summaryrefslogtreecommitdiff
path: root/src/gallium/drivers/llvmpipe/lp_state_fs.c
blob: 9b0e7cdd37c77eb45d60b812030267936fd48035 (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
/**************************************************************************
 * 
 * Copyright 2009 VMware, Inc.
 * 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.
 * 
 **************************************************************************/

#include "pipe/p_defines.h"
#include "util/u_memory.h"
#include "util/u_format.h"
#include "util/u_debug_dump.h"
#include "pipe/internal/p_winsys_screen.h"
#include "pipe/p_shader_tokens.h"
#include "draw/draw_context.h"
#include "tgsi/tgsi_dump.h"
#include "tgsi/tgsi_scan.h"
#include "tgsi/tgsi_parse.h"
#include "lp_bld_type.h"
#include "lp_bld_conv.h"
#include "lp_bld_logic.h"
#include "lp_bld_depth.h"
#include "lp_bld_tgsi.h"
#include "lp_bld_alpha.h"
#include "lp_bld_blend.h"
#include "lp_bld_swizzle.h"
#include "lp_bld_flow.h"
#include "lp_bld_debug.h"
#include "lp_screen.h"
#include "lp_context.h"
#include "lp_state.h"
#include "lp_quad.h"


static const unsigned char quad_offset_x[4] = {0, 1, 0, 1};
static const unsigned char quad_offset_y[4] = {0, 0, 1, 1};


static void
generate_pos(LLVMBuilderRef builder,
             LLVMValueRef x,
             LLVMValueRef y,
             LLVMValueRef a0_ptr,
             LLVMValueRef dadx_ptr,
             LLVMValueRef dady_ptr,
             LLVMValueRef *pos)
{
   LLVMTypeRef int_elem_type = LLVMInt32Type();
   LLVMTypeRef int_vec_type = LLVMVectorType(int_elem_type, QUAD_SIZE);
   LLVMTypeRef elem_type = LLVMFloatType();
   LLVMTypeRef vec_type = LLVMVectorType(elem_type, QUAD_SIZE);
   LLVMValueRef x_offsets[QUAD_SIZE];
   LLVMValueRef y_offsets[QUAD_SIZE];
   unsigned chan;
   unsigned i;

   x = lp_build_broadcast(builder, int_vec_type, x);
   y = lp_build_broadcast(builder, int_vec_type, y);

   for(i = 0; i < QUAD_SIZE; ++i) {
      x_offsets[i] = LLVMConstInt(int_elem_type, quad_offset_x[i], 0);
      y_offsets[i] = LLVMConstInt(int_elem_type, quad_offset_y[i], 0);
   }

   x = LLVMBuildAdd(builder, x, LLVMConstVector(x_offsets, QUAD_SIZE), "");
   y = LLVMBuildAdd(builder, y, LLVMConstVector(y_offsets, QUAD_SIZE), "");

   x = LLVMBuildSIToFP(builder, x, vec_type, "");
   y = LLVMBuildSIToFP(builder, y, vec_type, "");

   pos[0] = x;
   pos[1] = y;

   for(chan = 2; chan < NUM_CHANNELS; ++chan) {
      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), chan, 0);
      LLVMValueRef a0   = LLVMBuildLoad(builder, LLVMBuildGEP(builder, a0_ptr,   &index, 1, ""), "");
      LLVMValueRef dadx = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dadx_ptr, &index, 1, ""), "");
      LLVMValueRef dady = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dady_ptr, &index, 1, ""), "");
      LLVMValueRef res;
      a0   = lp_build_broadcast(builder, vec_type, a0);
      dadx = lp_build_broadcast(builder, vec_type, dadx);
      dady = lp_build_broadcast(builder, vec_type, dady);
      res = a0;
      res = LLVMBuildAdd(builder, res, LLVMBuildMul(builder, dadx, x, ""), "");
      res = LLVMBuildAdd(builder, res, LLVMBuildMul(builder, dady, y, ""), "");
      pos[chan] = res;
   }

   for(chan = 0; chan < NUM_CHANNELS; ++chan)
      lp_build_name(pos[chan], "pos.%c", "xyzw"[chan]);
}


static void
generate_depth(struct llvmpipe_context *lp,
               LLVMBuilderRef builder,
               const struct pipe_depth_state *state,
               union lp_type src_type,
               struct lp_build_mask_context *mask,
               LLVMValueRef src,
               LLVMValueRef dst_ptr)
{
   const struct util_format_description *format_desc;
   union lp_type dst_type;

   if(!lp->framebuffer.zsbuf)
      return;

   format_desc = util_format_description(lp->framebuffer.zsbuf->format);
   assert(format_desc);

   dst_type = lp_depth_type(format_desc, src_type.width*src_type.length);

   assert(dst_type.width == src_type.width);
   assert(dst_type.length == src_type.length);

#if 1
   src = lp_build_clamped_float_to_unsigned_norm(builder,
                                                 src_type,
                                                 dst_type.width,
                                                 src);
#else
   lp_build_conv(builder, src_type, dst_type, &src, 1, &src, 1);
#endif

   lp_build_depth_test(builder,
                       state,
                       dst_type,
                       format_desc,
                       mask,
                       src,
                       dst_ptr);
}


/**
 * Generate the fragment shader, depth/stencil and alpha tests.
 */
static void
generate_fs(struct llvmpipe_context *lp,
            struct lp_fragment_shader *shader,
            const struct lp_fragment_shader_variant_key *key,
            LLVMBuilderRef builder,
            union lp_type type,
            unsigned i,
            LLVMValueRef x,
            LLVMValueRef y,
            LLVMValueRef a0_ptr,
            LLVMValueRef dadx_ptr,
            LLVMValueRef dady_ptr,
            LLVMValueRef consts_ptr,
            LLVMValueRef *pmask,
            LLVMValueRef *color,
            LLVMValueRef depth_ptr,
            LLVMValueRef samplers_ptr)
{
   const struct tgsi_token *tokens = shader->base.tokens;
   LLVMTypeRef elem_type;
   LLVMTypeRef vec_type;
   LLVMTypeRef int_vec_type;
   LLVMValueRef pos[NUM_CHANNELS];
   LLVMValueRef outputs[PIPE_MAX_SHADER_OUTPUTS][NUM_CHANNELS];
   struct lp_build_mask_context mask;
   boolean early_depth_test;
   unsigned attrib;
   unsigned chan;

   elem_type = lp_build_elem_type(type);
   vec_type = lp_build_vec_type(type);
   int_vec_type = lp_build_int_vec_type(type);

   generate_pos(builder, x, y, a0_ptr, dadx_ptr, dady_ptr, pos);

   lp_build_mask_begin(&mask, builder, type, *pmask);

   early_depth_test =
      lp->depth_stencil->depth.enabled &&
      lp->framebuffer.zsbuf &&
      !lp->depth_stencil->alpha.enabled &&
      !lp->fs->info.uses_kill &&
      !lp->fs->info.writes_z;

   if(early_depth_test)
      generate_depth(lp, builder, &key->depth,
                          type, &mask,
                          pos[2], depth_ptr);

   memset(outputs, 0, sizeof outputs);

   lp_build_tgsi_soa(builder, tokens, type, &mask,
                     pos, a0_ptr, dadx_ptr, dady_ptr,
                     consts_ptr, outputs, samplers_ptr);

   for (attrib = 0; attrib < shader->info.num_outputs; ++attrib) {
      for(chan = 0; chan < NUM_CHANNELS; ++chan) {
         if(outputs[attrib][chan]) {
            lp_build_name(outputs[attrib][chan], "output%u.%u.%c", i, attrib, "xyzw"[chan]);

            switch (shader->info.output_semantic_name[attrib]) {
            case TGSI_SEMANTIC_COLOR:
               {
                  unsigned cbuf = shader->info.output_semantic_index[attrib];

                  lp_build_name(outputs[attrib][chan], "color%u.%u.%c", i, attrib, "rgba"[chan]);

                  /* Alpha test */
                  /* XXX: should the alpha reference value be passed separately? */
                  if(cbuf == 0 && chan == 3)
                     lp_build_alpha_test(builder, &key->alpha, type,
                                         &mask,
                                         outputs[attrib][chan]);

                  if(cbuf == 0)
                     color[chan] = outputs[attrib][chan];

                  break;
               }

            case TGSI_SEMANTIC_POSITION:
               if(chan == 2)
                  pos[2] = outputs[attrib][chan];
               break;
            }
         }
      }
   }

   if(!early_depth_test)
      generate_depth(lp, builder, &key->depth,
                          type, &mask,
                          pos[2], depth_ptr);

   lp_build_mask_end(&mask);

   *pmask = mask.value;

}


/**
 * Generate blending code according to blend->base state.
 * The blend function will look like:
 *    blend(mask, src_color, constant color, dst_color)
 * dst_color will be modified and contain the result of the blend func.
 */
static void
generate_blend(const struct pipe_blend_state *blend,
               LLVMBuilderRef builder,
               union lp_type type,
               LLVMValueRef mask,
               LLVMValueRef *src,
               LLVMValueRef const_ptr,
               LLVMValueRef dst_ptr)
{
   struct lp_build_context bld;
   LLVMTypeRef vec_type;
   LLVMTypeRef int_vec_type;
   LLVMValueRef con[4];
   LLVMValueRef dst[4];
   LLVMValueRef res[4];
   unsigned chan;

   vec_type = lp_build_vec_type(type);
   int_vec_type = lp_build_int_vec_type(type);

   lp_build_context_init(&bld, builder, type);

   for(chan = 0; chan < 4; ++chan) {
      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), chan, 0);

      if(const_ptr)
         con[chan] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, const_ptr, &index, 1, ""), "");
      else
         con[chan] = LLVMGetUndef(vec_type); /* FIXME */

      dst[chan] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, dst_ptr, &index, 1, ""), "");

      lp_build_name(con[chan], "con.%c", "rgba"[chan]);
      lp_build_name(dst[chan], "dst.%c", "rgba"[chan]);
   }

   lp_build_blend_soa(builder, blend, type, src, dst, con, res);

   for(chan = 0; chan < 4; ++chan) {
      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), chan, 0);
      lp_build_name(res[chan], "res.%c", "rgba"[chan]);
      res[chan] = lp_build_select(&bld, mask, res[chan], dst[chan]);
      LLVMBuildStore(builder, res[chan], LLVMBuildGEP(builder, dst_ptr, &index, 1, ""));
   }
}


static struct lp_fragment_shader_variant *
generate_fragment(struct llvmpipe_context *lp,
                  struct lp_fragment_shader *shader,
                  const struct lp_fragment_shader_variant_key *key)
{
   struct llvmpipe_screen *screen = llvmpipe_screen(lp->pipe.screen);
   struct lp_fragment_shader_variant *variant;
   union lp_type fs_type;
   union lp_type blend_type;
   LLVMTypeRef fs_elem_type;
   LLVMTypeRef fs_vec_type;
   LLVMTypeRef fs_int_vec_type;
   LLVMTypeRef blend_vec_type;
   LLVMTypeRef blend_int_vec_type;
   LLVMTypeRef arg_types[10];
   LLVMTypeRef func_type;
   LLVMValueRef x;
   LLVMValueRef y;
   LLVMValueRef a0_ptr;
   LLVMValueRef dadx_ptr;
   LLVMValueRef dady_ptr;
   LLVMValueRef consts_ptr;
   LLVMValueRef mask_ptr;
   LLVMValueRef color_ptr;
   LLVMValueRef depth_ptr;
   LLVMValueRef samplers_ptr;
   LLVMBasicBlockRef block;
   LLVMBuilderRef builder;
   LLVMValueRef fs_mask[LP_MAX_VECTOR_LENGTH];
   LLVMValueRef fs_out_color[NUM_CHANNELS][LP_MAX_VECTOR_LENGTH];
   LLVMValueRef blend_mask;
   LLVMValueRef blend_in_color[NUM_CHANNELS];
   LLVMValueRef fetch_texel;
   unsigned num_fs;
   unsigned i;
   unsigned chan;

#ifdef DEBUG
   tgsi_dump(shader->base.tokens, 0);
   if(key->depth.enabled) {
      debug_printf("depth.func = %s\n", debug_dump_func(key->depth.func, TRUE));
      debug_printf("depth.writemask = %u\n", key->depth.writemask);
      debug_printf("depth.occlusion_count = %u\n", key->depth.occlusion_count);
   }
   if(key->alpha.enabled) {
      debug_printf("alpha.func = %s\n", debug_dump_func(key->alpha.func, TRUE));
      debug_printf("alpha.ref_value = %f\n", key->alpha.ref_value);
   }
   if(key->blend.logicop_enable) {
      debug_printf("blend.logicop_func = %u\n", key->blend.logicop_func);
   }
   else if(key->blend.blend_enable) {
      debug_printf("blend.rgb_func = %s\n",   debug_dump_blend_func  (key->blend.rgb_func, TRUE));
      debug_printf("rgb_src_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_src_factor, TRUE));
      debug_printf("rgb_dst_factor = %s\n",   debug_dump_blend_factor(key->blend.rgb_dst_factor, TRUE));
      debug_printf("alpha_func = %s\n",       debug_dump_blend_func  (key->blend.alpha_func, TRUE));
      debug_printf("alpha_src_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_src_factor, TRUE));
      debug_printf("alpha_dst_factor = %s\n", debug_dump_blend_factor(key->blend.alpha_dst_factor, TRUE));
   }
   debug_printf("blend.colormask = 0x%x\n", key->blend.colormask);
#endif

   variant = CALLOC_STRUCT(lp_fragment_shader_variant);
   if(!variant)
      return NULL;

   variant->shader = shader;
   memcpy(&variant->key, key, sizeof *key);

   fs_type.value = 0;
   fs_type.floating = TRUE; /* floating point values */
   fs_type.sign = TRUE;     /* values are signed */
   fs_type.norm = FALSE;    /* values are not limited to [0,1] or [-1,1] */
   fs_type.width = 32;      /* 32-bit float */
   fs_type.length = 4;      /* 4 element per vector */
   num_fs = 4;

   blend_type.value = 0;
   blend_type.floating = FALSE; /* values are integers */
   blend_type.sign = FALSE;     /* values are unsigned */
   blend_type.norm = TRUE;      /* values are in [0,1] or [-1,1] */
   blend_type.width = 8;        /* 8-bit ubyte values */
   blend_type.length = 16;      /* 16 elements per vector */

   fs_elem_type = lp_build_elem_type(fs_type);
   fs_vec_type = lp_build_vec_type(fs_type);
   fs_int_vec_type = lp_build_int_vec_type(fs_type);

   blend_vec_type = lp_build_vec_type(blend_type);
   blend_int_vec_type = lp_build_int_vec_type(blend_type);

   arg_types[0] = LLVMInt32Type();                     /* x */
   arg_types[1] = LLVMInt32Type();                     /* y */
   arg_types[2] = LLVMPointerType(fs_elem_type, 0);    /* a0 */
   arg_types[3] = LLVMPointerType(fs_elem_type, 0);    /* dadx */
   arg_types[4] = LLVMPointerType(fs_elem_type, 0);    /* dady */
   arg_types[5] = LLVMPointerType(fs_elem_type, 0);    /* consts */
   arg_types[6] = LLVMPointerType(fs_int_vec_type, 0); /* mask */
   arg_types[7] = LLVMPointerType(blend_vec_type, 0);  /* color */
   arg_types[8] = LLVMPointerType(fs_int_vec_type, 0); /* depth */
   arg_types[9] = LLVMPointerType(LLVMInt8Type(), 0);  /* samplers */

   func_type = LLVMFunctionType(LLVMVoidType(), arg_types, Elements(arg_types), 0);

   variant->function = LLVMAddFunction(screen->module, "shader", func_type);
   LLVMSetFunctionCallConv(variant->function, LLVMCCallConv);
   for(i = 0; i < Elements(arg_types); ++i)
      if(LLVMGetTypeKind(arg_types[i]) == LLVMPointerTypeKind)
         LLVMAddAttribute(LLVMGetParam(variant->function, i), LLVMNoAliasAttribute);

   x            = LLVMGetParam(variant->function, 0);
   y            = LLVMGetParam(variant->function, 1);
   a0_ptr       = LLVMGetParam(variant->function, 2);
   dadx_ptr     = LLVMGetParam(variant->function, 3);
   dady_ptr     = LLVMGetParam(variant->function, 4);
   consts_ptr   = LLVMGetParam(variant->function, 5);
   mask_ptr     = LLVMGetParam(variant->function, 6);
   color_ptr    = LLVMGetParam(variant->function, 7);
   depth_ptr    = LLVMGetParam(variant->function, 8);
   samplers_ptr = LLVMGetParam(variant->function, 9);

   lp_build_name(x, "x");
   lp_build_name(y, "y");
   lp_build_name(a0_ptr, "a0");
   lp_build_name(dadx_ptr, "dadx");
   lp_build_name(dady_ptr, "dady");
   lp_build_name(consts_ptr, "consts");
   lp_build_name(mask_ptr, "mask");
   lp_build_name(color_ptr, "color");
   lp_build_name(depth_ptr, "depth");
   lp_build_name(samplers_ptr, "samplers");

   block = LLVMAppendBasicBlock(variant->function, "entry");
   builder = LLVMCreateBuilder();
   LLVMPositionBuilderAtEnd(builder, block);

   for(i = 0; i < num_fs; ++i) {
      LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
      LLVMValueRef out_color[NUM_CHANNELS];
      LLVMValueRef x_i;
      LLVMValueRef depth_ptr_i;

      /* TODO: Reuse position interpolation */
      x_i = LLVMBuildAdd(builder, x, LLVMConstInt(LLVMInt32Type(), 2*i, 0), "");

      fs_mask[i] = LLVMBuildLoad(builder, LLVMBuildGEP(builder, mask_ptr, &index, 1, ""), "");
      depth_ptr_i = LLVMBuildGEP(builder, depth_ptr, &index, 1, "");

      generate_fs(lp,
                  shader,
                  key,
                  builder,
                  fs_type,
                  i,
                  x_i,
                  y,
                  a0_ptr,
                  dadx_ptr,
                  dady_ptr,
                  consts_ptr,
                  &fs_mask[i],
                  out_color,
                  depth_ptr_i,
                  samplers_ptr);

      for(chan = 0; chan < NUM_CHANNELS; ++chan)
         fs_out_color[chan][i] = out_color[chan];
   }

   for(chan = 0; chan < NUM_CHANNELS; ++chan) {
      lp_build_conv(builder, fs_type, blend_type,
                    fs_out_color[chan], num_fs,
                    &blend_in_color[chan], 1);
      lp_build_name(blend_in_color[chan], "color.%c", "rgba"[chan]);
   }

   lp_build_conv_mask(builder, fs_type, blend_type,
                               fs_mask, num_fs,
                               &blend_mask, 1);

   generate_blend(&key->blend,
                  builder,
                  blend_type,
                  blend_mask,
                  blend_in_color,
                  NULL /* FIXME: blend_const_color */,
                  color_ptr);

   LLVMBuildRetVoid(builder);;

   LLVMDisposeBuilder(builder);

   LLVMRunFunctionPassManager(screen->pass, variant->function);

#ifdef DEBUG
   LLVMDumpValue(variant->function);
   debug_printf("\n");
#endif

   if(LLVMVerifyFunction(variant->function, LLVMPrintMessageAction)) {
      LLVMDumpValue(variant->function);
      abort();
   }

   fetch_texel = LLVMGetNamedFunction(screen->module, "fetch_texel");
   if(fetch_texel) {
      static boolean first_time = TRUE;
      if(first_time) {
         LLVMAddGlobalMapping(screen->engine, fetch_texel, lp_build_tgsi_fetch_texel_soa);
         first_time = FALSE;
      }
   }

   variant->jit_function = (lp_shader_fs_func)LLVMGetPointerToGlobal(screen->engine, variant->function);

#ifdef DEBUG
   lp_disassemble(variant->jit_function);
#endif

   variant->next = shader->variants;
   shader->variants = variant;

   return variant;
}


void *
llvmpipe_create_fs_state(struct pipe_context *pipe,
                         const struct pipe_shader_state *templ)
{
   struct lp_fragment_shader *shader;

   shader = CALLOC_STRUCT(lp_fragment_shader);
   if (!shader)
      return NULL;

   /* get/save the summary info for this shader */
   tgsi_scan_shader(templ->tokens, &shader->info);

   /* we need to keep a local copy of the tokens */
   shader->base.tokens = tgsi_dup_tokens(templ->tokens);

   return shader;
}


void
llvmpipe_bind_fs_state(struct pipe_context *pipe, void *fs)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);

   llvmpipe->fs = (struct lp_fragment_shader *) fs;

   llvmpipe->dirty |= LP_NEW_FS;
}


void
llvmpipe_delete_fs_state(struct pipe_context *pipe, void *fs)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
   struct llvmpipe_screen *screen = llvmpipe_screen(pipe->screen);
   struct lp_fragment_shader *shader = fs;
   struct lp_fragment_shader_variant *variant;

   assert(fs != llvmpipe->fs);

   variant = shader->variants;
   while(variant) {
      struct lp_fragment_shader_variant *next = variant->next;

      if(variant->function) {
         if(variant->jit_function)
            LLVMFreeMachineCodeForFunction(screen->engine, variant->function);
         LLVMDeleteFunction(variant->function);
      }

      FREE(variant);

      variant = next;
   }

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


void *
llvmpipe_create_vs_state(struct pipe_context *pipe,
                         const struct pipe_shader_state *templ)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);
   struct lp_vertex_shader *state;

   state = CALLOC_STRUCT(lp_vertex_shader);
   if (state == NULL ) 
      goto fail;

   /* copy shader tokens, the ones passed in will go away.
    */
   state->shader.tokens = tgsi_dup_tokens(templ->tokens);
   if (state->shader.tokens == NULL)
      goto fail;

   state->draw_data = draw_create_vertex_shader(llvmpipe->draw, templ);
   if (state->draw_data == NULL) 
      goto fail;

   return state;

fail:
   if (state) {
      FREE( (void *)state->shader.tokens );
      FREE( state->draw_data );
      FREE( state );
   }
   return NULL;
}


void
llvmpipe_bind_vs_state(struct pipe_context *pipe, void *vs)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);

   llvmpipe->vs = (const struct lp_vertex_shader *)vs;

   draw_bind_vertex_shader(llvmpipe->draw,
                           (llvmpipe->vs ? llvmpipe->vs->draw_data : NULL));

   llvmpipe->dirty |= LP_NEW_VS;
}


void
llvmpipe_delete_vs_state(struct pipe_context *pipe, void *vs)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);

   struct lp_vertex_shader *state =
      (struct lp_vertex_shader *)vs;

   draw_delete_vertex_shader(llvmpipe->draw, state->draw_data);
   FREE( state );
}



void
llvmpipe_set_constant_buffer(struct pipe_context *pipe,
                             uint shader, uint index,
                             const struct pipe_constant_buffer *buf)
{
   struct llvmpipe_context *llvmpipe = llvmpipe_context(pipe);

   assert(shader < PIPE_SHADER_TYPES);
   assert(index == 0);

   /* note: reference counting */
   pipe_buffer_reference(&llvmpipe->constants[shader].buffer,
			 buf ? buf->buffer : NULL);

   llvmpipe->dirty |= LP_NEW_CONSTANTS;
}


void llvmpipe_update_fs(struct llvmpipe_context *lp)
{
   struct lp_fragment_shader *shader = lp->fs;
   struct lp_fragment_shader_variant_key key;
   struct lp_fragment_shader_variant *variant;

   memset(&key, 0, sizeof key);
   memcpy(&key.depth, &lp->depth_stencil->depth, sizeof &key.depth);
   memcpy(&key.alpha, &lp->depth_stencil->alpha, sizeof &key.alpha);
   memcpy(&key.blend, &lp->blend->base, sizeof &key.blend);

   variant = shader->variants;
   while(variant) {
      if(memcmp(&variant->key, &key, sizeof key) == 0)
         break;

      variant = variant->next;
   }

   if(!variant)
      variant = generate_fragment(lp, shader, &key);

   shader->current = variant;
}