summaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
blob: 160dd2f6c624a7054523b2ad19f08ec5f893a373 (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
/*
 Copyright (C) Intel Corp.  2006.  All Rights Reserved.
 Intel funded Tungsten Graphics to
 develop this 3D driver.

 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, sublicense, 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 NONINFRINGEMENT.
 IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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.

 **********************************************************************/
 /*
  * Authors:
  *   Keith Whitwell <keithw@vmware.com>
  */


#include "main/context.h"
#include "main/blend.h"
#include "main/mtypes.h"
#include "main/samplerobj.h"
#include "program/prog_parameter.h"

#include "intel_mipmap_tree.h"
#include "intel_batchbuffer.h"
#include "intel_tex.h"
#include "intel_fbo.h"
#include "intel_buffer_objects.h"

#include "brw_context.h"
#include "brw_state.h"
#include "brw_defines.h"
#include "brw_wm.h"

GLuint
translate_tex_target(GLenum target)
{
   switch (target) {
   case GL_TEXTURE_1D:
   case GL_TEXTURE_1D_ARRAY_EXT:
      return BRW_SURFACE_1D;

   case GL_TEXTURE_RECTANGLE_NV:
      return BRW_SURFACE_2D;

   case GL_TEXTURE_2D:
   case GL_TEXTURE_2D_ARRAY_EXT:
   case GL_TEXTURE_EXTERNAL_OES:
   case GL_TEXTURE_2D_MULTISAMPLE:
   case GL_TEXTURE_2D_MULTISAMPLE_ARRAY:
      return BRW_SURFACE_2D;

   case GL_TEXTURE_3D:
      return BRW_SURFACE_3D;

   case GL_TEXTURE_CUBE_MAP:
   case GL_TEXTURE_CUBE_MAP_ARRAY:
      return BRW_SURFACE_CUBE;

   default:
      unreachable("not reached");
   }
}

uint32_t
brw_get_surface_tiling_bits(uint32_t tiling)
{
   switch (tiling) {
   case I915_TILING_X:
      return BRW_SURFACE_TILED;
   case I915_TILING_Y:
      return BRW_SURFACE_TILED | BRW_SURFACE_TILED_Y;
   default:
      return 0;
   }
}


uint32_t
brw_get_surface_num_multisamples(unsigned num_samples)
{
   if (num_samples > 1)
      return BRW_SURFACE_MULTISAMPLECOUNT_4;
   else
      return BRW_SURFACE_MULTISAMPLECOUNT_1;
}

void
brw_configure_w_tiled(const struct intel_mipmap_tree *mt,
                      bool is_render_target,
                      unsigned *width, unsigned *height,
                      unsigned *pitch, uint32_t *tiling, unsigned *format)
{
   static const unsigned halign_stencil = 8;

   /* In Y-tiling row is twice as wide as in W-tiling, and subsequently
    * there are half as many rows.
    * In addition, mip-levels are accessed manually by the program and
    * therefore the surface is setup to cover all the mip-levels for one slice.
    * (Hardware is still used to access individual slices).
    */
   *tiling = I915_TILING_Y;
   *pitch = mt->pitch * 2;
   *width = ALIGN(mt->total_width, halign_stencil) * 2;
   *height = (mt->total_height / mt->physical_depth0) / 2;

   if (is_render_target) {
      *format = BRW_SURFACEFORMAT_R8_UINT;
   }
}


/**
 * Compute the combination of DEPTH_TEXTURE_MODE and EXT_texture_swizzle
 * swizzling.
 */
int
brw_get_texture_swizzle(const struct gl_context *ctx,
                        const struct gl_texture_object *t)
{
   const struct gl_texture_image *img = t->Image[0][t->BaseLevel];

   int swizzles[SWIZZLE_NIL + 1] = {
      SWIZZLE_X,
      SWIZZLE_Y,
      SWIZZLE_Z,
      SWIZZLE_W,
      SWIZZLE_ZERO,
      SWIZZLE_ONE,
      SWIZZLE_NIL
   };

   if (img->_BaseFormat == GL_DEPTH_COMPONENT ||
       img->_BaseFormat == GL_DEPTH_STENCIL) {
      GLenum depth_mode = t->DepthMode;

      /* In ES 3.0, DEPTH_TEXTURE_MODE is expected to be GL_RED for textures
       * with depth component data specified with a sized internal format.
       * Otherwise, it's left at the old default, GL_LUMINANCE.
       */
      if (_mesa_is_gles3(ctx) &&
          img->InternalFormat != GL_DEPTH_COMPONENT &&
          img->InternalFormat != GL_DEPTH_STENCIL) {
         depth_mode = GL_RED;
      }

      switch (depth_mode) {
      case GL_ALPHA:
         swizzles[0] = SWIZZLE_ZERO;
         swizzles[1] = SWIZZLE_ZERO;
         swizzles[2] = SWIZZLE_ZERO;
         swizzles[3] = SWIZZLE_X;
         break;
      case GL_LUMINANCE:
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_X;
         swizzles[2] = SWIZZLE_X;
         swizzles[3] = SWIZZLE_ONE;
         break;
      case GL_INTENSITY:
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_X;
         swizzles[2] = SWIZZLE_X;
         swizzles[3] = SWIZZLE_X;
         break;
      case GL_RED:
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_ZERO;
         swizzles[2] = SWIZZLE_ZERO;
         swizzles[3] = SWIZZLE_ONE;
         break;
      }
   }

   GLenum datatype = _mesa_get_format_datatype(img->TexFormat);

   /* If the texture's format is alpha-only, force R, G, and B to
    * 0.0. Similarly, if the texture's format has no alpha channel,
    * force the alpha value read to 1.0. This allows for the
    * implementation to use an RGBA texture for any of these formats
    * without leaking any unexpected values.
    */
   switch (img->_BaseFormat) {
   case GL_ALPHA:
      swizzles[0] = SWIZZLE_ZERO;
      swizzles[1] = SWIZZLE_ZERO;
      swizzles[2] = SWIZZLE_ZERO;
      break;
   case GL_LUMINANCE:
      if (t->_IsIntegerFormat || datatype == GL_SIGNED_NORMALIZED) {
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_X;
         swizzles[2] = SWIZZLE_X;
         swizzles[3] = SWIZZLE_ONE;
      }
      break;
   case GL_LUMINANCE_ALPHA:
      if (datatype == GL_SIGNED_NORMALIZED) {
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_X;
         swizzles[2] = SWIZZLE_X;
         swizzles[3] = SWIZZLE_W;
      }
      break;
   case GL_INTENSITY:
      if (datatype == GL_SIGNED_NORMALIZED) {
         swizzles[0] = SWIZZLE_X;
         swizzles[1] = SWIZZLE_X;
         swizzles[2] = SWIZZLE_X;
         swizzles[3] = SWIZZLE_X;
      }
      break;
   case GL_RED:
   case GL_RG:
   case GL_RGB:
      if (_mesa_get_format_bits(img->TexFormat, GL_ALPHA_BITS) > 0)
         swizzles[3] = SWIZZLE_ONE;
      break;
   }

   return MAKE_SWIZZLE4(swizzles[GET_SWZ(t->_Swizzle, 0)],
                        swizzles[GET_SWZ(t->_Swizzle, 1)],
                        swizzles[GET_SWZ(t->_Swizzle, 2)],
                        swizzles[GET_SWZ(t->_Swizzle, 3)]);
}

static void
gen4_emit_buffer_surface_state(struct brw_context *brw,
                               uint32_t *out_offset,
                               drm_intel_bo *bo,
                               unsigned buffer_offset,
                               unsigned surface_format,
                               unsigned buffer_size,
                               unsigned pitch,
                               bool rw)
{
   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
                                    6 * 4, 32, out_offset);
   memset(surf, 0, 6 * 4);

   surf[0] = BRW_SURFACE_BUFFER << BRW_SURFACE_TYPE_SHIFT |
             surface_format << BRW_SURFACE_FORMAT_SHIFT |
             (brw->gen >= 6 ? BRW_SURFACE_RC_READ_WRITE : 0);
   surf[1] = (bo ? bo->offset64 : 0) + buffer_offset; /* reloc */
   surf[2] = (buffer_size & 0x7f) << BRW_SURFACE_WIDTH_SHIFT |
             ((buffer_size >> 7) & 0x1fff) << BRW_SURFACE_HEIGHT_SHIFT;
   surf[3] = ((buffer_size >> 20) & 0x7f) << BRW_SURFACE_DEPTH_SHIFT |
             (pitch - 1) << BRW_SURFACE_PITCH_SHIFT;

   /* Emit relocation to surface contents.  The 965 PRM, Volume 4, section
    * 5.1.2 "Data Cache" says: "the data cache does not exist as a separate
    * physical cache.  It is mapped in hardware to the sampler cache."
    */
   if (bo) {
      drm_intel_bo_emit_reloc(brw->batch.bo, *out_offset + 4,
                              bo, buffer_offset,
                              I915_GEM_DOMAIN_SAMPLER,
                              (rw ? I915_GEM_DOMAIN_SAMPLER : 0));
   }
}

void
brw_update_buffer_texture_surface(struct gl_context *ctx,
                                  unsigned unit,
                                  uint32_t *surf_offset)
{
   struct brw_context *brw = brw_context(ctx);
   struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
   struct intel_buffer_object *intel_obj =
      intel_buffer_object(tObj->BufferObject);
   uint32_t size = tObj->BufferSize;
   drm_intel_bo *bo = NULL;
   mesa_format format = tObj->_BufferObjectFormat;
   uint32_t brw_format = brw_format_for_mesa_format(format);
   int texel_size = _mesa_get_format_bytes(format);

   if (intel_obj) {
      size = MIN2(size, intel_obj->Base.Size);
      bo = intel_bufferobj_buffer(brw, intel_obj, tObj->BufferOffset, size);
   }

   if (brw_format == 0 && format != MESA_FORMAT_RGBA_FLOAT32) {
      _mesa_problem(NULL, "bad format %s for texture buffer\n",
		    _mesa_get_format_name(format));
   }

   brw->vtbl.emit_buffer_surface_state(brw, surf_offset, bo,
                                       tObj->BufferOffset,
                                       brw_format,
                                       size / texel_size,
                                       texel_size,
                                       false /* rw */);
}

static void
brw_update_texture_surface(struct gl_context *ctx,
                           unsigned unit,
                           uint32_t *surf_offset,
                           bool for_gather)
{
   struct brw_context *brw = brw_context(ctx);
   struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current;
   struct intel_texture_object *intelObj = intel_texture_object(tObj);
   struct intel_mipmap_tree *mt = intelObj->mt;
   struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit);
   uint32_t *surf;

   /* BRW_NEW_TEXTURE_BUFFER */
   if (tObj->Target == GL_TEXTURE_BUFFER) {
      brw_update_buffer_texture_surface(ctx, unit, surf_offset);
      return;
   }

   surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE,
			  6 * 4, 32, surf_offset);

   uint32_t tex_format = translate_tex_format(brw, mt->format,
                                              sampler->sRGBDecode);

   if (for_gather) {
      /* Sandybridge's gather4 message is broken for integer formats.
       * To work around this, we pretend the surface is UNORM for
       * 8 or 16-bit formats, and emit shader instructions to recover
       * the real INT/UINT value.  For 32-bit formats, we pretend
       * the surface is FLOAT, and simply reinterpret the resulting
       * bits.
       */
      switch (tex_format) {
      case BRW_SURFACEFORMAT_R8_SINT:
      case BRW_SURFACEFORMAT_R8_UINT:
         tex_format = BRW_SURFACEFORMAT_R8_UNORM;
         break;

      case BRW_SURFACEFORMAT_R16_SINT:
      case BRW_SURFACEFORMAT_R16_UINT:
         tex_format = BRW_SURFACEFORMAT_R16_UNORM;
         break;

      case BRW_SURFACEFORMAT_R32_SINT:
      case BRW_SURFACEFORMAT_R32_UINT:
         tex_format = BRW_SURFACEFORMAT_R32_FLOAT;
         break;

      default:
         break;
      }
   }

   surf[0] = (translate_tex_target(tObj->Target) << BRW_SURFACE_TYPE_SHIFT |
	      BRW_SURFACE_MIPMAPLAYOUT_BELOW << BRW_SURFACE_MIPLAYOUT_SHIFT |
	      BRW_SURFACE_CUBEFACE_ENABLES |
	      tex_format << BRW_SURFACE_FORMAT_SHIFT);

   surf[1] = mt->bo->offset64 + mt->offset; /* reloc */

   surf[2] = ((intelObj->_MaxLevel - tObj->BaseLevel) << BRW_SURFACE_LOD_SHIFT |
	      (mt->logical_width0 - 1) << BRW_SURFACE_WIDTH_SHIFT |
	      (mt->logical_height0 - 1) << BRW_SURFACE_HEIGHT_SHIFT);

   surf[3] = (brw_get_surface_tiling_bits(mt->tiling) |
	      (mt->logical_depth0 - 1) << BRW_SURFACE_DEPTH_SHIFT |
	      (mt->pitch - 1) << BRW_SURFACE_PITCH_SHIFT);

   surf[4] = (brw_get_surface_num_multisamples(mt->num_samples) |
              SET_FIELD(tObj->BaseLevel - mt->first_level, BRW_SURFACE_MIN_LOD));

   surf[5] = mt->align_h == 4 ? BRW_SURFACE_VERTICAL_ALIGN_ENABLE : 0;

   /* Emit relocation to surface contents */
   drm_intel_bo_emit_reloc(brw->batch.bo,
                           *surf_offset + 4,
                           mt->bo,
                           surf[1] - mt->bo->offset64,
                           I915_GEM_DOMAIN_SAMPLER, 0);
}

/**
 * Create the constant buffer surface.  Vertex/fragment shader constants will be
 * read from this buffer with Data Port Read instructions/messages.
 */
void
brw_create_constant_surface(struct brw_context *brw,
			    drm_intel_bo *bo,
			    uint32_t offset,
			    uint32_t size,
			    uint32_t *out_offset,
                            bool dword_pitch)
{
   uint32_t stride = dword_pitch ? 4 : 16;
   uint32_t elements = ALIGN(size, stride) / stride;

   brw->vtbl.emit_buffer_surface_state(brw, out_offset, bo, offset,
                                       BRW_SURFACEFORMAT_R32G32B32A32_FLOAT,
                                       elements, stride, false);
}

/**
 * Set up a binding table entry for use by stream output logic (transform
 * feedback).
 *
 * buffer_size_minus_1 must be less than BRW_MAX_NUM_BUFFER_ENTRIES.
 */
void
brw_update_sol_surface(struct brw_context *brw,
                       struct gl_buffer_object *buffer_obj,
                       uint32_t *out_offset, unsigned num_vector_components,
                       unsigned stride_dwords, unsigned offset_dwords)
{
   struct intel_buffer_object *intel_bo = intel_buffer_object(buffer_obj);
   uint32_t offset_bytes = 4 * offset_dwords;
   drm_intel_bo *bo = intel_bufferobj_buffer(brw, intel_bo,
                                             offset_bytes,
                                             buffer_obj->Size - offset_bytes);
   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 6 * 4, 32,
                                    out_offset);
   uint32_t pitch_minus_1 = 4*stride_dwords - 1;
   size_t size_dwords = buffer_obj->Size / 4;
   uint32_t buffer_size_minus_1, width, height, depth, surface_format;

   /* FIXME: can we rely on core Mesa to ensure that the buffer isn't
    * too big to map using a single binding table entry?
    */
   assert((size_dwords - offset_dwords) / stride_dwords
          <= BRW_MAX_NUM_BUFFER_ENTRIES);

   if (size_dwords > offset_dwords + num_vector_components) {
      /* There is room for at least 1 transform feedback output in the buffer.
       * Compute the number of additional transform feedback outputs the
       * buffer has room for.
       */
      buffer_size_minus_1 =
         (size_dwords - offset_dwords - num_vector_components) / stride_dwords;
   } else {
      /* There isn't even room for a single transform feedback output in the
       * buffer.  We can't configure the binding table entry to prevent output
       * entirely; we'll have to rely on the geometry shader to detect
       * overflow.  But to minimize the damage in case of a bug, set up the
       * binding table entry to just allow a single output.
       */
      buffer_size_minus_1 = 0;
   }
   width = buffer_size_minus_1 & 0x7f;
   height = (buffer_size_minus_1 & 0xfff80) >> 7;
   depth = (buffer_size_minus_1 & 0x7f00000) >> 20;

   switch (num_vector_components) {
   case 1:
      surface_format = BRW_SURFACEFORMAT_R32_FLOAT;
      break;
   case 2:
      surface_format = BRW_SURFACEFORMAT_R32G32_FLOAT;
      break;
   case 3:
      surface_format = BRW_SURFACEFORMAT_R32G32B32_FLOAT;
      break;
   case 4:
      surface_format = BRW_SURFACEFORMAT_R32G32B32A32_FLOAT;
      break;
   default:
      unreachable("Invalid vector size for transform feedback output");
   }

   surf[0] = BRW_SURFACE_BUFFER << BRW_SURFACE_TYPE_SHIFT |
      BRW_SURFACE_MIPMAPLAYOUT_BELOW << BRW_SURFACE_MIPLAYOUT_SHIFT |
      surface_format << BRW_SURFACE_FORMAT_SHIFT |
      BRW_SURFACE_RC_READ_WRITE;
   surf[1] = bo->offset64 + offset_bytes; /* reloc */
   surf[2] = (width << BRW_SURFACE_WIDTH_SHIFT |
	      height << BRW_SURFACE_HEIGHT_SHIFT);
   surf[3] = (depth << BRW_SURFACE_DEPTH_SHIFT |
              pitch_minus_1 << BRW_SURFACE_PITCH_SHIFT);
   surf[4] = 0;
   surf[5] = 0;

   /* Emit relocation to surface contents. */
   drm_intel_bo_emit_reloc(brw->batch.bo,
			   *out_offset + 4,
			   bo, offset_bytes,
			   I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER);
}

/* Creates a new WM constant buffer reflecting the current fragment program's
 * constants, if needed by the fragment program.
 *
 * Otherwise, constants go through the CURBEs using the brw_constant_buffer
 * state atom.
 */
static void
brw_upload_wm_pull_constants(struct brw_context *brw)
{
   struct brw_stage_state *stage_state = &brw->wm.base;
   /* BRW_NEW_FRAGMENT_PROGRAM */
   struct brw_fragment_program *fp =
      (struct brw_fragment_program *) brw->fragment_program;
   /* BRW_NEW_FS_PROG_DATA */
   struct brw_stage_prog_data *prog_data = &brw->wm.prog_data->base;

   /* _NEW_PROGRAM_CONSTANTS */
   brw_upload_pull_constants(brw, BRW_NEW_SURFACES, &fp->program.Base,
                             stage_state, prog_data, true);
}

const struct brw_tracked_state brw_wm_pull_constants = {
   .dirty = {
      .mesa = _NEW_PROGRAM_CONSTANTS,
      .brw = BRW_NEW_BATCH |
             BRW_NEW_FRAGMENT_PROGRAM |
             BRW_NEW_FS_PROG_DATA,
   },
   .emit = brw_upload_wm_pull_constants,
};

/**
 * Creates a null renderbuffer surface.
 *
 * This is used when the shader doesn't write to any color output.  An FB
 * write to target 0 will still be emitted, because that's how the thread is
 * terminated (and computed depth is returned), so we need to have the
 * hardware discard the target 0 color output..
 */
static void
brw_emit_null_surface_state(struct brw_context *brw,
                            unsigned width,
                            unsigned height,
                            unsigned samples,
                            uint32_t *out_offset)
{
   /* From the Sandy bridge PRM, Vol4 Part1 p71 (Surface Type: Programming
    * Notes):
    *
    *     A null surface will be used in instances where an actual surface is
    *     not bound. When a write message is generated to a null surface, no
    *     actual surface is written to. When a read message (including any
    *     sampling engine message) is generated to a null surface, the result
    *     is all zeros. Note that a null surface type is allowed to be used
    *     with all messages, even if it is not specificially indicated as
    *     supported. All of the remaining fields in surface state are ignored
    *     for null surfaces, with the following exceptions:
    *
    *     - [DevSNB+]: Width, Height, Depth, and LOD fields must match the
    *       depth buffer’s corresponding state for all render target surfaces,
    *       including null.
    *
    *     - Surface Format must be R8G8B8A8_UNORM.
    */
   unsigned surface_type = BRW_SURFACE_NULL;
   drm_intel_bo *bo = NULL;
   unsigned pitch_minus_1 = 0;
   uint32_t multisampling_state = 0;
   uint32_t *surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 6 * 4, 32,
                                    out_offset);

   if (samples > 1) {
      /* On Gen6, null render targets seem to cause GPU hangs when
       * multisampling.  So work around this problem by rendering into dummy
       * color buffer.
       *
       * To decrease the amount of memory needed by the workaround buffer, we
       * set its pitch to 128 bytes (the width of a Y tile).  This means that
       * the amount of memory needed for the workaround buffer is
       * (width_in_tiles + height_in_tiles - 1) tiles.
       *
       * Note that since the workaround buffer will be interpreted by the
       * hardware as an interleaved multisampled buffer, we need to compute
       * width_in_tiles and height_in_tiles by dividing the width and height
       * by 16 rather than the normal Y-tile size of 32.
       */
      unsigned width_in_tiles = ALIGN(width, 16) / 16;
      unsigned height_in_tiles = ALIGN(height, 16) / 16;
      unsigned size_needed = (width_in_tiles + height_in_tiles - 1) * 4096;
      brw_get_scratch_bo(brw, &brw->wm.multisampled_null_render_target_bo,
                         size_needed);
      bo = brw->wm.multisampled_null_render_target_bo;
      surface_type = BRW_SURFACE_2D;
      pitch_minus_1 = 127;
      multisampling_state = brw_get_surface_num_multisamples(samples);
   }

   surf[0] = (surface_type << BRW_SURFACE_TYPE_SHIFT |
	      BRW_SURFACEFORMAT_B8G8R8A8_UNORM << BRW_SURFACE_FORMAT_SHIFT);
   if (brw->gen < 6) {
      surf[0] |= (1 << BRW_SURFACE_WRITEDISABLE_R_SHIFT |
		  1 << BRW_SURFACE_WRITEDISABLE_G_SHIFT |
		  1 << BRW_SURFACE_WRITEDISABLE_B_SHIFT |
		  1 << BRW_SURFACE_WRITEDISABLE_A_SHIFT);
   }
   surf[1] = bo ? bo->offset64 : 0;
   surf[2] = ((width - 1) << BRW_SURFACE_WIDTH_SHIFT |
              (height - 1) << BRW_SURFACE_HEIGHT_SHIFT);

   /* From Sandy bridge PRM, Vol4 Part1 p82 (Tiled Surface: Programming
    * Notes):
    *
    *     If Surface Type is SURFTYPE_NULL, this field must be TRUE
    */
   surf[3] = (BRW_SURFACE_TILED | BRW_SURFACE_TILED_Y |
              pitch_minus_1 << BRW_SURFACE_PITCH_SHIFT);
   surf[4] = multisampling_state;
   surf[5] = 0;

   if (bo) {
      drm_intel_bo_emit_reloc(brw->batch.bo,
                              *out_offset + 4,
                              bo, 0,
                              I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER);
   }
}

/**
 * Sets up a surface state structure to point at the given region.
 * While it is only used for the front/back buffer currently, it should be
 * usable for further buffers when doing ARB_draw_buffer support.
 */
static uint32_t
brw_update_renderbuffer_surface(struct brw_context *brw,
                                struct gl_renderbuffer *rb,
                                bool layered, unsigned unit,
                                uint32_t surf_index)
{
   struct gl_context *ctx = &brw->ctx;
   struct intel_renderbuffer *irb = intel_renderbuffer(rb);
   struct intel_mipmap_tree *mt = irb->mt;
   uint32_t *surf;
   uint32_t tile_x, tile_y;
   uint32_t format = 0;
   uint32_t offset;
   /* _NEW_BUFFERS */
   mesa_format rb_format = _mesa_get_render_format(ctx, intel_rb_format(irb));
   /* BRW_NEW_FS_PROG_DATA */

   assert(!layered);

   if (rb->TexImage && !brw->has_surface_tile_offset) {
      intel_renderbuffer_get_tile_offsets(irb, &tile_x, &tile_y);

      if (tile_x != 0 || tile_y != 0) {
	 /* Original gen4 hardware couldn't draw to a non-tile-aligned
	  * destination in a miptree unless you actually setup your renderbuffer
	  * as a miptree and used the fragile lod/array_index/etc. controls to
	  * select the image.  So, instead, we just make a new single-level
	  * miptree and render into that.
	  */
	 intel_renderbuffer_move_to_temp(brw, irb, false);
	 mt = irb->mt;
      }
   }

   intel_miptree_used_for_rendering(irb->mt);

   surf = brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, 6 * 4, 32, &offset);

   format = brw->render_target_format[rb_format];
   if (unlikely(!brw->format_supported_as_render_target[rb_format])) {
      _mesa_problem(ctx, "%s: renderbuffer format %s unsupported\n",
                    __func__, _mesa_get_format_name(rb_format));
   }

   surf[0] = (BRW_SURFACE_2D << BRW_SURFACE_TYPE_SHIFT |
	      format << BRW_SURFACE_FORMAT_SHIFT);

   /* reloc */
   assert(mt->offset % mt->cpp == 0);
   surf[1] = (intel_renderbuffer_get_tile_offsets(irb, &tile_x, &tile_y) +
	      mt->bo->offset64 + mt->offset);

   surf[2] = ((rb->Width - 1) << BRW_SURFACE_WIDTH_SHIFT |
	      (rb->Height - 1) << BRW_SURFACE_HEIGHT_SHIFT);

   surf[3] = (brw_get_surface_tiling_bits(mt->tiling) |
	      (mt->pitch - 1) << BRW_SURFACE_PITCH_SHIFT);

   surf[4] = brw_get_surface_num_multisamples(mt->num_samples);

   assert(brw->has_surface_tile_offset || (tile_x == 0 && tile_y == 0));
   /* Note that the low bits of these fields are missing, so
    * there's the possibility of getting in trouble.
    */
   assert(tile_x % 4 == 0);
   assert(tile_y % 2 == 0);
   surf[5] = ((tile_x / 4) << BRW_SURFACE_X_OFFSET_SHIFT |
	      (tile_y / 2) << BRW_SURFACE_Y_OFFSET_SHIFT |
	      (mt->align_h == 4 ? BRW_SURFACE_VERTICAL_ALIGN_ENABLE : 0));

   if (brw->gen < 6) {
      /* _NEW_COLOR */
      if (!ctx->Color.ColorLogicOpEnabled &&
	  (ctx->Color.BlendEnabled & (1 << unit)))
	 surf[0] |= BRW_SURFACE_BLEND_ENABLED;

      if (!ctx->Color.ColorMask[unit][0])
	 surf[0] |= 1 << BRW_SURFACE_WRITEDISABLE_R_SHIFT;
      if (!ctx->Color.ColorMask[unit][1])
	 surf[0] |= 1 << BRW_SURFACE_WRITEDISABLE_G_SHIFT;
      if (!ctx->Color.ColorMask[unit][2])
	 surf[0] |= 1 << BRW_SURFACE_WRITEDISABLE_B_SHIFT;

      /* As mentioned above, disable writes to the alpha component when the
       * renderbuffer is XRGB.
       */
      if (ctx->DrawBuffer->Visual.alphaBits == 0 ||
	  !ctx->Color.ColorMask[unit][3]) {
	 surf[0] |= 1 << BRW_SURFACE_WRITEDISABLE_A_SHIFT;
      }
   }

   drm_intel_bo_emit_reloc(brw->batch.bo,
                           offset + 4,
                           mt->bo,
                           surf[1] - mt->bo->offset64,
                           I915_GEM_DOMAIN_RENDER,
                           I915_GEM_DOMAIN_RENDER);

   return offset;
}

/**
 * Construct SURFACE_STATE objects for renderbuffers/draw buffers.
 */
void
brw_update_renderbuffer_surfaces(struct brw_context *brw,
                                 const struct gl_framebuffer *fb,
                                 uint32_t render_target_start,
                                 uint32_t *surf_offset)
{
   GLuint i;

   /* Update surfaces for drawing buffers */
   if (fb->_NumColorDrawBuffers >= 1) {
      for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
         const uint32_t surf_index = render_target_start + i;

	 if (intel_renderbuffer(fb->_ColorDrawBuffers[i])) {
            surf_offset[surf_index] = 
               brw->vtbl.update_renderbuffer_surface(
                  brw, fb->_ColorDrawBuffers[i],
                  fb->MaxNumLayers > 0, i, surf_index);
	 } else {
            brw->vtbl.emit_null_surface_state(
               brw, fb->Width, fb->Height, fb->Visual.samples,
               &surf_offset[surf_index]);
	 }
      }
   } else {
      const uint32_t surf_index = render_target_start;
      brw->vtbl.emit_null_surface_state(
         brw, fb->Width, fb->Height, fb->Visual.samples,
         &surf_offset[surf_index]);
   }
}

static void
update_renderbuffer_surfaces(struct brw_context *brw)
{
   const struct gl_context *ctx = &brw->ctx;

   /* _NEW_BUFFERS | _NEW_COLOR */
   const struct gl_framebuffer *fb = ctx->DrawBuffer;
   brw_update_renderbuffer_surfaces(
      brw, fb,
      brw->wm.prog_data->binding_table.render_target_start,
      brw->wm.base.surf_offset);
   brw->ctx.NewDriverState |= BRW_NEW_SURFACES;
}

const struct brw_tracked_state brw_renderbuffer_surfaces = {
   .dirty = {
      .mesa = _NEW_BUFFERS |
              _NEW_COLOR,
      .brw = BRW_NEW_BATCH |
             BRW_NEW_FS_PROG_DATA,
   },
   .emit = update_renderbuffer_surfaces,
};

const struct brw_tracked_state gen6_renderbuffer_surfaces = {
   .dirty = {
      .mesa = _NEW_BUFFERS,
      .brw = BRW_NEW_BATCH,
   },
   .emit = update_renderbuffer_surfaces,
};


static void
update_stage_texture_surfaces(struct brw_context *brw,
                              const struct gl_program *prog,
                              struct brw_stage_state *stage_state,
                              bool for_gather)
{
   if (!prog)
      return;

   struct gl_context *ctx = &brw->ctx;

   uint32_t *surf_offset = stage_state->surf_offset;

   /* BRW_NEW_*_PROG_DATA */
   if (for_gather)
      surf_offset += stage_state->prog_data->binding_table.gather_texture_start;
   else
      surf_offset += stage_state->prog_data->binding_table.texture_start;

   unsigned num_samplers = _mesa_fls(prog->SamplersUsed);
   for (unsigned s = 0; s < num_samplers; s++) {
      surf_offset[s] = 0;

      if (prog->SamplersUsed & (1 << s)) {
         const unsigned unit = prog->SamplerUnits[s];

         /* _NEW_TEXTURE */
         if (ctx->Texture.Unit[unit]._Current) {
            brw->vtbl.update_texture_surface(ctx, unit, surf_offset + s, for_gather);
         }
      }
   }
}


/**
 * Construct SURFACE_STATE objects for enabled textures.
 */
static void
brw_update_texture_surfaces(struct brw_context *brw)
{
   /* BRW_NEW_VERTEX_PROGRAM */
   struct gl_program *vs = (struct gl_program *) brw->vertex_program;

   /* BRW_NEW_GEOMETRY_PROGRAM */
   struct gl_program *gs = (struct gl_program *) brw->geometry_program;

   /* BRW_NEW_FRAGMENT_PROGRAM */
   struct gl_program *fs = (struct gl_program *) brw->fragment_program;

   /* _NEW_TEXTURE */
   update_stage_texture_surfaces(brw, vs, &brw->vs.base, false);
   update_stage_texture_surfaces(brw, gs, &brw->gs.base, false);
   update_stage_texture_surfaces(brw, fs, &brw->wm.base, false);

   /* emit alternate set of surface state for gather. this
    * allows the surface format to be overriden for only the
    * gather4 messages. */
   if (brw->gen < 8) {
      if (vs && vs->UsesGather)
         update_stage_texture_surfaces(brw, vs, &brw->vs.base, true);
      if (gs && gs->UsesGather)
         update_stage_texture_surfaces(brw, gs, &brw->gs.base, true);
      if (fs && fs->UsesGather)
         update_stage_texture_surfaces(brw, fs, &brw->wm.base, true);
   }

   brw->ctx.NewDriverState |= BRW_NEW_SURFACES;
}

const struct brw_tracked_state brw_texture_surfaces = {
   .dirty = {
      .mesa = _NEW_TEXTURE,
      .brw = BRW_NEW_BATCH |
             BRW_NEW_FRAGMENT_PROGRAM |
             BRW_NEW_FS_PROG_DATA |
             BRW_NEW_GEOMETRY_PROGRAM |
             BRW_NEW_GS_PROG_DATA |
             BRW_NEW_TEXTURE_BUFFER |
             BRW_NEW_VERTEX_PROGRAM |
             BRW_NEW_VS_PROG_DATA,
   },
   .emit = brw_update_texture_surfaces,
};

void
brw_upload_ubo_surfaces(struct brw_context *brw,
			struct gl_shader *shader,
                        struct brw_stage_state *stage_state,
                        struct brw_stage_prog_data *prog_data,
                        bool dword_pitch)
{
   struct gl_context *ctx = &brw->ctx;

   if (!shader)
      return;

   uint32_t *surf_offsets =
      &stage_state->surf_offset[prog_data->binding_table.ubo_start];

   for (int i = 0; i < shader->NumUniformBlocks; i++) {
      struct gl_uniform_buffer_binding *binding;
      struct intel_buffer_object *intel_bo;

      binding = &ctx->UniformBufferBindings[shader->UniformBlocks[i].Binding];
      intel_bo = intel_buffer_object(binding->BufferObject);
      drm_intel_bo *bo =
         intel_bufferobj_buffer(brw, intel_bo,
                                binding->Offset,
                                binding->BufferObject->Size - binding->Offset);

      /* Because behavior for referencing outside of the binding's size in the
       * glBindBufferRange case is undefined, we can just bind the whole buffer
       * glBindBufferBase wants and be a correct implementation.
       */
      brw_create_constant_surface(brw, bo, binding->Offset,
                                  bo->size - binding->Offset,
                                  &surf_offsets[i],
                                  dword_pitch);
   }

   if (shader->NumUniformBlocks)
      brw->ctx.NewDriverState |= BRW_NEW_SURFACES;
}

static void
brw_upload_wm_ubo_surfaces(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   /* _NEW_PROGRAM */
   struct gl_shader_program *prog = ctx->_Shader->_CurrentFragmentProgram;

   if (!prog)
      return;

   /* BRW_NEW_FS_PROG_DATA */
   brw_upload_ubo_surfaces(brw, prog->_LinkedShaders[MESA_SHADER_FRAGMENT],
                           &brw->wm.base, &brw->wm.prog_data->base, true);
}

const struct brw_tracked_state brw_wm_ubo_surfaces = {
   .dirty = {
      .mesa = _NEW_PROGRAM,
      .brw = BRW_NEW_BATCH |
             BRW_NEW_FS_PROG_DATA |
             BRW_NEW_UNIFORM_BUFFER,
   },
   .emit = brw_upload_wm_ubo_surfaces,
};

void
brw_upload_abo_surfaces(struct brw_context *brw,
			struct gl_shader_program *prog,
                        struct brw_stage_state *stage_state,
                        struct brw_stage_prog_data *prog_data)
{
   struct gl_context *ctx = &brw->ctx;
   uint32_t *surf_offsets =
      &stage_state->surf_offset[prog_data->binding_table.abo_start];

   for (int i = 0; i < prog->NumAtomicBuffers; i++) {
      struct gl_atomic_buffer_binding *binding =
         &ctx->AtomicBufferBindings[prog->AtomicBuffers[i].Binding];
      struct intel_buffer_object *intel_bo =
         intel_buffer_object(binding->BufferObject);
      drm_intel_bo *bo = intel_bufferobj_buffer(
         brw, intel_bo, binding->Offset, intel_bo->Base.Size - binding->Offset);

      brw->vtbl.emit_buffer_surface_state(brw, &surf_offsets[i], bo,
                                          binding->Offset, BRW_SURFACEFORMAT_RAW,
                                          bo->size - binding->Offset, 1, true);
   }

   if (prog->NumAtomicBuffers)
      brw->ctx.NewDriverState |= BRW_NEW_SURFACES;
}

static void
brw_upload_wm_abo_surfaces(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   /* _NEW_PROGRAM */
   struct gl_shader_program *prog = ctx->Shader._CurrentFragmentProgram;

   if (prog) {
      /* BRW_NEW_FS_PROG_DATA */
      brw_upload_abo_surfaces(brw, prog, &brw->wm.base,
                              &brw->wm.prog_data->base);
   }
}

const struct brw_tracked_state brw_wm_abo_surfaces = {
   .dirty = {
      .mesa = _NEW_PROGRAM,
      .brw = BRW_NEW_ATOMIC_BUFFER |
             BRW_NEW_BATCH |
             BRW_NEW_FS_PROG_DATA,
   },
   .emit = brw_upload_wm_abo_surfaces,
};

static void
brw_upload_cs_abo_surfaces(struct brw_context *brw)
{
   struct gl_context *ctx = &brw->ctx;
   /* _NEW_PROGRAM */
   struct gl_shader_program *prog =
      ctx->_Shader->CurrentProgram[MESA_SHADER_COMPUTE];

   if (prog) {
      /* BRW_NEW_CS_PROG_DATA */
      brw_upload_abo_surfaces(brw, prog, &brw->cs.base,
                              &brw->cs.prog_data->base);
   }
}

const struct brw_tracked_state brw_cs_abo_surfaces = {
   .dirty = {
      .mesa = _NEW_PROGRAM,
      .brw = BRW_NEW_ATOMIC_BUFFER |
             BRW_NEW_BATCH |
             BRW_NEW_CS_PROG_DATA,
   },
   .emit = brw_upload_cs_abo_surfaces,
};

void
gen4_init_vtable_surface_functions(struct brw_context *brw)
{
   brw->vtbl.update_texture_surface = brw_update_texture_surface;
   brw->vtbl.update_renderbuffer_surface = brw_update_renderbuffer_surface;
   brw->vtbl.emit_null_surface_state = brw_emit_null_surface_state;
   brw->vtbl.emit_buffer_surface_state = gen4_emit_buffer_surface_state;
}