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
|
/*
* Copyright © 2014 Broadcom
* Copyright (C) 2012 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 "pipe/p_defines.h"
#include "util/u_blit.h"
#include "util/u_memory.h"
#include "util/u_format.h"
#include "util/u_inlines.h"
#include "util/u_surface.h"
#include "util/u_transfer_helper.h"
#include "util/u_upload_mgr.h"
#include "drm_fourcc.h"
#include "vc4_drm.h"
#include "vc4_screen.h"
#include "vc4_context.h"
#include "vc4_resource.h"
#include "vc4_tiling.h"
static bool
vc4_resource_bo_alloc(struct vc4_resource *rsc)
{
struct pipe_resource *prsc = &rsc->base;
struct pipe_screen *pscreen = prsc->screen;
struct vc4_bo *bo;
if (vc4_debug & VC4_DEBUG_SURFACE) {
fprintf(stderr, "alloc %p: size %d + offset %d -> %d\n",
rsc,
rsc->slices[0].size,
rsc->slices[0].offset,
rsc->slices[0].offset +
rsc->slices[0].size +
rsc->cube_map_stride * (prsc->array_size - 1));
}
bo = vc4_bo_alloc(vc4_screen(pscreen),
rsc->slices[0].offset +
rsc->slices[0].size +
rsc->cube_map_stride * (prsc->array_size - 1),
"resource");
if (bo) {
vc4_bo_unreference(&rsc->bo);
rsc->bo = bo;
return true;
} else {
return false;
}
}
static void
vc4_resource_transfer_unmap(struct pipe_context *pctx,
struct pipe_transfer *ptrans)
{
struct vc4_context *vc4 = vc4_context(pctx);
struct vc4_transfer *trans = vc4_transfer(ptrans);
if (trans->map) {
struct vc4_resource *rsc = vc4_resource(ptrans->resource);
struct vc4_resource_slice *slice = &rsc->slices[ptrans->level];
if (ptrans->usage & PIPE_TRANSFER_WRITE) {
vc4_store_tiled_image(rsc->bo->map + slice->offset +
ptrans->box.z * rsc->cube_map_stride,
slice->stride,
trans->map, ptrans->stride,
slice->tiling, rsc->cpp,
&ptrans->box);
}
free(trans->map);
}
pipe_resource_reference(&ptrans->resource, NULL);
slab_free(&vc4->transfer_pool, ptrans);
}
static void *
vc4_resource_transfer_map(struct pipe_context *pctx,
struct pipe_resource *prsc,
unsigned level, unsigned usage,
const struct pipe_box *box,
struct pipe_transfer **pptrans)
{
struct vc4_context *vc4 = vc4_context(pctx);
struct vc4_resource *rsc = vc4_resource(prsc);
struct vc4_transfer *trans;
struct pipe_transfer *ptrans;
enum pipe_format format = prsc->format;
char *buf;
/* Upgrade DISCARD_RANGE to WHOLE_RESOURCE if the whole resource is
* being mapped.
*/
if ((usage & PIPE_TRANSFER_DISCARD_RANGE) &&
!(usage & PIPE_TRANSFER_UNSYNCHRONIZED) &&
!(prsc->flags & PIPE_RESOURCE_FLAG_MAP_PERSISTENT) &&
prsc->last_level == 0 &&
prsc->width0 == box->width &&
prsc->height0 == box->height &&
prsc->depth0 == box->depth &&
prsc->array_size == 1 &&
rsc->bo->private) {
usage |= PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE;
}
if (usage & PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE) {
if (vc4_resource_bo_alloc(rsc)) {
/* If it might be bound as one of our vertex buffers,
* make sure we re-emit vertex buffer state.
*/
if (prsc->bind & PIPE_BIND_VERTEX_BUFFER)
vc4->dirty |= VC4_DIRTY_VTXBUF;
} else {
/* If we failed to reallocate, flush users so that we
* don't violate any syncing requirements.
*/
vc4_flush_jobs_reading_resource(vc4, prsc);
}
} else if (!(usage & PIPE_TRANSFER_UNSYNCHRONIZED)) {
/* If we're writing and the buffer is being used by the CL, we
* have to flush the CL first. If we're only reading, we need
* to flush if the CL has written our buffer.
*/
if (usage & PIPE_TRANSFER_WRITE)
vc4_flush_jobs_reading_resource(vc4, prsc);
else
vc4_flush_jobs_writing_resource(vc4, prsc);
}
if (usage & PIPE_TRANSFER_WRITE) {
rsc->writes++;
rsc->initialized_buffers = ~0;
}
trans = slab_alloc(&vc4->transfer_pool);
if (!trans)
return NULL;
/* XXX: Handle DONTBLOCK, DISCARD_RANGE, PERSISTENT, COHERENT. */
/* slab_alloc_st() doesn't zero: */
memset(trans, 0, sizeof(*trans));
ptrans = &trans->base;
pipe_resource_reference(&ptrans->resource, prsc);
ptrans->level = level;
ptrans->usage = usage;
ptrans->box = *box;
if (usage & PIPE_TRANSFER_UNSYNCHRONIZED)
buf = vc4_bo_map_unsynchronized(rsc->bo);
else
buf = vc4_bo_map(rsc->bo);
if (!buf) {
fprintf(stderr, "Failed to map bo\n");
goto fail;
}
*pptrans = ptrans;
struct vc4_resource_slice *slice = &rsc->slices[level];
if (rsc->tiled) {
/* No direct mappings of tiled, since we need to manually
* tile/untile.
*/
if (usage & PIPE_TRANSFER_MAP_DIRECTLY)
return NULL;
if (format == PIPE_FORMAT_ETC1_RGB8) {
/* ETC1 is arranged as 64-bit blocks, where each block
* is 4x4 pixels. Texture tiling operates on the
* 64-bit block the way it would an uncompressed
* pixels.
*/
assert(!(ptrans->box.x & 3));
assert(!(ptrans->box.y & 3));
ptrans->box.x >>= 2;
ptrans->box.y >>= 2;
ptrans->box.width = (ptrans->box.width + 3) >> 2;
ptrans->box.height = (ptrans->box.height + 3) >> 2;
}
ptrans->stride = ptrans->box.width * rsc->cpp;
ptrans->layer_stride = ptrans->stride * ptrans->box.height;
trans->map = malloc(ptrans->layer_stride * ptrans->box.depth);
if (usage & PIPE_TRANSFER_READ) {
vc4_load_tiled_image(trans->map, ptrans->stride,
buf + slice->offset +
ptrans->box.z * rsc->cube_map_stride,
slice->stride,
slice->tiling, rsc->cpp,
&ptrans->box);
}
return trans->map;
} else {
ptrans->stride = slice->stride;
ptrans->layer_stride = ptrans->stride;
return buf + slice->offset +
ptrans->box.y / util_format_get_blockheight(format) * ptrans->stride +
ptrans->box.x / util_format_get_blockwidth(format) * rsc->cpp +
ptrans->box.z * rsc->cube_map_stride;
}
fail:
vc4_resource_transfer_unmap(pctx, ptrans);
return NULL;
}
static void
vc4_texture_subdata(struct pipe_context *pctx,
struct pipe_resource *prsc,
unsigned level,
unsigned usage,
const struct pipe_box *box,
const void *data,
unsigned stride,
unsigned layer_stride)
{
struct vc4_resource *rsc = vc4_resource(prsc);
struct vc4_resource_slice *slice = &rsc->slices[level];
/* For a direct mapping, we can just take the u_transfer path. */
if (!rsc->tiled ||
box->depth != 1 ||
(usage & PIPE_TRANSFER_DISCARD_WHOLE_RESOURCE)) {
return u_default_texture_subdata(pctx, prsc, level, usage, box,
data, stride, layer_stride);
}
/* Otherwise, map and store the texture data directly into the tiled
* texture.
*/
void *buf;
if (usage & PIPE_TRANSFER_UNSYNCHRONIZED)
buf = vc4_bo_map_unsynchronized(rsc->bo);
else
buf = vc4_bo_map(rsc->bo);
vc4_store_tiled_image(buf + slice->offset +
box->z * rsc->cube_map_stride,
slice->stride,
(void *)data, stride,
slice->tiling, rsc->cpp,
box);
}
static void
vc4_resource_destroy(struct pipe_screen *pscreen,
struct pipe_resource *prsc)
{
struct vc4_screen *screen = vc4_screen(pscreen);
struct vc4_resource *rsc = vc4_resource(prsc);
vc4_bo_unreference(&rsc->bo);
if (rsc->scanout)
renderonly_scanout_destroy(rsc->scanout, screen->ro);
free(rsc);
}
static boolean
vc4_resource_get_handle(struct pipe_screen *pscreen,
struct pipe_context *pctx,
struct pipe_resource *prsc,
struct winsys_handle *whandle,
unsigned usage)
{
struct vc4_screen *screen = vc4_screen(pscreen);
struct vc4_resource *rsc = vc4_resource(prsc);
whandle->stride = rsc->slices[0].stride;
whandle->offset = 0;
/* If we're passing some reference to our BO out to some other part of
* the system, then we can't do any optimizations about only us being
* the ones seeing it (like BO caching or shadow update avoidance).
*/
rsc->bo->private = false;
if (rsc->tiled)
whandle->modifier = DRM_FORMAT_MOD_BROADCOM_VC4_T_TILED;
else
whandle->modifier = DRM_FORMAT_MOD_LINEAR;
switch (whandle->type) {
case WINSYS_HANDLE_TYPE_SHARED:
if (screen->ro) {
/* This could probably be supported, assuming that a
* control node was used for pl111.
*/
fprintf(stderr, "flink unsupported with pl111\n");
return FALSE;
}
return vc4_bo_flink(rsc->bo, &whandle->handle);
case WINSYS_HANDLE_TYPE_KMS:
if (screen->ro && renderonly_get_handle(rsc->scanout, whandle))
return TRUE;
whandle->handle = rsc->bo->handle;
return TRUE;
case WINSYS_HANDLE_TYPE_FD:
/* FDs are cross-device, so we can export directly from vc4.
*/
whandle->handle = vc4_bo_get_dmabuf(rsc->bo);
return whandle->handle != -1;
}
return FALSE;
}
static void
vc4_setup_slices(struct vc4_resource *rsc, const char *caller)
{
struct pipe_resource *prsc = &rsc->base;
uint32_t width = prsc->width0;
uint32_t height = prsc->height0;
if (prsc->format == PIPE_FORMAT_ETC1_RGB8) {
width = (width + 3) >> 2;
height = (height + 3) >> 2;
}
uint32_t pot_width = util_next_power_of_two(width);
uint32_t pot_height = util_next_power_of_two(height);
uint32_t offset = 0;
uint32_t utile_w = vc4_utile_width(rsc->cpp);
uint32_t utile_h = vc4_utile_height(rsc->cpp);
for (int i = prsc->last_level; i >= 0; i--) {
struct vc4_resource_slice *slice = &rsc->slices[i];
uint32_t level_width, level_height;
if (i == 0) {
level_width = width;
level_height = height;
} else {
level_width = u_minify(pot_width, i);
level_height = u_minify(pot_height, i);
}
if (!rsc->tiled) {
slice->tiling = VC4_TILING_FORMAT_LINEAR;
if (prsc->nr_samples > 1) {
/* MSAA (4x) surfaces are stored as raw tile buffer contents. */
level_width = align(level_width, 32);
level_height = align(level_height, 32);
} else {
level_width = align(level_width, utile_w);
}
} else {
if (vc4_size_is_lt(level_width, level_height,
rsc->cpp)) {
slice->tiling = VC4_TILING_FORMAT_LT;
level_width = align(level_width, utile_w);
level_height = align(level_height, utile_h);
} else {
slice->tiling = VC4_TILING_FORMAT_T;
level_width = align(level_width,
4 * 2 * utile_w);
level_height = align(level_height,
4 * 2 * utile_h);
}
}
slice->offset = offset;
slice->stride = (level_width * rsc->cpp *
MAX2(prsc->nr_samples, 1));
slice->size = level_height * slice->stride;
offset += slice->size;
if (vc4_debug & VC4_DEBUG_SURFACE) {
static const char tiling_chars[] = {
[VC4_TILING_FORMAT_LINEAR] = 'R',
[VC4_TILING_FORMAT_LT] = 'L',
[VC4_TILING_FORMAT_T] = 'T'
};
fprintf(stderr,
"rsc %s %p (format %s: vc4 %d), %dx%d: "
"level %d (%c) -> %dx%d, stride %d@0x%08x\n",
caller, rsc,
util_format_short_name(prsc->format),
rsc->vc4_format,
prsc->width0, prsc->height0,
i, tiling_chars[slice->tiling],
level_width, level_height,
slice->stride, slice->offset);
}
}
/* The texture base pointer that has to point to level 0 doesn't have
* intra-page bits, so we have to align it, and thus shift up all the
* smaller slices.
*/
uint32_t page_align_offset = (align(rsc->slices[0].offset, 4096) -
rsc->slices[0].offset);
if (page_align_offset) {
for (int i = 0; i <= prsc->last_level; i++)
rsc->slices[i].offset += page_align_offset;
}
/* Cube map faces appear as whole miptrees at a page-aligned offset
* from the first face's miptree.
*/
if (prsc->target == PIPE_TEXTURE_CUBE) {
rsc->cube_map_stride = align(rsc->slices[0].offset +
rsc->slices[0].size, 4096);
}
}
static struct vc4_resource *
vc4_resource_setup(struct pipe_screen *pscreen,
const struct pipe_resource *tmpl)
{
struct vc4_resource *rsc = CALLOC_STRUCT(vc4_resource);
if (!rsc)
return NULL;
struct pipe_resource *prsc = &rsc->base;
*prsc = *tmpl;
pipe_reference_init(&prsc->reference, 1);
prsc->screen = pscreen;
if (prsc->nr_samples <= 1)
rsc->cpp = util_format_get_blocksize(tmpl->format);
else
rsc->cpp = sizeof(uint32_t);
assert(rsc->cpp);
return rsc;
}
static enum vc4_texture_data_type
get_resource_texture_format(struct pipe_resource *prsc)
{
struct vc4_resource *rsc = vc4_resource(prsc);
uint8_t format = vc4_get_tex_format(prsc->format);
if (!rsc->tiled) {
if (prsc->nr_samples > 1) {
return ~0;
} else {
if (format == VC4_TEXTURE_TYPE_RGBA8888)
return VC4_TEXTURE_TYPE_RGBA32R;
else
return ~0;
}
}
return format;
}
static bool
find_modifier(uint64_t needle, const uint64_t *haystack, int count)
{
int i;
for (i = 0; i < count; i++) {
if (haystack[i] == needle)
return true;
}
return false;
}
static struct pipe_resource *
vc4_resource_create_with_modifiers(struct pipe_screen *pscreen,
const struct pipe_resource *tmpl,
const uint64_t *modifiers,
int count)
{
struct vc4_screen *screen = vc4_screen(pscreen);
struct vc4_resource *rsc = vc4_resource_setup(pscreen, tmpl);
struct pipe_resource *prsc = &rsc->base;
bool linear_ok = find_modifier(DRM_FORMAT_MOD_LINEAR, modifiers, count);
/* Use a tiled layout if we can, for better 3D performance. */
bool should_tile = true;
/* VBOs/PBOs are untiled (and 1 height). */
if (tmpl->target == PIPE_BUFFER)
should_tile = false;
/* MSAA buffers are linear. */
if (tmpl->nr_samples > 1)
should_tile = false;
/* No tiling when we're sharing with another device (pl111). */
if (screen->ro && (tmpl->bind & PIPE_BIND_SCANOUT))
should_tile = false;
/* Cursors are always linear, and the user can request linear as well.
*/
if (tmpl->bind & (PIPE_BIND_LINEAR | PIPE_BIND_CURSOR))
should_tile = false;
/* No shared objects with LT format -- the kernel only has T-format
* metadata. LT objects are small enough it's not worth the trouble to
* give them metadata to tile.
*/
if ((tmpl->bind & (PIPE_BIND_SHARED | PIPE_BIND_SCANOUT)) &&
vc4_size_is_lt(prsc->width0, prsc->height0, rsc->cpp))
should_tile = false;
/* If we're sharing or scanning out, we need the ioctl present to
* inform the kernel or the other side.
*/
if ((tmpl->bind & (PIPE_BIND_SHARED |
PIPE_BIND_SCANOUT)) && !screen->has_tiling_ioctl)
should_tile = false;
/* No user-specified modifier; determine our own. */
if (count == 1 && modifiers[0] == DRM_FORMAT_MOD_INVALID) {
linear_ok = true;
rsc->tiled = should_tile;
} else if (should_tile &&
find_modifier(DRM_FORMAT_MOD_BROADCOM_VC4_T_TILED,
modifiers, count)) {
rsc->tiled = true;
} else if (linear_ok) {
rsc->tiled = false;
} else {
fprintf(stderr, "Unsupported modifier requested\n");
return NULL;
}
if (tmpl->target != PIPE_BUFFER)
rsc->vc4_format = get_resource_texture_format(prsc);
vc4_setup_slices(rsc, "create");
if (!vc4_resource_bo_alloc(rsc))
goto fail;
if (screen->has_tiling_ioctl) {
uint64_t modifier;
if (rsc->tiled)
modifier = DRM_FORMAT_MOD_BROADCOM_VC4_T_TILED;
else
modifier = DRM_FORMAT_MOD_LINEAR;
struct drm_vc4_set_tiling set_tiling = {
.handle = rsc->bo->handle,
.modifier = modifier,
};
int ret = vc4_ioctl(screen->fd, DRM_IOCTL_VC4_SET_TILING,
&set_tiling);
if (ret != 0)
goto fail;
}
/* Set up the "scanout resource" (the dmabuf export of our buffer to
* the KMS handle) if the buffer might ever have
* resource_get_handle(WINSYS_HANDLE_TYPE_KMS) called on it.
* create_with_modifiers() doesn't give us usage flags, so we have to
* assume that all calls with modifiers are scanout-possible.
*/
if (screen->ro &&
((tmpl->bind & PIPE_BIND_SCANOUT) ||
!(count == 1 && modifiers[0] == DRM_FORMAT_MOD_INVALID))) {
rsc->scanout =
renderonly_scanout_for_resource(prsc, screen->ro, NULL);
if (!rsc->scanout)
goto fail;
}
vc4_bo_label(screen, rsc->bo, "%sresource %dx%d@%d/%d",
(tmpl->bind & PIPE_BIND_SCANOUT) ? "scanout " : "",
tmpl->width0, tmpl->height0,
rsc->cpp * 8, prsc->last_level);
return prsc;
fail:
vc4_resource_destroy(pscreen, prsc);
return NULL;
}
struct pipe_resource *
vc4_resource_create(struct pipe_screen *pscreen,
const struct pipe_resource *tmpl)
{
const uint64_t mod = DRM_FORMAT_MOD_INVALID;
return vc4_resource_create_with_modifiers(pscreen, tmpl, &mod, 1);
}
static struct pipe_resource *
vc4_resource_from_handle(struct pipe_screen *pscreen,
const struct pipe_resource *tmpl,
struct winsys_handle *whandle,
unsigned usage)
{
struct vc4_screen *screen = vc4_screen(pscreen);
struct vc4_resource *rsc = vc4_resource_setup(pscreen, tmpl);
struct pipe_resource *prsc = &rsc->base;
struct vc4_resource_slice *slice = &rsc->slices[0];
if (!rsc)
return NULL;
switch (whandle->type) {
case WINSYS_HANDLE_TYPE_SHARED:
rsc->bo = vc4_bo_open_name(screen, whandle->handle);
break;
case WINSYS_HANDLE_TYPE_FD:
rsc->bo = vc4_bo_open_dmabuf(screen, whandle->handle);
break;
default:
fprintf(stderr,
"Attempt to import unsupported handle type %d\n",
whandle->type);
}
if (!rsc->bo)
goto fail;
struct drm_vc4_get_tiling get_tiling = {
.handle = rsc->bo->handle,
};
int ret = vc4_ioctl(screen->fd, DRM_IOCTL_VC4_GET_TILING, &get_tiling);
if (ret != 0) {
whandle->modifier = DRM_FORMAT_MOD_LINEAR;
} else if (whandle->modifier == DRM_FORMAT_MOD_INVALID) {
whandle->modifier = get_tiling.modifier;
} else if (whandle->modifier != get_tiling.modifier) {
fprintf(stderr,
"Modifier 0x%llx vs. tiling (0x%llx) mismatch\n",
(long long)whandle->modifier, get_tiling.modifier);
goto fail;
}
switch (whandle->modifier) {
case DRM_FORMAT_MOD_LINEAR:
rsc->tiled = false;
break;
case DRM_FORMAT_MOD_BROADCOM_VC4_T_TILED:
rsc->tiled = true;
break;
default:
fprintf(stderr,
"Attempt to import unsupported modifier 0x%llx\n",
(long long)whandle->modifier);
goto fail;
}
rsc->vc4_format = get_resource_texture_format(prsc);
vc4_setup_slices(rsc, "import");
if (whandle->offset != 0) {
if (rsc->tiled) {
fprintf(stderr,
"Attempt to import unsupported "
"winsys offset %u\n",
whandle->offset);
goto fail;
}
rsc->slices[0].offset += whandle->offset;
if (rsc->slices[0].offset + rsc->slices[0].size >
rsc->bo->size) {
fprintf(stderr, "Attempt to import "
"with overflowing offset (%d + %d > %d)\n",
whandle->offset,
rsc->slices[0].size,
rsc->bo->size);
goto fail;
}
}
if (screen->ro) {
/* Make sure that renderonly has a handle to our buffer in the
* display's fd, so that a later renderonly_get_handle()
* returns correct handles or GEM names.
*/
rsc->scanout =
renderonly_create_gpu_import_for_resource(prsc,
screen->ro,
NULL);
if (!rsc->scanout)
goto fail;
}
if (rsc->tiled && whandle->stride != slice->stride) {
static bool warned = false;
if (!warned) {
warned = true;
fprintf(stderr,
"Attempting to import %dx%d %s with "
"unsupported stride %d instead of %d\n",
prsc->width0, prsc->height0,
util_format_short_name(prsc->format),
whandle->stride,
slice->stride);
}
goto fail;
} else if (!rsc->tiled) {
slice->stride = whandle->stride;
}
return prsc;
fail:
vc4_resource_destroy(pscreen, prsc);
return NULL;
}
static struct pipe_surface *
vc4_create_surface(struct pipe_context *pctx,
struct pipe_resource *ptex,
const struct pipe_surface *surf_tmpl)
{
struct vc4_surface *surface = CALLOC_STRUCT(vc4_surface);
struct vc4_resource *rsc = vc4_resource(ptex);
if (!surface)
return NULL;
assert(surf_tmpl->u.tex.first_layer == surf_tmpl->u.tex.last_layer);
struct pipe_surface *psurf = &surface->base;
unsigned level = surf_tmpl->u.tex.level;
pipe_reference_init(&psurf->reference, 1);
pipe_resource_reference(&psurf->texture, ptex);
psurf->context = pctx;
psurf->format = surf_tmpl->format;
psurf->width = u_minify(ptex->width0, level);
psurf->height = u_minify(ptex->height0, level);
psurf->u.tex.level = level;
psurf->u.tex.first_layer = surf_tmpl->u.tex.first_layer;
psurf->u.tex.last_layer = surf_tmpl->u.tex.last_layer;
surface->offset = (rsc->slices[level].offset +
psurf->u.tex.first_layer * rsc->cube_map_stride);
surface->tiling = rsc->slices[level].tiling;
return &surface->base;
}
static void
vc4_surface_destroy(struct pipe_context *pctx, struct pipe_surface *psurf)
{
pipe_resource_reference(&psurf->texture, NULL);
FREE(psurf);
}
static void
vc4_dump_surface_non_msaa(struct pipe_surface *psurf)
{
struct pipe_resource *prsc = psurf->texture;
struct vc4_resource *rsc = vc4_resource(prsc);
uint32_t *map = vc4_bo_map(rsc->bo);
uint32_t stride = rsc->slices[0].stride / 4;
uint32_t width = psurf->width;
uint32_t height = psurf->height;
uint32_t chunk_w = width / 79;
uint32_t chunk_h = height / 40;
uint32_t found_colors[10];
uint32_t num_found_colors = 0;
if (rsc->vc4_format != VC4_TEXTURE_TYPE_RGBA32R) {
fprintf(stderr, "%s: Unsupported format %s\n",
__func__, util_format_short_name(psurf->format));
return;
}
for (int by = 0; by < height; by += chunk_h) {
for (int bx = 0; bx < width; bx += chunk_w) {
int all_found_color = -1; /* nothing found */
for (int y = by; y < MIN2(height, by + chunk_h); y++) {
for (int x = bx; x < MIN2(width, bx + chunk_w); x++) {
uint32_t pix = map[y * stride + x];
int i;
for (i = 0; i < num_found_colors; i++) {
if (pix == found_colors[i])
break;
}
if (i == num_found_colors &&
num_found_colors <
ARRAY_SIZE(found_colors)) {
found_colors[num_found_colors++] = pix;
}
if (i < num_found_colors) {
if (all_found_color == -1)
all_found_color = i;
else if (i != all_found_color)
all_found_color = ARRAY_SIZE(found_colors);
}
}
}
/* If all pixels for this chunk have a consistent
* value, then print a character for it. Either a
* fixed name (particularly common for piglit tests),
* or a runtime-generated number.
*/
if (all_found_color >= 0 &&
all_found_color < ARRAY_SIZE(found_colors)) {
static const struct {
uint32_t val;
const char *c;
} named_colors[] = {
{ 0xff000000, "█" },
{ 0x00000000, "█" },
{ 0xffff0000, "r" },
{ 0xff00ff00, "g" },
{ 0xff0000ff, "b" },
{ 0xffffffff, "w" },
};
int i;
for (i = 0; i < ARRAY_SIZE(named_colors); i++) {
if (named_colors[i].val ==
found_colors[all_found_color]) {
fprintf(stderr, "%s",
named_colors[i].c);
break;
}
}
/* For unnamed colors, print a number and the
* numbers will have values printed at the
* end.
*/
if (i == ARRAY_SIZE(named_colors)) {
fprintf(stderr, "%c",
'0' + all_found_color);
}
} else {
/* If there's no consistent color, print this.
*/
fprintf(stderr, ".");
}
}
fprintf(stderr, "\n");
}
for (int i = 0; i < num_found_colors; i++) {
fprintf(stderr, "color %d: 0x%08x\n", i, found_colors[i]);
}
}
static uint32_t
vc4_surface_msaa_get_sample(struct pipe_surface *psurf,
uint32_t x, uint32_t y, uint32_t sample)
{
struct pipe_resource *prsc = psurf->texture;
struct vc4_resource *rsc = vc4_resource(prsc);
uint32_t tile_w = 32, tile_h = 32;
uint32_t tiles_w = DIV_ROUND_UP(psurf->width, 32);
uint32_t tile_x = x / tile_w;
uint32_t tile_y = y / tile_h;
uint32_t *tile = (vc4_bo_map(rsc->bo) +
VC4_TILE_BUFFER_SIZE * (tile_y * tiles_w + tile_x));
uint32_t subtile_x = x % tile_w;
uint32_t subtile_y = y % tile_h;
uint32_t quad_samples = VC4_MAX_SAMPLES * 4;
uint32_t tile_stride = quad_samples * tile_w / 2;
return *((uint32_t *)tile +
(subtile_y >> 1) * tile_stride +
(subtile_x >> 1) * quad_samples +
((subtile_y & 1) << 1) +
(subtile_x & 1) +
sample);
}
static void
vc4_dump_surface_msaa_char(struct pipe_surface *psurf,
uint32_t start_x, uint32_t start_y,
uint32_t w, uint32_t h)
{
bool all_same_color = true;
uint32_t all_pix = 0;
for (int y = start_y; y < start_y + h; y++) {
for (int x = start_x; x < start_x + w; x++) {
for (int s = 0; s < VC4_MAX_SAMPLES; s++) {
uint32_t pix = vc4_surface_msaa_get_sample(psurf,
x, y,
s);
if (x == start_x && y == start_y)
all_pix = pix;
else if (all_pix != pix)
all_same_color = false;
}
}
}
if (all_same_color) {
static const struct {
uint32_t val;
const char *c;
} named_colors[] = {
{ 0xff000000, "█" },
{ 0x00000000, "█" },
{ 0xffff0000, "r" },
{ 0xff00ff00, "g" },
{ 0xff0000ff, "b" },
{ 0xffffffff, "w" },
};
int i;
for (i = 0; i < ARRAY_SIZE(named_colors); i++) {
if (named_colors[i].val == all_pix) {
fprintf(stderr, "%s",
named_colors[i].c);
return;
}
}
fprintf(stderr, "x");
} else {
fprintf(stderr, ".");
}
}
static void
vc4_dump_surface_msaa(struct pipe_surface *psurf)
{
uint32_t tile_w = 32, tile_h = 32;
uint32_t tiles_w = DIV_ROUND_UP(psurf->width, tile_w);
uint32_t tiles_h = DIV_ROUND_UP(psurf->height, tile_h);
uint32_t char_w = 140, char_h = 60;
uint32_t char_w_per_tile = char_w / tiles_w - 1;
uint32_t char_h_per_tile = char_h / tiles_h - 1;
fprintf(stderr, "Surface: %dx%d (%dx MSAA)\n",
psurf->width, psurf->height, psurf->texture->nr_samples);
for (int x = 0; x < (char_w_per_tile + 1) * tiles_w; x++)
fprintf(stderr, "-");
fprintf(stderr, "\n");
for (int ty = 0; ty < psurf->height; ty += tile_h) {
for (int y = 0; y < char_h_per_tile; y++) {
for (int tx = 0; tx < psurf->width; tx += tile_w) {
for (int x = 0; x < char_w_per_tile; x++) {
uint32_t bx1 = (x * tile_w /
char_w_per_tile);
uint32_t bx2 = ((x + 1) * tile_w /
char_w_per_tile);
uint32_t by1 = (y * tile_h /
char_h_per_tile);
uint32_t by2 = ((y + 1) * tile_h /
char_h_per_tile);
vc4_dump_surface_msaa_char(psurf,
tx + bx1,
ty + by1,
bx2 - bx1,
by2 - by1);
}
fprintf(stderr, "|");
}
fprintf(stderr, "\n");
}
for (int x = 0; x < (char_w_per_tile + 1) * tiles_w; x++)
fprintf(stderr, "-");
fprintf(stderr, "\n");
}
}
/** Debug routine to dump the contents of an 8888 surface to the console */
void
vc4_dump_surface(struct pipe_surface *psurf)
{
if (!psurf)
return;
if (psurf->texture->nr_samples > 1)
vc4_dump_surface_msaa(psurf);
else
vc4_dump_surface_non_msaa(psurf);
}
static void
vc4_flush_resource(struct pipe_context *pctx, struct pipe_resource *resource)
{
/* All calls to flush_resource are followed by a flush of the context,
* so there's nothing to do.
*/
}
void
vc4_update_shadow_baselevel_texture(struct pipe_context *pctx,
struct pipe_sampler_view *pview)
{
struct vc4_sampler_view *view = vc4_sampler_view(pview);
struct vc4_resource *shadow = vc4_resource(view->texture);
struct vc4_resource *orig = vc4_resource(pview->texture);
assert(view->texture != pview->texture);
if (shadow->writes == orig->writes && orig->bo->private)
return;
perf_debug("Updating %dx%d@%d shadow texture due to %s\n",
orig->base.width0, orig->base.height0,
pview->u.tex.first_level,
pview->u.tex.first_level ? "base level" : "raster layout");
for (int i = 0; i <= shadow->base.last_level; i++) {
unsigned width = u_minify(shadow->base.width0, i);
unsigned height = u_minify(shadow->base.height0, i);
struct pipe_blit_info info = {
.dst = {
.resource = &shadow->base,
.level = i,
.box = {
.x = 0,
.y = 0,
.z = 0,
.width = width,
.height = height,
.depth = 1,
},
.format = shadow->base.format,
},
.src = {
.resource = &orig->base,
.level = pview->u.tex.first_level + i,
.box = {
.x = 0,
.y = 0,
.z = 0,
.width = width,
.height = height,
.depth = 1,
},
.format = orig->base.format,
},
.mask = ~0,
};
pctx->blit(pctx, &info);
}
shadow->writes = orig->writes;
}
/**
* Converts a 4-byte index buffer to 2 bytes.
*
* Since GLES2 only has support for 1 and 2-byte indices, the hardware doesn't
* include 4-byte index support, and we have to shrink it down.
*
* There's no fallback support for when indices end up being larger than 2^16,
* though it will at least assertion fail. Also, if the original index data
* was in user memory, it would be nice to not have uploaded it to a VBO
* before translating.
*/
struct pipe_resource *
vc4_get_shadow_index_buffer(struct pipe_context *pctx,
const struct pipe_draw_info *info,
uint32_t offset,
uint32_t count,
uint32_t *shadow_offset)
{
struct vc4_context *vc4 = vc4_context(pctx);
struct vc4_resource *orig = vc4_resource(info->index.resource);
perf_debug("Fallback conversion for %d uint indices\n", count);
void *data;
struct pipe_resource *shadow_rsc = NULL;
u_upload_alloc(vc4->uploader, 0, count * 2, 4,
shadow_offset, &shadow_rsc, &data);
uint16_t *dst = data;
struct pipe_transfer *src_transfer = NULL;
const uint32_t *src;
if (info->has_user_indices) {
src = info->index.user;
} else {
src = pipe_buffer_map_range(pctx, &orig->base,
offset,
count * 4,
PIPE_TRANSFER_READ, &src_transfer);
}
for (int i = 0; i < count; i++) {
uint32_t src_index = src[i];
assert(src_index <= 0xffff);
dst[i] = src_index;
}
if (src_transfer)
pctx->transfer_unmap(pctx, src_transfer);
return shadow_rsc;
}
static const struct u_transfer_vtbl transfer_vtbl = {
.resource_create = vc4_resource_create,
.resource_destroy = vc4_resource_destroy,
.transfer_map = vc4_resource_transfer_map,
.transfer_unmap = vc4_resource_transfer_unmap,
.transfer_flush_region = u_default_transfer_flush_region,
};
void
vc4_resource_screen_init(struct pipe_screen *pscreen)
{
struct vc4_screen *screen = vc4_screen(pscreen);
pscreen->resource_create = vc4_resource_create;
pscreen->resource_create_with_modifiers =
vc4_resource_create_with_modifiers;
pscreen->resource_from_handle = vc4_resource_from_handle;
pscreen->resource_destroy = u_resource_destroy_vtbl;
pscreen->resource_get_handle = vc4_resource_get_handle;
pscreen->resource_destroy = vc4_resource_destroy;
pscreen->transfer_helper = u_transfer_helper_create(&transfer_vtbl,
false, false,
false, true);
/* Test if the kernel has GET_TILING; it will return -EINVAL if the
* ioctl does not exist, but -ENOENT if we pass an impossible handle.
* 0 cannot be a valid GEM object, so use that.
*/
struct drm_vc4_get_tiling get_tiling = {
.handle = 0x0,
};
int ret = vc4_ioctl(screen->fd, DRM_IOCTL_VC4_GET_TILING, &get_tiling);
if (ret == -1 && errno == ENOENT)
screen->has_tiling_ioctl = true;
}
void
vc4_resource_context_init(struct pipe_context *pctx)
{
pctx->transfer_map = u_transfer_helper_transfer_map;
pctx->transfer_flush_region = u_transfer_helper_transfer_flush_region;
pctx->transfer_unmap = u_transfer_helper_transfer_unmap;
pctx->buffer_subdata = u_default_buffer_subdata;
pctx->texture_subdata = vc4_texture_subdata;
pctx->create_surface = vc4_create_surface;
pctx->surface_destroy = vc4_surface_destroy;
pctx->resource_copy_region = util_resource_copy_region;
pctx->blit = vc4_blit;
pctx->flush_resource = vc4_flush_resource;
}
|