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
|
/*
* 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 opt_algebraic.cpp
*
* Takes advantage of association, commutivity, and other algebraic
* properties to simplify expressions.
*/
#include "ir.h"
#include "ir_visitor.h"
#include "ir_rvalue_visitor.h"
#include "ir_optimization.h"
#include "ir_builder.h"
#include "glsl_types.h"
using namespace ir_builder;
namespace {
/**
* Visitor class for replacing expressions with ir_constant values.
*/
class ir_algebraic_visitor : public ir_rvalue_visitor {
public:
ir_algebraic_visitor(bool native_integers,
const struct gl_shader_compiler_options *options)
: options(options)
{
this->progress = false;
this->mem_ctx = NULL;
this->native_integers = native_integers;
}
virtual ~ir_algebraic_visitor()
{
}
ir_rvalue *handle_expression(ir_expression *ir);
void handle_rvalue(ir_rvalue **rvalue);
bool reassociate_constant(ir_expression *ir1,
int const_index,
ir_constant *constant,
ir_expression *ir2);
void reassociate_operands(ir_expression *ir1,
int op1,
ir_expression *ir2,
int op2);
ir_rvalue *swizzle_if_required(ir_expression *expr,
ir_rvalue *operand);
const struct gl_shader_compiler_options *options;
void *mem_ctx;
bool native_integers;
bool progress;
};
} /* unnamed namespace */
static inline bool
is_vec_zero(ir_constant *ir)
{
return (ir == NULL) ? false : ir->is_zero();
}
static inline bool
is_vec_one(ir_constant *ir)
{
return (ir == NULL) ? false : ir->is_one();
}
static inline bool
is_vec_two(ir_constant *ir)
{
return (ir == NULL) ? false : ir->is_value(2.0, 2);
}
static inline bool
is_vec_negative_one(ir_constant *ir)
{
return (ir == NULL) ? false : ir->is_negative_one();
}
static inline bool
is_vec_basis(ir_constant *ir)
{
return (ir == NULL) ? false : ir->is_basis();
}
static inline bool
is_valid_vec_const(ir_constant *ir)
{
if (ir == NULL)
return false;
if (!ir->type->is_scalar() && !ir->type->is_vector())
return false;
return true;
}
static inline bool
is_less_than_one(ir_constant *ir)
{
if (!is_valid_vec_const(ir))
return false;
unsigned component = 0;
for (int c = 0; c < ir->type->vector_elements; c++) {
if (ir->get_float_component(c) < 1.0f)
component++;
}
return (component == ir->type->vector_elements);
}
static inline bool
is_greater_than_zero(ir_constant *ir)
{
if (!is_valid_vec_const(ir))
return false;
unsigned component = 0;
for (int c = 0; c < ir->type->vector_elements; c++) {
if (ir->get_float_component(c) > 0.0f)
component++;
}
return (component == ir->type->vector_elements);
}
static void
update_type(ir_expression *ir)
{
if (ir->operands[0]->type->is_vector())
ir->type = ir->operands[0]->type;
else
ir->type = ir->operands[1]->type;
}
/* Recognize (v.x + v.y) + (v.z + v.w) as dot(v, 1.0) */
static ir_expression *
try_replace_with_dot(ir_expression *expr0, ir_expression *expr1, void *mem_ctx)
{
if (expr0 && expr0->operation == ir_binop_add &&
expr0->type->is_float() &&
expr1 && expr1->operation == ir_binop_add &&
expr1->type->is_float()) {
ir_swizzle *x = expr0->operands[0]->as_swizzle();
ir_swizzle *y = expr0->operands[1]->as_swizzle();
ir_swizzle *z = expr1->operands[0]->as_swizzle();
ir_swizzle *w = expr1->operands[1]->as_swizzle();
if (!x || x->mask.num_components != 1 ||
!y || y->mask.num_components != 1 ||
!z || z->mask.num_components != 1 ||
!w || w->mask.num_components != 1) {
return NULL;
}
bool swiz_seen[4] = {false, false, false, false};
swiz_seen[x->mask.x] = true;
swiz_seen[y->mask.x] = true;
swiz_seen[z->mask.x] = true;
swiz_seen[w->mask.x] = true;
if (!swiz_seen[0] || !swiz_seen[1] ||
!swiz_seen[2] || !swiz_seen[3]) {
return NULL;
}
if (x->val->equals(y->val) &&
x->val->equals(z->val) &&
x->val->equals(w->val)) {
return dot(x->val, new(mem_ctx) ir_constant(1.0f, 4));
}
}
return NULL;
}
void
ir_algebraic_visitor::reassociate_operands(ir_expression *ir1,
int op1,
ir_expression *ir2,
int op2)
{
ir_rvalue *temp = ir2->operands[op2];
ir2->operands[op2] = ir1->operands[op1];
ir1->operands[op1] = temp;
/* Update the type of ir2. The type of ir1 won't have changed --
* base types matched, and at least one of the operands of the 2
* binops is still a vector if any of them were.
*/
update_type(ir2);
this->progress = true;
}
/**
* Reassociates a constant down a tree of adds or multiplies.
*
* Consider (2 * (a * (b * 0.5))). We want to send up with a * b.
*/
bool
ir_algebraic_visitor::reassociate_constant(ir_expression *ir1, int const_index,
ir_constant *constant,
ir_expression *ir2)
{
if (!ir2 || ir1->operation != ir2->operation)
return false;
/* Don't want to even think about matrices. */
if (ir1->operands[0]->type->is_matrix() ||
ir1->operands[1]->type->is_matrix() ||
ir2->operands[0]->type->is_matrix() ||
ir2->operands[1]->type->is_matrix())
return false;
ir_constant *ir2_const[2];
ir2_const[0] = ir2->operands[0]->constant_expression_value();
ir2_const[1] = ir2->operands[1]->constant_expression_value();
if (ir2_const[0] && ir2_const[1])
return false;
if (ir2_const[0]) {
reassociate_operands(ir1, const_index, ir2, 1);
return true;
} else if (ir2_const[1]) {
reassociate_operands(ir1, const_index, ir2, 0);
return true;
}
if (reassociate_constant(ir1, const_index, constant,
ir2->operands[0]->as_expression())) {
update_type(ir2);
return true;
}
if (reassociate_constant(ir1, const_index, constant,
ir2->operands[1]->as_expression())) {
update_type(ir2);
return true;
}
return false;
}
/* When eliminating an expression and just returning one of its operands,
* we may need to swizzle that operand out to a vector if the expression was
* vector type.
*/
ir_rvalue *
ir_algebraic_visitor::swizzle_if_required(ir_expression *expr,
ir_rvalue *operand)
{
if (expr->type->is_vector() && operand->type->is_scalar()) {
return new(mem_ctx) ir_swizzle(operand, 0, 0, 0, 0,
expr->type->vector_elements);
} else
return operand;
}
ir_rvalue *
ir_algebraic_visitor::handle_expression(ir_expression *ir)
{
ir_constant *op_const[4] = {NULL, NULL, NULL, NULL};
ir_expression *op_expr[4] = {NULL, NULL, NULL, NULL};
unsigned int i;
assert(ir->get_num_operands() <= 4);
for (i = 0; i < ir->get_num_operands(); i++) {
if (ir->operands[i]->type->is_matrix())
return ir;
op_const[i] = ir->operands[i]->constant_expression_value();
op_expr[i] = ir->operands[i]->as_expression();
}
if (this->mem_ctx == NULL)
this->mem_ctx = ralloc_parent(ir);
switch (ir->operation) {
case ir_unop_bit_not:
if (op_expr[0] && op_expr[0]->operation == ir_unop_bit_not)
return op_expr[0]->operands[0];
break;
case ir_unop_abs:
if (op_expr[0] == NULL)
break;
switch (op_expr[0]->operation) {
case ir_unop_abs:
case ir_unop_neg:
return abs(op_expr[0]->operands[0]);
default:
break;
}
break;
case ir_unop_neg:
if (op_expr[0] == NULL)
break;
if (op_expr[0]->operation == ir_unop_neg) {
return op_expr[0]->operands[0];
}
break;
case ir_unop_exp:
if (op_expr[0] == NULL)
break;
if (op_expr[0]->operation == ir_unop_log) {
return op_expr[0]->operands[0];
}
break;
case ir_unop_log:
if (op_expr[0] == NULL)
break;
if (op_expr[0]->operation == ir_unop_exp) {
return op_expr[0]->operands[0];
}
break;
case ir_unop_exp2:
if (op_expr[0] == NULL)
break;
if (op_expr[0]->operation == ir_unop_log2) {
return op_expr[0]->operands[0];
}
if (!options->EmitNoPow && op_expr[0]->operation == ir_binop_mul) {
for (int log2_pos = 0; log2_pos < 2; log2_pos++) {
ir_expression *log2_expr =
op_expr[0]->operands[log2_pos]->as_expression();
if (log2_expr && log2_expr->operation == ir_unop_log2) {
return new(mem_ctx) ir_expression(ir_binop_pow,
ir->type,
log2_expr->operands[0],
op_expr[0]->operands[1 - log2_pos]);
}
}
}
break;
case ir_unop_log2:
if (op_expr[0] == NULL)
break;
if (op_expr[0]->operation == ir_unop_exp2) {
return op_expr[0]->operands[0];
}
break;
case ir_unop_logic_not: {
enum ir_expression_operation new_op = ir_unop_logic_not;
if (op_expr[0] == NULL)
break;
switch (op_expr[0]->operation) {
case ir_binop_less: new_op = ir_binop_gequal; break;
case ir_binop_greater: new_op = ir_binop_lequal; break;
case ir_binop_lequal: new_op = ir_binop_greater; break;
case ir_binop_gequal: new_op = ir_binop_less; break;
case ir_binop_equal: new_op = ir_binop_nequal; break;
case ir_binop_nequal: new_op = ir_binop_equal; break;
case ir_binop_all_equal: new_op = ir_binop_any_nequal; break;
case ir_binop_any_nequal: new_op = ir_binop_all_equal; break;
default:
/* The default case handler is here to silence a warning from GCC.
*/
break;
}
if (new_op != ir_unop_logic_not) {
return new(mem_ctx) ir_expression(new_op,
ir->type,
op_expr[0]->operands[0],
op_expr[0]->operands[1]);
}
break;
}
case ir_binop_add:
if (is_vec_zero(op_const[0]))
return ir->operands[1];
if (is_vec_zero(op_const[1]))
return ir->operands[0];
/* Reassociate addition of constants so that we can do constant
* folding.
*/
if (op_const[0] && !op_const[1])
reassociate_constant(ir, 0, op_const[0], op_expr[1]);
if (op_const[1] && !op_const[0])
reassociate_constant(ir, 1, op_const[1], op_expr[0]);
/* Recognize (v.x + v.y) + (v.z + v.w) as dot(v, 1.0) */
if (options->OptimizeForAOS) {
ir_expression *expr = try_replace_with_dot(op_expr[0], op_expr[1],
mem_ctx);
if (expr)
return expr;
}
/* Replace (-x + y) * a + x and commutative variations with lrp(x, y, a).
*
* (-x + y) * a + x
* (x * -a) + (y * a) + x
* x + (x * -a) + (y * a)
* x * (1 - a) + y * a
* lrp(x, y, a)
*/
for (int mul_pos = 0; mul_pos < 2; mul_pos++) {
ir_expression *mul = op_expr[mul_pos];
if (!mul || mul->operation != ir_binop_mul)
continue;
/* Multiply found on one of the operands. Now check for an
* inner addition operation.
*/
for (int inner_add_pos = 0; inner_add_pos < 2; inner_add_pos++) {
ir_expression *inner_add =
mul->operands[inner_add_pos]->as_expression();
if (!inner_add || inner_add->operation != ir_binop_add)
continue;
/* Inner addition found on one of the operands. Now check for
* one of the operands of the inner addition to be the negative
* of x_operand.
*/
for (int neg_pos = 0; neg_pos < 2; neg_pos++) {
ir_expression *neg =
inner_add->operands[neg_pos]->as_expression();
if (!neg || neg->operation != ir_unop_neg)
continue;
ir_rvalue *x_operand = ir->operands[1 - mul_pos];
if (!neg->operands[0]->equals(x_operand))
continue;
ir_rvalue *y_operand = inner_add->operands[1 - neg_pos];
ir_rvalue *a_operand = mul->operands[1 - inner_add_pos];
if (x_operand->type != y_operand->type ||
x_operand->type != a_operand->type)
continue;
return lrp(x_operand, y_operand, a_operand);
}
}
}
break;
case ir_binop_sub:
if (is_vec_zero(op_const[0]))
return neg(ir->operands[1]);
if (is_vec_zero(op_const[1]))
return ir->operands[0];
break;
case ir_binop_mul:
if (is_vec_one(op_const[0]))
return ir->operands[1];
if (is_vec_one(op_const[1]))
return ir->operands[0];
if (is_vec_zero(op_const[0]) || is_vec_zero(op_const[1]))
return ir_constant::zero(ir, ir->type);
if (is_vec_negative_one(op_const[0]))
return neg(ir->operands[1]);
if (is_vec_negative_one(op_const[1]))
return neg(ir->operands[0]);
/* Reassociate multiplication of constants so that we can do
* constant folding.
*/
if (op_const[0] && !op_const[1])
reassociate_constant(ir, 0, op_const[0], op_expr[1]);
if (op_const[1] && !op_const[0])
reassociate_constant(ir, 1, op_const[1], op_expr[0]);
break;
case ir_binop_div:
if (is_vec_one(op_const[0]) && ir->type->base_type == GLSL_TYPE_FLOAT) {
return new(mem_ctx) ir_expression(ir_unop_rcp,
ir->operands[1]->type,
ir->operands[1],
NULL);
}
if (is_vec_one(op_const[1]))
return ir->operands[0];
break;
case ir_binop_dot:
if (is_vec_zero(op_const[0]) || is_vec_zero(op_const[1]))
return ir_constant::zero(mem_ctx, ir->type);
if (is_vec_basis(op_const[0])) {
unsigned component = 0;
for (unsigned c = 0; c < op_const[0]->type->vector_elements; c++) {
if (op_const[0]->value.f[c] == 1.0)
component = c;
}
return new(mem_ctx) ir_swizzle(ir->operands[1], component, 0, 0, 0, 1);
}
if (is_vec_basis(op_const[1])) {
unsigned component = 0;
for (unsigned c = 0; c < op_const[1]->type->vector_elements; c++) {
if (op_const[1]->value.f[c] == 1.0)
component = c;
}
return new(mem_ctx) ir_swizzle(ir->operands[0], component, 0, 0, 0, 1);
}
for (int i = 0; i < 2; i++) {
if (!op_const[i])
continue;
unsigned components[4] = { 0 }, count = 0;
for (unsigned c = 0; c < op_const[i]->type->vector_elements; c++) {
if (op_const[i]->value.f[c] == 0.0)
continue;
components[count] = c;
count++;
}
/* No channels had zero values; bail. */
if (count >= op_const[i]->type->vector_elements)
break;
/* Swizzle both operands to remove the channels that were zero. */
return new(mem_ctx)
ir_expression(ir_binop_dot, glsl_type::float_type,
new(mem_ctx) ir_swizzle(ir->operands[0],
components, count),
new(mem_ctx) ir_swizzle(ir->operands[1],
components, count));
}
break;
case ir_binop_less:
case ir_binop_lequal:
case ir_binop_greater:
case ir_binop_gequal:
case ir_binop_equal:
case ir_binop_nequal:
for (int add_pos = 0; add_pos < 2; add_pos++) {
ir_expression *add = op_expr[add_pos];
if (!add || add->operation != ir_binop_add)
continue;
ir_constant *zero = op_const[1 - add_pos];
if (!is_vec_zero(zero))
continue;
return new(mem_ctx) ir_expression(ir->operation,
add->operands[0],
neg(add->operands[1]));
}
break;
case ir_binop_rshift:
case ir_binop_lshift:
/* 0 >> x == 0 */
if (is_vec_zero(op_const[0]))
return ir->operands[0];
/* x >> 0 == x */
if (is_vec_zero(op_const[1]))
return ir->operands[0];
break;
case ir_binop_logic_and:
if (is_vec_one(op_const[0])) {
return ir->operands[1];
} else if (is_vec_one(op_const[1])) {
return ir->operands[0];
} else if (is_vec_zero(op_const[0]) || is_vec_zero(op_const[1])) {
return ir_constant::zero(mem_ctx, ir->type);
} else if (op_expr[0] && op_expr[0]->operation == ir_unop_logic_not &&
op_expr[1] && op_expr[1]->operation == ir_unop_logic_not) {
/* De Morgan's Law:
* (not A) and (not B) === not (A or B)
*/
return logic_not(logic_or(op_expr[0]->operands[0],
op_expr[1]->operands[0]));
} else if (ir->operands[0]->equals(ir->operands[1])) {
/* (a && a) == a */
return ir->operands[0];
}
break;
case ir_binop_logic_xor:
if (is_vec_zero(op_const[0])) {
return ir->operands[1];
} else if (is_vec_zero(op_const[1])) {
return ir->operands[0];
} else if (is_vec_one(op_const[0])) {
return logic_not(ir->operands[1]);
} else if (is_vec_one(op_const[1])) {
return logic_not(ir->operands[0]);
} else if (ir->operands[0]->equals(ir->operands[1])) {
/* (a ^^ a) == false */
return ir_constant::zero(mem_ctx, ir->type);
}
break;
case ir_binop_logic_or:
if (is_vec_zero(op_const[0])) {
return ir->operands[1];
} else if (is_vec_zero(op_const[1])) {
return ir->operands[0];
} else if (is_vec_one(op_const[0]) || is_vec_one(op_const[1])) {
ir_constant_data data;
for (unsigned i = 0; i < 16; i++)
data.b[i] = true;
return new(mem_ctx) ir_constant(ir->type, &data);
} else if (op_expr[0] && op_expr[0]->operation == ir_unop_logic_not &&
op_expr[1] && op_expr[1]->operation == ir_unop_logic_not) {
/* De Morgan's Law:
* (not A) or (not B) === not (A and B)
*/
return logic_not(logic_and(op_expr[0]->operands[0],
op_expr[1]->operands[0]));
} else if (ir->operands[0]->equals(ir->operands[1])) {
/* (a || a) == a */
return ir->operands[0];
}
break;
case ir_binop_pow:
/* 1^x == 1 */
if (is_vec_one(op_const[0]))
return op_const[0];
/* x^1 == x */
if (is_vec_one(op_const[1]))
return ir->operands[0];
/* pow(2,x) == exp2(x) */
if (is_vec_two(op_const[0]))
return expr(ir_unop_exp2, ir->operands[1]);
if (is_vec_two(op_const[1])) {
ir_variable *x = new(ir) ir_variable(ir->operands[1]->type, "x",
ir_var_temporary);
base_ir->insert_before(x);
base_ir->insert_before(assign(x, ir->operands[0]));
return mul(x, x);
}
break;
case ir_binop_min:
case ir_binop_max:
if (ir->type->base_type != GLSL_TYPE_FLOAT)
break;
/* Replace min(max) operations and its commutative combinations with
* a saturate operation
*/
for (int op = 0; op < 2; op++) {
ir_expression *minmax = op_expr[op];
ir_constant *outer_const = op_const[1 - op];
ir_expression_operation op_cond = (ir->operation == ir_binop_max) ?
ir_binop_min : ir_binop_max;
if (!minmax || !outer_const || (minmax->operation != op_cond))
continue;
/* Found a min(max) combination. Now try to see if its operands
* meet our conditions that we can do just a single saturate operation
*/
for (int minmax_op = 0; minmax_op < 2; minmax_op++) {
ir_rvalue *inner_val_a = minmax->operands[minmax_op];
ir_rvalue *inner_val_b = minmax->operands[1 - minmax_op];
if (!inner_val_a || !inner_val_b)
continue;
/* Found a {min|max} ({max|min} (x, 0.0), 1.0) operation and its variations */
if ((outer_const->is_one() && inner_val_a->is_zero()) ||
(inner_val_a->is_one() && outer_const->is_zero()))
return saturate(inner_val_b);
/* Found a {min|max} ({max|min} (x, 0.0), b) where b < 1.0
* and its variations
*/
if (is_less_than_one(outer_const) && inner_val_b->is_zero())
return expr(ir_binop_min, saturate(inner_val_a), outer_const);
if (!inner_val_b->as_constant())
continue;
if (is_less_than_one(inner_val_b->as_constant()) && outer_const->is_zero())
return expr(ir_binop_min, saturate(inner_val_a), inner_val_b);
/* Found a {min|max} ({max|min} (x, b), 1.0), where b > 0.0
* and its variations
*/
if (outer_const->is_one() && is_greater_than_zero(inner_val_b->as_constant()))
return expr(ir_binop_max, saturate(inner_val_a), inner_val_b);
if (inner_val_b->as_constant()->is_one() && is_greater_than_zero(outer_const))
return expr(ir_binop_max, saturate(inner_val_a), outer_const);
}
}
break;
case ir_unop_rcp:
if (op_expr[0] && op_expr[0]->operation == ir_unop_rcp)
return op_expr[0]->operands[0];
/* While ir_to_mesa.cpp will lower sqrt(x) to rcp(rsq(x)), it does so at
* its IR level, so we can always apply this transformation.
*/
if (op_expr[0] && op_expr[0]->operation == ir_unop_rsq)
return sqrt(op_expr[0]->operands[0]);
/* As far as we know, all backends are OK with rsq. */
if (op_expr[0] && op_expr[0]->operation == ir_unop_sqrt) {
return rsq(op_expr[0]->operands[0]);
}
break;
case ir_triop_fma:
/* Operands are op0 * op1 + op2. */
if (is_vec_zero(op_const[0]) || is_vec_zero(op_const[1])) {
return ir->operands[2];
} else if (is_vec_zero(op_const[2])) {
return mul(ir->operands[0], ir->operands[1]);
} else if (is_vec_one(op_const[0])) {
return add(ir->operands[1], ir->operands[2]);
} else if (is_vec_one(op_const[1])) {
return add(ir->operands[0], ir->operands[2]);
}
break;
case ir_triop_lrp:
/* Operands are (x, y, a). */
if (is_vec_zero(op_const[2])) {
return ir->operands[0];
} else if (is_vec_one(op_const[2])) {
return ir->operands[1];
} else if (ir->operands[0]->equals(ir->operands[1])) {
return ir->operands[0];
} else if (is_vec_zero(op_const[0])) {
return mul(ir->operands[1], ir->operands[2]);
} else if (is_vec_zero(op_const[1])) {
unsigned op2_components = ir->operands[2]->type->vector_elements;
ir_constant *one = new(mem_ctx) ir_constant(1.0f, op2_components);
return mul(ir->operands[0], add(one, neg(ir->operands[2])));
}
break;
case ir_triop_csel:
if (is_vec_one(op_const[0]))
return ir->operands[1];
if (is_vec_zero(op_const[0]))
return ir->operands[2];
break;
default:
break;
}
return ir;
}
void
ir_algebraic_visitor::handle_rvalue(ir_rvalue **rvalue)
{
if (!*rvalue)
return;
ir_expression *expr = (*rvalue)->as_expression();
if (!expr || expr->operation == ir_quadop_vector)
return;
ir_rvalue *new_rvalue = handle_expression(expr);
if (new_rvalue == *rvalue)
return;
/* If the expr used to be some vec OP scalar returning a vector, and the
* optimization gave us back a scalar, we still need to turn it into a
* vector.
*/
*rvalue = swizzle_if_required(expr, new_rvalue);
this->progress = true;
}
bool
do_algebraic(exec_list *instructions, bool native_integers,
const struct gl_shader_compiler_options *options)
{
ir_algebraic_visitor v(native_integers, options);
visit_list_elements(&v, instructions);
return v.progress;
}
|