summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/r300/r300_state.c
blob: 8abc65a2b8e3460ee88d7235f26c1f5f080a4805 (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
/*
 * Copyright 2008 Corbin Simpson <MostAwesomeDude@gmail.com>
 * Copyright 2009 Marek Olšák <maraeo@gmail.com>
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * on the rights to use, copy, modify, merge, publish, distribute, sub
 * license, and/or sell copies of the Software, and to permit persons to whom
 * the Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
 * USE OR OTHER DEALINGS IN THE SOFTWARE. */

#include "draw/draw_context.h"

#include "util/u_blitter.h"
#include "util/u_math.h"
#include "util/u_mm.h"
#include "util/u_memory.h"
#include "util/u_pack_color.h"

#include "tgsi/tgsi_parse.h"

#include "pipe/p_config.h"

#include "r300_cb.h"
#include "r300_context.h"
#include "r300_emit.h"
#include "r300_reg.h"
#include "r300_screen.h"
#include "r300_screen_buffer.h"
#include "r300_state_inlines.h"
#include "r300_fs.h"
#include "r300_texture.h"
#include "r300_vs.h"
#include "r300_winsys.h"
#include "r300_hyperz.h"

/* r300_state: Functions used to intialize state context by translating
 * Gallium state objects into semi-native r300 state objects. */

#define UPDATE_STATE(cso, atom) \
    if (cso != atom.state) { \
        atom.state = cso;    \
        atom.dirty = TRUE;   \
    }

static boolean blend_discard_if_src_alpha_0(unsigned srcRGB, unsigned srcA,
                                            unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_ALPHA == 0, and the following state is set, the colorbuffer
     * will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_SRC_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_SRC_COLOR ||
            srcA == PIPE_BLENDFACTOR_SRC_ALPHA ||
            srcA == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE ||
            srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            dstA == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            dstA == PIPE_BLENDFACTOR_ONE);
}

static boolean blend_discard_if_src_alpha_1(unsigned srcRGB, unsigned srcA,
                                            unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_ALPHA == 1, and the following state is set, the colorbuffer
     * will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            srcA == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_SRC_ALPHA ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_SRC_COLOR ||
            dstA == PIPE_BLENDFACTOR_SRC_ALPHA ||
            dstA == PIPE_BLENDFACTOR_ONE);
}

static boolean blend_discard_if_src_color_0(unsigned srcRGB, unsigned srcA,
                                            unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_COLOR == (0,0,0), and the following state is set, the colorbuffer
     * will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_SRC_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_ONE);
}

static boolean blend_discard_if_src_color_1(unsigned srcRGB, unsigned srcA,
                                            unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_COLOR == (1,1,1), and the following state is set, the colorbuffer
     * will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_SRC_COLOR ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_ONE);
}

static boolean blend_discard_if_src_alpha_color_0(unsigned srcRGB, unsigned srcA,
                                                  unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_ALPHA_COLOR == (0,0,0,0), and the following state is set,
     * the colorbuffer will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_SRC_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_SRC_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_SRC_COLOR ||
            srcA == PIPE_BLENDFACTOR_SRC_ALPHA ||
            srcA == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE ||
            srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            dstRGB == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            dstA == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            dstA == PIPE_BLENDFACTOR_ONE);
}

static boolean blend_discard_if_src_alpha_color_1(unsigned srcRGB, unsigned srcA,
                                                  unsigned dstRGB, unsigned dstA)
{
    /* If the blend equation is ADD or REVERSE_SUBTRACT,
     * SRC_ALPHA_COLOR == (1,1,1,1), and the following state is set,
     * the colorbuffer will not be changed.
     * Notice that the dst factors are the src factors inverted. */
    return (srcRGB == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_ZERO) &&
           (srcA == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
            srcA == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
            srcA == PIPE_BLENDFACTOR_ZERO) &&
           (dstRGB == PIPE_BLENDFACTOR_SRC_COLOR ||
            dstRGB == PIPE_BLENDFACTOR_SRC_ALPHA ||
            dstRGB == PIPE_BLENDFACTOR_ONE) &&
           (dstA == PIPE_BLENDFACTOR_SRC_COLOR ||
            dstA == PIPE_BLENDFACTOR_SRC_ALPHA ||
            dstA == PIPE_BLENDFACTOR_ONE);
}

static unsigned bgra_cmask(unsigned mask)
{
    /* Gallium uses RGBA color ordering while R300 expects BGRA. */

    return ((mask & PIPE_MASK_R) << 2) |
           ((mask & PIPE_MASK_B) >> 2) |
           (mask & (PIPE_MASK_G | PIPE_MASK_A));
}

/* Create a new blend state based on the CSO blend state.
 *
 * This encompasses alpha blending, logic/raster ops, and blend dithering. */
static void* r300_create_blend_state(struct pipe_context* pipe,
                                     const struct pipe_blend_state* state)
{
    struct r300_screen* r300screen = r300_screen(pipe->screen);
    struct r300_blend_state* blend = CALLOC_STRUCT(r300_blend_state);
    uint32_t blend_control = 0;       /* R300_RB3D_CBLEND: 0x4e04 */
    uint32_t alpha_blend_control = 0; /* R300_RB3D_ABLEND: 0x4e08 */
    uint32_t color_channel_mask = 0;  /* R300_RB3D_COLOR_CHANNEL_MASK: 0x4e0c */
    uint32_t rop = 0;                 /* R300_RB3D_ROPCNTL: 0x4e18 */
    uint32_t dither = 0;              /* R300_RB3D_DITHER_CTL: 0x4e50 */
    CB_LOCALS;

    if (state->rt[0].blend_enable)
    {
        unsigned eqRGB = state->rt[0].rgb_func;
        unsigned srcRGB = state->rt[0].rgb_src_factor;
        unsigned dstRGB = state->rt[0].rgb_dst_factor;

        unsigned eqA = state->rt[0].alpha_func;
        unsigned srcA = state->rt[0].alpha_src_factor;
        unsigned dstA = state->rt[0].alpha_dst_factor;

        /* despite the name, ALPHA_BLEND_ENABLE has nothing to do with alpha,
         * this is just the crappy D3D naming */
        blend_control = R300_ALPHA_BLEND_ENABLE |
            r300_translate_blend_function(eqRGB) |
            ( r300_translate_blend_factor(srcRGB) << R300_SRC_BLEND_SHIFT) |
            ( r300_translate_blend_factor(dstRGB) << R300_DST_BLEND_SHIFT);

        /* Optimization: some operations do not require the destination color.
         *
         * When SRC_ALPHA_SATURATE is used, colorbuffer reads must be enabled,
         * otherwise blending gives incorrect results. It seems to be
         * a hardware bug. */
        if (eqRGB == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MIN ||
            eqRGB == PIPE_BLEND_MAX || eqA == PIPE_BLEND_MAX ||
            dstRGB != PIPE_BLENDFACTOR_ZERO ||
            dstA != PIPE_BLENDFACTOR_ZERO ||
            srcRGB == PIPE_BLENDFACTOR_DST_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_DST_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_INV_DST_COLOR ||
            srcRGB == PIPE_BLENDFACTOR_INV_DST_ALPHA ||
            srcA == PIPE_BLENDFACTOR_DST_COLOR ||
            srcA == PIPE_BLENDFACTOR_DST_ALPHA ||
            srcA == PIPE_BLENDFACTOR_INV_DST_COLOR ||
            srcA == PIPE_BLENDFACTOR_INV_DST_ALPHA ||
            srcRGB == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE) {
            /* Enable reading from the colorbuffer. */
            blend_control |= R300_READ_ENABLE;

            if (r300screen->caps.is_r500) {
                /* Optimization: Depending on incoming pixels, we can
                 * conditionally disable the reading in hardware... */
                if (eqRGB != PIPE_BLEND_MIN && eqA != PIPE_BLEND_MIN &&
                    eqRGB != PIPE_BLEND_MAX && eqA != PIPE_BLEND_MAX) {
                    /* Disable reading if SRC_ALPHA == 0. */
                    if ((dstRGB == PIPE_BLENDFACTOR_SRC_ALPHA ||
                         dstRGB == PIPE_BLENDFACTOR_ZERO) &&
                        (dstA == PIPE_BLENDFACTOR_SRC_COLOR ||
                         dstA == PIPE_BLENDFACTOR_SRC_ALPHA ||
                         dstA == PIPE_BLENDFACTOR_ZERO)) {
                         blend_control |= R500_SRC_ALPHA_0_NO_READ;
                    }

                    /* Disable reading if SRC_ALPHA == 1. */
                    if ((dstRGB == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
                         dstRGB == PIPE_BLENDFACTOR_ZERO) &&
                        (dstA == PIPE_BLENDFACTOR_INV_SRC_COLOR ||
                         dstA == PIPE_BLENDFACTOR_INV_SRC_ALPHA ||
                         dstA == PIPE_BLENDFACTOR_ZERO)) {
                         blend_control |= R500_SRC_ALPHA_1_NO_READ;
                    }
                }
            }
        }

        /* Optimization: discard pixels which don't change the colorbuffer.
         *
         * The code below is non-trivial and some math is involved.
         *
         * Discarding pixels must be disabled when FP16 AA is enabled.
         * This is a hardware bug. Also, this implementation wouldn't work
         * with FP blending enabled and equation clamping disabled.
         *
         * Equations other than ADD are rarely used and therefore won't be
         * optimized. */
        if ((eqRGB == PIPE_BLEND_ADD || eqRGB == PIPE_BLEND_REVERSE_SUBTRACT) &&
            (eqA == PIPE_BLEND_ADD || eqA == PIPE_BLEND_REVERSE_SUBTRACT)) {
            /* ADD: X+Y
             * REVERSE_SUBTRACT: Y-X
             *
             * The idea is:
             * If X = src*srcFactor = 0 and Y = dst*dstFactor = 1,
             * then CB will not be changed.
             *
             * Given the srcFactor and dstFactor variables, we can derive
             * what src and dst should be equal to and discard appropriate
             * pixels.
             */
            if (blend_discard_if_src_alpha_0(srcRGB, srcA, dstRGB, dstA)) {
                blend_control |= R300_DISCARD_SRC_PIXELS_SRC_ALPHA_0;
            } else if (blend_discard_if_src_alpha_1(srcRGB, srcA,
                                                    dstRGB, dstA)) {
                blend_control |= R300_DISCARD_SRC_PIXELS_SRC_ALPHA_1;
            } else if (blend_discard_if_src_color_0(srcRGB, srcA,
                                                    dstRGB, dstA)) {
                blend_control |= R300_DISCARD_SRC_PIXELS_SRC_COLOR_0;
            } else if (blend_discard_if_src_color_1(srcRGB, srcA,
                                                    dstRGB, dstA)) {
                blend_control |= R300_DISCARD_SRC_PIXELS_SRC_COLOR_1;
            } else if (blend_discard_if_src_alpha_color_0(srcRGB, srcA,
                                                          dstRGB, dstA)) {
                blend_control |=
                    R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_0;
            } else if (blend_discard_if_src_alpha_color_1(srcRGB, srcA,
                                                          dstRGB, dstA)) {
                blend_control |=
                    R300_DISCARD_SRC_PIXELS_SRC_ALPHA_COLOR_1;
            }
        }

        /* separate alpha */
        if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) {
            blend_control |= R300_SEPARATE_ALPHA_ENABLE;
            alpha_blend_control =
                r300_translate_blend_function(eqA) |
                (r300_translate_blend_factor(srcA) << R300_SRC_BLEND_SHIFT) |
                (r300_translate_blend_factor(dstA) << R300_DST_BLEND_SHIFT);
        }
    }

    /* PIPE_LOGICOP_* don't need to be translated, fortunately. */
    if (state->logicop_enable) {
        rop = R300_RB3D_ROPCNTL_ROP_ENABLE |
                (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT;
    }

    /* Color channel masks for all MRTs. */
    color_channel_mask = bgra_cmask(state->rt[0].colormask);
    if (r300screen->caps.is_r500 && state->independent_blend_enable) {
        if (state->rt[1].blend_enable) {
            color_channel_mask |= bgra_cmask(state->rt[1].colormask) << 4;
        }
        if (state->rt[2].blend_enable) {
            color_channel_mask |= bgra_cmask(state->rt[2].colormask) << 8;
        }
        if (state->rt[3].blend_enable) {
            color_channel_mask |= bgra_cmask(state->rt[3].colormask) << 12;
        }
    }

    /* Neither fglrx nor classic r300 ever set this, regardless of dithering
     * state. Since it's an optional implementation detail, we can leave it
     * out and never dither.
     *
     * This could be revisited if we ever get quality or conformance hints.
     *
    if (state->dither) {
        dither = R300_RB3D_DITHER_CTL_DITHER_MODE_LUT |
                        R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_LUT;
    }
    */

    /* Build a command buffer. */
    BEGIN_CB(blend->cb, 8);
    OUT_CB_REG(R300_RB3D_ROPCNTL, rop);
    OUT_CB_REG_SEQ(R300_RB3D_CBLEND, 3);
    OUT_CB(blend_control);
    OUT_CB(alpha_blend_control);
    OUT_CB(color_channel_mask);
    OUT_CB_REG(R300_RB3D_DITHER_CTL, dither);
    END_CB;

    /* The same as above, but with no colorbuffer reads and writes. */
    BEGIN_CB(blend->cb_no_readwrite, 8);
    OUT_CB_REG(R300_RB3D_ROPCNTL, rop);
    OUT_CB_REG_SEQ(R300_RB3D_CBLEND, 3);
    OUT_CB(0);
    OUT_CB(0);
    OUT_CB(0);
    OUT_CB_REG(R300_RB3D_DITHER_CTL, dither);
    END_CB;

    return (void*)blend;
}

/* Bind blend state. */
static void r300_bind_blend_state(struct pipe_context* pipe,
                                  void* state)
{
    struct r300_context* r300 = r300_context(pipe);

    UPDATE_STATE(state, r300->blend_state);
}

/* Free blend state. */
static void r300_delete_blend_state(struct pipe_context* pipe,
                                    void* state)
{
    FREE(state);
}

/* Convert float to 10bit integer */
static unsigned float_to_fixed10(float f)
{
    return CLAMP((unsigned)(f * 1023.9f), 0, 1023);
}

/* Set blend color.
 * Setup both R300 and R500 registers, figure out later which one to write. */
static void r300_set_blend_color(struct pipe_context* pipe,
                                 const struct pipe_blend_color* color)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_blend_color_state* state =
        (struct r300_blend_color_state*)r300->blend_color_state.state;
    CB_LOCALS;

    if (r300->screen->caps.is_r500) {
        /* XXX if FP16 blending is enabled, we should use the FP16 format */
        BEGIN_CB(state->cb, 3);
        OUT_CB_REG_SEQ(R500_RB3D_CONSTANT_COLOR_AR, 2);
        OUT_CB(float_to_fixed10(color->color[0]) |
               (float_to_fixed10(color->color[3]) << 16));
        OUT_CB(float_to_fixed10(color->color[2]) |
               (float_to_fixed10(color->color[1]) << 16));
        END_CB;
    } else {
        union util_color uc;
        util_pack_color(color->color, PIPE_FORMAT_B8G8R8A8_UNORM, &uc);

        BEGIN_CB(state->cb, 2);
        OUT_CB_REG(R300_RB3D_BLEND_COLOR, uc.ui);
        END_CB;
    }

    r300->blend_color_state.dirty = TRUE;
}

static void r300_set_clip_state(struct pipe_context* pipe,
                                const struct pipe_clip_state* state)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_clip_state *clip =
            (struct r300_clip_state*)r300->clip_state.state;
    CB_LOCALS;

    clip->clip = *state;

    if (r300->screen->caps.has_tcl) {
        r300->clip_state.size = 2 + !!state->nr * 3 + state->nr * 4;

        BEGIN_CB(clip->cb, r300->clip_state.size);
        if (state->nr) {
           OUT_CB_REG(R300_VAP_PVS_VECTOR_INDX_REG,
                   (r300->screen->caps.is_r500 ?
                    R500_PVS_UCP_START : R300_PVS_UCP_START));
           OUT_CB_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, state->nr * 4);
           OUT_CB_TABLE(state->ucp, state->nr * 4);
        }
        OUT_CB_REG(R300_VAP_CLIP_CNTL, ((1 << state->nr) - 1) |
                R300_PS_UCP_MODE_CLIP_AS_TRIFAN |
                (state->depth_clamp ? R300_CLIP_DISABLE : 0));
        END_CB;

        r300->clip_state.dirty = TRUE;
    } else {
        draw_set_clip_state(r300->draw, state);
    }
}

static void
r300_set_sample_mask(struct pipe_context *pipe,
                     unsigned sample_mask)
{
}


/* Create a new depth, stencil, and alpha state based on the CSO dsa state.
 *
 * This contains the depth buffer, stencil buffer, alpha test, and such.
 * On the Radeon, depth and stencil buffer setup are intertwined, which is
 * the reason for some of the strange-looking assignments across registers. */
static void*
        r300_create_dsa_state(struct pipe_context* pipe,
                              const struct pipe_depth_stencil_alpha_state* state)
{
    struct r300_capabilities *caps = &r300_screen(pipe->screen)->caps;
    struct r300_dsa_state* dsa = CALLOC_STRUCT(r300_dsa_state);
    CB_LOCALS;

    dsa->dsa = *state;

    /* Depth test setup. - separate write mask depth for decomp flush */
    if (state->depth.writemask) {
        dsa->z_buffer_control |= R300_Z_WRITE_ENABLE;
    }

    if (state->depth.enabled) {
        dsa->z_buffer_control |= R300_Z_ENABLE;

        dsa->z_stencil_control |=
            (r300_translate_depth_stencil_function(state->depth.func) <<
                R300_Z_FUNC_SHIFT);
    }

    /* Stencil buffer setup. */
    if (state->stencil[0].enabled) {
        dsa->z_buffer_control |= R300_STENCIL_ENABLE;
        dsa->z_stencil_control |=
            (r300_translate_depth_stencil_function(state->stencil[0].func) <<
                R300_S_FRONT_FUNC_SHIFT) |
            (r300_translate_stencil_op(state->stencil[0].fail_op) <<
                R300_S_FRONT_SFAIL_OP_SHIFT) |
            (r300_translate_stencil_op(state->stencil[0].zpass_op) <<
                R300_S_FRONT_ZPASS_OP_SHIFT) |
            (r300_translate_stencil_op(state->stencil[0].zfail_op) <<
                R300_S_FRONT_ZFAIL_OP_SHIFT);

        dsa->stencil_ref_mask =
                (state->stencil[0].valuemask << R300_STENCILMASK_SHIFT) |
                (state->stencil[0].writemask << R300_STENCILWRITEMASK_SHIFT);

        if (state->stencil[1].enabled) {
            dsa->two_sided = TRUE;

            dsa->z_buffer_control |= R300_STENCIL_FRONT_BACK;
            dsa->z_stencil_control |=
            (r300_translate_depth_stencil_function(state->stencil[1].func) <<
                R300_S_BACK_FUNC_SHIFT) |
            (r300_translate_stencil_op(state->stencil[1].fail_op) <<
                R300_S_BACK_SFAIL_OP_SHIFT) |
            (r300_translate_stencil_op(state->stencil[1].zpass_op) <<
                R300_S_BACK_ZPASS_OP_SHIFT) |
            (r300_translate_stencil_op(state->stencil[1].zfail_op) <<
                R300_S_BACK_ZFAIL_OP_SHIFT);

            dsa->stencil_ref_bf =
                (state->stencil[1].valuemask << R300_STENCILMASK_SHIFT) |
                (state->stencil[1].writemask << R300_STENCILWRITEMASK_SHIFT);

            if (caps->is_r500) {
                dsa->z_buffer_control |= R500_STENCIL_REFMASK_FRONT_BACK;
            } else {
                dsa->two_sided_stencil_ref =
                  (state->stencil[0].valuemask != state->stencil[1].valuemask ||
                   state->stencil[0].writemask != state->stencil[1].writemask);
            }
        }
    }

    /* Alpha test setup. */
    if (state->alpha.enabled) {
        dsa->alpha_function =
            r300_translate_alpha_function(state->alpha.func) |
            R300_FG_ALPHA_FUNC_ENABLE;

        /* We could use 10bit alpha ref but who needs that? */
        dsa->alpha_function |= float_to_ubyte(state->alpha.ref_value);

        if (caps->is_r500)
            dsa->alpha_function |= R500_FG_ALPHA_FUNC_8BIT;
    }

    BEGIN_CB(&dsa->cb_begin, 8);
    OUT_CB_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function);
    OUT_CB_REG_SEQ(R300_ZB_CNTL, 3);
    OUT_CB(dsa->z_buffer_control);
    OUT_CB(dsa->z_stencil_control);
    OUT_CB(dsa->stencil_ref_mask);
    OUT_CB_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf);
    END_CB;

    BEGIN_CB(dsa->cb_no_readwrite, 8);
    OUT_CB_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function);
    OUT_CB_REG_SEQ(R300_ZB_CNTL, 3);
    OUT_CB(0);
    OUT_CB(0);
    OUT_CB(0);
    OUT_CB_REG(R500_ZB_STENCILREFMASK_BF, 0);
    END_CB;

    return (void*)dsa;
}

static void r300_dsa_inject_stencilref(struct r300_context *r300)
{
    struct r300_dsa_state *dsa =
            (struct r300_dsa_state*)r300->dsa_state.state;

    if (!dsa)
        return;

    dsa->stencil_ref_mask =
        (dsa->stencil_ref_mask & ~R300_STENCILREF_MASK) |
        r300->stencil_ref.ref_value[0];
    dsa->stencil_ref_bf =
        (dsa->stencil_ref_bf & ~R300_STENCILREF_MASK) |
        r300->stencil_ref.ref_value[1];
}

/* Bind DSA state. */
static void r300_bind_dsa_state(struct pipe_context* pipe,
                                void* state)
{
    struct r300_context* r300 = r300_context(pipe);

    if (!state) {
        return;
    }

    UPDATE_STATE(state, r300->dsa_state);

    r300->hyperz_state.dirty = TRUE; /* Will be updated before the emission. */
    r300_dsa_inject_stencilref(r300);
}

/* Free DSA state. */
static void r300_delete_dsa_state(struct pipe_context* pipe,
                                  void* state)
{
    FREE(state);
}

static void r300_set_stencil_ref(struct pipe_context* pipe,
                                 const struct pipe_stencil_ref* sr)
{
    struct r300_context* r300 = r300_context(pipe);

    r300->stencil_ref = *sr;

    r300_dsa_inject_stencilref(r300);
    r300->dsa_state.dirty = TRUE;
}

static void r300_tex_set_tiling_flags(struct r300_context *r300,
                                      struct r300_texture *tex, unsigned level)
{
    /* Check if the macrotile flag needs to be changed.
     * Skip changing the flags otherwise. */
    if (tex->desc.macrotile[tex->surface_level] !=
        tex->desc.macrotile[level]) {
        /* Tiling determines how DRM treats the buffer data.
         * We must flush CS when changing it if the buffer is referenced. */
        if (r300->rws->cs_is_buffer_referenced(r300->cs,
                                               tex->buffer, R300_REF_CS))
            r300->context.flush(&r300->context, 0, NULL);

        r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
                tex->desc.microtile, tex->desc.macrotile[level],
                tex->desc.stride_in_bytes[0]);

        tex->surface_level = level;
    }
}

/* This switcheroo is needed just because of goddamned MACRO_SWITCH. */
static void r300_fb_set_tiling_flags(struct r300_context *r300,
                               const struct pipe_framebuffer_state *state)
{
    unsigned i;

    /* Set tiling flags for new surfaces. */
    for (i = 0; i < state->nr_cbufs; i++) {
        r300_tex_set_tiling_flags(r300,
                                  r300_texture(state->cbufs[i]->texture),
                                  state->cbufs[i]->level);
    }
    if (state->zsbuf) {
        r300_tex_set_tiling_flags(r300,
                                  r300_texture(state->zsbuf->texture),
                                  state->zsbuf->level);
    }
}

static void r300_print_fb_surf_info(struct pipe_surface *surf, unsigned index,
                                    const char *binding)
{
    struct pipe_resource *tex = surf->texture;
    struct r300_texture *rtex = r300_texture(tex);

    fprintf(stderr,
            "r300:   %s[%i] Dim: %ix%i, Offset: %i, ZSlice: %i, "
            "Face: %i, Level: %i, Format: %s\n"

            "r300:     TEX: Macro: %s, Micro: %s, Pitch: %i, "
            "Dim: %ix%ix%i, LastLevel: %i, Format: %s\n",

            binding, index, surf->width, surf->height, surf->offset,
            surf->zslice, surf->face, surf->level,
            util_format_short_name(surf->format),

            rtex->desc.macrotile[0] ? "YES" : " NO",
            rtex->desc.microtile ? "YES" : " NO",
            rtex->desc.stride_in_pixels[0],
            tex->width0, tex->height0, tex->depth0,
            tex->last_level, util_format_short_name(tex->format));
}

void r300_mark_fb_state_dirty(struct r300_context *r300,
                              enum r300_fb_state_change change)
{
    struct pipe_framebuffer_state *state = r300->fb_state.state;
    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);

    /* What is marked as dirty depends on the enum r300_fb_state_change. */
    r300->gpu_flush.dirty = TRUE;
    r300->fb_state.dirty = TRUE;
    r300->hyperz_state.dirty = TRUE;

    if (change == R300_CHANGED_FB_STATE) {
        r300->aa_state.dirty = TRUE;
        r300->fb_state_pipelined.dirty = TRUE;
    }

    /* Now compute the fb_state atom size. */
    r300->fb_state.size = 2 + (8 * state->nr_cbufs);

    if (r300->cbzb_clear)
        r300->fb_state.size += 10;
    else if (state->zsbuf) {
        r300->fb_state.size += 10;
        if (has_hyperz)
            r300->fb_state.size += r300->screen->caps.hiz_ram ? 8 : 4;
    }

    /* The size of the rest of atoms stays the same. */
}

static void
    r300_set_framebuffer_state(struct pipe_context* pipe,
                               const struct pipe_framebuffer_state* state)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_aa_state *aa = (struct r300_aa_state*)r300->aa_state.state;
    struct pipe_framebuffer_state *old_state = r300->fb_state.state;
    boolean has_hyperz = r300->rws->get_value(r300->rws, R300_CAN_HYPERZ);
    unsigned max_width, max_height, i;
    uint32_t zbuffer_bpp = 0;
    int blocksize;

    if (r300->screen->caps.is_r500) {
        max_width = max_height = 4096;
    } else if (r300->screen->caps.is_r400) {
        max_width = max_height = 4021;
    } else {
        max_width = max_height = 2560;
    }

    if (state->width > max_width || state->height > max_height) {
        fprintf(stderr, "r300: Implementation error: Render targets are too "
        "big in %s, refusing to bind framebuffer state!\n", __FUNCTION__);
        return;
    }

    /* If nr_cbufs is changed from zero to non-zero or vice versa... */
    if (!!old_state->nr_cbufs != !!state->nr_cbufs) {
        r300->blend_state.dirty = TRUE;
    }
    /* If zsbuf is set from NULL to non-NULL or vice versa.. */
    if (!!old_state->zsbuf != !!state->zsbuf) {
        r300->dsa_state.dirty = TRUE;
    }

    /* The tiling flags are dependent on the surface miplevel, unfortunately. */
    r300_fb_set_tiling_flags(r300, state);

    util_assign_framebuffer_state(r300->fb_state.state, state);

    r300_mark_fb_state_dirty(r300, R300_CHANGED_FB_STATE);

    r300->hiz_enable = false;
    r300->z_fastfill = false;
    r300->z_compression = false;
    
    if (state->zsbuf) {
        blocksize = util_format_get_blocksize(state->zsbuf->texture->format);
        switch (blocksize) {
        case 2:
            zbuffer_bpp = 16;
            break;
        case 4:
            zbuffer_bpp = 24;
            break;
        }
        if (has_hyperz) {
            struct r300_surface *zs_surf = r300_surface(state->zsbuf);
            struct r300_texture *tex;
            int compress = r300->screen->caps.is_rv350 ? RV350_Z_COMPRESS_88 : R300_Z_COMPRESS_44;
            int level = zs_surf->base.level;

            tex = r300_texture(zs_surf->base.texture);

            /* work out whether we can support hiz on this buffer */
            r300_hiz_alloc_block(r300, zs_surf);
        
            /* work out whether we can support zmask features on this buffer */
            r300_zmask_alloc_block(r300, zs_surf, compress);

            if (tex->hiz_mem[level]) {
                r300->hiz_enable = 1;
            }

            if (tex->zmask_mem[level]) {
                r300->z_fastfill = 1;
                /* compression causes hangs on 16-bit */
                if (zbuffer_bpp == 24)
                    r300->z_compression = compress;
            }
            DBG(r300, DBG_HYPERZ,
                "hyper-z features: hiz: %d @ %08x z-compression: %d z-fastfill: %d @ %08x\n", r300->hiz_enable,
                tex->hiz_mem[level] ? tex->hiz_mem[level]->ofs : 0xdeadbeef,
                r300->z_compression, r300->z_fastfill,
                tex->zmask_mem[level] ? tex->zmask_mem[level]->ofs : 0xdeadbeef);
        }

        /* Polygon offset depends on the zbuffer bit depth. */
        if (r300->zbuffer_bpp != zbuffer_bpp) {
            r300->zbuffer_bpp = zbuffer_bpp;

            if (r300->polygon_offset_enabled)
                r300->rs_state.dirty = TRUE;
        }
    }

    /* Set up AA config. */
    if (r300->rws->get_value(r300->rws, R300_VID_DRM_2_3_0)) {
        if (state->nr_cbufs && state->cbufs[0]->texture->nr_samples > 1) {
            aa->aa_config = R300_GB_AA_CONFIG_AA_ENABLE;

            switch (state->cbufs[0]->texture->nr_samples) {
                case 2:
                    aa->aa_config |= R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_2;
                    break;
                case 3:
                    aa->aa_config |= R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_3;
                    break;
                case 4:
                    aa->aa_config |= R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_4;
                    break;
                case 6:
                    aa->aa_config |= R300_GB_AA_CONFIG_NUM_AA_SUBSAMPLES_6;
                    break;
            }
        } else {
            aa->aa_config = 0;
        }
    }

    if (DBG_ON(r300, DBG_FB)) {
        fprintf(stderr, "r300: set_framebuffer_state:\n");
        for (i = 0; i < state->nr_cbufs; i++) {
            r300_print_fb_surf_info(state->cbufs[i], i, "CB");
        }
        if (state->zsbuf) {
            r300_print_fb_surf_info(state->zsbuf, 0, "ZB");
        }
    }
}

/* Create fragment shader state. */
static void* r300_create_fs_state(struct pipe_context* pipe,
                                  const struct pipe_shader_state* shader)
{
    struct r300_fragment_shader* fs = NULL;

    fs = (struct r300_fragment_shader*)CALLOC_STRUCT(r300_fragment_shader);

    /* Copy state directly into shader. */
    fs->state = *shader;
    fs->state.tokens = tgsi_dup_tokens(shader->tokens);

    return (void*)fs;
}

void r300_mark_fs_code_dirty(struct r300_context *r300)
{
    struct r300_fragment_shader* fs = r300_fs(r300);

    r300->fs.dirty = TRUE;
    r300->fs_rc_constant_state.dirty = TRUE;
    r300->fs_constants.dirty = TRUE;
    r300->fs.size = fs->shader->cb_code_size;

    if (r300->screen->caps.is_r500) {
        r300->fs_rc_constant_state.size = fs->shader->rc_state_count * 7;
        r300->fs_constants.size = fs->shader->externals_count * 4 + 3;
    } else {
        r300->fs_rc_constant_state.size = fs->shader->rc_state_count * 5;
        r300->fs_constants.size = fs->shader->externals_count * 4 + 1;
    }
}

/* Bind fragment shader state. */
static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;

    if (fs == NULL) {
        r300->fs.state = NULL;
        return;
    }

    r300->fs.state = fs;
    r300_pick_fragment_shader(r300);
    r300_mark_fs_code_dirty(r300);

    r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
}

/* Delete fragment shader state. */
static void r300_delete_fs_state(struct pipe_context* pipe, void* shader)
{
    struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader;
    struct r300_fragment_shader_code *tmp, *ptr = fs->first;

    while (ptr) {
        tmp = ptr;
        ptr = ptr->next;
        rc_constants_destroy(&tmp->code.constants);
        FREE(tmp->cb_code);
        FREE(tmp);
    }
    FREE((void*)fs->state.tokens);
    FREE(shader);
}

static void r300_set_polygon_stipple(struct pipe_context* pipe,
                                     const struct pipe_poly_stipple* state)
{
    /* XXX no idea how to set this up, but not terribly important */
}

/* Create a new rasterizer state based on the CSO rasterizer state.
 *
 * This is a very large chunk of state, and covers most of the graphics
 * backend (GB), geometry assembly (GA), and setup unit (SU) blocks.
 *
 * In a not entirely unironic sidenote, this state has nearly nothing to do
 * with the actual block on the Radeon called the rasterizer (RS). */
static void* r300_create_rs_state(struct pipe_context* pipe,
                                  const struct pipe_rasterizer_state* state)
{
    struct r300_rs_state* rs = CALLOC_STRUCT(r300_rs_state);
    int i;
    float psiz;
    uint32_t vap_control_status;    /* R300_VAP_CNTL_STATUS: 0x2140 */
    uint32_t point_size;            /* R300_GA_POINT_SIZE: 0x421c */
    uint32_t point_minmax;          /* R300_GA_POINT_MINMAX: 0x4230 */
    uint32_t line_control;          /* R300_GA_LINE_CNTL: 0x4234 */
    uint32_t polygon_offset_enable; /* R300_SU_POLY_OFFSET_ENABLE: 0x42b4 */
    uint32_t cull_mode;             /* R300_SU_CULL_MODE: 0x42b8 */
    uint32_t line_stipple_config;   /* R300_GA_LINE_STIPPLE_CONFIG: 0x4328 */
    uint32_t line_stipple_value;    /* R300_GA_LINE_STIPPLE_VALUE: 0x4260 */
    uint32_t polygon_mode;          /* R300_GA_POLY_MODE: 0x4288 */
    uint32_t clip_rule;             /* R300_SC_CLIP_RULE: 0x43D0 */

    /* Specifies top of Raster pipe specific enable controls,
     * i.e. texture coordinates stuffing for points, lines, triangles */
    uint32_t stuffing_enable;       /* R300_GB_ENABLE: 0x4008 */

    /* Point sprites texture coordinates, 0: lower left, 1: upper right */
    float point_texcoord_left;      /* R300_GA_POINT_S0: 0x4200 */
    float point_texcoord_bottom = 0;/* R300_GA_POINT_T0: 0x4204 */
    float point_texcoord_right;     /* R300_GA_POINT_S1: 0x4208 */
    float point_texcoord_top = 0;   /* R300_GA_POINT_T1: 0x420c */
    CB_LOCALS;

    /* Copy rasterizer state. */
    rs->rs = *state;
    rs->rs_draw = *state;

    /* Override some states for Draw. */
    rs->rs_draw.sprite_coord_enable = 0; /* We can do this in HW. */

#ifdef PIPE_ARCH_LITTLE_ENDIAN
    vap_control_status = R300_VC_NO_SWAP;
#else
    vap_control_status = R300_VC_32BIT_SWAP;
#endif

    /* If no TCL engine is present, turn off the HW TCL. */
    if (!r300_screen(pipe->screen)->caps.has_tcl) {
        vap_control_status |= R300_VAP_TCL_BYPASS;
    }

    /* Point size width and height. */
    point_size =
        pack_float_16_6x(state->point_size) |
        (pack_float_16_6x(state->point_size) << R300_POINTSIZE_X_SHIFT);

    /* Point size clamping. */
    if (state->point_size_per_vertex) {
        /* Per-vertex point size.
         * Clamp to [0, max FB size] */
        psiz = pipe->screen->get_paramf(pipe->screen,
                                        PIPE_CAP_MAX_POINT_WIDTH);
        point_minmax =
            pack_float_16_6x(psiz) << R300_GA_POINT_MINMAX_MAX_SHIFT;
    } else {
        /* We cannot disable the point-size vertex output,
         * so clamp it. */
        psiz = state->point_size;
        point_minmax =
            (pack_float_16_6x(psiz) << R300_GA_POINT_MINMAX_MIN_SHIFT) |
            (pack_float_16_6x(psiz) << R300_GA_POINT_MINMAX_MAX_SHIFT);
    }

    /* Line control. */
    line_control = pack_float_16_6x(state->line_width) |
        R300_GA_LINE_CNTL_END_TYPE_COMP;

    /* Enable polygon mode */
    polygon_mode = 0;
    if (state->fill_front != PIPE_POLYGON_MODE_FILL ||
        state->fill_back != PIPE_POLYGON_MODE_FILL) {
        polygon_mode = R300_GA_POLY_MODE_DUAL;
    }

    /* Front face */
    if (state->front_ccw) 
        cull_mode = R300_FRONT_FACE_CCW;
    else
        cull_mode = R300_FRONT_FACE_CW;

    /* Polygon offset */
    polygon_offset_enable = 0;
    if (util_get_offset(state, state->fill_front)) {
       polygon_offset_enable |= R300_FRONT_ENABLE;
    }
    if (util_get_offset(state, state->fill_back)) {
       polygon_offset_enable |= R300_BACK_ENABLE;
    }

    rs->polygon_offset_enable = polygon_offset_enable != 0;

    /* Polygon mode */
    if (polygon_mode) {
       polygon_mode |=
          r300_translate_polygon_mode_front(state->fill_front);
       polygon_mode |=
          r300_translate_polygon_mode_back(state->fill_back);
    }

    if (state->cull_face & PIPE_FACE_FRONT) {
        cull_mode |= R300_CULL_FRONT;
    }
    if (state->cull_face & PIPE_FACE_BACK) {
        cull_mode |= R300_CULL_BACK;
    }

    if (state->line_stipple_enable) {
        line_stipple_config =
            R300_GA_LINE_STIPPLE_CONFIG_LINE_RESET_LINE |
            (fui((float)state->line_stipple_factor) &
                R300_GA_LINE_STIPPLE_CONFIG_STIPPLE_SCALE_MASK);
        /* XXX this might need to be scaled up */
        line_stipple_value = state->line_stipple_pattern;
    } else {
        line_stipple_config = 0;
        line_stipple_value = 0;
    }

    if (state->flatshade) {
        rs->color_control = R300_SHADE_MODEL_FLAT;
    } else {
        rs->color_control = R300_SHADE_MODEL_SMOOTH;
    }

    clip_rule = state->scissor ? 0xAAAA : 0xFFFF;

    /* Point sprites */
    stuffing_enable = 0;
    if (state->sprite_coord_enable) {
        stuffing_enable = R300_GB_POINT_STUFF_ENABLE;
	for (i = 0; i < 8; i++) {
	    if (state->sprite_coord_enable & (1 << i))
                stuffing_enable |=
                    R300_GB_TEX_ST << (R300_GB_TEX0_SOURCE_SHIFT + (i*2));
	}

        point_texcoord_left = 0.0f;
        point_texcoord_right = 1.0f;

        switch (state->sprite_coord_mode) {
            case PIPE_SPRITE_COORD_UPPER_LEFT:
                point_texcoord_top = 0.0f;
                point_texcoord_bottom = 1.0f;
                break;
            case PIPE_SPRITE_COORD_LOWER_LEFT:
                point_texcoord_top = 1.0f;
                point_texcoord_bottom = 0.0f;
                break;
        }
    }

    /* Build the main command buffer. */
    BEGIN_CB(rs->cb_main, 25);
    OUT_CB_REG(R300_VAP_CNTL_STATUS, vap_control_status);
    OUT_CB_REG(R300_GA_POINT_SIZE, point_size);
    OUT_CB_REG_SEQ(R300_GA_POINT_MINMAX, 2);
    OUT_CB(point_minmax);
    OUT_CB(line_control);
    OUT_CB_REG_SEQ(R300_SU_POLY_OFFSET_ENABLE, 2);
    OUT_CB(polygon_offset_enable);
    rs->cull_mode_index = 9;
    OUT_CB(cull_mode);
    OUT_CB_REG(R300_GA_LINE_STIPPLE_CONFIG, line_stipple_config);
    OUT_CB_REG(R300_GA_LINE_STIPPLE_VALUE, line_stipple_value);
    OUT_CB_REG(R300_GA_POLY_MODE, polygon_mode);
    OUT_CB_REG(R300_SC_CLIP_RULE, clip_rule);
    OUT_CB_REG(R300_GB_ENABLE, stuffing_enable);
    OUT_CB_REG_SEQ(R300_GA_POINT_S0, 4);
    OUT_CB_32F(point_texcoord_left);
    OUT_CB_32F(point_texcoord_bottom);
    OUT_CB_32F(point_texcoord_right);
    OUT_CB_32F(point_texcoord_top);
    END_CB;

    /* Build the two command buffers for polygon offset setup. */
    if (polygon_offset_enable) {
        float scale = state->offset_scale * 12;
        float offset = state->offset_units * 4;

        BEGIN_CB(rs->cb_poly_offset_zb16, 5);
        OUT_CB_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
        OUT_CB_32F(scale);
        OUT_CB_32F(offset);
        OUT_CB_32F(scale);
        OUT_CB_32F(offset);
        END_CB;

        offset = state->offset_units * 2;

        BEGIN_CB(rs->cb_poly_offset_zb24, 5);
        OUT_CB_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE, 4);
        OUT_CB_32F(scale);
        OUT_CB_32F(offset);
        OUT_CB_32F(scale);
        OUT_CB_32F(offset);
        END_CB;
    }

    return (void*)rs;
}

/* Bind rasterizer state. */
static void r300_bind_rs_state(struct pipe_context* pipe, void* state)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_rs_state* rs = (struct r300_rs_state*)state;
    int last_sprite_coord_enable = r300->sprite_coord_enable;
    boolean last_two_sided_color = r300->two_sided_color;

    if (r300->draw && rs) {
        draw_set_rasterizer_state(r300->draw, &rs->rs_draw, state);
    }

    if (rs) {
        r300->polygon_offset_enabled = rs->polygon_offset_enable;
        r300->sprite_coord_enable = rs->rs.sprite_coord_enable;
        r300->two_sided_color = rs->rs.light_twoside;
    } else {
        r300->polygon_offset_enabled = FALSE;
        r300->sprite_coord_enable = 0;
        r300->two_sided_color = FALSE;
    }

    UPDATE_STATE(state, r300->rs_state);
    r300->rs_state.size = 25 + (r300->polygon_offset_enabled ? 5 : 0);

    if (last_sprite_coord_enable != r300->sprite_coord_enable ||
        last_two_sided_color != r300->two_sided_color) {
        r300->rs_block_state.dirty = TRUE;
    }
}

/* Free rasterizer state. */
static void r300_delete_rs_state(struct pipe_context* pipe, void* state)
{
    FREE(state);
}

static void*
        r300_create_sampler_state(struct pipe_context* pipe,
                                  const struct pipe_sampler_state* state)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_sampler_state* sampler = CALLOC_STRUCT(r300_sampler_state);
    boolean is_r500 = r300->screen->caps.is_r500;
    int lod_bias;
    union util_color uc;

    sampler->state = *state;

    /* r300 doesn't handle CLAMP and MIRROR_CLAMP correctly when either MAG
     * or MIN filter is NEAREST. Since texwrap produces same results
     * for CLAMP and CLAMP_TO_EDGE, we use them instead. */
    if (sampler->state.min_img_filter == PIPE_TEX_FILTER_NEAREST ||
        sampler->state.mag_img_filter == PIPE_TEX_FILTER_NEAREST) {
        /* Wrap S. */
        if (sampler->state.wrap_s == PIPE_TEX_WRAP_CLAMP)
            sampler->state.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
        else if (sampler->state.wrap_s == PIPE_TEX_WRAP_MIRROR_CLAMP)
            sampler->state.wrap_s = PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE;

        /* Wrap T. */
        if (sampler->state.wrap_t == PIPE_TEX_WRAP_CLAMP)
            sampler->state.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
        else if (sampler->state.wrap_t == PIPE_TEX_WRAP_MIRROR_CLAMP)
            sampler->state.wrap_t = PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE;

        /* Wrap R. */
        if (sampler->state.wrap_r == PIPE_TEX_WRAP_CLAMP)
            sampler->state.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
        else if (sampler->state.wrap_r == PIPE_TEX_WRAP_MIRROR_CLAMP)
            sampler->state.wrap_r = PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE;
    }

    sampler->filter0 |=
        (r300_translate_wrap(sampler->state.wrap_s) << R300_TX_WRAP_S_SHIFT) |
        (r300_translate_wrap(sampler->state.wrap_t) << R300_TX_WRAP_T_SHIFT) |
        (r300_translate_wrap(sampler->state.wrap_r) << R300_TX_WRAP_R_SHIFT);

    sampler->filter0 |= r300_translate_tex_filters(state->min_img_filter,
                                                   state->mag_img_filter,
                                                   state->min_mip_filter,
                                                   state->max_anisotropy > 0);

    sampler->filter0 |= r300_anisotropy(state->max_anisotropy);

    /* Unfortunately, r300-r500 don't support floating-point mipmap lods. */
    /* We must pass these to the merge function to clamp them properly. */
    sampler->min_lod = MAX2((unsigned)state->min_lod, 0);
    sampler->max_lod = MAX2((unsigned)ceilf(state->max_lod), 0);

    lod_bias = CLAMP((int)(state->lod_bias * 32 + 1), -(1 << 9), (1 << 9) - 1);

    sampler->filter1 |= (lod_bias << R300_LOD_BIAS_SHIFT) & R300_LOD_BIAS_MASK;

    /* This is very high quality anisotropic filtering for R5xx.
     * It's good for benchmarking the performance of texturing but
     * in practice we don't want to slow down the driver because it's
     * a pretty good performance killer. Feel free to play with it. */
    if (DBG_ON(r300, DBG_ANISOHQ) && is_r500) {
        sampler->filter1 |= r500_anisotropy(state->max_anisotropy);
    }

    util_pack_color(state->border_color, PIPE_FORMAT_B8G8R8A8_UNORM, &uc);
    sampler->border_color = uc.ui;

    /* R500-specific fixups and optimizations */
    if (r300->screen->caps.is_r500) {
        sampler->filter1 |= R500_BORDER_FIX;
    }

    return (void*)sampler;
}

static void r300_bind_sampler_states(struct pipe_context* pipe,
                                     unsigned count,
                                     void** states)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_textures_state* state =
        (struct r300_textures_state*)r300->textures_state.state;
    unsigned tex_units = r300->screen->caps.num_tex_units;

    if (count > tex_units) {
        return;
    }

    memcpy(state->sampler_states, states, sizeof(void*) * count);
    state->sampler_state_count = count;

    r300->textures_state.dirty = TRUE;
}

static void r300_lacks_vertex_textures(struct pipe_context* pipe,
                                       unsigned count,
                                       void** states)
{
}

static void r300_delete_sampler_state(struct pipe_context* pipe, void* state)
{
    FREE(state);
}

static uint32_t r300_assign_texture_cache_region(unsigned index, unsigned num)
{
    /* This looks like a hack, but I believe it's suppose to work like
     * that. To illustrate how this works, let's assume you have 5 textures.
     * From docs, 5 and the successive numbers are:
     *
     * FOURTH_1     = 5
     * FOURTH_2     = 6
     * FOURTH_3     = 7
     * EIGHTH_0     = 8
     * EIGHTH_1     = 9
     *
     * First 3 textures will get 3/4 of size of the cache, divived evenly
     * between them. The last 1/4 of the cache must be divided between
     * the last 2 textures, each will therefore get 1/8 of the cache.
     * Why not just to use "5 + texture_index" ?
     *
     * This simple trick works for all "num" <= 16.
     */
    if (num <= 1)
        return R300_TX_CACHE(R300_TX_CACHE_WHOLE);
    else
        return R300_TX_CACHE(num + index);
}

static void r300_set_fragment_sampler_views(struct pipe_context* pipe,
                                            unsigned count,
                                            struct pipe_sampler_view** views)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_textures_state* state =
        (struct r300_textures_state*)r300->textures_state.state;
    struct r300_texture *texture;
    unsigned i, real_num_views = 0, view_index = 0;
    unsigned tex_units = r300->screen->caps.num_tex_units;
    boolean dirty_tex = FALSE;

    if (count > tex_units) {
        return;
    }

    /* Calculate the real number of views. */
    for (i = 0; i < count; i++) {
        if (views[i])
            real_num_views++;
    }

    for (i = 0; i < count; i++) {
        if (&state->sampler_views[i]->base != views[i]) {
            pipe_sampler_view_reference(
                    (struct pipe_sampler_view**)&state->sampler_views[i],
                    views[i]);

            if (!views[i]) {
                continue;
            }

            /* A new sampler view (= texture)... */
            dirty_tex = TRUE;

            /* Set the texrect factor in the fragment shader.
             * Needed for RECT and NPOT fallback. */
            texture = r300_texture(views[i]->texture);
            if (texture->desc.is_npot) {
                r300->fs_rc_constant_state.dirty = TRUE;
            }

            state->sampler_views[i]->texcache_region =
                r300_assign_texture_cache_region(view_index, real_num_views);
            view_index++;
        }
    }

    for (i = count; i < tex_units; i++) {
        if (state->sampler_views[i]) {
            pipe_sampler_view_reference(
                    (struct pipe_sampler_view**)&state->sampler_views[i],
                    NULL);
        }
    }

    state->sampler_view_count = count;

    r300->textures_state.dirty = TRUE;

    if (dirty_tex) {
        r300->texture_cache_inval.dirty = TRUE;
    }
}

static struct pipe_sampler_view *
r300_create_sampler_view(struct pipe_context *pipe,
                         struct pipe_resource *texture,
                         const struct pipe_sampler_view *templ)
{
    struct r300_sampler_view *view = CALLOC_STRUCT(r300_sampler_view);
    struct r300_texture *tex = r300_texture(texture);
    boolean is_r500 = r300_screen(pipe->screen)->caps.is_r500;

    if (view) {
        view->base = *templ;
        view->base.reference.count = 1;
        view->base.context = pipe;
        view->base.texture = NULL;
        pipe_resource_reference(&view->base.texture, texture);

        view->swizzle[0] = templ->swizzle_r;
        view->swizzle[1] = templ->swizzle_g;
        view->swizzle[2] = templ->swizzle_b;
        view->swizzle[3] = templ->swizzle_a;

        view->format = tex->tx_format;
        view->format.format1 |= r300_translate_texformat(templ->format,
                                                         view->swizzle,
                                                         is_r500);
        if (is_r500) {
            view->format.format2 |= r500_tx_format_msb_bit(templ->format);
        }
    }

    return (struct pipe_sampler_view*)view;
}

static void
r300_sampler_view_destroy(struct pipe_context *pipe,
                          struct pipe_sampler_view *view)
{
   pipe_resource_reference(&view->texture, NULL);
   FREE(view);
}

static void r300_set_scissor_state(struct pipe_context* pipe,
                                   const struct pipe_scissor_state* state)
{
    struct r300_context* r300 = r300_context(pipe);

    memcpy(r300->scissor_state.state, state,
        sizeof(struct pipe_scissor_state));

    r300->scissor_state.dirty = TRUE;
}

static void r300_set_viewport_state(struct pipe_context* pipe,
                                    const struct pipe_viewport_state* state)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_viewport_state* viewport =
        (struct r300_viewport_state*)r300->viewport_state.state;

    r300->viewport = *state;

    if (r300->draw) {
        draw_set_viewport_state(r300->draw, state);
        viewport->vte_control = R300_VTX_XY_FMT | R300_VTX_Z_FMT;
        return;
    }

    /* Do the transform in HW. */
    viewport->vte_control = R300_VTX_W0_FMT;

    if (state->scale[0] != 1.0f) {
        viewport->xscale = state->scale[0];
        viewport->vte_control |= R300_VPORT_X_SCALE_ENA;
    }
    if (state->scale[1] != 1.0f) {
        viewport->yscale = state->scale[1];
        viewport->vte_control |= R300_VPORT_Y_SCALE_ENA;
    }
    if (state->scale[2] != 1.0f) {
        viewport->zscale = state->scale[2];
        viewport->vte_control |= R300_VPORT_Z_SCALE_ENA;
    }
    if (state->translate[0] != 0.0f) {
        viewport->xoffset = state->translate[0];
        viewport->vte_control |= R300_VPORT_X_OFFSET_ENA;
    }
    if (state->translate[1] != 0.0f) {
        viewport->yoffset = state->translate[1];
        viewport->vte_control |= R300_VPORT_Y_OFFSET_ENA;
    }
    if (state->translate[2] != 0.0f) {
        viewport->zoffset = state->translate[2];
        viewport->vte_control |= R300_VPORT_Z_OFFSET_ENA;
    }

    r300->viewport_state.dirty = TRUE;
    if (r300->fs.state && r300_fs(r300)->shader->inputs.wpos != ATTR_UNUSED) {
        r300->fs_rc_constant_state.dirty = TRUE;
    }
}

static void r300_set_vertex_buffers(struct pipe_context* pipe,
                                    unsigned count,
                                    const struct pipe_vertex_buffer* buffers)
{
    struct r300_context* r300 = r300_context(pipe);
    struct pipe_vertex_buffer *vbo;
    unsigned i, max_index = (1 << 24) - 1;
    boolean any_user_buffer = FALSE;

    if (count == r300->vertex_buffer_count &&
        memcmp(r300->vertex_buffer, buffers,
            sizeof(struct pipe_vertex_buffer) * count) == 0) {
        return;
    }

    if (r300->screen->caps.has_tcl) {
        /* HW TCL. */
        r300->incompatible_vb_layout = FALSE;

        /* Check if the strides and offsets are aligned to the size of DWORD. */
        for (i = 0; i < count; i++) {
            if (buffers[i].buffer) {
                if (buffers[i].stride % 4 != 0 ||
                    buffers[i].buffer_offset % 4 != 0) {
                    r300->incompatible_vb_layout = TRUE;
                    break;
                }
            }
        }

        for (i = 0; i < count; i++) {
            /* Why, yes, I AM casting away constness. How did you know? */
            vbo = (struct pipe_vertex_buffer*)&buffers[i];

            /* Skip NULL buffers */
            if (!buffers[i].buffer) {
                continue;
            }

            if (r300_buffer_is_user_buffer(vbo->buffer)) {
                any_user_buffer = TRUE;
            }

            if (vbo->max_index == ~0) {
                /* if no VBO stride then only one vertex value so max index is 1 */
                /* should think about converting to VS constants like svga does */
                if (!vbo->stride)
                    vbo->max_index = 1;
                else
                    vbo->max_index =
                             (vbo->buffer->width0 - vbo->buffer_offset) / vbo->stride;
            }

            max_index = MIN2(vbo->max_index, max_index);
        }

        r300->any_user_vbs = any_user_buffer;
        r300->vertex_buffer_max_index = max_index;

    } else {
        /* SW TCL. */
        draw_set_vertex_buffers(r300->draw, count, buffers);
    }

    /* Common code. */
    for (i = 0; i < count; i++) {
        /* Reference our buffer. */
        pipe_resource_reference(&r300->vertex_buffer[i].buffer, buffers[i].buffer);
    }
    for (; i < r300->vertex_buffer_count; i++) {
        /* Dereference any old buffers. */
        pipe_resource_reference(&r300->vertex_buffer[i].buffer, NULL);
    }

    memcpy(r300->vertex_buffer, buffers,
        sizeof(struct pipe_vertex_buffer) * count);
    r300->vertex_buffer_count = count;
}

static void r300_set_index_buffer(struct pipe_context* pipe,
                                  const struct pipe_index_buffer *ib)
{
    struct r300_context* r300 = r300_context(pipe);

    if (ib) {
        pipe_resource_reference(&r300->index_buffer.buffer, ib->buffer);
        memcpy(&r300->index_buffer, ib, sizeof(r300->index_buffer));
    }
    else {
        pipe_resource_reference(&r300->index_buffer.buffer, NULL);
        memset(&r300->index_buffer, 0, sizeof(r300->index_buffer));
    }

    /* TODO make this more like a state */
}

/* Initialize the PSC tables. */
static void r300_vertex_psc(struct r300_vertex_element_state *velems)
{
    struct r300_vertex_stream_state *vstream = &velems->vertex_stream;
    uint16_t type, swizzle;
    enum pipe_format format;
    unsigned i;

    if (velems->count > 16) {
        fprintf(stderr, "r300: More than 16 vertex elements are not supported,"
                " requested %i, using 16.\n", velems->count);
        velems->count = 16;
    }

    /* Vertex shaders have no semantics on their inputs,
     * so PSC should just route stuff based on the vertex elements,
     * and not on attrib information. */
    for (i = 0; i < velems->count; i++) {
        format = velems->hw_format[i];

        type = r300_translate_vertex_data_type(format);
        if (type == R300_INVALID_FORMAT) {
            fprintf(stderr, "r300: Bad vertex format %s.\n",
                    util_format_short_name(format));
            assert(0);
            abort();
        }

        type |= i << R300_DST_VEC_LOC_SHIFT;
        swizzle = r300_translate_vertex_data_swizzle(format);

        if (i & 1) {
            vstream->vap_prog_stream_cntl[i >> 1] |= type << 16;
            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle << 16;
        } else {
            vstream->vap_prog_stream_cntl[i >> 1] |= type;
            vstream->vap_prog_stream_cntl_ext[i >> 1] |= swizzle;
        }
    }

    /* Set the last vector in the PSC. */
    if (i) {
        i -= 1;
    }
    vstream->vap_prog_stream_cntl[i >> 1] |=
        (R300_LAST_VEC << (i & 1 ? 16 : 0));

    vstream->count = (i >> 1) + 1;
}

#define FORMAT_REPLACE(what, withwhat) \
    case PIPE_FORMAT_##what: *format = PIPE_FORMAT_##withwhat; break

static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
                                               unsigned count,
                                               const struct pipe_vertex_element* attribs)
{
    struct r300_vertex_element_state *velems;
    unsigned i;
    enum pipe_format *format;

    assert(count <= PIPE_MAX_ATTRIBS);
    velems = CALLOC_STRUCT(r300_vertex_element_state);
    if (velems != NULL) {
        velems->count = count;
        memcpy(velems->velem, attribs, sizeof(struct pipe_vertex_element) * count);

        if (r300_screen(pipe->screen)->caps.has_tcl) {
            /* Set the best hw format in case the original format is not
             * supported by hw. */
            for (i = 0; i < count; i++) {
                velems->hw_format[i] = velems->velem[i].src_format;
                format = &velems->hw_format[i];

                /* This is basically the list of unsupported formats.
                 * For now we don't care about the alignment, that's going to
                 * be sorted out after the PSC setup. */
                switch (*format) {
                    FORMAT_REPLACE(R64_FLOAT,           R32_FLOAT);
                    FORMAT_REPLACE(R64G64_FLOAT,        R32G32_FLOAT);
                    FORMAT_REPLACE(R64G64B64_FLOAT,     R32G32B32_FLOAT);
                    FORMAT_REPLACE(R64G64B64A64_FLOAT,  R32G32B32A32_FLOAT);

                    FORMAT_REPLACE(R32_UNORM,           R32_FLOAT);
                    FORMAT_REPLACE(R32G32_UNORM,        R32G32_FLOAT);
                    FORMAT_REPLACE(R32G32B32_UNORM,     R32G32B32_FLOAT);
                    FORMAT_REPLACE(R32G32B32A32_UNORM,  R32G32B32A32_FLOAT);

                    FORMAT_REPLACE(R32_USCALED,         R32_FLOAT);
                    FORMAT_REPLACE(R32G32_USCALED,      R32G32_FLOAT);
                    FORMAT_REPLACE(R32G32B32_USCALED,   R32G32B32_FLOAT);
                    FORMAT_REPLACE(R32G32B32A32_USCALED,R32G32B32A32_FLOAT);

                    FORMAT_REPLACE(R32_SNORM,           R32_FLOAT);
                    FORMAT_REPLACE(R32G32_SNORM,        R32G32_FLOAT);
                    FORMAT_REPLACE(R32G32B32_SNORM,     R32G32B32_FLOAT);
                    FORMAT_REPLACE(R32G32B32A32_SNORM,  R32G32B32A32_FLOAT);

                    FORMAT_REPLACE(R32_SSCALED,         R32_FLOAT);
                    FORMAT_REPLACE(R32G32_SSCALED,      R32G32_FLOAT);
                    FORMAT_REPLACE(R32G32B32_SSCALED,   R32G32B32_FLOAT);
                    FORMAT_REPLACE(R32G32B32A32_SSCALED,R32G32B32A32_FLOAT);

                    FORMAT_REPLACE(R32_FIXED,           R32_FLOAT);
                    FORMAT_REPLACE(R32G32_FIXED,        R32G32_FLOAT);
                    FORMAT_REPLACE(R32G32B32_FIXED,     R32G32B32_FLOAT);
                    FORMAT_REPLACE(R32G32B32A32_FIXED,  R32G32B32A32_FLOAT);

                    default:;
                }

                velems->incompatible_layout =
                        velems->incompatible_layout ||
                        velems->velem[i].src_format != velems->hw_format[i] ||
                        velems->velem[i].src_offset % 4 != 0;
            }

            /* Now setup PSC.
             * The unused components will be replaced by (..., 0, 1). */
            r300_vertex_psc(velems);

            /* Align the formats to the size of DWORD.
             * We only care about the blocksizes of the formats since
             * swizzles are already set up.
             * Also compute the vertex size. */
            for (i = 0; i < count; i++) {
                /* This is OK because we check for aligned strides too. */
                velems->hw_format_size[i] =
                    align(util_format_get_blocksize(velems->hw_format[i]), 4);
                velems->vertex_size_dwords += velems->hw_format_size[i] / 4;
            }
        }
    }
    return velems;
}

static void r300_bind_vertex_elements_state(struct pipe_context *pipe,
                                            void *state)
{
    struct r300_context *r300 = r300_context(pipe);
    struct r300_vertex_element_state *velems = state;

    if (velems == NULL) {
        return;
    }

    r300->velems = velems;

    if (r300->draw) {
        draw_set_vertex_elements(r300->draw, velems->count, velems->velem);
        return;
    }

    UPDATE_STATE(&velems->vertex_stream, r300->vertex_stream_state);
    r300->vertex_stream_state.size = (1 + velems->vertex_stream.count) * 2;
}

static void r300_delete_vertex_elements_state(struct pipe_context *pipe, void *state)
{
   FREE(state);
}

static void* r300_create_vs_state(struct pipe_context* pipe,
                                  const struct pipe_shader_state* shader)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_vertex_shader* vs = CALLOC_STRUCT(r300_vertex_shader);

    /* Copy state directly into shader. */
    vs->state = *shader;
    vs->state.tokens = tgsi_dup_tokens(shader->tokens);

    if (r300->screen->caps.has_tcl) {
        r300_init_vs_outputs(vs);
        r300_translate_vertex_shader(r300, vs);
    } else {
        r300_draw_init_vertex_shader(r300->draw, vs);
    }

    return vs;
}

static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;

    if (vs == NULL) {
        r300->vs_state.state = NULL;
        return;
    }
    if (vs == r300->vs_state.state) {
        return;
    }
    r300->vs_state.state = vs;

    /* The majority of the RS block bits is dependent on the vertex shader. */
    r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */

    if (r300->screen->caps.has_tcl) {
        unsigned fc_op_dwords = r300->screen->caps.is_r500 ? 3 : 2;
        r300->vs_state.dirty = TRUE;
        r300->vs_state.size =
                vs->code.length + 9 +
                (vs->immediates_count ? vs->immediates_count * 4 + 3 : 0) +
        (vs->code.num_fc_ops ? vs->code.num_fc_ops * fc_op_dwords + 4 : 0);

        if (vs->externals_count) {
            r300->vs_constants.dirty = TRUE;
            r300->vs_constants.size = vs->externals_count * 4 + 3;
        } else {
            r300->vs_constants.size = 0;
        }

        r300->pvs_flush.dirty = TRUE;
    } else {
        draw_bind_vertex_shader(r300->draw,
                (struct draw_vertex_shader*)vs->draw_vs);
    }
}

static void r300_delete_vs_state(struct pipe_context* pipe, void* shader)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_vertex_shader* vs = (struct r300_vertex_shader*)shader;

    if (r300->screen->caps.has_tcl) {
        rc_constants_destroy(&vs->code.constants);
    } else {
        draw_delete_vertex_shader(r300->draw,
                (struct draw_vertex_shader*)vs->draw_vs);
    }

    FREE((void*)vs->state.tokens);
    FREE(shader);
}

static void r300_set_constant_buffer(struct pipe_context *pipe,
                                     uint shader, uint index,
                                     struct pipe_resource *buf)
{
    struct r300_context* r300 = r300_context(pipe);
    struct r300_constant_buffer *cbuf;
    uint32_t *mapped = r300_buffer(buf)->user_buffer;
    int max_size = 0, max_size_bytes = 0, clamped_size = 0;

    switch (shader) {
        case PIPE_SHADER_VERTEX:
            cbuf = (struct r300_constant_buffer*)r300->vs_constants.state;
            max_size = 256;
            break;
        case PIPE_SHADER_FRAGMENT:
            cbuf = (struct r300_constant_buffer*)r300->fs_constants.state;
            if (r300->screen->caps.is_r500) {
                max_size = 256;
            } else {
                max_size = 32;
            }
            break;
        default:
            assert(0);
            return;
    }
    max_size_bytes = max_size * 4 * sizeof(float);

    if (buf == NULL || buf->width0 == 0 ||
        (mapped = r300_buffer(buf)->constant_buffer) == NULL) {
        cbuf->count = 0;
        return;
    }

    if (shader == PIPE_SHADER_FRAGMENT ||
        (shader == PIPE_SHADER_VERTEX && r300->screen->caps.has_tcl)) {
        assert((buf->width0 % (4 * sizeof(float))) == 0);

        /* Check the size of the constant buffer. */
        /* XXX Subtract immediates and RC_STATE_* variables. */
        if (buf->width0 > max_size_bytes) {
            fprintf(stderr, "r300: Max size of the constant buffer is "
                          "%i*4 floats.\n", max_size);
        }

        clamped_size = MIN2(buf->width0, max_size_bytes);
        cbuf->count = clamped_size / (4 * sizeof(float));
        cbuf->ptr = mapped;
    }

    if (shader == PIPE_SHADER_VERTEX) {
        if (r300->screen->caps.has_tcl) {
            if (r300->vs_constants.size) {
                r300->vs_constants.dirty = TRUE;
            }
            r300->pvs_flush.dirty = TRUE;
        } else if (r300->draw) {
            draw_set_mapped_constant_buffer(r300->draw, PIPE_SHADER_VERTEX,
                0, mapped, buf->width0);
        }
    } else if (shader == PIPE_SHADER_FRAGMENT) {
        r300->fs_constants.dirty = TRUE;
    }
}

void r300_init_state_functions(struct r300_context* r300)
{
    r300->context.create_blend_state = r300_create_blend_state;
    r300->context.bind_blend_state = r300_bind_blend_state;
    r300->context.delete_blend_state = r300_delete_blend_state;

    r300->context.set_blend_color = r300_set_blend_color;

    r300->context.set_clip_state = r300_set_clip_state;
    r300->context.set_sample_mask = r300_set_sample_mask;

    r300->context.set_constant_buffer = r300_set_constant_buffer;

    r300->context.create_depth_stencil_alpha_state = r300_create_dsa_state;
    r300->context.bind_depth_stencil_alpha_state = r300_bind_dsa_state;
    r300->context.delete_depth_stencil_alpha_state = r300_delete_dsa_state;

    r300->context.set_stencil_ref = r300_set_stencil_ref;

    r300->context.set_framebuffer_state = r300_set_framebuffer_state;

    r300->context.create_fs_state = r300_create_fs_state;
    r300->context.bind_fs_state = r300_bind_fs_state;
    r300->context.delete_fs_state = r300_delete_fs_state;

    r300->context.set_polygon_stipple = r300_set_polygon_stipple;

    r300->context.create_rasterizer_state = r300_create_rs_state;
    r300->context.bind_rasterizer_state = r300_bind_rs_state;
    r300->context.delete_rasterizer_state = r300_delete_rs_state;

    r300->context.create_sampler_state = r300_create_sampler_state;
    r300->context.bind_fragment_sampler_states = r300_bind_sampler_states;
    r300->context.bind_vertex_sampler_states = r300_lacks_vertex_textures;
    r300->context.delete_sampler_state = r300_delete_sampler_state;

    r300->context.set_fragment_sampler_views = r300_set_fragment_sampler_views;
    r300->context.create_sampler_view = r300_create_sampler_view;
    r300->context.sampler_view_destroy = r300_sampler_view_destroy;

    r300->context.set_scissor_state = r300_set_scissor_state;

    r300->context.set_viewport_state = r300_set_viewport_state;

    r300->context.set_vertex_buffers = r300_set_vertex_buffers;
    r300->context.set_index_buffer = r300_set_index_buffer;

    r300->context.create_vertex_elements_state = r300_create_vertex_elements_state;
    r300->context.bind_vertex_elements_state = r300_bind_vertex_elements_state;
    r300->context.delete_vertex_elements_state = r300_delete_vertex_elements_state;

    r300->context.create_vs_state = r300_create_vs_state;
    r300->context.bind_vs_state = r300_bind_vs_state;
    r300->context.delete_vs_state = r300_delete_vs_state;
}