summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/llvmpipe/lp_rast.c
blob: 9020cf9cec2c9ae5df8d95cd376f0fd730101ee3 (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
/**************************************************************************
 *
 * Copyright 2009 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.
 *
 **************************************************************************/

#include "util/u_memory.h"
#include "util/u_math.h"
#include "util/u_cpu_detect.h"
#include "util/u_surface.h"

#include "lp_bin_queue.h"
#include "lp_debug.h"
#include "lp_state.h"
#include "lp_rast.h"
#include "lp_rast_priv.h"
#include "lp_tile_soa.h"
#include "lp_bld_debug.h"
#include "lp_bin.h"


/**
 * Begin the rasterization phase.
 * Map the framebuffer surfaces.  Initialize the 'rast' state.
 */
static boolean
lp_rast_begin( struct lp_rasterizer *rast,
               const struct pipe_framebuffer_state *fb,
               boolean write_color,
               boolean write_zstencil )
{
   struct pipe_screen *screen = rast->screen;
   struct pipe_surface *cbuf, *zsbuf;

   LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);

   util_copy_framebuffer_state(&rast->state.fb, fb);

   rast->state.write_zstencil = write_zstencil;
   rast->state.write_color = write_color;

   rast->check_for_clipped_tiles = (fb->width % TILE_SIZE != 0 ||
                                    fb->height % TILE_SIZE != 0);

   /* XXX support multiple color buffers here */
   cbuf = rast->state.fb.cbufs[0];
   if (cbuf) {
      rast->cbuf_transfer = screen->get_tex_transfer(rast->screen,
                                                     cbuf->texture,
                                                     cbuf->face,
                                                     cbuf->level,
                                                     cbuf->zslice,
                                                     PIPE_TRANSFER_READ_WRITE,
                                                     0, 0,
                                                     fb->width, fb->height);
      if (!rast->cbuf_transfer)
         return FALSE;

      rast->cbuf_map = screen->transfer_map(rast->screen, 
                                            rast->cbuf_transfer);
      if (!rast->cbuf_map)
         return FALSE;
   }

   zsbuf = rast->state.fb.zsbuf;
   if (zsbuf) {
      rast->zsbuf_transfer = screen->get_tex_transfer(rast->screen,
                                                      zsbuf->texture,
                                                      zsbuf->face,
                                                      zsbuf->level,
                                                      zsbuf->zslice,
                                                      PIPE_TRANSFER_READ_WRITE,
                                                      0, 0,
                                                      fb->width, fb->height);
      if (!rast->zsbuf_transfer)
         return FALSE;

      rast->zsbuf_map = screen->transfer_map(rast->screen, 
                                            rast->zsbuf_transfer);
      if (!rast->zsbuf_map)
         return FALSE;
   }

   return TRUE;
}


/**
 * Finish the rasterization phase.
 * Unmap framebuffer surfaces.
 */
static void
lp_rast_end( struct lp_rasterizer *rast )
{
   struct pipe_screen *screen = rast->screen;

   if (rast->cbuf_map) 
      screen->transfer_unmap(screen, rast->cbuf_transfer);

   if (rast->zsbuf_map) 
      screen->transfer_unmap(screen, rast->zsbuf_transfer);

   if (rast->cbuf_transfer)
      screen->tex_transfer_destroy(rast->cbuf_transfer);

   if (rast->zsbuf_transfer)
      screen->tex_transfer_destroy(rast->zsbuf_transfer);

   rast->cbuf_transfer = NULL;
   rast->zsbuf_transfer = NULL;
   rast->cbuf_map = NULL;
   rast->zsbuf_map = NULL;
}


/**
 * Begining rasterization of a tile.
 * \param x  window X position of the tile, in pixels
 * \param y  window Y position of the tile, in pixels
 */
static void
lp_rast_start_tile( struct lp_rasterizer *rast,
                    unsigned thread_index,
                    unsigned x, unsigned y )
{
   LP_DBG(DEBUG_RAST, "%s %d,%d\n", __FUNCTION__, x, y);

   rast->tasks[thread_index].x = x;
   rast->tasks[thread_index].y = y;
}


/**
 * Clear the rasterizer's current color tile.
 * This is a bin command called during bin processing.
 */
void lp_rast_clear_color( struct lp_rasterizer *rast,
                          unsigned thread_index,
                          const union lp_rast_cmd_arg arg )
{
   const uint8_t *clear_color = arg.clear_color;
   uint8_t *color_tile = rast->tasks[thread_index].tile.color;
   
   LP_DBG(DEBUG_RAST, "%s 0x%x,0x%x,0x%x,0x%x\n", __FUNCTION__, 
              clear_color[0],
              clear_color[1],
              clear_color[2],
              clear_color[3]);

   if (clear_color[0] == clear_color[1] &&
       clear_color[1] == clear_color[2] &&
       clear_color[2] == clear_color[3]) {
      memset(color_tile, clear_color[0], TILE_SIZE * TILE_SIZE * 4);
   }
   else {
      unsigned x, y, chan;
      for (y = 0; y < TILE_SIZE; y++)
         for (x = 0; x < TILE_SIZE; x++)
            for (chan = 0; chan < 4; ++chan)
               TILE_PIXEL(color_tile, x, y, chan) = clear_color[chan];
   }
}


/**
 * Clear the rasterizer's current z/stencil tile.
 * This is a bin command called during bin processing.
 */
void lp_rast_clear_zstencil( struct lp_rasterizer *rast,
                             unsigned thread_index,
                             const union lp_rast_cmd_arg arg)
{
   unsigned i, j;
   uint32_t *depth_tile = rast->tasks[thread_index].tile.depth;
   
   LP_DBG(DEBUG_RAST, "%s 0x%x\n", __FUNCTION__, arg.clear_zstencil);

   for (i = 0; i < TILE_SIZE; i++)
      for (j = 0; j < TILE_SIZE; j++)
	 depth_tile[i*TILE_SIZE + j] = arg.clear_zstencil;
}


/**
 * Load tile color from the framebuffer surface.
 * This is a bin command called during bin processing.
 */
void lp_rast_load_color( struct lp_rasterizer *rast,
                         unsigned thread_index,
                         const union lp_rast_cmd_arg arg)
{
   struct lp_rasterizer_task *task = &rast->tasks[thread_index];
   const unsigned x = task->x;
   const unsigned y = task->y;
   int w = TILE_SIZE;
   int h = TILE_SIZE;

   LP_DBG(DEBUG_RAST, "%s at %u, %u\n", __FUNCTION__, x, y);

   if (x + w > rast->state.fb.width)
      w -= x + w - rast->state.fb.width;

   if (y + h > rast->state.fb.height)
      h -= y + h - rast->state.fb.height;

   assert(w >= 0);
   assert(h >= 0);
   assert(w <= TILE_SIZE);
   assert(h <= TILE_SIZE);

   lp_tile_read_4ub(rast->cbuf_transfer->format,
                     rast->tasks[thread_index].tile.color,
                     rast->cbuf_map, 
                     rast->cbuf_transfer->stride,
                     x, y,
                     w, h);
}


/**
 * Load tile z/stencil from the framebuffer surface.
 * This is a bin command called during bin processing.
 */
void lp_rast_load_zstencil( struct lp_rasterizer *rast,
                            unsigned thread_index,
                            const union lp_rast_cmd_arg arg )
{
   LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);

   /* call u_tile func to load depth (and stencil?) from surface */
}


void lp_rast_set_state( struct lp_rasterizer *rast,
                        unsigned thread_index,
                        const union lp_rast_cmd_arg arg )
{
   const struct lp_rast_state *state = arg.set_state;

   LP_DBG(DEBUG_RAST, "%s %p\n", __FUNCTION__, (void *) state);

   /* just set the current state pointer for this rasterizer */
   rast->tasks[thread_index].current_state = state;
}



/* Within a tile:
 */

/**
 * Run the shader on all blocks in a tile.  This is used when a tile is
 * completely contained inside a triangle.
 * This is a bin command called during bin processing.
 */
void lp_rast_shade_tile( struct lp_rasterizer *rast,
                         unsigned thread_index,
                         const union lp_rast_cmd_arg arg )
{
   const struct lp_rast_shader_inputs *inputs = arg.shade_tile;
   const unsigned tile_x = rast->tasks[thread_index].x;
   const unsigned tile_y = rast->tasks[thread_index].y;
   const unsigned mask = ~0;
   unsigned x, y;

   LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);

   /* Use the existing preference for 4x4 (four quads) shading:
    */
   for (y = 0; y < TILE_SIZE; y += 4)
      for (x = 0; x < TILE_SIZE; x += 4)
         lp_rast_shade_quads( rast,
                              thread_index,
                              inputs,
                              tile_x + x,
                              tile_y + y,
                              mask);
}


/**
 * Compute shading for a 4x4 block of pixels.
 * This is a bin command called during bin processing.
 */
void lp_rast_shade_quads( struct lp_rasterizer *rast,
                          unsigned thread_index,
                          const struct lp_rast_shader_inputs *inputs,
                          unsigned x, unsigned y,
                          unsigned mask)
{
#if 1
   const struct lp_rast_state *state = rast->tasks[thread_index].current_state;
   struct lp_rast_tile *tile = &rast->tasks[thread_index].tile;
   void *color;
   void *depth;
   uint32_t ALIGN16_ATTRIB masks[2][2][2][2];
   unsigned ix, iy;
   int block_offset;

   assert(state);

   /* Sanity checks */
   assert(x % TILE_VECTOR_WIDTH == 0);
   assert(y % TILE_VECTOR_HEIGHT == 0);

   /* mask: the rasterizer wants to treat pixels in 4x4 blocks, but
    * the pixel shader wants to swizzle them into 4 2x2 quads.
    * 
    * Additionally, the pixel shader wants masks as full dword ~0,
    * while the rasterizer wants to pack per-pixel bits tightly.
    */
#if 0
   unsigned qx, qy;
   for (qy = 0; qy < 2; ++qy)
      for (qx = 0; qx < 2; ++qx)
	 for (iy = 0; iy < 2; ++iy)
	    for (ix = 0; ix < 2; ++ix)
	       masks[qy][qx][iy][ix] = mask & (1 << (qy*8+iy*4+qx*2+ix)) ? ~0 : 0;
#else
   masks[0][0][0][0] = mask & (1 << (0*8+0*4+0*2+0)) ? ~0 : 0;
   masks[0][0][0][1] = mask & (1 << (0*8+0*4+0*2+1)) ? ~0 : 0;
   masks[0][0][1][0] = mask & (1 << (0*8+1*4+0*2+0)) ? ~0 : 0;
   masks[0][0][1][1] = mask & (1 << (0*8+1*4+0*2+1)) ? ~0 : 0;
   masks[0][1][0][0] = mask & (1 << (0*8+0*4+1*2+0)) ? ~0 : 0;
   masks[0][1][0][1] = mask & (1 << (0*8+0*4+1*2+1)) ? ~0 : 0;
   masks[0][1][1][0] = mask & (1 << (0*8+1*4+1*2+0)) ? ~0 : 0;
   masks[0][1][1][1] = mask & (1 << (0*8+1*4+1*2+1)) ? ~0 : 0;

   masks[1][0][0][0] = mask & (1 << (1*8+0*4+0*2+0)) ? ~0 : 0;
   masks[1][0][0][1] = mask & (1 << (1*8+0*4+0*2+1)) ? ~0 : 0;
   masks[1][0][1][0] = mask & (1 << (1*8+1*4+0*2+0)) ? ~0 : 0;
   masks[1][0][1][1] = mask & (1 << (1*8+1*4+0*2+1)) ? ~0 : 0;
   masks[1][1][0][0] = mask & (1 << (1*8+0*4+1*2+0)) ? ~0 : 0;
   masks[1][1][0][1] = mask & (1 << (1*8+0*4+1*2+1)) ? ~0 : 0;
   masks[1][1][1][0] = mask & (1 << (1*8+1*4+1*2+0)) ? ~0 : 0;
   masks[1][1][1][1] = mask & (1 << (1*8+1*4+1*2+1)) ? ~0 : 0;
#endif

   assert((x % 2) == 0);
   assert((y % 2) == 0);

   ix = x % TILE_SIZE;
   iy = y % TILE_SIZE;

   /* offset of the 16x16 pixel block within the tile */
   block_offset = ((iy/4)*(16*16) + (ix/4)*16);

   /* color buffer */
   color = tile->color + 4 * block_offset;

   /* depth buffer */
   depth = tile->depth + block_offset;

   /* XXX: This will most likely fail on 32bit x86 without -mstackrealign */
   assert(lp_check_alignment(masks, 16));

   assert(lp_check_alignment(depth, 16));
   assert(lp_check_alignment(color, 16));
   assert(lp_check_alignment(state->jit_context.blend_color, 16));

   /* run shader */
   state->jit_function( &state->jit_context,
                        x, y,
                        inputs->a0,
                        inputs->dadx,
                        inputs->dady,
                        &masks[0][0][0][0],
                        color,
                        depth);
#else
   struct lp_rast_tile *tile = &rast->tile;
   unsigned chan_index;
   unsigned q, ix, iy;

   x %= TILE_SIZE;
   y %= TILE_SIZE;

   /* mask */
   for (q = 0; q < 4; ++q)
      for(iy = 0; iy < 2; ++iy)
         for(ix = 0; ix < 2; ++ix)
            if(masks[q] & (1 << (iy*2 + ix)))
               for (chan_index = 0; chan_index < NUM_CHANNELS; ++chan_index)
                  TILE_PIXEL(tile->color, x + q*2 + ix, y + iy, chan_index) = 0xff;

#endif
}


/* End of tile:
 */


/**
 * Write the rasterizer's color tile to the framebuffer.
 */
static void lp_rast_store_color( struct lp_rasterizer *rast,
                                 unsigned thread_index)
{
   const unsigned x = rast->tasks[thread_index].x;
   const unsigned y = rast->tasks[thread_index].y;
   int w = TILE_SIZE;
   int h = TILE_SIZE;

   if (x + w > rast->state.fb.width)
      w -= x + w - rast->state.fb.width;

   if (y + h > rast->state.fb.height)
      h -= y + h - rast->state.fb.height;

   assert(w >= 0);
   assert(h >= 0);
   assert(w <= TILE_SIZE);
   assert(h <= TILE_SIZE);

   LP_DBG(DEBUG_RAST, "%s [%u] %d,%d %dx%d\n", __FUNCTION__,
          thread_index, x, y, w, h);

   lp_tile_write_4ub(rast->cbuf_transfer->format,
                     rast->tasks[thread_index].tile.color,
                     rast->cbuf_map, 
                     rast->cbuf_transfer->stride,
                     x, y,
                     w, h);
}


static void
lp_tile_write_z32(const uint32_t *src, uint8_t *dst, unsigned dst_stride,
                  unsigned x0, unsigned y0, unsigned w, unsigned h)
{
   unsigned x, y;
   uint8_t *dst_row = dst + y0*dst_stride;
   for (y = 0; y < h; ++y) {
      uint32_t *dst_pixel = (uint32_t *)(dst_row + x0*4);
      for (x = 0; x < w; ++x) {
         *dst_pixel++ = *src++;
      }
      dst_row += dst_stride;
   }
}

/**
 * Write the rasterizer's z/stencil tile to the framebuffer.
 */
static void lp_rast_store_zstencil( struct lp_rasterizer *rast,
                                    unsigned thread_index )
{
   const unsigned x = rast->tasks[thread_index].x;
   const unsigned y = rast->tasks[thread_index].y;
   unsigned w = TILE_SIZE;
   unsigned h = TILE_SIZE;

   if (x + w > rast->state.fb.width)
      w -= x + w - rast->state.fb.width;

   if (y + h > rast->state.fb.height)
      h -= y + h - rast->state.fb.height;

   LP_DBG(DEBUG_RAST, "%s %d,%d %dx%d\n", __FUNCTION__, x, y, w, h);

   assert(rast->zsbuf_transfer->format == PIPE_FORMAT_Z32_UNORM);
   lp_tile_write_z32(rast->tasks[thread_index].tile.depth,
                     rast->zsbuf_map, 
                     rast->zsbuf_transfer->stride,
                     x, y, w, h);
}


/**
 * Write the rasterizer's tiles to the framebuffer.
 */
static void
lp_rast_end_tile( struct lp_rasterizer *rast,
                  unsigned thread_index )
{
   LP_DBG(DEBUG_RAST, "%s\n", __FUNCTION__);

   if (rast->state.write_color)
      lp_rast_store_color(rast, thread_index);

   if (rast->state.write_zstencil)
      lp_rast_store_zstencil(rast, thread_index);
}


/**
 * When all the threads are done rasterizing a bin, one thread will
 * call this function to reset the bin and put it onto the empty queue.
 */
static void
release_bins( struct lp_rasterizer *rast,
              struct lp_bins *bins )
{
   util_unreference_framebuffer_state( &bins->fb );

   lp_reset_bins( bins );
   lp_bins_enqueue( rast->empty_bins, bins );
   rast->curr_bins = NULL;
}


/**
 * Rasterize commands for a single bin.
 * \param x, y  position of the bin's tile in the framebuffer
 * Must be called between lp_rast_begin() and lp_rast_end().
 * Called per thread.
 */
static void
rasterize_bin( struct lp_rasterizer *rast,
               unsigned thread_index,
               const struct cmd_bin *bin,
               int x, int y)
{
   const struct cmd_block_list *commands = &bin->commands;
   struct cmd_block *block;
   unsigned k;

   lp_rast_start_tile( rast, thread_index, x, y );

   /* simply execute each of the commands in the block list */
   for (block = commands->head; block; block = block->next) {
      for (k = 0; k < block->count; k++) {
         block->cmd[k]( rast, thread_index, block->arg[k] );
      }
   }

   lp_rast_end_tile( rast, thread_index );
}


/**
 * Rasterize/execute all bins.
 * Called per thread.
 */
static void
rasterize_bins( struct lp_rasterizer *rast,
                unsigned thread_index,
                struct lp_bins *bins,
                bool write_depth )
{
   /* loop over tile bins, rasterize each */
#if 0
   {
      unsigned i, j;
      for (i = 0; i < bins->tiles_x; i++) {
         for (j = 0; j < bins->tiles_y; j++) {
            struct cmd_bin *bin = lp_get_bin(bins, i, j);
            rasterize_bin( rast, thread_index,
                           bin, i * TILE_SIZE, j * TILE_SIZE );
         }
      }
   }
#else
   {
      struct cmd_bin *bin;
      int x, y;

      assert(bins);
      while ((bin = lp_bin_iter_next(bins, &x, &y))) {
         rasterize_bin( rast, thread_index, bin, x * TILE_SIZE, y * TILE_SIZE);
      }
   }
#endif
}


/**
 * Called by setup module when it has something for us to render.
 */
void
lp_rasterize_bins( struct lp_rasterizer *rast,
                   struct lp_bins *bins,
                   const struct pipe_framebuffer_state *fb,
                   bool write_depth )
{
   boolean debug = false;

   LP_DBG(DEBUG_SETUP, "%s\n", __FUNCTION__);

   if (debug) {
      unsigned x, y;
      printf("rasterize bins:\n");
      printf("  data size: %u\n", lp_bin_data_size(bins));
      for (y = 0; y < bins->tiles_y; y++) {
         for (x = 0; x < bins->tiles_x; x++) {
            printf("  bin %u, %u size: %u\n", x, y,
                   lp_bin_cmd_size(bins, x, y));
         }
      }
   }

   /* save framebuffer state in the bin */
   util_copy_framebuffer_state(&bins->fb, fb);
   bins->write_depth = write_depth;

   if (rast->num_threads == 0) {
      /* no threading */

      lp_rast_begin( rast, fb,
                     fb->cbufs[0]!= NULL,
                     fb->zsbuf != NULL && write_depth );

      lp_bin_iter_begin( bins );
      rasterize_bins( rast, 0, bins, write_depth );

      release_bins( rast, bins );

      lp_rast_end( rast );
   }
   else {
      /* threaded rendering! */
      unsigned i;

      lp_bins_enqueue( rast->full_bins, bins );

      /* signal the threads that there's work to do */
      for (i = 0; i < rast->num_threads; i++) {
         pipe_semaphore_signal(&rast->tasks[i].work_ready);
      }

      /* wait for work to complete */
      for (i = 0; i < rast->num_threads; i++) {
         pipe_semaphore_wait(&rast->tasks[i].work_done);
      }
   }

   LP_DBG(DEBUG_SETUP, "%s done \n", __FUNCTION__);
}


/**
 * This is the thread's main entrypoint.
 * It's a simple loop:
 *   1. wait for work
 *   2. do work
 *   3. signal that we're done
 */
static void *
thread_func( void *init_data )
{
   struct lp_rasterizer_task *task = (struct lp_rasterizer_task *) init_data;
   struct lp_rasterizer *rast = task->rast;
   boolean debug = false;

   while (1) {
      /* wait for work */
      if (debug)
         debug_printf("thread %d waiting for work\n", task->thread_index);
      pipe_semaphore_wait(&task->work_ready);

      if (task->thread_index == 0) {
         /* thread[0]:
          *  - get next set of bins to rasterize
          *  - map the framebuffer surfaces
          */
         const struct pipe_framebuffer_state *fb;
         boolean write_depth;

         rast->curr_bins = lp_bins_dequeue( rast->full_bins );

         lp_bin_iter_begin( rast->curr_bins );

         fb = &rast->curr_bins->fb;
         write_depth = rast->curr_bins->write_depth;

         lp_rast_begin( rast, fb,
                        fb->cbufs[0] != NULL,
                        fb->zsbuf != NULL && write_depth );
      }

      /* Wait for all threads to get here so that threads[1+] don't
       * get a null rast->curr_bins pointer.
       */
      pipe_barrier_wait( &rast->barrier );

      /* do work */
      if (debug)
         debug_printf("thread %d doing work\n", task->thread_index);
      rasterize_bins(rast, task->thread_index,
                     rast->curr_bins, rast->curr_bins->write_depth);
      
      /* wait for all threads to finish with this set of bins */
      pipe_barrier_wait( &rast->barrier );

      if (task->thread_index == 0) {
         /* thread[0]:
          * - release the bins object
          * - unmap the framebuffer surfaces
          */
         release_bins( rast, rast->curr_bins );
         lp_rast_end( rast );
      }

      /* signal done with work */
      if (debug)
         debug_printf("thread %d done working\n", task->thread_index);
      pipe_semaphore_signal(&task->work_done);
   }

   return NULL;
}


/**
 * Initialize semaphores and spawn the threads.
 */
static void
create_rast_threads(struct lp_rasterizer *rast)
{
   unsigned i;

   rast->num_threads = util_cpu_caps.nr_cpus;
   rast->num_threads = debug_get_num_option("LP_NUM_THREADS", rast->num_threads);
   rast->num_threads = MIN2(rast->num_threads, MAX_THREADS);

   /* NOTE: if num_threads is zero, we won't use any threads */
   for (i = 0; i < rast->num_threads; i++) {
      pipe_semaphore_init(&rast->tasks[i].work_ready, 0);
      pipe_semaphore_init(&rast->tasks[i].work_done, 0);
      rast->threads[i] = pipe_thread_create(thread_func,
                                            (void *) &rast->tasks[i]);
   }
}



/**
 * Create new lp_rasterizer.
 * \param empty  the queue to put empty bins on after we've finished
 *               processing them.
 */
struct lp_rasterizer *
lp_rast_create( struct pipe_screen *screen, struct lp_bins_queue *empty )
{
   struct lp_rasterizer *rast;
   unsigned i;

   rast = CALLOC_STRUCT(lp_rasterizer);
   if(!rast)
      return NULL;

   rast->screen = screen;

   rast->empty_bins = empty;
   rast->full_bins = lp_bins_queue_create();

   for (i = 0; i < Elements(rast->tasks); i++) {
      rast->tasks[i].tile.color = align_malloc( TILE_SIZE*TILE_SIZE*4, 16 );
      rast->tasks[i].tile.depth = align_malloc( TILE_SIZE*TILE_SIZE*4, 16 );
      rast->tasks[i].rast = rast;
      rast->tasks[i].thread_index = i;
   }

   create_rast_threads(rast);

   /* for synchronizing rasterization threads */
   pipe_barrier_init( &rast->barrier, rast->num_threads );

   return rast;
}


/* Shutdown:
 */
void lp_rast_destroy( struct lp_rasterizer *rast )
{
   unsigned i;

   util_unreference_framebuffer_state(&rast->state.fb);

   for (i = 0; i < Elements(rast->tasks); i++) {
      align_free(rast->tasks[i].tile.depth);
      align_free(rast->tasks[i].tile.color);
   }

   /* for synchronizing rasterization threads */
   pipe_barrier_destroy( &rast->barrier );

   FREE(rast);
}