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
|
/*
* Copyright (c) 2014 Scott Mansell
* Copyright © 2014 Broadcom
*
* 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 <stdio.h>
#include <inttypes.h>
#include "pipe/p_state.h"
#include "util/u_format.h"
#include "util/u_hash_table.h"
#include "util/u_hash.h"
#include "util/u_memory.h"
#include "tgsi/tgsi_parse.h"
#include "tgsi/tgsi_dump.h"
#include "vc4_context.h"
#include "vc4_qpu.h"
#include "vc4_qir.h"
struct tgsi_to_qir {
struct tgsi_parse_context parser;
struct qcompile *c;
struct qreg *temps;
struct qreg *inputs;
struct qreg *outputs;
struct qreg *uniforms;
struct qreg *consts;
uint32_t num_consts;
struct vc4_shader_state *shader_state;
struct vc4_fs_key *fs_key;
struct vc4_vs_key *vs_key;
uint32_t *uniform_data;
enum quniform_contents *uniform_contents;
uint32_t num_uniforms;
uint32_t num_outputs;
};
struct vc4_key {
struct vc4_shader_state *shader_state;
};
struct vc4_fs_key {
struct vc4_key base;
enum pipe_format color_format;
};
struct vc4_vs_key {
struct vc4_key base;
enum pipe_format attr_formats[8];
};
static struct qreg
add_uniform(struct tgsi_to_qir *trans,
enum quniform_contents contents,
uint32_t data)
{
uint32_t uniform = trans->num_uniforms++;
struct qreg u = { QFILE_UNIF, uniform };
trans->uniform_contents[uniform] = contents;
trans->uniform_data[uniform] = data;
return u;
}
static struct qreg
get_temp_for_uniform(struct tgsi_to_qir *trans, enum quniform_contents contents,
uint32_t data)
{
struct qcompile *c = trans->c;
for (int i = 0; i < trans->num_uniforms; i++) {
if (trans->uniform_contents[i] == contents &&
trans->uniform_data[i] == data)
return trans->uniforms[i];
}
struct qreg u = add_uniform(trans, contents, data);
struct qreg t = qir_MOV(c, u);
trans->uniforms[u.index] = t;
return t;
}
static struct qreg
qir_uniform_ui(struct tgsi_to_qir *trans, uint32_t ui)
{
return get_temp_for_uniform(trans, QUNIFORM_CONSTANT, ui);
}
static struct qreg
qir_uniform_f(struct tgsi_to_qir *trans, float f)
{
return qir_uniform_ui(trans, fui(f));
}
static struct qreg
get_src(struct tgsi_to_qir *trans, struct tgsi_src_register *src, int i)
{
struct qcompile *c = trans->c;
struct qreg r = c->undef;
uint32_t s = i;
switch (i) {
case TGSI_SWIZZLE_X:
s = src->SwizzleX;
break;
case TGSI_SWIZZLE_Y:
s = src->SwizzleY;
break;
case TGSI_SWIZZLE_Z:
s = src->SwizzleZ;
break;
case TGSI_SWIZZLE_W:
s = src->SwizzleW;
break;
default:
abort();
}
assert(!src->Indirect);
switch (src->File) {
case TGSI_FILE_NULL:
return r;
case TGSI_FILE_TEMPORARY:
r = trans->temps[src->Index * 4 + s];
break;
case TGSI_FILE_IMMEDIATE:
r = trans->consts[src->Index * 4 + s];
break;
case TGSI_FILE_CONSTANT:
r = get_temp_for_uniform(trans, QUNIFORM_UNIFORM,
src->Index * 4 + s);
break;
case TGSI_FILE_INPUT:
r = trans->inputs[src->Index * 4 + s];
break;
default:
fprintf(stderr, "unknown src file %d\n", src->File);
abort();
}
if (src->Absolute)
r = qir_FMAXABS(c, r, r);
if (src->Negate)
r = qir_FSUB(c, qir_uniform_f(trans, 0), r);
return r;
};
static void
update_dst(struct tgsi_to_qir *trans, struct tgsi_full_instruction *tgsi_inst,
int i, struct qreg val)
{
struct tgsi_dst_register *tgsi_dst = &tgsi_inst->Dst[0].Register;
assert(!tgsi_dst->Indirect);
switch (tgsi_dst->File) {
case TGSI_FILE_TEMPORARY:
trans->temps[tgsi_dst->Index * 4 + i] = val;
break;
case TGSI_FILE_OUTPUT:
trans->outputs[tgsi_dst->Index * 4 + i] = val;
trans->num_outputs = MAX2(trans->num_outputs,
tgsi_dst->Index * 4 + i + 1);
break;
default:
fprintf(stderr, "unknown dst file %d\n", tgsi_dst->File);
abort();
}
};
static struct qreg
tgsi_to_qir_alu(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
struct qreg dst = qir_get_temp(c);
qir_emit(c, qir_inst4(op, dst,
src[0 * 4 + i],
src[1 * 4 + i],
src[2 * 4 + i],
c->undef));
return dst;
}
static struct qreg
tgsi_to_qir_mad(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
return qir_FADD(c,
qir_FMUL(c,
src[0 * 4 + i],
src[1 * 4 + i]),
src[2 * 4 + i]);
}
static struct qreg
tgsi_to_qir_lit(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
struct qreg x = src[0 * 4 + 0];
struct qreg y = src[0 * 4 + 1];
struct qreg w = src[0 * 4 + 3];
switch (i) {
case 0:
case 3:
return qir_uniform_f(trans, 1.0);
case 1:
return qir_FMAX(c, src[0 * 4 + 0], qir_uniform_f(trans, 0.0));
case 2: {
struct qreg zero = qir_uniform_f(trans, 0.0);
/* XXX: Clamp w to -128..128 */
return qir_CMP(c,
x,
zero,
qir_EXP2(c, qir_FMUL(c,
w,
qir_LOG2(c,
qir_FMAX(c,
y,
zero)))));
}
default:
assert(!"not reached");
return c->undef;
}
}
static struct qreg
tgsi_to_qir_lrp(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
struct qreg src0 = src[0 * 4 + i];
struct qreg src1 = src[1 * 4 + i];
struct qreg src2 = src[2 * 4 + i];
/* LRP is:
* src0 * src1 + (1 - src0) * src2.
* -> src0 * src1 + src2 - src0 * src2
* -> src2 + src0 * (src1 - src2)
*/
return qir_FADD(c, src2, qir_FMUL(c, src0, qir_FSUB(c, src1, src2)));
}
static struct qreg
tgsi_to_qir_pow(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
/* Note that this instruction replicates its result from the x channel
*/
return qir_EXP2(c, qir_FMUL(c,
src[1 * 4 + 0],
qir_LOG2(c, src[0 * 4 + 0])));
}
static struct qreg
tgsi_to_qir_trunc(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
return qir_ITOF(c, qir_FTOI(c, src[0 * 4 + i]));
}
static struct qreg
tgsi_to_qir_dp(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
int num, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
struct qreg sum = qir_FMUL(c, src[0 * 4 + 0], src[1 * 4 + 0]);
for (int j = 1; j < num; j++) {
sum = qir_FADD(c, sum, qir_FMUL(c,
src[0 * 4 + j],
src[1 * 4 + j]));
}
return sum;
}
static struct qreg
tgsi_to_qir_dp2(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
return tgsi_to_qir_dp(trans, tgsi_inst, 2, src, i);
}
static struct qreg
tgsi_to_qir_dp3(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
return tgsi_to_qir_dp(trans, tgsi_inst, 3, src, i);
}
static struct qreg
tgsi_to_qir_dp4(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
return tgsi_to_qir_dp(trans, tgsi_inst, 4, src, i);
}
static struct qreg
tgsi_to_qir_abs(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op, struct qreg *src, int i)
{
struct qcompile *c = trans->c;
struct qreg arg = src[0 * 4 + i];
return qir_FMAXABS(c, arg, arg);
}
static void
emit_tgsi_declaration(struct tgsi_to_qir *trans,
struct tgsi_full_declaration *decl)
{
struct qcompile *c = trans->c;
switch (decl->Declaration.File) {
case TGSI_FILE_INPUT:
for (int i = decl->Range.First * 4;
i < (decl->Range.Last + 1) * 4;
i++) {
if (c->stage == QSTAGE_FRAG) {
struct qreg vary = {
QFILE_VARY,
i
};
trans->inputs[i] =
qir_VARY_ADD_C(c, qir_MOV(c, vary));
} else {
struct qreg dst = qir_get_temp(c);
/* XXX: attribute type/size/count */
qir_emit(c, qir_inst(QOP_VPM_READ,
dst,
c->undef,
c->undef));
trans->inputs[i] = dst;
}
c->num_inputs++;
}
break;
}
}
static void
emit_tgsi_instruction(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst)
{
struct qcompile *c = trans->c;
struct {
enum qop op;
struct qreg (*func)(struct tgsi_to_qir *trans,
struct tgsi_full_instruction *tgsi_inst,
enum qop op,
struct qreg *src, int i);
} op_trans[] = {
[TGSI_OPCODE_MOV] = { QOP_MOV, tgsi_to_qir_alu },
[TGSI_OPCODE_ABS] = { 0, tgsi_to_qir_abs },
[TGSI_OPCODE_MUL] = { QOP_FMUL, tgsi_to_qir_alu },
[TGSI_OPCODE_ADD] = { QOP_FADD, tgsi_to_qir_alu },
[TGSI_OPCODE_SUB] = { QOP_FSUB, tgsi_to_qir_alu },
[TGSI_OPCODE_MIN] = { QOP_FMIN, tgsi_to_qir_alu },
[TGSI_OPCODE_MAX] = { QOP_FMAX, tgsi_to_qir_alu },
[TGSI_OPCODE_RSQ] = { QOP_RSQ, tgsi_to_qir_alu },
[TGSI_OPCODE_SEQ] = { QOP_SEQ, tgsi_to_qir_alu },
[TGSI_OPCODE_SNE] = { QOP_SNE, tgsi_to_qir_alu },
[TGSI_OPCODE_SGE] = { QOP_SGE, tgsi_to_qir_alu },
[TGSI_OPCODE_SLT] = { QOP_SLT, tgsi_to_qir_alu },
[TGSI_OPCODE_CMP] = { QOP_CMP, tgsi_to_qir_alu },
[TGSI_OPCODE_MAD] = { 0, tgsi_to_qir_mad },
[TGSI_OPCODE_DP2] = { 0, tgsi_to_qir_dp2 },
[TGSI_OPCODE_DP3] = { 0, tgsi_to_qir_dp3 },
[TGSI_OPCODE_DP4] = { 0, tgsi_to_qir_dp4 },
[TGSI_OPCODE_RCP] = { QOP_RCP, tgsi_to_qir_alu },
[TGSI_OPCODE_RSQ] = { QOP_RSQ, tgsi_to_qir_alu },
[TGSI_OPCODE_EX2] = { QOP_EXP2, tgsi_to_qir_alu },
[TGSI_OPCODE_LG2] = { QOP_LOG2, tgsi_to_qir_alu },
[TGSI_OPCODE_LIT] = { 0, tgsi_to_qir_lit },
[TGSI_OPCODE_LRP] = { 0, tgsi_to_qir_lrp },
[TGSI_OPCODE_POW] = { 0, tgsi_to_qir_pow },
[TGSI_OPCODE_TRUNC] = { 0, tgsi_to_qir_trunc },
};
static int asdf = 0;
uint32_t tgsi_op = tgsi_inst->Instruction.Opcode;
if (tgsi_op == TGSI_OPCODE_END)
return;
if (tgsi_op > ARRAY_SIZE(op_trans) || !op_trans[tgsi_op].func) {
fprintf(stderr, "unknown tgsi inst: ");
tgsi_dump_instruction(tgsi_inst, asdf++);
fprintf(stderr, "\n");
abort();
}
struct qreg src_regs[12];
for (int s = 0; s < 3; s++) {
for (int i = 0; i < 4; i++) {
src_regs[4 * s + i] =
get_src(trans, &tgsi_inst->Src[s].Register, i);
}
}
for (int i = 0; i < 4; i++) {
if (!(tgsi_inst->Dst[0].Register.WriteMask & (1 << i)))
continue;
struct qreg result;
result = op_trans[tgsi_op].func(trans, tgsi_inst,
op_trans[tgsi_op].op,
src_regs, i);
if (tgsi_inst->Instruction.Saturate) {
float low = (tgsi_inst->Instruction.Saturate ==
TGSI_SAT_MINUS_PLUS_ONE ? -1.0 : 0.0);
result = qir_FMAX(c,
qir_FMIN(c,
result,
qir_uniform_f(trans, 1.0)),
qir_uniform_f(trans, low));
}
update_dst(trans, tgsi_inst, i, result);
}
}
static void
parse_tgsi_immediate(struct tgsi_to_qir *trans, struct tgsi_full_immediate *imm)
{
for (int i = 0; i < 4; i++) {
unsigned n = trans->num_consts++;
trans->consts[n] = qir_uniform_ui(trans, imm->u[i].Uint);
}
}
static void
emit_frag_end(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
struct qreg t = qir_get_temp(c);
const struct util_format_description *format_desc =
util_format_description(trans->fs_key->color_format);
struct qreg swizzled_outputs[4] = {
trans->outputs[format_desc->swizzle[0]],
trans->outputs[format_desc->swizzle[1]],
trans->outputs[format_desc->swizzle[2]],
trans->outputs[format_desc->swizzle[3]],
};
qir_emit(c, qir_inst4(QOP_PACK_COLORS, t,
swizzled_outputs[0],
swizzled_outputs[1],
swizzled_outputs[2],
swizzled_outputs[3]));
qir_emit(c, qir_inst(QOP_TLB_COLOR_WRITE, c->undef,
t, c->undef));
}
static void
emit_scaled_viewport_write(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
struct qreg xyi[2];
for (int i = 0; i < 2; i++) {
struct qreg scale =
add_uniform(trans, QUNIFORM_VIEWPORT_X_SCALE + i, 0);
xyi[i] = qir_FTOI(c, qir_FMUL(c, trans->outputs[i], scale));
}
qir_VPM_WRITE(c, qir_PACK_SCALED(c, xyi[0], xyi[1]));
}
static void
emit_zs_write(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
/* XXX: rescale */
qir_VPM_WRITE(c, trans->outputs[2]);
}
static void
emit_1_wc_write(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
/* XXX: RCP */
qir_VPM_WRITE(c, trans->outputs[3]);
}
static void
emit_vert_end(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
emit_scaled_viewport_write(trans);
emit_zs_write(trans);
emit_1_wc_write(trans);
for (int i = 4; i < trans->num_outputs; i++) {
qir_VPM_WRITE(c, trans->outputs[i]);
}
}
static void
emit_coord_end(struct tgsi_to_qir *trans)
{
struct qcompile *c = trans->c;
for (int i = 0; i < 4; i++)
qir_VPM_WRITE(c, trans->outputs[i]);
emit_scaled_viewport_write(trans);
emit_zs_write(trans);
emit_1_wc_write(trans);
}
static struct tgsi_to_qir *
vc4_shader_tgsi_to_qir(struct vc4_compiled_shader *shader, enum qstage stage,
struct vc4_key *key)
{
struct tgsi_to_qir *trans = CALLOC_STRUCT(tgsi_to_qir);
struct qcompile *c;
int ret;
c = qir_compile_init();
c->stage = stage;
memset(trans, 0, sizeof(*trans));
/* XXX sizing */
trans->temps = calloc(sizeof(struct qreg), 1024);
trans->inputs = calloc(sizeof(struct qreg), 8 * 4);
trans->outputs = calloc(sizeof(struct qreg), 1024);
trans->uniforms = calloc(sizeof(struct qreg), 1024);
trans->consts = calloc(sizeof(struct qreg), 1024);
trans->uniform_data = calloc(sizeof(uint32_t), 1024);
trans->uniform_contents = calloc(sizeof(enum quniform_contents), 1024);
trans->shader_state = key->shader_state;
trans->c = c;
ret = tgsi_parse_init(&trans->parser, trans->shader_state->base.tokens);
assert(ret == TGSI_PARSE_OK);
if (vc4_debug & VC4_DEBUG_TGSI) {
fprintf(stderr, "TGSI:\n");
tgsi_dump(trans->shader_state->base.tokens, 0);
}
switch (stage) {
case QSTAGE_FRAG:
trans->fs_key = (struct vc4_fs_key *)key;
break;
case QSTAGE_VERT:
trans->vs_key = (struct vc4_vs_key *)key;
break;
case QSTAGE_COORD:
trans->vs_key = (struct vc4_vs_key *)key;
break;
}
while (!tgsi_parse_end_of_tokens(&trans->parser)) {
tgsi_parse_token(&trans->parser);
switch (trans->parser.FullToken.Token.Type) {
case TGSI_TOKEN_TYPE_DECLARATION:
emit_tgsi_declaration(trans,
&trans->parser.FullToken.FullDeclaration);
break;
case TGSI_TOKEN_TYPE_INSTRUCTION:
emit_tgsi_instruction(trans,
&trans->parser.FullToken.FullInstruction);
break;
case TGSI_TOKEN_TYPE_IMMEDIATE:
parse_tgsi_immediate(trans,
&trans->parser.FullToken.FullImmediate);
break;
}
}
switch (stage) {
case QSTAGE_FRAG:
emit_frag_end(trans);
break;
case QSTAGE_VERT:
emit_vert_end(trans);
break;
case QSTAGE_COORD:
emit_coord_end(trans);
break;
}
tgsi_parse_free(&trans->parser);
free(trans->temps);
qir_optimize(c);
if (vc4_debug & VC4_DEBUG_QIR) {
fprintf(stderr, "QIR:\n");
qir_dump(c);
}
vc4_generate_code(c);
if (vc4_debug & VC4_DEBUG_SHADERDB) {
fprintf(stderr, "SHADER-DB: %s: %d instructions\n",
qir_get_stage_name(c->stage), c->qpu_inst_count);
fprintf(stderr, "SHADER-DB: %s: %d uniforms\n",
qir_get_stage_name(c->stage), trans->num_uniforms);
}
return trans;
}
static void *
vc4_shader_state_create(struct pipe_context *pctx,
const struct pipe_shader_state *cso)
{
struct vc4_shader_state *so = CALLOC_STRUCT(vc4_shader_state);
if (!so)
return NULL;
so->base.tokens = tgsi_dup_tokens(cso->tokens);
return so;
}
static void
copy_uniform_state_to_shader(struct vc4_compiled_shader *shader,
int shader_index,
struct tgsi_to_qir *trans)
{
int count = trans->num_uniforms;
struct vc4_shader_uniform_info *uinfo = &shader->uniforms[shader_index];
uinfo->count = count;
uinfo->data = malloc(count * sizeof(*uinfo->data));
memcpy(uinfo->data, trans->uniform_data,
count * sizeof(*uinfo->data));
uinfo->contents = malloc(count * sizeof(*uinfo->contents));
memcpy(uinfo->contents, trans->uniform_contents,
count * sizeof(*uinfo->contents));
}
static void
vc4_fs_compile(struct vc4_context *vc4, struct vc4_compiled_shader *shader,
struct vc4_fs_key *key)
{
struct tgsi_to_qir *trans = vc4_shader_tgsi_to_qir(shader, QSTAGE_FRAG,
&key->base);
shader->num_inputs = trans->c->num_inputs;
copy_uniform_state_to_shader(shader, 0, trans);
shader->bo = vc4_bo_alloc_mem(vc4->screen, trans->c->qpu_insts,
trans->c->qpu_inst_count * sizeof(uint64_t),
"fs_code");
qir_compile_destroy(trans->c);
free(trans);
}
static void
vc4_vs_compile(struct vc4_context *vc4, struct vc4_compiled_shader *shader,
struct vc4_vs_key *key)
{
struct tgsi_to_qir *vs_trans = vc4_shader_tgsi_to_qir(shader,
QSTAGE_VERT,
&key->base);
copy_uniform_state_to_shader(shader, 0, vs_trans);
struct tgsi_to_qir *cs_trans = vc4_shader_tgsi_to_qir(shader,
QSTAGE_COORD,
&key->base);
copy_uniform_state_to_shader(shader, 1, cs_trans);
uint32_t vs_size = vs_trans->c->qpu_inst_count * sizeof(uint64_t);
uint32_t cs_size = cs_trans->c->qpu_inst_count * sizeof(uint64_t);
shader->coord_shader_offset = vs_size; /* XXX: alignment? */
shader->bo = vc4_bo_alloc(vc4->screen,
shader->coord_shader_offset + cs_size,
"vs_code");
void *map = vc4_bo_map(shader->bo);
memcpy(map, vs_trans->c->qpu_insts, vs_size);
memcpy(map + shader->coord_shader_offset,
cs_trans->c->qpu_insts, cs_size);
qir_compile_destroy(vs_trans->c);
qir_compile_destroy(cs_trans->c);
}
static void
vc4_update_compiled_fs(struct vc4_context *vc4)
{
struct vc4_fs_key local_key;
struct vc4_fs_key *key = &local_key;
memset(key, 0, sizeof(*key));
key->base.shader_state = vc4->prog.bind_fs;
if (vc4->framebuffer.cbufs[0])
key->color_format = vc4->framebuffer.cbufs[0]->format;
vc4->prog.fs = util_hash_table_get(vc4->fs_cache, key);
if (vc4->prog.fs)
return;
key = malloc(sizeof(*key));
memcpy(key, &local_key, sizeof(*key));
struct vc4_compiled_shader *shader = CALLOC_STRUCT(vc4_compiled_shader);
vc4_fs_compile(vc4, shader, key);
util_hash_table_set(vc4->fs_cache, key, shader);
vc4->prog.fs = shader;
}
static void
vc4_update_compiled_vs(struct vc4_context *vc4)
{
struct vc4_vs_key local_key;
struct vc4_vs_key *key = &local_key;
memset(key, 0, sizeof(*key));
key->base.shader_state = vc4->prog.bind_vs;
vc4->prog.vs = util_hash_table_get(vc4->vs_cache, key);
if (vc4->prog.vs)
return;
key = malloc(sizeof(*key));
memcpy(key, &local_key, sizeof(*key));
struct vc4_compiled_shader *shader = CALLOC_STRUCT(vc4_compiled_shader);
vc4_vs_compile(vc4, shader, key);
util_hash_table_set(vc4->vs_cache, key, shader);
vc4->prog.vs = shader;
}
void
vc4_update_compiled_shaders(struct vc4_context *vc4)
{
vc4_update_compiled_fs(vc4);
vc4_update_compiled_vs(vc4);
}
static unsigned
fs_cache_hash(void *key)
{
return util_hash_crc32(key, sizeof(struct vc4_fs_key));
}
static unsigned
vs_cache_hash(void *key)
{
return util_hash_crc32(key, sizeof(struct vc4_vs_key));
}
static int
fs_cache_compare(void *key1, void *key2)
{
return memcmp(key1, key2, sizeof(struct vc4_fs_key));
}
static int
vs_cache_compare(void *key1, void *key2)
{
return memcmp(key1, key2, sizeof(struct vc4_vs_key));
}
struct delete_state {
struct vc4_context *vc4;
struct vc4_shader_state *shader_state;
};
static enum pipe_error
fs_delete_from_cache(void *in_key, void *in_value, void *data)
{
struct delete_state *del = data;
struct vc4_fs_key *key = in_key;
struct vc4_compiled_shader *shader = in_value;
if (key->base.shader_state == data) {
util_hash_table_remove(del->vc4->fs_cache, key);
vc4_bo_unreference(&shader->bo);
free(shader);
}
return 0;
}
static enum pipe_error
vs_delete_from_cache(void *in_key, void *in_value, void *data)
{
struct delete_state *del = data;
struct vc4_vs_key *key = in_key;
struct vc4_compiled_shader *shader = in_value;
if (key->base.shader_state == data) {
util_hash_table_remove(del->vc4->vs_cache, key);
vc4_bo_unreference(&shader->bo);
free(shader);
}
return 0;
}
static void
vc4_shader_state_delete(struct pipe_context *pctx, void *hwcso)
{
struct vc4_context *vc4 = vc4_context(pctx);
struct vc4_shader_state *so = hwcso;
struct delete_state del;
del.vc4 = vc4;
del.shader_state = so;
util_hash_table_foreach(vc4->fs_cache, fs_delete_from_cache, &del);
util_hash_table_foreach(vc4->vs_cache, vs_delete_from_cache, &del);
free((void *)so->base.tokens);
free(so);
}
void
vc4_get_uniform_bo(struct vc4_context *vc4, struct vc4_compiled_shader *shader,
struct vc4_constbuf_stateobj *cb,
int shader_index, struct vc4_bo **out_bo,
uint32_t *out_offset)
{
struct vc4_shader_uniform_info *uinfo = &shader->uniforms[shader_index];
struct vc4_bo *ubo = vc4_bo_alloc(vc4->screen,
MAX2(1, uinfo->count * 4), "ubo");
uint32_t *map = vc4_bo_map(ubo);
for (int i = 0; i < uinfo->count; i++) {
switch (uinfo->contents[i]) {
case QUNIFORM_CONSTANT:
map[i] = uinfo->data[i];
break;
case QUNIFORM_UNIFORM:
map[i] = ((uint32_t *)cb->cb[0].user_buffer)[uinfo->data[i]];
break;
case QUNIFORM_VIEWPORT_X_SCALE:
map[i] = fui(vc4->framebuffer.width * 16.0f / 2.0f);
break;
case QUNIFORM_VIEWPORT_Y_SCALE:
map[i] = fui(vc4->framebuffer.height * -16.0f / 2.0f);
break;
}
#if 0
fprintf(stderr, "%p/%d: %d: 0x%08x (%f)\n",
shader, shader_index, i, map[i], uif(map[i]));
#endif
}
*out_bo = ubo;
*out_offset = 0;
}
static void
vc4_fp_state_bind(struct pipe_context *pctx, void *hwcso)
{
struct vc4_context *vc4 = vc4_context(pctx);
vc4->prog.bind_fs = hwcso;
vc4->prog.dirty |= VC4_SHADER_DIRTY_FP;
vc4->dirty |= VC4_DIRTY_PROG;
}
static void
vc4_vp_state_bind(struct pipe_context *pctx, void *hwcso)
{
struct vc4_context *vc4 = vc4_context(pctx);
vc4->prog.bind_vs = hwcso;
vc4->prog.dirty |= VC4_SHADER_DIRTY_VP;
vc4->dirty |= VC4_DIRTY_PROG;
}
void
vc4_program_init(struct pipe_context *pctx)
{
struct vc4_context *vc4 = vc4_context(pctx);
pctx->create_vs_state = vc4_shader_state_create;
pctx->delete_vs_state = vc4_shader_state_delete;
pctx->create_fs_state = vc4_shader_state_create;
pctx->delete_fs_state = vc4_shader_state_delete;
pctx->bind_fs_state = vc4_fp_state_bind;
pctx->bind_vs_state = vc4_vp_state_bind;
vc4->fs_cache = util_hash_table_create(fs_cache_hash, fs_cache_compare);
vc4->vs_cache = util_hash_table_create(vs_cache_hash, vs_cache_compare);
}
|