summaryrefslogtreecommitdiff
path: root/src/mesa/pipe/softpipe/sp_prim_setup.c
blob: 2e27d00acfb02c4fec960caa3fbe63a67e38fd09 (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
/**************************************************************************
 * 
 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
 * 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 TUNGSTEN GRAPHICS 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 <keith@tungstengraphics.com>
 * \author  Brian Paul
 */


#include "sp_context.h"
#include "sp_headers.h"
#include "sp_quad.h"
#include "sp_prim_setup.h"
#include "pipe/draw/draw_private.h"
#include "pipe/draw/draw_vertex.h"
#include "pipe/p_util.h"

#define DEBUG_VERTS 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 */
};


/**
 * Triangle setup info (derived from draw_stage).
 * Also used for line drawing (taking some liberties).
 */
struct setup_stage {
   struct draw_stage stage; /**< This must be first (base class) */

   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 struct vertex_header *vmax;
   const struct vertex_header *vmid;
   const struct vertex_header *vmin;
   const struct vertex_header *vprovoke;

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

   float oneoverarea;

   struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS];
   struct quad_header quad; 

   struct {
      int left[2];   /**< [0] = row0, [1] = row1 */
      int right[2];
      int y;
      unsigned y_flags;
      unsigned mask;     /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */
   } span;
};



/**
 * Basically a cast wrapper.
 */
static INLINE struct setup_stage *setup_stage( struct draw_stage *stage )
{
   return (struct setup_stage *)stage;
}


/**
 * Clip setup->quad against the scissor/surface bounds.
 */
static INLINE void
quad_clip(struct setup_stage *setup)
{
   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 (setup->quad.x0 >= maxx ||
       setup->quad.y0 >= maxy ||
       setup->quad.x0 + 1 < minx ||
       setup->quad.y0 + 1 < miny) {
      /* totally clipped */
      setup->quad.mask = 0x0;
      return;
   }
   if (setup->quad.x0 < minx)
      setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
   if (setup->quad.y0 < miny)
      setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT);
   if (setup->quad.x0 == maxx - 1)
      setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
   if (setup->quad.y0 == maxy - 1)
      setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
}


/**
 * Emit a quad (pass to next stage) with clipping.
 */
static INLINE void
clip_emit_quad(struct setup_stage *setup)
{
   quad_clip(setup);
   if (setup->quad.mask) {
      struct softpipe_context *sp = setup->softpipe;
      sp->quad.first->run(sp->quad.first, &setup->quad);
   }
}


/**
 * Emit a quad (pass to next stage).  No clipping is done.
 */
static INLINE void
emit_quad( struct setup_stage *setup, int x, int y, unsigned mask )
{
   struct softpipe_context *sp = setup->softpipe;
   setup->quad.x0 = x;
   setup->quad.y0 = y;
   setup->quad.mask = mask;
   sp->quad.first->run(sp->quad.first, &setup->quad);
}


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


/**
 * Compute mask which indicates which pixels in the 2x2 quad are actually inside
 * the triangle's bounds.
 *
 * this is pretty nasty...  may need to rework flush_spans again to
 * fix it, if possible.
 */
static unsigned calculate_mask( struct setup_stage *setup,
			    int x )
{
   unsigned mask = 0;

   if (x >= setup->span.left[0] && x < setup->span.right[0]) 
      mask |= MASK_BOTTOM_LEFT;

   if (x >= setup->span.left[1] && x < setup->span.right[1]) 
      mask |= MASK_TOP_LEFT;
      
   if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) 
      mask |= MASK_BOTTOM_RIGHT;

   if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) 
      mask |= MASK_TOP_RIGHT;

   return mask;
}


/**
 * Render a horizontal span of quads
 */
static void flush_spans( struct setup_stage *setup )
{
   int minleft, maxright;
   int x;

   switch (setup->span.y_flags) {      
   case 3:
      minleft = MIN2(setup->span.left[0], setup->span.left[1]);
      maxright = MAX2(setup->span.right[0], setup->span.right[1]);
      break;

   case 1:
      minleft = setup->span.left[0];
      maxright = setup->span.right[0];
      break;

   case 2:
      minleft = setup->span.left[1];
      maxright = setup->span.right[1];
      break;

   default:
      return;
   }


   for (x = block(minleft); x <= block(maxright); )
   {
      emit_quad( setup, x, setup->span.y, 
                 calculate_mask( setup, x ) );
      x += 2;
   }

   setup->span.y = 0;
   setup->span.y_flags = 0;
   setup->span.right[0] = 0;
   setup->span.right[1] = 0;
}

#if DEBUG_VERTS
static void print_vertex(const struct setup_stage *setup,
                         const struct vertex_header *v)
{
   int i;
   fprintf(stderr, "Vertex: (%p)\n", v);
   for (i = 0; i < setup->quad.nr_attrs; i++) {
      fprintf(stderr, "  %d: %f %f %f %f\n",  i, 
              v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]);
   }
}
#endif

static boolean setup_sort_vertices( struct setup_stage *setup,
				      const struct prim_header *prim )
{
   const struct vertex_header *v0 = prim->v[0];
   const struct vertex_header *v1 = prim->v[1];
   const struct vertex_header *v2 = prim->v[2];

#if DEBUG_VERTS
   fprintf(stderr, "Triangle:\n");
   print_vertex(setup, v0);
   print_vertex(setup, v1);
   print_vertex(setup, v2);
#endif

   setup->vprovoke = v2;

   /* determine bottom to top order of vertices */
   {
      float y0 = v0->data[0][1];
      float y1 = v1->data[0][1];
      float y2 = v2->data[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->data[0][0] - setup->vmin->data[0][0];
   setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1];
   setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0];
   setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1];
   setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0];
   setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[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;
      /*
      _mesa_printf("%s one-over-area %f  area %f  det %f\n",
                   __FUNCTION__, setup->oneoverarea, area, prim->det );
      */
   }

   /* 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
    */
   setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW);

   return TRUE;
}


/**
 * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
 * The value value comes from vertex->data[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_stage *setup,
			 unsigned slot,
			 unsigned i )
{
   assert(slot < PIPE_MAX_SHADER_INPUTS);
   assert(i <= 3);

   setup->coef[slot].dadx[i] = 0;
   setup->coef[slot].dady[i] = 0;

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


/**
 * Compute a0, dadx and dady for a linearly interpolated coefficient,
 * for a triangle.
 */
static void tri_linear_coeff( struct setup_stage *setup,
                              unsigned slot,
                              unsigned i)
{
   float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i];
   float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i];
   float a = setup->ebot.dy * majda - botda * setup->emaj.dy;
   float b = setup->emaj.dx * botda - majda * setup->ebot.dx;
   
   assert(slot < PIPE_MAX_SHADER_INPUTS);
   assert(i <= 3);

   setup->coef[slot].dadx[i] = a * setup->oneoverarea;
   setup->coef[slot].dady[i] = b * setup->oneoverarea;

   /* 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 (0.5, 0.5).
    *
    * 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.
    */
   setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - 
			    (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + 
			     setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f)));

   /*
   _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n",
		slot, "xyzw"[i], 
		setup->coef[slot].a0[i],
		setup->coef[slot].dadx[i],
		setup->coef[slot].dady[i]);
   */
}


/**
 * 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.
 */
static void tri_persp_coeff( struct setup_stage *setup,
                             unsigned slot,
                             unsigned i )
{
   /* premultiply by 1/w:
    */
   float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3];
   float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3];
   float maxa = setup->vmax->data[slot][i] * setup->vmax->data[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;
      
   /*
   printf("tri persp %d,%d: %f %f %f\n", slot, i,
          setup->vmin->data[slot][i],
          setup->vmid->data[slot][i],
          setup->vmax->data[slot][i]
          );
   */

   assert(slot < PIPE_MAX_SHADER_INPUTS);
   assert(i <= 3);

   setup->coef[slot].dadx[i] = a * setup->oneoverarea;
   setup->coef[slot].dady[i] = b * setup->oneoverarea;
   setup->coef[slot].a0[i] = (mina - 
			    (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + 
			     setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f)));
}


/**
 * 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_stage *setup )
{
   const interp_mode *interp = setup->softpipe->vertex_info.interp_mode;
   unsigned slot, j;

   /* z and w are done by linear interpolation:
    */
   tri_linear_coeff(setup, 0, 2);
   tri_linear_coeff(setup, 0, 3);

   /* setup interpolation for all the remaining attributes:
    */
   for (slot = 1; slot < setup->quad.nr_attrs; slot++) {
      switch (interp[slot]) {
      case INTERP_CONSTANT:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    const_coeff(setup, slot, j);
	 break;
      
      case INTERP_LINEAR:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    tri_linear_coeff(setup, slot, j);
	 break;

      case INTERP_PERSPECTIVE:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    tri_persp_coeff(setup, slot, j);
	 break;

      default:
         /* invalid interp mode */
         assert(0);
      }
   }
}



static void setup_tri_edges( struct setup_stage *setup )
{
   float vmin_x = setup->vmin->data[0][0] + 0.5f;
   float vmid_x = setup->vmid->data[0][0] + 0.5f;

   float vmin_y = setup->vmin->data[0][1] - 0.5f;
   float vmid_y = setup->vmid->data[0][1] - 0.5f;
   float vmax_y = setup->vmax->data[0][1] - 0.5f;

   setup->emaj.sy = ceilf(vmin_y);
   setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy);
   setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy;
   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.dx / setup->etop.dy;
   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.dx / setup->ebot.dy;
   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_stage *setup,
			 struct edge *eleft,
			 struct edge *eright,
			 unsigned 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);

   /* clip top/bottom */
   start_y = sy;
   finish_y = sy + lines;

   if (start_y < miny)
      start_y = miny;

   if (finish_y > maxy)
      finish_y = maxy;

   start_y -= sy;
   finish_y -= sy;

   /*
   _mesa_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;
         setup->span.y_flags |= 1<<(_y&1);
      }
   }


   /* 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;
}


/**
 * Do setup for triangle rasterization, then render the triangle.
 */
static void setup_tri( struct draw_stage *stage,
		       struct prim_header *prim )
{
   struct setup_stage *setup = setup_stage( stage );

   /*
   _mesa_printf("%s\n", __FUNCTION__ );
   */

   setup_sort_vertices( setup, prim );
   setup_tri_coefficients( setup );
   setup_tri_edges( setup );

   setup->quad.prim = PRIM_TRI;

   setup->span.y = 0;
   setup->span.y_flags = 0;
   setup->span.right[0] = 0;
   setup->span.right[1] = 0;
   /*   setup->span.z_mode = tri_z_mode( setup->ctx ); */

   /*   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 );
}



/**
 * Compute a0, dadx and dady for a linearly interpolated coefficient,
 * for a line.
 */
static void
line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i)
{
   const float dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i];
   const float dadx = dz * setup->emaj.dx * setup->oneoverarea;
   const float dady = dz * setup->emaj.dy * setup->oneoverarea;
   setup->coef[slot].dadx[i] = dadx;
   setup->coef[slot].dady[i] = dady;
   setup->coef[slot].a0[i]
      = (setup->vmin->data[slot][i] - 
         (dadx * (setup->vmin->data[0][0] - 0.5f) + 
          dady * (setup->vmin->data[0][1] - 0.5f)));
}


/**
 * Compute a0, dadx and dady for a perspective-corrected interpolant,
 * for a line.
 */
static void
line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i)
{
   /* XXX to do */
   line_linear_coeff(setup, slot, i); /* XXX temporary */
}


/**
 * Compute the setup->coef[] array dadx, dady, a0 values.
 * Must be called after setup->vmin,vmax are initialized.
 */
static INLINE void
setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim)
{
   const interp_mode *interp = setup->softpipe->vertex_info.interp_mode;
   unsigned slot, j;

   /* use setup->vmin, vmax to point to vertices */
   setup->vprovoke = prim->v[1];
   setup->vmin = prim->v[0];
   setup->vmax = prim->v[1];

   setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0];
   setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1];
   /* NOTE: this is not really 1/area */
   setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx +
                                setup->emaj.dy * setup->emaj.dy);

   /* z and w are done by linear interpolation:
    */
   line_linear_coeff(setup, 0, 2);
   line_linear_coeff(setup, 0, 3);

   /* setup interpolation for all the remaining attributes:
    */
   for (slot = 1; slot < setup->quad.nr_attrs; slot++) {
      switch (interp[slot]) {
      case INTERP_CONSTANT:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    const_coeff(setup, slot, j);
	 break;
      
      case INTERP_LINEAR:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    line_linear_coeff(setup, slot, j);
	 break;

      case INTERP_PERSPECTIVE:
	 for (j = 0; j < NUM_CHANNELS; j++)
	    line_persp_coeff(setup, slot, j);
	 break;

      default:
         /* invalid interp mode */
         assert(0);
      }
   }
}


/**
 * Plot a pixel in a line segment.
 */
static INLINE void
plot(struct setup_stage *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.x0 || 
       quadY != setup->quad.y0) 
   {
      /* flush prev quad, start new quad */

      if (setup->quad.x0 != -1)
         clip_emit_quad(setup);

      setup->quad.x0 = quadX;
      setup->quad.y0 = quadY;
      setup->quad.mask = 0x0;
   }

   setup->quad.mask |= mask;
}


/**
 * Determine whether or not to emit a line fragment by checking
 * line stipple pattern.
 */
static INLINE unsigned
stipple_test(int counter, ushort pattern, int factor)
{
   int b = (counter / factor) & 0xf;
   return (1 << b) & pattern;
}


/**
 * Do setup for line rasterization, then render the line.
 * XXX single-pixel width, no stipple, etc
 */
static void
setup_line(struct draw_stage *stage, struct prim_header *prim)
{
   const struct vertex_header *v0 = prim->v[0];
   const struct vertex_header *v1 = prim->v[1];
   struct setup_stage *setup = setup_stage( stage );
   struct softpipe_context *sp = setup->softpipe;

   int x0 = (int) v0->data[0][0];
   int x1 = (int) v1->data[0][0];
   int y0 = (int) v0->data[0][1];
   int y1 = (int) v1->data[0][1];
   int dx = x1 - x0;
   int dy = y1 - y0;
   int xstep, ystep;

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

   setup_line_coefficients(setup, prim);

   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);

   setup->quad.x0 = setup->quad.y0 = -1;
   setup->quad.mask = 0x0;
   setup->quad.prim = PRIM_LINE;
   /* XXX temporary: set coverage to 1.0 so the line appears
    * if AA mode happens to be enabled.
    */
   setup->quad.coverage[0] =
   setup->quad.coverage[1] =
   setup->quad.coverage[2] =
   setup->quad.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++) {
         if (!sp->rasterizer->line_stipple_enable ||
             stipple_test(sp->line_stipple_counter,
                          sp->rasterizer->line_stipple_pattern,
                          sp->rasterizer->line_stipple_factor + 1)) {
             plot(setup, x0, y0);
         }

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

         sp->line_stipple_counter++;
      }
   }
   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++) {
         if (!sp->rasterizer->line_stipple_enable ||
             stipple_test(sp->line_stipple_counter,
                          sp->rasterizer->line_stipple_pattern,
                          sp->rasterizer->line_stipple_factor + 1)) {
            plot(setup, x0, y0);
         }

         y0 += ystep;

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

         sp->line_stipple_counter++;
      }
   }

   /* draw final quad */
   if (setup->quad.mask) {
      clip_emit_quad(setup);
   }
}


/**
 * Do setup for point rasterization, then render the point.
 * Round or square points...
 * XXX could optimize a lot for 1-pixel points.
 */
static void
setup_point(struct draw_stage *stage, struct prim_header *prim)
{
   struct setup_stage *setup = setup_stage( stage );
   const struct vertex_header *v0 = prim->v[0];
   const int sizeAttr = setup->softpipe->psize_slot;
   const float halfSize
      = sizeAttr > 0 ? (0.5f * v0->data[sizeAttr][0])
        : (0.5f * setup->softpipe->rasterizer->point_size);
   const boolean round = setup->softpipe->rasterizer->point_smooth;
   const float x = v0->data[0][0];  /* Note: data[0] is always position */
   const float y = v0->data[0][1];
   unsigned slot, j;

   /* 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 = prim->v[0];
   const_coeff(setup, 0, 2);
   const_coeff(setup, 0, 3);
   for (slot = 1; slot < setup->quad.nr_attrs; slot++) {
      for (j = 0; j < NUM_CHANNELS; j++)
         const_coeff(setup, slot, j);
   }

   setup->quad.prim = PRIM_POINT;

   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.x0 = (int) x - ix;
      setup->quad.y0 = (int) y - iy;
      setup->quad.mask = (1 << ix) << (2 * iy);
      clip_emit_quad(setup);
   }
   else {
      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));
      int ix, iy;

      if (round) {
         /* rounded points */
         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);

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

               setup->quad.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.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f);
                  setup->quad.mask |= MASK_BOTTOM_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.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f);
                  setup->quad.mask |= MASK_BOTTOM_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.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f);
                  setup->quad.mask |= MASK_TOP_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.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f);
                  setup->quad.mask |= MASK_TOP_RIGHT;
               }

               if (setup->quad.mask) {
                  setup->quad.x0 = ix;
                  setup->quad.y0 = iy;
                  clip_emit_quad(setup);
               }
            }
         }
      }
      else {
         /* square points */
         for (iy = iymin; iy <= iymax; iy += 2) {
            for (ix = ixmin; ix <= ixmax; ix += 2) {
               setup->quad.mask = 0xf;

               if (ix + 0.5 < x - halfSize) {
                  /* fragment is past left edge of point, turn off left bits */
                  setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_TOP_LEFT);
               }

               if (ix + 1.5 > x + halfSize) {
                  /* past the right edge */
                  setup->quad.mask &= ~(MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT);
               }

               if (iy + 0.5 < y - halfSize) {
                  /* below the bottom edge */
                  setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT);
               }

               if (iy + 1.5 > y + halfSize) {
                  /* above the top edge */
                  setup->quad.mask &= ~(MASK_TOP_LEFT | MASK_TOP_RIGHT);
               }

               if (setup->quad.mask) {
                  setup->quad.x0 = ix;
                  setup->quad.y0 = iy;
                  clip_emit_quad(setup);
               }
            }
         }
      }
   }
}



static void setup_begin( struct draw_stage *stage )
{
   struct setup_stage *setup = setup_stage(stage);
   struct softpipe_context *sp = setup->softpipe;

   setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs;

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


static void setup_end( struct draw_stage *stage )
{
}


static void reset_stipple_counter( struct draw_stage *stage )
{
   struct setup_stage *setup = setup_stage(stage);
   setup->softpipe->line_stipple_counter = 0;
}


/**
 * Create a new primitive setup/render stage.
 */
struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe )
{
   struct setup_stage *setup = CALLOC_STRUCT(setup_stage);

   setup->softpipe = softpipe;
   setup->stage.draw = softpipe->draw;
   setup->stage.begin = setup_begin;
   setup->stage.point = setup_point;
   setup->stage.line = setup_line;
   setup->stage.tri = setup_tri;
   setup->stage.end = setup_end;
   setup->stage.reset_stipple_counter = reset_stipple_counter;

   setup->quad.coef = setup->coef;

   return &setup->stage;
}


/* Recalculate det.  This is only used in the test harness below:
 */
static void calc_det( struct prim_header *header )
{
   /* Window coords: */
   const float *v0 = header->v[0]->data[0];
   const float *v1 = header->v[1]->data[0];
   const float *v2 = header->v[2]->data[0];

   /* edge vectors e = v0 - v2, f = v1 - v2 */
   const float ex = v0[0] - v2[0];
   const float ey = v0[1] - v2[1];
   const float fx = v1[0] - v2[0];
   const float fy = v1[1] - v2[1];
   
   /* det = cross(e,f).z */
   header->det = ex * fy - ey * fx;
}



/* Test harness - feed vertex buffer back into prim pipeline.
 *
 * The big issue at this point is that reset_stipple doesn't make it
 * through the interface.  Probably need to split primitives at reset
 * stipple, perhaps using the ~0 index marker.
 */
void sp_vbuf_setup_draw( struct pipe_context *pipe,
                         unsigned primitive,
                         const ushort *elements,
                         unsigned nr_elements,
                         const void *vertex_buffer,
                         unsigned nr_vertices )
{
   struct softpipe_context *softpipe = softpipe_context( pipe );
   struct setup_stage *setup = setup_stage( softpipe->setup );
   struct prim_header prim;
   unsigned vertex_size = setup->stage.draw->vertex_info.size * sizeof(float);
   int i, j;

   prim.det = 0;
   prim.reset_line_stipple = 0;
   prim.edgeflags = 0;
   prim.pad = 0;

   setup->stage.begin( &setup->stage );

   switch (primitive) {
   case PIPE_PRIM_TRIANGLES:
      for (i = 0; i < nr_elements; i += 3) {
         for (j = 0; j < 3; j++) 
            prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + 
                                                 elements[i+j] * vertex_size);
         
         calc_det(&prim);
         setup->stage.tri( &setup->stage, &prim );
      }
      break;

   case PIPE_PRIM_LINES:
      for (i = 0; i < nr_elements; i += 2) {
         for (j = 0; j < 2; j++) 
            prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + 
                                                 elements[i+j] * vertex_size);
         
         setup->stage.line( &setup->stage, &prim );
      }
      break;

   case PIPE_PRIM_POINTS:
      for (i = 0; i < nr_elements; i += 2) {
         prim.v[i] = (struct vertex_header *)((char *)vertex_buffer + 
                                              elements[i] * vertex_size);         
         setup->stage.point( &setup->stage, &prim );
      }
      break;
   }

   setup->stage.end( &setup->stage );
}