/
FramebufferManagerCommon.cpp
2546 lines (2227 loc) · 101 KB
/
FramebufferManagerCommon.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 (c) 2012- PPSSPP Project.
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation, version 2.0 or later versions.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License 2.0 for more details.
// A copy of the GPL 2.0 should have been included with the program.
// If not, see http://www.gnu.org/licenses/
// Official git repository and contact information can be found at
// https://github.com/hrydgard/ppsspp and http://www.ppsspp.org/.
#include <algorithm>
#include <sstream>
#include <cmath>
#include "Common/GPU/thin3d.h"
#include "Common/GPU/OpenGL/GLFeatures.h"
#include "Common/Data/Convert/ColorConv.h"
#include "Common/Data/Text/I18n.h"
#include "Common/Math/lin/matrix4x4.h"
#include "Common/Math/math_util.h"
#include "Common/System/Display.h"
#include "Common/CommonTypes.h"
#include "Core/Config.h"
#include "Core/ConfigValues.h"
#include "Core/Core.h"
#include "Core/CoreParameter.h"
#include "Core/Debugger/MemBlockInfo.h"
#include "Core/Host.h"
#include "Core/MIPS/MIPS.h"
#include "Core/Reporting.h"
#include "GPU/Common/DrawEngineCommon.h"
#include "GPU/Common/FramebufferManagerCommon.h"
#include "GPU/Common/PostShader.h"
#include "GPU/Common/PresentationCommon.h"
#include "GPU/Common/TextureCacheCommon.h"
#include "GPU/Common/ReinterpretFramebuffer.h"
#include "GPU/Debugger/Record.h"
#include "GPU/Debugger/Stepping.h"
#include "GPU/GPUInterface.h"
#include "GPU/GPUState.h"
FramebufferManagerCommon::FramebufferManagerCommon(Draw::DrawContext *draw)
: draw_(draw) {
presentation_ = new PresentationCommon(draw);
}
FramebufferManagerCommon::~FramebufferManagerCommon() {
DeviceLost();
DecimateFBOs();
for (auto vfb : vfbs_) {
DestroyFramebuf(vfb);
}
vfbs_.clear();
for (auto &tempFB : tempFBOs_) {
tempFB.second.fbo->Release();
}
tempFBOs_.clear();
// Do the same for ReadFramebuffersToMemory's VFBs
for (auto vfb : bvfbs_) {
DestroyFramebuf(vfb);
}
bvfbs_.clear();
delete presentation_;
}
void FramebufferManagerCommon::Init() {
// We may need to override the render size if the shader is upscaling or SSAA.
Resized();
}
bool FramebufferManagerCommon::UpdateSize() {
const bool newRender = renderWidth_ != (float)PSP_CoreParameter().renderWidth || renderHeight_ != (float)PSP_CoreParameter().renderHeight;
const bool newSettings = bloomHack_ != g_Config.iBloomHack || useBufferedRendering_ != (g_Config.iRenderingMode != FB_NON_BUFFERED_MODE);
renderWidth_ = (float)PSP_CoreParameter().renderWidth;
renderHeight_ = (float)PSP_CoreParameter().renderHeight;
renderScaleFactor_ = (float)PSP_CoreParameter().renderScaleFactor;
pixelWidth_ = PSP_CoreParameter().pixelWidth;
pixelHeight_ = PSP_CoreParameter().pixelHeight;
bloomHack_ = g_Config.iBloomHack;
useBufferedRendering_ = g_Config.iRenderingMode != FB_NON_BUFFERED_MODE;
presentation_->UpdateSize(pixelWidth_, pixelHeight_, renderWidth_, renderHeight_);
return newRender || newSettings;
}
void FramebufferManagerCommon::BeginFrame() {
DecimateFBOs();
currentRenderVfb_ = nullptr;
}
void FramebufferManagerCommon::SetDisplayFramebuffer(u32 framebuf, u32 stride, GEBufferFormat format) {
displayFramebufPtr_ = framebuf;
displayStride_ = stride;
displayFormat_ = format;
GPURecord::NotifyDisplay(framebuf, stride, format);
}
VirtualFramebuffer *FramebufferManagerCommon::GetVFBAt(u32 addr) const {
addr &= 0x3FFFFFFF;
VirtualFramebuffer *match = nullptr;
for (size_t i = 0; i < vfbs_.size(); ++i) {
VirtualFramebuffer *v = vfbs_[i];
if (v->fb_address == addr) {
// Could check w too but whatever (actually, might very well make sense to do so, depending on context).
if (!match || v->last_frame_render > match->last_frame_render) {
match = v;
}
}
}
return match;
}
u32 FramebufferManagerCommon::ColorBufferByteSize(const VirtualFramebuffer *vfb) const {
return vfb->fb_stride * vfb->height * (vfb->format == GE_FORMAT_8888 ? 4 : 2);
}
bool FramebufferManagerCommon::ShouldDownloadFramebuffer(const VirtualFramebuffer *vfb) const {
return PSP_CoreParameter().compat.flags().Force04154000Download && vfb->fb_address == 0x04154000;
}
// Heuristics to figure out the size of FBO to create.
// TODO: Possibly differentiate on whether through mode is used (since in through mode, viewport is meaningless?)
void FramebufferManagerCommon::EstimateDrawingSize(u32 fb_address, GEBufferFormat fb_format, int viewport_width, int viewport_height, int region_width, int region_height, int scissor_width, int scissor_height, int fb_stride, int &drawing_width, int &drawing_height) {
static const int MAX_FRAMEBUF_HEIGHT = 512;
// Games don't always set any of these. Take the greatest parameter that looks valid based on stride.
if (viewport_width > 4 && viewport_width <= fb_stride && viewport_height > 0) {
drawing_width = viewport_width;
drawing_height = viewport_height;
// Some games specify a viewport with 0.5, but don't have VRAM for 273. 480x272 is the buffer size.
if (viewport_width == 481 && region_width == 480 && viewport_height == 273 && region_height == 272) {
drawing_width = 480;
drawing_height = 272;
}
// Sometimes region is set larger than the VRAM for the framebuffer.
// However, in one game it's correctly set as a larger height (see #7277) with the same width.
// A bit of a hack, but we try to handle that unusual case here.
if (region_width <= fb_stride && (region_width > drawing_width || (region_width == drawing_width && region_height > drawing_height)) && region_height <= MAX_FRAMEBUF_HEIGHT) {
drawing_width = region_width;
drawing_height = std::max(drawing_height, region_height);
}
// Scissor is often set to a subsection of the framebuffer, so we pay the least attention to it.
if (scissor_width <= fb_stride && scissor_width > drawing_width && scissor_height <= MAX_FRAMEBUF_HEIGHT) {
drawing_width = scissor_width;
drawing_height = std::max(drawing_height, scissor_height);
}
} else {
// If viewport wasn't valid, let's just take the greatest anything regardless of stride.
drawing_width = std::min(std::max(region_width, scissor_width), fb_stride);
drawing_height = std::max(region_height, scissor_height);
}
if (scissor_width == 481 && region_width == 480 && scissor_height == 273 && region_height == 272) {
drawing_width = 480;
drawing_height = 272;
}
// Assume no buffer is > 512 tall, it couldn't be textured or displayed fully if so.
if (drawing_height >= MAX_FRAMEBUF_HEIGHT) {
if (region_height < MAX_FRAMEBUF_HEIGHT) {
drawing_height = region_height;
} else if (scissor_height < MAX_FRAMEBUF_HEIGHT) {
drawing_height = scissor_height;
}
}
if (viewport_width != region_width) {
// The majority of the time, these are equal. If not, let's check what we know.
u32 nearest_address = 0xFFFFFFFF;
for (size_t i = 0; i < vfbs_.size(); ++i) {
const u32 other_address = vfbs_[i]->fb_address & 0x3FFFFFFF;
if (other_address > fb_address && other_address < nearest_address) {
nearest_address = other_address;
}
}
// Unless the game is using overlapping buffers, the next buffer should be far enough away.
// This catches some cases where we can know this.
// Hmm. The problem is that we could only catch it for the first of two buffers...
const u32 bpp = fb_format == GE_FORMAT_8888 ? 4 : 2;
int avail_height = (nearest_address - fb_address) / (fb_stride * bpp);
if (avail_height < drawing_height && avail_height == region_height) {
drawing_width = std::min(region_width, fb_stride);
drawing_height = avail_height;
}
// Some games draw buffers interleaved, with a high stride/region/scissor but default viewport.
if (fb_stride == 1024 && region_width == 1024 && scissor_width == 1024) {
drawing_width = 1024;
}
}
DEBUG_LOG(G3D, "Est: %08x V: %ix%i, R: %ix%i, S: %ix%i, STR: %i, THR:%i, Z:%08x = %ix%i", fb_address, viewport_width,viewport_height, region_width, region_height, scissor_width, scissor_height, fb_stride, gstate.isModeThrough(), gstate.isDepthWriteEnabled() ? gstate.getDepthBufAddress() : 0, drawing_width, drawing_height);
}
void GetFramebufferHeuristicInputs(FramebufferHeuristicParams *params, const GPUgstate &gstate) {
params->fb_address = (gstate.getFrameBufRawAddress() & 0x3FFFFFFF) | 0x04000000; // GetFramebufferHeuristicInputs is only called from rendering, and thus, it's VRAM.
params->fb_stride = gstate.FrameBufStride();
params->z_address = (gstate.getDepthBufRawAddress() & 0x3FFFFFFF) | 0x04000000;
params->z_stride = gstate.DepthBufStride();
if (params->z_address == params->fb_address) {
// Probably indicates that the game doesn't care about Z for this VFB.
// Let's avoid matching it for Z copies and other shenanigans.
params->z_address = 0;
params->z_stride = 0;
}
params->fmt = gstate_c.framebufFormat;
params->isClearingDepth = gstate.isModeClear() && gstate.isClearModeDepthMask();
// Technically, it may write depth later, but we're trying to detect it only when it's really true.
if (gstate.isModeClear()) {
// Not quite seeing how this makes sense..
params->isWritingDepth = !gstate.isClearModeDepthMask() && gstate.isDepthWriteEnabled();
} else {
params->isWritingDepth = gstate.isDepthWriteEnabled();
}
params->isDrawing = !gstate.isModeClear() || !gstate.isClearModeColorMask() || !gstate.isClearModeAlphaMask();
params->isModeThrough = gstate.isModeThrough();
const bool alphaBlending = gstate.isAlphaBlendEnabled();
const bool logicOpBlending = gstate.isLogicOpEnabled() && gstate.getLogicOp() != GE_LOGIC_CLEAR && gstate.getLogicOp() != GE_LOGIC_COPY;
params->isBlending = alphaBlending || logicOpBlending;
// Viewport-X1 and Y1 are not the upper left corner, but half the width/height. A bit confusing.
float vpx = gstate.getViewportXScale();
float vpy = gstate.getViewportYScale();
// Work around problem in F1 Grand Prix, where it draws in through mode with a bogus viewport.
// We set bad values to 0 which causes the framebuffer size heuristic to rely on the other parameters instead.
if (std::isnan(vpx) || vpx > 10000000.0f) {
vpx = 0.f;
}
if (std::isnan(vpy) || vpy > 10000000.0f) {
vpy = 0.f;
}
params->viewportWidth = (int)(fabsf(vpx) * 2.0f);
params->viewportHeight = (int)(fabsf(vpy) * 2.0f);
params->regionWidth = gstate.getRegionX2() + 1;
params->regionHeight = gstate.getRegionY2() + 1;
params->scissorWidth = gstate.getScissorX2() + 1;
params->scissorHeight = gstate.getScissorY2() + 1;
if (gstate.getRegionRateX() != 0x100 || gstate.getRegionRateY() != 0x100) {
WARN_LOG_REPORT_ONCE(regionRate, G3D, "Drawing region rate add non-zero: %04x, %04x of %04x, %04x", gstate.getRegionRateX(), gstate.getRegionRateY(), gstate.getRegionX2(), gstate.getRegionY2());
}
}
VirtualFramebuffer *FramebufferManagerCommon::DoSetRenderFrameBuffer(const FramebufferHeuristicParams ¶ms, u32 skipDrawReason) {
gstate_c.Clean(DIRTY_FRAMEBUF);
// Collect all parameters. This whole function has really become a cesspool of heuristics...
// but it appears that's what it takes, unless we emulate VRAM layout more accurately somehow.
// As there are no clear "framebuffer width" and "framebuffer height" registers,
// we need to infer the size of the current framebuffer somehow.
int drawing_width, drawing_height;
EstimateDrawingSize(params.fb_address, params.fmt, params.viewportWidth, params.viewportHeight, params.regionWidth, params.regionHeight, params.scissorWidth, params.scissorHeight, std::max(params.fb_stride, (u16)4), drawing_width, drawing_height);
gstate_c.SetCurRTOffset(0, 0);
bool vfbFormatChanged = false;
if (params.fb_address == params.z_address) {
// Most likely Z will not be used in this pass, as that would wreak havoc (undefined behavior for sure)
// We probably don't need to do anything about that, but let's log it.
WARN_LOG_ONCE(color_equal_z, G3D, "Framebuffer bound with color addr == z addr, likely will not use Z in this pass: %08x", params.fb_address);
}
// Find a matching framebuffer
VirtualFramebuffer *vfb = nullptr;
for (size_t i = 0; i < vfbs_.size(); ++i) {
VirtualFramebuffer *v = vfbs_[i];
const u32 bpp = v->format == GE_FORMAT_8888 ? 4 : 2;
if (params.fb_address == v->fb_address) {
vfb = v;
// Update fb stride in case it changed
if (vfb->fb_stride != params.fb_stride) {
vfb->fb_stride = params.fb_stride;
vfbFormatChanged = true;
}
if (vfb->format != params.fmt) {
vfb->format = params.fmt;
vfbFormatChanged = true;
}
if (vfb->z_address == 0 && vfb->z_stride == 0 && params.z_stride != 0) {
// Got one that was created by CreateRAMFramebuffer. Since it has no depth buffer,
// we just recreate it immediately.
ResizeFramebufFBO(vfb, vfb->width, vfb->height, true);
}
// Keep track, but this isn't really used.
vfb->z_stride = params.z_stride;
// Heuristic: In throughmode, a higher height could be used. Let's avoid shrinking the buffer.
if (params.isModeThrough && (int)vfb->width <= params.fb_stride) {
vfb->width = std::max((int)vfb->width, drawing_width);
vfb->height = std::max((int)vfb->height, drawing_height);
} else {
vfb->width = drawing_width;
vfb->height = drawing_height;
}
break;
} else if (v->fb_stride == params.fb_stride && v->format == params.fmt) {
u32 v_fb_first_line_end_ptr = v->fb_address + v->fb_stride * 4; // This should be * bpp, but leaving like this until after 1.13 to be safe. The God of War games use this for shadows.
u32 v_fb_end_ptr = v->fb_address + v->fb_stride * v->height * bpp;
if (params.fb_address > v->fb_address && params.fb_address < v_fb_first_line_end_ptr) {
const int x_offset = (params.fb_address - v->fb_address) / bpp;
if (x_offset < params.fb_stride && v->height >= drawing_height) {
// Pretty certainly a pure render-to-X-offset.
WARN_LOG_REPORT_ONCE(renderoffset, HLE, "Rendering to framebuffer offset: %08x +%dx%d", v->fb_address, x_offset, 0);
vfb = v;
gstate_c.SetCurRTOffset(x_offset, 0);
vfb->width = std::max((int)vfb->width, x_offset + drawing_width);
// To prevent the newSize code from being confused.
drawing_width += x_offset;
break;
}
} else if (params.fb_address > v->fb_address && params.fb_address < v_fb_end_ptr && PSP_CoreParameter().compat.flags().AllowLargeFBTextureOffsets) {
// Fixes Juiced 2, though causes a lot of copying due to self-texturing. A better solution
// would be to copy from the overlapping framebuffer on bind.
if (params.fb_address % params.fb_stride == v->fb_address % params.fb_stride) {
// Framebuffers are overlapping on the Y axis.
const int y_offset = (params.fb_address - v->fb_address) / (bpp * params.fb_stride);
vfb = v;
gstate_c.SetCurRTOffset(0, y_offset);
// To prevent the newSize code from being confused.
drawing_height += y_offset;
break;
}
} else {
// We ignore this match.
// TODO: We can allow X/Y overlaps too, but haven't seen any so safer to not.
}
}
}
if (vfb) {
if ((drawing_width != vfb->bufferWidth || drawing_height != vfb->bufferHeight)) {
// Even if it's not newly wrong, if this is larger we need to resize up.
if (vfb->width > vfb->bufferWidth || vfb->height > vfb->bufferHeight) {
ResizeFramebufFBO(vfb, vfb->width, vfb->height);
} else if (vfb->newWidth != drawing_width || vfb->newHeight != drawing_height) {
// If it's newly wrong, or changing every frame, just keep track.
vfb->newWidth = drawing_width;
vfb->newHeight = drawing_height;
vfb->lastFrameNewSize = gpuStats.numFlips;
} else if (vfb->lastFrameNewSize + FBO_OLD_AGE < gpuStats.numFlips) {
// Okay, it's changed for a while (and stayed that way.) Let's start over.
// But only if we really need to, to avoid blinking.
bool needsRecreate = vfb->bufferWidth > params.fb_stride;
needsRecreate = needsRecreate || vfb->newWidth > vfb->bufferWidth || vfb->newWidth * 2 < vfb->bufferWidth;
needsRecreate = needsRecreate || vfb->newHeight > vfb->bufferHeight || vfb->newHeight * 2 < vfb->bufferHeight;
if (needsRecreate) {
ResizeFramebufFBO(vfb, vfb->width, vfb->height, true);
// Let's discard this information, might be wrong now.
vfb->safeWidth = 0;
vfb->safeHeight = 0;
} else {
// Even though we won't resize it, let's at least change the size params.
vfb->width = drawing_width;
vfb->height = drawing_height;
}
}
} else {
// It's not different, let's keep track of that too.
vfb->lastFrameNewSize = gpuStats.numFlips;
}
}
// None found? Create one.
if (!vfb) {
gstate_c.usingDepth = false; // reset depth buffer tracking
vfb = new VirtualFramebuffer{};
vfb->fbo = nullptr;
vfb->fb_address = params.fb_address;
vfb->fb_stride = params.fb_stride;
vfb->z_address = params.z_address;
vfb->z_stride = params.z_stride;
// The other width/height parameters are set in ResizeFramebufFBO below.
vfb->width = drawing_width;
vfb->height = drawing_height;
vfb->newWidth = drawing_width;
vfb->newHeight = drawing_height;
vfb->lastFrameNewSize = gpuStats.numFlips;
vfb->format = params.fmt;
vfb->drawnFormat = params.fmt;
vfb->usageFlags = FB_USAGE_RENDER_COLOR;
u32 byteSize = ColorBufferByteSize(vfb);
if (Memory::IsVRAMAddress(params.fb_address) && params.fb_address + byteSize > framebufRangeEnd_) {
framebufRangeEnd_ = params.fb_address + byteSize;
}
// This is where we actually create the framebuffer. The true is "force".
ResizeFramebufFBO(vfb, drawing_width, drawing_height, true);
NotifyRenderFramebufferCreated(vfb);
// Note that we do not even think about depth right now.
SetColorUpdated(vfb, skipDrawReason);
INFO_LOG(FRAMEBUF, "Creating FBO for %08x (z: %08x) : %d x %d x %s", vfb->fb_address, vfb->z_address, vfb->width, vfb->height, GeBufferFormatToString(vfb->format));
vfb->last_frame_render = gpuStats.numFlips;
frameLastFramebufUsed_ = gpuStats.numFlips;
vfbs_.push_back(vfb);
currentRenderVfb_ = vfb;
// Assume that if we're clearing right when switching to a new framebuffer, we don't need to upload.
if (useBufferedRendering_ && !g_Config.bDisableSlowFramebufEffects && params.isDrawing) {
gpu->PerformMemoryUpload(params.fb_address, byteSize);
// Alpha was already done by PerformMemoryUpload.
PerformStencilUpload(params.fb_address, byteSize, StencilUpload::STENCIL_IS_ZERO | StencilUpload::IGNORE_ALPHA);
// TODO: Is it worth trying to upload the depth buffer (only if it wasn't copied above..?)
}
// Let's check for depth buffer overlap. Might be interesting.
bool sharingReported = false;
for (size_t i = 0, end = vfbs_.size(); i < end; ++i) {
if (vfbs_[i]->z_stride != 0 && params.fb_address == vfbs_[i]->z_address) {
// If it's clearing it, most likely it just needs more video memory.
// Technically it could write something interesting and the other might not clear, but that's not likely.
if (params.isDrawing) {
if (params.fb_address != params.z_address && vfbs_[i]->fb_address != vfbs_[i]->z_address) {
WARN_LOG_REPORT(SCEGE, "FBO created from existing depthbuffer as color, %08x/%08x and %08x/%08x", params.fb_address, params.z_address, vfbs_[i]->fb_address, vfbs_[i]->z_address);
}
}
} else if (params.z_stride != 0 && params.z_address == vfbs_[i]->fb_address) {
// If it's clearing it, then it's probably just the reverse of the above case.
if (params.isWritingDepth) {
WARN_LOG_REPORT(SCEGE, "FBO using existing buffer as depthbuffer, %08x/%08x and %08x/%08x", params.fb_address, params.z_address, vfbs_[i]->fb_address, vfbs_[i]->z_address);
}
} else if (vfbs_[i]->z_stride != 0 && params.z_address == vfbs_[i]->z_address && params.fb_address != vfbs_[i]->fb_address && !sharingReported) {
// This happens a lot, but virtually always it's cleared.
// It's possible the other might not clear, but when every game is reported it's not useful.
if (params.isWritingDepth && (vfbs_[i]->usageFlags & FB_USAGE_RENDER_DEPTH)) {
WARN_LOG(SCEGE, "FBO reusing depthbuffer, c=%08x/d=%08x and c=%08x/d=%08x", params.fb_address, params.z_address, vfbs_[i]->fb_address, vfbs_[i]->z_address);
sharingReported = true;
}
}
}
// We already have it!
} else if (vfb != currentRenderVfb_) {
// Use it as a render target.
DEBUG_LOG(FRAMEBUF, "Switching render target to FBO for %08x: %d x %d x %d ", vfb->fb_address, vfb->width, vfb->height, vfb->format);
vfb->usageFlags |= FB_USAGE_RENDER_COLOR;
vfb->last_frame_render = gpuStats.numFlips;
frameLastFramebufUsed_ = gpuStats.numFlips;
vfb->dirtyAfterDisplay = true;
if ((skipDrawReason & SKIPDRAW_SKIPFRAME) == 0)
vfb->reallyDirtyAfterDisplay = true;
VirtualFramebuffer *prev = currentRenderVfb_;
currentRenderVfb_ = vfb;
NotifyRenderFramebufferSwitched(prev, vfb, params.isClearingDepth);
gstate_c.usingDepth = false; // reset depth buffer tracking
} else {
// Something changed, but we still got the same framebuffer we were already rendering to.
// Might not be a lot to do here, we check in NotifyRenderFramebufferUpdated
vfb->last_frame_render = gpuStats.numFlips;
frameLastFramebufUsed_ = gpuStats.numFlips;
vfb->dirtyAfterDisplay = true;
if ((skipDrawReason & SKIPDRAW_SKIPFRAME) == 0)
vfb->reallyDirtyAfterDisplay = true;
NotifyRenderFramebufferUpdated(vfb, vfbFormatChanged);
}
vfb->colorBindSeq = GetBindSeqCount();
gstate_c.curRTWidth = vfb->width;
gstate_c.curRTHeight = vfb->height;
gstate_c.curRTRenderWidth = vfb->renderWidth;
gstate_c.curRTRenderHeight = vfb->renderHeight;
return vfb;
}
// Called on the first use of depth in a render pass.
void FramebufferManagerCommon::SetDepthFrameBuffer() {
if (!currentRenderVfb_) {
return;
}
// "Resolve" the depth buffer, by copying from any overlapping buffers with fresher content.
CopyToDepthFromOverlappingFramebuffers(currentRenderVfb_);
currentRenderVfb_->usageFlags |= FB_USAGE_RENDER_DEPTH;
currentRenderVfb_->depthBindSeq = GetBindSeqCount();
}
void FramebufferManagerCommon::CopyToDepthFromOverlappingFramebuffers(VirtualFramebuffer *dest) {
struct CopySource {
VirtualFramebuffer *vfb;
RasterChannel channel;
int seq() const {
return channel == RASTER_DEPTH ? vfb->depthBindSeq : vfb->colorBindSeq;
}
bool operator < (const CopySource &other) const {
return seq() < other.seq();
}
};
std::vector<CopySource> sources;
for (auto src : vfbs_) {
if (src == dest)
continue;
if (src->fb_address == dest->z_address && src->fb_stride == dest->z_stride && src->format == GE_FORMAT_565) {
if (src->colorBindSeq > dest->depthBindSeq) {
// Source has newer data than the current buffer, use it.
sources.push_back(CopySource{ src, RASTER_COLOR });
}
} else if (src->z_address == dest->z_address && src->z_stride == dest->z_stride && src->depthBindSeq > dest->depthBindSeq) {
// Don't bother if the buffer was from another frame. This heuristic is old.
sources.push_back(CopySource{ src, RASTER_DEPTH });
} else {
// TODO: Do more detailed overlap checks here.
}
}
std::sort(sources.begin(), sources.end());
// TODO: A full copy will overwrite anything else. So we can eliminate
// anything that comes before such a copy.
// For now, let's just do the last thing, if there are multiple.
// for (auto &source : sources) {
if (sources.size()) {
auto &source = sources.back();
if (source.channel == RASTER_DEPTH) {
// Good old depth->depth copy.
BlitFramebufferDepth(source.vfb, dest);
gpuStats.numDepthCopies++;
dest->last_frame_depth_updated = gpuStats.numFlips;
} else if (source.channel == RASTER_COLOR) {
VirtualFramebuffer *src = source.vfb;
// Copying color to depth.
BlitUsingRaster(src->fbo, 0.0f, 0.0f, src->renderWidth, src->renderHeight, dest->fbo, 0.0f, 0.0f, dest->renderWidth, dest->renderHeight, false, DRAW2D_565_TO_DEPTH, "565_to_depth");
}
}
gstate_c.Dirty(DIRTY_TEXTURE_IMAGE | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_RASTER_STATE | DIRTY_DEPTHSTENCIL_STATE | DIRTY_BLEND_STATE);
}
void FramebufferManagerCommon::DestroyFramebuf(VirtualFramebuffer *v) {
// Notify the texture cache of both the color and depth buffers.
textureCache_->NotifyFramebuffer(v, NOTIFY_FB_DESTROYED);
if (v->fbo) {
v->fbo->Release();
v->fbo = nullptr;
}
// Wipe some pointers
if (currentRenderVfb_ == v)
currentRenderVfb_ = nullptr;
if (displayFramebuf_ == v)
displayFramebuf_ = nullptr;
if (prevDisplayFramebuf_ == v)
prevDisplayFramebuf_ = nullptr;
if (prevPrevDisplayFramebuf_ == v)
prevPrevDisplayFramebuf_ = nullptr;
delete v;
}
void FramebufferManagerCommon::BlitFramebufferDepth(VirtualFramebuffer *src, VirtualFramebuffer *dst) {
_dbg_assert_(src && dst);
_dbg_assert_(src != dst);
// Check that the depth address is even the same before actually blitting.
bool matchingDepthBuffer = src->z_address == dst->z_address && src->z_stride != 0 && dst->z_stride != 0;
bool matchingSize = (src->width == dst->width || (src->width == 512 && dst->width == 480) || (src->width == 480 && dst->width == 512)) && src->height == dst->height;
if (!matchingDepthBuffer || !matchingSize) {
return;
}
// Copy depth value from the previously bound framebuffer to the current one.
bool hasNewerDepth = src->last_frame_depth_render != 0 && src->last_frame_depth_render >= dst->last_frame_depth_updated;
if (!src->fbo || !dst->fbo || !useBufferedRendering_ || !hasNewerDepth) {
// If depth wasn't updated, then we're at least "two degrees" away from the data.
// This is an optimization: it probably doesn't need to be copied in this case.
return;
}
bool useCopy = draw_->GetDeviceCaps().framebufferSeparateDepthCopySupported || (!draw_->GetDeviceCaps().framebufferDepthBlitSupported && draw_->GetDeviceCaps().framebufferCopySupported);
bool useBlit = draw_->GetDeviceCaps().framebufferDepthBlitSupported;
bool useRaster = draw_->GetDeviceCaps().fragmentShaderDepthWriteSupported;
// Could do an attempt at optimization - if destination already bound, draw depth using raster.
// Let's experiment later, commented out for now. Currently we fall back to raster as a last resort here.
/*
if (currentRenderVfb_ == dst) {
useCopy = false;
useBlit = false;
}
*/
int w = std::min(src->renderWidth, dst->renderWidth);
int h = std::min(src->renderHeight, dst->renderHeight);
// TODO: It might even be advantageous on some GPUs to do this copy using a fragment shader that writes to Z, that way upcoming commands can just continue that render pass.
// Some GPUs can copy depth but only if stencil gets to come along for the ride. We only want to use this if there is no blit functionality.
if (useCopy) {
draw_->CopyFramebufferImage(src->fbo, 0, 0, 0, 0, dst->fbo, 0, 0, 0, 0, w, h, 1, Draw::FB_DEPTH_BIT, "BlitFramebufferDepth");
RebindFramebuffer("After BlitFramebufferDepth");
} else if (useBlit) {
// We'll accept whether we get a separate depth blit or not...
draw_->BlitFramebuffer(src->fbo, 0, 0, w, h, dst->fbo, 0, 0, w, h, Draw::FB_DEPTH_BIT, Draw::FB_BLIT_NEAREST, "BlitFramebufferDepth");
RebindFramebuffer("After BlitFramebufferDepth");
} else if (useRaster) {
BlitUsingRaster(src->fbo, 0, 0, w, h, dst->fbo, 0, 0, w, h, false, Draw2DShader::DRAW2D_COPY_DEPTH, "BlitDepthRaster");
}
draw_->InvalidateCachedState();
}
void FramebufferManagerCommon::NotifyRenderFramebufferCreated(VirtualFramebuffer *vfb) {
if (!useBufferedRendering_) {
// Let's ignore rendering to targets that have not (yet) been displayed.
gstate_c.skipDrawReason |= SKIPDRAW_NON_DISPLAYED_FB;
} else if (currentRenderVfb_) {
DownloadFramebufferOnSwitch(currentRenderVfb_);
}
textureCache_->NotifyFramebuffer(vfb, NOTIFY_FB_CREATED);
// Ugly...
if (gstate_c.curRTWidth != vfb->width || gstate_c.curRTHeight != vfb->height) {
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_CULLRANGE);
}
if (gstate_c.curRTRenderWidth != vfb->renderWidth || gstate_c.curRTRenderHeight != vfb->renderHeight) {
gstate_c.Dirty(DIRTY_PROJMATRIX);
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX);
}
}
void FramebufferManagerCommon::NotifyRenderFramebufferUpdated(VirtualFramebuffer *vfb, bool vfbFormatChanged) {
if (vfbFormatChanged) {
textureCache_->NotifyFramebuffer(vfb, NOTIFY_FB_UPDATED);
if (vfb->drawnFormat != vfb->format) {
ReinterpretFramebuffer(vfb, vfb->drawnFormat, vfb->format);
}
}
// ugly...
if (gstate_c.curRTWidth != vfb->width || gstate_c.curRTHeight != vfb->height) {
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_CULLRANGE);
}
if (gstate_c.curRTRenderWidth != vfb->renderWidth || gstate_c.curRTRenderHeight != vfb->renderHeight) {
gstate_c.Dirty(DIRTY_PROJMATRIX);
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX);
}
}
void FramebufferManagerCommon::NotifyRenderFramebufferSwitched(VirtualFramebuffer *prevVfb, VirtualFramebuffer *vfb, bool isClearingDepth) {
if (ShouldDownloadFramebuffer(vfb) && !vfb->memoryUpdated) {
ReadFramebufferToMemory(vfb, 0, 0, vfb->width, vfb->height);
vfb->usageFlags = (vfb->usageFlags | FB_USAGE_DOWNLOAD | FB_USAGE_FIRST_FRAME_SAVED) & ~FB_USAGE_DOWNLOAD_CLEAR;
} else {
DownloadFramebufferOnSwitch(prevVfb);
}
textureCache_->ForgetLastTexture();
shaderManager_->DirtyLastShader();
if (vfb->drawnFormat != vfb->format) {
ReinterpretFramebuffer(vfb, vfb->drawnFormat, vfb->format);
}
if (useBufferedRendering_) {
if (vfb->fbo) {
shaderManager_->DirtyLastShader();
draw_->BindFramebufferAsRenderTarget(vfb->fbo, { Draw::RPAction::KEEP, Draw::RPAction::KEEP, Draw::RPAction::KEEP }, "FramebufferSwitch");
} else {
// This should only happen very briefly when toggling useBufferedRendering_.
ResizeFramebufFBO(vfb, vfb->width, vfb->height, true);
}
} else {
if (vfb->fbo) {
// This should only happen very briefly when toggling useBufferedRendering_.
textureCache_->NotifyFramebuffer(vfb, NOTIFY_FB_DESTROYED);
vfb->fbo->Release();
vfb->fbo = nullptr;
}
// Let's ignore rendering to targets that have not (yet) been displayed.
if (vfb->usageFlags & FB_USAGE_DISPLAYED_FRAMEBUFFER) {
gstate_c.skipDrawReason &= ~SKIPDRAW_NON_DISPLAYED_FB;
} else {
gstate_c.skipDrawReason |= SKIPDRAW_NON_DISPLAYED_FB;
}
}
textureCache_->NotifyFramebuffer(vfb, NOTIFY_FB_UPDATED);
// ugly... is all this needed?
if (gstate_c.curRTWidth != vfb->width || gstate_c.curRTHeight != vfb->height) {
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_CULLRANGE);
}
if (gstate_c.curRTRenderWidth != vfb->renderWidth || gstate_c.curRTRenderHeight != vfb->renderHeight) {
gstate_c.Dirty(DIRTY_PROJMATRIX);
gstate_c.Dirty(DIRTY_PROJTHROUGHMATRIX);
}
}
void FramebufferManagerCommon::NotifyVideoUpload(u32 addr, int size, int width, GEBufferFormat fmt) {
// Note: UpdateFromMemory() is still called later.
// This is a special case where we have extra information prior to the invalidation.
// TODO: Could possibly be an offset...
VirtualFramebuffer *vfb = GetVFBAt(addr);
if (vfb) {
if (vfb->format != fmt || vfb->drawnFormat != fmt) {
DEBUG_LOG(ME, "Changing format for %08x from %d to %d", addr, vfb->drawnFormat, fmt);
vfb->format = fmt;
vfb->drawnFormat = fmt;
// Let's count this as a "render". This will also force us to use the correct format.
vfb->last_frame_render = gpuStats.numFlips;
}
if (vfb->fb_stride < width) {
DEBUG_LOG(ME, "Changing stride for %08x from %d to %d", addr, vfb->fb_stride, width);
const int bpp = fmt == GE_FORMAT_8888 ? 4 : 2;
ResizeFramebufFBO(vfb, width, size / (bpp * width));
// Resizing may change the viewport/etc.
gstate_c.Dirty(DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_CULLRANGE);
vfb->fb_stride = width;
// This might be a bit wider than necessary, but we'll redetect on next render.
vfb->width = width;
}
}
}
void FramebufferManagerCommon::UpdateFromMemory(u32 addr, int size, bool safe) {
// Take off the uncached flag from the address. Not to be confused with the start of VRAM.
addr &= 0x3FFFFFFF;
// TODO: Could go through all FBOs, but probably not important?
// TODO: Could also check for inner changes, but video is most important.
bool isDisplayBuf = addr == DisplayFramebufAddr() || addr == PrevDisplayFramebufAddr();
if (isDisplayBuf || safe) {
// TODO: Deleting the FBO is a heavy hammer solution, so let's only do it if it'd help.
if (!Memory::IsValidAddress(displayFramebufPtr_))
return;
for (size_t i = 0; i < vfbs_.size(); ++i) {
VirtualFramebuffer *vfb = vfbs_[i];
if (vfb->fb_address == addr) {
FlushBeforeCopy();
if (useBufferedRendering_ && vfb->fbo) {
GEBufferFormat fmt = vfb->format;
if (vfb->last_frame_render + 1 < gpuStats.numFlips && isDisplayBuf) {
// If we're not rendering to it, format may be wrong. Use displayFormat_ instead.
fmt = displayFormat_;
}
DrawPixels(vfb, 0, 0, Memory::GetPointer(addr), fmt, vfb->fb_stride, vfb->width, vfb->height);
SetColorUpdated(vfb, gstate_c.skipDrawReason);
} else {
INFO_LOG(FRAMEBUF, "Invalidating FBO for %08x (%i x %i x %i)", vfb->fb_address, vfb->width, vfb->height, vfb->format);
DestroyFramebuf(vfb);
vfbs_.erase(vfbs_.begin() + i--);
}
}
}
RebindFramebuffer("RebindFramebuffer - UpdateFromMemory");
}
// TODO: Necessary?
gstate_c.Dirty(DIRTY_FRAGMENTSHADER_STATE);
}
void FramebufferManagerCommon::DrawPixels(VirtualFramebuffer *vfb, int dstX, int dstY, const u8 *srcPixels, GEBufferFormat srcPixelFormat, int srcStride, int width, int height) {
textureCache_->ForgetLastTexture();
shaderManager_->DirtyLastShader(); // On GL, important that this is BEFORE drawing
float u0 = 0.0f, u1 = 1.0f;
float v0 = 0.0f, v1 = 1.0f;
DrawTextureFlags flags;
if (useBufferedRendering_ && vfb && vfb->fbo) {
flags = DRAWTEX_LINEAR;
draw_->BindFramebufferAsRenderTarget(vfb->fbo, { Draw::RPAction::KEEP, Draw::RPAction::KEEP, Draw::RPAction::KEEP }, "DrawPixels");
gstate_c.Dirty(DIRTY_VIEWPORTSCISSOR_STATE);
SetViewport2D(0, 0, vfb->renderWidth, vfb->renderHeight);
draw_->SetScissorRect(0, 0, vfb->renderWidth, vfb->renderHeight);
} else {
// We are drawing directly to the back buffer so need to flip.
// Should more of this be handled by the presentation engine?
if (needBackBufferYSwap_)
std::swap(v0, v1);
flags = g_Config.iBufFilter == SCALE_LINEAR ? DRAWTEX_LINEAR : DRAWTEX_NEAREST;
flags = flags | DRAWTEX_TO_BACKBUFFER;
FRect frame = GetScreenFrame(pixelWidth_, pixelHeight_);
FRect rc;
CenterDisplayOutputRect(&rc, 480.0f, 272.0f, frame, ROTATION_LOCKED_HORIZONTAL);
SetViewport2D(rc.x, rc.y, rc.w, rc.h);
draw_->SetScissorRect(0, 0, pixelWidth_, pixelHeight_);
}
Draw::Texture *pixelsTex = MakePixelTexture(srcPixels, srcPixelFormat, srcStride, width, height);
if (pixelsTex) {
draw_->BindTextures(0, 1, &pixelsTex);
DrawActiveTexture(dstX, dstY, width, height, vfb->bufferWidth, vfb->bufferHeight, u0, v0, u1, v1, ROTATION_LOCKED_HORIZONTAL, flags);
gpuStats.numUploads++;
pixelsTex->Release();
draw_->InvalidateCachedState();
gstate_c.Dirty(DIRTY_BLEND_STATE | DIRTY_RASTER_STATE | DIRTY_DEPTHSTENCIL_STATE | DIRTY_VIEWPORTSCISSOR_STATE | DIRTY_TEXTURE_IMAGE | DIRTY_TEXTURE_PARAMS);
}
}
bool FramebufferManagerCommon::BindFramebufferAsColorTexture(int stage, VirtualFramebuffer *framebuffer, int flags) {
if (!framebuffer->fbo || !useBufferedRendering_) {
draw_->BindTexture(stage, nullptr);
gstate_c.skipDrawReason |= SKIPDRAW_BAD_FB_TEXTURE;
return false;
}
// currentRenderVfb_ will always be set when this is called, except from the GE debugger.
// Let's just not bother with the copy in that case.
bool skipCopy = !(flags & BINDFBCOLOR_MAY_COPY) || GPUStepping::IsStepping();
// Currently rendering to this framebuffer. Need to make a copy.
if (!skipCopy && framebuffer == currentRenderVfb_) {
// TODO: Maybe merge with bvfbs_? Not sure if those could be packing, and they're created at a different size.
Draw::Framebuffer *renderCopy = GetTempFBO(TempFBO::COPY, framebuffer->renderWidth, framebuffer->renderHeight);
if (renderCopy) {
VirtualFramebuffer copyInfo = *framebuffer;
copyInfo.fbo = renderCopy;
CopyFramebufferForColorTexture(©Info, framebuffer, flags);
RebindFramebuffer("After BindFramebufferAsColorTexture");
draw_->BindFramebufferAsTexture(renderCopy, stage, Draw::FB_COLOR_BIT, 0);
} else {
draw_->BindFramebufferAsTexture(framebuffer->fbo, stage, Draw::FB_COLOR_BIT, 0);
}
return true;
} else if (framebuffer != currentRenderVfb_ || (flags & BINDFBCOLOR_FORCE_SELF) != 0) {
draw_->BindFramebufferAsTexture(framebuffer->fbo, stage, Draw::FB_COLOR_BIT, 0);
return true;
} else {
ERROR_LOG_REPORT_ONCE(vulkanSelfTexture, G3D, "Attempting to texture from target (src=%08x / target=%08x / flags=%d)", framebuffer->fb_address, currentRenderVfb_->fb_address, flags);
// To do this safely in Vulkan, we need to use input attachments.
// Actually if the texture region and render regions don't overlap, this is safe, but we need
// to transition to GENERAL image layout which will take some trickery.
// Badness on D3D11 to bind the currently rendered-to framebuffer as a texture.
draw_->BindTexture(stage, nullptr);
gstate_c.skipDrawReason |= SKIPDRAW_BAD_FB_TEXTURE;
return false;
}
}
void FramebufferManagerCommon::CopyFramebufferForColorTexture(VirtualFramebuffer *dst, VirtualFramebuffer *src, int flags) {
int x = 0;
int y = 0;
int w = src->drawnWidth;
int h = src->drawnHeight;
// If max is not > min, we probably could not detect it. Skip.
// See the vertex decoder, where this is updated.
if ((flags & BINDFBCOLOR_MAY_COPY_WITH_UV) == BINDFBCOLOR_MAY_COPY_WITH_UV && gstate_c.vertBounds.maxU > gstate_c.vertBounds.minU) {
x = std::max(gstate_c.vertBounds.minU, (u16)0);
y = std::max(gstate_c.vertBounds.minV, (u16)0);
w = std::min(gstate_c.vertBounds.maxU, src->drawnWidth) - x;
h = std::min(gstate_c.vertBounds.maxV, src->drawnHeight) - y;
// If we bound a framebuffer, apply the byte offset as pixels to the copy too.
if (flags & BINDFBCOLOR_APPLY_TEX_OFFSET) {
x += gstate_c.curTextureXOffset;
y += gstate_c.curTextureYOffset;
}
// We'll have to reapply these next time since we cropped to UV.
gstate_c.Dirty(DIRTY_TEXTURE_PARAMS);
}
if (x < src->drawnWidth && y < src->drawnHeight && w > 0 && h > 0) {
BlitFramebuffer(dst, x, y, src, x, y, w, h, 0, "Blit_CopyFramebufferForColorTexture");
}
}
Draw::Texture *FramebufferManagerCommon::MakePixelTexture(const u8 *srcPixels, GEBufferFormat srcPixelFormat, int srcStride, int width, int height) {
// TODO: We can just change the texture format and flip some bits around instead of this.
// Could share code with the texture cache perhaps.
auto generateTexture = [&](uint8_t *data, const uint8_t *initData, uint32_t w, uint32_t h, uint32_t d, uint32_t byteStride, uint32_t sliceByteStride) {
for (int y = 0; y < height; y++) {
const u16_le *src16 = (const u16_le *)srcPixels + srcStride * y;
const u32_le *src32 = (const u32_le *)srcPixels + srcStride * y;
u32 *dst = (u32 *)(data + byteStride * y);
switch (srcPixelFormat) {
case GE_FORMAT_565:
if (preferredPixelsFormat_ == Draw::DataFormat::B8G8R8A8_UNORM)
ConvertRGB565ToBGRA8888(dst, src16, width);
else
ConvertRGB565ToRGBA8888(dst, src16, width);
break;
case GE_FORMAT_5551:
if (preferredPixelsFormat_ == Draw::DataFormat::B8G8R8A8_UNORM)
ConvertRGBA5551ToBGRA8888(dst, src16, width);
else
ConvertRGBA5551ToRGBA8888(dst, src16, width);
break;
case GE_FORMAT_4444:
if (preferredPixelsFormat_ == Draw::DataFormat::B8G8R8A8_UNORM)
ConvertRGBA4444ToBGRA8888(dst, src16, width);
else
ConvertRGBA4444ToRGBA8888(dst, src16, width);
break;
case GE_FORMAT_8888:
if (preferredPixelsFormat_ == Draw::DataFormat::B8G8R8A8_UNORM)
ConvertRGBA8888ToBGRA8888(dst, src32, width);
// This means use original pointer as-is. May avoid or optimize a copy.
else if (srcStride == width)
return false;
else
memcpy(dst, src32, width * 4);
break;
case GE_FORMAT_INVALID:
case GE_FORMAT_DEPTH16:
_dbg_assert_msg_(false, "Invalid pixelFormat passed to DrawPixels().");
break;
}
}
return true;
};
Draw::TextureDesc desc{
Draw::TextureType::LINEAR2D,
preferredPixelsFormat_,
width,
height,
1,
1,
false,
"DrawPixels",
{ (uint8_t *)srcPixels },
generateTexture,
};
// Hot Shots Golf (#12355) does tons of these in a frame in some situations! So creating textures
// better be fast.
Draw::Texture *tex = draw_->CreateTexture(desc);
if (!tex)
ERROR_LOG(G3D, "Failed to create drawpixels texture");
return tex;
}
void FramebufferManagerCommon::DrawFramebufferToOutput(const u8 *srcPixels, GEBufferFormat srcPixelFormat, int srcStride) {
textureCache_->ForgetLastTexture();
shaderManager_->DirtyLastShader();
float u0 = 0.0f, u1 = 480.0f / 512.0f;
float v0 = 0.0f, v1 = 1.0f;
Draw::Texture *pixelsTex = MakePixelTexture(srcPixels, srcPixelFormat, srcStride, 512, 272);
if (!pixelsTex)
return;
int uvRotation = useBufferedRendering_ ? g_Config.iInternalScreenRotation : ROTATION_LOCKED_HORIZONTAL;
OutputFlags flags = g_Config.iBufFilter == SCALE_LINEAR ? OutputFlags::LINEAR : OutputFlags::NEAREST;
if (needBackBufferYSwap_) {
flags |= OutputFlags::BACKBUFFER_FLIPPED;
}
// DrawActiveTexture reverses these, probably to match "up".
if (GetGPUBackend() == GPUBackend::DIRECT3D9 || GetGPUBackend() == GPUBackend::DIRECT3D11) {
flags |= OutputFlags::POSITION_FLIPPED;
}
presentation_->UpdateUniforms(textureCache_->VideoIsPlaying());
presentation_->SourceTexture(pixelsTex, 512, 272);
presentation_->CopyToOutput(flags, uvRotation, u0, v0, u1, v1);
pixelsTex->Release();