summaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_nir.c
blob: ed836bf36bc1863240db2e1ab7e4a356186fb1e4 (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
/*
 * Copyright © 2014 Intel Corporation
 *
 * 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 AUTHORS OR COPYRIGHT HOLDERS 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 "brw_nir.h"
#include "brw_shader.h"
#include "compiler/nir/glsl_to_nir.h"
#include "compiler/nir/nir_builder.h"
#include "program/prog_to_nir.h"

static bool
is_input(nir_intrinsic_instr *intrin)
{
   return intrin->intrinsic == nir_intrinsic_load_input ||
          intrin->intrinsic == nir_intrinsic_load_per_vertex_input;
}

static bool
is_output(nir_intrinsic_instr *intrin)
{
   return intrin->intrinsic == nir_intrinsic_load_output ||
          intrin->intrinsic == nir_intrinsic_load_per_vertex_output ||
          intrin->intrinsic == nir_intrinsic_store_output ||
          intrin->intrinsic == nir_intrinsic_store_per_vertex_output;
}

/**
 * In many cases, we just add the base and offset together, so there's no
 * reason to keep them separate.  Sometimes, combining them is essential:
 * if a shader only accesses part of a compound variable (such as a matrix
 * or array), the variable's base may not actually exist in the VUE map.
 *
 * This pass adds constant offsets to instr->const_index[0], and resets
 * the offset source to 0.  Non-constant offsets remain unchanged - since
 * we don't know what part of a compound variable is accessed, we allocate
 * storage for the entire thing.
 */
struct add_const_offset_to_base_params {
   nir_builder b;
   nir_variable_mode mode;
};

static bool
add_const_offset_to_base_block(nir_block *block, void *closure)
{
   struct add_const_offset_to_base_params *params = closure;
   nir_builder *b = &params->b;

   nir_foreach_instr_safe(block, instr) {
      if (instr->type != nir_instr_type_intrinsic)
         continue;

      nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);

      if ((params->mode == nir_var_shader_in && is_input(intrin)) ||
          (params->mode == nir_var_shader_out && is_output(intrin))) {
         nir_src *offset = nir_get_io_offset_src(intrin);
         nir_const_value *const_offset = nir_src_as_const_value(*offset);

         if (const_offset) {
            intrin->const_index[0] += const_offset->u[0];
            b->cursor = nir_before_instr(&intrin->instr);
            nir_instr_rewrite_src(&intrin->instr, offset,
                                  nir_src_for_ssa(nir_imm_int(b, 0)));
         }
      }
   }
   return true;
}

static void
add_const_offset_to_base(nir_shader *nir, nir_variable_mode mode)
{
   struct add_const_offset_to_base_params params = { .mode = mode };

   nir_foreach_function(nir, f) {
      if (f->impl) {
         nir_builder_init(&params.b, f->impl);
         nir_foreach_block(f->impl, add_const_offset_to_base_block, &params);
      }
   }
}

static bool
remap_vs_attrs(nir_block *block, void *closure)
{
   GLbitfield64 inputs_read = *((GLbitfield64 *) closure);

   nir_foreach_instr(block, instr) {
      if (instr->type != nir_instr_type_intrinsic)
         continue;

      nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);

      if (intrin->intrinsic == nir_intrinsic_load_input) {
         /* Attributes come in a contiguous block, ordered by their
          * gl_vert_attrib value.  That means we can compute the slot
          * number for an attribute by masking out the enabled attributes
          * before it and counting the bits.
          */
         int attr = intrin->const_index[0];
         int slot = _mesa_bitcount_64(inputs_read & BITFIELD64_MASK(attr));

         intrin->const_index[0] = 4 * slot;
      }
   }
   return true;
}

static bool
remap_inputs_with_vue_map(nir_block *block, void *closure)
{
   const struct brw_vue_map *vue_map = closure;

   nir_foreach_instr(block, instr) {
      if (instr->type != nir_instr_type_intrinsic)
         continue;

      nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);

      if (intrin->intrinsic == nir_intrinsic_load_input ||
          intrin->intrinsic == nir_intrinsic_load_per_vertex_input) {
         int vue_slot = vue_map->varying_to_slot[intrin->const_index[0]];
         assert(vue_slot != -1);
         intrin->const_index[0] = vue_slot;
      }
   }
   return true;
}

struct remap_patch_urb_offsets_state {
   nir_builder b;
   struct brw_vue_map vue_map;
};

static bool
remap_patch_urb_offsets(nir_block *block, void *closure)
{
   struct remap_patch_urb_offsets_state *state = closure;

   nir_foreach_instr_safe(block, instr) {
      if (instr->type != nir_instr_type_intrinsic)
         continue;

      nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr);

      gl_shader_stage stage = state->b.shader->stage;

      if ((stage == MESA_SHADER_TESS_CTRL && is_output(intrin)) ||
          (stage == MESA_SHADER_TESS_EVAL && is_input(intrin))) {
         int vue_slot = state->vue_map.varying_to_slot[intrin->const_index[0]];
         assert(vue_slot != -1);
         intrin->const_index[0] = vue_slot;

         nir_src *vertex = nir_get_io_vertex_index_src(intrin);
         if (vertex) {
            nir_const_value *const_vertex = nir_src_as_const_value(*vertex);
            if (const_vertex) {
               intrin->const_index[0] += const_vertex->u[0] *
                                         state->vue_map.num_per_vertex_slots;
            } else {
               state->b.cursor = nir_before_instr(&intrin->instr);

               /* Multiply by the number of per-vertex slots. */
               nir_ssa_def *vertex_offset =
                  nir_imul(&state->b,
                           nir_ssa_for_src(&state->b, *vertex, 1),
                           nir_imm_int(&state->b,
                                       state->vue_map.num_per_vertex_slots));

               /* Add it to the existing offset */
               nir_src *offset = nir_get_io_offset_src(intrin);
               nir_ssa_def *total_offset =
                  nir_iadd(&state->b, vertex_offset,
                           nir_ssa_for_src(&state->b, *offset, 1));

               nir_instr_rewrite_src(&intrin->instr, offset,
                                     nir_src_for_ssa(total_offset));
            }
         }
      }
   }
   return true;
}

static void
brw_nir_lower_vs_inputs(nir_shader *nir,
                        const struct brw_device_info *devinfo,
                        bool is_scalar,
                        bool use_legacy_snorm_formula,
                        const uint8_t *vs_attrib_wa_flags)
{
   /* Start with the location of the variable's base. */
   foreach_list_typed(nir_variable, var, node, &nir->inputs) {
      var->data.driver_location = var->data.location;
   }

   /* Now use nir_lower_io to walk dereference chains.  Attribute arrays
    * are loaded as one vec4 per element (or matrix column), so we use
    * type_size_vec4 here.
    */
   nir_lower_io(nir, nir_var_shader_in, type_size_vec4);

   /* This pass needs actual constants */
   nir_opt_constant_folding(nir);

   add_const_offset_to_base(nir, nir_var_shader_in);

   brw_nir_apply_attribute_workarounds(nir, use_legacy_snorm_formula,
                                       vs_attrib_wa_flags);

   if (is_scalar) {
      /* Finally, translate VERT_ATTRIB_* values into the actual registers.
       *
       * Note that we can use nir->info.inputs_read instead of
       * key->inputs_read since the two are identical aside from Gen4-5
       * edge flag differences.
       */
      GLbitfield64 inputs_read = nir->info.inputs_read;

      nir_foreach_function(nir, function) {
         if (function->impl) {
            nir_foreach_block(function->impl, remap_vs_attrs, &inputs_read);
         }
      }
   }
}

static void
brw_nir_lower_vue_inputs(nir_shader *nir,
                         const struct brw_device_info *devinfo,
                         bool is_scalar)
{
   if (!is_scalar && nir->stage == MESA_SHADER_GEOMETRY) {
      foreach_list_typed(nir_variable, var, node, &nir->inputs) {
         var->data.driver_location = var->data.location;
      }
      nir_lower_io(nir, nir_var_shader_in, type_size_vec4);
   } else {
      /* The GLSL linker will have already matched up GS inputs and
       * the outputs of prior stages.  The driver does extend VS outputs
       * in some cases, but only for legacy OpenGL or Gen4-5 hardware,
       * neither of which offer geometry shader support.  So we can
       * safely ignore that.
       *
       * For SSO pipelines, we use a fixed VUE map layout based on variable
       * locations, so we can rely on rendezvous-by-location to make this
       * work.
       *
       * However, we need to ignore VARYING_SLOT_PRIMITIVE_ID, as it's not
       * written by previous stages and shows up via payload magic.
       */
      struct brw_vue_map input_vue_map;
      GLbitfield64 inputs_read =
         nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID;
      brw_compute_vue_map(devinfo, &input_vue_map, inputs_read,
                          nir->info.separate_shader ||
                          nir->stage == MESA_SHADER_TESS_CTRL);

      foreach_list_typed(nir_variable, var, node, &nir->inputs) {
         var->data.driver_location = var->data.location;
      }

      /* Inputs are stored in vec4 slots, so use type_size_vec4(). */
      nir_lower_io(nir, nir_var_shader_in, type_size_vec4);

      /* This pass needs actual constants */
      nir_opt_constant_folding(nir);

      add_const_offset_to_base(nir, nir_var_shader_in);

      nir_foreach_function(nir, function) {
         if (function->impl) {
            nir_foreach_block(function->impl, remap_inputs_with_vue_map,
                              &input_vue_map);
         }
      }
   }
}

static void
brw_nir_lower_tes_inputs(nir_shader *nir)
{
   struct remap_patch_urb_offsets_state state;
   brw_compute_tess_vue_map(&state.vue_map,
                            nir->info.inputs_read & ~VARYING_BIT_PRIMITIVE_ID,
                            nir->info.patch_inputs_read);

   foreach_list_typed(nir_variable, var, node, &nir->inputs) {
      var->data.driver_location = var->data.location;
   }

   nir_lower_io(nir, nir_var_shader_in, type_size_vec4);

   /* This pass needs actual constants */
   nir_opt_constant_folding(nir);

   add_const_offset_to_base(nir, nir_var_shader_in);

   nir_foreach_function(nir, function) {
      if (function->impl) {
         nir_builder_init(&state.b, function->impl);
         nir_foreach_block(function->impl, remap_patch_urb_offsets, &state);
      }
   }
}

static void
brw_nir_lower_fs_inputs(nir_shader *nir)
{
   nir_assign_var_locations(&nir->inputs, &nir->num_inputs, type_size_scalar);
   nir_lower_io(nir, nir_var_shader_in, type_size_scalar);
}

static void
brw_nir_lower_inputs(nir_shader *nir,
                     const struct brw_device_info *devinfo,
                     bool is_scalar,
                     bool use_legacy_snorm_formula,
                     const uint8_t *vs_attrib_wa_flags)
{
   switch (nir->stage) {
   case MESA_SHADER_VERTEX:
      brw_nir_lower_vs_inputs(nir, devinfo, is_scalar, use_legacy_snorm_formula,
                              vs_attrib_wa_flags);
      break;
   case MESA_SHADER_TESS_CTRL:
   case MESA_SHADER_GEOMETRY:
      brw_nir_lower_vue_inputs(nir, devinfo, is_scalar);
      break;
   case MESA_SHADER_TESS_EVAL:
      brw_nir_lower_tes_inputs(nir);
      break;
   case MESA_SHADER_FRAGMENT:
      assert(is_scalar);
      brw_nir_lower_fs_inputs(nir);
      break;
   case MESA_SHADER_COMPUTE:
      /* Compute shaders have no inputs. */
      assert(exec_list_is_empty(&nir->inputs));
      break;
   default:
      unreachable("unsupported shader stage");
   }
}

static void
brw_nir_lower_vue_outputs(nir_shader *nir,
                          bool is_scalar)
{
   if (is_scalar) {
      nir_assign_var_locations(&nir->outputs, &nir->num_outputs,
                               type_size_vec4_times_4);
      nir_lower_io(nir, nir_var_shader_out, type_size_vec4_times_4);
   } else {
      nir_foreach_variable(var, &nir->outputs)
         var->data.driver_location = var->data.location;
      nir_lower_io(nir, nir_var_shader_out, type_size_vec4);
   }
}

static void
brw_nir_lower_tcs_outputs(nir_shader *nir)
{
   struct remap_patch_urb_offsets_state state;
   brw_compute_tess_vue_map(&state.vue_map, nir->info.outputs_written,
                            nir->info.patch_outputs_written);

   nir_foreach_variable(var, &nir->outputs) {
      var->data.driver_location = var->data.location;
   }

   nir_lower_io(nir, nir_var_shader_out, type_size_vec4);

   /* This pass needs actual constants */
   nir_opt_constant_folding(nir);

   add_const_offset_to_base(nir, nir_var_shader_out);

   nir_foreach_function(nir, function) {
      if (function->impl) {
         nir_builder_init(&state.b, function->impl);
         nir_foreach_block(function->impl, remap_patch_urb_offsets, &state);
      }
   }
}

static void
brw_nir_lower_fs_outputs(nir_shader *nir)
{
   nir_assign_var_locations(&nir->outputs, &nir->num_outputs,
                            type_size_scalar);
   nir_lower_io(nir, nir_var_shader_out, type_size_scalar);
}

static void
brw_nir_lower_outputs(nir_shader *nir, bool is_scalar)
{
   switch (nir->stage) {
   case MESA_SHADER_VERTEX:
   case MESA_SHADER_TESS_EVAL:
   case MESA_SHADER_GEOMETRY:
      brw_nir_lower_vue_outputs(nir, is_scalar);
      break;
   case MESA_SHADER_TESS_CTRL:
      brw_nir_lower_tcs_outputs(nir);
      break;
   case MESA_SHADER_FRAGMENT:
      brw_nir_lower_fs_outputs(nir);
      break;
   case MESA_SHADER_COMPUTE:
      /* Compute shaders have no outputs. */
      assert(exec_list_is_empty(&nir->outputs));
      break;
   default:
      unreachable("unsupported shader stage");
   }
}

static int
type_size_scalar_bytes(const struct glsl_type *type)
{
   return type_size_scalar(type) * 4;
}

static int
type_size_vec4_bytes(const struct glsl_type *type)
{
   return type_size_vec4(type) * 16;
}

static void
brw_nir_lower_uniforms(nir_shader *nir, bool is_scalar)
{
   if (is_scalar) {
      nir_assign_var_locations(&nir->uniforms, &nir->num_uniforms,
                               type_size_scalar_bytes);
      nir_lower_io(nir, nir_var_uniform, type_size_scalar_bytes);
   } else {
      nir_assign_var_locations(&nir->uniforms, &nir->num_uniforms,
                               type_size_vec4_bytes);
      nir_lower_io(nir, nir_var_uniform, type_size_vec4_bytes);
   }
}

#define OPT(pass, ...) ({                                  \
   bool this_progress = false;                             \
   NIR_PASS(this_progress, nir, pass, ##__VA_ARGS__);      \
   if (this_progress)                                      \
      progress = true;                                     \
   this_progress;                                          \
})

#define OPT_V(pass, ...) NIR_PASS_V(nir, pass, ##__VA_ARGS__)

static nir_shader *
nir_optimize(nir_shader *nir, bool is_scalar)
{
   bool progress;
   do {
      progress = false;
      OPT_V(nir_lower_vars_to_ssa);

      if (is_scalar) {
         OPT_V(nir_lower_alu_to_scalar);
      }

      OPT(nir_copy_prop);

      if (is_scalar) {
         OPT_V(nir_lower_phis_to_scalar);
      }

      OPT(nir_copy_prop);
      OPT(nir_opt_dce);
      OPT(nir_opt_cse);
      OPT(nir_opt_peephole_select);
      OPT(nir_opt_algebraic);
      OPT(nir_opt_constant_folding);
      OPT(nir_opt_dead_cf);
      OPT(nir_opt_remove_phis);
      OPT(nir_opt_undef);
   } while (progress);

   return nir;
}

/* Does some simple lowering and runs the standard suite of optimizations
 *
 * This is intended to be called more-or-less directly after you get the
 * shader out of GLSL or some other source.  While it is geared towards i965,
 * it is not at all generator-specific except for the is_scalar flag.  Even
 * there, it is safe to call with is_scalar = false for a shader that is
 * intended for the FS backend as long as nir_optimize is called again with
 * is_scalar = true to scalarize everything prior to code gen.
 */
nir_shader *
brw_preprocess_nir(nir_shader *nir, bool is_scalar)
{
   bool progress; /* Written by OPT and OPT_V */
   (void)progress;

   if (nir->stage == MESA_SHADER_GEOMETRY)
      OPT(nir_lower_gs_intrinsics);

   static const nir_lower_tex_options tex_options = {
      .lower_txp = ~0,
   };

   OPT(nir_lower_tex, &tex_options);
   OPT(nir_normalize_cubemap_coords);

   OPT(nir_lower_global_vars_to_local);

   OPT(nir_split_var_copies);

   nir = nir_optimize(nir, is_scalar);

   if (is_scalar) {
      OPT_V(nir_lower_load_const_to_scalar);
   }

   /* Lower a bunch of stuff */
   OPT_V(nir_lower_var_copies);

   /* Get rid of split copies */
   nir = nir_optimize(nir, is_scalar);

   OPT(nir_remove_dead_variables);

   return nir;
}

/** Lower input and output loads and stores for i965. */
nir_shader *
brw_nir_lower_io(nir_shader *nir,
                 const struct brw_device_info *devinfo,
                 bool is_scalar,
                 bool use_legacy_snorm_formula,
                 const uint8_t *vs_attrib_wa_flags)
{
   bool progress; /* Written by OPT and OPT_V */
   (void)progress;

   OPT_V(brw_nir_lower_inputs, devinfo, is_scalar,
         use_legacy_snorm_formula, vs_attrib_wa_flags);
   OPT_V(brw_nir_lower_outputs, is_scalar);

   return nir;
}

/* Prepare the given shader for codegen
 *
 * This function is intended to be called right before going into the actual
 * backend and is highly backend-specific.  Also, once this function has been
 * called on a shader, it will no longer be in SSA form so most optimizations
 * will not work.
 */
nir_shader *
brw_postprocess_nir(nir_shader *nir,
                    const struct brw_device_info *devinfo,
                    bool is_scalar)
{
   bool debug_enabled =
      (INTEL_DEBUG & intel_debug_flag_for_shader_stage(nir->stage));

   bool progress; /* Written by OPT and OPT_V */
   (void)progress;

   nir = nir_optimize(nir, is_scalar);

   if (devinfo->gen >= 6) {
      /* Try and fuse multiply-adds */
      OPT(brw_nir_opt_peephole_ffma);
   }

   OPT(nir_opt_algebraic_late);

   OPT(nir_lower_locals_to_regs);

   OPT_V(nir_lower_to_source_mods);
   OPT(nir_copy_prop);
   OPT(nir_opt_dce);

   if (unlikely(debug_enabled)) {
      /* Re-index SSA defs so we print more sensible numbers. */
      nir_foreach_function(nir, function) {
         if (function->impl)
            nir_index_ssa_defs(function->impl);
      }

      fprintf(stderr, "NIR (SSA form) for %s shader:\n",
              _mesa_shader_stage_to_string(nir->stage));
      nir_print_shader(nir, stderr);
   }

   OPT_V(nir_convert_from_ssa, true);

   if (!is_scalar) {
      OPT_V(nir_move_vec_src_uses_to_dest);
      OPT(nir_lower_vec_to_movs);
   }

   /* This is the last pass we run before we start emitting stuff.  It
    * determines when we need to insert boolean resolves on Gen <= 5.  We
    * run it last because it stashes data in instr->pass_flags and we don't
    * want that to be squashed by other NIR passes.
    */
   if (devinfo->gen <= 5)
      brw_nir_analyze_boolean_resolves(nir);

   nir_sweep(nir);

   if (unlikely(debug_enabled)) {
      fprintf(stderr, "NIR (final form) for %s shader:\n",
              _mesa_shader_stage_to_string(nir->stage));
      nir_print_shader(nir, stderr);
   }

   return nir;
}

nir_shader *
brw_create_nir(struct brw_context *brw,
               const struct gl_shader_program *shader_prog,
               const struct gl_program *prog,
               gl_shader_stage stage,
               bool is_scalar)
{
   struct gl_context *ctx = &brw->ctx;
   const nir_shader_compiler_options *options =
      ctx->Const.ShaderCompilerOptions[stage].NirOptions;
   bool progress;
   nir_shader *nir;

   /* First, lower the GLSL IR or Mesa IR to NIR */
   if (shader_prog) {
      nir = glsl_to_nir(shader_prog, stage, options);
   } else {
      nir = prog_to_nir(prog, options);
      OPT_V(nir_convert_to_ssa); /* turn registers into SSA */
   }
   nir_validate_shader(nir);

   (void)progress;

   nir = brw_preprocess_nir(nir, is_scalar);

   OPT(nir_lower_system_values);
   OPT_V(brw_nir_lower_uniforms, is_scalar);

   if (shader_prog) {
      OPT_V(nir_lower_samplers, shader_prog);
      OPT_V(nir_lower_atomics, shader_prog);
   }

   return nir;
}

nir_shader *
brw_nir_apply_sampler_key(nir_shader *nir,
                          const struct brw_device_info *devinfo,
                          const struct brw_sampler_prog_key_data *key_tex,
                          bool is_scalar)
{
   nir_lower_tex_options tex_options = { 0 };

   /* Iron Lake and prior require lowering of all rectangle textures */
   if (devinfo->gen < 6)
      tex_options.lower_rect = true;

   /* Prior to Broadwell, our hardware can't actually do GL_CLAMP */
   if (devinfo->gen < 8) {
      tex_options.saturate_s = key_tex->gl_clamp_mask[0];
      tex_options.saturate_t = key_tex->gl_clamp_mask[1];
      tex_options.saturate_r = key_tex->gl_clamp_mask[2];
   }

   /* Prior to Haswell, we have to fake texture swizzle */
   for (unsigned s = 0; s < MAX_SAMPLERS; s++) {
      if (key_tex->swizzles[s] == SWIZZLE_NOOP)
         continue;

      tex_options.swizzle_result |= (1 << s);
      for (unsigned c = 0; c < 4; c++)
         tex_options.swizzles[s][c] = GET_SWZ(key_tex->swizzles[s], c);
   }

   if (nir_lower_tex(nir, &tex_options)) {
      nir_validate_shader(nir);
      nir = nir_optimize(nir, is_scalar);
   }

   return nir;
}

enum brw_reg_type
brw_type_for_nir_type(nir_alu_type type)
{
   switch (type) {
   case nir_type_uint:
      return BRW_REGISTER_TYPE_UD;
   case nir_type_bool:
   case nir_type_int:
      return BRW_REGISTER_TYPE_D;
   case nir_type_float:
      return BRW_REGISTER_TYPE_F;
   default:
      unreachable("unknown type");
   }

   return BRW_REGISTER_TYPE_F;
}

/* Returns the glsl_base_type corresponding to a nir_alu_type.
 * This is used by both brw_vec4_nir and brw_fs_nir.
 */
enum glsl_base_type
brw_glsl_base_type_for_nir_type(nir_alu_type type)
{
   switch (type) {
   case nir_type_float:
      return GLSL_TYPE_FLOAT;

   case nir_type_int:
      return GLSL_TYPE_INT;

   case nir_type_uint:
      return GLSL_TYPE_UINT;

   default:
      unreachable("bad type");
   }
}