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
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
|
/*
* Copyright © 2010 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.
*/
/**
* \file ir_to_mesa.cpp
*
* Translates the IR to ARB_fragment_program text if possible,
* printing the result
*/
#include <stdio.h>
#include "ir.h"
#include "ir_visitor.h"
#include "ir_print_visitor.h"
#include "ir_expression_flattening.h"
#include "glsl_types.h"
extern "C" {
#include "shader/prog_instruction.h"
#include "shader/prog_print.h"
}
/**
* This struct is a corresponding struct to Mesa prog_src_register, with
* wider fields.
*/
typedef struct ir_to_mesa_src_reg {
int file; /**< PROGRAM_* from Mesa */
int index; /**< temporary index, VERT_ATTRIB_*, FRAG_ATTRIB_*, etc. */
int swizzle; /**< SWIZZLE_XYZWONEZERO swizzles from Mesa. */
int negate; /**< NEGATE_XYZW mask from mesa */
bool reladdr; /**< Register index should be offset by address reg. */
} ir_to_mesa_src_reg;
typedef struct ir_to_mesa_dst_reg {
int file; /**< PROGRAM_* from Mesa */
int index; /**< temporary index, VERT_ATTRIB_*, FRAG_ATTRIB_*, etc. */
int writemask; /**< Bitfield of WRITEMASK_[XYZW] */
} ir_to_mesa_dst_reg;
extern ir_to_mesa_src_reg ir_to_mesa_undef;
class ir_to_mesa_instruction : public exec_node {
public:
enum prog_opcode op;
ir_to_mesa_dst_reg dst_reg;
ir_to_mesa_src_reg src_reg[3];
/** Pointer to the ir source this tree came from for debugging */
ir_instruction *ir;
};
class temp_entry : public exec_node {
public:
temp_entry(ir_variable *var, int file, int index)
: file(file), index(index), var(var)
{
/* empty */
}
int file;
int index;
ir_variable *var; /* variable that maps to this, if any */
};
class ir_to_mesa_visitor : public ir_visitor {
public:
ir_to_mesa_visitor();
int next_temp;
int next_constant;
int next_uniform;
temp_entry *find_variable_storage(ir_variable *var);
ir_to_mesa_src_reg get_temp(const glsl_type *type);
struct ir_to_mesa_src_reg src_reg_for_float(float val);
/**
* \name Visit methods
*
* As typical for the visitor pattern, there must be one \c visit method for
* each concrete subclass of \c ir_instruction. Virtual base classes within
* the hierarchy should not have \c visit methods.
*/
/*@{*/
virtual void visit(ir_variable *);
virtual void visit(ir_loop *);
virtual void visit(ir_loop_jump *);
virtual void visit(ir_function_signature *);
virtual void visit(ir_function *);
virtual void visit(ir_expression *);
virtual void visit(ir_swizzle *);
virtual void visit(ir_dereference_variable *);
virtual void visit(ir_dereference_array *);
virtual void visit(ir_dereference_record *);
virtual void visit(ir_assignment *);
virtual void visit(ir_constant *);
virtual void visit(ir_call *);
virtual void visit(ir_return *);
virtual void visit(ir_texture *);
virtual void visit(ir_if *);
/*@}*/
struct ir_to_mesa_src_reg result;
/** List of temp_entry */
exec_list variable_storage;
/** List of ir_to_mesa_instruction */
exec_list instructions;
ir_to_mesa_instruction *ir_to_mesa_emit_op1(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0);
ir_to_mesa_instruction *ir_to_mesa_emit_op2(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0,
ir_to_mesa_src_reg src1);
ir_to_mesa_instruction *ir_to_mesa_emit_op3(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0,
ir_to_mesa_src_reg src1,
ir_to_mesa_src_reg src2);
void ir_to_mesa_emit_scalar_op1(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0);
/* talloc context (the ) */
void *ctx;
};
ir_to_mesa_src_reg ir_to_mesa_undef = {
PROGRAM_UNDEFINED, 0, SWIZZLE_NOOP, NEGATE_NONE, false,
};
ir_to_mesa_dst_reg ir_to_mesa_undef_dst = {
PROGRAM_UNDEFINED, 0, SWIZZLE_NOOP
};
ir_to_mesa_dst_reg ir_to_mesa_address_reg = {
PROGRAM_ADDRESS, 0, WRITEMASK_X
};
static int swizzle_for_size(int size)
{
int size_swizzles[4] = {
MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X),
MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y),
MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_Z),
MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W),
};
return size_swizzles[size - 1];
}
/* This list should match up with builtin_variables.h */
static const struct {
const char *name;
int file;
int index;
} builtin_var_to_mesa_reg[] = {
/* core_vs */
{"gl_Position", PROGRAM_OUTPUT, VERT_RESULT_HPOS},
{"gl_PointSize", PROGRAM_OUTPUT, VERT_RESULT_PSIZ},
/* core_fs */
{"gl_FragCoord", PROGRAM_INPUT, FRAG_ATTRIB_WPOS},
{"gl_FrontFacing", PROGRAM_INPUT, FRAG_ATTRIB_FACE},
{"gl_FragColor", PROGRAM_OUTPUT, FRAG_ATTRIB_COL0},
{"gl_FragDepth", PROGRAM_UNDEFINED, FRAG_ATTRIB_WPOS}, /* FINISHME: WPOS.z */
/* 110_deprecated_fs */
{"gl_Color", PROGRAM_INPUT, FRAG_ATTRIB_COL0},
{"gl_SecondaryColor", PROGRAM_INPUT, FRAG_ATTRIB_COL1},
{"gl_FogFragCoord", PROGRAM_INPUT, FRAG_ATTRIB_FOGC},
{"gl_TexCoord", PROGRAM_INPUT, FRAG_ATTRIB_TEX0}, /* array */
/* 110_deprecated_vs */
{"gl_Vertex", PROGRAM_INPUT, VERT_ATTRIB_POS},
{"gl_Normal", PROGRAM_INPUT, VERT_ATTRIB_NORMAL},
{"gl_Color", PROGRAM_INPUT, VERT_ATTRIB_COLOR0},
{"gl_SecondaryColor", PROGRAM_INPUT, VERT_ATTRIB_COLOR1},
{"gl_MultiTexCoord0", PROGRAM_INPUT, VERT_ATTRIB_TEX0},
{"gl_MultiTexCoord1", PROGRAM_INPUT, VERT_ATTRIB_TEX1},
{"gl_MultiTexCoord2", PROGRAM_INPUT, VERT_ATTRIB_TEX2},
{"gl_MultiTexCoord3", PROGRAM_INPUT, VERT_ATTRIB_TEX3},
{"gl_MultiTexCoord4", PROGRAM_INPUT, VERT_ATTRIB_TEX4},
{"gl_MultiTexCoord5", PROGRAM_INPUT, VERT_ATTRIB_TEX5},
{"gl_MultiTexCoord6", PROGRAM_INPUT, VERT_ATTRIB_TEX6},
{"gl_MultiTexCoord7", PROGRAM_INPUT, VERT_ATTRIB_TEX7},
{"gl_TexCoord", PROGRAM_OUTPUT, VERT_RESULT_TEX0}, /* array */
{"gl_FogCoord", PROGRAM_INPUT, VERT_RESULT_FOGC},
/*{"gl_ClipVertex", PROGRAM_OUTPUT, VERT_ATTRIB_FOGC},*/ /* FINISHME */
{"gl_FrontColor", PROGRAM_OUTPUT, VERT_RESULT_COL0},
{"gl_BackColor", PROGRAM_OUTPUT, VERT_RESULT_BFC0},
{"gl_FrontSecondaryColor", PROGRAM_OUTPUT, VERT_RESULT_COL1},
{"gl_BackSecondaryColor", PROGRAM_OUTPUT, VERT_RESULT_BFC1},
{"gl_FogFragCoord", PROGRAM_OUTPUT, VERT_RESULT_FOGC},
/* 130_vs */
/*{"gl_VertexID", PROGRAM_INPUT, VERT_ATTRIB_FOGC},*/ /* FINISHME */
{"gl_FragData", PROGRAM_OUTPUT, FRAG_RESULT_DATA0}, /* array */
};
ir_to_mesa_instruction *
ir_to_mesa_visitor::ir_to_mesa_emit_op3(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0,
ir_to_mesa_src_reg src1,
ir_to_mesa_src_reg src2)
{
ir_to_mesa_instruction *inst = new(ctx) ir_to_mesa_instruction();
inst->op = op;
inst->dst_reg = dst;
inst->src_reg[0] = src0;
inst->src_reg[1] = src1;
inst->src_reg[2] = src2;
inst->ir = ir;
this->instructions.push_tail(inst);
return inst;
}
ir_to_mesa_instruction *
ir_to_mesa_visitor::ir_to_mesa_emit_op2(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0,
ir_to_mesa_src_reg src1)
{
return ir_to_mesa_emit_op3(ir, op, dst, src0, src1, ir_to_mesa_undef);
}
ir_to_mesa_instruction *
ir_to_mesa_visitor::ir_to_mesa_emit_op1(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0)
{
return ir_to_mesa_emit_op3(ir, op, dst,
src0, ir_to_mesa_undef, ir_to_mesa_undef);
}
inline ir_to_mesa_dst_reg
ir_to_mesa_dst_reg_from_src(ir_to_mesa_src_reg reg)
{
ir_to_mesa_dst_reg dst_reg;
dst_reg.file = reg.file;
dst_reg.index = reg.index;
dst_reg.writemask = WRITEMASK_XYZW;
return dst_reg;
}
/**
* Emits Mesa scalar opcodes to produce unique answers across channels.
*
* Some Mesa opcodes are scalar-only, like ARB_fp/vp. The src X
* channel determines the result across all channels. So to do a vec4
* of this operation, we want to emit a scalar per source channel used
* to produce dest channels.
*/
void
ir_to_mesa_visitor::ir_to_mesa_emit_scalar_op1(ir_instruction *ir,
enum prog_opcode op,
ir_to_mesa_dst_reg dst,
ir_to_mesa_src_reg src0)
{
int i, j;
int done_mask = ~dst.writemask;
/* Mesa RCP is a scalar operation splatting results to all channels,
* like ARB_fp/vp. So emit as many RCPs as necessary to cover our
* dst channels.
*/
for (i = 0; i < 4; i++) {
int this_mask = (1 << i);
ir_to_mesa_instruction *inst;
ir_to_mesa_src_reg src = src0;
if (done_mask & this_mask)
continue;
int src_swiz = GET_SWZ(src.swizzle, i);
for (j = i + 1; j < 4; j++) {
if (!(done_mask & (1 << j)) && GET_SWZ(src.swizzle, j) == src_swiz) {
this_mask |= (1 << j);
}
}
src.swizzle = MAKE_SWIZZLE4(src_swiz, src_swiz,
src_swiz, src_swiz);
inst = ir_to_mesa_emit_op1(ir, op,
dst,
src);
inst->dst_reg.writemask = this_mask;
done_mask |= this_mask;
}
}
struct ir_to_mesa_src_reg
ir_to_mesa_visitor::src_reg_for_float(float val)
{
ir_to_mesa_src_reg src_reg;
/* FINISHME: This will end up being _mesa_add_unnamed_constant,
* which handles sharing values and sharing channels of vec4
* constants for small values.
*/
/* FINISHME: Do something with the constant values for now.
*/
(void)val;
src_reg.file = PROGRAM_CONSTANT;
src_reg.index = this->next_constant++;
src_reg.swizzle = SWIZZLE_NOOP;
return src_reg;
}
/**
* In the initial pass of codegen, we assign temporary numbers to
* intermediate results. (not SSA -- variable assignments will reuse
* storage). Actual register allocation for the Mesa VM occurs in a
* pass over the Mesa IR later.
*/
ir_to_mesa_src_reg
ir_to_mesa_visitor::get_temp(const glsl_type *type)
{
ir_to_mesa_src_reg src_reg;
int swizzle[4];
int i;
assert(!type->is_array());
src_reg.file = PROGRAM_TEMPORARY;
src_reg.index = type->matrix_columns;
src_reg.reladdr = false;
for (i = 0; i < type->vector_elements; i++)
swizzle[i] = i;
for (; i < 4; i++)
swizzle[i] = type->vector_elements - 1;
src_reg.swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1],
swizzle[2], swizzle[3]);
return src_reg;
}
static int
type_size(const struct glsl_type *type)
{
unsigned int i;
int size;
switch (type->base_type) {
case GLSL_TYPE_UINT:
case GLSL_TYPE_INT:
case GLSL_TYPE_FLOAT:
case GLSL_TYPE_BOOL:
if (type->is_matrix()) {
return 4; /* FINISHME: Not all matrices are 4x4. */
} else {
/* Regardless of size of vector, it gets a vec4. This is bad
* packing for things like floats, but otherwise arrays become a
* mess. Hopefully a later pass over the code can pack scalars
* down if appropriate.
*/
return 1;
}
case GLSL_TYPE_ARRAY:
return type_size(type->fields.array) * type->length;
case GLSL_TYPE_STRUCT:
size = 0;
for (i = 0; i < type->length; i++) {
size += type_size(type->fields.structure[i].type);
}
return size;
default:
assert(0);
}
}
temp_entry *
ir_to_mesa_visitor::find_variable_storage(ir_variable *var)
{
temp_entry *entry;
foreach_iter(exec_list_iterator, iter, this->variable_storage) {
entry = (temp_entry *)iter.get();
if (entry->var == var)
return entry;
}
return NULL;
}
void
ir_to_mesa_visitor::visit(ir_variable *ir)
{
(void)ir;
}
void
ir_to_mesa_visitor::visit(ir_loop *ir)
{
assert(!ir->from);
assert(!ir->to);
assert(!ir->increment);
assert(!ir->counter);
ir_to_mesa_emit_op1(NULL, OPCODE_BGNLOOP,
ir_to_mesa_undef_dst, ir_to_mesa_undef);
visit_exec_list(&ir->body_instructions, this);
ir_to_mesa_emit_op1(NULL, OPCODE_ENDLOOP,
ir_to_mesa_undef_dst, ir_to_mesa_undef);
}
void
ir_to_mesa_visitor::visit(ir_loop_jump *ir)
{
switch (ir->mode) {
case ir_loop_jump::jump_break:
ir_to_mesa_emit_op1(NULL, OPCODE_BRK,
ir_to_mesa_undef_dst, ir_to_mesa_undef);
break;
case ir_loop_jump::jump_continue:
ir_to_mesa_emit_op1(NULL, OPCODE_CONT,
ir_to_mesa_undef_dst, ir_to_mesa_undef);
break;
}
}
void
ir_to_mesa_visitor::visit(ir_function_signature *ir)
{
assert(0);
(void)ir;
}
void
ir_to_mesa_visitor::visit(ir_function *ir)
{
/* Ignore function bodies other than main() -- we shouldn't see calls to
* them since they should all be inlined before we get to ir_to_mesa.
*/
if (strcmp(ir->name, "main") == 0) {
const ir_function_signature *sig;
exec_list empty;
sig = ir->matching_signature(&empty);
assert(sig);
foreach_iter(exec_list_iterator, iter, sig->body) {
ir_instruction *ir = (ir_instruction *)iter.get();
ir->accept(this);
}
}
}
void
ir_to_mesa_visitor::visit(ir_expression *ir)
{
unsigned int operand;
struct ir_to_mesa_src_reg op[2];
struct ir_to_mesa_src_reg result_src;
struct ir_to_mesa_dst_reg result_dst;
const glsl_type *vec4_type = glsl_type::get_instance(GLSL_TYPE_FLOAT, 4, 1);
const glsl_type *vec3_type = glsl_type::get_instance(GLSL_TYPE_FLOAT, 3, 1);
const glsl_type *vec2_type = glsl_type::get_instance(GLSL_TYPE_FLOAT, 2, 1);
for (operand = 0; operand < ir->get_num_operands(); operand++) {
this->result.file = PROGRAM_UNDEFINED;
ir->operands[operand]->accept(this);
if (this->result.file == PROGRAM_UNDEFINED) {
ir_print_visitor v;
printf("Failed to get tree for expression operand:\n");
ir->operands[operand]->accept(&v);
exit(1);
}
op[operand] = this->result;
/* Only expression implemented for matrices yet */
assert(!ir->operands[operand]->type->is_matrix() ||
ir->operation == ir_binop_mul);
}
this->result.file = PROGRAM_UNDEFINED;
/* Storage for our result. Ideally for an assignment we'd be using
* the actual storage for the result here, instead.
*/
result_src = get_temp(ir->type);
/* convenience for the emit functions below. */
result_dst = ir_to_mesa_dst_reg_from_src(result_src);
/* Limit writes to the channels that will be used by result_src later.
* This does limit this temp's use as a temporary for multi-instruction
* sequences.
*/
result_dst.writemask = (1 << ir->type->vector_elements) - 1;
switch (ir->operation) {
case ir_unop_logic_not:
ir_to_mesa_emit_op2(ir, OPCODE_SEQ, result_dst,
op[0], src_reg_for_float(0.0));
break;
case ir_unop_neg:
op[0].negate = ~op[0].negate;
result_src = op[0];
break;
case ir_unop_exp:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_EXP, result_dst, op[0]);
break;
case ir_unop_exp2:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_EX2, result_dst, op[0]);
break;
case ir_unop_log:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_LOG, result_dst, op[0]);
break;
case ir_unop_log2:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_LG2, result_dst, op[0]);
break;
case ir_unop_sin:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_SIN, result_dst, op[0]);
break;
case ir_unop_cos:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_COS, result_dst, op[0]);
break;
case ir_binop_add:
ir_to_mesa_emit_op2(ir, OPCODE_ADD, result_dst, op[0], op[1]);
break;
case ir_binop_sub:
ir_to_mesa_emit_op2(ir, OPCODE_SUB, result_dst, op[0], op[1]);
break;
case ir_binop_mul:
if (ir->operands[0]->type->is_matrix() &&
!ir->operands[1]->type->is_matrix()) {
if (ir->operands[0]->type->is_scalar()) {
ir_to_mesa_dst_reg dst_column = result_dst;
ir_to_mesa_src_reg src_column = op[0];
for (int i = 0; i < ir->operands[0]->type->matrix_columns; i++) {
ir_to_mesa_emit_op2(ir, OPCODE_MUL,
dst_column, src_column, op[1]);
dst_column.index++;
src_column.index++;
}
} else {
ir_to_mesa_dst_reg dst_chan = result_dst;
ir_to_mesa_src_reg src_column = op[0];
ir_to_mesa_src_reg src_chan = op[1];
for (int i = 0; i < ir->operands[0]->type->matrix_columns; i++) {
dst_chan.writemask = (1 << i);
src_chan.swizzle = MAKE_SWIZZLE4(i, i, i, i);
ir_to_mesa_emit_op2(ir, OPCODE_MUL,
dst_chan, src_column, src_chan);
src_column.index++;
}
}
} else {
assert(!ir->operands[0]->type->is_matrix());
assert(!ir->operands[1]->type->is_matrix());
ir_to_mesa_emit_op2(ir, OPCODE_MUL, result_dst, op[0], op[1]);
}
break;
case ir_binop_div:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_RCP, result_dst, op[1]);
ir_to_mesa_emit_op2(ir, OPCODE_MUL, result_dst, op[0], result_src);
break;
case ir_binop_less:
ir_to_mesa_emit_op2(ir, OPCODE_SLT, result_dst, op[0], op[1]);
break;
case ir_binop_greater:
ir_to_mesa_emit_op2(ir, OPCODE_SGT, result_dst, op[0], op[1]);
break;
case ir_binop_lequal:
ir_to_mesa_emit_op2(ir, OPCODE_SLE, result_dst, op[0], op[1]);
break;
case ir_binop_gequal:
ir_to_mesa_emit_op2(ir, OPCODE_SGE, result_dst, op[0], op[1]);
break;
case ir_binop_equal:
ir_to_mesa_emit_op2(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
break;
case ir_binop_logic_xor:
case ir_binop_nequal:
ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst, op[0], op[1]);
break;
case ir_binop_logic_or:
/* This could be a saturated add and skip the SNE. */
ir_to_mesa_emit_op2(ir, OPCODE_ADD,
result_dst,
op[0], op[1]);
ir_to_mesa_emit_op2(ir, OPCODE_SNE,
result_dst,
result_src, src_reg_for_float(0.0));
break;
case ir_binop_logic_and:
/* the bool args are stored as float 0.0 or 1.0, so "mul" gives us "and". */
ir_to_mesa_emit_op2(ir, OPCODE_MUL,
result_dst,
op[0], op[1]);
break;
case ir_binop_dot:
if (ir->operands[0]->type == vec4_type) {
assert(ir->operands[1]->type == vec4_type);
ir_to_mesa_emit_op2(ir, OPCODE_DP4,
result_dst,
op[0], op[1]);
} else if (ir->operands[0]->type == vec3_type) {
assert(ir->operands[1]->type == vec3_type);
ir_to_mesa_emit_op2(ir, OPCODE_DP3,
result_dst,
op[0], op[1]);
} else if (ir->operands[0]->type == vec2_type) {
assert(ir->operands[1]->type == vec2_type);
ir_to_mesa_emit_op2(ir, OPCODE_DP2,
result_dst,
op[0], op[1]);
}
break;
case ir_unop_sqrt:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_RSQ, result_dst, op[0]);
ir_to_mesa_emit_op1(ir, OPCODE_RCP, result_dst, result_src);
break;
case ir_unop_rsq:
ir_to_mesa_emit_scalar_op1(ir, OPCODE_RSQ, result_dst, op[0]);
break;
case ir_unop_i2f:
/* Mesa IR lacks types, ints are stored as truncated floats. */
result_src = op[0];
break;
case ir_unop_f2i:
ir_to_mesa_emit_op1(ir, OPCODE_TRUNC, result_dst, op[0]);
break;
case ir_unop_f2b:
ir_to_mesa_emit_op2(ir, OPCODE_SNE, result_dst,
result_src, src_reg_for_float(0.0));
break;
case ir_unop_trunc:
ir_to_mesa_emit_op1(ir, OPCODE_TRUNC, result_dst, op[0]);
break;
case ir_unop_ceil:
op[0].negate = ~op[0].negate;
ir_to_mesa_emit_op1(ir, OPCODE_FLR, result_dst, op[0]);
result_src.negate = ~result_src.negate;
break;
case ir_unop_floor:
ir_to_mesa_emit_op1(ir, OPCODE_FLR, result_dst, op[0]);
break;
case ir_binop_min:
ir_to_mesa_emit_op2(ir, OPCODE_MIN, result_dst, op[0], op[1]);
break;
case ir_binop_max:
ir_to_mesa_emit_op2(ir, OPCODE_MAX, result_dst, op[0], op[1]);
break;
default:
ir_print_visitor v;
printf("Failed to get tree for expression:\n");
ir->accept(&v);
exit(1);
break;
}
this->result = result_src;
}
void
ir_to_mesa_visitor::visit(ir_swizzle *ir)
{
ir_to_mesa_src_reg src_reg;
int i;
int swizzle[4];
/* Note that this is only swizzles in expressions, not those on the left
* hand side of an assignment, which do write masking. See ir_assignment
* for that.
*/
ir->val->accept(this);
src_reg = this->result;
assert(src_reg.file != PROGRAM_UNDEFINED);
for (i = 0; i < 4; i++) {
if (i < ir->type->vector_elements) {
switch (i) {
case 0:
swizzle[i] = ir->mask.x;
break;
case 1:
swizzle[i] = ir->mask.y;
break;
case 2:
swizzle[i] = ir->mask.z;
break;
case 3:
swizzle[i] = ir->mask.w;
break;
}
} else {
/* If the type is smaller than a vec4, replicate the last
* channel out.
*/
swizzle[i] = ir->type->vector_elements - 1;
}
}
src_reg.swizzle = MAKE_SWIZZLE4(swizzle[0],
swizzle[1],
swizzle[2],
swizzle[3]);
this->result = src_reg;
}
void
ir_to_mesa_visitor::visit(ir_dereference_variable *ir)
{
ir_to_mesa_src_reg src_reg;
temp_entry *entry = find_variable_storage(ir->var);
unsigned int i;
bool var_in;
if (!entry) {
switch (ir->var->mode) {
case ir_var_uniform:
entry = new(ctx) temp_entry(ir->var, PROGRAM_UNIFORM,
this->next_uniform);
this->variable_storage.push_tail(entry);
this->next_uniform += type_size(ir->var->type);
break;
case ir_var_in:
case ir_var_out:
case ir_var_inout:
var_in = (ir->var->mode == ir_var_in ||
ir->var->mode == ir_var_inout);
for (i = 0; i < ARRAY_SIZE(builtin_var_to_mesa_reg); i++) {
bool in = builtin_var_to_mesa_reg[i].file == PROGRAM_INPUT;
if (strcmp(ir->var->name, builtin_var_to_mesa_reg[i].name) == 0 &&
!(var_in ^ in))
break;
}
if (i == ARRAY_SIZE(builtin_var_to_mesa_reg)) {
printf("Failed to find builtin for %s variable %s\n",
var_in ? "in" : "out",
ir->var->name);
abort();
}
entry = new(ctx) temp_entry(ir->var,
builtin_var_to_mesa_reg[i].file,
builtin_var_to_mesa_reg[i].index);
break;
case ir_var_auto:
entry = new(ctx) temp_entry(ir->var, PROGRAM_TEMPORARY,
this->next_temp);
this->variable_storage.push_tail(entry);
next_temp += type_size(ir->var->type);
break;
}
if (!entry) {
printf("Failed to make storage for %s\n", ir->var->name);
exit(1);
}
}
src_reg.file = entry->file;
src_reg.index = entry->index;
/* If the type is smaller than a vec4, replicate the last channel out. */
src_reg.swizzle = swizzle_for_size(ir->var->type->vector_elements);
src_reg.reladdr = false;
src_reg.negate = 0;
this->result = src_reg;
}
void
ir_to_mesa_visitor::visit(ir_dereference_array *ir)
{
ir_constant *index;
ir_to_mesa_src_reg src_reg;
index = ir->array_index->constant_expression_value();
/* By the time we make it to this stage, matrices should be broken down
* to vectors.
*/
assert(!ir->type->is_matrix());
ir->array->accept(this);
src_reg = this->result;
if (src_reg.file == PROGRAM_INPUT ||
src_reg.file == PROGRAM_OUTPUT) {
assert(index); /* FINISHME: Handle variable indexing of builtins. */
src_reg.index += index->value.i[0];
} else {
if (index) {
src_reg.index += index->value.i[0];
} else {
ir_to_mesa_src_reg array_base = this->result;
/* Variable index array dereference. It eats the "vec4" of the
* base of the array and an index that offsets the Mesa register
* index.
*/
ir->array_index->accept(this);
/* FINISHME: This doesn't work when we're trying to do the LHS
* of an assignment.
*/
src_reg.reladdr = true;
ir_to_mesa_emit_op1(ir, OPCODE_ARL, ir_to_mesa_address_reg,
this->result);
this->result = get_temp(ir->type);
ir_to_mesa_emit_op1(ir, OPCODE_MOV,
ir_to_mesa_dst_reg_from_src(this->result),
src_reg);
}
}
/* If the type is smaller than a vec4, replicate the last channel out. */
src_reg.swizzle = swizzle_for_size(ir->type->vector_elements);
this->result = src_reg;
}
void
ir_to_mesa_visitor::visit(ir_dereference_record *ir)
{
unsigned int i;
const glsl_type *struct_type = ir->record->type;
int offset = 0;
ir->record->accept(this);
for (i = 0; i < struct_type->length; i++) {
if (strcmp(struct_type->fields.structure[i].name, ir->field) == 0)
break;
offset += type_size(struct_type->fields.structure[i].type);
}
this->result.index += offset;
}
/**
* We want to be careful in assignment setup to hit the actual storage
* instead of potentially using a temporary like we might with the
* ir_dereference handler.
*
* Thanks to ir_swizzle_swizzle, and ir_vec_index_to_swizzle, we
* should only see potentially one variable array index of a vector,
* and one swizzle, before getting to actual vec4 storage. So handle
* those, then go use ir_dereference to handle the rest.
*/
static struct ir_to_mesa_dst_reg
get_assignment_lhs(ir_instruction *ir, ir_to_mesa_visitor *v)
{
struct ir_to_mesa_dst_reg dst_reg;
ir_dereference *deref;
ir_swizzle *swiz;
/* Use the rvalue deref handler for the most part. We'll ignore
* swizzles in it and write swizzles using writemask, though.
*/
ir->accept(v);
dst_reg = ir_to_mesa_dst_reg_from_src(v->result);
if ((deref = ir->as_dereference())) {
ir_dereference_array *deref_array = ir->as_dereference_array();
assert(!deref_array || deref_array->array->type->is_array());
ir->accept(v);
} else if ((swiz = ir->as_swizzle())) {
dst_reg.writemask = 0;
if (swiz->mask.num_components >= 1)
dst_reg.writemask |= (1 << swiz->mask.x);
if (swiz->mask.num_components >= 2)
dst_reg.writemask |= (1 << swiz->mask.y);
if (swiz->mask.num_components >= 3)
dst_reg.writemask |= (1 << swiz->mask.z);
if (swiz->mask.num_components >= 4)
dst_reg.writemask |= (1 << swiz->mask.w);
}
return dst_reg;
}
void
ir_to_mesa_visitor::visit(ir_assignment *ir)
{
struct ir_to_mesa_dst_reg l;
struct ir_to_mesa_src_reg r;
assert(!ir->lhs->type->is_matrix());
assert(!ir->lhs->type->is_array());
assert(ir->lhs->type->base_type != GLSL_TYPE_STRUCT);
l = get_assignment_lhs(ir->lhs, this);
ir->rhs->accept(this);
r = this->result;
assert(l.file != PROGRAM_UNDEFINED);
assert(r.file != PROGRAM_UNDEFINED);
if (ir->condition) {
ir_constant *condition_constant;
condition_constant = ir->condition->constant_expression_value();
assert(condition_constant && condition_constant->value.b[0]);
}
ir_to_mesa_emit_op1(ir, OPCODE_MOV, l, r);
}
void
ir_to_mesa_visitor::visit(ir_constant *ir)
{
ir_to_mesa_src_reg src_reg;
assert(ir->type->base_type == GLSL_TYPE_FLOAT ||
ir->type->base_type == GLSL_TYPE_UINT ||
ir->type->base_type == GLSL_TYPE_INT ||
ir->type->base_type == GLSL_TYPE_BOOL);
/* FINISHME: This will end up being _mesa_add_unnamed_constant,
* which handles sharing values and sharing channels of vec4
* constants for small values.
*/
/* FINISHME: Do something with the constant values for now.
*/
src_reg.file = PROGRAM_CONSTANT;
src_reg.index = this->next_constant;
src_reg.swizzle = SWIZZLE_NOOP;
src_reg.reladdr = false;
src_reg.negate = 0;
this->next_constant += type_size(ir->type);
this->result = src_reg;
}
void
ir_to_mesa_visitor::visit(ir_call *ir)
{
printf("Can't support call to %s\n", ir->callee_name());
exit(1);
}
void
ir_to_mesa_visitor::visit(ir_texture *ir)
{
assert(0);
ir->coordinate->accept(this);
}
void
ir_to_mesa_visitor::visit(ir_return *ir)
{
assert(0);
ir->get_value()->accept(this);
}
void
ir_to_mesa_visitor::visit(ir_if *ir)
{
ir_to_mesa_instruction *if_inst, *else_inst = NULL;
ir->condition->accept(this);
assert(this->result.file != PROGRAM_UNDEFINED);
if_inst = ir_to_mesa_emit_op1(ir->condition,
OPCODE_IF, ir_to_mesa_undef_dst,
this->result);
this->instructions.push_tail(if_inst);
visit_exec_list(&ir->then_instructions, this);
if (!ir->else_instructions.is_empty()) {
else_inst = ir_to_mesa_emit_op1(ir->condition, OPCODE_ELSE,
ir_to_mesa_undef_dst,
ir_to_mesa_undef);
visit_exec_list(&ir->then_instructions, this);
}
if_inst = ir_to_mesa_emit_op1(ir->condition, OPCODE_ENDIF,
ir_to_mesa_undef_dst, ir_to_mesa_undef);
}
ir_to_mesa_visitor::ir_to_mesa_visitor()
{
result.file = PROGRAM_UNDEFINED;
next_temp = 1;
next_constant = 0;
next_uniform = 0;
}
static struct prog_src_register
mesa_src_reg_from_ir_src_reg(ir_to_mesa_src_reg reg)
{
struct prog_src_register mesa_reg;
mesa_reg.File = reg.file;
assert(reg.index < (1 << INST_INDEX_BITS) - 1);
mesa_reg.Index = reg.index;
mesa_reg.Swizzle = reg.swizzle;
mesa_reg.RelAddr = reg.reladdr;
return mesa_reg;
}
static void
set_branchtargets(struct prog_instruction *mesa_instructions,
int num_instructions)
{
int if_count = 0, loop_count;
int *if_stack, *loop_stack;
int if_stack_pos = 0, loop_stack_pos = 0;
int i, j;
for (i = 0; i < num_instructions; i++) {
switch (mesa_instructions[i].Opcode) {
case OPCODE_IF:
if_count++;
break;
case OPCODE_BGNLOOP:
loop_count++;
break;
case OPCODE_BRK:
case OPCODE_CONT:
mesa_instructions[i].BranchTarget = -1;
break;
default:
break;
}
}
if_stack = (int *)calloc(if_count, sizeof(*if_stack));
loop_stack = (int *)calloc(loop_count, sizeof(*loop_stack));
for (i = 0; i < num_instructions; i++) {
switch (mesa_instructions[i].Opcode) {
case OPCODE_IF:
if_stack[if_stack_pos] = i;
if_stack_pos++;
break;
case OPCODE_ELSE:
mesa_instructions[if_stack[if_stack_pos - 1]].BranchTarget = i;
if_stack[if_stack_pos - 1] = i;
break;
case OPCODE_ENDIF:
mesa_instructions[if_stack[if_stack_pos - 1]].BranchTarget = i;
if_stack_pos--;
break;
case OPCODE_BGNLOOP:
loop_stack[loop_stack_pos] = i;
loop_stack_pos++;
break;
case OPCODE_ENDLOOP:
loop_stack_pos--;
/* Rewrite any breaks/conts at this nesting level (haven't
* already had a BranchTarget assigned) to point to the end
* of the loop.
*/
for (j = loop_stack[loop_stack_pos]; j < i; j++) {
if (mesa_instructions[j].Opcode == OPCODE_BRK ||
mesa_instructions[j].Opcode == OPCODE_CONT) {
if (mesa_instructions[j].BranchTarget == -1) {
mesa_instructions[j].BranchTarget = i;
}
}
}
/* The loop ends point at each other. */
mesa_instructions[i].BranchTarget = loop_stack[loop_stack_pos];
mesa_instructions[loop_stack[loop_stack_pos]].BranchTarget = i;
default:
break;
}
}
free(if_stack);
}
static void
print_program(struct prog_instruction *mesa_instructions,
ir_instruction **mesa_instruction_annotation,
int num_instructions)
{
ir_instruction *last_ir = NULL;
int i;
for (i = 0; i < num_instructions; i++) {
struct prog_instruction *mesa_inst = mesa_instructions + i;
ir_instruction *ir = mesa_instruction_annotation[i];
if (last_ir != ir && ir) {
ir_print_visitor print;
ir->accept(&print);
printf("\n");
last_ir = ir;
}
_mesa_print_instruction(mesa_inst);
}
}
void
do_ir_to_mesa(exec_list *instructions)
{
ir_to_mesa_visitor v;
struct prog_instruction *mesa_instructions, *mesa_inst;
ir_instruction **mesa_instruction_annotation;
int i;
v.ctx = talloc_new(NULL);
visit_exec_list(instructions, &v);
int num_instructions = 0;
foreach_iter(exec_list_iterator, iter, v.instructions) {
num_instructions++;
}
mesa_instructions =
(struct prog_instruction *)calloc(num_instructions,
sizeof(*mesa_instructions));
mesa_instruction_annotation =
(ir_instruction **)calloc(num_instructions,
sizeof(*mesa_instruction_annotation));
mesa_inst = mesa_instructions;
i = 0;
foreach_iter(exec_list_iterator, iter, v.instructions) {
ir_to_mesa_instruction *inst = (ir_to_mesa_instruction *)iter.get();
mesa_inst->Opcode = inst->op;
mesa_inst->DstReg.File = inst->dst_reg.file;
mesa_inst->DstReg.Index = inst->dst_reg.index;
mesa_inst->DstReg.CondMask = COND_TR;
mesa_inst->DstReg.WriteMask = inst->dst_reg.writemask;
mesa_inst->SrcReg[0] = mesa_src_reg_from_ir_src_reg(inst->src_reg[0]);
mesa_inst->SrcReg[1] = mesa_src_reg_from_ir_src_reg(inst->src_reg[1]);
mesa_inst->SrcReg[2] = mesa_src_reg_from_ir_src_reg(inst->src_reg[2]);
mesa_instruction_annotation[i] = inst->ir;
mesa_inst++;
i++;
}
set_branchtargets(mesa_instructions, num_instructions);
print_program(mesa_instructions, mesa_instruction_annotation, num_instructions);
free(mesa_instruction_annotation);
talloc_free(v.ctx);
}
|