summaryrefslogtreecommitdiffstats
path: root/src/mesa/program/prog_to_nir.c
blob: cb1c19e9dfa2f74528719600f744a00576e902b4 (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
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
/*
 * Copyright © 2015 Intel Corporation
 * Copyright © 2014-2015 Broadcom
 * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
 *
 * 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 "compiler/nir/nir.h"
#include "compiler/nir/nir_builder.h"
#include "compiler/glsl/list.h"
#include "main/imports.h"
#include "main/mtypes.h"
#include "util/ralloc.h"

#include "prog_to_nir.h"
#include "prog_instruction.h"
#include "prog_parameter.h"
#include "prog_print.h"
#include "program.h"

/**
 * \file prog_to_nir.c
 *
 * A translator from Mesa IR (prog_instruction.h) to NIR.  This is primarily
 * intended to support ARB_vertex_program, ARB_fragment_program, and fixed-function
 * vertex processing.  Full GLSL support should use glsl_to_nir instead.
 */

struct ptn_compile {
   const struct gl_program *prog;
   nir_builder build;
   bool error;

   nir_variable *parameters;
   nir_variable *input_vars[VARYING_SLOT_MAX];
   nir_variable *output_vars[VARYING_SLOT_MAX];
   nir_variable *sysval_vars[SYSTEM_VALUE_MAX];
   nir_variable *sampler_vars[32]; /* matches number of bits in TexSrcUnit */
   nir_register **output_regs;
   nir_register **temp_regs;

   nir_register *addr_reg;
};

#define SWIZ(X, Y, Z, W) \
   (unsigned[4]){ SWIZZLE_##X, SWIZZLE_##Y, SWIZZLE_##Z, SWIZZLE_##W }
#define ptn_channel(b, src, ch) nir_swizzle(b, src, SWIZ(ch, ch, ch, ch), 1, true)

static nir_ssa_def *
ptn_src_for_dest(struct ptn_compile *c, nir_alu_dest *dest)
{
   nir_builder *b = &c->build;

   nir_alu_src src;
   memset(&src, 0, sizeof(src));

   if (dest->dest.is_ssa)
      src.src = nir_src_for_ssa(&dest->dest.ssa);
   else {
      assert(!dest->dest.reg.indirect);
      src.src = nir_src_for_reg(dest->dest.reg.reg);
      src.src.reg.base_offset = dest->dest.reg.base_offset;
   }

   for (int i = 0; i < 4; i++)
      src.swizzle[i] = i;

   return nir_fmov_alu(b, src, 4);
}

static nir_alu_dest
ptn_get_dest(struct ptn_compile *c, const struct prog_dst_register *prog_dst)
{
   nir_alu_dest dest;

   memset(&dest, 0, sizeof(dest));

   switch (prog_dst->File) {
   case PROGRAM_TEMPORARY:
      dest.dest.reg.reg = c->temp_regs[prog_dst->Index];
      break;
   case PROGRAM_OUTPUT:
      dest.dest.reg.reg = c->output_regs[prog_dst->Index];
      break;
   case PROGRAM_ADDRESS:
      assert(prog_dst->Index == 0);
      dest.dest.reg.reg = c->addr_reg;
      break;
   case PROGRAM_UNDEFINED:
      break;
   }

   dest.write_mask = prog_dst->WriteMask;
   dest.saturate = false;

   assert(!prog_dst->RelAddr);

   return dest;
}

static nir_ssa_def *
ptn_get_src(struct ptn_compile *c, const struct prog_src_register *prog_src)
{
   nir_builder *b = &c->build;
   nir_alu_src src;

   memset(&src, 0, sizeof(src));

   switch (prog_src->File) {
   case PROGRAM_UNDEFINED:
      return nir_imm_float(b, 0.0);
   case PROGRAM_TEMPORARY:
      assert(!prog_src->RelAddr && prog_src->Index >= 0);
      src.src.reg.reg = c->temp_regs[prog_src->Index];
      break;
   case PROGRAM_INPUT: {
      /* ARB_vertex_program doesn't allow relative addressing on vertex
       * attributes; ARB_fragment_program has no relative addressing at all.
       */
      assert(!prog_src->RelAddr);

      assert(prog_src->Index >= 0 && prog_src->Index < VARYING_SLOT_MAX);

      nir_variable *var = c->input_vars[prog_src->Index];
      src.src = nir_src_for_ssa(nir_load_var(b, var));
      break;
   }
   case PROGRAM_SYSTEM_VALUE: {
      assert(!prog_src->RelAddr);

      assert(prog_src->Index >= 0 && prog_src->Index < SYSTEM_VALUE_MAX);

      nir_variable *var = c->sysval_vars[prog_src->Index];
      src.src = nir_src_for_ssa(nir_load_var(b, var));
      break;
   }
   case PROGRAM_STATE_VAR:
   case PROGRAM_CONSTANT: {
      /* We actually want to look at the type in the Parameters list for this,
       * because it lets us upload constant builtin uniforms as actual
       * constants.
       */
      struct gl_program_parameter_list *plist = c->prog->Parameters;
      gl_register_file file = prog_src->RelAddr ? prog_src->File :
         plist->Parameters[prog_src->Index].Type;

      switch (file) {
      case PROGRAM_CONSTANT:
         if ((c->prog->arb.IndirectRegisterFiles &
              (1 << PROGRAM_CONSTANT)) == 0) {
            unsigned pvo = plist->ParameterValueOffset[prog_src->Index];
            float *v = (float *) plist->ParameterValues + pvo;
            src.src = nir_src_for_ssa(nir_imm_vec4(b, v[0], v[1], v[2], v[3]));
            break;
         }
         /* FALLTHROUGH */
      case PROGRAM_STATE_VAR: {
         assert(c->parameters != NULL);

         nir_deref_instr *deref = nir_build_deref_var(b, c->parameters);

         nir_ssa_def *index = nir_imm_int(b, prog_src->Index);
         if (prog_src->RelAddr)
            index = nir_iadd(b, index, nir_load_reg(b, c->addr_reg));
         deref = nir_build_deref_array(b, deref, nir_channel(b, index, 0));

         src.src = nir_src_for_ssa(nir_load_deref(b, deref));
         break;
      }
      default:
         fprintf(stderr, "bad uniform src register file: %s (%d)\n",
                 _mesa_register_file_name(file), file);
         abort();
      }
      break;
   }
   default:
      fprintf(stderr, "unknown src register file: %s (%d)\n",
              _mesa_register_file_name(prog_src->File), prog_src->File);
      abort();
   }

   nir_ssa_def *def;
   if (!HAS_EXTENDED_SWIZZLE(prog_src->Swizzle) &&
       (prog_src->Negate == NEGATE_NONE || prog_src->Negate == NEGATE_XYZW)) {
      /* The simple non-SWZ case. */
      for (int i = 0; i < 4; i++)
         src.swizzle[i] = GET_SWZ(prog_src->Swizzle, i);

      def = nir_fmov_alu(b, src, 4);

      if (prog_src->Negate)
         def = nir_fneg(b, def);
   } else {
      /* The SWZ instruction allows per-component zero/one swizzles, and also
       * per-component negation.
       */
      nir_ssa_def *chans[4];
      for (int i = 0; i < 4; i++) {
         int swizzle = GET_SWZ(prog_src->Swizzle, i);
         if (swizzle == SWIZZLE_ZERO) {
            chans[i] = nir_imm_float(b, 0.0);
         } else if (swizzle == SWIZZLE_ONE) {
            chans[i] = nir_imm_float(b, 1.0);
         } else {
            assert(swizzle != SWIZZLE_NIL);
            nir_alu_instr *mov = nir_alu_instr_create(b->shader, nir_op_fmov);
            nir_ssa_dest_init(&mov->instr, &mov->dest.dest, 1, 32, NULL);
            mov->dest.write_mask = 0x1;
            mov->src[0] = src;
            mov->src[0].swizzle[0] = swizzle;
            nir_builder_instr_insert(b, &mov->instr);

            chans[i] = &mov->dest.dest.ssa;
         }

         if (prog_src->Negate & (1 << i))
            chans[i] = nir_fneg(b, chans[i]);
      }
      def = nir_vec4(b, chans[0], chans[1], chans[2], chans[3]);
   }

   return def;
}

static void
ptn_alu(nir_builder *b, nir_op op, nir_alu_dest dest, nir_ssa_def **src)
{
   unsigned num_srcs = nir_op_infos[op].num_inputs;
   nir_alu_instr *instr = nir_alu_instr_create(b->shader, op);
   unsigned i;

   for (i = 0; i < num_srcs; i++)
      instr->src[i].src = nir_src_for_ssa(src[i]);

   instr->dest = dest;
   nir_builder_instr_insert(b, &instr->instr);
}

static void
ptn_move_dest_masked(nir_builder *b, nir_alu_dest dest,
                     nir_ssa_def *def, unsigned write_mask)
{
   if (!(dest.write_mask & write_mask))
      return;

   nir_alu_instr *mov = nir_alu_instr_create(b->shader, nir_op_fmov);
   if (!mov)
      return;

   mov->dest = dest;
   mov->dest.write_mask &= write_mask;
   mov->src[0].src = nir_src_for_ssa(def);
   for (unsigned i = def->num_components; i < 4; i++)
      mov->src[0].swizzle[i] = def->num_components - 1;
   nir_builder_instr_insert(b, &mov->instr);
}

static void
ptn_move_dest(nir_builder *b, nir_alu_dest dest, nir_ssa_def *def)
{
   ptn_move_dest_masked(b, dest, def, WRITEMASK_XYZW);
}

static void
ptn_arl(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_f2i32(b, nir_ffloor(b, src[0])));
}

/* EXP - Approximate Exponential Base 2
 *  dst.x = 2^{\lfloor src.x\rfloor}
 *  dst.y = src.x - \lfloor src.x\rfloor
 *  dst.z = 2^{src.x}
 *  dst.w = 1.0
 */
static void
ptn_exp(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   nir_ssa_def *srcx = ptn_channel(b, src[0], X);

   ptn_move_dest_masked(b, dest, nir_fexp2(b, nir_ffloor(b, srcx)), WRITEMASK_X);
   ptn_move_dest_masked(b, dest, nir_fsub(b, srcx, nir_ffloor(b, srcx)), WRITEMASK_Y);
   ptn_move_dest_masked(b, dest, nir_fexp2(b, srcx), WRITEMASK_Z);
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_W);
}

/* LOG - Approximate Logarithm Base 2
 *  dst.x = \lfloor\log_2{|src.x|}\rfloor
 *  dst.y = |src.x| * 2^{-\lfloor\log_2{|src.x|}\rfloor}}
 *  dst.z = \log_2{|src.x|}
 *  dst.w = 1.0
 */
static void
ptn_log(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   nir_ssa_def *abs_srcx = nir_fabs(b, ptn_channel(b, src[0], X));
   nir_ssa_def *log2 = nir_flog2(b, abs_srcx);
   nir_ssa_def *floor_log2 = nir_ffloor(b, log2);

   ptn_move_dest_masked(b, dest, floor_log2, WRITEMASK_X);
   ptn_move_dest_masked(b, dest,
                        nir_fmul(b, abs_srcx,
                                 nir_fexp2(b, nir_fneg(b, floor_log2))),
                        WRITEMASK_Y);
   ptn_move_dest_masked(b, dest, log2, WRITEMASK_Z);
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_W);
}

/* DST - Distance Vector
 *   dst.x = 1.0
 *   dst.y = src0.y \times src1.y
 *   dst.z = src0.z
 *   dst.w = src1.w
 */
static void
ptn_dst(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_X);
   ptn_move_dest_masked(b, dest, nir_fmul(b, src[0], src[1]), WRITEMASK_Y);
   ptn_move_dest_masked(b, dest, nir_fmov(b, src[0]), WRITEMASK_Z);
   ptn_move_dest_masked(b, dest, nir_fmov(b, src[1]), WRITEMASK_W);
}

/* LIT - Light Coefficients
 *  dst.x = 1.0
 *  dst.y = max(src.x, 0.0)
 *  dst.z = (src.x > 0.0) ? max(src.y, 0.0)^{clamp(src.w, -128.0, 128.0))} : 0
 *  dst.w = 1.0
 */
static void
ptn_lit(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_XW);

   ptn_move_dest_masked(b, dest, nir_fmax(b, ptn_channel(b, src[0], X),
                                          nir_imm_float(b, 0.0)), WRITEMASK_Y);

   if (dest.write_mask & WRITEMASK_Z) {
      nir_ssa_def *src0_y = ptn_channel(b, src[0], Y);
      nir_ssa_def *wclamp = nir_fmax(b, nir_fmin(b, ptn_channel(b, src[0], W),
                                                 nir_imm_float(b, 128.0)),
                                     nir_imm_float(b, -128.0));
      nir_ssa_def *pow = nir_fpow(b, nir_fmax(b, src0_y, nir_imm_float(b, 0.0)),
                                  wclamp);

      nir_ssa_def *z;
      if (b->shader->options->native_integers) {
         z = nir_bcsel(b,
                       nir_fge(b, nir_imm_float(b, 0.0), ptn_channel(b, src[0], X)),
                       nir_imm_float(b, 0.0),
                       pow);
      } else {
         z = nir_fcsel(b,
                       nir_sge(b, nir_imm_float(b, 0.0), ptn_channel(b, src[0], X)),
                       nir_imm_float(b, 0.0),
                       pow);
      }

      ptn_move_dest_masked(b, dest, z, WRITEMASK_Z);
   }
}

/* SCS - Sine Cosine
 *   dst.x = \cos{src.x}
 *   dst.y = \sin{src.x}
 *   dst.z = 0.0
 *   dst.w = 1.0
 */
static void
ptn_scs(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest_masked(b, dest, nir_fcos(b, ptn_channel(b, src[0], X)),
                        WRITEMASK_X);
   ptn_move_dest_masked(b, dest, nir_fsin(b, ptn_channel(b, src[0], X)),
                        WRITEMASK_Y);
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 0.0), WRITEMASK_Z);
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_W);
}

/**
 * Emit SLT.  For platforms with integers, prefer b2f(flt(...)).
 */
static void
ptn_slt(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   if (b->shader->options->native_integers) {
      ptn_move_dest(b, dest, nir_b2f32(b, nir_flt(b, src[0], src[1])));
   } else {
      ptn_move_dest(b, dest, nir_slt(b, src[0], src[1]));
   }
}

/**
 * Emit SGE.  For platforms with integers, prefer b2f(fge(...)).
 */
static void
ptn_sge(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   if (b->shader->options->native_integers) {
      ptn_move_dest(b, dest, nir_b2f32(b, nir_fge(b, src[0], src[1])));
   } else {
      ptn_move_dest(b, dest, nir_sge(b, src[0], src[1]));
   }
}

static void
ptn_xpd(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest_masked(b, dest,
                        nir_fsub(b,
                                 nir_fmul(b,
                                          nir_swizzle(b, src[0], SWIZ(Y, Z, X, W), 3, true),
                                          nir_swizzle(b, src[1], SWIZ(Z, X, Y, W), 3, true)),
                                 nir_fmul(b,
                                          nir_swizzle(b, src[1], SWIZ(Y, Z, X, W), 3, true),
                                          nir_swizzle(b, src[0], SWIZ(Z, X, Y, W), 3, true))),
                        WRITEMASK_XYZ);
   ptn_move_dest_masked(b, dest, nir_imm_float(b, 1.0), WRITEMASK_W);
}

static void
ptn_dp2(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_fdot2(b, src[0], src[1]));
}

static void
ptn_dp3(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_fdot3(b, src[0], src[1]));
}

static void
ptn_dp4(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_fdot4(b, src[0], src[1]));
}

static void
ptn_dph(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_fdph(b, src[0], src[1]));
}

static void
ptn_cmp(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   if (b->shader->options->native_integers) {
      ptn_move_dest(b, dest, nir_bcsel(b,
                                       nir_flt(b, src[0], nir_imm_float(b, 0.0)),
                                       src[1], src[2]));
   } else {
      ptn_move_dest(b, dest, nir_fcsel(b,
                                       nir_slt(b, src[0], nir_imm_float(b, 0.0)),
                                       src[1], src[2]));
   }
}

static void
ptn_lrp(nir_builder *b, nir_alu_dest dest, nir_ssa_def **src)
{
   ptn_move_dest(b, dest, nir_flrp(b, src[2], src[1], src[0]));
}

static void
ptn_kil(nir_builder *b, nir_ssa_def **src)
{
   nir_ssa_def *cmp = b->shader->options->native_integers ?
      nir_bany(b, nir_flt(b, src[0], nir_imm_float(b, 0.0))) :
      nir_fany_nequal4(b, nir_slt(b, src[0], nir_imm_float(b, 0.0)), nir_imm_float(b, 0.0));

   nir_intrinsic_instr *discard =
      nir_intrinsic_instr_create(b->shader, nir_intrinsic_discard_if);
   discard->src[0] = nir_src_for_ssa(cmp);
   nir_builder_instr_insert(b, &discard->instr);
}

static void
ptn_tex(struct ptn_compile *c, nir_alu_dest dest, nir_ssa_def **src,
        struct prog_instruction *prog_inst)
{
   nir_builder *b = &c->build;
   nir_tex_instr *instr;
   nir_texop op;
   unsigned num_srcs;

   switch (prog_inst->Opcode) {
   case OPCODE_TEX:
      op = nir_texop_tex;
      num_srcs = 1;
      break;
   case OPCODE_TXB:
      op = nir_texop_txb;
      num_srcs = 2;
      break;
   case OPCODE_TXD:
      op = nir_texop_txd;
      num_srcs = 3;
      break;
   case OPCODE_TXL:
      op = nir_texop_txl;
      num_srcs = 2;
      break;
   case OPCODE_TXP:
      op = nir_texop_tex;
      num_srcs = 2;
      break;
   default:
      fprintf(stderr, "unknown tex op %d\n", prog_inst->Opcode);
      abort();
   }

   /* Deref sources */
   num_srcs += 2;

   if (prog_inst->TexShadow)
      num_srcs++;

   instr = nir_tex_instr_create(b->shader, num_srcs);
   instr->op = op;
   instr->dest_type = nir_type_float;
   instr->is_shadow = prog_inst->TexShadow;

   switch (prog_inst->TexSrcTarget) {
   case TEXTURE_1D_INDEX:
      instr->sampler_dim = GLSL_SAMPLER_DIM_1D;
      break;
   case TEXTURE_2D_INDEX:
      instr->sampler_dim = GLSL_SAMPLER_DIM_2D;
      break;
   case TEXTURE_3D_INDEX:
      instr->sampler_dim = GLSL_SAMPLER_DIM_3D;
      break;
   case TEXTURE_CUBE_INDEX:
      instr->sampler_dim = GLSL_SAMPLER_DIM_CUBE;
      break;
   case TEXTURE_RECT_INDEX:
      instr->sampler_dim = GLSL_SAMPLER_DIM_RECT;
      break;
   default:
      fprintf(stderr, "Unknown texture target %d\n", prog_inst->TexSrcTarget);
      abort();
   }

   switch (instr->sampler_dim) {
   case GLSL_SAMPLER_DIM_1D:
   case GLSL_SAMPLER_DIM_BUF:
      instr->coord_components = 1;
      break;
   case GLSL_SAMPLER_DIM_2D:
   case GLSL_SAMPLER_DIM_RECT:
   case GLSL_SAMPLER_DIM_EXTERNAL:
   case GLSL_SAMPLER_DIM_MS:
      instr->coord_components = 2;
      break;
   case GLSL_SAMPLER_DIM_3D:
   case GLSL_SAMPLER_DIM_CUBE:
      instr->coord_components = 3;
      break;
   case GLSL_SAMPLER_DIM_SUBPASS:
   case GLSL_SAMPLER_DIM_SUBPASS_MS:
      unreachable("can't reach");
   }

   nir_variable *var = c->sampler_vars[prog_inst->TexSrcUnit];
   if (!var) {
      const struct glsl_type *type =
         glsl_sampler_type(instr->sampler_dim, false, false, GLSL_TYPE_FLOAT);
      var = nir_variable_create(b->shader, nir_var_uniform, type, "sampler");
      var->data.binding = prog_inst->TexSrcUnit;
      var->data.explicit_binding = true;
      c->sampler_vars[prog_inst->TexSrcUnit] = var;
   }

   nir_deref_instr *deref = nir_build_deref_var(b, var);

   unsigned src_number = 0;

   instr->src[src_number].src = nir_src_for_ssa(&deref->dest.ssa);
   instr->src[src_number].src_type = nir_tex_src_texture_deref;
   src_number++;
   instr->src[src_number].src = nir_src_for_ssa(&deref->dest.ssa);
   instr->src[src_number].src_type = nir_tex_src_sampler_deref;
   src_number++;

   instr->src[src_number].src =
      nir_src_for_ssa(nir_swizzle(b, src[0], SWIZ(X, Y, Z, W),
                                  instr->coord_components, true));
   instr->src[src_number].src_type = nir_tex_src_coord;
   src_number++;

   if (prog_inst->Opcode == OPCODE_TXP) {
      instr->src[src_number].src = nir_src_for_ssa(ptn_channel(b, src[0], W));
      instr->src[src_number].src_type = nir_tex_src_projector;
      src_number++;
   }

   if (prog_inst->Opcode == OPCODE_TXB) {
      instr->src[src_number].src = nir_src_for_ssa(ptn_channel(b, src[0], W));
      instr->src[src_number].src_type = nir_tex_src_bias;
      src_number++;
   }

   if (prog_inst->Opcode == OPCODE_TXL) {
      instr->src[src_number].src = nir_src_for_ssa(ptn_channel(b, src[0], W));
      instr->src[src_number].src_type = nir_tex_src_lod;
      src_number++;
   }

   if (instr->is_shadow) {
      if (instr->coord_components < 3)
         instr->src[src_number].src = nir_src_for_ssa(ptn_channel(b, src[0], Z));
      else
         instr->src[src_number].src = nir_src_for_ssa(ptn_channel(b, src[0], W));

      instr->src[src_number].src_type = nir_tex_src_comparator;
      src_number++;
   }

   assert(src_number == num_srcs);

   nir_ssa_dest_init(&instr->instr, &instr->dest, 4, 32, NULL);
   nir_builder_instr_insert(b, &instr->instr);

   /* Resolve the writemask on the texture op. */
   ptn_move_dest(b, dest, &instr->dest.ssa);
}

static const nir_op op_trans[MAX_OPCODE] = {
   [OPCODE_NOP] = 0,
   [OPCODE_ABS] = nir_op_fabs,
   [OPCODE_ADD] = nir_op_fadd,
   [OPCODE_ARL] = 0,
   [OPCODE_CMP] = 0,
   [OPCODE_COS] = 0,
   [OPCODE_DDX] = nir_op_fddx,
   [OPCODE_DDY] = nir_op_fddy,
   [OPCODE_DP2] = 0,
   [OPCODE_DP3] = 0,
   [OPCODE_DP4] = 0,
   [OPCODE_DPH] = 0,
   [OPCODE_DST] = 0,
   [OPCODE_END] = 0,
   [OPCODE_EX2] = 0,
   [OPCODE_EXP] = 0,
   [OPCODE_FLR] = nir_op_ffloor,
   [OPCODE_FRC] = nir_op_ffract,
   [OPCODE_LG2] = 0,
   [OPCODE_LIT] = 0,
   [OPCODE_LOG] = 0,
   [OPCODE_LRP] = 0,
   [OPCODE_MAD] = 0,
   [OPCODE_MAX] = nir_op_fmax,
   [OPCODE_MIN] = nir_op_fmin,
   [OPCODE_MOV] = nir_op_fmov,
   [OPCODE_MUL] = nir_op_fmul,
   [OPCODE_POW] = 0,
   [OPCODE_RCP] = 0,

   [OPCODE_RSQ] = 0,
   [OPCODE_SCS] = 0,
   [OPCODE_SGE] = 0,
   [OPCODE_SIN] = 0,
   [OPCODE_SLT] = 0,
   [OPCODE_SSG] = nir_op_fsign,
   [OPCODE_SUB] = nir_op_fsub,
   [OPCODE_SWZ] = 0,
   [OPCODE_TEX] = 0,
   [OPCODE_TRUNC] = nir_op_ftrunc,
   [OPCODE_TXB] = 0,
   [OPCODE_TXD] = 0,
   [OPCODE_TXL] = 0,
   [OPCODE_TXP] = 0,
   [OPCODE_XPD] = 0,
};

static void
ptn_emit_instruction(struct ptn_compile *c, struct prog_instruction *prog_inst)
{
   nir_builder *b = &c->build;
   unsigned i;
   const unsigned op = prog_inst->Opcode;

   if (op == OPCODE_END)
      return;

   nir_ssa_def *src[3];
   for (i = 0; i < 3; i++) {
      src[i] = ptn_get_src(c, &prog_inst->SrcReg[i]);
   }
   nir_alu_dest dest = ptn_get_dest(c, &prog_inst->DstReg);
   if (c->error)
      return;

   switch (op) {
   case OPCODE_RSQ:
      ptn_move_dest(b, dest,
                    nir_frsq(b, nir_fabs(b, ptn_channel(b, src[0], X))));
      break;

   case OPCODE_RCP:
      ptn_move_dest(b, dest, nir_frcp(b, ptn_channel(b, src[0], X)));
      break;

   case OPCODE_EX2:
      ptn_move_dest(b, dest, nir_fexp2(b, ptn_channel(b, src[0], X)));
      break;

   case OPCODE_LG2:
      ptn_move_dest(b, dest, nir_flog2(b, ptn_channel(b, src[0], X)));
      break;

   case OPCODE_POW:
      ptn_move_dest(b, dest, nir_fpow(b,
                                      ptn_channel(b, src[0], X),
                                      ptn_channel(b, src[1], X)));
      break;

   case OPCODE_COS:
      ptn_move_dest(b, dest, nir_fcos(b, ptn_channel(b, src[0], X)));
      break;

   case OPCODE_SIN:
      ptn_move_dest(b, dest, nir_fsin(b, ptn_channel(b, src[0], X)));
      break;

   case OPCODE_ARL:
      ptn_arl(b, dest, src);
      break;

   case OPCODE_EXP:
      ptn_exp(b, dest, src);
      break;

   case OPCODE_LOG:
      ptn_log(b, dest, src);
      break;

   case OPCODE_LRP:
      ptn_lrp(b, dest, src);
      break;

   case OPCODE_MAD:
      ptn_move_dest(b, dest, nir_fadd(b, nir_fmul(b, src[0], src[1]), src[2]));
      break;

   case OPCODE_DST:
      ptn_dst(b, dest, src);
      break;

   case OPCODE_LIT:
      ptn_lit(b, dest, src);
      break;

   case OPCODE_XPD:
      ptn_xpd(b, dest, src);
      break;

   case OPCODE_DP2:
      ptn_dp2(b, dest, src);
      break;

   case OPCODE_DP3:
      ptn_dp3(b, dest, src);
      break;

   case OPCODE_DP4:
      ptn_dp4(b, dest, src);
      break;

   case OPCODE_DPH:
      ptn_dph(b, dest, src);
      break;

   case OPCODE_KIL:
      ptn_kil(b, src);
      break;

   case OPCODE_CMP:
      ptn_cmp(b, dest, src);
      break;

   case OPCODE_SCS:
      ptn_scs(b, dest, src);
      break;

   case OPCODE_SLT:
      ptn_slt(b, dest, src);
      break;

   case OPCODE_SGE:
      ptn_sge(b, dest, src);
      break;

   case OPCODE_TEX:
   case OPCODE_TXB:
   case OPCODE_TXD:
   case OPCODE_TXL:
   case OPCODE_TXP:
      ptn_tex(c, dest, src, prog_inst);
      break;

   case OPCODE_SWZ:
      /* Extended swizzles were already handled in ptn_get_src(). */
      ptn_alu(b, nir_op_fmov, dest, src);
      break;

   case OPCODE_NOP:
      break;

   default:
      if (op_trans[op] != 0) {
         ptn_alu(b, op_trans[op], dest, src);
      } else {
         fprintf(stderr, "unknown opcode: %s\n", _mesa_opcode_string(op));
         abort();
      }
      break;
   }

   if (prog_inst->Saturate) {
      assert(prog_inst->Saturate);
      assert(!dest.dest.is_ssa);
      ptn_move_dest(b, dest, nir_fsat(b, ptn_src_for_dest(c, &dest)));
   }
}

/**
 * Puts a NIR intrinsic to store of each PROGRAM_OUTPUT value to the output
 * variables at the end of the shader.
 *
 * We don't generate these incrementally as the PROGRAM_OUTPUT values are
 * written, because there's no output load intrinsic, which means we couldn't
 * handle writemasks.
 */
static void
ptn_add_output_stores(struct ptn_compile *c)
{
   nir_builder *b = &c->build;

   nir_foreach_variable(var, &b->shader->outputs) {
      nir_ssa_def *src = nir_load_reg(b, c->output_regs[var->data.location]);
      if (c->prog->Target == GL_FRAGMENT_PROGRAM_ARB &&
          var->data.location == FRAG_RESULT_DEPTH) {
         /* result.depth has this strange convention of being the .z component of
          * a vec4 with undefined .xyw components.  We resolve it to a scalar, to
          * match GLSL's gl_FragDepth and the expectations of most backends.
          */
         src = nir_channel(b, src, 2);
      }
      if (c->prog->Target == GL_VERTEX_PROGRAM_ARB &&
          var->data.location == VARYING_SLOT_FOGC) {
         /* result.fogcoord is a single component value */
         src = nir_channel(b, src, 0);
      }
      unsigned num_components = glsl_get_vector_elements(var->type);
      nir_store_var(b, var, src, (1 << num_components) - 1);
   }
}

static void
setup_registers_and_variables(struct ptn_compile *c)
{
   nir_builder *b = &c->build;
   struct nir_shader *shader = b->shader;

   /* Create input variables. */
   uint64_t inputs_read = c->prog->info.inputs_read;
   while (inputs_read) {
      const int i = u_bit_scan64(&inputs_read);

      nir_variable *var =
         nir_variable_create(shader, nir_var_shader_in, glsl_vec4_type(),
                             ralloc_asprintf(shader, "in_%d", i));
      var->data.location = i;
      var->data.index = 0;

      if (c->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
         if (i == VARYING_SLOT_FOGC) {
            /* fogcoord is defined as <f, 0.0, 0.0, 1.0>.  Make the actual
             * input variable a float, and create a local containing the
             * full vec4 value.
             */
            var->type = glsl_float_type();

            nir_variable *fullvar =
               nir_local_variable_create(b->impl, glsl_vec4_type(),
                                         "fogcoord_tmp");

            nir_store_var(b, fullvar,
                          nir_vec4(b, nir_load_var(b, var),
                                   nir_imm_float(b, 0.0),
                                   nir_imm_float(b, 0.0),
                                   nir_imm_float(b, 1.0)),
                          WRITEMASK_XYZW);

            /* We inserted the real input into the list so the driver has real
             * inputs, but we set c->input_vars[i] to the temporary so we use
             * the splatted value.
             */
            c->input_vars[i] = fullvar;
            continue;
         }
      }

      c->input_vars[i] = var;
   }

   /* Create system value variables */
   uint64_t system_values_read = c->prog->info.system_values_read;
   while (system_values_read) {
      const int i = u_bit_scan64(&system_values_read);

      nir_variable *var =
         nir_variable_create(shader, nir_var_system_value, glsl_vec4_type(),
                             ralloc_asprintf(shader, "sv_%d", i));
      var->data.location = i;
      var->data.index = 0;

      c->sysval_vars[i] = var;
   }

   /* Create output registers and variables. */
   int max_outputs = util_last_bit(c->prog->info.outputs_written);
   c->output_regs = rzalloc_array(c, nir_register *, max_outputs);

   uint64_t outputs_written = c->prog->info.outputs_written;
   while (outputs_written) {
      const int i = u_bit_scan64(&outputs_written);

      /* Since we can't load from outputs in the IR, we make temporaries
       * for the outputs and emit stores to the real outputs at the end of
       * the shader.
       */
      nir_register *reg = nir_local_reg_create(b->impl);
      reg->num_components = 4;

      nir_variable *var = rzalloc(shader, nir_variable);
      if ((c->prog->Target == GL_FRAGMENT_PROGRAM_ARB && i == FRAG_RESULT_DEPTH) ||
          (c->prog->Target == GL_VERTEX_PROGRAM_ARB && i == VARYING_SLOT_FOGC))
         var->type = glsl_float_type();
      else
         var->type = glsl_vec4_type();
      var->data.mode = nir_var_shader_out;
      var->name = ralloc_asprintf(var, "out_%d", i);

      var->data.location = i;
      var->data.index = 0;

      c->output_regs[i] = reg;

      exec_list_push_tail(&shader->outputs, &var->node);
      c->output_vars[i] = var;
   }

   /* Create temporary registers. */
   c->temp_regs = rzalloc_array(c, nir_register *,
                                c->prog->arb.NumTemporaries);

   nir_register *reg;
   for (unsigned i = 0; i < c->prog->arb.NumTemporaries; i++) {
      reg = nir_local_reg_create(b->impl);
      if (!reg) {
         c->error = true;
         return;
      }
      reg->num_components = 4;
      c->temp_regs[i] = reg;
   }

   /* Create the address register (for ARB_vertex_program). */
   reg = nir_local_reg_create(b->impl);
   if (!reg) {
      c->error = true;
      return;
   }
   reg->num_components = 1;
   c->addr_reg = reg;
}

struct nir_shader *
prog_to_nir(const struct gl_program *prog,
            const nir_shader_compiler_options *options)
{
   struct ptn_compile *c;
   struct nir_shader *s;
   gl_shader_stage stage = _mesa_program_enum_to_shader_stage(prog->Target);

   c = rzalloc(NULL, struct ptn_compile);
   if (!c)
      return NULL;
   c->prog = prog;

   nir_builder_init_simple_shader(&c->build, NULL, stage, options);

   /* Copy the shader_info from the gl_program */
   c->build.shader->info = prog->info;

   s = c->build.shader;

   if (prog->Parameters->NumParameters > 0) {
      const struct glsl_type *type =
         glsl_array_type(glsl_vec4_type(), prog->Parameters->NumParameters, 0);
      c->parameters =
         nir_variable_create(s, nir_var_uniform, type,
                             prog->Parameters->Parameters[0].Name);
   }

   setup_registers_and_variables(c);
   if (unlikely(c->error))
      goto fail;

   for (unsigned int i = 0; i < prog->arb.NumInstructions; i++) {
      ptn_emit_instruction(c, &prog->arb.Instructions[i]);

      if (unlikely(c->error))
         break;
   }

   ptn_add_output_stores(c);

   s->info.name = ralloc_asprintf(s, "ARB%d", prog->Id);
   s->info.num_textures = util_last_bit(prog->SamplersUsed);
   s->info.num_ubos = 0;
   s->info.num_abos = 0;
   s->info.num_ssbos = 0;
   s->info.num_images = 0;
   s->info.uses_texture_gather = false;
   s->info.clip_distance_array_size = 0;
   s->info.cull_distance_array_size = 0;
   s->info.separate_shader = false;

fail:
   if (c->error) {
      ralloc_free(s);
      s = NULL;
   }
   ralloc_free(c);
   return s;
}