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
|
/*
* Copyright © 2015 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.
*/
#include <assert.h>
#include <stdbool.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include "anv_private.h"
/** \file anv_batch_chain.c
*
* This file contains functions related to anv_cmd_buffer as a data
* structure. This involves everything required to create and destroy
* the actual batch buffers as well as link them together and handle
* relocations and surface state. It specifically does *not* contain any
* handling of actual vkCmd calls beyond vkCmdExecuteCommands.
*/
/*-----------------------------------------------------------------------*
* Functions related to anv_reloc_list
*-----------------------------------------------------------------------*/
static VkResult
anv_reloc_list_init_clone(struct anv_reloc_list *list,
struct anv_device *device,
const struct anv_reloc_list *other_list)
{
if (other_list) {
list->num_relocs = other_list->num_relocs;
list->array_length = other_list->array_length;
} else {
list->num_relocs = 0;
list->array_length = 256;
}
list->relocs =
anv_device_alloc(device, list->array_length * sizeof(*list->relocs), 8,
VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (list->relocs == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
list->reloc_bos =
anv_device_alloc(device, list->array_length * sizeof(*list->reloc_bos), 8,
VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (list->reloc_bos == NULL) {
anv_device_free(device, list->relocs);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
if (other_list) {
memcpy(list->relocs, other_list->relocs,
list->array_length * sizeof(*list->relocs));
memcpy(list->reloc_bos, other_list->reloc_bos,
list->array_length * sizeof(*list->reloc_bos));
}
return VK_SUCCESS;
}
VkResult
anv_reloc_list_init(struct anv_reloc_list *list, struct anv_device *device)
{
return anv_reloc_list_init_clone(list, device, NULL);
}
void
anv_reloc_list_finish(struct anv_reloc_list *list, struct anv_device *device)
{
anv_device_free(device, list->relocs);
anv_device_free(device, list->reloc_bos);
}
static VkResult
anv_reloc_list_grow(struct anv_reloc_list *list, struct anv_device *device,
size_t num_additional_relocs)
{
if (list->num_relocs + num_additional_relocs <= list->array_length)
return VK_SUCCESS;
size_t new_length = list->array_length * 2;
while (new_length < list->num_relocs + num_additional_relocs)
new_length *= 2;
struct drm_i915_gem_relocation_entry *new_relocs =
anv_device_alloc(device, new_length * sizeof(*list->relocs), 8,
VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (new_relocs == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
struct anv_bo **new_reloc_bos =
anv_device_alloc(device, new_length * sizeof(*list->reloc_bos), 8,
VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (new_relocs == NULL) {
anv_device_free(device, new_relocs);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
memcpy(new_relocs, list->relocs, list->num_relocs * sizeof(*list->relocs));
memcpy(new_reloc_bos, list->reloc_bos,
list->num_relocs * sizeof(*list->reloc_bos));
anv_device_free(device, list->relocs);
anv_device_free(device, list->reloc_bos);
list->array_length = new_length;
list->relocs = new_relocs;
list->reloc_bos = new_reloc_bos;
return VK_SUCCESS;
}
uint64_t
anv_reloc_list_add(struct anv_reloc_list *list, struct anv_device *device,
uint32_t offset, struct anv_bo *target_bo, uint32_t delta)
{
struct drm_i915_gem_relocation_entry *entry;
int index;
anv_reloc_list_grow(list, device, 1);
/* TODO: Handle failure */
/* XXX: Can we use I915_EXEC_HANDLE_LUT? */
index = list->num_relocs++;
list->reloc_bos[index] = target_bo;
entry = &list->relocs[index];
entry->target_handle = target_bo->gem_handle;
entry->delta = delta;
entry->offset = offset;
entry->presumed_offset = target_bo->offset;
entry->read_domains = 0;
entry->write_domain = 0;
return target_bo->offset + delta;
}
static void
anv_reloc_list_append(struct anv_reloc_list *list, struct anv_device *device,
struct anv_reloc_list *other, uint32_t offset)
{
anv_reloc_list_grow(list, device, other->num_relocs);
/* TODO: Handle failure */
memcpy(&list->relocs[list->num_relocs], &other->relocs[0],
other->num_relocs * sizeof(other->relocs[0]));
memcpy(&list->reloc_bos[list->num_relocs], &other->reloc_bos[0],
other->num_relocs * sizeof(other->reloc_bos[0]));
for (uint32_t i = 0; i < other->num_relocs; i++)
list->relocs[i + list->num_relocs].offset += offset;
list->num_relocs += other->num_relocs;
}
/*-----------------------------------------------------------------------*
* Functions related to anv_batch
*-----------------------------------------------------------------------*/
void *
anv_batch_emit_dwords(struct anv_batch *batch, int num_dwords)
{
if (batch->next + num_dwords * 4 > batch->end)
batch->extend_cb(batch, batch->user_data);
void *p = batch->next;
batch->next += num_dwords * 4;
assert(batch->next <= batch->end);
return p;
}
uint64_t
anv_batch_emit_reloc(struct anv_batch *batch,
void *location, struct anv_bo *bo, uint32_t delta)
{
return anv_reloc_list_add(batch->relocs, batch->device,
location - batch->start, bo, delta);
}
void
anv_batch_emit_batch(struct anv_batch *batch, struct anv_batch *other)
{
uint32_t size, offset;
size = other->next - other->start;
assert(size % 4 == 0);
if (batch->next + size > batch->end)
batch->extend_cb(batch, batch->user_data);
assert(batch->next + size <= batch->end);
VG(VALGRIND_CHECK_MEM_IS_DEFINED(other->start, size));
memcpy(batch->next, other->start, size);
offset = batch->next - batch->start;
anv_reloc_list_append(batch->relocs, batch->device,
other->relocs, offset);
batch->next += size;
}
/*-----------------------------------------------------------------------*
* Functions related to anv_batch_bo
*-----------------------------------------------------------------------*/
static VkResult
anv_batch_bo_create(struct anv_device *device, struct anv_batch_bo **bbo_out)
{
VkResult result;
struct anv_batch_bo *bbo =
anv_device_alloc(device, sizeof(*bbo), 8, VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (bbo == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
result = anv_bo_pool_alloc(&device->batch_bo_pool, &bbo->bo);
if (result != VK_SUCCESS)
goto fail_alloc;
result = anv_reloc_list_init(&bbo->relocs, device);
if (result != VK_SUCCESS)
goto fail_bo_alloc;
*bbo_out = bbo;
return VK_SUCCESS;
fail_bo_alloc:
anv_bo_pool_free(&device->batch_bo_pool, &bbo->bo);
fail_alloc:
anv_device_free(device, bbo);
return result;
}
static VkResult
anv_batch_bo_clone(struct anv_device *device,
const struct anv_batch_bo *other_bbo,
struct anv_batch_bo **bbo_out)
{
VkResult result;
struct anv_batch_bo *bbo =
anv_device_alloc(device, sizeof(*bbo), 8, VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (bbo == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
result = anv_bo_pool_alloc(&device->batch_bo_pool, &bbo->bo);
if (result != VK_SUCCESS)
goto fail_alloc;
result = anv_reloc_list_init_clone(&bbo->relocs, device, &other_bbo->relocs);
if (result != VK_SUCCESS)
goto fail_bo_alloc;
bbo->length = other_bbo->length;
memcpy(bbo->bo.map, other_bbo->bo.map, other_bbo->length);
*bbo_out = bbo;
return VK_SUCCESS;
fail_bo_alloc:
anv_bo_pool_free(&device->batch_bo_pool, &bbo->bo);
fail_alloc:
anv_device_free(device, bbo);
return result;
}
static void
anv_batch_bo_start(struct anv_batch_bo *bbo, struct anv_batch *batch,
size_t batch_padding)
{
batch->next = batch->start = bbo->bo.map;
batch->end = bbo->bo.map + bbo->bo.size - batch_padding;
batch->relocs = &bbo->relocs;
bbo->relocs.num_relocs = 0;
}
static void
anv_batch_bo_continue(struct anv_batch_bo *bbo, struct anv_batch *batch,
size_t batch_padding)
{
batch->start = bbo->bo.map;
batch->next = bbo->bo.map + bbo->length;
batch->end = bbo->bo.map + bbo->bo.size - batch_padding;
batch->relocs = &bbo->relocs;
}
static void
anv_batch_bo_finish(struct anv_batch_bo *bbo, struct anv_batch *batch)
{
assert(batch->start == bbo->bo.map);
bbo->length = batch->next - batch->start;
VG(VALGRIND_CHECK_MEM_IS_DEFINED(batch->start, bbo->length));
}
static void
anv_batch_bo_destroy(struct anv_batch_bo *bbo, struct anv_device *device)
{
anv_reloc_list_finish(&bbo->relocs, device);
anv_bo_pool_free(&device->batch_bo_pool, &bbo->bo);
anv_device_free(device, bbo);
}
static VkResult
anv_batch_bo_list_clone(const struct list_head *list, struct anv_device *device,
struct list_head *new_list)
{
VkResult result = VK_SUCCESS;
list_inithead(new_list);
struct anv_batch_bo *prev_bbo = NULL;
list_for_each_entry(struct anv_batch_bo, bbo, list, link) {
struct anv_batch_bo *new_bbo;
result = anv_batch_bo_clone(device, bbo, &new_bbo);
if (result != VK_SUCCESS)
break;
list_addtail(&new_bbo->link, new_list);
if (prev_bbo) {
/* As we clone this list of batch_bo's, they chain one to the
* other using MI_BATCH_BUFFER_START commands. We need to fix up
* those relocations as we go. Fortunately, this is pretty easy
* as it will always be the last relocation in the list.
*/
uint32_t last_idx = prev_bbo->relocs.num_relocs - 1;
assert(prev_bbo->relocs.reloc_bos[last_idx] == &bbo->bo);
prev_bbo->relocs.reloc_bos[last_idx] = &new_bbo->bo;
}
prev_bbo = new_bbo;
}
if (result != VK_SUCCESS) {
list_for_each_entry_safe(struct anv_batch_bo, bbo, new_list, link)
anv_batch_bo_destroy(bbo, device);
}
return result;
}
/*-----------------------------------------------------------------------*
* Functions related to anv_batch_bo
*-----------------------------------------------------------------------*/
static inline struct anv_batch_bo *
anv_cmd_buffer_current_batch_bo(struct anv_cmd_buffer *cmd_buffer)
{
return LIST_ENTRY(struct anv_batch_bo, cmd_buffer->batch_bos.prev, link);
}
static inline struct anv_batch_bo *
anv_cmd_buffer_current_surface_bbo(struct anv_cmd_buffer *cmd_buffer)
{
return LIST_ENTRY(struct anv_batch_bo, cmd_buffer->surface_bos.prev, link);
}
struct anv_bo *
anv_cmd_buffer_current_surface_bo(struct anv_cmd_buffer *cmd_buffer)
{
return &anv_cmd_buffer_current_surface_bbo(cmd_buffer)->bo;
}
struct anv_reloc_list *
anv_cmd_buffer_current_surface_relocs(struct anv_cmd_buffer *cmd_buffer)
{
return &anv_cmd_buffer_current_surface_bbo(cmd_buffer)->relocs;
}
static void
cmd_buffer_chain_to_batch_bo(struct anv_cmd_buffer *cmd_buffer,
struct anv_batch_bo *bbo)
{
struct anv_batch *batch = &cmd_buffer->batch;
struct anv_batch_bo *current_bbo =
anv_cmd_buffer_current_batch_bo(cmd_buffer);
/* We set the end of the batch a little short so we would be sure we
* have room for the chaining command. Since we're about to emit the
* chaining command, let's set it back where it should go.
*/
batch->end += GEN8_MI_BATCH_BUFFER_START_length * 4;
assert(batch->end == current_bbo->bo.map + current_bbo->bo.size);
anv_batch_emit(batch, GEN8_MI_BATCH_BUFFER_START,
GEN8_MI_BATCH_BUFFER_START_header,
._2ndLevelBatchBuffer = _1stlevelbatch,
.AddressSpaceIndicator = ASI_PPGTT,
.BatchBufferStartAddress = { &bbo->bo, 0 },
);
anv_batch_bo_finish(current_bbo, batch);
}
static VkResult
anv_cmd_buffer_chain_batch(struct anv_batch *batch, void *_data)
{
struct anv_cmd_buffer *cmd_buffer = _data;
struct anv_batch_bo *new_bbo;
VkResult result = anv_batch_bo_create(cmd_buffer->device, &new_bbo);
if (result != VK_SUCCESS)
return result;
struct anv_batch_bo **seen_bbo = anv_vector_add(&cmd_buffer->seen_bbos);
if (seen_bbo == NULL) {
anv_batch_bo_destroy(new_bbo, cmd_buffer->device);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
*seen_bbo = new_bbo;
cmd_buffer_chain_to_batch_bo(cmd_buffer, new_bbo);
list_addtail(&new_bbo->link, &cmd_buffer->batch_bos);
anv_batch_bo_start(new_bbo, batch, GEN8_MI_BATCH_BUFFER_START_length * 4);
return VK_SUCCESS;
}
struct anv_state
anv_cmd_buffer_alloc_surface_state(struct anv_cmd_buffer *cmd_buffer,
uint32_t size, uint32_t alignment)
{
struct anv_bo *surface_bo =
anv_cmd_buffer_current_surface_bo(cmd_buffer);
struct anv_state state;
state.offset = align_u32(cmd_buffer->surface_next, alignment);
if (state.offset + size > surface_bo->size)
return (struct anv_state) { 0 };
state.map = surface_bo->map + state.offset;
state.alloc_size = size;
cmd_buffer->surface_next = state.offset + size;
assert(state.offset + size <= surface_bo->size);
return state;
}
struct anv_state
anv_cmd_buffer_alloc_dynamic_state(struct anv_cmd_buffer *cmd_buffer,
uint32_t size, uint32_t alignment)
{
return anv_state_stream_alloc(&cmd_buffer->dynamic_state_stream,
size, alignment);
}
VkResult
anv_cmd_buffer_new_surface_state_bo(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_batch_bo *new_bbo, *old_bbo =
anv_cmd_buffer_current_surface_bbo(cmd_buffer);
/* Finish off the old buffer */
old_bbo->length = cmd_buffer->surface_next;
VkResult result = anv_batch_bo_create(cmd_buffer->device, &new_bbo);
if (result != VK_SUCCESS)
return result;
struct anv_batch_bo **seen_bbo = anv_vector_add(&cmd_buffer->seen_bbos);
if (seen_bbo == NULL) {
anv_batch_bo_destroy(new_bbo, cmd_buffer->device);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
*seen_bbo = new_bbo;
cmd_buffer->surface_next = 1;
list_addtail(&new_bbo->link, &cmd_buffer->surface_bos);
return VK_SUCCESS;
}
VkResult
anv_cmd_buffer_init_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_batch_bo *batch_bo, *surface_bbo;
struct anv_device *device = cmd_buffer->device;
VkResult result;
list_inithead(&cmd_buffer->batch_bos);
list_inithead(&cmd_buffer->surface_bos);
result = anv_batch_bo_create(device, &batch_bo);
if (result != VK_SUCCESS)
return result;
list_addtail(&batch_bo->link, &cmd_buffer->batch_bos);
cmd_buffer->batch.device = device;
cmd_buffer->batch.extend_cb = anv_cmd_buffer_chain_batch;
cmd_buffer->batch.user_data = cmd_buffer;
anv_batch_bo_start(batch_bo, &cmd_buffer->batch,
GEN8_MI_BATCH_BUFFER_START_length * 4);
result = anv_batch_bo_create(device, &surface_bbo);
if (result != VK_SUCCESS)
goto fail_batch_bo;
list_addtail(&surface_bbo->link, &cmd_buffer->surface_bos);
int success = anv_vector_init(&cmd_buffer->seen_bbos,
sizeof(struct anv_bo *),
8 * sizeof(struct anv_bo *));
if (!success)
goto fail_surface_bo;
*(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) = batch_bo;
*(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) = surface_bbo;
/* Start surface_next at 1 so surface offset 0 is invalid. */
cmd_buffer->surface_next = 1;
cmd_buffer->execbuf2.objects = NULL;
cmd_buffer->execbuf2.bos = NULL;
cmd_buffer->execbuf2.array_length = 0;
return VK_SUCCESS;
fail_surface_bo:
anv_batch_bo_destroy(surface_bbo, device);
fail_batch_bo:
anv_batch_bo_destroy(batch_bo, device);
return result;
}
void
anv_cmd_buffer_fini_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_device *device = cmd_buffer->device;
anv_vector_finish(&cmd_buffer->seen_bbos);
/* Destroy all of the batch buffers */
list_for_each_entry_safe(struct anv_batch_bo, bbo,
&cmd_buffer->batch_bos, link) {
anv_batch_bo_destroy(bbo, device);
}
/* Destroy all of the surface state buffers */
list_for_each_entry_safe(struct anv_batch_bo, bbo,
&cmd_buffer->surface_bos, link) {
anv_batch_bo_destroy(bbo, device);
}
anv_device_free(device, cmd_buffer->execbuf2.objects);
anv_device_free(device, cmd_buffer->execbuf2.bos);
}
void
anv_cmd_buffer_reset_batch_bo_chain(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_device *device = cmd_buffer->device;
/* Delete all but the first batch bo */
assert(!list_empty(&cmd_buffer->batch_bos));
while (cmd_buffer->batch_bos.next != cmd_buffer->batch_bos.prev) {
struct anv_batch_bo *bbo = anv_cmd_buffer_current_batch_bo(cmd_buffer);
list_del(&bbo->link);
anv_batch_bo_destroy(bbo, device);
}
assert(!list_empty(&cmd_buffer->batch_bos));
anv_batch_bo_start(anv_cmd_buffer_current_batch_bo(cmd_buffer),
&cmd_buffer->batch,
GEN8_MI_BATCH_BUFFER_START_length * 4);
/* Delete all but the first batch bo */
assert(!list_empty(&cmd_buffer->batch_bos));
while (cmd_buffer->surface_bos.next != cmd_buffer->surface_bos.prev) {
struct anv_batch_bo *bbo = anv_cmd_buffer_current_surface_bbo(cmd_buffer);
list_del(&bbo->link);
anv_batch_bo_destroy(bbo, device);
}
assert(!list_empty(&cmd_buffer->batch_bos));
anv_cmd_buffer_current_surface_bbo(cmd_buffer)->relocs.num_relocs = 0;
cmd_buffer->surface_next = 1;
/* Reset the list of seen buffers */
cmd_buffer->seen_bbos.head = 0;
cmd_buffer->seen_bbos.tail = 0;
*(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) =
anv_cmd_buffer_current_batch_bo(cmd_buffer);
*(struct anv_batch_bo **)anv_vector_add(&cmd_buffer->seen_bbos) =
anv_cmd_buffer_current_surface_bbo(cmd_buffer);
}
void
anv_cmd_buffer_end_batch_buffer(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_batch_bo *batch_bo = anv_cmd_buffer_current_batch_bo(cmd_buffer);
struct anv_batch_bo *surface_bbo =
anv_cmd_buffer_current_surface_bbo(cmd_buffer);
if (cmd_buffer->level == VK_CMD_BUFFER_LEVEL_PRIMARY) {
anv_batch_emit(&cmd_buffer->batch, GEN8_MI_BATCH_BUFFER_END);
/* Round batch up to an even number of dwords. */
if ((cmd_buffer->batch.next - cmd_buffer->batch.start) & 4)
anv_batch_emit(&cmd_buffer->batch, GEN8_MI_NOOP);
cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_PRIMARY;
}
anv_batch_bo_finish(batch_bo, &cmd_buffer->batch);
surface_bbo->length = cmd_buffer->surface_next;
if (cmd_buffer->level == VK_CMD_BUFFER_LEVEL_SECONDARY) {
/* If this is a secondary command buffer, we need to determine the
* mode in which it will be executed with vkExecuteCommands. We
* determine this statically here so that this stays in sync with the
* actual ExecuteCommands implementation.
*/
if ((cmd_buffer->batch_bos.next == cmd_buffer->batch_bos.prev) &&
(anv_cmd_buffer_current_batch_bo(cmd_buffer)->length <
ANV_CMD_BUFFER_BATCH_SIZE / 2)) {
/* If the secondary has exactly one batch buffer in its list *and*
* that batch buffer is less than half of the maximum size, we're
* probably better of simply copying it into our batch.
*/
cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_EMIT;
} else if (cmd_buffer->opt_flags &
VK_CMD_BUFFER_OPTIMIZE_NO_SIMULTANEOUS_USE_BIT) {
cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_CHAIN;
/* For chaining mode, we need to increment the number of
* relocations. This is because, when we chain, we need to add
* an MI_BATCH_BUFFER_START command. Adding this command will
* also add a relocation. In order to handle theis we'll
* increment it here and decrement it right before adding the
* MI_BATCH_BUFFER_START command.
*/
anv_cmd_buffer_current_batch_bo(cmd_buffer)->relocs.num_relocs++;
} else {
cmd_buffer->exec_mode = ANV_CMD_BUFFER_EXEC_MODE_COPY_AND_CHAIN;
}
}
}
static inline VkResult
anv_cmd_buffer_add_seen_bbos(struct anv_cmd_buffer *cmd_buffer,
struct list_head *list)
{
list_for_each_entry(struct anv_batch_bo, bbo, list, link) {
struct anv_batch_bo **bbo_ptr = anv_vector_add(&cmd_buffer->seen_bbos);
if (bbo_ptr == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
*bbo_ptr = bbo;
}
return VK_SUCCESS;
}
void
anv_cmd_buffer_add_secondary(struct anv_cmd_buffer *primary,
struct anv_cmd_buffer *secondary)
{
switch (secondary->exec_mode) {
case ANV_CMD_BUFFER_EXEC_MODE_EMIT:
anv_batch_emit_batch(&primary->batch, &secondary->batch);
break;
case ANV_CMD_BUFFER_EXEC_MODE_CHAIN: {
struct anv_batch_bo *first_bbo =
list_first_entry(&secondary->batch_bos, struct anv_batch_bo, link);
struct anv_batch_bo *last_bbo =
list_last_entry(&secondary->batch_bos, struct anv_batch_bo, link);
anv_batch_emit(&primary->batch, GEN8_MI_BATCH_BUFFER_START,
GEN8_MI_BATCH_BUFFER_START_header,
._2ndLevelBatchBuffer = _1stlevelbatch,
.AddressSpaceIndicator = ASI_PPGTT,
.BatchBufferStartAddress = { &first_bbo->bo, 0 },
);
struct anv_batch_bo *this_bbo = anv_cmd_buffer_current_batch_bo(primary);
assert(primary->batch.start == this_bbo->bo.map);
uint32_t offset = primary->batch.next - primary->batch.start;
struct GEN8_MI_BATCH_BUFFER_START ret = {
GEN8_MI_BATCH_BUFFER_START_header,
._2ndLevelBatchBuffer = _1stlevelbatch,
.AddressSpaceIndicator = ASI_PPGTT,
.BatchBufferStartAddress = { &this_bbo->bo, offset },
};
/* The pack function below is going to insert a relocation. In order
* to allow us to splice this secondary into a primary multiple
* times, we can't have relocations from previous splices in this
* splice. In order to deal with this, we simply decrement the
* relocation count prior to inserting the next one. In order to
* handle the base case, num_relocs was artificially incremented in
* end_batch_buffer().
*/
last_bbo->relocs.num_relocs--;
GEN8_MI_BATCH_BUFFER_START_pack(&secondary->batch,
last_bbo->bo.map + last_bbo->length,
&ret);
anv_cmd_buffer_add_seen_bbos(primary, &secondary->batch_bos);
break;
}
case ANV_CMD_BUFFER_EXEC_MODE_COPY_AND_CHAIN: {
struct list_head copy_list;
VkResult result = anv_batch_bo_list_clone(&secondary->batch_bos,
secondary->device,
©_list);
if (result != VK_SUCCESS)
return; /* FIXME */
anv_cmd_buffer_add_seen_bbos(primary, ©_list);
struct anv_batch_bo *first_bbo =
list_first_entry(©_list, struct anv_batch_bo, link);
struct anv_batch_bo *last_bbo =
list_last_entry(©_list, struct anv_batch_bo, link);
cmd_buffer_chain_to_batch_bo(primary, first_bbo);
list_splicetail(©_list, &primary->batch_bos);
anv_batch_bo_continue(last_bbo, &primary->batch,
GEN8_MI_BATCH_BUFFER_START_length * 4);
anv_cmd_buffer_emit_state_base_address(primary);
break;
}
default:
assert(!"Invalid execution mode");
}
/* Mark the surface buffer from the secondary as seen */
anv_cmd_buffer_add_seen_bbos(primary, &secondary->surface_bos);
}
static VkResult
anv_cmd_buffer_add_bo(struct anv_cmd_buffer *cmd_buffer,
struct anv_bo *bo,
struct anv_reloc_list *relocs)
{
struct drm_i915_gem_exec_object2 *obj = NULL;
if (bo->index < cmd_buffer->execbuf2.bo_count &&
cmd_buffer->execbuf2.bos[bo->index] == bo)
obj = &cmd_buffer->execbuf2.objects[bo->index];
if (obj == NULL) {
/* We've never seen this one before. Add it to the list and assign
* an id that we can use later.
*/
if (cmd_buffer->execbuf2.bo_count >= cmd_buffer->execbuf2.array_length) {
uint32_t new_len = cmd_buffer->execbuf2.objects ?
cmd_buffer->execbuf2.array_length * 2 : 64;
struct drm_i915_gem_exec_object2 *new_objects =
anv_device_alloc(cmd_buffer->device, new_len * sizeof(*new_objects),
8, VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (new_objects == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
struct anv_bo **new_bos =
anv_device_alloc(cmd_buffer->device, new_len * sizeof(*new_bos),
8, VK_SYSTEM_ALLOC_TYPE_INTERNAL);
if (new_objects == NULL) {
anv_device_free(cmd_buffer->device, new_objects);
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
if (cmd_buffer->execbuf2.objects) {
memcpy(new_objects, cmd_buffer->execbuf2.objects,
cmd_buffer->execbuf2.bo_count * sizeof(*new_objects));
memcpy(new_bos, cmd_buffer->execbuf2.bos,
cmd_buffer->execbuf2.bo_count * sizeof(*new_bos));
}
cmd_buffer->execbuf2.objects = new_objects;
cmd_buffer->execbuf2.bos = new_bos;
cmd_buffer->execbuf2.array_length = new_len;
}
assert(cmd_buffer->execbuf2.bo_count < cmd_buffer->execbuf2.array_length);
bo->index = cmd_buffer->execbuf2.bo_count++;
obj = &cmd_buffer->execbuf2.objects[bo->index];
cmd_buffer->execbuf2.bos[bo->index] = bo;
obj->handle = bo->gem_handle;
obj->relocation_count = 0;
obj->relocs_ptr = 0;
obj->alignment = 0;
obj->offset = bo->offset;
obj->flags = 0;
obj->rsvd1 = 0;
obj->rsvd2 = 0;
}
if (relocs != NULL && obj->relocation_count == 0) {
/* This is the first time we've ever seen a list of relocations for
* this BO. Go ahead and set the relocations and then walk the list
* of relocations and add them all.
*/
obj->relocation_count = relocs->num_relocs;
obj->relocs_ptr = (uintptr_t) relocs->relocs;
for (size_t i = 0; i < relocs->num_relocs; i++)
anv_cmd_buffer_add_bo(cmd_buffer, relocs->reloc_bos[i], NULL);
}
return VK_SUCCESS;
}
static void
anv_cmd_buffer_process_relocs(struct anv_cmd_buffer *cmd_buffer,
struct anv_reloc_list *list)
{
struct anv_bo *bo;
/* If the kernel supports I915_EXEC_NO_RELOC, it will compare offset in
* struct drm_i915_gem_exec_object2 against the bos current offset and if
* all bos haven't moved it will skip relocation processing alltogether.
* If I915_EXEC_NO_RELOC is not supported, the kernel ignores the incoming
* value of offset so we can set it either way. For that to work we need
* to make sure all relocs use the same presumed offset.
*/
for (size_t i = 0; i < list->num_relocs; i++) {
bo = list->reloc_bos[i];
if (bo->offset != list->relocs[i].presumed_offset)
cmd_buffer->execbuf2.need_reloc = true;
list->relocs[i].target_handle = bo->index;
}
}
void
anv_cmd_buffer_prepare_execbuf(struct anv_cmd_buffer *cmd_buffer)
{
struct anv_batch *batch = &cmd_buffer->batch;
cmd_buffer->execbuf2.bo_count = 0;
cmd_buffer->execbuf2.need_reloc = false;
/* First, we walk over all of the bos we've seen and add them and their
* relocations to the validate list.
*/
struct anv_batch_bo **bbo;
anv_vector_foreach(bbo, &cmd_buffer->seen_bbos)
anv_cmd_buffer_add_bo(cmd_buffer, &(*bbo)->bo, &(*bbo)->relocs);
struct anv_batch_bo *first_batch_bo =
list_first_entry(&cmd_buffer->batch_bos, struct anv_batch_bo, link);
/* The kernel requires that the last entry in the validation list be the
* batch buffer to execute. We can simply swap the element
* corresponding to the first batch_bo in the chain with the last
* element in the list.
*/
if (first_batch_bo->bo.index != cmd_buffer->execbuf2.bo_count - 1) {
uint32_t idx = first_batch_bo->bo.index;
struct drm_i915_gem_exec_object2 tmp_obj =
cmd_buffer->execbuf2.objects[idx];
assert(cmd_buffer->execbuf2.bos[idx] == &first_batch_bo->bo);
cmd_buffer->execbuf2.objects[idx] =
cmd_buffer->execbuf2.objects[cmd_buffer->execbuf2.bo_count - 1];
cmd_buffer->execbuf2.bos[idx] =
cmd_buffer->execbuf2.bos[cmd_buffer->execbuf2.bo_count - 1];
cmd_buffer->execbuf2.bos[idx]->index = idx;
cmd_buffer->execbuf2.objects[cmd_buffer->execbuf2.bo_count - 1] = tmp_obj;
cmd_buffer->execbuf2.bos[cmd_buffer->execbuf2.bo_count - 1] =
&first_batch_bo->bo;
first_batch_bo->bo.index = cmd_buffer->execbuf2.bo_count - 1;
}
/* Now we go through and fixup all of the relocation lists to point to
* the correct indices in the object array. We have to do this after we
* reorder the list above as some of the indices may have changed.
*/
anv_vector_foreach(bbo, &cmd_buffer->seen_bbos)
anv_cmd_buffer_process_relocs(cmd_buffer, &(*bbo)->relocs);
cmd_buffer->execbuf2.execbuf = (struct drm_i915_gem_execbuffer2) {
.buffers_ptr = (uintptr_t) cmd_buffer->execbuf2.objects,
.buffer_count = cmd_buffer->execbuf2.bo_count,
.batch_start_offset = 0,
.batch_len = batch->next - batch->start,
.cliprects_ptr = 0,
.num_cliprects = 0,
.DR1 = 0,
.DR4 = 0,
.flags = I915_EXEC_HANDLE_LUT | I915_EXEC_RENDER,
.rsvd1 = cmd_buffer->device->context_id,
.rsvd2 = 0,
};
if (!cmd_buffer->execbuf2.need_reloc)
cmd_buffer->execbuf2.execbuf.flags |= I915_EXEC_NO_RELOC;
}
|