summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/softpipe/sp_setup.c
blob: ff3cb9fe5e1fc9d9a5798b0749397b98894a34a2 (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
/**************************************************************************
 *
 * Copyright 2007 VMware, Inc.
 * All Rights Reserved.
 *
 * 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, 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 VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 **************************************************************************/

/**
 * \brief  Primitive rasterization/rendering (points, lines, triangles)
 *
 * \author  Keith Whitwell <keithw@vmware.com>
 * \author  Brian Paul
 */

#include "sp_context.h"
#include "sp_quad.h"
#include "sp_quad_pipe.h"
#include "sp_setup.h"
#include "sp_state.h"
#include "draw/draw_context.h"
#include "draw/draw_vertex.h"
#include "pipe/p_shader_tokens.h"
#include "util/u_math.h"
#include "util/u_memory.h"


#define DEBUG_VERTS 0
#define DEBUG_FRAGS 0


/**
 * Triangle edge info
 */
struct edge {
   float dx;		/**< X(v1) - X(v0), used only during setup */
   float dy;		/**< Y(v1) - Y(v0), used only during setup */
   float dxdy;		/**< dx/dy */
   float sx, sy;	/**< first sample point coord */
   int lines;		/**< number of lines on this edge */
};


/**
 * Max number of quads (2x2 pixel blocks) to process per batch.
 * This can't be arbitrarily increased since we depend on some 32-bit
 * bitmasks (two bits per quad).
 */
#define MAX_QUADS 16


/**
 * Triangle setup info.
 * Also used for line drawing (taking some liberties).
 */
struct setup_context {
   struct softpipe_context *softpipe;

   /* Vertices are just an array of floats making up each attribute in
    * turn.  Currently fixed at 4 floats, but should change in time.
    * Codegen will help cope with this.
    */
   const float (*vmax)[4];
   const float (*vmid)[4];
   const float (*vmin)[4];
   const float (*vprovoke)[4];

   struct edge ebot;
   struct edge etop;
   struct edge emaj;

   float oneoverarea;
   int facing;

   float pixel_offset;
   unsigned max_layer;

   struct quad_header quad[MAX_QUADS];
   struct quad_header *quad_ptrs[MAX_QUADS];
   unsigned count;

   struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS];
   struct tgsi_interp_coef posCoef;  /* For Z, W */

   struct {
      int left[2];   /**< [0] = row0, [1] = row1 */
      int right[2];
      int y;
   } span;

#if DEBUG_FRAGS
   uint numFragsEmitted;  /**< per primitive */
   uint numFragsWritten;  /**< per primitive */
#endif

   unsigned cull_face;		/* which faces cull */
   unsigned nr_vertex_attrs;
};







/**
 * Clip setup->quad against the scissor/surface bounds.
 */
static inline void
quad_clip(struct setup_context *setup, struct quad_header *quad)
{
   const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect;
   const int minx = (int) cliprect->minx;
   const int maxx = (int) cliprect->maxx;
   const int miny = (int) cliprect->miny;
   const int maxy = (int) cliprect->maxy;

   if (quad->input.x0 >= maxx ||
       quad->input.y0 >= maxy ||
       quad->input.x0 + 1 < minx ||
       quad->input.y0 + 1 < miny) {
      /* totally clipped */
      quad->inout.mask = 0x0;
      return;
   }
   if (quad->input.x0 < minx)
      quad->inout.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
   if (quad->input.y0 < miny)
      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
   if (quad->input.x0 == maxx - 1)
      quad->inout.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
   if (quad->input.y0 == maxy - 1)
      quad->inout.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
}


/**
 * Emit a quad (pass to next stage) with clipping.
 */
static inline void
clip_emit_quad(struct setup_context *setup, struct quad_header *quad)
{
   quad_clip( setup, quad );

   if (quad->inout.mask) {
      struct softpipe_context *sp = setup->softpipe;

#if DEBUG_FRAGS
      setup->numFragsEmitted += util_bitcount(quad->inout.mask);
#endif

      sp->quad.first->run( sp->quad.first, &quad, 1 );
   }
}



/**
 * Given an X or Y coordinate, return the block/quad coordinate that it
 * belongs to.
 */
static inline int
block(int x)
{
   return x & ~(2-1);
}


static inline int
block_x(int x)
{
   return x & ~(16-1);
}


/**
 * Render a horizontal span of quads
 */
static void
flush_spans(struct setup_context *setup)
{
   const int step = MAX_QUADS;
   const int xleft0 = setup->span.left[0];
   const int xleft1 = setup->span.left[1];
   const int xright0 = setup->span.right[0];
   const int xright1 = setup->span.right[1];
   struct quad_stage *pipe = setup->softpipe->quad.first;

   const int minleft = block_x(MIN2(xleft0, xleft1));
   const int maxright = MAX2(xright0, xright1);
   int x;

   /* process quads in horizontal chunks of 16 */
   for (x = minleft; x < maxright; x += step) {
      unsigned skip_left0 = CLAMP(xleft0 - x, 0, step);
      unsigned skip_left1 = CLAMP(xleft1 - x, 0, step);
      unsigned skip_right0 = CLAMP(x + step - xright0, 0, step);
      unsigned skip_right1 = CLAMP(x + step - xright1, 0, step);
      unsigned lx = x;
      unsigned q = 0;

      unsigned skipmask_left0 = (1U << skip_left0) - 1U;
      unsigned skipmask_left1 = (1U << skip_left1) - 1U;

      /* These calculations fail when step == 32 and skip_right == 0.
       */
      unsigned skipmask_right0 = ~0U << (unsigned)(step - skip_right0);
      unsigned skipmask_right1 = ~0U << (unsigned)(step - skip_right1);

      unsigned mask0 = ~skipmask_left0 & ~skipmask_right0;
      unsigned mask1 = ~skipmask_left1 & ~skipmask_right1;

      if (mask0 | mask1) {
         do {
            unsigned quadmask = (mask0 & 3) | ((mask1 & 3) << 2);
            if (quadmask) {
               setup->quad[q].input.x0 = lx;
               setup->quad[q].input.y0 = setup->span.y;
               setup->quad[q].input.facing = setup->facing;
               setup->quad[q].inout.mask = quadmask;
               setup->quad_ptrs[q] = &setup->quad[q];
               q++;
#if DEBUG_FRAGS
               setup->numFragsEmitted += util_bitcount(quadmask);
#endif
            }
            mask0 >>= 2;
            mask1 >>= 2;
            lx += 2;
         } while (mask0 | mask1);

         pipe->run( pipe, setup->quad_ptrs, q );
      }
   }


   setup->span.y = 0;
   setup->span.right[0] = 0;
   setup->span.right[1] = 0;
   setup->span.left[0] = 1000000;     /* greater than right[0] */
   setup->span.left[1] = 1000000;     /* greater than right[1] */
}


#if DEBUG_VERTS
static void
print_vertex(const struct setup_context *setup,
             const float (*v)[4])
{
   int i;
   debug_printf("   Vertex: (%p)\n", (void *) v);
   for (i = 0; i < setup->nr_vertex_attrs; i++) {
      debug_printf("     %d: %f %f %f %f\n",  i,
              v[i][0], v[i][1], v[i][2], v[i][3]);
      if (util_is_inf_or_nan(v[i][0])) {
         debug_printf("   NaN!\n");
      }
   }
}
#endif


/**
 * Sort the vertices from top to bottom order, setting up the triangle
 * edge fields (ebot, emaj, etop).
 * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise
 */
static boolean
setup_sort_vertices(struct setup_context *setup,
                    float det,
                    const float (*v0)[4],
                    const float (*v1)[4],
                    const float (*v2)[4])
{
   if (setup->softpipe->rasterizer->flatshade_first)
      setup->vprovoke = v0;
   else
      setup->vprovoke = v2;

   /* determine bottom to top order of vertices */
   {
      float y0 = v0[0][1];
      float y1 = v1[0][1];
      float y2 = v2[0][1];
      if (y0 <= y1) {
	 if (y1 <= y2) {
	    /* y0<=y1<=y2 */
	    setup->vmin = v0;
	    setup->vmid = v1;
	    setup->vmax = v2;
	 }
	 else if (y2 <= y0) {
	    /* y2<=y0<=y1 */
	    setup->vmin = v2;
	    setup->vmid = v0;
	    setup->vmax = v1;
	 }
	 else {
	    /* y0<=y2<=y1 */
	    setup->vmin = v0;
	    setup->vmid = v2;
	    setup->vmax = v1;
	 }
      }
      else {
	 if (y0 <= y2) {
	    /* y1<=y0<=y2 */
	    setup->vmin = v1;
	    setup->vmid = v0;
	    setup->vmax = v2;
	 }
	 else if (y2 <= y1) {
	    /* y2<=y1<=y0 */
	    setup->vmin = v2;
	    setup->vmid = v1;
	    setup->vmax = v0;
	 }
	 else {
	    /* y1<=y2<=y0 */
	    setup->vmin = v1;
	    setup->vmid = v2;
	    setup->vmax = v0;
	 }
      }
   }

   setup->ebot.dx = setup->vmid[0][0] - setup->vmin[0][0];
   setup->ebot.dy = setup->vmid[0][1] - setup->vmin[0][1];
   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];
   setup->etop.dx = setup->vmax[0][0] - setup->vmid[0][0];
   setup->etop.dy = setup->vmax[0][1] - setup->vmid[0][1];

   /*
    * Compute triangle's area.  Use 1/area to compute partial
    * derivatives of attributes later.
    *
    * The area will be the same as prim->det, but the sign may be
    * different depending on how the vertices get sorted above.
    *
    * To determine whether the primitive is front or back facing we
    * use the prim->det value because its sign is correct.
    */
   {
      const float area = (setup->emaj.dx * setup->ebot.dy -
			    setup->ebot.dx * setup->emaj.dy);

      setup->oneoverarea = 1.0f / area;

      /*
      debug_printf("%s one-over-area %f  area %f  det %f\n",
                   __FUNCTION__, setup->oneoverarea, area, det );
      */
      if (util_is_inf_or_nan(setup->oneoverarea))
         return FALSE;
   }

   /* We need to know if this is a front or back-facing triangle for:
    *  - the GLSL gl_FrontFacing fragment attribute (bool)
    *  - two-sided stencil test
    * 0 = front-facing, 1 = back-facing
    */
   setup->facing = 
      ((det < 0.0) ^ 
       (setup->softpipe->rasterizer->front_ccw));

   {
      unsigned face = setup->facing == 0 ? PIPE_FACE_FRONT : PIPE_FACE_BACK;

      if (face & setup->cull_face)
	 return FALSE;
   }


   /* Prepare pixel offset for rasterisation:
    *  - pixel center (0.5, 0.5) for GL, or
    *  - assume (0.0, 0.0) for other APIs.
    */
   if (setup->softpipe->rasterizer->half_pixel_center) {
      setup->pixel_offset = 0.5f;
   } else {
      setup->pixel_offset = 0.0f;
   }

   return TRUE;
}


/* Apply cylindrical wrapping to v0, v1, v2 coordinates, if enabled.
 * Input coordinates must be in [0, 1] range, otherwise results are undefined.
 * Some combinations of coordinates produce invalid results,
 * but this behaviour is acceptable.
 */
static void
tri_apply_cylindrical_wrap(float v0,
                           float v1,
                           float v2,
                           uint cylindrical_wrap,
                           float output[3])
{
   if (cylindrical_wrap) {
      float delta;

      delta = v1 - v0;
      if (delta > 0.5f) {
         v0 += 1.0f;
      }
      else if (delta < -0.5f) {
         v1 += 1.0f;
      }

      delta = v2 - v1;
      if (delta > 0.5f) {
         v1 += 1.0f;
      }
      else if (delta < -0.5f) {
         v2 += 1.0f;
      }

      delta = v0 - v2;
      if (delta > 0.5f) {
         v2 += 1.0f;
      }
      else if (delta < -0.5f) {
         v0 += 1.0f;
      }
   }

   output[0] = v0;
   output[1] = v1;
   output[2] = v2;
}


/**
 * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
 * The value value comes from vertex[slot][i].
 * The result will be put into setup->coef[slot].a0[i].
 * \param slot  which attribute slot
 * \param i  which component of the slot (0..3)
 */
static void
const_coeff(struct setup_context *setup,
            struct tgsi_interp_coef *coef,
            uint vertSlot, uint i)
{
   assert(i <= 3);

   coef->dadx[i] = 0;
   coef->dady[i] = 0;

   /* need provoking vertex info!
    */
   coef->a0[i] = setup->vprovoke[vertSlot][i];
}


/**
 * Compute a0, dadx and dady for a linearly interpolated coefficient,
 * for a triangle.
 * v[0], v[1] and v[2] are vmin, vmid and vmax, respectively.
 */
static void
tri_linear_coeff(struct setup_context *setup,
                 struct tgsi_interp_coef *coef,
                 uint i,
                 const float v[3])
{
   float botda = v[1] - v[0];
   float majda = v[2] - v[0];
   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
   float dadx = a * setup->oneoverarea;
   float dady = b * setup->oneoverarea;

   assert(i <= 3);

   coef->dadx[i] = dadx;
   coef->dady[i] = dady;

   /* calculate a0 as the value which would be sampled for the
    * fragment at (0,0), taking into account that we want to sample at
    * pixel centers, in other words (pixel_offset, pixel_offset).
    *
    * this is neat but unfortunately not a good way to do things for
    * triangles with very large values of dadx or dady as it will
    * result in the subtraction and re-addition from a0 of a very
    * large number, which means we'll end up loosing a lot of the
    * fractional bits and precision from a0.  the way to fix this is
    * to define a0 as the sample at a pixel center somewhere near vmin
    * instead - i'll switch to this later.
    */
   coef->a0[i] = (v[0] -
                  (dadx * (setup->vmin[0][0] - setup->pixel_offset) +
                   dady * (setup->vmin[0][1] - setup->pixel_offset)));
}


/**
 * Compute a0, dadx and dady for a perspective-corrected interpolant,
 * for a triangle.
 * We basically multiply the vertex value by 1/w before computing
 * the plane coefficients (a0, dadx, dady).
 * Later, when we compute the value at a particular fragment position we'll
 * divide the interpolated value by the interpolated W at that fragment.
 * v[0], v[1] and v[2] are vmin, vmid and vmax, respectively.
 */
static void
tri_persp_coeff(struct setup_context *setup,
                struct tgsi_interp_coef *coef,
                uint i,
                const float v[3])
{
   /* premultiply by 1/w  (v[0][3] is always W):
    */
   float mina = v[0] * setup->vmin[0][3];
   float mida = v[1] * setup->vmid[0][3];
   float maxa = v[2] * setup->vmax[0][3];
   float botda = mida - mina;
   float majda = maxa - mina;
   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
   float dadx = a * setup->oneoverarea;
   float dady = b * setup->oneoverarea;

   assert(i <= 3);

   coef->dadx[i] = dadx;
   coef->dady[i] = dady;
   coef->a0[i] = (mina -
                  (dadx * (setup->vmin[0][0] - setup->pixel_offset) +
                   dady * (setup->vmin[0][1] - setup->pixel_offset)));
}


/**
 * Special coefficient setup for gl_FragCoord.
 * X and Y are trivial, though Y may have to be inverted for OpenGL.
 * Z and W are copied from posCoef which should have already been computed.
 * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask.
 */
static void
setup_fragcoord_coeff(struct setup_context *setup, uint slot)
{
   const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
   boolean origin_lower_left =
         fsInfo->properties[TGSI_PROPERTY_FS_COORD_ORIGIN];
   boolean pixel_center_integer =
         fsInfo->properties[TGSI_PROPERTY_FS_COORD_PIXEL_CENTER];

   /*X*/
   setup->coef[slot].a0[0] = pixel_center_integer ? 0.0f : 0.5f;
   setup->coef[slot].dadx[0] = 1.0f;
   setup->coef[slot].dady[0] = 0.0f;
   /*Y*/
   setup->coef[slot].a0[1] =
		   (origin_lower_left ? setup->softpipe->framebuffer.height-1 : 0)
		   + (pixel_center_integer ? 0.0f : 0.5f);
   setup->coef[slot].dadx[1] = 0.0f;
   setup->coef[slot].dady[1] = origin_lower_left ? -1.0f : 1.0f;
   /*Z*/
   setup->coef[slot].a0[2] = setup->posCoef.a0[2];
   setup->coef[slot].dadx[2] = setup->posCoef.dadx[2];
   setup->coef[slot].dady[2] = setup->posCoef.dady[2];
   /*W*/
   setup->coef[slot].a0[3] = setup->posCoef.a0[3];
   setup->coef[slot].dadx[3] = setup->posCoef.dadx[3];
   setup->coef[slot].dady[3] = setup->posCoef.dady[3];
}



/**
 * Compute the setup->coef[] array dadx, dady, a0 values.
 * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized.
 */
static void
setup_tri_coefficients(struct setup_context *setup)
{
   struct softpipe_context *softpipe = setup->softpipe;
   const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
   uint fragSlot;
   float v[3];

   /* z and w are done by linear interpolation:
    */
   v[0] = setup->vmin[0][2];
   v[1] = setup->vmid[0][2];
   v[2] = setup->vmax[0][2];
   tri_linear_coeff(setup, &setup->posCoef, 2, v);

   v[0] = setup->vmin[0][3];
   v[1] = setup->vmid[0][3];
   v[2] = setup->vmax[0][3];
   tri_linear_coeff(setup, &setup->posCoef, 3, v);

   /* setup interpolation for all the remaining attributes:
    */
   for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
      uint j;

      switch (vinfo->attrib[fragSlot].interp_mode) {
      case INTERP_CONSTANT:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++)
            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
         break;
      case INTERP_LINEAR:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
            tri_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
                                       setup->vmid[vertSlot][j],
                                       setup->vmax[vertSlot][j],
                                       fsInfo->input_cylindrical_wrap[fragSlot] & (1 << j),
                                       v);
            tri_linear_coeff(setup, &setup->coef[fragSlot], j, v);
         }
         break;
      case INTERP_PERSPECTIVE:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
            tri_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
                                       setup->vmid[vertSlot][j],
                                       setup->vmax[vertSlot][j],
                                       fsInfo->input_cylindrical_wrap[fragSlot] & (1 << j),
                                       v);
            tri_persp_coeff(setup, &setup->coef[fragSlot], j, v);
         }
         break;
      case INTERP_POS:
         setup_fragcoord_coeff(setup, fragSlot);
         break;
      default:
         assert(0);
      }

      if (fsInfo->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
         /* convert 0 to 1.0 and 1 to -1.0 */
         setup->coef[fragSlot].a0[0] = setup->facing * -2.0f + 1.0f;
         setup->coef[fragSlot].dadx[0] = 0.0;
         setup->coef[fragSlot].dady[0] = 0.0;
      }

      if (0) {
         for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
            debug_printf("attr[%d].%c: a0:%f dx:%f dy:%f\n",
                         fragSlot, "xyzw"[j],
                         setup->coef[fragSlot].a0[j],
                         setup->coef[fragSlot].dadx[j],
                         setup->coef[fragSlot].dady[j]);
         }
      }
   }
}


static void
setup_tri_edges(struct setup_context *setup)
{
   float vmin_x = setup->vmin[0][0] + setup->pixel_offset;
   float vmid_x = setup->vmid[0][0] + setup->pixel_offset;

   float vmin_y = setup->vmin[0][1] - setup->pixel_offset;
   float vmid_y = setup->vmid[0][1] - setup->pixel_offset;
   float vmax_y = setup->vmax[0][1] - setup->pixel_offset;

   setup->emaj.sy = ceilf(vmin_y);
   setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy);
   setup->emaj.dxdy = setup->emaj.dy ? setup->emaj.dx / setup->emaj.dy : .0f;
   setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy;

   setup->etop.sy = ceilf(vmid_y);
   setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy);
   setup->etop.dxdy = setup->etop.dy ? setup->etop.dx / setup->etop.dy : .0f;
   setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy;

   setup->ebot.sy = ceilf(vmin_y);
   setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy);
   setup->ebot.dxdy = setup->ebot.dy ? setup->ebot.dx / setup->ebot.dy : .0f;
   setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy;
}


/**
 * Render the upper or lower half of a triangle.
 * Scissoring/cliprect is applied here too.
 */
static void
subtriangle(struct setup_context *setup,
            struct edge *eleft,
            struct edge *eright,
            int lines)
{
   const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect;
   const int minx = (int) cliprect->minx;
   const int maxx = (int) cliprect->maxx;
   const int miny = (int) cliprect->miny;
   const int maxy = (int) cliprect->maxy;
   int y, start_y, finish_y;
   int sy = (int)eleft->sy;

   assert((int)eleft->sy == (int) eright->sy);
   assert(lines >= 0);

   /* clip top/bottom */
   start_y = sy;
   if (start_y < miny)
      start_y = miny;

   finish_y = sy + lines;
   if (finish_y > maxy)
      finish_y = maxy;

   start_y -= sy;
   finish_y -= sy;

   /*
   debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y);
   */

   for (y = start_y; y < finish_y; y++) {

      /* avoid accumulating adds as floats don't have the precision to
       * accurately iterate large triangle edges that way.  luckily we
       * can just multiply these days.
       *
       * this is all drowned out by the attribute interpolation anyway.
       */
      int left = (int)(eleft->sx + y * eleft->dxdy);
      int right = (int)(eright->sx + y * eright->dxdy);

      /* clip left/right */
      if (left < minx)
         left = minx;
      if (right > maxx)
         right = maxx;

      if (left < right) {
         int _y = sy + y;
         if (block(_y) != setup->span.y) {
            flush_spans(setup);
            setup->span.y = block(_y);
         }

         setup->span.left[_y&1] = left;
         setup->span.right[_y&1] = right;
      }
   }


   /* save the values so that emaj can be restarted:
    */
   eleft->sx += lines * eleft->dxdy;
   eright->sx += lines * eright->dxdy;
   eleft->sy += lines;
   eright->sy += lines;
}


/**
 * Recalculate prim's determinant.  This is needed as we don't have
 * get this information through the vbuf_render interface & we must
 * calculate it here.
 */
static float
calc_det(const float (*v0)[4],
         const float (*v1)[4],
         const float (*v2)[4])
{
   /* edge vectors e = v0 - v2, f = v1 - v2 */
   const float ex = v0[0][0] - v2[0][0];
   const float ey = v0[0][1] - v2[0][1];
   const float fx = v1[0][0] - v2[0][0];
   const float fy = v1[0][1] - v2[0][1];

   /* det = cross(e,f).z */
   return ex * fy - ey * fx;
}


/**
 * Do setup for triangle rasterization, then render the triangle.
 */
void
sp_setup_tri(struct setup_context *setup,
             const float (*v0)[4],
             const float (*v1)[4],
             const float (*v2)[4])
{
   float det;
   uint layer = 0;
#if DEBUG_VERTS
   debug_printf("Setup triangle:\n");
   print_vertex(setup, v0);
   print_vertex(setup, v1);
   print_vertex(setup, v2);
#endif

   if (setup->softpipe->no_rast || setup->softpipe->rasterizer->rasterizer_discard)
      return;
   
   det = calc_det(v0, v1, v2);
   /*
   debug_printf("%s\n", __FUNCTION__ );
   */

#if DEBUG_FRAGS
   setup->numFragsEmitted = 0;
   setup->numFragsWritten = 0;
#endif

   if (!setup_sort_vertices( setup, det, v0, v1, v2 ))
      return;

   setup_tri_coefficients( setup );
   setup_tri_edges( setup );

   assert(setup->softpipe->reduced_prim == PIPE_PRIM_TRIANGLES);

   setup->span.y = 0;
   setup->span.right[0] = 0;
   setup->span.right[1] = 0;
   /*   setup->span.z_mode = tri_z_mode( setup->ctx ); */
   if (setup->softpipe->layer_slot > 0) {
      layer = *(unsigned *)v1[setup->softpipe->layer_slot];
      layer = MIN2(layer, setup->max_layer);
   }
   setup->quad[0].input.layer = layer;

   /*   init_constant_attribs( setup ); */

   if (setup->oneoverarea < 0.0) {
      /* emaj on left:
       */
      subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines );
      subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines );
   }
   else {
      /* emaj on right:
       */
      subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines );
      subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines );
   }

   flush_spans( setup );

   if (setup->softpipe->active_statistics_queries) {
      setup->softpipe->pipeline_statistics.c_primitives++;
   }

#if DEBUG_FRAGS
   printf("Tri: %u frags emitted, %u written\n",
          setup->numFragsEmitted,
          setup->numFragsWritten);
#endif
}


/* Apply cylindrical wrapping to v0, v1 coordinates, if enabled.
 * Input coordinates must be in [0, 1] range, otherwise results are undefined.
 */
static void
line_apply_cylindrical_wrap(float v0,
                            float v1,
                            uint cylindrical_wrap,
                            float output[2])
{
   if (cylindrical_wrap) {
      float delta;

      delta = v1 - v0;
      if (delta > 0.5f) {
         v0 += 1.0f;
      }
      else if (delta < -0.5f) {
         v1 += 1.0f;
      }
   }

   output[0] = v0;
   output[1] = v1;
}


/**
 * Compute a0, dadx and dady for a linearly interpolated coefficient,
 * for a line.
 * v[0] and v[1] are vmin and vmax, respectively.
 */
static void
line_linear_coeff(const struct setup_context *setup,
                  struct tgsi_interp_coef *coef,
                  uint i,
                  const float v[2])
{
   const float da = v[1] - v[0];
   const float dadx = da * setup->emaj.dx * setup->oneoverarea;
   const float dady = da * setup->emaj.dy * setup->oneoverarea;
   coef->dadx[i] = dadx;
   coef->dady[i] = dady;
   coef->a0[i] = (v[0] -
                  (dadx * (setup->vmin[0][0] - setup->pixel_offset) +
                   dady * (setup->vmin[0][1] - setup->pixel_offset)));
}


/**
 * Compute a0, dadx and dady for a perspective-corrected interpolant,
 * for a line.
 * v[0] and v[1] are vmin and vmax, respectively.
 */
static void
line_persp_coeff(const struct setup_context *setup,
                 struct tgsi_interp_coef *coef,
                 uint i,
                 const float v[2])
{
   const float a0 = v[0] * setup->vmin[0][3];
   const float a1 = v[1] * setup->vmax[0][3];
   const float da = a1 - a0;
   const float dadx = da * setup->emaj.dx * setup->oneoverarea;
   const float dady = da * setup->emaj.dy * setup->oneoverarea;
   coef->dadx[i] = dadx;
   coef->dady[i] = dady;
   coef->a0[i] = (a0 -
                  (dadx * (setup->vmin[0][0] - setup->pixel_offset) +
                   dady * (setup->vmin[0][1] - setup->pixel_offset)));
}


/**
 * Compute the setup->coef[] array dadx, dady, a0 values.
 * Must be called after setup->vmin,vmax are initialized.
 */
static boolean
setup_line_coefficients(struct setup_context *setup,
                        const float (*v0)[4],
                        const float (*v1)[4])
{
   struct softpipe_context *softpipe = setup->softpipe;
   const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
   uint fragSlot;
   float area;
   float v[2];

   /* use setup->vmin, vmax to point to vertices */
   if (softpipe->rasterizer->flatshade_first)
      setup->vprovoke = v0;
   else
      setup->vprovoke = v1;
   setup->vmin = v0;
   setup->vmax = v1;

   setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0];
   setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1];

   /* NOTE: this is not really area but something proportional to it */
   area = setup->emaj.dx * setup->emaj.dx + setup->emaj.dy * setup->emaj.dy;
   if (area == 0.0f || util_is_inf_or_nan(area))
      return FALSE;
   setup->oneoverarea = 1.0f / area;

   /* z and w are done by linear interpolation:
    */
   v[0] = setup->vmin[0][2];
   v[1] = setup->vmax[0][2];
   line_linear_coeff(setup, &setup->posCoef, 2, v);

   v[0] = setup->vmin[0][3];
   v[1] = setup->vmax[0][3];
   line_linear_coeff(setup, &setup->posCoef, 3, v);

   /* setup interpolation for all the remaining attributes:
    */
   for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
      uint j;

      switch (vinfo->attrib[fragSlot].interp_mode) {
      case INTERP_CONSTANT:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++)
            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
         break;
      case INTERP_LINEAR:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
            line_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
                                        setup->vmax[vertSlot][j],
                                        fsInfo->input_cylindrical_wrap[fragSlot] & (1 << j),
                                        v);
            line_linear_coeff(setup, &setup->coef[fragSlot], j, v);
         }
         break;
      case INTERP_PERSPECTIVE:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++) {
            line_apply_cylindrical_wrap(setup->vmin[vertSlot][j],
                                        setup->vmax[vertSlot][j],
                                        fsInfo->input_cylindrical_wrap[fragSlot] & (1 << j),
                                        v);
            line_persp_coeff(setup, &setup->coef[fragSlot], j, v);
         }
         break;
      case INTERP_POS:
         setup_fragcoord_coeff(setup, fragSlot);
         break;
      default:
         assert(0);
      }

      if (fsInfo->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
         /* convert 0 to 1.0 and 1 to -1.0 */
         setup->coef[fragSlot].a0[0] = setup->facing * -2.0f + 1.0f;
         setup->coef[fragSlot].dadx[0] = 0.0;
         setup->coef[fragSlot].dady[0] = 0.0;
      }
   }
   return TRUE;
}


/**
 * Plot a pixel in a line segment.
 */
static inline void
plot(struct setup_context *setup, int x, int y)
{
   const int iy = y & 1;
   const int ix = x & 1;
   const int quadX = x - ix;
   const int quadY = y - iy;
   const int mask = (1 << ix) << (2 * iy);

   if (quadX != setup->quad[0].input.x0 ||
       quadY != setup->quad[0].input.y0)
   {
      /* flush prev quad, start new quad */

      if (setup->quad[0].input.x0 != -1)
         clip_emit_quad( setup, &setup->quad[0] );

      setup->quad[0].input.x0 = quadX;
      setup->quad[0].input.y0 = quadY;
      setup->quad[0].inout.mask = 0x0;
   }

   setup->quad[0].inout.mask |= mask;
}


/**
 * Do setup for line rasterization, then render the line.
 * Single-pixel width, no stipple, etc.  We rely on the 'draw' module
 * to handle stippling and wide lines.
 */
void
sp_setup_line(struct setup_context *setup,
              const float (*v0)[4],
              const float (*v1)[4])
{
   int x0 = (int) v0[0][0];
   int x1 = (int) v1[0][0];
   int y0 = (int) v0[0][1];
   int y1 = (int) v1[0][1];
   int dx = x1 - x0;
   int dy = y1 - y0;
   int xstep, ystep;
   uint layer = 0;

#if DEBUG_VERTS
   debug_printf("Setup line:\n");
   print_vertex(setup, v0);
   print_vertex(setup, v1);
#endif

   if (setup->softpipe->no_rast || setup->softpipe->rasterizer->rasterizer_discard)
      return;

   if (dx == 0 && dy == 0)
      return;

   if (!setup_line_coefficients(setup, v0, v1))
      return;

   assert(v0[0][0] < 1.0e9);
   assert(v0[0][1] < 1.0e9);
   assert(v1[0][0] < 1.0e9);
   assert(v1[0][1] < 1.0e9);

   if (dx < 0) {
      dx = -dx;   /* make positive */
      xstep = -1;
   }
   else {
      xstep = 1;
   }

   if (dy < 0) {
      dy = -dy;   /* make positive */
      ystep = -1;
   }
   else {
      ystep = 1;
   }

   assert(dx >= 0);
   assert(dy >= 0);
   assert(setup->softpipe->reduced_prim == PIPE_PRIM_LINES);

   setup->quad[0].input.x0 = setup->quad[0].input.y0 = -1;
   setup->quad[0].inout.mask = 0x0;
   if (setup->softpipe->layer_slot > 0) {
      layer = *(unsigned *)v1[setup->softpipe->layer_slot];
      layer = MIN2(layer, setup->max_layer);
   }
   setup->quad[0].input.layer = layer;

   /* XXX temporary: set coverage to 1.0 so the line appears
    * if AA mode happens to be enabled.
    */
   setup->quad[0].input.coverage[0] =
   setup->quad[0].input.coverage[1] =
   setup->quad[0].input.coverage[2] =
   setup->quad[0].input.coverage[3] = 1.0;

   if (dx > dy) {
      /*** X-major line ***/
      int i;
      const int errorInc = dy + dy;
      int error = errorInc - dx;
      const int errorDec = error - dx;

      for (i = 0; i < dx; i++) {
         plot(setup, x0, y0);

         x0 += xstep;
         if (error < 0) {
            error += errorInc;
         }
         else {
            error += errorDec;
            y0 += ystep;
         }
      }
   }
   else {
      /*** Y-major line ***/
      int i;
      const int errorInc = dx + dx;
      int error = errorInc - dy;
      const int errorDec = error - dy;

      for (i = 0; i < dy; i++) {
         plot(setup, x0, y0);

         y0 += ystep;
         if (error < 0) {
            error += errorInc;
         }
         else {
            error += errorDec;
            x0 += xstep;
         }
      }
   }

   /* draw final quad */
   if (setup->quad[0].inout.mask) {
      clip_emit_quad( setup, &setup->quad[0] );
   }
}


static void
point_persp_coeff(const struct setup_context *setup,
                  const float (*vert)[4],
                  struct tgsi_interp_coef *coef,
                  uint vertSlot, uint i)
{
   assert(i <= 3);
   coef->dadx[i] = 0.0F;
   coef->dady[i] = 0.0F;
   coef->a0[i] = vert[vertSlot][i] * vert[0][3];
}


/**
 * Do setup for point rasterization, then render the point.
 * Round or square points...
 * XXX could optimize a lot for 1-pixel points.
 */
void
sp_setup_point(struct setup_context *setup,
               const float (*v0)[4])
{
   struct softpipe_context *softpipe = setup->softpipe;
   const struct tgsi_shader_info *fsInfo = &setup->softpipe->fs_variant->info;
   const int sizeAttr = setup->softpipe->psize_slot;
   const float size
      = sizeAttr > 0 ? v0[sizeAttr][0]
      : setup->softpipe->rasterizer->point_size;
   const float halfSize = 0.5F * size;
   const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth;
   const float x = v0[0][0];  /* Note: data[0] is always position */
   const float y = v0[0][1];
   const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe);
   uint fragSlot;
   uint layer = 0;
#if DEBUG_VERTS
   debug_printf("Setup point:\n");
   print_vertex(setup, v0);
#endif

   if (setup->softpipe->no_rast || setup->softpipe->rasterizer->rasterizer_discard)
      return;

   assert(setup->softpipe->reduced_prim == PIPE_PRIM_POINTS);

   if (setup->softpipe->layer_slot > 0) {
      layer = *(unsigned *)v0[setup->softpipe->layer_slot];
      layer = MIN2(layer, setup->max_layer);
   }
   setup->quad[0].input.layer = layer;

   /* For points, all interpolants are constant-valued.
    * However, for point sprites, we'll need to setup texcoords appropriately.
    * XXX: which coefficients are the texcoords???
    * We may do point sprites as textured quads...
    *
    * KW: We don't know which coefficients are texcoords - ultimately
    * the choice of what interpolation mode to use for each attribute
    * should be determined by the fragment program, using
    * per-attribute declaration statements that include interpolation
    * mode as a parameter.  So either the fragment program will have
    * to be adjusted for pointsprite vs normal point behaviour, or
    * otherwise a special interpolation mode will have to be defined
    * which matches the required behaviour for point sprites.  But -
    * the latter is not a feature of normal hardware, and as such
    * probably should be ruled out on that basis.
    */
   setup->vprovoke = v0;

   /* setup Z, W */
   const_coeff(setup, &setup->posCoef, 0, 2);
   const_coeff(setup, &setup->posCoef, 0, 3);

   for (fragSlot = 0; fragSlot < fsInfo->num_inputs; fragSlot++) {
      const uint vertSlot = vinfo->attrib[fragSlot].src_index;
      uint j;

      switch (vinfo->attrib[fragSlot].interp_mode) {
      case INTERP_CONSTANT:
         /* fall-through */
      case INTERP_LINEAR:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++)
            const_coeff(setup, &setup->coef[fragSlot], vertSlot, j);
         break;
      case INTERP_PERSPECTIVE:
         for (j = 0; j < TGSI_NUM_CHANNELS; j++)
            point_persp_coeff(setup, setup->vprovoke,
                              &setup->coef[fragSlot], vertSlot, j);
         break;
      case INTERP_POS:
         setup_fragcoord_coeff(setup, fragSlot);
         break;
      default:
         assert(0);
      }

      if (fsInfo->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FACE) {
         /* convert 0 to 1.0 and 1 to -1.0 */
         setup->coef[fragSlot].a0[0] = setup->facing * -2.0f + 1.0f;
         setup->coef[fragSlot].dadx[0] = 0.0;
         setup->coef[fragSlot].dady[0] = 0.0;
      }
   }


   if (halfSize <= 0.5 && !round) {
      /* special case for 1-pixel points */
      const int ix = ((int) x) & 1;
      const int iy = ((int) y) & 1;
      setup->quad[0].input.x0 = (int) x - ix;
      setup->quad[0].input.y0 = (int) y - iy;
      setup->quad[0].inout.mask = (1 << ix) << (2 * iy);
      clip_emit_quad( setup, &setup->quad[0] );
   }
   else {
      if (round) {
         /* rounded points */
         const int ixmin = block((int) (x - halfSize));
         const int ixmax = block((int) (x + halfSize));
         const int iymin = block((int) (y - halfSize));
         const int iymax = block((int) (y + halfSize));
         const float rmin = halfSize - 0.7071F;  /* 0.7071 = sqrt(2)/2 */
         const float rmax = halfSize + 0.7071F;
         const float rmin2 = MAX2(0.0F, rmin * rmin);
         const float rmax2 = rmax * rmax;
         const float cscale = 1.0F / (rmax2 - rmin2);
         int ix, iy;

         for (iy = iymin; iy <= iymax; iy += 2) {
            for (ix = ixmin; ix <= ixmax; ix += 2) {
               float dx, dy, dist2, cover;

               setup->quad[0].inout.mask = 0x0;

               dx = (ix + 0.5f) - x;
               dy = (iy + 0.5f) - y;
               dist2 = dx * dx + dy * dy;
               if (dist2 <= rmax2) {
                  cover = 1.0F - (dist2 - rmin2) * cscale;
                  setup->quad[0].input.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
                  setup->quad[0].inout.mask |= MASK_TOP_LEFT;
               }

               dx = (ix + 1.5f) - x;
               dy = (iy + 0.5f) - y;
               dist2 = dx * dx + dy * dy;
               if (dist2 <= rmax2) {
                  cover = 1.0F - (dist2 - rmin2) * cscale;
                  setup->quad[0].input.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
                  setup->quad[0].inout.mask |= MASK_TOP_RIGHT;
               }

               dx = (ix + 0.5f) - x;
               dy = (iy + 1.5f) - y;
               dist2 = dx * dx + dy * dy;
               if (dist2 <= rmax2) {
                  cover = 1.0F - (dist2 - rmin2) * cscale;
                  setup->quad[0].input.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
                  setup->quad[0].inout.mask |= MASK_BOTTOM_LEFT;
               }

               dx = (ix + 1.5f) - x;
               dy = (iy + 1.5f) - y;
               dist2 = dx * dx + dy * dy;
               if (dist2 <= rmax2) {
                  cover = 1.0F - (dist2 - rmin2) * cscale;
                  setup->quad[0].input.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
                  setup->quad[0].inout.mask |= MASK_BOTTOM_RIGHT;
               }

               if (setup->quad[0].inout.mask) {
                  setup->quad[0].input.x0 = ix;
                  setup->quad[0].input.y0 = iy;
                  clip_emit_quad( setup, &setup->quad[0] );
               }
            }
         }
      }
      else {
         /* square points */
         const int xmin = (int) (x + 0.75 - halfSize);
         const int ymin = (int) (y + 0.25 - halfSize);
         const int xmax = xmin + (int) size;
         const int ymax = ymin + (int) size;
         /* XXX could apply scissor to xmin,ymin,xmax,ymax now */
         const int ixmin = block(xmin);
         const int ixmax = block(xmax - 1);
         const int iymin = block(ymin);
         const int iymax = block(ymax - 1);
         int ix, iy;

         /*
         debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax);
         */
         for (iy = iymin; iy <= iymax; iy += 2) {
            uint rowMask = 0xf;
            if (iy < ymin) {
               /* above the top edge */
               rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
            }
            if (iy + 1 >= ymax) {
               /* below the bottom edge */
               rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
            }

            for (ix = ixmin; ix <= ixmax; ix += 2) {
               uint mask = rowMask;

               if (ix < xmin) {
                  /* fragment is past left edge of point, turn off left bits */
                  mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
               }
               if (ix + 1 >= xmax) {
                  /* past the right edge */
                  mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
               }

               setup->quad[0].inout.mask = mask;
               setup->quad[0].input.x0 = ix;
               setup->quad[0].input.y0 = iy;
               clip_emit_quad( setup, &setup->quad[0] );
            }
         }
      }
   }
}


/**
 * Called by vbuf code just before we start buffering primitives.
 */
void
sp_setup_prepare(struct setup_context *setup)
{
   struct softpipe_context *sp = setup->softpipe;
   int i;
   unsigned max_layer = ~0;
   if (sp->dirty) {
      softpipe_update_derived(sp, sp->reduced_api_prim);
   }

   /* Note: nr_attrs is only used for debugging (vertex printing) */
   setup->nr_vertex_attrs = draw_num_shader_outputs(sp->draw);

   /*
    * Determine how many layers the fb has (used for clamping layer value).
    * OpenGL (but not d3d10) permits different amount of layers per rt, however
    * results are undefined if layer exceeds the amount of layers of ANY
    * attachment hence don't need separate per cbuf and zsbuf max.
    */
   for (i = 0; i < setup->softpipe->framebuffer.nr_cbufs; i++) {
      struct pipe_surface *cbuf = setup->softpipe->framebuffer.cbufs[i];
      if (cbuf) {
         max_layer = MIN2(max_layer,
                          cbuf->u.tex.last_layer - cbuf->u.tex.first_layer);

      }
   }

   setup->max_layer = max_layer;

   sp->quad.first->begin( sp->quad.first );

   if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES &&
       sp->rasterizer->fill_front == PIPE_POLYGON_MODE_FILL &&
       sp->rasterizer->fill_back == PIPE_POLYGON_MODE_FILL) {
      /* we'll do culling */
      setup->cull_face = sp->rasterizer->cull_face;
   }
   else {
      /* 'draw' will do culling */
      setup->cull_face = PIPE_FACE_NONE;
   }
}


void
sp_setup_destroy_context(struct setup_context *setup)
{
   FREE( setup );
}


/**
 * Create a new primitive setup/render stage.
 */
struct setup_context *
sp_setup_create_context(struct softpipe_context *softpipe)
{
   struct setup_context *setup = CALLOC_STRUCT(setup_context);
   unsigned i;

   setup->softpipe = softpipe;

   for (i = 0; i < MAX_QUADS; i++) {
      setup->quad[i].coef = setup->coef;
      setup->quad[i].posCoef = &setup->posCoef;
   }

   setup->span.left[0] = 1000000;     /* greater than right[0] */
   setup->span.left[1] = 1000000;     /* greater than right[1] */

   return setup;
}