/
GSRendererOGL.cpp
1482 lines (1239 loc) · 47.2 KB
/
GSRendererOGL.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) 2011-2011 Gregory hainaut
* Copyright (C) 2007-2009 Gabest
*
* 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; either version 2, or (at your option)
* any later version.
*
* 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 for more details.
*
* You should have received a copy of the GNU General Public License
* along with GNU Make; see the file COPYING. If not, write to
* the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA USA.
* http://www.gnu.org/copyleft/gpl.html
*
*/
#include "stdafx.h"
#include "GSRendererOGL.h"
#include "GSRenderer.h"
GSRendererOGL::GSRendererOGL()
: GSRendererHW(new GSTextureCacheOGL(this))
{
m_accurate_date = theApp.GetConfigB("accurate_date");
m_sw_blending = theApp.GetConfigI("accurate_blending_unit");
// Hope nothing requires too many draw calls.
m_drawlist.reserve(2048);
UserHacks_TCOffset = theApp.GetConfigI("UserHacks_TCOffset");
UserHacks_TCO_x = (UserHacks_TCOffset & 0xFFFF) / -1000.0f;
UserHacks_TCO_y = ((UserHacks_TCOffset >> 16) & 0xFFFF) / -1000.0f;
UserHacks_safe_fbmask = theApp.GetConfigB("UserHacks_safe_fbmask");
UserHacks_merge_sprite = theApp.GetConfigB("UserHacks_merge_pp_sprite");
UserHacks_unscale_pt_ln = theApp.GetConfigB("UserHacks_unscale_point_line");
m_prim_overlap = PRIM_OVERLAP_UNKNOW;
ResetStates();
if (!theApp.GetConfigB("UserHacks")) {
UserHacks_TCOffset = 0;
UserHacks_TCO_x = 0;
UserHacks_TCO_y = 0;
UserHacks_safe_fbmask = false;
UserHacks_merge_sprite = false;
UserHacks_unscale_pt_ln = false;
}
}
bool GSRendererOGL::CreateDevice(GSDevice* dev)
{
if (!GSRenderer::CreateDevice(dev))
return false;
return true;
}
void GSRendererOGL::Lines2Sprites()
{
ASSERT(m_vt.m_primclass == GS_SPRITE_CLASS);
// each sprite converted to quad needs twice the space
while(m_vertex.tail * 2 > m_vertex.maxcount)
{
GrowVertexBuffer();
}
// assume vertices are tightly packed and sequentially indexed (it should be the case)
if (m_vertex.next >= 2)
{
size_t count = m_vertex.next;
int i = (int)count * 2 - 4;
GSVertex* s = &m_vertex.buff[count - 2];
GSVertex* q = &m_vertex.buff[count * 2 - 4];
uint32* RESTRICT index = &m_index.buff[count * 3 - 6];
for(; i >= 0; i -= 4, s -= 2, q -= 4, index -= 6)
{
GSVertex v0 = s[0];
GSVertex v1 = s[1];
v0.RGBAQ = v1.RGBAQ;
v0.XYZ.Z = v1.XYZ.Z;
v0.FOG = v1.FOG;
q[0] = v0;
q[3] = v1;
// swap x, s, u
uint16 x = v0.XYZ.X;
v0.XYZ.X = v1.XYZ.X;
v1.XYZ.X = x;
float s = v0.ST.S;
v0.ST.S = v1.ST.S;
v1.ST.S = s;
uint16 u = v0.U;
v0.U = v1.U;
v1.U = u;
q[1] = v0;
q[2] = v1;
index[0] = i + 0;
index[1] = i + 1;
index[2] = i + 2;
index[3] = i + 1;
index[4] = i + 2;
index[5] = i + 3;
}
m_vertex.head = m_vertex.tail = m_vertex.next = count * 2;
m_index.tail = count * 3;
}
}
void GSRendererOGL::SetupIA(const float& sx, const float& sy)
{
GL_PUSH("IA");
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
if (UserHacks_WildHack && !isPackedUV_HackFlag && PRIM->TME && PRIM->FST) {
for(unsigned int i = 0; i < m_vertex.next; i++)
m_vertex.buff[i].UV &= 0x3FEF3FEF;
}
GLenum t = 0;
bool unscale_hack = UserHacks_unscale_pt_ln & (GetUpscaleMultiplier() > 1) & GLLoader::found_geometry_shader;
switch(m_vt.m_primclass)
{
case GS_POINT_CLASS:
if (unscale_hack) {
m_gs_sel.point = 1;
vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy);
}
t = GL_POINTS;
break;
case GS_LINE_CLASS:
if (unscale_hack) {
m_gs_sel.line = 1;
vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy);
}
t = GL_LINES;
break;
case GS_SPRITE_CLASS:
if (GLLoader::found_geometry_shader) {
m_gs_sel.sprite = 1;
t = GL_LINES;
} else {
Lines2Sprites();
t = GL_TRIANGLES;
}
break;
case GS_TRIANGLE_CLASS:
t = GL_TRIANGLES;
break;
default:
__assume(0);
}
dev->IASetVertexBuffer(m_vertex.buff, m_vertex.next);
dev->IASetIndexBuffer(m_index.buff, m_index.tail);
dev->IASetPrimitiveTopology(t);
}
void GSRendererOGL::EmulateAtst(const int pass, const GSTextureCache::Source* tex)
{
static const uint32 inverted_atst[] = {ATST_ALWAYS, ATST_NEVER, ATST_GEQUAL, ATST_GREATER, ATST_NOTEQUAL, ATST_LESS, ATST_LEQUAL, ATST_EQUAL};
int atst = (pass == 2) ? inverted_atst[m_context->TEST.ATST] : m_context->TEST.ATST;
if (!m_context->TEST.ATE) return;
switch (atst) {
case ATST_LESS:
if (tex && tex->m_spritehack_t) {
m_ps_sel.atst = 0;
} else {
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF - 0.1f;
m_ps_sel.atst = 1;
}
break;
case ATST_LEQUAL:
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF - 0.1f + 1.0f;
m_ps_sel.atst = 1;
break;
case ATST_GEQUAL:
// Maybe a -1 trick multiplication factor could be used to merge with ATST_LEQUAL case
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF - 0.1f;
m_ps_sel.atst = 2;
break;
case ATST_GREATER:
// Maybe a -1 trick multiplication factor could be used to merge with ATST_LESS case
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF - 0.1f + 1.0f;
m_ps_sel.atst = 2;
break;
case ATST_EQUAL:
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF;
m_ps_sel.atst = 3;
break;
case ATST_NOTEQUAL:
ps_cb.FogColor_AREF.a = (float)m_context->TEST.AREF;
m_ps_sel.atst = 4;
break;
case ATST_NEVER: // Draw won't be done so no need to implement it in shader
case ATST_ALWAYS:
default:
m_ps_sel.atst = 0;
break;
}
}
void GSRendererOGL::EmulateZbuffer()
{
if (m_context->TEST.ZTE) {
m_om_dssel.ztst = m_context->TEST.ZTST;
m_om_dssel.zwe = !m_context->ZBUF.ZMSK;
} else {
m_om_dssel.ztst = ZTST_ALWAYS;
}
uint32 max_z;
if (m_context->ZBUF.PSM == PSM_PSMZ32) {
max_z = 0xFFFFFFFF;
} else if (m_context->ZBUF.PSM == PSM_PSMZ24) {
max_z = 0xFFFFFF;
} else {
max_z = 0xFFFF;
}
// The real GS appears to do no masking based on the Z buffer format and writing larger Z values
// than the buffer supports seems to be an error condition on the real GS, causing it to crash.
// We are probably receiving bad coordinates from VU1 in these cases.
vs_cb.DepthMask = GSVector2i(max_z, max_z);
if (m_om_dssel.ztst >= ZTST_ALWAYS && m_om_dssel.zwe && (m_context->ZBUF.PSM != PSM_PSMZ32)) {
if (m_vt.m_max.p.z > max_z) {
ASSERT(m_vt.m_min.p.z > max_z); // sfex capcom logo
// Fixme :Following conditional fixes some dialog frame in Wild Arms 3, but may not be what was intended.
if (m_vt.m_min.p.z > max_z) {
GL_INS("Bad Z size on %s buffers", psm_str(m_context->ZBUF.PSM));
m_om_dssel.ztst = ZTST_ALWAYS;
}
}
}
GSVertex* v = &m_vertex.buff[0];
// Minor optimization of a corner case (it allow to better emulate some alpha test effects)
if (m_om_dssel.ztst == ZTST_GEQUAL && (m_vt.m_eq.xyzf & 0x4) && v[0].XYZ.Z == max_z) {
GL_INS("Optimize Z test GEQUAL to ALWAYS (%s)", psm_str(m_context->ZBUF.PSM));
m_om_dssel.ztst = ZTST_ALWAYS;
}
}
void GSRendererOGL::EmulateTextureShuffleAndFbmask()
{
if (m_texture_shuffle) {
m_ps_sel.shuffle = 1;
m_ps_sel.dfmt = 0;
const GIFRegXYOFFSET& o = m_context->XYOFFSET;
GSVertex* v = &m_vertex.buff[0];
size_t count = m_vertex.next;
// vertex position is 8 to 16 pixels, therefore it is the 16-31 bits of the colors
int pos = (v[0].XYZ.X - o.OFX) & 0xFF;
bool write_ba = (pos > 112 && pos < 136);
// Read texture is 8 to 16 pixels (same as above)
float tw = (float)(1u << m_context->TEX0.TW);
int tex_pos = (PRIM->FST) ? v[0].U : (int)(tw * v[0].ST.S);
tex_pos &= 0xFF;
m_ps_sel.read_ba = (tex_pos > 112 && tex_pos < 144);
// Convert the vertex info to a 32 bits color format equivalent
if (PRIM->FST) {
GL_INS("First vertex is P: %d => %d T: %d => %d", v[0].XYZ.X, v[1].XYZ.X, v[0].U, v[1].U);
for(size_t i = 0; i < count; i += 2) {
if (write_ba)
v[i].XYZ.X -= 128u;
else
v[i+1].XYZ.X += 128u;
if (m_ps_sel.read_ba)
v[i].U -= 128u;
else
v[i+1].U += 128u;
// Height is too big (2x).
int tex_offset = v[i].V & 0xF;
GSVector4i offset(o.OFY, tex_offset, o.OFY, tex_offset);
GSVector4i tmp(v[i].XYZ.Y, v[i].V, v[i+1].XYZ.Y, v[i+1].V);
tmp = GSVector4i(tmp - offset).srl32(1) + offset;
v[i].XYZ.Y = (uint16)tmp.x;
v[i].V = (uint16)tmp.y;
v[i+1].XYZ.Y = (uint16)tmp.z;
v[i+1].V = (uint16)tmp.w;
}
} else {
const float offset_8pix = 8.0f / tw;
GL_INS("First vertex is P: %d => %d T: %f => %f (offset %f)", v[0].XYZ.X, v[1].XYZ.X, v[0].ST.S, v[1].ST.S, offset_8pix);
for(size_t i = 0; i < count; i += 2) {
if (write_ba)
v[i].XYZ.X -= 128u;
else
v[i+1].XYZ.X += 128u;
if (m_ps_sel.read_ba)
v[i].ST.S -= offset_8pix;
else
v[i+1].ST.S += offset_8pix;
// Height is too big (2x).
GSVector4i offset(o.OFY, o.OFY);
GSVector4i tmp(v[i].XYZ.Y, v[i+1].XYZ.Y);
tmp = GSVector4i(tmp - offset).srl32(1) + offset;
//fprintf(stderr, "Before %d, After %d\n", v[i+1].XYZ.Y, tmp.y);
v[i].XYZ.Y = (uint16)tmp.x;
v[i].ST.T /= 2.0f;
v[i+1].XYZ.Y = (uint16)tmp.y;
v[i+1].ST.T /= 2.0f;
}
}
// If date is enabled you need to test the green channel instead of the
// alpha channel. Only enable this code in DATE mode to reduce the number
// of shader.
m_ps_sel.write_rg = !write_ba && m_context->TEST.DATE;
// Please bang my head against the wall!
// 1/ Reduce the frame mask to a 16 bit format
const uint32& m = m_context->FRAME.FBMSK;
uint32 fbmask = ((m >> 3) & 0x1F) | ((m >> 6) & 0x3E0) | ((m >> 9) & 0x7C00) | ((m >> 16) & 0x8000);
// FIXME GSVector will be nice here
uint8 rg_mask = fbmask & 0xFF;
uint8 ba_mask = (fbmask >> 8) & 0xFF;
m_om_csel.wrgba = 0;
// 2 Select the new mask (Please someone put SSE here)
if (rg_mask != 0xFF) {
if (write_ba) {
GL_INS("Color shuffle %s => B", m_ps_sel.read_ba ? "B" : "R");
m_om_csel.wb = 1;
} else {
GL_INS("Color shuffle %s => R", m_ps_sel.read_ba ? "B" : "R");
m_om_csel.wr = 1;
}
if (rg_mask)
m_ps_sel.fbmask = 1;
}
if (ba_mask != 0xFF) {
if (write_ba) {
GL_INS("Color shuffle %s => A", m_ps_sel.read_ba ? "A" : "G");
m_om_csel.wa = 1;
} else {
GL_INS("Color shuffle %s => G", m_ps_sel.read_ba ? "A" : "G");
m_om_csel.wg = 1;
}
if (ba_mask)
m_ps_sel.fbmask = 1;
}
if (m_ps_sel.fbmask && m_sw_blending) {
GL_INS("FBMASK SW emulated fb_mask:%x on tex shuffle", fbmask);
ps_cb.FbMask.r = rg_mask;
ps_cb.FbMask.g = rg_mask;
ps_cb.FbMask.b = ba_mask;
ps_cb.FbMask.a = ba_mask;
m_require_full_barrier = true;
} else {
m_ps_sel.fbmask = 0;
}
} else {
m_ps_sel.dfmt = GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt;
GSVector4i fbmask_v = GSVector4i::load((int)m_context->FRAME.FBMSK);
int ff_fbmask = fbmask_v.eq8(GSVector4i::xffffffff()).mask();
int zero_fbmask = fbmask_v.eq8(GSVector4i::zero()).mask();
m_om_csel.wrgba = ~ff_fbmask; // Enable channel if at least 1 bit is 0
m_ps_sel.fbmask = m_sw_blending && (~ff_fbmask & ~zero_fbmask & 0xF);
if (m_ps_sel.fbmask) {
ps_cb.FbMask = fbmask_v.u8to32();
// Only alpha is special here, I think we can take a very unsafe shortcut
// Alpha isn't blended on the GS but directly copyied into the RT.
//
// Behavior is clearly undefined however there is a high probability that
// it will work. Masked bit will be constant and normally the same everywhere
// RT/FS output/Cached value.
//
// Just to be sure let's add a new safe hack for unsafe access :)
//
// Here the GL spec quote to emphasize the unexpected behavior.
/*
- If a texel has been written, then in order to safely read the result
a texel fetch must be in a subsequent Draw separated by the command
void TextureBarrier(void);
TextureBarrier() will guarantee that writes have completed and caches
have been invalidated before subsequent Draws are executed.
*/
if (!(~ff_fbmask & ~zero_fbmask & 0x7) && !UserHacks_safe_fbmask) {
GL_INS("FBMASK Unsafe SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK,
(GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32);
m_require_one_barrier = true;
} else {
// The safe and accurate path (but slow)
GL_INS("FBMASK SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK,
(GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32);
m_require_full_barrier = true;;
}
}
}
}
void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::Source* tex)
{
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
// Uncomment to disable HLE emulation (allow to trace the draw call)
// m_channel_shuffle = false;
// First let's check we really have a channel shuffle effect
if (m_channel_shuffle) {
if (m_game.title == CRC::GT4 || m_game.title == CRC::GT3 || m_game.title == CRC::GTConcept || m_game.title == CRC::TouristTrophy) {
GL_INS("Gran Turismo RGB Channel");
m_ps_sel.channel = 7;
m_context->TEX0.TFX = TFX_DECAL;
*rt = tex->m_from_target;
} else if (m_game.title == CRC::Tekken5) {
GL_INS("Tekken5 RGB Channel");
m_ps_sel.channel = 7;
m_context->FRAME.FBMSK = 0xFF000000;
// 12 pages: 2 calls by channel, 3 channels, 1 blit
// Minus current draw call
m_skip = 12 * (3 + 3 + 1) - 1;
*rt = tex->m_from_target;
} else if ((tex->m_texture->GetType() == GSTexture::DepthStencil) && !(tex->m_32_bits_fmt)) {
// So far 2 games hit this code path. Urban Chaos and Tales of Abyss
// UC: will copy depth to green channel
// ToA: will copy depth to alpha channel
if ((m_context->FRAME.FBMSK & 0xFF0000) == 0xFF0000) {
// Green channel is masked
GL_INS("Tales Of Abyss Crazyness (MSB 16b depth to Alpha)");
m_ps_sel.tales_of_abyss_hle = 1;
} else {
GL_INS("Urban Chaos Crazyness (Green extraction)");
m_ps_sel.urban_chaos_hle = 1;
}
} else if (m_index.tail <= 64 && m_context->CLAMP.WMT == 3) {
// Blood will tell. I think it is channel effect too but again
// implemented in a different way. I don't want to add more CRC stuff. So
// let's disable channel when the signature is different
//
// Note: Tales Of Abyss and Tekken5 could hit this path too. Those games are
// handled above.
GL_INS("maybe not a channel!");
m_channel_shuffle = false;
} else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MAXU & 0x8) == 8)) {
// Read either blue or Alpha. Let's go for Blue ;)
// MGS3/Kill Zone
GL_INS("Blue channel");
m_ps_sel.channel = 3;
} else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MINU & 0x8) == 0)) {
// Read either Red or Green. Let's check the V coordinate. 0-1 is likely top so
// red. 2-3 is likely bottom so green (actually depends on texture base pointer offset)
bool green = PRIM->FST && (m_vertex.buff[0].V & 32);
if (green && (m_context->FRAME.FBMSK & 0x00FFFFFF) == 0x00FFFFFF) {
// Typically used in Terminator 3
int blue_mask = m_context->FRAME.FBMSK >> 24;
int green_mask = ~blue_mask & 0xFF;
int blue_shift = -1;
// Note: potentially we could also check the value of the clut
switch (m_context->FRAME.FBMSK >> 24) {
case 0xFF: ASSERT(0); break;
case 0xFE: blue_shift = 1; break;
case 0xFC: blue_shift = 2; break;
case 0xF8: blue_shift = 3; break;
case 0xF0: blue_shift = 4; break;
case 0xE0: blue_shift = 5; break;
case 0xC0: blue_shift = 6; break;
case 0x80: blue_shift = 7; break;
default: ASSERT(0); break;
}
int green_shift = 8 - blue_shift;
dev->SetupCBMisc(GSVector4i(blue_mask, blue_shift, green_mask, green_shift));
if (blue_shift >= 0) {
GL_INS("Green/Blue channel (%d, %d)", blue_shift, green_shift);
m_ps_sel.channel = 6;
m_context->FRAME.FBMSK = 0x00FFFFFF;
} else {
GL_INS("Green channel (wrong mask) (fbmask %x)", m_context->FRAME.FBMSK >> 24);
m_ps_sel.channel = 2;
}
} else if (green) {
GL_INS("Green channel");
m_ps_sel.channel = 2;
} else {
// Pop
GL_INS("Red channel");
m_ps_sel.channel = 1;
}
} else {
GL_INS("channel not supported");
m_channel_shuffle = false;
ASSERT(0);
}
}
// Effect is really a channel shuffle effect so let's cheat a little
if (m_channel_shuffle) {
dev->PSSetShaderResource(4, tex->m_from_target);
m_require_one_barrier = true;
// Replace current draw with a fullscreen sprite
//
// Performance GPU note: it could be wise to reduce the size to
// the rendered size of the framebuffer
GSVertex* s = &m_vertex.buff[0];
s[0].XYZ.X = (uint16)(m_context->XYOFFSET.OFX + 0);
s[1].XYZ.X = (uint16)(m_context->XYOFFSET.OFX + 16384);
s[0].XYZ.Y = (uint16)(m_context->XYOFFSET.OFY + 0);
s[1].XYZ.Y = (uint16)(m_context->XYOFFSET.OFY + 16384);
m_vertex.head = m_vertex.tail = m_vertex.next = 2;
m_index.tail = 2;
} else {
#ifdef ENABLE_OGL_DEBUG
dev->PSSetShaderResource(4, NULL);
#endif
}
}
void GSRendererOGL::EmulateBlending(bool DATE_GL42)
{
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
const GIFRegALPHA& ALPHA = m_context->ALPHA;
bool sw_blending = false;
// No blending so early exit
if (!(PRIM->ABE || PRIM->AA1 && m_vt.m_primclass == GS_LINE_CLASS)) {
#ifdef ENABLE_OGL_DEBUG
if (m_env.PABE.PABE) {
GL_INS("!!! ENV PABE without ABE !!!");
}
#endif
dev->OMSetBlendState();
return;
}
if (m_env.PABE.PABE)
{
GL_INS("!!! ENV PABE not supported !!!");
if (m_sw_blending >= ACC_BLEND_CCLIP_DALPHA) {
m_ps_sel.pabe = 1;
m_require_full_barrier |= (ALPHA.C == 1);
sw_blending = true;
}
//Breath of Fire Dragon Quarter triggers this in battles. Graphics are fine though.
//ASSERT(0);
}
// Compute the blending equation to detect special case
uint8 blend_index = uint8(((ALPHA.A * 3 + ALPHA.B) * 3 + ALPHA.C) * 3 + ALPHA.D);
int blend_flag = GSDeviceOGL::m_blendMapOGL[blend_index].bogus;
// SW Blend is (nearly) free. Let's use it.
bool impossible_or_free_blend = (blend_flag & (BLEND_NO_BAR|BLEND_A_MAX|BLEND_ACCU)) // Blend doesn't requires the costly barrier
|| (m_prim_overlap == PRIM_OVERLAP_NO) // Blend can be done in a single draw
|| (m_require_full_barrier); // Another effect (for example fbmask) already requires a full barrier
// Do the multiplication in shader for blending accumulation: Cs*As + Cd or Cs*Af + Cd
bool accumulation_blend = !!(blend_flag & BLEND_ACCU);
// Warning no break on purpose
switch (m_sw_blending) {
case ACC_BLEND_ULTRA: sw_blending |= true;
case ACC_BLEND_FULL: if (!m_vt.m_alpha.valid && (ALPHA.C == 0)) GetAlphaMinMax();
sw_blending |= (ALPHA.A != ALPHA.B) &&
((ALPHA.C == 0 && m_vt.m_alpha.max > 128) || (ALPHA.C == 2 && ALPHA.FIX > 128u));
case ACC_BLEND_CCLIP_DALPHA: sw_blending |= (ALPHA.C == 1) || (m_env.COLCLAMP.CLAMP == 0);
// Initial idea was to enable accurate blending for sprite rendering to handle
// correctly post-processing effect. Some games (ZoE) use tons of sprites as particles.
// In order to keep it fast, let's limit it to smaller draw call.
case ACC_BLEND_SPRITE: sw_blending |= m_vt.m_primclass == GS_SPRITE_CLASS && m_drawlist.size() < 100;
case ACC_BLEND_FREE: sw_blending |= impossible_or_free_blend;
default: /*sw_blending |= accumulation_blend*/;
}
// SW Blending
// GL42 interact very badly with sw blending. GL42 uses the primitiveID to find the primitive
// that write the bad alpha value. Sw blending will force the draw to run primitive by primitive
// (therefore primitiveID will be constant to 1)
sw_blending &= !DATE_GL42;
// Color clip
if (m_env.COLCLAMP.CLAMP == 0) {
if (m_prim_overlap == PRIM_OVERLAP_NO) {
// The fastest algo that requires a single pass
GL_INS("COLCLIP Free mode ENABLED");
m_ps_sel.colclip = 1;
//ASSERT(sw_blending);
sw_blending = true;
accumulation_blend = false; // disable the HDR algo
} else if (accumulation_blend) {
// A fast algo that requires 2 passes
GL_INS("COLCLIP Fast HDR mode ENABLED");
m_ps_sel.hdr = 1;
sw_blending = true; // Enable sw blending for the HDR algo
} else if (sw_blending) {
// A slow algo that could requires several passes (barely used)
GL_INS("COLCLIP SW ENABLED (blending is %d/%d/%d/%d)", ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D);
m_ps_sel.colclip = 1;
} else {
// Speed hack skip previous slow algo
GL_INS("Sorry colclip isn't supported");
}
}
// Seriously don't expect me to support this kind of crazyness.
// No mix of COLCLIP + accumulation_blend + DATE GL42
// Neither fbmask and GL42
ASSERT(!(m_ps_sel.hdr && DATE_GL42));
ASSERT(!(m_ps_sel.fbmask && DATE_GL42));
// For stat to optimize accurate option
#if 0
GL_INS("BLEND_INFO: %d/%d/%d/%d. Clamp:%d. Prim:%d number %d (drawlist %d) (sw %d)",
ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D, m_env.COLCLAMP.CLAMP, m_vt.m_primclass, m_vertex.next, m_drawlist.size(), sw_blending);
#endif
if (sw_blending) {
m_ps_sel.blend_a = ALPHA.A;
m_ps_sel.blend_b = ALPHA.B;
m_ps_sel.blend_c = ALPHA.C;
m_ps_sel.blend_d = ALPHA.D;
if (accumulation_blend) {
// Keep HW blending to do the addition/subtraction
dev->OMSetBlendState(blend_index, 0, false, true);
if (ALPHA.A == 2) {
// The blend unit does a reverse subtraction so it means
// the shader must output a positive value.
// Replace 0 - Cs by Cs - 0
m_ps_sel.blend_a = ALPHA.B;
m_ps_sel.blend_b = 2;
}
// Remove the addition/substraction from the SW blending
m_ps_sel.blend_d = 2;
// Note accumulation_blend doesn't require a barrier
} else {
// Disable HW blending
dev->OMSetBlendState();
m_require_full_barrier |= !(blend_flag & BLEND_NO_BAR);
}
// Require the fix alpha vlaue
if (ALPHA.C == 2) {
ps_cb.TA_Af.a = (float)ALPHA.FIX / 128.0f;
}
} else {
m_ps_sel.clr1 = !!(blend_flag & BLEND_C_CLR);
if (m_ps_sel.dfmt == 1 && ALPHA.C == 1) {
// 24 bits doesn't have an alpha channel so use 1.0f fix factor as equivalent
uint8 hacked_blend_index = blend_index + 3; // +3 <=> +1 on C
dev->OMSetBlendState(hacked_blend_index, 128, true);
} else {
dev->OMSetBlendState(blend_index, ALPHA.FIX, (ALPHA.C == 2));
}
}
}
void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
{
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
// Warning fetch the texture PSM format rather than the context format. The latter could have been corrected in the texture cache for depth.
//const GSLocalMemory::psm_t &psm = GSLocalMemory::m_psm[m_context->TEX0.PSM];
const GSLocalMemory::psm_t &psm = GSLocalMemory::m_psm[tex->m_TEX0.PSM];
const GSLocalMemory::psm_t &cpsm = psm.pal > 0 ? GSLocalMemory::m_psm[m_context->TEX0.CPSM] : psm;
const uint8 wms = m_context->CLAMP.WMS;
const uint8 wmt = m_context->CLAMP.WMT;
bool complex_wms_wmt = !!((wms | wmt) & 2);
bool bilinear = m_filter == 2 ? m_vt.IsLinear() : m_filter != 0;
bool simple_sample = !tex->m_palette && cpsm.fmt == 0 && !complex_wms_wmt && !psm.depth;
// Don't force extra filtering on sprite (it creates various upscaling issue)
bilinear &= !((m_vt.m_primclass == GS_SPRITE_CLASS) && m_userhacks_round_sprite_offset && !m_vt.IsLinear());
// 1 and 0 are equivalent
m_ps_sel.wms = (wms & 2) ? wms : 0;
m_ps_sel.wmt = (wmt & 2) ? wmt : 0;
// Depth + bilinear filtering isn't done yet (And I'm not sure we need it anyway but a game will prove me wrong)
// So of course, GTA set the linear mode, but sampling is done at texel center so it is equivalent to nearest sampling
ASSERT(!(psm.depth && m_vt.IsLinear()));
// Performance note:
// 1/ Don't set 0 as it is the default value
// 2/ Only keep aem when it is useful (avoid useless shader permutation)
if (m_ps_sel.shuffle) {
// Force a 32 bits access (normally shuffle is done on 16 bits)
// m_ps_sel.tex_fmt = 0; // removed as an optimization
m_ps_sel.aem = m_env.TEXA.AEM;
ASSERT(tex->m_target);
// Require a float conversion if the texure is a depth otherwise uses Integral scaling
if (psm.depth) {
m_ps_sel.depth_fmt = (tex->m_texture->GetType() != GSTexture::DepthStencil) ? 3 : 1;
}
// Shuffle is a 16 bits format, so aem is always required
GSVector4 ta(m_env.TEXA & GSVector4i::x000000ff());
ta /= 255.0f;
// FIXME rely on compiler for the optimization
ps_cb.TA_Af.x = ta.x;
ps_cb.TA_Af.y = ta.y;
// The purpose of texture shuffle is to move color channel. Extra interpolation is likely a bad idea.
bilinear &= m_vt.IsLinear();
} else if (tex->m_target) {
// Use an old target. AEM and index aren't resolved it must be done
// on the GPU
// Select the 32/24/16 bits color (AEM)
m_ps_sel.tex_fmt = cpsm.fmt;
m_ps_sel.aem = m_env.TEXA.AEM;
// Don't upload AEM if format is 32 bits
if (cpsm.fmt) {
GSVector4 ta(m_env.TEXA & GSVector4i::x000000ff());
ta /= 255.0f;
// FIXME rely on compiler for the optimization
ps_cb.TA_Af.x = ta.x;
ps_cb.TA_Af.y = ta.y;
}
// Select the index format
if (tex->m_palette) {
// FIXME Potentially improve fmt field in GSLocalMemory
if (m_context->TEX0.PSM == PSM_PSMT4HL)
m_ps_sel.tex_fmt |= 1 << 2;
else if (m_context->TEX0.PSM == PSM_PSMT4HH)
m_ps_sel.tex_fmt |= 2 << 2;
else
m_ps_sel.tex_fmt |= 3 << 2;
// Alpha channel of the RT is reinterpreted as an index. Star
// Ocean 3 uses it to emulate a stencil buffer. It is a very
// bad idea to force bilinear filtering on it.
bilinear &= m_vt.IsLinear();
}
// Depth format
if (tex->m_texture->GetType() == GSTexture::DepthStencil) {
// Require a float conversion if the texure is a depth format
m_ps_sel.depth_fmt = (psm.bpp == 16) ? 2 : 1;
// Don't force interpolation on depth format
bilinear &= m_vt.IsLinear();
} else if (psm.depth) {
// Use Integral scaling
m_ps_sel.depth_fmt = (tex->m_texture->GetType() != GSTexture::DepthStencil) ? 3 :
// Don't force interpolation on depth format
bilinear &= m_vt.IsLinear();
}
} else if (tex->m_palette) {
// Use a standard 8 bits texture. AEM is already done on the CLUT
// Therefore you only need to set the index
// m_ps_sel.aem = 0; // removed as an optimization
// Note 4 bits indexes are converted to 8 bits
m_ps_sel.tex_fmt = 3 << 2;
} else {
// Standard texture. Both index and AEM expansion were already done by the CPU.
// m_ps_sel.tex_fmt = 0; // removed as an optimization
// m_ps_sel.aem = 0; // removed as an optimization
}
if (m_context->TEX0.TFX == TFX_MODULATE && m_vt.m_eq.rgba == 0xFFFF && m_vt.m_min.c.eq(GSVector4i(128))) {
// Micro optimization that reduces GPU load (removes 5 instructions on the FS program)
m_ps_sel.tfx = TFX_DECAL;
} else {
m_ps_sel.tfx = m_context->TEX0.TFX;
}
m_ps_sel.tcc = m_context->TEX0.TCC;
m_ps_sel.ltf = bilinear && !simple_sample;
int w = tex->m_texture->GetWidth();
int h = tex->m_texture->GetHeight();
int tw = (int)(1 << m_context->TEX0.TW);
int th = (int)(1 << m_context->TEX0.TH);
GSVector4 WH(tw, th, w, h);
m_ps_sel.fst = !!PRIM->FST;
ps_cb.WH = WH;
ps_cb.HalfTexel = GSVector4(-0.5f, 0.5f).xxyy() / WH.zwzw();
if (complex_wms_wmt) {
ps_cb.MskFix = GSVector4i(m_context->CLAMP.MINU, m_context->CLAMP.MINV, m_context->CLAMP.MAXU, m_context->CLAMP.MAXV);
ps_cb.MinMax = GSVector4(ps_cb.MskFix) / WH.xyxy();
}
// TC Offset Hack
m_ps_sel.tcoffsethack = !!UserHacks_TCOffset;
ps_cb.TC_OH_TS = GSVector4(1/16.0f, 1/16.0f, UserHacks_TCO_x, UserHacks_TCO_y) / WH.xyxy();
// Only enable clamping in CLAMP mode. REGION_CLAMP will be done manually in the shader
m_ps_ssel.tau = (wms != CLAMP_CLAMP);
m_ps_ssel.tav = (wmt != CLAMP_CLAMP);
m_ps_ssel.ltf = bilinear && simple_sample;
m_ps_ssel.aniso = simple_sample;
// Setup Texture ressources
dev->SetupSampler(m_ps_ssel);
dev->PSSetShaderResources(tex->m_texture, tex->m_palette);
}
GSRendererOGL::PRIM_OVERLAP GSRendererOGL::PrimitiveOverlap()
{
// Either 1 triangle or 1 line or 3 POINTs
// It is bad for the POINTs but low probability that they overlap
if (m_vertex.next < 4)
return PRIM_OVERLAP_NO;
if (m_vt.m_primclass != GS_SPRITE_CLASS)
return PRIM_OVERLAP_UNKNOW; // maybe, maybe not
// Check intersection of sprite primitive only
size_t count = m_vertex.next;
PRIM_OVERLAP overlap = PRIM_OVERLAP_NO;
GSVertex* v = m_vertex.buff;
m_drawlist.clear();
size_t i = 0;
while (i < count) {
// In order to speed up comparison a bounding-box is accumulated. It removes a
// loop so code is much faster (check game virtua fighter). Besides it allow to check
// properly the Y order.
// .x = min(v[i].XYZ.X, v[i+1].XYZ.X)
// .y = min(v[i].XYZ.Y, v[i+1].XYZ.Y)
// .z = max(v[i].XYZ.X, v[i+1].XYZ.X)
// .w = max(v[i].XYZ.Y, v[i+1].XYZ.Y)
GSVector4i all = GSVector4i(v[i].m[1]).upl16(GSVector4i(v[i+1].m[1])).upl16().xzyw();
all = all.xyxy().blend(all.zwzw(), all > all.zwxy());
size_t j = i + 2;
while (j < count) {
GSVector4i sprite = GSVector4i(v[j].m[1]).upl16(GSVector4i(v[j+1].m[1])).upl16().xzyw();
sprite = sprite.xyxy().blend(sprite.zwzw(), sprite > sprite.zwxy());
// Be sure to get vertex in good order, otherwise .r* function doesn't
// work as expected.
ASSERT(sprite.x <= sprite.z);
ASSERT(sprite.y <= sprite.w);
ASSERT(all.x <= all.z);
ASSERT(all.y <= all.w);
if (all.rintersect(sprite).rempty()) {
all = all.runion_ordered(sprite);
} else {
overlap = PRIM_OVERLAP_YES;
break;
}
j += 2;
}
m_drawlist.push_back((j - i) >> 1); // Sprite count
i = j;
}
#if 0
// Old algo: less constraint but O(n^2) instead of O(n) as above
// You have no guarantee on the sprite order, first vertex can be either top-left or bottom-left
// There is a high probability that the draw call will uses same ordering for all vertices.
// In order to keep a small performance impact only the first sprite will be checked
//
// Some safe-guard will be added in the outer-loop to avoid corruption with a limited perf impact
if (v[1].XYZ.Y < v[0].XYZ.Y) {
// First vertex is Top-Left
for(size_t i = 0; i < count; i += 2) {
if (v[i+1].XYZ.Y > v[i].XYZ.Y) {
return PRIM_OVERLAP_UNKNOW;
}
GSVector4i vi(v[i].XYZ.X, v[i+1].XYZ.Y, v[i+1].XYZ.X, v[i].XYZ.Y);
for (size_t j = i+2; j < count; j += 2) {
GSVector4i vj(v[j].XYZ.X, v[j+1].XYZ.Y, v[j+1].XYZ.X, v[j].XYZ.Y);
GSVector4i inter = vi.rintersect(vj);
if (!inter.rempty()) {
return PRIM_OVERLAP_YES;
}
}
}
} else {
// First vertex is Bottom-Left
for(size_t i = 0; i < count; i += 2) {
if (v[i+1].XYZ.Y < v[i].XYZ.Y) {
return PRIM_OVERLAP_UNKNOW;
}
GSVector4i vi(v[i].XYZ.X, v[i].XYZ.Y, v[i+1].XYZ.X, v[i+1].XYZ.Y);
for (size_t j = i+2; j < count; j += 2) {
GSVector4i vj(v[j].XYZ.X, v[j].XYZ.Y, v[j+1].XYZ.X, v[j+1].XYZ.Y);
GSVector4i inter = vi.rintersect(vj);
if (!inter.rempty()) {
return PRIM_OVERLAP_YES;
}
}
}
}
#endif
//fprintf(stderr, "%d: Yes, code can be optimized (draw of %d vertices)\n", s_n, count);
return overlap;
}
GSVector4i GSRendererOGL::ComputeBoundingBox(const GSVector2& rtscale, const GSVector2i& rtsize)
{
GSVector4 scale = GSVector4(rtscale.x, rtscale.y);
GSVector4 offset = GSVector4(-1.0f, 1.0f); // Round value
GSVector4 box = m_vt.m_min.p.xyxy(m_vt.m_max.p) + offset.xxyy();
return GSVector4i(box * scale.xyxy()).rintersect(GSVector4i(0, 0, rtsize.x, rtsize.y));
}
void GSRendererOGL::SendDraw()
{
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
if (!m_require_full_barrier && m_require_one_barrier) {
// Need only a single barrier
glTextureBarrier();
dev->DrawIndexedPrimitive();
} else if (!m_require_full_barrier) {
// Don't need any barrier
dev->DrawIndexedPrimitive();
} else if (m_prim_overlap == PRIM_OVERLAP_NO) {
// Need full barrier but a single barrier will be enough
glTextureBarrier();
dev->DrawIndexedPrimitive();
} else if (m_vt.m_primclass == GS_SPRITE_CLASS) {
size_t nb_vertex = (GLLoader::found_geometry_shader) ? 2 : 6;
GL_PUSH("Split the draw (SPRITE)");
#if defined(_DEBUG)
// Check how draw call is split.
map<size_t, size_t> frequency;
for (const auto& it: m_drawlist)
++frequency[it];