urb_entry_output_length, 1);
   }
}

/**
 * Encode most of 3DSTATE_PS and 3DSTATE_PS_EXTRA based on the shader.
 */
static void
iris_store_fs_state(struct iris_context *ice,
                    const struct gen_device_info *devinfo,
                    struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_wm_prog_data *wm_prog_data = (void *) shader->prog_data;

   uint32_t *ps_state = (void *) shader->derived_data;
   uint32_t *psx_state = ps_state + GENX(3DSTATE_PS_length);

   iris_pack_command(GENX(3DSTATE_PS), ps_state, ps) {
      ps.VectorMaskEnable = true;
      //ps.SamplerCount = ...
      // XXX: WABTPPrefetchDisable, see above, drop at C0
      ps.BindingTableEntryCount = GEN_GEN == 11 ? 0 :
         prog_data->binding_table.size_bytes / 4;
      ps.FloatingPointMode = prog_data->use_alt_mode;
      ps.MaximumNumberofThreadsPerPSD = 64 - (GEN_GEN == 8 ? 2 : 1);

      ps.PushConstantEnable = prog_data->nr_params > 0 ||
                              prog_data->ubo_ranges[0].length > 0;

      /* From the documentation for this packet:
       * "If the PS kernel does not need the Position XY Offsets to
       *  compute a Position Value, then this field should be programmed
       *  to POSOFFSET_NONE."
       *
       * "SW Recommendation: If the PS kernel needs the Position Offsets
       *  to compute a Position XY value, this field should match Position
       *  ZW Interpolation Mode to ensure a consistent position.xyzw
       *  computation."
       *
       * We only require XY sample offsets. So, this recommendation doesn't
       * look useful at the moment.  We might need this in future.
       */
      ps.PositionXYOffsetSelect =
         wm_prog_data->uses_pos_offset ? POSOFFSET_SAMPLE : POSOFFSET_NONE;
      ps._8PixelDispatchEnable = wm_prog_data->dispatch_8;
      ps._16PixelDispatchEnable = wm_prog_data->dispatch_16;
      ps._32PixelDispatchEnable = wm_prog_data->dispatch_32;

      // XXX: Disable SIMD32 with 16x MSAA

      ps.DispatchGRFStartRegisterForConstantSetupData0 =
         brw_wm_prog_data_dispatch_grf_start_reg(wm_prog_data, ps, 0);
      ps.DispatchGRFStartRegisterForConstantSetupData1 =
         brw_wm_prog_data_dispatch_grf_start_reg(wm_prog_data, ps, 1);
      ps.DispatchGRFStartRegisterForConstantSetupData2 =
         brw_wm_prog_data_dispatch_grf_start_reg(wm_prog_data, ps, 2);

      ps.KernelStartPointer0 =
         KSP(shader) + brw_wm_prog_data_prog_offset(wm_prog_data, ps, 0);
      ps.KernelStartPointer1 =
         KSP(shader) + brw_wm_prog_data_prog_offset(wm_prog_data, ps, 1);
      ps.KernelStartPointer2 =
         KSP(shader) + brw_wm_prog_data_prog_offset(wm_prog_data, ps, 2);

      if (prog_data->total_scratch) {
         uint32_t scratch_addr =
            iris_get_scratch_space(ice, prog_data->total_scratch,
                                   MESA_SHADER_FRAGMENT);
         ps.PerThreadScratchSpace = ffs(prog_data->total_scratch) - 11;
         ps.ScratchSpaceBasePointer = rw_bo(NULL, scratch_addr);
      }
   }

   iris_pack_command(GENX(3DSTATE_PS_EXTRA), psx_state, psx) {
      psx.PixelShaderValid = true;
      psx.PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode;
      psx.PixelShaderKillsPixel = wm_prog_data->uses_kill;
      psx.AttributeEnable = wm_prog_data->num_varying_inputs != 0;
      psx.PixelShaderUsesSourceDepth = wm_prog_data->uses_src_depth;
      psx.PixelShaderUsesSourceW = wm_prog_data->uses_src_w;
      psx.PixelShaderIsPerSample = wm_prog_data->persample_dispatch;

      if (wm_prog_data->uses_sample_mask) {
         /* TODO: conservative rasterization */
         if (wm_prog_data->post_depth_coverage)
            psx.InputCoverageMaskState = ICMS_DEPTH_COVERAGE;
         else
            psx.InputCoverageMaskState = ICMS_NORMAL;
      }

      psx.oMaskPresenttoRenderTarget = wm_prog_data->uses_omask;
      psx.PixelShaderPullsBary = wm_prog_data->pulls_bary;
      psx.PixelShaderComputesStencil = wm_prog_data->computed_stencil;

      // XXX: UAV bit
   }
}

/**
 * Compute the size of the derived data (shader command packets).
 *
 * This must match the data written by the iris_store_xs_state() functions.
 */
static void
iris_store_cs_state(struct iris_context *ice,
                    const struct gen_device_info *devinfo,
                    struct iris_compiled_shader *shader)
{
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_cs_prog_data *cs_prog_data = (void *) shader->prog_data;
   void *map = shader->derived_data;

   iris_pack_state(GENX(INTERFACE_DESCRIPTOR_DATA), map, desc) {
      desc.KernelStartPointer = KSP(shader);
      desc.ConstantURBEntryReadLength = cs_prog_data->push.per_thread.regs;
      desc.NumberofThreadsinGPGPUThreadGroup = cs_prog_data->threads;
      desc.SharedLocalMemorySize =
         encode_slm_size(GEN_GEN, prog_data->total_shared);
      desc.BarrierEnable = cs_prog_data->uses_barrier;
      desc.CrossThreadConstantDataReadLength =
         cs_prog_data->push.cross_thread.regs;
   }
}

static unsigned
iris_derived_program_state_size(enum iris_program_cache_id cache_id)
{
   assert(cache_id <= IRIS_CACHE_BLORP);

   static const unsigned dwords[] = {
      [IRIS_CACHE_VS] = GENX(3DSTATE_VS_length),
      [IRIS_CACHE_TCS] = GENX(3DSTATE_HS_length),
      [IRIS_CACHE_TES] = GENX(3DSTATE_TE_length) + GENX(3DSTATE_DS_length),
      [IRIS_CACHE_GS] = GENX(3DSTATE_GS_length),
      [IRIS_CACHE_FS] =
         GENX(3DSTATE_PS_length) + GENX(3DSTATE_PS_EXTRA_length),
      [IRIS_CACHE_CS] = GENX(INTERFACE_DESCRIPTOR_DATA_length),
      [IRIS_CACHE_BLORP] = 0,
   };

   return sizeof(uint32_t) * dwords[cache_id];
}

/**
 * Create any state packets corresponding to the given shader stage
 * (i.e. 3DSTATE_VS) and save them as "derived data" in the shader variant.
 * This means that we can look up a program in the in-memory cache and
 * get most of the state packet without having to reconstruct it.
 */
static void
iris_store_derived_program_state(struct iris_context *ice,
                                 enum iris_program_cache_id cache_id,
                                 struct iris_compiled_shader *shader)
{
   struct iris_screen *screen = (void *) ice->ctx.screen;
   const struct gen_device_info *devinfo = &screen->devinfo;

   switch (cache_id) {
   case IRIS_CACHE_VS:
      iris_store_vs_state(ice, devinfo, shader);
      break;
   case IRIS_CACHE_TCS:
      iris_store_tcs_state(ice, devinfo, shader);
      break;
   case IRIS_CACHE_TES:
      iris_store_tes_state(ice, devinfo, shader);
      break;
   case IRIS_CACHE_GS:
      iris_store_gs_state(ice, devinfo, shader);
      break;
   case IRIS_CACHE_FS:
      iris_store_fs_state(ice, devinfo, shader);
      break;
   case IRIS_CACHE_CS:
      iris_store_cs_state(ice, devinfo, shader);
   case IRIS_CACHE_BLORP:
      break;
   default:
      break;
   }
}

/* ------------------------------------------------------------------- */

/**
 * Configure the URB.
 *
 * XXX: write a real comment.
 */
static void
iris_upload_urb_config(struct iris_context *ice, struct iris_batch *batch)
{
   const struct gen_device_info *devinfo = &batch->screen->devinfo;
   const unsigned push_size_kB = 32;
   unsigned entries[4];
   unsigned start[4];
   unsigned size[4];

   for (int i = MESA_SHADER_VERTEX; i <= MESA_SHADER_GEOMETRY; i++) {
      if (!ice->shaders.prog[i]) {
         size[i] = 1;
      } else {
         struct brw_vue_prog_data *vue_prog_data =
            (void *) ice->shaders.prog[i]->prog_data;
         size[i] = vue_prog_data->urb_entry_size;
      }
      assert(size[i] != 0);
   }

   gen_get_urb_config(devinfo, 1024 * push_size_kB,
                      1024 * ice->shaders.urb_size,
                      ice->shaders.prog[MESA_SHADER_TESS_EVAL] != NULL,
                      ice->shaders.prog[MESA_SHADER_GEOMETRY] != NULL,
                      size, entries, start);

   for (int i = MESA_SHADER_VERTEX; i <= MESA_SHADER_GEOMETRY; i++) {
      iris_emit_cmd(batch, GENX(3DSTATE_URB_VS), urb) {
         urb._3DCommandSubOpcode += i;
         urb.VSURBStartingAddress     = start[i];
         urb.VSURBEntryAllocationSize = size[i] - 1;
         urb.VSNumberofURBEntries     = entries[i];
      }
   }
}

static const uint32_t push_constant_opcodes[] = {
   [MESA_SHADER_VERTEX]    = 21,
   [MESA_SHADER_TESS_CTRL] = 25, /* HS */
   [MESA_SHADER_TESS_EVAL] = 26, /* DS */
   [MESA_SHADER_GEOMETRY]  = 22,
   [MESA_SHADER_FRAGMENT]  = 23,
   [MESA_SHADER_COMPUTE]   = 0,
};

static uint32_t
use_null_surface(struct iris_batch *batch, struct iris_context *ice)
{
   struct iris_bo *state_bo = iris_resource_bo(ice->state.unbound_tex.res);

   iris_use_pinned_bo(batch, state_bo, false);

   return ice->state.unbound_tex.offset;
}

static uint32_t
use_null_fb_surface(struct iris_batch *batch, struct iris_context *ice)
{
   /* If set_framebuffer_state() was never called, fall back to 1x1x1 */
   if (!ice->state.null_fb.res)
      return use_null_surface(batch, ice);

   struct iris_bo *state_bo = iris_resource_bo(ice->state.null_fb.res);

   iris_use_pinned_bo(batch, state_bo, false);

   return ice->state.null_fb.offset;
}

/**
 * Add a surface to the validation list, as well as the buffer containing
 * the corresponding SURFACE_STATE.
 *
 * Returns the binding table entry (offset to SURFACE_STATE).
 */
static uint32_t
use_surface(struct iris_batch *batch,
            struct pipe_surface *p_surf,
            bool writeable)
{
   struct iris_surface *surf = (void *) p_surf;

   iris_use_pinned_bo(batch, iris_resource_bo(p_surf->texture), writeable);
   iris_use_pinned_bo(batch, iris_resource_bo(surf->surface_state.res), false);

   return surf->surface_state.offset;
}

static uint32_t
use_sampler_view(struct iris_batch *batch, struct iris_sampler_view *isv)
{
   iris_use_pinned_bo(batch, isv->res->bo, false);
   iris_use_pinned_bo(batch, iris_resource_bo(isv->surface_state.res), false);

   return isv->surface_state.offset;
}

static uint32_t
use_const_buffer(struct iris_batch *batch,
                 struct iris_context *ice,
                 struct iris_const_buffer *cbuf)
{
   if (!cbuf->surface_state.res)
      return use_null_surface(batch, ice);

   iris_use_pinned_bo(batch, iris_resource_bo(cbuf->data.res), false);
   iris_use_pinned_bo(batch, iris_resource_bo(cbuf->surface_state.res), false);

   return cbuf->surface_state.offset;
}

static uint32_t
use_ssbo(struct iris_batch *batch, struct iris_context *ice,
         struct iris_shader_state *shs, int i)
{
   if (!shs->ssbo[i])
      return use_null_surface(batch, ice);

   struct iris_state_ref *surf_state = &shs->ssbo_surface_state[i];

   iris_use_pinned_bo(batch, iris_resource_bo(shs->ssbo[i]), true);
   iris_use_pinned_bo(batch, iris_resource_bo(surf_state->res), false);

   return surf_state->offset;
}

static uint32_t
use_image(struct iris_batch *batch, struct iris_context *ice,
          struct iris_shader_state *shs, int i)
{
   if (!shs->image[i].res)
      return use_null_surface(batch, ice);

   struct iris_state_ref *surf_state = &shs->image[i].surface_state;

   iris_use_pinned_bo(batch, iris_resource_bo(shs->image[i].res),
                      shs->image[i].access & PIPE_IMAGE_ACCESS_WRITE);
   iris_use_pinned_bo(batch, iris_resource_bo(surf_state->res), false);

   return surf_state->offset;
}

#define push_bt_entry(addr) \
   assert(addr >= binder_addr); \
   if (!pin_only) bt_map[s++] = (addr) - binder_addr;

/**
 * Populate the binding table for a given shader stage.
 *
 * This fills out the table of pointers to surfaces required by the shader,
 * and also adds those buffers to the validation list so the kernel can make
 * resident before running our batch.
 */
static void
iris_populate_binding_table(struct iris_context *ice,
                            struct iris_batch *batch,
                            gl_shader_stage stage,
                            bool pin_only)
{
   const struct iris_binder *binder = &ice->state.binder;
   struct iris_compiled_shader *shader = ice->shaders.prog[stage];
   if (!shader)
      return;

   struct iris_shader_state *shs = &ice->state.shaders[stage];
   uint32_t binder_addr = binder->bo->gtt_offset;

   //struct brw_stage_prog_data *prog_data = (void *) shader->prog_data;
   uint32_t *bt_map = binder->map + binder->bt_offset[stage];
   int s = 0;

   const struct shader_info *info = iris_get_shader_info(ice, stage);
   if (!info) {
      /* TCS passthrough doesn't need a binding table. */
      assert(stage == MESA_SHADER_TESS_CTRL);
      return;
   }

   if (stage == MESA_SHADER_COMPUTE) {
      /* surface for gl_NumWorkGroups */
      struct iris_state_ref *grid_data = &ice->state.grid_size;
      struct iris_state_ref *grid_state = &ice->state.grid_surf_state;
      iris_use_pinned_bo(batch, iris_resource_bo(grid_data->res),  false);
      iris_use_pinned_bo(batch, iris_resource_bo(grid_state->res), false);
      push_bt_entry(grid_state->offset);
   }

   if (stage == MESA_SHADER_FRAGMENT) {
      struct pipe_framebuffer_state *cso_fb = &ice->state.framebuffer;
      /* Note that cso_fb->nr_cbufs == fs_key->nr_color_regions. */
      if (cso_fb->nr_cbufs) {
         for (unsigned i = 0; i < cso_fb->nr_cbufs; i++) {
            uint32_t addr =
               cso_fb->cbufs[i] ? use_surface(batch, cso_fb->cbufs[i], true)
                                : use_null_fb_surface(batch, ice);
            push_bt_entry(addr);
         }
      } else {
         uint32_t addr = use_null_fb_surface(batch, ice);
         push_bt_entry(addr);
      }
   }

   //assert(prog_data->binding_table.texture_start ==
          //(ice->state.num_textures[stage] ? s : 0xd0d0d0d0));

   for (int i = 0; i < shs->num_textures; i++) {
      struct iris_sampler_view *view = shs->textures[i];
      uint32_t addr = view ? use_sampler_view(batch, view)
                           : use_null_surface(batch, ice);
      push_bt_entry(addr);
   }

   for (int i = 0; i < info->num_images; i++) {
      uint32_t addr = use_image(batch, ice, shs, i);
      push_bt_entry(addr);
   }

   const int num_ubos = iris_get_shader_num_ubos(ice, stage);

   for (int i = 0; i < num_ubos; i++) {
      uint32_t addr = use_const_buffer(batch, ice, &shs->constbuf[i]);
      push_bt_entry(addr);
   }

   /* XXX: st is wasting 16 binding table slots for ABOs.  Should add a cap
    * for changing nir_lower_atomics_to_ssbos setting and buffer_base offset
    * in st_atom_storagebuf.c so it'll compact them into one range, with
    * SSBOs starting at info->num_abos.  Ideally it'd reset num_abos to 0 too
    */
   if (info->num_abos + info->num_ssbos > 0) {
      for (int i = 0; i < IRIS_MAX_ABOS + info->num_ssbos; i++) {
         uint32_t addr = use_ssbo(batch, ice, shs, i);
         push_bt_entry(addr);
      }
   }

#if 0
      // XXX: not implemented yet
      assert(prog_data->binding_table.plane_start[1] == 0xd0d0d0d0);
      assert(prog_data->binding_table.plane_start[2] == 0xd0d0d0d0);
#endif
}

static void
iris_use_optional_res(struct iris_batch *batch,
                      struct pipe_resource *res,
                      bool writeable)
{
   if (res) {
      struct iris_bo *bo = iris_resource_bo(res);
      iris_use_pinned_bo(batch, bo, writeable);
   }
}

/* ------------------------------------------------------------------- */

/**
 * Pin any BOs which were installed by a previous batch, and restored
 * via the hardware logical context mechanism.
 *
 * We don't need to re-emit all state every batch - the hardware context
 * mechanism will save and restore it for us.  This includes pointers to
 * various BOs...which won't exist unless we ask the kernel to pin them
 * by adding them to the validation list.
 *
 * We can skip buffers if we've re-emitted those packets, as we're
 * overwriting those stale pointers with new ones, and don't actually
 * refer to the old BOs.
 */
static void
iris_restore_render_saved_bos(struct iris_context *ice,
                              struct iris_batch *batch,
                              const struct pipe_draw_info *draw)
{
   // XXX: whack IRIS_SHADER_DIRTY_BINDING_TABLE on new batch

   const uint64_t clean = ~ice->state.dirty;

   if (clean & IRIS_DIRTY_CC_VIEWPORT) {
      iris_use_optional_res(batch, ice->state.last_res.cc_vp, false);
   }

   if (clean & IRIS_DIRTY_SF_CL_VIEWPORT) {
      iris_use_optional_res(batch, ice->state.last_res.sf_cl_vp, false);
   }

   if (clean & IRIS_DIRTY_BLEND_STATE) {
      iris_use_optional_res(batch, ice->state.last_res.blend, false);
   }

   if (clean & IRIS_DIRTY_COLOR_CALC_STATE) {
      iris_use_optional_res(batch, ice->state.last_res.color_calc, false);
   }

   if (clean & IRIS_DIRTY_SCISSOR_RECT) {
      iris_use_optional_res(batch, ice->state.last_res.scissor, false);
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (!(clean & (IRIS_DIRTY_CONSTANTS_VS << stage)))
         continue;

      struct iris_shader_state *shs = &ice->state.shaders[stage];
      struct iris_compiled_shader *shader = ice->shaders.prog[stage];

      if (!shader)
         continue;

      struct brw_stage_prog_data *prog_data = (void *) shader->prog_data;

      for (int i = 0; i < 4; i++) {
         const struct brw_ubo_range *range = &prog_data->ubo_ranges[i];

         if (range->length == 0)
            continue;

         struct iris_const_buffer *cbuf = &shs->constbuf[range->block];
         struct iris_resource *res = (void *) cbuf->data.res;

         if (res)
            iris_use_pinned_bo(batch, res->bo, false);
         else
            iris_use_pinned_bo(batch, batch->screen->workaround_bo, false);
      }
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (clean & (IRIS_DIRTY_BINDINGS_VS << stage)) {
         /* Re-pin any buffers referred to by the binding table. */
         iris_populate_binding_table(ice, batch, stage, true);
      }
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      struct iris_shader_state *shs = &ice->state.shaders[stage];
      struct pipe_resource *res = shs->sampler_table.res;
      if (res)
         iris_use_pinned_bo(batch, iris_resource_bo(res), false);
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (clean & (IRIS_DIRTY_VS << stage)) {
         struct iris_compiled_shader *shader = ice->shaders.prog[stage];
         if (shader) {
            struct iris_bo *bo = iris_resource_bo(shader->assembly.res);
            iris_use_pinned_bo(batch, bo, false);
         }

         // XXX: scratch buffer
      }
   }

   if (clean & IRIS_DIRTY_DEPTH_BUFFER) {
      struct pipe_framebuffer_state *cso_fb = &ice->state.framebuffer;

      if (cso_fb->zsbuf) {
         struct iris_resource *zres, *sres;
         iris_get_depth_stencil_resources(cso_fb->zsbuf->texture,
                                          &zres, &sres);
         // XXX: might not be writable...
         if (zres)
            iris_use_pinned_bo(batch, zres->bo, true);
         if (sres)
            iris_use_pinned_bo(batch, sres->bo, true);
      }
   }

   if (draw->index_size == 0 && ice->state.last_res.index_buffer) {
      /* This draw didn't emit a new index buffer, so we are inheriting the
       * older index buffer.  This draw didn't need it, but future ones may.
       */
      struct iris_bo *bo = iris_resource_bo(ice->state.last_res.index_buffer);
      iris_use_pinned_bo(batch, bo, false);
   }

   if (clean & IRIS_DIRTY_VERTEX_BUFFERS) {
      struct iris_vertex_buffer_state *cso = &ice->state.genx->vertex_buffers;
      for (unsigned i = 0; i < cso->num_buffers; i++) {
         struct iris_resource *res = (void *) cso->resources[i];
         iris_use_pinned_bo(batch, res->bo, false);
      }
   }
}

static void
iris_restore_compute_saved_bos(struct iris_context *ice,
                               struct iris_batch *batch,
                               const struct pipe_grid_info *grid)
{
   const uint64_t clean = ~ice->state.dirty;

   const int stage = MESA_SHADER_COMPUTE;
   struct iris_shader_state *shs = &ice->state.shaders[stage];

   if (clean & IRIS_DIRTY_CONSTANTS_CS) {
      struct iris_compiled_shader *shader = ice->shaders.prog[stage];

      if (shader) {
         struct brw_stage_prog_data *prog_data = (void *) shader->prog_data;
         const struct brw_ubo_range *range = &prog_data->ubo_ranges[0];

         if (range->length > 0) {
            struct iris_const_buffer *cbuf = &shs->constbuf[range->block];
            struct iris_resource *res = (void *) cbuf->data.res;

            if (res)
               iris_use_pinned_bo(batch, res->bo, false);
            else
               iris_use_pinned_bo(batch, batch->screen->workaround_bo, false);
           }
      }
   }

   if (clean & IRIS_DIRTY_BINDINGS_CS) {
      /* Re-pin any buffers referred to by the binding table. */
      iris_populate_binding_table(ice, batch, stage, true);
   }

   struct pipe_resource *sampler_res = shs->sampler_table.res;
   if (sampler_res)
      iris_use_pinned_bo(batch, iris_resource_bo(sampler_res), false);

   if (clean & IRIS_DIRTY_CS) {
      struct iris_compiled_shader *shader = ice->shaders.prog[stage];
      if (shader) {
         struct iris_bo *bo = iris_resource_bo(shader->assembly.res);
         iris_use_pinned_bo(batch, bo, false);
      }

      // XXX: scratch buffer
   }
}

/**
 * Possibly emit STATE_BASE_ADDRESS to update Surface State Base Address.
 */
static void
iris_update_surface_base_address(struct iris_batch *batch,
                                 struct iris_binder *binder)
{
   if (batch->last_surface_base_address == binder->bo->gtt_offset)
      return;

   flush_for_state_base_change(batch);

   iris_emit_cmd(batch, GENX(STATE_BASE_ADDRESS), sba) {
      // XXX: sba.SurfaceStateMemoryObjectControlState = MOCS_WB;
      sba.SurfaceStateBaseAddressModifyEnable = true;
      sba.SurfaceStateBaseAddress = ro_bo(binder->bo, 0);
   }

   batch->last_surface_base_address = binder->bo->gtt_offset;
}

static void
iris_upload_dirty_render_state(struct iris_context *ice,
                               struct iris_batch *batch,
                               const struct pipe_draw_info *draw)
{
   const uint64_t dirty = ice->state.dirty;

   if (!(dirty & IRIS_ALL_DIRTY_FOR_RENDER))
      return;

   struct iris_genx_state *genx = ice->state.genx;
   struct iris_binder *binder = &ice->state.binder;
   struct brw_wm_prog_data *wm_prog_data = (void *)
      ice->shaders.prog[MESA_SHADER_FRAGMENT]->prog_data;

   if (dirty & IRIS_DIRTY_CC_VIEWPORT) {
      const struct iris_rasterizer_state *cso_rast = ice->state.cso_rast;
      uint32_t cc_vp_address;

      /* XXX: could avoid streaming for depth_clip [0,1] case. */
      uint32_t *cc_vp_map =
         stream_state(batch, ice->state.dynamic_uploader,
                      &ice->state.last_res.cc_vp,
                      4 * ice->state.num_viewports *
                      GENX(CC_VIEWPORT_length), 32, &cc_vp_address);
      for (int i = 0; i < ice->state.num_viewports; i++) {
         float zmin, zmax;
         util_viewport_zmin_zmax(&ice->state.viewports[i],
                                 cso_rast->clip_halfz, &zmin, &zmax);
         if (cso_rast->depth_clip_near)
            zmin = 0.0;
         if (cso_rast->depth_clip_far)
            zmax = 1.0;

         iris_pack_state(GENX(CC_VIEWPORT), cc_vp_map, ccv) {
            ccv.MinimumDepth = zmin;
            ccv.MaximumDepth = zmax;
         }

         cc_vp_map += GENX(CC_VIEWPORT_length);
      }

      iris_emit_cmd(batch, GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC), ptr) {
         ptr.CCViewportPointer = cc_vp_address;
      }
   }

   if (dirty & IRIS_DIRTY_SF_CL_VIEWPORT) {
      iris_emit_cmd(batch, GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP), ptr) {
         ptr.SFClipViewportPointer =
            emit_state(batch, ice->state.dynamic_uploader,
                       &ice->state.last_res.sf_cl_vp,
                       genx->sf_cl_vp, 4 * GENX(SF_CLIP_VIEWPORT_length) *
                       ice->state.num_viewports, 64);
      }
   }

   /* XXX: L3 State */

   // XXX: this is only flagged at setup, we assume a static configuration
   if (dirty & IRIS_DIRTY_URB) {
      iris_upload_urb_config(ice, batch);
   }

   if (dirty & IRIS_DIRTY_BLEND_STATE) {
      struct iris_blend_state *cso_blend = ice->state.cso_blend;
      struct pipe_framebuffer_state *cso_fb = &ice->state.framebuffer;
      struct iris_depth_stencil_alpha_state *cso_zsa = ice->state.cso_zsa;
      const int header_dwords = GENX(BLEND_STATE_length);
      const int rt_dwords = cso_fb->nr_cbufs * GENX(BLEND_STATE_ENTRY_length);
      uint32_t blend_offset;
      uint32_t *blend_map =
         stream_state(batch, ice->state.dynamic_uploader,
                      &ice->state.last_res.blend,
                      4 * (header_dwords + rt_dwords), 64, &blend_offset);

      uint32_t blend_state_header;
      iris_pack_state(GENX(BLEND_STATE), &blend_state_header, bs) {
         bs.AlphaTestEnable = cso_zsa->alpha.enabled;
         bs.AlphaTestFunction = translate_compare_func(cso_zsa->alpha.func);
      }

      blend_map[0] = blend_state_header | cso_blend->blend_state[0];
      memcpy(&blend_map[1], &cso_blend->blend_state[1], 4 * rt_dwords);

      iris_emit_cmd(batch, GENX(3DSTATE_BLEND_STATE_POINTERS), ptr) {
         ptr.BlendStatePointer = blend_offset;
         ptr.BlendStatePointerValid = true;
      }
   }

   if (dirty & IRIS_DIRTY_COLOR_CALC_STATE) {
      struct iris_depth_stencil_alpha_state *cso = ice->state.cso_zsa;
      uint32_t cc_offset;
      void *cc_map =
         stream_state(batch, ice->state.dynamic_uploader,
                      &ice->state.last_res.color_calc,
                      sizeof(uint32_t) * GENX(COLOR_CALC_STATE_length),
                      64, &cc_offset);
      iris_pack_state(GENX(COLOR_CALC_STATE), cc_map, cc) {
         cc.AlphaTestFormat = ALPHATEST_FLOAT32;
         cc.AlphaReferenceValueAsFLOAT32 = cso->alpha.ref_value;
         cc.BlendConstantColorRed   = ice->state.blend_color.color[0];
         cc.BlendConstantColorGreen = ice->state.blend_color.color[1];
         cc.BlendConstantColorBlue  = ice->state.blend_color.color[2];
         cc.BlendConstantColorAlpha = ice->state.blend_color.color[3];
      }
      iris_emit_cmd(batch, GENX(3DSTATE_CC_STATE_POINTERS), ptr) {
         ptr.ColorCalcStatePointer = cc_offset;
         ptr.ColorCalcStatePointerValid = true;
      }
   }

   /* Upload constants for TCS passthrough. */
   if ((dirty & IRIS_DIRTY_CONSTANTS_TCS) &&
       ice->shaders.prog[MESA_SHADER_TESS_CTRL] &&
       !ice->shaders.uncompiled[MESA_SHADER_TESS_CTRL]) {
      struct iris_compiled_shader *tes_shader = ice->shaders.prog[MESA_SHADER_TESS_EVAL];
      assert(tes_shader);

      /* Passthrough always copies 2 vec4s, so when uploading data we ensure
       * it is in the right layout for TES.
       */
      float hdr[8] = {};
      struct brw_tes_prog_data *tes_prog_data = (void *) tes_shader->prog_data;
      switch (tes_prog_data->domain) {
      case BRW_TESS_DOMAIN_QUAD:
         for (int i = 0; i < 4; i++)
            hdr[7 - i] = ice->state.default_outer_level[i];
         hdr[3] = ice->state.default_inner_level[0];
         hdr[2] = ice->state.default_inner_level[1];
         break;
      case BRW_TESS_DOMAIN_TRI:
         for (int i = 0; i < 3; i++)
            hdr[7 - i] = ice->state.default_outer_level[i];
         hdr[4] = ice->state.default_inner_level[0];
         break;
      case BRW_TESS_DOMAIN_ISOLINE:
         hdr[7] = ice->state.default_outer_level[1];
         hdr[6] = ice->state.default_outer_level[0];
         break;
      }

      struct iris_shader_state *shs = &ice->state.shaders[MESA_SHADER_TESS_CTRL];
      struct iris_const_buffer *cbuf = &shs->constbuf[0];
      u_upload_data(ice->ctx.const_uploader, 0, sizeof(hdr), 32,
                    &hdr[0], &cbuf->data.offset,
                    &cbuf->data.res);
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (!(dirty & (IRIS_DIRTY_CONSTANTS_VS << stage)))
         continue;

      struct iris_shader_state *shs = &ice->state.shaders[stage];
      struct iris_compiled_shader *shader = ice->shaders.prog[stage];

      if (!shader)
         continue;

      struct brw_stage_prog_data *prog_data = (void *) shader->prog_data;

      iris_emit_cmd(batch, GENX(3DSTATE_CONSTANT_VS), pkt) {
         pkt._3DCommandSubOpcode = push_constant_opcodes[stage];
         if (prog_data) {
            /* The Skylake PRM contains the following restriction:
             *
             *    "The driver must ensure The following case does not occur
             *     without a flush to the 3D engine: 3DSTATE_CONSTANT_* with
             *     buffer 3 read length equal to zero committed followed by a
             *     3DSTATE_CONSTANT_* with buffer 0 read length not equal to
             *     zero committed."
             *
             * To avoid this, we program the buffers in the highest slots.
             * This way, slot 0 is only used if slot 3 is also used.
             */
            int n = 3;

            for (int i = 3; i >= 0; i--) {
               const struct brw_ubo_range *range = &prog_data->ubo_ranges[i];

               if (range->length == 0)
                  continue;

               struct iris_const_buffer *cbuf = &shs->constbuf[range->block];
               struct iris_resource *res = (void *) cbuf->data.res;

               assert(cbuf->data.offset % 32 == 0);

               pkt.ConstantBody.ReadLength[n] = range->length;
               pkt.ConstantBody.Buffer[n] =
                  res ? ro_bo(res->bo, range->start * 32 + cbuf->data.offset)
                      : ro_bo(batch->screen->workaround_bo, 0);
               n--;
            }
         }
      }
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (dirty & (IRIS_DIRTY_BINDINGS_VS << stage)) {
         iris_emit_cmd(batch, GENX(3DSTATE_BINDING_TABLE_POINTERS_VS), ptr) {
            ptr._3DCommandSubOpcode = 38 + stage;
            ptr.PointertoVSBindingTable = binder->bt_offset[stage];
         }
      }
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (dirty & (IRIS_DIRTY_BINDINGS_VS << stage)) {
         iris_populate_binding_table(ice, batch, stage, false);
      }
   }

   if (ice->state.need_border_colors)
      iris_use_pinned_bo(batch, ice->state.border_color_pool.bo, false);

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (!(dirty & (IRIS_DIRTY_SAMPLER_STATES_VS << stage)) ||
          !ice->shaders.prog[stage])
         continue;

      struct iris_shader_state *shs = &ice->state.shaders[stage];
      struct pipe_resource *res = shs->sampler_table.res;
      if (res)
         iris_use_pinned_bo(batch, iris_resource_bo(res), false);

      iris_emit_cmd(batch, GENX(3DSTATE_SAMPLER_STATE_POINTERS_VS), ptr) {
         ptr._3DCommandSubOpcode = 43 + stage;
         ptr.PointertoVSSamplerState = shs->sampler_table.offset;
      }
   }

   if (dirty & IRIS_DIRTY_MULTISAMPLE) {
      iris_emit_cmd(batch, GENX(3DSTATE_MULTISAMPLE), ms) {
         ms.PixelLocation =
            ice->state.cso_rast->half_pixel_center ? CENTER : UL_CORNER;
         if (ice->state.framebuffer.samples > 0)
            ms.NumberofMultisamples = ffs(ice->state.framebuffer.samples) - 1;
      }
   }

   if (dirty & IRIS_DIRTY_SAMPLE_MASK) {
      iris_emit_cmd(batch, GENX(3DSTATE_SAMPLE_MASK), ms) {
         ms.SampleMask = MAX2(ice->state.sample_mask, 1);
      }
   }

   for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
      if (!(dirty & (IRIS_DIRTY_VS << stage)))
         continue;

      struct iris_compiled_shader *shader = ice->shaders.prog[stage];

      if (shader) {
         struct iris_resource *cache = (void *) shader->assembly.res;
         iris_use_pinned_bo(batch, cache->bo, false);
         iris_batch_emit(batch, shader->derived_data,
                         iris_derived_program_state_size(stage));
      } else {
         if (stage == MESA_SHADER_TESS_EVAL) {
            iris_emit_cmd(batch, GENX(3DSTATE_HS), hs);
            iris_emit_cmd(batch, GENX(3DSTATE_TE), te);
            iris_emit_cmd(batch, GENX(3DSTATE_DS), ds);
         } else if (stage == MESA_SHADER_GEOMETRY) {
            iris_emit_cmd(batch, GENX(3DSTATE_GS), gs);
         }
      }
   }

   if (ice->state.streamout_active) {
      if (dirty & IRIS_DIRTY_SO_BUFFERS) {
         iris_batch_emit(batch, genx->so_buffers,
                         4 * 4 * GENX(3DSTATE_SO_BUFFER_length));
         for (int i = 0; i < 4; i++) {
            struct iris_stream_output_target *tgt =
               (void *) ice->state.so_target[i];
            if (tgt) {
               iris_use_pinned_bo(batch, iris_resource_bo(tgt->base.buffer),
                                  true);
               iris_use_pinned_bo(batch, iris_resource_bo(tgt->offset.res),
                                  true);
            }
         }
      }

      if ((dirty & IRIS_DIRTY_SO_DECL_LIST) && ice->state.streamout) {
         uint32_t *decl_list =
            ice->state.streamout + GENX(3DSTATE_STREAMOUT_length);
         iris_batch_emit(batch, decl_list, 4 * ((decl_list[0] & 0xff) + 2));
      }

      if (dirty & IRIS_DIRTY_STREAMOUT) {
         const struct iris_rasterizer_state *cso_rast = ice->state.cso_rast;

         uint32_t dynamic_sol[GENX(3DSTATE_STREAMOUT_length)];
         iris_pack_command(GENX(3DSTATE_STREAMOUT), dynamic_sol, sol) {
            sol.SOFunctionEnable = true;
            sol.SOStatisticsEnable = true;

            sol.RenderingDisable = cso_rast->rasterizer_discard &&
                                   !ice->state.prims_generated_query_active;
            sol.ReorderMode = cso_rast->flatshade_first ? LEADING : TRAILING;
         }

         assert(ice->state.streamout);

         iris_emit_merge(batch, ice->state.streamout, dynamic_sol,
                         GENX(3DSTATE_STREAMOUT_length));
      }
   } else {
      if (dirty & IRIS_DIRTY_STREAMOUT) {
         iris_emit_cmd(batch, GENX(3DSTATE_STREAMOUT), sol);
      }
   }

   if (dirty & IRIS_DIRTY_CLIP) {
      struct iris_rasterizer_state *cso_rast = ice->state.cso_rast;
      struct pipe_framebuffer_state *cso_fb = &ice->state.framebuffer;

      uint32_t dynamic_clip[GENX(3DSTATE_CLIP_length)];
      iris_pack_command(GENX(3DSTATE_CLIP), &dynamic_clip, cl) {
         if (wm_prog_data->barycentric_interp_modes &
             BRW_BARYCENTRIC_NONPERSPECTIVE_BITS)
            cl.NonPerspectiveBarycentricEnable = true;

         cl.ForceZeroRTAIndexEnable = cso_fb->layers == 0;
         cl.MaximumVPIndex = ice->state.num_viewports - 1;
      }
      iris_emit_merge(batch, cso_rast->clip, dynamic_clip,
                      ARRAY_SIZE(cso_rast->clip));
   }

   if (dirty & IRIS_DIRTY_RASTER) {
      struct iris_rasterizer_state *cso = ice->state.cso_rast;
      iris_batch_emit(batch, cso->raster, sizeof(cso->raster));
      iris_batch_emit(batch, cso->sf, sizeof(cso->sf));

   }

   /* XXX: FS program updates needs to flag IRIS_DIRTY_WM */
   if (dirty & IRIS_DIRTY_WM) {
      struct iris_rasterizer_state *cso = ice->state.cso_rast;
      uint32_t dynamic_wm[GENX(3DSTATE_WM_length)];

      iris_pack_command(GENX(3DSTATE_WM), &dynamic_wm, wm) {
         wm.StatisticsEnable = ice->state.statistics_counters_enabled;

         wm.BarycentricInterpolationMode =
            wm_prog_data->barycentric_interp_modes;

         if (wm_prog_data->early_fragment_tests)
            wm.EarlyDepthStencilControl = EDSC_PREPS;
         else if (wm_prog_data->has_side_effects)
            wm.EarlyDepthStencilControl = EDSC_PSEXEC;
      }
      iris_emit_merge(batch, cso->wm, dynamic_wm, ARRAY_SIZE(cso->wm));
   }

   if (dirty & IRIS_DIRTY_SBE) {
      iris_emit_sbe(batch, ice);
   }

   if (dirty & IRIS_DIRTY_PS_BLEND) {
      struct iris_blend_state *cso_blend = ice->state.cso_blend;
      struct iris_depth_stencil_alpha_state *cso_zsa = ice->state.cso_zsa;
      uint32_t dynamic_pb[GENX(3DSTATE_PS_BLEND_length)];
      iris_pack_command(GENX(3DSTATE_PS_BLEND), &dynamic_pb, pb) {
         pb.HasWriteableRT = true; // XXX: comes from somewhere :(
         pb.AlphaTestEnable = cso_zsa->alpha.enabled;
      }

      iris_emit_merge(batch, cso_blend->ps_blend, dynamic_pb,
                      ARRAY_SIZE(cso_blend->ps_blend));
   }

   if (dirty & IRIS_DIRTY_WM_DEPTH_STENCIL) {
      struct iris_depth_stencil_alpha_state *cso = ice->state.cso_zsa;
      struct pipe_stencil_ref *p_stencil_refs = &ice->state.stencil_ref;

      uint32_t stencil_refs[GENX(3DSTATE_WM_DEPTH_STENCIL_length)];
      iris_pack_command(GENX(3DSTATE_WM_DEPTH_STENCIL), &stencil_refs, wmds) {
         wmds.StencilReferenceValue = p_stencil_refs->ref_value[0];
         wmds.BackfaceStencilReferenceValue = p_stencil_refs->ref_value[1];
      }
      iris_emit_merge(batch, cso->wmds, stencil_refs, ARRAY_SIZE(cso->wmds));
   }

   if (dirty & IRIS_DIRTY_SCISSOR_RECT) {
      uint32_t scissor_offset =
         emit_state(batch, ice->state.dynamic_uploader,
                    &ice->state.last_res.scissor,
                    ice->state.scissors,
                    sizeof(struct pipe_scissor_state) *
                    ice->state.num_viewports, 32);

      iris_emit_cmd(batch, GENX(3DSTATE_SCISSOR_STATE_POINTERS), ptr) {
         ptr.ScissorRectPointer = scissor_offset;
      }
   }

   if (dirty & IRIS_DIRTY_DEPTH_BUFFER) {
      struct pipe_framebuffer_state *cso_fb = &ice->state.framebuffer;
      struct iris_depth_buffer_state *cso_z = &ice->state.genx->depth_buffer;

      iris_batch_emit(batch, cso_z->packets, sizeof(cso_z->packets));

      if (cso_fb->zsbuf) {
         struct iris_resource *zres = (void *) cso_fb->zsbuf->texture;
         // XXX: depth might not be writable...
         iris_use_pinned_bo(batch, zres->bo, true);
      }
   }

   if (dirty & IRIS_DIRTY_POLYGON_STIPPLE) {
      iris_emit_cmd(batch, GENX(3DSTATE_POLY_STIPPLE_PATTERN), poly) {
         for (int i = 0; i < 32; i++) {
            poly.PatternRow[i] = ice->state.poly_stipple.stipple[i];
         }
      }
   }

   if (dirty & IRIS_DIRTY_LINE_STIPPLE) {
      struct iris_rasterizer_state *cso = ice->state.cso_rast;
      iris_batch_emit(batch, cso->line_stipple, sizeof(cso->line_stipple));
   }

   if (dirty & IRIS_DIRTY_VF_TOPOLOGY) {
      iris_emit_cmd(batch, GENX(3DSTATE_VF_TOPOLOGY), topo) {
         topo.PrimitiveTopologyType =
            translate_prim_type(draw->mode, draw->vertices_per_patch);
      }
   }

   if (dirty & IRIS_DIRTY_VERTEX_BUFFERS) {
      struct iris_vertex_buffer_state *cso = &ice->state.genx->vertex_buffers;
      const unsigned vb_dwords = GENX(VERTEX_BUFFER_STATE_length);

      if (cso->num_buffers > 0) {
         iris_batch_emit(batch, cso->vertex_buffers, sizeof(uint32_t) *
                         (1 + vb_dwords * cso->num_buffers));

         for (unsigned i = 0; i < cso->num_buffers; i++) {
            struct iris_resource *res = (void *) cso->resources[i];
            if (res)
               iris_use_pinned_bo(batch, res->bo, false);
         }
      }
   }

   if (dirty & IRIS_DIRTY_VERTEX_ELEMENTS) {
      struct iris_vertex_element_state *cso = ice->state.cso_vertex_elements;
      const unsigned entries = MAX2(cso->count, 1);
      iris_batch_emit(batch, cso->vertex_elements, sizeof(uint32_t) *
                      (1 + entries * GENX(VERTEX_ELEMENT_STATE_length)));
      iris_batch_emit(batch, cso->vf_instancing, sizeof(uint32_t) *
                      entries * GENX(3DSTATE_VF_INSTANCING_length));
   }

   if (dirty & IRIS_DIRTY_VF_SGVS) {
      const struct brw_vs_prog_data *vs_prog_data = (void *)
         ice->shaders.prog[MESA_SHADER_VERTEX]->prog_data;
      struct iris_vertex_element_state *cso = ice->state.cso_vertex_elements;

      iris_emit_cmd(batch, GENX(3DSTATE_VF_SGVS), sgv) {
         if (vs_prog_data->uses_vertexid) {
            sgv.VertexIDEnable = true;
            sgv.VertexIDComponentNumber = 2;
            sgv.VertexIDElementOffset = cso->count;
         }

         if (vs_prog_data->uses_instanceid) {
            sgv.InstanceIDEnable = true;
            sgv.InstanceIDComponentNumber = 3;
            sgv.InstanceIDElementOffset = cso->count;
         }
      }
   }

   if (dirty & IRIS_DIRTY_VF) {
      iris_emit_cmd(batch, GENX(3DSTATE_VF), vf) {
         if (draw->primitive_restart) {
            vf.IndexedDrawCutIndexEnable = true;
            vf.CutIndex = draw->restart_index;
         }
      }
   }

   // XXX: Gen8 - PMA fix
}

static void
iris_upload_render_state(struct iris_context *ice,
                         struct iris_batch *batch,
                         const struct pipe_draw_info *draw)
{
   /* Always pin the binder.  If we're emitting new binding table pointers,
    * we need it.  If not, we're probably inheriting old tables via the
    * context, and need it anyway.  Since true zero-bindings cases are
    * practically non-existent, just pin it and avoid last_res tracking.
    */
   iris_use_pinned_bo(batch, ice->state.binder.bo, false);

   iris_upload_dirty_render_state(ice, batch, draw);

   if (draw->index_size > 0) {
      unsigned offset;

      if (draw->has_user_indices) {
         u_upload_data(ice->ctx.stream_uploader, 0,
                       draw->count * draw->index_size, 4, draw->index.user,
                       &offset, &ice->state.last_res.index_buffer);
      } else {
         pipe_resource_reference(&ice->state.last_res.index_buffer,
                                 draw->index.resource);
         offset = 0;
      }

      struct iris_bo *bo = iris_resource_bo(ice->state.last_res.index_buffer);

      iris_emit_cmd(batch, GENX(3DSTATE_INDEX_BUFFER), ib) {
         ib.IndexFormat = draw->index_size >> 1;
         ib.MOCS = MOCS_WB;
         ib.BufferSize = bo->size;
         ib.BufferStartingAddress = ro_bo(bo, offset);
      }
   }

#define _3DPRIM_END_OFFSET          0x2420
#define _3DPRIM_START_VERTEX        0x2430
#define _3DPRIM_VERTEX_COUNT        0x2434
#define _3DPRIM_INSTANCE_COUNT      0x2438
#define _3DPRIM_START_INSTANCE      0x243C
#define _3DPRIM_BASE_VERTEX         0x2440

   if (draw->indirect) {
      /* We don't support this MultidrawIndirect. */
      assert(!draw->indirect->indirect_draw_count);

      struct iris_bo *bo = iris_resource_bo(draw->indirect->buffer);
      assert(bo);

      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = _3DPRIM_VERTEX_COUNT;
         lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 0);
      }
      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = _3DPRIM_INSTANCE_COUNT;
         lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 4);
      }
      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = _3DPRIM_START_VERTEX;
         lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 8);
      }
      if (draw->index_size) {
         iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
            lrm.RegisterAddress = _3DPRIM_BASE_VERTEX;
            lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 12);
         }
         iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
            lrm.RegisterAddress = _3DPRIM_START_INSTANCE;
            lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 16);
         }
      } else {
         iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
            lrm.RegisterAddress = _3DPRIM_START_INSTANCE;
            lrm.MemoryAddress = ro_bo(bo, draw->indirect->offset + 12);
         }
         iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
            lri.RegisterOffset = _3DPRIM_BASE_VERTEX;
            lri.DataDWord = 0;
         }
      }
   }

   iris_emit_cmd(batch, GENX(3DPRIMITIVE), prim) {
      prim.StartInstanceLocation = draw->start_instance;
      prim.InstanceCount = draw->instance_count;
      prim.VertexCountPerInstance = draw->count;
      prim.VertexAccessType = draw->index_size > 0 ? RANDOM : SEQUENTIAL;

      // XXX: this is probably bonkers.
      prim.StartVertexLocation = draw->start;

      prim.IndirectParameterEnable = draw->indirect != NULL;

      if (draw->index_size) {
         prim.BaseVertexLocation += draw->index_bias;
      } else {
         prim.StartVertexLocation += draw->index_bias;
      }

      //prim.BaseVertexLocation = ...;
   }

   if (!batch->contains_draw) {
      iris_restore_render_saved_bos(ice, batch, draw);
      batch->contains_draw = true;
   }
}

static void
iris_upload_compute_state(struct iris_context *ice,
                          struct iris_batch *batch,
                          const struct pipe_grid_info *grid)
{
   const uint64_t dirty = ice->state.dirty;
   struct iris_screen *screen = batch->screen;
   const struct gen_device_info *devinfo = &screen->devinfo;
   struct iris_binder *binder = &ice->state.binder;
   struct iris_shader_state *shs = &ice->state.shaders[MESA_SHADER_COMPUTE];
   struct iris_compiled_shader *shader =
      ice->shaders.prog[MESA_SHADER_COMPUTE];
   struct brw_stage_prog_data *prog_data = shader->prog_data;
   struct brw_cs_prog_data *cs_prog_data = (void *) prog_data;

   // XXX: L3 configuration not set up for SLM
   assert(prog_data->total_shared == 0);

   if (dirty & IRIS_DIRTY_BINDINGS_CS)
      iris_populate_binding_table(ice, batch, MESA_SHADER_COMPUTE, false);

   iris_use_optional_res(batch, shs->sampler_table.res, false);
   iris_use_pinned_bo(batch, iris_resource_bo(shader->assembly.res), false);

   if (ice->state.need_border_colors)
      iris_use_pinned_bo(batch, ice->state.border_color_pool.bo, false);

   if (dirty & IRIS_DIRTY_CS) {
      /* The MEDIA_VFE_STATE documentation for Gen8+ says:
       *
       *   "A stalling PIPE_CONTROL is required before MEDIA_VFE_STATE unless
       *    the only bits that are changed are scoreboard related: Scoreboard
       *    Enable, Scoreboard Type, Scoreboard Mask, Scoreboard Delta.  For
       *    these scoreboard related states, a MEDIA_STATE_FLUSH is
       *    sufficient."
       */
      iris_emit_pipe_control_flush(batch, PIPE_CONTROL_CS_STALL);

      iris_emit_cmd(batch, GENX(MEDIA_VFE_STATE), vfe) {
         if (prog_data->total_scratch) {
            uint32_t scratch_addr =
               iris_get_scratch_space(ice, prog_data->total_scratch,
                                      MESA_SHADER_COMPUTE);
            vfe.PerThreadScratchSpace = ffs(prog_data->total_scratch) - 11;
            vfe.ScratchSpaceBasePointer = rw_bo(NULL, scratch_addr);
         }

         vfe.MaximumNumberofThreads =
            devinfo->max_cs_threads * screen->subslice_total - 1;
#if GEN_GEN < 11
         vfe.ResetGatewayTimer =
            Resettingrelativetimerandlatchingtheglobaltimestamp;
#endif

         vfe.NumberofURBEntries = 2;
         vfe.URBEntryAllocationSize = 2;

         // XXX: Use Indirect Payload Storage?
         vfe.CURBEAllocationSize =
            ALIGN(cs_prog_data->push.per_thread.regs * cs_prog_data->threads +
                  cs_prog_data->push.cross_thread.regs, 2);
      }
   }

   // XXX: hack iris_set_constant_buffers to upload these thread counts
   // XXX: along with regular uniforms for compute shaders, somehow.

   uint32_t curbe_data_offset = 0;
   // TODO: Move subgroup-id into uniforms ubo so we can push uniforms
   assert(cs_prog_data->push.cross_thread.dwords == 0 &&
          cs_prog_data->push.per_thread.dwords == 1 &&
          cs_prog_data->base.param[0] == BRW_PARAM_BUILTIN_SUBGROUP_ID);
   struct pipe_resource *curbe_data_res = NULL;
   uint32_t *curbe_data_map =
      stream_state(batch, ice->state.dynamic_uploader, &curbe_data_res,
                   ALIGN(cs_prog_data->push.total.size, 64), 64,
                   &curbe_data_offset);
   assert(curbe_data_map);
   memset(curbe_data_map, 0x5a, ALIGN(cs_prog_data->push.total.size, 64));
   iris_fill_cs_push_const_buffer(cs_prog_data, curbe_data_map);

   if (dirty & IRIS_DIRTY_CONSTANTS_CS) {
      iris_emit_cmd(batch, GENX(MEDIA_CURBE_LOAD), curbe) {
         curbe.CURBETotalDataLength =
            ALIGN(cs_prog_data->push.total.size, 64);
         curbe.CURBEDataStartAddress = curbe_data_offset;
      }
   }

   if (dirty & (IRIS_DIRTY_SAMPLER_STATES_CS |
                IRIS_DIRTY_BINDINGS_CS |
                IRIS_DIRTY_CONSTANTS_CS |
                IRIS_DIRTY_CS)) {
      struct pipe_resource *desc_res = NULL;
      uint32_t desc[GENX(INTERFACE_DESCRIPTOR_DATA_length)];

      iris_pack_state(GENX(INTERFACE_DESCRIPTOR_DATA), desc, idd) {
         idd.SamplerStatePointer = shs->sampler_table.offset;
         idd.BindingTablePointer = binder->bt_offset[MESA_SHADER_COMPUTE];
         idd.ConstantURBEntryReadLength = cs_prog_data->push.per_thread.regs;
         idd.CrossThreadConstantDataReadLength =
            cs_prog_data->push.cross_thread.regs;
      }

      for (int i = 0; i < GENX(INTERFACE_DESCRIPTOR_DATA_length); i++)
         desc[i] |= ((uint32_t *) shader->derived_data)[i];

      iris_emit_cmd(batch, GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD), load) {
         load.InterfaceDescriptorTotalLength =
            GENX(INTERFACE_DESCRIPTOR_DATA_length) * sizeof(uint32_t);
         load.InterfaceDescriptorDataStartAddress =
            emit_state(batch, ice->state.dynamic_uploader,
                       &desc_res, desc, sizeof(desc), 32);
      }

      pipe_resource_reference(&desc_res, NULL);
   }

   uint32_t group_size = grid->block[0] * grid->block[1] * grid->block[2];
   uint32_t remainder = group_size & (cs_prog_data->simd_size - 1);
   uint32_t right_mask;

   if (remainder > 0)
      right_mask = ~0u >> (32 - remainder);
   else
      right_mask = ~0u >> (32 - cs_prog_data->simd_size);

#define GPGPU_DISPATCHDIMX 0x2500
#define GPGPU_DISPATCHDIMY 0x2504
#define GPGPU_DISPATCHDIMZ 0x2508

   if (grid->indirect) {
      struct iris_state_ref *grid_size = &ice->state.grid_size;
      struct iris_bo *bo = iris_resource_bo(grid_size->res);
      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = GPGPU_DISPATCHDIMX;
         lrm.MemoryAddress = ro_bo(bo, grid_size->offset + 0);
      }
      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = GPGPU_DISPATCHDIMY;
         lrm.MemoryAddress = ro_bo(bo, grid_size->offset + 4);
      }
      iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
         lrm.RegisterAddress = GPGPU_DISPATCHDIMZ;
         lrm.MemoryAddress = ro_bo(bo, grid_size->offset + 8);
      }
   }

   iris_emit_cmd(batch, GENX(GPGPU_WALKER), ggw) {
      ggw.IndirectParameterEnable    = grid->indirect != NULL;
      ggw.SIMDSize                   = cs_prog_data->simd_size / 16;
      ggw.ThreadDepthCounterMaximum  = 0;
      ggw.ThreadHeightCounterMaximum = 0;
      ggw.ThreadWidthCounterMaximum  = cs_prog_data->threads - 1;
      ggw.ThreadGroupIDXDimension    = grid->grid[0];
      ggw.ThreadGroupIDYDimension    = grid->grid[1];
      ggw.ThreadGroupIDZDimension    = grid->grid[2];
      ggw.RightExecutionMask         = right_mask;
      ggw.BottomExecutionMask        = 0xffffffff;
   }

   iris_emit_cmd(batch, GENX(MEDIA_STATE_FLUSH), msf);

   if (!batch->contains_draw) {
      iris_restore_compute_saved_bos(ice, batch, grid);
      batch->contains_draw = true;
   }
}

/**
 * State module teardown.
 */
static void
iris_destroy_state(struct iris_context *ice)
{
   iris_free_vertex_buffers(&ice->state.genx->vertex_buffers);

   // XXX: unreference resources/surfaces.
   for (unsigned i = 0; i < ice->state.framebuffer.nr_cbufs; i++) {
      pipe_surface_reference(&ice->state.framebuffer.cbufs[i], NULL);
   }
   pipe_surface_reference(&ice->state.framebuffer.zsbuf, NULL);

   for (int stage = 0; stage < MESA_SHADER_STAGES; stage++) {
      struct iris_shader_state *shs = &ice->state.shaders[stage];
      pipe_resource_reference(&shs->sampler_table.res, NULL);
   }
   free(ice->state.genx);

   pipe_resource_reference(&ice->state.last_res.cc_vp, NULL);
   pipe_resource_reference(&ice->state.last_res.sf_cl_vp, NULL);
   pipe_resource_reference(&ice->state.last_res.color_calc, NULL);
   pipe_resource_reference(&ice->state.last_res.scissor, NULL);
   pipe_resource_reference(&ice->state.last_res.blend, NULL);
   pipe_resource_reference(&ice->state.last_res.index_buffer, NULL);
}

/* ------------------------------------------------------------------- */

static void
iris_load_register_imm32(struct iris_batch *batch, uint32_t reg,
                         uint32_t val)
{
   _iris_emit_lri(batch, reg, val);
}

static void
iris_load_register_imm64(struct iris_batch *batch, uint32_t reg,
                         uint64_t val)
{
   _iris_emit_lri(batch, reg + 0, val & 0xffffffff);
   _iris_emit_lri(batch, reg + 4, val >> 32);
}

/**
 * Emit MI_LOAD_REGISTER_MEM to load a 32-bit MMIO register from a buffer.
 */
static void
iris_load_register_mem32(struct iris_batch *batch, uint32_t reg,
                         struct iris_bo *bo, uint32_t offset)
{
   iris_emit_cmd(batch, GENX(MI_LOAD_REGISTER_MEM), lrm) {
      lrm.RegisterAddress = reg;
      lrm.MemoryAddress = ro_bo(bo, offset);
   }
}

/**
 * Load a 64-bit value from a buffer into a MMIO register via
 * two MI_LOAD_REGISTER_MEM commands.
 */
static void
iris_load_register_mem64(struct iris_batch *batch, uint32_t reg,
                         struct iris_bo *bo, uint32_t offset)
{
   iris_load_register_mem32(batch, reg + 0, bo, offset + 0);
   iris_load_register_mem32(batch, reg + 4, bo, offset + 4);
}

static void
iris_store_register_mem32(struct iris_batch *batch, uint32_t reg,
                          struct iris_bo *bo, uint32_t offset,
                          bool predicated)
{
   iris_emit_cmd(batch, GENX(MI_STORE_REGISTER_MEM), srm) {
      srm.RegisterAddress = reg;
      srm.MemoryAddress = rw_bo(bo, offset);
      srm.PredicateEnable = predicated;
   }
}

static void
iris_store_register_mem64(struct iris_batch *batch, uint32_t reg,
                          struct iris_bo *bo, uint32_t offset,
                          bool predicated)
{
   iris_store_register_mem32(batch, reg + 0, bo, offset + 0, predicated);
   iris_store_register_mem32(batch, reg + 4, bo, offset + 4, predicated);
}

static void
iris_store_data_imm32(struct iris_batch *batch,
                      struct iris_bo *bo, uint32_t offset,
                      uint32_t imm)
{
   iris_emit_cmd(batch, GENX(MI_STORE_DATA_IMM), sdi) {
      sdi.Address = rw_bo(bo, offset);
      sdi.ImmediateData = imm;
   }
}

static void
iris_store_data_imm64(struct iris_batch *batch,
                      struct iris_bo *bo, uint32_t offset,
                      uint64_t imm)
{
   /* Can't use iris_emit_cmd because MI_STORE_DATA_IMM has a length of
    * 2 in genxml but it's actually variable length and we need 5 DWords.
    */
   void *map = iris_get_command_space(batch, 4 * 5);
   _iris_pack_command(batch, GENX(MI_STORE_DATA_IMM), map, sdi) {
      sdi.DWordLength = 5 - 2;
      sdi.Address = rw_bo(bo, offset);
      sdi.ImmediateData = imm;
   }
}

static void
iris_copy_mem_mem(struct iris_batch *batch,
                  struct iris_bo *dst_bo, uint32_t dst_offset,
                  struct iris_bo *src_bo, uint32_t src_offset,
                  unsigned bytes)
{
   /* MI_COPY_MEM_MEM operates on DWords. */
   assert(bytes % 4 == 0);
   assert(dst_offset % 4 == 0);
   assert(src_offset % 4 == 0);

   for (unsigned i = 0; i < bytes; i += 4) {
      iris_emit_cmd(batch, GENX(MI_COPY_MEM_MEM), cp) {
         cp.DestinationMemoryAddress = rw_bo(dst_bo, dst_offset + i);
         cp.SourceMemoryAddress = ro_bo(src_bo, src_offset + i);
      }
   }
}

/* ------------------------------------------------------------------- */

static unsigned
flags_to_post_sync_op(uint32_t flags)
{
   if (flags & PIPE_CONTROL_WRITE_IMMEDIATE)
      return WriteImmediateData;

   if (flags & PIPE_CONTROL_WRITE_DEPTH_COUNT)
      return WritePSDepthCount;

   if (flags & PIPE_CONTROL_WRITE_TIMESTAMP)
      return WriteTimestamp;

   return 0;
}

/**
 * Do the given flags have a Post Sync or LRI Post Sync operation?
 */
static enum pipe_control_flags
get_post_sync_flags(enum pipe_control_flags flags)
{
   flags &= PIPE_CONTROL_WRITE_IMMEDIATE |
            PIPE_CONTROL_WRITE_DEPTH_COUNT |
            PIPE_CONTROL_WRITE_TIMESTAMP |
            PIPE_CONTROL_LRI_POST_SYNC_OP;

   /* Only one "Post Sync Op" is allowed, and it's mutually exclusive with
    * "LRI Post Sync Operation".  So more than one bit set would be illegal.
    */
   assert(util_bitcount(flags) <= 1);

   return flags;
}

// XXX: compute support
#define IS_COMPUTE_PIPELINE(batch) (batch->engine != I915_EXEC_RENDER)

/**
 * Emit a series of PIPE_CONTROL commands, taking into account any
 * workarounds necessary to actually accomplish the caller's request.
 *
 * Unless otherwise noted, spec quotations in this function come from:
 *
 * Synchronization of the 3D Pipeline > PIPE_CONTROL Command > Programming
 * Restrictions for PIPE_CONTROL.
 *
 * You should not use this function directly.  Use the helpers in
 * iris_pipe_control.c instead, which may split the pipe control further.
 */
static void
iris_emit_raw_pipe_control(struct iris_batch *batch, uint32_t flags,
                           struct iris_bo *bo, uint32_t offset, uint64_t imm)
{
   UNUSED const struct gen_device_info *devinfo = &batch->screen->devinfo;
   enum pipe_control_flags post_sync_flags = get_post_sync_flags(flags);
   enum pipe_control_flags non_lri_post_sync_flags =
      post_sync_flags & ~PIPE_CONTROL_LRI_POST_SYNC_OP;

   /* Recursive PIPE_CONTROL workarounds --------------------------------
    * (http://knowyourmeme.com/memes/xzibit-yo-dawg)
    *
    * We do these first because we want to look at the original operation,
    * rather than any workarounds we set.
    */
   if (GEN_GEN == 9 && (flags & PIPE_CONTROL_VF_CACHE_INVALIDATE)) {
      /* The PIPE_CONTROL "VF Cache Invalidation Enable" bit description
       * lists several workarounds:
       *
       *    "Project: SKL, KBL, BXT
       *
       *     If the VF Cache Invalidation Enable is set to a 1 in a
       *     PIPE_CONTROL, a separate Null PIPE_CONTROL, all bitfields
       *     sets to 0, with the VF Cache Invalidation Enable set to 0
       *     needs to be sent prior to the PIPE_CONTROL with VF Cache
       *     Invalidation Enable set to a 1."
       */
      iris_emit_raw_pipe_control(batch, 0, NULL, 0, 0);
   }

   if (GEN_GEN == 9 && IS_COMPUTE_PIPELINE(batch) && post_sync_flags) {
      /* Project: SKL / Argument: LRI Post Sync Operation [23]
       *
       * "PIPECONTROL command with “Command Streamer Stall Enable” must be
       *  programmed prior to programming a PIPECONTROL command with "LRI
       *  Post Sync Operation" in GPGPU mode of operation (i.e when
       *  PIPELINE_SELECT command is set to GPGPU mode of operation)."
       *
       * The same text exists a few rows below for Post Sync Op.
       */
      iris_emit_raw_pipe_control(batch, PIPE_CONTROL_CS_STALL, bo, offset, imm);
   }

   if (GEN_GEN == 10 && (flags & PIPE_CONTROL_RENDER_TARGET_FLUSH)) {
      /* Cannonlake:
       * "Before sending a PIPE_CONTROL command with bit 12 set, SW must issue
       *  another PIPE_CONTROL with Render Target Cache Flush Enable (bit 12)
       *  = 0 and Pipe Control Flush Enable (bit 7) = 1"
       */
      iris_emit_raw_pipe_control(batch, PIPE_CONTROL_FLUSH_ENABLE, bo,
                                 offset, imm);
   }

   /* "Flush Types" workarounds ---------------------------------------------
    * We do these now because they may add post-sync operations or CS stalls.
    */

   if (GEN_GEN < 11 && flags & PIPE_CONTROL_VF_CACHE_INVALIDATE) {
      /* Project: BDW, SKL+ (stopping at CNL) / Argument: VF Invalidate
       *
       * "'Post Sync Operation' must be enabled to 'Write Immediate Data' or
       *  'Write PS Depth Count' or 'Write Timestamp'."
       */
      if (!bo) {
         flags |= PIPE_CONTROL_WRITE_IMMEDIATE;
         post_sync_flags |= PIPE_CONTROL_WRITE_IMMEDIATE;
         non_lri_post_sync_flags |= PIPE_CONTROL_WRITE_IMMEDIATE;
         bo = batch->screen->workaround_bo;
      }
   }

   /* #1130 from Gen10 workarounds page:
    *
    *    "Enable Depth Stall on every Post Sync Op if Render target Cache
    *     Flush is not enabled in same PIPE CONTROL and Enable Pixel score
    *     board stall if Render target cache flush is enabled."
    *
    * Applicable to CNL B0 and C0 steppings only.
    *
    * The wording here is unclear, and this workaround doesn't look anything
    * like the internal bug report recommendations, but leave it be for now...
    */
   if (GEN_GEN == 10) {
      if (flags & PIPE_CONTROL_RENDER_TARGET_FLUSH) {
         flags |= PIPE_CONTROL_STALL_AT_SCOREBOARD;
      } else if (flags & non_lri_post_sync_flags) {
         flags |= PIPE_CONTROL_DEPTH_STALL;
      }
   }

   if (flags & PIPE_CONTROL_DEPTH_STALL) {
      /* From the PIPE_CONTROL instruction table, bit 13 (Depth Stall Enable):
       *
       *    "This bit must be DISABLED for operations other than writing
       *     PS_DEPTH_COUNT."
       *
       * This seems like nonsense.  An Ivybridge workaround requires us to
       * emit a PIPE_CONTROL with a depth stall and write immediate post-sync
       * operation.  Gen8+ requires us to emit depth stalls and depth cache
       * flushes together.  So, it's hard to imagine this means anything other
       * than "we originally intended this to be used for PS_DEPTH_COUNT".
       *
       * We ignore the supposed restriction and do nothing.
       */
   }

   if (flags & (PIPE_CONTROL_RENDER_TARGET_FLUSH |
                PIPE_CONTROL_STALL_AT_SCOREBOARD)) {
      /* From the PIPE_CONTROL instruction table, bit 12 and bit 1:
       *
       *    "This bit must be DISABLED for End-of-pipe (Read) fences,
       *     PS_DEPTH_COUNT or TIMESTAMP queries."
       *
       * TODO: Implement end-of-pipe checking.
       */
      assert(!(post_sync_flags & (PIPE_CONTROL_WRITE_DEPTH_COUNT |
                                  PIPE_CONTROL_WRITE_TIMESTAMP)));
   }

   if (GEN_GEN < 11 && (flags & PIPE_CONTROL_STALL_AT_SCOREBOARD)) {
      /* From the PIPE_CONTROL instruction table, bit 1:
       *
       *    "This bit is ignored if Depth Stall Enable is set.
       *     Further, the render cache is not flushed even if Write Cache
       *     Flush Enable bit is set."
       *
       * We assert that the caller doesn't do this combination, to try and
       * prevent mistakes.  It shouldn't hurt the GPU, though.
       *
       * We skip this check on Gen11+ as the "Stall at Pixel Scoreboard"
       * and "Render Target Flush" combo is explicitly required for BTI
       * update workarounds.
       */
      assert(!(flags & (PIPE_CONTROL_DEPTH_STALL |
                        PIPE_CONTROL_RENDER_TARGET_FLUSH)));
   }

   /* PIPE_CONTROL page workarounds ------------------------------------- */

   if (GEN_GEN <= 8 && (flags & PIPE_CONTROL_STATE_CACHE_INVALIDATE)) {
      /* From the PIPE_CONTROL page itself:
       *
       *    "IVB, HSW, BDW
       *     Restriction: Pipe_control with CS-stall bit set must be issued
       *     before a pipe-control command that has the State Cache
       *     Invalidate bit set."
       */
      flags |= PIPE_CONTROL_CS_STALL;
   }

   if (flags & PIPE_CONTROL_FLUSH_LLC) {
      /* From the PIPE_CONTROL instruction table, bit 26 (Flush LLC):
       *
       *    "Project: ALL
       *     SW must always program Post-Sync Operation to "Write Immediate
       *     Data" when Flush LLC is set."
       *
       * For now, we just require the caller to do it.
       */
      assert(flags & PIPE_CONTROL_WRITE_IMMEDIATE);
   }

   /* "Post-Sync Operation" workarounds -------------------------------- */

   /* Project: All / Argument: Global Snapshot Count Reset [19]
    *
    * "This bit must not be exercised on any product.
    *  Requires stall bit ([20] of DW1) set."
    *
    * We don't use this, so we just assert that it isn't used.  The
    * PIPE_CONTROL instruction page indicates that they intended this
    * as a debug feature and don't think it is useful in production,
    * but it may actually be usable, should we ever want to.
    */
   assert((flags & PIPE_CONTROL_GLOBAL_SNAPSHOT_COUNT_RESET) == 0);

   if (flags & (PIPE_CONTROL_MEDIA_STATE_CLEAR |
                PIPE_CONTROL_INDIRECT_STATE_POINTERS_DISABLE)) {
      /* Project: All / Arguments:
       *
       * - Generic Media State Clear [16]
       * - Indirect State Pointers Disable [16]
       *
       *    "Requires stall bit ([20] of DW1) set."
       *
       * Also, the PIPE_CONTROL instruction table, bit 16 (Generic Media
       * State Clear) says:
       *
       *    "PIPECONTROL command with “Command Streamer Stall Enable” must be
       *     programmed prior to programming a PIPECONTROL command with "Media
       *     State Clear" set in GPGPU mode of operation"
       *
       * This is a subset of the earlier rule, so there's nothing to do.
       */
      flags |= PIPE_CONTROL_CS_STALL;
   }

   if (flags & PIPE_CONTROL_STORE_DATA_INDEX) {
      /* Project: All / Argument: Store Data Index
       *
       * "Post-Sync Operation ([15:14] of DW1) must be set to something other
       *  than '0'."
       *
       * For now, we just assert that the caller does this.  We might want to
       * automatically add a write to the workaround BO...
       */
      assert(non_lri_post_sync_flags != 0);
   }

   if (flags & PIPE_CONTROL_SYNC_GFDT) {
      /* Project: All / Argument: Sync GFDT
       *
       * "Post-Sync Operation ([15:14] of DW1) must be set to something other
       *  than '0' or 0x2520[13] must be set."
       *
       * For now, we just assert that the caller does this.
       */
      assert(non_lri_post_sync_flags != 0);
   }

   if (flags & PIPE_CONTROL_TLB_INVALIDATE) {
      /* Project: IVB+ / Argument: TLB inv
       *
       *    "Requires stall bit ([20] of DW1) set."
       *
       * Also, from the PIPE_CONTROL instruction table:
       *
       *    "Project: SKL+
       *     Post Sync Operation or CS stall must be set to ensure a TLB
       *     invalidation occurs.  Otherwise no cycle will occur to the TLB
       *     cache to invalidate."
       *
       * This is not a subset of the earlier rule, so there's nothing to do.
       */
      flags |= PIPE_CONTROL_CS_STALL;
   }

   if (GEN_GEN == 9 && devinfo->gt == 4) {
      /* TODO: The big Skylake GT4 post sync op workaround */
   }

   /* "GPGPU specific workarounds" (both post-sync and flush) ------------ */

   if (IS_COMPUTE_PIPELINE(batch)) {
      if (GEN_GEN >= 9 && (flags & PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE)) {
         /* Project: SKL+ / Argument: Tex Invalidate
          * "Requires stall bit ([20] of DW) set for all GPGPU Workloads."
          */
         flags |= PIPE_CONTROL_CS_STALL;
      }

      if (GEN_GEN == 8 && (post_sync_flags ||
                           (flags & (PIPE_CONTROL_NOTIFY_ENABLE |
                                     PIPE_CONTROL_DEPTH_STALL |
                                     PIPE_CONTROL_RENDER_TARGET_FLUSH |
                                     PIPE_CONTROL_DEPTH_CACHE_FLUSH |
                                     PIPE_CONTROL_DATA_CACHE_FLUSH)))) {
         /* Project: BDW / Arguments:
          *
          * - LRI Post Sync Operation   [23]
          * - Post Sync Op              [15:14]
          * - Notify En                 [8]
          * - Depth Stall               [13]
          * - Render Target Cache Flush [12]
          * - Depth Cache Flush         [0]
          * - DC Flush Enable           [5]
          *
          *    "Requires stall bit ([20] of DW) set for all GPGPU and Media
          *     Workloads."
          */
         flags |= PIPE_CONTROL_CS_STALL;

         /* Also, from the PIPE_CONTROL instruction table, bit 20:
          *
          *    "Project: BDW
          *     This bit must be always set when PIPE_CONTROL command is
          *     programmed by GPGPU and MEDIA workloads, except for the cases
          *     when only Read Only Cache Invalidation bits are set (State
          *     Cache Invalidation Enable, Instruction cache Invalidation
          *     Enable, Texture Cache Invalidation Enable, Constant Cache
          *     Invalidation Enable). This is to WA FFDOP CG issue, this WA
          *     need not implemented when FF_DOP_CG is disable via "Fixed
          *     Function DOP Clock Gate Disable" bit in RC_PSMI_CTRL register."
          *
          * It sounds like we could avoid CS stalls in some cases, but we
          * don't currently bother.  This list isn't exactly the list above,
          * either...
          */
      }
   }

   /* "Stall" workarounds ----------------------------------------------
    * These have to come after the earlier ones because we may have added
    * some additional CS stalls above.
    */

   if (GEN_GEN < 9 && (flags & PIPE_CONTROL_CS_STALL)) {
      /* Project: PRE-SKL, VLV, CHV
       *
       * "[All Stepping][All SKUs]:
       *
       *  One of the following must also be set:
       *
       *  - Render Target Cache Flush Enable ([12] of DW1)
       *  - Depth Cache Flush Enable ([0] of DW1)
       *  - Stall at Pixel Scoreboard ([1] of DW1)
       *  - Depth Stall ([13] of DW1)
       *  - Post-Sync Operation ([13] of DW1)
       *  - DC Flush Enable ([5] of DW1)"
       *
       * If we don't already have one of those bits set, we choose to add
       * "Stall at Pixel Scoreboard".  Some of the other bits require a
       * CS stall as a workaround (see above), which would send us into
       * an infinite recursion of PIPE_CONTROLs.  "Stall at Pixel Scoreboard"
       * appears to be safe, so we choose that.
       */
      const uint32_t wa_bits = PIPE_CONTROL_RENDER_TARGET_FLUSH |
                               PIPE_CONTROL_DEPTH_CACHE_FLUSH |
                               PIPE_CONTROL_WRITE_IMMEDIATE |
                               PIPE_CONTROL_WRITE_DEPTH_COUNT |
                               PIPE_CONTROL_WRITE_TIMESTAMP |
                               PIPE_CONTROL_STALL_AT_SCOREBOARD |
                               PIPE_CONTROL_DEPTH_STALL |
                               PIPE_CONTROL_DATA_CACHE_FLUSH;
      if (!(flags & wa_bits))
         flags |= PIPE_CONTROL_STALL_AT_SCOREBOARD;
   }

   /* Emit --------------------------------------------------------------- */

   iris_emit_cmd(batch, GENX(PIPE_CONTROL), pc) {
      pc.LRIPostSyncOperation = NoLRIOperation;
      pc.PipeControlFlushEnable = flags & PIPE_CONTROL_FLUSH_ENABLE;
      pc.DCFlushEnable = flags & PIPE_CONTROL_DATA_CACHE_FLUSH;
      pc.StoreDataIndex = 0;
      pc.CommandStreamerStallEnable = flags & PIPE_CONTROL_CS_STALL;
      pc.GlobalSnapshotCountReset =
         flags & PIPE_CONTROL_GLOBAL_SNAPSHOT_COUNT_RESET;
      pc.TLBInvalidate = flags & PIPE_CONTROL_TLB_INVALIDATE;
      pc.GenericMediaStateClear = flags & PIPE_CONTROL_MEDIA_STATE_CLEAR;
      pc.StallAtPixelScoreboard = flags & PIPE_CONTROL_STALL_AT_SCOREBOARD;
      pc.RenderTargetCacheFlushEnable =
         flags & PIPE_CONTROL_RENDER_TARGET_FLUSH;
      pc.DepthCacheFlushEnable = flags & PIPE_CONTROL_DEPTH_CACHE_FLUSH;
      pc.StateCacheInvalidationEnable =
         flags & PIPE_CONTROL_STATE_CACHE_INVALIDATE;
      pc.VFCacheInvalidationEnable = flags & PIPE_CONTROL_VF_CACHE_INVALIDATE;
      pc.ConstantCacheInvalidationEnable =
         flags & PIPE_CONTROL_CONST_CACHE_INVALIDATE;
      pc.PostSyncOperation = flags_to_post_sync_op(flags);
      pc.DepthStallEnable = flags & PIPE_CONTROL_DEPTH_STALL;
      pc.InstructionCacheInvalidateEnable =
         flags & PIPE_CONTROL_INSTRUCTION_INVALIDATE;
      pc.NotifyEnable = flags & PIPE_CONTROL_NOTIFY_ENABLE;
      pc.IndirectStatePointersDisable =
         flags & PIPE_CONTROL_INDIRECT_STATE_POINTERS_DISABLE;
      pc.TextureCacheInvalidationEnable =
         flags & PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE;
      pc.Address = rw_bo(bo, offset);
      pc.ImmediateData = imm;
   }
}

void
genX(init_state)(struct iris_context *ice)
{
   struct pipe_context *ctx = &ice->ctx;
   struct iris_screen *screen = (struct iris_screen *)ctx->screen;

   ctx->create_blend_state = iris_create_blend_state;
   ctx->create_depth_stencil_alpha_state = iris_create_zsa_state;
   ctx->create_rasterizer_state = iris_create_rasterizer_state;
   ctx->create_sampler_state = iris_create_sampler_state;
   ctx->create_sampler_view = iris_create_sampler_view;
   ctx->create_surface = iris_create_surface;
   ctx->create_vertex_elements_state = iris_create_vertex_elements;
   ctx->bind_blend_state = iris_bind_blend_state;
   ctx->bind_depth_stencil_alpha_state = iris_bind_zsa_state;
   ctx->bind_sampler_states = iris_bind_sampler_states;
   ctx->bind_rasterizer_state = iris_bind_rasterizer_state;
   ctx->bind_vertex_elements_state = iris_bind_vertex_elements_state;
   ctx->delete_blend_state = iris_delete_state;
   ctx->delete_depth_stencil_alpha_state = iris_delete_state;
   ctx->delete_fs_state = iris_delete_state;
   ctx->delete_rasterizer_state = iris_delete_state;
   ctx->delete_sampler_state = iris_delete_state;
   ctx->delete_vertex_elements_state = iris_delete_state;
   ctx->delete_tcs_state = iris_delete_state;
   ctx->delete_tes_state = iris_delete_state;
   ctx->delete_gs_state = iris_delete_state;
   ctx->delete_vs_state = iris_delete_state;
   ctx->set_blend_color = iris_set_blend_color;
   ctx->set_clip_state = iris_set_clip_state;
   ctx->set_constant_buffer = iris_set_constant_buffer;
   ctx->set_shader_buffers = iris_set_shader_buffers;
   ctx->set_shader_images = iris_set_shader_images;
   ctx->set_sampler_views = iris_set_sampler_views;
   ctx->set_tess_state = iris_set_tess_state;
   ctx->set_framebuffer_state = iris_set_framebuffer_state;
   ctx->set_polygon_stipple = iris_set_polygon_stipple;
   ctx->set_sample_mask = iris_set_sample_mask;
   ctx->set_scissor_states = iris_set_scissor_states;
   ctx->set_stencil_ref = iris_set_stencil_ref;
   ctx->set_vertex_buffers = iris_set_vertex_buffers;
   ctx->set_viewport_states = iris_set_viewport_states;
   ctx->sampler_view_destroy = iris_sampler_view_destroy;
   ctx->surface_destroy = iris_surface_destroy;
   ctx->draw_vbo = iris_draw_vbo;
   ctx->launch_grid = iris_launch_grid;
   ctx->create_stream_output_target = iris_create_stream_output_target;
   ctx->stream_output_target_destroy = iris_stream_output_target_destroy;
   ctx->set_stream_output_targets = iris_set_stream_output_targets;

   ice->vtbl.destroy_state = iris_destroy_state;
   ice->vtbl.init_render_context = iris_init_render_context;
   ice->vtbl.init_compute_context = iris_init_compute_context;
   ice->vtbl.upload_render_state = iris_upload_render_state;
   ice->vtbl.update_surface_base_address = iris_update_surface_base_address;
   ice->vtbl.upload_compute_state = iris_upload_compute_state;
   ice->vtbl.emit_raw_pipe_control = iris_emit_raw_pipe_control;
   ice->vtbl.load_register_imm32 = iris_load_register_imm32;
   ice->vtbl.load_register_imm64 = iris_load_register_imm64;
   ice->vtbl.load_register_mem32 = iris_load_register_mem32;
   ice->vtbl.load_register_mem64 = iris_load_register_mem64;
   ice->vtbl.store_register_mem32 = iris_store_register_mem32;
   ice->vtbl.store_register_mem64 = iris_store_register_mem64;
   ice->vtbl.store_data_imm32 = iris_store_data_imm32;
   ice->vtbl.store_data_imm64 = iris_store_data_imm64;
   ice->vtbl.copy_mem_mem = iris_copy_mem_mem;
   ice->vtbl.derived_program_state_size = iris_derived_program_state_size;
   ice->vtbl.store_derived_program_state = iris_store_derived_program_state;
   ice->vtbl.create_so_decl_list = iris_create_so_decl_list;
   ice->vtbl.populate_vs_key = iris_populate_vs_key;
   ice->vtbl.populate_tcs_key = iris_populate_tcs_key;
   ice->vtbl.populate_tes_key = iris_populate_tes_key;
   ice->vtbl.populate_gs_key = iris_populate_gs_key;
   ice->vtbl.populate_fs_key = iris_populate_fs_key;
   ice->vtbl.populate_cs_key = iris_populate_cs_key;

   ice->state.dirty = ~0ull;

   ice->state.statistics_counters_enabled = true;

   ice->state.sample_mask = 0xffff;
   ice->state.num_viewports = 1;
   ice->state.genx = calloc(1, sizeof(struct iris_genx_state));

   /* Make a 1x1x1 null surface for unbound textures */
   void *null_surf_map =
      upload_state(ice->state.surface_uploader, &ice->state.unbound_tex,
                   4 * GENX(RENDER_SURFACE_STATE_length), 64);
   isl_null_fill_state(&screen->isl_dev, null_surf_map, isl_extent3d(1, 1, 1));
   ice->state.unbound_tex.offset +=
      iris_bo_offset_from_base_address(iris_resource_bo(ice->state.unbound_tex.res));

   /* Default all scissor rectangles to be empty regions. */
   for (int i = 0; i < IRIS_MAX_VIEWPORTS; i++) {
      ice->state.scissors[i] = (struct pipe_scissor_state) {
         .minx = 1, .maxx = 0, .miny = 1, .maxy = 0,
      };
   }
}