summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/llvmpipe/lp_setup_tri.c
blob: 0557d35f8b152879f65f68365889a615d7f4d275 (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
/**************************************************************************
 *
 * 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.
 *
 **************************************************************************/

/*
 * Binning code for triangles
 */

#include "util/u_math.h"
#include "util/u_memory.h"
#include "lp_perf.h"
#include "lp_setup_context.h"
#include "lp_rast.h"
#include "lp_state_fs.h"

#define NUM_CHANNELS 4


/**
 * Compute a0 for a constant-valued coefficient (GL_FLAT shading).
 */
static void constant_coef( struct lp_setup_context *setup,
                           struct lp_rast_triangle *tri,
                           unsigned slot,
			   const float value,
                           unsigned i )
{
   tri->inputs.a0[slot][i] = value;
   tri->inputs.dadx[slot][i] = 0.0f;
   tri->inputs.dady[slot][i] = 0.0f;
}


/**
 * Compute a0, dadx and dady for a linearly interpolated coefficient,
 * for a triangle.
 */
static void linear_coef( struct lp_setup_context *setup,
                         struct lp_rast_triangle *tri,
                         float oneoverarea,
                         unsigned slot,
                         const float (*v1)[4],
                         const float (*v2)[4],
                         const float (*v3)[4],
                         unsigned vert_attr,
                         unsigned i)
{
   float a1 = v1[vert_attr][i];
   float a2 = v2[vert_attr][i];
   float a3 = v3[vert_attr][i];

   float da12 = a1 - a2;
   float da31 = a3 - a1;
   float dadx = (da12 * tri->dy31 - tri->dy12 * da31) * oneoverarea;
   float dady = (da31 * tri->dx12 - tri->dx31 * da12) * oneoverarea;

   tri->inputs.dadx[slot][i] = dadx;
   tri->inputs.dady[slot][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 (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.
    */
   tri->inputs.a0[slot][i] = (a1 -
                              (dadx * (v1[0][0] - setup->pixel_offset) +
                               dady * (v1[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.
 */
static void perspective_coef( struct lp_setup_context *setup,
                              struct lp_rast_triangle *tri,
                              float oneoverarea,
                              unsigned slot,
			      const float (*v1)[4],
			      const float (*v2)[4],
			      const float (*v3)[4],
			      unsigned vert_attr,
                              unsigned i)
{
   /* premultiply by 1/w  (v[0][3] is always 1/w):
    */
   float a1 = v1[vert_attr][i] * v1[0][3];
   float a2 = v2[vert_attr][i] * v2[0][3];
   float a3 = v3[vert_attr][i] * v3[0][3];
   float da12 = a1 - a2;
   float da31 = a3 - a1;
   float dadx = (da12 * tri->dy31 - tri->dy12 * da31) * oneoverarea;
   float dady = (da31 * tri->dx12 - tri->dx31 * da12) * oneoverarea;

   tri->inputs.dadx[slot][i] = dadx;
   tri->inputs.dady[slot][i] = dady;
   tri->inputs.a0[slot][i] = (a1 -
                              (dadx * (v1[0][0] - setup->pixel_offset) +
                               dady * (v1[0][1] - setup->pixel_offset)));
}


/**
 * Special coefficient setup for gl_FragCoord.
 * X and Y are trivial
 * Z and W are copied from position_coef 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_coef(struct lp_setup_context *setup,
                     struct lp_rast_triangle *tri,
                     float oneoverarea,
                     unsigned slot,
                     const float (*v1)[4],
                     const float (*v2)[4],
                     const float (*v3)[4],
                     unsigned usage_mask)
{
   /*X*/
   if (usage_mask & TGSI_WRITEMASK_X) {
      tri->inputs.a0[slot][0] = 0.0;
      tri->inputs.dadx[slot][0] = 1.0;
      tri->inputs.dady[slot][0] = 0.0;
   }

   /*Y*/
   if (usage_mask & TGSI_WRITEMASK_Y) {
      tri->inputs.a0[slot][1] = 0.0;
      tri->inputs.dadx[slot][1] = 0.0;
      tri->inputs.dady[slot][1] = 1.0;
   }

   /*Z*/
   if (usage_mask & TGSI_WRITEMASK_Z) {
      linear_coef(setup, tri, oneoverarea, slot, v1, v2, v3, 0, 2);
   }

   /*W*/
   if (usage_mask & TGSI_WRITEMASK_W) {
      linear_coef(setup, tri, oneoverarea, slot, v1, v2, v3, 0, 3);
   }
}


/**
 * Setup the fragment input attribute with the front-facing value.
 * \param frontface  is the triangle front facing?
 */
static void setup_facing_coef( struct lp_setup_context *setup,
                               struct lp_rast_triangle *tri,
                               unsigned slot,
                               boolean frontface,
                               unsigned usage_mask)
{
   /* convert TRUE to 1.0 and FALSE to -1.0 */
   if (usage_mask & TGSI_WRITEMASK_X)
      constant_coef( setup, tri, slot, 2.0f * frontface - 1.0f, 0 );

   if (usage_mask & TGSI_WRITEMASK_Y)
      constant_coef( setup, tri, slot, 0.0f, 1 ); /* wasted */

   if (usage_mask & TGSI_WRITEMASK_Z)
      constant_coef( setup, tri, slot, 0.0f, 2 ); /* wasted */

   if (usage_mask & TGSI_WRITEMASK_W)
      constant_coef( setup, tri, slot, 0.0f, 3 ); /* wasted */
}


/**
 * Compute the tri->coef[] array dadx, dady, a0 values.
 */
static void setup_tri_coefficients( struct lp_setup_context *setup,
				    struct lp_rast_triangle *tri,
                                    float oneoverarea,
				    const float (*v1)[4],
				    const float (*v2)[4],
				    const float (*v3)[4],
				    boolean frontface)
{
   unsigned fragcoord_usage_mask = TGSI_WRITEMASK_XYZ;
   unsigned slot;

   /* setup interpolation for all the remaining attributes:
    */
   for (slot = 0; slot < setup->fs.nr_inputs; slot++) {
      unsigned vert_attr = setup->fs.input[slot].src_index;
      unsigned usage_mask = setup->fs.input[slot].usage_mask;
      unsigned i;

      switch (setup->fs.input[slot].interp) {
      case LP_INTERP_CONSTANT:
         if (setup->flatshade_first) {
            for (i = 0; i < NUM_CHANNELS; i++)
               if (usage_mask & (1 << i))
                  constant_coef(setup, tri, slot+1, v1[vert_attr][i], i);
         }
         else {
            for (i = 0; i < NUM_CHANNELS; i++)
               if (usage_mask & (1 << i))
                  constant_coef(setup, tri, slot+1, v3[vert_attr][i], i);
         }
         break;

      case LP_INTERP_LINEAR:
         for (i = 0; i < NUM_CHANNELS; i++)
            if (usage_mask & (1 << i))
               linear_coef(setup, tri, oneoverarea, slot+1, v1, v2, v3, vert_attr, i);
         break;

      case LP_INTERP_PERSPECTIVE:
         for (i = 0; i < NUM_CHANNELS; i++)
            if (usage_mask & (1 << i))
               perspective_coef(setup, tri, oneoverarea, slot+1, v1, v2, v3, vert_attr, i);
         fragcoord_usage_mask |= TGSI_WRITEMASK_W;
         break;

      case LP_INTERP_POSITION:
         /*
          * The generated pixel interpolators will pick up the coeffs from
          * slot 0, so all need to ensure that the usage mask is covers all
          * usages.
          */
         fragcoord_usage_mask |= usage_mask;
         break;

      case LP_INTERP_FACING:
         setup_facing_coef(setup, tri, slot+1, frontface, usage_mask);
         break;

      default:
         assert(0);
      }
   }

   /* The internal position input is in slot zero:
    */
   setup_fragcoord_coef(setup, tri, oneoverarea, 0, v1, v2, v3,
                        fragcoord_usage_mask);
}



static INLINE int subpixel_snap( float a )
{
   return util_iround(FIXED_ONE * a - (FIXED_ONE / 2));
}



/**
 * Alloc space for a new triangle plus the input.a0/dadx/dady arrays
 * immediately after it.
 * The memory is allocated from the per-scene pool, not per-tile.
 * \param tri_size  returns number of bytes allocated
 * \param nr_inputs  number of fragment shader inputs
 * \return pointer to triangle space
 */
static INLINE struct lp_rast_triangle *
alloc_triangle(struct lp_scene *scene, unsigned nr_inputs, unsigned *tri_size)
{
   unsigned input_array_sz = NUM_CHANNELS * (nr_inputs + 1) * sizeof(float);
   struct lp_rast_triangle *tri;
   unsigned bytes;
   char *inputs;

   assert(sizeof(*tri) % 16 == 0);

   bytes = sizeof(*tri) + (3 * input_array_sz);

   tri = lp_scene_alloc_aligned( scene, bytes, 16 );

   if (tri) {
      inputs = (char *) (tri + 1);
      tri->inputs.a0   = (float (*)[4]) inputs;
      tri->inputs.dadx = (float (*)[4]) (inputs + input_array_sz);
      tri->inputs.dady = (float (*)[4]) (inputs + 2 * input_array_sz);

      *tri_size = bytes;
   }

   return tri;
}


/**
 * Print triangle vertex attribs (for debug).
 */
static void
print_triangle(struct lp_setup_context *setup,
               const float (*v1)[4],
               const float (*v2)[4],
               const float (*v3)[4])
{
   uint i;

   debug_printf("llvmpipe triangle\n");
   for (i = 0; i < setup->fs.nr_inputs; i++) {
      debug_printf("  v1[%d]:  %f %f %f %f\n", i,
                   v1[i][0], v1[i][1], v1[i][2], v1[i][3]);
   }
   for (i = 0; i < setup->fs.nr_inputs; i++) {
      debug_printf("  v2[%d]:  %f %f %f %f\n", i,
                   v2[i][0], v2[i][1], v2[i][2], v2[i][3]);
   }
   for (i = 0; i < setup->fs.nr_inputs; i++) {
      debug_printf("  v3[%d]:  %f %f %f %f\n", i,
                   v3[i][0], v3[i][1], v3[i][2], v3[i][3]);
   }
}


/**
 * Do basic setup for triangle rasterization and determine which
 * framebuffer tiles are touched.  Put the triangle in the scene's
 * bins for the tiles which we overlap.
 */
static void 
do_triangle_ccw(struct lp_setup_context *setup,
		const float (*v1)[4],
		const float (*v2)[4],
		const float (*v3)[4],
		boolean frontfacing )
{
   /* x/y positions in fixed point */
   const int x1 = subpixel_snap(v1[0][0] + 0.5 - setup->pixel_offset);
   const int x2 = subpixel_snap(v2[0][0] + 0.5 - setup->pixel_offset);
   const int x3 = subpixel_snap(v3[0][0] + 0.5 - setup->pixel_offset);
   const int y1 = subpixel_snap(v1[0][1] + 0.5 - setup->pixel_offset);
   const int y2 = subpixel_snap(v2[0][1] + 0.5 - setup->pixel_offset);
   const int y3 = subpixel_snap(v3[0][1] + 0.5 - setup->pixel_offset);

   struct lp_scene *scene = lp_setup_get_current_scene(setup);
   struct lp_rast_triangle *tri;
   int area;
   float oneoverarea;
   int minx, maxx, miny, maxy;
   unsigned tri_bytes;

   if (0)
      print_triangle(setup, v1, v2, v3);

   tri = alloc_triangle(scene, setup->fs.nr_inputs, &tri_bytes);
   if (!tri)
      return;

#ifdef DEBUG
   tri->v[0][0] = v1[0][0];
   tri->v[1][0] = v2[0][0];
   tri->v[2][0] = v3[0][0];
   tri->v[0][1] = v1[0][1];
   tri->v[1][1] = v2[0][1];
   tri->v[2][1] = v3[0][1];
#endif

   tri->dx12 = x1 - x2;
   tri->dx23 = x2 - x3;
   tri->dx31 = x3 - x1;

   tri->dy12 = y1 - y2;
   tri->dy23 = y2 - y3;
   tri->dy31 = y3 - y1;

   area = (tri->dx12 * tri->dy31 - tri->dx31 * tri->dy12);

   LP_COUNT(nr_tris);

   /* Cull non-ccw and zero-sized triangles. 
    *
    * XXX: subject to overflow??
    */
   if (area <= 0) {
      lp_scene_putback_data( scene, tri_bytes );
      LP_COUNT(nr_culled_tris);
      return;
   }

   /* Bounding rectangle (in pixels) */
   minx = (MIN3(x1, x2, x3) + (FIXED_ONE-1)) >> FIXED_ORDER;
   maxx = (MAX3(x1, x2, x3) + (FIXED_ONE-1)) >> FIXED_ORDER;
   miny = (MIN3(y1, y2, y3) + (FIXED_ONE-1)) >> FIXED_ORDER;
   maxy = (MAX3(y1, y2, y3) + (FIXED_ONE-1)) >> FIXED_ORDER;
   
   if (setup->scissor_test) {
      minx = MAX2(minx, setup->scissor.current.minx);
      maxx = MIN2(maxx, setup->scissor.current.maxx);
      miny = MAX2(miny, setup->scissor.current.miny);
      maxy = MIN2(maxy, setup->scissor.current.maxy);
   }

   if (miny == maxy || 
       minx == maxx) {
      lp_scene_putback_data( scene, tri_bytes );
      LP_COUNT(nr_culled_tris);
      return;
   }

   /* 
    */
   oneoverarea = ((float)FIXED_ONE) / (float)area;

   /* Setup parameter interpolants:
    */
   setup_tri_coefficients( setup, tri, oneoverarea, v1, v2, v3, frontfacing );

   tri->inputs.facing = frontfacing ? 1.0F : -1.0F;

   /* half-edge constants, will be interated over the whole render target.
    */
   tri->c1 = tri->dy12 * x1 - tri->dx12 * y1;
   tri->c2 = tri->dy23 * x2 - tri->dx23 * y2;
   tri->c3 = tri->dy31 * x3 - tri->dx31 * y3;

   /* correct for top-left fill convention:
    */
   if (tri->dy12 < 0 || (tri->dy12 == 0 && tri->dx12 > 0)) tri->c1++;
   if (tri->dy23 < 0 || (tri->dy23 == 0 && tri->dx23 > 0)) tri->c2++;
   if (tri->dy31 < 0 || (tri->dy31 == 0 && tri->dx31 > 0)) tri->c3++;

   tri->dy12 *= FIXED_ONE;
   tri->dy23 *= FIXED_ONE;
   tri->dy31 *= FIXED_ONE;

   tri->dx12 *= FIXED_ONE;
   tri->dx23 *= FIXED_ONE;
   tri->dx31 *= FIXED_ONE;

   /* find trivial reject offsets for each edge for a single-pixel
    * sized block.  These will be scaled up at each recursive level to
    * match the active blocksize.  Scaling in this way works best if
    * the blocks are square.
    */
   tri->eo1 = 0;
   if (tri->dy12 < 0) tri->eo1 -= tri->dy12;
   if (tri->dx12 > 0) tri->eo1 += tri->dx12;

   tri->eo2 = 0;
   if (tri->dy23 < 0) tri->eo2 -= tri->dy23;
   if (tri->dx23 > 0) tri->eo2 += tri->dx23;

   tri->eo3 = 0;
   if (tri->dy31 < 0) tri->eo3 -= tri->dy31;
   if (tri->dx31 > 0) tri->eo3 += tri->dx31;

   /* Calculate trivial accept offsets from the above.
    */
   tri->ei1 = tri->dx12 - tri->dy12 - tri->eo1;
   tri->ei2 = tri->dx23 - tri->dy23 - tri->eo2;
   tri->ei3 = tri->dx31 - tri->dy31 - tri->eo3;

   /* Fill in the inputs.step[][] arrays.
    * We've manually unrolled some loops here.
    */
   {
      const int xstep1 = -tri->dy12;
      const int xstep2 = -tri->dy23;
      const int xstep3 = -tri->dy31;
      const int ystep1 = tri->dx12;
      const int ystep2 = tri->dx23;
      const int ystep3 = tri->dx31;

#define SETUP_STEP(i, x, y)                                \
      do {                                                 \
         tri->inputs.step[0][i] = x * xstep1 + y * ystep1; \
         tri->inputs.step[1][i] = x * xstep2 + y * ystep2; \
         tri->inputs.step[2][i] = x * xstep3 + y * ystep3; \
      } while (0)

      SETUP_STEP(0, 0, 0);
      SETUP_STEP(1, 1, 0);
      SETUP_STEP(2, 0, 1);
      SETUP_STEP(3, 1, 1);

      SETUP_STEP(4, 2, 0);
      SETUP_STEP(5, 3, 0);
      SETUP_STEP(6, 2, 1);
      SETUP_STEP(7, 3, 1);

      SETUP_STEP(8, 0, 2);
      SETUP_STEP(9, 1, 2);
      SETUP_STEP(10, 0, 3);
      SETUP_STEP(11, 1, 3);

      SETUP_STEP(12, 2, 2);
      SETUP_STEP(13, 3, 2);
      SETUP_STEP(14, 2, 3);
      SETUP_STEP(15, 3, 3);
#undef STEP
   }

   /*
    * All fields of 'tri' are now set.  The remaining code here is
    * concerned with binning.
    */

   /* Convert to tile coordinates:
    */
   minx = minx / TILE_SIZE;
   miny = miny / TILE_SIZE;
   maxx = maxx / TILE_SIZE;
   maxy = maxy / TILE_SIZE;

   /*
    * Clamp to framebuffer size
    */
   minx = MAX2(minx, 0);
   miny = MAX2(miny, 0);
   maxx = MIN2(maxx, scene->tiles_x - 1);
   maxy = MIN2(maxy, scene->tiles_y - 1);

   /* Determine which tile(s) intersect the triangle's bounding box
    */
   if (miny == maxy && minx == maxx)
   {
      /* Triangle is contained in a single tile:
       */
      lp_scene_bin_command( scene, minx, miny, lp_rast_triangle, 
			    lp_rast_arg_triangle(tri) );
   }
   else 
   {
      int c1 = (tri->c1 + 
                tri->dx12 * miny * TILE_SIZE - 
                tri->dy12 * minx * TILE_SIZE);
      int c2 = (tri->c2 + 
                tri->dx23 * miny * TILE_SIZE -
                tri->dy23 * minx * TILE_SIZE);
      int c3 = (tri->c3 +
                tri->dx31 * miny * TILE_SIZE -
                tri->dy31 * minx * TILE_SIZE);

      int ei1 = tri->ei1 << TILE_ORDER;
      int ei2 = tri->ei2 << TILE_ORDER;
      int ei3 = tri->ei3 << TILE_ORDER;

      int eo1 = tri->eo1 << TILE_ORDER;
      int eo2 = tri->eo2 << TILE_ORDER;
      int eo3 = tri->eo3 << TILE_ORDER;

      int xstep1 = -(tri->dy12 << TILE_ORDER);
      int xstep2 = -(tri->dy23 << TILE_ORDER);
      int xstep3 = -(tri->dy31 << TILE_ORDER);

      int ystep1 = tri->dx12 << TILE_ORDER;
      int ystep2 = tri->dx23 << TILE_ORDER;
      int ystep3 = tri->dx31 << TILE_ORDER;
      int x, y;


      /* Test tile-sized blocks against the triangle.
       * Discard blocks fully outside the tri.  If the block is fully
       * contained inside the tri, bin an lp_rast_shade_tile command.
       * Else, bin a lp_rast_triangle command.
       */
      for (y = miny; y <= maxy; y++)
      {
	 int cx1 = c1;
	 int cx2 = c2;
	 int cx3 = c3;
	 boolean in = FALSE;  /* are we inside the triangle? */

	 for (x = minx; x <= maxx; x++)
	 {
	    if (cx1 + eo1 < 0 || 
		cx2 + eo2 < 0 ||
		cx3 + eo3 < 0) 
	    {
	       /* do nothing */
               LP_COUNT(nr_empty_64);
	       if (in)
		  break;  /* exiting triangle, all done with this row */
	    }
	    else if (cx1 + ei1 > 0 &&
		     cx2 + ei2 > 0 &&
		     cx3 + ei3 > 0) 
	    {
               /* triangle covers the whole tile- shade whole tile */
               LP_COUNT(nr_fully_covered_64);
	       in = TRUE;
	       if (setup->fs.current.variant->opaque) {
	          lp_scene_bin_reset( scene, x, y );
	          lp_scene_bin_command( scene, x, y,
	                                lp_rast_set_state,
	                                lp_rast_arg_state(setup->fs.stored) );
	       }
               lp_scene_bin_command( scene, x, y,
				     lp_rast_shade_tile,
				     lp_rast_arg_inputs(&tri->inputs) );
	    }
	    else 
	    { 
               /* rasterizer/shade partial tile */
               LP_COUNT(nr_partially_covered_64);
	       in = TRUE;
               lp_scene_bin_command( scene, x, y,
				     lp_rast_triangle, 
				     lp_rast_arg_triangle(tri) );
	    }

	    /* Iterate cx values across the region:
	     */
	    cx1 += xstep1;
	    cx2 += xstep2;
	    cx3 += xstep3;
	 }
      
	 /* Iterate c values down the region:
	  */
	 c1 += ystep1;
	 c2 += ystep2;
	 c3 += ystep3;    
      }
   }
}


/**
 * Draw triangle if it's CW, cull otherwise.
 */
static void triangle_cw( struct lp_setup_context *setup,
			 const float (*v0)[4],
			 const float (*v1)[4],
			 const float (*v2)[4] )
{
   do_triangle_ccw( setup, v1, v0, v2, !setup->ccw_is_frontface );
}


/**
 * Draw triangle if it's CCW, cull otherwise.
 */
static void triangle_ccw( struct lp_setup_context *setup,
			 const float (*v0)[4],
			 const float (*v1)[4],
			 const float (*v2)[4] )
{
   do_triangle_ccw( setup, v0, v1, v2, setup->ccw_is_frontface );
}



/**
 * Draw triangle whether it's CW or CCW.
 */
static void triangle_both( struct lp_setup_context *setup,
			   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 */
   if (ex * fy - ey * fx < 0.0f) 
      triangle_ccw( setup, v0, v1, v2 );
   else
      triangle_cw( setup, v0, v1, v2 );
}


static void triangle_nop( struct lp_setup_context *setup,
			  const float (*v0)[4],
			  const float (*v1)[4],
			  const float (*v2)[4] )
{
}


void 
lp_setup_choose_triangle( struct lp_setup_context *setup )
{
   switch (setup->cullmode) {
   case PIPE_FACE_NONE:
      setup->triangle = triangle_both;
      break;
   case PIPE_FACE_BACK:
      setup->triangle = setup->ccw_is_frontface ? triangle_ccw : triangle_cw;
      break;
   case PIPE_FACE_FRONT:
      setup->triangle = setup->ccw_is_frontface ? triangle_cw : triangle_ccw;
      break;
   default:
      setup->triangle = triangle_nop;
      break;
   }
}