1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
|
// Copyright 2019 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <algorithm>
#include <array>
#include <deque>
#include <memory>
#include <mutex>
#include <span>
#include <unordered_map>
#include <vector>
#include <boost/container/small_vector.hpp>
#include "common/common_types.h"
#include "common/div_ceil.h"
#include "common/microprofile.h"
#include "common/scope_exit.h"
#include "core/memory.h"
#include "core/settings.h"
#include "video_core/buffer_cache/buffer_base.h"
#include "video_core/delayed_destruction_ring.h"
#include "video_core/dirty_flags.h"
#include "video_core/engines/kepler_compute.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/memory_manager.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/texture_cache/slot_vector.h"
#include "video_core/texture_cache/types.h"
namespace VideoCommon {
MICROPROFILE_DECLARE(GPU_PrepareBuffers);
MICROPROFILE_DECLARE(GPU_BindUploadBuffers);
MICROPROFILE_DECLARE(GPU_DownloadMemory);
using BufferId = SlotId;
constexpr u32 NUM_VERTEX_BUFFERS = 32;
constexpr u32 NUM_TRANSFORM_FEEDBACK_BUFFERS = 4;
constexpr u32 NUM_GRAPHICS_UNIFORM_BUFFERS = 18;
constexpr u32 NUM_COMPUTE_UNIFORM_BUFFERS = 8;
constexpr u32 NUM_STORAGE_BUFFERS = 16;
constexpr u32 NUM_STAGES = 5;
template <typename P>
class BufferCache {
// Page size for caching purposes.
// This is unrelated to the CPU page size and it can be changed as it seems optimal.
static constexpr u32 PAGE_BITS = 16;
static constexpr u64 PAGE_SIZE = u64{1} << PAGE_BITS;
static constexpr bool IS_OPENGL = P::IS_OPENGL;
static constexpr bool HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS =
P::HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS;
static constexpr bool HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT =
P::HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT;
static constexpr bool NEEDS_BIND_UNIFORM_INDEX = P::NEEDS_BIND_UNIFORM_INDEX;
static constexpr bool NEEDS_BIND_STORAGE_INDEX = P::NEEDS_BIND_STORAGE_INDEX;
static constexpr bool USE_MEMORY_MAPS = P::USE_MEMORY_MAPS;
static constexpr BufferId NULL_BUFFER_ID{0};
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
using Runtime = typename P::Runtime;
using Buffer = typename P::Buffer;
struct Empty {};
struct Binding {
VAddr cpu_addr{};
u32 size{};
BufferId buffer_id;
};
static constexpr Binding NULL_BINDING{
.cpu_addr = 0,
.size = 0,
.buffer_id = NULL_BUFFER_ID,
};
public:
static constexpr size_t SKIP_CACHE_SIZE = 4096;
explicit BufferCache(VideoCore::RasterizerInterface& rasterizer_,
Tegra::Engines::Maxwell3D& maxwell3d_,
Tegra::Engines::KeplerCompute& kepler_compute_,
Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
Runtime& runtime_);
void TickFrame();
void WriteMemory(VAddr cpu_addr, u64 size);
void CachedWriteMemory(VAddr cpu_addr, u64 size);
void DownloadMemory(VAddr cpu_addr, u64 size);
void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size);
void UpdateGraphicsBuffers(bool is_indexed);
void UpdateComputeBuffers();
void BindHostGeometryBuffers(bool is_indexed);
void BindHostStageBuffers(size_t stage);
void BindHostComputeBuffers();
void SetEnabledUniformBuffers(size_t stage, u32 enabled);
void SetEnabledComputeUniformBuffers(u32 enabled);
void UnbindGraphicsStorageBuffers(size_t stage);
void BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
bool is_written);
void UnbindComputeStorageBuffers();
void BindComputeStorageBuffer(size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
bool is_written);
void FlushCachedWrites();
/// Return true when there are uncommitted buffers to be downloaded
[[nodiscard]] bool HasUncommittedFlushes() const noexcept;
/// Return true when the caller should wait for async downloads
[[nodiscard]] bool ShouldWaitAsyncFlushes() const noexcept;
/// Commit asynchronous downloads
void CommitAsyncFlushes();
/// Pop asynchronous downloads
void PopAsyncFlushes();
/// Return true when a CPU region is modified from the GPU
[[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size);
std::mutex mutex;
private:
template <typename Func>
static void ForEachEnabledBit(u32 enabled_mask, Func&& func) {
for (u32 index = 0; enabled_mask != 0; ++index, enabled_mask >>= 1) {
const int disabled_bits = std::countr_zero(enabled_mask);
index += disabled_bits;
enabled_mask >>= disabled_bits;
func(index);
}
}
template <typename Func>
void ForEachBufferInRange(VAddr cpu_addr, u64 size, Func&& func) {
const u64 page_end = Common::DivCeil(cpu_addr + size, PAGE_SIZE);
for (u64 page = cpu_addr >> PAGE_BITS; page < page_end;) {
const BufferId buffer_id = page_table[page];
if (!buffer_id) {
++page;
continue;
}
Buffer& buffer = slot_buffers[buffer_id];
func(buffer_id, buffer);
const VAddr end_addr = buffer.CpuAddr() + buffer.SizeBytes();
page = Common::DivCeil(end_addr, PAGE_SIZE);
}
}
static bool IsRangeGranular(VAddr cpu_addr, size_t size) {
return (cpu_addr & ~Core::Memory::PAGE_MASK) ==
((cpu_addr + size) & ~Core::Memory::PAGE_MASK);
}
void BindHostIndexBuffer();
void BindHostVertexBuffers();
void BindHostGraphicsUniformBuffers(size_t stage);
void BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32 binding_index, bool needs_bind);
void BindHostGraphicsStorageBuffers(size_t stage);
void BindHostTransformFeedbackBuffers();
void BindHostComputeUniformBuffers();
void BindHostComputeStorageBuffers();
void DoUpdateGraphicsBuffers(bool is_indexed);
void DoUpdateComputeBuffers();
void UpdateIndexBuffer();
void UpdateVertexBuffers();
void UpdateVertexBuffer(u32 index);
void UpdateUniformBuffers(size_t stage);
void UpdateStorageBuffers(size_t stage);
void UpdateTransformFeedbackBuffers();
void UpdateTransformFeedbackBuffer(u32 index);
void UpdateComputeUniformBuffers();
void UpdateComputeStorageBuffers();
void MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size);
[[nodiscard]] BufferId FindBuffer(VAddr cpu_addr, u32 size);
[[nodiscard]] BufferId CreateBuffer(VAddr cpu_addr, u32 wanted_size);
void Register(BufferId buffer_id);
void Unregister(BufferId buffer_id);
template <bool insert>
void ChangeRegister(BufferId buffer_id);
void SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
void SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size);
void UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
std::span<BufferCopy> copies);
void ImmediateUploadMemory(Buffer& buffer, u64 largest_copy,
std::span<const BufferCopy> copies);
void MappedUploadMemory(Buffer& buffer, u64 total_size_bytes,
std::span<const BufferCopy> copies);
void DeleteBuffer(BufferId buffer_id);
void ReplaceBufferDownloads(BufferId old_buffer_id, BufferId new_buffer_id);
void NotifyBufferDeletion();
[[nodiscard]] Binding StorageBufferBinding(GPUVAddr ssbo_addr) const;
[[nodiscard]] std::span<const u8> ImmediateBufferWithData(VAddr cpu_addr, size_t size);
[[nodiscard]] std::span<u8> ImmediateBuffer(size_t wanted_capacity);
[[nodiscard]] bool HasFastUniformBufferBound(size_t stage, u32 binding_index) const noexcept;
VideoCore::RasterizerInterface& rasterizer;
Tegra::Engines::Maxwell3D& maxwell3d;
Tegra::Engines::KeplerCompute& kepler_compute;
Tegra::MemoryManager& gpu_memory;
Core::Memory::Memory& cpu_memory;
Runtime& runtime;
SlotVector<Buffer> slot_buffers;
DelayedDestructionRing<Buffer, 8> delayed_destruction_ring;
u32 last_index_count = 0;
Binding index_buffer;
std::array<Binding, NUM_VERTEX_BUFFERS> vertex_buffers;
std::array<std::array<Binding, NUM_GRAPHICS_UNIFORM_BUFFERS>, NUM_STAGES> uniform_buffers;
std::array<std::array<Binding, NUM_STORAGE_BUFFERS>, NUM_STAGES> storage_buffers;
std::array<Binding, NUM_TRANSFORM_FEEDBACK_BUFFERS> transform_feedback_buffers;
std::array<Binding, NUM_COMPUTE_UNIFORM_BUFFERS> compute_uniform_buffers;
std::array<Binding, NUM_STORAGE_BUFFERS> compute_storage_buffers;
std::array<u32, NUM_STAGES> enabled_uniform_buffers{};
u32 enabled_compute_uniform_buffers = 0;
std::array<u32, NUM_STAGES> enabled_storage_buffers{};
std::array<u32, NUM_STAGES> written_storage_buffers{};
u32 enabled_compute_storage_buffers = 0;
u32 written_compute_storage_buffers = 0;
std::array<u32, NUM_STAGES> fast_bound_uniform_buffers{};
bool has_deleted_buffers = false;
std::conditional_t<HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS, std::array<u32, NUM_STAGES>, Empty>
dirty_uniform_buffers{};
std::vector<BufferId> cached_write_buffer_ids;
// TODO: This data structure is not optimal and it should be reworked
std::vector<BufferId> uncommitted_downloads;
std::deque<std::vector<BufferId>> committed_downloads;
size_t immediate_buffer_capacity = 0;
std::unique_ptr<u8[]> immediate_buffer_alloc;
std::array<BufferId, ((1ULL << 39) >> PAGE_BITS)> page_table;
};
template <class P>
BufferCache<P>::BufferCache(VideoCore::RasterizerInterface& rasterizer_,
Tegra::Engines::Maxwell3D& maxwell3d_,
Tegra::Engines::KeplerCompute& kepler_compute_,
Tegra::MemoryManager& gpu_memory_, Core::Memory::Memory& cpu_memory_,
Runtime& runtime_)
: rasterizer{rasterizer_}, maxwell3d{maxwell3d_}, kepler_compute{kepler_compute_},
gpu_memory{gpu_memory_}, cpu_memory{cpu_memory_}, runtime{runtime_} {
// Ensure the first slot is used for the null buffer
void(slot_buffers.insert(runtime, NullBufferParams{}));
}
template <class P>
void BufferCache<P>::TickFrame() {
delayed_destruction_ring.Tick();
}
template <class P>
void BufferCache<P>::WriteMemory(VAddr cpu_addr, u64 size) {
ForEachBufferInRange(cpu_addr, size, [&](BufferId, Buffer& buffer) {
buffer.MarkRegionAsCpuModified(cpu_addr, size);
});
}
template <class P>
void BufferCache<P>::CachedWriteMemory(VAddr cpu_addr, u64 size) {
ForEachBufferInRange(cpu_addr, size, [&](BufferId buffer_id, Buffer& buffer) {
if (!buffer.HasCachedWrites()) {
cached_write_buffer_ids.push_back(buffer_id);
}
buffer.CachedCpuWrite(cpu_addr, size);
});
}
template <class P>
void BufferCache<P>::DownloadMemory(VAddr cpu_addr, u64 size) {
ForEachBufferInRange(cpu_addr, size, [&](BufferId, Buffer& buffer) {
boost::container::small_vector<BufferCopy, 1> copies;
u64 total_size_bytes = 0;
u64 largest_copy = 0;
buffer.ForEachDownloadRange(cpu_addr, size, [&](u64 range_offset, u64 range_size) {
copies.push_back(BufferCopy{
.src_offset = range_offset,
.dst_offset = total_size_bytes,
.size = range_size,
});
total_size_bytes += range_size;
largest_copy = std::max(largest_copy, range_size);
});
if (total_size_bytes == 0) {
return;
}
MICROPROFILE_SCOPE(GPU_DownloadMemory);
if constexpr (USE_MEMORY_MAPS) {
auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
const u8* const mapped_memory = download_staging.mapped_span.data();
const std::span<BufferCopy> copies_span(copies.data(), copies.data() + copies.size());
runtime.CopyBuffer(download_staging.buffer, buffer, copies_span);
runtime.Finish();
for (const BufferCopy& copy : copies) {
const VAddr copy_cpu_addr = buffer.CpuAddr() + copy.src_offset;
const u8* copy_mapped_memory = mapped_memory + copy.dst_offset;
cpu_memory.WriteBlockUnsafe(copy_cpu_addr, copy_mapped_memory, copy.size);
}
} else {
const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
for (const BufferCopy& copy : copies) {
buffer.ImmediateDownload(copy.src_offset, immediate_buffer.subspan(0, copy.size));
const VAddr copy_cpu_addr = buffer.CpuAddr() + copy.src_offset;
cpu_memory.WriteBlockUnsafe(copy_cpu_addr, immediate_buffer.data(), copy.size);
}
}
});
}
template <class P>
void BufferCache<P>::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr,
u32 size) {
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
uniform_buffers[stage][index] = NULL_BINDING;
return;
}
const Binding binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = BufferId{},
};
uniform_buffers[stage][index] = binding;
}
template <class P>
void BufferCache<P>::UpdateGraphicsBuffers(bool is_indexed) {
MICROPROFILE_SCOPE(GPU_PrepareBuffers);
do {
has_deleted_buffers = false;
DoUpdateGraphicsBuffers(is_indexed);
} while (has_deleted_buffers);
}
template <class P>
void BufferCache<P>::UpdateComputeBuffers() {
MICROPROFILE_SCOPE(GPU_PrepareBuffers);
do {
has_deleted_buffers = false;
DoUpdateComputeBuffers();
} while (has_deleted_buffers);
}
template <class P>
void BufferCache<P>::BindHostGeometryBuffers(bool is_indexed) {
MICROPROFILE_SCOPE(GPU_BindUploadBuffers);
if (is_indexed) {
BindHostIndexBuffer();
} else if constexpr (!HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT) {
const auto& regs = maxwell3d.regs;
if (regs.draw.topology == Maxwell::PrimitiveTopology::Quads) {
runtime.BindQuadArrayIndexBuffer(regs.vertex_buffer.first, regs.vertex_buffer.count);
}
}
BindHostVertexBuffers();
BindHostTransformFeedbackBuffers();
}
template <class P>
void BufferCache<P>::BindHostStageBuffers(size_t stage) {
MICROPROFILE_SCOPE(GPU_BindUploadBuffers);
BindHostGraphicsUniformBuffers(stage);
BindHostGraphicsStorageBuffers(stage);
}
template <class P>
void BufferCache<P>::BindHostComputeBuffers() {
MICROPROFILE_SCOPE(GPU_BindUploadBuffers);
BindHostComputeUniformBuffers();
BindHostComputeStorageBuffers();
}
template <class P>
void BufferCache<P>::SetEnabledUniformBuffers(size_t stage, u32 enabled) {
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
if (enabled_uniform_buffers[stage] != enabled) {
dirty_uniform_buffers[stage] = ~u32{0};
}
}
enabled_uniform_buffers[stage] = enabled;
}
template <class P>
void BufferCache<P>::SetEnabledComputeUniformBuffers(u32 enabled) {
enabled_compute_uniform_buffers = enabled;
}
template <class P>
void BufferCache<P>::UnbindGraphicsStorageBuffers(size_t stage) {
enabled_storage_buffers[stage] = 0;
written_storage_buffers[stage] = 0;
}
template <class P>
void BufferCache<P>::BindGraphicsStorageBuffer(size_t stage, size_t ssbo_index, u32 cbuf_index,
u32 cbuf_offset, bool is_written) {
enabled_storage_buffers[stage] |= 1U << ssbo_index;
written_storage_buffers[stage] |= (is_written ? 1U : 0U) << ssbo_index;
const auto& cbufs = maxwell3d.state.shader_stages[stage];
const GPUVAddr ssbo_addr = cbufs.const_buffers[cbuf_index].address + cbuf_offset;
storage_buffers[stage][ssbo_index] = StorageBufferBinding(ssbo_addr);
}
template <class P>
void BufferCache<P>::UnbindComputeStorageBuffers() {
enabled_compute_storage_buffers = 0;
written_compute_storage_buffers = 0;
}
template <class P>
void BufferCache<P>::BindComputeStorageBuffer(size_t ssbo_index, u32 cbuf_index, u32 cbuf_offset,
bool is_written) {
enabled_compute_storage_buffers |= 1U << ssbo_index;
written_compute_storage_buffers |= (is_written ? 1U : 0U) << ssbo_index;
const auto& launch_desc = kepler_compute.launch_description;
ASSERT(((launch_desc.const_buffer_enable_mask >> cbuf_index) & 1) != 0);
const auto& cbufs = launch_desc.const_buffer_config;
const GPUVAddr ssbo_addr = cbufs[cbuf_index].Address() + cbuf_offset;
compute_storage_buffers[ssbo_index] = StorageBufferBinding(ssbo_addr);
}
template <class P>
void BufferCache<P>::FlushCachedWrites() {
for (const BufferId buffer_id : cached_write_buffer_ids) {
slot_buffers[buffer_id].FlushCachedWrites();
}
cached_write_buffer_ids.clear();
}
template <class P>
bool BufferCache<P>::HasUncommittedFlushes() const noexcept {
return !uncommitted_downloads.empty();
}
template <class P>
bool BufferCache<P>::ShouldWaitAsyncFlushes() const noexcept {
return !committed_downloads.empty() && !committed_downloads.front().empty();
}
template <class P>
void BufferCache<P>::CommitAsyncFlushes() {
// This is intentionally passing the value by copy
committed_downloads.push_front(uncommitted_downloads);
uncommitted_downloads.clear();
}
template <class P>
void BufferCache<P>::PopAsyncFlushes() {
if (committed_downloads.empty()) {
return;
}
auto scope_exit_pop_download = detail::ScopeExit([this] { committed_downloads.pop_back(); });
const std::span<const BufferId> download_ids = committed_downloads.back();
if (download_ids.empty()) {
return;
}
MICROPROFILE_SCOPE(GPU_DownloadMemory);
boost::container::small_vector<std::pair<BufferCopy, BufferId>, 1> downloads;
u64 total_size_bytes = 0;
u64 largest_copy = 0;
for (const BufferId buffer_id : download_ids) {
slot_buffers[buffer_id].ForEachDownloadRange([&](u64 range_offset, u64 range_size) {
downloads.push_back({
BufferCopy{
.src_offset = range_offset,
.dst_offset = total_size_bytes,
.size = range_size,
},
buffer_id,
});
total_size_bytes += range_size;
largest_copy = std::max(largest_copy, range_size);
});
}
if (downloads.empty()) {
return;
}
if constexpr (USE_MEMORY_MAPS) {
auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
for (const auto [copy, buffer_id] : downloads) {
const std::array copies{copy};
runtime.CopyBuffer(download_staging.buffer, slot_buffers[buffer_id], copies);
}
runtime.Finish();
for (const auto [copy, buffer_id] : downloads) {
const Buffer& buffer = slot_buffers[buffer_id];
const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
const u8* read_mapped_memory = download_staging.mapped_span.data() + copy.dst_offset;
cpu_memory.WriteBlockUnsafe(cpu_addr, read_mapped_memory, copy.size);
}
} else {
const std::span<u8> immediate_buffer = ImmediateBuffer(largest_copy);
for (const auto [copy, buffer_id] : downloads) {
Buffer& buffer = slot_buffers[buffer_id];
buffer.ImmediateDownload(copy.src_offset, immediate_buffer.subspan(0, copy.size));
const VAddr cpu_addr = buffer.CpuAddr() + copy.src_offset;
cpu_memory.WriteBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
}
}
}
template <class P>
bool BufferCache<P>::IsRegionGpuModified(VAddr addr, size_t size) {
const u64 page_end = Common::DivCeil(addr + size, PAGE_SIZE);
for (u64 page = addr >> PAGE_BITS; page < page_end;) {
const BufferId image_id = page_table[page];
if (!image_id) {
++page;
continue;
}
Buffer& buffer = slot_buffers[image_id];
if (buffer.IsRegionGpuModified(addr, size)) {
return true;
}
const VAddr end_addr = buffer.CpuAddr() + buffer.SizeBytes();
page = Common::DivCeil(end_addr, PAGE_SIZE);
}
return false;
}
template <class P>
void BufferCache<P>::BindHostIndexBuffer() {
Buffer& buffer = slot_buffers[index_buffer.buffer_id];
const u32 offset = buffer.Offset(index_buffer.cpu_addr);
const u32 size = index_buffer.size;
SynchronizeBuffer(buffer, index_buffer.cpu_addr, size);
if constexpr (HAS_FULL_INDEX_AND_PRIMITIVE_SUPPORT) {
runtime.BindIndexBuffer(buffer, offset, size);
} else {
runtime.BindIndexBuffer(maxwell3d.regs.draw.topology, maxwell3d.regs.index_array.format,
maxwell3d.regs.index_array.first, maxwell3d.regs.index_array.count,
buffer, offset, size);
}
}
template <class P>
void BufferCache<P>::BindHostVertexBuffers() {
auto& flags = maxwell3d.dirty.flags;
for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
const Binding& binding = vertex_buffers[index];
Buffer& buffer = slot_buffers[binding.buffer_id];
SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
if (!flags[Dirty::VertexBuffer0 + index]) {
continue;
}
flags[Dirty::VertexBuffer0 + index] = false;
const u32 stride = maxwell3d.regs.vertex_array[index].stride;
const u32 offset = buffer.Offset(binding.cpu_addr);
runtime.BindVertexBuffer(index, buffer, offset, binding.size, stride);
}
}
template <class P>
void BufferCache<P>::BindHostGraphicsUniformBuffers(size_t stage) {
u32 dirty = ~0U;
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
dirty = std::exchange(dirty_uniform_buffers[stage], 0);
}
u32 binding_index = 0;
ForEachEnabledBit(enabled_uniform_buffers[stage], [&](u32 index) {
const bool needs_bind = ((dirty >> index) & 1) != 0;
BindHostGraphicsUniformBuffer(stage, index, binding_index, needs_bind);
if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
++binding_index;
}
});
}
template <class P>
void BufferCache<P>::BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32 binding_index,
bool needs_bind) {
const Binding& binding = uniform_buffers[stage][index];
const VAddr cpu_addr = binding.cpu_addr;
const u32 size = binding.size;
Buffer& buffer = slot_buffers[binding.buffer_id];
if constexpr (IS_OPENGL) {
if (size <= SKIP_CACHE_SIZE && !buffer.IsRegionGpuModified(cpu_addr, size)) {
if (runtime.HasFastBufferSubData()) {
// Fast path for Nvidia
if (!HasFastUniformBufferBound(stage, binding_index)) {
// We only have to bind when the currently bound buffer is not the fast version
fast_bound_uniform_buffers[stage] |= 1U << binding_index;
runtime.BindFastUniformBuffer(stage, binding_index, size);
}
const auto span = ImmediateBufferWithData(cpu_addr, size);
runtime.PushFastUniformBuffer(stage, binding_index, span);
} else {
// Stream buffer path to avoid stalling on non-Nvidia drivers
const auto span = runtime.BindMappedUniformBuffer(stage, binding_index, size);
cpu_memory.ReadBlockUnsafe(cpu_addr, span.data(), size);
}
return;
}
}
// Classic cached path
SynchronizeBuffer(buffer, cpu_addr, size);
if (!needs_bind && !HasFastUniformBufferBound(stage, binding_index)) {
// Skip binding if it's not needed and if the bound buffer is not the fast version
// This exists to avoid instances where the fast buffer is bound and a GPU write happens
return;
}
fast_bound_uniform_buffers[stage] &= ~(1U << binding_index);
const u32 offset = buffer.Offset(cpu_addr);
if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
runtime.BindUniformBuffer(stage, binding_index, buffer, offset, size);
} else {
runtime.BindUniformBuffer(buffer, offset, size);
}
}
template <class P>
void BufferCache<P>::BindHostGraphicsStorageBuffers(size_t stage) {
u32 binding_index = 0;
ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
const Binding& binding = storage_buffers[stage][index];
Buffer& buffer = slot_buffers[binding.buffer_id];
const u32 size = binding.size;
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
const bool is_written = ((written_storage_buffers[stage] >> index) & 1) != 0;
if constexpr (NEEDS_BIND_STORAGE_INDEX) {
runtime.BindStorageBuffer(stage, binding_index, buffer, offset, size, is_written);
++binding_index;
} else {
runtime.BindStorageBuffer(buffer, offset, size, is_written);
}
});
}
template <class P>
void BufferCache<P>::BindHostTransformFeedbackBuffers() {
if (maxwell3d.regs.tfb_enabled == 0) {
return;
}
for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
const Binding& binding = transform_feedback_buffers[index];
Buffer& buffer = slot_buffers[binding.buffer_id];
const u32 size = binding.size;
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
runtime.BindTransformFeedbackBuffer(index, buffer, offset, size);
}
}
template <class P>
void BufferCache<P>::BindHostComputeUniformBuffers() {
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
// Mark all uniform buffers as dirty
dirty_uniform_buffers.fill(~u32{0});
}
u32 binding_index = 0;
ForEachEnabledBit(enabled_compute_uniform_buffers, [&](u32 index) {
const Binding& binding = compute_uniform_buffers[index];
Buffer& buffer = slot_buffers[binding.buffer_id];
const u32 size = binding.size;
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
if constexpr (NEEDS_BIND_UNIFORM_INDEX) {
runtime.BindComputeUniformBuffer(binding_index, buffer, offset, size);
++binding_index;
} else {
runtime.BindUniformBuffer(buffer, offset, size);
}
});
}
template <class P>
void BufferCache<P>::BindHostComputeStorageBuffers() {
u32 binding_index = 0;
ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
const Binding& binding = compute_storage_buffers[index];
Buffer& buffer = slot_buffers[binding.buffer_id];
const u32 size = binding.size;
SynchronizeBuffer(buffer, binding.cpu_addr, size);
const u32 offset = buffer.Offset(binding.cpu_addr);
const bool is_written = ((written_compute_storage_buffers >> index) & 1) != 0;
if constexpr (NEEDS_BIND_STORAGE_INDEX) {
runtime.BindComputeStorageBuffer(binding_index, buffer, offset, size, is_written);
++binding_index;
} else {
runtime.BindStorageBuffer(buffer, offset, size, is_written);
}
});
}
template <class P>
void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
if (is_indexed) {
UpdateIndexBuffer();
}
UpdateVertexBuffers();
UpdateTransformFeedbackBuffers();
for (size_t stage = 0; stage < NUM_STAGES; ++stage) {
UpdateUniformBuffers(stage);
UpdateStorageBuffers(stage);
}
}
template <class P>
void BufferCache<P>::DoUpdateComputeBuffers() {
UpdateComputeUniformBuffers();
UpdateComputeStorageBuffers();
}
template <class P>
void BufferCache<P>::UpdateIndexBuffer() {
// We have to check for the dirty flags and index count
// The index count is currently changed without updating the dirty flags
const auto& index_array = maxwell3d.regs.index_array;
auto& flags = maxwell3d.dirty.flags;
if (!flags[Dirty::IndexBuffer] && last_index_count == index_array.count) {
return;
}
flags[Dirty::IndexBuffer] = false;
last_index_count = index_array.count;
const GPUVAddr gpu_addr_begin = index_array.StartAddress();
const GPUVAddr gpu_addr_end = index_array.EndAddress();
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr_begin);
const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
const u32 draw_size = index_array.count * index_array.FormatSizeInBytes();
const u32 size = std::min(address_size, draw_size);
if (size == 0 || !cpu_addr) {
index_buffer = NULL_BINDING;
return;
}
index_buffer = Binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = FindBuffer(*cpu_addr, size),
};
}
template <class P>
void BufferCache<P>::UpdateVertexBuffers() {
auto& flags = maxwell3d.dirty.flags;
if (!maxwell3d.dirty.flags[Dirty::VertexBuffers]) {
return;
}
flags[Dirty::VertexBuffers] = false;
for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
UpdateVertexBuffer(index);
}
}
template <class P>
void BufferCache<P>::UpdateVertexBuffer(u32 index) {
if (!maxwell3d.dirty.flags[Dirty::VertexBuffer0 + index]) {
return;
}
const auto& array = maxwell3d.regs.vertex_array[index];
const auto& limit = maxwell3d.regs.vertex_array_limit[index];
const GPUVAddr gpu_addr_begin = array.StartAddress();
const GPUVAddr gpu_addr_end = limit.LimitAddress() + 1;
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr_begin);
const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
const u32 size = address_size; // TODO: Analyze stride and number of vertices
if (array.enable == 0 || size == 0 || !cpu_addr) {
vertex_buffers[index] = NULL_BINDING;
return;
}
vertex_buffers[index] = Binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = FindBuffer(*cpu_addr, size),
};
}
template <class P>
void BufferCache<P>::UpdateUniformBuffers(size_t stage) {
ForEachEnabledBit(enabled_uniform_buffers[stage], [&](u32 index) {
Binding& binding = uniform_buffers[stage][index];
if (binding.buffer_id) {
// Already updated
return;
}
// Mark as dirty
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
dirty_uniform_buffers[stage] |= 1U << index;
}
// Resolve buffer
binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
});
}
template <class P>
void BufferCache<P>::UpdateStorageBuffers(size_t stage) {
const u32 written_mask = written_storage_buffers[stage];
ForEachEnabledBit(enabled_storage_buffers[stage], [&](u32 index) {
// Resolve buffer
Binding& binding = storage_buffers[stage][index];
const BufferId buffer_id = FindBuffer(binding.cpu_addr, binding.size);
binding.buffer_id = buffer_id;
// Mark buffer as written if needed
if (((written_mask >> index) & 1) != 0) {
MarkWrittenBuffer(buffer_id, binding.cpu_addr, binding.size);
}
});
}
template <class P>
void BufferCache<P>::UpdateTransformFeedbackBuffers() {
if (maxwell3d.regs.tfb_enabled == 0) {
return;
}
for (u32 index = 0; index < NUM_TRANSFORM_FEEDBACK_BUFFERS; ++index) {
UpdateTransformFeedbackBuffer(index);
}
}
template <class P>
void BufferCache<P>::UpdateTransformFeedbackBuffer(u32 index) {
const auto& binding = maxwell3d.regs.tfb_bindings[index];
const GPUVAddr gpu_addr = binding.Address() + binding.buffer_offset;
const u32 size = binding.buffer_size;
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr);
if (binding.buffer_enable == 0 || size == 0 || !cpu_addr) {
transform_feedback_buffers[index] = NULL_BINDING;
return;
}
const BufferId buffer_id = FindBuffer(*cpu_addr, size);
transform_feedback_buffers[index] = Binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = buffer_id,
};
MarkWrittenBuffer(buffer_id, *cpu_addr, size);
}
template <class P>
void BufferCache<P>::UpdateComputeUniformBuffers() {
ForEachEnabledBit(enabled_compute_uniform_buffers, [&](u32 index) {
Binding& binding = compute_uniform_buffers[index];
binding = NULL_BINDING;
const auto& launch_desc = kepler_compute.launch_description;
if (((launch_desc.const_buffer_enable_mask >> index) & 1) != 0) {
const auto& cbuf = launch_desc.const_buffer_config[index];
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(cbuf.Address());
if (cpu_addr) {
binding.cpu_addr = *cpu_addr;
binding.size = cbuf.size;
}
}
binding.buffer_id = FindBuffer(binding.cpu_addr, binding.size);
});
}
template <class P>
void BufferCache<P>::UpdateComputeStorageBuffers() {
ForEachEnabledBit(enabled_compute_storage_buffers, [&](u32 index) {
// Resolve buffer
Binding& binding = compute_storage_buffers[index];
const BufferId buffer_id = FindBuffer(binding.cpu_addr, binding.size);
binding.buffer_id = buffer_id;
// Mark as written if needed
if (((written_compute_storage_buffers >> index) & 1) != 0) {
MarkWrittenBuffer(buffer_id, binding.cpu_addr, binding.size);
}
});
}
template <class P>
void BufferCache<P>::MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size) {
Buffer& buffer = slot_buffers[buffer_id];
buffer.MarkRegionAsGpuModified(cpu_addr, size);
const bool is_accuracy_high = Settings::IsGPULevelHigh();
const bool is_async = Settings::values.use_asynchronous_gpu_emulation.GetValue();
if (!is_accuracy_high || !is_async) {
return;
}
if (std::ranges::find(uncommitted_downloads, buffer_id) != uncommitted_downloads.end()) {
// Already inserted
return;
}
uncommitted_downloads.push_back(buffer_id);
}
template <class P>
BufferId BufferCache<P>::FindBuffer(VAddr cpu_addr, u32 size) {
if (cpu_addr == 0) {
return NULL_BUFFER_ID;
}
const u64 page = cpu_addr >> PAGE_BITS;
const BufferId buffer_id = page_table[page];
if (!buffer_id) {
return CreateBuffer(cpu_addr, size);
}
const Buffer& buffer = slot_buffers[buffer_id];
if (buffer.IsInBounds(cpu_addr, size)) {
return buffer_id;
}
return CreateBuffer(cpu_addr, size);
}
template <class P>
BufferId BufferCache<P>::CreateBuffer(VAddr cpu_addr, u32 wanted_size) {
std::vector<BufferId> overlap_ids;
VAddr cpu_addr_begin = cpu_addr;
VAddr cpu_addr_end = cpu_addr + wanted_size;
for (; cpu_addr >> PAGE_BITS < Common::DivCeil(cpu_addr_end, PAGE_SIZE);
cpu_addr += PAGE_SIZE) {
const BufferId overlap_id = page_table[cpu_addr >> PAGE_BITS];
if (!overlap_id) {
continue;
}
Buffer& overlap = slot_buffers[overlap_id];
if (overlap.IsPicked()) {
continue;
}
overlap.Pick();
overlap_ids.push_back(overlap_id);
const VAddr overlap_cpu_addr = overlap.CpuAddr();
if (overlap_cpu_addr < cpu_addr_begin) {
cpu_addr = cpu_addr_begin = overlap_cpu_addr;
}
cpu_addr_end = std::max(cpu_addr_end, overlap_cpu_addr + overlap.SizeBytes());
}
const u32 size = static_cast<u32>(cpu_addr_end - cpu_addr_begin);
const BufferId new_buffer_id = slot_buffers.insert(runtime, rasterizer, cpu_addr_begin, size);
Buffer& new_buffer = slot_buffers[new_buffer_id];
for (const BufferId overlap_id : overlap_ids) {
Buffer& overlap = slot_buffers[overlap_id];
overlap.Unpick();
std::vector<BufferCopy> copies;
const size_t dst_base_offset = overlap.CpuAddr() - new_buffer.CpuAddr();
overlap.ForEachDownloadRange([&](u64 begin, u64 range_size) {
copies.push_back(BufferCopy{
.src_offset = begin,
.dst_offset = dst_base_offset + begin,
.size = range_size,
});
new_buffer.UnmarkRegionAsCpuModified(begin, range_size);
new_buffer.MarkRegionAsGpuModified(begin, range_size);
});
if (!copies.empty()) {
runtime.CopyBuffer(slot_buffers[new_buffer_id], overlap, copies);
}
ReplaceBufferDownloads(overlap_id, new_buffer_id);
DeleteBuffer(overlap_id);
}
Register(new_buffer_id);
return new_buffer_id;
}
template <class P>
void BufferCache<P>::Register(BufferId buffer_id) {
ChangeRegister<true>(buffer_id);
}
template <class P>
void BufferCache<P>::Unregister(BufferId buffer_id) {
ChangeRegister<false>(buffer_id);
}
template <class P>
template <bool insert>
void BufferCache<P>::ChangeRegister(BufferId buffer_id) {
const Buffer& buffer = slot_buffers[buffer_id];
const VAddr cpu_addr_begin = buffer.CpuAddr();
const VAddr cpu_addr_end = cpu_addr_begin + buffer.SizeBytes();
const u64 page_begin = cpu_addr_begin / PAGE_SIZE;
const u64 page_end = Common::DivCeil(cpu_addr_end, PAGE_SIZE);
for (u64 page = page_begin; page != page_end; ++page) {
if constexpr (insert) {
page_table[page] = buffer_id;
} else {
page_table[page] = BufferId{};
}
}
}
template <class P>
void BufferCache<P>::SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size) {
if (buffer.CpuAddr() == 0) {
return;
}
SynchronizeBufferImpl(buffer, cpu_addr, size);
}
template <class P>
void BufferCache<P>::SynchronizeBufferImpl(Buffer& buffer, VAddr cpu_addr, u32 size) {
boost::container::small_vector<BufferCopy, 4> copies;
u64 total_size_bytes = 0;
u64 largest_copy = 0;
buffer.ForEachUploadRange(cpu_addr, size, [&](u64 range_offset, u64 range_size) {
copies.push_back(BufferCopy{
.src_offset = total_size_bytes,
.dst_offset = range_offset,
.size = range_size,
});
total_size_bytes += range_size;
largest_copy = std::max(largest_copy, range_size);
});
if (total_size_bytes == 0) {
return;
}
const std::span<BufferCopy> copies_span(copies.data(), copies.size());
UploadMemory(buffer, total_size_bytes, largest_copy, copies_span);
}
template <class P>
void BufferCache<P>::UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
std::span<BufferCopy> copies) {
if constexpr (USE_MEMORY_MAPS) {
MappedUploadMemory(buffer, total_size_bytes, copies);
} else {
ImmediateUploadMemory(buffer, largest_copy, copies);
}
}
template <class P>
void BufferCache<P>::ImmediateUploadMemory(Buffer& buffer, u64 largest_copy,
std::span<const BufferCopy> copies) {
std::span<u8> immediate_buffer;
for (const BufferCopy& copy : copies) {
std::span<const u8> upload_span;
const VAddr cpu_addr = buffer.CpuAddr() + copy.dst_offset;
if (IsRangeGranular(cpu_addr, copy.size)) {
upload_span = std::span(cpu_memory.GetPointer(cpu_addr), copy.size);
} else {
if (immediate_buffer.empty()) {
immediate_buffer = ImmediateBuffer(largest_copy);
}
cpu_memory.ReadBlockUnsafe(cpu_addr, immediate_buffer.data(), copy.size);
upload_span = immediate_buffer.subspan(0, copy.size);
}
buffer.ImmediateUpload(copy.dst_offset, upload_span);
}
}
template <class P>
void BufferCache<P>::MappedUploadMemory(Buffer& buffer, u64 total_size_bytes,
std::span<const BufferCopy> copies) {
auto upload_staging = runtime.UploadStagingBuffer(total_size_bytes);
const std::span<u8> staging_pointer = upload_staging.mapped_span;
for (const BufferCopy& copy : copies) {
const VAddr cpu_addr = buffer.CpuAddr() + copy.dst_offset;
u8* const src_pointer = staging_pointer.data() + copy.src_offset;
cpu_memory.ReadBlockUnsafe(cpu_addr, src_pointer, copy.size);
}
runtime.CopyBuffer(buffer, upload_staging.buffer, copies);
}
template <class P>
void BufferCache<P>::DeleteBuffer(BufferId buffer_id) {
const auto scalar_replace = [buffer_id](Binding& binding) {
if (binding.buffer_id == buffer_id) {
binding.buffer_id = BufferId{};
}
};
const auto replace = [scalar_replace](std::span<Binding> bindings) {
std::ranges::for_each(bindings, scalar_replace);
};
scalar_replace(index_buffer);
replace(vertex_buffers);
std::ranges::for_each(uniform_buffers, replace);
std::ranges::for_each(storage_buffers, replace);
replace(transform_feedback_buffers);
replace(compute_uniform_buffers);
replace(compute_storage_buffers);
std::erase(cached_write_buffer_ids, buffer_id);
// Mark the whole buffer as CPU written to stop tracking CPU writes
Buffer& buffer = slot_buffers[buffer_id];
buffer.MarkRegionAsCpuModified(buffer.CpuAddr(), buffer.SizeBytes());
Unregister(buffer_id);
delayed_destruction_ring.Push(std::move(slot_buffers[buffer_id]));
NotifyBufferDeletion();
}
template <class P>
void BufferCache<P>::ReplaceBufferDownloads(BufferId old_buffer_id, BufferId new_buffer_id) {
const auto replace = [old_buffer_id, new_buffer_id](std::vector<BufferId>& buffers) {
std::ranges::replace(buffers, old_buffer_id, new_buffer_id);
if (auto it = std::ranges::find(buffers, new_buffer_id); it != buffers.end()) {
buffers.erase(std::remove(it + 1, buffers.end(), new_buffer_id), buffers.end());
}
};
replace(uncommitted_downloads);
std::ranges::for_each(committed_downloads, replace);
}
template <class P>
void BufferCache<P>::NotifyBufferDeletion() {
if constexpr (HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS) {
dirty_uniform_buffers.fill(~u32{0});
}
auto& flags = maxwell3d.dirty.flags;
flags[Dirty::IndexBuffer] = true;
flags[Dirty::VertexBuffers] = true;
for (u32 index = 0; index < NUM_VERTEX_BUFFERS; ++index) {
flags[Dirty::VertexBuffer0 + index] = true;
}
has_deleted_buffers = true;
}
template <class P>
typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr ssbo_addr) const {
const GPUVAddr gpu_addr = gpu_memory.Read<u64>(ssbo_addr);
const u32 size = gpu_memory.Read<u32>(ssbo_addr + 8);
const std::optional<VAddr> cpu_addr = gpu_memory.GpuToCpuAddress(gpu_addr);
if (!cpu_addr || size == 0) {
return NULL_BINDING;
}
const Binding binding{
.cpu_addr = *cpu_addr,
.size = size,
.buffer_id = BufferId{},
};
return binding;
}
template <class P>
std::span<const u8> BufferCache<P>::ImmediateBufferWithData(VAddr cpu_addr, size_t size) {
u8* const base_pointer = cpu_memory.GetPointer(cpu_addr);
if (IsRangeGranular(cpu_addr, size) ||
base_pointer + size == cpu_memory.GetPointer(cpu_addr + size)) {
return std::span(base_pointer, size);
} else {
const std::span<u8> span = ImmediateBuffer(size);
cpu_memory.ReadBlockUnsafe(cpu_addr, span.data(), size);
return span;
}
}
template <class P>
std::span<u8> BufferCache<P>::ImmediateBuffer(size_t wanted_capacity) {
if (wanted_capacity > immediate_buffer_capacity) {
immediate_buffer_capacity = wanted_capacity;
immediate_buffer_alloc = std::make_unique<u8[]>(wanted_capacity);
}
return std::span<u8>(immediate_buffer_alloc.get(), wanted_capacity);
}
template <class P>
bool BufferCache<P>::HasFastUniformBufferBound(size_t stage, u32 binding_index) const noexcept {
if constexpr (IS_OPENGL) {
return ((fast_bound_uniform_buffers[stage] >> binding_index) & 1) != 0;
} else {
// Only OpenGL has fast uniform buffers
return false;
}
}
} // namespace VideoCommon
|