-
Notifications
You must be signed in to change notification settings - Fork 2.6k
/
TextureCacheBase.cpp
2937 lines (2579 loc) · 110 KB
/
TextureCacheBase.cpp
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
// Copyright 2010 Dolphin Emulator Project
// Licensed under GPLv2+
// Refer to the license.txt file included.
#include <algorithm>
#include <cmath>
#include <cstring>
#include <memory>
#include <string>
#include <utility>
#include <vector>
#if defined(_M_X86) || defined(_M_X86_64)
#include <pmmintrin.h>
#endif
#include "Common/Align.h"
#include "Common/Assert.h"
#include "Common/ChunkFile.h"
#include "Common/CommonTypes.h"
#include "Common/FileUtil.h"
#include "Common/Hash.h"
#include "Common/Logging/Log.h"
#include "Common/MathUtil.h"
#include "Common/MemoryUtil.h"
#include "Common/StringUtil.h"
#include "Core/Config/GraphicsSettings.h"
#include "Core/ConfigManager.h"
#include "Core/FifoPlayer/FifoPlayer.h"
#include "Core/FifoPlayer/FifoRecorder.h"
#include "Core/HW/Memmap.h"
#include "VideoCommon/AbstractFramebuffer.h"
#include "VideoCommon/AbstractStagingTexture.h"
#include "VideoCommon/BPMemory.h"
#include "VideoCommon/FramebufferManager.h"
#include "VideoCommon/HiresTextures.h"
#include "VideoCommon/PixelShaderManager.h"
#include "VideoCommon/RenderBase.h"
#include "VideoCommon/SamplerCommon.h"
#include "VideoCommon/ShaderCache.h"
#include "VideoCommon/Statistics.h"
#include "VideoCommon/TextureCacheBase.h"
#include "VideoCommon/TextureConversionShader.h"
#include "VideoCommon/TextureConverterShaderGen.h"
#include "VideoCommon/TextureDecoder.h"
#include "VideoCommon/VertexManagerBase.h"
#include "VideoCommon/VideoCommon.h"
#include "VideoCommon/VideoConfig.h"
static const u64 TEXHASH_INVALID = 0;
// Sonic the Fighters (inside Sonic Gems Collection) loops a 64 frames animation
static const int TEXTURE_KILL_THRESHOLD = 64;
static const int TEXTURE_POOL_KILL_THRESHOLD = 3;
std::unique_ptr<TextureCacheBase> g_texture_cache;
std::bitset<8> TextureCacheBase::valid_bind_points;
TextureCacheBase::TCacheEntry::TCacheEntry(std::unique_ptr<AbstractTexture> tex,
std::unique_ptr<AbstractFramebuffer> fb)
: texture(std::move(tex)), framebuffer(std::move(fb))
{
}
TextureCacheBase::TCacheEntry::~TCacheEntry()
{
for (auto& reference : references)
reference->references.erase(this);
}
void TextureCacheBase::CheckTempSize(size_t required_size)
{
if (required_size <= temp_size)
return;
temp_size = required_size;
Common::FreeAlignedMemory(temp);
temp = static_cast<u8*>(Common::AllocateAlignedMemory(temp_size, 16));
}
TextureCacheBase::TextureCacheBase()
{
SetBackupConfig(g_ActiveConfig);
temp_size = 2048 * 2048 * 4;
temp = static_cast<u8*>(Common::AllocateAlignedMemory(temp_size, 16));
TexDecoder_SetTexFmtOverlayOptions(backup_config.texfmt_overlay,
backup_config.texfmt_overlay_center);
HiresTexture::Init();
Common::SetHash64Function();
InvalidateAllBindPoints();
}
TextureCacheBase::~TextureCacheBase()
{
// Clear pending EFB copies first, so we don't try to flush them.
m_pending_efb_copies.clear();
HiresTexture::Shutdown();
Invalidate();
Common::FreeAlignedMemory(temp);
temp = nullptr;
}
bool TextureCacheBase::Initialize()
{
if (!CreateUtilityTextures())
{
PanicAlert("Failed to create utility textures.");
return false;
}
return true;
}
void TextureCacheBase::Invalidate()
{
FlushEFBCopies();
InvalidateAllBindPoints();
for (size_t i = 0; i < bound_textures.size(); ++i)
{
bound_textures[i] = nullptr;
}
for (auto& tex : textures_by_address)
{
delete tex.second;
}
textures_by_address.clear();
textures_by_hash.clear();
texture_pool.clear();
}
void TextureCacheBase::OnConfigChanged(VideoConfig& config)
{
if (config.bHiresTextures != backup_config.hires_textures ||
config.bCacheHiresTextures != backup_config.cache_hires_textures)
{
HiresTexture::Update();
}
// TODO: Invalidating texcache is really stupid in some of these cases
if (config.iSafeTextureCache_ColorSamples != backup_config.color_samples ||
config.bTexFmtOverlayEnable != backup_config.texfmt_overlay ||
config.bTexFmtOverlayCenter != backup_config.texfmt_overlay_center ||
config.bHiresTextures != backup_config.hires_textures ||
config.bEnableGPUTextureDecoding != backup_config.gpu_texture_decoding ||
config.bDisableCopyToVRAM != backup_config.disable_vram_copies ||
config.bArbitraryMipmapDetection != backup_config.arbitrary_mipmap_detection)
{
Invalidate();
TexDecoder_SetTexFmtOverlayOptions(g_ActiveConfig.bTexFmtOverlayEnable,
g_ActiveConfig.bTexFmtOverlayCenter);
}
SetBackupConfig(config);
}
void TextureCacheBase::Cleanup(int _frameCount)
{
TexAddrCache::iterator iter = textures_by_address.begin();
TexAddrCache::iterator tcend = textures_by_address.end();
while (iter != tcend)
{
if (iter->second->tmem_only)
{
iter = InvalidateTexture(iter);
}
else if (iter->second->frameCount == FRAMECOUNT_INVALID)
{
iter->second->frameCount = _frameCount;
++iter;
}
else if (_frameCount > TEXTURE_KILL_THRESHOLD + iter->second->frameCount)
{
if (iter->second->IsCopy())
{
// Only remove EFB copies when they wouldn't be used anymore(changed hash), because EFB
// copies living on the
// host GPU are unrecoverable. Perform this check only every TEXTURE_KILL_THRESHOLD for
// performance reasons
if ((_frameCount - iter->second->frameCount) % TEXTURE_KILL_THRESHOLD == 1 &&
iter->second->hash != iter->second->CalculateHash())
{
iter = InvalidateTexture(iter);
}
else
{
++iter;
}
}
else
{
iter = InvalidateTexture(iter);
}
}
else
{
++iter;
}
}
TexPool::iterator iter2 = texture_pool.begin();
TexPool::iterator tcend2 = texture_pool.end();
while (iter2 != tcend2)
{
if (iter2->second.frameCount == FRAMECOUNT_INVALID)
{
iter2->second.frameCount = _frameCount;
}
if (_frameCount > TEXTURE_POOL_KILL_THRESHOLD + iter2->second.frameCount)
{
iter2 = texture_pool.erase(iter2);
}
else
{
++iter2;
}
}
}
bool TextureCacheBase::TCacheEntry::OverlapsMemoryRange(u32 range_address, u32 range_size) const
{
if (addr + size_in_bytes <= range_address)
return false;
if (addr >= range_address + range_size)
return false;
return true;
}
void TextureCacheBase::SetBackupConfig(const VideoConfig& config)
{
backup_config.color_samples = config.iSafeTextureCache_ColorSamples;
backup_config.texfmt_overlay = config.bTexFmtOverlayEnable;
backup_config.texfmt_overlay_center = config.bTexFmtOverlayCenter;
backup_config.hires_textures = config.bHiresTextures;
backup_config.cache_hires_textures = config.bCacheHiresTextures;
backup_config.stereo_3d = config.stereo_mode != StereoMode::Off;
backup_config.efb_mono_depth = config.bStereoEFBMonoDepth;
backup_config.gpu_texture_decoding = config.bEnableGPUTextureDecoding;
backup_config.disable_vram_copies = config.bDisableCopyToVRAM;
backup_config.arbitrary_mipmap_detection = config.bArbitraryMipmapDetection;
}
TextureCacheBase::TCacheEntry*
TextureCacheBase::ApplyPaletteToEntry(TCacheEntry* entry, u8* palette, TLUTFormat tlutfmt)
{
DEBUG_ASSERT(g_ActiveConfig.backend_info.bSupportsPaletteConversion);
const AbstractPipeline* pipeline = g_shader_cache->GetPaletteConversionPipeline(tlutfmt);
if (!pipeline)
{
ERROR_LOG(VIDEO, "Failed to get conversion pipeline for format 0x%02X",
static_cast<u32>(tlutfmt));
return nullptr;
}
TextureConfig new_config = entry->texture->GetConfig();
new_config.levels = 1;
new_config.flags |= AbstractTextureFlag_RenderTarget;
TCacheEntry* decoded_entry = AllocateCacheEntry(new_config);
if (!decoded_entry)
return nullptr;
decoded_entry->SetGeneralParameters(entry->addr, entry->size_in_bytes, entry->format,
entry->should_force_safe_hashing);
decoded_entry->SetDimensions(entry->native_width, entry->native_height, 1);
decoded_entry->SetHashes(entry->base_hash, entry->hash);
decoded_entry->frameCount = FRAMECOUNT_INVALID;
decoded_entry->should_force_safe_hashing = false;
decoded_entry->SetNotCopy();
decoded_entry->may_have_overlapping_textures = entry->may_have_overlapping_textures;
g_renderer->BeginUtilityDrawing();
const u32 palette_size = entry->format == TextureFormat::I4 ? 32 : 512;
u32 texel_buffer_offset;
if (g_vertex_manager->UploadTexelBuffer(palette, palette_size,
TexelBufferFormat::TEXEL_BUFFER_FORMAT_R16_UINT,
&texel_buffer_offset))
{
struct Uniforms
{
float multiplier;
u32 texel_buffer_offset;
u32 pad[2];
};
static_assert(std::is_standard_layout<Uniforms>::value);
Uniforms uniforms = {};
uniforms.multiplier = entry->format == TextureFormat::I4 ? 15.0f : 255.0f;
uniforms.texel_buffer_offset = texel_buffer_offset;
g_vertex_manager->UploadUtilityUniforms(&uniforms, sizeof(uniforms));
g_renderer->SetAndDiscardFramebuffer(decoded_entry->framebuffer.get());
g_renderer->SetViewportAndScissor(decoded_entry->texture->GetRect());
g_renderer->SetPipeline(pipeline);
g_renderer->SetTexture(1, entry->texture.get());
g_renderer->SetSamplerState(1, RenderState::GetPointSamplerState());
g_renderer->Draw(0, 3);
g_renderer->EndUtilityDrawing();
decoded_entry->texture->FinishedRendering();
}
else
{
ERROR_LOG(VIDEO, "Texel buffer upload of %u bytes failed", palette_size);
g_renderer->EndUtilityDrawing();
}
textures_by_address.emplace(decoded_entry->addr, decoded_entry);
return decoded_entry;
}
TextureCacheBase::TCacheEntry* TextureCacheBase::ReinterpretEntry(const TCacheEntry* existing_entry,
TextureFormat new_format)
{
const AbstractPipeline* pipeline =
g_shader_cache->GetTextureReinterpretPipeline(existing_entry->format.texfmt, new_format);
if (!pipeline)
{
ERROR_LOG(VIDEO,
"Failed to obtain texture reinterpreting pipeline from format 0x%02X to 0x%02X",
static_cast<u32>(existing_entry->format.texfmt), static_cast<u32>(new_format));
return nullptr;
}
TextureConfig new_config = existing_entry->texture->GetConfig();
new_config.levels = 1;
new_config.flags |= AbstractTextureFlag_RenderTarget;
TCacheEntry* reinterpreted_entry = AllocateCacheEntry(new_config);
if (!reinterpreted_entry)
return nullptr;
reinterpreted_entry->SetGeneralParameters(existing_entry->addr, existing_entry->size_in_bytes,
new_format, existing_entry->should_force_safe_hashing);
reinterpreted_entry->SetDimensions(existing_entry->native_width, existing_entry->native_height,
1);
reinterpreted_entry->SetHashes(existing_entry->base_hash, existing_entry->hash);
reinterpreted_entry->frameCount = existing_entry->frameCount;
reinterpreted_entry->SetNotCopy();
reinterpreted_entry->is_efb_copy = existing_entry->is_efb_copy;
reinterpreted_entry->may_have_overlapping_textures =
existing_entry->may_have_overlapping_textures;
g_renderer->BeginUtilityDrawing();
g_renderer->SetAndDiscardFramebuffer(reinterpreted_entry->framebuffer.get());
g_renderer->SetViewportAndScissor(reinterpreted_entry->texture->GetRect());
g_renderer->SetPipeline(pipeline);
g_renderer->SetTexture(0, existing_entry->texture.get());
g_renderer->SetSamplerState(1, RenderState::GetPointSamplerState());
g_renderer->Draw(0, 3);
g_renderer->EndUtilityDrawing();
reinterpreted_entry->texture->FinishedRendering();
textures_by_address.emplace(reinterpreted_entry->addr, reinterpreted_entry);
return reinterpreted_entry;
}
void TextureCacheBase::ScaleTextureCacheEntryTo(TextureCacheBase::TCacheEntry* entry, u32 new_width,
u32 new_height)
{
if (entry->GetWidth() == new_width && entry->GetHeight() == new_height)
{
return;
}
const u32 max = g_ActiveConfig.backend_info.MaxTextureSize;
if (max < new_width || max < new_height)
{
ERROR_LOG(VIDEO, "Texture too big, width = %d, height = %d", new_width, new_height);
return;
}
const TextureConfig newconfig(new_width, new_height, 1, entry->GetNumLayers(), 1,
AbstractTextureFormat::RGBA8, AbstractTextureFlag_RenderTarget);
std::optional<TexPoolEntry> new_texture = AllocateTexture(newconfig);
if (!new_texture)
{
ERROR_LOG(VIDEO, "Scaling failed due to texture allocation failure");
return;
}
// No need to convert the coordinates here since they'll be the same.
g_renderer->ScaleTexture(new_texture->framebuffer.get(),
new_texture->texture->GetConfig().GetRect(), entry->texture.get(),
entry->texture->GetConfig().GetRect());
entry->texture.swap(new_texture->texture);
entry->framebuffer.swap(new_texture->framebuffer);
// At this point new_texture has the old texture in it,
// we can potentially reuse this, so let's move it back to the pool
auto config = new_texture->texture->GetConfig();
texture_pool.emplace(
config, TexPoolEntry(std::move(new_texture->texture), std::move(new_texture->framebuffer)));
}
bool TextureCacheBase::CheckReadbackTexture(u32 width, u32 height, AbstractTextureFormat format)
{
if (m_readback_texture && m_readback_texture->GetConfig().width >= width &&
m_readback_texture->GetConfig().height >= height &&
m_readback_texture->GetConfig().format == format)
{
return true;
}
TextureConfig staging_config(std::max(width, 128u), std::max(height, 128u), 1, 1, 1, format, 0);
m_readback_texture.reset();
m_readback_texture =
g_renderer->CreateStagingTexture(StagingTextureType::Readback, staging_config);
return m_readback_texture != nullptr;
}
void TextureCacheBase::SerializeTexture(AbstractTexture* tex, const TextureConfig& config,
PointerWrap& p)
{
// If we're in measure mode, skip the actual readback to save some time.
const bool skip_readback = p.GetMode() == PointerWrap::MODE_MEASURE;
p.DoPOD(config);
std::vector<u8> texture_data;
if (skip_readback || CheckReadbackTexture(config.width, config.height, config.format))
{
// Save out each layer of the texture to the staging texture, and then
// append it onto the end of the vector. This gives us all the sub-images
// in one single buffer which can be written out to the save state.
for (u32 layer = 0; layer < config.layers; layer++)
{
for (u32 level = 0; level < config.levels; level++)
{
u32 level_width = std::max(config.width >> level, 1u);
u32 level_height = std::max(config.height >> level, 1u);
auto rect = tex->GetConfig().GetMipRect(level);
if (!skip_readback)
m_readback_texture->CopyFromTexture(tex, rect, layer, level, rect);
size_t stride = AbstractTexture::CalculateStrideForFormat(config.format, level_width);
size_t size = stride * level_height;
size_t start = texture_data.size();
texture_data.resize(texture_data.size() + size);
if (!skip_readback)
m_readback_texture->ReadTexels(rect, &texture_data[start], static_cast<u32>(stride));
}
}
}
else
{
PanicAlert("Failed to create staging texture for serialization");
}
p.Do(texture_data);
}
std::optional<TextureCacheBase::TexPoolEntry> TextureCacheBase::DeserializeTexture(PointerWrap& p)
{
TextureConfig config;
p.Do(config);
std::vector<u8> texture_data;
p.Do(texture_data);
if (p.GetMode() != PointerWrap::MODE_READ || texture_data.empty())
return std::nullopt;
auto tex = AllocateTexture(config);
if (!tex)
{
PanicAlert("Failed to create texture for deserialization");
return std::nullopt;
}
size_t start = 0;
for (u32 layer = 0; layer < config.layers; layer++)
{
for (u32 level = 0; level < config.levels; level++)
{
u32 level_width = std::max(config.width >> level, 1u);
u32 level_height = std::max(config.height >> level, 1u);
size_t stride = AbstractTexture::CalculateStrideForFormat(config.format, level_width);
size_t size = stride * level_height;
if ((start + size) > texture_data.size())
{
ERROR_LOG(VIDEO, "Insufficient texture data for layer %u level %u", layer, level);
return tex;
}
tex->texture->Load(level, level_width, level_height, level_width, &texture_data[start], size);
start += size;
}
}
return tex;
}
void TextureCacheBase::DoState(PointerWrap& p)
{
// Flush all pending XFB copies before either loading or saving.
FlushEFBCopies();
p.Do(last_entry_id);
if (p.GetMode() == PointerWrap::MODE_WRITE || p.GetMode() == PointerWrap::MODE_MEASURE)
DoSaveState(p);
else
DoLoadState(p);
}
void TextureCacheBase::DoSaveState(PointerWrap& p)
{
std::map<const TCacheEntry*, u32> entry_map;
std::vector<TCacheEntry*> entries_to_save;
auto ShouldSaveEntry = [](const TCacheEntry* entry) {
// We skip non-copies as they can be decoded from RAM when the state is loaded.
// Storing them would duplicate data in the save state file, adding to decompression time.
return entry->IsCopy();
};
auto AddCacheEntryToMap = [&entry_map, &entries_to_save, &p](TCacheEntry* entry) -> u32 {
auto iter = entry_map.find(entry);
if (iter != entry_map.end())
return iter->second;
// Since we are sequentially allocating texture entries, we need to save the textures in the
// same order they were collected. This is because of iterating both the address and hash maps.
// Therefore, the map is used for fast lookup, and the vector for ordering.
u32 id = static_cast<u32>(entry_map.size());
entry_map.emplace(entry, id);
entries_to_save.push_back(entry);
return id;
};
auto GetCacheEntryId = [&entry_map](const TCacheEntry* entry) -> std::optional<u32> {
auto iter = entry_map.find(entry);
return iter != entry_map.end() ? std::make_optional(iter->second) : std::nullopt;
};
// Transform the textures_by_address and textures_by_hash maps to a mapping
// of address/hash to entry ID.
std::vector<std::pair<u32, u32>> textures_by_address_list;
std::vector<std::pair<u64, u32>> textures_by_hash_list;
if (Config::Get(Config::GFX_SAVE_TEXTURE_CACHE_TO_STATE))
{
for (const auto& it : textures_by_address)
{
if (ShouldSaveEntry(it.second))
{
u32 id = AddCacheEntryToMap(it.second);
textures_by_address_list.push_back(std::make_pair(it.first, id));
}
}
for (const auto& it : textures_by_hash)
{
if (ShouldSaveEntry(it.second))
{
u32 id = AddCacheEntryToMap(it.second);
textures_by_hash_list.push_back(std::make_pair(it.first, id));
}
}
}
// Save the texture cache entries out in the order the were referenced.
u32 size = static_cast<u32>(entries_to_save.size());
p.Do(size);
for (TCacheEntry* entry : entries_to_save)
{
g_texture_cache->SerializeTexture(entry->texture.get(), entry->texture->GetConfig(), p);
entry->DoState(p);
}
p.DoMarker("TextureCacheEntries");
// Save references for each cache entry.
// As references are circular, we need to have everything created before linking entries.
std::set<std::pair<u32, u32>> reference_pairs;
for (const auto& it : entry_map)
{
const TCacheEntry* entry = it.first;
auto id1 = GetCacheEntryId(entry);
if (!id1)
continue;
for (const TCacheEntry* referenced_entry : entry->references)
{
auto id2 = GetCacheEntryId(referenced_entry);
if (!id2)
continue;
auto refpair1 = std::make_pair(*id1, *id2);
auto refpair2 = std::make_pair(*id2, *id1);
if (reference_pairs.count(refpair1) == 0 && reference_pairs.count(refpair2) == 0)
reference_pairs.insert(refpair1);
}
}
size = static_cast<u32>(reference_pairs.size());
p.Do(size);
for (const auto& it : reference_pairs)
{
p.Do(it.first);
p.Do(it.second);
}
size = static_cast<u32>(textures_by_address_list.size());
p.Do(size);
for (const auto& it : textures_by_address_list)
{
p.Do(it.first);
p.Do(it.second);
}
size = static_cast<u32>(textures_by_hash_list.size());
p.Do(size);
for (const auto& it : textures_by_hash_list)
{
p.Do(it.first);
p.Do(it.second);
}
// Free the readback texture to potentially save host-mapped GPU memory, depending on where
// the driver mapped the staging buffer.
m_readback_texture.reset();
}
void TextureCacheBase::DoLoadState(PointerWrap& p)
{
// Helper for getting a cache entry from an ID.
std::map<u32, TCacheEntry*> id_map;
auto GetEntry = [&id_map](u32 id) {
auto iter = id_map.find(id);
return iter == id_map.end() ? nullptr : iter->second;
};
// Only clear out state when actually restoring/loading.
// Since we throw away entries when not in loading mode now, we don't need to check
// before inserting entries into the cache, as GetEntry will always return null.
const bool commit_state = p.GetMode() == PointerWrap::MODE_READ;
if (commit_state)
Invalidate();
// Preload all cache entries.
u32 size = 0;
p.Do(size);
for (u32 i = 0; i < size; i++)
{
// Even if the texture isn't valid, we still need to create the cache entry object
// to update the point in the state state. We'll just throw it away if it's invalid.
auto tex = g_texture_cache->DeserializeTexture(p);
TCacheEntry* entry = new TCacheEntry(std::move(tex->texture), std::move(tex->framebuffer));
entry->textures_by_hash_iter = g_texture_cache->textures_by_hash.end();
entry->DoState(p);
if (entry->texture && commit_state)
id_map.emplace(i, entry);
else
delete entry;
}
p.DoMarker("TextureCacheEntries");
// Link all cache entry references.
p.Do(size);
for (u32 i = 0; i < size; i++)
{
u32 id1 = 0, id2 = 0;
p.Do(id1);
p.Do(id2);
TCacheEntry* e1 = GetEntry(id1);
TCacheEntry* e2 = GetEntry(id2);
if (e1 && e2)
e1->CreateReference(e2);
}
// Fill in address map.
p.Do(size);
for (u32 i = 0; i < size; i++)
{
u32 addr = 0;
u32 id = 0;
p.Do(addr);
p.Do(id);
TCacheEntry* entry = GetEntry(id);
if (entry)
textures_by_address.emplace(addr, entry);
}
// Fill in hash map.
p.Do(size);
for (u32 i = 0; i < size; i++)
{
u64 hash = 0;
u32 id = 0;
p.Do(hash);
p.Do(id);
TCacheEntry* entry = GetEntry(id);
if (entry)
entry->textures_by_hash_iter = textures_by_hash.emplace(hash, entry);
}
}
void TextureCacheBase::TCacheEntry::DoState(PointerWrap& p)
{
p.Do(addr);
p.Do(size_in_bytes);
p.Do(base_hash);
p.Do(hash);
p.Do(format);
p.Do(memory_stride);
p.Do(is_efb_copy);
p.Do(is_custom_tex);
p.Do(may_have_overlapping_textures);
p.Do(tmem_only);
p.Do(has_arbitrary_mips);
p.Do(should_force_safe_hashing);
p.Do(is_xfb_copy);
p.Do(is_xfb_container);
p.Do(id);
p.Do(reference_changed);
p.Do(native_width);
p.Do(native_height);
p.Do(native_levels);
p.Do(frameCount);
}
TextureCacheBase::TCacheEntry*
TextureCacheBase::DoPartialTextureUpdates(TCacheEntry* entry_to_update, u8* palette,
TLUTFormat tlutfmt)
{
// If the flag may_have_overlapping_textures is cleared, there are no overlapping EFB copies,
// which aren't applied already. It is set for new textures, and for the affected range
// on each EFB copy.
if (!entry_to_update->may_have_overlapping_textures)
return entry_to_update;
entry_to_update->may_have_overlapping_textures = false;
const bool isPaletteTexture = IsColorIndexed(entry_to_update->format.texfmt);
// EFB copies are excluded from these updates, until there's an example where a game would
// benefit from updating. This would require more work to be done.
if (entry_to_update->IsCopy())
return entry_to_update;
u32 block_width = TexDecoder_GetBlockWidthInTexels(entry_to_update->format.texfmt);
u32 block_height = TexDecoder_GetBlockHeightInTexels(entry_to_update->format.texfmt);
u32 block_size = block_width * block_height *
TexDecoder_GetTexelSizeInNibbles(entry_to_update->format.texfmt) / 2;
u32 numBlocksX = (entry_to_update->native_width + block_width - 1) / block_width;
auto iter = FindOverlappingTextures(entry_to_update->addr, entry_to_update->size_in_bytes);
while (iter.first != iter.second)
{
TCacheEntry* entry = iter.first->second;
if (entry != entry_to_update && entry->IsCopy() && !entry->tmem_only &&
entry->references.count(entry_to_update) == 0 &&
entry->OverlapsMemoryRange(entry_to_update->addr, entry_to_update->size_in_bytes) &&
entry->memory_stride == numBlocksX * block_size)
{
if (entry->hash == entry->CalculateHash())
{
// If the texture formats are not compatible or convertible, skip it.
if (!IsCompatibleTextureFormat(entry_to_update->format.texfmt, entry->format.texfmt))
{
if (!CanReinterpretTextureOnGPU(entry_to_update->format.texfmt, entry->format.texfmt))
{
++iter.first;
continue;
}
TCacheEntry* reinterpreted_entry =
ReinterpretEntry(entry, entry_to_update->format.texfmt);
if (reinterpreted_entry)
entry = reinterpreted_entry;
}
if (isPaletteTexture)
{
TCacheEntry* decoded_entry = ApplyPaletteToEntry(entry, palette, tlutfmt);
if (decoded_entry)
{
// Link the efb copy with the partially updated texture, so we won't apply this partial
// update again
entry->CreateReference(entry_to_update);
// Mark the texture update as used, as if it was loaded directly
entry->frameCount = FRAMECOUNT_INVALID;
entry = decoded_entry;
}
else
{
++iter.first;
continue;
}
}
u32 src_x, src_y, dst_x, dst_y;
// Note for understanding the math:
// Normal textures can't be strided, so the 2 missing cases with src_x > 0 don't exist
if (entry->addr >= entry_to_update->addr)
{
u32 block_offset = (entry->addr - entry_to_update->addr) / block_size;
u32 block_x = block_offset % numBlocksX;
u32 block_y = block_offset / numBlocksX;
src_x = 0;
src_y = 0;
dst_x = block_x * block_width;
dst_y = block_y * block_height;
}
else
{
u32 block_offset = (entry_to_update->addr - entry->addr) / block_size;
u32 block_x = (~block_offset + 1) % numBlocksX;
u32 block_y = (block_offset + block_x) / numBlocksX;
src_x = 0;
src_y = block_y * block_height;
dst_x = block_x * block_width;
dst_y = 0;
}
u32 copy_width =
std::min(entry->native_width - src_x, entry_to_update->native_width - dst_x);
u32 copy_height =
std::min(entry->native_height - src_y, entry_to_update->native_height - dst_y);
// If one of the textures is scaled, scale both with the current efb scaling factor
if (entry_to_update->native_width != entry_to_update->GetWidth() ||
entry_to_update->native_height != entry_to_update->GetHeight() ||
entry->native_width != entry->GetWidth() || entry->native_height != entry->GetHeight())
{
ScaleTextureCacheEntryTo(entry_to_update,
g_renderer->EFBToScaledX(entry_to_update->native_width),
g_renderer->EFBToScaledY(entry_to_update->native_height));
ScaleTextureCacheEntryTo(entry, g_renderer->EFBToScaledX(entry->native_width),
g_renderer->EFBToScaledY(entry->native_height));
src_x = g_renderer->EFBToScaledX(src_x);
src_y = g_renderer->EFBToScaledY(src_y);
dst_x = g_renderer->EFBToScaledX(dst_x);
dst_y = g_renderer->EFBToScaledY(dst_y);
copy_width = g_renderer->EFBToScaledX(copy_width);
copy_height = g_renderer->EFBToScaledY(copy_height);
}
// If the source rectangle is outside of what we actually have in VRAM, skip the copy.
// The backend doesn't do any clamping, so if we don't, we'd pass out-of-range coordinates
// to the graphics driver, which can cause GPU resets.
if (static_cast<u32>(src_x + copy_width) > entry->GetWidth() ||
static_cast<u32>(src_y + copy_height) > entry->GetHeight() ||
static_cast<u32>(dst_x + copy_width) > entry_to_update->GetWidth() ||
static_cast<u32>(dst_y + copy_height) > entry_to_update->GetHeight())
{
++iter.first;
continue;
}
MathUtil::Rectangle<int> srcrect, dstrect;
srcrect.left = src_x;
srcrect.top = src_y;
srcrect.right = (src_x + copy_width);
srcrect.bottom = (src_y + copy_height);
dstrect.left = dst_x;
dstrect.top = dst_y;
dstrect.right = (dst_x + copy_width);
dstrect.bottom = (dst_y + copy_height);
// If one copy is stereo, and the other isn't... not much we can do here :/
const u32 layers_to_copy = std::min(entry->GetNumLayers(), entry_to_update->GetNumLayers());
for (u32 layer = 0; layer < layers_to_copy; layer++)
{
entry_to_update->texture->CopyRectangleFromTexture(entry->texture.get(), srcrect, layer,
0, dstrect, layer, 0);
}
if (isPaletteTexture)
{
// Remove the temporary converted texture, it won't be used anywhere else
// TODO: It would be nice to convert and copy in one step, but this code path isn't common
iter.first = InvalidateTexture(iter.first);
continue;
}
else
{
// Link the two textures together, so we won't apply this partial update again
entry->CreateReference(entry_to_update);
// Mark the texture update as used, as if it was loaded directly
entry->frameCount = FRAMECOUNT_INVALID;
}
}
else
{
// If the hash does not match, this EFB copy will not be used for anything, so remove it
iter.first = InvalidateTexture(iter.first);
continue;
}
}
++iter.first;
}
return entry_to_update;
}
void TextureCacheBase::DumpTexture(TCacheEntry* entry, std::string basename, unsigned int level,
bool is_arbitrary)
{
std::string szDir = File::GetUserPath(D_DUMPTEXTURES_IDX) + SConfig::GetInstance().GetGameID();
// make sure that the directory exists
if (!File::IsDirectory(szDir))
File::CreateDir(szDir);
if (is_arbitrary)
{
basename += "_arb";
}
if (level > 0)
{
basename += StringFromFormat("_mip%i", level);
}
std::string filename = szDir + "/" + basename + ".png";
if (!File::Exists(filename))
entry->texture->Save(filename, level);
}
static u32 CalculateLevelSize(u32 level_0_size, u32 level)
{
return std::max(level_0_size >> level, 1u);
}
static void SetSamplerState(u32 index, float custom_tex_scale, bool custom_tex,
bool has_arbitrary_mips)
{
const FourTexUnits& tex = bpmem.tex[index / 4];
const TexMode0& tm0 = tex.texMode0[index % 4];
SamplerState state = {};
state.Generate(bpmem, index);
// Force texture filtering config option.
if (g_ActiveConfig.bForceFiltering)
{
state.min_filter = SamplerState::Filter::Linear;
state.mag_filter = SamplerState::Filter::Linear;
state.mipmap_filter = SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0) ?
SamplerState::Filter::Linear :
SamplerState::Filter::Point;
}
// Custom textures may have a greater number of mips
if (custom_tex)
state.max_lod = 255;
// Anisotropic filtering option.
if (g_ActiveConfig.iMaxAnisotropy != 0 && !SamplerCommon::IsBpTexMode0PointFiltering(tm0))
{
// https://www.opengl.org/registry/specs/EXT/texture_filter_anisotropic.txt
// For predictable results on all hardware/drivers, only use one of:
// GL_LINEAR + GL_LINEAR (No Mipmaps [Bilinear])
// GL_LINEAR + GL_LINEAR_MIPMAP_LINEAR (w/ Mipmaps [Trilinear])
// Letting the game set other combinations will have varying arbitrary results;
// possibly being interpreted as equal to bilinear/trilinear, implicitly
// disabling anisotropy, or changing the anisotropic algorithm employed.
state.min_filter = SamplerState::Filter::Linear;
state.mag_filter = SamplerState::Filter::Linear;
if (SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0))
state.mipmap_filter = SamplerState::Filter::Linear;
state.anisotropic_filtering = 1;
}
else
{
state.anisotropic_filtering = 0;
}
if (has_arbitrary_mips && SamplerCommon::AreBpTexMode0MipmapsEnabled(tm0))
{
// Apply a secondary bias calculated from the IR scale to pull inwards mipmaps
// that have arbitrary contents, eg. are used for fog effects where the
// distance they kick in at is important to preserve at any resolution.
// Correct this with the upscaling factor of custom textures.
s64 lod_offset = std::log2(g_renderer->GetEFBScale() / custom_tex_scale) * 256.f;
state.lod_bias = std::clamp<s64>(state.lod_bias + lod_offset, -32768, 32767);
// Anisotropic also pushes mips farther away so it cannot be used either
state.anisotropic_filtering = 0;
}
g_renderer->SetSamplerState(index, state);
}
void TextureCacheBase::BindTextures()
{