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
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
|
// 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 <list>
#include <memory>
#include <mutex>
#include <set>
#include <tuple>
#include <unordered_map>
#include <vector>
#include <boost/container/small_vector.hpp>
#include <boost/icl/interval_map.hpp>
#include <boost/range/iterator_range.hpp>
#include "common/assert.h"
#include "common/common_types.h"
#include "common/math_util.h"
#include "core/core.h"
#include "core/memory.h"
#include "core/settings.h"
#include "video_core/dirty_flags.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/engines/maxwell_3d.h"
#include "video_core/gpu.h"
#include "video_core/memory_manager.h"
#include "video_core/rasterizer_interface.h"
#include "video_core/surface.h"
#include "video_core/texture_cache/copy_params.h"
#include "video_core/texture_cache/format_lookup_table.h"
#include "video_core/texture_cache/surface_base.h"
#include "video_core/texture_cache/surface_params.h"
#include "video_core/texture_cache/surface_view.h"
namespace Tegra::Texture {
struct FullTextureInfo;
}
namespace VideoCore {
class RasterizerInterface;
}
namespace VideoCommon {
using VideoCore::Surface::PixelFormat;
using VideoCore::Surface::SurfaceTarget;
using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig;
template <typename TSurface, typename TView>
class TextureCache {
using VectorSurface = boost::container::small_vector<TSurface, 1>;
public:
void InvalidateRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
for (const auto& surface : GetSurfacesInRegion(addr, size)) {
Unregister(surface);
}
}
void OnCPUWrite(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
for (const auto& surface : GetSurfacesInRegion(addr, size)) {
if (surface->IsMemoryMarked()) {
UnmarkMemory(surface);
surface->SetSyncPending(true);
marked_for_unregister.emplace_back(surface);
}
}
}
void SyncGuestHost() {
std::lock_guard lock{mutex};
for (const auto& surface : marked_for_unregister) {
if (surface->IsRegistered()) {
surface->SetSyncPending(false);
Unregister(surface);
}
}
marked_for_unregister.clear();
}
/**
* Guarantees that rendertargets don't unregister themselves if the
* collide. Protection is currently only done on 3D slices.
*/
void GuardRenderTargets(bool new_guard) {
guard_render_targets = new_guard;
}
void GuardSamplers(bool new_guard) {
guard_samplers = new_guard;
}
void FlushRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
auto surfaces = GetSurfacesInRegion(addr, size);
if (surfaces.empty()) {
return;
}
std::sort(surfaces.begin(), surfaces.end(), [](const TSurface& a, const TSurface& b) {
return a->GetModificationTick() < b->GetModificationTick();
});
for (const auto& surface : surfaces) {
mutex.unlock();
FlushSurface(surface);
mutex.lock();
}
}
bool MustFlushRegion(VAddr addr, std::size_t size) {
std::lock_guard lock{mutex};
const auto surfaces = GetSurfacesInRegion(addr, size);
return std::any_of(surfaces.cbegin(), surfaces.cend(),
[](const TSurface& surface) { return surface->IsModified(); });
}
TView GetTextureSurface(const Tegra::Texture::TICEntry& tic,
const VideoCommon::Shader::Sampler& entry) {
std::lock_guard lock{mutex};
const auto gpu_addr{tic.Address()};
if (!gpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
if (!IsTypeCompatible(tic.texture_type, entry)) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
const auto params{SurfaceParams::CreateForTexture(format_lookup_table, tic, entry)};
const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, true, false);
if (guard_samplers) {
sampled_textures.push_back(surface);
}
return view;
}
TView GetImageSurface(const Tegra::Texture::TICEntry& tic,
const VideoCommon::Shader::Image& entry) {
std::lock_guard lock{mutex};
const auto gpu_addr{tic.Address()};
if (!gpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
}
const auto params{SurfaceParams::CreateForImage(format_lookup_table, tic, entry)};
const auto [surface, view] = GetSurface(gpu_addr, *cpu_addr, params, true, false);
if (guard_samplers) {
sampled_textures.push_back(surface);
}
return view;
}
bool TextureBarrier() {
const bool any_rt =
std::any_of(sampled_textures.begin(), sampled_textures.end(),
[](const auto& surface) { return surface->IsRenderTarget(); });
sampled_textures.clear();
return any_rt;
}
TView GetDepthBufferSurface(bool preserve_contents) {
std::lock_guard lock{mutex};
auto& maxwell3d = system.GPU().Maxwell3D();
if (!maxwell3d.dirty.flags[VideoCommon::Dirty::ZetaBuffer]) {
return depth_buffer.view;
}
maxwell3d.dirty.flags[VideoCommon::Dirty::ZetaBuffer] = false;
const auto& regs{maxwell3d.regs};
const auto gpu_addr{regs.zeta.Address()};
if (!gpu_addr || !regs.zeta_enable) {
SetEmptyDepthBuffer();
return {};
}
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
SetEmptyDepthBuffer();
return {};
}
const auto depth_params{SurfaceParams::CreateForDepthBuffer(system)};
auto surface_view = GetSurface(gpu_addr, *cpu_addr, depth_params, preserve_contents, true);
if (depth_buffer.target)
depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
depth_buffer.target = surface_view.first;
depth_buffer.view = surface_view.second;
if (depth_buffer.target)
depth_buffer.target->MarkAsRenderTarget(true, DEPTH_RT);
return surface_view.second;
}
TView GetColorBufferSurface(std::size_t index, bool preserve_contents) {
std::lock_guard lock{mutex};
ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets);
auto& maxwell3d = system.GPU().Maxwell3D();
if (!maxwell3d.dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index]) {
return render_targets[index].view;
}
maxwell3d.dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index] = false;
const auto& regs{maxwell3d.regs};
if (index >= regs.rt_control.count || regs.rt[index].Address() == 0 ||
regs.rt[index].format == Tegra::RenderTargetFormat::NONE) {
SetEmptyColorBuffer(index);
return {};
}
const auto& config{regs.rt[index]};
const auto gpu_addr{config.Address()};
if (!gpu_addr) {
SetEmptyColorBuffer(index);
return {};
}
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
SetEmptyColorBuffer(index);
return {};
}
auto surface_view =
GetSurface(gpu_addr, *cpu_addr, SurfaceParams::CreateForFramebuffer(system, index),
preserve_contents, true);
if (render_targets[index].target) {
auto& surface = render_targets[index].target;
surface->MarkAsRenderTarget(false, NO_RT);
const auto& cr_params = surface->GetSurfaceParams();
if (!cr_params.is_tiled && Settings::values.use_asynchronous_gpu_emulation) {
AsyncFlushSurface(surface);
}
}
render_targets[index].target = surface_view.first;
render_targets[index].view = surface_view.second;
if (render_targets[index].target)
render_targets[index].target->MarkAsRenderTarget(true, static_cast<u32>(index));
return surface_view.second;
}
void MarkColorBufferInUse(std::size_t index) {
if (auto& render_target = render_targets[index].target) {
render_target->MarkAsModified(true, Tick());
}
}
void MarkDepthBufferInUse() {
if (depth_buffer.target) {
depth_buffer.target->MarkAsModified(true, Tick());
}
}
void SetEmptyDepthBuffer() {
if (depth_buffer.target == nullptr) {
return;
}
depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
depth_buffer.target = nullptr;
depth_buffer.view = nullptr;
}
void SetEmptyColorBuffer(std::size_t index) {
if (render_targets[index].target == nullptr) {
return;
}
render_targets[index].target->MarkAsRenderTarget(false, NO_RT);
render_targets[index].target = nullptr;
render_targets[index].view = nullptr;
}
void DoFermiCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
const Tegra::Engines::Fermi2D::Regs::Surface& dst_config,
const Tegra::Engines::Fermi2D::Config& copy_config) {
std::lock_guard lock{mutex};
SurfaceParams src_params = SurfaceParams::CreateForFermiCopySurface(src_config);
SurfaceParams dst_params = SurfaceParams::CreateForFermiCopySurface(dst_config);
const GPUVAddr src_gpu_addr = src_config.Address();
const GPUVAddr dst_gpu_addr = dst_config.Address();
DeduceBestBlit(src_params, dst_params, src_gpu_addr, dst_gpu_addr);
const std::optional<VAddr> dst_cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(dst_gpu_addr);
const std::optional<VAddr> src_cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(src_gpu_addr);
std::pair<TSurface, TView> dst_surface =
GetSurface(dst_gpu_addr, *dst_cpu_addr, dst_params, true, false);
std::pair<TSurface, TView> src_surface =
GetSurface(src_gpu_addr, *src_cpu_addr, src_params, true, false);
ImageBlit(src_surface.second, dst_surface.second, copy_config);
dst_surface.first->MarkAsModified(true, Tick());
}
TSurface TryFindFramebufferSurface(VAddr addr) {
if (!addr) {
return nullptr;
}
const VAddr page = addr >> registry_page_bits;
std::vector<TSurface>& list = registry[page];
for (auto& surface : list) {
if (surface->GetCpuAddr() == addr) {
return surface;
}
}
return nullptr;
}
u64 Tick() {
return ++ticks;
}
void CommitAsyncFlushes() {
committed_flushes.push_back(uncommitted_flushes);
uncommitted_flushes.reset();
}
bool HasUncommittedFlushes() const {
return uncommitted_flushes != nullptr;
}
bool ShouldWaitAsyncFlushes() const {
return !committed_flushes.empty() && committed_flushes.front() != nullptr;
}
void PopAsyncFlushes() {
if (committed_flushes.empty()) {
return;
}
auto& flush_list = committed_flushes.front();
if (!flush_list) {
committed_flushes.pop_front();
return;
}
for (TSurface& surface : *flush_list) {
FlushSurface(surface);
}
committed_flushes.pop_front();
}
protected:
explicit TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
bool is_astc_supported)
: system{system}, is_astc_supported{is_astc_supported}, rasterizer{rasterizer} {
for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
SetEmptyColorBuffer(i);
}
SetEmptyDepthBuffer();
staging_cache.SetSize(2);
const auto make_siblings = [this](PixelFormat a, PixelFormat b) {
siblings_table[static_cast<std::size_t>(a)] = b;
siblings_table[static_cast<std::size_t>(b)] = a;
};
std::fill(siblings_table.begin(), siblings_table.end(), PixelFormat::Invalid);
make_siblings(PixelFormat::Z16, PixelFormat::R16U);
make_siblings(PixelFormat::Z32F, PixelFormat::R32F);
make_siblings(PixelFormat::Z32FS8, PixelFormat::RG32F);
sampled_textures.reserve(64);
}
~TextureCache() = default;
virtual TSurface CreateSurface(GPUVAddr gpu_addr, const SurfaceParams& params) = 0;
virtual void ImageCopy(TSurface& src_surface, TSurface& dst_surface,
const CopyParams& copy_params) = 0;
virtual void ImageBlit(TView& src_view, TView& dst_view,
const Tegra::Engines::Fermi2D::Config& copy_config) = 0;
// Depending on the backend, a buffer copy can be slow as it means deoptimizing the texture
// and reading it from a separate buffer.
virtual void BufferCopy(TSurface& src_surface, TSurface& dst_surface) = 0;
void ManageRenderTargetUnregister(TSurface& surface) {
auto& dirty = system.GPU().Maxwell3D().dirty;
const u32 index = surface->GetRenderTarget();
if (index == DEPTH_RT) {
dirty.flags[VideoCommon::Dirty::ZetaBuffer] = true;
} else {
dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index] = true;
}
dirty.flags[VideoCommon::Dirty::RenderTargets] = true;
}
void Register(TSurface surface) {
const GPUVAddr gpu_addr = surface->GetGpuAddr();
const std::size_t size = surface->GetSizeInBytes();
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
LOG_CRITICAL(HW_GPU, "Failed to register surface with unmapped gpu_address 0x{:016x}",
gpu_addr);
return;
}
surface->SetCpuAddr(*cpu_addr);
RegisterInnerCache(surface);
surface->MarkAsRegistered(true);
surface->SetMemoryMarked(true);
rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1);
}
void UnmarkMemory(TSurface surface) {
if (!surface->IsMemoryMarked()) {
return;
}
const std::size_t size = surface->GetSizeInBytes();
const VAddr cpu_addr = surface->GetCpuAddr();
rasterizer.UpdatePagesCachedCount(cpu_addr, size, -1);
surface->SetMemoryMarked(false);
}
void Unregister(TSurface surface) {
if (guard_render_targets && surface->IsProtected()) {
return;
}
if (!guard_render_targets && surface->IsRenderTarget()) {
ManageRenderTargetUnregister(surface);
}
UnmarkMemory(surface);
if (surface->IsSyncPending()) {
marked_for_unregister.remove(surface);
surface->SetSyncPending(false);
}
UnregisterInnerCache(surface);
surface->MarkAsRegistered(false);
ReserveSurface(surface->GetSurfaceParams(), surface);
}
TSurface GetUncachedSurface(const GPUVAddr gpu_addr, const SurfaceParams& params) {
if (const auto surface = TryGetReservedSurface(params); surface) {
surface->SetGpuAddr(gpu_addr);
return surface;
}
// No reserved surface available, create a new one and reserve it
auto new_surface{CreateSurface(gpu_addr, params)};
return new_surface;
}
Core::System& system;
const bool is_astc_supported;
private:
enum class RecycleStrategy : u32 {
Ignore = 0,
Flush = 1,
BufferCopy = 3,
};
enum class DeductionType : u32 {
DeductionComplete,
DeductionIncomplete,
DeductionFailed,
};
struct Deduction {
DeductionType type{DeductionType::DeductionFailed};
TSurface surface{};
bool Failed() const {
return type == DeductionType::DeductionFailed;
}
bool Incomplete() const {
return type == DeductionType::DeductionIncomplete;
}
bool IsDepth() const {
return surface->GetSurfaceParams().IsPixelFormatZeta();
}
};
/**
* Takes care of selecting a proper strategy to deal with a texture recycle.
*
* @param overlaps The overlapping surfaces registered in the cache.
* @param params The parameters on the new surface.
* @param gpu_addr The starting address of the new surface.
* @param untopological Indicates to the recycler that the texture has no way
* to match the overlaps due to topological reasons.
**/
RecycleStrategy PickStrategy(VectorSurface& overlaps, const SurfaceParams& params,
const GPUVAddr gpu_addr, const MatchTopologyResult untopological) {
if (Settings::IsGPULevelExtreme()) {
return RecycleStrategy::Flush;
}
// 3D Textures decision
if (params.block_depth > 1 || params.target == SurfaceTarget::Texture3D) {
return RecycleStrategy::Flush;
}
for (const auto& s : overlaps) {
const auto& s_params = s->GetSurfaceParams();
if (s_params.block_depth > 1 || s_params.target == SurfaceTarget::Texture3D) {
return RecycleStrategy::Flush;
}
}
// Untopological decision
if (untopological == MatchTopologyResult::CompressUnmatch) {
return RecycleStrategy::Flush;
}
if (untopological == MatchTopologyResult::FullMatch && !params.is_tiled) {
return RecycleStrategy::Flush;
}
return RecycleStrategy::Ignore;
}
/**
* Used to decide what to do with textures we can't resolve in the cache It has 2 implemented
* strategies: Ignore and Flush.
*
* - Ignore: Just unregisters all the overlaps and loads the new texture.
* - Flush: Flushes all the overlaps into memory and loads the new surface from that data.
*
* @param overlaps The overlapping surfaces registered in the cache.
* @param params The parameters for the new surface.
* @param gpu_addr The starting address of the new surface.
* @param preserve_contents Indicates that the new surface should be loaded from memory or left
* blank.
* @param untopological Indicates to the recycler that the texture has no way to match the
* overlaps due to topological reasons.
**/
std::pair<TSurface, TView> RecycleSurface(VectorSurface& overlaps, const SurfaceParams& params,
const GPUVAddr gpu_addr, const bool preserve_contents,
const MatchTopologyResult untopological) {
const bool do_load = preserve_contents && Settings::IsGPULevelExtreme();
for (auto& surface : overlaps) {
Unregister(surface);
}
switch (PickStrategy(overlaps, params, gpu_addr, untopological)) {
case RecycleStrategy::Ignore: {
return InitializeSurface(gpu_addr, params, do_load);
}
case RecycleStrategy::Flush: {
std::sort(overlaps.begin(), overlaps.end(),
[](const TSurface& a, const TSurface& b) -> bool {
return a->GetModificationTick() < b->GetModificationTick();
});
for (auto& surface : overlaps) {
FlushSurface(surface);
}
return InitializeSurface(gpu_addr, params, preserve_contents);
}
case RecycleStrategy::BufferCopy: {
auto new_surface = GetUncachedSurface(gpu_addr, params);
BufferCopy(overlaps[0], new_surface);
return {new_surface, new_surface->GetMainView()};
}
default: {
UNIMPLEMENTED_MSG("Unimplemented Texture Cache Recycling Strategy!");
return InitializeSurface(gpu_addr, params, do_load);
}
}
}
/**
* Takes a single surface and recreates into another that may differ in
* format, target or width alignment.
*
* @param current_surface The registered surface in the cache which we want to convert.
* @param params The new surface params which we'll use to recreate the surface.
* @param is_render Whether or not the surface is a render target.
**/
std::pair<TSurface, TView> RebuildSurface(TSurface current_surface, const SurfaceParams& params,
bool is_render) {
const auto gpu_addr = current_surface->GetGpuAddr();
const auto& cr_params = current_surface->GetSurfaceParams();
TSurface new_surface;
if (cr_params.pixel_format != params.pixel_format && !is_render &&
GetSiblingFormat(cr_params.pixel_format) == params.pixel_format) {
SurfaceParams new_params = params;
new_params.pixel_format = cr_params.pixel_format;
new_params.type = cr_params.type;
new_surface = GetUncachedSurface(gpu_addr, new_params);
} else {
new_surface = GetUncachedSurface(gpu_addr, params);
}
const auto& final_params = new_surface->GetSurfaceParams();
if (cr_params.type != final_params.type) {
if (Settings::IsGPULevelExtreme()) {
BufferCopy(current_surface, new_surface);
}
} else {
std::vector<CopyParams> bricks = current_surface->BreakDown(final_params);
for (auto& brick : bricks) {
ImageCopy(current_surface, new_surface, brick);
}
}
Unregister(current_surface);
Register(new_surface);
new_surface->MarkAsModified(current_surface->IsModified(), Tick());
return {new_surface, new_surface->GetMainView()};
}
/**
* Takes a single surface and checks with the new surface's params if it's an exact
* match, we return the main view of the registered surface. If its formats don't
* match, we rebuild the surface. We call this last method a `Mirage`. If formats
* match but the targets don't, we create an overview View of the registered surface.
*
* @param current_surface The registered surface in the cache which we want to convert.
* @param params The new surface params which we want to check.
* @param is_render Whether or not the surface is a render target.
**/
std::pair<TSurface, TView> ManageStructuralMatch(TSurface current_surface,
const SurfaceParams& params, bool is_render) {
const bool is_mirage = !current_surface->MatchFormat(params.pixel_format);
const bool matches_target = current_surface->MatchTarget(params.target);
const auto match_check = [&]() -> std::pair<TSurface, TView> {
if (matches_target) {
return {current_surface, current_surface->GetMainView()};
}
return {current_surface, current_surface->EmplaceOverview(params)};
};
if (!is_mirage) {
return match_check();
}
if (!is_render && GetSiblingFormat(current_surface->GetFormat()) == params.pixel_format) {
return match_check();
}
return RebuildSurface(current_surface, params, is_render);
}
/**
* Unlike RebuildSurface where we know whether or not registered surfaces match the candidate
* in some way, we have no guarantees here. We try to see if the overlaps are sublayers/mipmaps
* of the new surface, if they all match we end up recreating a surface for them,
* else we return nothing.
*
* @param overlaps The overlapping surfaces registered in the cache.
* @param params The parameters on the new surface.
* @param gpu_addr The starting address of the new surface.
**/
std::optional<std::pair<TSurface, TView>> TryReconstructSurface(VectorSurface& overlaps,
const SurfaceParams& params,
const GPUVAddr gpu_addr) {
if (params.target == SurfaceTarget::Texture3D) {
return {};
}
bool modified = false;
TSurface new_surface = GetUncachedSurface(gpu_addr, params);
u32 passed_tests = 0;
for (auto& surface : overlaps) {
const SurfaceParams& src_params = surface->GetSurfaceParams();
if (src_params.is_layered || src_params.num_levels > 1) {
// We send this cases to recycle as they are more complex to handle
return {};
}
const std::size_t candidate_size = surface->GetSizeInBytes();
auto mipmap_layer{new_surface->GetLayerMipmap(surface->GetGpuAddr())};
if (!mipmap_layer) {
continue;
}
const auto [layer, mipmap] = *mipmap_layer;
if (new_surface->GetMipmapSize(mipmap) != candidate_size) {
continue;
}
modified |= surface->IsModified();
// Now we got all the data set up
const u32 width = SurfaceParams::IntersectWidth(src_params, params, 0, mipmap);
const u32 height = SurfaceParams::IntersectHeight(src_params, params, 0, mipmap);
const CopyParams copy_params(0, 0, 0, 0, 0, layer, 0, mipmap, width, height, 1);
passed_tests++;
ImageCopy(surface, new_surface, copy_params);
}
if (passed_tests == 0) {
return {};
// In Accurate GPU all tests should pass, else we recycle
} else if (Settings::IsGPULevelExtreme() && passed_tests != overlaps.size()) {
return {};
}
for (const auto& surface : overlaps) {
Unregister(surface);
}
new_surface->MarkAsModified(modified, Tick());
Register(new_surface);
return {{new_surface, new_surface->GetMainView()}};
}
/**
* Takes care of managing 3D textures and its slices. Does HLE methods for reconstructing the 3D
* textures within the GPU if possible. Falls back to LLE when it isn't possible to use any of
* the HLE methods.
*
* @param overlaps The overlapping surfaces registered in the cache.
* @param params The parameters on the new surface.
* @param gpu_addr The starting address of the new surface.
* @param cpu_addr The starting address of the new surface on physical memory.
* @param preserve_contents Indicates that the new surface should be loaded from memory or
* left blank.
*/
std::optional<std::pair<TSurface, TView>> Manage3DSurfaces(VectorSurface& overlaps,
const SurfaceParams& params,
const GPUVAddr gpu_addr,
const VAddr cpu_addr,
bool preserve_contents) {
if (params.target == SurfaceTarget::Texture3D) {
bool failed = false;
if (params.num_levels > 1) {
// We can't handle mipmaps in 3D textures yet, better fallback to LLE approach
return std::nullopt;
}
TSurface new_surface = GetUncachedSurface(gpu_addr, params);
bool modified = false;
for (auto& surface : overlaps) {
const SurfaceParams& src_params = surface->GetSurfaceParams();
if (src_params.target != SurfaceTarget::Texture2D) {
failed = true;
break;
}
if (src_params.height != params.height) {
failed = true;
break;
}
if (src_params.block_depth != params.block_depth ||
src_params.block_height != params.block_height) {
failed = true;
break;
}
const u32 offset = static_cast<u32>(surface->GetCpuAddr() - cpu_addr);
const auto offsets = params.GetBlockOffsetXYZ(offset);
const auto z = std::get<2>(offsets);
modified |= surface->IsModified();
const CopyParams copy_params(0, 0, 0, 0, 0, z, 0, 0, params.width, params.height,
1);
ImageCopy(surface, new_surface, copy_params);
}
if (failed) {
return std::nullopt;
}
for (const auto& surface : overlaps) {
Unregister(surface);
}
new_surface->MarkAsModified(modified, Tick());
Register(new_surface);
auto view = new_surface->GetMainView();
return {{std::move(new_surface), view}};
} else {
for (const auto& surface : overlaps) {
if (!surface->MatchTarget(params.target)) {
if (overlaps.size() == 1 && surface->GetCpuAddr() == cpu_addr) {
if (Settings::IsGPULevelExtreme()) {
return std::nullopt;
}
Unregister(surface);
return InitializeSurface(gpu_addr, params, preserve_contents);
}
return std::nullopt;
}
if (surface->GetCpuAddr() != cpu_addr) {
continue;
}
if (surface->MatchesStructure(params) == MatchStructureResult::FullMatch) {
return {{surface, surface->GetMainView()}};
}
}
return InitializeSurface(gpu_addr, params, preserve_contents);
}
}
/**
* Gets the starting address and parameters of a candidate surface and tries
* to find a matching surface within the cache. This is done in 3 big steps:
*
* 1. Check the 1st Level Cache in order to find an exact match, if we fail, we move to step 2.
*
* 2. Check if there are any overlaps at all, if there are none, we just load the texture from
* memory else we move to step 3.
*
* 3. Consists of figuring out the relationship between the candidate texture and the
* overlaps. We divide the scenarios depending if there's 1 or many overlaps. If
* there's many, we just try to reconstruct a new surface out of them based on the
* candidate's parameters, if we fail, we recycle. When there's only 1 overlap then we
* have to check if the candidate is a view (layer/mipmap) of the overlap or if the
* registered surface is a mipmap/layer of the candidate. In this last case we reconstruct
* a new surface.
*
* @param gpu_addr The starting address of the candidate surface.
* @param params The parameters on the candidate surface.
* @param preserve_contents Indicates that the new surface should be loaded from memory or
* left blank.
* @param is_render Whether or not the surface is a render target.
**/
std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const VAddr cpu_addr,
const SurfaceParams& params, bool preserve_contents,
bool is_render) {
// Step 1
// Check Level 1 Cache for a fast structural match. If candidate surface
// matches at certain level we are pretty much done.
if (const auto iter = l1_cache.find(cpu_addr); iter != l1_cache.end()) {
TSurface& current_surface = iter->second;
const auto topological_result = current_surface->MatchesTopology(params);
if (topological_result != MatchTopologyResult::FullMatch) {
VectorSurface overlaps{current_surface};
return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
topological_result);
}
const auto struct_result = current_surface->MatchesStructure(params);
if (struct_result != MatchStructureResult::None) {
const auto& old_params = current_surface->GetSurfaceParams();
const bool not_3d = params.target != SurfaceTarget::Texture3D &&
old_params.target != SurfaceTarget::Texture3D;
if (not_3d || current_surface->MatchTarget(params.target)) {
if (struct_result == MatchStructureResult::FullMatch) {
return ManageStructuralMatch(current_surface, params, is_render);
} else {
return RebuildSurface(current_surface, params, is_render);
}
}
}
}
// Step 2
// Obtain all possible overlaps in the memory region
const std::size_t candidate_size = params.GetGuestSizeInBytes();
auto overlaps{GetSurfacesInRegion(cpu_addr, candidate_size)};
// If none are found, we are done. we just load the surface and create it.
if (overlaps.empty()) {
return InitializeSurface(gpu_addr, params, preserve_contents);
}
// Step 3
// Now we need to figure the relationship between the texture and its overlaps
// we do a topological test to ensure we can find some relationship. If it fails
// immediately recycle the texture
for (const auto& surface : overlaps) {
const auto topological_result = surface->MatchesTopology(params);
if (topological_result != MatchTopologyResult::FullMatch) {
return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
topological_result);
}
}
// Check if it's a 3D texture
if (params.block_depth > 0) {
auto surface =
Manage3DSurfaces(overlaps, params, gpu_addr, cpu_addr, preserve_contents);
if (surface) {
return *surface;
}
}
// Split cases between 1 overlap or many.
if (overlaps.size() == 1) {
TSurface current_surface = overlaps[0];
// First check if the surface is within the overlap. If not, it means
// two things either the candidate surface is a supertexture of the overlap
// or they don't match in any known way.
if (!current_surface->IsInside(gpu_addr, gpu_addr + candidate_size)) {
if (current_surface->GetGpuAddr() == gpu_addr) {
std::optional<std::pair<TSurface, TView>> view =
TryReconstructSurface(overlaps, params, gpu_addr);
if (view) {
return *view;
}
}
return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
MatchTopologyResult::FullMatch);
}
// Now we check if the candidate is a mipmap/layer of the overlap
std::optional<TView> view =
current_surface->EmplaceView(params, gpu_addr, candidate_size);
if (view) {
const bool is_mirage = !current_surface->MatchFormat(params.pixel_format);
if (is_mirage) {
// On a mirage view, we need to recreate the surface under this new view
// and then obtain a view again.
SurfaceParams new_params = current_surface->GetSurfaceParams();
const u32 wh = SurfaceParams::ConvertWidth(
new_params.width, new_params.pixel_format, params.pixel_format);
const u32 hh = SurfaceParams::ConvertHeight(
new_params.height, new_params.pixel_format, params.pixel_format);
new_params.width = wh;
new_params.height = hh;
new_params.pixel_format = params.pixel_format;
std::pair<TSurface, TView> pair =
RebuildSurface(current_surface, new_params, is_render);
std::optional<TView> mirage_view =
pair.first->EmplaceView(params, gpu_addr, candidate_size);
if (mirage_view)
return {pair.first, *mirage_view};
return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
MatchTopologyResult::FullMatch);
}
return {current_surface, *view};
}
} else {
// If there are many overlaps, odds are they are subtextures of the candidate
// surface. We try to construct a new surface based on the candidate parameters,
// using the overlaps. If a single overlap fails, this will fail.
std::optional<std::pair<TSurface, TView>> view =
TryReconstructSurface(overlaps, params, gpu_addr);
if (view) {
return *view;
}
}
// We failed all the tests, recycle the overlaps into a new texture.
return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
MatchTopologyResult::FullMatch);
}
/**
* Gets the starting address and parameters of a candidate surface and tries to find a
* matching surface within the cache that's similar to it. If there are many textures
* or the texture found if entirely incompatible, it will fail. If no texture is found, the
* blit will be unsuccessful.
*
* @param gpu_addr The starting address of the candidate surface.
* @param params The parameters on the candidate surface.
**/
Deduction DeduceSurface(const GPUVAddr gpu_addr, const SurfaceParams& params) {
const std::optional<VAddr> cpu_addr =
system.GPU().MemoryManager().GpuToCpuAddress(gpu_addr);
if (!cpu_addr) {
Deduction result{};
result.type = DeductionType::DeductionFailed;
return result;
}
if (const auto iter = l1_cache.find(*cpu_addr); iter != l1_cache.end()) {
TSurface& current_surface = iter->second;
const auto topological_result = current_surface->MatchesTopology(params);
if (topological_result != MatchTopologyResult::FullMatch) {
Deduction result{};
result.type = DeductionType::DeductionFailed;
return result;
}
const auto struct_result = current_surface->MatchesStructure(params);
if (struct_result != MatchStructureResult::None &&
current_surface->MatchTarget(params.target)) {
Deduction result{};
result.type = DeductionType::DeductionComplete;
result.surface = current_surface;
return result;
}
}
const std::size_t candidate_size = params.GetGuestSizeInBytes();
auto overlaps{GetSurfacesInRegion(*cpu_addr, candidate_size)};
if (overlaps.empty()) {
Deduction result{};
result.type = DeductionType::DeductionIncomplete;
return result;
}
if (overlaps.size() > 1) {
Deduction result{};
result.type = DeductionType::DeductionFailed;
return result;
} else {
Deduction result{};
result.type = DeductionType::DeductionComplete;
result.surface = overlaps[0];
return result;
}
}
/**
* Gets a null surface based on a target texture.
* @param target The target of the null surface.
*/
TView GetNullSurface(SurfaceTarget target) {
const u32 i_target = static_cast<u32>(target);
if (const auto it = invalid_cache.find(i_target); it != invalid_cache.end()) {
return it->second->GetMainView();
}
SurfaceParams params{};
params.target = target;
params.is_tiled = false;
params.srgb_conversion = false;
params.is_layered = false;
params.block_width = 0;
params.block_height = 0;
params.block_depth = 0;
params.tile_width_spacing = 1;
params.width = 1;
params.height = 1;
params.depth = 1;
if (target == SurfaceTarget::TextureCubemap || target == SurfaceTarget::TextureCubeArray) {
params.depth = 6;
}
params.pitch = 4;
params.num_levels = 1;
params.emulated_levels = 1;
params.pixel_format = VideoCore::Surface::PixelFormat::R8U;
params.type = VideoCore::Surface::SurfaceType::ColorTexture;
auto surface = CreateSurface(0ULL, params);
invalid_memory.resize(surface->GetHostSizeInBytes(), 0U);
surface->UploadTexture(invalid_memory);
surface->MarkAsModified(false, Tick());
invalid_cache.emplace(i_target, surface);
return surface->GetMainView();
}
/**
* Gets the a source and destination starting address and parameters,
* and tries to deduce if they are supposed to be depth textures. If so, their
* parameters are modified and fixed into so.
*
* @param src_params The parameters of the candidate surface.
* @param dst_params The parameters of the destination surface.
* @param src_gpu_addr The starting address of the candidate surface.
* @param dst_gpu_addr The starting address of the destination surface.
**/
void DeduceBestBlit(SurfaceParams& src_params, SurfaceParams& dst_params,
const GPUVAddr src_gpu_addr, const GPUVAddr dst_gpu_addr) {
auto deduced_src = DeduceSurface(src_gpu_addr, src_params);
auto deduced_dst = DeduceSurface(src_gpu_addr, src_params);
if (deduced_src.Failed() || deduced_dst.Failed()) {
return;
}
const bool incomplete_src = deduced_src.Incomplete();
const bool incomplete_dst = deduced_dst.Incomplete();
if (incomplete_src && incomplete_dst) {
return;
}
const bool any_incomplete = incomplete_src || incomplete_dst;
if (!any_incomplete) {
if (!(deduced_src.IsDepth() && deduced_dst.IsDepth())) {
return;
}
} else {
if (incomplete_src && !(deduced_dst.IsDepth())) {
return;
}
if (incomplete_dst && !(deduced_src.IsDepth())) {
return;
}
}
const auto inherit_format = [](SurfaceParams& to, TSurface from) {
const SurfaceParams& params = from->GetSurfaceParams();
to.pixel_format = params.pixel_format;
to.type = params.type;
};
// Now we got the cases where one or both is Depth and the other is not known
if (!incomplete_src) {
inherit_format(src_params, deduced_src.surface);
} else {
inherit_format(src_params, deduced_dst.surface);
}
if (!incomplete_dst) {
inherit_format(dst_params, deduced_dst.surface);
} else {
inherit_format(dst_params, deduced_src.surface);
}
}
std::pair<TSurface, TView> InitializeSurface(GPUVAddr gpu_addr, const SurfaceParams& params,
bool preserve_contents) {
auto new_surface{GetUncachedSurface(gpu_addr, params)};
Register(new_surface);
if (preserve_contents) {
LoadSurface(new_surface);
}
return {new_surface, new_surface->GetMainView()};
}
void LoadSurface(const TSurface& surface) {
staging_cache.GetBuffer(0).resize(surface->GetHostSizeInBytes());
surface->LoadBuffer(system.GPU().MemoryManager(), staging_cache);
surface->UploadTexture(staging_cache.GetBuffer(0));
surface->MarkAsModified(false, Tick());
}
void FlushSurface(const TSurface& surface) {
if (!surface->IsModified()) {
return;
}
staging_cache.GetBuffer(0).resize(surface->GetHostSizeInBytes());
surface->DownloadTexture(staging_cache.GetBuffer(0));
surface->FlushBuffer(system.GPU().MemoryManager(), staging_cache);
surface->MarkAsModified(false, Tick());
}
void RegisterInnerCache(TSurface& surface) {
const VAddr cpu_addr = surface->GetCpuAddr();
VAddr start = cpu_addr >> registry_page_bits;
const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits;
l1_cache[cpu_addr] = surface;
while (start <= end) {
registry[start].push_back(surface);
start++;
}
}
void UnregisterInnerCache(TSurface& surface) {
const VAddr cpu_addr = surface->GetCpuAddr();
VAddr start = cpu_addr >> registry_page_bits;
const VAddr end = (surface->GetCpuAddrEnd() - 1) >> registry_page_bits;
l1_cache.erase(cpu_addr);
while (start <= end) {
auto& reg{registry[start]};
reg.erase(std::find(reg.begin(), reg.end(), surface));
start++;
}
}
VectorSurface GetSurfacesInRegion(const VAddr cpu_addr, const std::size_t size) {
if (size == 0) {
return {};
}
const VAddr cpu_addr_end = cpu_addr + size;
VAddr start = cpu_addr >> registry_page_bits;
const VAddr end = (cpu_addr_end - 1) >> registry_page_bits;
VectorSurface surfaces;
while (start <= end) {
std::vector<TSurface>& list = registry[start];
for (auto& surface : list) {
if (!surface->IsPicked() && surface->Overlaps(cpu_addr, cpu_addr_end)) {
surface->MarkAsPicked(true);
surfaces.push_back(surface);
}
}
start++;
}
for (auto& surface : surfaces) {
surface->MarkAsPicked(false);
}
return surfaces;
}
void ReserveSurface(const SurfaceParams& params, TSurface surface) {
surface_reserve[params].push_back(std::move(surface));
}
TSurface TryGetReservedSurface(const SurfaceParams& params) {
auto search{surface_reserve.find(params)};
if (search == surface_reserve.end()) {
return {};
}
for (auto& surface : search->second) {
if (!surface->IsRegistered()) {
return surface;
}
}
return {};
}
constexpr PixelFormat GetSiblingFormat(PixelFormat format) const {
return siblings_table[static_cast<std::size_t>(format)];
}
/// Returns true the shader sampler entry is compatible with the TIC texture type.
static bool IsTypeCompatible(Tegra::Texture::TextureType tic_type,
const VideoCommon::Shader::Sampler& entry) {
const auto shader_type = entry.type;
switch (tic_type) {
case Tegra::Texture::TextureType::Texture1D:
case Tegra::Texture::TextureType::Texture1DArray:
return shader_type == Tegra::Shader::TextureType::Texture1D;
case Tegra::Texture::TextureType::Texture1DBuffer:
// TODO(Rodrigo): Assume as valid for now
return true;
case Tegra::Texture::TextureType::Texture2D:
case Tegra::Texture::TextureType::Texture2DNoMipmap:
return shader_type == Tegra::Shader::TextureType::Texture2D;
case Tegra::Texture::TextureType::Texture2DArray:
return shader_type == Tegra::Shader::TextureType::Texture2D ||
shader_type == Tegra::Shader::TextureType::TextureCube;
case Tegra::Texture::TextureType::Texture3D:
return shader_type == Tegra::Shader::TextureType::Texture3D;
case Tegra::Texture::TextureType::TextureCubeArray:
case Tegra::Texture::TextureType::TextureCubemap:
if (shader_type == Tegra::Shader::TextureType::TextureCube) {
return true;
}
return shader_type == Tegra::Shader::TextureType::Texture2D && entry.is_array;
}
UNREACHABLE();
return true;
}
struct FramebufferTargetInfo {
TSurface target;
TView view;
};
void AsyncFlushSurface(TSurface& surface) {
if (!uncommitted_flushes) {
uncommitted_flushes = std::make_shared<std::list<TSurface>>();
}
uncommitted_flushes->push_back(surface);
}
VideoCore::RasterizerInterface& rasterizer;
FormatLookupTable format_lookup_table;
u64 ticks{};
// Guards the cache for protection conflicts.
bool guard_render_targets{};
bool guard_samplers{};
// The siblings table is for formats that can inter exchange with one another
// without causing issues. This is only valid when a conflict occurs on a non
// rendering use.
std::array<PixelFormat, static_cast<std::size_t>(PixelFormat::Max)> siblings_table;
// The internal Cache is different for the Texture Cache. It's based on buckets
// of 1MB. This fits better for the purpose of this cache as textures are normaly
// large in size.
static constexpr u64 registry_page_bits{20};
static constexpr u64 registry_page_size{1 << registry_page_bits};
std::unordered_map<VAddr, std::vector<TSurface>> registry;
static constexpr u32 DEPTH_RT = 8;
static constexpr u32 NO_RT = 0xFFFFFFFF;
// The L1 Cache is used for fast texture lookup before checking the overlaps
// This avoids calculating size and other stuffs.
std::unordered_map<VAddr, TSurface> l1_cache;
/// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
/// previously been used. This is to prevent surfaces from being constantly created and
/// destroyed when used with different surface parameters.
std::unordered_map<SurfaceParams, std::vector<TSurface>> surface_reserve;
std::array<FramebufferTargetInfo, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets>
render_targets;
FramebufferTargetInfo depth_buffer;
std::vector<TSurface> sampled_textures;
/// This cache stores null surfaces in order to be used as a placeholder
/// for invalid texture calls.
std::unordered_map<u32, TSurface> invalid_cache;
std::vector<u8> invalid_memory;
std::list<TSurface> marked_for_unregister;
std::shared_ptr<std::list<TSurface>> uncommitted_flushes{};
std::list<std::shared_ptr<std::list<TSurface>>> committed_flushes;
StagingCache staging_cache;
std::recursive_mutex mutex;
};
} // namespace VideoCommon
|